81163201.Pdf

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

provided by Elsevier - Publisher Connector

Structure, Vol. 10, 1127–1137, August, 2002, 2002 Elsevier Science Ltd. All rights reserved. PII S0969-2126(02)00802-X Structure of Methylene-Tetrahydromethanopterin Dehydrogenase from Methylobacterium extorquens AM1

1,3 2 Ulrich Ermler, Christoph H. Hagemeier, H4MPT, the latter in a dephospho form, could be shown Annette Roth,1 Ulrike Demmer,1 in Methylobacterium extorquens AM1 [5, 6]. Dephos- 1,2,4 1 Wolfgang Grabarse, Eberhard Warkentin, pho-H4MPT lacks the terminal ␣-hydroxyglutaryl phos- and Julia A. Vorholt2,3,5 phate unit described for the cofactor from the methano- 1Max-Planck-Institut fu¨ r Biophysik genic archaeon Methanothermobacter marburgensis Heinrich-Hoffmann-Strasse 7 (formerly called Methanobacterium thermoautotro- D-60528 Frankfurt am Main phicum) [7].

Germany M. extorquens AM1 possesses two independent C1 2 Max-Planck-Institut fu¨ r Terrestrische pathways based on the use of the two pterin cofactors.

Mikrobiologie Experimental evidence is available that the H4MPT- Karl-von-Frisch-Strasse dependent pathway has a catabolic function in the oxi- D-35043 Marburg dation of reduced one-carbon substrates, such as meth- Germany anol or formaldehyde, and that the one-carbon units

bound to H4F are used as the source of biosynthetic carbon for purine and formyl methionine-tRNA synthesis 10 Summary from N -formyl-H4F [6, 8, 9]. The key enzyme of the serine cycle for formaldehyde assimilation, serine hy- 5 10 NADP-dependent methylene-H4MPT dehydrogenase, droxymethyltransferase, uses N ,N -methylene-H4Fas MtdA, from Methylobacterium extorquens AM1 cata- substrate [10]. This substrate is also required for thymi- lyzes the dehydrogenation of methylene-tetrahydro- dylate synthesis. All the H4F- and H4MPT-dependent en- methanopterin and methylene-tetrahydrofolate with zymes involved are essential for growth on methanol [5]. -NADP؉ as cosubstrate. The X-ray structure of MtdA In mitochondria and most bacteria, the N5,N10-methyl with and without NADP bound was established at 1.9 A˚ ene-H4F dehydrogenase and cyclohydrolase functions resolution. The enzyme is present as a homotrimer. are combined within a single bifunctional enzyme [2]. The ␣,␤ fold of the monomer is related to that of meth- Such a dual function enzyme is apparently absent in M.

ylene-H4F dehydrogenases, suggesting a common extorquens AM1. Instead, the proteobacterium contains evolutionary origin. The position of the active site is two monofunctional dehydrogenases, MtdA and MtdB, located within a large crevice built up by the two do- and two cyclohydrolases [6, 9, 11]. Both dehydroge- mains of one subunit and one domain of a second nases use pyridine nucleotides as cosubstrates, and

subunit. Methylene-H4MPT could be modeled into the aspects of the substrate specificities of these enzymes cleft, and crucial active site residues such as Phe18, are unusual. MtdA shows absolute specificity for NADP, Lys256, His260, and Thr102 were identified. The mo- but can catalyze both the dehydrogenation of methyl- ⌬ ЊϭϪ lecular basis of the different substrate specificities ene-H4MPT ( G 13 kJ/mol) and the reversible dehy- ⌬ Њϭϩ and different catalytic demands of MtdA compared to drogenation of methylene-H4F( G 3.5 kJ/mol), al-

methylene-H4F dehydrogenases are discussed. beit with a 20-fold lower catalytic efficiency as the H4MPT-dependent reaction (Figure 1) [6]. By contrast, Introduction MtdB is specific for the pterin substrate methylene- H4MPT but uses NAD and NADP as cosubstrate. It has been discussed that MtdB is involved in formaldehyde Tetrahydrofolate (H4F) and tetrahydromethanopterin oxidation to CO2 and generation of NADH, whereas MtdA (H4MPT) are cofactors able to carry one-carbon units between the formyl and methyl oxidation level. The one- is primarily an H4F-dependent enzyme in vivo [11]. Inter- carbon units are bound to the N5 and/or N10 nitrogen estingly, the mtdA gene is located directly upstream of fchA, which encodes methenyl-H4F cyclohydrolase. The atoms of the reduced pterins. H4F and H4MPT are structural analogs with respect to the pterin moiety but have gene mtdB is located in a gene cluster together with other H4MPT-dependent enzymes for formaldehyde oxi- distinct side chains: H4F has an electron-withdrawing carbonyl group in conjugation with the N10 atom via the dation to CO2, such as the formaldehyde-activating enzyme Fae, the methenyl-H4MPT cyclohydrolase Mch, aromatic ring, whereas H4MPT has an electron-donating and the formyl methanofuran:tetrahydromethanopterin methylene group in this position [1]. H4F is present in formyltransferase complex (Ftr) [8, 9, 12]. almost all organisms [2], whereas H4MPT was previously thought to be restricted to anaerobic methanogenic MtdA and MtdB have several properties in common archaea [3] but is now also found within aerobic methylo- with the two kinds of methylene-H4MPT dehydrogenases found in methanogenic archaea, the F420-dependent [13] trophic proteobacteria [4]. The occurrence of H4F and and H2-forming methylene-H4MPT dehydrogenases [14], as well as with the methylene-H F dehydrogenases from 3 4 Correspondence: [email protected] (U.E.), bacteria and eucarya [2]. All the dehydrogenases are [email protected] (J.A.V.) 4 Present address: GVCIC, BASF AG, 67065 Ludwigshafen, composed of only one type of subunit with a molecular Germany. 5 Present address: INRA/CNRS, BP27 Chemin de Borde Rouge, Key words: crystal structure; conformational change; NADP; tetra-

31326 Castanet-Tolosan, France. hydromethanopterin; tetrahydrofolate; C1 metabolism Structure 1128

Figure 1. Reaction of MtdA

Methylene-H4MPT (-H4F) is oxidized to meth- ϩ enyl-H4MPT (-H4F) with NADP as cosubstrate.

H4MPT is composed of a 7-methyl-6-ethyl- pterin, an aminobenzyl, a 1-desoxyribose, a ribose, a phosphate, and a 2-hydroxygluta-

rate moiety. H4F consists of a 6-methyl-pterin, a p-aminobenzoate, and a glutamate moiety. The two methyl groups painted in blue are

only present in H4MPT. Methylene and methenyl groups are indicated in red.

mass between 30 and 40 kDa, lack chromophoric pros- (see below), is novel. Whereas the NADP domains pro- thetic groups, exhibit a ternary complex catalytic mech- trude into the solvent, the three pterin domains assem- anism, and are pro-R specific with respect to the methyl- ble close to the noncrystallographic 3-fold axis (Figure ene cofactor [15]. However, analysis of the primary 3A). The interface between the pterin domains is pre- structure revealed that the dehydrogenases fall into four dominantly nonpolar (69% nonpolar, 31% polar) and different classes, the sequence identity between them buries 9% of the surface area per monomer [22]. The being only around 15%, whereas within one family the center of the trimer is closed by the three side chains sequence identity of the dehydrogenases is 30% or of His30 which point from the three subunits toward the higher. The three-dimensional structure of a methylene- pseudosymmetry axis.

H4MPT dehydrogenase from a methanogenic archaeon The pterin domain of MtdA is composed of an has not been solved yet. N-terminal segment (residues 2–105) with a ␤␣␤␣␤␣␤ Here we describe the structure of NADP-dependent folding unit and a C-terminal segment (residues 257– methylene-H4MPT/H4F dehydrogenase MtdA with the 288) that includes strand ␤12 and in part helices ␣9 and cofactor NADPϩ bound and we present a model for the ␣10 (Figures 3B and 3C). Both segments join to a five- binding of the pterin cofactors. The active site geometry stranded parallel ␤ sheet flanked by four ␣ helices on is discussed with respect to that of mono- and bifunc- both sides. A similar fold has been observed in several tional methylene-H4F dehydrogenases [16–18], illustrat- proteins (DALI) [23]. The most similar structure is the ing the interplay between conservation and specific transcriptional regulatory protein FixJ [24]. Strands ␤1 functional adaptation on an atomic ground. and ␤2 as well as helices ␣1 and ␣2 are involved in forming the interface between the subunits. Results and Discussion The fold of the NADP domain (residues 106–256) es- sentially corresponds to that found in the widespread Structure of MtdA dinucleotide binding family [25]. The major difference is MtdA from M. extorquens AM1 heterologously ex- that strand ␤10 connects strands ␤9 and ␤11 instead pressed in Escherichia coli BL21 was crystallized using of an ␣ helix (Figures 3B and 3C). The most similar PEG1500 as precipitant. Its structure was solved by members of the dinucleotide binding family are alanine the multiple isomorphous replacement method at 2.9 A˚ dehydrogenase [26] and methylene-H4F dehydroge- resolution and refined at 1.9 A˚ resolution (see Table 1). In nase/cyclohydrolase [16], with rms deviations com- addition, the enzyme-NADPϩ complex was structurally pared to MtdA of 2.4 A˚ and 2.2 A˚ , using more than 70% characterized at 1.9 A˚ resolution using the molecular of the C␣ atoms for superposition [23]. replacement method for phase determination. Figure 2 The two domains are arranged in such a way that illustrates the quality of the final electron density map a small cleft is formed at the interface between them of the latter structure. (Figures 3B and 3C). Moreover, the ␤ sheets of the do- In agreement with gel filtration experiments [6], the mains are oriented perpendicular to each other and their crystal structure revealed that MtdA is a homotrimer C-terminal ends point toward the interface. The domains (Figure 3A). Each of the subunits has a size of about are linked by the two long helices ␣4 and ␣9, the latter 63 A˚ ϫ 40 A˚ ϫ 35 A˚ and consists of two domains which being bent, and by several noncovalent interactions in- are designated pterin and NADP (binding) domains (Fig- cluding the ion pairs Arg167-Glu278, Arg135-Glu74, and ures 3B and 3C), both adopting an ␣/␤ twisted open Glu143-Lys285. Nevertheless, the NADP domain is mo- sheet structure. The fold of the subunits exhibits signifi- bile relative to the pterin domain. The superposition of cant similarities to that of the malic enzyme [19], the the three monomers in the asymmetric unit results in aspartate and the ornithine transcarbamoylase [20, 21], an rms deviation of around 0.5 A˚ between monomers, and the methylene-H4F dehydrogenase [16–18], but the compared to about 0.3 A˚ between individual domains. oligomeric organization, which is functionally essential Moreover, the NADP domain of one subunit is present in Structure of Methylene-H4MPT Dehydrogenase 1129

Table 1. Structure Determination

1 1 1 2 Data set MtdA1 MtdA2 MtdA3 HgAc2 Thimerosal K2PtCI4 TERPI Wavelength 1.05 0.84 1.54 1.54 1.54 1.54 1.54 Resolution (A˚ ) 1.9 1.9 2.9 2.9 2.9 2.9 2.9 Completeness (%) 93.8 95.3 93.6 93.1 85.3 78.4 63.1 Multiplicity 3.0 3.4 2.3 1.8 2.9 2.1 2.3 3 Rsym (%) 6.4 9.1 5.2 8.8 11.1 8.5 13.4 Phasing power4 (cent) 1.5 1.4 1.4 2.1 (acent) 1.8 1.9 1.5 1.9 5 RCullis (%) (cent) 0.77 0.75 0.84 0.7 (acent) 0.77 0.75 0.82 0.75 6 Rcryst (%) 19.0 18.6 7 Rfree (%) 22.9 22.2 Rmsd bond length (A˚ ) 0.014 0.011 Rmsd bond angles (Њ) 1.5 1.31

two distinct conformations; the rotation and translation 3B and 3C), similar to the other members of the dinucleo- parameters between them are 4Њ and 1.7 A˚ , respectively. tide binding family [25]. The average B factor is 30.2 A˚ 2 for NADPϩ compared to 27.0 A˚ 2 for the polypeptide chain, which is compatible with a complete occupancy NADP Binding ϩ ϩ of NADP . When considering the 3-fold noncrystallo- Analysis of crystals of MtdA harboring NADP revealed ϩ that the cosubstrate is located at the C-terminal end of graphic symmetry axis as fixed, NADP binding changes the central ␤ sheet of the NADP binding domain (Figures the conformation of the NADP domain relative to the pterin domain (Figure 4A). The rotation of about 6Њ causes an opening of the interdomain crevice. As men- tioned above, a fraction of one NADP domain in the MtdA structure without NADPϩ significantly deviates from the conformations of the others (for details, see Experimen- tal Procedures) and is similar to that found in the enzyme-NADPϩ complex. Therefore, we assume that in this system, NADPϩ binding freezes one of the conformations present rather than induces a conformational change. The protein in the conformation after NADPϩ

binding presumably forms the H4MPT binding site, so ϩ NADP has to bind prior to methylene-H4MPT. Interest- ingly, kinetic data of the homologous MtdB indicate a faster reaction when started with NADPϩ instead of

methylene-H4MPT [11]. Conformational changes upon NADPϩ binding were already expected, as crystals of form 1 crack upon soaking with NADPϩ. The elongated NADPϩ is multiply anchored to the protein matrix. In particular, the phosphate oxygens of the adenine ribose phosphate group (Figure 4B) form six hydrogen bonds to the side chains of Arg152, Lys156, and Thr129, creating a site that is specific for NADPϩ. These residues are not conserved in MtdB (Figure 5), which can bind both NADϩ and NADPϩ. Notably, the replacement of Arg152 by histidine is an instructive example for sophisticated protein design because both side chains are capable of holding the adenine ring in its position by aromatic interactions and are able to interact with the phosphate oxygens. However, the larger distance between the histidine side chain and the Figure 2. Section of the 2FoϪFc Electron Density Map of MtdA phosphate oxygens weakens the interactions between The view is focused on the segment of the second subunit, which is an essential constituent of the substrate binding cleft. The map them, thereby favoring the observed dual specificity in was calculated at a resolution of 1.9 A˚ and is contoured at 1.7 ␴. This MtdB. The pyrophosphate group is bound to the classi- figure was generated with BOBSCRIPT [42]. cal glycine-rich loop with the recognition sequence Structure 1130

Figure 3. Structure of MtdA (A) Ribbon diagram of the MtdA trimer. Subunits are drawn in green, red, and blue. The three pterin binding domains are arranged close to the noncrystallographic 3-fold axis, whereas the NADP binding domains point toward the solvent. The substrate binding sites are located between the subunits. (B) Stereoview of the C␣ trace of an MtdA monomer. The course of the polypeptide chain from the amino- to the carboxy-terminal end is shown in a “paint ramp” representation. Every twentieth residue is labeled. (C) Ribbon diagram of the MtdA monomer. Each monomer consists of an NADP (green) and a pterin (red) domain. NADPϩ is bound at the C-terminal end of the ␤ sheet; the nicotinamide group is embedded in the crevice between the two domains. This figure was generated with MOLSCRIPT [43] and RASTER3D [44].

GXGXXG [27], which is known to interact with adenine crevice and linked with residues of helix ␣5 of the NADP ribose phosphate and nicotinamide (Figure 4B). Interest- domain and the two interdomain helices ␣9 and ␣10 ingly, this canonical sequence is not conserved in the (Figure 3C). The Si side of the nicotinamide ring is related H4F-dependent dehydrogenases (see below), al- packed against the side chains of Thr102 and Val132 though the conformation of this loop is strictly main- which are strictly conserved in MtdA and MtdB from tained. M. extorquens AM1 [11] and in putative proteins from The nicotinamide ring is embedded in the interdomain Xanthobacter autotrophicus [4] and Methylococcus Structure of Methylene-H4MPT Dehydrogenase 1131

Figure 4. NADP Binding (A) Conformational change of the NADP binding domain with NADPϩ (green/aquamarine) and without NADPϩ (red/orange). The pterin domains on the left side superimpose almost perfectly, whereas the NADP domain moves after NADPϩ binding in a way which opens the interdomain cleft. (B) Polar interactions between NADPϩ and the protein matrix. The specificity of MtdA for NADP but not for NAD is achieved by multiple interactions between the ribose phosphate group and the polypeptide chain. Notably, the amide oxygen of the nicotinamide group forms a hydrogen bond to the amino group of Lys256. The model is depicted in a ball- and-stick representation (carbon for NADPϩ in light blue, carbon for the polypeptide in gold, oxygen in red, nitrogen in blue, and phosphorus in yellow). This figure was generated with MOLSCRIPT [43] and RASTER3D [44].

capsulatus (Figure 5) [47]. The Re side of the nicotin- posed (Figure 2). Flexible, solvent-exposed aromatic amide ring is solvent exposed and is most likely the site residues coating a hydrophobic cleft are reminiscent of for the binding of methylene-H4MPT (see below). The other H4MPT-dependent enzymes such as methenyl- pyridine part of nicotinamide is planar, as expected for H4MPT cyclohydrolase [28] and methylene-H4MPT re- oxidized NADP. However, a significant bend of the am- ductase [29]. ide group is induced by a hydrogen bond between the Modeling of methylene-H4MPT/H4F into its binding oxygen of the amide group and the positively charged site is possible because one knows the position of the ammonium group of Lys256 (Figure 4B). Interestingly, conserved residues within the cleft (Figure 6B), the dis- the hydride-transferring atom C4 of the nicotinamide tance of ca. 3.5 A˚ between the hydride-transferring ϩ ring is in van der Waals distance to the carbonyl oxygen atoms C4 and C14a of NADP and methylene-H4MPT, of Asn97 (3.4 A˚ ) and the hydroxyl group of Thr102 (3.4 A˚ ). respectively, and the stereochemistry of the hydride transfer. 1H-NMR studies revealed that the pro-R hydro-

H4MPT and H4F Binding gen of methylene-H4MPT is transferred into the pro-R ϩ A complex between methylene-H4MPT/H4F and MtdA position of NADP and vice versa, such that the Re side ϩ could not be structurally analyzed thus far. Neverthe- of H4MPT faces the Re side of NADP [15]. The Re side ϩ less, the binding site of the pterin can be unambiguously orientation of NADP toward H4MPT is consistent with located at the bottom of a deep, rather hydrophobic the presented structural data. Moreover, model building cleft in the vicinity of the observed nicotinamide position of the substrate took into account that the pterin and

(Figure 6A). This cleft, 20 A˚ long and 8 A˚ wide, is formed aminobenzoyl rings of the product, methenyl-H4MPT/ by the NADP and pterin domains of one subunit and the H4F, are planarly arranged due to the large conjugated pterin domain of a second subunit. The importance of system. Based on these restraints, methylene-(dephos- the second subunit for binding methylene-H4MPT is pho-)H4MPT was incorporated into the substrate binding underlined by the high degree of conservation of the cleft using the program O [30]. The energy of the ternary cleft-forming residues within the MtdA/MtdB family. complex was subsequently minimized within the pro- These residues, including Tyr51, Thr52, Phe85, and gram CNS (Figure 6A) [31]. The cleft is sufficiently large Phe88, are nonpolar, relatively flexible, and solvent ex- to harbor the entire molecule without interfering with Structure 1132 Structure of Methylene-H4MPT Dehydrogenase 1133

ranged such that the planar system extends from the pyrimidine group of the pterin at the cleft bottom to the aminobenzoyl group touching the loop following strand ␤9. After a sharp kink, the ribitol and ribose groups of

H4MPT extend linearly to the opening of the cleft (Figure 6A). The ␣-hydroxyglutaryl phosphate unit, present in

H4MPT but not in dephospho-H4MPT, protrudes into the bulk solvent and does not bind to the protein. This is in agreement with the similar kinetic parameters for de-

phospho-H4MPT and H4MPT [6]. The pterin ring is mainly anchored to the protein matrix by aromatic interactions to the parallelly oriented benzyl side chain of Phe18. The strictly conserved Phe18, which is part of a hydrophobic patch involving Val22, Thr52,

and Leu257 in the H4MPT free enzyme, presumably ro- tates to the Si side of the pterin ring upon methylene-

H4MPT/H4F binding. Attractive candidates for forming polar protein-pterin interactions are Asp19 and Lys256, the side chains of which are in close contact to the amine and carbonyl substituents of the pyrimidine group

of methylene-H4MPT/H4F. These interactions might also be responsible for the observed Re, Re stereospecificity of the hydride transfer, as an alternative Si face orienta-

tion of H4MPT toward the nicotinamide is spatially possible.

The aminobenzoyl group of methylene-H4MPT/H4Fis flanked by Ile199 and Tyr51. Gln224 might be involved

in binding the amide group of methylene-H4Forthe

ribitol hydroxy groups of methylene-H4MPT. Interest- ingly, Ile199 and Gln224, the latter located in a conserved region, are not conserved between MtdA and MtdB (Fig- ure 5), which might affect the different specificity be-

tween MtdA and MtdB for methylene-H4F (Figure 5). In MtdA, the major fraction of the glutamate tail of

methylene-H4F is probably directed toward the solvent and does not interact with the second subunit. In contrast, the ribitol and the ribose groups of methylene-

H4MPT are sandwiched between the conserved side chains of Ile50 on one side and of Phe85 and Phe88 (protruding from the second subunit) on the other side Figure 6. Substrate Binding Sites and Active Site of MtdA of the cleft. However, the distances between these resi-

(A) The substrate binding cleft is built up by the NADP (molecular dues and the modeled methylene-(dephospho-)H4MPT surface colored in blue) and pterin (light blue) domains of one subunit are too large for forming van der Waals contacts (Figure and the pterin domain (yellow) of the second subunit. The position ϩ 6A). Therefore, methylene-H4MPT binding might induce of NADP (light brown) was determined experimentally, and that of relative domain movements to decrease the diameter methylene-H4MPT (red) is modeled. (B) Conservation of the residues on the protein surface within the of the cleft. MtdA/MtdB family. The surface was colored in blue when the equivalent residues in the five aligned sequences (Figure 5) were identical. Catalytic Reaction ϩ NADP and methylene-H4MPT bind in the regions with the highest As observed for other NAD(P)-dependent oxidoreduc- sequence identity. tases, the active site of MtdA is embedded between two This figure was generated with GRASP [45]. domains, one being involved in the binding of NADP

and the other mainly in the binding of H4MPT and H4F. any main chain atom. However, the side chains of Phe18 The protein scaffold shields the site of the hydride trans- and Asn97 have to evade the modeled methylene- fer from bulk solvent and fixes the catalytically relevant

(dephospho-)H4MPT. Methylene-H4MPT/H4F are ar- nicotinamide and pterin groups in parallel and within

Figure 5. Primary Structure Alignment within the MtdA/MtdB Family The multiple sequence alignment includes MtdA and MtdB from M. extorquens [6, 11], and putative MtdB proteins from Xanthobacter autotrophicus [4] and from the unfinished genome of Methylococcus capsulatus [47]. The alignment of the methylene-H4F dehydrogenases and MtdA is based on a structural superposition using GA-FIT [32]. Secondary structures were assigned for MtdA and the human bifunctional methylene-H4F dehydrogenase using the program DSSP [46] and are indicated by arrows and bars below the sequences. This figure was generated with GRASP [45]. Structure 1134

Figure 7. Comparison of the Active Site Regions of Methylene-H4MPT and Methylene-H4F Dehydrogenases The most pronounced difference is the positioning of a polypeptide segment which is an important constituent of the active site of MtdA

(magenta) but not of methylene-H4F dehydrogenase/cyclohydrolase (yellow). The different design of the active site might be responsible for the different protonation state for the essential Lys256 in MtdA and Lys56 in the human methylene-H4F dehydrogenase/cyclohydrolase. This figure was generated with MOLSCRIPT [43].

van der Waals contacts to each other (Figure 6A). This Comparison between Methylene-H4F- and -H4MPT- is achieved by tightly packing the side chain of Phe18 Dependent Dehydrogenases to the Si side of methylene-H4MPT and the side chains of A sequence alignment of mono- or bifunctional NADP- ϩ Val132 and Thr102 to the Si side of NADP . In agreement dependent methylene-H4F dehydrogenases [16–18] and with kinetic data [6], the presented structural data indi- methylene-H4MPT dehydrogenases (MtdA and MtdB) cate a ternary catalytic mechanism and a direct hydride did not indicate a structural relationship between these transfer between the substrates. enzymes [6, 11]. However, the structural comparison The protein matrix does not only function as a scaffold revealed a common fold among NADP-dependent pterin for positioning the reaction partners appropriately, it dehydrogenases as documented by an rms deviation also participates directly in catalysis. For example, the between MtdA and the human, E. coli, and yeast methyl- ˚ ˚ ˚ carbonyl oxygen of Asn97 and the hydroxyl group of ene-H4F dehydrogenases of 3.6 A, 3.7 A, and 3.9 A, ␣ Thr102 may stabilize the reduced NADP; in particular, respectively, based on around 65% of the C atoms [23]. the Re hydrogen of the C4 atom can favorably interact Due to a different relative orientation of the two domains, with the carbonyl oxygen of Asn97 (Figure 4B). A key the rms deviations between the individual domains are ˚ ˚ ˚ residue in substrate binding and catalysis appears to much lower with 1.4 A, 1.3 A, and 1.5 A for the NADP ˚ ˚ ˚ be the conserved Lys256, which was observed in two domain and 1.9 A, 1.8 A, and 1.9 A, respectively, for the alternate conformations. In the MtdA structure without pterin domain [32]. The nearly identical rms deviation NADPϩ bound, the amino group of Lys256 is hydrogen between MtdA and each of the three structurally known bonded to the imidazole nitrogen of His260, which methylene-H4F dehydrogenases indicates a similar structural difference and a similar evolutionary distance seems to be uncharged due to this interaction and to between them. In other words, the close structural rela- its hydrophobic environment (Phe64, Ile264, and Val22). tionship among mono- and bifunctional methylene- Thus, His260 acts as a hydrogen acceptor and not as H F dehydrogenases [18] suggests a single common an acid/base catalyst, which is further supported by the 4 branchpoint from the MtdA/MtdB family. However, their fact that His260 is replaced by a glutamine in MtdB common ancestor needs not be a precursor NADP- (Figure 5). In a second conformation, the side chain of dependent pterin dehydrogenase because other en- Lys256 is directed toward the cleft bottom and interacts ϩ zymes, like the malic enzyme [19] and aspartate and with NADP and potentially with methylene-H4MPT/H4F ornithine transcarbamoylases [20, 21] with completely (Figure 7). The hydrogen bond between the positively different substrate specificities, are also structurally re- charged ammonium group of Lys256 and the amide lated to MtdA. Thus, a convergent evolution, from differ- ϩ oxygen of NADP leads to a withdrawal of negative ent precursor proteins adopting the described fold, can- charge from the nicotinamide ring system and to a rota- not be excluded. tion of the amide group out of the ring plane (Figure 4B), Besides the obvious structural relationship that de- which weakens the conjugated double bond system. fines a superfamily of NAD(P)-dependent pterin dehy- Both effects destabilize the aromatic NADPϩ state and drogenases, there are remarkable differences between ϩ favor the reduction process from NADP to NADPH by methylene-H4F and -H4MPT dehydrogenases: oxidizing methylene-H4MPT. (1) MtdA contains a larger C-terminal extension com- Structure of Methylene-H4MPT Dehydrogenase 1135

pared to methylene-H4F dehydrogenases (Figure 5) [16– drogenases [18]. In contrast, Lys256 of the NADP do-

18], whereas the latter enzymes contain a large N-ter- main in MtdA is closer to NADP than to H4MPT, and minal extension compared to MtdA. This N-terminal destabilizes oxidized NADP by its positive charge as extension consists of a ␤ strand attached to the central postulated above (Figure 4B). sheet of the NADP domain and a long ␣ helix connecting the NADP and pterin domains. In contrast, the C-ter- Biological Implications minal extension of MtdA contains strand ␤12, attached to the central sheet of the pterin domain, and an ␣ helix, Heterotrophic methylotrophic proteobacteria like Meth- located between the domains. Both extensions bridge ylobacterium extorquens AM1 convert methanol to CO2 the two domains and perhaps reflect different strategies for energy conservation by using H4MPT. Previously, for fine tuning their relative mobility and relative orienta- H4MPT-dependent reactions were thought to only be tions. Indeed, a superposition of MtdA and the human involved in energy metabolism of methanogenic and methylene-H4F dehydrogenase/cyclohydrolase requires sulfate-reducing archaea. The dehydrogenation of a rotation and a translation of the NADP domain toward methylene-H4MPT to methenyl-H4MPT, one reaction of the pterin domain of 23.6Њ and 2.3 A˚ , respectively, this pathway, is performed in M. extorquens AM1 by thereby reducing the diameter of the cleft. As described two homologous NAD(P)-dependent methylene-H4MPT above, a similar movement of the NADP domain of MtdA dehydrogenases, MtdA and MtdB. MtdA shows a dual could take place upon methylene-H4MPT binding to op- specificity for methylene-H4MPT and methylene-H4F, in timize substrate binding and to bury the hydrophobic contrast to MtdB, which only reacts with methylene- side chains. H4MPT. (2) The size and character of the substrate binding We determined the crystal structure of MtdA at 1.9 A˚ ϩ cleft differ significantly between methylene-H4F dehy- resolution with and without the cosubstrate NADP ; the drogenases and MtdA. First, the cleft is built by one substrate methylene-H4MPT could be modeled ac- subunit in methylene-H4F dehydrogenases and by two cording to several biochemical and geometric restraints. subunits in MtdA. The participation of the second sub- The enzyme functions as a homotrimer. Each subunit is unit elongates the cleft (Figure 6A) and increases its composed of two domains which move relative to each hydrophobic character, thus providing an improved other upon substrate binding. Although no significant binding situation for the ribitol and ribose groups of sequence identity between the MtdA/MtdB and methyl-

H4MPT but not for the negatively charged glutamate ene-H4F dehydrogenase families could be identified, group of H4F. In other words, the inability of the homodi- their monomers are structurally similar such that a diver- meric methylene-H4F dehydrogenases to bind methyl- gent evolutionary development from a common ances- ene-H4MPT seems to be due to the absence of a properly tor could be envisaged. positioned second subunit. Thus, trimer formation in Substrate binding and catalysis of MtdA and methyl-

MtdA is a prerequisite for binding H4MPT. This is remark- ene-H4F dehydrogenases are related on the one hand, able, as the involvement of a second subunit is a highly but different substrate specificities and thermodynamic complicated strategy to realize substrate specificity. It demands require substantial structural adaptations on requires a series of amino acid exchanges far away from the other. First, the capability of MtdA and methylene- the substrate binding site. Second, the character of the H4F dehydrogenases to preferably bind methylene- active site is markedly influenced by a strictly conserved H4MPT and methylene-H4F, respectively, is primarily due segment between helices ␣2 and ␣3 at the bottom of to their different oligomeric state. The participation of a the cleft, which is directly involved in substrate binding second subunit in the trimeric MtdA is essential for and catalysis in the case of MtdA (Figure 7) but far away H4MPT binding. Second, a loop segment is a central from the active site in methylene-H4F dehydrogenases. constituent of the active site in MtdA but not in methyl-

In MtdA, this segment increases the polarity at the bot- ene-H4F dehydrogenases. Due to this segment, the ac- tom of the cleft (Figure 7), whereas in methylene-H4F tive site of MtdA is polar in contrast to that of methylene- dehydrogenases, large nonpolar side chains point to H4F dehydrogenases, which is largely hydrophobic. This this region [16]. Moreover, the position of the loop in is of interest because a catalytically crucial lysine of

MtdA shortens the binding site for the pterin and amino- methylene-H4F dehydrogenase/cyclohydrolase is pos- benzoyl moieties, which might cause a steric problem tulated to function via its uncharged state, whereas the for the longer planar system of H4F, and which could positively charged state of Lys256 of MtdA destabilizes be exploited by MtdB to prevent H4F binding by small oxidized NADP. This destabilization facilitates the re- conformational changes. duction of NADPϩ and presumably contributes kinetic (3) A comparative analysis of putative active site resi- enhancement to the thermodynamically favored reduc- dues revealed that two lysines play an important but tion of NADP by methylene-H4MPT, which is important different catalytic role. In human methylene-H F dehy- 4 for the C1 metabolism of M. extorquens AM1. drogenase/cyclohydrolase, Lys56 protrudes from the pterin domain toward the cleft bottom and is closer to Experimental Procedures ϩ H4MPT than to NADP (Figure 7). Lys56 is considered to be neutral due to its nonpolar environment and uses Crystallization and Data Collection its amino group for nucleophilic attack to the pterin MtdA was overproduced in E. coli BL21 and purified as described [11]. Crystallization trials were performed with an enzyme solution ring. Notably, Lys56 presumably has a more important consisting of 15 mg/ml protein in 100 mM HEPES (pH 7.5), using function for the cyclohydrolase than for the dehydroge- the hanging drop vapor diffusion method. Crystals of form 1 were nase reaction [33]. This is supported by the absence of obtained at a temperature of 4ЊC when 1 ␮l protein solution was an equivalent lysine in monofunctional yeast H4F dehy- mixed with 1 ␮l reservoir solution containing 20% PEG1500, 20 mM Structure 1136

Tris (pH 8.5). Their space group was C2 and their lattice parameters References were a ϭ 133.1 A˚ ,bϭ 85.2 A˚ ,cϭ 92.4 A˚ and ␤ϭ113.8Њ best compatible with one trimer in the asymmetric unit. Crystals of form 1. Maden, B.E.H. (2000). Tetrahydrofolate and tetrahydrometha- 2 grew with a reservoir solution of 17% PEG8000, 50 mM MES (pH nopterin compared: functionally distinct carriers in C-1 metabo- 6.5), in the presence of 5 mM NADPϩ. These crystals belong to the lism. Biochem. J. 350, 609–629. 2. MacKenzie, R.E. (1984). Biogenesis and interconversion of sub- space group P212121 with lattice parameters of a ϭ 75.9 A˚ ,bϭ 79.9 A˚ , and c ϭ 154.3 A˚ . The asymmetric unit contains one trimer. stituted tetrahydrofolates. In Folates and Pterins. Chemistry and Data sets MtdA1 and MtdA2 (Table 1) were collected at the Max- Biochemistry of Folates, Volume 1, R.L. Blakley and S.J. Ben- Planck beamline BW6 and the EMBL beamline X11 under flash- kovic, eds. (New York: John Wiley & Sons), pp. 256–306. 3. DiMarco, A.A., Bobik, T.A., and Wolfe, R.S. (1990). Unusual co- frozen conditions. Data sets MtdA3, HgAc2, thimerosal, K2PtCl4, and TERPI (platinum (II) 2,2Ј:6Ј,2″ terpyridine chloride) were measured enzymes of methanogenesis. Annu. Rev. Biochem. 59, 355–394. at 18ЊC using a Rigaku rotating Cu anode and a MarResearch image 4. Vorholt, J.A., Chistoserdova, L., Stolyar, S.M., Thauer, R.K., and plate detector. Processing and scaling of the reflections were per- Lidstrom, M.E. (1999). Distribution of tetrahydromethanopterin- formed with the HKL suite [34] and the CCP4 package [35]. The dependent enzymes in methylotrophic bacteria and phylogeny statistics of the data sets are summarized in Table 1. of methenyl tetrahydromethanopterin cyclohydrolases. J. Bac- teriol. 181, 5750–5757. 5. Chistoserdova, L., Vorholt, J.A., Thauer, R.K., and Lidstrom, Phase Determination and Model Building M.E. (1998). C transfer enzymes and coenzymes linking methyl- Phases of MtdA crystals of form 1 were determined with the method 1 otrophic bacteria and methanogenic archaea. Science 281, of multiple isomorphous replacement on the basis of five heavy 99–102. atom derivatives (Table 1). The difference Patterson map of the 6. Vorholt, J.A., Chistoserdova, L., Lidstrom, M.E., and Thauer, derivative HgAc could be interpreted by the program SOLVE [36]. 2 R.K. (1998). The NADP-dependent methylene tetrahydrometha- Further positional refinement and final phase determination was nopterin dehydrogenase in Methylobacterium extorquens AM1. performed with the program SHARP (see Table 1) [37]. The solvent- J. Bacteriol. 180, 5351–5356. flattened electron density map at 2.9 A˚ resolution allowed the sepa- 7. van Beelen, P., Stassen, A.P., Bosch, J.W., Vogels, G.D., Guijt, ration of the three monomers in the asymmetric unit and the calcula- W., and Haasnoot, C.A. (1984). Elucidation of the structure of tion of the noncrystallographic symmetry operators from the heavy methanopterin, a coenzyme from Methanobacterium thermoau- atom positions in order to 3-fold average the density using DM [38]. totrophicum, using two-dimensional nuclear-magnetic-reso- Afterwards, the polypeptide chain could be incorporated into the nance techniques. Eur. J. Biochem. 138, 563–571. electron density using O [30] and associated programs [39] and the 8. Vorholt, J.A., Marx, C.J., Lidstrom, M.E., and Thauer, R.K. (2000). known primary structure of MtdA [6]. Phases of crystals of form Novel formaldehyde-activating enzyme in Methylobacterium 2 containing the MtdA-NADP complex were determined with the extorquens AM1 required for growth on methanol. J. Bacteriol. method of molecular replacement using AMoRe [40]. 182, 6645–6650. 9. Pomper, B.K., Vorholt, J.A., Chistoserdova, L., Lidstrom, M.E., Refinement and Thauer, R.K. (1999). A methenyl tetrahydromethanopterin MtdA of crystal form 1 was refined with CNS [31] using the recom- cyclohydrolase and a methenyl tetrahydrofolate cyclohydrolase mended standard protocols. Water molecules were built in when in Methylobacterium extorquens AM1. Eur. J. Biochem. 261, the 2FoϪFc and the FoϪFc electron density map were higher than 1 475–480. ␴ and 3 ␴, respectively. Moreover, electron density peaks had to 10. Chistoserdova, L.V., and Lidstrom, M.E. (1994). Genetics of the be roughly spherical and be positioned within 3.3 A˚ from the nearest serine cycle in Methylobacterium extorquens AM1: identifica- hydrogen bond donor/acceptor. Water molecules with temperature tion, sequence, and mutation of three new genes involved in factors greater than 55 A˚ 2 were removed from the model. After C-1 assimilation, orf4, mtkA, and mtkB. J. Bacteriol. 176, 7398– several rounds of computational and manual model corrections the 7404.

Rcryst and Rfree factors converged to 19.0% and 22.9%, respectively 11. Hagemeier, C.H., Chistoserdova, L., Lidstrom, M.E., Thauer, (Table 1). The final model comprises residues 2–288 for all three R.K., and Vorholt, J.A. (2000). Characterization of a second subunits and 558 solvent molecules. Alternate conformations were methylene tetrahydromethanopterin dehydrogenase from Meth- found for 5 single residues and, surprisingly, in large parts of the ylobacterium extorquens AM1. Eur. J. Biochem. 267, 3762– NADP domain of one of the monomers comprising residues 110– 3769. 235. Allowing for two conformations in this segment (with con- 12. Pomper, B.K., and Vorholt, J.A. (2001). Characterization of the strained noncrystallographic symmetry) in a ratio of 0.6:0.4 improved formyltransferase from Methylobacterium extorquens AM1. Eur. the Rfree factor by around 1.3%. MtdA of crystal form 2 was equiva- J. Biochem. 268, 4769–4775. lently refined to Rcryst and Rfree factors of 18.6% and 22.2%, respec- 13. Klein, A.R., and Thauer, R.K. (1997). Overexpression of the coen- 5 10 tively (Table 1). The resulting model contains residues 2–288 for all zyme F420-dependent N ,N -methylenetetrahydromethanop- three subunits and 712 solvent molecules. Alternate conformations terin dehydrogenase gene from the hyperthermophilic Metha- were found for 16 residues. Model errors were assessed using PRO- nopyrus kandleri. Eur. J. Biochem. 245, 386–391. CHECK [41], which verifies deviations of geometric parameters from 14. Thauer, R.K., Klein, A.R., and Hartmann, G.C. (1996). Reactions ideal values and of dihedral angles from allowed regions for the with molecular hydrogen in microorganisms. Evidence for a main and side chain atoms. The Ramachandran plot indicates that purely organic hydrogenation catalyst. Chem. Rev. 96, 3031– all nonglycine residues have favorable dihedral angles. 3042. 15. Hagemeier, C.H., Bartoschek, S., Griesinger, C., Thauer, R.K., and Vorholt, J.A. (2001). Re-face stereospecificity of NADP- Acknowledgments dependent methylenetetrahydromethanopterin dehydrogenase from Methylobacterium extorquens AM1 as determined by NMR This work was supported by the Max-Planck-Gesellschaft and the spectroscopy. FEBS Lett. 494, 95–98. Peter und Traudl Engelhorn-Stiftung (scholarship to C.H.H.). Se- 16. Allaire, M., Li, Y., MacKenzie, R.E., and Cygler, M. (1998). The quencing of M. capsulatus by TIGR was accomplished with the 3-D structure of a folate-dependent dehydrogenase/cyclohy- support of DOE. We thank Rudolf K. Thauer for discussion, Hartmut drolase bifunctional enzyme at 1.5 A˚ resolution. Structure 6, Michel for continuous support, Erica Lyon for reading the manu- 173–182. script, and the staff of the Max-Planck Institute and EMBL beamlines 17. Shen, B.W., Dyer, D.H., Huang, J.Y., D’Ari, L., Rabinowitz, J., for help during data collection. and Stoddard, B.L. (1999). The crystal structure of a bacterial, bifunctional 5,10 methylene-tetrahydrofolate dehydrogenase Received: January 28, 2002 cyclohydrolase. Protein Sci. 8, 1342–1349. Revised: May 10, 2002 18. Monzingo, A.F., Breksa, A., Ernst, S., Appling, D.R., and Rober- Accepted: May 13, 2002 tus, J.D. (2000). The X-ray structure of the NAD-dependent 5,10- Structure of Methylene-H4MPT Dehydrogenase 1137

methylenetetrahydrofolate dehydrogenase from Saccharo- structure determination of bovine mitochondrial F-1 ATPase. myces cerevisiae. Protein Sci. 9, 1374–1381. Acta Crystallogr. D 52, 30–42. 19. Xu, Y., Bhargava, G., Wu, H., Loeber, G., and Tong, L. (1999). 39. Kleywegt, G.J., and Jones, T.A. (1996). xdlMAPMAN and xdlDA- Crystal structure of human mitochondrial NAD(P)(ϩ)-dependent TAMAN—programs for reformatting, analysis and manipulation malic enzyme: a new class of oxidative decarboxylases. Struc- of biomacromolecular electron-density maps and reflection ture 7, 877–889. data sets. Acta Crystallogr. D 52, 826–828. 20. Honzatko, R.B., Crawford, J.L., Monaco, H.L., Ladner, J.E., 40. Navaza, J. (1994). AMoRe—an automated package for molecu- Ewards, B.F., Evans, D.R., Warren, S.G., Wiley, D.C., Ladner, lar replacement. Acta Crystallogr. A 50, 157–163. R.C., and Lipscomb, W.N. (1982). Crystal and molecular struc- 41. Laskowski, R.A., Macarthur, M.W., Moss, D.S., and Thornton, tures of native and CTP-liganded aspartate carbamoyltrans- J.M. (1993). PROCHECK—a program to check the stereochemi- ferase from Escherichia coli. J. Mol. Biol. 160, 219–263. cal quality of protein structures. J. Appl. Crystallogr. 26, 21. Villeret, V., Tricot, C., Stalon, V., and Dideberg, O. (1995). Crystal 283–291. structure of Pseudomonas aeruginosa catabolic ornithine trans- 42. Esnouf, R.M. (1999). Further additions to MOLSCRIPT version carbamoylase at 3.0-A˚ resolution: a different oligomeric organi- 1.4, including reading and contouring of electron-density maps. zation in the transcarbamoylase family. Proc. Natl. Acad. Sci. Acta Crystallogr. D 55, 938–940. USA 92, 10762–10766. 43. Kraulis, P.J. (1991). MOLSCRIPT—a program to produce both 22. Hubbard, S.J., Campbell, S.F., and Thornton, J.M. (1991). Mo- detailed and schematic plots of protein structures. J. Appl. Crys- lecular recognition. Conformational analysis of limited proteo- tallogr. 24, 946–950. lytic sites and serine proteinase protein inhibitors. J. Mol. Biol. 44. Merritt, E.A., and Murphy, M.E.P. (1994). Raster3D version- 220, 507–530. 2.0—a program for photorealistic molecular graphics. Acta 23. Holm, L., and Sander, C. (1993). Protein structure comparison Crystallogr. D 50, 869–873. by alignment of distance matrices. J. Mol. Biol. 233, 123–138. 45. Nicholls, A., Sharp, K.A., and Honig, B. (1991). Protein folding 24. Gouet, P., Fabry, B., Guillet, V., Birck, C., Mourey, L., Kahn, and association: insights from the interfacial and thermody- D., and Samama, J.P. (1999). Structural transitions in the FixJ namic properties of hydrocarbons. Proteins 11, 281–296. receiver domain. Structure 7, 1517–1526. 46. Kabsch, W., and Sander, C. (1983). Dictionary of protein second- 25. Carugo, O., and Argos, P. (1997). NADP-dependent enzymes. ary structure—pattern-recognition of hydrogen-bonded and II: Evolution of the mono- and dinucleotide binding domains. geometrical features. Biopolymers 22, 2577–2637. Proteins 28, 29–40. 47. The Institute of Genomic Research, http://www.tigr.org. 26. Baker, P.J., Sawa, Y., Shibata, H., Sedelnikova, S.E., and Rice, D.W. (1998). Analysis of the structure and substrate binding of Accession Numbers Phormidium lapideum alanine dehydrogenase. Nat. Struct. Biol. 5, 561–567. The atomic coordinates for methylene-H4MPT dehydrogenase MtdA 27. Wierenga, R.K., Terpstra, P., and Hol, W.G.J. (1986). Prediction with and without NADPϩ have been deposited in the Protein Data of the occurrence of the ADP-binding ␤␣␤-fold in proteins, using Bank under ID codes 1LUA and 1LU9, respectively. an amino acid sequence fingerprint. J. Mol. Biol. 187, 101–107. 28. Grabarse, W., Vaupel, M., Vorholt, J.A., Shima, S., Thauer, R.K., Wittershagen, A., Bourenkov, G., Bartunik, H.D., and Ermler, U. (1999). The crystal structure of methenyltetrahydromethanop- terin cyclohydrolase from the hyperthermophilic archaeon Methanopyrus kandleri. Structure 7, 1257–1268. 29. Shima, S., Warkentin, E., Grabarse, W., Sordel, M., Wicke, M.,

Thauer, R.K., and Ermler, U. (2000). Structure of coenzyme F420 dependent methylenetetrahydromethanopterin reductase from two methanogenic archaea. J. Mol. Biol. 300, 935–950. 30. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved methods for binding protein models in electron density maps and the location of errors in these models. Acta Crys- tallogr. A 47, 110–119. 31. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D 54, 905–921. 32. May, A.C., and Johnson, M.S. (1994). Protein structure compari- sons using a combination of a genetic algorithm, dynamic pro- gramming and least-squares minimization. Protein Eng. 7, 475–485. 33. Schmidt, A., Wu, H.P., MacKenzie, R.E., Chen, V.J., Bewly, J.R., Ray, J.E., Toth, J.E., and Cygler, M. (2000). Structures of three inhibitor complexes provide insight into the reaction mechanism of the human methylenetetrahydrofolate dehydrogenase/cyclohydrolase. Biochemistry 39, 6325–6335. 34. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. 35. Bailey, S. (1994). The CCP4 suite—programs for protein crystallography. Acta Crystallogr. D 50, 760–763. 36. Terwilliger, T.C., and Berendzen, J. (1999). Automated MAD and MIR structure solution. Acta Crystallogr. D 55, 849–861. 37. de La Fortelle, E., and Bricogne, G. (1997). Maximum-likelihood heavy-atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472–494. 38. Abrahams, J.P., and Leslie, A.G.W. (1996). Methods used in the