structural and functional studies on e.coli diacylglycerol

Dissertation zur Erlangung des Doktorgrades der Naturwissenschaften

vorgelegt beim Fachbereich 14

der Johann Wolfgang Goethe – Universität in Frankfurt am Main

Structural and Functional Studies

on E.coli Diacylglycerol Kinase

by MAS NMR Spectroscopy

von

Kristin Möbius

aus Riesa

Frankfurt am Main, 2018

(D30)

vom Fachbereich 14

der Johann Wolfgang Goethe – Universität als Dissertation angenommen

Dekan: Prof. Dr. Clemens Glaubitz

Erster Gutachter: Prof. Dr. Clemens Glaubitz

Zweiter Gutachter: Prof. Dr. Volker Dötsch

Datum der Disputation:

Summary

Summary

The focus of this thesis is the integral membrane protein Escherichia coli diacylglycerol

kinase (DGK). It is located within the inner membrane, where it catalyzes the ATP-

dependent phosphorylation of diacylglycerol (DAG) to phosphatic acid (PA). It plays an

important role in recycling DAG during the biosynthesis of membrane-derived

oligosaccharides (MDOs) [1, 2] and lipopolysaccharides (LPSs) [3]. DGK is a unique

enzyme, which does not share any sequence homology with typical kinases. With

43 kDa, it is the smallest known kinase [4]. It features a notable complexity in structure

and function [4-7] as well as a remarkable stability [8, 9]. It is a homotrimer, in which

each monomer contains three transmembrane helices and one N-terminal amphiphilic

surface helix [5]. The trimer features three active sites arranged around the

membrane/cytoplasm interface [5]. Each active site is formed by two adjacent

protomers, leading to an unusual catalytic site architecture of the composite shared site

model [5, 7, 10]. Several mutational studies [6, 7, 10], MD simulations [6, 11], solution

NMR studies in detergent micelles [7] and particularly the 3D structure determination

by X-ray crystallography [5, 6] offer insights into a possible catalytic mechanism of

DGK. However, important long-standing questions remain unsolved.

The aim of this thesis is the investigation of DGK’s structure and function at an atomic

level directly within the native-like lipid bilayer using MAS NMR. This way, a deeper

understanding of DGK’s catalytic mechanism should be obtained.

First, the preparation of DGK was optimized, in order to achieve high amounts of pure,

active, stable and homogeneously reconstituted protein, which provides well-resolved

MAS NMR spectra (chapter 3). For this purpose, the preparation protocol that has been

used before in this lab was optimized successively, defining the DGK construct, type of

detergent, lipid composition, reconstitution method and protein-to-lipid ratio. The quality

of the sample was characterized by SDS- and BN-PAGE, SEC, LILBID-MS, sucrose

density gradient centrifugation, the coupled activity assay and MAS NMR. The

optimization resulted in a efficient protocol that provides a DGK sample, which is more

native than in previous studies and, in addition, features MAS NMR spectra of high

resolution and sensitivity.

The high quality MAS NMR spectra formed the foundation for the second step, the

resonance assignment of DGK’s backbone and side chains (chapter 4). The

assignment was performed at high magnetic field (1H frequency 850 MHz). The

Summary

sequential assignment of immobile domains was carried out using a combination of

dipolar coupling based 3D experiments, NCACX, NCOCX and CONCA. The

measurement time could be reduced by paramagnetic doping with Gd3+-DOTA [12] in

combination with a custom-built E-free probehead (Bruker). The sequential assignment

was mainly performed using a uniformly labelled sample (U-13C,15N-DGK). Residual

ambiguities could be resolved by reverse labelling of isoleucine, leucine and valine (U-

13C,15N-DGK-I,L,V). The carbon and nitrogen resonances could be assigned for

approximately 82% of the residues, from which 74% were completely assigned. For

validation, ssFLYA was applied, which is a generally applicable algorithm for the

automatic assignment of protein solid state NMR spectra [13]. Its principal applicability

for demanding systems as membrane proteins could be proven in this study for the first

time. Overall, 91.5% of the backbone and 89.1% of all (backbone + side chains)

manually obtained assignments could be confirmed by ssFLYA. For the completion of

DGK’s assignment, scalar coupling based 2D experiments, 1H-13C/15N HETCOR and

13C-13C TOBSY, were carried out to detect highly mobile residues. This way, residues

of the two termini and the cytosolic loop, which were not detectable by dipolar coupling

based experiments, could be assigned tentatively. Whereupon, peaks for arginine and

lysine were assigned unambiguously to Arg9 and Lys12. Overall, 84% of the residues

located in transmembrane and extramembranous regions could be assigned with the

used NMR strategy.

During the assignment procedure, no systematic peak doublets or triplets were

detected, indicating that the DGK trimer adopts a symmetric conformation in its apo

state. This is in contrast to the X-ray structure, which shows asymmetries between the

three subunits. The differences are most likely attributed to different experimental

conditions. Especially, crystal packing may be a potential source for structural

asymmetries.

Based on the resonance assignment, a secondary structure analysis was carried out. It

showed substantial similarities between wild-type DGK, its thermostable mutant [14]

determined both by MAS NMR and the crystal structure of wtDGK [6]. However, there

are few differences. In contrast to both MAS NMR secondary structures, the crystal

structure shows small deviations around the flexible regions. The largest discrepancy

occurs at the cytosolic loop (CL) concerning its length and position. It is shifted from the

residues 83-87 in the MAS NMR structure of wtDGK to the residues 83–90 (subunit A),

86-91 (subunit B) and 82-87 (subunit C) of the X-ray structure.

In addition to 13C/15N detected experiments at a moderate MAS rate, very first 1H

detected experiments in combination with ultra-fast MAS at 111 kHz were carried out

as well, demonstrating its general applicability on fully protonated DGK in lipid bilayers.

Summary

On the basis of the nearly complete assignment of DGK, the apo state was compared

with the substrate bound states (chapter 5). Perturbations in peak position and intensity

of the substrate bound states were analysed for each of the 101 assigned residues in

3D and 2D heteronuclear correlation spectra. The nucleotide-bound state was

emulated by adenylylmethylenediphosphonate (AMP-PCP), a non-hydrolysable ATP

analogue, whereas the DAG-bound state was mimicked by 1,2-dioctanoyl-sn-glycerol

(DOG, chain length n = 8). For finding saturation conditions, a competitive Mg*ATP

inhibition assay was performed by monitoring the ATPase activity as a function of

Mg*AMP-PCP concentration. Additionally, the binding of AMP-PCP and DOG was

verified by 31P-cross polarization (CP) experiments.

Significant chemical shift perturbations and altered peak intensities could be observed

in both the AMP-PCP and DOG bound state. These data provide evidence that all

three active sites are occupied at the same time. Additionally, it could be demonstrated

that the nucleotide substrate induces a substantial conformational change. This most

likely supports the enzyme in binding of the lipid substrate, indicating positive

heteroallostery. For DGK bound with either AMP-PCP + DOG or only AMP-PCP, a

similar spectral fingerprint was observed. This implies that binding of the nucleotide

seems to set the enzyme into a catalytic active state, triggering the actual phosphoryl

transfer reaction.

The investigation of DGK’s remarkable stability and the cross-talk between its subunits

forms the last part of this thesis (chapter 6). This demands for the identification of key

intra- and interprotomer contacts, which are of structural or functional importance. For

this purpose, 13C-13C DARR and 2D NCOCX spectra with long mixing times were

recorded using high field MAS NMR. Additionally, DNP-enhanced 13C−15N TEDOR

experiments were conducted on mixed labelled DGK trimers to enable the visualization

of interprotomer contacts. In order to generate mixed labelled trimers, a procedure was

established using SDS. With the applied NMR strategy, functionally relevant intra-

(Arg32 - Trp25/ Glu28/ Ala29 and Trp112 - Ser61) and interprotomer (ArgNn,e -

AspCg/ GluCd/ AsnCg) long-range interactions could be identified. Based on the

crystal structure, the interprotomer contacts can be most likely attributed to Arg81-

Glu88 and Arg92-Asn27/Glu28. The identified interactions may stabilize the active sites

and/or transmit information about substrate binding or changes of the surrounding lipid

bilayer within and/or between protomers.

Zusammenfassung

Zusammenfassung

Im Fokus dieser Dissertation steht das integrale Membranprotein Escherichia coli

Diacylglycerolkinase (DGK). Es befindet sich in der inneren Membran, wo es die ATP-

abhängige Phosphorylierung von Diacylglycerol (DAG) zu Phosphatidylsäure (PA)

katalysiert. Es spielt eine wichtige Rolle im Recycling von DAG während der

Biosynthese von Membran-abgeleiteten Oligosacchariden (MDOs) [1, 2] und

Lipopolysacchariden (LPSs) [3]. MDOs werden beispielsweise unter Bedingungen

geringer Osmolarität in großen Mengen produziert [15, 16], während LPSs die

Hauptbestandteile der bakteriellen Außenmembran darstellen [17]. Beide

Komponenten dienen dem Schutz des gram-negativen Bakteriums. E.coli DGK ist ein

einzigartiges Enzym, welches weder Struktur- noch Sequenzähnlichkeiten zu anderen

Kinasen aufweist. Mit 43 kDa ist es die kleinste bekannte Kinase [4]. DGK verfügt über

eine bemerkenswerte Komplexität in seiner Struktur und Funktion [4-7] sowie in seiner

Stabilität [8, 9]. Li et al. bestimmten 2013 die 3D Kristallstruktur von DGK in lipidisch-

kubischen Phasen bestehend aus Monoacylglycerolen, die auch als Substrat fungieren

[5]. Die Kristallstruktur zeigt DGK als Homotrimer, in welchem jedes Monomer drei

Transmembranhelices und eine N-terminale amphiphile Oberflächenhelix besitzt. Das

Trimer hat drei aktive Stellen, welche um die Membran/Zytoplasma-Kontaktfläche

angeordnet sind. Jede aktive Stelle wird von den Transmembranhelices einer

Untereinheit und der Oberflächenhelix einer benachbarten Untereinheit gebildet.

Daraus ergibt sich eine ungewöhnliche Architektur der katalytisch-aktiven Stellen

basierend auf dem sogenannten „composite shared site model“ [5, 7, 10].

Verschiedene Mutationsstudien [6, 7, 10], MD Simulationen [6, 11], Lösungs-NMR-

Studien in Detergenz-Mizellen [7] und vor allem die 3D Strukturbestimmung durch

Röntgenkristallographie [5, 6] liefern Einblicke in einen möglichen katalytischen

Mechanismus von DGK. Allerdings bleiben wichtige, seit Langem bestehende Fragen

unbeantwortet. Es ist beispielsweise unklar, ob das DGK-Trimer eine symmetrische

oder asymmetrische Konformation einnimmt. Außerdem ist noch nicht bekannt, ob sich

alle drei aktiven Zentren von DGK während der Katalyse im selben oder in

unterschiedlichen Zuständen befinden. Desweiteren ist noch nicht erwiesen, ob DGK

eine entscheidende Konformationsänderung vor dem eigentlichen Phoshoryltransfer

durchläuft. Eine Vermutung diesbezüglich wurde bereits 1997 von Badola und Sanders

angestellt [18], konnte allerdings bisher noch nicht belegt werden. Betrachtet man die

hohe Stabilität des Enzyms und die Tatsache, dass jedes aktive Zentrum durch zwei

benachbarte Untereinheiten gebildet wird, ergibt sich zudem die Frage, ob spezifische

Zusammenfassung

Intra- und Interprotomerkontakte existieren. Die vorliegende Dissertation widmet sich

diesen Fragen unter Verwendung von Festkörper-Nuklearmagnetischer

Resonanzspektroskopie (FK-NMR). Im Detail wurde MAS NMR Spektroskopie

angewandt, welche ein Drehen der Probe im sogenannten magischen Winkel von

54.74° impliziert. Hierbei handelt es sich um eine etablierte Methode für Studien an

Membranproteinen, welche zunehmend an Bedeutung in der Strukturbiologie gewinnt

und wesentliche komplementäre Daten zur Röntgenkristallographie,

Kryoelektronenmikroskopie und Lösungs-NMR liefert. Basierend auf der hohen

Sensitivität der NMR-Signale bzgl. der lokalen Umgebung können selbst schwache

Ligandenbindungen durch Änderungen in der chemischen Verschiebung analysiert

werden, was wiederum Struktur-Aktivitäts-Korrelationen erlaubt. MAS NMR bietet

zudem den Vorteil, dass es nicht die strukturelle Plastizität des entsprechenden

Membranproteins einschränkt, welche in den meisten Fällen von funktionaler

Bedeutung ist. Weiterhin bietet MAS NMR die Möglichkeit Membranproteine direkt in

der Lipiddoppelschicht zu untersuchen. Damit wird das zu untersuchende System

physiologischen Bedingungen näher gebracht als in anderen Membran-imitierenden

Umgebungen wie beispielsweise in Detergenz-Mizellen. Die Lipiddoppelschicht stellt

einen wichtigen strukturellen Faktor dar und ist zudem in den meisten Fällen direkt mit

der katalytischen Aktivität des Membranproteins verbunden [11, 19-21].

Das Ziel dieser Dissertation ist die Untersuchung der Struktur und Funktion von DGK

auf atomarem Level, direkt in der Lipiddoppelschicht mittels MAS NMR. Auf diesem

Weg soll ein tieferes Verständnis von DGK‘s katalytischem Mechanismus erlangt

werden.

Hierfür wurde zunächst die Präparation von DGK optimiert, um hohe Ausbeuten an

reinem, aktivem, stabilem und homogen-rekonstituiertem Protein zu erzielen, welches

gut-aufgelöste FK-NMR Spektren liefert (Kapitel 3). Dazu wurde das

Präparationsprotokoll schrittweise optimiert, welches zuvor in dieser Arbeitsgruppe

verwendet wurde. Hierbei wurden das DGK-Konstrukt, die Art des Detergenzes, die

Lipidzusammensetzung, die Rekonstitutionsmethode und das Protein-Lipid-Verhältnis

neu definiert: Es wurde Wildtyp-DGK, dessen hohe Stabilität bereits nachgewiesen

worden ist, anstelle der thermostabilen Quadrupelmutante (Δ4-DGK: I53C, I70L, M96L,

V107D) verwendet. Außerdem wurde das schonende, nicht-ionische Detergenz n-

Dodecyl-β-d-Maltopyranosid (DDM) anstatt des harschen zwitter-ionischen

Detergenzes Dodecyl Phosphocholin (DPC) genutzt. Die Reinheit von Wildtyp-DGK in

Detergenz-Mizellen wurde durch SDS-PAGE und SEC charakterisiert, während der

Zusammenfassung

oligomere Zustand mittels BN-PAGE und LILBID-MS bewertet wurde. Die Ausbeute an

DGK konnte um 50% von 20-30 mg DGK pro Liter Zellkultur auf 30-45 mg/l erhöht

werden. Weiterhin wurde die Rekonstitutionsmethode geändert. Statt einer

langwierigen Dialyse wurde zur hydrophoben Absorption mittels BioBeads gewechselt,

was den Zeitaufwand von zwei Wochen auf zwei Tage reduzierte. Zudem wurde die

Lipidkomposition modifiziert. Anstelle von 67.3mol% DMPC/ 32.7mol% Cholesterol

wurde die dem System besser entsprechende Lipidzusammensetzung aus 90mol%

DMPC/ 10mol% DMPA verwendet, die zudem zu einer höheren spektralen Auflösung

führte. Außerdem wurde das molare Protein-Lipid-Verhältnis von 1:80 auf 1:50 erhöht.

Somit konnten 30% mehr Protein in den Rotor gepackt werden, was die spektrale

Sensitivität verbesserte. Die Sucrose-Dichtegradienten-Zentrifugation wurde genutzt,

um eine homogene Proteinrekonstitution in die Liposomen zu überprüfen. Die Qualität

der optimierten Proteoliposomprobe wurde außerdem mit Hilfe eines gekoppelten

Aktivitätsassays und MAS NMR charakterisiert. Die Optimierung führte zu einem

effizienten Protokoll, welches DGK-Proben liefert, die nativer sind als in bisherigen

Studien und zudem FK-NMR-Spektren von hoher Auflösung und hoher Sensitivität

aufweisen.

Die qualitativ-hochwertigen FK-NMR-Spektren bildeten die Grundlage für den zweiten

wichtigen Schritt, die Resonanz-Zuordnung von DGK’s Rückgrat und Seitenketten

(Kapitel 4). Die Zuordnung wurde mittels multidimensionaler FK-NMR bei hohem

Magnetfeld (1H Frequenz von 850 MHz) durchgeführt. Die sequentielle Zuordnung der

immobilen Domänen erfolgte durch eine Kombination aus 3D Experimenten (NCACX,

NCOCX, CONCA), die auf der dipolaren Kopplung basieren. Die Messzeit konnte

durch paramagnetisches Doping mit Gd3+-DOTA in Kombination mit einem speziell

angefertigten E-freien Probenkopf (Bruker) reduziert werden. Die Zuordnung erfolgte

hauptsächlich an einer uniform-markierten Probe (U-13C,15N-DGK). Verbleibende

Ambiguitäten konnten mittels reverse labelling von Isoleucin, Leucin und Valin (U-

13C,15N-DGK-I,L,V) beseitigt werden. Die 13C- und 15N-Resonanzen konnten für ca.

82% der Reste zugeordnet werden. Davon wurden 74% vollständig zugeordnet. Zur

Validierung wurde ssFLYA angewandt. Hierbei handelt es sich um einen allgemein

anwendbaren Algorithmus zur automatischen Zuordnung von Protein-FK-NMR-

Spektren. Seine prinzipielle Anwendbarkeit für anspruchsvolle Systeme wie

Membranproteine konnte mittels dieser Studie nachgewiesen werden. Insgesamt

wurden 91,5% der manuell erhaltenen Rückgratzuordnungen und 89,1% aller

(Rückgrat und Seitenketten) manuellen Zuordungen durch ssFLYA bestätigt. Zur

Vervollständigung der Zuordnung von DGK wurden zudem J-Kopplung basierte 2D

Zusammenfassung

Experimente (1H-13C/15N HETCOR und 13C-13C TOBSY) zur Detektion von mobilen

Resten durchgeführt. Hierbei konnten Reste der beiden Termini und des zytosolischen

Loops, welche durch Experimente, die auf der dipolaren Kopplung basieren, nicht

detektierbar sind, tentativ zugeordnet werden. Wobei Signale für Arginin und Lysin

eindeutig Arg9 und Lys12 zugeordnet wurden. Insgesamt konnten 84% der Reste,

welche sich in den Transmembran- und Extramembran-Regionen befinden, durch die

angewandte NMR-Strategie zugeordnet werden.

Während des Zuordnungsprozesses wurden keine systematischen Signal-Dubletts

bzw. Tripletts detektiert, was darauf schließen lässt, dass das DGK-Trimer eine

symmetrische Konformation im Apo-Zustand einnimmt. Das steht im Gegensatz zur

Kristallstruktur, welche Asymmetrien zwischen den Untereinheiten aufweist [5]. Die

Abweichungen können auf unterschiedliche experimentelle Bedingungen zurückgeführt

werden. Hierbei scheint vor allem die dichte Packung der Kristalle eine mögliche

Quelle für strukturelle Asymmetrien zu sein.

Die Sekundärstrukturanalyse zeigte substantielle Ähnlichkeiten zwischen Wildtyp-DGK,

seiner thermostabilen Mutante, wobei beide Sekundärstrukturen durch FK-NMR

bestimmt wurden, und der Kristallstruktur von Wildtyp-DGK (PDB 3ZE4). Allerdings

konnten auch einige Unterschiede festgestellt werden. Im Gegensatz zu beiden FK-

NMR-Sekundärstrukturen, zeigt die Kristallstruktur geringe Abweichungen im Bereich

der flexiblen Regionen. Der größte Unterschied tritt im zytoplasmatischen Loop

bezüglich Position und Länge auf. Er ist von den Resten 83-87 in der MAS NMR

Struktur des Wildtyps zu den Resten 83–90 (Untereinheit A), 86-91 (Untereinheit B)

und 82-87 (Untereinheit C) in der Kristallstruktur verschoben.

Neben 13C/15N-detektierten Experimenten bei einer moderaten MAS-Rate, wurden

erste 1H-detektierte Experimente in Kombination mit ultra-schnellem MAS bei 111 kHz

(0.7 mm Rotor) durchgeführt. Die Ergebnisse indizieren eine erfolgsversprechende

Basis für zukünftige Experimente dieser Art an vollständig-protoniertem DGK in

Lipiddoppelschichten.

Auf der Basis der nahezu vollständigen Zuordnung von DGK, wurde der Apo-Zustand

mit den Substrat-gebundenen Zuständen verglichen (Kapitel 5). Es wurden

Änderungen in Peakposition und -intensität der Substrat-gebundenen Zustände für

jeden der 101 zugeordneten Reste in 3D und 2D heteronuklearen Korrelationsspektren

analysiert. Der Nukleotid-gebundene Zustand wurde durch Adenylyl

Methylendiphosphonat (AMP-PCP), einem nicht-hydrolysierbaren ATP-Analogon,

emuliert, während der DAG-gebundene Zustand durch 1,2-Dioctanoyl-sn-glycerol

(DOG, Kettenlänge n=8) imitiert wurde. Zur Ermittlung geeigneter

Zusammenfassung

Sättigungsbedingungen wurde ein kompetitiver Mg*ATP-Inhibierungsassay

durchgeführt, in dem die ATPase Aktivität als Funktion der Mg*AMP-PCP-

Konzentration beobachtet wurde. Außerdem wurde die Bindung von AMP-PCP und

DOG durch 31P-Kreuzpolarisationsexperimente überprüft.

Es konnten sowohl im AMP-PCP- als auch im DOG-gebundenen Zustand eindeutige

Änderungen der chemischen Verschiebung sowie der Peakintensitäten beobachtet

werden. Diese Daten liefern Hinweise darauf, dass alle drei aktiven Stellen gleichzeitig

besetzt sind. Außerdem konnte gezeigt werden, dass das Nukleotidsubstrat eine

weitreichende Konformationsänderung hervorruft. Diese dient sehr wahrscheinlich der

Bindung des Lipidsubstrates an das Enzym und indiziert somit eine positive

Heteroallosterie. Zudem weisen der AMP-PCP+DOG-gebundene und der

ausschließlich AMP-PCP-gebundene Zustand den gleichen spektralen Fingerabdruck

auf. Das deutet daraufhin, dass das Nukleotid das Enzym in einen katalytisch-aktiven

Zustand zu versetzen scheint, welcher die eigentliche Phosphoryltransferreaktion

einleitet.

Die Untersuchung der hohen Stabilität von DGK sowie der Kommunikation zwischen

den Untereinheiten bildet den letzten Teil der Dissertation (Kapitel 6). Dies erforderte

die Identifizierung von entscheidenden Intra- und Interprotomerkontakten, welche eine

strukturelle und funktionelle Bedeutung haben. Hierfür wurden 13C-13C DARR- und 2D

NCOCX-Spektren mit langen Mischzeiten unter Verwendung von Hochfeld-NMR

aufgenommen. Außerdem wurden DNP-verstärkte 13C−15N TEDOR-Experimente

durchgeführt, um Interprotomerkontakte in gemischt-markierten DGK-Trimeren

nachzuweisen. Zur Erzeugung von gemischt-markierten DGK-Trimeren wurde ein

Verfahren etabliert, welches SDS nutzt. Mit Hilfe der angewandten NMR-Strategie

konnten so funktional-relevante Intra (Arg32 - Trp25/ Glu28/ Ala29 and Trp112 -

Ser61)- und Inter (ArgNn,e - AspCg/ GluCd/ AsnCg)-protomerinteraktionen identifiziert

werden. Basierend auf der Kristallstruktur können die Interprotomerkontakte sehr

wahrscheinlich Arg81-Glu88 und Arg92-Asn27/Glu28 zugeordnet werden [5]. Die

identifizierten Interaktionen stablisieren hierbei möglicherweise die aktiven Zentren

und/oder übermitteln Informationen über die Substratbindung bzw. über Änderungen in

der umgebenden Lipiddoppelschicht innerhalb und/oder zwischen den einzelnen

Protomeren.

Table of Contents

1

Table of Contents

1 Introduction ................................................................................ 5

1.1 Diacylglycerol kinases in Eukaryotes and Prokaryotes.......................... 6

1.1.1 Escherichia coli diacylglycerol kinase ............................................................... 9

1.1.1.1 Location and function ................................................................................................ 9

1.1.1.2 Enzymology ............................................................................................................. 10

1.1.1.3 Stability/ folding and misfolding ............................................................................... 12

1.1.1.4 Structure .................................................................................................................. 14

1.1.1.5 Mapping of the active site........................................................................................ 19

1.1.1.6 Proposed catalytic mechanism ............................................................................... 24

1.2 Aim of thesis ............................................................................................. 27

1.3 Solid state nuclear magnetic resonance (ssNMR) spectroscopy ........ 29

1.3.1 Theoretical background .................................................................................. 29

1.3.1.1 Interactions in ssNMR ............................................................................................. 29

1.3.1.1.1 Chemical shift and dipolar coupling ................................................................. 29

1.3.1.2 Magic angle spinning (MAS) ................................................................................... 31

1.3.1.3 Cross polarization .................................................................................................... 32

1.3.1.4 Multidimensional NMR experiments ........................................................................ 34

1.3.2 High-field MAS NMR ...................................................................................... 35

1.3.2.1 Dipolar coupling based experiments for the detection of immobile residues .......... 35

1.3.2.1.1 Homonuclear correlation experiments based on proton driven spin diffusion . 35

1.3.2.1.2 Heteronuclear correlation experiments for the sequential assignment ............ 37

1.3.2.1.2.1 3D NCACX and NCOCX ........................................................................... 37

1.3.2.1.2.2 3D CONCA ................................................................................................ 38

1.3.2.2 Scalar coupling based experiments for the detection and tentative assignment of

highly mobile residues ............................................................................................ 39

1.3.2.2.1 2D 1H-

13C/

15N HETCOR ................................................................................... 39

1.3.2.2.2 2D 13

C-13

C TOBSY ........................................................................................... 40

1.3.3 Dynamic nuclear polarization (DNP)-enhanced MAS NMR ............................. 40

1.3.3.1 13

C-15

N TEDOR experiments for the dectection of interprotomer contacts ............. 42

2 Materials and Methods ............................................................. 45

2.1 Constructs and cells ................................................................................ 45

2.2 Molecular Cloning .................................................................................... 46

2.2.1 Transformation ............................................................................................... 48

2.2.2 Glycerol stocks ............................................................................................... 48

Table of Contents

2

2.3 Protein expression and purification ....................................................... 48

2.3.1 Samples for high field MAS NMR ................................................................... 48

2.3.2 Mixed labelled samples for DNP-enhanced MAS NMR .................................. 51

2.4 Protein reconstitution .............................................................................. 53

2.4.1 Liposome preparation ..................................................................................... 53

2.4.2 Reconstitution via BioBeads ........................................................................... 54

2.5 Sample characterization .......................................................................... 54

2.5.1 SDS-PAGE ..................................................................................................... 54

2.5.2 SEC ................................................................................................................ 55

2.5.3 BN-PAGE ....................................................................................................... 55

2.5.4 LILBID-MS ...................................................................................................... 56

2.5.5 Sucrose density gradient centrifugation .......................................................... 57

2.5.6 Coupled activity assay .................................................................................... 57

2.6 Preparing substrate-bound states of DGK ............................................. 58

2.7 MAS NMR .................................................................................................. 59

2.7.1 MAS NMR at high field ................................................................................... 59

2.7.1.1 Manual resonance assignment ............................................................................... 59

2.7.1.2 Substrate bound states ........................................................................................... 60

2.7.1.3 Data analysis ........................................................................................................... 60

2.7.2 Automatic resonance assignment by ssFLYA ................................................. 61

2.7.3 DNP-enhanced MAS NMR ............................................................................. 63

3 Sample optimization ................................................................. 65

3.1 Introduction .............................................................................................. 65

3.2 Results and Discussion ........................................................................... 66

3.2.1 DGK construct: Quadruple mutant (Δ4-DGK) vs. wild-type DGK .................... 66

3.2.2 Type of detergent: DPC vs. DDM ................................................................... 67

3.2.3 Characterizing the yield, purity and oligomeric state of wtDGK in DDM .......... 68

3.2.4 Reconstitution method: Dialysis vs. BioBeads ................................................ 69

3.2.5 Lipid composition, protein-to-lipid ratio and functional characterization ........... 70

3.2.6 Evaluating the optimized DGK proteoliposomes for MAS NMR application .... 75

3.3 Summary ................................................................................................... 76

4 Resonance assignment ............................................................ 77

4.1 Introduction .............................................................................................. 77

Table of Contents

3

4.1.1 Applied isotope labelling strategy ................................................................... 79

4.1.2 Applied strategy for improvements of the sensitivity and resolution ................ 81

4.1.2.1 High magnetic fields ................................................................................................ 81

4.1.2.2 Paramagnetic doping in combination with an E-free probehead ............................ 81

4.1.3 Applied assignment procedure ....................................................................... 82

4.1.3.1 Sequential assignment of immobile domains .......................................................... 83

4.1.3.1.1 Automatic assignment of immobile domains by ssFLYA ................................. 84

4.2 Results and Discussion ........................................................................... 85

4.2.1 Spectral resolution and isotope labelling ......................................................... 85

4.2.2 Paramagnetic doping in combination with an E-free probehead ..................... 86

4.2.3 Sequential assignment of immobile domains .................................................. 87

4.2.3.1 Automatic assignment of immobile domains by ssFLYA ........................................ 90

4.2.4 Tentative assignment of highly mobile regions ............................................... 92

4.2.5 Summary of the assignment ........................................................................... 94

4.2.6 Secondary structure analysis .......................................................................... 95

4.2.7 DGK forms a symmetric trimer in its apo state ................................................ 96

4.3 Outlook ...................................................................................................... 97

4.3.1 Further labelling strategies ............................................................................. 97

4.3.2 Perspective: 1H detection in combination with ultra-fast MAS ......................... 98

5 Functional studies based on chemical shift perturbations . 103

5.1 Introduction ............................................................................................ 103

5.2 Results .................................................................................................... 104

5.2.1 Establishing nucleotide- and DAG-bound states of DGK for NMR analysis .. 104

5.2.2 DGK forms a symmetric trimer in its substrate bound states ......................... 106

5.2.3 Substrate-induced chemical shift and peak intensity perturbations ............... 107

5.2.3.1 AMP-PCP bound state .......................................................................................... 107

5.2.3.2 DOG bound state .................................................................................................. 109

5.2.3.3 AMP-PCP + DOG bound state .............................................................................. 112

5.3 Discussion .............................................................................................. 112

5.3.1 DGK forms a symmetric trimer in its substrate-bound states ........................ 112

5.3.2 AMP-PCP bound state ................................................................................. 113

5.3.2.1 Comparison of the AMP-PCP bound state with solution NMR data ..................... 114

5.3.3 DOG bound state ......................................................................................... 115

5.4 Summary and Outlook ........................................................................... 116

Table of Contents

4

6 Long-range contacts .............................................................. 117

6.1 Introduction ............................................................................................ 117

6.2 Intraprotomer contacts visualized by high field MAS NMR ................ 118

6.2.1 Results and Discussion ................................................................................ 118

6.3 Interprotomer contacts visualized by DNP-enhanced MAS NMR ...... 119

6.3.1 Introduction .................................................................................................. 119

6.3.2 Results ......................................................................................................... 121

6.3.2.1 Creating mixed labelled trimers of DGK ................................................................ 121

6.3.2.2 Validation of the application of AMUPol as biradical ............................................. 124

6.3.2.3 DNP-enhanced 15

N−13

C TEDOR experiments ...................................................... 125

6.3.2.3.1 Finding the best mixing time using 1D TEDOR spectra ................................. 125

6.3.2.3.2 Visualizing interprotomer contacts using 2D TEDOR spectra ....................... 127

6.3.2.3.3 Attemps to assign the cross-peak by RxA-mutants ....................................... 130

6.3.2.3.4 AMP-PCP bound state of mixed labelled DGK .............................................. 133

6.3.3 Discussion .................................................................................................... 133

6.3.3.1 Creating mixed labelled trimers of DGK ................................................................ 133

6.3.3.2 Statistical analysis of unique interfaces in mixed labelled DGK ........................... 134

6.3.3.3 DNP-enhanced 15

N-13

C TEDOR experiments ....................................................... 136

6.3.3.4 Attemps to assign the cross-peak by RxA-mutants .............................................. 136

6.3.3.4.1 Drawback of mutations ................................................................................... 137

6.3.3.5 Assessing the interprotomer contacts during nucleotide binding .......................... 138

6.4 Summary and Outlook ........................................................................... 139

Appendix .................................................................................... 141

Supplementary tables .................................................................................. 141

List of abbreviations .................................................................................... 157

List of figures ................................................................................................ 160

List of tables ................................................................................................. 172

References ................................................................................. 174

Declaration of contributions ..................................................... 192

Acknowledgements ...................... Fehler! Textmarke nicht definiert.

Curriculum vitae ........................... Fehler! Textmarke nicht definiert.

Chapter 1: Introduction

5

1 Introduction

Membrane proteins represent between 20 to 30% of the genes in most organisms [22,

23]. They are essential for both cellular life and human health. In detail, they are critical

to cell physiology, playing roles in signalling, trafficking, transport, adhesion, and

recognition. Due to their importance, membrane proteins are major targets of

biomedical research. An analysis performed by Overington et al. concluded that

membrane proteins provide more than 60% of drug targets [24]. Thus, drug discovery

efforts aim to understand their biological functions and take advantage of their

therapeutic potential. In order to achieve this goal, molecular structure determination is

necessary. In recent years, structural biology of membrane proteins has progressed

remarkably. X-ray crystallography, electron microscopy (EM) and nuclear magnetic

resonance (NMR) have all contributed fundamental and complementary structural data

[25-28], each with individual advantages and particular challenges. X-ray

crystallography has made considerable contributions to membrane protein structural

biology, with remarkable achievements in the field of G protein coupled receptors

(GPCRs) [29-31]. EM has long been used to investigate the structures of membrane

proteins in proteolipid two-dimensional (2D) crystals [32]. The recent development of

single-particle cryo-EM [33-35] provides higher resolution structures of membrane

proteins without the need of the preparation of large, well-ordered crystalline samples

[26, 36]. NMR has a long history as a key technology in enhancing our understanding

of the structural, chemical, and dynamical characteristics of lipid bilayer membranes.

Early NMR studies offered elementary information about the structures and dynamics

of phospholipid formations, and the impact of membrane proteins and several other

membrane components on lipid bilayers [37-40]. NMR plays also a fundamental role in

membrane protein structural biology, providing methods for the investigation of

membrane proteins in a large variety of samples, including soluble detergent micelles,

detergent-free lipid bilayers, and native cell envelope preparations [7, 28, 41-45]. The

wide range of sample types mirrors the versatility of NMR as a tool for characterizing

the structures, dynamics, and functional interactions of biomolecules. NMR is also

versed at exploring intrinsically disordered regions of proteins [46]. Additionally, based

on the high sensitivity of NMR signals to the local environment, they are extremely

useful for analysing even weak ligand binding by chemical shift changes, providing

structure activity correlations for binding processes or conformational changes [47].

Another considerable advantage of NMR as a technique for structural analysis is that it

enables the investigation of membrane proteins without attenuating its structural

plasticity that is in most cases integral to the biological function. This is contrary to X-


6

ray and single particle cryo-EM studies, which stabilize a single molecular conformation

by the crystallization process itself and/or by the application of cryogenic temperatures.

In addition, they often demand elaborate sample engineering, such as antibody

stabilization and protein mutations, truncations, insertions and modifications [26, 36,

48-52]. Moreover, regarding EM, three-dimensional (3D) reconstruction persists

demanding for proteins smaller than ~100 kDa. Recent progress in NMR structural

studies of membrane proteins reflect large developments in the areas of recombinant

protein expression, sample preparation, pulse sequences for high resolution

spectroscopy, radio-frequency probes, high-field magnets and computational methods.

All enable monitoring single atomic sites in membrane proteins with ever-expanding

accuracy. Thus, far-reaching and detailed information addressing structural and

dynamical changes can be obtained. Solution NMR methods can be used for structure

determination of membrane proteins in detergent micelles or detergent/lipid mixed

micelles [7, 41, 42]. Whereas solid state NMR, in particular MAS (magic angle

spinning) NMR, offers the possibility to explore membrane proteins in detergent-free

lipid bilayers, which brings the investigated system closer to physiological conditions

compared to other membrane mimicking environments such as detergent micelles. The

membrane environment is of key importance as it is a strong structural factor. In most

cases, it is also directly linked to the catalytic activity of the membrane protein [11, 19-

21].

In this study, the structure and dynamics of the membrane protein diacylglycerol kinase

(DGK) embedded into the lipid bilayer is investigated by MAS NMR.

1.1 Diacylglycerol kinases in Eukaryotes and Prokaryotes

Diacylglycerol kinases (DGKs) are members of a conserved family of intracellular lipid

kinases that catalyze the ATP-dependent phosphorylation of diacylglycerol (DAG) to

phosphatic acid (PA) [53, 54]. DAG and PA are intermediates in lipid biosynthetic

pathways and two main signalling molecules. DAG is a common second messenger.

Its cellular levels are increased through hydrolysis of phosphoinositides by

phospholipase C (PLC) in response to a variety of extracellular stimuli including growth

factors and hormones. DAG is most known as an activator of protein kinase C (PKC)

[55], which plays a key role in biological processes like cell proliferation and

differentiation [56]. DAG also interacts with other effector proteins, such as α- and β-

chimaerins (having Rac-GAP activity) [57], guanyl nucleotide-exchange factors for Ras

and Rap, i.e., RasGRP [57-60] and CalDAG-GEFI [60], respectively, as well as


7

nonselective cation channels (TRPC6 and -3) [61]. Thus, DGK regulates the presence

of PKC at the membrane [62], and/or terminates receptor-induced PKC activation and

thereby the signalling pathways downstream of PKC. Moreover, DGK indirectly

regulates small molecular weight G proteins via their nucleotide exchange factors or

GTPase-activating proteins (e.g. chimaerins for Rac), which are activated by DAG. Not

only DGK’s substrate, but also its product, PA, serves as a second messenger. PA

interacts with various target proteins, including Raf-1 kinase [63], PKC-ζ [64],

phosphatidylinositol-4-phosphate-5-kinase [65, 66] and protein tyrosine-phosphatase

[67, 68]. Because of their importance, it is crucial that the intracellular levels of DAG

and PA are tightly regulated for maintenance of normal physiological conditions, which

is accomplished by the diacylglycerol kinases. Multiple forms of DGK are found in most

eukaryotic organisms [69, 70]. In mammalian species, 10 different water soluble

isozymes of DGK have been identified. They are denoted as α, β, γ, δ, ε, z, η, θ, ι as

well as κ, and differ in their biochemical properties, tissue distributions, and lengths

(ranging from 567 aa to >1150 aa) [56, 69]. Based on sequence similarities between

these isozymes and the presence or absence of specific functional domains, they have

been grouped into five different classes [70]. All of these isozymes feature a large

catalytic domain, which is sometimes divided into two parts - catalytic and accessory

domain. Additionally, they have at least two cysteine rich domains (CRDs), which

enable to recruit the protein to the membrane and to bind the lipid substrate, which is

present in the membrane [70]. The simplest and shortest (567 aa) of these isozymes is

DGK-ε, the sole member of class III. It contains only the commonly shared catalytic

domain and two CRDs. In addition, DGK-ε also features a transmembrane helix near

its N-terminal end consisting of 20 - 40 amino acids [71], which most likely plays a role

in membrane interaction [72]. DGKs of class I (α, β and γ isozymes), contain in addition

to the commonly shared domains two EF-hand motifs and a conserved domain of

unknown function near the N-terminal end [70], while DGKs of class II (δ, η and κ

isozymes) feature a plecstrin homology (PH) domain, a sterile α motif (SAM) domain

and a large insert within the catalytic domain separating it into two parts. DGKs of class

IV (z and ι isozymes) are distinguished from the others by a sequence homologous to

the MARCKS phosphorylation site domain and four ankyrin repeats near the C-terminal

end. Lastly, DGK-θ is the sole member of class V. It contains three C1 domains, a Gly-

Pro rich domain and a PH-domain-like region with an overlapping RAS-associating

domain. All these unique domains among the DGK isozymes represent a large

structural diversity, indicating a distinct mechanism of regulation for each isoform.

Recent studies have demonstrated that DGK isozymes play crucial roles in a wide

variety of mammalian signal transduction pathways conducting growth factor/cytokine-


8

dependent cell proliferation and motility, seizure activity, immune responses,

cardiovascular responses and insulin receptor-mediated glucose metabolism [73].

Thus, it is suggested that several DGK isozymes can act as potential drug targets for

cancer, epilepsy, autoimmunity, cardiac hypertrophy, hypertension and type II diabetes

[53, 73, 74].

DGKs are also found in other eukaryotic organisms, such as Drosophila melanogaster

[75], Caenorhabditis elegans [76], Arabidopsis thaliana [77] and yeast [78, 79] as well

as in Gram-positive bacteria encoded by the dgkB gene [80]. While DGKs from all

these different organisms that are typically water soluble do have at least one

conserved catalytic domain, there is a prokaryotic membrane integral DGK encoded by

the dgkA gene, which does not contain any of the canonical sequence features

(catalytic domain and cysteine-rich C1 domains) [56, 81-85]. It does not share

considerable homology with other known kinases and it does not feature any typical

kinase sequence motifs, such as P-loop (Gly-X-X-X-X-Gly-Lys-Thr/Ser) or any other

structural or functional motif. These observations strongly suggest that the prokaryotic

DGK encoded by the dgkA gene is evolutionarily unrelated to the eukaryotic

counterparts, and that the proteins exhibiting DGK activity have evolved independently

in bacteria [56, 70].

The dgkA gene codes for integral membrane DGK present in the Gram-negative

bacterium Escherchia coli (E.coli) and undecaprenol kinase (UDPK) located in Gram-

positive bacteria, such as Staphylococcus aureus, Streptococcus mutants and Bacillus

subtilis [84]. Genes homologous to dgkA are widely spread in the eubacterial domain.

However, they are not omnipresent. Mycobacteria, for example, do not have a dgkA

homolog. In Eukaryota, dgkA-like genes are rarely found. They could only be

determined in the California poplar tree (Populus trichocarpa), the castor oil plant

(Ricinus communis), and the freshwater amoeba (Paulinella chromatophora). There,

the function of the protein, which is encoded by the rare dgkA-like gene, is unknown so

far. Given its shortage in eukaryotes, this gene is likely an evolutionary relict, possibly

emanated from endosymbiosis with cyanobacteria [86-88].

In this study, E.coli diacylglycerol kinase is the object of interest, denoted, hereafter, by

DGK. It is a unique enzyme, which exhibits a high stability [10, 89] as well as a

remarkable complexity in structure and function [5-7]. The high complexity in

combination with a convenient experimental handling has set DGK into focus as a

model system for investigations of membrane protein structure, function and folding

[10, 18, 81, 84, 89-92] as demonstrated below.


9

1.1.1 Escherichia coli diacylglycerol kinase

1.1.1.1 Location and function

DGK is a homotrimeric enzyme located within the inner membrane of E.coli, where it

catalyzes the ATP-dependent phosphorylation of DAG to PA at the

membrane/cytoplasm interface. In 1978, Raetz and Newman mapped the location of

the dgkA gene and used genetic and biochemical experiments demonstrating that

DGK’s major function is the participation in the membrane-derived oligosaccharide

(MDO) cycle [2, 54] as shown in Figure 1. Large quantities of MDOs are produced in

the periplasm of E. coli under conditions of environmental stress such as low

osmolarity [15, 16]. During the MDO biosynthetic pathway, phosphoglycerol is

transferred from phosphatidylglycerol (PG) in the outer leaflet of the plasma membrane

(PM) to the nascent MDOs, generating DAG as a byproduct. DAG is known to be

potentially membrane-disruptive because of its preference for non-bilayer lipid phases.

DGK recycles DAG, after it transverses the PM to reach the inner leaflet, into non-toxic

PA, which is a central intermediate of the glycerophospholipid biosynthesis in bacteria.

DGK thereby provides the basis for restoring PG that is consumed during the MDO

biosynthesis.

Figure 1. Physiological role of E.coli diacylglycerol kinase (DGK) in recycling during the

biosynthesis of membrane-derived oligosaccharides (MDOs) [1, 2] that are largely generated in

response to environmental stress, such as low osmolarity [15, 16]. DGK is located within the


10

inner membrane, where it catalyzes the ATP-dependent phosphorylation of potentially

membrane-disruptive diacylglycerol (DAG) to non-toxic phosphatic acid (PA), providing the

basis for restoring phosphatidylglycerol (PG), which is consumed in the MDO cycle. The cartoon

is based on the X-ray structure using the PDB ID 4UXX [6]. The figure is adapted from Van

Horn et al. [84].

The second major function of DGK is the participation in the lipopolysaccharide (LPS)

biosynthesis, which is the main constituent of the bacterial outer membrane [17]. One

step of this pathway is the transfer of phosphoethanolamine from the outer leaflet of the

PM to lipid A, an LPS biosynthetic intermediate. This process generates DAG as a

byproduct, which is then recycled by DGK [84].

1.1.1.2 Enzymology

DGK belongs to the first integral membrane enzymes, which were subject of detailed

enzyme kinetic studies. First, it was solubilized in Cutscum [93], known to be a harsh

detergent. Over the past 40 years, most kinetic studies on DGK have been performed

in mixed detergent-lipid micelles, e.g. in Triton X-100/ lipid mixed micelles [94]. At that

time, DGK was purified into organic solvents, which is now known to cause protein

unfolding leading to low specific activities (<1 U/mg) [95]. Bell and co-workers

overcame this problem by solubilizing DGK in octylglucoside (OG), using

OG/phospholipid mixed micelles for kinetic experiments, which offered a clearly

increased activity of ~28 U/mg [96, 97]. With these early studies, Bell and co-workers

could demonstrate that DGK requires next to Mg2+ complexed to ATP (MgATP) a free

second Mg2+ ion for activation with the preference for Mg2+ as divalent ATP counterion

[98]. Additionally, it could be shown that the enzyme features an absolute requirement

for a lipid co-factor in micelles [97]. Cardiolipin turned out to be a particularly effective

activator in OG-micelles [97]. Furthermore, Bell and co-workers found out that DGK

does not feature lipid substrate specificity regarding its lipid substrate especially

concerning the fatty acyl chains [97]. This promiscuity for the lipid substrate is in

contrast to the detected specificity with respect to the nucleotide substrate. DGK

strongly favours ATP as the phosphate donor over other nucleotides [18, 99]. It could

be shown that ADP is a very weak phosphoryl donor, whereas adenosine

tetraphosphate features a remarkably reduced affinity relative to either ADP or ATP.

The ribose and adenine moiety specificities of DGK were tested. Measured kcat values

of DGK for guanosine triphosphate (GTP) and inosine triphosphate (ITP) were

decreased significantly compared to ATP, suggesting a purine selectivity.


11

Steady state kinetic studies were carried out by Badola and Sanders [18]. They were

consistent with a random equilibrium mechanism, in which ATP and DAG binding to the

enzyme can occur in both the presence and absence of the other substrate.

Additionally, these data suggest a direct, in-line phosphoryl transfer from ATP to DAG.

This is supported by the fact that no enzyme-phosphate covalent intermediate could be

detected. Furthermore, it could be demonstrated that the bisubstrate analogue

adenosine 5’-tetraphosphoryl-3-O-(1,2-dihexanoyl)-sn-glycerol inhibits the enzyme. It is

a better inhibitor than the bisubstrate itself. In addition, it was shown that DGK features

a higher affinity for ATP than for DAG: The KM was determined to be 1.2 ± 0.5 mM and

5.0 ± 2.2 mol% for ATP and DAG, respectively [18]. The substrate cooperativity factor,

α, was calculated to be 0.48 ± 0.17, implying a modest degree of positive

heteroallostery between the two substrates, which means that the binding of one

substrate enhances the affinity for the other substrate [18].

While DGK’s kcat/KM,ATP is with 104 M-1s-1 modest in comparison to the efficiency of

various water-soluble enzymes, DGK nevertheless seems to be an evolutionarily

optimized biocatalyst in a way that its reaction rate approaches the substrate diffusion-

controlled limit. Under physiological conditions, binding of ATP to DGK is unlikely the

rate-limiting step. Rather, the transbilayer diffusion of DAG appears to limit the rate of

DGK’s reaction in vivo. DAG has to perform a flip-flop through the inner membrane

from its site of production on the outer leaflet to the cytoplasmic site on the inner leaflet,

where DGK’s putative active site is located. However, it cannot be excluded that a

transporter or permease elevates the rate of DAG transbilayer diffusion. Evidence for

this assumption is given with the flip-flop rate of ~50 s-1 measured in lipid vesicles,

which is for a hydrophobic molecule as DAG rather rapid [100, 101]. The flip-flop rate is

remarkably similar to DGK’s kcat of ~26 s-1. Furthermore, DGK’s KM for DAG is with

~5.0 mol% on the same order as the concentration of DAG in the E. coli inner

membrane under conditions of an active MDO cycle [18]. Consequently, DGK seems to

be able to catalyze phosphorylation of DAG under physiological conditions at a rate

that is on the same order as the rate by which DAG can diffuse to its active site. Thus,

DGK was argued to satisfy Knowles’ classic definition of a “perfect enzyme” [102],

since its chemical step seems to be able to keep up with the rate of substrate diffusion

to the active site.

The overexpression of DGK with an N-terminal His6-tag by Bowie and co-workers [9]

provided the facile purification of DGK into either detergent micelles or into several

model membrane systems, such as bicelles or amphipols. DGK featured a maximal

activity of ~110 U/mg in mixed n-decyl-β-maltopyranoside (DM)/phospholipid micelles

at 30°C, which is in the same range as the activity in lipid vesicles of optimal


12

composition. Similar activity was determined for DGK in 3-([3-

cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate (CHAPSO)-1,2-

ditetradecanoyl-sn-glycero-3-phosphocholine (DMPC) or CHAPSO-1,2-

dihexadecanoyl-sn-glycero-3-phosphocholine (DPPC) bicelles [103]. High activity was

also observed for DGK in lyso-myristoyl-phosphatidylcholine (LMPC) detergent

micelles in the absence of lipids [8], in amphipols [104], and in the lipidic cubic

mesophase formed by monoacylglycerols [105]. Kinetic studies demonstrated that

DGK is more active, when it is reconstituted into lipid vesicles composed exclusively of

phosphatidylcholine (PC) than in vesicles that also contain 20 mol% of anionic

phospholipids, or in lipid vesicles emulating the plasma membrane of E. coli [106].

Previous work in this lab applied time-resolved 31P MAS NMR for monitoring

simultaneously ATP hydrolysis taking place in the aqueous phase and DAG

phosphorylation proceeding in the membrane phase [107]. The enzymatic activity of

DGK was investigated with different lipid substrates, 1,2-dibutyrylglycerol (DBG, chain

length n=4, water soluble) and 1,2-dioctanoyl-sn-glycerol (DOG, chain length n=8,

amphiphilic), as well as ATP analogues. It was found out that DGK has a high basal

ATPase activity using DBG as substrate. Only ~70% of all hydrolyzed ATP molecules

were used for DBG phosphorylation. The ATPase activity could be demonstrated to

increase by a rising concentration of DBG, showing a strong, positive cooperativity.

DOG, a lipid substrate with longer acyl chains, was, in contrast, phosphorylated more

efficiently. Furthermore, this study showed that the transition state analogue ADP*VO4,

a complex of ADP with orthovanadate (VO4), decreased the ATP hydrolysis rate to

35% and uncoupled the ATP hydrolysis reaction from the DAG phosphorylation [107].

Using the ATP analogue ATPγS, it was found that DGK is able to transfer the

thiophosphoryl group of ATPγS to DBG, implying a certain plasticity of the active site

[107]. The rate of ATPγS hydrolysis was observed to be faster than of ATP, but no

further stimulation by the addition of lipid substrate could be determined. DGK was

observed to follow a random-equilibrium mechanism, which is in good agreement with

studies in detergent micelles [18].

1.1.1.3 Stability/ folding and misfolding

DGK has been demonstrated to be highly stable in native membranes, being resistant

to irreversible activation upon incubation for a few minutes at 100°C [93, 108]. Even in

detergent micelles, it was observed to be quite stable: In decylmaltoside (DM) micelles

at 70°C and pH 6.5, the t1/2 for irreversible inactivation was determined to be on the


13

order of several hours [8]. This high stability is reflected by a general tolerance of wild-

type DGK to mutations [109]. Lau and Bowie developed a method for quantitating the

thermodynamic stability of DGK under micellar conditions [9]. Folded DGK in DM

micelles was treated with sodium dodecylsulfate (SDS), which is known to be a harsh

denaturing agent. The obtained data offered for the transmembrane domain an

impressive unfolding free energy of 16 kcal/mol and for the cytoplasmic domain

6 kcal/mol, indicating that even the cytosolic domain exhibits a respectable stability [9].

However, it is reported that DGK is much less stable in detergent micelles than in

native membranes [99, 108]. Misfolding of DGK can normally be avoided or corrected,

applying reconstitutive refolding. During this procedure, a detergent is utilized to

solubilize the protein and is then subsequently removed after mixing with lipid-

containing micelles [110]. It could be demonstrated that DGK refolds during the final

steps of detergent removal and vesicle formation. This procedure allows to produce

properly folded single-Cys mutants of DGK, which were applied for kinetic studies to

identify active site residues and to perform disulfide mapping of DGK’s oligomeric

interface [7, 91]. Studies with more than 20 mutants exhibited that the overall rate of

folding and insertion of DGK into lipid vesicles relates to folding efficiency: The risk of

becoming trapped in misfolded states is for mutants higher, which insert and/or fold

slowly [90, 111-113]. A strong correlation between protein stability and folding

efficiency could be demonstrated [90, 113]. Mainly mutations, which cause a

significantly reduced folding efficiency, are mutations that destabilize DGK [90, 113].

Next to these mutants, which are characterized by destabilization promoted-misfolding,

a set of DGK mutants has been observed, which do not seem to be highly destabilized,

but are prone to aggregation and tend to express only at low levels in E. coli. Notably,

these mutations are all present in or near the active site [7].

Booth and co-workers demonstrated that misfolding takes place before association with

vesicles and that it is proposed to cause aggregation [114]. Additionally, it was shown

that irreversible misfolding occurs at the level of monomer [90].

Furthermore, it could be demonstrated that the rate and efficiency of DGK’s association

with 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) bilayers can be modulated by

changing the lipid composition. The folding rate and efficiency was increased in the

presence of anionic 1,2-dioleoyl-sn-glycero-3-phosphoglycerol (DOPG) and decreased

with lyso-OPC [115], most likely induced by lowering membrane curvature stress

and/or the increased lateral pressure in the head group region.

Bowie and co-workers showed that about 10% of the purified mutants possessed a

higher stability in detergent micelles than wild-type DGK [10]. They combined several

mutations into a single DGK mutant, causing partly additivity of the stabilization


14

enhancements detected for the parent single-site mutants [89], leading to a

thermostable quadruple-mutant form of DGK, Δ4-DGK (I53C, I70L, M96L, V107D)

(Figure 2). Δ4-DGK showed a functional half-life at 80°C in OG micelles of 35 min,

which is in contrast to ≪1 minute for the wild type. Δ4-DGK was also used, next to Δ7-

DGK (A41C, C46A, I53C, I70L, M96L, V107D, C113A) (Figure 2), for crystallization

purposes [5, 6]. Both mutants were still fully active after 10 min at 95°C, whereas the

wild type was already inactivated [5].

Figure 2. Sequence alignment of wild-type DGK and the two thermostable mutants, Δ4- and

Δ7-DGK [5]. The mutations in Δ4- and Δ7-DGK are labelled green. The N-terminal tag is

highlighted orange.

1.1.1.4 Structure

With 43 kDa (121 residues per monomer), DGK is the smallest known kinase [84]. In

spite of its small size, it exhibits a notable complexity in structure and function [5-7, 84].

DGK forms a homotrimer [5, 7, 91], in which each monomer contains three

transmembrane helices (H1-3) (Figure 3). H1 is the shortest and preceded by an N-

terminal amphiphilic surface helix (SH) [5, 7]. H2 and H3 extend into the cytoplasm and

are connected by a cytosolic loop (CL), likely to be quite mobile. On the other side of

the membrane, H1 and H2 are linked by a short periplasmic loop (PL).


15

Figure 3. Topology plot of wild-type DGK. The plot was created based on the DGK X-ray

structure [5] and refined by the CSI values obtained from chemical shifts in this study (Table

S5). The membrane is indicated by two solid black lines as calculated in the PPM server [116].

The secondary structure elements of DGK are denoted as: CL, cytoplasmic loop; H1-3, helices

1-3; PL, periplasmic loop and SH, surface helix.

The trimer contains three active sites, which are centred around the

membrane/cytoplasm interface. Each one is a composite of two subunits according to

the shared site model [5, 7, 10].

So far, two 3D structures have been published for DGK shown in Figure 4: one

obtained by solution NMR in dodecylphosphocholine (DPC) micelles [7] and one by 3D

crystallization in lipidic cubic phases (LCP) composed of monoacylglycerols (MAGs),

which also act as lipid substrates [5]. The 3D structures are distinct from other kinases,

including the water soluble DGKs. The solution NMR structure is based on a backbone

assignment of 90% of wild-type DGK and was revealed by a variety of structural

restraints including paramagnetic relaxation enhancement (PRE)-derived long range

distances, residual dipolar coupling (RDC)-based orientational restraints, and distance

restraints from biochemical disulfide mapping [7]. The structure shows a three-fold

symmetry, which is created around a left-handed parallel three-helix bundle built by

helix 2 (H2) from each of the three subunits [7]. This structural core represents the

centre of three overlapping four-helix bundles. Each four-helix bundle involves helices

from all three subunits. The SH, which is N-terminal to H1, is suggested to point away

from the protein core into the bulk solvent. Its first 25 N-terminal residues are observed

to be motional disordered. A key feature of the solution NMR structure is domain

swapping: H3 of one subunit interacts with H1 and H2 from the adjacent subunit in a

domain swapped fashion. The interface between each four-helix bundle is a membrane

immersed cavity that is described by the authors as portico. Each portico is bound by

H1 and H3 (“pillars”) and is covered by the loop linking H2 and H3 (“cornice”). The


16

portico comprises the majority of DGK’s highly conserved residues. Each active site is

formed by H2 of one subunit and H1 and H3 of the adjacent subunit according the

composite shared site model [7]. For the crystal structure, the lipidic cubic phase (LCP,

in meso) method was used [105]. A crystal structure for wild-type DGK and the

thermostable mutants Δ4-DGK (I53C, I70L, M96L, V107D) and Δ7-DGK (A41C, C46A,

I53C, I70L, M96L, V107D, C113A) was obtained with a resolution of 3.70 Å, 3.10 Å and

2.05 Å, respectively [5]. The crystal structure of all three constructs presents a trimer

with layered packing [117]. Δ4-DGK features the most complete model and is used for

the structure description. As the solution NMR structure, the X-ray structure shows an

approximate three-fold symmetry axis that passes through the centre of the trimer

normal to the membrane plane. The core of the trimer is formed by helix 2 (H2) from

each of the three subunits, creating a parallel three-helix bundle. Extending away from

the core, H1 and H3 of each subunit build the sides of an equilateral triangle with H1 at

the apex position. Viewed from the cytosol, the surface helix (SH), angles away from

the trimer core, tangenting H3 from an adjacent subunit. It is suggested to reside on the

cytosolic side of the membrane, anchoring the protein at the membrane interface. Each

active site is formed by the polar/apolar regions of H1–H3 of one subunit and the

surface helix of an adjacent subunit [5], leading to an unusual catalytic site architecture

of the composite shared site model [5]. Accordingly, the crystal structures for DGK

deviate substantially from the solution NMR version. In the solution NMR model, H3 is

domain swapped and contacts H1 of the adjacent subunit. This is based on the cross-

linking data, which indicate that residue 33, located in the cytosolic region of H1, is

linked to the residues 96 and 97 of H3 of another subunit. These two disulfide cross-

linking distance restraints are not compatible with the crystal structure, for which the

pairs of residues are significantly above the cut-off distance of 12 Å. As a result of the

different quaternary structures, the architecture and chemistry of the active sites

deviate notably between these two models. Though they are consistent with the shared

sites model, but for different reasons. The highest discrepancy correlates with domain

swapping, which is present in the solution but not in the crystal structure. This is

accompanied by the fact that the active site is ill-defined for the solution NMR structure,

whereas it is nearly complete for the X-ray structure. This is related to the SH, which is

assumed to be part of the ATP binding site [84]. In the X-ray structure, the SH is nearly

complete and well-defined, at least for two subunits, residing at the cytosolic part of the

membrane. In the solution NMR structure, it is motional disordered over its first 25 N-

terminal residues, pointing into the bulk solvent. If we take both structures as correct,

then the monitored differences occur presumably due to the manner, in which the

proteins are present at the time of structure determination. In one case, the protein is


17

embedded in a DPC micelle at pH 7.8 and 45°C. In the other case, it is incorporated in

the lattice of a LCP crystal composed of MAGs at pH 5.6 and -173°C. Thus, reasons

for these differences could be i.e. the shorter chain length, the very high curvature and

the small hydrophobic thickness of DPC micelles. Additionally, detergents are known to

be rather poor mimics of the lipid bilayer, in which membrane proteins have been

optimized for folding [118]. Thus, the observed motional disorder over the first 25 N-

terminal residues might be a consequence of destabilization of DGK by DPC micelles.

Also the fact that the solution NMR structure is a backbone-only model without side

chain restraints has to be considered. Whereas the structure of DGK in 3D crystals

might be affected by protein-protein crystal contacts and by properties of the lipid cubic

phases.

Figure 4. Comparison of the solution NMR (PDB 2KDC, wild-type DGK) and crystal structure of

DGK (PDB 3ZE5, Δ4-DGK). A view of the crystal (right) and solution NMR (left) structure from

the cytoplasm (a) and the membrane (b) plane.


18

The mentioned discrepancies between the two structures highlight the need for

studying DGK directly in detergent-free lipid bilayers by MAS NMR, which simulate the

physiological environment of the protein to a high extent. So far, a secondary structure

of a thermostable mutant of DGK with multiple mutations embedded into E.coli total

lipids was published by Yang and co-workers using MAS NMR [14]. Figure 5 illustrates

that the overall secondary structure and topology of DGK obtained by ssNMR is in

agreement with the data from solution NMR and X-ray studies. All three structures

show a high α-helical content. However, few notable differences exist, most likely

reflecting the impact of the environment, in which the membrane protein was

embedded (DPC micelles/ LCP consisting of MAGs/ E.coli lipid bilayers). In contrast to

the ssNMR data, the solution NMR structure features two small distortions in Y16

within SH and I70/L70 of H2. In addition, there are small deviations around the

interhelical turn (T) between H1 and SH and around the periplasmic loop (PL) between

H1 and H2 as well as the cytoplasmic loop (CL) between H2 and H3, which are 3, 1

and 6 residues longer in DPC micelles compared to E.coli lipid bilayers. Furthermore,

H/D exchange studies by ssNMR [14] demonstrated that there are residues in the SH

that are not water-accessible, indicating that these residues must be shielded from the

solvent either by close contact with the cytoplasmic surface of the membrane and/or

through close protein–protein contacts. This is contrary to the solution NMR structure,

which suggests that the SH points away from the protein core into the bulk solvent, a

situation in which water accessibility would be expected for all residues of the SH.

Another difference is detected for H2, for which the H/D exchange data suggest that it

is completely membrane embedded, whereas more than 10 residues are solvent

exposed at the cytosolic side in DPC micelles, although this helix has a similar length in

both preparations. One reason could be the shorter chain length and the high curvature

of the DPC micelles (30–40 Å diameters [119], which may interfere with the

approximate diameter of the DGK trimer (~100 Å) [7], which is in contrast to almost

planar lipid bilayers. Additionally, the N-terminal disorder caused by dynamics might be

a consequence of destabilization of DGK by DPC micelles, as mentioned above [118].

Generally, the ssNMR-derived secondary structure agrees better with the crystal

structure (PDB 3ZE5, Δ4-DGK), though the X-ray structure features asymmetries in the

secondary structure between the three subunits concerning the flexible regions, PL and

CL. The length of T, PL and CL is comparable. In the X-ray structure, T is only one

residue shorter. Deviations concerning PL are small as well. It is one residue shorter in

subunit A and C and one residue longer in subunit B. The comparatively highest

difference occurs with respect to the position of CL, which is shifted from residues 81–

85 in the ssNMR structure to residues 87–91 (subunit A), 86–91 (subunit B), and 83–87


19

(subunit C) of the X-ray structure. In addition, fewer residues of H2 and H3 are solvent

exposed in E.coli lipids. The structure and topology of DGK in 3D crystals might be

affected by protein–protein crystal contacts and by the properties of the LCPs. The

influence of the hydrophobic environment on the structure of membrane proteins has

already been extensively investigated by comparing the structures of the single-TM

protein influenza virus AM2 in three different solubilization environments [20, 21, 120-

122]. In terms of DGK, it should be noted that the variations in the secondary structures

in different environments all arise in catalytically critical regions, such as the SH,

cytosolic regions of H2 and H3 and the CL. The conformational plasticity of these

regions in different environments is likely a requirement for the catalytic activity of DGK.

However, further factors, such as mutations, different pH and temperatures may be

additional sources for structure perturbations as well. Summing up, the X-ray structure

of DGK clearly features more similarities to the solid state NMR structure concerning

secondary structure and topology than the solution NMR structure.

Figure 5. Comparison of the DGK secondary structures obtained from solution NMR (PDB

2KDC, wild-type DGK, chain A), solid state NMR [14], and X-ray crystallography (PDB 3ZE5,

Δ4-DGK, chain A): Rectangles symbolize α-helical regions, whereas solid lines reflect

deviations from helicity. Residues that were not resolved are illustrated by dashed lines. The

differences between the secondary structures are highlighted in green. Both the ssNMR and the

X-ray studies used a thermostabilized mutant, whereas the wild type was used for solution NMR

structure determination.

1.1.1.5 Mapping of the active site

So far, several studies were carried out using different techniques to map the active

site. They could demonstrate that the DGK trimer contains three active sites, which are

centred around the membrane/cytoplasm interface. Each one is a composite of two

subunits according the shared site model [5, 7, 10].

Bowie and co-workers identified five likely active-site mutants, such as A14Q, N72S,

E76L, K94L, and D95N (Table 1, Row 1) [10]. Mixtures of either A14Q or E76L with

N72S or K94L were observed to exhibit much greater activity than the mutant proteins

by themselves, indicating that Ala14 and Glu76 may be located on one half of the

active site, while Asn72 and Lys94 are on another half-site.


20

Sanders and co-workers characterized the active site by extensive cysteine scanning

mutational studies (Table 1, Row 2) [7]. Next to these mutational studies, they carried

out solution NMR based titration experiments. Here, wild-type DGK was titrated with

Mg*AMP-PCP or dibutyrylglycerol (DBG) at concentrations of 0 mM, 2 mM, 4 mM, 8

mM and 16 mM. Additionally, DGK was titrated by both substrates (16 mM and 40 mM,

respectively). Backbone chemical shift perturbations (CSPs) were determined as a

function of increased substrate concentration (Table 1, Row 8 and 9). All indications

were found to point to key residues on the cytosolic side of H2 and H3 and to a lesser

extent on H1. The structure and functional studies also illuminate the conclusion of

previous mutagenesis that DGK’s active site must lie at the interface between subunits

[10], involving H1 and H3 from one subunit and H2 from an adjacent subunit [7, 10].

Caffrey and co-workers, mapped the active site by solving the X-ray structure of Δ4-

DGK co-crystallized with the non-hydrolysable ATP analogue, adenylylmethylenedi-

phosphonate (AMP-PCP) (PDB 4UXX) [6]. Here, Δ4-DGK was crystallized in the LCP

composed of MAGs, which double as lipid substrates. Additionally, it was co-

crystallized with AMP-PCP, using a concentration of 10 mM. As divalent counterion,

zinc was used. In the complex, one active site contains two Zn atoms, one AMP-PCP

molecule and two lipid substrates, whereas the other two active sites are nucleotide-

free. They demonstrated that the active site is formed by the polar/apolar regions of

H1–H3 of one subunit and the surface helix of an adjacent subunit [5], leading to a

catalytic site architecture of the composite shared site model [5]. Next to crystallization

studies, Caffrey and co-workers investigated the impact of several residues on the

kinase activity by site-specific mutations monitored both by a coupled enzyme assay

and molecular dynamics simulations (MDS) [6] (Table 1, Row 3). Several residues,

which were identified by mutational studies [6, 7, 10] to have an impact on catalysis,

were found to be proximal (≤ 5 Å) to the nucleotide and lipid substrates in the X-ray

structure (Table 1, Row 4 and 5) [6]. In the following, the observations by Li et al. [6]

concerning possible directly interacting residues are described in detail (Figure 6):

Zn*AMP-PCP was found to be nearly completely stretched and localized by

electrostatic, hydrogen bonding and hydrophobic interactions along its length on the

comparatively flat, cytosol-exposed surface of H2 and H3 of one subunit. The two zinc

atoms bound to AMP-PCP were observed to be coordinated by the phosphates of

AMP-PCP on one side and Glu28 and Glu76 on the other side. The β- and γ-

phosphate groups, in turn, are assumed to be additionally coordinated by Arg9 and

Asn72, respectively. The two hydroxyls of the ribose in AMP-PCP were determined to

be tightly hydrogen bonded to the side-chain carboxyl group of Asp95, which was

found to interact with Gly91 via an H-bond. Additionally, the ribose of AMP-PCP was


21

observed to be hold in place by hydrophobic interactions with the methylenes of Lys94.

Furthermore, the ε-amino group of Lys94 was determined to interact with the N7 of the

purine and with the α-phosphate of AMP-PCP. Lys94 itself was found to be kept in

place by a salt bridge with Asp80. The purine ring of AMP-PCP was observed to be

oriented by hydrogen bonds between N1 and N6 of adenine with His87 and Glu85,

respectively, both located in the CL. The tyrosyl ring of Tyr86, which is present in the

CL between Glu85 and His87, is assumed to cover the adenyl of AMP-PCP. It is

suggested to lock adenine firmly against DGK by π-π stacking. The active site includes

additionally two lipid substrates (MAGs). MAG1 has been modelled into the electron

density map with its headgroup deep in the protein, located at the membrane/cytosol

interface with its reactive 1-OH next to the γ-phosphate of the nucleotide. The two

entities were found to be ~4 Å apart, which is consistent with a direct, in-line

phosphoryl transfer mechanism, excluding the formation of an enzyme phosphate

intermediate with subsequent phosphotransfer to the lipid substrate. This is in

agreement with kinetic and biochemical data mentioned above (1.1.1.2 Enzymology)

[18, 107]. MAG2 was observed to appear in the putative lipid substrate-binding pocket,

with its headgroup ~4 Å from MAG1. Their acyl chains were determined to extend into

the membrane along the hydrophobic surface of the protein. They were monitored to

reside within or close to a three-walled hydrophobic pocket created by the

transmembrane regions of H1-3. The electron density for MAGs in the complex was

observed to be variable and partly discontinuous, reflecting acyl chain flexibility. One of

the two other active sites in the complex, which lack nucleotide, was found to contain

two MAGs orientated similar to those in the active site with Zn*AMP-PCP. The other

active site without nucleotide was observed to contain just one, distant MAG, since its

SH is not visible in density until Ser17, leading to an active site wide open. The 1-OH of

MAG1 was monitored to be close to the side chain carboxyl group of Glu69, whereas

the 2-OH of MAG1 was found to be proximal to Ser98. The carbonyl oxygen of the

ester linkage in MAG1 was detected to be close to Ser17, allowing hydrogen-bonding.

However, it should be noted that the glycerol headgroup of MAG1 could not be oriented

unambiguously in the active site, based on the available electron density maps, making

the contribution of Glu69, Ser98 and Ser17 concerning lipid substrate interaction

assailable.


22

Figure 6. Substrate-binding sites determined in the X-ray structure (Δ4-DGK, PDB 4UXX) [6].

(a) Structure-based and possible interactions with the non-hydrolysable ATP analogue

adenylylmethylenediphosphonate (AMP-PCP, blue) and its two counterions (Zn, orange). The

figure is adapted and modified from Li et al. [6]. (b) Possible interactions of Ser17, Glu69 and

Ser98 with the lipid substrate monoacylglycerol (MAG, yellow).

Data from molecular dynamics (MD) simulations by Jia and co-workers are in good

agreement with the X-ray structure concerning the impact of Arg9, Glu28, Asn72,

Glu76, Lys94 and Asp95 during nucleotide binding and Arg9, Ser17, Glu69 and Ser98

according lipid substrate binding (Table 1, Row 6 and 7) [11]. Additionally, Jia and co-

workers found evidence for Lys12 to form hydrogen bonds with β- and γ-phosphate of

ATP. It could been shown that Lys12 leads to subsequent protonation of the β-

phosphate oxygen. Together with Arg9, it is reported to support the stabilization of the

transition state by dispersion of negative charges on γ-phosphate [11]. Furthermore,

Jia and co-workers could demonstrate that the phosphoryl transfer reaction applies a

dissociative mechanism [11].

Some of the mentioned identified active site residues were found to closely correlate

with the degree of residue conservation observed among homologs [5, 7, 10] (Table 1,

bold residues).


23

Table 1. Mapping of the active site through the identification of functionally relevant residues by

mutational studies, X-ray crystallization, MD simulations and solution NMR

Mutational studies1

X-ray structure (LCPs) [6]

MD simulation (POPE/POPG bilayer) [11]

solution NMR (DPC micelles) [7]

reduced activity when mutated

close proximity to the respective substrate

directly interacting with the respective

substrate

significant backbone CSPs

2

DM micelles

[10]

DPC micelles

[7]

LCPs [6]

AMP-PCP binding

lipid binding (MAG)

ATP binding

lipid binding (DOG)

AMP-PCP binding

lipid binding (DBG)

T8 T8

R9 R9 R9 R9 R9 R9

K12

A13 A13

A14

S17 S17

S17

S17

G20 G20

N27

E28 E28 E28

E28

A30 A30

F31

R32

Q33

Q33

E34 E34

E34

(indirect) E34 E34

G35

D51

I59

S60

V62

M63

V65

M66

E69 E69 E69

E69

E69 E69 E69

I70

N72 N72 N72 N72

N72

N72 N72

S73 S73

I75

E76 E76 E76 E76

E76

D80 D80

D80 (indirect)

D80 (indirect)

R81


24

G83 G83

G83

E85

Y86

H87

L89

S90 S90

G91

(indirect) G91

A93

A93

K94 K94 K94 K94

K94

K94

D95 D95 D95 D95

D95

D95

G97 G97

S98

S98

S98

S98

A99

A99

A100 A100

T112

C113

1 DGK mutants that exhibit ≤25% of wild type activity in the respective environment.

2 CSPs represent only the backbone. They are not determined for the side chains.

bold residues: >99% sequence identity across homologues [6]

italic residues: A direct interaction with the respective substrate is possible due to the X-ray

structure, but is not mentioned anywhere.

indirect: structure-based interaction with a catalytically important residues (D80->K94, G91-

>D95)

yellow labelling: All (dark yellow) or most (light yellow) of the mentioned studies agree

concerning the catalytic impact of this residue

1.1.1.6 Proposed catalytic mechanism

Based on the X-ray complex (PDB 4UXX) mentioned above, Caffrey and co-workers

suggested that Glu69 initiates the reaction by abstracting a proton from the primary

hydroxyl, since its side chain carboxyl is in hydrogen-bonding distance to the 1-OH of

MAG1, [6]. This is consistent with MD simulation (MDS) studies by Jia and co-workers

[11]. Additionally, a structure of nucleotide-free Δ7-DGK was obtained by serial

femtosecond crystallography (SFX) with an X-ray free electron laser (XFEL) at room

temperature (RT), monitoring two conformations for the following three critical residues:

Glu34, Glu69 and Glu76. In MD simulations, the two alternative conformers of the side

chains were observed to convert into one another by protonation or deprotonation of

the side chain carboxyl group [6]. Based on these observations from the X-ray

structures in combination with mutational studies [6, 7, 10] and MD simulations [6],

Caffrey and co-workers proposed following catalytic mechanism for DGK [6]:


25

In the nucleotide-free state, the catalytic site is hydrated. Glu34 is predicted to have a

pKa of 7.53 and is, thus, present in its protonated state at a neutral pH, forming a

hydrogen bond with the deprotonated Glu69. This was found to be a stable

configuration in the MD simulations. The coordination of the substrates to the active

site causes then a conformational change of Glu34, which allows its deprotonation by

water. This induces destabilizing electrostatic interactions between deprotonated Glu34

and Glu69. Due to the hydrophobic nature of the catalytic site, deprotonation of Glu34

will increase the pKa of Glu69, making it a stronger base for proton abstraction of the

lipid substrate. Thus, Caffrey and co-workers hypothesised that deprotonation of Glu34

promotes the deprotonation of the lipid substrate by Glu69, leading to a formation of a

nucleophilic alkoxide. The alkoxide, in turn, induces a nucleophilic attack on the γ-

phosphate of the nucleotide, which causes the formation of a pentahedral intermediate,

stabilized by Asn72 and/or Arg9. MD simulations could show that protonation of Glu69

induces a side-chain switch. This alternate conformation was found to extend deeper

into the membrane and to have a predicted pKa value of 8.82. Additionally, protonated

Glu69 presumably interacts with deprotonated Glu34 via a stabilizing hydrogen-bond.

Thus, the hydrogen-bonding is now inversed compared to the nucleotide-free state.

After the pentahedral intermediate is collapsed, which causes a breaking of the β-γ

linkage, the ADP and phosphorylated lipid substrate are formed.

This is followed by the diffusion of the products from the active site. The release of the

phosphorylated lipid, which is relatively bulky and negatively charged, from the active

site is expected to occur via the opening between SH and H1. This may be facilitated

electrostatically by Glu69 and Glu76 in H2, creating a push, and by Arg9 and Lys12 in

the SH, inducing a pull. In addition, Glu69 is deprotonated by Glu34. After product

release, the active site is reset for another round of catalysis: Water molecules diffuse

into the active site, hydrating it and Glu34 returns into its nucleotide-free state

conformation. Density functional theory (DFT) simulations on MDS-sampled

configurations, monitoring close MAG1 contacts, suggest that 1-OH proton abstraction

by Glu69 is kinetically more likely than phosphate cleavage.


27

1.2 Aim of thesis

Despite many years of research including solution NMR and X-ray crystallization

studies, important long-standing questions regarding DGK’s catalytic mechanism

remain unsolved. It is not clear yet, if the DGK trimer adopts a symmetric or

asymmetric conformation. Additionally, it is unknown whether the three active sites of

DGK are in same or different states during catalysis and whether DGK undergoes a

substantial conformational change prior to the actual phosphoryl transfer. Taking into

account that the DGK trimer exhibits a remarkable stability and that each active site is

built by components of two protomers based on the composite shared site model, the

question arises whether specific long-range intra- and interprotomer interactions exist.

In this study, these questions will be addressed by multidimensional high field as well

as dynamic nuclear polarization (DNP)-enhanced 13C,15N MAS NMR.

Preparing membrane protein samples for MAS NMR is still a challenge. In order to

answer these questions by MAS NMR defined above, it is required to achieve high

amounts of pure, active, stable and homogeneously reconstituted protein, which

provides well-resolved MAS NMR spectra. This task will be accomplished by a

stepwise optimization of the preparation protocol, defining the DGK construct,

detergent, lipid composition, reconstitution method and protein-to-lipid ratio (Chapter

3).

For the investigation of structural and dynamical changes defining the catalytic

mechanism of a protein, the assignment of its backbone and side chains is mandatory.

However, the assignment of membrane proteins by MAS NMR is still a highly

demanding task, challenging from sample preparation over experiment planning and

performance until data analysis. Thus, a NMR strategy has to be established, which

helps to diminish obstacles and enables a nearly complete assignment of wild-type

DGK in lipid bilayers (chapter 4). In detail, an appropriate isotope labelling strategy and

assignment procedure have to be defined. A strategy for improvements of the

sensitivity and resolution has to be found and an automatic assignment algorithm will

be tested to make the assignment faster and more reliable.

In order to highlight changes in structure and dynamics within the catalytic hotspots of

DGK, the apo state of DGK will be compared with the substrate bound states. On the

basis of the assignment, perturbations in peak position and intensity of the substrate

bound states have to be analysed in 3D and 2D heteronuclear correlation spectra.


28

Therefore, nucleotide and DAG-bound states of DGK have to be established for NMR

analysis (chapter 5).

In order to understand DGK’s remarkable stability and the cross-talk between its

subunits, key intra- and interprotomer contacts have to be identified. For the detection

of intraprotomer contacts, 13C-13C DARR and 2D NCOCX spectra with long mixing

times will be recorded using high field MAS NMR. For the detection of interprotomer

contacts, mixed labelled trimers have to be created, which contain 13C and 15N residue

contacts between adjacent protomers. These could then be visualized by DNP-

enhanced TEDOR experiments. For this purpose, a procedure has to be established

for the production of mixed labelled DGK trimers. Firstly, conditions have to be tested,

in which DGK trimers can be disassembled into monomers or dimers. If this is

accomplished, the next step will be to initiate a reassembling of the monomers back to

active trimers. The third step will be to find conditions for a sufficient signal

enhancement for DNP-enhanced TEDOR experiments, allowing the identification of

interprotomer contacts. Finally, the detected contacts have to be assigned. In addition,

it has to be found out, if these contacts are involved in nucleotide binding (chapter 6).

If this schedule is successful, not only a deeper understanding of DGK’s mechanism

can be obtained, it would also lead to an optimized and reproducible protocol for the

sample preparation for high field and DNP-based NMR experiments. Additionally, the

assignment of DGK would provide a valuable basis for all kinds of experiments in

future. All in all, this would be highly advantageous for subsequent research on DGK.


29

1.3 Solid state nuclear magnetic resonance (ssNMR) spectroscopy

1.3.1 Theoretical background

1.3.1.1 Interactions in ssNMR

Each nuclear spin interacts with its environment. These interactions are specific and

provide information about the local structure and dynamic of a membrane protein. They

can be provoked by external and internal fields. Internal interactions are shielding

interactions (chemical shift, CS), dipolar couplings (D), J-couplings (J) that are also

referred to as scalar couplings, and quadrupolar interactions (Q). The intrinsic

Hamiltonian (Hint) is defined as a sum of these interactions:

�̂�𝑖𝑛𝑡 = �̂�𝐶𝑆 + �̂�𝐷 + �̂�𝐽 + �̂�𝑄 (1)

While J-couplings are isotropic, quadrupolar interactions, chemical shielding and

dipolar couplings are anisotropic. Quadrupolar interactions occur only in nuclei with

spin quantum numbers >½. Since only nuclei featuring a spin quantum number of ½

are of interest in this study, quadrupolar interactions are not further discussed.

Chemical shielding and dipolar couplings are described in more detail in the following.

1.3.1.1.1 Chemical shift and dipolar coupling

The chemical shift characterizes the local, magnetic micro environment of a nuclear

spin. It is affected by diamagnetic and paramagnetic influences of the electron shell,

ring currents in aromatic residues, the solvent effect e.g. from surrounding buffers or

lipid bilayers, and distinct couplings between nucleus and electron within the molecule.

Soluble molecules tumble isotropically, leading to an isotropic chemical shift (δiso),

whereas membrane proteins in lipid bilayers are constrained in their mobility. Thus,

they usually adopt different orientations with respect to the magnetic field B0. This

results in a so-called powder spectrum. The dependence of the chemical shift on the

orientation of the molecule is defined as chemical shift anisotropy (CSA), which is

determined by a second rank tensor with three parameters: δxx, δyy, δzz.

The isotropic chemical shift (δiso) can be obtained as the mean value of these three

parameters:

𝛿𝑖𝑠𝑜 =

1

3 (𝛿𝑥𝑥 + 𝛿𝑦𝑦 + 𝛿𝑧𝑧) (2)


30

|𝛿𝑍𝑍 − 𝛿𝑖𝑠𝑜| ≥ |𝛿𝑋𝑋 − 𝛿𝑖𝑠𝑜| ≥ |𝛿𝑌𝑌 − 𝛿𝑖𝑠𝑜| (3)

Whereas, the anisotropic chemical shift (δaniso) characterizes the width of the

anisotropic peak:

𝛿𝑎𝑛𝑖𝑠𝑜 = 𝛿𝑍𝑍 − 𝛿𝑖𝑠𝑜 (4)

Additionally, the asymmetry of the chemical shift (η) describes the shape of the peak or

the symmetry of a spectrum:

𝜂 = 𝛿𝑌𝑌 − 𝛿𝑋𝑋

𝛿𝑎𝑛𝑖𝑠𝑜 (5)

The asymmetry factor (η) leads to a value between 0 and 1. In the case of an axial

symmetry, the value is zero. In the case of asymmetry, the value is one. In a static

sample, η ≠ 0 and anisotropic interactions are prevalent. By rotating the sample around

an axis with a rate larger that δaniso, the anisotropic interactions are cancelled and the

asymmetry parameter η = 0. The chemical shift δ is characterized by isotropic and

anisotropic contributions:

𝛿(𝛼, 𝛽) = 𝛿𝑖𝑠𝑜 +1

2 𝛿𝑎𝑛𝑖𝑠𝑜(3𝑐𝑜𝑠2𝛽 − 1 − 𝜂𝑠𝑖𝑛2𝛽𝑐𝑜𝑠2𝛼) (6)

It is dependent from the Euler angles α and β, describing the orientation towards B0.

The Hamiltonian for the chemical shift can be defined as:

�̂�𝐶𝑆 = �̂�𝛿𝑖𝑠𝑜+ �̂�𝛿𝑎𝑛𝑖𝑠𝑜

(7)

Homo- and heteronuclear dipolar coupling also add to peak broadening. The

Hamiltonians for the homo (equation 8, 10)- and heteronuclear (equation 9, 10) dipolar

couplings can be written as:

�̂�𝐼𝑆 = 𝑑𝐼𝑆 ∙ (3 ∙ 𝐼𝑍𝑆𝑍 − 𝐼 ∙ 𝑆) (8)

�̂�𝐼𝑆 = 𝑑𝐼𝑆 ∙ 2 ∙ 𝐼𝑍 ∙ 𝑆𝑍 (9)

with the dipolar coupling constant (dIS) defined as:

𝑑𝐼𝑆(𝑟𝐼𝑆, 𝛽𝐼𝑆) = −𝛾𝐼𝛾𝑆𝜇0ħ

8𝜋2𝑟𝐼𝑆3 ∙ (3𝑐𝑜𝑠2𝛽𝐼𝑆 − 1) (10)


31

The dipolar coupling constant (dIS) characterizes the direct interaction through space

between two spins. It is determined by the distance between those spins (r-3) as well as

the angle of their internuclear vector and the magnetic field (βIS). γI and γS are the

gyromagnetic ratios of the coupling spins. Dipolar couplings occur in ssNMR powder

spectra with the so-called PAKE pattern. The two signals correspond to energy

differences, depending on the parallel or antiparallel alignment of the I spin with respect

to the S spin. The IzSz term in equation 9 gives positive (parallel spins) and negative

(antiparallel spins) energies. The two maxima of the PAKE doublet relates to the

situation when the internuclear vector is perpendicular to the magnetic field (βIS = 90°).

This situation is the most frequent one, which explains the higher intensity. The

distance between the two maxima of the PAKE doublet correlates with the dipolar

coupling constant (dIS). The "feet" of the PAKE doublet correspond to the situation

when the internuclear vector is parallel to the magnetic field. There is only one possible

orientation of the dipolar vector, explaining the lower intensity (βIS = 0°).

1.3.1.2 Magic angle spinning (MAS)

Chemical shift anisotropy (CSA) and dipolar couplings cause severe line broadening,

leading to substantial signal overlap and thus to ambiguous data. In order to obtain

well-resolved spectra with a reasonable small line width, one makes use of the term

3cos2β-1, of which both interactions depend on. In detail, the sample is spun around

the angle β = 54,74° with respect to the external magnetic field B0. According equation

6 and 10, respectively, CSA becomes zero and dipolar couplings disappear.

Additionally, the asymmetry parameter η in the term ηsin2βcos2α is cancelled out by

sample rotation. This method is called magic angle spinning (MAS) (Figure 7). In order

to reduce dipolar coupling effectively by MAS, the spinning rate of the sample has to be

faster than the coupling between the spins. Slow rates generate visible spinning

sidebands, which appear in addition to the isotropic signal with the distance of the

spinning frequency.


32

Figure 7. Impact of magic angle spinning (MAS) at 54.74° on solid state NMR spectra. (a)

Depiction of a MAS rotor that is tilted in the magic angle β = 54.74° with respect to the magnetic

field B0. 15

N (b)- and 1H (c)-NMR spectra of the microcrystalline tri-peptide N-formyl-Met-Leu-

Phe-OH under static (blue) and MAS (red) conditions. In the static 15

N-NMR spectrum, the

isotropic (δiso) and the anisotropic (δaniso) chemical shift as well as the CSA parameters δxx, δyy

and δzz are labelled accordingly. Comparing the 15

N- and 1H-NMR spectra under MAS of

25 kHz, it becomes obvious that higher spinning speeds are needed to obtain well-resolved 1H-

NMR spectra (see chapter 4, outlook), whereas the 15

N-NMR spectrum features already a good

resolution at 25 kHz. The figures are adapted from the lecture script “Solid state NMR”,

prepared by Prof. Clemens Glaubitz, Goethe University Frankfurt am Main, summer semester

2015.

In addition to MAS, proton dipolar couplings can be suppressed by applying decoupling

sequences during mixing and acquisition time, which all together cause well-resolved

spectra.

For some studies, information about the distance of the nuclei and/or the bond angle

are relevant, which can be extracted from the dipolar coupling constant. In a rotational

resonance experiment, the spinning frequency relates to a multiple of the isotropic

chemical shift difference between the interacting nuclei. This way, the coupling is

restored for one particular spin interaction.

1.3.1.3 Cross polarization

Solution NMR is typically based on the detection of 1H nuclei. They offer a high

detection sensitivity due to a natural abundance of more than 99.9% and a high

gyromagnetic ratio. In MAS ssNMR, the linewidths of protons remain broad at

moderate MAS frequencies (10–20 kHz), because of the strong inter-proton dipolar

couplings. Hence, 13C and 15N nuclei are usually detected. They have smaller

gyromagnetic ratios than protons, leading to weaker bulk magnetizations and thus to

weaker signal intensities. To elevate the signals from these nuclei, cross polarization

(CP) is applied according to the Hartmann-Hahn condition, in which magnetization is


33

transferred from highly abundant protons (I spin) to dilute 13C or 15N (S spins), when

both spin systems are in contact. The exchange of magnetization is induced by the

simultaneous application of two continuous radiofrequency (RF) fields, B1, on the

respective resonance frequencies of I and S spin. The nutation frequency (ω1) of the

two spin systems (I, S) depends on their gyromagnetic ratio (γ) as well as the strength

of the RF field (B1).

𝜔1𝐼 = 𝛾𝐼𝐵1

𝐼 or 𝜔1𝑆 = 𝛾𝑆𝐵1

𝑆 (11)

If the nutation frequency (ω1) is equal for the I and S spin, a dipolar interaction between

the two spin systems is introduced, allowing the polarization transfer:

Static 𝜔1𝐼 = 𝜔1

𝑆 (12)

𝛾𝐼𝐵1𝐼 = 𝛾𝑆𝐵1

𝑆 (13)

MAS 𝜔1𝐼 = 𝜔1

𝑆 ± 𝜂𝜔𝑟𝑜𝑡 (14)

The CP pulse sequence (Figure 8) starts with a 1H 90°pulse that tilts the proton

magnetization from the z-axis into the xy-plane. Then, a spin lock field on both nuclei

channels is applied with a distinct contact time (tCP). During the CP time, typically

between 100 μs and 10 ms, the magnetization of the X nuclei increases, depending on

the strength of the dipolar coupling between 1H and X. The magnetization builds up

until a steady-state equilibrium is reached. At the same time, the magnetization at 1H

and X decreases due to spin-lattice relaxation in the rotating frame. The signal

enhancement under CP conditions is limited by the quotient of the gyromagnetic ratio

of the two spin systems (e.g. γ1H/γ13C ~ 4). After the CP step, the FID of the X spins is

detected, while proton decoupling is applied. Although CP experiments are performed

under MAS, strong 1H-X dipolar couplings cannot be completely removed. Therefore,

heteronuclear decoupling is used to exclude the couplings during signal acquisition.

This is carried out by either continuous RF irradiation in the proton channel (continuous

wave (CW) decoupling) or by decoupling pulse sequences, such as the two pulse

phase modulation (TPPM).


34

Figure 8. Pulse sequence of a cross polarization (CP) experiment according the Hartmann-

Hahn condition, in which magnetization is transferred from highly abundant I spins to dilute S

spins.

1.3.1.4 Multidimensional NMR experiments

For large and complex systems as membrane proteins, 1D spectra are usually not

sufficient to resolve unique signals due to high spectral crowding. This obstacle can be

resolved by the application of multidimensional spectra. With 2D and 3D NMR

experiments, further information about spin correlations and a better resolution of

signals in the additional spectral dimension(s) can be obtained. The basic 2D

experiment consists of a preparation, evolution (t1), mixing and detection period (t2).

During the preparation period, the spins are tilted into the xy plane by 90° pulses or a

cross polarization step, generating transverse magnetization, which then evolves

during the evolution period (t1). During this time, relaxation takes place under the

influence of various nuclear spin interactions. Thereafter, the mixing period follows,

during which a second pulse flips the y component onto z axis. The idea of the mixing

step is to allow spin communication (e.g. via cross-relaxation, exchange or spin

couplings) for a fixed period, under the application of specific pulse sequences. Then,

the magnetization is prepared for detection. During the detection period, the signal is

recorded as a free induction decay or FID at regular time intervals as a function of t2

(S(t)).

In 2D NMR spectroscopy, the signal is detected as a function of two time increments, t1

and t2. The resulting data are Fourier transformed twice to obtain a spectrum, which is

a function of two frequency variables. In detail, the t1 time is elongated successively by

a distinct increment from zero to an upper limit. The signal amplitude, which is recorded

during t2, depends on the t1 time, oscillating as a cosine function. This oscillation of the

signal amplitude in the indirect dimension provides a second FID. Thus, 2D

experiments are assembled 1D experiments that evolve into the second dimension via


35

incrementing the t1 time. A drawback of multidimensional NMR experiments is the

extended experimental time. Data as a function of t1 is time consuming to record, since

for each t1-increment the whole pulse sequence has to be executed.

1.3.2 High-field MAS NMR

1.3.2.1 Dipolar coupling based experiments for the detection of immobile

residues

Immobile domains, usually transmembrane regions, are molecular segments with

smaller amplitude motions. They are not sufficient to fully average anisotropic

interactions (such as HH, HC, or HN dipole couplings) and can be observed by dipolar

coupling based cross-polarization (CP) as described above.

1.3.2.1.1 Homonuclear correlation experiments based on proton driven spin

diffusion

One conventional 2D experiment in solid state NMR is based on proton driven spin

diffusion (PDSD). It is typically conducted to evaluate structural homogeneity,

resolution and secondary structure of the protein sample (as shown in chapter 3). Spin

diffusion is a homonuclear transfer of magnetization between 13C- or 15N-nuclei through

the spin network of heteronuclear coupled protons. In this study, only a 13C-based

PDSD was carried out. The pulse sequence is shown in Figure 9. The first step, the

preparation period, involves a CP transfer from 1H to 13C. Thus, a magnetization on the

13C nuclei is generated in the x,y-plane. During the evolution period, t1, the 13C spins

evolve under the influence of the Hamilton operator of the chemical shift, while proton

decoupling is applied. During this time, no transfer between spins occurs. The evolution

period is terminated with a 90° pulse on the 13C spins, rotating them at the z-axis. This

is followed by a mixing step, in which the spin diffusion between 13C nuclei takes place

through space by flip-flop interactions. The mixing time typically lasts from 10 ms to 3 s.

During this time, the protons are not decoupled. The exchange through spin diffusion is

principally related to the distance between two spins by a factor of r-6. Thus, the length

of the mixing time defines the distance of the magnetization transfer. With short mixing

times (10 - 20 ms), the magnetization is mostly transferred between neighbouring

atoms, causing peaks within one residue. Longer mixing times (100 ms – 3 s) enable

more distant interactions, leading to inter-residue connectivities. After the mixing, a

second 90° pulse turns the magnetization from the z-axis into the transversal x,y-plane,


36

where it is detected. During acquisition the heteronuclear decoupling is applied. Since

the PDSD experiment is homonuclear, a diagonal crossing the 2D spectrum is

observable. The diagonal accords to the 1D spectrum. All off-diagonal peaks are called

cross peaks and correspond to the chemical shifts of those nuclei that undergo spin

diffusion.

Figure 9. Pulse sequence of a 2D 13

C-13

C PDSD experiment. During the preparation time, the

magnetization is transferred from 1H to

13C via CP step. This is followed by the t1 period, when

the 13

C chemical shift evolves, while 1H nuclei are decoupled. Thereafter, the magnetization is

transferred back on the z-axis through a 90° pulse on the 13

C spins and the mixing step takes

place, in which the proton driven spin diffusion between 13

C nuclei occur through space by flip-

flop interactions. For detection, the 13

C spins are transferred back in the x,y-plane and the FID is

recorded under 1H decoupling.

In a PDSD experiment, the proton decoupling is switched off during the mixing time to

reintroduce the dipolar coupling between 13C and 1H nuclei. Through homogenous line

broadening based on dipolar 1H-13C and 1H-1H coupling, a 13C spin pair can overlap

with different chemical shifts, allowing proton driven spin diffusion between the 13C

nuclei. However, at higher MAS rates, the heteronuclear dipolar 1H-13C coupling is

scaled down. Dipolar assisted rotational resonance (DARR) can be applied, in which

the 1H-13C dipolar coupling is recovered by CW irradiation on 1H, whereupon the 1H RF

field intensity satisfies the condition of rotary-resonance: ω1H = ωMAS. Applying this

experiment with long mixing times, higher magnetic fields and/or higher MAS rates

enables a more efficient magnetization transfer. This in turn leads to stronger cross

peaks as compared to a PDSD spectrum.


37

1.3.2.1.2 Heteronuclear correlation experiments for the sequential assignment

The sequential assignment process includes spin system identification, assignment of

the spin system to the amino acid type, linking of the spin systems, and mapping them

to the protein amino acid sequence. This is achieved by the analysis of three dipolar

coupling based heteronuclear 3D correlation experiments, NCACX, NCOCX, and

CONCA [123, 124]. Here, C-N connections are needed. They are gained by a

heteronuclear band-selective cross-polarization (CP) transfer between the 15N amide

and the 13Cα (NCA transfer) or the 13CO (NCO/CON transfer). The NCA transfer is

intra-residue, since it links 15N[i] with 13C[i], whereas the NCO/CON transfer is inter-

residue, as it connects the 15N[i] with 13CO[i-1] through the peptide bond.

1.3.2.1.2.1 3D NCACX and NCOCX

In the first CP step, the magnetization is transferred from 1H[i] to 15N[i] under the

Hartmann-Hahn matching condition. Then, the evolution of the 15N[i] chemical shift

takes place. Afterwards, the magnetization is selectively transferred to 13Cα[i] or 13CO[i-

1] during a specific second CP step. The 15N/13C CP steps are very sensitive to the

matching conditions for the spin lock fields, reliant on RF field amplitude, chemical shift

and MAS frequency [125]. The matching condition is gained with the MAS rate ωr:

ωI,eff – ωS,eff = n * ωr I = 13C, S = 15N (15)

The 15N frequency is set on resonance to the amide backbone (~ 120 ppm) and the 15N

RF amplitude (~ 2.5 * ωr) is chosen accordingly. Concurrently, the 13C frequency is set

for the favoured transfer to Cα (~ 55 ppm) or CO (~ 170 ppm) and the matching

condition is calculated with respect to the 13C amplitude (~ 1.5 * ωr). For the 15N-13C

transfer, spin lock fields are optimized to improve the efficiency and to compensate RF

inhomogeneity. 13C, 15N amplitudes, frequencies, CP spin lock fields and decoupling

are optimized in incremental steps to obtain the best conditions for each sample. After

the double cross polarization (DCP), evolution on 13Cα/13CO takes place, which is

followed by a non-specific 13C-13C transfer step that transfers the magnetization to any

other proximate 13C nuclei (CX) through a DARR period [126, 127]. Longer DARR

mixing times enable more distant interactions, leading to inter-residue connectivities,

which can be useful in the assignment process. Summing up, the chemical shift is

evolved on 15N and 13Cα/13CO and then detected on 13Cx, leading to a 3D spectrum.

The pulse sequence of the NCACX and NCOCX experiment and the respective

polarization transfer pathways are illustrated in Figure 10.


38

Figure 10. (a) Pulse sequence of the NCACX and NCOCX experiment. Both are 15

N-13

C

correlation transfer experiments with a subsequent 13

C-13

C mixing step. During the preparation

period, a broad-band 1H

15N-CP step is used to generate

15N polarization that evolves during t1

under proton decoupling. For the 15

N-13

C transfer, optimized spin lock fields on the 15

N and 13

C

channel are applied under proton decoupling. The 13

C off-set is centered in the Cα region for

NCA and in the CO region for NCO. After the double cross polarization (DCP), evolution on 13

Cα/13

CO takes place under proton decoupling during t2. Subsequently, a DARR step follows,

which transfers the magnetization to any other proximate 13

C nuclei. Therefore, two 90° pulses

at an off-set of 100 ppm were applied for excitation and reconversion of longitudinal

magnetization. The detection of 13

C magnetization (t3) represents the final step, during which

the protons are decoupled. (b) Polarization transfer pathway for the NCACX (left) and NCOCX

(right) pulse sequence, schematically illustrated for a di-peptide. The selected off-set

frequencies on Cα or CO enable a magnetization transfer within the same amino acid [i] along

the side chain, resulting in cross peaks in the NCACX spectrum, or along the side chain of the

previous amino acid [i-1], leading to cross peaks in the NCOCX spectrum, respectively (black

arrows). Additional through-space dipolar-assisted pathways are possible as well (grey arrows).

1.3.2.1.2.2 3D CONCA

Magnetization is generated on 1H and transferred to 13C via the first CP step. 13C

chemical shift evolves during t1. Band selective 13CO[i-1] to 15N[i] is obtained via spin

lock on 15N and ramping through the specific CP matching on the 13C channel. The

magnetization then evolves on 15N during t2, followed by a transfer to 13Cα[i] and


39

detection. The pulse sequence of the CONCA experiment and the respective

polarization transfer pathway are shown in Figure 11.

Figure 11. (a) Pulse sequence of the CONCA experiment. It is a 15

N-13

C correlation transfer

experiment accomplished by three CP steps. During the preparation period, a selective 1H

13C-

CP step is used to generate 13

C polarization that evolves during t1 under proton decoupling.

This is followed by a second selective CP step, which transfers magnetization from 13

CO[i-1] to 15

N[i]. The 13

C off-set is centered in the CO region for CON. The magnetization then evolves on 15

N[i] during t2 under proton decoupling. Subsequently, the third CP step takes place,

transferring magnetization from 15

N[i] to 13

Cα[i]. The 13

C off-set is centered in the Cα region for

NCA. The detection of 13

C magnetization (t3) represents the final step, during which the protons

are decoupled. (b) Polarization transfer pathway for the CONCA pulse sequence, schematically

illustrated for a di-peptide. The selected off-set frequencies on CO and later Cα enable a

magnetization transfer from CO of the previous amino acid [i-1] via N towards Cα (black arrow)

and sometimes Cβ (grey arrow) of the following amino acid [i], leading to cross peaks in the

CONCA spectrum (black arrows).

1.3.2.2 Scalar coupling based experiments for the detection and tentative

assignment of highly mobile residues

Highly mobile regions, usually termini or loops, feature fast and large amplitude

fluctuations on time scales of <10-5 s. They can be selected by a refocused INEPT

(insensitive nuclei enhanced by polarization transfer) [128] step based on scalar

couplings, leading to solution state-like spectra due to efficient molecular averaging of

anisotropic interactions.

1.3.2.2.1 2D 1H-13C/15N HETCOR

The magnetization is transferred from 1H to attached 13C or 15N nuclei through a

refocused INEPT (insensitive nuclei enhanced by polarization transfer) step [128]


40

based on J-couplings. After the initial 90º 1H pulse, 1H chemical shift evolution during

the variable t1 period takes place. The evolution delay is fixed to achieve antiphase 1H

magnetization with respect to 13C via JHC (JHC ~200 Hz) or 15N via (JHN ~90 Hz). The

magnetization is transferred to 13C/15N by applying simultaneous 90º 1H and 13C/15N

pulses. The 13C/15N chemical shift evolution during the variable t2 period takes place

and is then detected on 13C/15N nuclei.

1.3.2.2.2 2D 13C-13C TOBSY

The magnetization is transferred from 1H to attached 13C nuclei through a refocused

INEPT step based on J-couplings as described above. This is followed by isotropic 13C

mixing using TOBSY (through-bond correlation spectroscopy). The 13C-13C correlation

is established through carbon-carbon J-couplings (JCC ~35-53 Hz). The detection takes

place on 13C. The pulse sequence of the TOBSY experiment is illustrated in Figure 12.

Figure 12. Pulse sequence of the 13

C-13

C TOBSY experiment. The magnetization is transferred

from 1H to attached

13C nuclei through a refocused INEPT step based on J-couplings: After the

initial 90º 1H pulse,

1H chemical shift evolution during the variable t1 period takes place. The

evolution delay is fixed to achieve antiphase 1H magnetization with respect to

13C via JHC (JHC

~200 Hz). The magnetization is transferred to 13

C by applying simultaneous 90º 1H and

13C

pulses. The 13

C chemical shift evolution during the variable t2 period takes place. This is

followed by isotropic 13

C mixing using TOBSY. The 13

C-13

C correlation is established through

carbon-carbon J-couplings (JCC ~35-53 Hz). The detection of 13

C magnetization represents the

final step (t3), during which the protons are decoupled.

1.3.3 Dynamic nuclear polarization (DNP)-enhanced MAS NMR

Though conventional MAS NMR is capable to answer a large number of biological

questions, the inherent low sensitivity can limit its applicability for certain issues due to


41

very long measurement times. Especially the detection of long-range interprotomer

contacts exhibts a challenge for conventional MAS NMR [129].

DNP has turned out to be perfect to overcome the drawback of low sensitivity. It

increases the sensitivity by a microwave-driven magnetization transfer from unpaired

electrons. This process relies on the fact that electrons feature a significantly higher

magnetization compared to nuclei due to a much higher gyromagnetic ratio of the

electron spins. During microwave irradiation, the transitions of the electron spin energy

become saturated, leading to a transfer of the electron polarization to proximate

nuclear spins, which in turn causes an almost homogeneous nuclear hyperpolarization.

It elevates the sensitivity in NMR spectra by up to ∼102 or lowers the acquisition time in

multidimensional experiments by up to ∼104 [130]. The theoretical enhancement (ε) for

protons is defined as the ratio of the gyromagnetic ratio (γ) of the electron (e-) and

proton (1H):

휀𝑚𝑎𝑥 ≈𝛾𝑒−

𝛾1𝐻 ≈ 660 (16)

Nuclear enhancement can be caused by different mechanisms: the Overhauser effect

(OE), the solid effect (SE), the cross effect (CE) and thermal mixing (TM) [131]. The

Overhauser effect (OE) is connected with relaxation-driven processes. It takes place,

when the interaction between electron and nucleus is time/motion-dependent. Thus, it

is primarily important for signal enhancement of solution NMR samples [132]. Thermal

mixing (TM) is closely related to the cross effect, with the difference that more dipolar

coupled electrons are used in the sample. Additionally, it is mediated by interactions at

low temperatures (T<10 K). Therefore, the conditions for both the OE and the TM are

not optimal during MAS-DNP for biological samples. Mainly the cross effect (CE) and to

some extent the solid effect (SE) contribute to signal enhancement. The SE takes

place in electron-nuclear spin interactions, which are motion independent, as in frozen

solutions or solids. It is a two spin process, including one electron and one nucleus. If

the electron spin system is irradiated at:

ω = ωe ± ωn, (17)

with ωe as electron and ωn as nuclear Larmor frequency, a simultaneous flipping of

electron and nuclear spins occurs. This results in reallocation in populations among the

electron-nucleus sublevels, leading to enhancement. The signal enhancement due to

SE is inversely proportional to the magnetic field. The cross effect (CE) is a three-spin


42

process, including an electron-electron dipole and one nucleus. It takes place under

following condition [132]:

ωn = ωe1 – ωe2, (18)

with ωe1 and ωe1 as the frequencies of the dipole and ωn as nuclear Larmor frequency.

In this study, signal enhancement is obtained by the CE using a biradical agent,

serving as a source for unpaired electrons (Figure 13). One such biradical is AMUPol

(Figure 13a) containing two linked nitroxide radicals [133]. It has been widely used for

biological samples and could been shown to be optimal for membrane proteins [134]. It

is used in this study as well. It is water soluble due to its PEG-ylated linker and can be

dissolved in a mixture containing water, glycerol (cryoprotectant) and D2O (slows

proton relaxation). Maximal enhancements can be achieved at low concentrations of

10-20 mM [134].

Figure 13. The expected low number of interacting spin pairs make dynamic nuclear

polarization for signal enhancement essential, which is obtained by the three spin cross effect

using the biradical AMUPol as a source for unpaired electrons. (a) Reconstituted mixed labelled

DGK doped with AMUPol [133] is depicted. (b) It is subjected to continuous wave microwave

irradiation, resulting in polarization transfer from electrons via protons to the sites of interest.

1.3.3.1 13C-15N TEDOR experiments for the dectection of interprotomer contacts

Specific interprotomer contacts can be detected by 15N−13C transferred echo double

resonance (TEDOR) spectroscopy [135, 136]. In the TEDOR experiment [135], the

heteronuclear 13C-15N coupling is re-introduced during MAS. It builds up on the


43

rotational echo double resonance (REDOR) experiment [137] that is used to obtain 13C-

15N distances, depending on their dipolar coupling [138]. The extent of dipolar

interactions alters with the angle of the internuclear vector towards the external

magnetic field, B0. During MAS, a single rotor period features positive and negative

values for heteronuclear coupling, depending on the orientation. Thus, in MAS NMR,

the average value of dipolar coupling for each rotor period is zero. In order to re-

introduce 13C-15N interactions, the REDOR pulse sequence includes a mixing period

with rotor synchronized 180° pulses in the 15N channel, which are applied after the 1H-

13C CP step. Thereby, the 15N magnetization is each flipped by 180°, changing the sign

of the Hamiltonian of the 13C-15N coupling. This way, the heteronuclear interaction is

not averaged to zero during MAS. Two 180° pulses per rotor period effectively rebuild

~70% of the coupling [139]. In the middle of the mixing time (tmix/2), a 180° pulse in the

13C channel is irradiated to refocus the 13C magnetization. In absence of 15N 180°

pulses, a 13C reference signal (S0) is detected, which decreases with elongated mixing

times due to homogeneous factors. In order to monitor inter-nuclear distances with

REDOR, the experiments are performed with increasing numbers of 15N 180° pulses.

These experiments are compared to the respective reference spectra (S/S0). The 13C

signal intensities (S/S0) are charted against the conducted mixing times, resulting in a

curve that can be fitted to gain the 13C-15N dipolar coupling [140]. TEDOR experiments

provide 1D or 2D spectra with all 13C-15N correlations revealed at once. The pulse

sequence (Figure 14) starts with a CP step from 1H to 13C nuclei in the preparation

period, followed by a first mixing time involving the REDOR sequence (tmix/2), which re-

introduces heteronuclear dipolar couplings. Then, one 90° pulse is irradiated each in

the 13C and 15N channel. These two pulses are applied time-delayed, separated by a z-

filter period (Δ). They result in a coherence transfer to 15N spins that evolve during the

evolution period (t1). Then, a pair of 90° pulses is irradiated in the 13C and 15N channel,

resulting in a transfer of the spin coherence back to 13C. Thereafter, a second mixing

step represented by the REDOR sequence takes place, followed by another z-filter

period (Δ) enclosed by two 90° pulses in the 13C channel. Finally, the FID detection

period (t2) occurs on the 13C spins. The z-filter periods are necessitated to compensate

13C-13C J-couplings, which affect the detection of weak 13C-15N dipolar couplings.

These J-couplings can cause wrong cross peaks and phase twisted 2D signals [136].

2D Fourier transformation leads to 2D spectra with cross peaks appearing at 13C and

15N frequencies of the interacting spins. The cross peak intensity depends on the

heteronuclear distance, thus on the degree of dipolar coupling. Additionally, it depends

on the length of the used mixing time (tmix), which relates to the number of 180° pulses

in the 15N channel during tmix/4 (l0). Short mixing times disclose short distance


44

correlations, such as between covalently bound 13C and 15N nuclei e.g. in peptide

bonds of proteins. While longer mixing times lead to long-ranging distances.

Figure 14. TEDOR pulse sequence [136]. The sequence starts with a CP-step, transferring

magnetization from 1H to

13C nuclei. This is followed by two REDOR-steps (tmix/2) to reintroduce

heteronuclear dipolar couplings, which are enclosed by 90° pulses. L0 describes the number of

180° pulses during tmix/4. The evolution time (t1) takes place in-between the two REDOR steps.

The detection of 13

C magnetization (t2) represents the final step, during which the protons are

decoupled.

Chapter 2: Materials and Methods

45

2 Materials and Methods

The chemicals, materials and equipment that were used in this study are listed in Table

S1, Table S2, Table S3.

2.1 Constructs and cells

The synthetic gene coding for wild-type diacylglycerol kinase (UniProtKB accession

code: P0ABN1) was cloned into the plasmid vector pSD005, a derivative of pTrcHisB.

The plasmid vector pSD005 was a generous gift from Prof. Dr. C. R. Sanders

(Vanderbilt University, Nashville, USA) and has been already used before in this lab.

The vector contains an ampicillin resistance sequence. The synthetic DGK gene is

localized between the unique cleavage sites NcoI and HindIII and is expressed from

the strong isopropyl-D-galactosidase (IPTG) inducible promoter Ptrc. The encoded

protein incorporates an N-terminal leader sequence containing a hexahistidine-tag for

purification. The gene product has a size of 14.3 kDa. Next to the DGK wild-type

variant, a quadruple mutant (Δ4-DGK) that has been used in this lab before, was

applied for comparison, including following mutations: I53C, I70L, M96L and V107D

[89]. The sequences for wtDGK and Δ4-DGK are shown in Figure 15.

Figure 15. Sequence alignment of wild-type DGK and the thermostable mutant, Δ4-DGK [89].

The mutations in Δ4-DGK are labelled green. The N-terminal tag is highlighted orange.

The DH5α E.coli strain (New England Biolabs, Frankfurt, Germany) featuring a high

copy number was used for transformation after PCR-based site-directed mutagenesis,


46

whereas the T7 Express E.coli strain (NewEngland Biolabs, Frankfurt am Main,

Germany) was applied for protein expression.

2.2 Molecular Cloning

All single-site mutations were introduced to the wtDGK template vector by polymerase

chain reaction (PCR) amplification with overlapping mutagenic primers. Therefore,

primers were designed, which are complementary to the parental DNA and contain the

desired mutation. In Table 2 all designed primer sequences used in this study are

illustrated:

Table 2. Primer sequences for single-site mutations

mutation primer sequence

R9A 5'- CC GGT TTC ACC GCT ATC ATC AAA G -3'

5'- GC TTT GAT GAT AGC GGT GAA ACC -3'

R22A 5'- GG AAA GGC CTG GCT GCT GCT TGG ATC -3'

5'- CCA AGC AGC AGC CAG GCC TTT C -3'

R32A 5'- GCT GCA TTC GCT CAG GAA GGT GTT GC -3'

5'- C ACC TTC CTG AGC GAA TGC AGC TTC -3'

R55A 5'- GCT ATC ACC GCT GTT CTG CTG ATC -3'

5'- CAG CAG AAC AGC GGT GAT AGC GTC -3'

R81A 5'- C GAA GCT GTT GTT GAC GCT ATC GGA TCC GAA TAC CAC -3'

5' - G GTA TTC GGA TCC GAT AGC GTC AAC AAC AGC TTC GAT A - 3'

R92A 5'- C CAC GAA CTG AGC GGC GCC GCT AAA GAC ATG GG -3'

5'- CC CAT GTC TTT AGC GGC GCC GCT CAG TTC GTG G -3'

The primers bind to the complementary sequence on the template DNA around the

mutation site, from where the DNA polymerase begins to extend the single strand

primer complimentary to the template. The elongation of the primer generates a

plasmid containing the desired mutation. The following reaction mixture (Table 3) and

program (Table 4) was used for the PCR-based site-directed mutagenesis.


47

Table 3. Components of PCR reaction mixture

Components amount

template DNA (50 ng/μl) 1 μl

5x Phusion/ GC buffer 10 μl

forward primer (0.1 μg/µl) 1 μl

reverse primer (0.1 μg/µl) 1 μl

dNTPs (10 mM) 1 μl

Phusion polymerase 0.5 μl

100% DMSO (optional) 1.5 μl (3%)

ddH2O 34 μl

Total 50 μl

Table 4. Standard PCR program used for mutagenesis of DGK

PCR-steps Temperature time cycles

pre-heating 98°C 30 s 1

denaturation 98°C 10 s

30 annealing 55°C 30 s

elongation 72°C 5 min

final elongation step 68°C 10 min 1

The annealing temperature was adapted to the melting temperature of the respective

primer. Additionally, the number of cycles was optimized for each mutagenesis.

The next step was accomplished, in order to exclude non-mutated parental from newly

mutated DNA. DNA isolated from nearly all E. coli strains is dam-methylated, whereas

the newly amplified DNA including the desired mutation contains no methyl-groups.

Therefore, the DNA was treated with DpnI endonuclease (target sequence: 5’-

Gm6ATC-3’), which is specific for methylated and hemimethylated DNA and digests

the non-mutated parental DNA. In detail, 1 µl DpnI were added to the DNA, followed by

an incubation at 37°C for 1 h.

In order to obtain high-quality, pure plasmid DNA for routine molecular biology

applications, a PCR purification kit (Macherey-Nagel NucleoSpin Plasmid) was used

according the instructions of the manufacturer. Thereafter, the mutated plasmid was

transformed into competent DH5α cells.


48

2.2.1 Transformation

For the transformation, plasmid DNA and cell suspension were thawed on ice. 2 μl

DNA (~100 ng μl-1) were added to 50 μl cells and incubated on ice for 30 min. Cells

incubated with 2 μl water acted as referee. After a heat shock at 42°C for 90 s was

applied, the cells were kept on ice for 5 min. Subsequently, the cells were resuspended

in 200 μl sterile LB medium and incubated for 1 h at 37°C and 550 rpm. The cell

suspension was plated on LB agar plates containing 0.1 mg/ml ampicillin and

incubated overnight at 37°C. The following day, sterile LB medium with 0.1 mg/ml

ampicillin was inoculated with a bacterial colony from the LB agar plate and incubated

overnight at 37°C and 220 rpm.

The plasmid DNA was isolated from 5 ml bacterial culture using a DNA extraction kit

(Macherey-Nagel NucleoSpin Plasmid) according to the manufacturer instructions. The

DNA was eluted with ddH2O and the plasmid concentration was determined at the

spectrophotometer (Thermo Scientific NanoDrop 1000). Sequences of wtDGK and all

single-site mutant constructs were verified at Eurofins MWG Operon.

2.2.2 Glycerol stocks

For the production of bacterial glycerol stocks, 100 ml sterile LB medium with

0.1 mg/ml ampicillin were inoculated with a bacterial colony and incubated overnight at

37°C and 220 rpm. 500 µl of the bacterial culture were added to the glycerol stock

(Roth), which was then inverted 10x. Afterwards, the supernatant was completely

removed. Subsequently, the glycerol stock beads with the attached E. coli cells were

shock-frozen immediately in liquid nitrogen and kept at -80°C.

2.3 Protein expression and purification

2.3.1 Samples for high field MAS NMR

For expression of DGK in E.coli T7 express cells, pre-cultures were prepared.

Therefore, 100 ml sterile LB medium with 0.1 mg/ml ampicillin were inoculated with a

bacterial colony from the LB agar plate or with one bead from the glycerol stock and

incubated overnight for ~20 h at 27°C and 220 rpm. Main cultures were inoculated with

1% of the pre-culture and grown in M9 minimal medium with [U-13C]glucose and

[15N]ammonium chloride. The salt components of the used M9 minimal medium are

listed according to the respective labelling Table 5a, Table 6a, Table 7a. They were


49

dissolved in ddH2O and the pH was adjusted to 7.0. In order to reduce spectral overlap,

unlabelled amino acids were added to the M9 medium to suppress isotope labelling of

these residues. In this study, a reverse labelled sample was prepared, in which

isoleucine (Ile), leucine (Leu) and valine (Val) were unlabelled (U-13C,15N-DGK-I,L,V)

(Table 7). For expression, 0.5 l of M9 medium was poured into a 2.5 l expression flask

and autoclaved at 121°C for 20 min. Before inoculation, autoclaved or filter-sterilized

solutions were added as listed in Table 5b, Table 6b, Table 7b.

Table 5. Composition of M9 minimal medium for the expression of unlabelled DGK

Components amount for 0.5 l way of sterilization

a Na2HPO4 (anhydrous) 3 g

autoclaving at 121°C for 20 min KH2PO4 (anhydrous) 1.5 g

NaCl 0.5 g

NH4Cl 0.5 g

added before inoculation:

b 1 M CaCl2 50 µl

autoclaving at 121°C for 20 min 1 M MgSO4 x 7 H2O 500 µl

40% glucose solution 5 ml

vitamin mix (see below) 1 ml filter sterilization via 0.2 µm filter

ampicillin (100 mg/ml) 500 µl

Table 6. Composition of M9 minimal medium for the expression of U-13

C,15

N-DGK




NaCl 0.5 g


b 1 M CaCl2 50 µl autoclaving at 121°C for 20 min

1 M MgSO4 x 7 H2O 500 µl 15

NH4Cl 0.5 g

filter sterilization via 0.2 µm filter U-13

C-glucose 2 g

vitamin mix (see below) 1 ml



50


C,15

N-DGK-I,L,V



autoclaving at 121°C for 20 min

KH2PO4 (anhydrous) 1.5 g

NaCl 0.5 g

isoleucine (unlabelled) 0.12 g

leucine (unlabelled) 0.12 g

valine (unlabelled) 0.12 g



1 M MgSO4 x 7 H2O 500 µl 15

NH4Cl 0.5 g

filter sterilization via 0.2 µm filter U-13

C-glucose 2 g

vitamin mix (see below) 1 ml


To prepare the vitamin mix, 1.5 g of vitamin tablets (Centrum, Pfizer Consumer

Healthcare GmbH, Berlin, Germany) were pulverized and dissolved in 20 ml sterile,

deionized H2O by vortexing. The suspension was centrifuged at 8000 rpm for 20 min at

4°C. The supernatant was filter-sterilized with a 0.2 μm filter and used for the

expression medium.

For inoculation of the M9 medium, 10 ml pre-culture were pelleted by centrifugation for

10 min at 6000 rpm. The supernatant was discarded and the pelleted cells were

washed with sterile LB medium. This was followed by a second centrifugation step.

Then, the pellet was resuspended in 10 ml of the respective M9 medium and the cell

suspension was added to the expression flasks. E.coli cells were grown until an

OD600 = 0.6 - 0.8 was reached, whereupon protein expression was induced by the

addition of 500 µl of IPTG (200 mg/l). Cells were harvested after 16 h of protein

expression at 27°C and 220 rpm. The cell harvesting was performed by centrifugation

at 6000 rpm for 15 min at 4°C. The supernatant was discarded and the pelleted cells

were washed with buffer A (300 mM NaCl, 50 mM HEPES, pH 7.5). This was followed

by a second centrifugation step. Then, the pellet was resuspended in 30 ml of buffer A

and stored at -80°C.

The cell suspension was thawed. One tablet of “complete protease inhibitor”, 3% (w/v)

OG, a spatula tip of DNase I and MgCl2 were added. The solubilization of wild-type


51

DGK was performed by stirring the cell suspension for 2 h at 4°C, whereas for Δ4-DGK

an extended incubation time of ~16 h at 4 °C was necessary.

The solubilized protein, which remains in the supernatant, was separated from the non-

solubilized material by centrifugation at 10 000 rpm for 30 min at 4°C.

Then, DGK was purified by immobilized metal ion affinity chromatography (IMAC) using

Ni-NTA as resin. In detail, 1.5 ml Ni-NTA and 0.75 ml imidazole (2 M) were added to

the supernatant and incubated for 1 h at 4°C under gentle stirring. Due to its N-terminal

hexahistidine-tag, the protein binds to Ni-NTA. The bound protein was washed with

75 ml buffer A containing 1.5% (w/v) OG and 50 mM imidazole. Then, the detergent

was exchanged by washing with 50 ml of buffer A containing 0.05% DDM. The protein

was finally eluted with buffer A containing 400 mM imidazole and 0.05% DDM until an

OD280<0.05 was reached. The protein concentration was determined by absorption

spectroscopy at 280 nm using the law of Lambert-Beer:

𝐴280 = 휀 ∙ 𝑐 ∙ 𝑙 (19)

A280 = absorbance at 280 nm, c = concentration of the sample (mg/ml), l = pathlength

(0.3 cm), ε = extinction coefficient (DGK: 30 480 M-1 cm-1)

2.3.2 Mixed labelled samples for DNP-enhanced MAS NMR

The expression and purification of mixed labelled DGK was in general performed as

described above with the difference that 13C- and 15N-labelled samples were expressed

separately to create mixed samples that only exhibit interprotomer and no

intraprotomer 13C−15N contacts (Table 8, Table 9). For the selectively labelled sample,

15N-labelled arginine and lysine (15Nh1/2-Arg, 15Nz-Lys) were added to M9 minimal

medium (Table 10). 12C-enriched glucose (99.5%) was used in 15N-labelled samples

instead of normal glucose to suppress 13C natural abundance within a protomer.


52


C-DGK

components amount for 0.5 l way of sterilization



NaCl 0.5 g

NH4Cl 0.5 g



1 M MgSO4 x 7 H2O 500 µl

U-13

C-glucose 2 g filter sterilization via 0.2 µm filter

vitamin mix 1 ml



C,15

N-DGK




NaCl 0.5 g




U-12

C-glucose 2 g

filter sterilization via 0.2 µm filter 15

NH4Cl 0.5 g

vitamin mix 1 ml



C,15

NArg,Lys-DGK




NaCl 0.5 g

NH4Cl 0.5 g




U-12

C-glucose 2 g filter sterilization via 0.2 µm filter

15Nh1/2-arginine 0.20 g


53

15Nz-lysine 0.21 g

vitamin mix 1 ml


In order to create mixed labelled complexes, DGK trimers had to be disassembled into

monomers, which was performed by SDS after the purification step. Differently labelled

monomers were then mixed. Subsequently, SDS had to be removed to allow

reassembling of the monomers to mixed labelled trimers, which was carried out by

washing with buffer A including DDM. In order to perform the washing step, the protein

containing a hexahistidine-tag was immobilized by binding to the Ni-NTA resin.

Therefore, the imidazole, which was used to elute DGK from the Ni-NTA resin during

the previously performed IMAC step, had to be removed from the protein solution. This

was carried out via a PD-10 desalting column, which was used according to the

instructions of the manufacturer. Therefore, in turn, it was necessary to concentrate the

protein solution to a small volume, which was performed via an Amicon centrifugal filter

with a molecular weight cutoff of 10 kDa according to the instruction manual.

Afterwards, the protein concentration was determined by absorption spectroscopy at

280 nm as described above. Then, the desalted protein solution was diluted to a

concentration of 0.2 mg/ml and was laced with 2% SDS, in order to disassemble the

DGK trimers. A low concentration of 0.2 mg/ml had to be used for disruption, since

DGK is more stable at higher protein concentrations [141]. Differently labelled protein,

for instance U-13C-DGK and U-12C,15N-DGK, were mixed in a 1:1 ratio. The protein was

incubated in SDS overnight at room temperature under gentle stirring. Additionally Ni-

NTA was added, to which the protein binds due to its histidine-tag. The next day, the

bound DGK monomers were washed with 300 ml buffer A containing containing 0.05%

DDM to remove SDS and to regain trimeric DGK. Then, the protein was eluted with

buffer A containing 400 mM imidazole and 0.05% (w/v) DDM until an OD280<0.05 was

reached and the protein concentration was determined. Using this approach, samples

were prepared, in which DGK consists of U13C- and U15N,12C-protomers ([CN]-DGK) or

U13C- and U15Nh1/2-Arg,15Nz-Lys,12C-protomers ([CN(Arg,Lys)]-DGK).

2.4 Protein reconstitution

2.4.1 Liposome preparation

The lipids DMPC and DMPA were weighted, dissolved in chloroform/ methanol (2:1)

and mixed in a 90mol%-to-10mol%-ratio, respectively. The solvent was evaporated

under a stream of nitrogen and the lipids were then further dried overnight by vacuum


54

rotary evaporation with a pressure of 40 mbar. The next day, the lipids were rehydrated

with lipid buffer (50 mM HEPES, 300 mM NaCl, 1 mM EDTA, pH 8) at a concentration

of 0.045 mmol/ml. For the preparation of liposomes, DDM was added to enable a

homogeneous insertion of DGK during the reconstitution. In order to produce small

multilamellar liposomes, 6 - 8 freeze-thaw cycles were carried out by the application of

liquid nitrogen and sonication.

2.4.2 Reconstitution via BioBeads

The protein was mixed with the pre-softened liposomes and incubated for 1 h at 22°C.

Detergent removal was done with autoclaved SM-2 BioBeads. In detail, 4-times 80 mg

BioBeads/ml were added and incubated primarily overnight at 4°C and subsequently 3-

times for each 2 h at 22°C. Finally, the BioBeads were removed by filtration.

For all NMR experiments, reconstituted DGK was washed 6-7 times with NMR buffer

(20 mM HEPES, 3 mM MgCl2, pH 7.2).

2.5 Sample characterization

2.5.1 SDS-PAGE

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was used for

the qualitative characterization of the sample purity. 13 μl sample were mixed with 7 μl

loading buffer (RunBlue LDS Sample Buffer 4x: 40% glycerol, 4% LDS, 0.8 M

triethanolamine-chloride pH 7.6, 4% Ficoll-400, 0.025% Phenol Red, 0.025%

Coomassie Brilliant Blue G250, 2 mM EDTA) containing, additionally, a concentration

of 10% β-mercaptoethanol. A total of 15 μl of each sample and the protein marker

(AppliChem Protein-Marker III 6.5-200 kDa) were loaded on precast gels from

ExpedeonTM (RunBlue SDS Gel 4-20%, 20 μl loading wells). The gel was fixed in the

SDS-PAGE apparatus. The inner cathode chamber was filled with cathode buffer

(0.1 M Tris, 0.1 M Tricine, 0.1% SDS, pH 8.25) and the outer anode chamber was filled

with anode buffer (0.2 M Tris, pH 8.9). Electrophoresis took place at 180 V for ~30 min

until the dye front reached the base of the gel. The, the gels were stained with 0.025%

Coomasie Brilliant Blue G250 for 30 min and subsequently destained in 50% methanol,

40% acetic acid and 10% ddH2O.


55

2.5.2 SEC

In order to characterize the homogeneity of the protein sample, size exclusion

chromatography (SEC) was carried out, during which the proteins are separated

according to their hydrodynamic volume. This way aggregates or co-purified proteins

can be separated.

For SEC analysis of DGK, the ÄKTA-Explorer System with a Superdex 75 10/300 GL

column with a separation range between 3-70 kDa (GE Healthcare) was used at room

temperature. The column had a bed volume of 24 ml. Before loading the protein

sample, the column was washed with 50 ml sterile filtered and degassed ddH2O. Then,

it was equilibrated with 50 ml filtered and degassed SEC buffer (100 mM NaCl, 50 mM

HEPES, pH 7.2) with the respective detergent. The washing steps were performed at

slow flow rates of 0.2 ml min-1. The protein sample was filtered with Pall Life Sciences

500 μl tubes (0.2 μm filter pore size), before it was loaded onto the column. The, the

sample was filled into the 500 µl sample loop via a syringe. The run was performed at

higher flow rates of 0.5 ml min-1. During the run, the absorption at 280 nm was detected

against the elution volume. After the run was completed, the column was washed with

ddH2O and afterwards with 20% ethanol.

2.5.3 BN-PAGE

Blue native-polyacrylamide gel electrophoresis (BN-PAGE) was used to assess the

oligomeric state of DGK. In contrast to the SDS-PAGE, which has a denaturing effect

on proteins, the native state of the proteins is retained during the BN-PAGE by using

Coomassie Brilliant Blue G250 as the charge shift molecule. Coomassie Brilliant Blue

G250 binds to the hydrophobic regions of the protein, replacing the detergent micelle

[142]. Thus, different oligomeric states of a protein can be determined by BN-PAGE.

Samples were prepared as described in Table 11 with BN-PAGE loading buffer 4x

(200 mM BisTris, 64 mM HCl, 200 mM NaCl, 40% (w/v) glycerol, 0.004% (w/v)

Ponceau S, pH 7.2) and 5% Coomassie Brilliant Blue G250 solution. The concentration

of Coomassie Brilliant Blue G250 was adjusted to one quarter of the detergent

concentration in the sample.

Table 11. Sample preparation for BN-PAGE

Component volume

protein in detergent micelles x µl (3.5-4 µg protein)

BN-PAGE loading buffer 4x 6.25 µl


56

10% DDM solution 0.75 µl

5% Coomassie Brilliant Blue G250 solution 0.1-1 µl

ddH2O adjusted to 25 µl

The samples were loaded onto Precast NativePAGE Novex Bis-Tris 4-10% gels from

InvitrogenTM, which were fixed in the BN-PAGE apparatus. The outer anode chamber

was filled with BN-PAGE running buffer (50 mM BisTris, 50 mM Tricine, pH 6.8) and

the inner cathode chamber was filled with dark blue BN-PAGE cathode buffer (50 mM

BisTris, 50 mM Tricine, 0.02% Coomassie Brilliant Blue G250, pH 6.8). Afterwards,

25 µl of each sample and 5 µl of the protein marker (NativeMarkTM unstained protein

standard from InvitrogenTM: 20-1200 kDa) were loaded onto the gel. The

electrophoresis was then carried out at 150 V for ~2 h at room temperature until the

dye front reached the base of the gel. Finally, the gel was incubated in a fixing solution

(40% ethanol and 10% acetic acid), microwaved for ~45 s and shaken at room

temperature for 15 min. The same procedure was repeated with destaining solution

(8% acetic acid). Subsequently, the gel was incubated in the destaining solution until

the bands became clearly visible for analysis.

2.5.4 LILBID-MS

Laser induced liquid bead ion desorption-mass spectrometry (LILBID-MS) was used to

probe the oligomeric state of DGK in different detergents, verifying the results obtained

by BN-PAGE. All LILBID-MS measurements on DGK were carried out by Oliver Peetz

of the research group of Prof. Dr. Nina Morgner (Institute of Physical and Theoretical

Chemistry, Goethe University Frankfurt am Main). Since LILBID is highly tolerant

concerning salts and detergents, it enables the investigation of biomolecular complexes

in native-like environments, such as membrane proteins. LILBID-MS is soft enough to

prevent fragmentation. However, fragmentation can be induced on demand by

elevating the laser intensity, which allows the determination of subunit compositions.

During LILBID, microdroplets of an aqueous solution containing buffer, salt and further

additive components (e.g. detergent) next to the analyte (e.g. membrane protein) are

injected into the vacuum and irradiated one-by-one by mid-IR laser pulses. For the

LILBID-MS experiments on DGK in different detergents, the protein was transferred

into a salt-free buffer containing 50 mM ammonium acetate and the respective

detergent. This was carried out via a PD-10 desalting column according to the

instructions of the manufacturer. In detail, samples containing 20 µM DGK in 50 mM

ammonium acetate buffer with 0.025% DDM and samples containing 4 µM DGK in


57

50 mM ammonium acetate buffer with 0.1% SDS were prepared. The LILBID-MS

measurements were performed using previously published standard settings [143]. A

piezo-driven droplet dispenser (MD-K-130 by Microdrop Technologies GmbH,

Norderstedt, Germany) generated droplets of 50 μm diameter at a repetition rate of

10 Hz. The droplets were transferred into a two-stage differential vacuum chamber.

Beginning at ~0.3 bar, they were transferred into high vacuum (10-5 mbar) via apertures

that reduce pressure. There, the droplets were irradiated one-by-one by mid-IR laser

pulses produced by a home-build Nd:Yag pumped LiNbO3 optical parametric oscillator

(OPO) [144, 145]. The IR pulses were tuned to the absorption wavelength of water at

2.94 µm. Consequently, the IR radiation directly excite the symmetric and asymmetric

stretch vibration of H2O. At this frequency, the penetration depth of the IR beam in H2O

is just ~1 µm. The excited stretch vibrations were shown to relax in bulk water within a

few hundreds of femtoseconds [146], leading to a production of heat. Thus, by applying

a pulse, causing an absorption of multiple photons, a high temperature is generated.

Due to the fulminating strong temperature increase, an elevated pressure is induced,

leading to a supercritical state, in which the droplets expand explosively and the

analyte ion is released into the gas phase. The ions are accelerated by a pulsed

electric field and mass-analyzed by a reflectron time-of-flight (TOF) mass spectrometer.

The mass spectra were recorded with a 8-bit digitiser card (Aquiris). The hardware was

controlled by a home-written software using LabView that enables the timing of the

measurements and data accumulation. The software Massign was used for data

processing, including signal calibration, smoothing, and background subtraction [147].

2.5.5 Sucrose density gradient centrifugation

A sucrose gradient (40% - 10%) was carried out to verify a homogenous reconstitution

of the protein into liposomes. Therefore, 1 ml of each solution, starting with 40%, was

carefully layered into a 5 ml centrifuge tube and left to equilibrate for 1 h at room

temperature. Then, 0.2 ml of a 0.4 mg ml-1 solution of empty liposomes or

proteoliposomes were carefully layered on top. The centrifugation was performed at

33 000 rpm in a SW50.1 rotor overnight at 4°C.

2.5.6 Coupled activity assay

The activity of reconstituted DGK was determined using a coupled enzyme assay [18].

The formation of ADP during the catalytic reaction of DGK was measured in a two-step


58

process. In the first step, the produced ADP and phosphoenolpyruvate were converted

to ATP and pyruvate by pyruvate kinase (PK). In the second step, pyruvate was

converted to lactate by lactate dehydrogenase (LDH), oxidizing NADH to NAD+. The

decrease of NADH absorbance was monitored at 340 nm.

The measurement was done in a Tecan Reader Infinite M200 at 30°C. The assay

buffer (100 µl per measurement), provided on a 96well plate, was based on 25 mM

PIPES at pH 6.8 and contained 1 mM phosphoenolpyruvate, 0.5 mM NADH, 3 mM

MgATP, 15 mM MgCl2, 50 mM LiCl and 0.1 mM EDTA. 5 µl of an enzyme mix of

pyruvate kinase (PK) (18.2 units) and lactate dehydrogenase (LDH) (7.5 units) was

added to the assay buffer. Everything was preincubated at 30°C until the remaining

ADP was dissipated. The reaction was then initiated by the addition of 1-2 µl of protein.

The activity was stimulated by 2.63 µl of the water soluble lipid substrate analogue 1,2-

dibutyrylglycerol (DBG) (100 mM). The synthesis of DBG was carried out by Andreas

Jakob of the research group of Prof. Dr. Alexander Heckel (Institute of Organic

Chemistry and Chemical Biology, Goethe University Frankfurt am Main).

Concentrations of the different constituents were adapted so that the reaction catalyzed

by DGK was rate-limiting. The activity was calculated from the linear decrease of the

NADH absorption over time.

2.6 Preparing substrate-bound states of DGK

In order to saturate DGK with nucleotide substrate, reconstituted DGK was incubated

with a 14-fold molar excess of the ATP analogue adenylylmethylenediphosphonate

(AMP-PCP) and a 28-fold molar excess of MgCl2 overnight at 4°C. AMP-PCP and

MgCl2 were dissolved in 20 mM HEPES (pH 7.2). The accessibility of the nucleotide to

all active sites of DGK was enhanced by a 5 min sonication step in water bath (followed

by an incubation on ice to prevent over-heating) prior to the incubation time at 4°C. In

order to saturate DGK with the lipid substrate in a 10-fold molar excess, it was

reconstituted in a molar protein-to-lipid ratio of 1:50 into liposomes consisting of

80 mol% DMPC/DMPA (9:1) and 20 mol% 1,2-dioctanoyl-sn-glycerol (DOG, n=8). For

the production of DGK saturated with both the nucleotide and the lipid substrate, it was

firstly reconstituted into 80 mol% DMPC/DMPA (9:1) and 20 mol% DOG (n=8) in a

molar protein-to-lipid ratio of 1:50 and then incubated with a 14-fold molar excess of

AMP-PCP and a 28-fold molar excess of MgCl2 overnight at 4°C.


59

2.7 MAS NMR

2.7.1 MAS NMR at high field

For the NMR experiments at high field, all proteoliposome samples were sedimented

by ultracentrifugation at 55 000 rpm for 1 h at 4°C and packed into a 3.2 thin wall rotor.

Approximately 18 mg of DGK could be loaded into the thin wall rotor. The NMR

experiments were then carried out on a Bruker wide bore Avance III solid state NMR

spectrometer with a 1H frequency of 850.32 MHz. A sample spinning rate of 15.2 kHz

was applied in each case. All samples were adjusted to a temperature of 275 K and

pH 7.2. The NMR time of dipolar coupling based experiments was reduced by

paramagnetic doping with 2 mM Gd3+-DOTA [12] in combination with an E-free 3.2 mm

triple-resonance HCN MAS probehead, which enabled using a recycle delay of 0.8 s.

Thereby, ~3x of the measurement time could be saved compared to the standard

probehead (recycle delay of 2.5 s). The E-free probehead was custom-built by Bruker.

2.7.1.1 Manual resonance assignment

NMR resonance assignment is based on multidimensional spectra that correlate

nuclear spins, leading to cross peaks. These nuclear correlation experiments are

chosen to complement each other, linking spin systems and mapping them to the

protein amino acid sequence, causing a network of peaks.

For the sequential assignment of the immobile domains of DGK, a combination of

dipolar coupling based 3D experiments (NCACX, NCOCX, CONCA) [123, 124] was

carried out. 13C and 15N assignments were mainly performed using uniformly labelled

samples (U-13C,15N-DGK). Residual ambiguities were resolved by reverse labelling of

Isoleucine, Leucine and Valine (U-13C,15N-DGK-I,L,V). The experiments were either

performed with an E-free or standard 3.2 mm triple-resonance HCN MAS probehead

(Bruker). All experimental parameters are listed in Table S4.

For the tentative assignment of highly mobile residues of DGK, scalar coupling based

2D experiments (13C-13C TOBSY, 1H-13C HETCOR, 1H-15N HETCOR) [46, 148, 149]

were applied. Therefore, only the uniformly labelled sample (U-13C,15N-DGK) was used.

The experiments were performed with a standard 3.2 mm triple-resonance HCN MAS

probehead (Bruker). Typical 90° pulse lengths were 2.5 µs (1H), 4.5 µs (13C) and 6 µs

(15N). A recycle delay of 2 s and a SPINAL64 1H decoupling of 100 kHz were used. For

all heteronuclear transfer steps a refocused INEPT (insensitive nuclei enhanced by

polarization transfer) [128] was applied with scalar couplings of 200 Hz (HC) and 90 Hz


60

(HN). For the 13C-13C homonuclear polarization transfer, TOBSY (through-bond

correlation spectroscopy) [148] was applied with a 3.75 ms P916 mixing sequence.

2.7.1.2 Substrate bound states

The binding of the substrate(s) to DGK was verified for each case (AMP-PCP,

DOG+ATP, DOG+AMP-PCP) by 31P-CP (cross polarization) and 31P-DP (direct

polarization) experiments. Therefore, the standard 3.2 mm double-resonance HX MAS

probehead (Bruker) was used. Typical 90° pulse lengths were 3 µs (1H) and 4 µs (31P).

A recycle delay of 3 s and a SPINAL64 1H decoupling [150] of 83.3 kHz were used.

The CP contact time was 5 ms. All CP spectra were recorded with 16 000 scans, while

the DP spectra were carried out with 2000-4000 scans.

Scalar and dipolar coupling based experiments of DGK in its apo state, saturated with

AMP-PCP, DOG and with AMP-PCP + DOG were carried out for the analysis of CSPs

and peak intensities during substrate(s) binding. The experiments were either

performed with an E-free or standard 3.2 mm triple-resonance HCN MAS probehead

(Bruker). All experimental parameters are listed in Table S4.

2.7.1.3 Data analysis

All spectra were processed in TopSpin 3.5.b.91. pl 7 (Bruker). For comparison, the

respective spectra were processed the same way. Data analysis of multi-dimensional

experiments and resonance assignment were performed using CCPN 2.4.2 [151].

For the first 2D NCA spectrum, 13C chemical shift referencing was carried out with

respect to DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid) through alanine

(179.85 ppm). 15N chemical shifts were indirectly referenced to liquid NH3 at 0 ppm

through the 15N/13C gyromagnetic ratio of ~0.4. All other 2D and 3D spectra were

subsequently referenced to an isolated resonance in this spectrum (112TrpCa). 31P

chemical shift referencing was performed with respect to 10% phosphoric acid at

0 ppm through crystalline triethylphosphine (TEP) (58.62 ppm).

Concerning the analysis of chemical shift perturbations (CSPs) in 3D NCACX spectra,

weighted chemical shift changes were calculated according to [152]:

∆𝛿(𝐶𝑥, 𝐶𝛼, 𝑁) = √(∆𝛿𝐶𝑥)2 + (∆𝛿𝐶𝛼)2 + (∆𝛿𝑁

2.48)² (20)


61

where ΔδCx and ΔδCα and ΔδN represent chemical shift perturbations in the 13Cx,

13Cα and 15N dimension, respectively. All CSPs ≥ 0.2 ppm were counted as significant.

2.7.2 Automatic resonance assignment by ssFLYA

The ssFLYA algorithm [13] is predicated on FLYA, an automated resonance

assignment algorithm for solution NMR [153] and integrated in the software package

CYANA [154, 155]. In this study, ssFLYA was conducted on DGK by Dr. Sina Kazemi

of the research group of Prof. Dr. Peter Güntert (Institute for Biophysical Chemistry,

Goethe University Frankfurt am Main). ssFLYA assigns the frequencies to the spins by

mapping the network of expected peaks with unknown positions to the unassigned

detected peaks with known position. As input, it used solely the sequence of DGK and

the unassigned peak lists from dipolar coupling based 3D experiments (NCACX,

NCOCX, CONCA) obtained with the uniform (U-13C,15N-DGK) and reverse (U-13C,15N-

DGK-I,L,V) labelled sample. Partial assignments like grouping chemical shifts to a spin

system or setting them to atom types (N, Ca, Cb, Cg, etc.) within one residue were not

necessary. Peak lists were gained by manual peak picking, guided by input information

gathered from the manual assignment. The N-terminal region (-9Met–14Ala) of the

protein sequence, which is known to be highly mobile and not detectable, was

excluded in the calculations. Identified spectral artifacts like folded peaks were

removed from the input peak lists as well. For the calculations including the peak lists

of the reverse (U-13C,15N-DGK-I,L,V) labelled sample, the residue types Ile, Leu and

Val were excluded.

The entire experimental data set was used simultaneously to obviate possible

entrapments, in which results of an already performed step stay fixed for subsequent

steps. Instead of dictating a specific assignment strategy, ssFLYA creates peaks that

are expected in a given spectrum by applying a set of rules for through-bond or

through-space polarization transfer. Then, it constructs an optimal mapping of the

expected peaks that are assigned by definition but with unknown positions with the

detected peaks that are initially unassigned but with known positions in the spectrum

[153, 156-158]. Expected peaks are predicated on covalent connections between

atoms. Most of these ssNMR experiments include a relatively unspecific 13C–13C

transfer, leading to additional signals from neighbouring carbons with a high probability

for directly bound ones. This effect was considered by adding covalent bond patterns

with lower detection probability. An evolutionary optimization procedure was applied

that functions with a population of individuals, each typifying an assignment solution for


62

the protein. The search area of an expected peak was defined by BMRB chemical shift

statistics [159]. Detected peaks were assigned within a given tolerance. Just one

detected peak could be mapped by one expected peak. The first generation of

assignment solutions was obtained randomly, but subjected to these conditions. In

each generation a local optimization algorithm revoked little parts of a mapping and

reassigned the expected peaks for a defined number of iterations with 15 000 as

default. Then, the obtained different solutions of one generation were recombined into

a new generation. Individuals and particular parts of an individual that are conducive to

a new individual were defined by a scoring function (equation 21). The solution, which

developed this function to a maximum, represents the final assignment. The global

score for complete assignment solutions assessed four criteria of an assignment

solution: the distribution of chemical shift values with regard to the chemical shift

statistics, the assimilation of peaks assigned to the same atom, the totality of the

assignment, and a penalty for chemical shift degeneracy. The global score G was

calculated as following [13]:

𝐺 = ∑ [𝑤1(𝑎)𝑄1(𝑎) + ∑ 𝑤2(𝑎, 𝑛)𝑄2(𝑎, 𝑛)/𝑏(𝑛)]𝑛∈𝑁′𝑎𝑎∈𝐴

∑ [𝑤1(𝑎) + ∑ 𝑤2(𝑎, 𝑛)]𝑛∈𝑁𝑎𝑎∈𝐴0 (21)

A0 describes the set of all atoms for which expected peaks are present, whereas A ⊆

A0 represents the set of assigned atoms, Na the set of expected peaks for atom a, and

N’a ⊆ Na the subset of expected peaks, which were mapped to a detected peak. b(n)

relates to the ambiguity of the assignment and aligns the number of expected peaks

that were assigned to the same detected peak as expected peak n. Unassigned atoms

and unmapped peaks were conducive via the normalization by the denominator. The

weighting factors were set to w1(a) = 4 and w2(a, n) = 1 for all calculations in this

study. The measure of quality Q1(a) refers to the agreement of the average chemical

shift ϖ(a) in the chemical shift list of atom a with the respective chemical shift statistics.

Likewise, Q2(a,n) quantifies the agreement between the chemical shift of atom a

determined from the detected peak, mapped by the expected peak n, and the average

frequency of the atom in the assigned peaks of the respective spectrum [13, 153]. For

a perfect match, the measures of quality Q = 1. In all other cases Q < 1. If Q = 0 or Q =

-∞, the assignment was considered as insufficient and equates to no assignment.

Thus, the global score G is 1 for a theoretical perfect assignment, and G < 1 in all other

cases. To enhance and evaluate the accuracy of the assignment, 20 independent runs

of the algorithm were carried out with several random seeds. For each atom a

consensus chemical shift was generated from the values determined in the individual


63

runs [13, 153, 160, 161]. The consensus chemical shift ϖ(a) for an atom a is the value,

which developed this function to a maximum [13]:

µ(𝜔) = 1

𝑚 ∑ 𝑒𝑥𝑝 (−

1

2(

𝜔 − 𝜛𝑗(𝑎)

휀(𝑎))

2

)𝑚

𝑗=1 (22)

where ϖj(a) is the chemical shift value determined for atom a in run j, and ε(a) is the

chemical shift tolerance, which was set to 0.55 ppm for all calculations. The maximum

value of this function, µ(ϖ(a)), is a measure of the consensus of the chemical shift

values gained in the individual runs. It can be determined without the knowledge of

reference assignments. If all chemical shift values are identical, then µ(ϖ(a)) = 1.

Assignments with µ(ϖ(a)) ≥ 0.8 were classified as “strong”, all others as ‘‘weak’’. Weak

assignments were considered as tentative and needed further verification, since 39-

72% turned out to be erroneous with respect to manual assignments of already

performed automatic assignments by ssFLYA [13]. The assignment calculation could

be performed within approximately 10 min, if 20 CPU cores were availabe.

2.7.3 DNP-enhanced MAS NMR

Reconstituted protein samples were doped with a polarizing agent in order to achieve

DNP signal enhancement. Two proteoliposome pellets, each ~20 µl, were covered with

∼20 μL of a 20 mM AMUPol [133] solution (60% D2O, 30% glycerol-d8, 10% H2O) and

incubated for 20 h at 4°C. The solution was completely removed before the sample

was packed into a 3.2 mm ZrO2 rotor.

DNP-enhanced MAS NMR spectra were recorded on a Bruker 400 DNP system

consisting of a 400 MHz WB Avance II NMR spectrometer, a 263 GHz Gyrotron as

microwave source, and a 3.2 mm HCN Cryo MAS probe. All experiments were

conducted with 8 kHz MAS, and the microwave power at the probe was 12.5 W. During

DNP experiments, the temperature was kept at 105 K. For all experiments, a

SPINAL64 1H decoupling [150] of 100 kHz was applied during acquisition. A recycle

delay of 2.2 s was used. 2D 15N−13C correlation spectra were acquired using the z-

filtered TEDOR sequence [136]. Typical pulse lengths of 2.5 µs (1H 90°), 4.0 µs

(13C 90°), 8.0 µs (13C 180°), 7.5 µs (15N 90°) and 15 µs (15N 180°) were applied. The

CP contact time was 1.1 ms. A mixing time of 6.25 ms (24 rotor cycles) was used for all

experiments. The 2D-spectra for [CC]DGK and [CN(Arg,Lys)]DGK were acquired with

1504 scans in the direct dimension and 60 increments of 125 µs each in the indirect

dimension. The FID acquisition time in the direct dimension was 10 ms. The 15N pulse


64

carrier was set to 54 ppm and the 13C pulse offset was set to 174 ppm. The 2D-spectra

for [CN]DGK and [CN]DGK-RxA were acquired with 2880 scans in the direct dimension

and 25 increments of 250 µs each in the indirect dimension. The FID acquisition time in

the direct dimension was 10 ms. The 15N pulse carrier was set to 100 ppm and the

13C pulse offset was set to 174 ppm.

All spectra were processed in TopSpin 3.5.b.91. pl 7 (Bruker). For comparison, the

respective spectra were processed the same way. Data analysis of 2D experiments

was performed using CCPN 2.4.2 [151].

Chapter 3: Sample optimization

65

3 Sample optimization

3.1 Introduction

E. coli DGK serves as a model system for membrane proteins since decades starting

1965 [93]. One good reason is its convenient experimental handling. Consequently, its

molecular cloning, expression and purification are well established. It can be easily

expressed recombinantly in E.coli cells, solubilized in octylglucoside (OG) [96, 97] and

purified by immobilized metal ion affinity chromatography (IMAC) using a hexa-His

tagged form [9]. Additionally, DGK features an easily assayable function. Badola and

Sanders designed an assay system [18], which is based on central aspects of the

mixed micellar assay developed by Bell and co-workers [97]. It is combined with the

classic pyruvate kinase/lactic dehydrogenase reaction coupling system that was long

utilized in studies of water-soluble kinases, in which rate-limiting ADP production by

DGK is coupled to NADH oxidation [103, 110].

However, the preparation of a membrane protein sample for MAS NMR is generally

considered to be a challenge, since high amounts of pure, isotope labelled protein are

required due to the inherently low sensitivity of the technique. Furthermore, in terms of

membrane proteins, it is highly beneficial, if they are embedded into lipid bilayers, since

they are strongly affected by interactions with the surrounding membrane. Additionally,

the catalytic mechanism of numerous membrane proteins is substantially linked to the

lipid bilayer [11, 20, 21, 162]. Thus, the reconstitution step is of key importance for the

investigation of a native-like membrane protein. It is a process, in which a purified,

typically detergent solubilized membrane protein is incorporated into an artificial lipid

bilayer. In this connection, a homogenous reconstitution of the protein into lipid

membrane is necessary to obtain well-resolved MAS NMR spectra. Likewise important

is the stability of the reconstituted sample, since the measurements can take up to two

weeks.

Summing up, the general aim of membrane protein preparation for MAS NMR

experiments is the production of high amounts of pure, active, stable and

homogeneously reconstituted protein, which provides well-resolved MAS NMR spectra.

Previously in this lab, efforts have been made to prepare a DGK sample that features

good spectral resolution for structural and functional studies by MAS NMR [163]. In

detail, the thermostable quadruple mutant (I53C, I70L, M96L, V107D) of DGK was

used. As detergent environment and for lipid softening, dodecyl phosphocholine (DPC)

and as lipid components 67.3mol% DMPC/ 32.7mol% cholesterol were utilized. The


66

reconstitution step was performed by dialysis and the molar protein-to-lipid ratio was

1:80.

This chapter demonstrates a stepwise optimization of the whole preparation protocol,

leading not only to a more native-like sample, but also to high quality MAS NMR

spectra and a very efficient preparation process.

3.2 Results and Discussion

3.2.1 DGK construct: Quadruple mutant (Δ4-DGK) vs. wild-type DGK

Since the main aim of the sample optimization was to elaborate a most native-like

sample, wild-type DGK and its thermostable mutant Δ4-DGK (CLLD-DGK), used in

preliminary studies in this lab, were compared. The quadruple mutant was introduced

by Zhou and Bowie [89]. It features a high thermostability, proven to be advantageous

for crystallization purposes [5, 6]. However, the wild type has been shown to be itself

remarkable stable: In native membranes, it is resistant for a few minutes to irreversible

inactivation at 100°C [93, 108]. Even in detergent micelles, DGK features a high

stability. The half time for irreversible inactivation at 70°C and pH 6.5 is on the order of

hours [8, 164]. Lau and Bowie worked out a method for quantitating the thermodynamic

stability of DGK [9]. Denaturing sodium dodecylsulfate (SDS) was titrated to folded

DGK in decylmaltoside micelles. This way, an impressive unfolding free energy of 16

kcal/mol for the transmembrane domain and even 6 kcal/mol for the cytoplasmic

domain could be demonstrated. These data indicate that wild-type DGK should be

stable enough for functional studies by MAS NMR.

In this study, wild-type DGK and its thermostable mutant were compared according

their activity (Figure 16a). Using the coupled enzyme assay, it could be shown that the

wild type is 61% more active than the Δ4-mutant, indicating that the four inserted

mutations most likely lead to unfavorable structural alterations in the enzyme. This

remarkable difference clearly highlights the need to use wild-type DGK for structural

and functional studies. Also the high quality MAS NMR spectra of wtDGK, shown in

3.2.6, were convincing (Figure 21b). Thus, further optimization was done with the wild

type.


67

3.2.2 Type of detergent: DPC vs. DDM

The next step was to find a perfect detergent. After the expression, DGK has to be

solubilized by the transfer into detergent micelles to remove impurities and to enable

manipulations in solution. The purification on a Ni-NTA column allows the transfer of

the bound protein into different detergent environments by washing the column with

another detergent prior to elution. The type of detergent influences the stability and

activity of membrane proteins and affects the homogeneous incorporation into

liposomes [165, 166]. Not only the protein itself but also the preformed liposomes are

saturated with detergent to disrupt lipid-lipid interactions, resulting in a more permeable

bilayer for protein uptake.

In order to check the purity and stability of wtDGK in the respective detergent, size

exclusion chromatography (SEC) was carried out immediately after the IMAC step. The

chromatogram of DGK in dodecyl phosphocholine (DPC) features a small peak

indicating aggregation (Figure 16c2). In contrast, wtDGK eluted in n-dodecyl-β-d-

maltopyranoside (DDM) is stable over several weeks at 4°C without showing any

aggregates (Figure 16c1). Also the peak is clearly broader in DPC. This is reflected in

the activity as well. Reconstituted DGK shows a notable reduced activity in DPC

compared to DDM, which is decreased by 35% (Figure 16b). This can be explained by

the physico-chemical characteristics of the two detergents according to their head

group and hydrophobic chain. Non-ionic detergents, like DDM, bear an uncharged

hydrophilic glycosidic head group. They are known to be mild and comparably non-

denaturizing, since they break lipid-lipid and lipid-protein rather than protein-protein

interactions. DPC in contrast belongs to zwitterionic detergents, which are considered

to be harsher and more deactivating than non-ionic ones [167, 168]. Therefore, DDM is

the detergent of choice for all subsequent studies.


68

Figure 16. (a) Comparison of the activity of wt- (dark grey) and Δ4-DGK (red) reconstituted into

DMPC/DMPA. Both samples are prepared the same way. (b) Comparison of the activity of

reconstituted wtDGK, prepared in DDM (dark grey) and DPC (red). 100% activity corresponds to

the rate recorded with wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1

mg-1

.

All activity measurements were repeated three times. The activity was calculated as the mean

value. Error bars correspond to standard deviations. (c) Size exclusion chromatography (SEC)

of wtDGK in 0.05% DDM (c1) and 0.5% DPC (c2), performed after several weeks at 4°C or

immediately after the IMAC step, respectively.

3.2.3 Characterizing the yield, purity and oligomeric state of wtDGK in DDM

Protein expression and purification were performed as described in chapter 2. The

sample is characterized regarding yield, purity and oligomeric state. The yield for

wtDGK could be increased by 50%, from 20-30 mg of DGK per liter E. coli culture to

30-45 mg/L, most likely simply by an optimized handling of the sample. The inserted

mutations of the quadruple mutant seem not to play a role, since for both wild type and

Δ4-mutant similar yields were obtained. The purity of wtDGK in DDM micelles was

proved by SDS-PAGE, showing a pure protein sample after IMAC purification (Figure

17a). BN-PAGE and LILBID-MS analysis were used for a reliable assessment of the

oligomeric state. The BN-PAGE shows wtDGK exclusively in its trimeric form without

any aggregates observable (Figure 17b). LILBID-MS verifies the results from BN-

PAGE, demonstrating that wtDGK forms predominantly trimers in DDM micelles

(Figure 17c). LILBID-MS was carried out by Oliver Peetz of the research group of Prof.


69

Dr. Nina Morgner (Institute of Physical and Theoretical Chemistry, Goethe University

Frankfurt am Main).

Figure 17. Characterization of the purity and oligomeric state of wtDGK in DDM micelles. (a)

The SDS-PAGE verifies the purity of the protein solution after IMAC purification. (b) The BN-

PAGE offers a reliable assessment of the oligomeric state, clearly showing wtDGK as trimer in

DDM micelles. (c) The trimeric state of wtDGK in DDM micelles is confirmed by LILBID-MS: The

signals for the monomeric, dimeric and trimeric form of wtDGK are labelled by “1”, “2” and “3”,

respectively. They occur at charged states of −1 and −2. The LILBID mass spectrum was

recorded by Oliver Peetz of the research group of Prof. Dr. Nina Morgner (Institute of Physical

and Theoretical Chemistry, Goethe University Frankfurt am Main).

3.2.4 Reconstitution method: Dialysis vs. BioBeads

Since membrane proteins in general and DGK in particular are strongly affected by

interactions with the surrounding membrane [11, 20, 21, 162], DGK was reconstituted

into liposomes for its investigation by MAS NMR. Therefor, the detergent had to be

removed. For this purpose, different removal methods for detergents are available.

They take advantage of the features of the respective detergent, i.e. critical micellar

concentration (CMC), charge or aggregation number. Most important for the

reconstitution step is that the detergent has to be completely removed, since even

small impurities interfere with the NMR spectra. Furthermore, the membrane protein

should be distributed among the liposomes evenly to ensure well-resolved MAS NMR

spectra. Most common removal methods are dialysis and hydrophobic absorption via

BioBeads [167, 168]. During dialysis, the concentration of the detergent is diluted to

values below the CMC, leading to the decomposition of micelles to single detergent

monomers. These monomers can be easily eliminated by a concentration gradient over


70

a dialysis membrane with a certain cut-off. This process requires a detergent-free

buffer of about 1000-fold excess compared to the protein-liposome-detergent solution.

It is quite time-consuming, since it takes usually over one to two weeks. However, this

reconstitution method might be useful for certain proteins, which are prone to

aggregation during a fast detergent removal. Dialysis functions best with detergents of

a high CMC, with low molecular weight and small cross-sectional area. Non-ionic

detergents are difficult to eliminate by dialysis due to their low CMC. They can be

clearly better removed by hydrophobic absorption via BioBeads [165, 169], whereupon

amphiphilic detergents bind to the hydrophobic surface of the insoluble beads using

their hydrophobic tail. The beads are mixed with the solution containing the detergent,

incubated for one to two days under slow stirring and can then be easily removed by

centrifugation or filtration.

Since the type of detergent used in this study was altered from zwitterionic DPC to non-

ionic DDM, the method of reconstitution was changed from dialysis to hydrophobic

absorption via BioBeads. The sucrose gradient (40%-10%) in Figure 20 shows that

DGK could be homogenously reconstituted using Biobeads. Also the high quality MAS

NMR spectra of DGK reconstituted into liposomes via BioBeads were convincing

(Figure 21b). Additionally, the usage of BioBeads as reconstitution method instead of

dialysis had the positive side effect of reducing the expenditure of time remarkably from

two weeks to two days.

Table 12. Detergents used and compared in this dissertation. The classification, CMC and

concentration range are shown [163].

Detergent Classification CMC [mM] Conc.% [w/v]

DDM n-dodecyl-β-maltopyranoside non-ionic, alkyl maltoside 0.15-0.19 0.03-0.10

DPC dodecyl phosphocholine zwitterionic, alkylphosphocholine 1.10 0.20-0.60

3.2.5 Lipid composition, protein-to-lipid ratio and functional characterization

Another crucial point for the preparation of a membrane protein sample for MAS NMR

is the definition of the liposome composition. The model membrane should guarantee

both a perfect emulation of the native environment to maintain the functional native

state of the membrane protein and a high resolution as well as a high sensitivity of the

NMR spectra. The membrane affects the structure and activity of a membrane protein

through specific or unspecific interactions [21, 165, 166]. Biological membranes contain

a complex mixture of lipids, which are distinguished on the one side by their acyl chain,


71

according to length and double bonds, and on the other side by their head group,

concerning size and charge. Thus, membranes reveal particular physical properties

depending on the lipid composition, such as thickness, curvature, shape, lateral

pressure, hydration or dielectric constant. Attempts to create a proper model

membrane are guided by these properties along with the goal to maintain the lipid

composition as simple as possible particularly with regard to a better reproducibility.

Thus, synthetic lipids are typically chosen.

In this study, the liposome composition was changed from 67.3mol% DMPC/ 32.7mol%

cholesterol to 90mol% DMPC/ 10mol% DMPA. Both compositions have the zwitterionic

phospholipid 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC, di(C14:0)PC) in

common. Zwitterionic phospholipids are known to be the major lipid species in

membranes [170]. For preliminary work in this lab, cholesterol was used in addition to

modulate protein motions in the membrane [163]. However, it is rather uncommon as a

lipid component to mimic the E.coli membrane. It is known to be part of animal cell

membranes, where it is essential to maintain both membrane structural integrity and

fluidity. In this study, the anionic phospholipid 1,2-dimyristoyl-sn-glycero-3-phosphate

(DMPA, di(C14:0)PA) was used instead of cholesterol. As zwitterionic lipids, anionic

ones are known to be part of all membranes. The presence of anionic lipids is

especially important for the binding of peripheral membrane proteins or protein

segments to the membrane surface. Whereupon positively charged residues interact

electrostatically with the anionic lipids in the bilayer. Furthermore, it has to be

highlighted that 90mol% DMPC in combination with 10mol% DMPA is a widely used

liposome composition. It has been already successfully used for green proteorhodopsin

(GPR) [134, 171], krokinobactereikastus rhodopsin 2 (KR2) [172] and the ATP-binding

cassette (ABC) transporter MsbA [173]. It was also applied for the structure

determination of Anabaena sensory rhodopsin (ASR) by MAS NMR [174]. Especially

the latter is a good example that the usage of 90mol% DMPC/ 10mol% DMPA can

support a high spectral resolution that cannot be taken for granted in MAS NMR.

The two liposome compositions were compared according their fingerprint and

resolution in 2D 13C-13C PDSD spectra. Figure 18 shows a similar fingerprint for DGK in

both lipid compositions, indicating a similar secondary structure. Nonetheless,

differences in the spectral resolution can be observed as well. The enlargement of the

representative region in the 2D 13C-13C PDSD spectra (Figure 18) displays clearly a

reduced resolution for 67.3mol% DMPC/ 32.7mol% cholesterol. For instance, the

selected peak features a line width in F1 dimension of 138 Hz and 406 Hz and in F2

dimension of 574 Hz and 1006 Hz for 90mol% DMPC/ 10mol% DMPA and 67.3mol%

DMPC/ 32.7mol% cholesterol, respectively. The reduced resolution implies a lesser


72

homogenous sample in 67.3mol% DMPC/ 32.7mol% cholesterol, which might reflect

the disordering effect of cholesterol for the lipid bilayer in the gel phase. In contrast,

DMPA has possibly a stabilizing effect, especially on the amphiphilic surface helix

(SH), which contains several positively charged residues (Arg9, Lys12, Lys19 and

Arg22) [5] that most likely interact electrostatically with the anionic lipids in the bilayer.

However, kinetic studies by Lee and co-workers indicated that product analogues of

DGK’s physiological forward reaction, such as DMPA, possibly bind to the active site

[106]. Pilot et al. demonstrated that 1,2-dioleoyl-sn-glycero-3-phosphate (DOPA,

di(C18:1)PA) increases the Km of the lipid substrate 1,2-dihexanoylglycerol (DHG,

chain length n=6) [106]. This has been explained by a possible binding of the product

analogue to the active site in competition with the lipid substrate. However, the

superposition of the 2D 13C-13C PDSD spectra of DGK embedded into 90mol% DMPC/

10mol% DMPA and 67.3mol% DMPC/ 32.7mol% cholesterol shows the same

fingerprint with no chemical shift perturbations (CSPs) observable. Cholesterol, which

is anyway rather uncommon for E.coli membranes, is not reported to bind to the active

site of DGK. Thus, the same spectral fingerprint of the two liposome compositions

indicates that DMPA does not bind to DGK’s active site. If this would be the case, DGK

in complex with DMPA would have provided significant CSP’s compared to its pure apo

state. Based on the high sensitivity of NMR signals to the local environment, even

weak ligand binding can be analysed by chemical shift changes. In addition, NMR

features the advantage that changes can be investigated directly and specifically at an

atomic level.

The differences concerning the assumption of the binding of the product analogue to

the active site of DGK might arise from different experimental conditions. For the kinetic

studies by Lee and co-workers, DGK was reconstituted into 80mol% DOPC/ 20mol%

DOPA in a molar lipid-to-protein ratio of 6000:1. This implies a remarkable high molar

DOPA-to-protein ratio of 1200:1. In our study by MAS NMR, a clearly lower molar

DMPA-to-protein ratio of 5:1 was applied. Thus, a very high molar excess of DMPA

compared to DGK seems to be necessary to force its binding to the enzyme. Based on

our MAS NMR data, this is obviously not the case at a 5-fold molar excess of DMPA. In

addition, it has to be mentioned that Lee and co-workers carried out the activity

measurements on unsealed membrane fragments consisting of phospholipids and

detergent (cholate). They were obtained by dilution of the reconstition mixture into

buffer, decreasing the concentration of cholate below its critical micelle concentration.

In contrast, DGK was present in detergent-free liposomes during the MAS NMR

experiments.


73

Due to the high quality of the MAS NMR spectra, the widely and successfully used

90mol% DMPC/ 10mol% DMPA composition was used for all subsequent studies.

Figure 18. Superimposed 2D 13

C-13

C PDSD spectra of U-13

C,15

N-DGK embedded into 90mol%

DMPC/ 10mol% DMPA (black) or 67.3mol% DMPC/ 32.7mol% cholesterol (red), showing a

similar fingerprint. The enlargement of a representative region in the 2D 13

C-13

C PDSD spectra

displays clearly a reduced resolution for 67.3mol% DMPC/ 32.7mol% cholesterol (red). For

instance, the selected peak, P, features a line width in F1 dimension of 138 Hz and 406 Hz and

in F2 dimension of 574 Hz and 1006 Hz for 90mol% DMPC/ 10mol% DMPA (black) and

67.3mol% DMPC/ 32.7mol% cholesterol (red), respectively. The line widths were obtained from

CCPN analysis 2.4.1 [175].

In order to pack more protein into the rotor (volume ≤ 50 μl), which in turn elevates the

sensitivity, the molar protein-lipid ratio was increased stepwise from 1:80 to 1:20. The

starting point of 1:80 was given by previous studies in this lab [163]. However, an

increased protein-lipid ratio should be accompanied with a well-preserved activity of the

membrane protein, when it is incorporated into the liposomes. Thus, the activity was

checked for each sample. Figure 19 shows that a higher molar protein-lipid ratio up to

1:50 does not reduce the activity. If the ratio is further increased up to 1:20, the activity

decreases significantly, which indicates that most likely unfavorable protein-protein

interactions and adverse changes of the protein-lipid-interaction occur. Consequently,

the molar protein-lipid ratio of 1:50 was used for all subsequent studies, which allows to

pack 30% more protein into the rotor compared to the starting point of 1:80.


74

Figure 19. Comparison of the activity of wtDGK reconstituted into DMPC/DMPA with different

molar protein-to-lipid ratios increasing from 1:80 to 1:20. The ratio of 1:50 (grey) is used for all

subsequent studies. 100% activity corresponds to the rate recorded with wtDGK in 90mol%

DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1

mg-1

. Experiments were repeated three times.

The activity was calculated as the mean value. Error bars correspond to standard deviations.

Furthermore, a sucrose gradient (40%-10%) was performed to verify, if DGK is

homogenously reconstituted into the DMPC/DMPA liposomes, which is of key

importance for well-resolved MAS NMR spectra. The sucrose gradient clearly reveals a

homogenous size distribution of the proteoliposomes, since only one sharp band is

visible without any empty liposomes observable (Figure 20).

Figure 20. Sucrose density gradient (40%-10%) centrifugation of empty liposomes (left) and

wtDGK reconstituted in DMPC/DMPA in a molar protein-to-lipid ratio of 1:50 (right), revealing

homogeneous size distribution of proteoliposomes without any empty liposomes observable.


75

To verify the functionality of wtDGK in liposomes, we used a coupled enzyme assay as

described in chapter 2. An average activity of 90 (± 9.9) µmol min-1 mg-1 was obtained

for wtDGK in 90mol% DMPC/ 10mol% DMPA. The measured activities were highly

reproducible from sample to sample and are comparable to values published before [7,

107].

3.2.6 Evaluating the optimized DGK proteoliposomes for MAS NMR application

To further determine the quality of the proteoliposome sample, 1D 13C and 15N cross-

polarization (CP) MAS as well as 2D 13C–13C correlation spectra, namely proton driven

spin diffusion (PDSD) spectra, were carried out. In spite of spectral overlap

characteristically for α-helical proteins, a fine structure is clearly visible even in 1D

spectra, and some single resonances are resolved. For instance, a sharp band at

approximately 107.0 ppm in the 15N CP spectrum corresponds to the 15N glycine

signals (Figure 21a). The 2D 13C-13C PDSD was conducted with short carbon-carbon

mixing (20 ms) to yield one bond correlations between aliphatic atoms (Figure 21b).

The spectrum provides a fingerprint of the sample, evaluating structural homogeneity,

resolution and secondary structure. The high number of well resolved peaks

demonstrates a homogeneous sample preparation.

Figure 21. Evaluation of the proteoliposom sample by MAS NMR. 1D 13

C and 15

N cross

polarization (CP) spectra of U-13

C,15

N-wtDGK reconstituted into DMPC/DMPA, exhibit a good


76

spectral resolution (a). 2D 13

C-13

C PDSD spectrum of U-13

C,15

N-wtDGK reconstituted into

DMPC/DMPA. A mixing time of 20 ms was used to yield one bond correlations between

aliphatic atoms. The spectrum evaluates structural homogeneity, resolution and secondary

structure. The high number of well resolved peaks demonstrates a homogeneous sample

preparation (b).

3.3 Summary

The optimization resulted in a proteoliposome sample, which mimics physiological

conditions to a higher extent than before. The activity could be increased by using wild-

type DGK instead of the quadruple thermostable mutant and by the application of DDM

as detergent instead of DPC. The yield for DGK could be increased by 50%, from 20-

30 mg of DGK per liter E.coli culture to 30-45 mg/l. The purity of wtDGK in DDM

micelles was verified by SDS-PAGE and SEC. Additionally, BN-PAGE and LILBID-MS

showed that DGK appears exclusively in its trimeric form. Moreover, it could be proven

that DGK can be homogenously reconstituted into 90 mol% DMPC/ 10 mol% DMPA

using BioBeads, leading to a fully active protein and even better resolved spectra than

shown before in this lab. Using BioBeads instead of dialysis as reconstitution method

has the positive side effect of reducing the expenditure of time remarkably from two

weeks to two days. Furthermore, by increasing the molar protein-to-lipid ratio, 30%

more protein can be packed into the rotor, yielding to a higher sensitivity. Another

feature of the sample, which is worth to mention, is its long-term stability, being

especially advantageous for recording time-consuming 3D spectra regarding

assignment purposes. Hence, the new optimized preparation protocol is more efficient

and provides a more native-like sample, which delivers high quality MAS NMR spectra,

increasing the value of all subsequent studies. Especially, the good spectral resolution

provides the basis for an assignment of DGK, which is demonstrated in the following

chapter.

Chapter 4: Resonance assignment

77

4 Resonance assignment

4.1 Introduction

For the investigation of structural and dynamical changes defining the catalytic

mechanism of a protein, the assignment of its backbone and side chains is mandatory.

So far, an assignment of a thermostable mutant of DGK in E.coli total lipids was

published by Yang and co-workers using MAS ssNMR [14]. A comparison of the first

13C-13C correlation MAS ssNMR spectrum of wild-type DGK with the 13C chemical shifts

derived from the mutant (BMRB entry: 19754) [14] is shown in Figure 22. Upon first

inspection, it becomes obvious that a transfer of these assignments to the wild type

sample is not possible. The cross peaks from the mutant clearly do not match the

peaks from the wild type. Deviations are most likely caused by inserted mutations.

Additional sources might arise from different experimental conditions. Our experiments

were carried out on DGK reconstituted into DMPC/DMPA under a pH of 7.2 and full

hydration. The thermostable mutant was reconstituted into E.coli total lipids and

measured under a pH of 6.6 and 23 wt% H2O [14].

Figure 22. Comparison of MAS ssNMR data of the wild type with its thermostable mutant. 2D 13

C-13

C PDSD spectrum of U-13

C,15

N-wtDGK, recorded with a mixing time of 20 ms (left).

Enlargement of the selected region in the 2D 13

C-13

C PDSD spectrum (right). The spectrum is

compared with the assignment of the thermostable mutant of DGK gained by MAS NMR [14].

The comparison shows that a transfer of these assignments to the wild type sample is not

possible. The red cross peaks were obtained with CCPN analysis 2.4.1 [175].


78

Thus, we carried out sequential assignment of wild-type DGK. The capability of magic

angle spinning (MAS) solid state NMR for the assignment of membrane proteins in

phospholipids could be demonstrated for the light-harvesting 2 protein, LH2 [176, 177];

sensory rhodopsin II from Natronomonas pharaonis, NpSRII [178]; and green

proteorhodopsin, GPR [124] as well. For few membrane proteins, even complete 3D

structures could be determined in phospholipids, including: the sensory rhodopsin from

Anabaena, ASR [174]; the human chemokine receptor, CXCR1 [179]; the M2 1H

channel from influenza virus [180, 181]; the bacterial inner membrane protein DsbB

[182, 183]; the Mycobacterium tuberculosis cell division protein, CrgA [184] and the

membrane-inserted form of the fd bacteriophage major coat protein [185] (Figure 23).

Figure 23. Examples of membrane protein structures determined in phospholipids by solid state

NMR. (a) Anabaena sensory rhodopsin, ASR: PDB 2m3g, with bound retinal (yellow) [174]. (b)

human chemokine receptor, CXCR1: PDB 2lnl [179]. (c) M2 1H channel from influenza virus:

PDB 2l0j [180, 181]. (d) Bacterial inner membrane protein DsbB: PDB 2leg [182, 183]. (e)

Mycobacterium cell division protein, CrgA: PDB 2mmu [184]. (f) Membrane-inserted form of the

fd bacteriophage coat protein: PDB 1mzt [185]. The figure is adapted from Marassi and Opella

[186].

Prescind from these outstanding achievements, the assignment of membrane proteins

by MAS ssNMR is still a highly challenging task. A first obstacle to overcome is to

produce sufficient amounts of purified, stable and active protein, which is

homogenously reconstituted in an appropriate lipid environment, as demonstrated in

the previous chapter 3. The second obstacle arises from spectral crowding. Membrane

proteins are generally built up of residues that are all constrained in nearly the same

secondary structure, yielding a narrow chemical shift dispersion. A third obstacle

occurs due to the low intrinsic sensitivity of the technique. The intensity of an NMR

signal is proportional to the population difference between the spin states, specified by

Boltzmann statistics. The differences in the population between the spin states are

minor, since they are separated by only weak energy differences in the radio-frequency

range. This turns solid state NMR spectroscopy into a relatively insensitive technique


79

compared to other spectroscopic methods, requiring long signal accumulation times to

gain spectra with a sufficient signal-to-noise ratio. The fourth obstacle addresses the

data analysis for the achievement of a reliable assignment to large extent, which poses

a highly demanding and time-consuming task.

In this chapter, a strategy is presented, which helps to diminish these obstacles and

enables a nearly complete assignment of wild-type DGK in lipid bilayers.

4.1.1 Applied isotope labelling strategy

A repertory of isotope labelling techniques, ranging from uniform to site specific, were

developed for in vivo and in vitro expression systems. For uniform labelling, the

expression medium contains a sole carbon source, such as glucose or derivatives

(glycerol, acetate, pyruvate, succinate) and a sole nitrogen source, like ammonium

salts (chloride, nitrate, sulfate). It offers a maximum of isotope labelled sites in one

single sample. However, based on the presence of a dominating single type of

secondary structure (α-helix or β-strand) and the natural repetitiveness of hydrophobic

amino acids, the spectra suffer from spectral overlap, causing ambiguities in data

analysis. Spectral crowding can be, for instance, reduced by reverse labelling, which is

complementary to other labelling strategies, implying selective unlabelling of specific

amino acids. The expression medium, including 13C-glucose and 15NH4Cl, is

supplemented with unlabelled amino acids, containing 12C and 14N. The high presence

of the unlabelled amino acids leads to their direct incorporation into the protein

sequence during expression. The cells do not need to synthesize these amino acids

from the labelled precursors. This way, they are undetectable for NMR, resulting in a

significant decrease of spectral crowding. This labelling approach is rather inexpensive

compared to others. Additionally, same growth conditions as for uniform 13C/15N

labelling can be applied, simplifying its application. For the development of reverse

labelling, following aspects were taken into account: Based on the amino acid

sequence, a first selection was carried out on the relative abundance of each amino

acid type. Usually, it proves to be difficult to assign every amino acid of a type with high

abundance. Transmembrane segments are normally rife with hydrophobic amino acids,

for which spectral overlap is known. Thus, key hydrophobic residues are usually

chosen for unlabelling [178, 187]. In contrast, extramembrane loop regions typically

consisting of polar or charged amino acids are well resolved. Aromatic residues, which

are often located at the membrane interface, could be considered for reverse labelling

as well, since their resolution can be compromised by ring currents. One issue that has


80

to be taken into account during reverse labelling or selective unlabelling, is scrambling.

During the amino acid biosynthesis in E. coli, metabolic conversions occur, in which

amino acids are transformed into one another. During these metabolic reactions

labelled nuclei could be transferred to amino acids that should not be labelled, leading

to unwanted signals. This issue can be overcome by supplementing the culture

medium with high amounts of all twenty amino acids, since biosynthetic enzymes are

regulated by feedback inhibition. Another solution is provided by specifically modified

host strains, in which scrambling of certain amino acids is controlled with defective

enzymes. One such strain is e.g. E.coli CT19 featuring a reduced transaminase activity

[188]. The easiest option to avoid or at least reduce scrambling is the use of amino

acids, which are end products of the bacterial metabolic cycle and therefore not

converted into other amino acids. These amino acids are Arg, Cys, His, Ile, Leu, Lys,

Met, Phe, Pro, Trp, Tyr and Val. The other amino acids (Ala, Asn, Asp, Glu, Gln, Gly,

Ser and Thr) should be avoided due to the risk of scrambling [189]. In this study,

uniform 13C,15N and reverse labelling was applied. For reverse labelling, the most

frequent hydrophobic amino acids: Ile, Leu and Val (Table 13), which are not subjected

to scrambling, were chosen.

Table 13. Amino acid composition of wild-type DGK. The numbers in brackets belong to the

His6-tag and linker. The most frequent hydrophobic amino acids: Ile, Leu and Val, which were

chosen for reverse labelling, are highlighted orange.

amino acid type number

Ala 18

Arg 6

Asn 4

Asp 4

Cys 2

Gln 1

Glu 6 (+1)

Gly 8 (+1)

His 2 (+6)

Ile 15

Leu 12 (+1)


81

Lys 3

Met 3 (+ 1)

Phe 3

Ser 8

Thr 5

Trp 5

Tyr 2

Val 14

4.1.2 Applied strategy for improvements of the sensitivity and resolution

4.1.2.1 High magnetic fields

Improvements of the sensitivity can be already obtained by high magnetic fields, since

the population difference increases with the magnetic field strength (B0). In practice, the

signal-to-noise ratio of modern NMR spectrometers scales approximately with B01.5.

Thus, an enhancement in sensitivity of about 2 is gained, when a 18.7 T (800 MHz)

spectrometer is compared with a 11.7 T (500 MHz) instrument. With increasing field

strength, the spectral resolution improves as well, since line-broadening mechanisms

involving dipolar or scalar couplings only arise in higher-order perturbation terms.

Hence, a high magnetic field is a prerequisite for assignment studies. The assignment

of wtDGK, performed here, was carried out on a Bruker wide bore Avance III solid state

NMR spectrometer with a 1H frequency of 850 MHz.

4.1.2.2 Paramagnetic doping in combination with an E-free probehead

Further improvements can be achieved by the enhancement of the signal-to-noise ratio

per unit time. Nearly 90% of the measurement time is needed to restore the 1H

Boltzmann equilibrium after each cross-polarization (CP) step. The required recycle

delay matches 4–5-times the 1H longitudinal relaxation time (1H-T1), which is generally

needed to restrict the probehead duty cycle and to prevent surplus sample heating, as

well. One possible route for reducing 1H-T1 and thus enhancing the signal-to-noise

ratio per unit time is doping the sample with paramagnetic relaxation agents [190].

Paramagnetic agents feature unpaired electrons, which have a remarkable influence

on the NMR spectra due to the electronic magnetic moment, being about 650 times


82

higher compared to protons. Random variations of electron spin-nuclear spin

interactions cause paramagnetic relaxation enhancement (PRE), allowing faster data

acquisition schemes. For the use in protein NMR, the unpaired electrons must reside in

chemically non-reactive compounds that are stable in aqueous solution, e.g. in a

chelate complex. Previously in this lab, Ullrich et al. could demonstrate that Gd3+-DOTA

is the most potent relaxation agent for membrane proteins so far [12]. Gd3+ contains

seven unpaired electrons. It does not cause pseudo-contact shifts (PCS), since its

magnetic susceptibility tensor is isotropic, but the size of the isotropic tensor

component and its slow electronic relaxation rates cause large paramagnetic relaxation

enhancement (PRE). PRE effects by Gd3+ are larger than those reported for nitroxides

and comparable to Cu2+ and Mn2+ (reviewed in [191]).

However, higher probehead duty cycles could cause sample damage due to an

increased sample heating. Paramagnetic doping in combination with fast MAS (≥ 40

kHz) [192-194] or sample deuteration [195] with low power decoupling would represent

possible solutions. In this study, sample heating was avoided by using a MAS

probehead with reduced E-field [196], which was specifically designed for

measurements on a Bruker wide bore Avance III solid state NMR spectrometer with a

1H frequency of 850 MHz. It allows utilizing protonated samples, high power

decoupling, moderate sample spinning rates and MAS rotors of conventional size.

4.1.3 Applied assignment procedure

DGK consists of transmembrane α-helices, connected by a cytosolic and a periplasmic

loop, and an amphiphilic surface helix. This architecture results in very different

molecular dynamic time scales and amplitudes within the protein [149]. Motions at

different time scales and amplitudes are crucial for the function of membrane proteins,

which are embedded in lipid bilayers and exposed to water at the water-membrane

interface. Immobile domains, usually transmembrane regions, are molecular segments

with smaller amplitude motions. They are not sufficient to fully average anisotropic

interactions (such as HH, HC, or HN dipole couplings) and can be observed by dipolar

coupling based cross-polarization (CP). In this study, the sequential assignment of

immobile domains was performed by dipolar coupling based 3D heteronuclear

correlation experiments: NCACX, NCOCX and CONCA. In contrast, highly mobile

regions, usually termini or loops, feature fast and large amplitude fluctuations on time

scales of <10-5 s. They can be selected by a refocused INEPT (insensitive nuclei

enhanced by polarization transfer) [128] step based on scalar couplings, leading to


83

solution-state-like spectra due to efficient molecular averaging of anisotropic

interactions. Thus, for the completion of the assignment of wtDGK, scalar coupling

based 2D experiments, 1H-13C/15N HETCOR and 13C-13C TOBSY, were carried out to

detect highly mobile residues. All dipolar coupling based 3D and scalar coupling based

2D experiments are explained in detail in chapter 1 under the bullet 1.3.2.

4.1.3.1 Sequential assignment of immobile domains

The sequential assignment process includes spin system identification, assignment of

the spin system to the amino acid type, linking of the spin systems, and mapping them

to the protein amino acid sequence. For the identification of the spin systems, it has to

be taken into account that DGK contains a total of 121 residues, of which 41 are not

labelled in the reverse labelled DGK-I,L,V sample (15 isoleucines, 12 leucines, and 14

valines). Thus, a maximum of 121 and 80 spin systems in the uniformly and reverse

labelled sample, respectively, are expected. However, it has to be considered as well

that not all of them might be visible due to higher dynamics. Therefore, scalar coupling

based experiments represent a proper solution.

For the identification of the amino acid type, the BMRB database [159] was used. The

amino acid type was largely assigned through the measurement of the chemical shifts

of the side-chain carbon atoms. Here, glycines represent very good starting points for

the assignment procedure, since they have a characteristic Ca chemical shift of ~45

ppm. The amino acids Ala, Ser and Thr can be easily identified as well due to their

specific Cb chemical shift. For these residues (with short or unique side chains), short-

range one- and two-bond correlations, are sufficient to identify the amino acid type.

With respect to the average Ca and Cb chemical shift range, the other amino acids can

be distinguished into 4 groups:

Group 1 (Ca ~57 ppm, Cb ~30 ppm): Lys, Arg, Gln, Glu, Met, His, Trp, Cys (reduced)

Group 2 (Ca ~57 ppm, Cb ~40 ppm): Asp, Asn, Phe, Tyr, Leu, Cys (oxidized)

Group 3 (Ca ~60 ppm, Cb ~40 ppm): Ile (very specific side chain pattern)

Group 4 (Ca ~65 ppm, Cb ~30 ppm): Val

Their unambiguous identification requires the detection of three- or four-bond carbon–

carbon connectivities.


84

In this study, an 3D NCACX experiment with a DARR mixing time of 50 ms was used to

establish these long-range intra-residue correlations, enabling the identification of as

many amino acids by type as possible. For the 3D NCOCX experiment, even a DARR

mixing time of 100 ms was used to enable long-range inter-residue correlations. For

the linking of the spin systems and mapping them to the protein amino acid sequence,

the 3D NCOCX and 3D CONCA experiment are used, establishing inter-residue

correlations between nitrogen atoms and carbon atoms of the preceding residue. The

NCO/CON connects the 15N[i] with 13CO[i-1] through the peptide bond. The 3D CONCA

experiment is essential in the spectral assignment strategy, because it allows matching

cross-peaks in 3D NCACX and 3D NCOCX experiments based on at least two shifts,

greatly increasing the reliability of the matching.

4.1.3.1.1 Automatic assignment of immobile domains by ssFLYA

Manual resonance assignment is generally considered as very arduous. It requires a

considerable amount of time by an experienced spectroscopist. Therefore, automatic

assignment algorithms were developed to fasten the assignment procedure.

Additionally, they can be used as a verification of the already performed assignments.

However, manual resonance assignment persist the standard in solution NMR and the

almost exclusive procedure in solid state NMR due to features or imperfections of

experimental NMR spectra, such as broad linewidth, signal overlap, low sensitivity and

spectral artefacts. These imperfections are usually more pronounced in ssNMR

spectra, increasing difficulties concerning an automated assignment. Only very few

automated algorithms have been developed for solid state NMR resonance

assignment. One algorithm [197] is based on the AutoAssign package for solution NMR

[198] and was used to assign the backbone resonances of GB1 with peak lists from 3D

NCACX, CAN(CO)CA and 4D CANCOCX experiments as input. Another approach

[199, 200] enables the assignment of backbone and side chain resonances by

analysing random combinations of spectra with random dimensions. For the shown

examples, the algorithm uses input peak lists from NCACX, NCOCA and CONCA

spectra. However, this approach holds the drawback that signals must be grouped

together to a spin system and assigned to atom types (e.g., N, Ca, Cb, Cg).

Additionally, possible assignments of the amino acid type must be predefined before

running the algorithm. In this study, ssFLYA, a generally applicable algorithm for

automated backbone and side-chain resonance assignment of protein solid state NMR

spectra [13], is applied. It is predicated on FLYA, an automated resonance assignment


85

algorithm for solution NMR [153] and integrated in the software package CYANA [154,

155]. The ssFLYA algorithm uses solely the protein sequence and unassigned peak

lists from any combination of multidimensional ssNMR spectra. So far, only

microcrystals, such as ubiquitin and the Ure2 prion C-terminal domain as well as

amyloids like HET-s(218–289) and a-synuclein have been successfully assigned by

ssFLYA. In this study, we tested its principal applicability for membrane proteins and

used it for validation of the manually obtained assignments of DGK. For this purpose,

ssFLYA used the dipolar coupling based 3D experiments (NCACX, NCOCX, CONCA).

4.2 Results and Discussion

4.2.1 Spectral resolution and isotope labelling

With the help of the high magnetic field (1H frequency of 850 MHz) and an optimized

sample preparation (chapter 3), well-resolved NMR-spectra of high signal-to-noise ratio

were observed by MAS ssNMR. Thus, the 13C and 15N assignments were mainly

carried out using uniformly labelled samples (U-13C,15N-wtDGK). Here, 15N and 13Ca

signals in the 2D NCA spectra exhibit comparable good average linewidths of

approximately 105 and 185 Hz, respectively (Figure 24). Residual ambiguities could be

resolved by reverse labelling of isoleucine, leucine and valine (U-13C,15N-wtDGK-I,L,V).

Compared to the uniform 13C,15N-labelled sample, which offers a maximum of isotope

labelled sites with 132 labelled residues, the reverse labelled sample contains only 90

labelled residues. Thus, the number of signals in the spectrum of the reverse labelled

sample should be theoretically reduced by 32%, which would be valid for the case that

all labelled residues are visible in the spectrum. Figure 24 shows that in the spectrum

of the reverse labelled sample, the peaks for Ile, Leu and Val are clearly missing

compared to the uniform labelled sample. Apart from that, the two spectra do not

remarkably differ from each other. Overall, this indicates a successful reverse labelling.


86

Figure 24. Comparison of the 2D NCA spectra of uniform labelled U-13

C,15

N-wtDGK (black) and

reverse labelled U-13

C,15

N-wtDGK-I,L,V (green). The 15

N and 13

Ca signals of the uniform

labelled U-13

C,15

N-wtDGK exhibit already comparable good average linewidths of approximately

105 and 185 Hz, respectively. To resolve residual ambiguities, Ile, Leu and Val are specifically

unlabelled in the reverse labelled sample. The inscriptions for Ile, Leu and Val are labelled in

red and for the other amino acids in black. In the 2D NCA of the reverse labelled sample, the

peaks for Ile, Leu and Val are clearly missing as expected. Apart from that, the two NCA spectra

do not remarkably differ from each other.

4.2.2 Paramagnetic doping in combination with an E-free probehead

The obstacle of low intrinsic sensitivity of MAS ssNMR, leading to long signal

accumulation times to gain spectra with sufficient signal-to-noise ratio, could be

diminished by paramagnetic doping with Gd3+-DOTA (Ullrich, Holper et al. 2014) and

an E-free probehead. The E-free probehead enables short recycle delays of 0.8 s,

saving ~3x of the measurement time compared to the standard probehead with a

recycle delay of 2.5 s (Figure 25).


87

Figure 25. 15

N CP spectra of U-13

C,15

N-DGK. The spectra were recorded with an E-free (green)

and with a standard (black) 3.2 mm triple-resonance HCN MAS probehead (Bruker). In both

cases 128 scans were applied. The E-free probehead enables to use a recycle delay of 0.8 s,

saving ~3x of the measurement time compared to the standard probehead with a recycle delay

of 2.5 s. The E-free probehead was custom-built and is still under development.

Next to paramagnetic doping in combination with the E-free probehead, different

strategies for a faster signal build-up are available, such as optimum control pulse

sequences [201] and non-uniform sampling [202-204].

4.2.3 Sequential assignment of immobile domains

Sequential assignment of wild-type DGK was carried out using a combination of dipolar

coupling based 3D experiments, NCACX, NCOCX and CONCA. A representative

sequential walk linking I26 to A29 is shown in Figure 26. By analyzing the three

heteronuclear 3D correlation experiments, we are able to identify, assign and to link the

spin systems, which allow to map them to the protein amino acid sequence. Figure 26

shows 2D planes extracted from the 3D NCACX and NCOCX spectra with a DARR

mixing time of 50 ms and 100 ms, respectively, enabling the detection of long-range

intra- or inter-residue correlations of three- or four-bond carbon-carbon connectivities.

Furthermore, 2D planes of the 3D CONCA experiment are shown permitting the

matching of cross-peaks in the NCACX and NCOCX spectra based on two shifts. Thus,

the CONCA spectrum considerably increases the reliability of the matching. Each set of

three spectra stands for a Cx[i−1]–N[i]–Cx[i] spin system. For example, the N27

NCACX peaks are linked to the I26-N27 CONCA peak through the same N and Ca and


88

the I26 NCOCX peaks are connected to the I26–N27 CONCA peak via the same N and

CO, thus generating a Cx[i−1]–N[i]–Cx[i] system. It is associated with the prior system

through all carbon shifts of I26 that are visible in both NCACX and NCOCX spectra. In

addition to CO[i−1]–N[i]–CA[i] correlations in the CONCA spectrum, we also detected

CO[i−1]–N[i]–Cb[i] correlations. Although these correlations are not necessary for

building spin systems and for the linking process, they provide a quality control in

validating assignments. So-called shuffled peaks in the NCACX spectrum serve as a

validation as well. They originate from N[i]-Ca[i]-CO[i-1]-Cx[i-1] or N[i]-Ca[i]-CO[i]-

Ca[i+1]-Cx[i+1], such as 108AlaN–108AlaCa-107ValCO-107ValCa-107ValCb-

107ValCg1-107ValCg2 or 97GlyN–97GlyCa-97GlyCO-98SerCa-98SerCb-98SerCO,

among others. These peaks can be easily distinguished, since they are much weaker

than those emanating from the one- or two-bond N[i]-Ca[i]-Cb[i] transfers within the

same residue. In addition, they often originate from amino acids with short side chains,

such as glycines, alanines and serines. Shuffled peaks are also visible in the NCOCX

spectrum. They are from the N[i]-CO[i-1]-Ca[i]-Cx[i] type as it is e.g. observable in the

2D plane for E28, which includes, among the peaks for E28, the peaks for Ca and Cb

of A29 as well (Figure 26).


89

Figure 26. Resonance assignment of U-13

C,15

N-wtDGK based on a set of 3D NCACX, NCOCX

and CONCA spectra. A representative sequential walk from I26 to A29 is shown. Each set of

three spectra represents a Cx[i−1]–N[i]–Cx[i] spin system. For example, the N27 NCACX peaks

are connected to the I26-N27 CONCA peak via the same N and Ca. The I26 NCOCX peaks are

linked to the I26–N27 CONCA peak through the same N and CO, resulting in a Cx[i−1]–N[i]–

Cx[i] system, which is linked with the preceding system through all carbon shifts of I26 that are

visible in both NCACX and NCOCX spectra. The assignments are depicted by lines.

The backbone and most side chain carbon and nitrogen resonances could be assigned

for ~82% of the residues (= 99 residues), from which 73 residues (~74% of the

assigned residues) are completely assigned. Assigned resonances are highlighted in a

2D NCA spectrum (Figure 27).


90

Figure 27. 2D NCA spectrum of U-13

C,15

N-DGK with assigned peaks labelled.

4.2.3.1 Automatic assignment of immobile domains by ssFLYA

For validation, ssFLYA, a generally applicable algorithm for the automatic assignment

of protein solid state NMR spectra [13], was applied. ssFLYA was performed by Dr.

Sina Kazemi of the research group of Prof. Dr. Peter Güntert (Institute for Biophysical

Chemistry, Goethe University Frankfurt am Main). So far, it has been used for

microcrystals and amyloids [13]. In this study, its principal applicability for demanding

systems as membrane proteins could be demonstrated for the first time. Overall, 91.5%

of the backbone and 89.1% of all (backbone + side chains) assignments could be

confirmed by ssFLYA, verifying the manually obtained assignments. A full overview of

the assignment is provided in Figure 28. However, few incorrect assignments with

respect to the manual assignment appeared as well. They occur for residues close to

the termini or loops, which are known to be dynamic, yielding very weak signals (G15-

W18, S84, E88) or are probably not observable in the spectra at all (E85-H87, S118-

G121). Remote side chain atoms feature less intense signals and thus lead to incorrect

assignments as well. Further incorrect assignments could originate from spectral


91

artifacts or missing peaks due to overlapped regions. Thus, for the automatic

assignment of DGK by ssFLYA, careful manual peak picking with knowledge about the

spectra is a prerequisite. The percentage of strong assignments depended on the

quality of the input peak lists. Known spectral artifacts like folded peaks had to be

removed from the input peak lists, since the algorithm could not distinguish between

folded peaks and correct ones, leading to a disturbed calculation and hence to overall

incorrect assignments. Especially, the peaks for R32, R81, R92 and K94, arising from a

magnetization transfer starting from the side chains, which are additionally folded in,

led to incorrect assignments by ssFLYA. However, ssFLYA features the big advantage

of accelerating the time-consuming assignment process. It is able to perform an

assignment calculation within approximately 10 min, if 20 CPU cores are availabe.

Thus, an implementation of ssFLYA into the manual assignment procedure lead to a

faster and more reliable assignment.

Figure 28. Automated resonance assignment by ssFLYA confirms 91.5% of the backbone and

89.1% of all (backbone + side chains) assignments obtained manually. Assignments are

classified as strong, if ≥ 80% of the individual chemical shift values from 20 independent runs of

the algorithm differ by less than 0.55 ppm from the consensus value (strong colors). Other

assignments by ssFLYA are graded as weak (light colors). From other studies by ssFLYA, they

are known to be erroneous for 39 – 72% [13]. Each assignment for an atom is symbolized by a

colored rectangle: green - assignment by ssFLYA agrees with the manual reference assignment

within a tolerance of 0.55 ppm; red - assignment does not match with the reference; blue -

assigned by ssFLYA, but not manually; black – assigned manually, but not by ssFLYA. The

second row illustrates backbone assignments for N, Ca, and CO. The third to eighth row

represent the side chain assignments. For branched side chains, the relevant row is subdivided

into an upper part for one branch and a lower part for the other branch. ssFLYA was performed

by Dr. Sina Kazemi of the research group of Prof. Dr. Peter Güntert (Institute for Biophysical

Chemistry, Goethe University Frankfurt am Main). He also kindly provided this figure.


92

4.2.4 Tentative assignment of highly mobile regions

Since not all residues could be detected in CP-based experiments, scalar coupling

based spectra were recorded as well. Comparable to the data reported for the

thermostable DGK mutant [149], some mobile residues could so be monitored. 1H-15N

HETCOR, 1H-13C HETCOR and 13C-13C TOBSY spectra (Figure 29) were recorded,

showing well-resolved peaks of “solution-state”-like quality due to motions on the

submicrosecond time scale. Most of the peaks could be assigned to types of amino

acids based on the BMRB database [205], such as Ala, Arg, Asn, Gly, His, Ile, Leu,

Lys, Phe and Thr. Hence, these peaks could be tentatively assigned to the two termini

and the cytosolic loop between helix 2 and 3, where residues occur that could not be

detected by dipolar coupling based experiments. Peaks for arginine and lysine could be

assigned unambiguously to Arg9 and Lys12 by exclusion, as all other arginines and

lysines are already assigned by experiments based on dipolar coupling.


93

Figure 29. 2D scalar coupling based 1H-

15N HETCOR (a),

1H-

13C HETCOR (b) and

13C-

13C

TOBSY (c) of U-13

C,15

N-DGK with tentative assignments. All residues, which could not be

detected and assigned by dipolar coupling based experiments are considered as possible

candidates for detection by experiments based on scalar coupling. INEPT and TOBSY were

applied for 1H-

15N or

1H-

13C heteronuclear polarization and

13C-

13C homonuclear mixing,

respectively. Peaks for Arg9 and Lys12 are labelled green, as they could be assigned

unambiguously. Peaks for the aromatic rings were folded in the indirect dimension to save

measurement time. Amino acids that refer to the His-tag are labelled by ‘tag’.


94

4.2.5 Summary of the assignment

With the careful optimization of the sample preparation (chapter 3) and the used NMR

strategy, 84% of the residues (= 101 residues) located in transmembrane and

extramembranous regions could be assigned by dipolar and scalar coupling based

experiments. The assigned chemical shifts have been deposited to BioMagResBank

(BMRB entry 27570). All assignments are plotted in Figure 30a and the assigned

residues are mapped on the topology of DGK (Figure 30b). Additionally, they are

itemized in Table S5. The residues of the termini and the cytosolic loop were not or

only tentatively assigned, since they were too mobile for dipolar- but not mobile enough

for scalar coupling based experiments.

Figure 30. Resonance assignment of DGK. (a) Each assignment for an atom is symbolized by a

blue rectangle: The second row illustrates backbone assignments for N, Ca, and CO. The third

to eighth row represent the side chain assignments. For branched side chains, the relevant row

is subdivided into an upper part for one branch and a lower part for the other branch. This figure

was kindly provided by Dr. Sina Kazemi of the research group of Prof. Dr. Peter Güntert

(Institute for Biophysical Chemistry, Goethe University Frankfurt am Main). (b) The assigned

residues are mapped on the topology plot of DGK. The plot was created with respect to the X-


95

ray structure of DGK (PDB 3ZE5) [5] and refined by CSI values obtained from chemical shifts

(Table S5). The membrane is depicted by two solid black lines. 84% residues of DGK were

assigned by dipolar and scalar coupling based experiments.

4.2.6 Secondary structure analysis

Based on the resonance assignment, the secondary structure of DGK was calculated

by chemical shift index (CSI) analysis. Figure 31a displays the CSI Δδ as a function of

the residue number, in which Δδ stands for the deviation of the experimentally

determined MAS NMR chemical shifts (exp) for Ca and Cb from their random coil

standard chemical shifts (rc) according to the following equation [206]: Δδ =[δCa(exp)-

δCa(rc)] - [δCb(exp)-δCb(rc)]. For Gly residues and residues without any assignment of

Cb, only Ca secondary shifts were considered. Strongly positive (≥ 1.5 ppm) values of

the CSI indicate an α-helical structure, whereas negative or near-zero values imply

deviations from helicity. Figure 31a shows the comparison of the secondary structures

of wild-type DGK, its thermostable mutant [14] determined both by MAS NMR and the

crystal structure [6] of wtDGK (PDB 3ZE4, chain A). All three secondary structures

exhibit substantial similarities, particularly regarding the high α-helical content.

However, there are some differences, too. In contrast to both MAS NMR secondary

structures, the crystal structure features small deviations around the flexible regions:

the interhelical turn (T) between helix 1 (H1) and the surface helix (SH), the periplasmic

loop (PL) between helix 1 (H1) and helix 2 (H2), as well as the cytoplasmic loop (CL)

between helix 2 (H2) and helix 3 (H3). In subunit A of the crystal structure, the position

of T and PL is slightly displaced upstream by two residues in comparison with the MAS

NMR structures. Furthermore, T is one residue longer and PL one residue shorter in

the X-ray structure than in the MAS NMR structures. Regarding the position and/or

length of the CL, all three structures differ from each other. CL is shifted from the

residues 83-87 in the MAS NMR structure of wtDGK to the residues 81–85 of the

thermostable mutant and to the residues 83–90 (subunit A) of the X-ray structure.

However, it has to be endorsed that the positions and lengths of the non-helical

structures even vary between the three different subunits A, B and C within the X-ray

structure itself [5, 14]. Especially the CL ranges from residue 83 to 90 (subunit A), 86 to

91 (subunit B) and 82-87 (subunit C) within one trimer.


96

Figure 31. Secondary structure analysis based on the chemical shifts. The chemical shift index

(CSI) Δδ is derived from the difference between the experimentally determined MAS NMR

chemical shifts (exp) for Ca and Cb and their random coil standard chemical shifts (rc)

according to Δδ =[δCa(exp)-δCa(rc)] - [δCb(exp)-δCb(rc)] [206]. For Gly residues and residues

without any assignment of Cb, only Ca secondary shifts were used. Strongly positive

(≥ 1.5 ppm) values of the CSI imply an α-helical structure, whereas negative or near-zero values

indicate deviations from helicity. (a) The secondary structure of wild-type DGK determined by

MAS NMR is compared with the MAS NMR structure of the thermostable mutant [14] and the X-

ray structure of wtDGK (PDB 3ZE4, chain A) [6]. Rectangles represent α-helical regions

involving the surface helix (SH) and the three transmembrane helices (H1-3), whereas solid

lines symbolize deviations from helicity including the interhelical turn (T), the periplasmic (PL) as

well as the cytoplasmic loop (CL). Residues that were not resolved by ssNMR or by X-ray

crystallography are depicted by dashed lines. Disparities between the three secondary

structures are highlighted in green. (b) The 2D NCACX spectrum of U-13

C,15

N-DGK shows all

assigned glycines. The regions for helical and random coil (rc) secondary structure are

coloured. Gly83 and Gly91 are labelled bold, since the DGK X-ray structure exhibits

asymmetries for both residues: Both were observed within a helical and a random coil structure

[5]. These asymmetries are not detectable by MAS NMR.

4.2.7 DGK forms a symmetric trimer in its apo state

During the assignment procedure, no systematic peak doublets or triplets were

observed, indicating that the DGK trimer adopts a symmetric conformation in its apo

state. This is already to some extent visible in the 2D NCA spectrum, e.g. for well-

resolved peaks, such as Gly20, Asn27, Ala29, Glu88, Ala99 and Trp112 (Figure 27).

This is opposite to the X-ray structure of nucleotide-free DGK, possessing asymmetries

in the secondary structure between the three subunits as already mentioned above [5,


97

14]. The DGK X-ray structure [5] features asymmetries for instance for Gly83 and

Gly91, which were determined both within a helical and a random coil structure in the

DGK trimer. By MAS NMR, no hints for more than one peak for Gly83 and Gly91 are

visible (Figure 31b). The deviations regarding the conformational symmetry of DGK

between MAS NMR and X-ray data might occur due to different experimental

conditions. In this connection, particularly crystal packing may be a potential source for

structural asymmetries [207]. Additionally, radiation damage such as decarboxylation of

glutamates and aspartates could lead to the loss of salt bridges [208]. This in turn may

lead to partial instabilities and thus partial slow movements within the protein molecule

despite the frozen state [208], causing conformational inhomogeneities. A possible salt

bridge, which might be partly lost, would be between Arg81 and Glu88 as shown in

chapter 6 (Figure 42a), whereupon Glu88 is supposed to be in the cytosolic loop [5]. In

general, a high mobility of the loop regions, which is reflected by higher B-factors and

chain displacements [5] may cause ambiguities regarding the definition of loop position

and length.

4.3 Outlook

4.3.1 Further labelling strategies

For an unambiguous assignment of larger regions, suffering from spectral overlap,

and/or the determination of distance restraints, position-specific labelling might be a

proper solution. It is based on [1,3-13C]-glycerol or [2-13C]-glycerol, which replace

glucose in the expression medium [209]. Amino acids that originate from the glycolytic

and pentose phosphate pathways (Ala, Cys, Gly, His, Leu, Phe, Ser, Trp, Tyr, Val)

have all sites either 13C or 12C labelled in almost all cases according the “all-or-nothing”

principle. The remaining amino acids are synthesized from precursors that participate

in the citric acid pathway and cause non-random mixtures of isotopomers (mixed

labelling). This way, amino acids can be allocated to sub-groups, which show similar

labelling patterns. The assignment strategy uses these characteristic labelling patterns

for the different amino acid types to identify spin systems in the spectra of [1,3-13C]-

glycerol, [2-13C]-glycerol and [U-13C]-labelled samples. These spin systems are then

linked and mapped to the amino acid sequence of the respective protein. [1,3-13C]- and

[2-13C]-glycerol labelling reduces signal overlap and improves spectral line width. The

applicability of this assignment strategy has been demonstrated for the microcrystalline

chicken α-spectrin SH3 domain (62 residues) [209], the αB-crystallin (175 residues)

[210] and the outer membrane protein G, OmpG (281 residues) [211]. Other extensive


98

selective labelling schemes utilized [1,4-13C], [2,3-13C], and [1,2,3,4-13C] succinic acid

based samples [212, 213]. Fractional [U-13C]-glucose labelling [214, 215], labelling with

[1-13C]-glucose [216] or [2-13C]-glucose [217] represent alternative labelling approaches

as well. They might be suitable for proteins with specific amino acid type compositions.

For a detailed look on certain residues, still suffering from signal overlap, selective

labelling (including labelling of unique pairs) could be used. Here, single isotope

labelled amino acid(s) are added to the growth medium, which are directly incorporated

in the protein during expression in E.coli. One major drawback arises from amino acid

scrambling. This decreases the labelling efficiency on the target site and leads to

additional signals in the spectrum, complicating data analysis. Here, as mentioned

above, amino acids, which are end products of the bacterial metabolic cycle and

therefore not converted into other amino acids, have to be used to reduce scrambling

to a minimum. This restricts the applicability of selective labelling on following amino

acids: Arg, Cys, His, Ile, Leu, Lys, Met, Phe, Pro, Trp, Tyr and Val.

In order to still apply selective labelling on the other amino acids without the risk of

scrambling, E.coli based cell-free expression could be tested as an alternative to in

vivo expression [218, 219]. It allows almost any amino acid labelling scheme due to the

lack of a cellular metabolism as a source of amino acid scrambling [220]. Furthermore,

this in vitro system exhibits an attractive option, since high yields can be obtained while

only low amounts of expensive isotope labelled amino acids are needed. Additionally,

experimental parameters such as pH, redox potential or co-factor dependence can be

better controlled compared to in shake flask expression [218, 219]. Another advantage

of this method is its highly timesaving character: The cell-free produced protein

typically needs no further purification as compared to in vivo expression [220].

However, protein expression is a highly complex, evolutionarily optimized procedure.

Thus, there exist a risk for certain systems that only little amount can be obtained by

cell-free expression or the optimization process takes too much time. In these cases,

1H detection in combination with ultra-fast MAS might be an answer. It allows the

investigation of sub-milligram amounts of protein (0.7 mm probe, ~0.5 mg sample) as

explained in the following section.

4.3.2 Perspective: 1H detection in combination with ultra-fast MAS

Since molecular tumbling is suppressed in membrane proteins, their NMR spectra are

broadened by strong anisotropic interactions. Moderate magic angle spinning (MAS)

[221, 222] in combination with high power decoupling [223] is applied to average out


99

these interactions and to re-establish high resolution for low gamma nuclei like 13C or

15N. As demonstrated in this study, this approach is very successful and commonly

used. However, protons would offer a much greater detection sensitivity than those of

13C or 15N, since they feature a natural abundance of more than 99.9% and a high

gyromagnetic ratio. Unfortunately, the linewidths of protons remain immensely broad at

moderate MAS frequencies (10–20 kHz), because of the strong inter-proton dipolar

couplings. Remarkable engineering progress [224-228] and improvements of sample

preparations [229-232] have facilitated 1H-detection in many systems with limited

mobility, and have turned it to an important technique for increasing sensitivity and

resolution these days. 1H detection in ssNMR is enabled by very fast spinning rates

(>40 kHz) and high magnetic fields, in order to diminish homogeneous line broadening

by suppressing the large network of strong homonuclear 1H-1H dipolar couplings [181,

233-242]. Higher spectral resolution can be obtained partly by proton dilution strategies

as well, such as perdeuteration and back-protonation at the exchangeable sites [231].

Admittedly, perdeuteration might be problematic during protein expression, because of

anemic growth in deuterium oxide, which sometimes is even incompatible with protein

expression, like for example in mammalian cells. When feasible, it allows reintroduction

of 1H species only at sites that are exchangeable and accessible to solvent. This

excludes the large hydrophobic transmembrane regions from analyses by 1H-detected

experiments [243]. For certain systems, unfolding and refolding of membrane proteins,

would be a solution. This allows the incorporation of 1H species in transmembrane

regions. However, such protocols are rare and not applicable to all proteins [181, 234,

241, 244]. In order to overcome this obstacle at least in part, membrane proteins are

expressed in H2O in the presence of deuterated 13C glucose, leading to 1H/2H species,

which are homogeneously distributed in both water-accessible and inaccessible

regions [245]. But this labelling strategy holds the drawback of poorly resolved 13C

resonances from side-chain moieties, which are essential for structure determination.

Consequently, extensive deuteration may not always be a realizable approach.

However, this obstacle can be solved by radiofrequency probes with spinning rates

greater than 100 kHz [228], permitting NMR studies of fully protonated samples. With

increasing MAS, the proton linewidths in proteins decrease linearly, enhancing

resolution and sensitivity. With MAS rates >100 kHz, proton linewidths are sufficiently

narrow to resolve most resonances in a two-dimensional spectrum, involving a proton

dimension for detection. Increasing the spinning frequency to even higher values would

narrow the resonance lines further and would allow additional improvements in

spectroscopic methods and achievable resolution. It is assumed that at 250 kHz MAS

the homonuclear dipolar couplings are effectively averaged out for fully protonated


100

proteins and solution-state like spectra can be obtained [240]. 1H detection in

combination with fast MAS allow high resolution solid state NMR of membrane proteins

as well. A number of structural studies could demonstrate its applicability for both β-

barrel and α-helical membrane proteins in lipids and 2D crystal preparations [234, 241,

246-251]. Here, the work of Pintacuda and co-workers is outstanding [251]. They

demonstrated the applicability of 1H detection in combination with ultra-fast MAS at

100 kHz on the fully protonated hepta-helical membrane protein proteorhodopsin (PR)

in native-like lipids (DMPC/DMPA). Their approach provides 1H-based sequential

assignments and the identification of long-range interhelical 1H−1H contacts between

the side chains in transmembrane regions. Thus, this work represents a step towards

structure determination of membrane proteins by 1H detected ssNMR. Additionally,

ultra-fast MAS allows the investigation of sub-milligram amounts of protein (0.7 mm

probe, ~0.5 mg sample), which is especially beneficial for systems, suffering from low

yield. An additional benefit of fast spinning is the applicability of low-power pulse

sequences improving the electrical stability of the probeheads, as well as greatly

reducing rf heating of the sample.

For U-13C,15N-wtDGK in phospholipids (DMPC/DMPA), very first 1H detected

experiments were carried out as well. 2D 1H-15N correlation experiments (2D hNH)

[252] were performed on a Bruker 600 MHz spectrometer at ~4°C and a MAS rate of

111 kHz (Bruker 0.7 mm rotor, ~0.5 mg sample). They were conducted by Dr. Venita

Decker at Bruker BioSpin GmbH in Rheinstetten. The 2D hNH experiment is

comparable to a solution NMR HSQC (Heteronuclear Single Quantum Correlation)

experiment, but with the difference that it is based on CP. The initial 1H magnetization

is transferred to 15N via a HN-CP step. The 15N signal evolves under 1H decoupling.

Using an H/C/N triple channel probe, 13C-decoupling can be used as well. Water

suppression is carried out with a MISSISSIPPI scheme [253], which is applied on-

resonance with the water peak. The 15N magnetization is transferred back to 1H via a

NH-CP step, which is followed by 1H-detection under low power decoupling on 15N, and

possibly 13C. Hence, the 2D hNH experiment correlates the amide proton to its

nitrogen. Figure 32 demonstrates its principle applicability. It is noteworthy that with

only ~0.5 mg sample a comparatively high signal intensity can be obtained within

~12 h. This clearly highlights 1H detection at ultra-fast MAS as an important technique

for increasing sensitivity. For few, well-resolved peaks, the assignments from 13C/15N

detected experiments (Table S5) performed under moderate MAS (15.2 kHz) could be

transferred. However, further optimization steps, ranging from rotor packing to

appropriate pulse optimizations, are still needed to gain a higher signal-to-noise ratio

e.g. for recording three or higher dimensional spectra.


101

Figure 32. Solid state NMR 1H detected 2D

1H-

15N correlation spectrum (2D hNH) of fully

protonated U-13

C,15

N-DGK in phospholipid bilayers. The spectrum was recorded on a Bruker

600 MHz spectrometer at ~278.15-283.15 K and a MAS rate of 111 kHz (Bruker 0.7 mm rotor,

~0.5 mg sample). It was conducted with 400 scans and a duty cycle of 0.8 s. The total

measurement time was ~12 h. Some, well-resolved residues from the extramembrane regions

(green) could be assigned based on the assignments from 13

C/15

N detected experiments

conducted at a MAS rate of 15.2 kHz (Table S5). The spectrum was recorded by Dr. Venita

Decker at Bruker BioSpin GmbH in Rheinstetten.

If the further optimization turns out to be successful, high quality data could be

collected, allowing 1H-based sequential assignments and the identification of 1H-1H

proximities, as shown for PR [251]. This would help to define a 3D structure of DGK by

MAS NMR. Additionally, well-resolved spectra would allow automated data analysis as

already used for solution NMR studies, resulting in a reliable and especially in a faster

assignment and structure determination.

Direct proton detection might be useful to selectively observe the mobile entities in a

fully protonated sample as well. With the application of INEPT-based spectroscopy,

many protein resonances with intrinsically narrow linewidths could be already

selectively observed (Figure 29). However, these resonances could only be tentatively

assigned. With the implementation of direct proton detection, the sensitivity can be

increased. Sensitivity enhancements of up to tenfold have been already reported for 2D

1H–13C INEPT HSQC experiments, when proton detection is compared to carbon

detection [248]. In addition, Ward et al. could show that proton-detected experiments


102

can be easily extended to three dimensions through the incorporation of proton–proton

mixing. This step facilitates the detection of side chain protons, extending the spin

systems by the proton chemical shifts. Additionally, it provides the possibility of the

determination of inter-residue correlations, allowing a sequential assignment of the

mobile regions. Concerning DGK, especially the cytosolic loop, which is invisible in

dipolar coupling based experiments, would be of great interest, since it is reported to

be involved in nucleotide binding [6]. Changes in dynamics between apo and

nucleotide bound state might be observable.

In general, DGK could serve as a model system for 1H detection on α-helical

membrane proteins in phospholipids for further methodical developments of the

technique. Further tests would clearly benefit from DGK’s high stability.

Chapter 5: Functional studies

103

5 Functional studies based on chemical shift perturbations

5.1 Introduction

DGK catalyzes the ATP-dependent phosphorylation of diacylglycerol (DAG) to

phosphatic acid (PA) at the membrane/cytoplasm interface. Over the last decades, an

immense data set has been collected to find out, how this unique and complex enzyme

accomplishes this reaction:

Two 3D structures have been published for DGK: one obtained by solution NMR in

dodecylphosphocholine (DPC) micelles [7] and one by 3D crystallization in lipidic cubic

phases (LCP) composed of monoacylglycerols (MAGs) [5]. Additionally, several studies

were carried out using different techniques to map the active site. Functionally relevant

residues were identified by mutational studies [6, 7, 10], X-ray crystallization [6], MD

simulations [11] and solution NMR [7, 10]. Furthermore, based on observations from

the X-ray structures in combination with mutational studies [6, 7, 10] and MD

simulations [6], Caffrey and co-workers proposed a catalytic mechanism for DGK [6].

Despite this valuable data set, important long-standing questions regarding DGK’s

catalytic mechanism remain unsolved. It is unknown yet whether the three active sites

of DGK are in same or different states during catalysis and whether DGK undergoes a

substantial conformational change prior to the actual phosphoryl transfer.

In this chapter, these questions are addressed by multidimensional high field 13C,15N

MAS NMR. A considerable advantage of NMR as a technique for structural and

functional analysis is that it enables the investigation of membrane proteins without

attenuating its structural plasticity that is in most cases integral to the biological

function. This is contrary to X-ray studies on DGK, which stabilize a single molecular

conformation by the crystallization process itself and by the application of cryogenic

temperatures. Another advantage of solid state MAS NMR is the possibility of

performing experiments directly within the lipid bilayer [254], which brings it closer to

physiological conditions compared to other membrane mimicking environments such

as detergent micelles. The membrane environment is of key importance as it is a

strong structural factor [11, 19-21]. It is also directly linked to the catalytic activity of

DGK [11, 19]. Thus, MAS NMR can provide highly complementary data to X-ray

crystallization and solution NMR studies.

The apo state of DGK is compared with the substrate bound states on the basis of the

nearly complete assignment of DGK (84%), shown in chapter 4. Perturbations in peak

position and intensity of the substrate bound states were analysed for each of the

101 assigned residues in 3D and 2D heteronuclear correlation spectra. Substrate-


104

induced chemical shift perturbations indicate structural changes while alterations in

peak intensities can be interpreted qualitatively in terms of altered dynamics: An

increase in mobility causes a reduction in peak intensities in experiments based on

cross polarisation. The nucleotide-bound state is based on

adenylylmethylenediphosphonate (AMP-PCP), a suitable non-hydrolysable ATP

analogue, which has been already used in solution NMR and X-ray crystallization

studies [6, 7]. The DAG-bound state is emulated by 1,2-dioctanoyl-sn-glycerol (DOG,

chain length n=8), which has been reported to act as a lipid substrate for DGK [107,

255].

5.2 Results

5.2.1 Establishing nucleotide- and DAG-bound states of DGK for NMR analysis

In order to find saturation conditions for DGK with Mg*AMP-PCP, a competitive

Mg*ATP inhibition assay was performed by monitoring the ATPase activity as a

function of Mg*AMP-PCP concentration (Figure 33a). It turned out that a concentration

of at least 10 mM of Mg*AMP-PCP (10-fold molar excess) is necessary to reduce

DGK’s activity below 10%, leading to saturation. The comparatively high Mg*AMP-

PCP-to-protein ratio, needed to gain saturation, is indicative of low-affinity nucleotide

binding. This finding is consistent with titration studies on DGK by solution NMR [7].

Additionally, 1H-31P cross polarization (CP) experiments were carried out, which show

that AMP-PCP binds to DGK under these conditions (Figure 33b). Furthermore, it could

be demonstrated that the fully saturated system is stable over at least 30 d at 4°C

without any significant evidence of degradation (Figure 33c). This ensures long MAS

NMR experiments such as 3D NCACX experiments with measurement times of up to

14 d.


105

Figure 33. DGK in the AMP-PCP bound state. (a) Competitive inhibition assay verifies the

binding of Mg*AMP-PCP to the active sites of DGK. DGK proteoliposomes were incubated with

4 to 16 mM of Mg*AMP-PCP, equating 4 to 16-fold molar excess compared to DGK. Mg*ATP

(3 mM) was present in each sample. A concentration of at least 10 mM of Mg*AMP-PCP (10-

fold molar excess) is needed to decrease the activity of DGK below 10%, resulting in a fully

saturated system. 100% activity corresponds to the rate recorded with wtDGK in 90mol%


mg-1



(b) The 31

P-CP MAS spectrum confirms the binding of AMP-PCP. For this purpose,

proteoliposomes were incubated with 14 mM Mg*AMP-PCP (pH 7.2). (c) The 2D NCA spectra

of U-13

C,15

N-DGK-I,L,V incubated with 14 mM Mg*AMP-PCP (pH 7.2), recorded immediately

after the incubation (black) and after 30 d (green), show that the fully saturated system is stable

over a long period of time without any significant evidence of degradation.

In order to generate a state of DGK with bound lipid substrate, the protein was

reconstituted into DMPC/DMPA liposomes, including 20 mol% DOG [107, 255].

Empirically, it turned out that 20 mol% of DOG is an applicable amount to keep the

liposomes still intact. In order to test, whether DOG can reach the active site, MgATP

was added to the sample. This way, DGK could transfer the phosphoryl group from

ATP to DOG, yielding DOG-PA. Phosphorylated DOG should then become visible in

31P spectra. Indeed, the DOG-PA signal could be monitored in 31P-cross polarization

(CP) and direct polarization (DP) spectra (Figure 34a). The DP spectrum shows all 31P-

species such as α- and β-ADP as well as DOG-PA. The CP spectrum on the other side

only displays membrane-bound 31P-species such as DOG-PA and phospholipids. Both


106

the DP and CP spectra prove that DOG reaches the active sites of DGK in our

preparations. The same conditions were used to prepare a sample, in which DGK was

saturated with both Mg*AMP-PCP and DOG (80 mol% DMPC/DMPA and 20 mol%

DOG). In this case, DOG could not be phosphorylated, but bound Mg*AMP-PCP could

be observed as described above (Figure 34b).

Figure 34. (a) DGK in the DOG bound state. DGK was reconstituted into 80 mol%

DMPC/DMPA and 20 mol% DOG and incubated with 14 mM Mg*ATP (pH 7.2). DGK

phosphorylates DOG to DOG-PA, which can be observed by 31

P-MAS NMR, both by cross- and

direct polarization. The spectra prove that DOG can reach the active site of DGK under the here

applied experimental conditions. (b) DGK in the DOG+AMP-PCP bound state. 31

P-CP spectrum

of DGK reconstituted into 80 mol% DMPC/DMPA and 20 mol% DOG and incubated with 14 mM

Mg*AMP-PCP (pH 7.2). It illustrates 31

P species of the bound AMP-PCP, which demonstrates a

binding of the nucleotide to DGK.

5.2.2 DGK forms a symmetric trimer in its substrate bound states

Under the applied experimental conditions, no peak splitting could be detected for the

AMP-PCP and the DOG bound state during the assignment process. This is, as for the

apo state, already partly observable in the 2D NCA spectra, e.g. for well-resolved

peaks, such as Gly20, Asn27, Ala29, Glu88, Ala99 and Trp112 (Figure 35a, b). This

implies a symmetric conformation of DGK in its substrate bound states. Especially,

Gly83 and Gly91, framing the cytosolic loop, which is reported to be involved in

nucleotide binding [6], act as perfect probes for conformational asymmetries induced

by AMP-PCP. Clearly, both do not split in the AMP-PCP bound state (Figure 35a): Only

one peak for Gly83, indicating a random coil structure and one peak for Gly91,

suggesting a helical structure, could be detected. These findings support the


107

assumption that the DGK trimer adopts a symmetric conformation in the AMP-PCP

bound state.

Figure 35. The DGK trimer adopts a symmetric conformation in its substrate bound states. (a)

Superposition of 2D NCA spectra of apo DGK (black) and AMP-PCP-bound DGK (green).

Regions for helical and random coil (rc) secondary structure are highlighted for glycines. (b)

Superposition of 2D NCA spectra of apo DGK (black) and DOG-bound DGK (yellow). For the

AMP-PCP and the DOG bound state no peak splitting can be observed.

5.2.3 Substrate-induced chemical shift and peak intensity perturbations

5.2.3.1 AMP-PCP bound state

In the AMP-PCP bound state, significant chemical shift perturbations (CSPs ≥ 0.2 ppm)

and alterations in peak intensities are visible for 57% of the assigned residues,

including backbone and side chains (Figure 36a-c, Table S6). Previous mutational

studies have identified the following residues of DGK to be catalytically relevant: T8,

R9, A13, S17, G20, E28, A30, F31, R32, E34, E69, N72, S73, E76, D80, R81, G83,

L89, S90, A93, K94, D95, G97, S98 and A100 [7, 256]. With the studies by MAS NMR

using heteronuclear 2D and 3D correlation experiments, 14 out of these 25 residues

could be confirmed to be affected by nucleotide binding, featuring significant CSPs

and/or changes in peak intensities (labelled bold). Additionally, further 43 residues


108

could be identified: G15, Y16, K19, R22, A23, W25, I26, N27, A29, Q33, A37, L40,

A41, V43, A45, C46, W47, L48, D49, V50, D51, I53, R55, V56, L57, V62, V65, M66,

I67, I70, A74, A77, V79, I82, E88, G91, R92, A99, V101, L102, T111, I114 and L116.

Affected residues, showing significant CSPs and/or perturbations in peak intensity, are

highlighted in the DGK topology plot and mapped on the 3D structure in Figure 36d.

The picture illustrates that binding of the nucleotide has a higher impact on DGK as it

was assumed so far [7]. Largest CSPs ≥ 0.4 ppm are visible mainly for residues of the

cytosolic region: E28, F31, R32, Q33, E69, I70, V79, R81, I82, G91 and K94, but also

for the periplasmic loop: D49, I53 and for the transmembrane region: A45, V62. For

V62, D80 and K94. Representative sections from the 3D NCACX spectra are shown in

Figure 36c. Especially, glycines turned out to be sensitive to AMP-PCP binding.

Significant chemical shift perturbations could be detected for Gly15, Gly20, Gly91 and

Gly97 and an increased peak intensity was observed for Gly20 and Gly83 (Figure 36a,

b). Particularly, Gly83 located in the cytosolic loop (CL) features a notable increase in

signal intensity, suggesting a reduced mobility of CL. This is consistent with X-ray data,

implying a participation of the cytosolic loop in nucleotide binding via Glu85, Tyr86 and

His87 [6]. However, the most pronounced effect could be observed for Gly91, which is

present in the extramembranous part of helix 3. It shows a high shift of 1.7 ppm in the

N dimension to a lower ppm value. A reason could be the close proximity to the purine

ring of AMP-PCP [6], resulting in a higher shielding of 91GlyN and that in turn would

cause a shift to a lower ppm value in the N dimension. Overall, glycines act as perfect

sensors for nucleotide binding and accompanied conformational and dynamical

changes in the cytoplasmic region.


109

Figure 36. Effect of nucleotide binding on DGK. (a) Superposition of 2D NCA spectra of DGK’s

apo (black) and AMP-PCP-bound (green) state. Representative pronounced shifts are

illustrated in subsections of 2D NCACX (b) and 3D NCACX (c) spectra. (d) The topology and

ribbon model of the DGK monomer are shown with residues highlighted that are affected by

AMP-PCP. In the topology maps, alterations in peak intensity and different levels of weighted

CSPs are distinguished. In the ribbon model of monomeric DGK, residues, which show a

response on AMP-PCP binding, are highlighted in green. The ribbon model is obtained from the

OPM database [116], using the PDB ID 4UXX from the X-ray structure [6].

5.2.3.2 DOG bound state

In the DOG bound state, overall CSPs and dynamical changes are considerably less

pronounced compared to the AMP-PCP bound state. Only 14% of the assigned

residues and mainly their side chains are affected (Figure 37a-c, Table S6). CSPs


110

≥ 0.3 ppm are visible in side chain nuclei of K19, Q33, E69 and I70. Changes in peak

intensities appear more often, implying dynamical changes, which occur mainly in the

surface helix as shown in Figure 37d. These structural and dynamical changes might

take place due to a specific direct response on the lipid substrate in the active site, but

could be also caused by changes of the membrane due to 20 mol% of DOG, which is

inserted into the lipid bilayer. However, based on biochemical and structural data [6, 7,

11, 256], the perturbations concerning Arg9, Gln33 and Glu69 are likely related to

specific interactions with the lipid substrate bound to the active site. Glu69 is reported

to be conserved and relevant for the catalytic function of DGK. It is depicted to directly

interact with the lipid substrate [7, 256]. Its side chain peaks are raised during DOG

binding, which is especially observable for Cd. Furthermore, Cb shows a clear CSP

(Figure 37b). Both the signal increase and the CSP imply an interaction of the Glu69

side chain with the lipid substrate. Gln33, which features a significant CSP ≥ 0.3 ppm

for Cg (Figure 37b), is located in the active site [6] as well. Hence, it might be also a

possible interacting partner for DOG. Moreover, Arg9 and Lys12 that are located on the

surface helix are affected by DOG as well. Both, which are identified as highly mobile

residues in scalar coupling based experiments, are significantly reduced in the DOG

bound state compared to the apo state (Figure 37c), which is indicative for a decrease

of mobility. Arg9 has been already reported to be catalytically important, directly

interacting with the lipid substrate [6, 7, 11, 256].


111

Figure 37. Effect of DOG binding on DGK. (a) Superposition of 2D NCA spectra of DGK’s apo

(black) and DOG-bound (yellow) state. (b) Representative extractions from the 3D NCACX

illustrate shifts for Glu69 and Gln33. (c) Superposition of 15

N and 13

C INEPT-based experiments

of the apo (black) and DOG-bound (yellow) state of DGK. In the DOG bound state, the INEPT

signals are decreased compared to the apo state, indicating a reduction in mobility. Arg9 and

Lys12, which could be assigned unambiguously, are highlighted. (d) The topology and ribbon

model of the DGK monomer highlight residues that are affected by DOG. In the topology maps,

alterations in peak intensity and different levels of weighted CSPs are distinguished. In the

ribbon model of monomeric DGK, residues, which show a respond on DOG, are highlighted in

green. The ribbon model is obtained from the OPM database [116], using the PDB ID 4UXX

from the X-ray structure [6].


112

5.2.3.3 AMP-PCP + DOG bound state

The AMP-PCP + DOG bound state shows a similar fingerprint compared to the state

with only AMP-PCP bound (Figure 38).

Figure 38. Effect of AMP-PCP and DOG binding on DGK. (a) Superposition of 2D NCA spectra

of apo (black) and AMP-PCP-bound (green) DGK. (b) Superposition of 2D NCA spectra of

AMP-PCP-bound (green) and AMP-PCP+DOG-bound (pink) DGK. Both the AMP-PCP bound

and AMP-PCP+DOG bound states feature a similar fingerprint with significant alterations

compared to the apo state.

5.3 Discussion

5.3.1 DGK forms a symmetric trimer in its substrate-bound states

Under the applied experimental conditions, the DGK trimer adopts a symmetric

conformation in the AMP-PCP bound and the DOG bound state. This suggests that all

three active sites are occupied by the respective substrate at the same time. This in

turn indicates that under high substrate concentration, the three active sites are

possibly in the same state during catalysis, which is contrary to the hypothesis by

Caffrey and co-workers. The crystal structure of Δ4DGK co-crystallized with AMP-PCP

shows that only one active site is occupied by the nucleotide substrate, even though a

high concentration of AMP-PCP (10 mM) was used like in this study [6]. Differences in


113

the number of occupied active sites between NMR and X-ray data might occur due to

different experimental conditions. The investigations by MAS NMR were carried out on

wild-type DGK in DMPC/DMPA-liposomes under a physiological pH of 7.2 and at

~275 K. Crystallization of the thermostable Δ4-mutant took place under acidic

conditions (pH 5.6) in lipidic cubic phases formed by monoacylglycerols (MAGs), which

also act as lipid substrates. Its structure could be affected by crystallization contacts.

5.3.2 AMP-PCP bound state

This study demonstrates for the first time that not only the cytosolic part but also long

ranges of the transmembrane domains are affected by nucleotide binding Figure 36d.

The nucleotide most likely governs DGK into its catalytic active form. This is consistent

with the fact that DGK’s high nucleotide specificity is mainly observed in form of

reductions in kcat for ATP analogues [257]. It is also supported by the evidence that

the tetraphosphate-linked bisubstrate analogue served as a good inhibitor [257].

Furthermore, it is not surprising that nucleotide binding is noticed on a large scale in

this small kinase. With regard to the size of DGK, the nucleotide is comparatively bulky.

Additionally, the impact of AMP-PCP on DGK is likely even enhanced by the

circumstances that under the here applied experimental conditions three nucleotide

molecules are bound to one kinase molecule. These findings are accompanied by data

of Jia and co-workers, who demonstrated that the membrane is essential for stabilizing

the small structure of DGK, enabling the orientation of the substrates [11]. A similar

substantial conformational change was found for other kinases catalyzing direct

phosphoryl transfer as well [258-260].

Accordingly, binding of nucleotide induces an information transfer through the whole

enzyme towards the opposite site, which might occur in preparation of DGK for binding

the lipid substrate. The respective residues are possibly oriented for a proper binding of

the lipid. This implies positive heteroallostery, emanating from the nucleotide substrate,

which is in-line with kinetic studies of DGK, indicating that binding of the nucleotide

substrate does result in an enhanced affinity of the lipid substrate [4, 257].

For the bound states of DGK with either AMP-PCP + DOG or only AMP-PCP, a similar

fingerprint was detected (Figure 38). This suggests that the nucleotide substrate

induces a substantial conformational change, which is possibly required to trigger the

actual phosphoryl transfer reaction. Convincing kinetic and structural data support a

direct, in-line phosphoryl transfer based on a close proximity of the γ-phosphate of the

nucleotide and the 1-OH of the lipid substrate [6, 107, 257]. Respectively, our data


114

indicate that the specific conformation induced by the nucleotide is most likely the

catalytic active conformation of DGK, which is needed to bring both substrates in close

proximity.

5.3.2.1 Comparison of the AMP-PCP bound state with solution NMR data

Next to the investigation of the AMP-PCP bound state of DGK by solid state NMR,

titration studies of DGK with up to 16 mM AMP-PCP were carried out by solution NMR

[7]. Both studies provide significant chemical shift changes (CSPs) concerning the

backbone of several residues. For 4 residues, namely E69, N72, G91 and K94, both

studies are in agreement, showing significant CSPs, whereas they deviate for a large

number of other residues. For instance, the solution NMR study reveals for G83 the

highest CSP. This is in contrast to our data, which feature for G83 only changes in

peak intensity, but no significant CSP (Figure 36a, b). Furthermore, e.g. for G20, A29,

D49, I70, I82, E88 and G97 clearly significant CSPs could be observed by solid state

NMR (Figure 36a, b), which are not visible by solution NMR [7]. The observed

differences occur presumably due to the manner, in which the proteins are present at

the time of substrate binding. In one case, the protein is embedded in a DPC micelle at

45°C [7]. In the other case, it is incorporated in DMPC/DMPC liposomes at 4°C. The

applied pH and AMP-PCP concentration were in a similar range. The shorter chain

length, the very high curvature and the small hydrophobic thickness of DPC micelles

(30–40 Å diameters) [119] could interfere with the approximate diameter of the DGK

trimer (~100 Å) [7], which is in contrast to almost planar lipid bilayers. This is in

agreement with notable differences, which are already visible in the secondary

structure obtained by solution and solid state NMR: In contrast to the ssNMR data, the

solution NMR structure features two small distortions in Y16 within the surface helix

(SH) and I70/L70 of helix 2 (H2). In addition, there are small deviations around the

interhelical turn (T) between H1 and SH and around the periplasmic loop (PL) between

H1 and H2 as well as the cytoplasmic loop (CL) between H2 and H3, which are 3, 1

and 6 residues longer in DPC micelles compared to lipid bilayers. It should be noted

that the variations in the secondary structures in different environments all arise in

catalytically critical regions, such as the SH, cytosolic regions of H2 and H3 and the

CL. These differences most likely reflect the impact of the environment, in which the

membrane protein is embedded.


115

5.3.3 DOG bound state

In the DOG bound state, overall changes are significantly less pronounced. Especially,

the transmembrane domains do not show any reaction on the acyl chains of DOG

(Figure 37d). However, few significant perturbations could be determined at the

cytoplasm/membrane interface for Arg9, Lys12, Gln33 and Glu69 (Figure 37a-c).

Based on the crystal structure, all four of them are located in the active site in close

proximity to the headgroup of the lipid substrate [6]. Their side chains possibly interact

with the proximal OH group or the carbonyl oxygen at the ester linkages of DOG,

keeping the headgroup of the lipid substrate in position (Figure 39). Especially, Arg9

and Glu69 have been already reported to directly interact with the headgroup of the

lipid substrate via H-bonds [6, 11]. Hence, these data imply that the lipid substrate is

primarily recognized by its headgroup and not by its acyl chains, which is in agreement

with observations by Walsh et al. [255].

Figure 39. Enlarged view from the membrane plane, illustrating the crystal structure of Δ4 DGK

(PDB 3ZE5) [5], accommodating the lipid substrate (orange) in the hydrophobic pocket.

Possible interactions between the side chain of Arg9, Lys12, Gln33 and Glu69 with the proximal

OH group of the lipid substrate are depicted.

Changes in DGK induced by DOG might take place, as just mentioned, due to a

specific direct response on the lipid substrate in the active site, but could be also

provoked by the altered order parameter of proximate phospholipids due to inserted

DOG (20 mol%) into the lipid bilayer. It is known that DAGs with a short chain length

(DOG, chain length n = 8) induce an ordering effect close to the headgroup of the

phospholipid side chains [261]. The DOG molecules intercalate between the bulky

headgroups of the phospholipids and promote a tighter contact between their side

chains in the region close to the headgroups, thereby increasing their order


116

parameters. The lower segments of the phospholipid side chains, which cannot be

reached by DOG, are slightly more disordered in the presence of DOG.

5.4 Summary and Outlook

This study provides an almost complete resonance assignment of wtDGK within the

lipid bilayer in its apo-, nucleotide- and lipid substrate-bound states. This way, the

overall response of DGK towards substrate binding could be mapped. It could be

shown that all three active sites can be occupied concurrently by both AMP-PCP and

DOG. Under high substrate concentration, the three active sites are most likely in the

same state during catalysis. Additionally, it could be demonstrated for the first time that

not only the cytosolic region but also large parts of the transmembrane domains are

affected by nucleotide binding. This possibly supports the enzyme in binding of the lipid

substrate, implying positive heteroallostery. Furthermore, the substantial

conformational change induced by the nucleotide seems to set the enzyme into a

catalytic active state, triggering the actual phosphoryl transfer reaction.

In order to obtain further insights into DGK’s catalytic mechanism, the protonation state

of several residues could be investigated. Unfortunately, an unambiguous

determination at this stage was not feasible due to signal overlap and low signal

intensity in the 3D NMR spectra. For this purpose, a strategic labelling scheme to

follow these specific residues by 2D experiments has to be developed. This way, they

could be specifically monitored in the apo, AMP-PCP-, DOG- and AMP-PCP+DOG-

bound state.

Furthermore, nucleotide-protein and DAG-protein interactions could be investigated in

the AMP-PCP and DOG bound state using DNP-enhanced MAS NMR. Labelling

schemes could be decided with respect to the X-ray structure (PDB 4UXX) [6].

Additionally, the role of the surface helix in sensing osmolality and altered lateral

pressure in the lipid bilayer [5] could be probed by multidimensional high field MAS

NMR based on the almost complete resonance assignment shown in chapter 4.

Different concentrations of DAG could be inserted into the lipid bilayer, modulating the

membrane order parameter. In order to probe how DGK, especially its SH, responds to

changes of the membrane lateral pressure, chemical shift perturbations could be

examined and site-resolved order parameters and relaxation rates could be

determined.

Chapter 6: Long-range contacts

117

6 Long-range contacts

6.1 Introduction

Inter-residue contacts are the cement of a protein structure, stabilizing its fold. Thus,

they have been of great interest for the investigation of the mechanisms of protein

folding and stability [262, 263]. Contacts also provide a platform for crosstalk between

single residues and thus, in some cases between different domains or subunits. Hence,

they take up functional roles as well [264, 265]. Contacts in proteins can be of different

nature. Hydrogen bonds are formed between two electronegative atoms, such as

nitrogen and oxygen, by sharing a hydrogen atom [266]. It has been demonstrated that

even weak hydrogen bonds could provide inter-residue contacts [267, 268]. Ionic

bonds are based on interactions between oppositely charged groups of a molecule,

e.g. between the positively charged basic side chains of lysine and arginine, and the

negatively charged carboxyl groups of glutamate and aspartate [269]. In contrast, van

der Waals interactions are weak forces [270]. They act not only between polar

molecules but also between electrically neutral atoms and molecules. This is due to the

shifting of electrons in the outer shell of the atoms, which temporarily cause charge

movements and so-called polarization. Charged areas with different signs are then

attracted to one another and hence provide an attraction between two atoms, even if

they are electrically neutral overall.

DGK appears as a homotrimer and features a remarkable stability in native

membranes [8, 9]. Its oligomeric arrangement is of direct functional relevance, since its

complex forming monomers alone are not functional. Each active site is built by

components of two monomers, based on the composite shared site model [5-7, 10].

This implies that the substrate-bound state of one site is relayed to the other sites,

suggesting cooperativity in substrate binding [5, 6], which is not uncommon for many

oligomeric proteins [271]. The assumption of cooperativity is supported by the

presence of the amphiphilic surface helix, which suggests itself as a mediator

transmitting inter-subunit information [5, 6]. However, understanding DGK’s remarkable

stability and the cross-talk between its subunits demands the identification of key intra-

and interprotomer contacts, which are of structural or functional importance.


118

6.2 Intraprotomer contacts visualized by high field MAS NMR

6.2.1 Results and Discussion

In order to identify long-range contacts in the apo state of DGK, 13C-13C DARR and 2D

NCOCX spectra with long mixing times were recorded using high field MAS NMR. For

this purpose, the extensive side chain assignment obtained in chapter 4 could be used.

The spectra were analyzed with focus on long-range, non-sequential side chain-side

chain cross peaks. This way, some intraprotomer contacts could be determined:

Crosspeaks between Ca of Ser61 located in helix 2 and side chain carbons of Trp112

present in helix 3 (Figure 40a) reveal close proximity of both residues. This is

consistent with findings of Caffrey and co-workers, who suggested a hydrogen bond

between Ne of Trp112 and the OH-group of Ser61 (Figure 40b) [6]. The contact

between Trp112 and Ser61 has most likely a stabilizing effect on the transmembrane

region of each monomer (Figure 40b). This is in agreement with mutational studies by

Lau and Bowie, who demonstrated that the mutant of DGK, in which Trp112 is replaced

by Phe, was susceptible to aggregation and showed a low specific activity [9].

Figure 40. Intraprotomer interactions in the transmembrane region of DGK between Trp112 and

Ser61. (a) 2D 13

C-13

C DARR spectrum of U-13

C,15

N-DGK-ILV with 800 ms mixing time.

Crosspeaks appear between Ca of Ser61 and the side chain carbons of Trp112 revealing an

intraprotomer contact between helices 2 (Ser61) and 3 (Trp112). (b) Visualization of the

intraprotomer contact between Trp112 and Ser61 in the crystal structure of Δ4 DGK (PDB


119

4UXX) [6]. Enlarged view from the membrane plane, accommodating the lipid substrate (yellow)

in the hydrophobic pocket (left). View from the periplasm, depicting the three monomers in

different shades of grey (right). Trp112 (H3) is secured by a hydrogen bond to Ser61 (H2) in

the lower region of the hydrophobic pocket.

Furthermore, crosspeaks between side chain nitrogens, Ne and Nh1/2, of Arg32 with

25TrpCO, 28GluCb and 29AlaCa,Cb,CO were detected (Figure 41a). Arg32 is located

at the membrane/cytoplasm interface of helix 1, while Trp25, Glu28 and Ala29 are

present in the SH, interhelical turn and the cytoplasmic part of H1 (Figure 41b). These

contacts are described for the first time. Additionally, a R32A mutant was prepared,

which features a strongly reduced activity (Figure 49, grey bar). This is in agreement

with previous mutational studies [6, 7]. The functional relevance of Arg32 could be

explained by its role in forming the contacts with Trp25, Glu28 and Ala29, which have

most likely a strengthening effect on the joint between H1 and SH, stabilizing the SH

and thus the active site as well (Figure 41b).

Figure 41. Intraprotomer interactions in the cytoplasmic region of DGK between Arg32 and

Trp25/Glu28/Ala29. (a) 2D NCOCX spectrum of U-13

C,15

N-DGK with a 400ms DARR mixing

step. Crosspeaks between 32ArgNh1/2 and 32ArgNe with 25TrpCO, Glu28Cb and

29AlaCa/Cb/CO are determined, caused by an intraprotomer contact between these residues in

helix 1 and the surface helix. (b) Depiction of the intraprotomer contact involving Arg32 in the

crystal structure of Δ4 DGK (PDB 3ZE5) [5]. Enlarged view from the membrane plane,

illustrating the intraprotomer contacts for 32ArgNe/Nh1,2 with 25TrpCO, Glu28Cb and

29AlaCa/b/O.

6.3 Interprotomer contacts visualized by DNP-enhanced MAS NMR

6.3.1 Introduction

Unfortunately, no cross-protomer contacts could be detected by conventional high field

MAS NMR. The identification of interactions at protomer interfaces is in general a more

complex and challenging task. Most data available so far are from crystal structures of


120

membrane proteins. Crystal structures have been reported for DGK as well [5, 6],

suggesting several interprotomer contacts as illustrated in Figure 42.

Figure 42. Representative possible interprotomer contacts in DGK suggested by the crystal

structure of Δ4 DGK (PDB 3ZE5) [5]. Enlarged view from the cytoplasm (a) and membrane

plane (b, c), illustrating a possible interprotomer contact for Arg81 and Glu88 (a), Arg92 and

Asn27 (b) as well as Lys19 and Asp95 (c).

However, the predicted contacts are in general debatable due to incompleteness of the

electron density concerning relevant side chains (Arg81, Glu88, Arg92, Asn27 and

Lys19: PDB 3ZE5 electron density map). Additionally, contacts in crystal structures can

be assailable due to the difficulty to distinguish between biologically relevant

interactions from those induced by crystal contacts [272, 273]. This highlights the need

for complementary spectroscopic data in the membrane environment. Solid state NMR

is able to provide such interaction data. It could be shown that it is possible to

investigate protein−protein contacts by directly observing interpeptide dipole couplings

[274] or indirectly through paramagnetic relaxation enhancement [275]. Concerning

amyloid fibrils, mixed labelled samples were utilized to determine interprotein distance

constraints [276, 277].

By dis- and reassembling of trimeric DGK, mixed labelled 13C−15N complexes can be

gained, which feature a unique isotope labelling pattern at their protomer interfaces.

15N−13C transferred echo double resonance (TEDOR) spectroscopy [135, 136] enables

the detection of specific interprotomer contacts. Conventional ssNMR would be barely

suitable for the identification of interprotomer contacts [129], since reassembling of

different labelled protomers to the oligomeric state is based on a statistical distribution,

leading to a comparably low number of mixed labelled interfaces. Hence, sensitivity

enhancement is needed. Dynamic nuclear polarization (DNP) has been developed to

increase the sensitivity of MAS NMR by orders of magnitude [278]. There are several

studies under cryogenic conditions, which benefited from DNP, as presented for


121

various membrane proteins [134, 279-281]. Maciejko et al. could verify for the first time

that DNP-enhanced ssNMR can be used to detect interprotomer contacts within a

homo-oligomeric membrane protein embedded into lipid bilayers [129, 134]. This work

was done in this lab using pentameric green proteorhodopsin. In this chapter, it is

demonstrated that the basic principles of the work from Maciejko et al. can be applied

to trimeric DGK as well. An extensive study was carried out to establish a procedure for

dis- and reassembling of homo-trimeric DGK to produce active mixed labelled

complexes, which is verified by LILBID-MS [282], BN-PAGE [283] and the coupled

activity assay [18]. A labelling scheme is utilized, which enables the detection of

possible cross-protomer interactions (Arg/Lys−Asp/Glu/Asn), as predicted from the

crystal structures of DGK (Figure 42) [5, 6]. DNP-enhanced 13C−15N TEDOR

experiments are conducted to monitor these contacts, while single-site mutations are

inserted in order to assign them. Additionally, the mixed labelled trimers were saturated

with the ATP analogue, AMP-PCP, to determine, if these contacts are involved in

nucleotide binding.

The theoretical background of DNP and the TEDOR experiment is explained in detail in

section 1.3.3.

6.3.2 Results

6.3.2.1 Creating mixed labelled trimers of DGK

In order to monitor 13C−15N side chain contacts using DNP-enhanced TEDOR

experiments, mixed labelled complexes, consisting of neighbouring 13C and 15N

protomers, can be gained by dis- and reassembling of homo-trimeric DGK. As shown in

chapter 3, the BN-PAGE (Figure 17b) shows one main trimer population for DGK

surrounded by DDM micelles, which is in-line with literature [7, 18]. It is also confirmed

by LILBID-MS (Figure 17c), a well-tested and unambiguous method for determining the

mass of macromolecular complexes [284]. It verifies that BN-PAGE analysis offers a

reliable assessment of the oligomeric state. DGK in its trimeric form has been shown to

be remarkable stable [9, 84, 89, 93, 108]. Data by Lau and Bowie indicate an

impressive unfolding free energy of 16 kcal/mol for the transmembrane domain and

even 6 kcal/mol for the cytoplasmic domain [9]. Thus, harsh conditions are necessary

to disrupt the oligomeric state. Thermal denaturation of membrane proteins is usually

irreversible as shown for DGK [141] as well as many other membrane proteins, such as

bacteriorhodopsin [285-287], erythrocyte band 3 [288, 289], cytochrome c oxidase

[290, 291] and photosystem II [292-294]. In contrast, it is known from a number of


122

studies that numerous membrane proteins can be refolded after denaturation with

chemical denaturants, such as guanidine hydrochloride (GuHCl), urea or SDS [295-

299]. Unlike GuHCl and urea, SDS is able to build mixed micelles with other

detergents, providing an environment for the protein, in which the α-helical content of

membrane proteins is often preserved [300, 301]. Lau and Bowie could show that even

after complete denaturation of DGK by the anionic SDS, the helical content is retained.

Additionally, 90% of the enzymatic activity could be recovered by dilution of the SDS-

denatured DGK into a DM solution [9]. Consequently, the detergent of choice for

disassembling trimeric DGK was SDS, which should enable the creation of active

mixed labelled trimers. For this purpose, a complete separation of the trimers into

monomers is beneficial, since it results in the largest number of interprotomer 13C−15N

interfaces after reassembling. Thus, the optimal ratio of protein-to-SDS concentration

had to be determined. Since DGK is more stable at higher protein concentrations [141],

a low concentration of 0.2 mg/ml was used for disruption. The BN-PAGE (Figure 43b)

illustrates the effect of different detergent concentrations on disassembling of trimeric

DGK. The higher the SDS concentration, the higher is the degree of disassembly into

monomers. The disruption into monomers is incomplete for SDS concentrations from

0.5%-1.5%, since clearly a population of dimers is visible. Starting from 2% SDS, the

separation into monomers seems to be total, which is therefore, used for disassembly

of trimeric DGK in the subsequent studies, allowing sufficient mixing of differently

labelled protomers. LILBID-MS was applied as well, confirming the predominantly

monomeric state of DGK in SDS micelles (Figure 43c). The LILBID mass spectrum was

recorded by Oliver Peetz of the research group of Prof. Dr. Nina Morgner (Institute of

Physical and Theoretical Chemistry, Goethe University Frankfurt am Main).

Furthermore, it was tested, if the trimeric form of DGK can be regained, when SDS is

removed. The BN-PAGE (Figure 43d) shows that a transition from monomeric to

trimeric DGK can be obtained, when SDS is completely displaced by DDM through

extensive washing. A scheme for the creation of mixed labelled trimers is presented in

Figure 43a. In order to examine, whether the disrupting detergent affects the activity of

DGK, the coupled activity assay was carried out. For this purpose, DGK trimers in DDM

micelles (-SDS), DGK monomers in SDS micelles (+SDS) and DGK trimers in DDM

micelles after SDS treatment (±SDS) were reconstituted into lipid bilayers. As

reconstitution method BioBeads were applied, reported to be suitable not only for DDM,

but also for SDS removal [302]. Figure 43e illustrates that trimers reconstituted from

DDM micelles before and after SDS treatment feature a similar activity. Monomers from

SDS micelles lead to a reduced activity below 20%, indicating that monomers, known


123

to be separately not functional [5, 7, 10], do not reassemble to trimers upon

reconstitution.

Figure 43. Creation of active mixed labelled trimers of DGK. (a) Differently labelled trimers of

DGK are separately expressed, solubilized, purified and eluted in DDM. Subsequently, they are

disassembled into monomers or dimers by SDS and mixed in a 1:1 ratio. Then, SDS is removed

and replaced by DDM, resulting in mixed labelled DGK trimers, which can then be reconstituted

into liposomes. (b) BN-PAGE of DGK (0.2 mg/ml) in different SDS concentrations: (1) 0.5%, (2)

1%, (3) 1.5%, (4) 2%, (5) 3%. The higher the SDS concentration, the higher is the degree of

disassembly into monomers. (c) LILBID-MS confirms the predominantly monomeric state of

DGK in SDS micelles: The signals for the monomeric, dimeric and trimeric form of DGK are

labelled by “1”, “2” and “3”, respectively. They occur at a charged state of −1. The LILBID mass

spectrum was recorded by Oliver Peetz of the research group of Prof. Dr. Nina Morgner

(Institute of Physical and Theoretical Chemistry, Goethe University Frankfurt am Main). (d) The

BN-PAGE shows DGK as trimer in DDM micelles before (1) and after (3) SDS treatment and

mostly in its monomeric state in SDS micelles (2). (e) Activity of DGK reconstituted from

different detergent environments: DGK trimers from DDM micelles before SDS treatment (-SDS,

dark grey) and after SDS treatment (±SDS, green) as well as DGK monomers from SDS

micelles (+SDS, yellow) were reconstituted. 100% activity corresponds to the rate recorded with

wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1

mg-1

. Experiments were

repeated three times. The activity was calculated as the mean value. Error bars correspond to

standard deviations.


124

Using the procedure above, samples were prepared, in which the trimer is composed

of (i) 13C- and 15N-DGK protomers ([CN]-DGK), (ii) 13C-DGK and 15N-Arg-Lys-DGK

protomers ([CN(Arg,Lys)]-DGK) and (iii) 13C-DGK ([CC]-DGK). The last one serves as

control sample for the differentiation of 13C−15N contacts from naturally occurring 13C-

or 15N-isotopes. In order to decrease natural abundance signals, all 15N-labelled

protomers were 13C-depleted by applying [12C6]glucose (99.5%) as a carbon source,

leading to a 50% lowering of the 13C background.

6.3.2.2 Validation of the application of AMUPol as biradical

DNP-enhancement was obtained by doping the sample with the biradical polarizing

agent AMUPol, serving as a source for unpaired electrons [133]. The enhancement (ε)

is determined by the ratio of NMR signal intensities with and without microwave

irradiation [131]. 45-fold signal heightening could be achieved, when the

proteoliposome pellet (~40 µl) was split into two small ones (~20 µl) for ~20 h of

incubation each with ~20 µl AMUPol solution (Figure 44a). In contrast, if just the unsplit

pellet of ~40 µl was incubated with ~40 µl AMUPol, a signal enhancement of only 23-

fold was determined after ≥ 24 h of incubation. Figure 44b proves that the presence of

AMUPol around the protein has no influence on DGK’s activity, since a similar activity

with and without the biradical was reached.

Figure 44. Validation of the application of AMUPol. (a) DNP enhancement shown for a 13

C−CP

spectrum of DGK incubated with 20 mM AMUPol. Upon microwave irradiation, a 45-fold

sensitivity enhancement is reached. (b) Activity of DGK with (+) and without (-) AMUPol,


125

indicating that the presence of the biradical has no influence on the activity. 100% activity

corresponds to the rate recorded with wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9)

µmol min-1

mg-1

. Experiments were repeated three times. The activity was calculated as the

mean value. Error bars correspond to standard deviations.

6.3.2.3 DNP-enhanced 15N−13C TEDOR experiments

6.3.2.3.1 Finding the best mixing time using 1D TEDOR spectra

The achieved sensitivity enhancement facilitated 15N−13C TEDOR experiments on the

mixed labelled trimers of DGK. By using TEDOR spectra, through-space dipole−dipole

contacts between 13C and 15N spins can be identified. Firstly, several 1D-TEDOR

spectra were recorded with different rotor periods (L0 = 4 – 40) to find the best mixing

time for our experiments. L0 describes the number of 180° pulses during tmix/4. Figure

45 illustrates TEDOR spectra of the control sample ([CC]-DGK).

Figure 45. DNP-enhanced 1D-TEDOR spectra of the control sample ([CC]-DGK) at different

mixing times. The spectra were recorded with 4096 scans at a 400 MHz spectrometer, ~105 K,

pH 7.2 and a spinning speed of 8 kHz. The 263 GHz gyrotron was operated at a collector

current of 70 mA. Six spectra were recorded with L0 (rotor periods) = 4, 8, 16, 24, 32, and 40.

The spectrum with short mixing time (L0 = 4) features TEDOR signals in the CO (170-

185 ppm) and the Ca (50 - 70 ppm) range. The signal at ~96 ppm represents the


126

spinning sideband of the carbonyl peak. Concerning higher mixing times (L0 = 16 - 40),

both the CO and the Ca signals decrease. However, small signals arise with increasing

mixing times in the aliphatic side chain range (10 - 40 ppm). All these signals originate

from the natural abundance of 15N in 13C-protomers. The CO signals arise from the

backbone peptide bonds, linking covalently two consecutive amino acids, whereas the

Ca signals derive from covalently bonded neighbouring backbone atoms within the

same amino acid. With 121 residues, DGK has 120 peptide bonds. Thus, a natural 15N

abundance of ~0.4% leads to statistically ~0.5 15N per DGK monomer, which is

obviously enough to be detected through DNP enhancement. The TEDOR signals for

covalently bonded nuclei are dominant at shorter mixing times due to their short

distances (CO-N: 1.32 Å, N-Ca: 1.45 Å [303]). In contrast, with longer mixing times, i.e.

with more 180° pulses in the 15N channel, longer 15N-13C distances can be determined,

such as the amide 15N and the side chain carbons. These signals peak at L0 = 24 and

decrease slightly with higher L0s.

Mixed labelled [CN]-DGK shows a similar behaviour concerning the mixing time as the

control sample ([CC]-DGK) (Figure 46). It features signals in the CO and Ca range from

naturally appearing 15N-isotopes (~0.4%) in the 13C and 13C-isotopes (~0.5%) in the 15N

labelled protomers, which reduce with higher mixing times. Also the behaviour of the

TEDOR signals in the aliphatic side chain range is comparable to the control sample.

They peak with L0 = 24, with the difference that they appear clearly stronger, indicating

further 15N contacts to aliphatic carbons. They most likely arise from protein-protein

contacts between backbone 15N amide and aliphatic 13C side chains of neighbouring

protomers. These signals are additional to natural abundance signals, suggesting that

the preparation of mixed labelled trimers ([CN]-DGK) was successful, providing

interprotomer 13C-15N contacts. Since signals in the aliphatic side chain range peak

with L0=24 (6.25 ms), it turned out that 6.25 ms is the perfect total mixing time to

optimize the signal intensity of anticipated long-range interprotomer couplings in the

DGK sample.


127

Figure 46. DNP-enhanced 1D-TEDOR spectra of [CN]-DGK) at different mixing times. The

spectra were recorded with 3520 scans at a 400 MHz spectrometer, ~105 K, pH 7.2 and a

spinning speed of 8 kHz. The 263 GHz gyrotron was operated at a collector current of 70 mA.

Six spectra were recorded with L0 (rotor periods) = 4, 8, 16, 24, 32, and 40.

6.3.2.3.2 Visualizing interprotomer contacts using 2D TEDOR spectra

In order to monitor specific 13C- 15N contacts between the protomers, 2D-TEDOR

spectra were recorded (Figure 47). All three spectra exhibit certain natural abundance

signals originating from N−CO and N−Ca single bond contacts as well as from long-

range N−Cx couplings. In addition, arginine intraresidue 13C−15N contacts between

ArgNe,n and ArgCz are visible. These natural abundance signals appear, as already

implied for the 1D TEDOR spectra, from naturally arising 13C-isotopes (~0.5%) in the

15N labelled protomers that are still included in spite of depletion by 12C. 12C6-glucose is

unfortunately not available with purity higher than 99.5%. In addition, the 15N natural

abundance in 13C-labelled protomers amounts to ∼0.4%.

The comparison of the spectrum of mixed labelled [CN]-DGK with the [CC]-DGK

control spectrum offers one additional cross-peak, which implies at least one cross-

protomer 13C−15N contact. The peak appears in the 15N chemical shift range of

arginine, suggesting that this amino acid type participates in cross-protomer

interactions. This is supported by the selectively labelled [CN(Arg,Lys)]-DGK sample,

which illustrates the same cross-protomer contact as [CN]-DGK. The cross-peak


128

reveals a correlation between the 15N resonance of Arg-Ne,n with 13C resonances of

carboxyl groups from Asp-Cg and/or Glu-Cd and/or the carbonyl group from Asn-Cg,

indicating possible cross-protomer salt bridge or H-bond contacts between these

residues. The selectively labelled sample was not only arginine, but also lysine

labelled, since the crystal structures of DGK predict one interprotomer contact with a

distance of ~3.7 Å between 19Lys-Nz (34 ppm) and 95Asp-Cg (179 ppm) as well

(Figure 42c) [5, 6]. However, both spectra for [CN]-DGK and [CN(Arg,Lys)]-DGK do not

feature a significant cross peak involving lysine, implying that the distance between

these two residues is too large in the wild type to be detectable. Nevertheless, the data

concerning the cross peak involving arginine reveal that DNP-enhanced TEDOR

spectra enable investigating 13C−15N contacts across the protomer interface of trimeric

and not only of higher-oligomeric membrane proteins. Further data are required to

assign the cross-peak to distinct residues and to clarify its impact on structural stability

and functionality.


129

Figure 47. DNP-enhanced 15

N−13

C-TEDOR spectra (tmix = 6.25 ms) of [CN]-DGK,

[CN(Arg,Lys)]-DGK and the control sample, [CC]-DGK. All spectra reveal cross-peaks

originating from natural abundance intramolecular backbone 13

C−15

N-contacts (highlighted in

grey). Further cross-peaks (highlighted green) are detected in [CN]-DGK and [CN(Arg,Lys)]-

DGK. They can be assigned to cross-protomer contacts, reflecting a through-space correlation

between Arg and Asn/Asp/Glu. These cross-peaks demonstrate that salt bridges or H-bonds

between Asn/Asp/Glu and Arg must be present at the protomer interfaces.


130

6.3.2.3.3 Attemps to assign the cross-peak by RxA-mutants

Single-site mutations were chosen to investigate the impact of the respective arginine

on DGK’s oligomerization, structural stability and functionality. For creating single-site

mutants, alanine was used for substitution. It is usually the first choice for mutational

scanning [304-306]. It is non-bulky and chemically inert due to its methyl group. A

mutation to alanine is generally equivalent to simply truncating a side chain back to Cb,

the first side chain atom. It has the propensity to form α-helices, but can also occur in

β-sheets, since the position of Cb depends upon the backbone dihedral angles of the

polypeptide and is part of the main chain structure of the protein. Glycine, for instance,

which removes Cb, is unusually flexible and can adopt polypeptide backbone

conformations that are generally not allowed by other amino acids. Consequently, a

mutation to glycine would cause flexibility and possible conformational changes,

making interpretation more complex than for alanine. Replacing side chains with larger,

more constrained amino acids with a branched pattern, more polar, differently charged,

or more hydrophobic atoms might cause changes in structure and conformation along

with the side chain chemistry. Thus, they would complicate the analyses of results

more than alanine as well.

Since DGK contains in total six arginines, the single-site mutations R9A, R22A, R32A,

R55A, R81A and R92A were introduced into the protein. These RxA-mutations not only

enable to investigate, if the side chain of the respective arginine has an impact on

DGK’s oligomerization, stability and function, but also can be used for cross-peak

identification. The mixed labelled single mutants were prepared as the wild type and

analysed by BN-PAGE. Figure 48 shows for all mutants a similar oligomerization

behaviour compared to wtDGK: All mutants form trimers in DDM that can be disrupted

to mainly monomers by SDS and reassemble mostly to trimers, when SDS is replaced

by DDM. There are also small deviations visible. Some mutants feature a weak dimer

population in SDS and/or in DDM after SDS treatment, which cannot be exclusively

explained by the inserted mutation, since also small deviations in handling during the

preparation process might be possible. In order to clarify, if solely the mutation is the

reason for this slight difference, all mutants would have to be prepared for at least three

times.


131

Figure 48. Characterization of the oligomeric state of the single-site RxA mutants in comparison

to wtDGK by BN-PAGE. The BN-PAGES show DGK as trimer in DDM micelles before (1) and

after (3) SDS treatment and mainly as monomers in SDS micelles (2). All RxA mutants show a

similar oligomerization behavior as the wt, indicating that the respective arginines located in

extramembranous regions of DGK are not necessary for the trimer formation.

The next characterization step should verify whether the respective introduced

mutation and/or the SDS treatment affects the activity of DGK. Thus, the coupled

activity assay was performed. For this purpose, RxA-DGK trimers in DDM micelles

before and after SDS treatment were reconstituted into lipid bilayers. Figure 49

indicates that all six arginines in DGK play a functional role, since the activity is clearly

reduced for all RxA-mutants compared to the wild type. Especially, the mutation of

Arg9, Arg32, Arg81 and Arg92 caused a reduction of activity below 25%. The SDS

treatment of the RxA mutants led to a loss of activity of only up to 20%, which is

comparable to the wt. Thus, they are only slightly prone to SDS.


132

Figure 49. Kinase activity of DGK affected by single site RxA mutations, expressed as a

percentage of wt activity. DGK trimers from DDM micelles (dark grey) and DGK trimers from

DDM micelles after SDS treatment (green) were reconstituted into lipid bilayers and then

measured. The activity is clearly reduced for all RxA-mutants compared to the wild type. The

SDS treatment of the RxA mutants led to a loss of activity of only up to 20%, which is

comparable to the wt. 100% activity corresponds to the rate recorded with wtDGK in 90mol%


mg-1



Overall, the data suggest that the six arginines are not essential for the trimer formation

of DGK, but they are all more or less functionally relevant. Mixed labelled complexes of

all mutants ([CN]-DGK-RxA) were analysed by DNP-enhanced TEDOR experiments

and compared with the wild type spectrum as well as among each other in order to spot

any differences (Figure 50). Since only one cross-peak in the arginine range (Arg-

Asp/Glu/Asn) had been observed, the spectral width was reduced and the offset

adjusted to focus the peak of interest, which allows to use more scans, leading to a

better signal-to-noise ratio. Unfortunately, the cross-peak does not disappear for each

of the six possible arginine mutants ([CN]-DGK-RxA), indicating that multiple arginines

contribute to these interactions. However, a reduction of the cross peak intensities for

R81A- and R92A-DGK could be observed compared to the wild type, suggesting that

both arginines are involved in the cross-protomer interactions. The activity of both RxA-

mutants is also clearly decreased (Figure 49).


133

6.3.2.3.4 AMP-PCP bound state of mixed labelled DGK

Additionally, it was tried to determine, if bound nucleotide substrate affects the cross-

protomer contacts significantly. For this purpose, DGK was saturated with the ATP

analogue adenylylmethylenediphosphonate (AMP-PCP) as described in chapter 2 and

5. Its characterization revealed a fully saturated system, which is in addition highly

stable, enabling long MAS NMR experiments. However, an obvious chemical shift

perturbation or a (dis)appearance of cross-peaks is not observable (Figure 50).

Figure 50. DNP-enhanced 15

N−13

C-TEDOR spectra of mixed labelled trimers (U-13

C/U15

N12

C-

DGK). The interprotomer crosspeak does not disappear for each single-site RxA-mutant,

demonstrating that more than one Arg contributes to these interactions. Based on the 3D crystal

structure [6], only Arg81 and Arg92 have an appropriate location at the interface to be involved

in these interactions. This would be in-line with the observed reduction of the cross peak

intensities for R81A- and R92A-DGK. For the AMP-PCP bound state of DGK, no significant

changes of the cross-peak are observable.

6.3.3 Discussion

6.3.3.1 Creating mixed labelled trimers of DGK

The overall procedure to monitor interprotomer contacts is technically challenging. The

first obstacle is to find a proper method for reassembling differently labelled protomers

to active mixed labelled complexes, which includes firstly the separation of oligomers to

protomers. Here, BN-PAGE and LILBID-MS analysis confirmed that DGK is present in

its trimeric form in DDM micelles (Figure 17b, c). For disassembling of DGK, suitable

conditions could be carved out, which enabled to disrupt DGK mainly into monomers,


134

despite the high stability of its trimeric state. Therefore, harsh conditions including the

application of the anionic detergent SDS are necessary to disrupt the oligomeric state,

which however implicates a risk for irreversible denaturation. The conditions, such as

the ratio of protein-to-SDS concentration, have to be fine-tuned. Since DGK is more

stable at higher protein concentrations [141], a low concentration of 0.2 mg/ml had to

be used for disruption by 2% SDS. The concentration dependence can be explained by

the evidence that dissociation and unfolding occurs prior inactivation, which is provided

by the correlation between thermodynamic stability and kinetic stability [141]. If the

protein concentration is too high, the disruption would have been incomplete or

extreme high SDS concentrations would have been necessary. Most important, it could

be verified that the effects of 2% SDS on DGK are reversible. If SDS is displaced by

DDM, the trimeric form of DGK can be retrieved, shown by BN-PAGE analysis (Figure

43d). Additionally, 90% of the enzymatic activity could be recovered this way, illustrated

by the coupled activity assay (Figure 43e). These data are in-line with the findings of

Lau and Bowie, who reported that DGK can be reversibly unfolded by SDS and retains

much of its helical content [9]. Besides, this study shows for the first time that it is

necessary to eliminate SDS completely prior to the reconstitution step. DGK monomers

seem not to reassemble in correctly refolded and active trimers within the lipid bilayer.

Although BioBeads have been reported to be suitable as reconstitution method for SDS

removal [302], they might absorb SDS too fast, leading to aggregation of the unfolded

DGK before a proper refolding and trimer formation can occur. SDS removal via

dialysis instead of BioBeads might be a suitable alternative, since dialysis enables a

slower elimination of detergent. But this can take up to two weeks. The quickest and

more reliable procedure is replacing SDS completely by DDM to generate refolded

trimeric DGK before the reconstitution into lipids occurs, which leads to active

complexes. Taken all, this study demonstrates that mixed labelled oligomers with

preserved activity can be generated for comparably small oligomers like DGK (43 kDa).

6.3.3.2 Statistical analysis of unique interfaces in mixed labelled DGK

The second obstacle results from the intrinsically low sensitivity of MAS NMR. The low

number of cross-protomer spin pairs would be beneath the detection limit of

conventional ssNMR, even if large amounts of membrane protein were present.

Additionally, the number of contacts cannot be controlled during the mixing process,

since the assembly of mixed complexes is statistically defined. The association of 13C-


135

and 15N-labelled monomers to [CN]-DGK causes a certain number of configurations

that are unequally populated.

The population P for a specific configuration is calculated by:

𝑃(𝑁, 𝑘) =𝑁!

𝑘! (𝑁 − 𝑘)! (23)

where N represents the number of monomers (3) and k the number of 15N-labelled

protomers within the mixed labelled complex of [CN]-DGK (0, 1, 2, 3). Each

configuration leads to a certain number of 13C−15N interfaces, I(N,k) (0, 2, 2, 0). The

average number of interfaces per DGK trimer is determined by:

𝐼𝑎𝑣𝑔 = ∑ 𝑃(𝑁, 𝑘) 𝐼(𝑁, 𝑘)𝑁

𝑘=1

∑ 𝑃(𝑁, 𝑘)𝑁𝑘=1

= ∑ 𝑃(3, 𝑘) × 𝐼(3, 𝑘)3

𝑘=0

∑ 𝑃(3, 𝑘)3𝑘=0

= (1 × 0) + (3 × 2) + (3 × 2) + (1 × 0)

8

(24)

= 12

8= 1.5

For a trimer, four distinct configurations are possible, leading to an average number of

1.5 N−C interfaces per complex, of which only 50% are unique interfaces (N→C vs.

C→N) that are conducive to the 15N−13C TEDOR spectra:

𝐼𝑢𝑛𝑖𝑞𝑢𝑒𝑎𝑣𝑔

= ∑ 𝑃(3, 𝑘) × 𝐼𝑢𝑛𝑖𝑞𝑢𝑒(3, 𝑘)3

𝑘=0

∑ 𝑃(3, 𝑘)3𝑘=0

= (1 × 0) + (3 × 1) + (3 × 1) + (1 × 0)

8=

6

8= 0.75 (25)

Thus, just 0.75 specifically labelled interfaces are calculated per DGK trimer, leading to

only 0.75 TEDOR-active Arg-Asp/Glu/Asn interactions. Consequently, this work would

simply not be feasible based on conventional NMR. Fortunately, this hurdle can be

overcome by dynamic nuclear polarization (DNP).

N 3

k 0 1 2 3

P 1 3 3 1

I 0 2 2 0

Iunique 0 1 1 0


136

6.3.3.3 DNP-enhanced 15N-13C TEDOR experiments

Dynamic nuclear polarization using the biradical AMUPol offered a maximal signal

enhancement for DGK of 45-fold (Figure 44a). This level of enhancement could be

reached by an increase of the accessibility and a reduction of the diffusion time of

AMUPol via reducing the volume of the pelleted proteoliposome sample for the

incubation with the biradical. The improvement of the diffusion efficiency might be

necessary due to the comparably low molar protein-lipid-ratio of 1:50. Taken all, these

data confirm that AMUPol is suitable for DNP-enhanced ssNMR on membrane

proteins.

A drawback of DNP are signals from naturally appearing 15N- (~0.4%) as well as 13C-

isotopes (~0.5%), contributing to the spectra pattern of [CN]DGK. They can be

identified by using control experiments ([CC]DGK), which enable to distinguish specific

cross-protomer from natural abundance signals.

This study demonstrates that DNP-enhanced TEDOR experiments allow the detection

of 13C-15N contacts across the protomer interface of trimeric DGK. A cross-peak, which

exposes in the range of Arg-Ne,n, could be visualized (Figure 47).

6.3.3.4 Attemps to assign the cross-peak by RxA-mutants

In order to assign this cross-peak to specific residues, suitably chosen single-site

mutations were introduced into DGK. Therefore, six RxA-mutants were prepared,

covering all arginines (R9, R22, R32, R55, R81 and R92) in DGK. Mixed labelled

complexes of these mutants ([CN]-DGK-RxA) could be produced in the same way as

the wild type, since they feature a similar oligomerization behaviour (Figure 48): All

mutants form trimers in DDM that can be disrupted to mainly monomers by SDS and

reassemble mostly to trimers, when SDS is replaced by DDM. This suggests that the

respective arginines, which are all located in extramembranous parts of DGK, are not

essential for the trimer formation. The driving force for the oligomerization is most likely

the transmembrane domain, which is reported to be more stable than the

extramembranous regions. Lau and Bowie could demonstrate that the denaturation by

SDS occurs in two steps. The first one includes the extramembranous parts, leading to

a partially unfolded intermediate, whereas the second one affects the transmembrane

domain, resulting in a completely unfolded or denatured protein [9].

However, the observed cross-peak did not disappear for each of the six RxA-mutants

(Figure 50), which suggests more than just one interprotomer contact, involving several

arginines. Whereupon, R81A- and R92A-DGK show a reduced cross peak compared


137

to the wild type. Both, Arg81 and Arg92, could be shown by mutational studies to be

functionally relevant (Figure 49) [6, 7], though they are not reported to interact directly

with the nucleotide or lipid substrate [6]. A reason could be a participation in forming

interprotomer contacts, allowing a cross-talk between the protomers or simply by

stabilizing the active site. Due to the crystal structure, a possible interacting partner for

Arg81 is Glu88, located in the CL of the adjacent subunit (Figure 42a) [6]. This contact

might stabilize the loop, which is reported to participate in binding of the nucleotide [6].

It is suggested that the Tyr86 side chain of CL acts as a cover of the nucleotide binding

site in the nucleotide bound state [5]. The Arg81-Glu88 contact possibly keeps the

cover open in the apo state to facilitate binding of the nucleotide. For Arg92, a possible

interacting residue is Asn27, located in the interhelical turn of the adjacent subunit,

linking SH and H1 (Figure 42b) [6]. The possible contact most likely holds both the SH

and the H1 in close proximity towards H2 and H3, stabilizing the active site.

Unfortunately, the contacts cannot be distinguished from one another in the DNP-

enhanced TEDOR spectra (Figure 50). They overlap due to bad spectral resolution,

caused by DNP conditions, involving low field (1H frequency of 400 MHz) and low

temperature (~105 K).

6.3.3.4.1 Drawback of mutations

Not only the bad resolution of DNP spectra poses an obstacle, but also the negative

side-effects of mutations. All RxA-mutants of DGK, used to assign the interprotomer

contacts to specific residues and to elucidate their importance for oligomerization,

feature a significant reduction in activity. This suggests that these mutations have

conformational impact on the protein. Furthermore, it is likely that, if one contact is

omitted, the loss is compensated by another contact. Evidence for that is found in the

TEDOR spectra of R81A- and R92A-DGK with a higher spectral width, allowing the

detection of a possible cross-peak in the lysine range as well (Figure 51). R81A- and

R92A-DGK feature both a significant cross peak in the 15N chemical shift range of

lysine, indicating a participation of this amino acid type in forming an interprotomer

contact, which is not observable for the wild type. This suggests that if either Arg81 or

Arg92 is eliminated, an interprotomer contact involving lysine is formed or strengthened

to compensate the loss of the respective arginine. The arising cross-peak exposes

most likely a salt bridge between 19Lys-Nz and 95Asp-Cg at the protomer interface, as

predicted by the crystal structures (Figure 42c) [5, 6]. Because of the negative side-


138

effects of mutations, double mutants for the identification of cross peaks should not be

taken into account.

Figure 51. DNP-enhanced 2D-TEDOR spectra (tmix = 6.25 ms) of the RxA mutants [CN]-DGK-

R81A and [CN]-DGK-R92A compared to wild-type [CN]-DGK and [CN(Arg,Lys)]-DGK. [CN]-

DGK-R81A and [CN]-DGK-R92A feature both a significant cross peak in the 15

N chemical shift

range of lysine (*), indicating a participation of this amino acid type in forming an interprotomer

contact, which is not observable for the wild type.

6.3.3.5 Assessing the interprotomer contacts during nucleotide binding

Additionally, it was tried to determine, if bound nucleotide substrate affects the cross-

protomer contacts significantly. For this purpose, DGK was saturated with the ATP

analogue AMP-PCP, as described in chapter 5. However, an obvious chemical shift

perturbation or a (dis)appearance of cross-peaks is not observable (Figure 50),

indicating at the first sight no involvement of these Arg-contacts in nucleotide binding.

However, a more detailed look at higher field and temperature with better spectral

resolution would be necessary to judge, if the contacts are really excluded from

nucleotide binding, since a proper detection of conformational changes, identified upon

significant chemical shift changes, is not possible due to the worse line broadening,


139

caused by DNP conditions. Only drastic changes, such as appearance or

disappearance of signals, would lead to unambiguous information about the system

during substrate binding.

6.4 Summary and Outlook

Overall, functionally relevant intra (Arg32-Trp25/Glu28/Ala29; Trp112-Ser61)- and

interprotomer (Arg-contacts) long-range interactions could be detected, which possibly

stabilize the active sites and/or transmit information about substrate binding or changes

of the surrounding lipid bilayer within and between protomers.

Additionally, it could be proven that the highly stable trimeric form of DGK can be

disrupted into monomers by SDS and that, most notably, the effects of SDS on the

protein are reversible, enabling the preparation of fully active mixed labelled trimers.

Furthermore, this study shows that DNP-enhanced TEDOR experiments allow the

detection of 13C-15N contacts across the protomer interface of trimeric DGK. Here,

overlapping cross-peaks, which expose a correlation between Arg-Nn,e and Asp-Cg/

Glu-Cd/ Asn-Cg, could be visualized.

Moreover, this study includes an overall picture with valuable information about all

arginines in DGK, which illustrates that all of them are functionally relevant, but not

essential for the trimer formation.

Next to it, this work demonstrates that mutational studies might not always be the best

choice for the assignment of certain signals, since a mutation holds the risk of

unwanted conformational changes within the protein, leading to ambiguous results.

Unfortunately, working under DNP conditions, involving low field and low temperature,

goes along with bad spectral resolution. Thus, the cross peaks, involving most likely

more than one arginine, could not be assigned to specific residues, since they strongly

overlap. Additionally, a proper detection of conformational changes induced by

nucleotide binding is not possible due to worse line broadening. However, drastic

changes of the cross-peaks during nucleotide binding are not observable.

In order to overcome the issue of bad resolution caused by severe line broadening

associated with the necessity to perform experiments at cryogenic temperatures, DNP

NMR at high magnetic field (800 MHz) in combination with fast MAS (40 kHz) could be

applied. Jaudzems et al. could demonstrate that these conditions yield enhanced

resolution and long coherence lifetimes enabling the acquisition of resolved 2D

correlation spectra. This in turn would allow the detection of better resolved longrange

contacts that can not be observed at room temperature [307].


140

Another option to still assign the detected cross-peaks to specific residues by low field

DNP (400 MHz) and slow MAS (8 kHz), might be the application of more complex

labelling schemes. Here, specific labelling, in detail unique pair labelling, in

combination with a 2D NCOCX experiment could lead to the identification.

Appendix

141

Appendix

Supplementary tables

Table S1. List of chemicals. All were obtained with pro analysis (p.a.) quality.

Chemical Molecular formula Molar mass (g mol-1) Manufacturer

Adenine C5H5N5 135.13 AppliChem

Agar-agar -- -- Roth

Amino acids -- -- AppliChem

Ammonium chloride NH4Cl 53.49 AppliChem

15N-ammonium chloride

15NH4Cl 54.48 Cambridge Isotope Lab.

Ampicilin Natriumsalz C16H18N3NaO4S 371.39 Roth

AMUPol C36H62N4O11 726.90 Bruker

ATP C10H14N5Na2O13P3 551.10 AppliChem

BisTris C8H19NO5 209.24 AppliChem

Calcium chloride CaCl2 * 2H2O 147.01 Sigma-Aldrich

Chloroform CHCl3 119.38 Roth

Cholesterol C27H46O 386.67 Avanti

Coomassie Brilliant Blue G250 C47H48N3NaO7S2 840.01 AppliChem

Cytosine C4H5N3O 111.1 AppliChem

DDM C24H46O11 510.63 AppliChem

Disodium hydrogen phosphat Na2HPO4 141.96 AppliChem

DMPA C31H60O8PNa 614.76 Avanti

DMPC C36H72NO8P 677.93 Avanti

DNase I -- ~31000 AppliChem

DPC C17H38NO4P 351.5 Anatrace

EDTA C10H16N2O8 292.25 AppliChem

Ethanol (96%) C2H6O 46.07 Roth

Glukose C6H12O6 180.16 AppliChem

12C-glucose

12C6H12O6 180.09 Cambridge Isotope Lab.

13C-glucose

13C6H12O6 186.11 Cambridge Isotope Lab.

HEPES C8H18N2O4S 238.31 Roth

Imidazole C3H4N2 68.08 AppliChem

IPTG C9H18O5S 238.30 AppliChem

Isotope labelled amino acids -- -- Cambridge Isotope Lab.

LB-Medium -- -- Roth

LDS Sample Buffer 4X -- -- Thermofisher

Lithiumchlorid LiCl 42.39 abcr

Magnesium chloride MgCl2 * 6H2O 203.30 AppliChem

Magnesium sulfate MgSO4 * 7H2O 246.48 AppliChem

Methanol (99%) CH4O 32.04 Roth

Monopotassium phosphate KH2PO4 136.09 AppliChem

NADH C21H27N7Na2O14P2 709.41 AppliChem

Ni-NTA agarose -- -- Macherey-Nagel

OG C14H28O6 292.38 AppliChem

Phosphoenolpyruvate monopotassium salt C3H4KO6P 206.13 AppliChem

PIPES C8H18N2O6S2 302.37 Roth

Protease Inhibitor cOmplete -- -- Sigma Aldrich

Sodium chloride NaCl 58.44 AppliChem

Thymine C5H6N2O2 126.11 AppliChem

Appendix

142

Uracil C4H4N2O2 112.09 AppliChem

Centrum vitamin tablets -- -- Pfizer Consumer Healthcare

Table S2. List of consumable materials.

Material

Type Manufacturer

Amicon® Ultra-15 Ultracel

® - 10K Merck Millipore Ltd.

Bio-Beads™ SM-2 Adsorbent Media Bio-Rad

Blue Native PAGE Gel Native PAGE Novex BisTris Gel 4-10%, 10 wells Invitrogen

Falcon tube 50, 15 ml Sigma

Glycerol stock beads Roti-Store cryo vials Roth

Microfuge tube Polypropylene 1.5 ml Beckman Coulter®

PD10 column GE17-0851-01-ColumnPD10 GE Healthcare

Petri dishes 90 mm Ø Roth

Plasmid DNA extraction kit NucleoSpin Plasmid Macherey-Nagel

Pipette tips 5000 µl, 1000 µl, 200 µl, 20 µl, 2.5 µl Eppendorf

Plastic tubes 1.5 ml Eppendorf

SDS-PAGE gel RunBlue SDS Gel 4-20% Expedeon

Sterile filter 0.2 µm pore size Sartourius Stedim Biotech

Table S3. List of equipment.

Equipment Type Manufacturer

Autoclave V-75 (75 l volume) Systec

Blue Native PAGE equipment XCell SureLock Mini-Cell System Life Techonologies

Centrifuges

Allegra 21R

Beckman Coulter

Avanti J-E Beckman Coulter

Biofuge Pico Heraeus

GS-15R Beckman Coulter

Gyrotron 263 GHz Bruker

HPLC systems

Äkta Prime GE Healthcare

Äkta Purifier GE Healthcare

Incubators

Innova 44 New Brunswick Scientific

Thermomixer Compact Eppendorf

NMR spectrometers

850 MHz WB Avance III Bruker

400 MHz WB Avance II Bruker

pH meter SevenEasy Mettler Toledo

Rotary evaporator Rotavapor R-200 Büchi

SDS-PAGE equipment Mini-PROTEAN Tetra Biorad

Spectrophotometers V-550 Jasco

NanpDrop 1000 Thermo Scientific

Infinite M200 Tecan Reader

Ultracentrifuge Optima LE-80K Beckman Coulter

Appendix

143

Table S4. Experimental parameters for all multidimensional and dipolar coupling based spectra of DGK in its apo state (white), saturated with AMP-

PCP (light grey), DOG (middle grey) and with AMP-PCP + DOG (dark grey).

dimensionality

2D

experiment PDSD DARR NCA

sample U-13

C,15

N-DGK U-13

C,15

N-DGK-I,L,V U-13

C,15

N-DGK-I,L,V U-13

C,15

N-DGK U-13

C,15

N-DGK + AMP-PCP U-13

C,15

N-DGK + DOG U-13

C,15

N-DGK + AMP-PCP + DOG

figure 21, 22 40a 24 24, 27, 35a/b, 36a, 37a, 38a 33, 35a, 36a, 38a/b 35b, 37a 38b

probehead HCN E-free E-free HCN E-free E-free HCN

recycle delay [s] 2.5 0.8 0.8 3.5 1.0 0.8 2.5

transfer 1 HC-CP HC-CP HC-CP HN-CP HN-CP HN-CP HN-CP

field [kHz] 84(H) 55.6(C) 75.7(H) 55.6(C) 70.9(H) 55.6(C) 72.8(H) 41.7(N) 74.3(H) 41.7(N) 72.1(H) 41.7(N) 76.7(H) 41.7(N)

shape (ramp) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H)

contact time [ms] 1.25 1.20 1.30 1.40 1.40 1.50 1.20

carrier [ppm] 113 113 118.2 118.2 118.2 118.2 118.2

transfer 2 PDSD DARR NCA-DCP NCA-DCP NCA-DCP NCA-DCP NCA-DCP

field [kHz] - 11.8(H) 38(N) 22.8(C) 83.3(H) 38(N) 22.8(C) 100(H) 38(N) 22.8(C) 83.3(H) 38(N) 22.8(C) 83.3(H) 38(N) 22.8(C) 83.3(H)

shape (ramp) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C)

contact/mixing time [ms] 20 800 4.8 3.5 5.0 4.4 3.8

carrier [ppm] 60.9 60.9 60.9 60.9 60.9

T1 increments 1344 1344 176 160 165 176 160

spectral width [kHz] 55.6 55.6 3.8 3.8 3.8 3.8 3.8

aqu. time [ms] 12.0 12.0 23.2 21.0 21.7 23.2 21.1

T2 increments 3390 3390 3380 3390 3380 3380 3380

spectral width [kHz] 100 100 100 100 100 100 100

aqu. time [ms] 17 17 17 17 17 17 17

1H SPINAL decoupling [kHz] 100 83.3 83.3 100 83.3 83.3 83.3

number of scans 72 192 496 208 336 336 128

total measurement time 2d21h 4d21h 20h 1d9h 16h 14h 14h

Appendix

144

dimensionality 2D

experiment NCACX NCOCX

sample U-13

C,15

N-DGK U-13

C,15


C,15

N-DGK

figure 36b 36b 41a

probehead HCN E-free E-free

recycle delay [s] 3.0 1.0 1.2

transfer 1 HN-CP HN-CP HN-CP

field [kHz] 88(H) 41.7(N) 74.3(H) 41.7(N) 73.3(H) 41.7(N)

shape (ramp) 80.100 (H) 80.100 (H) 80.100 (H)

contact time [ms] 1.4 1.4 1.4

carrier [ppm] 118.2 118.2 99

transfer 2 NCA-DCP NCA-DCP NCO-DCP

field [kHz] 38(N) 22.8(C) 100(H) 38(N) 22.8(C) 83.3(H) 38(N) 53.2(C)

83.3(H)

shape (ramp) 90.100 (C) 90.100 (C) 90.100 (C)

contact/mixing time [ms] 4.6 4.2 4.6

carrier [ppm] 57.6 57.6 165

transfer 3 DARR DARR DARR

field [kHz] 14.1 (H) 11.9 (H) 11.8 (H)

contact/ mixing time [ms] 50 50 400

carrier [ppm] 57.6 57.6 165

t1 increments 128 120 104

spectral width [kHz] 3.04 3.04 7.6

aqu. time [ms] 21 19.7 6.8

t2 increments 3390 3390 3390

spectral width [kHz] 100 100 100

aqu. time [ms] 17 17 17

1H SPINAL decoupling [kHz] 100 83.3 83.3

number of scans 1464 1600 2960

total measurement time 6d17h 2d21h 7d6h

Appendix

145

dimensionality 3D

experiment NCACX NCOCX CONCA

sample U-13

C,15

N-DGK U-13

C,15

N-DGK-I,LV U-13

C,15


C,15

N-DGK + DOG U-13

C,15

N-DGK U-13

C,15

N-DGK-I,LV U-13

C,15

N-DGK U-13

C,15

N-DGK-I,LV

figure 26, 36c, 37b not shown 36c 37b 26 not shown 26 not shown

probehead HCN E-free E-free HCN HCN E-free HCN E-free

recycle delay [s] 2.5 0.8 1.0 2.5 2.5 1.0 3.0 1.0

transfer 1 HN-CP HN-CP HN-CP HN-CP HN-CP HN-CP HC-CP HC-CP

field [kHz] 74.1(H) 41.7(N) 60.4(H) 41.7(N) 74.3(H) 41.7(N) 74.7(H) 41.7(N) 72.8(H) 41.7(N) 60.4(H) 41.7(N) 74.1(H) 55.6© 77.1(H) 55.6©

shape (ramp) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H) 80.100 (H)

contact time [ms] 1.80 0.95 1.40 1.40 1.25 0.95 1.50 1.70

carrier [ppm] 118.2 118.2 118.2 118.2 118.2 118.2 176.6 176.6

transfer 2 NCA-DCP NCA-DCP NCA-DCP NCA-DCP NCO-DCP NCO-DCP CON-DCP CON-DCP

field [kHz] 38(N) 22.8© 100(H) 38(N) 22.8© 83.3(H) 38(N) 22.8© 83.3(H) 38(N) 22.8© 100(H) 38(N) 53.2© 100(H) 38(N) 53.2© 83.3(H) 38(N) 53.2© 100(H) 38(N) 53.2© 83.3(H)

shape (ramp) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C) 90.100 (C)

contact/mixing time [ms] 4.0 4.7 4.6 4.5 4.0 4.8 8.0 7.1

carrier [ppm] 57.6 57.6 57.6 57.6 176.6 176.6 118.2 118.2

transfer 3 DARR DARR DARR DARR DARR DARR NCA-DCP NCA-DCP

field [kHz] 12.5 (H) 11.8 (H) 11.9 (H) 13.6 (H) 12.5 (H) 11.8 (H) 38(N) 22.8© 100(H) 38(N) 22.8© 83.3(H)

shape (ramp) 90.100 (C) 90.100 (C)

contact/ mixing time [ms] 50 50 50 50 100 100 5.0 4.8

carrier [ppm] 57.6 57.6 57.6 57.6 176.6 176.6 57.6 57.6

t1 increments 80 74 80 80 64 64 40 46

spectral width [kHz] 3.04 3.04 3.04 3.04 3.04 3.04 2.5 2.5

aqu. time [ms] 13.2 12.2 13.2 13.2 10.5 10.5 7.9 9.1

t2 increments 64 94 72 64 50 50 48 50

spectral width [kHz] 5.1 6.6 5,07 5,07 2.5 2.5 2.5 2.5

aqu. time [ms] 6.3 7.1 7.1 6.3 9.9 9.9 9.5 9.9

t3 increments 3390 3390 3390 3390 3390 3390 2988 2988

spectral width [kHz] 100 100 100 100 100 100 100 100

aqu. time [ms] 17 17 17 17 17 17 15 15

1H SPINAL decoupling [kHz] 100 83.3 83.3 100 100 83.3 100 83.3

number of scans 64 136 88 88 72 152 96 240

total measurement time 9d19h 9d16h 6d9h 13d12h 7d1h 6d3h 6d12h 6d16h

Appendix

146

Table S5. Resonance assignments of wild-type DGK in DMPC/DMPA liposomes by MAS NMR experiments. Chemical shifts are given in ppm.

ResID N CO Ca Cb Cd Cd1 Cd2 Hd Ce Ce2 Ce3 Ne He Cg Cg1 Cg2 Ch2 Nh1/2 Cz Cz2 Cz3

1 Ala

2 Asn

3 Asn

4 Thr

5 Thr

6 Gly

7 Phe

8 Thr

9 Arg

43.30

3.16

84.50 7.35 27.24

71.69

10 Ile

11 Ile

12 Lys

29.21

42.05

2.98

13 Ala

14 Ala

174.70

15 Gly 113.27 177.97 47.07

16 Tyr 119.52 176.61 59.46

133.85

117.47

17 Ser

18 Trp

19 Lys 116.77 180.05 59.25 32.57

25.75

20 Gly 109.11 174.01 46.56

21 Leu 121.55 177.73 57.50 41.67

23.45

26.02

22 Arg 120.40 177.03 59.16 29.78 44.30

29.82

23 Ala 118.36 180.19 54.41 18.39

24 Ala 121.80 176.86 54.81 15.75

Appendix

147

25 Trp 116.31 176.51 60.75 29.28

126.59

110.67

26 Ile 113.33 177.81 63.94 38.49

13.28

28.84 17.01

27 Asn 111.21 174.96 55.36 41.22

176.90

28 Glu 117.46 174.69 53.42 28.63 180.87

32.91

29 Ala 130.72 178.95 55.33 18.24

30 Ala 116.25 179.25 54.83 18.01

31 Phe 114.63 176.77 61.36 39.31

131.20

139.33

32 Arg 116.47 178.04 58.93 31.50 43.94

83.32

27.90

69.98 158.75

33 Gln 115.09 179.52 59.36 27.86

33.84

34 Glu 115.68 177.02 59.05 26.95

34.39

35 Gly 107.72 174.59 46.99

36 Val 119.17 176.50 67.07 32.15

21.55 21.55

37 Ala 118.85 178.14 54.92 19.35

38 Val 117.04 178.16 67.30 31.08

39 Leu 117.44 178.32

40 Leu 117.35 177.74 58.25 41.14

26.94

41 Ala 117.24 179.25 55.11 18.01

42 Val 118.13 177.96 67.26 31.02

22.94 22.94

43 Val 119.61 177.79 67.72 30.75

22.26 22.26

44 Ile 118.74 177.77 65.96 37.56

13.33

29.75 17.01

45 Ala 120.51 179.15 54.95 18.83

46 Cys 113.21 172.57 63.16 27.62

47 Trp 121.36 176.97 58.17 30.72

48 Leu 116.03 176.56 54.88 44.48

23.35

26.37

49 Asp 123.00 176.03 52.59 39.28

50 Val 109.85 174.87 58.02 33.38

20.29 17.35

Appendix

148

51 Asp 117.78 174.53 52.71 41.67

180.04

52 Ala 121.13 178.54 55.39 19.18

53 Ile 116.01 177.28 65.31 37.14

13.35

29.67 18.89

54 Thr 115.95 175.45 67.45

21.12

55 Arg 119.97 177.62 60.87 29.37 42.70

27.97

56 Val 116.13 178.36 67.23 31.01

24.37 23.39

57 Leu 122.24 179.87 57.67 42.60

26.62

58 Leu 120.34 179.05 57.67 39.96

25.74 24.39

27.02

59 Ile 116.14 178.89 65.29 38.99

16.12

29.00 17.93

60 Ser 116.17 177.13 62.56

61 Ser 115.07 177.34 61.49 63.01

62 Val 114.40 177.57 64.24 30.41

23.31 19.38

63 Met 119.86 179.25 57.15 32.89

30.61

64 Leu 120.25 178.18 57.71 39.98

22.43

25.13

65 Val 115.30 176.57 66.77 31.18

21.99 20.26

66 Met 112.89 177.98 56.64 31.47

67 Ile 119.09 176.63 65.77 38.01

14.24

28.93 17.25

68 Val 116.36 177.23 66.70 31.02

23.23 22.87

69 Glu 120.91 179.11 58.01 29.52 181.23

35.33

70 Ile 122.38 177.26 66.32 37.18

14.83

26.56 19.48

71 Leu 121.20 178.28 57.95 41.68

22.05

26.56

72 Asn 118.20 176.83 56.50 38.45

178.26

73 Ser 115.97 176.06 62.25 63.04

74 Ala 126.91 178.33 55.37 17.54

75 Ile 118.39 177.28 65.09 36.97

12.81

29.34 16.61

76 Glu 119.75 176.88 59.99 28.77 182.67

35.65

Appendix

149

77 Ala 119.70 178.92 55.01 17.00

78 Val 116.34 177.02 66.27 30.86

23.53 22.96

79 Val 119.21 179.05 67.17 30.95

24.33 22.95

80 Asp 120.95 178.21 56.46 39.41

81 Arg 122.54 178.18 57.45

42.24

84.30

27.50

71.70 158.13

82 Ile 121.11 176.84 64.49 37.96

15.16

28.86 16.32

83 Gly 105.00 172.69 44.96

84 Ser 121.37 179.12 58.34

85 Glu

86 Tyr

87 His

88 Glu 129.84 179.21 59.55 29.52

35.36

89 Leu 121.29 179.02 57.31 41.66

27.08

90 Ser 116.49 175.47 62.30 62.56

91 Gly 106.74 174.75 47.04

92 Arg 120.47 177.89 59.02 31.36 44.27

80.77

25.16

76.04 159.39

93 Ala 119.23 179.12 55.82 18.80

94 Lys 114.08 179.99 59.41 32.43 29.86

41.71

26.55

95 Asp 122.43 178.72 57.58 39.11

96 Met 119.97 177.09 60.26 33.92

16.91

32.09

97 Gly 106.66 175.57 47.55

98 Ser 113.65 177.28 61.82 62.57

99 Ala 125.37 178.14 54.34 17.85

100 Ala 120.41 178.62 55.58 17.94

101 Val 116.25 177.43 66.17 31.30

21.17

102 Leu 119.23 178.20 58.16 40.74

23.80 22.02

27.04

Appendix

150

103 Ile 115.82 177.00 64.63 36.09

11.07

27.45 18.79

104 Ala 121.52 179.59 55.96 17.02

105 Ile 120.16 149.30 66.03 37.44

12.98

30.73 16.95

106 Ile 120.13 177.49 66.12 37.14

12.92

29.40 16.71

107 Val 118.35 178.58 67.08 31.06

22.88 22.88

108 Ala 126.54 178.22 56.07 16.67

109 Val 117.62 178.17 67.19 30.99

22.00 22.91

110 Ile 118.96 176.72 66.29 38.40

13.27

29.78 17.04

111 Thr 116.61 175.14 67.90 68.50

20.77

112 Trp 120.50 177.88 62.90 27.39

125.68 130.41

137.55 118.98 130.66

123.76

114.76 120.84

113 Cys 115.39 176.33 64.95 27.45

114 Ile 116.27 178.76 65.74 37.70

13.75

27.38

115 Leu 116.25 180.15 57.44 40.71

22.52

26.63

116 Leu 118.76 178.24 57.69 40.39

21.84

27.09

117 Trp 115.61 175.24 64.57 26.34

118 Ser 115.44 175.38

119 His

120 Phe

121 Gly

Appendix

151

Table S6. Summary of all significant perturbations in peak position and intensity

during the interaction of DGK with its substrates.

*Arg9 and Lys12 are detected in scalar coupling based experiments. The other

residues are observed in dipolar coupling based experiments.

AMP-PCP DOG

weighted CSP peak intensity weighted CSP peak intensity

9 Arg Cg x reduced*

Cd x reduced*

Ne x reduced*

Nh1/h2 x reduced*

12 Lys Cd x reduced*

Ce x reduced*

15 Gly Ca x disappeared x reduced

CO x disappeared x reduced

16 Tyr Ca x disappeared

Cb x disappeared x disappeared

Cd1 x disappeared

Ce2 x disappeared x disappeared

CO x disappeared x reduced

19 Lys Cb x disappeared

Cg 0.39 increased

CO x reduced

20 Gly Ca 0.22 increased

CO 0.22 increased

22 Arg Cd x disappeared

23 Ala Cb 0.35 x

CO 0.26 x

25 Trp Cg 0.38 x

26 Ile Cb 0.22 reduced

Cg1 0.22 x

Cg2 x increased

Appendix

152

Cd x increased x disappeared

27 Asn Cb 0.20 x

Cg 0.21 x

28 Glu Cg 0.40 reduced x reduced

Cd x disappeared

CO 0.21 x

29 Ala Ca 0.22 x

Cb 0.23 x

CO 0.31 x

31 Phe Cb 0.27 x

Cg 0.58 reduced x reduced

Cd 0.57 x x reduced

Ce x appears

32 Arg Cb 0.24 reduced

Cg 0.41 x

Cd 0.25 reduced

33 Gln Cb x reduced x reduced

Cg 0.54 reduced 0.31 reduced

CO 0.23 reduced

37 Ala Ca x reduced

Cb 0.21 reduced

CO x reduced

40 Leu Cd x appears

41 Ala Cb 0.30 x

43 Val Cb 0.35 x

Cgb 0.36 x

45 Ala Cb 0.22 x

CO 0.47 x

46 Cys Ca 0.23 reduced

Cb 0.27 reduced

Appendix

153

CO 0.22 reduced

47 Trp Cb x disappeared

48 Leu Cd 0.24 x

49 Asp Ca 0.34 x

Cb 0.38 increased x increased

CO 1.29 increased x increased

50 Val Ca x disappeared x increased

Cb x disappeared x increased

Cgb x disappeared x increased

Cga x disappeared x increased

CO x disappeared x increased

51 Asp Ca x reduced

Cb x reduced

Cg 0.33 reduced

CO x reduced

53 Ile Cg2 0.40 x

CO 0.28 x

55 Arg Cb 0.27 reduced

Cg x reduced

Cd 0.28 reduced

56 Val Cg1 0.32 x

57 Leu CO 0.22 reduced

62 Val Cb 0.38 x

Cg2 0.23 x

Cg1 0.43 x

65 Val CO 0.29 x

66 Met Ca 0.21 reduced

Cb x reduced

CO 0.27 x

Appendix

154

67 Ile Cg1 x reduced

Cd x reduced

69 Glu Cb 1.20 reduced 0.35 increased

Cg x disappeared

Cd x disappeared x increased

70 Ile Ca 0.28 x

Cb 0.38 x

Cg1 0.31 x 0.35 reduced

Cg2 0.39 x 0.24 x

Cd 0.47 increased 0.29 increased

CO 0.27 x

72 Asn Ca 0.20 x

74 Ala Cb 0.32 x

76 Glu Cb x reduced

Cg 0.26 reduced

Cd x disappeared x increased

77 Ala CO 0.22 x

79 Val Cb 0.28 x

Cg1 0.45 x

Cg2 0.28 x

CO 0.44 x

80 Asp Ca 0.36 x

Cb 0.25 x

CO 0.28 x

81 Arg Cg 0.39 x

Cd 0.62 x

Cz x reduced

82 Ile Ca 0.62 reduced

Cb 0.44 reduced

Cg1 0.43 reduced

Cg2 0.54 reduced

Cd2 0.48 reduced

Appendix

155

CO 0.54 reduced

83 Gly Ca x increased

CO x increased

88 Glu Ca 0.39 reduced

Cb 0.29 reduced

Cg x disappeared

CO 0.21 reduced

91 Gly Ca 0.70 x

CO 0.80 x

92 Arg Ca x reduced

Cb 0.24 reduced

Cg x reduced

Cd x reduced

CO x reduced

94 Lys Cb 0.27 x

Cg 0.28 x

Cd 0.76 x

Ce 0.20 x

95 Asp Cb x reduced

97 Gly Ca 0.23 x

CO 0.31 x

98 Ser Ca x reduced

Cb x reduced

CO x reduced

99 Ala Ca x reduced

Cb x reduced

CO x reduced

101 Val Cg1 0.30 x

102 Leu Ca 0.29 x

Cd1 0.21 x

Appendix

156

111 Thr CO 0.28 x

114 Ile Cg1 x increased

Cd x increased

116 Leu Ca x disappeared

Cb x reduced

Cg x reduced

Cd x disappeared

CO x disappeared

Appendix

157

List of abbreviations

ADP adenosine 5‘-diphosphate

AMP-PCP adenylylmethylenediphosphonate

ATP adenosine 5‘-triphosphate

a.u. arbitrary units

BMRB biological magnetic resonance bank

BN-PAGE blue native-polyacrylamid gel electrophoresis

BSA albumin from bovine serum

CE cross effect

CHAPSO 3-([3-cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate

CL cytosolic loop

CMC critical micelle concentration

CP cross polarization

Cryo-EM cryo electron microscopy

CSA chemical shift anisotropy

CW continuous wave

DAG diacylglycerol

DCP double cross polarization

ddH2O double-distilled water

DGK diacylglycerol kinase

DARR dipolar assisted rotational resonance

DBG 1,2-dibutyryl-sn-glycerol

DM n-decyl-β-glucopyranoside

DDM n-dodecyl-β-glucopyranoside

DMPA 1,2-dimyristoyl-sn-glycero-3-phosphate

DMPC 1,2-dimyristoyl-sn-glycero-3-phosphocholine

DNA deoxyribonucleic acid

DNP dynamic nuclear polarization

DOG 1,2-dioctanoyl-sn-glycerol

DOPC 1,2-dioleoyl-sn-glycero-3-phosphocholine

DOPG 1,2-dioleoyl-sn-glycero-3-phosphoglycerol

DPC n-dodecylphosphocholine

DPPC 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine

DSS 4,4-dimethyl-4-silapentane-1-sulfonic acid

EDTA ethylenediaminetetraacetic acid

FID free induction decay

Gd3+

-DOTA gadolinium 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid

GPCR G-protein coupled receptor

H1-3 helix 1-3

HEPES 4-(2-hydroxyethyl)-1-piperazine-1-ethanesulfonic acid

HETCOR heternuclear correlation experiment

HPLC high-performance liquid chromatography

IMAC immobilized metal affinity chromatography

IPTG isopropyl β-D-1-thiogalactopyranoside

KD dissociation constant

KM Michaelis-Menten constant

Appendix

158

LB lysogeny broth

LCP lipidic cubic phase

LDH lactate dehydrogenase

LDS lithium dodecyl sulfate

LILBID-MS liquid beam ionization/desorption mass spectrometry

M molecular weight

MAG monoacylglycerol

MAS magic angle spinning

MDO membrane derived oligosaccharides

MDS molecular dynamics simulation

NAD nicotinamide adenine dinucleotide

Ni-NTA Nickel-nitrilotriacetic acid

NMR nuclear magnetic resonance

ns number of scans

OD optical density

OE Overhauser effect

OG n-octyl-β-D-glucopyranoside

PA phosphatidic acid

PC phosphocholine

PCR polymerase chain reaction

PDB protein data bank

PDSD proton driven spin diffusion

PE phosphoethanolamine

PG phosphatidyl glycerol

Pi inorganic phosphorus

PK pyruvate kinase

PKC protein kinase C

PL periplasmic loop

ppm parts per million

PRE paramagnetic relaxation enhancement

REDOR rotational echo double resonance

RF radiofrequency

RT room temperature

S/N signal-to-noise ratio

SDS sodium dodecyl sulfate

SDS-PAGE SDS polyacrylamid gel electrophoresis

SE solid effect

SEC size exclusion chromatography

SH surface helix

SPINAL small phase incremental alternation

ssNMR solid state nuclear magnetic resonance

T interhelical turn

T1 spin lattice relaxation constant

T1ρ spin lattice relaxation constant in the rotating frame

T2 spin-spin relaxation constant

TEDOR Transferred echo Double Resonance

TEPS triethylphosphine

Appendix

159

TM transmembrane (helix)

TPPM two phase pulse modulation

Tris tris(hydroxymethyl)aminomethane

UV/VI ultraviolet/visible

v/v volume per volume

wt wild type

w/v weigth per volume

1D one dimensional

2D two dimensional

Appendix

160

List of figures

Figure 1. Physiological role of E.coli diacylglycerol kinase (DGK) in recycling during the

biosynthesis of membrane-derived oligosaccharides (MDOs) [1, 2] that are largely

generated in response to environmental stress, such as low osmolarity [15, 16]. DGK is

located within the inner membrane, where it catalyzes the ATP-dependent

phosphorylation of potentially membrane-disruptive diacylglycerol (DAG) to non-toxic

phosphatic acid (PA), providing the basis for restoring phosphatidylglycerol (PG), which

is consumed in the MDO cycle. The cartoon is based on the X-ray structure using the

PDB ID 4UXX [6]. The figure is adapted from Van Horn et al. [84]. ............................... 9

Figure 2. Sequence alignment of wild-type DGK and the two thermostable mutants,

Δ4- and Δ7-DGK [5]. The mutations in Δ4- and Δ7-DGK are labelled green. The N-

terminal tag is highlighted orange. .............................................................................. 14

Figure 3. Topology plot of wild-type DGK. The plot was created based on the DGK X-

ray structure [5] and refined by the CSI values obtained from chemical shifts in this

study (Table S5). The membrane is indicated by two solid black lines as calculated in

the PPM server [116]. The secondary structure elements of DGK are denoted as: CL,

cytoplasmic loop; H1-3, helices 1-3; PL, periplasmic loop and SH, surface helix. ....... 15

Figure 4. Comparison of the solution NMR (PDB 2KDC, wild-type DGK) and crystal

structure of DGK (PDB 3ZE5, Δ4-DGK). A view of the crystal (right) and solution NMR

(left) structure from the cytoplasm (a) and the membrane (b) plane. .......................... 17

Figure 5. Comparison of the DGK secondary structures obtained from solution NMR

(PDB 2KDC, wild-type DGK, chain A), solid state NMR [14], and X-ray crystallography

(PDB 3ZE5, Δ4-DGK, chain A): Rectangles symbolize α-helical regions, whereas solid

lines reflect deviations from helicity. Residues that were not resolved are illustrated by

dashed lines. The differences between the secondary structures are highlighted in

green. Both the ssNMR and the X-ray studies used a thermostabilized mutant, whereas

the wild type was used for solution NMR structure determination. ............................... 19

Figure 6. Substrate-binding sites determined in the X-ray structure (Δ4-DGK, PDB

4UXX) [6]. (a) Structure-based and possible interactions with the non-hydrolysable ATP

analogue adenylylmethylenediphosphonate (AMP-PCP, blue) and its two counterions

(Zn, orange). The figure is adapted and modified from Li et al. [6]. (b) Possible

interactions of Ser17, Glu69 and Ser98 with the lipid substrate monoacylglycerol (MAG,

yellow). ....................................................................................................................... 22

Appendix

161

Figure 7. Impact of magic angle spinning (MAS) at 54.74° on solid state NMR spectra.

(a) Depiction of a MAS rotor that is tilted in the magic angle β = 54.74° with respect to

the magnetic field B0. 15N (b)- and 1H (c)-NMR spectra of the microcrystalline tri-peptide

N-formyl-Met-Leu-Phe-OH under static (blue) and MAS (red) conditions. In the static

15N-NMR spectrum, the isotropic (δiso) and the anisotropic (δaniso) chemical shift as well

as the CSA parameters δxx, δyy and δzz are labelled accordingly. Comparing the 15N-

and 1H-NMR spectra under MAS of 25 kHz, it becomes obvious that higher spinning

speeds are needed to obtain well-resolved 1H-NMR spectra (see chapter 4, outlook),

whereas the 15N-NMR spectrum features already a good resolution at 25 kHz. The

figures are adapted from the lecture script “Solid state NMR”, prepared by Prof.

Clemens Glaubitz, Goethe University Frankfurt am Main, summer semester 2015. .... 32

Figure 8. Pulse sequence of a cross polarization (CP) experiment according the

Hartmann-Hahn condition, in which magnetization is transferred from highly abundant I

spins to dilute S spins. ................................................................................................ 34

Figure 9. Pulse sequence of a 2D 13C-13C PDSD experiment. During the preparation

time, the magnetization is transferred from 1H to 13C via CP step. This is followed by

the t1 period, when the 13C chemical shift evolves, while 1H nuclei are decoupled.

Thereafter, the magnetization is transferred back on the z-axis through a 90° pulse on

the 13C spins and the mixing step takes place, in which the proton driven spin diffusion

between 13C nuclei occur through space by flip-flop interactions. For detection, the 13C

spins are transferred back in the x,y-plane and the FID is recorded under 1H

decoupling. ................................................................................................................. 36

Figure 10. (a) Pulse sequence of the NCACX and NCOCX experiment. Both are 15N-

13C correlation transfer experiments with a subsequent 13C-13C mixing step. During the

preparation period, a broad-band 1H15N-CP step is used to generate 15N polarization

that evolves during t1 under proton decoupling. For the 15N-13C transfer, optimized spin

lock fields on the 15N and 13C channel are applied under proton decoupling. The 13C off-

set is centered in the Cα region for NCA and in the CO region for NCO. After the

double cross polarization (DCP), evolution on 13Cα/13CO takes place under proton

decoupling during t2. Subsequently, a DARR step follows, which transfers the

magnetization to any other proximate 13C nuclei. Therefore, two 90° pulses at an off-set

of 100 ppm were applied for excitation and reconversion of longitudinal magnetization.

The detection of 13C magnetization (t3) represents the final step, during which the

protons are decoupled. (b) Polarization transfer pathway for the NCACX (left) and

NCOCX (right) pulse sequence, schematically illustrated for a di-peptide. The selected

Appendix

162

off-set frequencies on Cα or CO enable a magnetization transfer within the same amino

acid [i] along the side chain, resulting in cross peaks in the NCACX spectrum, or along

the side chain of the previous amino acid [i-1], leading to cross peaks in the NCOCX

spectrum, respectively (black arrows). Additional through-space dipolar-assisted

pathways are possible as well (grey arrows). .............................................................. 38

Figure 11. (a) Pulse sequence of the CONCA experiment. It is a 15N-13C correlation

transfer experiment accomplished by three CP steps. During the preparation period, a

selective 1H13C-CP step is used to generate 13C polarization that evolves during t1

under proton decoupling. This is followed by a second selective CP step, which

transfers magnetization from 13CO[i-1] to 15N[i]. The 13C off-set is centered in the CO

region for CON. The magnetization then evolves on 15N[i] during t2 under proton

decoupling. Subsequently, the third CP step takes place, transferring magnetization

from 15N[i] to 13Cα[i]. The 13C off-set is centered in the Cα region for NCA. The

detection of 13C magnetization (t3) represents the final step, during which the protons

are decoupled. (b) Polarization transfer pathway for the CONCA pulse sequence,

schematically illustrated for a di-peptide. The selected off-set frequencies on CO and

later Cα enable a magnetization transfer from CO of the previous amino acid [i-1] via N

towards Cα (black arrow) and sometimes Cβ (grey arrow) of the following amino acid

[i], leading to cross peaks in the CONCA spectrum (black arrows). ............................. 39

Figure 12. Pulse sequence of the 13C-13C TOBSY experiment. The magnetization is

transferred from 1H to attached 13C nuclei through a refocused INEPT step based on J-

couplings: After the initial 90º 1H pulse, 1H chemical shift evolution during the variable

t1 period takes place. The evolution delay is fixed to achieve antiphase 1H

magnetization with respect to 13C via JHC (JHC ~200 Hz). The magnetization is

transferred to 13C by applying simultaneous 90º 1H and 13C pulses. The 13C chemical

shift evolution during the variable t2 period takes place. This is followed by isotropic 13C

mixing using TOBSY. The 13C-13C correlation is established through carbon-carbon J-

couplings (JCC ~35-53 Hz). The detection of 13C magnetization represents the final step

(t3), during which the protons are decoupled. .............................................................. 40

Figure 13. The expected low number of interacting spin pairs make dynamic nuclear

polarization for signal enhancement essential, which is obtained by the three spin cross

effect using the biradical AMUPol as a source for unpaired electrons. (a) Reconstituted

mixed labelled DGK doped with AMUPol [133] is depicted. (b) It is subjected to

continuous wave microwave irradiation, resulting in polarization transfer from electrons

via protons to the sites of interest................................................................................ 42

Appendix

163

Figure 14. TEDOR pulse sequence [136]. The sequence starts with a CP-step,

transferring magnetization from 1H to 13C nuclei. This is followed by two REDOR-steps

(tmix/2) to reintroduce heteronuclear dipolar couplings, which are enclosed by 90°

pulses. L0 describes the number of 180° pulses during tmix/4. The evolution time (t1)

takes place in-between the two REDOR steps. The detection of 13C magnetization (t2)

represents the final step, during which the protons are decoupled. ............................. 44

Figure 15. Sequence alignment of wild-type DGK and the thermostable mutant, Δ4-

DGK [89]. The mutations in Δ4-DGK are labelled green. The N-terminal tag is

highlighted orange. ..................................................................................................... 45

Figure 16. (a) Comparison of the activity of wt- (dark grey) and Δ4-DGK (red)

reconstituted into DMPC/DMPA. Both samples are prepared the same way. (b)

Comparison of the activity of reconstituted wtDGK, prepared in DDM (dark grey) and

DPC (red). 100% activity corresponds to the rate recorded with wtDGK in 90mol%

DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1 mg-1. All activity measurements were

repeated three times. The activity was calculated as the mean value. Error bars

correspond to standard deviations. (c) Size exclusion chromatography (SEC) of wtDGK

in 0.05% DDM (c1) and 0.5% DPC (c2), performed after several weeks at 4°C or

immediately after the IMAC step, respectively............................................................. 68

Figure 17. Characterization of the purity and oligomeric state of wtDGK in DDM

micelles. (a) The SDS-PAGE verifies the purity of the protein solution after IMAC

purification. (b) The BN-PAGE offers a reliable assessment of the oligomeric state,

clearly showing wtDGK as trimer in DDM micelles. (c) The trimeric state of wtDGK in

DDM micelles is confirmed by LILBID-MS: The signals for the monomeric, dimeric and

trimeric form of wtDGK are labelled by “1”, “2” and “3”, respectively. They occur at

charged states of −1 and −2. The LILBID mass spectrum was recorded by Oliver Peetz

of the research group of Prof. Dr. Nina Morgner (Institute of Physical and Theoretical

Chemistry, Goethe University Frankfurt am Main). ...................................................... 69

Figure 18. Superimposed 2D 13C-13C PDSD spectra of U-13C,15N-DGK embedded into

90mol% DMPC/ 10mol% DMPA (black) or 67.3mol% DMPC/ 32.7mol% cholesterol

(red), showing a similar fingerprint. The enlargement of a representative region in the

2D 13C-13C PDSD spectra displays clearly a reduced resolution for 67.3mol% DMPC/

32.7mol% cholesterol (red). For instance, the selected peak, P, features a line width in

F1 dimension of 138 Hz and 406 Hz and in F2 dimension of 574 Hz and 1006 Hz for

Appendix

164

90mol% DMPC/ 10mol% DMPA (black) and 67.3mol% DMPC/ 32.7mol% cholesterol

(red), respectively. The line widths were obtained from CCPN analysis 2.4.1 [175]. ... 73

Figure 19. Comparison of the activity of wtDGK reconstituted into DMPC/DMPA with

different molar protein-to-lipid ratios increasing from 1:80 to 1:20. The ratio of 1:50

(grey) is used for all subsequent studies. 100% activity corresponds to the rate

recorded with wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1 mg-1.

Experiments were repeated three times. The activity was calculated as the mean value.

Error bars correspond to standard deviations.............................................................. 74

Figure 20. Sucrose density gradient (40%-10%) centrifugation of empty liposomes

(left) and wtDGK reconstituted in DMPC/DMPA in a molar protein-to-lipid ratio of 1:50

(right), revealing homogeneous size distribution of proteoliposomes without any empty

liposomes observable. ................................................................................................ 74

Figure 21. Evaluation of the proteoliposom sample by MAS NMR. 1D 13C and 15N

cross polarization (CP) spectra of U-13C,15N-wtDGK reconstituted into DMPC/DMPA,

exhibit a good spectral resolution (a). 2D 13C-13C PDSD spectrum of U-13C,15N-wtDGK

reconstituted into DMPC/DMPA. A mixing time of 20 ms was used to yield one bond

correlations between aliphatic atoms. The spectrum evaluates structural homogeneity,

resolution and secondary structure. The high number of well resolved peaks

demonstrates a homogeneous sample preparation (b). .............................................. 75

Figure 22. Comparison of MAS ssNMR data of the wild type with its thermostable

mutant. 2D 13C-13C PDSD spectrum of U-13C,15N-wtDGK, recorded with a mixing time

of 20 ms (left). Enlargement of the selected region in the 2D 13C-13C PDSD spectrum

(right). The spectrum is compared with the assignment of the thermostable mutant of

DGK gained by MAS NMR [14]. The comparison shows that a transfer of these

assignments to the wild type sample is not possible. The red cross peaks were

obtained with CCPN analysis 2.4.1 [175]. ................................................................... 77

Figure 23. Examples of membrane protein structures determined in phospholipids by

solid state NMR. (a) Anabaena sensory rhodopsin, ASR: PDB 2m3g, with bound retinal

(yellow) [174]. (b) human chemokine receptor, CXCR1: PDB 2lnl [179]. (c) M2 1H

channel from influenza virus: PDB 2l0j [180, 181]. (d) Bacterial inner membrane protein

DsbB: PDB 2leg [182, 183]. (e) Mycobacterium cell division protein, CrgA: PDB 2mmu

[184]. (f) Membrane-inserted form of the fd bacteriophage coat protein: PDB 1mzt

[185]. The figure is adapted from Marassi and Opella [186]. ....................................... 78

Appendix

165

Figure 24. Comparison of the 2D NCA spectra of uniform labelled U-13C,15N-wtDGK

(black) and reverse labelled U-13C,15N-wtDGK-I,L,V (green). The 15N and 13Ca signals

of the uniform labelled U-13C,15N-wtDGK exhibit already comparable good average

linewidths of approximately 105 and 185 Hz, respectively. To resolve residual

ambiguities, Ile, Leu and Val are specifically unlabelled in the reverse labelled sample.

The inscriptions for Ile, Leu and Val are labelled in red and for the other amino acids in

black. In the 2D NCA of the reverse labelled sample, the peaks for Ile, Leu and Val are

clearly missing as expected. Apart from that, the two NCA spectra do not remarkably

differ from each other. ................................................................................................. 86

Figure 25. 15N CP spectra of U-13C,15N-DGK. The spectra were recorded with an E-

free (green) and with a standard (black) 3.2 mm triple-resonance HCN MAS probehead

(Bruker). In both cases 128 scans were applied. The E-free probehead enables to use

a recycle delay of 0.8 s, saving ~3x of the measurement time compared to the standard

probehead with a recycle delay of 2.5 s. The E-free probehead was custom-built and is

still under development. .............................................................................................. 87

Figure 26. Resonance assignment of U-13C,15N-wtDGK based on a set of 3D NCACX,

NCOCX and CONCA spectra. A representative sequential walk from I26 to A29 is

shown. Each set of three spectra represents a Cx[i−1]–N[i]–Cx[i] spin system. For

example, the N27 NCACX peaks are connected to the I26-N27 CONCA peak via the

same N and Ca. The I26 NCOCX peaks are linked to the I26–N27 CONCA peak

through the same N and CO, resulting in a Cx[i−1]–N[i]–Cx[i] system, which is linked

with the preceding system through all carbon shifts of I26 that are visible in both

NCACX and NCOCX spectra. The assignments are depicted by lines. ....................... 89

Figure 27. 2D NCA spectrum of U-13C,15N-DGK with assigned peaks labelled. .......... 90

Figure 28. Automated resonance assignment by ssFLYA confirms 91.5% of the

backbone and 89.1% of all (backbone + side chains) assignments obtained manually.

Assignments are classified as strong, if ≥ 80% of the individual chemical shift values

from 20 independent runs of the algorithm differ by less than 0.55 ppm from the

consensus value (strong colors). Other assignments by ssFLYA are graded as weak

(light colors). From other studies by ssFLYA, they are known to be erroneous for 39 –

72% [13]. Each assignment for an atom is symbolized by a colored rectangle: green -

assignment by ssFLYA agrees with the manual reference assignment within a

tolerance of 0.55 ppm; red - assignment does not match with the reference; blue -

assigned by ssFLYA, but not manually; black – assigned manually, but not by ssFLYA.

Appendix

166

The second row illustrates backbone assignments for N, Ca, and CO. The third to

eighth row represent the side chain assignments. For branched side chains, the

relevant row is subdivided into an upper part for one branch and a lower part for the

other branch. ssFLYA was performed by Dr. Sina Kazemi of the research group of

Prof. Dr. Peter Güntert (Institute for Biophysical Chemistry, Goethe University Frankfurt

am Main). He also kindly provided this figure. ............................................................. 91

Figure 29. 2D scalar coupling based 1H-15N HETCOR (a), 1H-13C HETCOR (b) and

13C-13C TOBSY (c) of U-13C,15N-DGK with tentative assignments. All residues, which

could not be detected and assigned by dipolar coupling based experiments are

considered as possible candidates for detection by experiments based on scalar

coupling. INEPT and TOBSY were applied for 1H-15N or 1H-13C heteronuclear

polarization and 13C-13C homonuclear mixing, respectively. Peaks for Arg9 and Lys12

are labelled green, as they could be assigned unambiguously. Peaks for the aromatic

rings were folded in the indirect dimension to save measurement time. Amino acids

that refer to the His-tag are labelled by ‘tag’. ............................................................... 93

Figure 30. Resonance assignment of DGK. (a) Each assignment for an atom is

symbolized by a blue rectangle: The second row illustrates backbone assignments for

N, Ca, and CO. The third to eighth row represent the side chain assignments. For

branched side chains, the relevant row is subdivided into an upper part for one branch

and a lower part for the other branch. This figure was kindly provided by Dr. Sina

Kazemi of the research group of Prof. Dr. Peter Güntert (Institute for Biophysical

Chemistry, Goethe University Frankfurt am Main). (b) The assigned residues are

mapped on the topology plot of DGK. The plot was created with respect to the X-ray

structure of DGK (PDB 3ZE5) [5] and refined by CSI values obtained from chemical

shifts (Table S5). The membrane is depicted by two solid black lines. 84% residues of

DGK were assigned by dipolar and scalar coupling based experiments...................... 94

Figure 31. Secondary structure analysis based on the chemical shifts. The chemical

shift index (CSI) Δδ is derived from the difference between the experimentally

determined MAS NMR chemical shifts (exp) for Ca and Cb and their random coil

standard chemical shifts (rc) according to Δδ =[δCa(exp)-δCa(rc)] - [δCb(exp)-δCb(rc)]

[206]. For Gly residues and residues without any assignment of Cb, only Ca secondary

shifts were used. Strongly positive (≥ 1.5 ppm) values of the CSI imply an α-helical

structure, whereas negative or near-zero values indicate deviations from helicity. (a)

The secondary structure of wild-type DGK determined by MAS NMR is compared with

the MAS NMR structure of the thermostable mutant [14] and the X-ray structure of

Appendix

167

wtDGK (PDB 3ZE4, chain A) [6]. Rectangles represent α-helical regions involving the

surface helix (SH) and the three transmembrane helices (H1-3), whereas solid lines

symbolize deviations from helicity including the interhelical turn (T), the periplasmic

(PL) as well as the cytoplasmic loop (CL). Residues that were not resolved by ssNMR

or by X-ray crystallography are depicted by dashed lines. Disparities between the three

secondary structures are highlighted in green. (b) The 2D NCACX spectrum of U-

13C,15N-DGK shows all assigned glycines. The regions for helical and random coil (rc)

secondary structure are coloured. Gly83 and Gly91 are labelled bold, since the DGK X-

ray structure exhibits asymmetries for both residues: Both were observed within a

helical and a random coil structure [5]. These asymmetries are not detectable by MAS

NMR. .......................................................................................................................... 96

Figure 32. Solid state NMR 1H detected 2D 1H-15N correlation spectrum (2D hNH) of

fully protonated U-13C,15N-DGK in phospholipid bilayers. The spectrum was recorded

on a Bruker 600 MHz spectrometer at ~278.15-283.15 K and a MAS rate of 111 kHz

(Bruker 0.7 mm rotor, ~0.5 mg sample). It was conducted with 400 scans and a duty

cycle of 0.8 s. The total measurement time was ~12 h. Some, well-resolved residues

from the extramembrane regions (green) could be assigned based on the assignments

from 13C/15N detected experiments conducted at a MAS rate of 15.2 kHz (Table S5).

The spectrum was recorded by Dr. Venita Decker at Bruker BioSpin GmbH in

Rheinstetten. ............................................................................................................ 101

Figure 33. DGK in the AMP-PCP bound state. (a) Competitive inhibition assay verifies

the binding of Mg*AMP-PCP to the active sites of DGK. DGK proteoliposomes were

incubated with 4 to 16 mM of Mg*AMP-PCP, equating 4 to 16-fold molar excess

compared to DGK. Mg*ATP (3 mM) was present in each sample. A concentration of at

least 10 mM of Mg*AMP-PCP (10-fold molar excess) is needed to decrease the activity

of DGK below 10%, resulting in a fully saturated system. 100% activity corresponds to

the rate recorded with wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9) µmol

min-1 mg-1. Experiments were repeated three times. The activity was calculated as the

mean value. Error bars correspond to standard deviations. (b) The 31P-CP MAS

spectrum confirms the binding of AMP-PCP. For this purpose, proteoliposomes were

incubated with 14 mM Mg*AMP-PCP (pH 7.2). (c) The 2D NCA spectra of U-13C,15N-

DGK-I,L,V incubated with 14 mM Mg*AMP-PCP (pH 7.2), recorded immediately after

the incubation (black) and after 30 d (green), show that the fully saturated system is

stable over a long period of time without any significant evidence of degradation. .... 105

Appendix

168

Figure 34. (a) DGK in the DOG bound state. DGK was reconstituted into 80 mol%

DMPC/DMPA and 20 mol% DOG and incubated with 14 mM Mg*ATP (pH 7.2). DGK

phosphorylates DOG to DOG-PA, which can be observed by 31P-MAS NMR, both by

cross- and direct polarization. The spectra prove that DOG can reach the active site of

DGK under the here applied experimental conditions. (b) DGK in the DOG+AMP-PCP

bound state. 31P-CP spectrum of DGK reconstituted into 80 mol% DMPC/DMPA and

20 mol% DOG and incubated with 14 mM Mg*AMP-PCP (pH 7.2). It illustrates 31P

species of the bound AMP-PCP, which demonstrates a binding of the nucleotide to

DGK. ......................................................................................................................... 106

Figure 35. The DGK trimer adopts a symmetric conformation in its substrate bound

states. (a) Superposition of 2D NCA spectra of apo DGK (black) and AMP-PCP-bound

DGK (green). Regions for helical and random coil (rc) secondary structure are

highlighted for glycines. (b) Superposition of 2D NCA spectra of apo DGK (black) and

DOG-bound DGK (yellow). For the AMP-PCP and the DOG bound state no peak

splitting can be observed. ......................................................................................... 107

Figure 36. Effect of nucleotide binding on DGK. (a) Superposition of 2D NCA spectra

of DGK’s apo (black) and AMP-PCP-bound (green) state. Representative pronounced

shifts are illustrated in subsections of 2D NCACX (b) and 3D NCACX (c) spectra. (d)

The topology and ribbon model of the DGK monomer are shown with residues

highlighted that are affected by AMP-PCP. In the topology maps, alterations in peak

intensity and different levels of weighted CSPs are distinguished. In the ribbon model of

monomeric DGK, residues, which show a response on AMP-PCP binding, are

highlighted in green. The ribbon model is obtained from the OPM database [116], using

the PDB ID 4UXX from the X-ray structure [6]. .......................................................... 109

Figure 37. Effect of DOG binding on DGK. (a) Superposition of 2D NCA spectra of

DGK’s apo (black) and DOG-bound (yellow) state. (b) Representative extractions from

the 3D NCACX illustrate shifts for Glu69 and Gln33. (c) Superposition of 15N and 13C

INEPT-based experiments of the apo (black) and DOG-bound (yellow) state of DGK. In

the DOG bound state, the INEPT signals are decreased compared to the apo state,

indicating a reduction in mobility. Arg9 and Lys12, which could be assigned

unambiguously, are highlighted. (d) The topology and ribbon model of the DGK

monomer highlight residues that are affected by DOG. In the topology maps,

alterations in peak intensity and different levels of weighted CSPs are distinguished. In

the ribbon model of monomeric DGK, residues, which show a respond on DOG, are

Appendix

169

highlighted in green. The ribbon model is obtained from the OPM database [116], using

the PDB ID 4UXX from the X-ray structure [6]. .......................................................... 111

Figure 38. Effect of AMP-PCP and DOG binding on DGK. (a) Superposition of 2D NCA

spectra of apo (black) and AMP-PCP-bound (green) DGK. (b) Superposition of 2D

NCA spectra of AMP-PCP-bound (green) and AMP-PCP+DOG-bound (pink) DGK.

Both the AMP-PCP bound and AMP-PCP+DOG bound states feature a similar

fingerprint with significant alterations compared to the apo state. ............................. 112

Figure 39. Enlarged view from the membrane plane, illustrating the crystal structure of

Δ4 DGK (PDB 3ZE5) [5], accommodating the lipid substrate (orange) in the

hydrophobic pocket. Possible interactions between the side chain of Arg9, Lys12,

Gln33 and Glu69 with the proximal OH group of the lipid substrate are depicted. ..... 115

Figure 40. Intraprotomer interactions in the transmembrane region of DGK between

Trp112 and Ser61. (a) 2D 13C-13C DARR spectrum of U-13C,15N-DGK-ILV with 800 ms

mixing time. Crosspeaks appear between Ca of Ser61 and the side chain carbons of

Trp112 revealing an intraprotomer contact between helices 2 (Ser61) and 3 (Trp112).

(b) Visualization of the intraprotomer contact between Trp112 and Ser61 in the crystal

structure of Δ4 DGK (PDB 4UXX) [6]. Enlarged view from the membrane plane,

accommodating the lipid substrate (yellow) in the hydrophobic pocket (left). View from

the periplasm, depicting the three monomers in different shades of grey (right). Trp112

(H3) is secured by a hydrogen bond to Ser61 (H2) in the lower region of the

hydrophobic pocket. .................................................................................................. 118

Figure 41. Intraprotomer interactions in the cytoplasmic region of DGK between Arg32

and Trp25/Glu28/Ala29. (a) 2D NCOCX spectrum of U-13C,15N-DGK with a 400ms

DARR mixing step. Crosspeaks between 32ArgNh1/2 and 32ArgNe with 25TrpCO,

Glu28Cb and 29AlaCa/Cb/CO are determined, caused by an intraprotomer contact

between these residues in helix 1 and the surface helix. (b) Depiction of the

intraprotomer contact involving Arg32 in the crystal structure of Δ4 DGK (PDB 3ZE5)

[5]. Enlarged view from the membrane plane, illustrating the intraprotomer contacts for

32ArgNe/Nh1,2 with 25TrpCO, Glu28Cb and 29AlaCa/b/O. ..................................... 119

Figure 42. Representative possible interprotomer contacts in DGK suggested by the

crystal structure of Δ4 DGK (PDB 3ZE5) [5]. Enlarged view from the cytoplasm (a) and

membrane plane (b, c), illustrating a possible interprotomer contact for Arg81 and

Glu88 (a), Arg92 and Asn27 (b) as well as Lys19 and Asp95 (c). ............................. 120

Appendix

170

Figure 43. Creation of active mixed labelled trimers of DGK. (a) Differently labelled

trimers of DGK are separately expressed, solubilized, purified and eluted in DDM.

Subsequently, they are disassembled into monomers or dimers by SDS and mixed in a

1:1 ratio. Then, SDS is removed and replaced by DDM, resulting in mixed labelled

DGK trimers, which can then be reconstituted into liposomes. (b) BN-PAGE of DGK

(0.2 mg/ml) in different SDS concentrations: (1) 0.5%, (2) 1%, (3) 1.5%, (4) 2%, (5) 3%.

The higher the SDS concentration, the higher is the degree of disassembly into

monomers. (c) LILBID-MS confirms the predominantly monomeric state of DGK in SDS

micelles: The signals for the monomeric, dimeric and trimeric form of DGK are labelled

by “1”, “2” and “3”, respectively. They occur at a charged state of −1. The LILBID mass

spectrum was recorded by Oliver Peetz of the research group of Prof. Dr. Nina

Morgner (Institute of Physical and Theoretical Chemistry, Goethe University Frankfurt

am Main). (d) The BN-PAGE shows DGK as trimer in DDM micelles before (1) and

after (3) SDS treatment and mostly in its monomeric state in SDS micelles (2). (e)

Activity of DGK reconstituted from different detergent environments: DGK trimers from

DDM micelles before SDS treatment (-SDS, dark grey) and after SDS treatment

(±SDS, green) as well as DGK monomers from SDS micelles (+SDS, yellow) were

reconstituted. 100% activity corresponds to the rate recorded with wtDGK in 90mol%

DMPC/ 10mol% DMPA of 90 (± 9.9) µmol min-1 mg-1. Experiments were repeated three

times. The activity was calculated as the mean value. Error bars correspond to

standard deviations. .................................................................................................. 123

Figure 44. Validation of the application of AMUPol. (a) DNP enhancement shown for a

13C−CP spectrum of DGK incubated with 20 mM AMUPol. Upon microwave irradiation,

a 45-fold sensitivity enhancement is reached. (b) Activity of DGK with (+) and without (-

) AMUPol, indicating that the presence of the biradical has no influence on the activity.

100% activity corresponds to the rate recorded with wtDGK in 90mol% DMPC/ 10mol%

DMPA of 90 (± 9.9) µmol min-1 mg-1. Experiments were repeated three times. The

activity was calculated as the mean value. Error bars correspond to standard

deviations. ................................................................................................................ 124

Figure 45. DNP-enhanced 1D-TEDOR spectra of the control sample ([CC]-DGK) at

different mixing times. The spectra were recorded with 4096 scans at a 400 MHz

spectrometer, ~105 K, pH 7.2 and a spinning speed of 8 kHz. The 263 GHz gyrotron

was operated at a collector current of 70 mA. Six spectra were recorded with L0 (rotor

periods) = 4, 8, 16, 24, 32, and 40. ........................................................................... 125

Appendix

171

Figure 46. DNP-enhanced 1D-TEDOR spectra of [CN]-DGK) at different mixing times.

The spectra were recorded with 3520 scans at a 400 MHz spectrometer, ~105 K,

pH 7.2 and a spinning speed of 8 kHz. The 263 GHz gyrotron was operated at a

collector current of 70 mA. Six spectra were recorded with L0 (rotor periods) = 4, 8, 16,

24, 32, and 40. .......................................................................................................... 127

Figure 47. DNP-enhanced 15N−13C-TEDOR spectra (tmix = 6.25 ms) of [CN]-DGK,

[CN(Arg,Lys)]-DGK and the control sample, [CC]-DGK. All spectra reveal cross-peaks

originating from natural abundance intramolecular backbone 13C−15N-contacts

(highlighted in grey). Further cross-peaks (highlighted green) are detected in [CN]-DGK

and [CN(Arg,Lys)]-DGK. They can be assigned to cross-protomer contacts, reflecting a

through-space correlation between Arg and Asn/Asp/Glu. These cross-peaks

demonstrate that salt bridges or H-bonds between Asn/Asp/Glu and Arg must be

present at the protomer interfaces. ........................................................................... 129

Figure 48. Characterization of the oligomeric state of the single-site RxA mutants in

comparison to wtDGK by BN-PAGE. The BN-PAGES show DGK as trimer in DDM

micelles before (1) and after (3) SDS treatment and mainly as monomers in SDS

micelles (2). All RxA mutants show a similar oligomerization behavior as the wt,

indicating that the respective arginines located in extramembranous regions of DGK

are not necessary for the trimer formation. ................................................................ 131

Figure 49. Kinase activity of DGK affected by single site RxA mutations, expressed as

a percentage of wt activity. DGK trimers from DDM micelles (dark grey) and DGK

trimers from DDM micelles after SDS treatment (green) were reconstituted into lipid

bilayers and then measured. The activity is clearly reduced for all RxA-mutants

compared to the wild type. The SDS treatment of the RxA mutants led to a loss of

activity of only up to 20%, which is comparable to the wt. 100% activity corresponds to

the rate recorded with wtDGK in 90mol% DMPC/ 10mol% DMPA of 90 (± 9.9) µmol

min-1 mg-1. Experiments were repeated three times. The activity was calculated as the

mean value. Error bars correspond to standard deviations. ...................................... 132

Figure 50. DNP-enhanced 15N−13C-TEDOR spectra of mixed labelled trimers (U-

13C/U15N12C-DGK). The interprotomer crosspeak does not disappear for each single-

site RxA-mutant, demonstrating that more than one Arg contributes to these

interactions. Based on the 3D crystal structure [6], only Arg81 and Arg92 have an

appropriate location at the interface to be involved in these interactions. This would be

in-line with the observed reduction of the cross peak intensities for R81A- and R92A-

Appendix

172

DGK. For the AMP-PCP bound state of DGK, no significant changes of the cross-peak

are observable. ......................................................................................................... 133

Figure 51. DNP-enhanced 2D-TEDOR spectra (tmix = 6.25 ms) of the RxA mutants

[CN]-DGK-R81A and [CN]-DGK-R92A compared to wild-type [CN]-DGK and

[CN(Arg,Lys)]-DGK. [CN]-DGK-R81A and [CN]-DGK-R92A feature both a significant

cross peak in the 15N chemical shift range of lysine (*), indicating a participation of this

amino acid type in forming an interprotomer contact, which is not observable for the

wild type. ................................................................................................................... 138

List of tables

Table 1. Mapping of the active site through the identification of functionally relevant

residues by mutational studies, X-ray crystallization, MD simulations and solution NMR

................................................................................................................................... 23

Table 2. Primer sequences for single-site mutations .................................................. 46

Table 3. Components of PCR reaction mixture ........................................................... 47

Table 4. Standard PCR program used for mutagenesis of DGK ................................. 47

Table 5. Composition of M9 minimal medium for the expression of unlabelled DGK .. 49

Table 6. Composition of M9 minimal medium for the expression of U-13C,15N-DGK .... 49

Table 7. Composition of M9 minimal medium for the expression of U-13C,15N-DGK-I,L,V

................................................................................................................................... 50

Table 8. Composition of M9 minimal medium for the expression of U-13C-DGK .......... 52

Table 9. Composition of M9 minimal medium for the expression of U-12C,15N-DGK .... 52

Table 10. Composition of M9 minimal medium for the expression of U-12C,15NArg,Lys-

DGK ............................................................................................................................ 52

Table 11. Sample preparation for BN-PAGE .............................................................. 55

Table 12. Detergents used and compared in this dissertation. The classification, CMC

and concentration range are shown. ........................................................................... 70

Table 13. Amino acid composition of wild-type DGK. The numbers in brackets belong

to the His6-tag and linker. The most frequent hydrophobic amino acids: Ile, Leu and

Val, which were chosen for reverse labelling, are highlighted orange. ........................ 80

Table S1. List of chemicals. All were obtained with pro analysis (p.a.) quality. ......... 141

Table S2. List of consumable materials. ................................................................... 142

Appendix

173

Table S3. List of equipment. ..................................................................................... 142

Table S4. Experimental parameters for all multidimensional and dipolar coupling based

spectra of DGK in its apo state (white), saturated with AMP-PCP (light grey), DOG

(middle grey) and with AMP-PCP + DOG (dark grey). .............................................. 143

Table S5. Resonance assignments of wild-type DGK in DMPC/DMPA liposomes by

MAS NMR experiments. Chemical shifts are given in ppm........................................ 146

Table S6. Summary of all significant perturbations in peak position and intensity ..... 151

References

174

References

1. Kennedy E.P., Osmotic regulation and the biosynthesis of membrane-derived oligosaccharides in Escherichia coli. Proc Natl Acad Sci USA, 1982. 79(4): p. 1092-5.

2. Raetz C.R. and Newman K.F., Diglyceride kinase mutants of Escherichia coli: inner membrane association of 1,2-diglyceride and its relation to synthesis of membrane-derived oligosaccharides. J of Bacteriol, 1979. 137(2): p. 860-8.

3. Zhang Y.M. and Rock C.O., Membrane lipid homeostasis in bacteria. Nat Rev Microbiol, 2008. 6(3): p. 222-33.

4. Van Horn W.D. and Sanders C.R., Prokaryotic diacylglycerol kinase and undecaprenol kinase. Ann. Rev. Biophys., 2012. 41: p. 81-101.

5. Li D., Lyons J.A., Pye V.E., Vogeley L., Aragao D., Kenyon C.P., Shah S.T., Doherty C., Aherne M., and Caffrey M., Crystal structure of the integral membrane diacylglycerol kinase. Nature, 2013. 497(7450): p. 521-4.

6. Li D., Stansfeld P.J., Sansom M.S., Keogh A., Vogeley L., Howe N., Lyons J.A., Aragao D., Fromme P., Fromme R., Basu S., Grotjohann I., Kupitz C., Rendek K., Weierstall U., Zatsepin N.A., Cherezov V., Liu W., Bandaru S., English N.J., Gati C., Barty A., Yefanov O., Chapman H.N., Diederichs K., Messerschmidt M., Boutet S., Williams G.J., Marvin Seibert M., and Caffrey M., Ternary structure reveals mechanism of a membrane diacylglycerol kinase. Nat Commun, 2015. 6: p. 10140.

7. Van Horn W.D., Kim H.J., Ellis C.D., Hadziselimovic A., Sulistijo E.S., Karra M.D., Tian C., Sonnichsen F.D., and Sanders C.R., Solution nuclear magnetic resonance structure of membrane-integral diacylglycerol kinase. Science, 2009. 324(5935): p. 1726-9.

8. Koehler J., Sulistijo E.S., Sakakura M., Kim H.J., Ellis C.D., and Sanders C.R., Lysophospholipid micelles sustain the stability and catalytic activity of diacylglycerol kinase in the absence of lipids. Biochemistry, 2010. 49(33): p. 7089-99.

9. Lau F.W. and Bowie J.U., A method for assessing the stability of a membrane protein. Biochemistry, 1997. 36(19): p. 5884-92.

10. Lau F.W., Chen X., and Bowie J.U., Active sites of diacylglycerol kinase from Escherichia coli are shared between subunits. Biochemistry, 1999. 38(17): p. 5521-7.

11. Jiang Y., Tan H., Zheng J., Li X., Chen G., and Jia Z., Phosphoryl transfer reaction catalyzed by membrane diacylglycerol kinase: a theoretical mechanism study. Phys Chem Chem Phys, 2015. 17(38): p. 25228-34.

12. Ullrich S.J., Holper S., and Glaubitz C., Paramagnetic doping of a 7TM membrane protein in lipid bilayers by Gd(3)(+)-complexes for solid-state NMR spectroscopy. J Biomol NMR, 2014. 58(1): p. 27-35.

13. Schmidt E., Gath J., Habenstein B., Ravotti F., Szekely K., Huber M., Buchner L., Bockmann A., Meier B.H., and Güntert P., Automated solid-state NMR resonance assignment of protein microcrystals and amyloids. J Biomol NMR, 2013. 56(3): p. 243-54.

14. Chen Y., Zhang Z., Tang X., Li J., Glaubitz C., and Yang J., Conformation and topology of diacylglycerol kinase in E.coli membranes revealed by solid-state NMR spectroscopy. Angew Chem, Int Edit, 2014. 53(22): p. 5624-8.

15. Chen P., Novak J., Qi F., and Caufield P.W., Diacylglycerol kinase is involved in regulation of expression of the lantibiotic mutacin II of Streptococcus mutans. J Bacteriol, 1998. 180(1): p. 167-70.

16. Yamashita Y., Takehara T., and Kuramitsu H.K., Molecular characterization of a STreptococcus mutans mutant altered in environmental stress responses. J Bacteriol, 1993. 175(19): p. 6220-8.

References

175

17. Zhang Y.M. and Rock C.O., Membrane lipid homeostasis in bacteria. Nat Rev Microbiol, 2008. 6(3): p. 222-33.

18. Badola P. and Sanders C.R., 2nd, Escherichia coli diacylglycerol kinase is an evolutionarily optimized membrane enzyme and catalyzes direct phosphoryl transfer. J Biol Chem, 1997. 272(39): p. 24176-82.

19. Murray D.T., Li C., Gao F.P., Qin H., and Cross T.A., Membrane protein structural validation by oriented sample solid-state NMR: diacylglycerol kinase. Biophys J, 2014. 106(8): p. 1559-69.

20. Cady S., Wang T., and Hong M., Membrane-dependent effects of a cytoplasmic helix on the structure and drug binding of the influenza virus M2 protein. J Am Chem Soc, 2011. 133(30): p. 11572-9.

21. Cross T.A., Sharma M., Yi M., and Zhou H.X., Influence of solubilizing environments on membrane protein structures. Trends Biochem Sci, 2011. 36(2): p. 117-25.

22. Wallin E. and Heijne G.V., Genome‐wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Sci, 1998. 7(4): p. 1029-38.

23. Stevens T.J. and Arkin I.T., Do more complex organisms have a greater proportion of membrane proteins in their genomes? Proteins: Structure, Function, and Bioinformatics, 2000. 39(4): p. 417-20.

24. Overington J.P., Al-Lazikani B., and Hopkins A.L., How many drug targets are there? Nat Rev Drug Discov, 2006. 5(12): p. 993.

25. Moraes I., Evans G., Sanchez-Weatherby J., Newstead S., and Stewart P.D.S., Membrane protein structure determination—the next generation. Biochim Biophys Acta (BBA)-Biomembranes, 2014. 1838(1): p. 78-87.

26. De Zorzi R., Mi W., Liao M., and Walz T., Single-particle electron microscopy in the study of membrane protein structure. J Electron Microsc, 2015. 65(1): p. 81-96.

27. Maslennikov I. and Choe S., Advances in NMR structures of integral membrane proteins. Curr Opin Struct Biol, 2013. 23(4): p. 555-62.

28. Cross T.A., Ekanayake V., Paulino J., and Wright A., Solid state NMR: the essential technology for helical membrane protein structural characterization. J Magn Reson, 2014. 239: p. 100-9.

29. Rosenbaum D.M., Rasmussen S.G., and Kobilka B.K., The structure and function of G-protein-coupled receptors. Nature, 2009. 459(7245): p. 356.

30. Katritch V., Cherezov V., and Stevens R.C., Structure-function of the G protein–coupled receptor superfamily. Annu Rev Pharmacol Toxicol, 2013. 53: p. 531-56.

31. Lodowski D.T., Angel T.E., and Palczewski K., Comparative analysis of GPCR crystal structures. Photochemistry and photobiology, 2009. 85(2): p. 425-430.

32. Raunser S. and Walz T., Electron crystallography as a technique to study the structure on membrane proteins in a lipidic environment. Annu Rev Biophys, 2009. 38: p. 89-105.

33. Kühlbrandt W., Microscopy: cryo-EM enters a new era. Elife, 2014. 3: p. e03678.

34. Grigorieff N., J Struct Biol: direct detection pays off for electron cryo-microscopy. Elife, 2013. 2: p. e00573.

35. Cheng Y., Single-particle cryo-EM at crystallographic resolution. Cell, 2015. 161(3): p. 450-7.

36. Liao M., Cao E., Julius D., and Cheng Y., Structure of the TRPV1 ion channel determined by electron cryo-microscopy. Nature, 2013. 504(7478): p. 107.

37. McLaughlin A., Cullis P., Hemminga M., Hoult D., Radda G., Ritchie G., Seeley P., and Richards R., Application of 31P NMR to model and biological membrane systems. FEBS letters, 1975. 57(2): p. 213-8.

References

176

38. Griffin R., Observation of the effect of water on the phosphorus-31 nuclear magnetic resonance spectra of dipalmitoyllecithin. J Am Chem Soc, 1976. 98(3): p. 851-3.

39. Seelig J., Deuterium magnetic resonance: theory and application to lipid membranes. Q Rev Biophys, 1977. 10(3): p. 353-418.

40. Macdonald P.M., Deuterium NMR and the topography of surface electrostatic charge. Acc Chem Res, 1997. 30(5): p. 196-203.

41. Nietlispach D. and Gautier A., Solution NMR studies of polytopic α-helical membrane proteins. Curr Opin Struct Biol, 2011. 21(4): p. 497-508.

42. Hiller S. and Wagner G., The role of solution NMR in the structure determinations of VDAC-1 and other membrane proteins. Curr Opin Struct Biol, 2009. 19(4): p. 396-401.

43. Wang S. and Ladizhansky V., Recent advances in magic angle spinning solid state NMR of membrane proteins. Prog Nucl Magn Reson Spectrosc, 2014. 82: p. 1-26.

44. Kaplan M., Pinto C., Houben K., and Baldus M., Nuclear magnetic resonance (NMR) applied to membrane–protein complexes. Q Rev Biophys, 2016. 49.

45. Ladizhansky V., Applications of solid-state NMR to membrane proteins. Biochim Biophys Acta (BBA)-Proteins and Proteomics, 2017.

46. Andronesi O.C., Becker S., Seidel K., Heise H., Young H.S., and Baldus M., Determination of membrane protein structure and dynamics by magic-angle-spinning solid-state NMR spectroscopy. J Am Chem Soc, 2005. 127(37): p. 12965-74.

47. Yao Y., Fujimoto L.M., Hirshman N., Bobkov A.A., Antignani A., Youle R.J., and Marassi F.M., Conformation of BCL-XL upon membrane integration. J Mol Biol, 2015. 427(13): p. 2262-70.

48. Landau E.M. and Rosenbusch J.P., Lipidic cubic phases: a novel concept for the crystallization of membrane proteins. Proc Natl Acad Sci, 1996. 93(25): p. 14532-5.

49. Loll P.J., Membrane proteins, detergents and crystals: what is the state of the art? Acta Crystallogr F: Structural Biology Communications, 2014. 70(12): p. 1576-1583.

50. Caffrey M., A comprehensive review of the lipid cubic phase or in meso method for crystallizing membrane and soluble proteins and complexes. Acta Crystallogr F: Structural Biology Communications, 2015. 71(1): p. 3-18.

51. Tate C.G. and Schertler G.F., Engineering G protein-coupled receptors to facilitate their structure determination. Curr Opin Struct Biol, 2009. 19(4): p. 386-95.

52. Rasmussen S.G., Choi H.-J., Fung J.J., Pardon E., Casarosa P., Chae P.S., DeVree B.T., Rosenbaum D.M., Thian F.S., and Kobilka T.S., Structure of a nanobody-stabilized active state of the β 2 adrenoceptor. Nature, 2011. 469(7329): p. 175.

53. Mérida I., Ávila-Flores A., and Merino E., Diacylglycerol kinases: at the hub of cell signalling. Biochem J, 2008. 409(1): p. 1-18.

54. Raetz C.R. and Newman K.F., Neutral lipid accumulation in the membranes of Escherichia coli mutants lacking diglyceride kinase. J Biol Chem, 1978. 253(11): p. 3882-7.

55. Tanaka C. and Nishizuka Y., The protein kinase C family for neuronal signaling. Annu Rev Neurosci, 1994. 17(1): p. 551-67.

56. van Blitterswijk W.J. and Houssa B., Properties and functions of diacylglycerol kinases. Cell Signalling, 2000. 12(9-10): p. 595-605.

57. Ron D. and Kazanietz M.G., New insights into the regulation of protein kinase C and novel phorbol ester receptors. FASEB J, 1999. 13(13): p. 1658-76.

References

177

58. Ebinu J.O., Bottorff D.A., Chan E.Y., Stang S.L., Dunn R.J., and Stone J.C., RasGRP, a Ras guanyl nucleotide-releasing protein with calcium-and diacylglycerol-binding motifs. Science, 1998. 280(5366): p. 1082-6.

59. Tognon C.E., Kirk H.E., Passmore L.A., Whitehead I.P., Der C.J., and Kay R.J., Regulation of RasGRP via a phorbol ester-responsive C1 domain. Mol Cell Biol, 1998. 18(12): p. 6995-7008.

60. Kawasaki H., Springett G.M., Toki S., Canales J.J., Harlan P., Blumenstiel J.P., Chen E.J., Bany I.A., Mochizuki N., and Ashbacher A., A Rap guanine nucleotide exchange factor enriched highly in the basal ganglia. Proc Natl Acad Sci, 1998. 95(22): p. 13278-83.

61. Hofmann T., Obukhov A.G., Schaefer M., Harteneck C., Gudermann T., and Schultz G., Direct activation of human TRPC6 and TRPC3 channels by diacylglycerol. Nature, 1999. 397(6716): p. 259.

62. Pettitt T.R. and Wakelam M.J., Diacylglycerol kinase ε, but not ζ, selectively removes polyunsaturated diacylglycerol, inducing altered protein kinase C distribution in vivo. J Biol Chem, 1999. 274(51): p. 36181-6.

63. Ghosh S., Strum J.C., Sciorra V.A., Daniel L., and Bell R.M., Raf-1 kinase possesses distinct binding domains for phosphatidylserine and phosphatidic acid Phosphatidic acid regulates the translocation of Raf-1 in 12-O-tetradecanoylphorbol-13-acetate-stimulated Madin-Darby canine kidney cells. J Biol Chem, 1996. 271(14): p. 8472-80.

64. Limatola C., Schaap D., Moolenaar W., and Van Blitterswijk W., Phosphatidic acid activation of protein kinase C-ζ overexpressed in COS cells: comparison with other protein kinase C isotypes and other acidic lipids. Biochem J, 1994. 304(3): p. 1001-8.

65. Moritz A., De Graan P., Gispen W.H., and Wirtz K., Phosphatidic acid is a specific activator of phosphatidylinositol-4-phosphate kinase. J Biol Chem, 1992. 267(11): p. 7207-10.

66. Honda A., Nogami M., Yokozeki T., Yamazaki M., Nakamura H., Watanabe H., Kawamoto K., Nakayama K., Morris A.J., and Frohman M.A., Phosphatidylinositol 4-phosphate 5-kinase α is a downstream effector of the small G protein ARF6 in membrane ruffle formation. Cell, 1999. 99(5): p. 521-532.

67. Zhao Z., Shen S.-H., and Fischer E.H., Stimulation by phospholipids of a protein-tyrosine-phosphatase containing two src homology 2 domains. Proc Natl Acad Sci, 1993. 90(9): p. 4251-5.

68. Lu B. and Benning C., A 25-amino acid sequence of the Arabidopsis TGD2 protein is sufficient for specific binding of phosphatidic acid. J Biol Chem, 2009: p. jbc. M109. 016014.

69. Shulga Y.V., Topham M.K., and Epand R.M., Regulation and functions of diacylglycerol kinases. Chem Rev, 2011. 111(10): p. 6186-208.

70. Gupta R.S. and Epand R.M., Phylogenetic analysis of the diacylglycerol kinase family of proteins and identification of multiple highly-specific conserved inserts and deletions within the catalytic domain that are distinctive characteristics of different classes of DGK homologs. PloS one, 2017. 12(8): p. e0182758.

71. Thirugnanam S., Topham M.K., and Epand R.M., Physiological implications of the contrasting modulation of the activities of the epsilon- and zeta-isoforms of diacylglycerol kinase. Biochemistry, 2001. 40(35): p. 10607-13.

72. Jennings W., Doshi S., D’Souza K., and Epand R.M., Molecular properties of diacylglycerol kinase-epsilon in relation to function. Chem Phys Lipids, 2015. 192: p. 100-8.

73. Sakane F., Imai S.-i., Kai M., Yasuda S., and Kanoh H., Diacylglycerol kinases as emerging potential drug targets for a variety of diseases. Curr Drug Targets, 2008. 9(8): p. 626-40.

References

178

74. Goto K., Hozumi Y., Nakano T., Saino-Saito S., and Martelli A.M., Lipid messenger, diacylglycerol, and its regulator, diacylglycerol kinase, in cells, organs, and animals: history and perspective. Tohoku J Exp Med, 2008. 214(3): p. 199-212.

75. Masai I., Hosoya T., Kojima S.-i., and Hotta Y., Molecular cloning of a Drosophila diacylglycerol kinase gene that is expressed in the nervous system and muscle. Proc Natl Acad Sci, 1992. 89(13): p. 6030-4.

76. Miller K.G., Emerson M.D., and Rand J.B., G o α and diacylglycerol kinase negatively regulate the G q α pathway in C. elegans. Neuron, 1999. 24(2): p. 323-33.

77. Katagiri T., Mizoguchi T., and Shinozaki K., Molecular cloning of a cDNA encoding diacylglycerol kinase (DGK) in Arabidopsis thaliana. Plant Mol Biol, 1996. 30(3): p. 647-53.

78. Han G.S., O'Hara L., Carman G.M., and Siniossoglou S., An unconventional diacylglycerol kinase that regulates phospholipid synthesis and nuclear membrane growth. J Biol Chem, 2008. 283(29): p. 20433-42.

79. Han G.S., O'Hara L., Siniossoglou S., and Carman G.M., Characterization of the yeast DGK1-encoded CTP-dependent diacylglycerol kinase. J Biol Chem, 2008. 283(29): p. 20443-53.

80. Miller D.J., Jerga A., Rock C.O., and White S.W., Analysis of the Staphylococcus aureus DgkB structure reveals a common catalytic mechanism for the soluble diacylglycerol kinases. Structure, 2008. 16(7): p. 1036-46.

81. Sanders C.R., 2nd, Czerski L., Vinogradova O., Badola P., Song D., and Smith S.O., Escherichia coli diacylglycerol kinase is an alpha-helical polytopic membrane protein and can spontaneously insert into preformed lipid vesicles. Biochemistry, 1996. 35(26): p. 8610-8.

82. Schaap D., van der Wal J., and van Blitterswijk W.J., Consensus sequences for ATP-binding sites in protein kinases do not apply to diacylglycerol kinases. Biochem J, 1994. 304 ( Pt 2): p. 661-2.

83. Smith C.A. and Rayment I., Active site comparisons highlight structural similarities between myosin and other P-loop proteins. Biophys J, 1996. 70(4): p. 1590-602.

84. Van Horn W.D. and Sanders C.R., Prokaryotic diacylglycerol kinase and undecaprenol kinase. Annu Rev Biophys, 2012. 41: p. 81-101.

85. Lightner V., Bell R., and Modrich P., The DNA sequences encoding plsB and dgk loci of Escherichia coli. J Biol Chem, 1983. 258(18): p. 10856-61.

86. Keeling P.J., Diversity and evolutionary history of plastids and their hosts. Am J Bot, 2004. 91(10): p. 1481-93.

87. Nowack E.C., Melkonian M., and Glöckner G., Chromatophore genome sequence of Paulinella sheds light on acquisition of photosynthesis by eukaryotes. Curr Biol, 2008. 18(6): p. 410-18.

88. Yoon H.S., Hackett J.D., Ciniglia C., Pinto G., and Bhattacharya D., A molecular timeline for the origin of photosynthetic eukaryotes. Mol Biol Evol, 2004. 21(5): p. 809-18.

89. Zhou Y. and Bowie J.U., Building a thermostable membrane protein. J Biol Chem, 2000. 275(10): p. 6975-9.

90. Mi D., Kim H.J., Hadziselimovic A., and Sanders C.R., Irreversible misfolding of diacylglycerol kinase is independent of aggregation and occurs prior to trimerization and membrane association. Biochemistry, 2006. 45(33): p. 10072-84.

91. Nagy J.K., Lau F.W., Bowie J.U., and Sanders C.R., Mapping the oligomeric interface of diacylglycerol kinase by engineered thiol cross-linking: homologous sites in the transmembrane domain. Biochemistry, 2000. 39(14): p. 4154-64.

References

179

92. Smith R.L., O'Toole J.F., Maguire M.E., and Sanders C.R., 2nd, Membrane topology of Escherichia coli diacylglycerol kinase. J Bacteriol, 1994. 176(17): p. 5459-65.

93. Pieringer R.A. and Kunnes R.S., The biosynthesis of phosphatic acid and lysophosphatic acid by glyceride phosphokinase pathways in Escherichia coli J Biol Chem, 1965. 240: p. 2833-8.

94. Bohnenberger E. and Sandermann H., Jr., Diglyceride kinase from Escherichia coli. Purification in organic solvent and some properties of the enzyme. Eur J Biochem, 1979. 94(2): p. 401-7.

95. Vinogradova O., Badola P., Czerski L., Sönnichsen F., and Sanders 2nd C., Escherichia coli diacylglycerol kinase: a case study in the application of solution NMR methods to an integral membrane protein. Biophys J, 1997. 72(6): p. 2688-701.

96. Loomis C.R., Walsh J., and Bell R., sn-1, 2-Diacylglycerol kinase of Escherichia coli. Purification, reconstitution, and partial amino-and carboxyl-terminal analysis. J Biol Chem, 1985. 260(7): p. 4091-7.

97. Walsh J.P. and Bell R.M., sn-1,2-Diacylglycerol kinase of Escherichia coli. Structural and kinetic analysis of the lipid cofactor dependence. J Biol Chem, 1986. 261(32): p. 15062-9.

98. Walsh J.P. and Bell R.M., sn-1,2-Diacylglycerol kinase of Escherichia coli. Mixed micellar analysis of the phospholipid cofactor requirement and divalent cation dependence. J Biol Chem, 1986. 261(14): p. 6239-47.

99. Schneider E.G. and Kennedy E.P., Partial purification and properties of diglyceride kinase from Escherichia coli. Biochim Biophys Acta, 1976. 441(2): p. 201-12.

100. Allan D., Thomas P., and Michell R.H., Rapid transbilayer diffusion of 1,2-diacylglycerol and its relevance to control of membrane curvature. Nature, 1978. 276(5685): p. 289-90.

101. Hamilton J.A., Bhamidipati S.P., Kodali D.R., and Small D.M., The interfacial conformation and transbilayer movement of diacylglycerols in phospholipid bilayers. J Biol Chem, 1991. 266(2): p. 1177-86.

102. Albery W.J. and Knowles J.R., Efficiency and evolution of enzyme catalysis. Angew Chem Int Ed, 1977. 16(5): p. 285-93.

103. Czerski L. and Sanders C.R., Functionality of a membrane protein in bicelles. Anal Biochem, 2000. 284(2): p. 327-33.

104. Gorzelle B.M., Hoffman A.K., Keyes M.H., Gray D.N., Ray D.G., and Sanders C.R., Amphipols can support the activity of a membrane enzyme. J Am Chem Soc, 2002. 124(39): p. 11594-5.

105. Li D. and Caffrey M., Lipid cubic phase as a membrane mimetic for integral membrane protein enzymes. Proc Natl Acad Sci, 2011. 108(21): p. 8639-44.

106. Pilot J.D., East J.M., and Lee A.G., Effects of phospholipid headgroup and phase on the activity of diacylglycerol kinase of Escherichia coli. Biochemistry, 2001. 40(49): p. 14891-7.

107. Ullrich S.J., Hellmich U.A., Ullrich S., and Glaubitz C., Interfacial enzyme kinetics of a membrane bound kinase analyzed by real-time MAS-NMR. Nat Chem Biol, 2011. 7(5): p. 263-70.

108. Russ E., Kaiser U., and Sandermann H., Jr., Lipid-dependent membrane enzymes. Purification to homogeneity and further characterization of diacylglycerol kinase from Escherichia coli. Eur J Biochem, 1988. 171(1-2): p. 335-42.

109. Zhou Y., Wen J., and Bowie J.U., A passive transmembrane helix. Nat Struct Mol Biol, 1997. 4(12): p. 986.

110. Gorzelle B.M., Nagy J.K., Oxenoid K., Lonzer W.L., Cafiso D.S., and Sanders C.R., Reconstitutive refolding of diacylglycerol kinase, an integral membrane protein. Biochemistry, 1999. 38(49): p. 16373-82.

References

180

111. Nagy J.K., Lonzer W.L., and Sanders C.R., Kinetic study of folding and misfolding of diacylglycerol kinase in model membranes. Biochemistry, 2001. 40(30): p. 8971-80.

112. Nagy J.K. and Sanders C.R., A critical residue in the folding pathway of an integral membrane protein. Biochemistry, 2002. 41(29): p. 9021-5.

113. Nagy J.K. and Sanders C.R., Destabilizing mutations promote membrane protein misfolding. Biochemistry, 2004. 43(1): p. 19-25.

114. Lorch M. and Booth P.J., Insertion kinetics of a denatured alpha helical membrane protein into phospholipid bilayer vesicles. J Mol Biol, 2004. 344(4): p. 1109-21.

115. Seddon A.M., Lorch M., Ces O., Templer R.H., Macrae F., and Booth P.J., Phosphatidylglycerol lipids enhance folding of an alpha helical membrane protein. J Mol Biol, 2008. 380(3): p. 548-56.

116. Lomize M.A., Pogozheva I.D., Joo H., Mosberg H.I., and Lomize A.L., OPM database and PPM web server: resources for positioning of proteins in membranes. Nucleic Acids Res, 2012. 40(Database issue): p. 370-6.

117. Caffrey M., Li D., and Dukkipati A., Membrane protein structure determination using crystallography and lipidic mesophases: recent advances and successes. Biochemistry, 2012. 51(32): p. 6266-88.

118. Lau F.W., Nauli S., Zhou Y., and Bowie J.U., Changing single side-chains can greatly enhance the resistance of a membrane protein to irreversible inactivation1. J Mol Biol, 1999. 290(2): p. 559-64.

119. Lazaridis T., Mallik B., and Chen Y., Implicit solvent simulations of DPC micelle formation. J Phys Chem B, 2005. 109(31): p. 15098-106.

120. Murray D.T., Das N., and Cross T.A., Solid state NMR strategy for characterizing native membrane protein structures. Acc Chem Res, 2013. 46(9): p. 2172-81.

121. Wang T., Cady S.D., and Hong M., NMR determination of protein partitioning into membrane domains with different curvatures and application to the influenza M2 peptide. Biophys J, 2012. 102(4): p. 787-94.

122. Hu F., Luo W., Cady S.D., and Hong M., Conformational plasticity of the influenza A M2 transmembrane helix in lipid bilayers under varying pH, drug binding, and membrane thickness. Biochim Biophys Acta (BBA)-Biomembranes, 2011. 1808(1): p. 415-23.

123. Pauli J., Baldus M., van Rossum B., de Groot H., and Oschkinat H., Backbone and side-chain C-13 and N-15 signal assignments of the alpha-spectrin SH3 domain by magic angle spinning solid-state NMR at 17.6 tesla. ChemBioChem, 2001. 2(4): p. 272-81.

124. Shi L., Ahmed M.A., Zhang W., Whited G., Brown L.S., and Ladizhansky V., Three-dimensional solid-state NMR study of a seven-helical integral membrane proton pump--structural insights. J Mol Biol, 2009. 386(4): p. 1078-93.

125. Rienstra C.M., Hohwy M., Hong M., and Griffin R.G., 2D and 3D 15N− 13C− 13C NMR Chemical Shift Correlation Spectroscopy of Solids: Assignment of MAS Spectra of Peptides. J Am Chem Soc, 2000. 122(44): p. 10979-90.

126. Takegoshi K., Nakamura S., and Terao T., 13 C–1 H dipolar-driven 13 C–13 C recoupling without 13 C rf irradiation in nuclear magnetic resonance of rotating solids. J Chem Phys, 2003. 118(5): p. 2325-41.

127. Pauli J., Baldus M., van Rossum B., de Groot H., and Oschkinat H., Backbone and Side‐Chain 13C and 15N Signal Assignments of the α‐Spectrin SH3

Domain by Magic Angle Spinning Solid‐State NMR at 17.6 Tesla. ChemBioChem, 2001. 2(4): p. 272-81.

128. Morris G.A. and Freeman R., Enhancement of nuclear magnetic resonance signals by polarization transfer. J Am Chem Soc, 1979. 101(3): p. 760-2.

References

181

129. Maciejko J., Identification of protein-protein contacts in Green Proteorhodopsin oligomers by DNP enhanced solid state NMR. Diploma Thesis 2014, Goethe University Frankfurt am Main

130. Griffin R.G. and Prisner T.F., High field dynamic nuclear polarization--the renaissance. Phys Chem Chem Phys, 2010. 12(22): p. 5737-40.

131. Hu K.N., Polarizing agents and mechanisms for high-field dynamic nuclear polarization of frozen dielectric solids. Solid State Nucl Magn Reson, 2011. 40(2): p. 31-41.

132. Hovav Y., Feintuch A., and Vega S., Theoretical aspects of dynamic nuclear polarization in the solid state–spin temperature and thermal mixing. Phys Chem Chem Phys, 2013. 15(1): p. 188-203.

133. Sauvee C., Rosay M., Casano G., Aussenac F., Weber R.T., Ouari O., and Tordo P., Highly efficient, water-soluble polarizing agents for dynamic nuclear polarization at high frequency. Angew Chem Int Ed Engl, 2013. 52(41): p. 10858-61.

134. Maciejko J., Mehler M., Kaur J., Lieblein T., Morgner N., Ouari O., Tordo P., Becker-Baldus J., and Glaubitz C., Visualizing Specific Cross-Protomer Interactions in the Homo-Oligomeric Membrane Protein Proteorhodopsin by Dynamic-Nuclear-Polarization-Enhanced Solid-State NMR. J Am Chem Soc, 2015. 137(28): p. 9032-43.

135. Hing A.W., Vega S., and Schaefer J., Transferred-echo double-resonance NMR. J Magn Reson, 1992. 96: p. 205-9.

136. Jaroniec C.P., Filip C., and Griffin R.G., 3D TEDOR NMR experiments for the simultaneous measurement of multiple carbon-nitrogen distances in uniformly (13)C,(15)N-labeled solids. J Am Chem Soc, 2002. 124(36): p. 10728-42.

137. Gullion T. and Schaefer. Detection of Weak Heteronuclear Dipolar Coupling by Rotational-Echo Double-Resonance Nuclear Magnetic Resonance. J Adv Magn Reson, 1989. 13: p. 57-83.

138. Newby Z.E., D O'Connell III J., Gruswitz F., Hays F.A., Harries W.E., Harwood I.M., Ho J.D., Lee J.K., Savage D.F., and Miercke L.J., A general protocol for the crystallization of membrane proteins for X-ray structural investigation. Nat Protoc, 2009. 4(5): p. 619.

139. Laws D.D., Bitter H.M.L., and Jerschow A., Solid‐state NMR spectroscopic methods in chemistry. Angew Chem Int Ed, 2002. 41(17): p. 3096-129.

140. Apperley D.C., Harris R.K., and Hodgkinson P., Solid-state NMR: Basic principles and practice. 2012: Momentum Press.

141. Zhou Y., Lau F.W., Nauli S., Yang D., and Bowie J.U., Inactivation mechanism of the membrane protein diacylglycerol kinase in detergent solution. Protein Sci, 2001. 10(2): p. 378-83.

142. Crichton P.G., Harding M., Ruprecht J.J., Lee Y., and Kunji E.R., Lipid, detergent and Coomassie dye affect the migration of small membrane proteins in blue native gels: mitochondrial carriers migrate as monomers not dimers. J Biol Chem, 2013.

143. Morgner N., Barth H.-D., and Brutschy B., A new way to detect noncovalently bonded complexes of biomolecules from liquid micro-droplets by laser mass spectrometry. Austr J Chem, 2006. 59(2): p. 109-14.

144. Giordmaine J. and Miller R.C., Tunable coherent parametric oscillation in LiNb O 3 at optical frequencies. Phys Rev Lett, 1965. 14(24): p. 973.

145. Huisken F., Kulcke A., Voelkel D., Laush C., and Lisy J., New infrared injection‐seeded optical parametric oscillator with high energy and narrow bandwidth output. Appl Phys Lett, 1993. 62(8): p. 805-7.

146. Kropman M.F., Nienhuys H.-K., Woutersen S., and Bakker H.J., Vibrational relaxation and hydrogen-bond dynamics of HDO: H2O. J Phys Chem A, 2001. 105(19): p. 4622-6.

References

182

147. Morgner N., Hoffmann J., Barth H.-D., Meier T., and Brutschy B., LILBID-mass spectrometry applied to the mass analysis of RNA polymerase II and an F1Fo-ATP synthase. Int J Mass Spectrom, 2008. 277(1-3): p. 309-13.

148. Baldus M. and Meier B., Total correlation spectroscopy in the solid state. The use of scalar couplings to determine the through-bond connectivity. 1996, Academic Press.

149. Zhang Z., Chen Y., Tang X., Li J., Wang L., and Yang J., Solid-state NMR shows that dynamically different domains of membrane proteins have different hydration dependence. J Phys Chem B, 2014. 118(32): p. 9553-64.

150. Fung B., Khitrin A., and Ermolaev K., An improved broadband decoupling sequence for liquid crystals and solids. J Magn Reson, 2000. 142(1): p. 97-101.

151. Vranken W.F., Boucher W., Stevens T.J., Fogh R.H., Pajon A., Llinas M., Ulrich E.L., Markley J.L., Ionides J., and Laue E.D., The CCPN data model for NMR spectroscopy: development of a software pipeline. Proteins, 2005. 59(4): p. 687-96.

152. Cavanagh J., Fairbrother W.J., Palmer III A.G., and Skelton N.J., Protein NMR spectroscopy: principles and practice. 1995: Academic Press.

153. Schmidt E. and Güntert P., A new algorithm for reliable and general NMR resonance assignment. J Am Chem Soc, 2012. 134(30): p. 12817-29.

154. Güntert P., Automated structure determination from NMR spectra. Eur Biophys J, 2009. 38(2): p. 129.

155. Güntert P., Mumenthaler C., and Wüthrich K., Torsion angle dynamics for NMR structure calculation with the new program Dyana1. J Mol Biol, 1997. 273(1): p. 283-98.

156. Bartels C., Billeter M., Güntert P., and Wüthrich K., Automated sequence-specific NMR assignment of homologous proteins using the program GARANT. J Biomol NMR, 1996. 7(3): p. 207-13.

157. Bartels C., Güntert P., Billeter M., and Wüthrich K., GARANT‐a general algorithm for resonance assignment of multidimensional nuclear magnetic resonance spectra. J Comput Chem, 1997. 18(1): p. 139-49.

158. Schmucki R., Yokoyama S., and Güntert P., Automated assignment of NMR chemical shifts using peak-particle dynamics simulation with the DYNASSIGN algorithm. J Biomol NMR, 2009. 43(2): p. 97-109.

159. Ulrich E.L., Akutsu H., Doreleijers J.F., Harano Y., Ioannidis Y.E., Lin J., Livny M., Mading S., Maziuk D., and Miller Z., BioMagResBank. Nucleic Acids Res, 2007. 36: p. 402-8.

160. López-Méndez B. and Güntert P., Automated protein structure determination from NMR spectra. J Am Chem Soc, 2006. 128(40): p. 13112-22.

161. Malmodin D., Papavoine C.H., and Billeter M., Fully automated sequence-specific resonance assignments of hetero-nuclear protein spectra. J Biomol NMR, 2003. 27(1): p. 69-79.

162. Oxenoid K. and Chou J.J., The structure of phospholamban pentamer reveals a channel-like architecture in membranes. Proc Natl Acad Sci USA, 2005. 102(31): p. 10870-5.

163. Ullrich S.J., Insights into a lipid regulator by solid-state MAS NMR: Kinetic and structure-functional studies on diacylglycerol kinase. PhD Thesis 2013, Goethe University: Frankfurt am Main.

164. Li Q., Mittal R., Huang L., Travis B., and Sanders C.R., Bolaamphiphile-class surfactants can stabilize and support the function of solubilized integral membrane proteins. Biochemistry, 2009. 48(49): p. 11606-8.

165. Seddon A.M., Curnow P., and Booth P.J., Membrane proteins, lipids and detergents: not just a soap opera. Biochim Biophys Acta (BBA)-Biomembranes, 2004. 1666(1): p. 105-17.

166. Warschawski D.E., Arnold A.A., Beaugrand M., Gravel A., Chartrand É., and Marcotte I., Choosing membrane mimetics for NMR structural studies of

References

183

transmembrane proteins. Biochim Biophys Acta (BBA)-Biomembranes, 2011. 1808(8): p. 1957-74.

167. Ollivon M., Lesieur S., Grabielle-Madelmont C., and Paternostre M.t., Vesicle reconstitution from lipid–detergent mixed micelles. Biochim Biophys Acta (BBA)-Biomembranes, 2000. 1508(1): p. 34-50.

168. Rigaud J.-L., Pitard B., and Levy D., Reconstitution of membrane proteins into liposomes: application to energy-transducing membrane proteins. Biochim Biophys Acta (BBA)-Bioenergetics, 1995. 1231(3): p. 223-46.

169. Rigaud J.-L., Mosser G., Lacapere J.-J., Olofsson A., Levy D., and Ranck J.-L., Bio-Beads: an efficient strategy for two-dimensional crystallization of membrane proteins. J Struct Biol, 1997. 118(3): p. 226-35.

170. Gurtovenko A.A. and Vattulainen I., Membrane potential and electrostatics of phospholipid bilayers with asymmetric transmembrane distribution of anionic lipids. J Phys Chem B, 2008. 112(15): p. 4629-34.

171. Mehler M., Scholz F., Ullrich S.J., Mao J., Braun M., Brown L.J., Brown R.C., Fiedler S.A., Becker-Baldus J., and Wachtveitl J., The EF loop in green proteorhodopsin affects conformation and photocycle dynamics. Biophys J, 2013. 105(2): p. 385-97.

172. Kaur J., Kriebel C.N., Eberhardt P., Jakdetchai O., Leeder A.J., Weber I., Brown L.J., Brown R.C., Becker-Baldus J., and Bamann C., Solid-state NMR analysis of the sodium pump Krokinobacter rhodopsin 2 and its H30A mutant. J Struct Biol, 2018.

173. Kaur H., Lakatos-Karoly A., Vogel R., Nöll A., Tampé R., and Glaubitz C., Coupled ATPase-adenylate kinase activity in ABC transporters. Nat Commun, 2016. 7: p. 13864.

174. Wang S., Munro R.A., Shi L., Kawamura I., Okitsu T., Wada A., Kim S.Y., Jung K.H., Brown L.S., and Ladizhansky V., Solid-state NMR spectroscopy structure determination of a lipid-embedded heptahelical membrane protein. Nat Methods, 2013. 10(10): p. 1007-12.

175. Vranken W.F., Boucher W., Stevens T.J., Fogh R.H., Pajon A., Llinas M., Ulrich E.L., Markley J.L., Ionides J., and Laue E.D., The CCPN data model for NMR spectroscopy: development of a software pipeline. Proteins: Structure, Function, and Bioinformatics, 2005. 59(4): p. 687-96.

176. Gammeren A.J.v., Hulsbergen F.B., Hollander J.G., and Groot H.J.M.d., Residual backbone and side-chain 13C and 15N resonance assignments of the intrinsic transmembrane light-harvesting 2 protein complex by solid-state Magic Angle Spinning NMR spectroscopy. J Biomol NMR, 2005. 31(4): p. 279-93.

177. Egorova-Zachernyuk T., Hollander J., Fraser N., Gast P., Hoff A., Cogdell R., De Groot H., and Baldus M., Heteronuclear 2D-correlations in a uniformly [13C, 15N] labeled membrane-protein complex at ultra-high magnetic fields. J Biomol NMR, 2001. 19(3): p. 243-53.

178. Etzkorn M., Martell S., Andronesi O.C., Seidel K., Engelhard M., and Baldus M., Secondary structure, dynamics, and topology of a seven-helix receptor in native membranes, studied by solid-state NMR spectroscopy. Angew Chem Int Ed Engl, 2007. 46(3): p. 459-62.

179. Park S.H., Das B.B., Casagrande F., Tian Y., Nothnagel H.J., Chu M., Kiefer H., Maier K., De Angelis A.A., Marassi F.M., and Opella S.J., Structure of the chemokine receptor CXCR1 in phospholipid bilayers. Nature, 2012. 491(7426): p. 779-83.

180. Sharma M., Yi M., Dong H., Qin H., Peterson E., Busath D.D., Zhou H.-X., and Cross T.A., Insight into the mechanism of the influenza A proton channel from a structure in a lipid bilayer. Science, 2010. 330(6003): p. 509-12.

181. Andreas L.B., Reese M., Eddy M.T., Gelev V., Ni Q.Z., Miller E.A., Emsley L., Pintacuda G., Chou J.J., and Griffin R.G., Structure and mechanism of the

References

184

influenza A M218–60 dimer of dimers. J Am Chem Soc, 2015. 137(47): p. 14877-86.

182. Tang M., Sperling L.J., Berthold D.A., Schwieters C.D., Nesbitt A.E., Nieuwkoop A.J., Gennis R.B., and Rienstra C.M., High-resolution membrane protein structure by joint calculations with solid-state NMR and X-ray experimental data. J Biomol NMR, 2011. 51(3): p. 227.

183. Tang M., Nesbitt A.E., Sperling L.J., Berthold D.A., Schwieters C.D., Gennis R.B., and Rienstra C.M., Structure of the disulfide bond generating membrane protein DsbB in the lipid bilayer. J Mol Biol, 2013. 425(10): p. 1670-82.

184. Das N., Dai J., Hung I., Rajagopalan M., Zhou H.-X., and Cross T.A., Structure of CrgA, a cell division structural and regulatory protein from Mycobacterium tuberculosis, in lipid bilayers. Proc Natl Acad Sci, 2015. 112(2): p. 119-26.

185. Marassi F.M. and Opella S.J., Simultaneous assignment and structure determination of a membrane protein from NMR orientational restraints. Protein Sci, 2003. 12(3): p. 403-11.

186. Opella S.J. and Marassi F.M., Applications of NMR to membrane proteins. Arch Biochem Biophys, 2017. 628: p. 92-101.

187. Schneider R., Ader C., Lange A., Giller K., Hornig S., Pongs O., Becker S., and Baldus M., Solid-state NMR spectroscopy applied to a chimeric potassium channel in lipid bilayers. J Am Chem Soc, 2008. 130(23): p. 7427-35.

188. Waugh D.S., Genetic tools for selective labeling of proteins with α-15 N-amino acids. J Biomol NMR, 1996. 8(2): p. 184-92.

189. Muchmore D.C., McIntosh L.P., Russell C.B., Anderson D.E., and Dahlquist F.W., [3] Expression and nitrogen-15 labeling of proteins for proton and nitrogen-15 nuclear magnetic resonance, in Methods Enzymol, 1989. 177: p. 44-73.

190. Ganapathy S., Naito A., and McDowell C., Paramagnetic doping as an aid in obtaining high-resolution carbon-13 NMR spectra of biomolecules in the solid state. J Am Chem Soc, 1981. 103(20): p. 6011-5.

191. Otting G., Protein NMR using paramagnetic ions. Annu Rev Biophys, 2010. 39: p. 387-405.

192. Wickramasinghe N.P., Kotecha M., Samoson A., Past J., and Ishii Y., Sensitivity enhancement in 13C solid-state NMR of protein microcrystals by use of paramagnetic metal ions for optimizing 1H T1 relaxation. J Magn Reson, 2007. 184(2): p. 350-6.

193. Wickramasinghe N.P., Shaibat M.A., Jones C.R., Casabianca L.B., De Dios A.C., Harwood J.S., and Ishii Y., Progress in C 13 and H 1 solid-state nuclear magnetic resonance for paramagnetic systems under very fast magic angle spinning. J Chem Phys, 2008. 128(5).

194. Wickramasinghe N.P., Parthasarathy S., Jones C.R., Bhardwaj C., Long F., Kotecha M., Mehboob S., Fung L.W., Past J., and Samoson A., Nanomole-scale protein solid-state NMR by breaking intrinsic 1 H T 1 boundaries. Nat Methods, 2009. 6(3): p. 215.

195. Linser R., Chevelkov V., Diehl A., and Reif B., Sensitivity enhancement using paramagnetic relaxation in MAS solid-state NMR of perdeuterated proteins. J Magn Reson, 2007. 189(2): p. 209-16.

196. Stringer J.A., Bronnimann C.E., Mullen C.G., Zhou D.H., Stellfox S.A., Li Y., Williams E.H., and Rienstra C.M., Reduction of RF-induced sample heating with a scroll coil resonator structure for solid-state NMR probes. J Magn Reson, 2005. 173(1): p. 40-8.

197. Moseley H.N., Sperling L.J., and Rienstra C.M., Automated protein resonance assignments of magic angle spinning solid-state NMR spectra of β1 immunoglobulin binding domain of protein G (GB1). J Biomol NMR, 2010. 48(3): p. 123-8.

References

185

198. Zimmerman D.E., Kulikowski C.A., Huang Y., Feng W., Tashiro M., Shimotakahara S., Chien C.-y., Powers R., and Montelione G.T., Automated analysis of protein NMR assignments using methods from artificial intelligence1. J Mol Biol, 1997. 269(4): p. 592-610.

199. Hu K.-N., Qiang W., and Tycko R., A general Monte Carlo/simulated annealing algorithm for resonance assignment in NMR of uniformly labeled biopolymers. J Biomol NMR, 2011. 50(3): p. 267-76.

200. Tycko R. and Hu K.-N., A Monte Carlo/simulated annealing algorithm for sequential resonance assignment in solid state NMR of uniformly labeled proteins with magic-angle spinning. J Magn Reson, 2010. 205(2): p. 304-14.

201. Tosner Z., Sarkar R., Becker-Baldus J., Glaubitz C., Wegner S., Engelke F., Glaser S.J., and Reif B., Overcoming volume selectivity of dipolar recoupling in biological solid-state NMR. Angew Chem Int Ed in English, 2018.

202. Barna J., Laue E., Mayger M., Skilling J., and Worrall S., Exponential sampling, an alternative method for sampling in two-dimensional NMR experiments. J Magn Reson (1969), 1987. 73(1): p. 69-77.

203. Hyberts S.G., Arthanari H., and Wagner G., Applications of non-uniform sampling and processing. Top Curr Chem, 2011. 316: p. 125-48.

204. Paramasivam S., Suiter C.L., Hou G., Sun S., Palmer M., Hoch J.C., Rovnyak D., and Polenova T., Enhanced sensitivity by nonuniform sampling enables multidimensional MAS NMR spectroscopy of protein assemblies. J Phys Chem B, 2012. 116(25): p. 7416-27.

205. Ulrich E.L., Akutsu H., Doreleijers J.F., Harano Y., Ioannidis Y.E., Lin J., Livny M., Mading S., Maziuk D., Miller Z., Nakatani E., Schulte C.F., Tolmie D.E., Kent Wenger R., Yao H., and Markley J.L., BioMagResBank. Nucleic Acids Res, 2008. 36(Database issue): p. D402-8.

206. Luca S., Filippov D.V., van Boom J.H., Oschkinat H., de Groot H.J., and Baldus M., Secondary chemical shifts in immobilized peptides and proteins: a qualitative basis for structure refinement under magic angle spinning. J Biomol NMR, 2001. 20(4): p. 325-31.

207. Tsuchiya Y., Nakamura H., and Kinoshita K., Discrimination between biological interfaces and crystal-packing contacts. Adv. Appl. Bioinform. Chem., 2008. 1: p. 99-113.

208. Burmeister W.P., Structural changes in a cryo-cooled protein crystal owing to radiation damage. Acta Crystallogr D: Biological Crystallography, 2000. 56(3): p. 328-41.

209. Higman V.A., Flinders J., Hiller M., Jehle S., Markovic S., Fiedler S., van Rossum B.-J., and Oschkinat H., Assigning large proteins in the solid state: a MAS NMR resonance assignment strategy using selectively and extensively 13 C-labelled proteins. J Biomol NMR, 2009. 44(4): p. 245-60.

210. Jehle S., van Rossum B., Stout J.R., Noguchi S.M., Falber K., Rehbein K., Oschkinat H., Klevit R.E., and Rajagopal P., αB-crystallin: a hybrid solid-state/solution-state NMR investigation reveals structural aspects of the heterogeneous oligomer. J Mol Biol, 2009. 385(5): p. 1481-97.

211. Hiller M., Higman V.A., Jehle S., van Rossum B.-J., Kühlbrandt W., and Oschkinat H., [2, 3-13C]-labeling of Aromatic Residues Getting a Head Start in the Magic-Angle-Spinning NMR Assignment of Membrane Proteins. J Am Chem Soc, 2008. 130(2): p. 408-9.

212. van Gammeren A., Hulsbergen F., Hollander J., and de Groot H., Biosynthetic site-specific 13 C labeling of the light-harvesting 2 protein complex: A model for solid state NMR structure determination of transmembrane proteins. J Biomol NMR, 2004. 30(3): p. 267-74.

213. Van Gammeren A., Hulsbergen F., Hollander J., and De Groot H., Residual backbone and side-chain 13 C and 15 N resonance assignments of the intrinsic

References

186

transmembrane light-harvesting 2 protein complex by solid-state Magic Angle Spinning NMR spectroscopy. J Biomol NMR, 2005. 31(4): p. 279-93.

214. Neri D., Szyperski T., Otting G., Senn H., and Wuethrich K., Stereospecific nuclear magnetic resonance assignments of the methyl groups of valine and leucine in the DNA-binding domain of the 434 repressor by biosynthetically directed fractional 13 C labeling, in Nmr In Structural Biology: A Collection of Papers by Kurt Wüthrich. 1995, World Scientific. p. 427-33.

215. Schubert M., Manolikas T., Rogowski M., and Meier B.H., Solid-state NMR spectroscopy of 10% 13 C labeled ubiquitin: spectral simplification and stereospecific assignment of isopropyl groups. J Biomol NMR, 2006. 35(3): p. 167-73.

216. Hong M., Determination of multiple φ-torsion angles in proteins by selective and extensive 13C labeling and two-dimensional solid-state NMR. J Magn Reson, 1999. 139(2): p. 389-401.

217. Lundström P., Teilum K., Carstensen T., Bezsonova I., Wiesner S., Hansen D.F., Religa T.L., Akke M., and Kay L.E., Fractional 13 C enrichment of isolated carbons using [1-13 C]-or [2-13 C]-glucose facilitates the accurate measurement of dynamics at backbone C α and side-chain methyl positions in proteins. J Biomol NMR, 2007. 38(3): p. 199-212.

218. Schwarz D., Dötsch V., and Bernhard F., Production of membrane proteins using cell‐free expression systems. Proteomics, 2008. 8(19): p. 3933-46.

219. Junge F., Schneider B., Reckel S., Schwarz D., Dötsch V., and Bernhard F., Large-scale production of functional membrane proteins. Cell Mol Life Sci, 2008. 65(11): p. 1729-55.

220. Maslennikov I., Klammt C., Hwang E., Kefala G., Okamura M., Esquivies L., Mörs K., Glaubitz C., Kwiatkowski W., and Jeon Y.H., Membrane domain structures of three classes of histidine kinase receptors by cell-free expression and rapid NMR analysis. Proc Natl Acad Sci, 2010. 107(24): p. 10902-7.

221. Andrew E.R., Bradbury A., and Eades R., Nuclear magnetic resonance spectra from a crystal rotated at high speed. Nature, 1958. 182(4650): p. 1659.

222. Lowe I., Free induction decays of rotating solids. Phys Rev Lett, 1959. 2(7): p. 285.

223. Mehring M., Pines A., Rhim W.K., and Waugh J., Spin‐Decoupling in the Resolution of Chemical Shifts in Solids by Pulsed NMR. J Chem Phys, 1971. 54(7): p. 3239-40.

224. Parthasarathy S., Nishiyama Y., and Ishii Y., Sensitivity and resolution enhanced solid-state NMR for paramagnetic systems and biomolecules under very fast magic angle spinning. Acc Chem Res, 2013. 46(9): p. 2127-35.

225. Kobayashi T., Mao K., Paluch P., Nowak‐Król A., Sniechowska J., Nishiyama Y., Gryko D.T., Potrzebowski M.J., and Pruski M., Study of Intermolecular Interactions in the Corrole Matrix by Solid‐State NMR under 100 kHz MAS and Theoretical Calculations. Angew Chem Int Ed, 2013. 52(52): p. 14108-11.

226. Ye Y.Q., Malon M., Martineau C., Taulelle F., and Nishiyama Y., Rapid measurement of multidimensional 1H solid-state NMR spectra at ultra-fast MAS frequencies. J Magn Reson, 2014. 239: p. 75-80.

227. Nishiyama Y., Zhang R., and Ramamoorthy A., Finite-pulse radio frequency driven recoupling with phase cycling for 2D 1H/1H correlation at ultrafast MAS frequencies. J Magn Reson, 2014. 243: p. 25-32.

228. Agarwal V., Penzel S., Szekely K., Cadalbert R., Testori E., Oss A., Past J., Samoson A., Ernst M., and Böckmann A., De Novo 3D Structure Determination from Sub‐milligram Protein Samples by Solid‐State 100 kHz MAS NMR Spectroscopy. Angew Chem Int Ed, 2014. 53(45): p. 12253-6.

229. McDermott A., Creuzet F., Kolbert A., and Griffin R., High-resolution magic-angle-spinning NMR spectra of protons in deuterated solids. J Magn Reson, 1992. 98: p. 408-13.

References

187

230. Reif B., Jaroniec C., Rienstra C., Hohwy M., and Griffin R., 1H–1H MAS correlation spectroscopy and distance measurements in a deuterated peptide. vier, 2001. 151(2): p. 320-7.

231. Paulson E.K., Morcombe C.R., Gaponenko V., Dancheck B., Byrd R.A., and Zilm K.W., Sensitive high resolution inverse detection NMR spectroscopy of proteins in the solid state. J Am Chem Soc, 2003. 125(51): p. 15831-6.

232. Chevelkov V., van Rossum B.J., Castellani F., Rehbein K., Diehl A., Hohwy M., Steuernagel S., Engelke F., Oschkinat H., and Reif B., H detection in MAS solid-state NMR spectroscopy of biomacromolecules employing pulsed field gradients for residual solvent suppression. J Am Chem Soc, 2003. 125(26): p. 7788-9.

233. Linser R., Dasari M., Hiller M., Higman V., Fink U., Lopez del Amo J.M.,

Markovic S., Handel L., Kessler B., and Schmieder P., Proton‐Detected Solid‐State NMR Spectroscopy of Fibrillar and Membrane Proteins. Angew Chem Int Ed, 2011. 50(19): p. 4508-12.

234. Barbet-Massin E., Pell A.J., Retel J.S., Andreas L.B., Jaudzems K., Franks W.T., Nieuwkoop A.J., Hiller M., Higman V., and Guerry P., Rapid proton-detected NMR assignment for proteins with fast magic angle spinning. J Am Chem Soc, 2014. 136(35): p. 12489-97.

235. Ishii Y. and Tycko R., Sensitivity enhancement in solid state 15N NMR by indirect detection with high-speed magic angle spinning. J Magn Res, 142: p. 199-204.

236. Ishii Y., Yesinowski J.P., and Tycko R., Sensitivity enhancement in solid-state 13C NMR of synthetic polymers and biopolymers by 1H NMR detection with high-speed magic angle spinning. J Am Chem Soc, 2001. 123(12): p. 2921-2.

237. Reif B. and Griffin R., 1H detected 1H, 15N correlation spectroscopy in rotating solids. J Magn Reson, 2003. 160(1): p. 78-83.

238. Chevelkov V., Rehbein K., Diehl A., and Reif B., Ultrahigh Resolution in Proton Solid‐State NMR Spectroscopy at High Levels of Deuteration. Angew Chem Int Ed, 2006. 45(23): p. 3878-81.

239. Park S.H., Yang C., Opella S.J., and Mueller L.J., Resolution and measurement of heteronuclear dipolar couplings of a noncrystalline protein immobilized in a biological supramolecular assembly by proton-detected MAS solid-state NMR spectroscopy. J Magn Reson, 2013. 237: p. 164-8.

240. Böckmann A., Ernst M., and Meier B.H., Spinning proteins, the faster, the better? J Magn Reson, 2015. 253: p. 71-9.

241. Eddy M.T., Su Y., Silvers R., Andreas L., Clark L., Wagner G., Pintacuda G., Emsley L., and Griffin R.G., Lipid bilayer-bound conformation of an integral membrane beta barrel protein by multidimensional MAS NMR. J Biomol NMR, 2015. 61(3-4): p. 299-310.

242. Zhang R., Mroue K.H., and Ramamoorthy A., Proton-based ultrafast magic angle spinning solid-state NMR spectroscopy. Acc Chem Res, 2017. 50(4): p. 1105-13.

243. Ward M.E., Shi L., Lake E., Krishnamurthy S., Hutchins H., Brown L.S., and Ladizhansky V., Proton-detected solid-state NMR reveals intramembrane polar networks in a seven-helical transmembrane protein proteorhodopsin. J Am Chem Soc, 2011. 133(43): p. 17434-43.

244. Zhou D.H., Nieuwkoop A.J., Berthold D.A., Comellas G., Sperling L.J., Tang M., Shah G.J., Brea E.J., Lemkau L.R., and Rienstra C.M., Solid-state NMR analysis of membrane proteins and protein aggregates by proton detected spectroscopy. J Biomol NMR, 2012. 54(3): p. 291-305.

245. Medeiros‐Silva J., Mance D., Daniëls M., Jekhmane S., Houben K., Baldus M., and Weingarth M., 1H‐Detected Solid‐State NMR Studies of Water‐Inaccessible Proteins In Vitro and In Situ. Angew Chem Int Ed, 2016. 55(43): p. 13606-10.

References

188

246. Saurel O., Iordanov I., Nars G., Demange P., Le Marchand T., Andreas L.B., Pintacuda G., and Milon A., Local and global dynamics in klebsiella pneumoniae outer membrane protein a in lipid bilayers probed at atomic resolution. J Am Chem Soc, 2017. 139(4): p. 1590-7.

247. Yao Y., Dutta S.K., Park S.H., Rai R., Fujimoto L.M., Bobkov A.A., Opella S.J., and Marassi F.M., High resolution solid-state NMR spectroscopy of the Yersinia pestis outer membrane protein Ail in lipid membranes. J Biomol NMR, 2017. 67(3): p. 179-90.

248. Ward M.E., Ritz E., Ahmed M.A., Bamm V.V., Harauz G., Brown L.S., and Ladizhansky V., Proton detection for signal enhancement in solid-state NMR experiments on mobile species in membrane proteins. J Biomol NMR, 2015. 63(4): p. 375-88.

249. Weingarth M., Van Der Cruijsen E.A., Ostmeyer J., Lievestro S., Roux B., and Baldus M., Quantitative analysis of the water occupancy around the selectivity filter of a K+ channel in different gating modes. J Am Chem Soc, 2014. 136(5): p. 2000-7.

250. Jain M.G., Lalli D., Stanek J., Gowda C., Prakash S., Schwarzer T.S., Schubeis T., Castiglione K., Andreas L.B., and Madhu P.K., Selective 1H–1H Distance Restraints in Fully Protonated Proteins by Very Fast Magic-Angle Spinning Solid-State NMR. J Phys Chem letters, 2017. 8(11): p. 2399-405.

251. Lalli D., Idso M.N., Andreas L.B., Hussain S., Baxter N., Han S., Chmelka B.F., and Pintacuda G., Proton-Based Structural Analysis of a Heptahelical Transmembrane Protein in Lipid Bilayers. J Am Chem Soc, 2017. 139(37): p. 13006-12.

252. Daebel V., Introduction to Proton Detection in Biological Samples under Ultra-Fast Magic Angle Spinning. Bruker App Note, 2015.

253. Zhou D.H. and Rienstra C.M., High-performance solvent suppression for proton detected solid-state NMR. J Magn Reson, 2008. 192(1): p. 167-72.

254. Ladizhansky V., Applications of solid-state NMR to membrane proteins. Biochim Biophys Acta-Proteins and Proteomics, 2017. 1865(11): p. 1577-86.

255. Walsh J.P., Fahrner L., and Bell R.M., sn-1,2-diacylglycerol kinase of Escherichia coli. Diacylglycerol analogues define specificity and mechanism. J Biol Chem, 1990. 265(8): p. 4374-81.

256. Lau F.W., Chen X., and Bowie J.U., Active sites of diacylglycerol kinase from Escherichia coli are shared between subunits. Biochemistry, 1999. 38(17): p. 5521-7.

257. Badola P. and Sanders C.R., Escherichia coli diacylglycerol kinase is an evolutionarily optimized membrane enzyme and catalyzes direct phosphoryl transfer. J Biol Chem, 1997. 272(39): p. 24176-82.

258. Schulz G.E., Muller C.W., and Diederichs K., Induced-fit movements in adenylate kinases. J Mol Biol, 1990. 213(4): p. 627-30.

259. Bennett W.S., Jr. and Steitz T.A., Glucose-induced conformational change in yeast hexokinase. Proc Natl Acad Sci U S A, 1978. 75(10): p. 4848-52.

260. Gerstein M., Lesk A.M., and Chothia C., Structural mechanisms for domain movements in proteins. Biochemistry, 1994. 33(22): p. 6739-49.

261. De Boeck H. and Zidovetzki R., Interactions of saturated diacylglycerols with phosphatidylcholine bilayers: a deuterium NMR study. Biochemistry, 1992. 31(2): p. 623-630.

262. Punta M. and Rost B., Protein folding rates estimated from contact predictions. J Mol Biol, 2005. 348(3): p. 507-12.

263. Gromiha M.M. and Selvaraj S., Importance of long-range interactions in protein folding. Biophys Chem, 2015. 7(1): p. 49-68.

264. Heringa J. and Argos P., Side-chain clusters in protein structures and their role in protein folding. J Mol Biol, 1991. 220(1): p. 151-71.

References

189

265. Kannan N., Selvaraj S., Gromiha M.M., and Vishveshwara S., Clusters in α/β barrel proteins: Implications for protein structure, function, and folding: A graph theoretical approach. Proteins: Structure, Function, and Bioinformatics, 2001. 43(2): p. 103-12.

266. Baker E. and Hubbard R., Hydrogen bonding in globular proteins. Prog Biophys Mol Biol, 1984. 44(2): p. 97-179.

267. Burley S. and Petsko G., Weakly polar interactions in proteins, in Advances in protein chemistry. Adv Protein Chem, 1988. 39: p. 125-89.

268. Brandl M., Weiss M.S., Jabs A., Sühnel J., and Hilgenfeld R., Ch⋯ π-interactions in proteins. J Mol Biol, 2001. 307(1): p. 357-77.

269. Mitchell J.B., Thornton J.M., Singh J., and Price S.L., Towards an understanding of the arginine-aspartate interaction. J Mol Biol, 1992. 226(1): p. 251-62.

270. Rose G.D. and Wolfenden R., Hydrogen bonding, hydrophobicity, packing, and protein folding. Annu Rev Biophys Biomol Struct, 1993. 22(1): p. 381-415.

271. Ali M.H. and Imperiali B., Protein oligomerization: how and why. Bioorg Med Chem, 2005. 13(17): p. 5013-20.

272. Duarte J.M., Biyani N., Baskaran K., and Capitani G., An analysis of oligomerization interfaces in transmembrane proteins. BMC Struct Biol, 2013. 13: p. 21.

273. Tsuchiya Y., Nakamura H., and Kinoshita K., Discrimination between biological interfaces and crystal-packing contacts. Adv Appl Bioinform Chem, 2008. 1: p. 99-113.

274. Smith S.O., Song D., Shekar S., Groesbeek M., Ziliox M., and Aimoto S., Structure of the transmembrane dimer interface of glycophorin A in membrane bilayers. Biochemistry, 2001. 40(22): p. 6553-8.

275. Wang S., Munro R.A., Kim S.Y., Jung K.H., Brown L.S., and Ladizhansky V., Paramagnetic relaxation enhancement reveals oligomerization interface of a membrane protein. J Am Chem Soc, 2012. 134(41): p. 16995-8.

276. Bayro M.J., Debelouchina G.T., Eddy M.T., Birkett N.R., MacPhee C.E., Rosay M., Maas W.E., Dobson C.M., and Griffin R.G., Intermolecular structure determination of amyloid fibrils with magic-angle spinning and dynamic nuclear polarization NMR. J Am Chem Soc, 2011. 133(35): p. 13967-74.

277. Van Melckebeke H., Wasmer C., Lange A., Ab E., Loquet A., Bockmann A., and Meier B.H., Atomic-resolution three-dimensional structure of HET-s(218-289) amyloid fibrils by solid-state NMR spectroscopy. J Am Chem Soc, 2010. 132(39): p. 13765-75.

278. Hall D.A., Maus D.C., Gerfen G.J., Inati S.J., Becerra L.R., Dahlquist F.W., and Griffin R.G., Polarization-enhanced NMR spectroscopy of biomolecules in frozen solution. Science, 1997. 276(5314): p. 930-2.

279. Bajaj V.S., Mak-Jurkauskas M.L., Belenky M., Herzfeld J., and Griffin R.G., Functional and shunt states of bacteriorhodopsin resolved by 250 GHz dynamic nuclear polarization-enhanced solid-state NMR. Proc Natl Acad Sci USA, 2009. 106(23): p. 9244-9.

280. Jacso T., Franks W.T., Rose H., Fink U., Broecker J., Keller S., Oschkinat H., and Reif B., Characterization of membrane proteins in isolated native cellular membranes by dynamic nuclear polarization solid-state NMR spectroscopy without purification and reconstitution. Angew Chem Int Ed Engl, 2012. 51(2): p. 432-5.

281. Ong Y.S., Lakatos A., Becker-Baldus J., Pos K.M., and Glaubitz C., Detecting substrates bound to the secondary multidrug efflux pump EmrE by DNP-enhanced solid-state NMR. J Am Chem Soc, 2013. 135(42): p. 15754-62.

282. Morgner N., Kleinschroth T., Barth H.D., Ludwig B., and Brutschy B., A novel approach to analyze membrane proteins by laser mass spectrometry: from

References

190

protein subunits to the integral complex. J Am Soc Mass Spectrom, 2007. 18(8): p. 1429-38.

283. Wittig I., Braun H.P., and Schagger H., Blue native PAGE. Nat Protoc, 2006. 1(1): p. 418-28.

284. Kleinekofort W., Avdiev J., and Brutschy B., A new method of laser desorption mass spectrometry for the study of biological macromolecules. Int J Mass Spectrom Ion Proc, 1996. 152(2-3): p. 135-42.

285. Brouillette C.G., Muccio D.D., and Finney T.K., pH dependence of bacteriorhodopsin thermal unfolding. Biochemistry, 1987. 26(23): p. 7431-8.

286. Brouillette C.G., McMichens R.B., Stern L.J., and Khorana H.G., Structure and thermal stability of monomeric bacteriorhodopsin in mixed phospholipid/detergent micelles. Proteins, 1989. 5(1): p. 38-46.

287. Kahn T.W., Sturtevant J.M., and Engelman D.M., Thermodynamic measurements of the contributions of helix-connecting loops and of retinal to the stability of bacteriorhodopsin. Biochemistry, 1992. 31(37): p. 8829-39.

288. Davio S.R. and Low P.S., Characterization of the calorimetric C transition of the human erythrocyte membrane. Biochemistry, 1982. 21(15): p. 3585-93.

289. Maneri L.R. and Low P.S., Structural stability of the erythrocyte anion transporter, band 3, in different lipid environments. A differential scanning calorimetric study. J Biol Chem, 1988. 263(31): p. 16170-8.

290. Rigell C.W., de Saussure C., and Freire E., Protein and lipid structural transitions in cytochrome c oxidase-dimyristoylphosphatidylcholine reconstitutions. Biochemistry, 1985. 24(20): p. 5638-46.

291. Morin P.E., Diggs D., and Freire E., Thermal stability of membrane-reconstituted yeast cytochrome c oxidase. Biochemistry, 1990. 29(3): p. 781-8.

292. Cramer W.A., Whitmarsh J., and Low P.S., Differential scanning calorimetry of chloroplast membranes: identification of an endothermic transition associated with the water-splitting complex of photosystem II. Biochemistry, 1981. 20(1): p. 157-62.

293. Thompson L.K., Sturtevant J.M., and Brudvig G.W., Differential scanning calorimetric studies of photosystem II: evidence for a structural role for cytochrome b559 in the oxygen-evolving complex. Biochemistry, 1986. 25(20): p. 6161-9.

294. Thompson L.K., Blaylock R., Sturtevant J.M., and Brudvig G.W., Molecular basis of the heat denaturation of photosystem II. Biochemistry, 1989. 28(16): p. 6686-95.

295. Huang K.S., Bayley H., Liao M.J., London E., and Khorana H.G., Refolding of an integral membrane protein. Denaturation, renaturation, and reconstitution of intact bacteriorhodopsin and two proteolytic fragments. J Biol Chem, 1981. 256(8): p. 3802-9.

296. London E. and Khorana H.G., Denaturation and renaturation of bacteriorhodopsin in detergents and lipid-detergent mixtures. J Biol Chem, 1982. 257(12): p. 7003-11.

297. Dornmair K., Kiefer H., and Jahnig F., Refolding of an integral membrane protein. OmpA of Escherichia coli. J Biol Chem, 1990. 265(31): p. 18907-11.

298. Eisele J.L. and Rosenbusch J.P., In vitro folding and oligomerization of a membrane protein. Transition of bacterial porin from random coil to native conformation. J Biol Chem, 1990. 265(18): p. 10217-20.

299. Klug C.S., Su W., Liu J., Klebba P.E., and Feix J.B., Denaturant unfolding of the ferric enterobactin receptor and ligand-induced stabilization studied by site-directed spin labeling. Biochemistry, 1995. 34(43): p. 14230-6.

300. Reynolds J.A. and Tanford C., The gross conformation of protein-sodium dodecyl sulfate complexes. J Biol Chem, 1970. 245(19): p. 5161-5.

References

191

301. Mattice W.L., Riser J.M., and Clark D.S., Conformational properties of the complexes formed by proteins and sodium dodecyl sulfate. Biochemistry, 1976. 15(19): p. 4264-72.

302. Fox J.L., Stevens S.E., Jr., Taylor C.P., Jr., and Poulsen L.L., SDS removal from protein by polystyrene beads. Anal Biochem, 1978. 87(1): p. 253-6.

303. Berg J., Tymoczko J., and Stryer L., Biochemistry, 5th edition. New York. 2002, New York: WH Freeman & Co.

304. Weiss G.A., Watanabe C.K., Zhong A., Goddard A., and Sidhu S.S., Rapid mapping of protein functional epitopes by combinatorial alanine scanning. Proc Natl Acad Sci USA, 2000. 97(16): p. 8950-4.

305. Simonsen S.M., Sando L., Rosengren K.J., Wang C.K., Colgrave M.L., Daly N.L., and Craik D.J., Alanine scanning mutagenesis of the prototypic cyclotide reveals a cluster of residues essential for bioactivity. J Biol Chem, 2008. 283(15): p. 9805-13.

306. Howlader M.T., Kagawa Y., Miyakawa A., Yamamoto A., Taniguchi T., Hayakawa T., and Sakai H., Alanine scanning analyses of the three major loops in domain II of Bacillus thuringiensis mosquitocidal toxin Cry4Aa. Appl Environ Microbiol, 2010. 76(3): p. 860-5.

307. Jaudzems K., Bertarello A., Chaudhari S.R., Pica A., Cala-De Paepe D., Barbet-Massin E., Pell A.J., Akopjana I., Kotelovica S., and Gajan D., Dynamic nuclear polarization enhanced biomolecular NMR spectroscopy at high magnetic field with fast magic‐angle spinning. Angew Chem, 2018. 57(25): p. 7458-62

Declaration of contributions

192


Except where stated otherwise by reference or acknowledgment, the work presented

was generated by myself under the supervision of my advisors during my doctoral

studies. All contributions from colleagues are explicitly referenced in the thesis. The

material listed below was obtained in the context of collaborative research:

Figure 17c

o Title

Characterization of the purity and oligomeric state of wtDGK in DDM micelles.

(c) The trimeric state of wtDGK in DDM micelles is confirmed by LILBID-MS.

o Collaboration partner

The LILBID mass spectrum was recorded by Oliver Peetz of the research group

of Prof. Dr. Nina Morgner, Institute of Physical and Theoretical Chemistry,

Goethe University Frankfurt am Main.

o My contribution

I provided the sample for the LILBID MS measurements and modified the

figure.

Figure 28

o Title

Automated resonance assignment by ssFLYA confirms 91.5% of the backbone

and 89.1% of all (backbone + side chains) assignments obtained manually.


ssFLYA was performed by Dr. Sina Kazemi of the research group of Prof. Dr.

Peter Güntert, Institute for Biophysical Chemistry, Goethe University Frankfurt

am Main. He also created this figure.

o My contribution

I provided the peak lists of the 3D NCACX, NCOCX and CONCA spectra of

uniform and reverse labelled DGK.

Figure 32

o Title

Solid state NMR 1H detected 2D 1H-15N correlation spectrum (2D hNH) of U-

13C,15N-DGK in phospholipid bilayers.


193


The 2D 1H-15N correlation (2D hNH) spectrum was recorded by Dr. Venita

Decker at Bruker BioSpin GmbH in Rheinstetten.

o My contribution

I provided the sample for the 1H detected 2D 1H-15N correlation experiment and

transferred the assignments from 13C/15N detected experiments on well-

resolved residues in this spectrum.

Figure 43c

o Title

Creation of active mixed labelled trimers of DGK. (c) LILBID-MS confirms the

predominantly monomeric state of DGK in SDS micelles.


The LILBID mass spectrum was recorded by Oliver Peetz of the research group

of Prof. Dr. Nina Morgner, Institute of Physical and Theoretical Chemistry,

Goethe University Frankfurt am Main.

o My contribution

I provided the sample for the LILBID MS measurements and modified the

figure.

1,2-Dibutyrylglycerol (DBG)


The synthesis of 1,2-dibutyrylglycerol (DBG) was carried out by Andreas Jakob

of the research group of Prof. Dr. Alexander Heckel, Institute of Organic

Chemistry and Chemical Biology, Goethe University Frankfurt am Main.

Publications

The sections 3.2.3, 3.2.5, 3.2.6, 4.1, 4.2.1, 4.2.2, 4.2.3, 4.2.3.1, 4.2.4, 4.2.5, 4.2.6,

4.2.7, 5.2, 5.3, 6.2, 6.3.2.1, 6.3.2.2, 6.3.2.3.2, 6.3.2.3.3 correspond (in part) to the

manuscript Möbius et al.: Global response of wild-type E. coli diacylglycerol kinase

towards nucleotide and lipid substrate binding observed by 3D and 2D MAS NMR.

Figure 1; Figure 17a,b; Figure 20; Figure 21a,b; Figure 22; Figure 24; Figure 26; Figure

27; Figure 28; Figure 29a,b,c; Figure 30b; Figure 31a; Figure 33a,b,c; Figure 34a,b;

Figure 35a,b; Figure 36a,b,c,d; Figure 37a,b,c,d; Figure 38a,b; Figure 39; Figure 40a,b;

Figure 41a,b; Figure 42a,b; Figure 43a,d; Figure 44a,b; Figure 47; Figure 48; Figure 49

and Figure 50 as well as Table S4; Table S5 and Table S6 are adapted from this

manuscript.

structural and functional studies on e.coli diacylglycerol

Documents