GABA production by glutamic acid decarboxylaseis regulated by a dynamic catalytic loopGustavo Fenalti1,6, Ruby H P Law1,6, Ashley M Buckle1,6, Christopher Langendorf1, Kellie Tuck2,Carlos J Rosado1, Noel G Faux1, Khalid Mahmood1, Christiane S Hampe3, J Paul Banga4, Matthew Wilce1,Jason Schmidberger1, Jamie Rossjohn1, Ossama El-Kabbani5, Robert N Pike1, A Ian Smith1, Ian R Mackay1,Merrill J Rowley1 & James C Whisstock1
Gamma-aminobutyric acid (GABA) is synthesized by two isoforms of the pyridoxal 5¢-phosphate–dependent enzyme glutamicacid decarboxylase (GAD65 and GAD67). GAD67 is constitutively active and is responsible for basal GABA production. Incontrast, GAD65, an autoantigen in type I diabetes, is transiently activated in response to the demand for extra GABA inneurotransmission, and cycles between an active holo form and an inactive apo form. We have determined the crystal structuresof N-terminal truncations of both GAD isoforms. The structure of GAD67 shows a tethered loop covering the active site,providing a catalytic environment that sustains GABA production. In contrast, the same catalytic loop is inherently mobile inGAD65. Kinetic studies suggest that mobility in the catalytic loop promotes a side reaction that results in cofactor release andGAD65 autoinactivation. These data reveal the molecular basis for regulation of GABA homeostasis.
Glutamic acid decarboxylase (GAD) is an essential enzyme present in alleukaryotes and in many prokaryotes1–3. In mammals, the two isoformsof GAD (termed GAD67 or GAD1, and GAD65 or GAD2, respectively)function to produce the inhibitory neurotransmitter GABA fromglutamate and control fundamental processes such as neurogenesis4,5,movement and tissue development6–8. Gene knockout studies in micehave helped delineate distinct roles for each isoform; Gad1–/– mice havesubstantially reduced GABA levels and die at birth of severe cleft palate7.In contrast, Gad2–/– mice have normal basal levels of GABA and appearnormal at birth, but develop fatal seizures and anxiety phenotypes9. Inhumans, a mutation in GAD1 results in spastic cerebral palsy10.
GAD is a pyridoxal 5¢-phosphate (PLP)-dependent enzyme. Mem-bers of this superfamily catalyze a range of important reactions,including a-decarboxylation, transamination, racemization and aldolcleavage11,12, and many of these enzymes catalyze more than onereaction. Two possible outcomes of glutamate decarboxylation byGAD have been described (Fig. 1a). Initial studies on preparationsof porcine enzyme revealed that the primary reaction of GADproduces GABA and active holoenzyme, whereas the side reactionyields pyridoxamine 5¢-phosphate (PMP), succinic semialdehyde(SSA) and inactive apo enzyme13 (Fig. 1a). Further studies onhuman recombinant GAD showed that GAD65 is at least 15 timesmore efficient than GAD67 at catalyzing the side reaction, which leadsto rapid loss of enzyme activity14. It has been suggested that this
represents a key mechanism for the control of GABA production13,14.Consistent with this, studies on rat brain GAD reveal that B80% ofGad1 isolated from cells exists in the active holo form, whereas B80%of Gad2 is in the inactive apo form15. Collectively, these data suggestthat GAD67 is responsible for production of a basal pool of GABA,whereas GAD65 is activated to produce extra GABA when required3,9.In this study, we set out to investigate the structural basis for thefunctional differences between the two GAD isoforms.
RESULTSCrystal structure of GAD67To address the structural basis of basal GABA production, wedetermined the 2.3-A structure of an N-terminally truncated formof GAD67 (amino acid residues 90–594; Fig. 1b and Fig. 2) that isenzymatically active16. GAD67 forms an obligate functional dimer17,with B6,800 A2 buried at the dimer interface. The monomeric unitcomprises three domains (N-terminal, PLP-binding and C-terminal;Fig. 1b and Fig. 2a). The N-terminal domain includes two parallela-helices that pack against the N-terminal and PLP-binding domainsof the other monomer. The PLP-binding domain adopts the type IPLP-dependent transferase–like fold and comprises nine a-helicessurrounding a nine-stranded, mainly parallel b-sheet (Fig. 2a). TheC-terminal domain contains three a-helices, together with a four-stranded antiparallel b-sheet.
Received 13 November 2006; accepted 7 March 2007; published online 25 March 2007; doi:10.1038/nsmb1228
1Department of Biochemistry and Molecular Biology, Monash University, Clayton, Melbourne, VIC 3800, Australia. 2School of Chemistry, Monash University,Melbourne VIC 3800, Australia. 3Department of Medicine, University of Washington School of Medicine, Seattle, Washington 98195, USA. 4King’s College LondonSchool of Medicine, Division of Gene and Cell Based Therapy, Denmark Hill Campus, London SE5 9PJ, UK. 5Victorian College of Pharmacy, Monash University,Parkville, VIC 3052, Australia. 6These authors contributed equally to this work. Correspondence should be addressed to J.C.W. ([email protected])or M.J.R. ([email protected]).
280 VOLUME 14 NUMBER 4 APRIL 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY
The two active sites are located in the center of the PLP-bindingdomain at the dimer interface (Fig. 2a). Recombinant GAD proteinwas extracted from cells with a buffer containing glutamate and PLP18,and examination of both active sites of GAD67 revealed electrondensity consistent with the presence of PLP as well as the productGABA (Fig. 2 and Supplementary Fig. 1 online). In each activesite, a Schiff base linkage (internal aldimine) between PLP and theactive site lysine (Lys405) is supported by the continuous electrondensity (Fig. 2b,c and Supplementary Fig. 1). The conservedbase-stacking residue His291 (ref. 19) is positioned on the top ofthe pyridine ring of PLP (Fig. 2b,c). In monomer A, the active sitecontains a single GABA molecule (Fig. 2b). In the active site ofmonomer B, two discretely disordered conformations of GABAwere observed (Fig. 2c). In one conformation, the carboxyl groupof GABA forms a salt bridge with Arg567 and a hydrogen bond withGln190. In the other conformation, continuous density between PLPand GABA reveals the presence of a trapped intermediate, modeled asa PLP-GABA moiety. The torsion angle around the C4-C4¢bond deviates markedly from coplanarity with the pyridine ring,suggesting that this moiety is favored over its unprotonated pre-cursor, the quinoid, which has a C4-C4¢ double bond (Fig. 2c).However, the resolution of the present data preclude a fullmechanistic interpretation.
Notably, each active site is substantially covered by an extendedloop (residues 432–442, termed the ‘catalytic loop’) contributed intrans by the other monomer (Fig. 2a,d and Supplementary Fig. 2online). The catalytic loop is well ordered in the electrondensity, buries B1,150 A2 at its interface and makes 12 hydrogenbonds, four water-mediated hydrogen bonds, two salt bridgesand three hydrophobic interactions with the body of the molecule(Supplementary Fig. 2 and Supplementary Table 1 online). More-over, the catalytic loop brings the conserved residue Tyr434 into closeproximity to the base-stacking His291 and GABA (Fig. 2b,c). Twoconformations of Tyr434 are observed. In the active site of monomer A,the side chain hydroxyl group of Tyr434 forms a hydrogen bond to theNe2 of the base-stacking His291 (Fig. 2b). In the active site ofmonomer B, Tyr434 is flipped to an alternative conformation wherethe hydroxyl group hydrogen-bonds to the backbone nitrogen ofTyr292 (Fig. 2c). Superpositions reveal that in the former conforma-tion (A), the hydroxyl group of Tyr434 would be 2.8 A from theprotonation site (Ca) of a PLP-GABA moiety. In this conformation,we observe free GABA in the active site. We therefore suggest thatTyr434 is directly responsible for protonating the Ca position and thatthe hydrogen bond between Tyr434 and His291 in GAD67 favorablylowers the pKa of the Tyr434 in the active site of monomer A (Fig. 2b).In the active site of monomer B (Fig. 2c), the hydroxyl group of
Quinoid
SSA + PMP
Sidereaction
Glutamate
CO2
GABA
Primaryreaction
PLP
H+ + H2O H+ + H2O
GADApo form(inactive)
GADHolo form(active)
GAD67GAD65
GAD67GAD65
GAD67GAD65
GAD67GAD65
GAD67GAD65
GAD67GAD65
GAD67GAD65
1 10 20 30 40 50 60 70 80 90
1 10 20 30 40 50 60 70 80 90
580
590
490 500 510 520 530 550 560 570
580570560550540530520510500
390 400 410 420 430 440 450 460 470 480
490
380370360350340330320310300290
390380370360350340330320310300
s7B s8B s9B α9 s5B α10
α7α6α5α4
s4Bα8
190 200 210 220 230 240 250 260 270 280
100
100 110 120
α1 α2 α3
130 140 150 160 170 180 190
110 120 130 140 150 160 170 180
200 210 220 230 240 250 260 270 280 290
480470460450442428420410400
s1C
s3B s2B α11 catalytic loop α12 α13
s6B
s1A
s1B
s2C s4C s3Cα14 α15
a
b
540
Figure 1 Reactions catalyzed by GAD and sequence alignment of both GAD isoforms. (a) GAD holoenzyme (PLP bound) catalyzes the decarboxylation of
glutamate. Two alternate pathways have been characterized13. (b) Sequence alignment of GAD67 and GAD65, labeled with secondary structure. Gray,
deleted region; blue, N-terminal region; cyan, PLP domain; green, C-terminal domain; magenta, catalytic loop; black stars, GAD65 residues that affect the
binding of monoclonal autoantibodies; black boxes, active site residues; red boxes, functional residues; green circles, residues mutated in this study.
Alignment was produced using ALSCRIPT41.
ART IC L E S
NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 4 APRIL 2007 281
Tyr434 would be unable to protonate the Ca of a PLP-GABA moiety,as it is 46 A away from the Ca. This is consistent with the presence ofthe PLP-GABA intermediate. Mobility of Tyr434 may also facilitatesubstrate ingress and product egress from the active site. Together,these observations provide a structural rationalization for GABAproduction by GAD67.
Crystal structure of GAD65Whereas most neurotransmitters are synthesized each by a singleenzyme, GABA is unusual in that its biosynthesis is catalyzed by
both GAD67 and GAD65 isoforms, with the majority of the latterenzyme existing in an inactivated state until it responds to an urgentdemand for extra GABA, for example in response to stress8,20. Tounderstand this behavior, we determined the 2.3-A crystal structure ofa truncated form of GAD65 (residues 84–585; Fig. 1b and Fig. 3).GAD65 is 71% identical to GAD67 in the region structurally chara-cterized (474 residues) and superposes on GAD67 with an r.m.s.deviation of 0.8 A (over 474 Ca atoms), confirming that GAD65 andGAD67 adopt the same fold and dimeric arrangement (Fig. 3 andSupplementary Tables 2 and 3 online).
PLP-GABA
a
b
dLys405-PLP
PLP-GABA
Tyr434
Figure 2 Dimeric structure of GAD67 within the asymmetric unit of the crystal. (a) The N-terminal (blue), PLP-binding (cyan) and C-terminal (green)
domains are labeled. Monomer A is colored slightly lighter than monomer B. In the active sites, the Lys405-PLP Schiff base and the PLP-GABA moiety areshown as orange spheres and bound GABA product as yellow spheres. The catalytic loop forms a flap over the active site of an adjacent monomer in trans
and is colored magenta in each monomer. (b,c) Active sites of GAD67, showing a close-up of Lys-PLP (shown as orange sticks), PLP-GABA (dark gray) and
noncovalently bound GABA (yellow) moieties. The 2Fo – Fc omit electron density of GAD67 contoured at 1 s is also shown (mesh). Hydrogen bonds are
shown as dotted lines, water molecules as red spheres. The side chain of Tyr434 from the catalytic loop of other monomer is shown in magenta. Protonation
sites C4¢ and Ca are labeled. Chemical structures of modeled moieties are shown in the key to structures; stereo representations of b and c can be found in
Supplementary Figure 1. (d) Surface representation of the GAD67 active site, colored as in a. Lys405-PLP (orange), PLP-GABA (gray) and GABA product
(yellow) are shown as sticks. The two conformations of Tyr434 are shown (purple from monomer A, mauve from monomer B).
Catalytic loop (GAD67)
PLP domain
432–442 (423–433)
197–472(188–463)
C-terminal domain473-593 (464–584)
C-terminal domains
N-terminaldomain
93-196 (88–187)
422
517521
434
GAD67 A monomer
GAD65GAD67 B monomer
Catalytic loops
517 517
521
521
a bFigure 3 Structural comparisons between GAD67
and GAD65. (a) Structural superposition between
GAD67 monomer A (green), GAD67 monomer B
(purple) and GAD65 (light brown). Disordered
regions in GAD65 are numbered. (b) Structural
superposition of GAD67 dimer (green) and
GAD65 dimer (light brown). PLP is shown as
orange spheres and GABA moieties bound in
active site are shown as yellow spheres. The
catalytic loops of GAD67 are colored magenta;
this region is not visible in electron density in
GAD65. Both figures highlight the structural
shifts in the C terminus.
ART IC L E S
282 VOLUME 14 NUMBER 4 APRIL 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY
In both active sites of GAD65, PLP is covalently attached to thecatalytic Lys396. GABA can also be modeled in two discretelydisordered conformations (Fig. 4a) similar to those seen in GAD67(Fig. 4b). However, no electron density was observed that suggestedthe presence of the covalent PLP-GABA intermediate (Fig. 4a).Notably, in contrast to GAD67, the region corresponding to thecatalytic loop in GAD65 is disordered, rendering the active sitecompletely exposed (Fig. 2d, Fig. 3 and Fig. 4c,d). A comparisonbetween the sequences and structures of GAD isoforms offers insightinto the structural basis for this difference. Most residues in contactwith the catalytic loop of GAD67 are identical in GAD65 (Fig. 1b,Fig. 4c, Supplementary Table 3 and Supplementary Fig. 3 online).The exception is Phe283 in GAD65 (position 292 in GAD67), which issubstituted by a tyrosine in GAD67 (Tyr292 contacts the catalyticTyr434 and forms a hydrogen bond to Cys431; Fig. 4c). Furthermore,the catalytic loop of GAD65 contains several substitutions including aGly-Ser substitution at position 424 (433 in GAD67) and aPro-Gln substitution at position 429 (438 in GAD67) (Fig. 4c). In
GAD67, Pro438 is centered on a sharp kink in the catalytic loop andGly433 adopts a +/+ conformation (that is, one in the positivequadrant of the Ramachandran plot) in monomer A (Fig. 4c); thestereochemical properties of both of these residues may be importantfor the loop conformation in GAD67.
Structural basis for GAD65 autoinactivationWe investigated how the differences in and around the active sites ofthe two isoforms may influence enzyme activity. Consistent with thepublished literature14, after preincubation with glutamate for 20 min,GAD67 retains full activity, whereas GAD65 loses B75% of activity(Fig. 5a and Supplementary Table 4 online), which can be restored byaddition of PLP (data not shown). Furthermore, mutation of theconserved Tyr434 (Tyr425 in GAD65) in the catalytic loop abolishedGABA production (Supplementary Fig. 4 online), which demon-strates that this residue is essential for catalysis.
To test the role of the catalytic loop in GAD65 autoinactivation, weinterchanged this region between the two isoforms, creating thechimeric constructs GAD65(67loop) and GAD67(65loop). Althoughthe extent of inactivation of GAD65(67loop) was similar to that ofGAD65, that of GAD67(65loop) was significantly greater (P ¼ 0.0015)(Fig. 5b,c and Supplementary Table 4).
We also investigated the role of GAD65 Phe283/GAD67 Tyr292 inGAD inactivation. The extent of inactivation of the mutants
a
c
b
dLys396-PLP
Phosphate
GABA
Monomer B
Monomer A
GAD65
GAD65
GAD65(F
283Y
)
GAD65(6
7loop
)
GAD65(6
7loop
_F28
3Y)
GAD67
GAD67(Y
292F
)
GAD67(6
5loop
)
GAD67(6
5loop
_Y29
2F)
GAD670
25
50
75
100
1250 min20 min
P = 5.11 × 10–7
P = 1.5 × 10–3
P = 1.56 × 10–6P = 2.2 × 10–3
P = 1.68 × 10–7
125
100
75
50
25
0
% r
esid
ual a
ctiv
ity
125
100
25
75
50
0
% r
esid
ual a
ctiv
ity
% r
esid
ual a
ctiv
ity
a
b c
Figure 5 Inactivation of GAD in the presence of glutamate. The activity of
each GAD variant after 20 min preincubation with glutamate is reported as apercentage of its activity at time 0, before preincubation. (a) Activity of wild-
type GAD65 and GAD67 before (0 min) and after (20 min) preincubation
with glutamate. (b,c) Activity of wild-type and mutant GAD65 (b) and
GAD67 (c) after preincubation. Each bar in a–c shows the mean of three
measurements; error bars show s.d.
Figure 4 Comparison of GAD67 and GAD65 active sites, in stereo.
(a) Active site of GAD65. 2Fo – Fc omit electron density contoured at 1 s is
also shown (atoms from Lys-PLP and GABA omitted from density
calculation). Lys-PLP moiety is shown as orange sticks; noncovalently bound
GABA molecules are colored yellow and cyan. (b) Superposition of active
site residues of GAD67 (green, monomer A; cyan, monomer B) and GAD65
(light brown). Both Tyr434 side chain conformations from the catalytic loops
of GAD67 are shown. Tyr434 from monomer B (in cyan) enters the active
site of monomer A. In this conformation, the hydroxyl group of Tyr434 is
2.8 A from the Ca atom. (c) Interactions between catalytic loop and
adjacent monomer. Residues that are different in GAD65 are colored orange.
The alternative conformation of Tyr434 (from monomer B) is shown as
mauve bonds. Phe283 of GAD65 is shown as yellow bonds. Residues in
GAD65 corresponding to GAD67 residues 432–442 are disordered. The PLP
moiety in the active site is shown as orange spheres, hydrogen bonds asdotted lines and water molecules as red spheres. (d) Surface representation
of active site of GAD65, in similar orientation to that in Figure 2d. Lys396-
PLP (orange), GABA product (yellow/cyan sticks).
ART IC L E S
NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 4 APRIL 2007 283
GAD65(F283Y) and GAD67(Y292F) did not differ from the respectivewild-type enzymes (Fig. 5b,c and Supplementary Table 4). However,the extent of inactivation of GAD65(67loop_F283Y) mutant wassignificantly less (Fig. 5b, P ¼ 1.68 � 10�7): this mutant lostB44% rather than B75% of activity after 20 min. Moreover,GAD67(65loop_Y292F) lost B58% of activity after 20 min (Fig. 5c,P ¼ 1.58 � 10�6). Thus, both the sequence of the catalytic loop andthe residues against which this region packs are importantfor modulating enzyme autoinactivation. In addition, the entireC-terminal domain, against which the catalytic loop packs, is moremobile in GAD65 and is shifted 1.5 A relative to its position in GAD67(Fig. 3 and Supplementary Fig. 5 online). We therefore suggest thatthe position and mobility of the C-terminal domain may furthercontribute to the dynamic role of the catalytic loop in GAD65autoinactivation. Notably, the structure of Escherichia coli GAD,which is distantly related to mammalian GAD, also reveals mobilityin the C-terminal region1,2.
GAD65 as an autoantigenGAD65 is autoantigenic, with autoantibodies characteristically detect-able in type I diabetes as well as neurological disorders that includestiff-person syndrome15,21 and cerebellar ataxia22; in contrast, despiteits high sequence identity with GAD65, GAD67 rarely is autoantigenic.Notably, certain autoantibodies, such as mAb b78, also inhibit GAD65catalytic activity in vitro23. We note that many of the point mutationsthat affect the binding of GAD65 autoantibodies (includingb78 (ref. 23)) map to regions of structural divergence from GAD67and of mobility, such as the C terminus (Fig. 3 and SupplementaryFigs. 5–7 online). However, the precise epitope of a GAD autoanti-body complex has not been structurally defined, precluding a moredetailed analysis.
DISCUSSIONOur structural and mutational data reveal that the dynamic confor-mation of the catalytic loop is key in the inactivation of GAD65.Structural studies on a related enzyme, DOPA decarboxylase (DDC),provide an attractive explanation for how the transient absence of thecatalytic tyrosine may result in enzyme inactivation. In DDC, as inGAD65, the region corresponding to the catalytic loop is disordered(Supplementary Fig. 8a online)24. Crucially, under anaerobic condi-tions, mutation of Tyr332 in DDC (the residue equivalent to thecatalytic Tyr434 in GAD67) results in protonation at the C4¢ positionand a decarboxylation-dependent transamination side reaction. Thisside reaction results in PMP release and enzyme inactivation25. It hastherefore been suggested that in DDC, Tyr332 performs the protona-tion of the Ca atom of the quinoid intermediate that is crucial for theprimary enzymatic pathway25 (Supplementary Figs. 8 and 9 online).Our data strongly support this hypothesis. For GAD67, our resultssuggest that the continuous presence of Tyr434 in the active site favorsprotonation of the Ca atom and uninterrupted GABA production(Supplementary Figs. 8 and 9). Furthermore, in GAD65, the resultssuggest that the transient absence of the catalytic tyrosine allowsprotonation at C4¢, perhaps by the Schiff base Lys396 (ref. 26). This inturn leads to SSA production, release of PMP and enzyme autoinacti-vation (Supplementary Fig. 9b).
The structure of the catalytic loop in GAD67 provides definitivegeneral insight into the role of the ordinarily flexible catalytic loop inthe mechanism of PLP-dependent decarboxylases. These data revealhow GAD67 is able to catalyze the steady production of GABA andhow short-term requirements for extra GABA can be satisfied bycofactor-controlled activation and subsequent autoinactivation of
GAD65. Insight into the latter process may prove important fordeveloping approaches to therapeutically stabilize GABA productionby GAD65. Finally, these data indicate a mechanism for GABA-glutamate homeostasis and offer a persuasive explanation for therequirement of two GAD isoforms in mammals.
METHODSProtein production and crystallization. The coding sequences of human
GAD65 and GAD67, residues 84–585 and 90–594 (Fig. 1b), respectively, were
expressed in Saccharomyces cerevisiae as fusions to a C-terminal hexahistidine
tag18. Recombinant proteins were purified from the cell lysate by immobilized
metal-affinity chromatography as described27 followed by size-exclusion chro-
matography. Before crystallization, purified holoenzymes were concentrated to
10 mg ml–1 and equal molar chelidonic acid was added. The proteins were
crystallized by the hanging drop method. GAD65 was crystallized in 20%
(v/v) ethanol, 100 mM MES (pH 6.2), 10 mM 2-mercaptoethanol and 20 mM
CaCl2, and GAD67 in 18% (w/v) PEG 8,000, 100 mM MES (pH 6.3), 10 mM
2-mercaptoethanol and 20 mM CaCl2, at 20 1C.
X-ray data collection, structure determination and refinement. Data were
collected at the IMCA-CAT beamline at the Advanced Photon Source (Table 1).
Both GAD65 and GAD67 crystals diffracted to 2.3-A resolution. GAD67
crystals belong to space group P21 and have unit cell dimensions of a ¼84.05 A, b ¼ 62.74 A, c ¼ 101.35 A and b ¼ 106.71, consistent with two
molecules per asymmetric unit. GAD65 crystals belong to space group C2221
and have unit cell dimensions of a ¼ 78.25 A, b ¼ 99.06 A and c ¼ 120.1 A,
consistent with one molecule per asymmetric unit. The data were merged and
processed using MOSFLM and SCALA28,29. Subsequent crystallographic and
structural analysis was done using the CCP4i interface30 to the CCP4 suite31,
unless stated otherwise. Five percent of the dataset was flagged for calculation
of Rfree, with neither a sigma nor a low-resolution cut-off applied to the data. A
summary of statistics is provided in Table 1.
Table 1 Data collection and refinement statistics
GAD67 GAD65
Data collection
Space group P21 C2221
Cell dimensions
a, b, c (A) 84.05, 62.74, 101.35 78.25, 99.05, 120.01
a, b, g (1) 90.00, 106.69, 90.00 90.00, 90.00, 90.00
Resolution (A) 97.1 (2.3) 54.6 (2.3)
Rmerge 4.5 (19.4) 3.5 (16.3)
I / sI 17.1 (4.4) 14.9 (4.7)
Completeness (%) 93.7 (69.2) 98.8 (98.8)
Redundancy 3.4 (2.3) 3.8 (3.8)
Refinement
Resolution (A) 2.3 2.3
No. reflections 42,284 20,717
Rwork / Rfree 17.8 / 21.4 19.5 / 25.1
No. atoms
Protein 8,292 3,881
Ligand 36 20
Water 359 93
B-factors
Protein 32.0 54.5
Ligand 33.2 53.7
Water 34.7 52.4
R.m.s. deviations
Bond lengths (A) 0.008 0.009
Bond angles (1) 1.2 1.3
Values in parentheses are for highest-resolution shell.
ART IC L E S
284 VOLUME 14 NUMBER 4 APRIL 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY
The structure of GAD67 was solved using the molecular replacement
method and PHASER32. A search model was constructed from the crystal
structure of pig dopa decarboxylase (DDC; PDB 1JS3)24, the closest structural
homolog identified using the FFAS server33 (sequence identity ¼ 20%). The
structure was trimmed to remove regions of high sequence divergence, leaving
predominantly residues belonging to the PLP-binding domain (representing
B60% of the total GAD67 structure). A ‘mixed’ model consisting of conserved
side chains (all other non–alanine/glycine residues truncated at Cg atom) was
then created using the SCRWL server34. Two outstanding solutions having
Z-scores of 12 and 10 were produced, and these packed well within the unit cell.
Together with the unbiased features in the initial electron density maps, the
correctness of the molecular replacement solution was confirmed.
Structure refinement and model building proceeded using one molecule in
the asymmetric unit (with the other noncrystallographic symmetry
(NCS)-related molecule generated using NCS operators). Maximum-likelihood
refinement using REFMAC35, incorporating translation, libration and screw-
rotation displacement (TLS) refinement was carried out, using a bulk solvent
correction (Babinet model with mask). Throughout most stages of refinement,
tight NCS restraints were imposed on all residues in the two molecules in the
asymmetric unit. All model building and structural validation was done
using COOT36. Water molecules were added to the model using ARP/wARP37
when the Rfree reached 30%. Solvent molecules were retained only if they
had acceptable hydrogen-bonding geometry contacts of 2.5–3.5 A with
protein atoms or with existing solvent and were in good 2Fo – Fc and
Fo – Fc electron density.
The structure of GAD65 was determined by molecular replacement
using PHASER32 and the refined GAD67 model. Refinement proceeded as
for GAD67.
Structural analysis. PYMOL38 was used to produce Figures 2–4 and Supple-
mentary Figures 1–3 and 5–8. Structures were superimposed using
MUSTANG39. Accessible surface areas were calculated using the CCP4 program
AREAIMOL31.
Mutagenesis. Amino acid substitutions were introduced into the GAD65 and
GAD67 sequence using the QuikChange mutagenesis kit (Stratagene) with the
primers listed in Supplementary Table 5 online.
Enzyme activity assays. Inactivation of GAD in the presence of glutamate was
measured using a method based on previously described work14,40, with the
following modifications. Radioactive glutamate (0.1 mCi) was added to each
reaction mix and enzyme activity was determined in 10-min assays. For each
variant, purified holoenzyme (at a final concentration of 20 mg ml–1) was
incubated at 30 1C for 20 min in the presence of 5 mM glutamate in buffer A
(0.1% (v/v) Triton X-100, 1 mM 2-mercaptoethanol, 1 mM 2-aminoethyli-
sothiouronium bromide, 100 mM K/NaPO4 (pH 7.2)). Initial velocity values
(Supplementary Table 4) of reaction mix before (time 0) and after glutamate
treatment were determined by adding L-[1-14C ]glutamate and incubating at
30 1C for 10 min (again in buffer A). The 14CO2 produced was trapped under
argon, using a benzethonium hydroxide–treated filter-paper plug placed in the
lid of the microfuge tubes. The residual activity after 20 min of treatment with
glutamate was calculated as a percentage of the enzyme activity at time 0. Each
bar is the mean of three measurements; error bars show s.d. Statistical
comparisons were done using two-tailed two-sample Student’s t-tests. All
assays were done under initial velocity conditions (Supplementary Fig. 10
online). GABA production was measured by 1H NMR spectroscopy on a
Bruker DPX 400 MHz spectrometer (Supplementary Fig. 4).
Accession codes. Protein Data Bank: Coordinates have been deposited with
accession codes 2OKJ (GAD67) and 2OKK (GAD65).
Note: Supplementary information is available on the Nature Structural & MolecularBiology website.
ACKNOWLEDGMENTSJ.C.W. is a National Health and Medical Research Council of Australia (NHMRC)Principal Research Fellow and a Monash University Senior Logan Fellow. A.M.B.and M.W. are NHMRC Senior Research Fellows. A.I.S. is an NHMRC PrincipalResearch Fellow. J.R. is an Australian Research Council Federation Fellow. G.F. is
a PhD scholar funded by the CAPES Foundation, subordinated to the Ministryof Education, Brazil. This work was supported by the NHMRC, the AustralianResearch Council and the State Government of Victoria (Australia). We thank thestaff at IMCA-CAT (The Advanced Photon Source) for technical assistance, theAustralian Synchrotron Research Program for support and M. Dunstone forcritical reading of the manuscript.
AUTHOR CONTRIBUTIONSG.F. purified protein, crystallized protein, collected and analyzed data, performedenzyme assays and wrote the paper. R.H.P.L. purified protein, crystallized protein,collected and analyzed data, determined structures, performed enzyme assays andwrote the paper. A.M.B. collected and processed data, determined and analyzedstructures and wrote the paper. C.L. cloned GAD67 and produced bothrecombinant GAD proteins. K.T. performed NMR and enzyme assays. C.J.R.made GAD mutants. N.G.F. assisted with structural analysis. K.M., assisted withkinetic analysis. C.S.H. and J.P.B. provided antibodies and immunological dataand analysis. M.W. and J.S. assisted with GAD67 data collection. J.R. assisted withGAD data collection and in writing the paper. O.E.-K. designed the GAD65expression construct. R.N.P. performed enzyme kinetic analysis. A.I.S. performedmass spectrometry experiments and N-terminal sequencing. I.R.M. analyzedimmunological data and provided critical review of the manuscript. M.J.R. co-ledthe research, analyzed immunological data and provided critical review of themanuscript. J.C.W. led the research, collected and analyzed data, performedstructural analysis and wrote the paper.
COMPETING INTERESTS STATEMENTThe authors declare no competing financial interests.
Published online at http://www.nature.com/nsmb/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Capitani, G. et al. Crystal structure and functional analysis of Escherichia coliglutamate decarboxylase. EMBO J. 22, 4027–4037 (2003).
2. Gut, H. et al. Escherichia coli acid resistance: pH-sensing, activation by chloride andautoinhibition in GadB. EMBO J. 25, 2643–2651 (2006).
3. Soghomonian, J.J. & Martin, D.L. Two isoforms of glutamate decarboxylase: why?Trends Pharmacol. Sci. 19, 500–505 (1998).
4. Fagiolini, M. et al. Specific GABAA circuits for visual cortical plasticity. Science 303,1681–1683 (2004).
5. Ge, S. et al. GABA regulates synaptic integration of newly generated neurons in theadult brain. Nature 439, 589–593 (2006).
6. Nakatsu, Y. et al. A cluster of three GABAA receptor subunit genes is deleted in aneurological mutant of the mouse p locus. Nature 364, 448–450 (1993).
7. Asada, H. et al. Cleft palate and decreased brain gamma-aminobutyric acid in micelacking the 67-kDa isoform of glutamic acid decarboxylase. Proc. Natl. Acad. Sci. USA94, 6496–6499 (1997).
8. Kash, S.F. et al. Epilepsy in mice deficient in the 65-kDa isoform of glutamic aciddecarboxylase. Proc. Natl. Acad. Sci. USA 94, 14060–14065 (1997).
9. Asada, H. et al. Mice lacking the 65 kDa isoform of glutamic acid decarboxylase(GAD65) maintain normal levels of GAD67 and GABA in their brains but aresusceptible to seizures. Biochem. Biophys. Res. Commun. 229, 891–895 (1996).
10. Lynex, C.N. et al. Homozygosity for a missense mutation in the 67 kDa isoform ofglutamate decarboxylase in a family with autosomal recessive spastic cerebral palsy:parallels with Stiff-Person Syndrome and other movement disorders. BMC Neurol. 4,20 (2004).
11. Jansonius, J.N. Structure, evolution and action of vitamin B6-dependent enzymes.Curr. Opin. Struct. Biol. 8, 759–769 (1998).
12. Percudani, R. & Peracchi, A. A genomic overview of pyridoxal-phosphate-dependentenzymes. EMBO Rep. 4, 850–854 (2003).
14. Battaglioli, G., Liu, H. & Martin, D.L. Kinetic differences between the isoforms ofglutamate decarboxylase: implications for the regulation of GABA synthesis.J. Neurochem. 86, 879–887 (2003).
15. Lernmark, A. Glutamic acid decarboxylase–gene to antigen to disease. J. Intern. Med.240, 259–277 (1996).
16. Battaglioli, G., Liu, H., Hauer, C.R. & Martin, D.L. Glutamate decarboxylase: loss of N-terminal segment does not affect homodimerization and determination of the oxidationstate of cysteine residues. Neurochem. Res. 30, 989–1001 (2005).
17. Martin, D.L., Liu, H., Martin, S.B. & Wu, S.J. Structural features and regulatoryproperties of the brain glutamate decarboxylases. Neurochem. Int. 37, 111–119(2000).
18. Law, R.H., Rowley, M.J., Mackay, I.R. & Corner, B. Expression in Saccharomycescerevisiae of antigenically and enzymatically active recombinant glutamic acid de-carboxylase. J. Biotechnol. 61, 57–68 (1998).
19. Qu, K., Martin, D.L. & Lawrence, C.E. Motifs and structural fold of the cofactorbinding site of human glutamate decarboxylase. Protein Sci. 7, 1092–1105(1998).
ART IC L E S
NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 4 APRIL 2007 285
20. Kash, S.F., Tecott, L.H., Hodge, C. & Baekkeskov, S. Increased anxiety and alteredresponses to anxiolytics in mice deficient in the 65-kDa isoform of glutamic aciddecarboxylase. Proc. Natl. Acad. Sci. USA 96, 1698–1703 (1999).
21. Baekkeskov, S. et al. Identification of the 64K autoantigen in insulin-dependentdiabetes as the GABA-synthesizing enzyme glutamic acid decarboxylase. Nature347, 151–156 (1990).
22. Pearce, D.A., Atkinson, M. & Tagle, D.A. Glutamic acid decarboxylase autoimmunity inBatten disease and other disorders. Neurology 63, 2001–2005 (2004).
23. Raju, R. et al. Analysis of GAD65 autoantibodies in Stiff-Person Syndrome patients.J. Immunol. 175, 7755–7762 (2005).
25. Bertoldi, M., Gonsalvi, M., Contestabile, R. & Voltattorni, C.B. Mutation of tyrosine332 to phenylalanine converts dopa decarboxylase into a decarboxylation-dependentoxidative deaminase. J. Biol. Chem. 277, 36357–36362 (2002).
27. Papakonstantinou, T., Law, R.H., Gardiner, P., Rowley, M.J. & Mackay, I.R. Compara-tive expression and purification of human glutamic acid decarboxylase from Sacchar-omyces cerevisiae and Pichia pastoris. Enzyme Microb. Technol. 26, 645–652(2000).
28. Evans, P. Scaling and assessment of data quality. Acta Crystallogr. D Biol. Crystallogr.62, 72–82 (2006).
29. Leslie, A.G.W. Joint CCP4 + ESF-EAMCB Newsletter on Protein Crystallography No. 26(1992).
30. Potterton, E., Briggs, P., Turkenburg, M. & Dodson, E. A graphical user interface to theCCP4 program suite. Acta Crystallogr. D Biol. Crystallogr. 59, 1131–1137 (2003).
31. Collaborative Computational Project, Number 4. The CCP4 suite: programs for proteincrystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763 (1994).
32. McCoy, A.J., Storoni, L.C. & Read, R.J. Simple algorithm for a maximum-likelihoodSAD function. Acta Crystallogr. D Biol. Crystallogr. 60, 1220–1228 (2004).
33. Jaroszewski, L., Rychlewski, L., Li, Z., Li, W. & Godzik, A. FFAS03: a server for profile–profile sequence alignments. Nucleic Acids Res. 33, W284–W288 (2005).
34. Canutescu, A.A., Shelenkov, A.A. & Dunbrack, R.L., Jr. A graph-theory algorithm forrapid protein side-chain prediction. Protein Sci. 12, 2001–2014 (2003).
35. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Refinement of macromolecular structuresby the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255(1997).
36. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. ActaCrystallogr. D Biol. Crystallogr. 60, 2126–2132 (2004).
37. Morris, R.J., Perrakis, A. & Lamzin, V.S. ARP/wARP and automatic interpretation ofprotein electron density maps. Methods Enzymol. 374, 229–244 (2003).
38. DeLano, W.L. The PyMOL User’s Manual (DeLano Scientific, San Carlos, California,USA, 2002).
40. Martin, S.B. & Martin, D.L. Stimulation by phosphate on the activation of glutamateapodecarboxylase by pyridoxyl-5¢-phosphate and its implications for the control ofGABA synthesis. J. Neurochem. 33, 1275–1283 (1979).
41. Barton, G.J. ALSCRIPT: a tool to format multiple sequence alignments. Protein Eng. 6,37–40 (1993).
ART IC L E S
286 VOLUME 14 NUMBER 4 APRIL 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY
