design of a phosphorylatable pdz domain with peptide ... · structure article design of a...

Structure

Article

Design of a Phosphorylatable PDZ Domainwith Peptide-Specific Affinity ChangesColin A. Smith,1,8,* Catherine A. Shi,2 Matthew K. Chroust,3 Thomas E. Bliska,6 Mark J.S. Kelly,4 Matthew P. Jacobson,4,7

and Tanja Kortemme5,7,*1Graduate Program in Biological and Medical Informatics2Graduate Group in Biophysics3School of Dentistry4Department of Pharmaceutical Chemistry5Department of Bioengineering and Therapeutic Sciences

University of California San Francisco, 1700 4th Street, San Francisco, CA 94158, USA6Williams College, 880 Main Street, Williamstown, MA 01267, USA7California Institute for Quantitative Biosciences, 1700 4th Street, San Francisco, CA 94158, USA8Present address: Department of Theoretical and Computational Biophysics, Max Planck Institute for Biophysical Chemistry,

Am Fassberg 11, Gottingen, Germany*Correspondence: [email protected] (C.A.S.), [email protected] (T.K.)

http://dx.doi.org/10.1016/j.str.2012.10.007

SUMMARY

Phosphorylation is one of the most common post-translational modifications controlling cellular pro-tein activity. Here, we describe a combined com-putational and experimental strategy to design newphosphorylation sites into globular proteins to regu-late their functions. We target a peptide recognitionprotein, the Erbin PDZ domain, to be phosphorylatedby cAMP-dependent protein kinase. Comparing thefive successful designs to the unsuccessful cases,we find a trade-off between protein stability and theability to be modified by phosphorylation. In twodesigns, Erbin’s peptide binding function is modifiedby phosphorylation, where the presence of the phos-phate group destabilizes peptide binding. One ofthese showed an additional switch in specificityby introducing favorable interactions between adesigned arginine in the peptide and phosphoserineon the PDZ domain. Because of the diversity of PDZdomains, this opens avenues for the design of re-lated phosphoswitchable domains to create a reper-toire of regulatable interaction parts for syntheticbiology.

INTRODUCTION

Rational protein design has been increasingly successful at pro-

ducing proteins with new functions. Many early results involved

adapting an existing function for new uses, such as changing

enzyme substrate specificity (Wells et al., 1987; Yoshikuni

et al., 2006), modifying protein interaction specificity (Shifman

and Mayo, 2003; Kortemme et al., 2004; Grigoryan et al., 2009;

Kapp et al., 2012), or altering fluorescent properties (Treynor

et al., 2007). More recent efforts have shown remarkable prog-

ress in creating proteins with entirely new functions, including

54 Structure 21, 54–64, January 8, 2013 ª2013 Elsevier Ltd All rights

enzymatic activity (Jiang et al., 2008; Rothlisberger et al.,

2008), de novo protein binding (Karanicolas et al., 2011), bio-

mineralization catalysis (Masica et al., 2010), and oxygen trans-

port (Koder et al., 2009). Proteins have also been engineered to

adapt and respond to their environment in different ways, such

as pH-sensitive antibodies (Murtaugh et al., 2011), proteins

that change folds in response to pH or other ions (Cerasoli

et al., 2005; Palmer et al., 2006; Ambroggio and Kuhlman,

2006), small domains that fold or unfold upon phosphorylation

(Balakrishnan and Zondlo, 2006; Riemen and Waters, 2009),

phosphorylation-induced oligomerization of alpha helices (Szilak

et al., 1997; Signarvic and DeGrado, 2003), and FRET-based

reporters of protein kinase activity (Zhang et al., 2001). Such

designs can form the basis of synthetic, posttranslational regula-

tory machinery for programming cellular behavior.

Nature has evolved many mechanisms for modifying and

regulating protein function, with one of the most pervasive being

phosphorylation. One of the simplest regulatory mechanisms is

disruption of a protein-protein interface via addition of a phos-

phate group. Other proteins create interactions upon phosphor-

ylation, with domain families, such as SH2 and 14-3-3, having

specifically evolved for this purpose. Phosphorylation can also

alter intramolecular properties that change protein activity, with

phosphorylation of the kinase ‘‘activation loop’’ being a canonical

example (Steinberg et al., 1993; Adams et al., 1995).

Phosphorylation sites can be readily introduced into proteins

by fusion of an existing phosphorylatable domain or disordered

region. One of the first reported instances of a phosphorylation

site introduced through site-specific mutation was the incorpo-

ration of a cAMP-dependent protein kinase (PKA) recognition

sequence in the C-terminal tail of human interferon a (IFN-a) (Li

et al., 1989), with the goal of creating radioactive 32P-labeled

proteins for research and clinical use. That work was followed

by other examples, including mutation of the C-terminal region

of a monoclonal antibody to be recognized by casein kinase I

(CKI) (Lin et al., 1999). Another phosphosite was designed in

the semiflexible linker region between the Fab and Fc regions

of a monoclonal antibody heavy chain (Wu et al., 2004). Beyond

producing radiolabeled proteins, other phosphosite design

reserved

mailto:[email protected]

mailto:[email protected]


AKQEIRVRVEKDPELGFSISGGVGGRGNPFRPDDDGIFVTRVQPEGPASKLLQPGDKIIQANGYSFINIEHGQAVSLLKTFQNTVELIIVREVSS

RRXSIKRXSI...

Scan Degenerate PKA Recognition Motif Over

91 PDZ Positions

Filter Destabilizing Mutations with

Rosetta

1235 Sequences91 Positions




Filter Sequences with pkaPS Phosphosite

Predictor

Filter by Visual Inspection

-18 PFRPDDDGIFVTRVQRRGSASKLLQPGDKIIQANGYSFINIE +23

Score 42 Residues Around Target Serine

Figure 1. Computational Phosphosite Design Protocol Applied and

Tested Here

Table 1. Scoring of Initial Phosphosite Designs

Protein

Mutant

Sequence

WT

Sequence

Rosetta D

Score

pkaPS

Score

S13 RKDSE EKDPE �5.50 0.96

S14 KRPSL KDPEL �1.54 1.31

S33 FRPSV FRPDD �0.44 1.09

S43 RRVSP TRVQP �0.12 1.76

S45 VRPSG VQPEG �3.17 1.08

S47 RRGSA PEGPA �0.72 2.06

S53 KRLSP KLLQP �0.50 1.37

S82 RRFSF KTFQN �0.58 1.15

Initial phosphosite designs based on having a negative change in Rosetta

score (relative to thewild-type) and a high pkaPS score.Mutated residues

are in bold.

Structure

Design of a Phosphoswitchable PDZ Domain

efforts have aimed at modulating protein-protein interactions

and protein regulation. Yeh and colleagues (Yeh et al., 2007)

mutated a peptide recognized by the syntrophin PDZ domain

so that it could also be phosphorylated by PKA. Upon phosphor-

ylation the PDZ-peptide interaction was destabilized, which

allowed activation of individual guanine exchange factors

(GEF) fused between the peptide and PDZ domain, whose inter-

action conformationally restricted and inactivated the GEF.

As illustrated above, most previous studies have focused on

introducing phosphorylation sites into sequences outside folded

protein domains or in triggering folding/unfolding transitions

upon phosphorylation (Balakrishnan and Zondlo, 2006; Riemen

and Waters, 2009). Here, we describe the work that involves

phosphorylation of a globular domain itself to modulate its func-

tion. This should enable new avenues for designing controllable

domain-domain interactions, enzyme activities, and conforma-

tional changes. In addition, characterization of design successes

and failures may provide clues about how phosphorylation

sites evolve in folded domains. We make use of the Erbin PDZ

domain, whose peptide interaction specificities have been ex-

tensively characterized with phage display (Laura et al., 2002;

Tonikian et al., 2008; Ernst et al., 2009). We first apply a compu-

tational design strategy to select mutations that are predicted to

not disrupt native binding activity in the unphosphorylated form

and to be phosphorylated by PKA. Guided by characteriza-

tion of thermostability, chemical shifts, and solvent exposure,

we then evaluate the successes and limitations of the design

strategy and generate several additional successful designs.

Finally, we characterize the changes in binding affinity and inter-

action specificity upon phosphorylation with a set of peptide

variants.

RESULTS

Initial Phosphosite DesignsWe first wanted to develop and test a computational strategy to

generate sites in the Erbin PDZ domain that could be phosphor-

ylated by cAMP dependent protein kinase (PKA), irrespective of

whether phosphorylation would affect peptide binding to the

PDZ domain. Exhaustive scanning of a degenerate PKA recogni-

tionmotif yielded over 1,000 potential sequences (see the Exper-

imental Procedures; Figure 1). We used the change in Rosetta

score (Smith and Kortemme, 2008) in the presence of the

peptide as a proxy for the extent of disruption in fold stability

or binding affinity. Our protocol incorporated limited amounts

of backbone flexibility shown to better predict side-chain con-

formations (Smith and Kortemme, 2008). To computationally

approximate the probability of phosphorylation, we used pkaPS

(Neuberger et al., 2007), a sequence-based predictor of phos-

phorylation sites optimized for PKA. Using those computational

tools and subsequent manual inspection of a reduced number of

candidates (Figure 1), we selected eight sequences for testing

(Figure 2; Table 1). Each designwas named for the serine residue

targeted for phosphorylation (e.g., S13 standing for serine 13).

Using a radiolabeled phosphotransfer assay, the eight designs

were tested for phosphorylation by PKA. S47 and S82 both

showed significant levels of phosphorylation, with a stronger

signal at S47 (Figure 3A). Using mass spectrometry, we con-

firmed that each domain was singly phosphorylated (i.e.,

Structure 21

M+80) and found that S47 was phosphorylated to completion

in the time used for the gel-based assays, whereas S82 was

not. Residue 47 is in the middle of a 15-residue loop between

the third and fourth beta strands of the Erbin PDZ domain (Fig-

ure 2B). Residue 82 is located on a short loop between the alpha

helix and the last beta strand. The pattern of phosphorylation did

not strictly follow the pkaPS scores: although S47 had the high-

est pkaPS score (2.1; Table 1), the designs with next three

highest scores (1.8–1.3) did not phosphorylate. However, S82,

the construct with the next lowest score (1.2), could be

phosphorylated.

Given that known phosphorylation sites are often found in

disordered regions of proteins (Iakoucheva et al., 2004), we

hypothesized that disorder and/or protein stability may play

a role in substrate phosphorylation. To determine whether this

was the case for our designs, we used apparent melting temper-

atures from thermal denaturation as a proxy for protein stability.

, 54–64, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 55

S82S53S47S45S43S33S14S13WT

10 12 14 30 32 34 36 38 40 42 44 46 48 50 52 54 78 80 82 84

VVVVVVVVV

EEEEEEERE

KKKKKKKKK

DDDDDDRDD

PPPPPPPSP

EEEEEESEE

LLLLLLLLL

FFFFFFFFF

RRRRRRRRR

PPPPPPPPP

DDDDDSDDD

DDDDDVDDD

DDDDDDDDD

GGGGGGGGG

IIIIIIIII

FFFFFFFFF

VVVVVVVVV

TTTTRTTTT

RRRRRRRRR

VVVVVVVVV

QQQRSQQQQ

PPRPPPPPP

EERSEEEEE

GGGGGGGGG

PPSPPPPPP

AAAAAAAAA

SSSSSSSSS

KKKKKKKKK

LRLLLLLLL

LLLLLLLLL

QSQQQQQQQ

PPPPPPPPP

LLLLLLLLL

RKKKKKKKK

RTTTTTTTT

FFFFFFFFF

SQQQQQQQQ

FNNNNNNNN

TTTTTTTTT

... ... ... ...

A

B

Figure 2. Initial Phosphosite Designs

(A) Protein sequences of wild-type Erbin PDZ

domain and eight designed phosphorylation

candidates. Amino acids are numbered such that

residue 1 is the first amino acid present in the PDB

structure 1MFG (Birrane et al., 2003) (i.e., residue

1277 in full-length Erbin). Serines targeted for

phosphorylation are shown in red, added arginines

are shown in blue, and added hydrophobic amino

acids are shown in orange. See also Figure S1 for

gel filtration chromatography to characterize the

oligomeric state of WT and S47.

(B) Structural distribution of target serines shown

on PDB 1MFG (Birrane et al., 2003) with C-alpha

atoms shown as spheres and the bound peptide

shown in purple. All structure figures were

rendered with PyMOL (Schrodinger, LLC, 2012).

Structure


Figure 3B shows melting curves measured by circular dichroism

for wild-type Erbin and the eight designs. S82 and S47, the two

variants that could be phosphorylated, also had the two lowest

melting temperatures of the initial eight designs (Table 2). These

lower melting temperatures relative to the wild-type protein were

unexpected (based on our design criterion to not destabilize the

protein) and indicate known difficulties in accurately estimating

the energetic effects of surface mutations. As discussed later,

these two designs were still functional and bound to all tested

peptides (Table 3). S47 showed a 4- to 6-fold reduction in

binding affinity over wild-type and S82 bound slightly tighter

than wild-type (less than a 2-fold increase). Reversible GuHCl

denaturation of WT and S47 mirrored the thermal denaturation

results and showed that S47 had lower fold stability than WT

(Figure S2 available online). The qualitative rates of phosphoryla-

tion do not necessarily correlate, as S82 was less thermostable

than the more rapidly phosphorylated S47.

Nuclear Magnetic Resonance CharacterizationTo investigate structural differences between wild-type and

S47, the mutant that showed the greatest rate of phos-

phorylation, we acquired HSQC spectra for it as well as

wild-type Erbin in the unbound form (Figure 4A). Although

assignments were previously determined for wild-type Erbin

bound to peptide TGWETWV (Skelton et al., 2003), there

56 Structure 21, 54–64, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

were significant differences between it

and the free form, preventing assign-

ment transfer. Therefore, we obtained

de novo 1H, 15N, 13CA, and 13CB reso-

nance assignments for both WT and

S47 in the unbound form. In Figure 4A,

peaks for equivalent residues in WT

and S47 are connected by lines. Also,

as indicated, the spectrum for S47 had

two additional 1H-15N peaks because

of the P44R and P47S mutations. Fig-

ure 4B shows atom-specific chemical

shift differences, as well as the overall

chemical shift difference normalized by

the standard deviation (s) of all reso-

nances found in a database of protein chemical shifts (Ulrich

et al., 2008). As expected, the most extensive chemical shift

changes (up to 1s) were for residues 42–52, which are next

to the mutated residues in sequence. Interestingly, a segment

of residues adjacent in the tertiary structure, 7–17, also had

large shift differences of up to 0.7s (Figure 4C), and a more

distant segment, 82–85, had smaller differences up to 0.3s.

These differences suggest that there was a change in structure

and/or dynamics that was more than would be expected for a

mutation of several solvent-exposed residues.

To accommodate recognition by PKA, there is likely some

degree of local or global structural opening/unfolding that is

either a transient or stable feature of the solution ensemble.

We reasoned that such opening is likely associated with an in-

creased solvent exchange of amide hydrogens. Using CLEANEX

hydrogen exchange experiments (Hwang et al., 1998) with

mixing times from 10–100 ms, we observed residues in all

three segments mentioned above having greater hydrogen

exchange in S47 than in WT (Figure 5). These changes in solvent

exchange are among the most significant observed over the

whole protein structure. Assuming the hydrogen exchange

profiles of the other nonphosphorylatable designs are similar to

WT, this supports a model in which S47 is shifted toward

a more open conformation at equilibrium, possibly contributing

to recognition and phosphorylation by PKA.

30 40 50 60 70 80

Nor

mal

ized

Elli

ptic

ity1.

00.

50.

0 S82S47S33S53S43S45S13S14WT

WT S13 S14 S33 S43 S45 S47 S53 S82A

B

Figure 3. Phosphorylation and Thermostability of Initial Designs

(A) SDS-PAGE gel (10%–20%) of Erbin PDZ domain designs phosphorylated

with 32P ATP.

(B) Circular dichroism signal at 218 nm showing thermal denaturation of

designs. Normalization is described in the Experimental Procedures. See also

Figure S2 for equilibrium GuHCl denaturation curves of WT and S47.

Table 2. ApparentMelting Temperatures of Phosphosite Designs

Initial Designs Secondary Designs

Protein Tm,app Protein Tm,app

S82 44 S13-A47 43

S47 46 S43-A47 44

S33 49 S14-A47 44

S53 50 S45-A47 48

S43 57 S47A 50

S45 58 S13-A47a 50

S13 60 S14-A47a 51

S14 61 pS13-A47a 55

WT 66 pS14-A47a 55

Temperatures (Tm,app) in degrees Celcius from irreversible thermal dena-

turation measured by circular dichroism signal at 218 nm.a6xHis tag removed. Phosphorylated designs are indicated by ‘‘p’’.

Table 3. Binding Affinities of Unphosphorylated and

Phosphorylated Designs

Peptides

Protein TGWETWV TGWETRV TGWETDV

S13-A47 0.022 ± 0.0025 1.4 ± 0.03 0.71 ± 0.11

pS13-A47 0.072 ± 0.0016 4.5 ± 0.61 3.2 ± 0.65

S14-A47 0.017 ± 0.0011 3.3 ± 0.07 0.54 ± 0.10

pS14-A47 0.17 ± 0.032 22 ± 2.3 4.7 ± 1.3

S43-A47 0.050 ± 0.0023 26 ± 0.3 16 ± 0.2

pS43-A47 0.45 ± 0.034 37 ± 2.8 123 ± 7.7

S47 0.094 ± 0.018 28 ± 0.6 21 ± 0.6

pS47 0.14 ± 0.003 25 ± 1.1 25 ± 0.8

S82 0.017 ± 0.0008 3.2 ± 0.03 1.9 ± 0.12

pS82 0.024 ± 0.0028 4.5 ± 0.22 3.1 ± 0.04

WT 0.022 ± 0.0016 5.2 ± 0.30 3.5 ± 0.18

Dissociation constants (Kd) in mM derived from fluorescence polarization

experiments for unphosphorylated and phosphorylated PDZ domains

with three peptides differing at the �1 position. Standard errors from

multiple independent experiments are given. See also Figure S5 for indi-

vidual binding curves. Phosphorylated designs are indicated by ‘‘p’’.

Structure


Second Round of Phosphosite DesignsAfter noting the chemical shift differences in segments 7–17 and

42–52 of S47, we hypothesized that the structural changes

enabling S47 to be phosphorylated might also rescue phosphor-

ylation at other sites located in the altered regions. Furthermore,

the increase in S47 hydrogen exchange at residues 15, 45, and

46 suggested that the changes might be related to structural

opening. Therefore, we designed a different template protein,

A47, which included the arginine mutations of S47 (P44R and

E45R) but incorporated an alanine at position 47 to prevent

phosphorylation. Using that template, we reintroduced the

mutations for the previous S13, S14, S43, and S45 designs (Fig-

ure 6A). These candidate phosphosites were chosen because

they were located on the structurally perturbed loops and in

regions near the peptide that were more likely to affect binding

affinity.

Both radiolabeled phosphotransfer assays and mass spec-

trometry indicated that A47 did not phosphorylate (Figure 6B),

as expected, showing that phosphorylation of S47 was specific

to serine 47 and that the A47 construct was suitable for addition

of other sites. The same assays showed that three out of the four

constructs (S13, S14, and S43) could now be phosphorylated in

the A47 background (Figure 6B). Determination of the thermo-

stability using circular dichroism showed that A47 was more

stable than S47 but still significantly less stable than WT. All of

the secondary phosphosite designs were less stable than the

A47 construct on which they were based, and the three phos-

phorylatable designs were less stable than S47.

Beyond stability (Table 2 lists apparent melting temperatures)

other sequence-based factors could play a role in the failure of

the other designs. S33, S45–A47, and S53 had apparent melting

temperatures in the 48�C–50�C range that were close to the

other phosphorylatable designs. However, although all have an

arginine residue two sequence positions before the serine,

none of them have an arginine three positions before, likely

Structure 21

reducing their recognition by PKA, which strongly prefers two

upstream arginines at the �3 and �2 positions (Kemp et al.,

1977; Hutti et al., 2004).

In the context of the present study, it is difficult to fully decon-

volute all of the sequence and structure-based aspects of phos-

phorylation, in particular the relative importance of local versus

global instability. Although thermal denaturation was performed

on all designs, more extensive nuclear magnetic resonance

(NMR) characterization of solvent exposure was limited to WT

and S47. Nevertheless, the locally destabilizing effects of the

S47 mutations, along with their ability to rescue the phosphory-

lation of S13, S14, and S43, suggests that local substrate desta-

bilization may play an important role in phosphorylation.

Binding Affinity Changes upon PhosphorylationTo determine how the engineered mutations and phosphoryla-

tion affected PDZ domain function, we measured dissociation


A

B

C Figure 4. Chemical Shift Differences

between WT and S47

(A) 2D 15N HSQC of wild-type (blue) and S47 (red)

Erbin PDZ domains with equivalent residues con-

nected by black lines and residues mutated from

proline given numbered labels. See also Figure S3

for spectra at multiple concentrations.

(B) Normalized overall (see the Experimental

Procedures) and atom-specific chemical shift

differences between wild-type and S47.

(C) Overall chemical shift differences displayed on

the PDB structure 1MFG (Birrane et al., 2003) with

C-alpha radii scaled by 50%–100% to reflect nor-

malized overall chemical shift difference. Mutated

residues, not contained in the plots, are shown in

purple. The peptide (not included in the NMR ex-

periment) is shown in purple for referencepurposes.

Structure


constants (Kd) for three synthetic peptides (Figure 7A). The first

peptide was a phage-display derived (Laura et al., 2002) high-

affinity sequence (TGWETWV), whose residue positions are

given decreasing numbers starting at zero for the C-terminal

valine. Because the side chain of the�1 position had the closest

proximity to the phosphosite positions (Figure 7B), we hypothe-

sized that electrostatic interactions with the phosphate group

may be possible. Therefore, we also tested peptides with an

arginine (TGWETRV) and aspartate (TGWETDV) at the �1 posi-

tion to evaluate potential electrostatic attraction and repulsion.

Wild-type Erbin had the highest affinity for TGWETWV (22 nM)

and bound approximately two orders of magnitude more weakly

to peptides with the arginine and aspartate residues at the �1

position (see Table 3 and all binding curves in Figure S5).

Depending on the position, mutations in the designed Erbin vari-

ants either increased or decreased the base (unphosphorylated)

58 Structure 21, 54–64, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

binding affinity to peptides. S43–A47 and

S47 bound the three peptides somewhat

weaker than wild-type Erbin by a factor

of up to six, whereas S13–A47, S14–

A47, and S82 all showed improvements

in binding affinity to the TGWETRV and

TGWETDV peptides by up to 6.5-fold.

The mutations in S13–A47, S14–A47,

and S82 were spatially clustered around

the peptide’s negatively charged C

terminus. Each involved the addition of

1–2 arginine residues, which could result

in electrostatic attraction and increased

binding. Subtle changes in the binding

site geometry could also account for the

observed affinity changes.

Upon phosphorylation, the S14–A47

and S43–A47 designs showed nearly

10-fold decreases in binding affinity (Fig-

ure 7A), corresponding to a change in

binding energy of 1.4 kcal/mol. With a

3- to 4-fold decrease, S13–A47 was less

perturbed by phosphorylation. S47 and

S82 both showed negligible decreases

in binding affinity. For S13–A47 and

S14–A47, which showed significant decreases in binding affinity

for all peptides, phosphorylation increased thermostability

(Figure 6D), suggesting that fold stability was not a factor in de-

stabilizing its binding interface. Instead, the degree of binding

affinity change appears to be associated with distance of

the phosphate group from the peptide with residues 14 and 43

being closer and residues 47 and 82being further (see Figure 7B).

The observed changes in binding affinity were similar for both

the tryptophan and aspartate peptides, indicating that there

was little additional electrostatic repulsion resulting from the

aspartate peptide. Double mutant cycle analysis in designed

beta-hairpin peptides has shown that interactions between adja-

cent phosphoserine/tryptophan residues were destabilizing,

whereas phosphoserine/valine were not (Riemen and Waters,

2009). The peptide binding affinities for S43–A47 show a similar

pattern of phosphoserine/tryptophan destabilization, along with

A C

B

Figure 5. Increased Amide Hydrogen Exchange around Phosphorylation Site

(A) Fitted CLEANEX (Hwang et al., 1998) hydrogen exchange curves (solid lines) showing increased solvent accessibility in the S47 design (red) over wild-type

(blue) for residues near the phosphorylation site. The ratio of peak volume (V) to reference peak volume (V0) is shown. Initial velocities are shownwith dashed lines

along with their standard errors (shaded regions).

(B) Difference in initial velocities (k) normalized by the sum of their standard errors (SEk). Points outside of the dashed region have nonoverlapping error bars.

Residues shown in (A) are boxed.

(C) Normalized initial velocity differences displayed on the PDB structure 1MFG (Birrane et al., 2003) with amide hydrogen radii scaled by 50%–100% to reflect

normalized differences in initial velocities. The peptide (not included in the experiment) is shown in purple for reference purposes.

Structure


equivalent interface destabilization from the phosphoserine/

aspartate interaction. However, for S43–A47, the arginine

peptide showed only a marginal decrease in binding affinity

upon phosphorylation. This may indicate that electrostatic or

hydrogen-bonding interactions between the phosphate and

arginine (Figure 7C) largely counteract any destabilizing effects

of phosphorylation. Hydrogen bonds between phosphorylated

amino acids and arginine have been theorized to be particularly

strong (Mandell et al., 2007). For the arginine and aspartate

peptides, this differential destabilization results in a change in

specificity from preferring aspartate in the unphosphorylated

form (R: 26 mM versus D: 16 mM) to arginine in the phosphory-

lated form (R: 37 mM versus D: 123 mM).

DISCUSSION

This study highlights the successful incorporation of phosphory-

lation sites into globular protein structures. The most efficiently

phosphorylated initial design (S47) had decreased thermosta-

bility and increased backbone solvent exposure. By transferring

destabilizing mutations from that design to others, phosphoryla-

tion activity was rescued in three other designed proteins. Upon

phosphorylation, several designs showed approximately 10-fold

reductions in binding affinity and one design exhibited a change

in peptide binding specificity.

Structure 21

The interactions between natural PDZ domains and their

target peptides are sometimes controlled by phosphorylation,

with the most prevalent mechanism being phosphorylation of

the recognized peptide (Ivarsson, 2012). In addition to cognate

peptide phosphorylation, phosphorylation of PDZ domains

themselves occurs naturally and is known to disrupt peptide

binding. One site was found in the first PDZ domain of human

disks large homolog 1 (DLG1-1), which is phosphorylated at

S232 (Gardoni et al., 2003) (aligning with Erbin PDZ P13 or

E14) by Ca2+/calmodulin-dependent protein kinase II (CaMKII).

Another disruptive phosphorylation site is located on the first

PDZ domain of Npt2a-binding protein sodium-hydrogen ex-

changer regulatory factor-1 (NHERF1-1) that is phosphorylated

at S77 (Weinman et al., 2007) (aligning with Erbin PDZ S76).

The Erbin PDZ domain used in this phosphosite design study

is not known to be naturally phosphorylated.

Phosphorylation has been traditionally thought of as a mecha-

nism to modulate a single function, either through binary on/off

switching or a graded response (Pufall et al., 2005). The phos-

phoswitchable PDZ domains described here suggest that it is

possible to design proteins that switch from one function to

another upon phosphorylation. Although there are several

examples of peptide phosphorylation resulting in changes in

binding from one peptide recognition domain to another (Akiva

et al., 2012), changes in specificity of the peptide recognition


30 40 50 60 70 80

Nor

mal

ized

Elli

ptic

ity1.

00.

50.

0 S13 A47S43 A47S14 A47S82S47S45 A47A47WT

A47S13A47

S14A47

S43A47

S45A47 S47 S82

S45 A47S43 A47S14 A47S13 A47

A4710 12 14 38 40 42 44 46 48

VVVVV

EEERE

KKKKK

DDRDD

PPPSP

EESEE

LLLLL

FFFFF

VVVVV

TRTTT

RRRRR

VVVVV

RSQQQ

RRRRR

SRRRR

GGGGG

AAAAA

AAAAA

... ... ...

B

C

A

30 40 50 60 70 80

Nor

mal

ized

Elli

ptic

ity1.

00.

50.

0 S13 A47 ( 6xHis)S14 A47 ( 6xHis)pS13 A47 ( 6xHis)pS14 A47 ( 6xHis)

D

Figure 6. Phosphorylation and Thermostability of Secondary

Designs

(A) Phosphorylation candidate designs based on combining a destabilized

alanine mutant of the S47 variant (A47) with previous designs (S13, S14, S43,

and S45; see Figure 2) containing the desired phosphorylation site near the

peptide-binding site.

(B) SDS-PAGE gel (10%–20%) showing phosphorylation of three of the

second round designs with 32P ATP.

(C) Circular dichroism signal at 218 nm during thermal denaturation shows that

phosphorylatable designs unfold at the same temperature or lower than S47.

(D) Thermal denaturation shows that phosphorylation stabilizes both S13-A47

and S14-A47. See also Figure S4 for thermal denaturation of His-tagged

S14-A47 incubated with TCEP.

Fol

d C

hang

e in

Kd

upon

Pho

spho

ryla

tion

02

46

810

S13 A47 S43 A47 S82S14 A47 S47

1 TGWETWVTGWETRVTGWETDV

A

B

C

R-1

pS43

Figure 7. Peptide-Dependent Changes in Binding Affinity upon

Phosphorylation

(A) Peptide binding affinity decreases upon phosphorylation shown as fold

changes in Kd upon phosphorylation (i.e., Kd,phosphorylated/Kd,unphosphorylated).

Three variants of a phage display derived high affinity peptide (Laura et al.,

2002) were tested, each with a different amino acid (W, R, or D) at the �1

position. Error bars reflect propagated standard errors from Table 3.

(B) C-alpha atoms of the phosphorylated residues and the�1 peptide position

are shown with spheres on the PDB structure 1MFG (Birrane et al., 2003).

(C) Computational model of phosphorylated S43-A47 bound to the TGWETRV

peptide built bymutating and optimizing side chains of the PDB structure 1N7T

(Skelton et al., 2003).

Structure


domains themselves has not been readily observed. The in vitro

change in specificity we demonstrate is relatively modest, and

increases in affinity modulation may be required to produce

observable in vivo effects. With further design iterations, it

may be possible to increase the degree of change in specificity

shown here.

We observed an apparent inverse relationship between

protein stability and phosphorylation. Considerable evidence

already exists that enzymes trade off fold stability in order to opti-

mize their active sites for function (Shoichet et al., 1995) and that


evolution of new functions requires either an excess buffer of

stability or compensatory stabilizing mutations (Tokuriki et al.,

2008). This study illustrates a similar stability-function trade-off

reserved

Structure


in enzyme substrates, where instability appears to be correlated

with function, that is, the ability to be phosphorylated. Moreover,

it has been theorized that evolutionary pressure exists against

protein overstabilization because it makes proteins difficult to

regulate through the cell’s natural degradation machinery

(DePristo et al., 2005). In the case of our designed PDZ domains,

overstabilization would prevent regulation through posttransla-

tional modification. It stands to reason that natural proteins

may have similar constraints to facilitate chemical modifications

required for regulatory or other functional purposes.

For several pairs of designs in this study (i.e., S13/S13-A47

and S14/S14-A47), the sequence differences that enabled phos-

phorylation were 30 residues away from the phosphorylation

site in primary sequence. Predictors of phosphorylation that

only incorporate knowledge about local sequence context and

domain boundaries are incapable of discriminating the rate of

phosphorylation in these pairs. Instead, the most obvious

discriminating factor appears to be protein stability. Without

invoking protein stability, an alternative explanation could be

that surface and/or net-charge effects resulting from two addi-

tional arginine residues near the phosphorylation site promote

recognition by PKA.

It is likely that future studies will benefit from making small

decreases in protein stability a goal of the design process. For

this study, the computational screening process aimed to avoid

destabilization of the globular protein structure. Neither the

computational predictions nor visual inspection of the primarily

solvent-exposed mutations suggested that the PDZ domains

would be significantly destabilized. Although it has been shown

that optimization of charge-charge interactions can stabilize

proteins (Strickler et al., 2006), the stabilizing or destabilizing

effect of adding surface charges is still not well understood.

An advantageous aspect of the protein design strategy used

here is that there is relatively little overlap between the rede-

signed residues and those residues thought to bemost important

for determining PDZ-peptide specificity (Tonikian et al., 2008).

Because of this, the mutations used to introduce a PKA phos-

phorylation site into the Erbin PDZ domain could likely be trans-

ferred to other PDZ domains without significantly perturbing their

specificities. Because of the large natural diversity of PDZ

domain specificities (Tonikian et al., 2008), and the ease with

which other specificities can be synthetically created (Ernst

et al., 2009), it may be possible to expand this work into a reper-

toire of regulatable protein interaction parts for application in

synthetic biology. Such PDZ domains could be incorporated

into protein scaffolds to posttranslationally control or divert

cellular signal transduction (Good et al., 2011) or metabolic flux

through biosynthetic pathways (Dueber et al., 2009).

EXPERIMENTAL PROCEDURES

Phosphosite Design

For the initial round of phosphosite selection, we targeted the entire Erbin PDZ

domain protein sequence (Figure 1). The cAMP-dependent protein kinase

(PKA) recognition motif was represented as a five-residue sequence with

degeneracies in the first and last positions, R/K/X-R-X-S-I/K/F/V/X, with X

being the wild-type amino acid. The 15 possible combinations of this motif

were scanned over the 95-residue sequence taken from the Protein Data

Bank (PDB) code 1MFG (Birrane et al., 2003). The 91 possible positions yielded

1,235 unique candidate sequences. To limit destabilization of both the protein

Structure 21

fold and peptide binding, we only selected sequences whose total Rosetta

score was less than wild-type (i.e., predicted to be no less stable), which

was determined by applying the Backrub point mutation protocol (Smith and

Kortemme, 2008) to PDB code 1MFG with the peptide included. This resulted

in 148 sequences (at 29 positions). To take the context of the surrounding

protein sequence into account, these sequences were then screened for the

likelihood of phosphorylation with the pkaPS (Neuberger et al., 2007)

sequence-based PKA substrate predictor. This left 71 sequences (at 23 posi-

tions) with a pksPS scores greater than one. The candidate sequence with the

best pkaPS score at each position was visually inspected, and of these, the

most promising eight sequences (balancing kinase accessibility and closeness

to the canonical PKA recognition motif) that required six or fewer base pair

changes were selected for experimental screening. Kinase accessibility was

judged primarily by side-chain solvent exposure and how convex (as opposed

to concave) the structure was around the motif residues. This was balanced

against the number of arginine residues at the �2 and �3 positions (i.e.,

more convex sites were allowed to have fewer arginines).

These eight designs were named for the serine position that was targeted for

phosphorylation (i.e., S13 for serine 13). In S13 and S14, lysine 11 was origi-

nally designed to be mutated to arginine. Because mutation of this conserved

lysine has been shown to disrupt both PDZ domain stability and peptide

binding affinity (Harris et al., 2003), it was reverted to wild-type in both designs.

This resulted in a pkaPS score slightly lower than the one for S13, indicating

a slightly reduced predicted probability of phosphorylation. The reduction in

the number of arginine residues and the largely buried lysine were allowed

because of the convex, almost wedge-like shape of the PDZ domain around

these two positions.

Cloning and Mutagenesis

The Erbin PDZ domain (base pairs 3943–4239 of the coding sequence) was

cloned out of full-length Homo sapiens Erbin (GenBank 358679310:311-

4549) that had been previously cloned into pRK5-myc (Borg et al., 2000)

(courtesy of J.P. Borg). This Erbin construct was inserted along with an

N-terminal ATG between the NotI and XhoI sites of a modified pET-47b(+)

expression vector (Novagen, Darmstadt, Germany) that had eight base pairs

(GGGTACCA) after the HRV-3C cleavage site removed (courtesy R.M. Stroud).

For NMR studies, a shortened construct was prepared by inserting it without

the additional N-terminal ATG between the EcoRI and XhoI sites of the modi-

fied pET-47b(+) expression vector.Mutations weremadewith theQuikChange

mutagenesis kit (Agilent, Santa Clara, CA, USA) and primers designed

with the online QuikChange primer design tool (http://www.agilent.com/

genomics/qcpd).

Protein Expression and Purification

All proteinswere expressed in BL21 cells at 25�C, inducedwith 0.5mM IPTG at

OD 0.5 (600 nm), and harvested after 6 hr. Following a standard protocol

(QIAGEN, 2003), cells were lysed after storage at �80�C in 300 mM NaCl,

10 mM imidazole, and 50 mM Na2HPO4 (pH 8) and then sonicated and spun

down. After incubation of the supernatant for 1 hr with Ni-NTA agarose resin

(QIAGEN, Hilden, Germany), beads were spun down and washed three times

with 300 mM NaCl, 20 mM imidazole, and 100 mM Na2HPO4 (pH 8), followed

by elution with 300 mM NaCl, 250 mM imidazole, and 50 mM Na2HPO4 (pH 8).

Protein concentrations were determined with the Pierce Coomassie Plus

Bradford Protein Assay (Thermo Fisher Scientific, Waltham, MA, USA) using

the included bovine serum albumin (BSA) as the standard.

Erbin constructs were expressed in Luria broth (LB) medium, purified, dia-

lyzed into 300 mM NaCl and 50 mM Na2HPO4 (pH 8), and then stored at

4�C. Mass spectrometry analysis of S43-A47 suggested that the five rare

AGA/AGC arginine codons were resulting in misincorporation of one lysine

(mass loss of 28 daltons, M-28; Calderone et al., 1996) in 30%–40% of the

expressed protein. Cotransformation of pET-47 S43-A47 with the pJY2

plasmid (You et al., 1999) (Enzo Life Sciences, Farmingdale, NY, USA) amelio-

rated the effect. For NMR studies, Erbin constructs were expressed in

M9 minimal medium (6 g/l Na2HPO4, 3 g/l KH2PO4, 0.5 g/l NaCl, 247 mg/l

MgSO4, and 14.7 mg/l CaCl2) with 1 g/l 15N NH4Cl (Isotec, St. Louis, MO,

USA) and 4 g/l glucose. Doubly labeled proteins were expressed by

substituting 1 g/l 13C glucose (Isotec). After purification, labeled proteins

were dialyzed into PBS at a pH of 6.5, similar to a previous NMR study (Skelton


http://www.agilent.com/genomics/qcpd

http://www.agilent.com/genomics/qcpd

Structure


et al., 2003), and stored at 4�C. The constructs used for NMR studies were

shown to be monomeric up to 750 mM by high-performance liquid chromatog-

raphy (HPLC) gel filtration (Figure S1) and comparison of one-dimensional

(1D)/two-dimensional (2D)NMRspectra at different concentrations (FigureS3).

The Mus musculus PKA catalytic subunit was expressed using a previously

constructed pET-15b vector (Narayana et al., 1997) (Addgene 14921). Purified

PKA was dialyzed into 200 mM KCl, 5 mM DTT, and 20 mM KH2PO4 (pH 6.5)

and then stored at 4�C.

Protein Constructs

After expression, the longer construct cloned into the NotI and XhoI sites

was 137 amino acids in length, with wild-type having the sequence of

AHHHHHHSAALEVLFQGPGSEFCTGLGAPDVRRQACGRMHELjAKQEIRVRV

EKDPELGFSISGGVGGRGNPFRPDDDGIFVTRVQPEGPASKLLQPGDKIIQANG

YSFINIEHGQAVSLLKTFQNTVELIIVREVSS (the vertical bar immediately

precedes the alanine assigned residue number 1). It was used for radiolabeled

phosphoassays and circular dichroism studies. For binding studies, the 6xHis

tag of this construct was cleaved (see below), giving the 121 amino acid

wild-type sequence GPGSEFCTGLGAPDVRRQACGRMHELjAKQEIRVRVE

KDPELGFSISGGVGGRGNPFRPDDDGIFVTRVQPEGPASKLLQPGDKIIQANG

YSFINIEHGQAVSLLKTFQNTVELIIVREVSS. For NMR studies, the shorter

construct cloned between EcoRI and XhoI yielded a 120 amino acid wild-

type sequence, AHHHHHHSAALEVLFQGPGSEFHELjAKQEIRVRVEKDPEL

GFSISGGVGGRGNPFRPDDDGIFVTRVQPEGPASKLLQPGDKIIQANGYSFINI

EHGQAVSLLKTFQNTVELIIVREVSS.

Phosphorylation Assays

Phosphorylation assays were conducted with 80 mM g-32P ATP (0.012 mCi/ml),

40 mMPDZ domain, 0.4 mMPKA, 2mMDTT, 10mMMgCl2, 300mMNaCl, and

50 mM Na2HPO4 (pH 8). After incubation at 33�C for 1 hr, the reactions were

quenched by addition of Laemmli sample buffer and boiling at 95�C. 10%–

20% Tris-HCl SDS-PAGE gels (BioRad, Hercules, CA, USA) were imaged

using a storage phosphor screen. The number of covalently attached phos-

phates for all proteins was determined via mass spectrometry using a Waters

Micromass LCT Premier LC-MS system.

Circular Dichroism Spectroscopy

Thermal denaturation was performed on a Jasco J-715 spectrometer using

a 1 mm cuvette. Erbin constructs were diluted to 40 mM with 300 mM NaCl

and 50 mM Na2HPO4 (pH 8). Data acquisition was done at 218 nm (the peak

of the native spectrum) using a bandwidth of 1 nm, response time of 16 s,

temperature slope of 1�C/minute, and data pitch of 0.5�C. Between samples,

cuvettes were washed and stored in aqua regia (1:3 solution of concentrated

nitric acid and hydrochloric acid). In other experiments, thermal denaturation

of wild-type Erbin was determined to be irreversible.

Melting data having temperature (t) and ellipticity signal (y) were fit in R

(R Development Core Team, 2012) with the equation

y =yhigh � ylow

1+ eðTm;app�tÞ=s + ylow + at:

The fitted parameters included apparent melting temperature (Tm,app), low

and high values of the signal (ylow and yhigh, respectively), scale of the unfolding

transition (s), and linear background slope (a). Because all curves displayed

similar linear slopes before and after the unfolding transition, the a slope

parameter was fit globally across all protein curves. For visual clarity, plots

were normalized such that ylow = �1, yhigh = 0, and a = 0.

NMR Spectroscopy1H-15N HSQC spectra for WT and S47 were acquired on a Bruker 800 MHz

spectrometer and processed with Bruker Topspin. For all experiments, the

target sample temperature was set to 300.7 K at the console. Complete reso-

nance assignments for both WT (Biological Magnetic Resonance Data Bank

[BMRB] entry 18785) and S47 (BMRB entry 18786) were done with a combina-

tion of HNCA, HNCACB, and CBCA(CO)NH experiments on a 500 MHz Bruker

spectrometer. Spectra were processed in NMRPipe, and assignments were

carried out in CcpNmr Analysis. To determine normalized overall chemical shift

differences between WT and S47, the differences were first divided by the

residue type and atom-specific standard deviations from the Biological


Magnetic Resonance Data Bank (BMRB) statistics stored in CcpNmr Analysis.

The overall difference was then calculated by taking the root-mean-square of

all normalized differences for a given residue.

CLEANEX hydrogen exchange spectra (Hwang et al., 1998) were acquired

on a Bruker 800 MHz spectrometer in separate experiments with mixing times

of 10, 20, 40, 60, 80, and 100 ms. To determine reference peak intensities and

ensure sample homogeneity, a standard Fast-HSQC (FHSQC) was acquired

immediately before and after the series of CLEANEX experiments. WT Erbin

showed negligible precipitation, so all mixing times were acquired in a single

day using a 900 mM sample with eight scans per phase cycle. To compensate

for an increased rate of S47 precipitation, especially at high concentrations,

the different mixing times were acquired over two days (10/40/80 and 20/60/

100 ms, respectively) using identical 400 mM samples with 16 scans per phase

cycle. Similar data was obtained with a 900 mM S47 sample and eight scans

per phase cycle, albeit with considerably more noise.

To control for sample precipitation, a 1D proton spectrum was acquired

before and after each 2D spectrum. Normalization factors for all 1D spectra

were determined by calculating the scaling factor that gave the best least-

squares superimposition of the first 1D spectrum onto all subsequent spectra

over the 9.6–6.6 and 2.6–0.1 ppm ranges. Normalization factors for each 2D

spectrum were determined by averaging the normalization factors of the 1D

spectra immediately before and after. S47 showed a decrease in intensity of

approximately 50% over the course of each day.

Spectra were processed with NMRPipe, and peak volumes were integrated

with CcpNmr Analysis. CLEANEX hydrogen exchange buildup data consisting

ofmixing times (tm), peak volumes (V), and reference peak volumes (V0) were fit

in R (R Development Core Team, 2012) with the equation (Hwang et al., 1998)

V

Vo

=k

R1A;app + k � R1B;app

�e�R1B;apptm � e�ðR1A;app + kÞtm �:

With very low exchange rates (k), the apparent longitudinal/transverse relax-

ation rate (R1A,app) does not significantly affect the shape of the buildup, which

leads to unstable curve fitting. For those amide protons, R1A,app was fixed at

0 s�1. The previously determined value (Hwang et al., 1998) of 0.6 s�1 was

used for the apparent relaxation rate of water (R1B,app).

Preparation of Phosphorylated Protein

Phosphorylated protein was prepared by addition of 5 mM ATP, 1 mM TCEP,

and 10 mM MgCl2. Phosphorylation during the reactions was monitored by

mass spectrometry. Depending on the phosphorylation rate, PKA was added

up to10%of the total proteinmassand incubated at 4�C–16�C.Uponcomplete

phosphorylation, several of the proteins showed two phosphorylation events

(M+160). Analysis of the protein sequences with pkaPS suggested the second

most probable PKA phosphorylation site to be a serine immediately after the

N-terminal 6xHis tag. Cleavage of the tag with HRV-3C protease (courtesy

R.M. Stroud) confirmed this hypothesis, so the 6xHis tags of all unphosphory-

lated and phosphorylated PDZ domains were cleaved prior to binding studies.

The 6xHis tag, His-tagged PKA, and HRV-3C protease were removed using

TALON His-Tag purification resin (Clontech, Madison, WI, USA).

Fluorescence Polarization

Three peptides, TGWETXV, each with a different amino acid at the X position

(W, R, or D) were synthesized with an N-terminal fluorescein label (GenMed

Synthesis, San Antonio, TX, USA; quality control using mass spectrometry

and purity >85% via C18 reverse-phase HPLC). Lyophilized peptide was

resuspended in 5 mM Na2HPO4 (pH 7.4) and the peptide concentration

was determined using the fluorescein extinction coefficient at 490 nm

(76,000 M�1 cm�1) (Mota et al., 1991).

Fluorescence polarization assays were performed by preparing solutions

with log-spaced concentrations of PDZ domain over two orders magnitude.

The assay solutions were then made up with 40% (by volume) of the PDZ dilu-

tions, 0.25%1 mMpeptide, 49.75%binding buffer (20mMDTT, 20mMHEPES

[pH 7.4]) (Lauffer et al., 2010), and 10% 1mg/ml BSA. This resulted in a 2.5 nM

final peptide concentration and 120 mM NaCl. Solutions were prepared on

96-well plates, and three aliquots per dilution were then transferred to

384-well plates for reading on an Analyst AD fluorometer (Molecular Devices,

Sunnyvale, CA, USA). To prevent protein precipitation, mixing was done in

a cold room, and plates kept on ice until immediately prior to reading after

reserved

Structure


equilibration to 25�C. All measurements were done in at least duplicate using

separate batches of expressed protein and separate phosphorylation reac-

tions. Binding measurements for pairs of unphosphorylated and phosphory-

lated proteins taken from same expression batch were done on the same

day, and replicates of pairs were done on different days.

Fluorescence polarization data were first normalized by subtracting the

mean polarization for a blank sample with no PDZ domain. The resulting

changes in polarization (Y) and concentrations (D) were fit in R (R Development

Core Team, 2012) with the following equation

Y =Y0

D

D+Kd

:

The fit parameters included the dissociation constant (Kd) and the maximum

change in polarization (Y0). For a given Kd (A) with standard deviation (sA) and

N replicates, the standard error (SEA) was determined, SEA = sA=ffiffiffiffiN

p. For

a given fold change between two Kd values (A/B) and correlation coefficient

between A and B (rAB), the standard error was calculated using typical prop-

agation of uncertainty,

SEA=B =A

B

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðsA=AÞ2 + ðsB=BÞ2�2rABsAsB=ðABÞ

qffiffiffiffiN

p :

Phosphoserine Modeling

The interaction between the phosphoserine of S43-A47 and the TGWETRV

peptide was modeled using PDB code 1N7T (Skelton et al., 2003), an NMR

structure of wild-type Erbin bound to peptide TGWETWV. PDZ residues 40,

43, 44, 45, and 47, as well as the peptide residue �1 were first changed

to the phosphorylated target sequence of S43-A47 and TGWETRV with

PyRosetta (Chaudhury et al., 2010). Those side chains, along with adjacent

residue R41, were then optimized with the Protein Local Optimization Program

(PLOP) (Groban et al., 2006) version 17.0, using the ‘‘side predict’’ command

with minimum overlap factor of 0.5.

ACCESSION NUMBERS

The BMRB entries for the chemical shift data reported in this paper are 18785

(Erbin WT) and 18786 (Erbin S47).

SUPPLEMENTAL INFORMATION

Supplemental Information includes five figures and PyMOL session scenes

and can be found with this article online at http://dx.doi.org/10.1016/j.str.

2012.10.007.

ACKNOWLEDGMENTS

The authors thank Eli Groban for assistance in setting up kinase assays; Ian

Harwood and the lab of Robert Stroud for generously providing the modified

pET-47 expression vector, HRV-3C protease, and associated supplies; and

Julie Zorn, Justin Rettenmaier, and the lab of Jim Wells for invaluable help

with mass spectrometry analysis. This work was supported by an award

from the National Science Foundation (NSF) to T.K. (NSF CAREER MCB-

0744541) and the Synthetic Biology Engineering Research Center (SynBERC,

NSF-EEC-0540879). C.A. Smith was additionally supported by a DOD NDSEG

fellowship and the NSF GRFP. C.A. Shi, M.K.C., and T.E.B. were supported by

the UCSF summer research training program and SynBERC.

Received: July 18, 2012

Revised: October 13, 2012

Accepted: October 18, 2012

Published online: November 15, 2012

REFERENCES

Adams, J.A., McGlone, M.L., Gibson, R., and Taylor, S.S. (1995).

Phosphorylation modulates catalytic function and regulation in the cAMP-

dependent protein kinase. Biochemistry 34, 2447–2454.

Structure 21

Akiva, E., Friedlander, G., Itzhaki, Z., andMargalit, H. (2012). A dynamic view of

domain-motif interactions. PLoS Comput. Biol. 8, e1002341.

Ambroggio, X.I., and Kuhlman, B. (2006). Computational design of a single

amino acid sequence that can switch between two distinct protein folds.

J. Am. Chem. Soc. 128, 1154–1161.

Balakrishnan, S., and Zondlo, N.J. (2006). Design of a protein kinase-inducible

domain. J. Am. Chem. Soc. 128, 5590–5591.

Birrane, G., Chung, J., and Ladias, J.A. (2003). Novel mode of ligand recogni-

tion by the Erbin PDZ domain. J. Biol. Chem. 278, 1399–1402.

Borg, J.P., Marchetto, S., Le Bivic, A., Ollendorff, V., Jaulin-Bastard, F., Saito,

H., Fournier, E., Adelaıde, J., Margolis, B., and Birnbaum, D. (2000). ERBIN:

a basolateral PDZ protein that interacts with the mammalian ERBB2/HER2

receptor. Nat. Cell Biol. 2, 407–414.

Calderone, T.L., Stevens, R.D., and Oas, T.G. (1996). High-level misincorpora-

tion of lysine for arginine at AGA codons in a fusion protein expressed in

Escherichia coli. J. Mol. Biol. 262, 407–412.

Cerasoli, E., Sharpe, B.K., and Woolfson, D.N. (2005). ZiCo: a peptide

designed to switch folded state upon binding zinc. J. Am. Chem. Soc. 127,

15008–15009.

Chaudhury, S., Lyskov, S., and Gray, J.J. (2010). PyRosetta: a script-based

interface for implementing molecular modeling algorithms using Rosetta.

Bioinformatics 26, 689–691.

DePristo, M.A.,Weinreich, D.M., and Hartl, D.L. (2005). Missensemeanderings

in sequence space: a biophysical view of protein evolution. Nat. Rev. Genet. 6,

678–687.

Dueber, J.E., Wu, G.C., Malmirchegini, G.R., Moon, T.S., Petzold, C.J., Ullal,

A.V., Prather, K.L.J., and Keasling, J.D. (2009). Synthetic protein scaffolds

provide modular control over metabolic flux. Nat. Biotechnol. 27, 753–759.

Ernst, A., Sazinsky, S.L., Hui, S., Currell, B., Dharsee, M., Seshagiri, S., Bader,

G.D., and Sidhu, S.S. (2009). Rapid evolution of functional complexity in

a domain family. Sci. Signal. 2, ra50.

Gardoni, F., Mauceri, D., Fiorentini, C., Bellone, C., Missale, C., Cattabeni, F.,

and Di Luca, M. (2003). CaMKII-dependent phosphorylation regulates SAP97/

NR2A interaction. J. Biol. Chem. 278, 44745–44752.

Good, M.C., Zalatan, J.G., and Lim, W.A. (2011). Scaffold proteins: hubs for

controlling the flow of cellular information. Science 332, 680–686.

Grigoryan, G., Reinke, A.W., and Keating, A.E. (2009). Design of protein-

interaction specificity gives selective bZIP-binding peptides. Nature 458,

859–864.

Groban, E.S., Narayanan, A., and Jacobson, M.P. (2006). Conformational

changes in protein loops and helices induced by post-translational phosphor-

ylation. PLoS Comput. Biol. 2, e32.

Harris, B.Z., Lau, F.W., Fujii, N., Guy, R.K., and Lim, W.A. (2003). Role of elec-

trostatic interactions in PDZ domain ligand recognition. Biochemistry 42,

2797–2805.

Hutti, J.E., Jarrell, E.T., Chang, J.D., Abbott, D.W., Storz, P., Toker, A., Cantley,

L.C., and Turk, B.E. (2004). A rapid method for determining protein kinase

phosphorylation specificity. Nat. Methods 1, 27–29.

Hwang, T.L., van Zijl, P.C., and Mori, S. (1998). Accurate quantitation of water-

amide proton exchange rates using the phase-modulated CLEAN chemical

EXchange (CLEANEX-PM) approach with a Fast-HSQC (FHSQC) detection

scheme. J. Biomol. NMR 11, 221–226.

Iakoucheva, L.M., Radivojac, P., Brown, C.J., O’Connor, T.R., Sikes, J.G.,

Obradovic, Z., and Dunker, A.K. (2004). The importance of intrinsic disorder

for protein phosphorylation. Nucleic Acids Res. 32, 1037–1049.

Ivarsson, Y. (2012). Plasticity of PDZ domains in ligand recognition and

signaling. FEBS Lett. 586, 2638–2647.

Jiang, L., Althoff, E.A., Clemente, F.R., Doyle, L., Rothlisberger, D., Zanghellini,

A., Gallaher, J.L., Betker, J.L., Tanaka, F., Barbas, C.F., 3rd., et al. (2008). De

novo computational design of retro-aldol enzymes. Science 319, 1387–1391.

Kapp, G.T., Liu, S., Stein, A., Wong, D.T., Remenyi, A., Yeh, B.J., Fraser, J.S.,

Taunton, J., Lim, W.A., and Kortemme, T. (2012). Control of protein signaling




Structure


using a computationally designed GTPase/GEF orthogonal pair. Proc. Natl.

Acad. Sci. USA 109, 5277–5282.

Karanicolas, J., Corn, J.E., Chen, I., Joachimiak, L.A., Dym, O., Peck, S.H.,

Albeck, S., Unger, T., Hu, W., Liu, G., et al. (2011). A de novo protein binding

pair by computational design and directed evolution. Mol. Cell 42, 250–260.

Kemp, B.E., Graves, D.J., Benjamini, E., and Krebs, E.G. (1977). Role of

multiple basic residues in determining the substrate specificity of cyclic

AMP-dependent protein kinase. J. Biol. Chem. 252, 4888–4894.

Koder, R.L., Anderson, J.L., Solomon, L.A., Reddy, K.S., Moser, C.C., and

Dutton, P.L. (2009). Design and engineering of an O(2) transport protein.

Nature 458, 305–309.

Kortemme, T., Joachimiak, L.A., Bullock, A.N., Schuler, A.D., Stoddard, B.L.,

and Baker, D. (2004). Computational redesign of protein-protein interaction

specificity. Nat. Struct. Mol. Biol. 11, 371–379.

Lauffer, B.E., Melero, C., Temkin, P., Lei, C., Hong, W., Kortemme, T., and von

Zastrow,M. (2010). SNX27mediates PDZ-directed sorting from endosomes to

the plasma membrane. J. Cell Biol. 190, 565–574.

Laura, R.P., Witt, A.S., Held, H.A., Gerstner, R., Deshayes, K., Koehler, M.F.,

Kosik, K.S., Sidhu, S.S., and Lasky, L.A. (2002). The Erbin PDZ domain binds

with high affinity and specificity to the carboxyl termini of delta-catenin and

ARVCF. J. Biol. Chem. 277, 12906–12914.

Li, B.L., Langer, J.A., Schwartz, B., and Pestka, S. (1989). Creation of phos-

phorylation sites in proteins: construction of a phosphorylatable human inter-

feron alpha. Proc. Natl. Acad. Sci. USA 86, 558–562.

Lin, L., Gillies, S.D., Schlom, J., and Pestka, S. (1999). Construction of phos-

phorylatable chimeric monoclonal antibody CC49with a casein kinase I recog-

nition site. Protein Expr. Purif. 15, 83–91.

Mandell, D.J., Chorny, I., Groban, E.S., Wong, S.E., Levine, E., Rapp, C.S., and

Jacobson, M.P. (2007). Strengths of hydrogen bonds involving phosphory-

lated amino acid side chains. J. Am. Chem. Soc. 129, 820–827.

Masica, D.L., Schrier, S.B., Specht, E.A., and Gray, J.J. (2010). De novo design

of peptide-calcite biomineralization systems. J. Am. Chem. Soc. 132, 12252–

12262.

Mota, M.C., Carvalho, P., Ramalho, J., and Leite, E. (1991).

Spectrophotometric analysis of sodium fluorescein aqueous solutions.

Determination of molar absorption coefficient. Int. Ophthalmol. 15, 321–326.

Murtaugh, M.L., Fanning, S.W., Sharma, T.M., Terry, A.M., and Horn, J.R.

(2011). A combinatorial histidine scanning library approach to engineer highly

pH-dependent protein switches. Protein Sci. 20, 1619–1631.

Narayana, N., Cox, S., Shaltiel, S., Taylor, S.S., and Xuong, N. (1997). Crystal

structure of a polyhistidine-tagged recombinant catalytic subunit of cAMP-

dependent protein kinase complexed with the peptide inhibitor PKI(5-24)

and adenosine. Biochemistry 36, 4438–4448.

Neuberger, G., Schneider, G., and Eisenhaber, F. (2007). pkaPS: prediction of

protein kinase A phosphorylation sites with the simplified kinase-substrate

binding model. Biol. Direct 2, 2.

Palmer, A.E., Giacomello, M., Kortemme, T., Hires, S.A., Lev-Ram, V., Baker,

D., and Tsien, R.Y. (2006). Ca2+ indicators based on computationally rede-

signed calmodulin-peptide pairs. Chem. Biol. 13, 521–530.

Pufall, M.A., Lee, G.M., Nelson, M.L., Kang, H.S., Velyvis, A., Kay, L.E.,

McIntosh, L.P., and Graves, B.J. (2005). Variable control of Ets-1 DNA binding

by multiple phosphates in an unstructured region. Science 309, 142–145.

QIAGEN. (June 2003). The QIAexpressionist, fifth edition (Hilden, Germany:

QIAGEN).

R Development Core Team. (2012). R: A Language and Environment for

Statistical Computing (Vienna, Austria: R Foundation for Statistical

Computing).

Riemen, A.J., and Waters, M.L. (2009). Controlling peptide folding with repul-

sive interactions between phosphorylated amino acids and tryptophan. J. Am.

Chem. Soc. 131, 14081–14087.

Rothlisberger, D., Khersonsky, O., Wollacott, A.M., Jiang, L., DeChancie, J.,

Betker, J., Gallaher, J.L., Althoff, E.A., Zanghellini, A., Dym, O., et al. (2008).


Kemp elimination catalysts by computational enzyme design. Nature 453,

190–195.

Schrodinger, LLC. (2012). The PyMOL Molecular Graphics System,

Version 1.5.

Shifman, J.M., and Mayo, S.L. (2003). Exploring the origins of binding speci-

ficity through the computational redesign of calmodulin. Proc. Natl. Acad.

Sci. USA 100, 13274–13279.

Shoichet, B.K., Baase,W.A., Kuroki, R., andMatthews, B.W. (1995). A relation-

ship between protein stability and protein function. Proc. Natl. Acad. Sci. USA

92, 452–456.

Signarvic, R.S., and DeGrado, W.F. (2003). De novo design of a molecular

switch: phosphorylation-dependent association of designed peptides.

J. Mol. Biol. 334, 1–12.

Skelton, N.J., Koehler, M.F., Zobel, K., Wong, W.L., Yeh, S., Pisabarro, M.T.,

Yin, J.P., Lasky, L.A., and Sidhu, S.S. (2003). Origins of PDZ domain ligand

specificity. Structure determination andmutagenesis of the Erbin PDZ domain.

J. Biol. Chem. 278, 7645–7654.

Smith, C.A., and Kortemme, T. (2008). Backrub-like backbone simulation reca-

pitulates natural protein conformational variability and improves mutant side-

chain prediction. J. Mol. Biol. 380, 742–756.

Steinberg, R.A., Cauthron, R.D., Symcox, M.M., and Shuntoh, H. (1993).

Autoactivation of catalytic (C alpha) subunit of cyclic AMP-dependent protein

kinase by phosphorylation of threonine 197. Mol. Cell. Biol. 13, 2332–2341.

Strickler, S.S., Gribenko, A.V., Gribenko, A.V., Keiffer, T.R., Tomlinson, J.,

Reihle, T., Loladze, V.V., and Makhatadze, G.I. (2006). Protein stability and

surface electrostatics: a charged relationship. Biochemistry 45, 2761–2766.

Szilak, L., Moitra, J., and Vinson, C. (1997). Design of a leucine zipper coiled

coil stabilized 1.4 kcal mol-1 by phosphorylation of a serine in the e position.

Protein Sci. 6, 1273–1283.

Tokuriki, N., Stricher, F., Serrano, L., and Tawfik, D.S. (2008). How protein

stability and new functions trade off. PLoS Comput. Biol. 4, e1000002.

Tonikian, R., Zhang, Y., Sazinsky, S.L., Currell, B., Yeh, J.H., Reva, B., Held,

H.A., Appleton, B.A., Evangelista, M., Wu, Y., et al. (2008). A specificity map

for the PDZ domain family. PLoS Biol. 6, e239.

Treynor, T.P., Vizcarra, C.L., Nedelcu, D., and Mayo, S.L. (2007).

Computationally designed libraries of fluorescent proteins evaluated by pres-

ervation and diversity of function. Proc. Natl. Acad. Sci. USA 104, 48–53.

Ulrich, E.L., Akutsu, H., Doreleijers, J.F., Harano, Y., Ioannidis, Y.E., Lin, J.,

Livny, M., Mading, S., Maziuk, D., Miller, Z., et al. (2008). BioMagResBank.

Nucleic Acids Res. 36(Database issue), D402–D408.

Weinman, E.J., Biswas, R.S., Peng, G., Shen, L., Turner, C.L., E, X., Steplock,

D., Shenolikar, S., and Cunningham, R. (2007). Parathyroid hormone inhibits

renal phosphate transport by phosphorylation of serine 77 of sodium-

hydrogen exchanger regulatory factor-1. J. Clin. Invest. 117, 3412–3420.

Wells, J.A., Powers, D.B., Bott, R.R., Graycar, T.P., and Estell, D.A. (1987).

Designing substrate specificity by protein engineering of electrostatic interac-

tions. Proc. Natl. Acad. Sci. USA 84, 1219–1223.

Wu, W., Kerrigan, J.E., Yadav, P., Schwartz, B., Izotova, L., Lavoie, T.B., and

Pestka, S. (2004). Design and construction of a phosphorylatable chimeric

monoclonal antibody with a highly stable phosphate. Oncol. Res. 14, 541–558.

Yeh, B.J., Rutigliano, R.J., Deb, A., Bar-Sagi, D., and Lim, W.A. (2007).

Rewiring cellular morphology pathways with synthetic guanine nucleotide

exchange factors. Nature 447, 596–600.

Yoshikuni, Y., Ferrin, T.E., and Keasling, J.D. (2006). Designed divergent

evolution of enzyme function. Nature 440, 1078–1082.

You, J., Cohen, R.E., and Pickart, C.M. (1999). Construct for high-level ex-

pression and low misincorporation of lysine for arginine during expression of

pET-encoded eukaryotic proteins in Escherichia coli. Biotechniques 27,

950–954.

Zhang, J., Ma, Y., Taylor, S.S., and Tsien, R.Y. (2001). Genetically encoded

reporters of protein kinase A activity reveal impact of substrate tethering.

Proc. Natl. Acad. Sci. U.S.A 98, 14997–15002.

reserved

design of a phosphorylatable pdz domain with peptide ... · structure article design of a...

Documents