How an Enzyme Binds the C1 Carrier Tetrahydromethanopterin

Tetrahydromethanopterin (H4 MPT) is a tetrahydrofolate analogue involved as a C1 carrier in the metabolism of various groups of microorganisms. How H4MPT is bound to the respective C1 unit converting enzymes remained elusive. We describe here the structure of the homopentameric formaldehyde-activating enzyme (Fae) from Methylobacterium extorquens AM1 established at 2.0 Å without and at 1.9 Å with methylene-H4MPT bound. Methylene-H4MPT is bound in an “S”-shaped conformation into the cleft formed between two adjacent subunits. Coenzyme binding is accompanied by side chain rearrangements up to 5 Å and leads to a rigidification of the C-terminal arm, a formation of a new hydrophobic cluster, and an inversion of the amide side chain of Gln88. Methylene-H4MPT in Fae shows a characteristic kink between the tetrahydropyrazine and the imidazolidine rings of 70° that is more pronounced than that reported for free methylene-H4MPT in solution (50°). Fae is an essential enzyme for energy metabolism and formaldehyde detoxification of this bacterium and catalyzes the formation of methylene-H4MPT from H4MPT and formaldehyde. The molecular mechanism ofthis reaction involving His22 as acid catalyst is discussed.

H 4 F has been well established, the presence of H 4 MPT in various groups of Archaea and bacteria is just emerging. Originally, after the discovery of H 4 MPT (2,3) it was thought to be restricted to methanogenic Archaea where it represents the essential cofactor of a series of enzymes that form the integral part of the process of methane formation from H 2 and CO 2 (4). However, in the last decade, the much wider occurrence and importance of H 4 MPT became evident; H 4 MPT was found in sulfate-reducing Archaea that are phylogenetically closely related to methanogenic Archaea (5,6). More surprisingly, H 4 MPT was later found to be present also outside the Archaea and shown to be an essential cofactor of the central metabolism of many methylotrophic ␣-, ␤-, and ␥-proteobacteria (7,8), a group of bacteria that also comprises methanotrophic bacteria. Very recently, it was found that not only aerobic methane oxidation relies on H 4 MPT but also anaerobic oxidation of methane is likely to depend on H 4 MPT (9), a process catalyzed by a group of Archaea closely related to the Methanosarcinales (10,11). All these organisms are highly specialized in C 1 metabolism and are of great ecological importance in the global carbon cycle (12,13). The recent documentation of functional H 4 MPT-dependent enzymes in the enigmatic bacterial group of Planctomycetes re-opened the debate of the evolution of H 4 MPT and H 4 MPT-dependent enzymes because phylogenetic analysis places the Planctomycetes sequences as distant from the archaeal counterparts as from their proteobacterial counterparts (14).
Functionally the most important difference between H 4 MPT and H 4 F is the electron-donating methylene group of H 4 MPT in position 1c (Fig. 1), which is conjugated to N 10 through the aromatic ring, whereas H 4 F has an electron withdrawing carbonyl group in this position (1,15). One consequence is that the redox potentials of the N 5 , N 10 -methenyl-H 4 MPT ϩ /N 5 , N 10methylene-H 4 MPT couple (Ϫ390 mV) and of the N 5 , N 10 -methylene-H 4 MPT/N 5 -methyl-H 4 MPT couple (Ϫ310 mV) are almost 100 mV more negative than the corresponding H 4 F couples. The structural and functional differences between H 4 MPT and H 4 F are reflected in the finding that most of the enzymes catalyzing the interconversion of their C 1 derivatives are highly specific for H 4 (20), and F 420 -dependent methylene-H 4 MPT reductase. So far none of these enzymes could be crystallized in complex with H 4 MPT or one of its derivatives. Only the conformation of methylene-H 4 MPT bound to H 2 -forming methylene-H 4 MPT dehydrogenase (Hmd) was determined by two-dimensional NMR spectroscopy (21).
Here we describe the structure of the formaldehyde-activating enzyme Fae from Methylobacterium extorquens AM1 with and without methylene-H 4 MPT bound. The enzyme catalyzes the condensation of formaldehyde with H 4 MPT to methylene-H 4 MPT (22). This reaction also proceeds spontaneously but only at a lower rate. Fae was discovered in M. extorquens AM1, which grows aerobically at the expense of methanol oxidation to CO 2 involving N 5 , N 10 -methylene-H 4 MPT, N 5 , N 10 -methenyl-H 4 MPT ϩ , and N 5 -formyl-H 4 MPT as intermediates (23). Fae appears to be specific for H 4 MPT; no formaldehyde-H 4 F condensing activity could be found with purified Fae (22). Fae minus mutants of M. extorquens AM1 are no longer capable of growth on methanol and are inhibited by trace amounts of formaldehyde indicating that Fae has a vital function in methylotrophic energy metabolism and formaldehyde detoxification (22). The importance of Fae for methylotrophy is also reflected by its high abundance in the cytoplasm of the cell (24) and its presence in diverse methylotrophic bacteria (23). Functional orthologs of Fae are also present in some methanogenic Archaea (22) and in Planctomyces species (14). The widespread occurrence of Fae orthologs suggests that formaldehyde may play an unknown but important role in a broad group of prokaryotes.

EXPERIMENTAL PROCEDURES
M. extorquens AM1 is the strain deposited under DSM 1338 in the Deutsche Sammlung von Mikroorganismen und Zellkulturen (Braunschweig, Germany).
Heterologous Overproduction of Fae and Purification-Amplification of the fae gene was achieved with Expand-DNA-polymerase (Roche Applied Science), the primers 5Ј-GAGACCCCATATGGCAAAAATCAC-CAAGGTTC-3Ј (sense, the NdeI site is underlined) and 5Ј-CTGCCCAG-GAATTCCTCCGATCTAAGCGTT-3Ј (antisense, the EcoRI site is underlined), and chromosomal DNA of M. extorquens AM1 as a template. The PCR product was digested with NdeI and EcoRI and ligated into the pET17b expression vector and then introduced into Escherichia coli BL21 (DE3) pLysS. Each transformant of E. coli BL21 (DE3) pLysS was grown aerobically at 37°C on minimal medium M9 (25) supplemented with ampicillin (100 g ml Ϫ1 ) and chloramphenicol (50 g ml Ϫ1 ). When the ⌬A 600 of the culture reached 0.5, cells were induced by 2 mM isopropyl-␤-D-thiogalactopyranoside. After 4 h, the cells were harvested by centrifugation at 4200 ϫ g at 4°C. Selenomethionine-labeled protein was produced using the method of metabolic inhibition (26).
Non-labeled Fae and selenomethionine-labeled protein were purified under aerobic conditions as described in Vorholt (45). H 4 MPT is similar to H 4 F in that both compounds consist of a reduced pterin linked to an arylamine via a methylene group with C 1 units binding at N 5 ,N 10 or both N 5 and N 10 . In methylene-H 4 MPT the C 1 unit bridges N 5 and N 10 forming an imidazolidine ring, which is condensed to the tetrahydropyrazine ring of the reduced pterin. To the left of the phenyl ring the structure is referred to as head group and to the right of the phenyl ring as the tail of H 4 MPT and H 4 F, respectively. marburgensis (DSM 2133) (27) and stored in 10 mM MOPS/KOH buffer pH 7. H 4 MPT from M. extorquens AM1 differs from that in M. marburgensis by lacking the phosphate and hydroxyglutarate group. It has been shown, however, that enzymes from M. extorquens AM1 are equally active with H 4 MPT and with the dephospho form (22).
Crystallization and Data Collection-Crystallization trials were performed with non-labeled and selenomethionine-labeled Fae at 4°C under aerobic conditions and with enzyme in the presence of H 4 MPT at 8°C under strictly anaerobic conditions. Within a hanging drop experiment each drop consisted of 1 l of enzyme solution (13 mg/ml) and 1 l of reservoir solution. Crystals of non-labeled Fae as well as selenomethionine-labeled Fae grew in a reservoir solution composed of 0.2 M calcium chloride ϫ 2H 2 O, 0.1 M sodium acetate ϫ 3H 2 O, pH 4.6, and 10 -20% (v/v) isopropanol. Their space group was P4 3 2 1 2, and the lattice parameters were a ϭ b ϭ 120.7 Å and c ϭ 205.9 Å. For crystallization in the presence of H 4 MPT, the enzyme solution (13 mg/ml) was supplemented with 5 mM H 4 MPT and then combined with the reservoir solution containing 0.1 M HEPES/NaOH, pH 7.5, and 20% (w/v) polyethylene glycol 10,000. The space group was P2 1 , and the lattice parameters were a ϭ 48.9 Å, b ϭ 112.6 Å, c ϭ 72.0 Å and ␣,␥ ϭ 90°, ␤ ϭ 91.0°. Data were collected at ID14-4 and ID29 beamlines at the European Synchrotron Radiation Facility, Grenoble, France (Table I). Processing and scaling were performed with XDS (28) and Denzo/ Scalepack (29).
Phase Determination and Refinement-Phases were determined using the multiple anomalous wavelength dispersion method with selenium as anomalous scatterer. The selenium sites were found using SHELXD (30) and further refined using SHARP (31). The phases were calculated with SHARP and improved by solvent flattening (32) assuming a solvent content of 70%. 5-fold molecular averaging within DM (33) resulted in an excellent electron density map where ϳ80% of the chain could be traced by the automated model building program MAID (34). Except for 10 residues at the C-terminal end the residual model could be manually incorporated using O (35). Iterative cycles of refinement and manual model building were carried out using the program package CNS (36) and O. The refinement statistics are given in Table I. The structure of the H 4 MPT-bound enzyme was solved by molecular replacement using the program EPMR (37) with the coordinates of Fae without bound substrate as the search model. After initial refinement the C-terminal amino acids disordered in the coenzyme free structure and H 4 MPT later replaced by methylene-H 4 MPT were modeled into the density. The results of the refinement are listed in Table I. The quality of the models was checked with PROCHECK (38).

Structure of Fae with and without Methylene-H 4 MPT
Bound-Formaldehyde-activating enzyme Fae in the absence and presence of methylene-H 4 MPT was structurally characterized in two crystal forms at a resolution of 2.0 and 1.9 Å, respectively ( Table I) Fae is organized as homopentameric protein complex with dimensions of about 70 Å ϫ 70 Å ϫ 40 Å ( Fig. 2A). Each monomer consists of one compact domain that belongs to the class of ␣/␤ proteins. The central sheet contains five strands (␤1-␤5) joined in the order ␤1, ␤2, ␤5, ␤4, and ␤3, with only ␤4 and ␤5 oriented in parallel. Helix ␣1 arranged after strand ␤2 packs against one side of the sheet; helices ␣2 (after strand 4) and ␣3 (after strand 5) pack against the other (Fig. 2B). This architecture is somehow reminiscent to that of the ribosomal protein S 5 domain 2-like family to which for example the elongation factor G (39), the ribosomal protein S 5 (40), and the galactokinase, homoserine kinase, mevalonate kinase, and phosphomevalonate kinase family belong. According to the program DALI (41), the root mean square deviations between Fae and elongation factor G and phosphomevalonate kinase (42) are 2.7 Å and 3.1 Å using ϳ60% of the C␣ positions for calculation. In comparison, the root mean square deviation between the five monomers in the asymmetric unit is ϳ0.15 Å that between the pentamers of the two crystal forms is 0.6 Å. A rare topological feature of this fold is the ␤4␣2␤5 left-handed cross-over linkage that appears to be crucial for the integrity of the fold (Fig. 2B). Compared with the other family members helix ␣2 in Fae is longer and part of the H 4 MPT binding site. The major difference between Fae and the other family members is an insertion between strand ␤2 and ␤4 (Fig. 2B) consisting of helix ␣1, strands ␤3, and an unusual protrusion at the end of strand ␤3 (see below).
The pentamer can be subdivided into three circular layers built up of an ␣-helical, a ␤-sheet, and again an ␣-helical region ( Fig. 2A). The inner ring is formed by the five tightly linked helices ␣1 of the insertion. The outer layer is formed by helices ␣2, ␣3, and ␣4 with the latter being connected to helix ␣2 of the next monomer. The central ring consists of the five fivestranded ␤-sheets with each of them oriented roughly in a perpendicular manner to the neighboring sheet. The hydrophobic core of each sheet is enlarged by helices ␣1 of the next monomer at the inner side and of the C-terminal segment of the previous monomer at the outer side. A channel crosses the entire pentamer along the 5-fold axis ( Fig. 2A) and is occupied with several solvent molecules and extra electron density that could not be assigned. The binding site for methylene-H 4 MPT is located in a 20 Å long, 8 Å wide, and 12 Å deep cleft at the interface between two adjacent subunits called A and B (Fig. 2) with the constituting residues highly conserved (Fig. 2C). Upon methylene-H 4 MPT binding the width of the cleft is slightly decreased because of a rotation of helix ␣2 of about 5°and because of a displacement of strands ␤1, ␤2, and ␤5 in the range of 0.3-0.5 Å. Additionally, the flexible C-terminal arm of subunit A (A160 -A166) is rigidified, and the protrusion of subunit B is shifted around 2 Å toward the coenzyme.

Conformation of Methylene-H 4 MPT When Bound to Fae-
Methylene-H 4 MPT binds to the binding cleft with a high occupancy of about 80%. However, the temperature factor increased dramatically from the pterin and imidazolidine rings (30 Å 2 ) via the benzene ring (38 Å 2 ), the ribitol group (50 Å 2 ), the ribose group (68 Å 2 ), to the phosphate group (78 Å 2 ) indicating only an excellent electron density of the functionally relevant head group. The 2-hydroxyglutarate group is located in the bulk solvent (Fig. 3) and is not visible in the electron density map. Note that Fae from M. extorquens AM1 was crystallized to-  Fig. 2, A and B, and Figs. 3 and 4 were generated with MOLSCRIPT (46) and RASTER3D (47). C, molecular surface representation of the Fae pentamer highlighting the five methylene-H 4 MPT binding clefts and the high degree of conservation of their constituting residues. The surface was colored in blue when the equivalent residues in at least eight of the nine aligned sequences were identical to Fae from M. extorquens AM1 (see supplemental data). The figure was generated with GRASP (48).
gether with H 4 MPT from M. marburgensis rather than with the shorter H 4 MPT from M. extorquens AM1 (7). H 4 MPT from M. marburgensis contains 11 asymmetric carbons (Fig. 1). The quality of the electron density map allowed us to deduce the stereoconfiguration of five of these, 7a, 6a, 11a, 2c, and 3c ( Fig.  1), which agreed with that determined previously by two-dimensional NMR spectroscopy (43).
Methylene-H 4 MPT is accommodated into its binding site in an "S"-shaped conformation with the S positioned in a perpendicular manner to the front side of the cleft (Fig. 3). The pterin ring points toward the channel bottom, the imidazolidine ring and the phenyl ring are attached roughly parallel to the length of the cleft, and the ribose and phosphate groups are directed toward the bulk solvent. The S shape of methylene-H 4 MPT is the result of two kinks (Fig. 3A). The first sharp kink of ϳ70°i s located between the pterin and the imidazolidine rings around the N 5 -C 6a bond. An additional small rotation between the imidazolidine and the phenyl rings results in a nearly perpendicular orientation between the pterin and the phenyl ring. The second kink of roughly 90°is performed within the ribitol group. This conformation of methylene-H 4 MPT implicates that solely the pterin ring is shielded from bulk solvent by the described roof. The rest of the methylene-H 4 MPT including the imidazolidine ring is at least partly solvent accessible.
The conformation of methylene-H 4 MPT in the Fae-methylene-H 4 MPT complex (as described by this work) is different from the conformations of methylene-H 4 MPT in solution or when bound to Hmd, which have previously been determined by two-dimensional NMR spectroscopy (21). The major conformational surprise of methylene-H 4 MPT in Fae is the large kink angle of about 70°between the pterin and the imidazolidine ring that is in solution only 50°and when bound to Hmd only 40° (Fig. 4). This remarkable change necessitates different conformations of the sp3-configured C 6a and C 7a atoms of the tetrahydropyrazine ring. The C 6a atom has to point to the Si face in Fae but to the Re face in solution and when bound to Hmd. Consequently, atom C 7a of methylene-H 4 MPT bound to Fae is oriented to the Re face that leads to a conformation of the C 13a atom perpendicular to the pterin ring, whereas the Si face orientation leads to an equatorial position as found in the free and Hmd bound form (Fig. 4). Obviously, the protein scaffold of Fae substantially influences the conformation of methylene-H 4 MPT upon binding. For example, methylene-H 4 MPT in the conformation found in solution would interfere with Phe A166 of Fae that could not evade because of its contact to Val B81 . A related conformational variability of the kink is expected for methylene-H 4 F dependent enzymes, although an enzyme-methylene-H 4 F complex is so far not structurally characterized.
Interactions between Fae and Methylene-H 4 MPT-The increasing flexibility along the elongated molecule is reflected in a parallel decrease of the protein-cofactor interactions. Only two hydrogen bonds and a few van der Waals contacts are formed between the protein matrix and the tail groups of H 4 MPT (Fig. 3A). In other words binding is essentially based on interactions between the polypeptide chain and the catalytically relevant head groups. The conserved residues Asp A24 , Lys A71 , Leu B53 , and Gln B88 form specific hydrogen bond interactions toward the N 1 , NH 2 2a , N 3 H, and O 4a atoms (Fig. 3A). The head groups including the imidazole group of His A22 are encircled by two hydrophobic belts one above and the other below the pterin ring. The first belt includes Phe A119 , Val A20 , Leu A13 , Pro B58 , Pro A165 , and Phe A166 ; the second belt contains Val A72 , Phe A119 , the hydrophobic portions of Lys A71 and Thr B50 , Leu B52 , Phe B84 , Val B81 , and Ala B80 . The two belts partly touch each other and are opened at the front of the cleft with the shortest distances of ϳ7.5 Å between Phe A119 and Phe A166 and between Val A72 and Ala B80 (Fig. 3A). This entrance provides access to the catalytically relevant N 5 , N 10 , and C 14a atoms of the imidazolidine ring. Of particular importance for the adjustment of the ring conformations are His A22 and Leu B52 sandwiching the pterin ring and Leu B52 and Phe B84 that are positioned in the groove between the pterin, imidazolidine, and phenyl rings (Fig. 3A). Interestingly, the methyl groups of methylene-H 4 MPT not present in methylene-H 4 F (Fig. 1) mainly interact with invariant hydrophobic side chains of the C-terminal arm (Fig. 3B). Atom C 12a is in contact with Phe A166 , and atom C 13a is in contact with His A164 , Phe A166 , and Pro B58 , with the latter protruding from the described protrusion at the end of strand ␤3.
Interestingly, the binding site of methylene-H 4 MPT cannot be considered as prebuilt. In the empty enzyme Phe A119 , Phe B84 and His A22 point into the coenyzme binding site but evade the arriving methylene-H 4 MPT by movements up to 5 Å thereby inducing many additional conformational changes. For example, the rotation of Phe A119 induces substantial rearrangements of the side chains of His A121 and Glu A123 and allows Lys A71 to partially fill out the generated free place. Likewise, Phe B84 swings toward the channel bottom accompanied by a shift of Leu B52 to the Re side of the pterin ring that induces a movement of Thr B50 value of 1.9 Å. Consequently, a modeling of methylene-H 4 MPT to the empty Fae structure would not be possible.
The mentioned residues are conserved in Fae from different organisms but not in any of the other H 4 MPT-specific enzymes. A common binding motif for the C 1 carrier was not found. This was also not to be expected because the known H 4 MPT specific enzymes are not similar on the sequence nor on the structural level. This also holds true for the different H 4 F specific enzymes, which also do not show a common H 4 F binding motif.
Selectivity of Fae for Methylene-H 4 MPT Rather Than for Methylene-H 4 F-Fae catalyzes the reaction of formaldehyde with H 4 MPT. Using the same assay a formation of methylene-H 4 F from formaldehyde and H 4 F was not observed (22). The specificity of Fae for H 4 MPT is also indicated by the finding that the rate of condensation of formaldehyde and H 4 MPT was not inhibited by the addition of an excess of H 4 F. 2 This finding needs discussion because except for two methyl groups H 4 MPT and H 4 F only differ in their tail groups, but as described the ribitol, ribose, and phosphate groups of methylene-H 4 MPT appear to contribute only slightly to binding (Fig.  3). Although these few interactions as well as an interference between the protein and the formylglutamate tail groups in H 4 F might be crucial for selectivity the structural data support The reaction can be subdivided in a nucleophilic addition and a nucleophilic substitution process. A key function in formaldehyde activation and catalysis is attributed to His A22 , which is strictly conserved. a binding mechanism that attributes a key function to the additional methyl groups (Fig. 3B). Accordingly, three of four van der Waals contacts between the methyl groups are formed to side chains of the C-terminal arm that might contribute to its partial fixation. A simultaneous conformational change of several side chains at the Re side of methylene-H 4 MPT generates a hydrophobic cluster composed of the tetrahydropyrazine and the imidazolidine rings, Leu A13 , His A164 , Pro A165 , Phe A166 , and Pro B58 and most interestingly triggers an inversion of the amide group of Gln B88 . This exchange of the amide oxygen and amine groups is the prerequisite to form two hydrogen bonds to methylene-H 4 MPT and one hydrogen bond to His A164 of the C-terminal arm (Fig. 3B). Thus, preferred binding of H 4 MPT against H 4 F is not only accomplished by the quantitatively small van der Waals interactions between the methyl groups and the protein but by an induced cooperative process that enhances both methylene-H 4 MPT binding and the fixation of the C-terminal arm. The interactions between the C-terminal arm and the methyl groups of methylene-H 4 MPT might additionally influence the kink angle between the pterin and imidazolidine rings and thus the binding energy between methylene-H 4 MPT and Fae.
Formaldehyde Binding Site and Enzymatic Mechanism-Attempts to determine a structure of Fae in complex with formaldehyde failed, but an attractive binding site of the substrate is offered by the structure of the Fae-methylene-H 4 MPT complex. At first glance, an activation of formaldehyde by the amine group of Lys A171 appears to be chemically plausible (Fig. 3A), but its fixation by a large number of interactions and the absence of space for formaldehyde binding without pushing H 4 MPT out of its binding site reject this possibility. More attractively, a site either occupied with a solvent molecule or an unknown molecule (depending on the considered cleft of the asymmetric unit) is positioned parallel to the imidazolidine ring and is accessible from bulk solvent. Furthermore, the oxygen atom of formaldehyde can be modeled into the site of the solvent molecule or into a protrusion of the electron density of the unknown molecule ϳ4 Å apart from the N 5 atom of methylene-H 4 MPT and 3 Å apart from the N␦ 1 atom of the highly conserved His A22 . Despite its fairly hydrophobic environment His A22 is probably protonated as its N⑀ 2 atom donates a hydrogen bond to the negatively charged O⑀ 1 atom of Glu A11 (Fig. 3A). Assuming the oxygen atom as fixed, the methylene group of formaldehyde can be placed in front of the N 5 atom the distance of 3 Å, which is optimal for a nucleophilic attack. In this way the formaldehyde binding site is approximately defined.
On the basis of the proposed formaldehyde position and of an assumed similarity between the conformation of H 4 MPT and the observed methylene-H 4 MPT a catalytic mechanism was outlined that essentially consists of a nucleophilic addition and a nucleophilic substitution reaction (Fig. 5). First, the nucleophilic N 5 atom attacks the carbonyl carbon of formaldehyde thereby forming a tetrahedral anionic intermediate state that become protonated by His A22 . The positively charged His A22 enhances the electrophilic properties of the carbonyl carbon of formaldehyde and serves as a general acid catalyst thereby playing a key role in the formaldehyde activation process. After formation of the hydroxymethylene-H 4 MPT adduct the N 5 nitrogen is presumably deprotonated (Fig. 5). Second, the N 10 atom nucleophilically attacks the hydroxymethylene carbon atom and the hydroxylate group is concomitantly released. His A22 might be already reprotonated prior to hydroxylate release and acts again as hydrogen donor. After deprotonation of the N 10 atom the product methylene-H 4 MPT is generated. Interestingly, the electron density suggests a flap of N 10 above the ring plane in agreement with the conformation of the imi-dazolidine ring of methylene-H 4 MPT in the Hmd but in contrast to that in solution (21) (Fig. 4).
This mechanism is in agreement with the results of kinetic experiments of spontaneous methylene-H 4 F formation from H 4 F and formaldehyde (44). The spontaneous reaction proceeds optimally under acidic conditions indicating that a protonation step is involved. Most likely, in the enzyme, the proton for this step is provided by protonated His A22 . Thus, the presented structure of the Fae-methylene-H 4 MPT complex is not only the prototype of how this cofactor binds to an enzyme, but it also provides insights into the mechanism of how the highly toxic intermediate formaldehyde is metabolized and thus detoxified.