document

letters

Crystal structure of abifunctional transformylaseand cyclohydrolase enzymein purine biosynthesisSamantha E. Greasley1, Patricia Horton1, Joseph Ramcharan2, G. Peter Beardsley3, Stephen J. Benkovic2 and Ian A. Wilson1

1Department of Molecular Biology and The Skaggs Institute for ChemicalBiology, The Scripps Research Institute, 10550 North Torrey Pines Road, LaJolla, California 92037, USA. 2Department of Chemistry, Pennsylvania StateUniversity, University Park, Pennsylvania 16802, USA. 3Departments ofPediatrics and Pharmacology, Yale University School of Medicine, 3094 LMP,333 Cedar Street, P.O. Box 208064, New Haven, Connecticut 06520-8064, USA.

ATIC, the product of the purH gene, is a 64 kDa bifunctionalenzyme that possesses the final two activities in de novo purinebiosynthesis, AICAR transformylase and IMP cyclohydrolase.The crystal structure of avian ATIC has been determined to1.75 Å resolution by the MAD method using a Se-methioninemodified enzyme. ATIC forms an intertwined dimer with anextensive interface of ∼ 5,000 Å2 per monomer. Each monomeris composed of two novel, separate functional domains. The N-terminal domain (up to residue 199) is responsible for theIMPCH activity, whereas the AICAR Tfase activity resides inthe C-terminal domain (200–593). The active sites of theIMPCH and AICAR Tfase domains are ∼ 50 Å apart, with nostructural evidence of a tunnel connecting the two active sites.The crystal structure of ATIC provides a framework to probeboth catalytic mechanisms and to design specific inhibitors foruse in cancer chemotherapy and inflammation.

ATIC is a bifunctional enzyme involved in thepurine biosynthesis pathway. One of the activi-ties of ATIC is AICAR transformylase (AICARTfase), which catalyzes the formylation of 5-aminoimidazole-4-carboxamide-ribonucleotide(AICAR) by N-10-formyl-tetrahydrofolate (10-formyl-THF) to produce formyl-AICAR(FAICAR) and THF (Fig. 1a). ATIC also acts asan IMP (inosine monophosphate) cyclohydro-lase (IMPCH), which converts FAICAR to IMPin the final step of de novo purine biosynthesis.Because cancer cells rely more heavily on thede novo pathway than normal cells, which favorthe salvage pathway1 as their main purine

source, ATIC is an attractive candidate for rational design of anti-cancer agents. In particular, ATIC and other folate-dependentenzymes in purine and pyrimidine biosynthesis — such as, glyci-namide ribonucleotide transformylase (GAR Tfase), dihydrofolatereductase (DHFR) and thymidylate synthase (TS) — are targetsfor design of novel antifolate compounds.

Methotrexate (MTX), an inhibitor of DHFR, has been in com-mon clinical use in the treatment of malignant diseases for morethan 50 years. A number of antifolates against GAR Tfase, DHFRand TS are currently in clinical trials, including 6R-dideazate-trahydrofolate (DDATHF, lometrexol)2,3 and LY231514 (a multi-targeted antifolate)4. Polyglutamate metabolites of MTX inhibitother folate-dependent enzymes in the metabolic pathways, ofmethionine biosynthesis, thymidylate synthesis and de novopurine biosynthesis, as well as inhibiting its primary targetDHFR. This further emphasizes the need for continued study ofthese folate-dependent enzymes.

ATIC has also been implicated as a target for nonsteroid anti-inflammatory drugs (NSAIDs), such as sulfasalazine5, and isresponsible for the anti-inflammatory effect of low doses ofpolyglutamated MTX6. These inhibitors of AICAR Tfase, NSAIDsand MTX, cause intracellular accumulation of AICAR. This accu-mulation inhibits adenosine deaminase and adenosine kinase.This inhibition leads to increased levels of adenosine and ulti-mately gives rise to an anti-inflammatory response7,8. Hence,ATIC inhibitors have potential use in treatment of inflammatorydiseases, such as rheumatoid arthritis and inflammatory boweldisease.

The ATIC crystal structure was undertaken to decipher thespatial relationships of the two putative active sites and, conse-quently, the overall mechanism of catalysis. Sequence compar-isons revealed no significant homologies, including none withGAR Tfase, DHFR and TS. The IMPCH and AICAR Tfase activ-ities can be expressed as separate fragments of the entire protein,1–200 and 201–593 (numbering of avian protein)9, respectively,

a

402 nature structural biology • volume 8 number 5 • may 2001

Fig. 1 Structure and function of ATIC. a, Reactionscatalyzed by ATIC and the folate cofactor (10-f-THF).b, Stereo view of the crystal structure of the avATICdimer. Monomer A is colored in blue arrows (β-strands) and green ribbons (helices) with yellowconnecting residues, while monomer B is coloredred (β-strands), orange (helices) and purple (con-necting). The two purple spheres located near theC-termini represent bound potassium ions. Residue199, the last residue in the IMPCH domain, is labeledalong with the N- and C-termini. The bound GMP isshown in a ball-and-stick representation, and * indi-cates the approximate position of the phosphate ofbound ligands in the IMPCH (bottom, black) andAICAR Tfase (top, black and blue) active sites, whichare separated by ∼ 50 Å (black *).

b

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

© 2001 Nature Publishing Group http://structbio.nature.com

letters

nature structural biology • volume 8 number 5 • may 2001 403

precluding earlier suggestions that both activities reside within acommon active site10. However, the proximity of the two activesites could still facilitate substrate channeling or tunneling. Insolution, ATIC has been reported as a homodimer11,12, as amonomer9 or in monomer–dimer equilibrium13. The crystalstructure of ATIC was determined to address the issues of assem-bly and mechanism and to further the overall goal of elucidatingstructures for all of the enzymes in the de novo purine biosynthe-sis pathway14–20.

Overall structureRecently, we described the crystallization of avian ATIC (avATIC)that was expressed in Escherichia coli as a GST fusion protein21.The crystal structure of SeMet-labeled avATIC (see Methods) wasdetermined by multiple wavelength anomalous dispersion(MAD) at 1.75 Å resolution (Table 1). The avATIC is an exten-sively intertwined dimer, with the monomers (A, B) related by anapproximate noncrystallographic two-fold axis of symmetryalong the long axis of the molecule (Fig. 1b). Each monomer iscomposed of two distinct functional domains, as predicted fromATIC truncation studies9. Based on the crystal structure, the N-terminal domain, responsible for the IMPCH activity ofATIC13, is composed of ordered residues 4–199, whereas the C-terminal domain (residues 200–593) constitutes the region ofAICAR Tfase activity. A small substructure, consisting of two β-hairpin motifs that form an unusual, interdigitating eight-stranded β-sheet, acts as a bridge between the IMPCH andAICAR Tfase domains. A search for similar folds for either theATIC dimer or the intact monomer in the Protein Data Base

(PDB) using Dali22 did not reveal any significantsimilarities, even for other folate-binding proteins.

Dimer interfaceThe dimer interface is extensive, burying 4,929 Å2

(19%) of molecular surface area per monomer(using a 1.4 Å probe in MS)23, 1,900 Å2 of which isfrom the IMPCH domain and 3,000 Å2 from theAICAR Tfase domain. The ATIC dimer interface isnotably larger than the average of 3,500 Å2 for a64 kDa protein in other homodimers24. The ATICinterface also exhibits a high degree of shape com-plementarity (SC)25,26, with an SC value of 0.74(using a 1.7 Å probe), which is typical for oligomer-

ic proteins and protein–protein inhibitor interfaces (0.70–0.76).Taken together with reports of homodimeric avian11 and yeast12

ATIC and the high association (Kd = 240 nM) of the human ATIChomodimer13,27, the dimer constitutes the biologically active formof the enzyme.

IMPCH domainThe IMPCH domain is a three-layered α/β/α structure com-posed of a central five-stranded parallel β-sheet, with a strandorder of 5, 4, 1, 2, 3, resembling a Rossmann fold topology(Fig. 2a,b). The β-sheet is surrounded on one side by three α-helices (α1, α2 and α9) and by seven α-helices on the other.The IMPCH domain was compared to all other structures in thePDB using the program Dali22. The top Z-score of 8.9 (<2.0 isstructurally dissimilar), was observed for methylglyoxal syn-thase28 (PDB code 1B93), which shares only 19% sequence iden-tity and superimposes with a root mean square (r.m.s.) deviationof 3.0 Å for 114 structurally equivalent Cα atoms. Although thetopology of the IMPCH domain resembles that of a nucleotidebinding domain, the sequence does not contain any of the typicalconsensus motifs, such as the GXXGXGK (where X signifies anyresidue) fingerprint for binding mononucleotides29.

IMPCH active siteDuring refinement, distinct electron density was observed inmonomer A for a bound ligand that clearly resembled a purinenucleotide (Fig. 2c). This was particularly surprising as nonucleotides were added either during purification or during crys-tallization, suggesting that the ligand was acquired from the bac-

Fig. 2 Structure of the IMPCH domain. a, The IMPCHdomain of monomer A, residues 4–199. β-strands (blue)are labeled β1–β5 and α-helices (green) α1–α10. The GMP(see text) is shown in red. b, Cartoon of the topology ofthe IMPCH domain generated by the TOPS web server(www3.ebi.ac.uk/tops). Helices are shown as circles and β-strands as triangles. c, Cα trace of a close up view of theIMPCH active site of monomer A. The labeled residuesinteract with the bound purine nucleotide (blue), mod-eled as GMP. The electron density represents a 3Fo – 2Fc

map after the first round of refinement and is contouredat 1.2 σ. d, Superposition of the two IMPCH active sites inthe ATIC dimer. Monomer B (unliganded) is colored gray,while monomer A (liganded) is colored according to ther.m.s. deviation (0.2 Å (green)–2.5 Å (red)) of individualmain chain atoms between monomer A and B as deter-mined in ProFit (Martin, A.C.R., SciTech Software). The sidechains of Tyr 105 and Lys 67 are also shown. e, f, Solventaccessible surface area of the IMPCH binding site ofmonomer A (e) and monomer B (f). The monomer A and Bsurfaces are colored green and blue respectively.Conformational differences in residues surrounding thebinding site results in a large cavity in the unligandedstructure.

a b

c

e f

d

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters


terial cells with an affinity high enough to be car-ried throughout the purification procedure. Theligand was modeled as a molecule of GMP(Fig. 2c,d), which is in good agreement with theelectron density. Intriguingly, the amide group atthe C2 position of the purine ring is bent slightlyout of the plane of the ring.

The putative purine nucleotide is firmlyembedded in the active site (Fig. 2e) by an exten-sive network of hydrogen bond interactions.Alignment of 14 ATIC sequences in SWISS-PROTrevealed that highly conserved residues are clus-tered around the purine binding site, includingthose that form hydrogen bonds with thenucleotide.

Superposition of monomer A (liganded) and monomer B(unliganded) revealed surprising differences in the ligand bindingsite (Fig. 2d). In monomer A, Tyr 105 and Lys 67 are positionedover the putative purine ring, closing off the pocket such that the

GMP nucleotide is 96% buried (Fig. 2e). In contrast, equivalentresidues in monomer B have moved further apart, with backbonechanges up to 3.8 Å (Pro 106). The largest change (14.1 Å) is forthe side chain hydroxyl of Tyr 105, which is flipped out of thebinding pocket in the unliganded structure. Thus, the active site

of the unliganded structure forms a large opencavity rather than the closed, tightly packed bind-ing pocket of the liganded monomer (Fig. 2e,f). Allof the ligand contact residues come frommonomer A, except for a water-mediated interac-tion with Lys 138 of monomer B that contributesto one edge of the binding pocket. Why only onemonomer of the ATIC dimer binds ligand is notyet clear, but may indicate the possibility of nega-tive cooperativity, as in the half site reactivity ofthymidylate synthase30.

AICAR Tfase domainAt first glance, the AICAR Tfase domain appearsto consist of a central 11-stranded β-sheet sur-rounded by α-helices and a small subdomain.However, the automated domain assignment pro-

Fig. 3 The AICAR Tfase domain. a, Stereo view of thestructure of the AICAR Tfase domain colored by subdo-mains 2–4, as assessed by DOMID (Guogusng Lu, LundUniversity). The labeled residues correspond to the lastresidue of each domain. The position of the bound potas-sium ion is shown as a purple sphere (not shown isdomain 1, the IMPCH domain). b, Superposition of theCα trace of domain 2 (green) and domain 4 (yellow)showing internal duplication in the AICAR Tfase region. * indicates the location of domain 3 which was omittedfor clarity. Conserved β-strands are labeled 1–5, and α-helices are labeled A–D. The equivalent main chainatoms of the core structures of domains 2 and 4 superim-pose with r.m.s. deviations of 1.53 Å (monomer A) and1.54 Å (monomer B). c, Cartoon of the topology (as inFig. 2b) of domains 2 and 4. The regions that are struc-turally homologous have been labeled and colored as in(a,b).

a

b c

Fig. 4 The potassium ion binding site. a, Stereo view ofthe potassium ion binding site in the AICAR Tfasedomain. For clarity, side chain atoms have been omitted,except for Asn 432, which is involved in a cis–peptidebond; the two C-terminal histidine resides; and residuesthat bond (dashed lines with bond lengths in Å) via themain chain or side chain to the potassium ion. b, Stereoview of the 2Fo − Fc electron density map around theAsn 432 cis–peptide calculated after the multiple modelaveraging routine, but before automated main chaintracing in ARP/wARP40. The map is contoured at 2 σ(blue) and 4.5 σ (magenta).

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters

nature structural biology • volume 8 number 5 • may 2001 405

gram, DOMID (Guogusng Lu, Lund University) proposes thatAICAR Tfase has three structural domains (domains 2–4,Fig. 3a), in addition to the N-terminal IMPCH domain (domain1). Domain 2 is composed of residues 200–374, whereas domain4 (375–593) is interrupted by insertion of a small domain 3(469–532). Domains 2 and 4 have a similar topology, despite anyobvious internal sequence homology (Fig. 3b,c). This suggestsdomain duplication, which is also found in cytidine deaminasethat is involved in pyrimidine catabolism and has a similar fold22

to these AICAR Tfase subdomains (Z-score = 5.0). Domains 2and 4 have a common central five-stranded β-sheet, of strandorder 5, 4, 3, 1, 2, with strand 1 running antiparallel to the otherβ-strands (Fig. 3c). The β-sheet is surrounded by three α-helices(αB, αC and αD) on one side, while a long α-helix, αA, abuts theopposite face (Fig. 3c). Both domains begin with β-hairpinmotifs that form the interdigitating bridging β-sheet in thedimer. Domain 2 has an additional β-strand, β1′, that runsantiparallel to strand β5 and adopts a loop, and a short 310-helixin domain 4. Furthermore, domain 2 possesses an extra α-helix,αB′, followed by a 310-helix, that are both inserted between

strand β2 and helix αB (Fig. 3c). Thesetwo structural motifs are related by a rota-tion of 162° (representing an approximatepseudo dyad) about an axis perpendicularto the plane of the β-sheet that creates theextended central β-sheet (Fig. 3a).Domain 3, a 64-residue insertion indomain 4, is composed of four α-helicesand two short β-strands and has a uniquetopology, as assessed by Dali22. This regionexhibits the highest degree of thermalmobility in the structure, as reflected bythe B-values (Table 1).

While adding waters during refine-ment, two strong peaks were refined to B-values of 2 Å2 with large positive peaksin an Fo − Fc electron density map. SinceATIC exhibits higher activity in bufferscontaining either sodium or potassiumchloride9, two potassium ions (one permonomer) were subsequently refinedwith average B-values of 12 Å2, equal tothe surrounding protein. The potassiumion binds to domain 4 in the loop betweenhelix αA and strand β1, interacting withthe main chain carbonyls of Val 426,Thr 429 and Leu 590, the hydroxyl ofSer 431 and the side chain carboxyl ofAsp 540 (C-terminal end of β3) (Fig. 4a).The ligands coordinated to the potassiumdo not form any regular geometry, consis-tent with other potassium binding sites inthe Cambridge Structural Database(CSD) and the Protein Data Bank (PDB).Analysis of all potassium–oxygen andsodium–oxygen bond lengths in the CSDrevealed average bond lengths of 2.8 Åand 2.4 Å, respectively. The equivalentdistances between the carbonyl and car-boxyl oxygens in ATIC to the bound ionrange from 2.7 Å–3.1 Å, providing furthersupport for its assignment as potassium(Fig. 4a). The potassium ion may play a

key role in stabilization of the tertiary structure, particularly inmediating interactions of the C-terminal residues with a loopbetween helix αA and strand β1 in domain 4 (Figs 3a, 4a). Thisloop contains a nonproline cis peptide bond, between Ser 431and Asn 432, which has well defined electron density (Fig. 4b).Although nonproline cis peptides are rare in proteins (5.2% forXaa-Pro but 0.03% for Xaa-non-Pro)31, they are frequentlyfound adjacent to enzyme active sites31,32.

AICAR Tfase active sitePreliminary data of a complex of avATIC and a multisubstrateadduct inhibitor (MAI), β-DADF (not shown), designed tomimic both the AICAR substrate and the folate cofactor33, hasallowed unambiguous identification of the AICAR Tfase activesite at the dimer interface (Fig. 1b). Although little is knownabout the mechanism of the formyl transfer reaction, the loca-tion of the AICAR Tfase active site makes it difficult to under-stand how the enzyme could function as a monomer because theactive site appears to be assembled from components of bothsubunits. Indeed, recent steady state kinetic data suggest that

Table 1 Summary of crystallographic data

Data collectionWavelength (Å) 0.9799 (f′′ ) 0.9801 (f′) 0.9648Resolution (Å)1 50–1.75 (1.78–1.75) 50–1.85 (1.87–1.85) 50–1.75 (1.78–1.75)#observations 460,370 423,566 472,323#unique reflections 229,603 208,901 234,781Completeness (%) 87.9 (47.3) 94.3 (65.3) 90.1 (51.7)Rmerge

2 (%) 5.8 (47.5) 6.3 (57.2) 6.3 (44.9)I/σ 15.8 (2.4) 14.1 (2.0) 13.9 (2.2)Overall figure of merit

Before solvent flattening 0.60After solvent flattening 0.82

Refinement statisticsResolution (Å) 50–1.75Number of reflections 119,759Number in test set 11,992Rcryst

3 (%) 20.0Rfree

4 (%) 21.6Number of residues 1,180Number of ligand atoms 24 (GMP), 2 (K+)Number of waters 744Average B-values5 (Å2)

IMPCH domain 1 21.0:21.1AICAR Tfase domain 2 13.3:14.3AICAR Tfase domain 3 20.9:38.8AICAR Tfase domain 4 15.6:17.3Ligand [GMP:K+] 22.5:12.0Waters 25.1

Ramachandran Plot (%) R.m.s. deviationsMost favored 93.1 Bond lengths (Å) 0.006Additional allowed 6.5 Angles (o) 1.3Generously allowed 0.2 Dihedral (o) 22.3Disallowed 0.2 Improper (o) 0.8

1Numbers in parenthesis refer to the highest resolution shell.2Rmerge = [ Σh Σi | Ii(h) − < I(h) > | / Σh Σi Ii(h)] × 100, where < I(h) > is the mean of the I(h) observationof reflection h.3Rcryst = Σhkl | Fo – Fc | / Σhkl | Fo |4Rfree was calculated as for Rcryst but on 10% of the data excluded from the refinement45.5Values are given for monomer A: monomer B.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


letters


B gave an r.m.s. deviation for main chain and all atoms, respectively,of 0.61 Å and 0.99 Å, 0.26 Å and 0.68 Å for the AICAR Tfase domainsand 0.94 Å and 1.39 Å for the IMPCH domains. The larger r.m.s. devi-ations for the IMPCH domain are due to conformational changesfrom ligand binding to monomer A. Figs 1b, 2a,c,d, 3a,b and 4 werecreated with BOBSCRIPT43 and RASTER3D44. Figs 2e,f were madewith INSIGHT (Molecular Simulations, Inc.).

Coordinates. Coordinates have been deposited in the Protein DataBank (accession code 1G8M).

AcknowledgmentsThis work was supported in part by NIH Grants to I.A.W., S.J.B. and G.P.B. Wethank X. Dai and A. Heine for help with data collection and advice duringstructure determination, M. Rudolph and R. Stanfield for valuable advice andassistance in computational analysis, J. Vergis and K. Bulock for helpfuldiscussions, and the ALS staff of beamline 5.0.2 for guidance during datacollection. This is publication 13714-MB from The Scripps Research Institute.

Correspondence should be addressed to I.A.W. email: [email protected]

Received 11 December, 2000; accepted 2 March, 2001.

1. Jackson, R.C. & Harkrader, R.J. In Nucleosides and cancer treatment (edsTattersall, M.H.N. & Fox, R.M.) 18–31 (Academic Press, Sydney; 1981).

2. Beardsley, G.P., Moroson, B.A., Taylor, E.C. & Moran, R.G. J. Biol. Chem. 264,328–333 (1989).

3. Erba, E., Sen, S., Sessa, C., Vikhanskaya, F.L. & D’Incalci, M. Br. J. Cancer 69,205–211 (1994).

4. Mendelsohn, L.G. et al. Semin. Oncol. 26, 42–47 (1999).5. Baggott, J.E., Morgan, S.L., Ha, T., Vaughn, W.H. & Hine, R.J. Biochem. J. 282,

197–202 (1992).6. Allegra, C.J., Drake, J.C., Jolivet, J. & Chabner, B.A. Proc. Natl. Acad. Sci. USA 82,

4881–4885 (1985).7. Cronstein, B.N., Naime, D. & Ostad, E. J. Clin. Invest. 92, 2675–2682 (1993).8. Gadangi, P. et al. J. Immunol. 156, 1937–1941 (1996).9. Rayl, E.A., Moroson, B.A. & Beardsley, G.P. J. Biol. Chem. 271, 2225–2233 (1996).

10. Mueller, W.T. & Benkovic, S.J. Biochemistry 20, 337–344 (1981).11. Ni, L., Guan, K., Zalkin, H. & Dixon, J.E. Gene 106, 197–205 (1991).12. Tibbetts, A.S. & Appling, D.R. J. Biol. Chem. 275, 20920–20927 (2000).13. Beardsley, G.P. et al. Adv. Exp. Med. Biol. 431, 221–226 (1998).14. Krahn, J.M. et al. Biochemistry 36, 11061–11068 (1997).15. Wang, W., Kappock, T.J., Stubbe, J. & Ealick, S.E. Biochemistry 37, 15647–15662

(1998).16. Chen, P. et al. J. Mol. Biol. 227, 283–292 (1992).17. Li, C., Kappock, T.J., Stubbe, J., Weaver, T.M. & Ealick, S.E. Structure Fold Des. 7,

1155–1166 (1999).18. Mathews, I.I., Kappock, T.J., Stubbe, J. & Ealick, S.E. Structure Fold Des. 7,

1395–1406 (1999).19. Levdikov, V.M. et al. Structure 6, 363–376 (1998).20. Toth, E.A. & Yeates, T.O. Structure Fold Des. 8, 163–174 (2000).21. Reyes, V.M., Greasley, S.E., Stura, E.A., Beardsley, G.P. & Wilson, I.A. Acta

Crystallogr. D 56, 1051–1054 (2000).22. Holm, L. & Sander, C. J. Mol. Biol. 233, 123–138 (1993).23. Connolly, M.L. Science 221, 709–713 (1983).24. Jones, S. & Thornton, J.M. Proc. Natl. Acad. Sci. USA 93, 13–20 (1996).25. Lawrence, M.C. & Colman, P.M. J. Mol. Biol. 234, 946–950 (1993).26. CCP4. Acta Crystallogr. D 50, 760–763 (1994).27. Vergis, J.M., Bulock, K.G., Fleming, K.G. & Beardsley, G.P. J. Biol. Chem. 276,

7726–7733 (2001).28. Saadat, D. & Harrison, D.H. Structure Fold Des. 7, 309–317 (1999).29. Schulz, G. Curr. Opin. Struct. Biol. 2, 61–67 (1992).30. Anderson, A.C., O’Neil, R.H., DeLano, W.L. & Stroud, R.M. Biochemistry 38,

13829–13836 (1999).31. Jabs, A., Weiss, M.S. & Hilgenfeld, R. J. Mol. Biol. 286, 291–304 (1999).32. Weiss, M.S., Jabs, A. & Hilgenfeld, R. Nature Struct. Biol. 5, 676 (1998).33. Wall, M., Shim, J.H. & Benkovic, S.J. J. Med. Chem. 42, 3421–3424 (1999).34. Wall, M., Shim, J.H. & Benkovic, S.J. Biochemistry 39, 11303–11311 (2000).35. Otwinowski, Z. & Minor, W. Methods Enzymol. 276, 307–326 (1997).36. Matthews, B.W. J. Mol. Biol. 33, 491–497 (1968).37. Terwilliger, T.C. & Berendzen, J. Acta Crystallogr. D 52, 749–757 (1996).38. LaFortelle, E.D. & Bricogne, G. Methods Enzymol. 276A, 472–494 (1997).39. Abrahams, J.P. & Leslie, A.G.W. Acta Crystallogr. D 52, 30–42 (1996).40. Lamzin, V.S. & Wilson, K.S. Acta Crystallogr. D 49, 129–149 (1993).41. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Acta Crystallogr. A 47,

110–119 (1991).42. Brünger, A.T. et al. Acta Crystallogr. D 54, 905–921 (1998).43. Kraulis, P.J. J. Applied Crystallogr. 24, 946–950 (1991).44. Merritt, E.A. & Bacon, D.J. Methods Enzymol. 277, 505–524 (1997).45. Brünger, A.T. Nature 355, 472–475 (1992).

dimer formation is required for AICAR Tfase activity, butIMPCH activity can be observed for both monomer and dimer27,consistent with the crystallographic evidence for two indepen-dent IMPCH binding sites.

A remaining question was whether or not FAICAR, the prod-uct of the AICAR Tfase activity, can be channeled to the IMPCHactive site for conversion to IMP. The active sites of IMPCH andAICAR Tfase are separated by ∼ 50 Å (Fig. 1b) and provide nostructural evidence of a connecting tunnel, consistent withkinetic evidence that product channeling does not occur34 (G.P.Beardsley, pers. comm.).

Now that the active sites of ATIC have been identified, we arepoised to probe the mechanism of catalysis of the IMPCH and,in particular, of the challenging AICAR Tfase reaction. Crystalstructures of ATIC in complex with substrates, cofactors andinhibitors are in progress and, in combination with site-directedmutagenesis, should expedite this important goal. Such mecha-nistic and structural insights should aid in the design of specificinhibitors of ATIC as anti-cancer and anti-inflammatory agents.

MethodsProtein expression and purification. The plasmid, pET28a-avATIC encoding the N-terminal hexa-His-tagged avian ATIC, wastransformed into a methionine auxotrophic strain, E. coli B834(DE3)(Novagen), that was grown at 37 °C in methionine-deficient medi-um supplemented with 0.3 mM Se-Met (Sigma). After reaching anOD of 0.4 at 600 nm, recombinant protein expression was inducedwith 1 mM IPTG overnight, and cells were harvested by centrifuga-tion. ATIC was purified to homogeneity by nickel affinity (Qiagen)and Superdex 200HR (Amersham Pharmacia Biotech) chromatogra-phy. Purified enzyme, in 25 mM Tris, pH 7.4, 150 mM NaCl, 50 mMKCl, 5 mM DTT and 5 mM EDTA, was concentrated to 10 mg ml–1.The Se-Met protein is fully active and incorporation of the SeMetwas 96% as assessed by MALDI mass spectrometry.

Crystallization and data collection. ATIC was crystallizedovernight from 15% (w/v) PEG 8000, 0.2 M imidazole, pH 7.2, and5 mM DTT at 22 °C using vapor diffusion. Data for MAD phasingwere collected on a single crystal at three wavelengths (Table 1) to1.75 Å resolution at −180 °C on beamline 5.0.2 at the AdvancedLight Source (ALS) in Berkeley. Translation of the crystal in the X-raybeam after the second wavelength restored the initial resolution(Table 1). Data were processed and scaled using the HKL package35.The space group is P21 with unit cell dimensions a = 65.1 Å, b =106.0 Å, c = 103.5 Å and β=108.0°. The Matthews coefficient36 (Vm=2.6 Å3 Da–1) suggested two monomers in the asymmetric unit with asolvent content of 53%.

Structure determination and refinement. The positions of theSe atoms were determined in SOLVE37 and refined in SHARP38,resulting in a figure of merit (FOM) of 0.60 and 0.57 for centric andacentric data, respectively. Density modification, performed inSOLOMON39, improved the FOM to 0.82. ARP/wARP40 was used forfurther phase improvement and initial automated main chain trac-ing. The resulting electron density (Fig. 4b) was of excellent qualityand facilitated tracing of residues 6–406 and 411–593 for eachmonomer using O41. Refinement to 1.75 Å was performed in CNS42.One molecule of GMP was included in the refinement after the sec-ond cycle of manual rebuilding. Water molecules were added auto-matically using CNS42 and verified by manual inspection in O41. Thefinal refined structure (Table 1) is composed of all residues in eachmonomer, except for Lys 163, which was refined as alanine, residues1–3 and the preceding 20 residues (containing the hexa-His-tag andthrombin cleavage site). Superposition of monomer A on monomer

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om


document

Documents