Structural definition of the lysine swing in Arabidopsisthaliana PDX1: Intermediate channeling facilitatingvitamin B6 biosynthesisGraham C. Robinsona, Markus Kaufmanna, Céline Rouxa, and Teresa B. Fitzpatricka,1

aDepartment of Botany and Plant Biology, University of Geneva, 1211 Geneva, Switzerland

Edited by Robert Haselkorn, University of Chicago, Chicago, IL, and approved August 5, 2016 (received for review May 20, 2016)

Vitamin B6 is indispensible for all organisms, notably as the coen-zyme form pyridoxal 5′-phosphate. Plants make the compound denovo using a relatively simple pathway comprising pyridoxine syn-thase (PDX1) and pyridoxine glutaminase (PDX2). PDX1 is remarkablegiven its multifaceted synthetic ability to carry out isomerization,imine formation, ammonia addition, aldol-type condensation, cycli-zation, and aromatization, all in the absence of coenzymes or recruit-ment of specialized domains. Two active sites (P1 and P2) facilitatethe plethora of reactions, but it is not known how the two are co-ordinated and, moreover, if intermediates are tunneled betweenactive sites. Here we present X-ray structures of PDX1.3 from Arabi-dopsis thaliana, the overall architecture of which is a dodecamer of(β/α)8 barrels, similar to the majority of its homologs. An apoenzymestructure revealed that features around the P1 active site in PDX1.3have adopted inward conformations consistent with a catalyticallyprimed state and delineated a substrate accessible cavity above thisactive site, not noted in other reported structures. Comparison withthe structure of PDX1.3 with an intermediate along the catalytictrajectory demonstrated that a lysine residue swings from the dis-tinct P2 site to the P1 site at this stage of catalysis and is held in placeby a molecular catch and pin, positioning it for transfer of servicedsubstrate back to P2. The study shows that a simple lysine swingingarm coordinates use of chemically disparate sites, dispensing withthe need for additional factors, and provides an elegant exampleof solving complex chemistry to generate an essential metabolite.

Arabidopsis thaliana | vitamin B6 biosynthesis | crystal structure |lysine swing

Vitamin B6 in its form as pyridoxal 5′-phosphate (PLP) is anessential coenzyme involved in over 140 biochemical reactions,

more than any other known nutrient. It is the most versatile coen-zyme known in nature being involved in transamination, de-carboxylation, elimination, racemization, and replacement reactions(1). It is biosynthesized de novo by microorganisms and plants. As anessential micronutrient, it must be taken in the diet of animals in-cluding humans. The absence of the pathway in animals renders thebiosynthesis de novo pathway a potential drug target in pathogenicorganisms (2). It is now established that there are two biochemicalroutes that lead to the biosynthesis de novo of PLP. The first to beresolved was the deoxyxylulose 5-phosphate (DXP)-dependent path-way, which requires seven enzymes and is found in a few selectmicroorganisms including Escherichia coli (reviewed in ref. 3). Thesecond route does not involve DXP (DXP-independent), only re-quires two enzymes, and is by far the predominant route being foundin most microorganisms and all plants (reviewed in ref. 3). The twoenzymes required for the DXP-independent route are pyridoxinesynthase (PDX1) and pyridoxine glutaminase (PDX2), which to-gether form a glutamine amidotransferase complex that uses ribose5-phosphate (R5P), glyceraldehyde 3-phosphate (G3P), and gluta-mine in a highly complicated sequence of reactions to produce PLP(4, 5) (Fig. 1). Although PDX2 is a classic glutaminase hydrolyzingglutamine to ammonia and glutamate (4–6), the remarkable reactionmechanism of PDX1 that involves pentose and triose isomerization,imine formation, ammonia addition, aldol-type condensation, cycli-

zation, and aromatization to produce the PLP molecule has receivedconsiderable attention but remains enigmatic (7–10). Both bio-chemical and structural studies have served to capture snapshots ofthe PDX1 enzyme in action at certain stages along the catalytictrajectory (10–18). Nonetheless, the sequence of events cannot yetbe stitched together and is essential to provide a complete overviewof one of nature’s most complicated enzymes from a fundamentalperspective, as well as its potential as a drug target.Multifunctional enzymes generally recruit distinct domains for

their reactions, or connect different sites by substrate channeling(19, 20), or remodel a single active site (21) to facilitate themultitude of transformations taking place. Substrate channelingcan be mediated by tunneling (22) or by transfer via a covalentlybound prosthetic group to the successive active site. Classic ex-amples of the latter are lipoyl-lysine residues of 2-oxo acid dehy-drogenases (23), biotinyl-lysine residues of carboxylases (24), andphosphopantetheinyl-serine residues of fatty acid synthases (25,26). Notably, the referred to prosthetic groups are all vitamin-derived coenzymes. The covalently attached prosthetic groupsgenerally have freedom to rotate within their respective activesites and thus serve as swinging arms ferrying substrates or inter-mediates to the subsequent active site (27). Enhancement ofcatalytic efficiency, channeling, and protection of stable interme-diates are hypothesized reasons for such approaches (28). None-theless, the concept has remained open to questions, whichinclude whether the decorated side chain swings or the wholeprotein domain, as well as the need for the posttranslational

Significance

Multifunctional enzymes have been shown to recruit distinct do-mains for their reactions, remodel active sites, or connect differentsites by substrate channeling to facilitate the multitude of trans-formations taking place. Pyridoxine synthase (PDX1) of the vitaminB6 biosynthesis machinery is a remarkable enzyme that alone hasa polymorphic catalytic ability designated to two active sites, thecoordination of which is unclear. Here structural snapshots allowus to describe a lysine swinging arm mechanism that facilitatesserviced substrate transfer and demonstrates how an enzyme cancouple distinct chemistry between active sites, dispensing withthe need for extra domains, substrate tunneling, or transfer ofcoenzyme bound intermediates. The work provides an elegantexample of simplicity at work in nature’s sea of complexity.

Author contributions: G.C.R. and T.B.F. designed research; G.C.R. and M.K. performed re-search; G.C.R., M.K., and C.R. contributed new reagents/analytic tools; G.C.R. and T.B.F.analyzed data; and G.C.R. and T.B.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: Crystallography, atomic coordinates, and structure factors have beendeposited in the RCSB Protein Data Bank [accession nos. 5K3V (apo-PDX1.3) and 5K2Z(PDX1.3-adduct)].1To whom correspondence should be addressed. Email: theresa.fitzpatrick@unige.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1608125113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1608125113 PNAS | Published online September 19, 2016 | E5821–E5829

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

modification to include the prosthetic group facilitating themultistep chemical reaction (28). The complex chemistry carriedout by PDX1 provides an intriguing example to investigate inthe context of multifunctional enzymes given that no prostheticgroup is required and the enzyme alone carries out the entireseries of reactions.All PDX1s are made up of two active sites referred to as P1 and

P2 (16, 17, 29) and together coordinate the sequence of reactionsresulting in PLP production (Fig. 1). The process begins with thebinding of R5P in the P1 site in a covalent interaction with an es-sential lysine residue [K97 in Arabidopsis thaliana (hereafter referredto as Arabidopsis) PDX1.3] (10, 17). Loss of phosphate and waterfrom R5P as well as the incorporation of ammonia (from glutaminein the presence of PDX2) result in a chromophoric intermediatelikely to be unique to PLP synthase, which has been observedspectroscopically (8–10) but has eluded precise structural charac-terization. The reaction can only proceed further in the presence ofG3P, which is assumed to take place in the P2 site and is where thefinal product PLP has been observed in X-ray crystal structures (10,12, 18). A pertinent question is how the intermediate is transferredfrom the P1 to the P2 site. A second lysine (K165 in ArabidopsisPDX1.3) is postulated to orchestrate the interaction between theactive sites (10, 17) but had only been observed pointing toward theP2 site until recently. The recent report captured PDX1 in a ratherheterogeneous population of different catalytic states in which thissecond lysine points into either the P1 or P2 site (15); however, themixture of states in each subunit precluded clarification of thetransition between the two states. Precise definition of the swingingof this essential lysine between the active sites would help to resolvethe obscure steps between the initiation of the PLP synthase reac-tion in P1 and its completion in P2.Here we have determined X-ray crystal structures of a PDX1

(PDX1.3) from Arabidopsis. Like its bacterial counterparts, theplant protein displays dodecameric architecture. In contrast toprevious structures, the Arabidopsis apoenzyme is in a closedconformation but poised for catalysis and serves to delineate acavity above the P1 active site that provides access to the R5Psubstrate. In the apoenzyme, the second active site lysine is ori-ented toward the P2 site defining the resting state, i.e., its positionat the beginning and end of the catalytic cycle. The structure ofArabidopsis PDX1.3 with the chromophoric intermediate solved to1.8 Å was also determined and can be used to corroborate aproposed structure for this adduct. Moreover, this structure clearlyplaced the second active site lysine in the P1 site at this stage of

catalysis. A molecular catch and pin mechanism explains the re-straint of this lysine in P1, release of which presumably facilitatesthe swing back to P2 and completion of PLP biosynthesis. Overall,these snapshots of the plant PDX1 provide key insights into theintricate workings of a remarkable enzyme essential for plantsurvival and more generally into multifunctional enzymes.

Results and DiscussionPDX1.3 from Arabidopsis Adopts a Dodecameric Architecture. Thestructure of apo-PDX1.3 from Arabidopsis has been determined at1.9 Å resolution (Rwork, 17.7%; Rfree, 20.3%; Table S1). It com-prises residues 20–296 with 13 residues not visible at the C ter-minus in addition to the 19 missing from the N terminus. TwelvePDX1.3 subunits that fold as (β/α)8 barrels form a double hex-americ ring interdigitating into the dodecameric structure(Fig. 2A). The dodecameric architecture is consistent with earlierestimations of the molecular mass of PDX1.3 (411,000 Da) by sizeexclusion chromatography coupled to static light scattering (30).The individual hexameric rings are 100 Å in diameter with acentral pore 40 Å in diameter (Fig. 2A). The two hexamers stacktogether to form a structure 90 Å high (Fig. 2A). The overall ar-chitecture is similar to homologs of PDX1.3 that have been crys-tallized from bacterial sources including Bacillus subtilis (2NV1and 2NV2) (16), Geobacillus stearothermophilus (1ZNN, 4WXY,4WXZ, and 4WY0) (15, 29), Thermotoga maritima (2ISS) (17),and Mycobacterium tuberculosis Rv2606c (4JDY) (31), as well asthose from apicomplexan Plasmodium species (4ADS, 4ADT, and4ADU) (11), all of which form dodecamers. Notably, structures ofthe homologous enzymes from the yeast Saccharomyces cerevisiae(3FEM, 3O05, 3O06, and 3O07) (13, 18) and the archaeonPyrococcus horikoshii (4FIQ and 4FIR) (32) crystallize as hex-amers. Labeling of individual α-helices and β-strands of PDX1.3was done according to the labeling introduced for the B. subtilishomolog (Fig. 2B) (16). Thus, contact between each adjacent(β/α)8 barrel in PDX1.3 is established by helix α8′′, which runsparallel to helix α8 near the C terminus (Fig. 2 A and B), as forall homologs. As for other dodecameric PDX1s, the interdigi-tation of the hexameric rings in PDX1.3 is established by helicesα6, α6′, and α6′′ of each subunit (Fig. 2 A and B). Notably, helixαN is present despite the absence of the glutaminase subunit.Previously, this helix has only been observed in complexes ofPDX1 with PDX2, with the exception of PDX1 from S. cer-evisiae (13, 16, 17).

The P1 Active Site of apo-PDX1.3 Displays a Catalytically PoisedArchitecture. Two active sites, annotated P1 and P2, have beendefinitively located in PDX1 enzymes (16, 17). These sites wereoriginally defined by bound chloride, phosphate, or sulfate ions inX-ray crystallographic structures (16, 17, 29) and more recently bythe bound substrate R5P (P1) and product PLP (P2) (11, 18, 32)and have been validated through several biochemical studies (10,12, 16). In PDX1.3, the P1 site is located at the C-terminal mouthof the β-barrel and contains density for a sulfate ion surrounded bythe characteristic phosphate-binding loop that includes Gly–Thr–Gly, as well as other established conserved residues of this activesite (16, 29) (Fig. 2B). The P2 site is located at the interface be-tween subunits of opposing hexamers and also contains density fora bound sulfate ion surrounded by the conserved residues char-acteristic of this region (Fig. 2B). As for the homologous coun-terparts of PDX1.3, the bound sulfates in P1 and P2 occupy thesame position as the phosphates of bound R5P and PLP, re-spectively (11, 15, 17, 18, 32). Markedly, except for coordinatedwaters, there is no other density observed in the P1 or P2 sites,such that this structural snapshot of the PDX1.3 enzyme can beconsidered substrate and product free, i.e., apo-PDX1.3.Although the overall architecture of PDX1.3 from Arabidopsis

is similar to that of its homologs (rmsd of 1.14 Å, 0.77 Å, and1.35 Å comparing 250, 262, and 242 corresponding Cα atoms of

OPO32-

OH

OHO

O NH2O

O

OH

O

OH

R5P

Pi + H2ONH3

K97

K165

K165

G3P

PLP

2H2O

P1 P2NH

HN

NH2

OPO32-

OPO32-

N

K165

NH

OPO32- O

OH

OH

OH

H2O

H2N

K97HN

Fig. 1. Scheme depicting selected reactions of PDX1 in the P1 and P2 activesites. Ribose 5-phosphate (R5P) enters the P1 active site of PDX1 and forms acovalent intermediate with K97 (Arabidopsis numbering). Loss of water andinorganic phosphate (Pi) from R5P and incorporation of ammonia (NH3)leads to formation of a covalent chromophoric adduct. The presence of K165in the P1 site at this stage of catalysis facilitates transfer of the chromophoricadduct to the P2 active site. Condensation with glyceraldehyde 3-phosphate(G3P) to form the product pyridoxal 5′-phosphate (PLP) completes one cat-alytic cycle. The interactions of the e-amino groups of the respective lysineswith either R5P or the derived chromophoric adduct are depicted in gray.

E5822 | www.pnas.org/cgi/doi/10.1073/pnas.1608125113 Robinson et al.

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

B. subtilis,G. stearothermophilus, and S. cerevisiae PDX1, respectively),we observed that the small segment of each subunit insertedbetween β1 and α2 within the lumen of each hexameric ring,comprising residues 65–71, adopts an ordered conformation, α2′,in PDX1.3 from Arabidopsis (Fig. 3A). In the reported bacterialPDX1 structures, the α2′ fold has only been observed in struc-tures of the ternary complex with the glutaminase PDX2 andwas, moreover, suggested to result from priming by a signal fromthe glutaminase subunit (15–17). However, α2′ has since beenobserved in the reported structure of PDX1 alone from thearchaeon P. horikoshii (4FIQ and 4FIR) (32), as well as in theeukaryotic PDX1 structures reported from S. cerevisiae (3FEM,3O05, 3O06, and 3O07) (13, 18) and Plasmodium berghei (4ADTand 4ADU) (11), all of which were crystallized in the absence ofPDX2. Furthermore, in the archaeal and eukaryotic structures,α2′ exists in either of two conformations (out or in), dependenton the absence or presence of R5P, respectively (11, 18, 32). Inthe presence of the substrate, the inward movement of α2′ par-tially covers the solvent accessible P1 active site. Here a com-parison with the B. subtilis PDX1/PDX2 complex (in which thesubstrate is absent) and substrate bound P. berghei PDX1revealed that α2′ of PDX1.3 is in the inward conformation de-spite the absence of substrate (Fig. 3B). In an analogous fashion,another short helix, α8′, adopts an outward and inward positionover the P1 site dependent on the absence or presence of sub-strate, respectively (11, 18, 32). Here we observe that this helix isin the inward conformation in PDX1.3 (Fig. 3 A and C), despitethe absence of substrate. The dipole of this short helix contrib-utes to coordination of bound sulfate in PDX1.3 (equivalent tothe phosphate of bound substrate). Additionally, a serine residue

located at the N terminus of this helix in PDX1.3, S252, forms ahydrogen bond with the amide of A66, located at the N terminusof α2′, as well as interacting with the dipole moment of α2′(Fig. 3C).In contrast to the other reported PDX1 structures, we also

observed that most of the C-terminal region (residues 286–296 outof 308) was resolved in PDX1.3 (Fig. 3A), even though the sub-strate is absent. The C-terminal region following on from α8′′extends from the outer surface of the subunit in the dodecamer tothe lumen across a cleft formed between adjacent subunits in thesame hexamer (Fig. 3A). Remarkably, the peptide backbone ofthis region is zippered into place by intrasubunit and intersubunitinteractions across this cleft (Fig. 3D). On one side, a series ofhydrogen bonds comprising the guanidinium group of R263 andamide groups of G253, G258, and T170 within the same subunitcoordinate the peptide backbone of the C-terminal region directlyor through coordination of waters. This is matched on the otherside by interaction with the side chains of R76 and D79 and theamide of V74 located in the loop connecting α2′ and α2 of theadjacent subunit, in addition to a network of structural waters(Fig. 3D).Taken together, the inward position of α2′ as well as α8′ and the

coordination of the C-terminal region define three importantstructural arrangements that have only been collectively seenpreviously in PDX1 structures in the presence of the R5P sub-strate. As these arrangements serve to limit access to the P1 site,they are assumed to represent a catalytically operating state.However, we have observed this arrangement in the absence ofsubstrate and must therefore conclude that this rather represents acatalytically poised state, at least in the case of PDX1.3. Moreover,the network of intrasubunit and intersubunit interactions zipper-ing the C-terminal region in place would appear to be poised tocoordinate cross-talk between two neighboring subunits. We havepreviously demonstrated that PDX1 displays high cooperativity inrelation to the binding of the R5P substrate for which the C ter-minus is essential (12, 14). Therefore, this structure of PDX1.3serves to capture the fundamental essence of the C terminus inlinking neighboring catalytically poised subunits.

A Catalytically Operational PDX1.3 Captures a Lysine Swinging ArmMechanism.To further test if the geometric arrangements observedin PDX1.3 represent a catalytically poised state rather than a stateclosed to substrate (possibly caused by bound sulfate mimickingsubstrate binding), we performed crystal-soaking experiments withR5P. Indeed, it must be mentioned that the P1 site is predomi-nantly hydrophobic, the majority of residue interactions only beingwith the phosphate moiety of the substrate. Because an am-monium salt is present in the crystallization solution, we an-ticipated that the enzyme would catalytically process R5Punder these conditions. Specifically, we have shown previouslythat PDX1 can transform covalently bound R5P into a chro-mophoric intermediate with an absorption maximum around315 nm that results from loss of water and phosphate in thepresence of a source of ammonia (10) (Fig. 1). We verified thatthis intermediate forms under the crystallization conditionsused here by performing the reaction in solution under cry-oprotection conditions (Fig. 4A). As expected, no absorbancecorresponding to the formation of the product PLP (414 nm)was observed, due to the absence of the second substrate G3P.The structure of PDX1.3 (PDX1.3-adduct) under these condi-tions was solved to 1.8 Å (Rwork, 18.0%; Rfree, 20.4%; Table S1)and contains the same resolvable areas as the apoenzyme, withthe addition of N297 that is resolved in this structure. Theglobal architecture of the PDX1.3-adduct obtained remainedvery similar to that observed in the absence of added substrate(rmsd 0.311 Å comparing 271 Cα atoms over the two struc-tures) (Fig. 4B). Nonetheless, key differences were observed inthe P1 site in particular. First, the essential catalytic aspartate

A

B

40 Å

100 Å

90 Å

90º

α1

β1

α8'

α7

α6''

α6'

α6α5

α4

α3

α2α8

α8''β3

β2

β4

β6β7β5

β8 αN

α2'

β1

β3

β2

β4

β6β7β5

β8

α5

β6

α6'

G169

T170

G171

D118

M59

M161

K97D40

F249

H131

K203R153

R154K165

P1 site

P2 site

P-loop

Fig. 2. Overall architecture of PDX1.3 from Arabidopsis. (A) PDX1.3 forms adodecamer composed of two interdigitated hexamers (green and light blue),shown from above and the side after rotation by 90°. Dimensions are as in-dicated. (B) The individual subunits of PDX1.3 adopt a (β/α)8 fold with addi-tional structural elements labeled as indicated. The P1 and P2 catalytic sites canbe clearly defined with residues characteristic of each site and sulfate ions(yellow and red sticks) shown. Note one residue in P2, lysine 203 (K203), iscontributed by α6′ of a subunit on the opposing hexamer (light blue). Watershave been omitted for clarity. The electron density is contoured to 1.0σ.

Robinson et al. PNAS | Published online September 19, 2016 | E5823

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

residue (D40 in PDX1.3) in P1 (10) that resides on the β1-strand and is oriented toward the center of the P1 site is shifted1.0 Å deeper into the active site (and toward the adduct mol-ecule; see below) compared with its position in the apo-PDX1.3structure (Fig. 4C). This advanced position of D40 in thepresence of substrate represents a conformation either poised toperform or performing catalytic attack, consistent with its pre-dicted role in shuffling protons during the formation of thechromophoric adduct (10). In the related structure of PDXSfrom G. stearothermophilus (Chain F; 4WY0), the catalytic as-partate does not adopt such an advanced position. Second, adramatic reorientation is observed with the second catalytic ly-sine residue, K165, which pointed toward the P2 site in theapostructure but has rotated through 120° to position itself in theP1 site in the adduct structure (Fig. 4C). The precise role andmechanism of this lysine residue has been a matter of debate (4,10, 17). Nevertheless, it is hypothesized that this residue couplesthe P1 and P2 active sites by a swinging arm mechanism pivotingbetween the two sites. However, with the exception of the mostrecently described structure of the G. stearothermophilus en-zyme (15), it has been exclusively observed oriented toward theP2 site, regardless of the presence or absence of substrate(R5P) or product (PLP) in both prokaryotic and eukaryoticstructures (11, 15, 17, 18, 32). Thus, orientation of this lysineresidue toward the P2 site presumably represents the restingstate. Although the rotated position of the equivalent catalyticlysine has been observed recently in the G. stearothermophilusenzyme (15), the heterogeneous nature of the P1 site with re-spect to the ligand obscured its function. We observed consis-tent additional density indicating a covalently bound molecule tothe catalytically essential K97 residue in all P1 sites in thePDX1.3-adduct structure. The chemical structure of the chro-mophoric intermediate proposed by Hanes et al. (8) from NMRstudies can be modeled into this density. The initial imineformed between K97 and R5P occurs at C1 of the latter(Fig. 1), but it has been proposed that this migrates to C5 duringthe course of the formation of the chromophore (8). However,the adduct observed here is poorly described by the electrondensity when oriented so as to bond to K97 via C5 (Fig. 4D). Inaddition, the mean Bfactor for the adduct in this orientation ishigh (76.0 Å2). By contrast, when oriented to bind K97 viaC1, the chromophore is described more clearly, albeit notcompletely, by the electron density (Fig. 4D) and has a reducedBfactor (58.8 Å2). A model of the same intermediate in astructure of the G. stearothermophilus enzyme proposed that theadduct is bonded via C5 [although the proposed chromo-phoric intermediate was observed in only one subunit of thedodecamer (15)]. Significantly, the model of the proposed chromo-phore in the electron density present in the P1 site fits very well to aspecies covalently bound to both K97 and K165 (Fig. 4D). It istempting to suggest that this represents a transfer intermediate andthus captures the mechanism by which this reaction intermediate istransferred from the P1 to the P2 site; that is, K165 binds the adductvia C5 in preparation for transfer to the P2 site. This would notnecessitate the C1 to C5 migration proposed by Hanes et al. (8),which was deduced from highly processed enzyme that had beentreated with acid and denatured with urea. Although this species hasnot been observed in mass spectrometry (MS) data (10, 15), it shouldbe noted that such a modification would not affect the subunit intactmass as determined by electrospray ionization–MS but may not beamenable to tryptic digestion and/or analysis by matrix assisted laserdesorption/ionization–time of flight–MS. Notably, the electron den-sity for the observed adduct is distinct from that of the anion boundto the phosphate binding loop (Fig. 4D). This observation validatedthe processing of R5P and demonstrated the catalytic competenceof PDX1.3 in the crystal. Additional density adjoining the adduct,but not describing the chromophore itself, should be noted. Itspresence may indicate multiple isomeric species, possibly as a result

A

D

B C

C-terminalregion

α8'

α8''

G258

R263

G253

T170

R76

D79

V74

β1

α8'

α2'

S252 A66

3.1Å

α2'

α2

α2'

α2'

α8''

α8''

α8'α8'

α2'

α2'

C-terminalregion

β2

β3

Fig. 3. Arabidopsis apo-PDX1.3 adopts a catalytically poised structure. (A)Neighboring subunits on the same hexameric ring. In apo-PDX1.3, struc-tural elements α2′ and α8′ (blue) are located at the top of the P1 site, andthe C-terminal region (blue) extends across the subunit interface. (B) Twoconformations can be distinguished for the α2′ helix. An outward con-formation is seen in the ternary PDX1/PDX2 complex from B. subtilis(2NV2; dark gray), which is in the absence of the R5P substrate. By contrast,an inward conformation is seen in the structure of P. berghei solved in thepresence of R5P (4ADU; light gray). The α2′ helix in Arabidopsis PDX1.3adopts an inward conformation despite the absence of substrate (green).(C ) In an analogous fashion, the α8′ helix can adopt either an outwardconformation, as seen in the absence of R5P in the ternary complex ofPDX1/PDX2 complex from B. subtilis (2NV2; dark gray), or an inward con-formation, as seen in the P. berghei PDX1 in the presence of R5P (4ADU;light gray). Arabidopsis PDX1.3 adopts the inward conformation despitethe absence of R5P and interacts via S252 with the N-terminal end of theα2′ helix (green) through its dipole and A66. (D) The C-terminal region(green) occupies a position in the lumen at the interface with a neigh-boring subunit (light green). It coordinates interactions with both subunitszippering it in place. On one side, a series of hydrogen bonds comprisingthe guanidinium group of R263 and amide groups of G253, G258, andT170 within the same subunit coordinate the peptide backbone of theC-terminal region directly or through coordination of waters (cyan). This ismatched on the other side, by interaction with the side chains of R76, D79,and the amide of V74 located in the loop connecting α2′ and α2 of theadjacent subunit, in addition to a network of structural waters.

E5824 | www.pnas.org/cgi/doi/10.1073/pnas.1608125113 Robinson et al.

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

of sample processing. A similar effect may account for the presenceof the chromophore in just 1 of the 12 subunits in a structure ofG. stearothermophilus PDX1 (15). Taken together, the comparisonof the apo-PDX1.3 and the PDX1.3-adduct structures presentedhere clearly defined a consistent P2 site orientation for K165 in theabsence of substrate and P1 site orientation in the presence of thechromophoric adduct. Therefore, we have captured the two keysnapshots of the swinging arm mechanism of this lysine residueoriginally postulated by Ealick and coworkers (17).

Mechanistic Insight into the Lysine Swinging Arm of PDX1.3. Furtherexamination of the apo- and PDX1.3-adduct structures allowed usto provide mechanistic insight into the movement of K165 from theP2 to the P1 site (Fig. 5). When oriented toward P2 (resting state),the amide group of K165 hydrogen bonds with the peptide back-bone of C145 (Fig. 5A). In this orientation, the hydroxyl group ofthe T164 residue neighboring K165 hydrogen bonds with the car-bonyl group of A228 on β7, and its peptide backbone interacts withR163 (on β6; Fig. 5A). In the PDX1.3-adduct structure, the amidenitrogen of K165 retains its interaction with C145, but the e-aminogroup is reoriented 120° from P2 to interact with the chromophoricadduct in P1, allowing the carbonyl of K165 to hydrogen bond withH179 (Fig. 5B). In parallel, the hydroxyl group of T164 hydrogenbonds with Q225, which has adopted a different rotamer positionto satisfy this interaction, and the amide carbonyl has rotated toform a hydrogen bond with A228 in the neighboring β-strand (β7).It should be noted that Q225 retains two hydrogen bond interac-tions in both structures (Fig. 5 A and B), and overall conservationof hydrogen bonds following formation of the chromophoric in-termediate indicates that the local rearrangement of this region isstable. Presumably, these conformational changes are driven by the

sequence of reactions that process covalently bound R5P to thechromophoric adduct because in the presence of unprocessed R5Pthis lysine residue points into P2.A notable barrier to the swing required by K165 to adopt the

P1 site orientation is a salt bridge formed between E121 andR163 (Fig. 5C). The proximity and pseudoplanar nature of thissalt bridge indicate that it is a strong interaction, and it is unlikelythat passage of K165 to the P1 site would break it. The corre-sponding residue in B. subtilis PDX1 is a glutamate also (E105),and its mutation to an aspartate increased the specificity constantfor formation of the chromophoric adduct approximately twofold(12). It was thus proposed that this salt bridge acts as a steric gatein coordination of P1 and P2 site activities (12). It is possible thatthe smaller aspartate residue would move the impeding salt bridgeaway from the P2 site-oriented lysine, increasing maneuverabilityof this catalytic residue. It therefore seems likely instead that thetrajectory of the K165 swing is through the opening that is linedwith P65, V122, and A168 and moreover may serve as a hydro-phobic gate between the active sites (Fig. 5C, Left and Right).Although the hydrophobic nature of this gate would seem to act asan obstacle to the swing from the P2 site, it may also serve aratchet-like function, preventing return of K165 to the P2 site untilits action is complete and further energy is provided—possiblythrough binding of the second substrate G3P.Remarkably, despite substantial reorientation of K165 and

T164 between the apo-PDX1.3 and PDX1.3-adduct structures,the conformational changes in this region are primarily restrictedto these two residues. This stability is due, in part, to theneighboring glycine residue, G166 (Fig. 5 A and B), which is ableto tolerate large deviations in backbone torsion angles (33). Acomparison with the recent G. stearothermophilus structures (15),

A

D

B

C

0.005

0.025

0.045

0.065

0.085

250 300 350 400 450 500

Abso

rban

ce (A

U)

Wavelength (nm)

0 min5 min15 min30 min60 min90 min

310 nm

414 nm

P1 site

P2 site

K97

D40

K165

D40

K97

K165

D40

K97 K165

T164

β1

β3

β6

1Å

120º

54º

Fig. 4. Capture of the swinging arm lysine in the structure of PDX1.3 containing a chromophoric intermediate. (A) Formation of the chromophoric intermediateby PDX1.3 in the presence of R5P in solution under cryoprotection conditions. A typical increase in absorbance at 310 nm over time in the presence of substrateand an ammonium source (ammonium sulfate from the crystallization solution) is observed. Difference spectra of PDX1.3 (74 μM) in the absence and presence ofR5P (6 mM) are shown. (B) The overall architecture of the apo-PDX1.3 (green) and PDX1.3-adduct (beige) structures are very similar. The catalytically importantlysine (K165) that pivots between the P1 and P2 sites upon comparison of both structures is indicated. (C) A close-up view of the comparison of the P1 and P2active sites in apo-PDX1.3 (green) and PDX1.3-adduct (beige) to indicate the inward movement of the catalytically important residue, D40, by 1 Å, as well as thepivoting through 120° of K165. Note the reorientation of the neighboring threonine residue (T164) by 54°. (D) Fo–Fc omit map (2.5 σ) of the proposed structure ofthe chromophoric intermediate (yellow), K97 (beige), K165 (beige), and bound anion (gray). Left shows that of the proposed chromophore covalently bound toK97 via C5, whereas Right shows the chromophore bound to K97 via C1. The electron density matches well with a transfer intermediate, where C5 of thechromophore is bound to K165. Note the density for the bound anion is distinct from that of the covalently bound species.

Robinson et al. PNAS | Published online September 19, 2016 | E5825

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

where the active site lysine is also seen in both orientations, indicatesthat there is much greater structural rigidity in the Arabidopsisenzyme than in its bacterial counterpart. This rigidity arisesfrom three key differences: First, in the structure of theG. stearothermophilus protein with the equivalent active site ly-sine (K149) pointing into the P2 site, a proline residue (P152)adopts a position oriented in toward the P1 site (Fig. 6A). Fol-lowing the swing of K149 from the P2 site to the P1 site, theunfavorable interaction that would result from positioning thee-amino group of this residue 2.7Å away from the proline iscircumvented by reorientation of P152 away from the lysine (Fig.6A). The reduced torsional flexibility of a proline residue meansthe polypeptide backbone moves to accommodate its shiftedorientation (Fig. 6A). In Arabidopsis PDX1.3, an alanine (A168)occupies the equivalent position of P152 in G. stearothermophilusPDX1, and therefore, its smaller side chain does not provide thesame obstruction. Consequently, repositioning of this loop in theArabidopsis PDX1.3 structure is not required (Fig. 6A). Second,the loop region connecting the β6 strand and the α6 helix inPDX1.3 is stabilized by interactions with the α2′ helix (Fig. 6B).The first turn of α2′ contains a proline (P65) and an alanine (A66),which form a hydrophobic pocket with residues A168 and G169,located at the top of the loop. The second turn of α2′ contains anarginine (R69), the guanidinium group of which forms hydrogen

bonds with the peptide backbone of A168, stabilizing the loop.Therefore, the α2′ helix acts as a catch, restricting movement inthe loop connecting the β6 strand and the α6 helix. The α2′ helixis disordered in the matching structure of the G. stearothermophilusenzyme, and so movement of the corresponding loop is notrestricted in this species. Third, in PDX1.3 a glutamate (E167)located in the loop forms hydrogen bonds with T170 and coor-dinates a water molecule with the peptide backbone of a regionnear the C terminus of the protein (Fig. 6C). This glutamate pinprovides additional stabilization to the loop connecting the β6strand and the α6 helix. The C-terminal region of the matchingG. stearothermophilus structure is not resolved; presumably, it isdisordered and unable to form hydrogen bond interactions.Consequently, there is no hindrance to movement of the corre-sponding glutamate: it adopts a rotated position and provides nostabilization of the loop (Fig. 6C). The structural rigidity con-ferred on the Arabidopsis PDX1.3 protein in the absence ofsubstrate and in the presence of the adduct is further emphasizedby the Bfactor values, which indicate a decrease in thermal motionin the adduct structure (Fig. 6D and Table S2). All of thiscombined serves to illustrate not only the pathway of reor-ientation of K165 from P2 to P1 but also the stabilized assemblyof key structural elements required from a catalytic poised to anoperational PDX1.3.

Substrate Access to the P1 Site. An open question is how thesubstrate R5P entered the catalytically primed state of the plantenzyme because the defined geometric rearrangements havepreviously been postulated to represent a closed state to sub-strate (11, 15). Specifically, in previous structures of PDX1, moreopen conformations are defined in the absence of substrate, inwhich α2′ is in the outward conformation or disordered and theC-terminal region is disordered also, and thus, the P1 site isaccessible from the lumen of the dodecamer via a water filledchannel (11, 15–18, 29, 32). However, in the Arabidopsis apo-PDX1.3 structure, the inward conformation of α2′ and α8′ andthe ordered conformation of the C-terminal region block accessto the P1 site via this route (Fig. 7A). We note here that the apo-PDX1.3 has a second, water-filled cavity, located at the surfaceof the enzyme above the P1 site that extends down to K97 (Fig.7 A and B). It is plausible that R5P can enter via this cavity, andmoreover, in forming the Schiff base with K97, the alkyl sidechain of this residue would block the end of the cavity (Fig. 7B).Furthermore, a comparison of the apo-PDX1.3 structure withthat of the PDX1.3-adduct reveals that β1, α1, and the inter-connecting loop have dropped down toward the P1 site (Fig. 7C).As well as moving D40 into position for catalytic attack, this shifthas the effect of closing the described cavity. Valine 42, located atthe apex of the β1/α1 loop, is moved into proximity to the peptidecarbonyls of L61 and E62, located on the loop connecting β2and α2′. To compensate for this unfavorable interaction, thissecond loop is turned away, and the side chain of E62 twists tocover the mouth of the cavity, where it forms a salt bridge withthe guanidinium group of R63 and hydrogen bonds with theamide nitrogen of R76 and a water, coordinated by polar andcharged groups, located at the entrance of the proposed cavity(Fig. 7 C and D). It is possible that this gating mechanismprevents substrate access to the P1 site before completion ofa round of PLP biosynthesis. Significantly, a similar cavity isobserved in all other structures of PDX1, which althoughslightly variable in the surface starting point all run down to theactive site lysine (Fig. 7E). Therefore, this cavity is likely toserve as an access point for R5P, particularly in organisms thatuse this catalytically poised configuration of the enzyme. Fur-thermore, the access cavity itself mirrors the channels inaquaglyceroporins (34) and lactose permease (35, 36), in termsof the residues present (conserved arginine at entrance withsome hydrophobic surfaces).

A B

C

90º

K165

β5

β6

β7

β5

β6

β7K165

K97

E121

G166C145T164

A228

Q225

G244

N184

R163

H179

E121

G166C145

T164

A228

Q225 G244

N184

R163

H179

α6α6

P1 site

P2 site

P1 site

P2 site

β6

α2'

K165K165

P65 P65

A168

A168V122

V122

E121

R163E121R163

K97

α2'

β6

P1 site

P2 site

P1 site

P2 site

Fig. 5. Defining the K165 swinging arm in PDX1.3. Close-up views of theinteractions between K165 and surrounding residues in (A) apo-PDX1.3(green) and (B) PDX1.3 adduct (beige). Nonspecific secondary structure in-teractions have been omitted for clarity. K165 is stabilized in a P2 orienta-tion through interactions with the peptide backbone of C145, in which thehydroxyl group of T164 hydrogen bonds with the carbonyl group of A228,and its peptide backbone interacts with R163 and the adjacent strand β7.K165 P1 orientation is stabilized by a hydrogen bond interaction between itscarbonyl group and H179. In this orientation, the C145 interaction is main-tained, but the hydroxyl group of T164 hydrogen bonds with Q225, whichhas adopted a different rotamer position, and the amide carbonyl has ro-tated to form a hydrogen bond with A228 in β7. (C) A salt bridge formedbetween E121 and R163 forces the swing of K165 from the P2 to the P1 sitethrough the opening lined with P65, V122 and A168 (compare Left andRight, oriented 90° from each other along the axis indicated).

E5826 | www.pnas.org/cgi/doi/10.1073/pnas.1608125113 Robinson et al.

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

ConclusionThe data presented here provide a structural view of PLP bio-synthesis de novo in plants. They reveal that in the absence ofsubstrate, PDX1.3—in contrast to its bacterial counterparts—adoptsa catalytically poised conformation, in which several structural fea-tures around the P1 active site are in place. A water-filled cavity isobserved above the catalytic lysine (K97) of the P1 site, which isplausible to allow the entry of R5P in the Arabidopsis enzyme. In thepresence of the pentose phosphate substrate, an intermediate cor-responding to the predicted chromophoric adduct molecule at theP1 site is observed, and the mouth of the cavity is closed off byreorientation of a glutamate residue. In this PDX1.3-adduct struc-ture, the catalytic aspartate, D40, is in an attacking conformation,extended toward the adduct at the center of the P1 site. Significantly,K165 previously thought to coordinate the P1 and P2 active sites haspivoted from its resting position in the P2 site to the P1 site andprovides a clear demonstration of a swinging arm mechanism. Alysine-mediated swinging arm mechanism has been described pre-viously for a variety of different protein families, including Class Ialdolases (21), 2-oxoacid dehydrogenases, and the H protein ofglycine decarboxylases (reviewed in ref. 28). However, they differfrom that described here in that they either occur within a singleactive center (class I aldolases) or mediate shuttling of coenzymebound intermediates. PDX1 makes use of the lysine swinging arm tocoordinate distal divergent reaction sites, facilitating enzyme chem-istry that notably has no need for a coenzyme intermediate. This thusprovides an elegant example of how a sequence of biochemical

reactions can be conducted by a single enzyme with no need to re-cruit distinct partners or remodel an active site.In summary, we have captured key snapshots of the plant PDX1

in action, helping to define the intricate workings of a highlycomplicated enzyme, essential for plant survival and important asa source of vitamin B6 to animals including humans.

MethodsProtein Expression and Purification. The construct pET–PDX1.3 containing thePDX1 homolog from Arabidopsis designed for expression of the protein witha C-terminal hexa-histidine tag as described previously (37) was used in thisstudy. Expression in E. coli BL21 (DE3) RIL cells was induced by the addition of1 mM IPTG, followed by growth for 6 h at 28 °C (30). For protein purification,the cell pellet from 1 L of expression culture was resuspended in 100 mMsodium phosphate buffer, pH 7.5, containing 300 mM NaCl, 10 mM imid-azole, 1 mM β-mercaptoethanol, and 1 mM PMSF (Buffer A). Lysozyme(0.3 mg·mL−1), DNaseI (50 μg·mL−1), and a protease inhibitor mixture (EDTA-free;Roche) were added to the cell suspension, and lysis was performed by son-ication. Insoluble material was removed by centrifugation (18,000 × g for 30 minat 4 °C), and the clarified lysate was applied to a gravity flow column con-taining 0.5 mL nickel-nitrilotriacetic acid affinity resin (Macherey–Nagel)preequilibrated with Buffer A. Lysate was applied to the column a total ofthree times, and bound material was washed with 25 mL Buffer A followedby 25 mL of Buffer A containing 50 mM imidazole. Protein was eluted fromthe column with Buffer A containing 300 mM imidazole and concentrated to500 μL with partial buffer exchange (1:1) to 20 mM Tris·HCl, pH 7.0, con-taining 200 mM potassium chloride and 10 mM DTT followed by applicationto a preequilibrated Superdex200 (10/300; GE Healthcare) at 0.4 mL·min−1.Fractions containing the dodecameric form of PDX1.3 (30) were collected

A

D

B C

K165

β6α6

P1 site P152

A1682.7 Å

K165β6

α6

P65

A168

A66G169

R69

α2'

K165

β6

α6

E151

T170

E167

C-terminalregion

K149

P2 site

α2'

C-terminalregion

α8'

α6

14.3 Å2 64.6 Å2

α2'

α8'

α6C-terminal

region

Fig. 6. Conformational stability of the PDX1.3-adduct structure parallels a catch and pin. (A–C) Three features define conformational stability in the Ara-bidopsis PDX1.3-adduct (beige) upon comparison with its bacterial counterpart from G. stearothermophilus (cyan and gray). A proline residue, P152, in theG. stearothermophilus enzyme is displaced by the P2 site lysine (K149) upon formation of the chromophoric adduct [compare gray (1ZNN) and cyan (4WY0)depictions, before and after formation of the chromophore, respectively]. Consequently, the path of the loop connecting β6 and α6 deviates from the groundstate position during catalysis in this species. The proline residue is an alanine in Arabidopsis (beige), and so a deviation is not required upon formation of thechromophoric adduct (A). The loop region connecting the β6 strand and the α6 helix in Arabidopsis PDX1.3 is stabilized by interactions with the α2′ helix,which acts like a catch (B). Movement of the loop is restricted by P65 and A66 in the first turn of α2′ forming a hydrophobic pocket with A168 and G169,located at the top of the loop. The guanidinium group of R69 in the second turn of α2′ forms a hydrogen bond with the peptide backbone of A168, furtherstabilizing the loop. The α2′ helix is disordered in the matching structure of the G. stearothermophilus enzyme, and so movement of the corresponding loop isnot restricted in this species. A glutamate residue (E167) in Arabidopsis PDX1.3 (beige) located in the loop connecting β6 and α6 forms hydrogen bonds withT170 and coordinates a water molecule with the peptide backbone of a region near the C terminus of the protein (C). This glutamate pin provides additionalstabilization to this loop. The C-terminal region of the matching G. stearothermophilus structure is not resolved, and there is no hindrance to movement ofthe corresponding glutamate: it adopts a rotated position providing no stabilization of the loop. (D) Bfactor coloration of the apo-PDX1.3 (Left) and PDX1.3-adduct structures (Right). An increase in thermal motion is indicated by a shift in color from blue to red.

Robinson et al. PNAS | Published online September 19, 2016 | E5827

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

and analyzed by SDS/PAGE, and protein concentration was measured byinfrared absorbance using the Direct Detect system (Merck Millipore).

Crystallization and Data Collection. Crystals were grown at 18 °C by sitting-drop vapor diffusion. Crystallization drops, consisting of 5 μL protein solu-tion (2 mg·mL−1) and 5 μL precipitant solution [100 mM Mes·OH, pH 6.5,containing 1.4 M ammonium sulfate and 10% (vol/vol) 1,4-dioxane], wereequilibrated against a 1-mL reservoir of precipitant solution. Crystals withrhombohedral morphology (space group H3; 146) and unit cell dimensions178.5 Å × 178.5 Å × 115.8 Å appeared after 3–5 d and reached maximumdimensions (100 μm × 50 μm × 50 μm) after 4 wk. Crystals were recoveredwith nylon loops (Molecular Dimensions) and transferred to crystallization solu-tion supplemented with 5% (vol/vol) ethylene glycol. This solution was

replaced in a stepwise manner to obtain a final solution containing 20%(vol/vol) ethylene glycol in 5% increments (3 min per step). Cryoprotectedcrystals were plunge-frozen in liquid nitrogen for storage and X-ray dif-fraction experiments. Crystals containing the chromophoric adduct wereobtained by preparing crystals of apo-PDX1.3 as above, but cryoprotectionsolutions were supplemented with 6 mM R5P, and crystals were equilibratedfor time intervals of 10 min to 27 h. A total of 22 diffraction datasets wereobtained from PXI at the Swiss Light Source (PSI) at a wavelength of 0.9786nm. Crystals of apo-PDX1.3 and PDX1.3-adduct diffracted to 1.46 Å and1.76 Å, respectively. For data collection statistics see Table S1.

PDX1.3 Chromophoric Intermediate Formation. To mimic formation of thepreviously described chromophoric intermediate in PDX1 (10, 38), formation

A

D

E

B

C

R76

1245º

45º

R76

β1

α1

E62

R63

V42

D40

L61

Lys-R5P

E62

R63

Lys-R5P

K97

β3

β4

P. berghei G. stearothermophilusP. horikoshii

2 1

45º

Fig. 7. An alternative solvent-exposed cavity providing access to R5P. (A) Surface representation of B. subtilis PDX1 (gray, 2NV2; Middle) and Arabidopsisapo-PDX1.3 (green; Right), overlaid on their respective chain representations. Positively charged and negatively charged residues are colored blue and red,respectively. The orientation of the molecule relative to the typical depiction of the (β/α)8 barrel is shown (Left). The bacterial enzyme possesses two solventexposed cavities, labeled 1 and 2. The poised conformation of apo-PDX1.3 obscures cavity 1, but cavity 2 is open to solvent. (B) Close-up of the water-filledaccess channel (Middle) proposed to be the site of R5P entry in apo-PDX1.3 (site 2). Coloring as in A. Right illustrates the blocking of the channel by the alkylgroup of the active site lysine following substrate binding and is overlaid by the covalently bound active site Lys-R5P intermediate (gray) from the P. bergheienzyme (4ADU) following alignment of the two structures. (C) A comparison of apo-PDX1.3 (green) and PDX1.3-adduct (beige) illustrates movement of β1, α1,and E62 in Arabidopsis PDX1.3 following formation of the chromophoric adduct. The Lys-R5P intermediate from P. berghei is depicted (gray) as in B, Right. Aswell as moving D40 into position for catalytic attack (short arrow), this shift has the effect of closing the described cavity. V42, located at the apex of the β1/α1loop, is moved into proximity to the peptide carbonyls of L61 and E62, located on the loop connecting β2 and α2′. To compensate for this unfavorable in-teraction, this second loop is turned away, and the side chain of E62 twists to cover the mouth of the cavity, where it forms hydrogen bonds with theguanidinium group of R63, the amide nitrogen of R76, and a water, coordinated by polar and charged groups, located at the entrance of the proposed cavity(D). In D, E62 from the PDX1.3-adduct structure is superimposed on the apo-PDX1.3 structure as shown in B. (E) A similar site 2 cavity is observed in all otherstructures of PDX1 as illustrated for P. berghei (4ADU), P. horikoshii (4FIR), and G. stearothermophilus (4ZXY).

E5828 | www.pnas.org/cgi/doi/10.1073/pnas.1608125113 Robinson et al.

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

of the adduct was assessed in solution under cryoprotection conditionsusing a 96-well microplate reader (Synergy2; BioTek). The stoichiometry ofthe chromophore relative to total protein was estimated from the absor-bance maxima at 282 nm and 310 nm for PDX1.3 and chromophore, re-spectively, and the respective extinction coefficients of 5,960 M−1·cm−1

(estimated from the PDX1.3 primary sequence) and 16,200 M−1·cm−1 (10).

Structure Solution and Refinement. Diffraction data were processed withprograms of the CCP4 suite (39). Diffraction images were analyzed and in-tegrated with MOSFLM (40, 41), and structure factors were obtained withSCALA (42). The PDX1.3 structure was determined by molecular replacementusing the program PHASER MR (43) using a model prepared with CHAINSAW(44) based on the B. subtilis PDX1 homolog (2NV2). Phases for the chro-mophoric intermediate were obtained using the apostructure as a searchmodel. Model building and refinement were performed with COOT (45) andREFMAC5 (46), respectively, and coordinates and chemical restraints for the

chromophoric intermediate were generated with JLIGAND (47). Validationwas performed with COOT (45), PDB_REDO (48), and MOLPROBITY (49), andstructural analysis was performed with PYMOL and BAVERAGE (39). Re-finement statistics are presented in Table S1. The resolvable region of theapostructure from Arabidopsis (residues 21–296, chain A) was aligned withstructures from S. cerevisiae (3O07, Chain A), B. subtilis (2NV2, Chain A), andG. stearothermophilus (4WXY, Chain A) (15, 16, 18). Alignment was per-formed with SUPER in PYMOL using Cα atoms and with no cycles of re-finement (50). Figs. 2–7 were prepared with PYMOL (50), and density mapsin Figs. 2 and 4 were prepared with PHENIX (51).

ACKNOWLEDGMENTS. We thank Vincent Olieric, Tomizaki Takashi, and allstaff at the Swiss Light Source for assistance, as well as Stéphane Thore (Uni-versity of Bordeaux) for critical analysis of structural refinements. Financialsupport is gratefully acknowledged from the Swiss National Science Founda-tion (Grant 31003A-141117/1 to T.B.F.) as well as the University of Geneva.

1. Eliot AC, Kirsch JF (2004) Pyridoxal phosphate enzymes: Mechanistic, structural, andevolutionary considerations. Annu Rev Biochem 73:383–415.

2. Müller S, Kappes B (2007) Vitamin and cofactor biosynthesis pathways in Plasmodiumand other apicomplexan parasites. Trends Parasitol 23(3):112–121.

3. Fitzpatrick TB, et al. (2007) Two independent routes of de novo vitamin B6 bio-synthesis: Not that different after all. Biochem J 407(1):1–13.

4. Burns KE, Xiang Y, Kinsland CL, McLafferty FW, Begley TP (2005) Reconstitution andbiochemical characterization of a new pyridoxal-5′-phosphate biosynthetic pathway.J Am Chem Soc 127(11):3682–3683.

5. Raschle T, Amrhein N, Fitzpatrick TB (2005) On the two components of pyridoxal5′-phosphate synthase from Bacillus subtilis. J Biol Chem 280(37):32291–32300.

6. Belitsky BR (2004) Physical and enzymological interaction of Bacillus subtilis proteins re-quired for de novo pyridoxal 5′-phosphate biosynthesis. J Bacteriol 186(4):1191–1196.

7. Hanes JW, et al. (2008) Mechanistic studies on pyridoxal phosphate synthase: Thereaction pathway leading to a chromophoric intermediate. J Am Chem Soc 130(10):3043–3052.

8. Hanes JW, Keresztes I, Begley TP (2008) 13C NMR snapshots of the complex reactioncoordinate of pyridoxal phosphate synthase. Nat Chem Biol 4(7):425–430.

9. Hanes JW, Keresztes I, Begley TP (2008) Trapping of a chromophoric intermediate inthe Pdx1-catalyzed biosynthesis of pyridoxal 5′-phosphate. Angew Chem Int Ed Engl47(11):2102–2105.

10. Raschle T, et al. (2007) Reaction mechanism of pyridoxal 5′-phosphate synthase. De-tection of an enzyme-bound chromophoric intermediate. J Biol Chem 282(9):6098–6105.

11. Guédez G, et al. (2012) Assembly of the eukaryotic PLP-synthase complex from Plas-modium and activation of the Pdx1 enzyme. Structure 20(1):172–184.

12. Moccand C, Kaufmann M, Fitzpatrick TB (2011) It takes two to tango: Defining anessential second active site in pyridoxal 5′-phosphate synthase. PLoS One 6(1):e16042.

13. Neuwirth M, et al. (2009) X-ray crystal structure of Saccharomyces cerevisiae Pdx1provides insights into the oligomeric nature of PLP synthases. FEBS Lett 583(13):2179–2186.

14. Raschle T, et al. (2009) Intersubunit cross-talk in pyridoxal 5′-phosphate synthase, co-ordinated by the C terminus of the synthase subunit. J Biol Chem 284(12):7706–7718.

15. Smith AM, Brown WC, Harms E, Smith JL (2015) Crystal structures capture three states inthe catalytic cycle of a pyridoxal phosphate (PLP) synthase. J Biol Chem 290(9):5226–5239.

16. Strohmeier M, et al. (2006) Structure of a bacterial pyridoxal 5′-phosphate synthasecomplex. Proc Natl Acad Sci USA 103(51):19284–19289.

17. Zein F, et al. (2006) Structural insights into the mechanism of the PLP synthase ho-loenzyme from Thermotoga maritima. Biochemistry 45(49):14609–14620.

18. Zhang X, et al. (2010) Structural insights into the catalytic mechanism of the yeastpyridoxal 5-phosphate synthase Snz1. Biochem J 432(3):445–450.

19. Dunn MF, Niks D, Ngo H, Barends TR, Schlichting I (2008) Tryptophan synthase: Theworkings of a channeling nanomachine. Trends Biochem Sci 33(6):254–264.

20. Mouilleron S, Golinelli-Pimpaneau B (2007) Conformational changes in ammonia-channeling glutamine amidotransferases. Curr Opin Struct Biol 17(6):653–664.

21. Du J, Say RF, Lü W, Fuchs G, Einsle O (2011) Active-site remodelling in the bifunctionalfructose-1,6-bisphosphate aldolase/phosphatase. Nature 478(7370):534–537.

22. Hilario E, et al. (2016) Visualizing the tunnel in tryptophan synthase with crystallog-raphy: Insights into a selective filter for accommodating indole and rejecting water.Biochim Biophys Acta 1864(3):268–279.

23. Reed LJ (1998) From lipoic acid to multi-enzyme complexes. Protein Sci 7(1):220–224.24. Knowles JR (1989) The mechanism of biotin-dependent enzymes. Annu Rev Biochem

58:195–221.25. Gipson P, et al. (2010) Direct structural insight into the substrate-shuttling mechanism

of yeast fatty acid synthase by electron cryomicroscopy. Proc Natl Acad Sci USA107(20):9164–9169.

26. Leibundgut M, Jenni S, Frick C, Ban N (2007) Structural basis for substrate delivery byacyl carrier protein in the yeast fatty acid synthase. Science 316(5822):288–290.

27. Perham RN, Reche PA (1998) Swinging arms in multifunctional enzymes and thespecificity of post-translational modification. Biochem Soc Trans 26(3):299–303.

28. Perham RN (2000) Swinging arms and swinging domains in multifunctional enzymes:Catalytic machines for multistep reactions. Annu Rev Biochem 69:961–1004.

29. Zhu J, Burgner JW, Harms E, Belitsky BR, Smith JL (2005) A new arrangement of (beta/alpha)8 barrels in the synthase subunit of PLP synthase. J Biol Chem 280(30):27914–27923.

30. Moccand C, et al. (2014) The pseudoenzyme PDX1.2 boosts vitamin B6 biosynthesisunder heat and oxidative stress in Arabidopsis. J Biol Chem 289(12):8203–8216.

31. Kim S, Kim KJ (2013) Crystal structure of Mycobacterium tuberculosis Rv2606c: Apyridoxal biosynthesis lyase. Biochem Biophys Res Commun 435(2):255–259.

32. Matsuura A, Yoon JY, Yoon HJ, Lee HH, Suh SW (2012) Crystal structure of pyridoxalbiosynthesis lyase PdxS from Pyrococcus horikoshii. Mol Cells 34(4):407–412.

33. Ramachandran GN, Sasisekharan V (1968) Conformation of polypeptides and pro-teins. Adv Protein Chem 23:283–438.

34. Newby ZER, et al. (2008) Crystal structure of the aquaglyceroporin PfAQP from themalarial parasite Plasmodium falciparum. Nat Struct Mol Biol 15(6):619–625.

35. Abramson J, et al. (2003) Structure and mechanism of the lactose permease of Es-cherichia coli. Science 301(5633):610–615.

36. Kumar H, Finer-Moore JS, Kaback HR, Stroud RM (2015) Structure of LacY with anα-substituted galactoside: Connecting the binding site to the protonation site. ProcNatl Acad Sci USA 112(29):9004–9009.

37. Tambasco-Studart M, et al. (2005) Vitamin B6 biosynthesis in higher plants. Proc NatlAcad Sci USA 102(38):13687–13692.

38. Tambasco-Studart M, Tews I, Amrhein N, Fitzpatrick TB (2007) Functional analysis ofPDX2 from Arabidopsis, a glutaminase involved in vitamin B6 biosynthesis. PlantPhysiol 144(2):915–925.

39. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. ActaCrystallogr D Biol Crystallogr 67(Pt 4):235–242.

40. Battye TG, Kontogiannis L, Johnson O, Powell HR, Leslie AG (2011) iMOSFLM: A newgraphical interface for diffraction-image processing with MOSFLM. Acta Crystallogr DBiol Crystallogr 67(Pt 4):271–281.

41. Leslie AG (2006) The integration of macromolecular diffraction data. Acta CrystallogrD Biol Crystallogr 62(Pt 1):48–57.

42. Evans PR (2011) An introduction to data reduction: Space-group determination,scaling and intensity statistics. Acta Crystallogr D Biol Crystallogr 67(Pt 4):282–292.

43. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Cryst 40(Pt 4):658–674.44. Stein N (2008) CHAINSAW: A program for mutating pdb files used as templates in

molecular replacement. J Appl Cryst 41(3):641–643.45. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.

Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.46. Murshudov GN, et al. (2011) REFMAC5 for the refinement of macromolecular crystal

structures. Acta Crystallogr D Biol Crystallogr 67(Pt 4):355–367.47. Lebedev AA, et al. (2012) JLigand: A graphical tool for the CCP4 template-restraint

library. Acta Crystallogr D Biol Crystallogr 68(Pt 4):431–440.48. Joosten RP, Long F, Murshudov GN, Perrakis A (2014) The PDB_REDO server for

macromolecular structure model optimization. IUCrJ 1(Pt 4):213–220.49. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecular

crystallography. Acta Crystallogr D Biol Crystallogr 66(Pt 1):12–21.50. DeLano WL (2002) The PyMOL Molecular Graphics System (Schrodinger LLC, San

Carlos, CA).51. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-

molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221.

Robinson et al. PNAS | Published online September 19, 2016 | E5829

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0