The Crystal Structure of Methenyltetrahydromethanopterin

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Research Article 1257

The crystal structure of methenyltetrahydromethanopterin cyclohydrolase from the hyperthermophilic archaeon Methanopyrus kandleri Wolfgang Grabarse1,2, Martin Vaupel1, Julia A Vorholt1, Seigo Shima1, Rudolf K Thauer1, Axel Wittershagen3, Gleb Bourenkov4, Hans D Bartunik4 and Ulrich Ermler2*

Background: The reduction of carbon dioxide to methane in methanogenic Addresses: 1Max-Planck-Institut für terrestrische archaea involves the tetrahydrofolate analogue tetrahydromethanopterin Mikrobiologie, Karl-von-Frisch-Straße, 35043 Marburg, Germany, 2Max-Planck-Institut für (H MPT) as a C unit carrier. In the third step of this reaction sequence, 4 1 Biophysik, Heinrich-Hoffmann-Straße 7, 60528 5 + N -formyl-H4MPT is converted to methenyl-H4MPT by the enzyme Frankfurt, Germany, 3Institut für Anorganische methenyltetrahydromethanopterin cyclohydrolase. The cyclohydrolase from the Chemie, AK Prof. Kolbesen, Marie-Curie-Straße 11, hyperthermophilic archaeon Methanopyrus kandleri (Mch) is extremely 60439 Frankfurt, Germany and 4Max-Planck- thermostable and adapted to a high intracellular concentration of lyotropic salts. Arbeitsgruppen für strukturelle Molekularbiologie, D-22603 Hamburg, Germany.

Results: Mch was crystallized and its structure solved at 2.0 Å resolution using *Corresponding author. a combination of the single isomorphous replacement (SIR) and multiple E-mail: [email protected] anomalous dispersion (MAD) techniques. The structure of the homotrimeric α/β Key words: halophilic enzyme, hyperthermophilic enzyme reveals a new fold that is composed of two domains forming a large enzyme, methanogenic archaea, Methanopyrus sequence-conserved pocket between them. Two phosphate ions were found in kandleri, methenyltetrahydromethanopterin and adjacent to this pocket, respectively; the latter is displaced by the cyclohydrolase phosphate moiety of the substrate formyl-H MPT according to a hypothetical 4 Received: 30 March 1999 model of the substrate binding. Revisions requested: 18 May 1999 Revisions received: 3 June 1999 Conclusions: Although the exact position of the substrate is not yet known, the Accepted: 7 June 1999 residues lining the active site of Mch could be tentatively assigned. Comparison Published: 30 September 1999 of Mch with the tetrahydrofolate-specific cyclohydrolase/dehydrogenase reveals similarities in domain arrangement and in some active-site residues, whereas Structure October 1999, 7:1257–1268 the fold appears to be different. The adaptation of Mch to high salt http://biomednet.com/elecref/0969212600701257 concentrations and high temperatures is reflected by the excess of acidic 0969-2126/99/$ – see front matter residues at the trimer surface and by the higher oligomerization state of Mch © 1999 Elsevier Science Ltd. All rights reserved. compared with its mesophilic counterparts.

Introduction Methanopyrus kandleri [9,10], Archaeoglobus fulgidus [2] and

Methenyl-tetrahydromethanopterin (methenyl-H4MPT) Methylobacterium extorquens AM1 [11] and the genes that cyclohydrolase is a cytoplasmic enzyme found in encode it have been cloned and sequenced. The methanogenic archaea [1], sulphate-reducing archaea [2] sequences at the protein level show a high degree of and methylotrophic bacteria [3]. It catalyzes the reversible similarity [9,11] (Figure 2). formation of N5,N10-methenyltetrahydromethanopterin + 5 (methenyl-H4MPT ) from N -formyltetrahydromethano- The methenyl-H4MPT cyclohydrolases from the differ- pterin (formyl-H4MPT) [4,5] (Figure 1). ent organisms are all composed of only one type of subunit with a molecular mass of approximately 35 kDa 5 + 5 10 + N -formyl-H4MPT+H N ,N -methenyl-H4MPT +H2O and are devoid of a chromophoric prosthetic group [9,2]. ∆G°′ = –5 kJ/mol [1] (1) They are reported to have a homodimeric quaternary structure [2,11] with the exception of the cyclohydrolase This reaction in the forward direction is involved in the from M. kandleri, which is a homotrimer (this paper). The

reduction of CO2 to methane and in autotrophic CO2 fixa- catalytic mechanism of the enzyme has not yet been elu- tion. In the reverse direction it is involved in C1 unit oxi- cidated. A very similar reaction is catalyzed by the dation to CO2. enzyme methenyltetrahydrofolate cyclohydrolase which is the reversible formation of N5,N10-methenyltetra- + 10 The enzyme has been purified from Methanobacterium hydrofolate (methenyl-H4F ) from N -tetrahydrofolate thermoautotrophicum [6,7], Methanosarcina barkeri [8], (formyl-H4F). 1258 Structure 1999, Vol 7 No 10

Figure 1

(a) Comparison of tetrahydromethanopterin 1-Desoxyribose (H MPT) and tetrahydrofolate (H F). 2-Hydroxyglutarate 4 4 (a) Structure of tetrahydromethanopterin Aniline OH Ribose (H4MPT) [53]. H4MPT can be subdivided into 7-Methyl-6-ethyl-pterin six moieties. The methylene group in para O O O position to the N10 is shaded in green. The OH OH COO two nitrogens carrying the C unit are OH HO 1 H O P O numbered in blue. (b) Structure of N10 COO O O 5,6,7,8-tetrahydrofolic acid (H4F). The H H H 10 CH carbonyl group in para position to the N is N 3 shaded in green. (c) Structures of H MPT HN 5 4 H carrying a C unit (shown in red) at the CH Phosphate 1 3 formate oxidation level. The C1 unit can be H2 N N N carried either on N5 or N10. N5-formyl-H MPT H H 4 is converted to methenyl-H MPT+ by the 5,6,7,8-Tetrahydromethanopterin (H4MPT) 4 methenyl-H4MPT cyclohydrolase. (b) O COO

N COO H H O N10 H N HN 5 H 5,6,7,8-Tetrahydrofolic acid (H4F) H 2 N N N H (c) O H R C H R N 10 O H NH N R O C O 10 O 10 C H CH3 CH3 N CH3 N N HN HN 5 H HN 5 H 5 H CH3 CH3 CH3 H N N H 2 N N H2 N N N 2 N H N H H H H H 5 10 Methenyl-H4MPT N -formyl-H4MPT N -formyl-H4MPT Structure

10 + 5 10 + N -formyl-H4F+H N ,N -methenyl-H4F +H2O NADPH. Although the crystal structure was solved with ∆G°′ = +6 kJ/mol [12] (2) NADP+ bound to the protein, no complex with formyl- + H4F, methenyl-H4F or methylene-H4F has been H4F and H4MPT are analogous coenzymes (Figures 1a,b) reported so far. A binding site for methenyl-H4F cyclo- that functionally differ mainly in the reactivity of N10, hydrolase was tentatively assigned and a sequence-con- which has a pKa of 2.4 in the case of H4MPT and of served YXXXK motif (using single-letter amino acid code) –1.2 in the case of H4F [13]. Interestingly, both was proposed to be involved in catalysis of the cyclohydro- + + methenyl-H4F and methenyl-H4MPT spontaneously lase step of the reaction [15]. hydrolyze to the N10-formyl derivative under alkaline conditions, whereas the enzyme-catalyzed reaction yields Here we report the crystal structure of methenyl-H4MPT 10 5 N -formyl-H4F [12] and N -formyl-H4MPT, respec- cyclohydrolase from M. kandleri (Mch). M. kandleri is a tively [5,11,14]. hyperthermophilic methanogenic archaeon that grows optimally at 98°C [16,17] and maintains a high intracellu- The crystal structure of a human methenyltetrahydro- lar concentration of cyclic 2,3-diphosphoglycerate (> 1 M) folate cyclohydrolase has been solved recently [15]. This and potassium ions (> 2 M) [16–18]. The enzymes in is a bifunctional enzyme with no sequence similarity to M. kandleri are therefore adapted to high temperatures and the methenyl-H4MPT cyclohydrolase; it catalyzes both high salt concentrations. Thus, Mch is completely stable + the formation of methenyl-H4F from formyl-H4F and the at 90°C and dependent on high concentrations of lyotropic + reduction of methenyl-H4F to methylene-H4F with salts for activity [10]. Research Article Methenyltetrahydromethanopterin cyclohydrolase Grabarse et al. 1259

Figure 2

Alignment of the known H4MPT cyclohydrolase primary structures and Mb. the ------VSVNIEA KKIVDR MI EGADDLKI S V D K LENGST VI DCGVN V Mc. jan ------MLSVNKKA LEIVNK MI ENKEEINID V I K LENGAT VLDCGVN V secondary structure assignment on the basis Ms. bar ------MI SVNEMG SNVI EE ML DWSEDLKTE V L K LNNGAT VI DCGVK A Ae. ful ------MLSVNEI A AEIVED ML DYEEELRI E S K K LENGAI VVDCGVN V of the structure of Mch. The amino acid Mtb. ex MSSNTSAPSLNALA GPLVES LVADAAKLRLI V A Q E- NGAR TVDAGAN A Mp. kan ------VSVNENA LPLVER MI ERAELLNVE V Q E LENGTT VI DCGVE A similarity as calculated by the program 210203040

ALSCRIPT [54] is shown in different shades α1 β1 β2 of blue. The amino acids Lys94 and Tyr190 have been marked with blue arrowheads. The Mb. the D GSI KAGELYTA VCLGG LAD V GI SI PGDLSERFALPSV KI KT DFPAI S numbering of the amino acids is according to Mc. jan P GSWKAGKLFTK ICLGG LAH V GI SL SPCECKGI TLPYV KI KT SHPAI A Ms. bar E GGYEAGMY LAR LCLAD LAD L --KY TTFDLNGLKWPAI QVAT DNPVI A the protein variant of Mch described in the Ae. ful P GSYDAGIMYTQ VCMGG LAD V --DI VVDTI NDVPFAFV TEYT DHPAI A Mtb. ex R GSI EAGRRIAE ICLGG LGT V --TI API GPVASWPYTV VVHS ADPVLA text. The organisms with their growth optimum Mp. kan A GGF EAGLLFSE VCMGG LAT V --EL TEFEHDGLCLPAV QVT T DHPAVS temperatures are Methanobacterium 50 60 70 80 thermoautotrophicum (Mb. the), 65°C, α2 β3 β4 α3 Methanococcus jannaschii (Mc. jan), 95°C,

Methanosarcina barkeri (Ms. bar), 37°C, Mb. the TLGAQKAGWS V SVGD----FFALGSGPARALALKPAETYEEI G YQDEA Mc. jan TLGAQKAGWA V KVGK----YFAMGSGPARALAKKPKKTYEEI G YEDDA Archaeoglobus fulgidus (Ae. ful), 95°C, Ms. bar CMASQYAGWR I SVGN----YFGMGSGPARALGLKPKELYEEI G YEDDF Methylobacterium extorquens (Mtb. ex), 30°C, Ae. ful CLGSQKAGWQ I KVDK----YFAMGSGPARALALKPKKTYERI E YEDDA Mtb. ex CL GSQYAGWS L ADEEGDSGFFALGSGPGRAVAVV- EELYKELG YRDNA and M. kandleri (Mp. kan), 95°C. The figure Mp. kan TLAAQKAGWQ V QVGD----YFAMGSGPARALALKPKETYEEI D YEDDA 90 100 110 120 130 was drawn using the program ALSCRIPT. The α β β5 β6 α4 helices and sheets are indicated by red tubes and green arrows, respectively. Mb. the DI AVLTLEADKLPGEDVTDKI AEEC D V S PENVYVLVAPTSSLVGSI QI Mc. jan DVAVLCLEASKLPNEEVAEYVAKEC G V E VENVYLLVAPTASLVGSI QI Ms. bar EAAVLVMESDKLPDEKVVEFI AKHC S V D PENVMI AVAPTASI AGSVQI Ae. ful DVAVI ALEANQLPDEKVMEFI AKEC D V D PENVYALVAPTASI VGSVQI Mtb. ex TTTALVLESGSAPPASVVNKVAAAT G L A PENVTFI YAPTQSLAGSTQV Mp. kan DVAI LCLESSELPDEDVAEHVADEC G V D PENLYLLVAPTASI VGSVQV 140 150 160 170 180

β7 α5 β8

Mb. the SGRVVENGTYKM L EAL H FDVNKV KYA A G I API APVDPDSL K AMGK TND Mc. jan SGRVVENGTYKM L EVL E FDVNKV KYA A G L API API IGDDF A MMGA TND Ms. bar SARVVETGI HK- F ESV G FDI NCI KSG Y G V API APVVGKDV Q CMGS TND Ae. ful SGRI VETAI FK- M NEI G YDPKLI VSG A G R CPI SPI LENDL K AMGS TND Mtb. ex VARVLEVALHK- A HTV G FDLHKI LDG I G S APLSPPHPDFI Q AMGR TND Mp. kan SARVVETGLYKL L EVL E YDVTRV KYA T G T API APVADDDG E AMGR TND 190 200 210 220

α6 β9 α7

Mb. the AVLFGGR TYYYI ESEEG- DDI KSLAEN LPSS ASEGYGK PFYD VFKEAD Mc. jan MV L Y GGI TYYYI KSDEN- DDI ESLCKA LPSC ASKDYGK PFME VFKAAD Ms. bar CVI YCGE TNYTVRFDGELAELEEFVKK VPST TSQDFGK PFYQ TFKEAN Ae. ful SMMYYGS VFLTVKKY------DEI LKN VPSC TSRDYGK PFYE IFKAAN Mtb. ex AIIYGGR VQLFVDADD- - ADAKQL AEQ IPST TSADHGA PFAE IFSRVN Mp. kan CI LYGGT VYLYVEGD---DELPEVVEE LPSE ASEDYGK PFMK IFEEAD 230 240 250 260 270

β10 α8 α9

Mb. the YDFYKI DK GMF APAE VVI NDLR TGEVFRAG F V NEELLMK SFGL---- Mc. jan YDFYKI DK GMF APAV VVI NDMT TGKVYRAG K V NAEVLKKSLGWTEL - Ms. bar FDFFKVDA GMF APAR LTVNDLN STKTI SSG G L YPEI LLQSFGI RNV- Ae. ful YDFYKI DP NLFAPAQ I AVNDLE TGKTYVHG K L NAEVLFQSYQI VLEE Mtb. ex GDF YKI DG ALFSPAE AI VTSVK TGKSFRGG R L EPQL VDASFV----- Mp. kan YDFYKI DP GVF APAR VVVNDLS TGKTYTAG E I NVDVLKESFGL---- 280 290 300 310

β11 β12 α10 Structure

Results results showed the enzyme to contain 0.5 mol calcium per Structure determination mol subunit Mch at pH 7.0. Heavy-atom derivatization Mch heterologously overproduced in Escherichia coli was was achieved with lutetium, an element that can replace crystallized using PEG 1500 as precipitant at pH 4.0. Three calcium ions in proteins [20] and that has exceptionally different crystal forms were obtained; the crystals with the good anomalous X-ray scattering properties. The quality space group P6322 were used for structure determination of the solvent-flattened electron-density map was suffi- (see the Materials and methods section for details). cient to build in the complete Cα trace of the Mch model.

The structure of Mch was solved using the method of The model of Mch, including all 316 amino acid residues, single isomorphous replacement (SIR) in combination three phosphate ions and 245 water molecules, was with multiple anomalous dispersion (MAD). This solution refined to an Rcryst of 19.8% and an Rfree of 21.8%. The strategy was stimulated by the result of a metal analysis sequence of the recombinant enzyme differs from the prior to structure determination using total reflection wild-type enzyme in two amino acids at the N terminus X-ray fluorescence spectrometry (TXRF) [19]. The and C terminus. The N terminus starts with the sequence 1260 Structure 1999, Vol 7 No 10

MVSV instead of VSV and the C terminus ends with FSL be subdivided into two smaller β sheets that are inclined instead of FGL [9]. Crystals of the enzyme with the origi- against each other by nearly 90°. Domain B, comprising nal N terminus belonging to the space group R3 were also residues 58–82 and 171–316, also consists of a six-stranded obtained. The structure of this crystal form was solved by mixed β sheet with the arrangement of the strands being molecular replacement and refined to 2.4 Å resolution. different to that of domain A (Figure 3c). Strands β3 and β4 were found to be longer than the other strands of the The monomer structure β sheet, which is noteworthy because these extensions The monomer of Mch is arranged in a globular form with participate in trimer formation. In contrast to domain A, 3 overall dimensions of 50 × 40 × 35 Å . The Cα trace of the the central β sheet of domain B is only covered by helices enzyme, the secondary structure elements and a fold α6, α7, α8 and α10 on its frontside but the back of the topology diagram are presented in Figures 3a–c. β sheet is involved in trimerization.

The fold of Mch can be analyzed in different ways. The extensive contacts between the two domains make According to the most familiar way of description the the subdivision into two domains somewhat artificial. The structure can be subdivided into two closely attached structure can alternatively be considered as an entity com- domains both adopting an α/β structure. The two domains, posed of a β/α/β layer. Whereas the β layers correspond to denoted A and B, are mutually arranged in such a manner the described central β sheets of the domains, the α layer as to create a large pocket between them. Domain A comprises helices α1, α2, α6, α7, α9 and α10 (Figure 3b). includes residues 1–57 and 83–170 and reveals the central folding motif to be a six-stranded mixed β sheet. The According to the program DALI [21], Mch possesses a β sheet is flanked by helices α1 and α2 on the back and new fold within the α/β family. Also, the separated helix α5 on the front, thus forming an α/β sandwich motif. domains A and B reveal no significant fold relationship to A more precise analysis of the β sheet indicates that it can any of the other structurally characterized proteins. Thus,

Figure 3

(a) Diagram of Mch Cα trace, fold and topology. 30 30 (a) Stereo drawing of the Cα trace of Mch 160 160 produced with the program SETOR [55]. 150 150 Domain A is shown in green, domain B in red. 20 20 The sidechains of three active-site residues are 10 10 shown in pink. (b) Ribbon diagram of the overall 40 100 40 100 130 130 fold of Mch; helices are shown in red, β strands 140 110 140 110 (c) 170 N 170 120 N in green. Fold topology plot of Mch. 50 Arg183 60 120 50Arg183 60 90 90 220 180 Lys94 220 180 Lys94 80 Glu18680 Glu186 190 190 210 200 210 200 230 230 280 280 C 310 240 C 310 240 260 290 260 290 70 250 70 250 300 300

(b) (c) α4 α3 β1 β2 α5 Domain A β 8 β8 β7 β6 β5 β7 β1 β2 α1 β6 β5 α5 α 2 N α1 α2 α6 N β3 α3 β α4 α7 4 α8 C β9 α6 α7 α9 α10 Domain B C β10 α9 β11 α10 β3 β4 β9 β10 β11 β12 β 12 α8 Structure Research Article Methenyltetrahydromethanopterin cyclohydrolase Grabarse et al. 1261

Figure 4

Domain arrangement and electrostatic (a) (b) potential of Mch. (a) Molecular surface representation of Mch including a theoretical + model of the substrate methenyl-H4MPT . The pterin and aminobenzoate moieties of the substrate are located in a large pocket between domain A (red) and domain B (green). The figure was created using the program GRASP [56]. A yellow arrow was Pi402 Pi401 drawn into the figure for alignment with Figure 5. (b) Electrostatic potential of Mch calculated at 1.5 M salt concentration (potential range –30 kT–10 kT). Two clusters of positively charged residues in and adjacent to the pocket serve as binding sites for the Electrostatic potential phosphate ions Pi401 and Pi402. Domain A Positive Domain B Negative Structure the functionally related tetrahydrofolate-dependent cyclo- formed by helices α7 and α9 and the following loop. Sur- hydrolase/dehydrogenase [15] adopts a different fold than prisingly, the residues of this loop are both solvent- Mch, but shows a similar overall architecture because, as exposed and hydrophobic (Figure 5a, yellow arrow). in Mch, there are two α/β domains forming a pocket Although these residues are well conserved throughout between them. the sequences of the H4MPT-dependent cyclohydrolases (Figure 5b) they are highly flexible, as indicated by the The profile of the protein surface creates a large pocket dramatically increased isotropic temperature factors. The of approximately 20 × 8 × 10 Å3 in size between domains solvent-exposed sidechains of Phe276, Tyr277 and A and B (Figure 4a). The pocket is characterized by a Phe284 (Figure 5a) are orientated towards the pocket and high degree of amino acid sequence conservation interact with the more buried unpolar residues Phe265, (Figure 5b) and is therefore highly attractive as a binding Phe269, Met223, Met266 and Ile231 in a large hydropho- + site for the bulky methenyl-H4MPT . The pocket size in bic cluster. These sidechains might also form hydrophobic + relation to the size of the substrate methenyl-H4MPT is and aromatic interactions with the large conjugated aro- + shown in Figure 4a. matic system of the substrate methenyl-H4MPT and with hydrophobic residues of the pocket wall from the The bottom of this pocket is formed by the long kinked domain A side in the enzyme–substrate complex. helix α6 whereas the side wall of the pocket from the domain A side is made up by helices α3 and α5, the loop In contrast, the bottom of the pocket and parts of the side connecting helix α3 to strand β5, and strands β5, β6 and wall from the domain A side are dominated by polar and β7. The side wall of the pocket from the domain B side is charged residues including Glu186, Ser109, Arg183, Lys94

Figure 5

Hydrophobicity and primary structure (a) (b) conservation of Mch. (a) Hydrophobicity of the surface amino acids of Mch. The pocket wall from the domain B side is dominated by hydrophobic residues. The solvent-exposed, highly flexible and hydrophobic loop is marked by a yellow arrow. (b) Primary structure similarity of the methenyl-H4MPT cyclohydrolases from archaea mapped onto the molecular surface of Mch as calculated by the program ALSCRIPT. The pocket and the hydrophobic loop from the domain B side (yellow arrows) are the most sequence- conserved regions of the protein. Amino acid sequence Amino acid hydrophobicity conservation Hydrophobic Identity Structure Hydrophilic Not conserved 1262 Structure 1999, Vol 7 No 10

Figure 6

Stereo picture of the nonhydrophobic residues (in single-letter amino acid code) in Q179 E140 Q179 E140 the pocket of Mch. The salt bridges of the two phosphate ions with the protein residues are indicated by dotted lines. H152 H152 S109 S109 R183 W97 R183 W97 Pi402 Pi402 Pi401 Pi401 K94 K94 E186 K118 E186 K118

Y190 Y190

Structure

and Tyr190. The short distance of 4 Å between atoms interface. In the P6322 crystal form, which is described OH-Tyr190 and Nε-Lys94 (Figure 6) is interesting here, these interfaces coincide with the crystallographic because a lysine and a tyrosine residue were proposed to threefold and twofold axes. The solvent-accessible surface be involved in catalysis of the cyclohydrolase step in the that is buried by trimer formation is 2198 Å2 (16% of the 2 H4F-dependent cyclohydrolase/dehydrogenase bifunc- monomer surface) compared with only 484 Å (3.7%) for tional enzyme [15] (see the Discussion section). the dimer interface. Thus, the accessible surface buried by the trimer interface is in the range usually observed for The electron density clearly showed the presence of three oligomeric proteins [22]. Given that the same trimer inter- phosphate or sulphate anions, although neither phosphate face is present in all crystal forms and the trimer is also nor sulphate was used during protein crystallization or much more compact compared with the dimer, we propose purification. Two of these anions, modelled as phosphates, that Mch is present in cells as a trimer (Figure 7). are located in and at the entrance of the pocket (Figure 4b). Each of the two phosphate-binding sites might constitute a Each monomer forms two contact areas, denoted I and II, binding site for the negatively charged α-hydroxyglu- to each of the neighboring subunits. Contact area I is built + β α tarylphosphate moiety of the substrate methenyl-H4MPT up by the loops connecting strand 7 and helix 7, strand given that there are no further solvent-exposed clusters of β2 and helix α2, helix α1 and strand β1, strand β8 and positively charged residues in the active-site region (see helix α6, by the C-terminal segment and by the prolonged β β the Discussion section). The phosphate ion Pi401 found at portions of strands 3 and 4. Contact area II is formed by the entrance of the pocket (Figure 4b) is linked to the the central β sheet at the backside of domain B and is protein by hydrogen bonds to residues Lys118, His152 and composed of the N-terminal segment, strand β3 and the β β β β β Trp97, whereas the phosphate ion Pi402 is located inside following loop, strands 4, 9, 10, 11 and 12 and the the pocket and contacts atoms Nε-Lys94, N-Ala95, connecting loop between strands β11 and β12. At the Nε1-Arg183 and Oε1-Glu186 (Figure 6). The two phos- trimer axis, the prolonged portions of strands β3 and β4 of phate ions Pi401 and Pi402 are approximately 17 Å apart the three monomers form the wall of a hole of about 4.5 Å and have temperature factors of 40 Å2 and 27 Å2, respec- diameter that is occupied by a solvent molecule positively, indicating a tighter binding or a higher occupancy of tioned at the trimer axis. the phosphate buried in the pocket. An overview of the nonhydrophobic residues in the pocket including the The trimer interface is basically built up of a large, mainly bound phosphate ions is presented in Figure 6. hydrophobic cluster that is surrounded by polar regions. The central hydrophobic cluster comprises residues The electron density did not reveal an unequivocal cation Phe46, Leu50, Leu65, Pro76, Ser173, Ile174 and Val215 of binding site. Even at the heavy atom binding site, which is one monomer and residues Thr81, Tyr205, Thr207, located between residues Asp146 and Asp148, no electron Tyr237 and Leu294 of the second monomer. No water density corresponding to a calcium ion was observed in the molecules were found inside this cluster. The surrounding native dataset. A possible explanation for this finding, polar region shields the nonpolar region of the interface which contradicts the TXRF results, could be a protonation from bulk solvent and contributes to oligomerization by of the aspartate residues at pH 4.0 used for crystallization. forming ten intersubunit hydrogen bonds and several polar interactions mediated by firmly bound water mol- The trimer interface ecules. Although four positively and five negatively In all three crystal forms of Mch the enzyme has a crystallo- charged residues are involved in hydrogen binding, only a graphic dimer interface and, independently of that, a trimer few ion pairs were found at the polar interface region. For Research Article Methenyltetrahydromethanopterin cyclohydrolase Grabarse et al. 1263

Figure 7

The Mch trimer. (a) Van der Waals’ sphere and (b) ribbon drawing of the Mch trimer viewed perpendicular to the trimer axis. The compactness of the trimer is achieved by the large interface area mainly formed by the three sixfold β sheets of domain B. The figures were created using the programs MOLSCRIPT [57], RASTER3D [58] and SETOR.

example, one ion-pair network formed between Glu67, accessible surface of Mch compared with only eight lysines, Glu69, His70 and Arg288 is located near the trimer axis. five arginines and three histidines per monomer. The frequency of glutamate residues, almost all of which are Interactions between adjacent monomers help to anchor solvent exposed, was found to be increased by 50% above the N termini and C termini of the protein. The N termi- the average of the H4MPT cyclohydrolases from other nus is fixed by a salt bridge between the N terminal amino organisms. In contrast, the aspartate frequency was not sig- group of Met1 and residue Glu20 of the adjacent nificantly increased. The negatively charged residues cover monomer. Additionally, an intramolecular hydrogen bond the entire trimer surface except for the substrate pocket, is formed between the peptide oxygen of Val2 and the thereby confirming the functional relevance of this surface hydroxyl group of Tyr128. From the separately deter- area. The hydrophobic surface fraction was found to be mined structure of Mch with the original N terminus Val2 decreased to 50% compared with 57% for an average it is known that a similar interaction is established mesophilic protein [23]. This trend also manifests itself in between Glu and the N terminus Val2. The fixation of the the lower frequency of hydrophobic amino acids compared

C terminus is enhanced by a hydrophobic intersubunit with other H4MPT-dependent cyclohydrolases [9]. interaction involving the sidechains of Leu316 and Phe314 with residues Thr299 and Thr301 of the adjacent Discussion monomer. Moreover, at the C terminus, a hydrogen bond Modelling of the substrate binding is present between the peptide nitrogen of residue Attempts to cocrystallize Mch with the substrate have, as Leu316 and the peptide oxygen of Lys311. yet, failed. A possible explanation for this finding is that the enzyme without bound substrate is stabilized by Sequence comparison studies indicate that the residues crystal-packing forces or, alternatively, that the substrate forming the hydrophobic cluster at the trimer interface binding affinity of the enzyme is reduced at pH4 (neces- tend to be either conserved or conservatively exchanged, sary for crystallization). whereas the charged residues at the polar part of the trimer interface are not maintained. Consequently, only Nevertheless, a model for the binding of methenyl- + one of the ten hydrogen bonds in the polar trimer inter- H4MPT to Mch can be derived from two characteristic face region was found to be sequence-conserved in all properties of the protein surface. The first property is the other H4MPT-dependent cyclohydrolases. A prediction of previously described sequence-conserved pocket the site of the dimer interface in the H4MPT cyclohydro- between domains A and B. Its partly hydrophobic charac- lases of M. thermoautotrophicum and M. barkeri is difficult ter, in particular the presence of the aromatic sidechains because the described trimeric oligomerization mode is lining up the wall of the pocket from the domain B side, not a suitable model for a dimer interface. and its size are consistent with the binding of the pterin and aminobenzoate moieties of the substrate inside the Surface of the trimer pocket. The second property is the presence of two clus- A striking feature of Mch is the abundance of negatively ters of positively charged amino acids in and adjacent to charged amino acid residues, especially glutamates, at the the pocket, each of them binding a phosphate ion. A func- solvent-accessible surface of the protein. There are 36 tional relevance of these clusters is likely because the solvent-accessible glutamates and 20 aspartates at the protein surface is predominantly covered by acidic 1264 Structure 1999, Vol 7 No 10

residues that are necessary for adaptation to high salt con- flexible, but highly sequence-conserved hydrophobic centrations. An obvious role for one of these clusters is to loop, which might be rigidified by hydrophobic interac- bind the threefold negatively charged α-hydroxyglu- tions between its aromatic sidechains and the tetra- tarylphosphate moiety of the substrate (Figure 1) hydropterin ring. Moreover, after the conformational whereby the observed phosphate ion might mimic the change most of the solvent-exposed hydrophobic surface binding mode of the phosphate moiety of the substrate at this pocket wall would be buried. Because it is not as + methenyl-H4MPT . The binding of phosphate ions yet known whether the phosphate ion Pi402 is present instead of the phosphate moiety of a substrate has been during substrate binding, and because the domain B side reported for several enzymes, such as phosphoribosyl- of the pocket is expected to undergo structural changes anthranilate isomerase/indoleglycerolphosphate synthase upon substrate binding, an accurate prediction of the [24], and recently for the NADPH-dependent dihydro- interactions between the protein and the substrate + folate reductase from Haloferax volcanii [25]. The cluster methenyl-H4MPT is not possible. Hence, a detailed of positively charged amino acids at the position of the analysis of the substrate binding has to await structure phosphate ion Pi401 is a highly attractive binding site for determination of an enzyme–substrate complex. It the phosphate moiety of the substrate as it not only allows should be noted, however, that if the proposed orienta- for the binding of the α-hydroxyglutarylphosphate moiety tion of the substrate is correct, the two nitrogen atoms N5 but also for the binding of the pterin head of the substrate and N10 and the attached formyl group can be reached by inside the pocket. The plausibility of this binding mode is the residues Lys94, Arg183 and Glu186 that contact the supported by the observation that residue Lys118, which buried phosphate ion Pi402. binds to the phosphate Pi401, is conserved in the archaeal but not in the bacterial H4MPT cyclohydrolases, in which Enzymatic function the substrate was reported to be devoid of the α-hydroxy- The mechanism of the base-catalyzed spontaneous glutarylphosphate moiety. In contrast, the binding of the hydrolysis of formamidinium cations analogous to both phosphate moiety of the substrate at the position of the H4F and H4MPT has been studied in great detail [27,28]. phosphate ion Pi402 appears to be unlikely because the These studies suggest that the spontaneous hydrolysis of pocket does not provide positively charged amino acids or the formamidinium cation proceeds via a tetrahedral imi- mainchain amide groups that could bind the two adjacent dazolidin-2-ol intermediate that might also be present in negatively charged carboxylate groups of the α-hydroxy- the enzyme-catalyzed reaction. The breakdown of the glutarate moiety. imidazolidin-2-ol to the N10-formyl compound via acid/base catalysis can involve either a positively charged We have built a hypothetical model of the enzyme–sub- ammonium derivative or a negatively charged oxyanion strate complex on the basis of the assumption that the derivative as transition states. phosphate moiety of the substrate exactly binds at the observed phosphate position Pi401 and the pterin moiety From our modelling studies we propose that the only polar binds inside the pocket (Figure 4a). The modelling was or charged residues that can possibly contact the atoms N5, 10 further restrained because of the assumption that the N and the carried C1 unit of the substrate are Lys94, + substrate is bound in the form of methenyl-H4MPT Glu186 and Arg183. Thus, the putative active site, which is 5 which, in contrast to the product N -formyl-H4MPT, pos- identical to the most sequence-conserved region of the sesses a large nearly planar conjugated system comprising protein, consists of charged residues that could stabilize the pyrimidine ring, the upper part of the tetrahydro- charged transition states. The finding that the putative pyrazin and the imidazol rings and the benzyl ring as active-site region is a phosphate-binding site might indicate observed in the structure of methenyltetrahydrofolate that an oxyanion intermediate is favoured by the enzyme. [26]. Assuming the binding of the large conjugated The presence of the observed phosphate ion in the system inside the pocket and the binding of the phos- enzyme–substrate complex, however, cannot be excluded. phate moiety at the observed phosphate-binding site, the substrate has to be kinked in its desoxyribosyl moiety by For the methenyl-H4F-cyclohydrolase/dehydrogenase nearly 90° at the entrance of the pocket (Figure 4a). The bifunctional enzyme, a tyrosine and a lysine residue were crystal structure clearly shows that the depth of the proposed to be involved in the catalysis of the cyclohydro- pocket is sufficient to accommodate this nearly planar, lase reaction [15]. Interestingly, the residue Lys94 found conformationally rigid system, but also indicates that the in the putative active site of Mch is also neighboured by a pocket is too broad to provide contact between tyrosine (Figure 6), although the geometric arrangement + methenyl-H4MPT and both pocket walls. We assume, of the residues is different. Because Tyr190 was found to therefore, a substrate-induced movement of the be exchanged against phenylalanine in A. fulgidus hydrophobic part of domain B by approximately 1.5 Å (Figure 2), a catalytic function of this residue as a hydro- towards the plane of the pterin and aniline rings. This gen donor is impossible for the H4MPT-dependent cyclo- assumption is supported by the presence of the described hydrolases. The residue Lys94 is, in contrast to residues Research Article Methenyltetrahydromethanopterin cyclohydrolase Grabarse et al. 1265

Glu186 and Arg183, not entirely sequence-conserved but Mch is not independently adapted to high temperatures exchanged against a tyrosine in M. barkeri and and high salt concentrations, but both properties are M. extorquens. As both tyrosine and lysine have been pro- strongly interconnected. Activity and stability at high tem- posed to be involved in acid/base-catalyzed reactions peratures can be achieved only at high concentrations of [29,30], a catalytic function of residue Lys94 cannot be lyotropic salts. This can be understood at a molecular level excluded. Site-directed mutagenesis experiments will be because high lyotropic salt concentrations enhance carried out to explore the catalytic function of the active- hydrophobic interactions in the trimer interface because of site residues. the salting out effect [45] and reduce the repulsion of the negatively charged amino acids at the trimer surface. As Structural determinants which reflect the adaptation to the the substrate binding itself involves hydrophobic interac- extreme environment of M. kandleri tions, the salt dependence of the activity of Mch might Biochemical and structural information is available for also reflect differences in substrate affinity. several proteins from either halophilic and hyperthermophilic organisms; however, M. kandleri is the only Biological implications organism from which proteins have been structurally Tetrahydromethanopterin (H4MPT) is a structural ana- characterized that have to withstand both temperatures logue of tetrahydrofolate (H4F) that is found in the near 100°C and salt concentrations of higher than 1 M. anaerobic methanogenic and sulphate-reducing archaea Comparison of halophilic proteins with their non- and in the aerobic methylotrophic bacterium Methylo- halophilic counterparts indicated three structural deter- bacterium extorquens. Both H4MPT and H4F are minants contributing to halotolerance: an increased involved in the conversion of C1 units from the formate fraction of acidic residues at the surface [25,31,32]; a to the methanol oxidation level. The enzyme from the decreased fraction of hydrophobic residues at the surface hyperthermophilic methanogenic archaeon Methanopy-

[33]; and an increased arginine to lysine ratio [25,34,35]. rus kandleri is involved in CO2 reduction to methane, a Indeed, in Mch, the frequency of glutamic acid was found metabolic pathway that is coupled with energy conserva- to be dramatically increased, whereas the hydrophobic tion. Methenyl-H4MPT cyclohydrolase catalyzes the + surface fraction was significantly decreased when com- reversible interconversion from methenyl-H4MPT to pared with other H4MPT-dependent cyclohydrolases. formyl-H4MPT, which is analogous to the reaction per- Both a decreased hydrophobic surface fraction and an formed by methenyl-H4F cyclohydrolase. The hydrolysis increased fraction of acidic residues are thought to products differ, however, because in the case of 5 prevent the enzyme from aggregating at high salt concen- methenyl-H4MPT cyclohydrolase, N -formyl-H4MPT is trations [36,37]. In contrast, an increase of the arginine to formed and in the case of methenyl-H4F cyclohydrolase, 10 lysine ratio was not observed in Mch, and this observation N -formyl-H4F is formed, which results in an important also applies for other proteins from M. kandleri such as difference in the thermodynamics of the reactions. 5 methyl coenzyme M reductase [38] and formyl- Whereas hydrolysis of H4MPT to N -formyl-H4MPT is methanofuran:H4MPT formyltransferase [39]. endergonic by +5 kJ/mol, that of methenyl-H4F to 10 N -formyl-H4F is exergonic by –6 kJ/mol. Adaptation strategies of hyperthermophilic proteins as derived from comparative structure analysis include the In this report, the crystal structure of methenyl-H4MPT stabilization of flexible regions such as loops or the cyclohydrolase from M. kandleri heterologously pro- N termini and C termini [40], an increased number of duced in Escherichia coli is presented. The structure of intrasubunit ion pairs [41] and a higher oligomerization the homotrimeric protein exhibits a new α/β fold that is state [42]. The most striking adaptation strategy in Mch is different to that of the folate-dependent dehydroge- the change in the oligomerization state from a dimer, as nase/cyclohydrolase. Both structures are similar, reported for mesophilic or even moderately thermophilic however, in that both have two α/β domains with a

H4MPT-dependent cyclohydrolases [2,6,8,11], to a trimer; large pocket between them. this substantially enlarges the buried surface area and increases the packing density. Because of the trimeriza- Although the structure of the enzyme–substrate complex tion not only the hydrophobic interactions between the is not known, the approximate binding site of the sub- + subunits are strengthened but several loops and also the strate methenyl-H4MPT can be predicted considering N termini and C termini are fixed by contacts to the two independent, but highly compatible structural fea- neighboring subunits. In contrast, the number of intrasub- tures. The first is a stretch of basic residues at the unit ion pairs per residue was not found to be higher than entrance of the pocket, which might bind to the nega- that of an average mesophilic protein [43] as reported in tively charged α-hydroxyglutarylphosphate moiety of the [39,41,44], but in fact to be decreased to 0.03. An substrate. The second feature is a large interdomain increased number of ion pairs might, therefore, not be a pocket that can accommodate the aminobenzoate and prerequisite for thermophilic adaptation. pterin moieties of the substrate. One of the pocket walls 1266 Structure 1999, Vol 7 No 10

includes a solvent-exposed flexible loop of hydrophobic Figure 8 amino acids that is predicted to undergo a conformational change upon substrate binding. Although it was impossible to predict specific protein–substrate interactions from modelling studies, the residues lining the active site could be identified and their function is now being investigated by site-directed mutagenesis studies.

Mch is active at temperatures above 90°C and in the presence of high concentrations of lyotropic salts. Major determinants of the thermostability of Mch are the higher oligomerization state compared with the mesophilic H4MPT-dependent cyclohydrolases and the fixture of the N and C termini. In contrast to other thermophilic proteins, no increase in the number of ion pairs was observed in Mch. The adaptation to high concentrations of lyotropic salts is reflected by the abundance of negatively charged amino acids and by a decreased hydrophobic fraction of residues at the surface of the enzyme. The observed salt dependency of the thermostability is probably because of the compensation of repul- sive forces between the negatively charged surface residues by inorganic cations and because of an increase Solvent-flattened SIRAS electron-density map with the final model of of the hydrophobic interactions within the core and the Mch. The map was contoured at 1σ. subunit interface.

method at a temperature of 4°C using a reservoir solution containing Materials and methods 28% PEG 1000, 30% glycerol and 0.1 M sodium acetate buffer at Total reflection X-ray fluorecence spectrometry (TXRF) pH 4.0. Droplets of 2 µl volume of the reservoir solution and of a solu- All TXRF measurements were performed using an EXTRA IIa TXRF tion containing 10 mg/ml protein in 50 mM Tris buffer at pH 7.5 were spectrometer (ATOMIKA Instruments, Oberschleissheim, Germany) thoroughly mixed. equipped with Mo, W(L) and W(brems) excitation units, a Si(Li) solid state detector, an automatic sample changer and a computer con- Three similar crystal forms were obtained under nearly identical condi- trolled multichannel analyzer system. Prior to analysis the enzyme was tions. Hexagonal crystals with the size 0.3 × 0.3 × 0.5 mm3, belonging dialyzed against a buffer solution containing 20 mM Tris-acetate pH 7. to the space group P6322 with unit-cell parameters of a=b=126.1 Å An internal standard of yttrium (100 µg/ml) was added to the prepared and c = 172.1 Å were used for structure determination. The two other µ enzyme solution. A sample of 5 l of this solution was transferred onto crystal forms belong to the space groups R3 and C2221 with cell con- a quartz glass carrier, dried to a thin film and measured for 1000 s stants of a=b=126.5 Å, c = 262.6 Å and a = 126.1 Å, b = 217.6 Å, using Mo radiation for excitation (voltage 50 kV, current 38 mA, dead c = 72.9 Å, respectively. The R3 crystal form contains the protein with time 20–30%). All samples were measured three times. the original N terminus.

Crystallization Data collection and phasing Mch was overexpressed in E. coli and purified as described [9]. Crys- Data sets Nat and Lu5 were collected in-house with a Rigaku RAXIS tals suitable for X-ray analysis were obtained by the hanging-drop 2C detector mounted on a Rigaku rotating Cu anode X-ray generator.

Table 1

Data sets used for structure determination and refinement.

Dataset Nat1 NatR3 Lu1 Lu2 Lu3 Lu4 Lu5

Wavelength (Å) 1.542 1.050 1.337 1.338 1.050 1.337 1.542 Resolution (Å) 2.0 2.4 2.6 2.6 2.6 2.8 2.7 Completeness (%) overall/outer shell 99.3/99.7 94.7/91.2 98.7/99.6 94.3/96.1 93.3/95.6 81.3/82.0 99.3/99.6

Rsym (%)* overall/outer shell 8.2/30.1 6.8/30.8 5.9/18.7 6.4/18.7 5.5/18.7 7.7/21.2 9.5/22.6 Phasing power (iso/ano)† – – 1.30/0.87 1.30/0.74 1.00/0.61 0.79/0.72 0.72/0.70

Σ Σ † *Rsym =( |–I|/ |I|, where the sum is over all symmetry-equivalent reflections. Phasing power = FH/E, where FH is the calculated heavy-atom structure factor and E is the residual lack of closure error. Research Article Methenyltetrahydromethanopterin cyclohydrolase Grabarse et al. 1267

Table 2 References 1. Thauer, R.K. (1998). Biochemistry of methanogenesis: a tribute to Refinement statistics. Marjory Stephenson. Microbiology 144, 2377-2406. 2. Klein, A.R., Breitung, J., Linder, D., Stetter, K.O. & Thauer, R.K. (1993). Data set Nat NatR3, original N terminus N5,N10-Methenyltetrahydromethanopterin cyclohydrolase from the extremely thermophilic sulfate reducing Archaeoglobus fulgidus: Resolution range (Å) 30–2.0 30–2.4 comparison of its properties with those of the cyclohydrolase from the extremely thermophilic Methanopyrus kandleri. Arch. Microbiol. Rcryst (%) 19.8 21.2 159, 213-219. Rfree (%) 21.9 26.2 3. Chistoserdova, L., Vorholt, J.A., Thauer, R.K. & Lidstrom, M.E. (1998). Number of protein atoms 2363 10,400 C1 transfer enzymes and coenzymes linking methylotrophic bacteria and methanogenic archaea. Science 281, 99-102. Number of water molecules 216 880 4. Donnelly, M.I., Escalante-Semerana, J.C., Rinehart, K.L. Jr. & Wolfe, Number of phosphate ions 3 12 R.S. (1985). Methenyl-tetrahydromethanopterin cyclohydrolase in cell Rmsd bond lengths (Å) 0.005 0.006 extracts of Methanobacterium. Arch. Biochem. Biophys. 242, 430-439. Rmsd bond angles (°) 0.8 0.9 5. Breitung, J. & Thauer, R.K. (1990). Formylmethanofuran: Mean B factor (Å2) tetrahydromethanopterin formyl-transferase from Methanosarcina barkeri: identification of N5-formyltetrahydromethanopterin as the protein 25.5 30.7 product. FEBS Lett. 275, 226-230. solvent 40.8 43.3 6. DiMarco, A.A., Donnelly, M.I. & Wolfe, R.S. (1986). Purification and properties of the 5,10-methenyltetrahydromethanopterin cyclohydrolase from Methanobacterium thermoautotrophicum. J. Bacteriol. 168, 1372-1377. 7. Vaupel, M., Dietz, H., Linder, D. & Thauer, R.K. (1996). Primary All measurements were performed at 90K with a cryostream cooler structure of cyclohydrolase (Mch) from Methanobacterium (Oxford Cryosystems) using the mother liquor of the crystals as cryo- thermoautotrophicum (strain Marburg) and functional expression of protectant. Lu derivatives of Mch were prepared by dissolving 10 µg the mch gene in Escherichia coli. Eur. J. Biochem. 236, 294-300. solid LuCl3 in the hanging drop since the crystals appeared to be very 8. Te Brömmelstroet, B.W., et al., & Vogels, G.D. (1990). Purification susceptible to changes in the mother liqour. The crystals for the native and properties of 5,10-methenyltetrahydromethanopterin data set Nat and the derivative data sets Lu1–Lu5 were originated from cyclohydrolase from Methanosarcina barkeri. J. Bacteriol. the same drop. 172, 564-571. 9. Vaupel, M., Vorholt, J.A. & Thauer, R.K. (1998). Overproduction and one-step purification of the N5,N10-methenyltetrahydromethanopterin MAD data and dataset NatR3 were collected at the BW6 beamline at cyclohydrolase (Mch) from the hyperthermophilic Methanopyrus DESY in Hamburg. After identifying the exact wavelength of the kandleri. Extremophiles 2, 15-22. Lu–LIII-edge by a fluorescence scan, two data sets Lu1 and Lu2 at 10. Breitung, J., Schmitz, R.A., Stetter, K.O. & Thauer, R.K. (1991). the peak wavelength and at the inflection point were measured in an N5,N10-Methenyl-tetrahydromethanopterin cyclohydrolase from the inverse beam experiment. Two data sets (Lu3 and Lu4) of a second extreme thermophile Methanopyrus kandleri: increase of catalytic crystal were recorded at the remote and the peak wavelength. The efficiency (kcat/KM) and thermostability in the presence of salts. Arch. statistics of data collection are summarized in Table 1. All data were Microbiol. 156, 517-524. integrated and scaled with the HKL [46] and the CCP4 [47] suites. 11. Pomper, B.K., Vorholt, J.A., Chistoserdova, L., Lidstrom, M.E. & Thauer, R.K. (1999). A methenyl-tetrahydromethanopterin cyclohydrolase (Mch) One lutetium-binding site was identified in the anomalous Patterson and a methenyl-tetrahydrofolate cyclohydrolase (Fch) in map of Lu1. Heavy-atom position and anomalous scattering coeffi- Methylobacterium extorquens AM1. Eur. J. Biochem. 261, 475-480. cients were refined and phases calculated using the program 12. Greenberg, D.M. (1963). Synthesis and transformations of folic SHARP [48]. SHARP was useful to handle the large anisotropy of coenzymes. II. Cyclohydrolase (5,10-methenyl to 10-formyl). In the heavy atom binding site in the z direction and to accurately esti- Methods in Enzymology Vol. VI (Colowick, S.P. and Kaplan, N.O., mate the anomalous scattering contribution from a refinement of the eds.), pp. 386-387, Academic Press, New York, USA. anomalous scattering factors. The phasing statistics of the data sets 13. Thauer, R.K., Klein, A.R. & Hartmann, G.C. (1996). Reactions with are summarized in Table 1. After solvent flattening with the program molecular hydrogen in microorganisms. Evidence for a purely organic SOLOMON [46], a readily interpretable electron density (Figure 8) hydrogenation catalyst. Chem. Rev. 96, 3031-3042. 14. DiMarco, A.A., Bobik, T.A. & Wolfe, R.S. (1990). Unusual coenzymes was obtained. of methanogenesis. Annu. Rev. Biochem. 59, 355-394. 15. Allaire, M., Li, Y.G., MacKenzie, R.E. & Cygler, M. (1998). The 3-D Model building and refinement structure of a folate-dependent dehydrogenase/cyclohydrolase A starting model containing all amino acids was built into the solvent- bifunctional enzyme at 1.5 Å resolution. Structure 6, 173-182. flattened electron-density map with the program O [49] and subjected to 16. Huber, R., Kurr, M., Jannasch, H.W. & Stetter, K.O. (1989). A novel a simulated annealing protocol using the maximum likelihood target of the group of abyssal methanogenic archaebacteria (Methanopyrus) program CNS [50]. Subsequent cycles of interactive model building and growing at 110°C. Nature 342, 833-834. positional and isotropic temperature-factor refinement resulted in a final 17. Kurr, M., et al., & Stetter, K.O. (1991). Methanopyrus kandleri, gen. and sp. nov. represents a novel group of hyperthermophilic model with the parameters as given in Table 2. The quality of the refine- methanogens, growing at 110°C. Arch. Microbiol. 156, 239-247. ment was checked using the program PROCHECK [51]. No outliers 18. Shima, S., Hérault, D.A., Berkessel, A. & Thauer, R.K. (1998). were found in the Ramachandran plot. The structure of the R3 crystal Activation and thermostabilization effects of cyclic 2,3- form was solved by molecular replacement using the program AMoRe diphosphoglycerate on enzymes from the hyperthermophilic [52] and refined using the same protocol as for the P6322 crystal form. Methanopyrus kandleri. Arch. Microbiol. 170, 469-472. 19. Klockenkaemper R. (1997). Total-Reflection X-ray Fluorescence Accession numbers Analysis. John Wiley & Sons, New York, USA. The coordinates of Mch have been deposited with the Brookhaven Protein 20. Yoder, M.D. & Jurnak, F. (1995). The refined three dimensional Data Bank, accession code 1QLM, and are available upon publication. structure of pectate lyase C from Erwinia chrysanthemi at 2.2 Å resolution-implications for an enzymatic mechanism. Plant Physiol. 107, 349-364. Acknowledgements 21. Holm, L. & Sander, C. (1993). Secondary structure comparison by We thank Hartmut Michel for generous support, Roy Lancaster for help alignment of distance matrices. J. Mol. Biol. 233, 123-138. with the program SHARP and Bernd Kolbesen for support with the 22. Miller, S., Lesk, A.M., Janin, J. & Chothia, C. (1987). The accessible TXRF measurements. surface area and stability of oligomeric proteins. Nature 328, 834-836. 1268 Structure 1999, Vol 7 No 10

23. Janin, J., Miller, S. & Chothia, C. (1988). Surface, subunit interfaces 46. Otwinowski, Z. (1993). Oscillation data reduction program. In Data and interior of oligomeric proteins. J. Mol. Biol. 204, 155-164. Collection and Processing (Sawyer, L., Isaacs, N. & Bailey S. eds.) 24. Wilmanns, M., Priestle, J.P., Niermann, T. & Jansonius, J.N. (1992). Three pp. 556-562. CLRC Daresbury Laboratory, Warrington, UK. dimensional structure of the bifunctional enzyme phosphoribo- 47. CCP4 (1994). The CCP4 suite: programs for protein crystallography. sylanthranilate isomerase/indoleglycerolphosphate synthase from Acta Crystallogr. D 50, 760-763. Escherichia coli refined at 2 Å resolution. J. Mol. Biol. 223, 477-508. 48. Bricogne, G. & de la Fortelle, E. (1997). SHARP: A maximum 25. Pieper, U., Kapadia, G., Mevarech, M. & Herzberg, O. (1998). likelihood heavy Atom Parameter Refinement Program for MIR and Structural features of halophilicity derived from the crystal structure of MAD Methods. In Methods in Enzymology (Sweet, R.M. & Carter, dihydrofolate reductase from the dead sea halophilic archaeon C.W. Jr. eds.) 276, pp. 472-492, Academic Press, New York, USA. Haloferax volcanii. Structure 6, 75-88. 49. Jones, T.A., Zhou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). 26. Fontecilla-Camps, J.C., et al., & Kisliuk, R.L. (1979). Absolute Improved methods for binding protein models in electron density configuration of biological tetrahydrofolates. A crystallographic maps and the location of errors in these models. Acta Crystallogr. A determination. J. Am. Chem. Soc. 101, 6114-6115. 47, 110-119. 27. Burdick, B.A., Benkovic, P.A. & Benkovic, S.J. (1976). Studies on 50. Brünger A., et al. & Warren G.L. (1998). Crystallography and NMR models for tetrahydrofolic acid 8: hydrolysis and methoxyaminolysis of system: a new software package for macromolecular structure amidines. J. Am. Chem. Soc. 99, 5716-5724. determinations. Acta Crystallogr. D 54, 905-921. 28. Benkovic, S.J., Bullard, W.P. & Benkovic, P.A. (1972). Studies on 51. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. models for tetrahydrofolic acid 3: hydrolytic interconversion of the (1993). PROCHECK a program to check the stereochemical quality tetrahydroquinoxaline analogs at the formate level of oxidation. of protein structures. J. Appl. Crystallogr. 26, 283-291. J. Am. Chem. Soc. 94, 7542-7549. 52. Navaza, J. (1994). AMoRe: an automated package for molecular 29. Schlegel, B.P., Ratnam, K. & Penning, T.M. (1998). Retention of NADPH- replacement. Acta Crystallogr. A 50, 157-163. linked quinone reductase activity in an aldo–keto reductase following 53. van Beelen, P., et al., & Haasnoot, C.A. (1984). Elucidation of the mutation of the catalytic tyrosine. Biochemistry 37, 11003-11011. structure of methanopterin, a coenzyme from Methanobacterium 30. Malashkevich, V.N., et al., & Jansonius, J.N. (1995). Structural basis for thermoautotrophicum, using two dimensional nuclear magnetic the catalytic activity of aspartate aminotransferase K258H lacking the resonance techniques. Eur. J. Biochem. 138, 563-571. pyridoxal 5′-phosphate-binding lysine residue. Biochemistry 34, 405-414. 54. Barton G. (1993). ALSCRIPT: a tool to format multiple sequence 31. Dym, O., Mevarech, M. & Sussman, J.L. (1995). Structural features alignments. Protein Eng. 6, 37-40. that stabilize halophilic malate dehydrogenase from an 55. Evans, S.V. (1993) SETOR: a hardware lighted three dimensional archaebacterium. Science 267, 1344-1346. solid model representation of macromolecules. J. Mol. Graph. 32. Frolow, F., Harel, M., Sussman, J.L., Mevarech, M. & Shoham, M. 11, 134-138. (1996). Insights into protein adaptation to a saturated salt 56. Nicholls, A., Bharadwaj, J. & Honig, B. (1993). GRASP: graphical environment from the crystal structure of a halophilic 2Fe–2S- representation and analysis of surface properties. Biophys. J. ferredoxin. Nat. Struct. Biol. 3, 452-458. 64, 166-170. 33. Russo, A., Rullo, R., Nitti, G., Masullo, M. & Bocchini V. (1997). Iron 57. Kraulis, P.J. (1991). MOLSCRIPT a program to produce both detailed superoxide dismutase from the archaeon Sulfolobus solfataricus: and schematic plots of protein structures. J. Appl. Crystallogr. average hydrophobicity and amino acid weight are involved in the 24, 946-950. adaptation of proteins to extreme environments. Biochim. Biophys. 58. Merritt, E.A. & Murphy, M.E. P. (1994). Raster3D version 2.0 A Acta 1343, 23-30. program for photorealistic molecular graphics. Acta Crystallogr. D 34. Jaenicke R. (1991). Protein stability and molecular adaptation to 50, 869-873. extreme conditions. Eur J. Biochem. 202, 715-728. 35. Cendrin, F., Chroboczek, J., Zaccai, G., Eisenberg, H. & Mevarech M. (1993). Cloning sequencing and expression in Escherichia coli of the gene coding for malate dehydrogenase of the extremely halophilic archaebacterium Haloarcula marismortui. Biochemistry 32, 4308-4313. 36. Elcock, A.H. & McCammon, J.A. (1998). Electrostatic contributions to the stability of halophilic proteins. J. Mol. Biol. 280, 731-748. 37. Danson, M. & Hough, D.W. (1997). The structural basis of protein halophilicity. Comp. Biochem. Physiol. 117A, 307-312. 38. Noelling, J., et al., & Reeve J.N.(1996). Phylogeny of Methanopyrus kandleri based on methyl coenzyme M reductase operons. Int. J. Syst. Bact. 46, 1170-1173. 39. Ermler, U., Merckel, M.C., Shima, S. & Thauer, R.K. (1997). Formyl- methanofuran:tetrahydromethanopterin formyltransferase from Methanopyrus kandleri: new insights into salt dependence and thermostability. Structure 5, 635-646. 40. Auerbach, G., et al., & Jacob, U. (1997). Closed structure of phosphoglycerate kinase from Thermotoga maritima reveals the catalytic mechanism and determinants of thermal stability. Structure 5, 1475-1483. 41. Yip, K.S.P., et al., & Rice, D.W. (1995). The structure of Pyrococcus furiosus glutamate dehydrogenase reveals a key role for ion pair networks in maintaining enzyme stability at extreme temperatures. Structure 3, 1147-1158. 42. Villeret, V., et al., & Van Beeumen, J. (1998). The crystal structure of Pyrococcus furiosus ornithinecarbamoyltransferase reveals a key role for oligomerization in enzyme stability at extremely high temperatures. Proc. Natl Acad. Sci. USA 95, 2801-2806. 43. Barlow, D.J. & Thornton, J.M. (1983). Ion pairs in proteins. J. Mol. Biol. 168, 857-885. 44. Auerbach,G. et al., & Jaenicke, R. (1998). Lactate dehydrogenase from the hyperthermophilic bacterium Thermotoga maritima: the Because Structure with Folding & Design operates a crystal structure at 2.1 Å resolution reveals strategies for intrinsic protein stabilization. Structure 6, 769-781. ‘Continuous Publication System’ for Research Papers, this 45. Shima, S., et al., & Thauer, R.K. (1998). Lyotropic-salt-induced changes paper has been published on the internet before being printed in monomer/dimer/ tetramer association equilibrium of formyltransferase from the hyperthermophilic Methanopyrus kandleri in relation to the (accessed from http://biomednet.com/cbiology/str). For activity and thermostability of the enzyme. Eur. J. Biochem. 258, 85-92. further information, see the explanation on the contents page.