x-ray structure of tmp kinase from mycobacterium...

doi:10.1006/jmbi.2001.4843 available online at http://www.idealibrary.com on J. Mol. Biol. (2001) 311, 87±100

X-ray Structure of TMP Kinase from Mycobacteriumtuberculosis Complexed with TMP at 1.95 AÊ Resolution

I. Li de la Sierra1, H. Munier-Lehmann2, A. M. Gilles2, O. BaÃrzu2

and M. Delarue1*

1UniteÂ de Biochimie Structuraleand2Laboratoire de ChimieStructurale desMacromoleÂcules, URA 2185 duC.N.R.S., Institut Pasteur, 28rue du Dr. Roux, 75724 ParisCedex 15, France

Present address: I. Li de la Sierrad'Enzymologie et de Biochimie StruBat. 34, Avenue de la Terrasse, 9119France.

Abbreviations used: TMPK, thymTMPKMtub, thymidylate kinase fromtuberculosis; TMPKEcoli, thymidylateEscherichia coli; TMPKYeast, thymidyyeast; CMPK, cytidylate kinase; UMkinase; AMPK, adenylate kinase; Nmonophosphate kinase; TMP, thymmonophosphate; TDP, thymidine dthymidine triphosphate; TB, tubercuimmunode®ciency virus; MIR, multreplacement; HSV, herpes simplex v

E-mail address of the [email protected]

0022-2836/01/010087±14 $35.00/0

The X-ray structure of Mycobacterium tuberculosis TMP kinase at 1.95 AÊ

resolution is described as a binary complex with its natural substrateTMP. Its main features involve: (i) a clear magnesium-binding site; (ii) analpha-helical conformation for the so-called LID region; and (iii) a highdensity of positive charges in the active site. There is a network of inter-actions involving highly conserved side-chains of the protein, the mag-nesium ion, a sulphate ion mimicking the b phosphate group of ATP andthe TMP molecule itself. All these interactions conspire in stabilizingwhat appears to be the closed form of the enzyme. A complete multia-lignement of all (32) known sequences of TMP kinases is presented.Subtle differences in the TMP binding site were noted, as compared tothe Escherichia coli, yeast and human enzyme structures, which have beenreported recently. These differences could be used to design speci®cinhibitors of this essential enzyme of nucleotide metabolism. Two casesof compensatory mutations were detected in the TMP binding site ofeukarotic and prokaryotic enzymes. In addition, an intriguing high valueof the electric ®eld is reported in the vicinity of the phosphate group ofTMP and the putative binding site of the g phosphate group of ATP.

# 2001 Academic Press

Keywords: crystal structure; rational drug design; AZTMP; thymidylatekinase; Mycobacterium tuberculosis
*Corresponding author
Introduction

The incidence of tuberculosis (TB) has beenincreasing during the last 20 years and it is nowthe ®rst cause of mortality among infectious dis-eases in the world, killing more than two millionpeople a year.1 Mycobacterium tuberculosis is the

, Laboratoirecturales, C.N.R.S.8 Gif/Yvette Cedex,

idylate kinase;Mycobacterium

kinase fromlate kinase fromPK, uridylate

MPK, nucleosideidineiphosphate; TTP,losis; HIV, humaniple isomorphousirus.

ing author:

principal microbial agent involved for humans.Tuberculosis is primarily transmitted via airborneaerosoled secretions. A peculiar aspect of its patho-genicity comes from the fact that it can remainquiescent and become active decades later. One ofthe most signi®cant risk factor for developingtuberculosis is human immunode®ciency virus(HIV) infection. The current treatment of active TBincludes four drugs (isoniazid, rifampicin, pyrazi-namide and ethambutol) for at least six months. Asigni®cant proportion of patients do not completethe therapy, especially in developing countries,and this has led to the appearance of resistantstrains of M. tuberculosis. Therefore, there is cur-rently a large effort to identify new potential tar-gets for inhibitors and to develop new antibiotics.

In this work, one essential enzyme of nucleotidemetabolism, namely thymidine monophosphatekinase (TMPK), is taken as a potential target fordeveloping rationally designed inhibitors.

TMPK (E.C.2.7.4.9, ATP:TMP phosphotransfer-ase) belongs to a large superfamily of nucleosidemonophosphate kinases (NMPK). It catalyses thephosphorylation of thymidine monophosphate

# 2001 Academic Press

http://www.idealibrary.commailto:[email protected]

88 X-ray Structure of M. tuberculosis Thymidylate Kinase

(TMP) to thymidine diphosphate (TDP) utilizingATP as its preferred phosphoryl donor.2 It lies atthe junction of the de novo and salvage pathwaysof thymidine triphosphate (TTP) metabolism and isthe last speci®c enzyme for its synthesis. Thesecharacteristics make the TMPK a good target forthe design of new antibiotics drugs.

The high-resolution structure of TMPK shouldbe a good starting point to devising novel inhibi-tory compounds using a structure-based drug-design approach. It may be worth mentioning herethat one of the most successful antiviral drugsagainst herpes simplex virus (aciclovir) is directedagainst thymidine kinase, which is responsible forthe synthesis of both TMP and TDP in cellsinfected by the virus.3,4

Here, we report the structure of the TMPK fromM. tuberculosis (TMPKMtub) bound to its naturalsubstrate, TMP, at 1.95 AÊ resolution. It is the ®rstNMPK reported structure from M. tuberculosis andfrom a pathogen in general. The structure, com-bined with a careful analysis of the alignment ofall known sequences annotated as TMPKs, allowsus to identify the residues involved in the TMP-binding site and those probably important for cata-lysis. The structure of the TMPKMtub catalytic siteis quite different from other bacterial or eukaryoticenzymes: arginine residues from both LID and P-loop regions are probably implicated in the phos-phoryl transfer, one magnesium ion is readily vis-ible in the active site and the LID region is in ahelical conformation characteristic of the closedform of the molecules of this family, even thoughthe second substrate is not present in the crystal.

Results and Discussion

Structure determination and overall description

The recombinant TMPKMtub in the presence ofTMP yielded crystals suitable for X-ray study.5 The

Table 1. Heavy-atom parameters, data-collection and phasing

nat1 EtHgPO4

Conc. (mM) - 2Soaking time (hrs) - 24X-ray source ESRF, ID14-3 LURE, DW21b

(wavelength (AÊ )) (0.94) (1.28)Detector MARCCD MAR IPUnit-cell parametersa b, c (AÊ ) 76.62, 134.38 76.84, 134.98

Resolution (AÊ ) 40-1.95 (2.0-1.95) 12-3.0 (4.0-3.0)Total number of refl. 184,676 22,643Total of unique refl. 17,694 (1209) 4571 (2630)Multiplicity 9.8 (8.2) 5.0 (4.8)Completeness (%) 100 (99.9) 90.7 (91.2)Rmerge

a 0.052 (0.227) 0.094 (0.120)I/s(I) 38.1 (8.6) 26.1 (19.5)Riso

b - 0.173Phasing power (Res. AÊ ) - 1.11 (9-3.8)

a Rmerge �h�ijIhi ÿ hIhij/�h�iIhi, where Ihi is the ith observation ofb Riso �jFPH ÿ FPj/�FP, where FPH and FP are the derivative and

crystal structure was determined by multiple iso-morphous replacement (MIR) using ®ve heavy-atom derivatives, with one of them giving reliablephase information up to 2.7 AÊ resolution (Table 1).The initial 2.7 AÊ MIR map was improved by den-sity modi®cation techniques and allowed the con-struction of the model without ambiguity. Themodel was re®ned to a crystallographic R-factor of21.6 % (Rfree 25 %; Table 2). The quality of the struc-ture was assessed using PROCHECK6 and gives anoverall G-factor of 0.24. TMPKMtub has 214 residuesand a molecular mass of 24 kDa. The re®nedstructure consists of 208 amino acid residues (thesix C-terminal residues were not observed in theelectron density map), one TMP molecule , onesulphate group and one magnesium ion bound tothe catalytic domain (Figure 1(a)).

The strictly conserved residue Arg95 (located inthe consensus sequence DR(Y/F/H), residues 94-96 in the M. tuberculosis amino acid sequence;Figure 2) is the only non-glycine residue lying out-side the allowed regions of the Ramanchandranplot. This particular conformation results from itslocation in the catalytic site, Arg95 being in directcontact with the phosphate moiety of the TMPmolecule in our complex structure (see below). TheM. tuberculosis enzyme also has one cis-residueconformation at the strictly conserved proline resi-due of motif (F/E)P at position 37, which is alsoobserved in the cis conformation in all TMPK struc-tures reported so far.7

The global folding of the M. tuberculosis enzymeis similar to that of the other TMPKs, despite thelow degree of similarity of their amino acidsequences (26 %, 25 % and 22 % sequence identityover about 200 aligned residues with TMPKEcoli,TMPKYeast and TMPKhuman, respectively).TMPKMtub has nine a-helices, surrounding a ®ve-stranded b-sheet core conserved in all TMPKs.Comparing the TMPKMtub-TMP structure with

statistics

SmCl3 K2PtCl4 UO2(acetate)25I-dUMP

12 1 1 272 20 24 24 � 3

ESRF, ID14-3 Rigaku ESRF, ID14-3 LURE, DW32(0.94) (1.5418) (0.94) (0.96)

MARCCD MAR IP MARCCD MAR IP

76.51, 134.84 75.71, 133.89 76.47, 134.55 76.46, 133.6540-1.91 (1.95-

1.91) 20-3 (3.5-3) 40-2.4 (2.46-2.4) 12-3.0 (3.11-3)19,1371 20,431 79,790 34,220

18729 (1214) 4797 (1718) 9333 (635) 4179 (405)10.2 (8.4) 4.3 (4.4) 8.6 (6.1) 8.2 (8.3)99.3 (99.9) 97.1 (96.8) 96.2 (100) 84.7 (86.5)

0.094 (0.242) 0.133 (0.286) 0.042 (0.328) 0.127 (0.159)37.2 (6.4) 14.1 (6.1) 39.2 (10.9) 10.6 (4.6)

0.132 (15-3.8) 0.075 (12-4) 0.166 (29-2.5) 0.119 (12-3.8)1.13 (15-3.8) 0.9 (10-4) 1.2 (15-2.7) 1.09 (15-3.8)

the re¯ection h, while hIhi is the mean intensity of re¯ection h.the native structure-factor amplitudes, respectively.

Table 2. Re®nement statistics

Resolution range (AÊ ) 20-1.95No. of reflections

Used for refinement 17,326Rfree calculation 889

No. of non-hydrogen atomsProtein 1543TMP 21SO4

2ÿ (two molecules) 10Mg2 1Water 150

Model parametersRfactor

a (%) 21.59Rfree

b (%) 25.01rmsd from ideality

Bond lengths (AÊ ) 0.005Bond angle distances (AÊ ) 0.017

Average temperature factorsMain-chain 24.4Side-chains 26.8TMP 21.8SO4

2 ÿ (molecule 2) 18.9 (23.4)Mg2 37.5Water 37.9

Ramanchandran plotResidues in most favoured regions (%) 96.6Residues in additional allowed regions (%) 2.8Overall G-factorc 0.24

a Rfactor � jjFoj ÿ jFcjj/jFoj.b Rfree was calculated with a small fraction (5 %) of randomly

selected re¯ections.c G-factor is the overall measure of structure quality from

PROCHECK.6

Figure 1. TMPKMtub-TMP binary complex structure.(a) Ribbon diagram of the protein complexed with TMP,the sulphate and the magnesium ions bound to theactive site, showing in red the LID and the P-loopregions. (b) Superposition of the Ca traces of TMPKMtub-TMP (red), TMPKYeast-TMP (green) and TMPKEcoli-TP5A(blue) structures. The LID region adopts a helical confor-mation in the TMPKMtub-TMP complex, in contrast to adisordered region in the TMPKYeast-TMP complex.

X-ray Structure of M. tuberculosis Thymidylate Kinase 89

both the similar structure from yeast (1tmk) or theTMPKYeast-TP5A (3tmk) and TMPKEcoli-TP5A(4tmk) structures results in good overlap of theb-sheet core (rmsd for ®ve strands, 23 Ca atoms,0.74 AÊ , 0.71 and 0.81 AÊ for 1tmk, 3tmk and 4tmk,respectively) but the surrounding helices and loopshave moved signi®cantly relative to each other (for163 Ca superposing atoms, the rmsd is 10.89 AÊ ,10.93 AÊ and 8.44 AÊ for 1tmk, 3tmk and 4tmk,respectively; see Figure 1(b)). This, together withthe low level of sequence identity conservation,could explain the failure of the molecular replace-ment attempts to solve TMPMtub structure usingboth yeast and Escherichia coli models and the needfor an MIR structure determination.

The TMP kinase family

All known TMPK sequences have been alignedtogether in Figure 2. This includes 17 bacterialenzymes, seven archaebacterial enzymes and eighteukaryotic enzymes. The alignment has beenadjusted manually, especially in the LID region, totake into account all the available structural infor-mation; it becomes less certain after residue 180 orso, rendering the identi®cation of any residueessential for adenine recognition dif®cult. Themultialignment provides enlightening functionalinformation by looking at strictly conserved resi-dues and at those that are strictly conserved insidesubfamilies and systematically changed from one

Figure 2 (legend shown on page 92)


subfamily to the other; they are coloured yellow inFigure 2. It then becomes necessary to look at allavailable three-dimensional structures to under-stand the origin of this phenomenon.

The three-dimensional structures of TMPK fromyeast,8,9 E. coli10 and more recently human7 havebeen solved and show a similar fold to otherNMPKs; namely, a core of ®ve-stranded parallel

Figure 2 (legend shown on page 92)


b-sheets surrounded by nine a-helices. Threeregions contain the essential residues for the func-tion of this enzyme (Figure 1(a)): ®rst, the P-loopmotif (consensus sequence GxxxxGKS/T),11 whichcontrols the positioning of the phosphoryl groups

of the phosphate donor; second, the loop contain-ing the strictly conserved arginine residue thatbrings the donor and the acceptor nucleotidestogether with consensus sequence DR(Y/H/F),12

and third the LID region, a ¯exible stretch that


closes on the phosphoryl donor when it binds (seealso Via et al.13 for a recent review of thesesequence motifs).

In addition to these functionally essential regionsjust mentioned, the most striking features revealedby the multialignment are the (F/E)P sequencemotif at the junction between strand b2 and helixa2, the KPD motif just before the b4 strand and thestrictly conserved serine residue at position 99 (seeFigure 2), which, to our knowledge, has remainedunnoticed so far.

On the basis of the location of the active-sitearginine residues in either the P-loop or LIDsequences, TMPKs were categorized into two type-s.10 Type I TMPKs (e.g. yeast and human) have, inaddition to the invariant lysine (residue 13 in theM. tuberculosis amino acid sequence), an invariantbasic residue at position 10 in their P-loopsequence that can interact with the g phosphategroup of ATP and lack such a positively chargedresidue in the LID region. In contrast, type IITMPKs (e.g. E. coli), have a glycine residue in theP-loop at position 10 and one additional basic resi-due in the LID region that interacts with ATP(Arg153); however, this last residue is not strictlyconserved in prokaryotes or in archaebacteria(Figure 2).

It came as a surprise, however, that the multia-lignment indicated that the M. tuberculosissequence is actually closer to the eukaryoticenzymes (including the viral sequences) and poss-ibly the archeal enzymes than to the bacterialenzymes. This is especially true in the region justupstream from motif KPD (a6-b4) and in the loopbetween helix a2 and strand b2 (motif FP in eukar-yotes and EP in prokaryotes), as well as in theregion just downstream from the LID region (helixa8, especially residue Q172). However, the LIDregion itself clearly resembles more closely the bac-terial ones, albeit with a characteristic arginine andaspartate-rich large insertion. The P-loop is quite

Figure 2. Multialignment of all known TMPK sequences.package GCG (version 9.1),37 with opening and extension gadjusted manually in the LID region and in the helix a8 rother known structures (E. coli, yeast, human TMPKs). Theaddition to the generic name kthy (which stands for thymidymyctu M. tuberculosis; mycle M. leprae; helpj, Helicobacter pmycpn, Mycoplasma pneumoniae; ecoli, E. coli; yerpe, Y. pestiBacillus subtilis; ricpr, Rickettsia prowazekii; chlpn, Chlamydia pcrescentus; aerpr, Aeropyrum pernix; syny3, Synechocystis sp.; aArcheoglobus fulgidus; metja, Methanococcus janaschii; pyrho,Methanobacterium thermoautotrophicum; schpo, Schizosaccharomvirus; variv, Variola virus; human, Homo sapiens; mouse, Mswine fever virus. Residues are boxed if more than 50 % idenitions are in bold red. Residues that are more than 80 % idengroup to the other are indicated in yellow. Four groups of setwo sequences of M. tuberculosis and M. leprae come ®rst, th15 bacterial sequences, then the last eukaryotic eight sequenary structures of the M. tuberculosis enzyme are indicated atgreen upper triangles for residues involved in TMP bindinbinding and blue crosses for the ones involved in the Mg2 b

peculiar: it does not have the positive charge of theeukaryotes in position 10 (it has glycine instead, asin prokaryotes), but it has another extra arginineresidue at position 14. Therefore, the M. tuberculosisenzyme appears, together with its closely relatedcousin Mycobacterium leprae, as unique among allknown TMPK sequences14 with a possible horizon-tal gene transfer event with eukaryotes. Whateverthe reality of this gene transfer, the same eventhappened in M. leprae.

Dimerization mode

TMPKMtub, as well as TMPKYeast or TMPKEcoli, isa homodimer in solution. The TMPKMtub com-plexed with TMP crystallizes with one moleculeper asymmetric unit; the functional dimer can berecovered by using the symmetry operator x ÿ y,ÿy, ÿz. The interface of the dimer consists of threepairs of helices (a2, a3 and a6) as observed inE. coli enzyme complexes, which also crystallizeswith one molecule per asymmetric unit.10 Thesame helices are observed in the interfacial zone ofTMPKYeast complexes, which crystallize either withone dimer8 or four dimers9 per asymmetric unit; infact, the E. coli and the yeast dimers are remark-ably superimposable. However, having best super-imposed the ®rst monomers of all three enzymes,the second monomer of M. tuberculosis appears tobe oriented upside-down as compared to either itsyeast or E. coli counterparts. In other words,whereas the 2-fold axis of TMPKEcoli or TMPKYeastdimer runs orthogonal to helix a3, across the dimerinterface, relating a2, a3 and a7 to their antiparallelequivalents, the 2-fold axis of TMPKMtub runs par-allel to a3, relating a2, a3 and a6 to a60, a30 anda20, respectively. The dimer interface contains aclosely packed hydrophobic core but also an ionpair Glu50-Arg127, which could explain the originof this different dimerization mode, since it is not

The alignment was done using the program PILEUP ofap penalties of 6 and 2, respectively. The alignment wasegion, taking into account the structural elements of all

Figure was generated with the program ESPript.22 Inlate kinase), the following abbreviations have been used:ylori J99; helpy, H. pylori; mycge, Mycoplama genitalis;s; haein, H. in¯uenzae; bucai, Buchnera aphidicola; bacsu,neumoniae; chltr, Chlamydia trachomatis; caucr, Caulobacterquae, Aquifex aeolicus; thema, Thermotoga maritama; arcfu,Pyrococcus horikoshii; sulso, Sulfolobus solfataricus; metth,yces pombe; yeast, Saccharomyces cerevisiae; vaccv, Vaccinia

us musculis; caee, Caernorhabditis elegans; asfb7, africantity was reached at this position. Strictly conserved pos-

tical in each group de®ned below, but different from onequences have been de®ned in the program ESPript:22 theen the seven archaebacterial sequences, then a group ofces as the fourth group. The numbering and the second-the ®rst line. Special symbols were added in the last line:g, red ellipses for residues putatively involved in ATPinding.

Figure 3. (a) The sulphate ion-binding site, showingthe bridge between the LID region (residues 149 and153) and the P-loop (residues 13 and 14). (b) The Mg2 -binding site, showing that the Mg2 ligands (residues166 and 9, the 50 phosphate oxygen and water mol-ecules) are arranged in an octahedral con®guration.


conserved in enzymes other than M. tuberculosis orM. leprae.

Stabilization of the structure of the LID region

One of the main characteristics of the TMPKMtub-TMP complex structure is that the LID region isobserved in an a-helix conformation, even thoughthe ATP binding site is unoccupied.

The LID segment is described in the otherknown NMPK structures as highly ¯exibleand undergoes substantial conformationalchanges when the ATP molecule is ®xed to theenzyme.15 ± 17 Similarily, there is a transitionbetween a coil to an helical conformation in theLID region of TMPKYeast when ATP is bound.

9 Inall known ternary complexes of TMPKEcoli, it isalso structured as an a-helix because ATP ana-logues were present in the active site (no structureof the E. coli enzyme with an empty ATP-bindingsite has been reported so far).

To explain the stability of the LID region in ourstructure, which lacks the phosphate donor, therole of a sulphate ion in the active site should beemphasized. Indeed, this sulphate ion partlyexplains the structural ordering of the LID region,because it provides a direct link between Arg153 ofthe LID and Lys13 of the P-loop (see Figures 3(a)and 4(a)); in a second shell of interactions (5-6 AÊ ),it is at the center of a galaxy of positive charges(Arg14, Arg95, Arg149 and Arg160). This sulphateion is in the position expected to be occupied bythe b phosphate group of ATP and this is a recur-rent situation that has been observed in manyNMPKs. In thymidine kinase, the presence of thissulphate ion has been proved to be linked directlyto the a-helix structuring of the LID region throughmolecular dynamics simulations (M. Orozco et al.,personal communication). There is yet another linkbetween the LID region and the P-loop throughthe magnesium ion, of which two of the ligandsare Glu166 and Asp9 (from the LID region and theP-loop, respectively); this is described in moredetail below (Figures 3(b) and 4(b)).

Magnesium ion

The second main structural characteristic of theM. tuberculosis structure is that it contains a mag-nesium site, even though no ADP or non-hydroly-sable analogue of ATP is present in the crystal. Thechemical nature of this positive peak in the initialelectron density maps was inferred from its octa-hedral coordination (all the distances between themagnesium ion and its oxygen ligands are in therange 2.2-2.25 AÊ ) and the fact that it is the onlydivalent cation present in millimolar amounts inthe mother liquor.

The magnesium ion observed here is a uniquefeature of the TMPKMtub structure. No magnesiumatom was reported for the yeast or the E. colienzyme.8 ± 10 For the human enzyme, one mag-nesium binding site has been reported but it is

located further away along the ATP-binding site,between the b and g phosphate groups.7

Three water molecules (W1009, W1018 andW1050) and an oxygen atom from the phosphorylgroup are the other ligands that coordinate theMg2 in an octahedral con®guration (Figures 3(b)and 4(b)), in addition to the carboxylate oxygenatoms of Asp9 and of Glu166 already mentioned.One of these water molecules is in direct contactwith a carboxylate oxygen atom of Asp163.

M. tuberculosis TMP kinase is in the fullyclosed conformation

NMPKs have been described in several states,especially the AMPKs:16 an open conformation isobserved without substrate, a partially closed con-formation with a single substrate and a fully closedconformation in the presence of both substrates.Both open and partially closed conformations were

Figure 4. (a) A schematic drawing of the TMP-bindingsite in the Mg2 and sulphate ion region (CHEM-DRAW), displaying all the residues in direct contactwith the substrate. (b) As in (a), with the base moiety ofthe TMP-binding site.

Figure 5. (a) A drawing of the thymidine moiety-binding site in TMPKMtub. (b) Network of (hydrogen-bond) interactions involved in the stabilization of 30OHgroup of the TMP molecule.


observed for the CMPK from E. coli.18 Fully closedconformations were observed for the UMPK-CMPK from yeast and Dictyostelium discoideum.19,20

Finally, partially closed8 and fully closed7,9,10 con-formations were observed for the TMPK fromE. coli, yeast and human, respectively.

The presence of the magnesium ion and of thesulphate ion already pointed to an explanation forthe structuring and the closing of the LID region.Here, we argue that the enzyme is in its fullyclosed conformation, even though the second sub-strate is not bound; indeed, the 30OH group of thesugar moiety of TMP is at the center of an exten-sive hydrogen bond network, which can bedescribed as follows. As mentioned earlier, Asp9 isessential in holding this 30OH in place, but theAsp9 carboxylate group is in turn hydrogenbonded to Tyr103 and Glu172 (Figure 4(a)), whichare all strictly conserved (Figure 2). The hydroxylgroup of the Tyr103 side-chain is in turn hydrogenbonded to Arg95, as in the human enzyme,7 andthe Gln172 side-chain OE1 atom is located 3.1 AÊ

away from the main-chain nitrogen atom of Asp9;Arg95 is itself maintained in place through thestrictly conserved Ser99, thus completing this intri-cate network (see Figure 5(b)). All these inter-actions conspire to make the TMPKMtub structurepresented here as most closely related to the fullyclosed TMPK structure.

The TMP-binding site

As mentioned earlier, TMP is essential for thecrystallization of TMPKMtub; the experimental elec-tron density of the TMP molecule could be seen inthe initial MIR-DM maps. Figure 6(a) shows theelectron density in a 2Fobs ÿ Fcalc map using amodel where the TMP molecule has been omitted.

There are three main interactions that character-ize the TMP binding (Figures 4 and 5): (i) a stack-ing interaction involving the pyrimidine ring andthe Phe70 side-chain; (ii) the interaction withTyr103, which helps select deoxy-ribonucleotidesversus ribonucleotides; (iii) and the hydrogen bondbetween the O4 atom in the base moiety and the

Figure 6. (a) Model of bound TMP in 2Fo ÿ Fc map;the TMP molecule was omitted from the model used tocalculate the phases. Contours are drawn at the 2s levelabove the mean electron density. (b) Fourier-differenceF5I-dUMP ÿ Fc map. Fc denotes structure factors ampli-tudes of the TMPKMtub model without TMP. The peakin the electron density map corresponding to an iodineatom in position 5 of the pyrimidine ring is observedwith a cut-off of 5s above the mean electron density.The Figure was drawn with BOBSCRIPT.38


Arg74 side-chain, which favours thymidine or ura-cil over cytosine (Figures 4(b) and 5(a)). Thesethree positions are almost universally conserved inall known TMPK sequences (see Figure 2). The roleof the last interaction in the kinase activity wascon®rmed recently in the case of the thymidinekinase from herpes simplex virus (HSV) type 1:steady-state kinetic studies showed that mutatingGln125 (Arg74 equivalent residue in the TMPKMtubsequence) into Glu, Asp or Asn in the thymidinekinase from HSV-type 1 has a devastating effect onthe phosphorylation of TMP, the thymidylatekinase activity of all three mutants being decreasedby over 90 %.21

Apart from forming an ion pair with Glu124, theArg74 side-chain is stabilized through stacking

interaction of its hydrophobic part with Phe36, aresidue strictly conserved in M. tuberculosis,M. leprae and eukaryotes and located just beforethe structurally important Pro37, which is strictlyconserved in all TMPKs and whose cis confor-mation forms the very bottom of the cavity respon-sible for the thymidine binding (Figure 5(a)). It isrewarding to ®nd that this hydrophobic interactionis replaced in prokaryotes by a (compensating)direct hydrogen bond between the NE atom ofArg74 and the carboxylate group of the strictlyconserved glutamate residue that replaces Phe36 inall prokaryotes, while the interaction with Glu124is replaced by an interaction with Thr105, Asp102and Tyr75 (E. coli numbering). Therefore, the situ-ations of Arg74 in prokarotes in eukaryotes arealmost mirror images of one another, withTMPKMtub behaving as a eukaryotic TMPK.

All direct contacts between TMP and TMPKMtubare shown in Figure 4(b). There are 11 residueslocated less than 3.9 AÊ away from the TMP mol-ecule, and six of these make hydrogen bonds withthe nucleotide: Asp9, Phe36, Tyr39, Arg74, Arg95and Asn100. Also Ser99, a strictly conserved resi-due that had hitherto escaped notice, is crucial inthe positioning of Arg95 (Figure 5(b)). Severalsubtle differences are observed when comparingwith other TMPKs. This concerns: (i) the hydrogenbond between the N3 atom of the pyrimidine ringwith the Asn100 side-chain (changed to a glycineresidue in both yeast and human, and to a threo-nine residue in the E. coli enzyme); (ii) the Tyr39residue makes two polar contacts with the TMP,one with a phosphate oxygen atom (3.1 AÊ ) and theother with the oxygen atom at the 50 position(3.4 AÊ ). Arginine and glycine residues replace thistyrosine in human and yeast enzymes respectively.(iii) The 30-hydroxyl group of the ribose moietymakes three polar contacts (Figure 4(b)), one witha water molecule (2.72 AÊ ) involved in the Mg2

coordination (see above) and two others (2.72 AÊ

and 3.3 AÊ ) with Asp9 of M. tuberculosis sequence.The direct interaction of the P-loop with the sugarmoiety of the monophosphate substrate is a uniquefeature of TMPKs.8 It is interesting to note thatresidues Asn100 and Tyr39 are precisely thosepainted in yellow in the multialignement drawnwith EPScript22 using the option of highlightingresidues conserved in subfamilies but changedfrom one subfamily to the other (Figure 2).

Interaction with othernucleoside monophosphates

TMPKMtub phosphorylates dUMP analogues,14 as

was observed for other TMPKs.8,23 Substitution ofthe methyl group of position 5 of the pyrimidinering by iodine (5I-dUMP) affects both the Km andthe Vm parameters of the enzyme: the Km value is3.5-fold higher than that of TMP and reaction rateis 70 % of that with TMP (see Table 3). Moreover,both 5I-dUMP and TMP bind to the phosphateacceptor binding site in a very similar fashion, as

Table 3. Steady-state kinetic parameters

Yeasta E. colib Y. pestisb M. tuberculosisc

Km for TMP (mM) 9 15 45 4.5Km for 5I-dUMP (mM) - - - 140Vm (ATP, 5I-dUMP) (mM/min mg) - - - 7.5Vm (ATP, TMP) (mM/min mg) 84 50 25 13kcat (s

ÿ1) with TMP and ATP 35 10.5 9.6 4.5Km for AZTMP (mM) 6 170 90 -kcat for TMP/kcat for AZTMP 200 2.5 96 -Ratio kcat/Km for TMP AZTMP 133 27.5 192.8 -KI for AZTMP (mM) - - - 10

Results for E. coli, Y. pestis and Yeast enzymes were extracted from the following publications:a Lavie et al.10b Chenal-Francisque et al.23c Munier-Lehmann et al.14


shown in the experimental X-ray structure of theTMPKMtub-5I-dUMP complex (Figure 6(b)),obtained by soaking the original crystals in a2 mM solution of 5I-dUMP.

AZTMP was observed to be a competitive inhibi-tor of TMPKMtub with a KI of 10 mM: the presenceof an azido group totally abolishes the catalysiswithout changing the af®nity. It is the ®rstreported TMPK that does not phosphorylate theAZTMP molecule, in contrast to other TMPKsfrom prokaryotes or eukaryotes.14 AZTMP is asubstrate for the E. coli, Yersinia pestis and yeastenzymes (with a reduction of kcat of only 2.5-fold,96-fold and 200-fold for the E. coli, Y. pestis andyeast enzymes, respectively; see Table 3), and forSalmonella typhi and Haemophilus in¯uenzae.23

S. typhi and E. coli TMPKs phosphorylate AZTMPat comparable rates, whereas H. in¯uenzae enzymewas more similar to the Y. pestis TMPK.

Even though we have not been able to exchangethe TMP for AZTMP in our original crystals in con-ditions similar to those leading to the exchange ofTMP with 5I-dUMP, nor to grow large enough co-crystals, we can still postulate a plausible expla-nation for the inhibitory effect of AZTMP. By look-ing at the structure of the active site, it is possibleto imagine a direct interaction between the mag-nesium ion and a modelled azido group ofAZTMP, as inspired by the superimposed structureof the AZTP5A-TMPKEcoli complex, because thedistance between the last nitrogen atom of theazido group and the cation can be reduced to only2.0 AÊ , after rotation around the C30-N30 bond. Thiswould displace one of the ligands of the mag-nesium ion and deeply perturb the geometry of theactive site, since this magnesium ion is involved inthe positioning of several key chemical groups forthe reaction, namely one of the phosphate oxygenatoms and the essential Asp9.

Inferring the ATP binding site

Despite the fact that TMPKMtub was co-crystal-lized as a binary complex with TMP, but in theabsence of ATP, two structural characteristicsobserved in the ternary complexes from yeast and

E. coli enzymes (when the ATP is ®xed) are alsoobserved in our structure. The side-chain of resi-due Arg149 in the LID helix is located as observedin the E. coli enzyme, namely opposite both theloop between b5 and a9 (comprising Leu 193 inE. coli) and the Thr15 residue (a highly conservedresidue in TMPKs), ready to interact by stackinginteraction with the adenine ring of ATP(Figure 7(a)). These three regions together form theapparent binding site of ATP in the complex withAZTP5A in E. coli (Figure 7(b)). Following both thestructural work on yeast and E. coli TMPK as wellas a recent article on sequence determinants map-ping the binding sites of both substrates inNMPKs,13 it is possible to identify a semi-invariantsmall residue (alanine, serine, glycine, threonine) atposition 196 in the loop between helix a9 andstrand b5, common to all TMPKs (see Figure 2).Except for the fact that the loop between b5 and a9is much shorter in M. tuberculosis than in E. coli,everything is in place to accommodate the adeninemoiety of ATP in our structure.

Electrostatic potential in the active site

A number of positive or negative charges arepointing towards the active site of TMPKMtub. Thehelical conformation of the LID region allows side-chains of Arg153, Arg156 and Arg160 to be locatedaround the P-loop segment, which itself places twopositive charges in this region, thereby contribut-ing to create a highly positive electrostatic potentialand mapping directly the binding of the tripho-sphate moiety of the ATP molecule.

In addition, the Lys13 and Asp163 residuesinteract with the phosphoryl group of TMP viawater molecules W1014 and W1009, respectively.

On the other hand, two acidic residues, Asp9and Glu166, create a negative electrostatic potentialnear the TMP phosphate binding site, which con-tributes to the binding of the magnesium ion. Thiscreates a high gradient of the electric potential (i.e.electric ®eld) in the vicinity of the a phosphategroup of TMP and the g phosphate group of ATP.In fact, it could be calculated that this gradient issuch that one goes from the 30 kT/e to the ÿ30

Figure 7. (a) Putative ATP-binding site inTMPKMtub and (b) the observed AZTP5A-binding site inTMPKEcoli.

10 Only the adenine moiety of ATP is rep-resented. The side-chains of the a-helical LID regionhave been omitted for the sake of clarity, except forR149. A small adjustment of the most extreme part ofthis arginine side-chain is necessary in the M. tuberculo-sis enzyme to accommodate (and stack under) the ade-nine ring. The Figures were drawn with MOLSCRIPT.39

Figure 8. (a) Electrostatic potential surfaces in theM. tuberculosis enzyme in the vicinity of the TMP-ATP-binding sites, calculated with program Delphi,35 exclud-ing the magnesium and sulphate ions, at zero ionicforce. The blue surface represents the 30 kT/e potentialsurface and the red one the ÿ30 kT/e surface. The TMPmolecule and the magnesium ion are in ball and stickrepresentation (b) As in (a) for the E. coli enzyme, exceptthat the isopotential surfaces are contoured at 15 kT/eand ÿ20 kT/e, respectively. The AZTP5A molecule isalso included.


kT/e potential surface in less than 8 AÊ inTMPKMtub, creating an electric ®eld as high as 10

7

V cmÿ1 (Figure 8(a)). This is calculated in theabsence of the magnesium and sulphate ion, atzero ionic strength. Interestingly, this effect isqualitatively maintained in the E. coli enzyme,albeit with an amplitude divided by a factor of 2(Figure 8(b)). This may be the reason why a mag-nesium ion is observed in the M. tuberculosisenzyme and not the E. coli enzyme, because bind-ing of this cation by and large supresses this veryhigh electric ®eld. Nevertheless, should the mag-nesium ion move during catalysis, the electric ®eldcould develop again and be used to break downthe covalent bond between the b and g phosphategroups of ATP, and attract the displaced electrons

suf®ciently close to the phosphate group of TMP tomake a new covalent bond.

Possible catalytic residues

In AMPK and UMPK enzymes, the arginine resi-dues located in the LID region have been shown toplay a role in catalyzing phosphoryl transfer.20,24 InTMPKEcoli, the arginine residues located in the LIDregion could also play a similar role, i.e. stabilizethe transition state,10 but in the eukaryoticenzymes one of the key basic residues is located inthe P-loop and not in the LID region. In the human


enzyme, there seems to be a different mechanismat work.7

Apart from the side-chains of Arg153, Arg156and Arg160 from the LID region, the active site ofTMPKMtub also contains the Arg14 residue (in theP-loop segment). This last residue is not conservedand is replaced by a threonine residue in E. coliand yeast or a serine residue in human TMPK; it islocated at the N terminus of the a-helix 1. Its side-chain is located in such a position that it could beengaged in the binding of the g phosphoryl groupof the ATP molecule, as inferred from thecomplexes of E. coli and yeast enzymes with thebisubstrate inhibitor TP5A.

In addition to these residues, the TMP phosphor-yl group of the TMPKMtub structure makes directpolar contacts with three residues (Asp9, Tyr39and Arg95) and solvent-mediated contacts with sixresidues (Gly12, Lys13, Phe36, Arg153, Asp163 andGlu166) of the enzyme molecule (see Figure 4).These tyrosine and arginine residues are good can-didates in assisting the transfer of a phosphorylgroup to TMP: indeed, an arginine residue in theposition of Arg95 plays this role in the yeast, E. coliand human enzymes.7 ± 10 It follows that Tyr39,which is unique to TMPKMtub structure andsequence, stands out as the second possible targetfor the design of inhibitors speci®c to TMPKMtub, inaddition to Asn100 as already mentioned. It isreplaced by an arginine residue at the samesequence position (39) in eukaryotes or at position47 (M. tuberculosis numbering) in prokaryotes,where it is strictly conserved among all but onearchaebacteria.

Conclusion

This study of TMPKMtub complexed to TMP pro-vides the ®rst structure of a pathogen TMPK andthe ®rst example of a TMPK with an LID domainstructured in a-helix in the absence of a boundATP molecule. The spatial con®guration describedas selective for the adenine ring is also in place inour X-ray structure. The interaction of theTMPKMtub with the TMP shows three differencesin the contacts when compared to the yeast,human or E. coli enzymes: Arg14 and Tyr39, whichinteract with the phosphate moiety and Asn100,with the base moiety. The side-chains from fourarginine residues, 14, 153, 156, 160, are observedaround the phosphate-binding site and could becatalytically important.

Both the location along the sequence (P-loop andLID region) of those residues supposed to play acatalytic role and the kinetic results suggest thatthe TMPKMtub is not similar to the other TMPKsreported until now and that its ®ne structure couldnot have been predicted accurately using state-of-the-art homology modelling methods.

The question of ascertaining whether one ofthese residues plays a role more important thanthe others or whether it is a collective effect best

described by the electrostatic potential (and itsassociated electric ®eld) will require more detailedtheoretical studies. The role of the magnesium iondeserves special care, because any movement ofthis cation during catalysis would develop an enor-mous electric ®eld in the vicinity of the chemicalbond to be broken.

Materials and Methods

Crystallization and data collection

Crystals of M. tuberculosis TMPK in complex withTMP were obtained as described.5 Brie¯y, a 6 ml drop ofa 1:1 mixture of the protein solution at 5-8 mg mlÿ1 andthe reservoir solution was equilibrated with a 35 %(w/v) ammonium sulphate solution, 0.1 Mes (pH 6.0),containing 2 % (w/v) PEG 600. Crystals appear as hexa-gonal bipyramids within three weeks. The space groupis P6522 with cell dimensions 76.6 AÊ , 76.6 AÊ and 134.4 AÊ .Numerous heavy-atom compounds were screened atdifferent concentrations and varying incubation times.The heavy-atom derivatives were prepared by transfer-ring the TMPKMtub-TMP crystals to a stabilisation sol-ution of 60 % (w/v) ammonium sulphate, 10 % (w/v)PEG 2000 and 100 mM Mes (pH 6.0) containing theheavy-atom reagents. The concentration and soakingtime for each of the isomorphous derivatives obtainedwith mercury, samarium, platinum, and uranium salts,and the TMPKMtub binary complex with the nucleotideanalogue 5I-dUMP are indicated in Table 1.

X-ray data were collected from cryo-cooled crystalsusing 25 % (w/v) glycerol as cryoprotectant. The X-raysources and detectors used for data collection are listedin Table 1. Diffraction data were processed using MARX-DS25,26 or DENZO/SCALEPACK27 packages. The CCP4package28 was used to calculate structure factors fromthe observed intensities (TRUNCATE) and scale nativeto derivative data (FHSCAL).

Structure determination

Resolution of the crystal structure of the TMPMtub-TMP complex was performed by the multiple isomor-phous replacement (MIR) method. The Patterson mapswere interpreted with the automated procedure devel-oped in the program HEAVY29 and checked by cross-Fourier differences. The heavy-atom positions were thenre®ned with MLPHARE (CCP4, 1994) and DM,30 in Solo-mon mode,31 was used for solvent ¯attening calculations.Details of the crystallographic data sets used for struc-ture solution and re®nement are given in Table 1. Theoverall ®gure-of-merit of MLPHARE phases increasedfrom 0.43 to 0.75 after solvent ¯attening at 2.7 AÊ resol-ution, assuming 45 % solvent. The resulting map was ofexcellent quality.

Model building and refinement

Model building and adjustments were done with theprogram O,32 ®rst into the solvent-¯attened MIR map,then into the SIGMAA-weighted maps.33 Re®nementwas performed up to 1.95 AÊ resolution with REFMAC(CCP4, 1994).34 Standards protocols, including maximumlikelihood target, bulk solvent correction and isotropic B-factors were used. The model was inspected manuallywith SIGMAA-weighted 2Fo ÿ Fc and Fo ÿ Fc maps, and


progress in the model re®nement was evaluated by thedecrease in the free R-factor. Re®nement statistics can befound in Table 2.

The current model includes 208 residues (1-208), onemolecule of TMP, two sulphate ions, one metal ion and150 water molecules. One of the sulphate ions is locatedat the interface between two symmetry-related mol-ecules, while the other one is in the active site. There isone cis-proline (Pro37). The following residues weremodeled as alanine residues, because their side-chaindensity was either poorly de®ned or non-existent: Arg86,Glu144, Leu145. The residues 209-214 of the C terminuswere not observed.

Electrostatic calculations

All calculations were performed using the programDelphi run on an SGI machine.35 Partial charges wereassigned according to the dictionary of AMBER. Theelectrostatic potential (in units of kT/e) was mappedonto the molecular surface using the program Grasp.36

The ionic strength of the buffer was set to zero, with theinterior dielectric constant set to 2-4 and that of the sol-vent to 80. The TMP molecule and the magnesium andsulphate ions were omitted from the calculation. Toachieve better accuracy, and especially to remove arte-facts at the border of the grid, a focussing technique wasused and performed in three steps, with the moleculeoccupying gradually more and more of the grid (25 %,50 % and 75 %). The ®nal grid spacing was 1.5 grid unit/AÊ . The ®nal map was interpolated into a 65 � 65 � 65grid to allow for visualization with Grasp, or, alterna-tively, was converted into an O-style map using thedescription of the map ®le format provided in the Delphimanual.

Protein Data Bank accession numbers

Coordinates of the TMPKMtub binary complex havebeen deposited in the RSCB Protein Databank withaccession number 1G3U.

Acknowledgements

We thank the staff of ID14 (ESRF, Grenoble) and ofLURE (Orsay) for excellent facility with X-ray data col-lection. This work was supported by grants from theEEC (BIO98 CT-0354), Institut Pasteur, INSERM andCNRS (URA 2185). We thank M. Orozco for communi-cating results prior to publication.

References

1. Stokstad, E. (2000). Drug-resistant TB on the rise.Science, 287, 2391.

2. Anderson, E. P. (1973). Nucleoside and nucleotidekinases. In The Enzymes (Boyer, P. D., ed.), 3rd edit.,vol. 8, pp. 49-96, Academic Press, New York.

3. Wild, K., Bohner, T., Folkers, G. & Schulz, G. E.(1997). The structures of thymidine kinase fromHerpes simplex virus type I in complex withsubstrates and a substrate analogue. Protein Sci. 6,2097-2106.

4. Champness, J. N., Bennett, M. S., Wien, F., Visse, R.,Summers, W. C. & Herdewijn, P. et al. (1998).Exploring the active site of herpes simplex virus

type-1 thymidine kinase by X-ray crystallography ofcomplexes with aciclovir and other ligands. Proteins:Struct. Funct. Genet. 32, 350-361.

5. Li de la Sierra, I., Munier-Lehmann, H., Gilles, A. M.,Barzu, O. & Delarue, M. (2000). Crystallization andpreliminary X-ray analysis of the thymidylate kinasefrom M. tuberculosis. Acta Crystallog. sect. D, 56, 226-228.

6. Laskowski, R. A., McArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK, a program toassess the validity of crystallographic models. J. Appl.Crystallog. 26, 283-291.

7. Ostermann, N., Schlichting, I., Brundiers, R.,Konrad, M., Reinstein, J. & Veit, T. et al. (2000).Insights into the phosphoryltransfer mechanism ofhuman thymidylate kinase gained from crystalstructures of enzyme complexes along the reactioncoordinate. Struct. Fold. Des. 8, 629-642.

8. Lavie, A., Vetter, I. R., Konrad, M., Goody, R. S.,Reinstein, J. & Schlichting, I. (1997). Structure ofthymidylate kinase reveals the cause behind thelimiting step in AZT activation. Nature Struct. Biol.4, 601-604.

9. Lavie, A., Konrad, M., Brundiers, R., Goody, R. S.,Schlichting, I. & Reinstein, J. (1998). Crystal structureof yeast thymidylate kinase complexed with thebisubstrate inhibitor P1-(50-adenosyl) P5-(50-thymi-dyl) pentaphosphate (TP5A) at 2.0 AÊ resolution:implications for catalysis and AZT activation.Biochemistry, 37, 3677-3686.

10. Lavie, A., Ostermann, N., Brundiers, R., Goody, R.,Reinstein, J., Konrad, M. & Schlichting, I. (1998).Structural basis for ef®cient phosphorylation of 30-azidothymidine monophosphate by Escherichia colithymidylate kinase. Proc. Natl Acad. Sci. USA, 95,14045-14050.

11. Saraste, M., Sibbald, P. R. & Wittinghofer, A. (1990).The P-loop-a common motif in ATP- and GTP-bind-ing proteins. Trends Biochem. Sci. 15, 430-434.

12. Reynes, J. P., Tiraby, M., Baron, M., Drocourt, D. &Tiraby, G. (1996). Escherichia coli thymidylate kinase:molecular cloning, nucleotide sequence, and geneticorganization of the corresponding tmk locus.J. Bacteriol. 178, 2804-2812.

13. Via, A., FerreÁ, F., Branetti, B., Valencia, A. &Helmer-Citerich, M. (2000). Three-dimensional viewof the surface motif associated with the P-loop struc-ture: cis and trans cases of convergent evolution.J. Mol. Biol. 303, 455-465.

14. Munier-Lehmann, H., Chaffotte, A., Pochet, S. &Labesse, G. (2001). Thymidylate kinase of M. tuber-culosis: a chimera sharing properties common toeukaryotic and bacterial enzymes. Protein Sci. 10,1195-1205.

15. Berry, M. B., Meador, B., Bilderback, T., Liang, P.,Glaser, M. & Phillips, G. N., Jr (1994). The closedconformation of a highly ¯exible protein: the struc-ture of E. coli adenylate kinase with bound AMPand AMPPNP. Proteins: Struct. Funct. Genet. 19, 183-198.

16. Vonrhein, C., Schlauderer, G. J. & Schulz, G. E.(1995). Movie of the structural changes during a cat-alytic cycle of nucleoside monophosphate kinases.Structure, 3, 483-490.

17. Muller, C. W., Schlauderer, G. J., Reinstein, J. &Schulz, G. E. (1996). Adenylate kinase motionsduring catalysis: an energetic counterweight balan-cing substrate binding. Structure, 4, 147-156.


18. Briozzo, P., Golinelli-Pimpaneau, B., Gilles, A. M.,Gaucher, J. F., Burlacu-Miron, S. & Sakamoto, H.et al. (1998). Structures of Escherichia coli CMP kinasealone and in complex with CDP: a new fold of thenucleoside monophosphate binding domain andinsights into cytosine nucleotide speci®city. Struc-ture, 6, 1517-1527.

19. MuÈ ller-Dieckmann, H. J. & Schulz, G. E. (1994). Thestructure of uridylate kinase with its substrates,showing the transition state geometry. J. Mol. Biol.236, 361-367.

20. Schlichting, I. & Reinstein, J. (1997). Structures ofactive conformations of UMP kinase from Dictyoste-lium discoideum suggest phosphoryl transfer isassociative. Biochemistry, 36, 9290-9296.

21. Hinds, T. A., Compadre, C., Hurlburt, B. K. &Drake, R. R. (2000). Conservative mutations of Gln-125 in HSV-1 thymidine kinase result in a ganciclo-vir with minimal deoxypyrimidine kinase activities.Biochemistry, 39, 4105-4111.

22. Gouet, P., Courcelle, E., Stuart, D. & Metoz, F.(1999). ESPript: multiple sequence alignments inPostscript. BioInformatics, 15, 305-308.

23. Chenal-Francisque, V., Tourneux, L., Carniel, E.,Christova, P., Li de la Sierra, I., Barzu, O. & Gilles,A. M. (1999). The highly similar TMP kinases ofYersinia pestis and Escherichia coli differ markedlyin their AZTMP phosphorylating activity. Eur. J. Bio-chem. 265, 112-119.

24. Abele, U. & Schulz, G. E. (1995). High resolutionstructures of adenylate kinase from yeast ligatedwith inhibitor AP5A showing the pathway of phos-phoryl transfer. Protein Sci. 4, 1262-1271.

25. Kabsch, W. (1988). Automatic indexing of rotationdiffraction patterns. J. Appl. Crystallog. 21, 67-71.

26. Kabsch, W. (1988). Evaluation of single-crystal X-raydiffraction data from a position-sensitive detector.J. Appl. Crystallog. 21, 916-924.

27. Otwinowski, Z. & Minor, W. (1997). Processing ofX-ray diffraction data collected in the oscillationmode. Methods Enzymol. 276, 307-326.

28. Collaborative Computing Project No. 4 (1994). TheCCP4 suite: programs for protein crystallography.Acta Crystallog. sect. D, 50, 760-763.

29. Terwilliger, T. C., Kim, S. H. & Eisenberg, D. (1987).Generalized method for determining heavy-atomsparameters using the difference Patterson function.Acta Crystallog. sect. A, 43, 1-5.

30. Cowtan, K. & Main, P. (1998). Miscellaneous algor-ithms for density modi®cation. Acta Crystallog. sect.D, 54, 487-93.

31. Abrahams, J. P. & Leslie, A. G. W. (1996). Methodsused in the structure determination of bovinemitochondrial F1-ATPase. Acta Crystallog. sect. D, 52,30-42.

32. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard,M. (1991). Improved methods for the building ofprotein models in electron density maps and thelocation of errors in these models. Acta Crystallog.sect. A, 47, 110-119.

33. Read, R. J. (1986). Improved Fourier coef®cients formaps using phases from partial structures witherrors. Acta Crystallog. sect. A, 42, 140-149.

34. Murshudov, G., Vagin, A. & Dodson, E. (1997).Re®nement of macromolecular structures by themaximum-likelihood method. Acta Crystallog. sect.D, 53, 240-255.

35. Honig, B. & Nicholls, A. (1995). Classical electro-statics in biology and chemistry. Science, 268, 1144-1149.

36. Nicholls, A., Sharp, K. & Honig, B. (1991). Proteinfolding and association: insights from the interfacialand thermodynamics properties of hydrocarbons.Proteins: Struct. Funct. Genet. 11, 281-291.

37. GCG (1997). Wisconsin Package Version 9.1, Gen-etics Computer Group, Madison, USA, WI.

38. Esnouf, R. (1999). Extension of the programMolscript to display electron density maps. ActaCrystallog. sect. D, 55, 938-940.

39. Kraulis, P. J. (1991). MOLSCRIPT: a program to pro-duce both detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946-950.

Edited by R. Huber

(Received 7 February 2001; received in revised form 23 May 2001; accepted 25 May 2001)

X-ray Structure of TMP Kinase from Mycobacterium tuberculosisComplexedwithTMPat1.95 Å ResolutionIntroductionResults and DiscussionFigure 1Figure 2 (legend shown on page 92)Figure 2 (legend shown on page 92)Figure 3Figure 4Figure 5Figure 6Figure 7Figure 8Table 1Table 2Table 3Structure determination and overall descriptionThe TMP kinase familyDimerization modeStabilization of the structure of the LID regionMagnesium ionM. tuberculosis TMP kinase is in the fully closed conformationThe TMP-binding siteInteraction with other nucleoside monophosphatesInferring the ATP binding siteElectrostatic potential in the active sitePossible catalytic residues

ConclusionMaterials and MethodsCrystallization and data collectionStructure determinationModel building and refinementElectrostatic calculationsProtein Data Bank accession numbers

AcknowledgementsReferences

x-ray structure of tmp kinase from mycobacterium...

Documents