Crystal Structure of Novel NADP-dependent 3-Hydroxyisobutyrate Dehydrogenase from Thermus thermophilus HB8

doi:10.1016/j.jmb.2005.07.068

J. Mol. Biol. (2005) 352, 905–917

Crystal Structure of Novel NADP-dependent 3-Hydroxyisobutyrate Dehydrogenase from Thermus thermophilus HB8 Neratur K. Lokanath1, Noriyasu Ohshima1, Koji Takio1, Ikuya Shiromizu1 Chizu Kuroishi1, Nobuo Okazaki1, Seiki Kuramitsu2 Shigeyuki Yokoyama2, Masashi Miyano3 and Naoki Kunishima1* 1

Highthroughput Factory RIKEN Harima Institute at SPring-8, 1-1-1 Kouto Mikazuki-cho, Sayo-gun, Hyogo 679-5148, Japan 2

Structurome Research Group RIKEN Harima Institute at SPring-8, 1-1-1 Kouto Mikazuki-cho, Sayo-gun, Hyogo 679-5148, Japan 3

Structural Biophysics Laboratory, RIKEN Harima Institute at SPring-8, 1-1-1 Kouto, Mikazuki-cho, Sayo-gun Hyogo 679-5148, Japan

3-Hydroxyisobutyrate, a central metabolite in the valine catabolic pathway, is reversibly oxidized to methylmalonate semialdehyde by a specific dehydrogenase belonging to the 3-hydroxyacid dehydrogenase family. To gain insight into the function of this enzyme at the atomic level, we have determined the first crystal structures of the 3-hydroxyisobutyrate dehydrogenase from Thermus thermophilus HB8: holo enzyme and sulfate ion complex. The crystal structures reveal a unique tetrameric oligomerization and a bound cofactor NADPC. This bacterial enzyme may adopt a novel cofactor-dependence on NADP, whereas NAD is preferred in eukaryotic enzymes. The protomer folds into two distinct domains with open/closed interdomain conformations. The cofactor NADPC with syn nicotinamide and the sulfate ion are bound to distinct sites located at the interdomain cleft of the protomer through an induced-fit domain closure upon cofactor binding. From the structural comparison with the crystal structure of 6-phosphogluconate dehydrogenase, another member of the 3-hydroxyacid dehydrogenase family, it is suggested that the observed sulfate ion and the substrate 3-hydroxyisobutyrate share the same binding pocket. The observed oligomeric state might be important for the catalytic function through forming the active site involving two adjacent subunits, which seems to be conserved in the 3-hydroxyacid dehydrogenases. A kinetic study confirms that this enzyme has strict substrate specificity for 3-hydroxyisobutyrate and serine, but it cannot distinguish the chirality of the substrates. Lys165 is likely the catalytic residue of the enzyme. q 2005 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: 3-hydroxyisobutyrate; 3-hydroxyisobutyrate dehydrogenase; induced-fit; oligomerization; crystal structure

Introduction 3-Hydroxyisobutyrate (HIBA), a central metabolite in the valine catabolic pathway, is reversibly oxidized to methylmalonate semialdehyde by a Present address: I. Shiromizu, Bioscience Laboratory, Mochida Pharma-ceuticals Co. Ltd., 722 Jimba-azauenohara, Gotemba, Shizuoka 412-8524, Japan. Abbreviations used: HIBA, 3-hydroxyisobutyrate; HIBADH, 3-hydroxyisobutyrate dehydrogenase; MAD, multiwavelength anomalous diffraction; r.m.s.d., rootmean-square deviation; 6PGDH, 6-phosphogluconate dehydrogenase; PDB, Protein Data Bank. E-mail address of the corresponding author: [email protected]

specific dehydrogenase (HIBADH; EC 1.1.1.31) with a cofactor dependent usually on NAD.1 HIBADH is distributed widely in yeasts, bacteria and mammalian tissues.2,3 The general catabolic pathway of valine is shown in Figure 1. The catabolism of valine is unique among the three branched-chain amino acids (leucine, isoleucine and valine) because of an essential step in which a coenzyme A ester is hydrolyzed to give a carboxylate. Whether a tissue catabolyzes or releases HIBA is dependent on the level of HIBADH. When valine catabolism is initiated in muscle, some of the carbon is released as 2-oxoisovalerate and some is metabolized through the branched-chain 2-oxoacid dehydrogenase and subsequent enzymes of valine catabolism.

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

906

Structure of NADP-dependent HIBADH

Figure 1. Catabolic pathway of valine. The step catalyzed by HIBADH is boxed.

However, the exact pathways of complete valine catabolism remain uncertain due to the paucity of description of the kinetics and substrate specificities of the enzymes involved. Compared to other coenzyme A ester intermediates of valine

catabolism, HIBA diffuses across the mitochondrial and plasma membranes. For this reason, HIBA has been suggested to be the most important valine catabolite for hepatic gluconeogenesis, being a better substrate than 2-oxo-isovalerate.4

907

Structure of NADP-dependent HIBADH

The molecular basis of the HIBADH function is not fully understood due to the lack of 3-D structural analysis. This enzyme does not require a metal ion for catalysis, and has a number of features in common with the short-chain alcohol dehydrogenases,5,6 and contains a highly conserved dinucleotide cofactor-binding region. The control of HIBADH is of interest because elevated levels of HIBA in serum and urine have been reported in several diseases, viz. hydroxyisobutyric aciduria, brain disorders and methymalonic acidaemia.7–9 If the level of HIBA in an organism is changed from its equilibrium, the process of valine catabolism is disturbed, leading to death of the organism. HIBADH is sensitive to inhibition by the reduced form of the cofactor, which may be relevant to the short-term physiological control of valine catabolism.10

In order to gain better insight into enzymatic function, and to overcome the ill effects of faulty valine catabolism, it is necessary to determine the crystal structure of HIBADH. We have reported the purification and the crystallization of HIBADH from Thermus thermophilus HB8.11 Here, we report ˚ resolution the structure of the holo enzyme at 1.8 A and its complex structure with sulfate ion. To our knowledge, this is the first detailed 3-D structural characterization of any HIBADH. Unexpectedly, we found that the T. thermophilus HIBADH bound NADP rather than NAD as a cofactor, indicating a novel cofactor-dependence of this enzyme. On the basis of the crystal structures and a kinetic study, we present a molecular basis for the cofactor/substrate recognition, and point out the importance of the domain rearrangement and oligomerization of the NADP-dependent HIBADH.

Table 1. Summary of data collection, phasing and refinement statistics A. Selenomethionyl HIBADH crystals for MAD phasing SeMetpeak ˚) Wavelength (A 0.97884 ˚) Resolution (A 40.0–2.20 (2.28–2.20) Space group Cell dimensions ˚) a (A ˚) b (A ˚) c (A Unique reflections 78,370 Completeness (%) 99.3 (97.8) 4.7 (8.0) Rmergea (%) 15.9 (12.4) hI/s(I)i Redundancy 4.44 Overall figure of merit Before density modification 0.66 After density modification 0.84

SeMetinflection

SeMetremote

0.97934 40.0–2.2 (2.28–2.20)

0.98400 40.0–2.20 (2.28–2.20)

85.81 106.28 168.41 77,365 98.1 (94.5) 2.7 (6.1) 34.1 (14.6) 3.87

77,059 97.7 (92.8) 2.6 (6.5) 37.5 (14.5) 3.75

P212121

B. Refined forms of HIBADH ˚) Resolution range used in refinement (A Space group Cell dimensions ˚) a (A ˚) b (A ˚) c (A No. observations No. unique reflections Completeness (%) Rmergea (%) hI/s(I)i Redundancy R-factor (%) Rfree (%) ˚ 2) Wilson B-factor (A ˚ 2) Average B-factor of protein atoms (A r.m.s.d. from ideal ˚) Bond lengths (A Bond angles (deg.) Ramachandran plot Most-favored region (%) Additionally allowed regions (%) Disallowed regions (%)

Holo HIBADH

Complex with sulfate ion

30.0–1.80 (1.86–1.80) P212121

40.0–2.00 (2.07–2.00) P212121

85.89 106.37 168.64 890,536 139,102 97.2 (92.0) 6.8 (43.4) 13.0 (4.9) 6.4 18.2 20.1 19.8 21.6

93.86 109.03 116.39 338,952 80,660 98.5 (93.2) 5.1 (25.3) 18.6 (6.7) 4.2 20.2 22.2 24.2 25.9

0.0047 1.27

0.0060 1.10

92.8 7.1 0.1

91.9 8.1 0.0

The numbersPin parentheses are for the Phighest resolution shell. P a RmergeZ hkl j|Ij(hkl)KhI(hkl)i|/ jI(hkl), Ij(hkl) and hI(hkl)i are the observed intensity of measurement j and the mean intensity for the reflection with indices hkl, respectively.

908

Results and Discussion Model quality The crystal structure of the NADP-dependent HIBADH from T. thermophilus HB8 in the holo form ˚ resolution by the has been determined at 1.8 A multiwavelength anomalous diffraction (MAD) method using selenomethionine (SeMet)-substi˚ resolution tuted protein crystals.12 Phases at 2.2 A were obtained using selenium atoms from MAD data collected at SPring-8, Japan. Subsequently, the sulfate ion complex structure was solved by the molecular replacement method using the holo enzyme coordinates as a search model and refined ˚ resolution. The asymmetric unit of the holo at 2.0 A form crystal contains an apparent tetramer of HIBADH protomers. The final atomic model for the holo enzyme contains residues 2–289 for each protomer (only the C protomer has 1–289), three NADPC molecules and 868 water molecules. The electron density of the putative active site of the D protomer was too weak for modeling the NADPC molecule. Analysis of stereochemistry using the program PROCHECK 13 shows that 92.8% of the residues are in the most-favored regions, and 7.1% are in additionally allowed regions of the Ramachandran plot. With the exception of one residue (Phe223) in the D chain, which fits neatly into well-defined electron density (fZ 38.08, jZ60.68), no other residue lies in the generously allowed or disallowed regions. The refined model has low R-values and excellent geometry (Table 1). Overall structure of HIBADH protomer The structure of the HIBADH protomer contains eight b-strands (named b1 through b8) and 13 a-helices (a1 through a13) (Figure 2(a)). The protomer has two distinct domains. Domain I with the dinucleotide-binding region comprises residues from the N-terminal a/b domain of the protein (residues 2–159). This typical Rossmann fold domain contains two a/b units: a six-stranded parallel b-sheet (b1-b6) covered by four helices (a1-a4) and followed by a mixed three-stranded b-sheet (b7-b9) covered by two helices (a5 and a6). Domain II (residues 160–289) consists of only helices (a7-a13) from the C-terminal segments of the protein. The two domains are connected by a long a-helix, a7 (residues 160–187). The NADPC molecule is bound at the cleft between the two domains. The root-mean-square deviation (r.m.s.d.) values between equivalent Ca atoms of the subunits A, B and C of the apparent tetramer in the ˚ , whereas the values asymmetric unit are about 0.2 A between the D subunit and the other subunits are ˚ . In fact, the D subunit that does not about 1.6 A contain a NADPC molecule shows the most open interdomain conformation compared to the other three subunits in the tetramer, suggesting an

Structure of NADP-dependent HIBADH

induced-fit domain rearrangement upon NADPC binding. The affinity to NADPC of the closed conformer might be higher than that of the open conformer. We calculated domain movements using LSQKAB,14 which shows maximum interdomain orientational difference of 128 between the open and closed conformers. Further domain movements might be restrained due to the limited flexibility of the long helix (a7) that separates the two domains. The interdomain interface includes invariant residues Asp113, Val116, Gly161, Lys165 and Lys234 (highlighted in Figure 3), which may contribute to the domain movement or the catalytic function of this enzyme. Tetrameric quaternary structure Crystallographic analyses reveal that HIBADH exists as an apparent tetramer in the crystal with extensive interfaces (Figure 2(b)). Notably, both the crystal forms contain globally identical tetramers. The observed tetramers are consistent with the results of dynamic light-scattering studies showing the tetrameric state of HIBADH in solution.11 These observations suggest that HIBADH functions as a homotetramer. Each subunit of the tetramer interacts with two neighboring subunits. We define the interface between the subunits A and B making the AB dimer as interface I, that between the subunits A and C making the AC dimer as interface II, and that between the subunits A and D as interface III (Figure 2(b)). Interface I shows a major buried ˚ 2, which corresponds 18.6% of surface area of 2456 A the total protomer surface area, whereas interfaces ˚ 2, ˚ 2 and 387 A II and III show minor areas of 648 A respectively. The major interface I is mediated by the domain I residues in the vicinity of the active site and the domain II residues, including those of the long interdomain helix a7. The interacting surfaces at interface I are essentially hydrophobic. The hydrophobic core is surrounded by 21 polar interactions. On the other hand, interfaces II and III are dominated by 12 and five hydrogen bonds, respectively, and they are mediated by only the domain II residues. We mapped the highly conserved residues in orthologues onto the molecular surface, which clearly revealed a conservation pattern (not shown). Interface I includes invariant residues of domain I, Ser117, Ser118, Thr129, Met131 and Gly159, and of domain II, Asn168 and Glu181 (highlighted in Figure 3). This result may indicate the biological importance of these residues in the catalytic function of this protein through maintaining the integrity of the AB dimer, as reported for instance for the dimeric dehydroquinate synthase.15 The other two interfaces do not contain invariant residues, suggesting that residue conservation is not necessarily required at these interfaces; for instance, they might contribute to the thermostability of proteins, as reported for the dimeric/tetrameric 2-deoxyribose-5-phosphate aldolase.16

Structure of NADP-dependent HIBADH

909

Figure 2. Overall ribbon representation of HIBADH. (a) Stereo view of the HIBADH protomer with secondary structure assignments. The two domains are labeled. The N and C termini are colored blue and red, respectively. The cofactor NADPC is shown as a ball-and-stick model. (b)Tetramer structure of HIBADH in three orthogonal views. Subunits A, B, C and D are colored yellow, light blue, orange and red, respectively. The perspective of the A subunit in the left tetramer model is similar to that of the protomer in (a). Molecular local 222 symmetry is depicted as crystallographic symbols. The three interfaces are indicated by dotted circles and are labeled. (c) Structural comparison

910

Structure of NADP-dependent HIBADH

Figure 3. Sequence alignment of HIBADH orthologues. The secondary structure elements in the crystal structure of T. thermophilus HIBADH are shown above the alignment. Strictly conserved residues are highlighted. The triangle mark below the sequence indicates the catalytic residue. The abbreviations used for the source organisms are: TT, T. thermophilus HB8; DR, Deinococcus radiodurans; RE, Rhizobium elti; BM, Bacillus melitensis; ML, Mesorhizobium loti; PA, Pseudomonas aeruginosa; VP, Vibrio parahaemolyticus; BH, Bacillus halodurans; HS, Homo sapiens; MM, Mus musculus. This Figure was drawn using ESPript.32 with three relevant proteins: dimeric 6-phosphogluconate dehydrogenase from Ovis aries (left; PDB, 2PGD), dimeric conserved hypothetical protein from M. thermoautotrophicum (middle; PDB 1I36), and KTN domain from B. subtilis (right; PDB, 1LSU). Ribbon models of the proteins are superimposed onto the HIBADH tetramer with the first perspective in (b). The colorings of the models are the same as those of the corresponding subunits of the HIBADH tetramer in (b). In the model of 6PGDH, the linker (residues 293–313) and the C-terminal tail (residues 435–473) have been removed for the clarity. Figures 2, 4(a) and 5 were drawn using MOLSCRIPT,29 BOBSCRIPT30 and Raster3D.31

Structure of NADP-dependent HIBADH

Structural resemblance to other proteins A 3-D structural similarity search using DALI server17 was performed between the refined model of the HIBADH protomer and the coordinates available in the Protein Data Bank (PDB). The search provides certain oxidoreductases showing considerable similarities with z scores of 21.0–14.1 ˚ . These proteins and with r.m.s.d. values of 3.1–3.9 A include sheep liver 6-phosphogluconate dehydrogenase (6PGDH; PDB, 2PGD),18,19 the conserved hypothetical protein from Methanobacterium thermoautotrophicum (PDB, 1I36), UDP-glucose dehydrogenase (PDB, 1DLI), GDP-mannose 6-dehydrogenase (PDB, 1MFZ) and l-3-hydroxyacyl-coenzyme A dehydrogenase (PDB, 2HDH), and share a common overall folding with HIBADH. Of these ˚, proteins, 6PGDH (z scoreZ21.0, r.m.s.d.Z3.5 A sequence identityZ23%) and the conserved hypo˚, thetical protein (z scoreZ17.4, r.m.s.d.Z3.9 A sequence identityZ16%) show a relatively higher level of similarity to HIBADH when compared to the other proteins. It is reported that HIBADH and 6PGDH may comprise a distinct subgroup of oxidoreductases referred to as the 3-hydroxyacid dehydrogenase family.20 Our result is consistent with the assumption that HIBADH, 6PGDH and the conserved hypothetical protein belong to the 3-hydroxyacid dehydrogenase family. We also tried to find the structural similarity of domains I and II of HIBADH separately with the available coordinates in PDB. The search shows a significant similarity between domain I and the tetrameric Bacillus subtilis KTN (PDB, 1LSU; z scoreZ11.0, ˚ , sequence identityZ14%) which is r.m.s.d.Z2.8 A involved in a certain potassium channel (right, Figure 2(c)).21 This KTN domain has a nucleotidebinding motif, resulting in the same fold as that of domain I of HIBADH in spite of their very different functions. The functional oligomerizations of both 6PGDH18 and the conserved hypothetical protein (S. V. Korolev et al., unpublished results) were reported to be dimers. The 6PGDH dimer can be superimposed globally on the AC dimer of HIBADH (left, Figure 2(c)); the z score is 22.0 and the r.m.s.d. ˚ for the Ca superposition. Curiously, value is 4.1 A 6PGDH has a longer sequence of 482 residues as compared with the 289 residues of HIBADH and the 264 residues of the conserved hypothetical protein, although they belong to the same family. From the structural comparison, we found that the C-terminal region of 6PGDH (residues 177–482) may include duplication of a domain that is equivalent to domain II of HIBADH. A DALI pairwise comparison between the two domain II homologous regions in 6PGDH (residues 177–292 and 313–434) revealed a clear structural similarity ˚ ), in contrast with an (z scoreZ11.4, r.m.s.d.Z1.9 A identity of 8%, which is not significant. Interestingly, considering the 2-fold symmetry relationships between the domain II equivalent parts, the 6PGDH protomer corresponds to the AB dimer of

911 HIBADH, suggesting that the 6PGDH dimer mimics the tetrameric structure of HIBADH (indicated by different coloring in Figure 2(c), left). On the other hand, the dimer of the conserved hypothetical protein can be superimposed on the AB dimer of HIBADH; the z score is 18.3 and the ˚ for the Ca superposition r.m.s.d. value is 5.0 A (Figure 2(c), middle). These structural comparisons suggest that the AB dimer is functionally more essential in 3-hydroxyacid dehydrogenases, although the AC dimer may be important for the 6PGDH function. Furthermore, the domain duplication of 6PGDH indicates that the tetrameric oligomerization seen in HIBADH could be considered as the evolutionary prototype of 6PGDH. Cofactor NADP binding at the interdomain cleft During sample preparation and crystallization, no cofactor was added to the HIBADH solution. However, the crystal structure reveals the presence of bound NADPC, which is reminiscent of a physiological cofactor. Many NAD/NADP-dependent dehydrogenases contain the Rossmann fold for binding the dinucleotide cofactor;22 the pyrophosphate group interacts with the GXGXX(G/A) sequence fingerprint motif found in the Rossmann fold. This characteristic glycine-rich fingerprint motif is present at the N terminus in domain I (Gly8-Gly13), which is essential for the nucleotide binding. In the holo enzyme structure, the NADPC molecule with a nicotinamide ring in the syn conformation has well-defined electron density in three of the four subunits of the holo enzyme tetramer (Figure 4(a)). Thus, we could identify the residues that contribute to cofactor recognition (Figure 4(b)). An NADPC molecule is bound at the interdomain cleft of each protomer. The adenine base moiety of NADPC resides in a groove formed by the side-chains of Arg31 and Glu68, showing a cation–p interaction and many non-polar interactions. The nicotinamide ribose moiety of NADPC is situated between the two domains of the protomer. The amide nitrogen atom of the nicotinamide ring is hydrogen bonded to the main-chain carbonyl oxygen atom of Thr226. The nicotinamide ring is sandwiched between the side-chains of invariant residues Met12 and Phe227. The 3 0 hydroxyl group of the nicotinamide ribose makes hydrogen bonds with domain I residues Leu63, Pro64 and Ser91 as well as with the domain II invariant residue Lys234, whereas, the 2 0 -hydroxyl group is hydrogen bonded only to the domain II residue Lys234. Oxygen atoms of the 5 0 -diphosphate make hydrogen bonds with the main-chain nitrogen atoms of residues Ala11 and Met12. The 3 0 -hydroxyl group of the adenosine base is hydrogen bonded to the side-chain of Asn30. The 2 0 -phosphate group of adenosine makes hydrogen bonds with the side-chains of Asn30, Arg31 and Thr32, and with the main chain of Thr32, suggesting the NADP-dependence of this enzyme.

912

Structure of NADP-dependent HIBADH

˚ resolution Figure 4. Cofactor binding in the active site pocket. (a) The composite omit map electron density at 1.8 A contoured at 2s for the NADPC molecule bound to the A subunit in the holo HIBADH tetramer. The final NADPC model is shown as a stick model. (b) Structural surroundings of the NADPC molecule in the A subunit of the holo form. This Figure was drawn using LIGPLOT.33

Strikingly, the residues responsible for cofactor recognition are not well conserved among the HIBADH orthologues, except Met12, Phe227 and Lys234 (Figure 3), although they seem to be important to define the interaction with NADPC. This suggests that the enzymes conforming to the consensus glycine motif sequences do not necessarily need to be strict in cofactor recognition. Our enzymatic assays (see below) and the present structure indicate that T. thermophilus HIBADH prefers NADP rather than NAD. The hydrogen bonds between the 2 0 -phosphate group of the adenosine base and residues Asn30, Arg31 and

Thr32, which are located at the end of the b2 strand of the Rossmann fold, may be important for cofactor specificity. From a sequence analysis, it was reported that NAD/NADP specificity of 3-hydroxyacid dehydrogenases may be determined by the region 18–20 residues C-terminal to the conserved GXGXXG motif,20 which corresponds to the Asn30Arg31-Thr32 triad of HIBADH (Figure 3). The NADP-dependent enzyme seems to contain a positively charged residue in the triad, in contrast to a negatively charged residue in the NADdependent enzyme. From this point of view, HIBADH orthologues could be classified into two

Structure of NADP-dependent HIBADH

913

Figure 5. Comparison of binding ligands between HIBADH and 6PGDH in stereo views. (a) NADPC/sulfate binding in the sulfate ion complex form of HIBADH. NADPC and important residues are depicted as ball-and-stick models and labeled. Residues of the domains I and II of a protomer are colored yellow and green, respectively. Residues from the other subunit are shown in light blue. (b) Close-up view of the sulfate binding adjacent to the nicotinamide ring of NADPC molecule. Important hydrogen bonds are indicated by broken lines. (c) Close-up view of the cofactor/substrate binding in 6PGDH with the presentation equivalent to (b). The active NADPC analogue nicotinamide-8-bromo-adenine dinucleotide phosphate (Nbr8ADPC), the substrate 6-phosphogluconate (6PG) and the catalytic general base Lys183 are shown. This picture is produced from a superposition of the cofactor molecule in the Nbr8ADPC–6PGDH complex (PDB, 1PGN) to the 6PG–6PGDH complex structure (PDB, 1PGP).

914

Structure of NADP-dependent HIBADH

groups: NAD-dependent enzymes and NADPdependent enzymes. Thus, T. thermophilus HIBADH may be recognized as a novel type of HIBADH with NADP dependence. Interestingly, the enzymes from bacteria and archaea seem to prefer the arginine residue in the triad, suggesting that NADP is used predominantly in these enzymes. On the other hand, in eukaryotic enzymes with N-terminal targeting signals to mitochondria, an aspartate residue, indicating NAD dependence, is well conserved in the triad. In spite of the observed difference in cofactor dependence, the NADPdependent T. thermophilus HIBADH shares the substrate specificity with the well-characterized mammal NAD-dependent enzymes, as shown in the later section. Probably, the cofactor of HIBADH was evolutionarily selected, depending upon the environment of the enzyme in the cell; the NADdependent eukaryotic enzymes may adapt to the NAD-rich environment in mitochondria, whereas the NADP-dependent enzymes may be preferable for the organisms without mitochondria, such as bacteria and archaea. Complex with a sulfate ion A sulfate ion was reported as a competitive inhibitor of 6PGDH.19 We analyzed the crystal structure of HIBADH complexed with a sulfate ion, determined by the molecular replacement ˚ resolution. The binding of the method at 2.0 A C NADP molecule is essentially identical in the holo enzyme and its complex with a sulfate ion. In the crystal of the sulfate ion complex, the asymmetric unit contains a tetramer of HIBADH. All four subunits contain cofactor molecules and sulfate ions in their active site. The best Ca superposition of the tetramers between the sulfate complex and the ˚ , indicating holo form has an r.m.s.d. value of 0.3 A that the tetramers of both forms share the same subunit/domain orientations in spite of their different packing in the crystals. In contrast to the holo form tetramer, all subunits of the sulfate complex tetramer have both the NADPC molecule and the sulfate ion at their active sites. Although the D subunit in the open conformation has weak electron density for the NADPC molecule, we could

model the cofactor. This observation suggests that a sulfate ion can stabilize cofactor binding to the open protomer. A sulfate ion is located at the active site near the nicotinamide ring of the NADPC molecule (Figure 5). Therefore, it is likely that the sulfate ion and the substrate HIBA share the same binding pocket, as observed in 6PGDH. The sulfate ion is associated mostly with one subunit (between two domains), and additionally with a few residues from domain II of another subunit. One of the oxygen atoms of the sulfate ion makes two hydrogen bonds with the main-chain nitrogen atoms of Gly118 and Gly119, and another oxygen atom makes three hydrogen bonds with the sidechains of Ser117, Lys165 and Ser204. The remaining two oxygen atoms make hydrogen bonds with water molecules. To determine the HIBA binding mode in the substrate-binding pocket unambiguously, the substrate HIBA was added to the mother liquor containing the holo form or the sulfate complex form of HIBADH crystals. However, the crystals were damaged severely upon addition of the substrate, implying substantial ligand-induced conformational changes. We also tried co-crystallization with the substrate under various conditions (unpublished results), but have been unsuccessful to date. HIBADH activity HIBADH activity was measured at 50 8C using S-HIBA as the substrate. HIBADH was active in the presence of both NADC and NADPC. The kcat (sK1) value for NADC was ten times higher than that for NADPC (Table 2). The Km (mM) value for NADPC was 30 times lower than that for NADC. Considering the kcat/Km values, it seems that the enzyme prefers NADPC, especially at the physiological low concentration of the cofactors. Thus, we conclude that NADP might be the physiological coenzyme of T. thermophilus HIBADH. In order to identify the catalytic residue, the pH-dependence of the HIBADH activity was examined at 50 8C (Figure 6). The reaction rate shows its maximum above pH 9.5, and the halfmaximum activity is observed at pH 7.8. This result suggests that the potential catalytic residue has a

Table 2. Enzymatic properties of T. thermophilus HIBADH A. Coenzyme specificity Coenzyme NAD NADPC C

B. Substrate specificity Substrate S-HIBA R-HIBA L-Serine D-Serine

Km (mM)

kcat (sK1)

kcat/Km (MK1 sK1)

0.24 0.0078

2.6 0.24

10.8!103 31.2!103

Km (mM)

kcat (sK1)

kcat/Km (MK1 sK1)

0.28 0.19 0.13 0.13

0.20 0.18 0.16 0.19

0.72!103 0.95!103 1.22!103 1.43!103

915

Structure of NADP-dependent HIBADH

The enzyme was inactive with respect to other carbohydrates such as lactic acid, maleic acid, S-3-hydroxybutyrate, R-3-hydroxybutyrate, L-threonine and D-threonine. Thus, T. thermophilus HIBADH has strict substrate specificity for HIBA and serine, although the enzyme does not distinguish the chirality of the substrates. Although the biological significance of the observed activity to serine is uncertain, HIBADH was suggested to be involved in serine metabolism in Pseudomonas putida,6 which is reminiscent of the importance of this additional activity.

Figure 6. pH-dependence of HIBADH activity at 50 8C. The reaction mixture contained 50 mM buffer, 1 mM NAD and 60 mg/ml of HIBADH in a final volume of 500 ml. The inset denotes the buffers used in the experiment. HIBADH activities were plotted versus pH values measured at 50 8C.

pKa value of 7.8 and its deprotonated species should be active. There is only one candidate for such an ionizable active residue in the vicinity of the substrate-binding site; that is Lys165. The pKa value for the 3-amino group of free lysine in solution is 10.5. However, the pKa value of an ionizable sidechain can be lowered if it is located in an environment of low dielectric constant, such as inside the surface pocket of protein molecules. Therefore, we concluded that Lys165 located at the substrate-binding pocket might be the catalytic residue. The pKa value of Lys165 may be lowered further at the growth temperature of T. thermophilus, around 75 8C, because of the reduced dielectric constant of water at the high temperature. Assuming a decrease of 0.03 in pKa for every 1 deg. C rise in the temperature, which is a typical value for an amino group, the pKa value of Lys165 at 75 8C can be extrapolated as 7.1, which is in the range of physiological pH, in agreement with the fact that T. thermophilus HIBADH is active under the growth conditions of the source organism. Interestingly, the catalytic residue of 6PGDH was also reported to be lysine, Lys183, which corresponds to Lys165 of HIBADH,19 suggesting that the catalytic lysine residue is conserved in the 3-hyroxyacid dehydrogenase family (Figure 5). It has been reported that the HIBADH from rabbit liver converts the optical R- and S- isomers of methylmalonate semialdehyde reversibly to the respective isomers of 3-hydroxyisobutyrate but the enzyme can use both isomers as substrates.10 The present study confirms that T. thermophilus HIBADH can accept both isomers as substrates (Table 2). Interestingly, T. thermophilus enzyme is active also with L-serine or D-serine in the presence of NADPC or NADC, indicating that the methyl group of HIBA can be replaced by the amino group.

Conclusions We determined the 3-D crystal structures of T. thermophilus HIBADH including its complex with sulfate ion for the first time. The cofactor NADPC molecule was bound at the cleft between the two domains of the protomer with the open/ closed interdomain conformations. It is likely that Lys165 functions as the general base for the oxidation reaction of HIBADH, as suggested by the pH-dependence experiment and the analogy from the crystal structure of 6PGDH. The observed intersubunit active site comprising both subunits of the AB dimer suggests the catalytic importance of the oligomeric state. For the oxidation/reduction of the substrate, the hydrogen-bonding network of the intersubunit active site residues seems to be essential. Upon binding NADPC to the enzyme, an induced-fit domain closure has been observed. In the closed protomer, the nicotinamide ring of NADPC is shielded from the solvent upon binding sulfate ion, although the ADP portion of NADPC is significantly exposed to the solvent. The closed interdomain conformation might be important for protecting the HIBADH reaction from free access to the external solvent. From the structural comparison between HIBADH and 6PGDH, it is likely that the observed sulfate ion and the substrate HIBA share the same binding pocket. The conformation of the syn nicotinamide ribose part of NADPC is defined predominantly by the hydrogen bonds with the main chain of Thr226 and the side-chain of invariant Lys234, as well as by the van der Waals stacking with the side-chains of invariant residues Met12 and Phe227, indicating that there is no space for accommodating the substrate HIBA in the vicinity of the nicotinamide ring except for the proposed substrate pocket. With this binding mode of NADPC, the hydride (HK) ion transfer can occur only on the si-face of the nicotinamide ring, indicating that HIBADH is a pro-S (B) dehydrogenase, as in 6PGDH.19 In conclusion, the evidence presented here successfully identified the active site and provided the molecular basis for understanding the catalytic mechanism of HIBADH. Further biochemical studies and the results from ongoing genomics and proteomics projects will most likely underline the importance of this enzyme in the T. thermophilus system.23

916

Materials and Methods Enzyme assays The S and R-isomers of 3-hydroxyisobutyrate were synthesized from their methylesters as described.10 The standard reaction mixture contained 50 mM Ches–NaOH (pH 9.5), 2 mM coenzymes (NADC or NADPC), 8.7 mg/ ml of HIBADH and various concentrations of substrates in a final volume of 500 ml. The reaction mixture was incubated in a cuvette at 50 8C. Reduction of the coenzyme was monitored by measuring the absorbance at 340 nm. The coenzyme specificity and the pHdependence were determined for substrates at a concentration of 5 mM. Sample preparation, crystallization and data collection HIBADH was purified and crystallized as described.11 For the preparation of SeMet-substituted protein, the BL21(DE3)Star (Invitrogen) cells were grown in M9 medium until they reached an absorbance at 600 nm (A600) of 0.4. At this point, 100 mg of L-lysine, 100 mg of L -phenylalanine, 100 mg of L -threonine, 50 mg of L-isoleucine, 50 mg of L-leucine and 60 mg of SeMet were added to 1 l of culture and the cells were grown at 37 8C for a further 1 h, before inducing the expression with 1 mM IPTG overnight at 25 8C. The SeMet-substituted HIBADH was crystallized in a way similar to that used for the native holo enzyme. Crystals of HIBADH in complex with a sulfate ion were obtained by the microbatch method at 23 8C in a way similar to that used for the holo enzyme, except for using a different precipitant solution comprising 0.2 M ammonium sulfate, 0.1 M Mes–NaOH (pH 6.5), 32% (w/v) polyethylene glycol monomethyl ether (PEG-MME) 5000. The sulfate complex crystals with typical dimensions of 0.15 mm! 0.10 mm!0.15 mm were flash-frozen in liquid nitrogen without using a cryoprotectant. They belong to the space ˚ , bZ group P212121, with cell parameters of aZ93.86 A ˚ ˚ 109.03 A, cZ116.39 A, and contain four chains of HIBADH in the asymmetric unit. Although the sulfate complex has the same space group as that of the holo enzyme form, their unit cell parameters are completely different due to their different crystal packing. For all the crystals, data collection was carried out at 100 K under a nitrogen stream. All the data sets were collected using synchrotron radiation at beamline BL26B1 of SPring-8, Japan with R-axis V image plate. An initial data set was collected on a crystal of the holo enzyme to a resolution of ˚ . For the structure determination, MAD data sets 1.8 A from a single crystal of the SeMet-substituted HIBADH ˚ resolution using the peak were collected up to 2.2 A ˚ ), edge (0.97934 A ˚ ), and remote (0.98400 A ˚) (0.97884 A absorption wavelengths of selenium. A complete data set was collected on crystals of HIBADH in complex with ˚ . All the diffraction sulfate ion to a resolution of 2.0 A images were processed using the HKL2000 package.24 Further details are given in Table 1. Structure determination and refinement SOLVE/RESOLVE was used for locating the selenium sites (which located 21 sites out of 24) and for phasing and automated model building, which placed 60% of the residues correctly in the four protomers.25,26 The remaining residues of the tetramer were built manually using

Structure of NADP-dependent HIBADH

QUANTA (Accelrys, San Diego, CA, USA) with the help of the CCP4 suite.27 A complete model, including the SeMet residues of each protomer, was built during the refinement using CNS.28 A 4-fold non-crystallographic symmetry restraint was applied during the initial stages of the refinement, and it was removed at later stages. The ˚ holo enzyme initial model was refined against the 1.80 A data set, first as a rigid body then using restrained refinement interspersed with manual rebuilding and the addition of solvent molecules using the program QUANTA. Several cycles of model building and refinement yielded the final model. The data processing, details of model obtained and refinement statistics are summarized in Table 1. The structure of HIBADH in complex with sulfate ion was solved by the molecular replacement program CNS,28 using the holo enzyme dimer structure as the search model. The cofactor and sulfate ions were built into difference electron density maps and the refinement was carried in the same way as that used for the holo enzyme. Final refinement statistics are given in Table 1. In both cases, the crystallographic R-factor and Rfree (calculated from a test set of 5% of the reflections) were monitored. The same set of Rfree reflections (from refinement of the holo form) was used for the refinement of the HIBDAH-sulfate ion complex. Structural evaluations of the final models of all forms were performed using PROCHECK.13 Protein Data Bank accession numbers The atomic coordinates and structure factors have been deposited in the RCSB Protein Data Bank with accession codes 1J3V (2CVZ) for the holo form and 1WP4 for the sulfate complex form.

Acknowledgements We thank the technical staff of the Highthroughput Factory for assistance with these experiments. We thank A. Hara and M.R.N. Murthy for critical reading of the manuscript, and beamline staff for assistance during data collection at beamline BL26B1 of SPring-8. This work (TT0368/ HTPF00007) was supported by “National Project on Protein Structural and Functional Analysis” funded by the MEXT of Japan.

References 1. Robinson, W. G. & Coon, M. J. (1957). The purification and properties of b-hydroxyisobutyric dehydrogenase. J. Biol. Chem. 225, 511–521. 2. Bannerjee, D., Sanders, L. E. & Sokatch, J. R. (1979). Properties of purified methylmalonate semialdehyde dehydrogenase of Pseudomonas aeruginosa. J. Biol. Chem. 245, 1828–1835. 3. Hasegawa, J. (1981). Distribution in organisms and sterospecificity of b-hydroxyisobutyrate dehydrogenase. Agric. Biol. Chem. 45, 2899–2901. 4. Letto, J., Brosnan, M. E. & Brosnan, J. T. (1986). Valine metabolism: gluconeogenesis from 3-hydroxyisobutyrate. Biochem. J. 240, 909–912. 5. Hawes, J. W., Crabb, D. W., Chan, R. M., Rougraff,

Structure of NADP-dependent HIBADH

6.

7.

8.

9.

10.

11.

12.

13.

14. 15.

16.

17. 18.

P. M. & Harris, R. A. (1995). Chemical modification and site-directed mutagenesis studies of rat 3-hydroxyisobutyrate dehydrogenase. Biochemistry, 34, 4231–4237. Chowdhury, E. K., Nagata, S. & Misono, H. (1996). 3-hydroxyisobutyrate dehydrogenase from Pseudomonas putida E23: purification and characterization. Biosci. Biotechnol. Biochem. 60, 2043–2047. Congdon, P. J., Haigh, D., Smith, R., Green, A. & Pollitt, R. J. (1981). Hypermethioninaemia and 3-hydroxyisobutyric aciduria in an apparently healthy baby. J. Inherited Metab. Dis. 4, 79–80. Landaas, S. (1975). Accumulation of 3-hydroxyisobutyric acid, 2-methyl-3-hydroxyisobutyric acid and 3-hydroxyisovaleric acid in ketoacidosis. Clin. Chim. Acta, 64, 143–154. Ko, F.-J., Nyhan, W. L., Wolff, J., Barshop, B. & Sweetman, L. (1991). 3-Hydroxyisobutyrate aciduria: an inborn error of valine catabolism. Pediatr. Res. 30, 322–326. Rougraff, P. M., Paxton, R., Kuntz, M. J., Crabb, D. W. & Harris, R. A. (1988). Purification and characterization of 3-hydroxyisobutyrate dehydrogenase from rabbit liver. J. Biol. Chem. 263, 327–331. Lokanath, N. K., Shiromizu, I., Nodake, Y., Sugahara, M., Yokoyama, S., Kuramitsu, S. et al. (2003). Crystallization and preliminary X-ray crystallographic studies of NADP-dependent 3-hydroxyisobutyrate dehydrogenase from Thermus thermophilus HB8. Acta Crystallog. sect. D, 59, 2294–2296. Hendrickson, W. A., Horton, J. R. & LeMaster, D. M. (1990). Selenomethionyl proteins produced for analysis by multiwavelength anomalous diffraction (MAD): a vehicle for direct determination of threedimensional structure. EMBO J. 9, 1665–1672. Laskowski, R. A., Macarthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the sterochemical quality of protein structures. J. Appl. Crystallog. 26, 283–291. Kabsch, W. (1976). A solution for the best rotation to relate two sets of vectors. Acta Crystallog. sect. A, 32, 922–923. Sugahara, M., Nodake, Y., Sugahara, M. & Kunishima, N. (2005). Crystal structure of dehydroquinate synthase from Thermus thermophilus HB8 showing functional importance of the dimeric state. Proteins: Struct. Funct. Genet. 58, 249–252. Lokanath, N. K., Shiromizu, I., Ohshima, N., Nodake, Y., Sugahara, M., Yokoyama, S. et al. (2004). Structure of aldolase from Thermus thermophilus HB8 showing the contribution of oligomeric state to thermostability. Acta Crystallog. sect. D, 60, 1816–1823. Holm, L. & Sander, C. (1998). Touring protein fold space with Dali/FSSP. Nucl. Acids Res. 26, 316–319. Adams, M. J., Gover, S., Leaback, R., Phillips, C. & Somers, D. (1991). The structure of 6-phosphogluco˚ resolution. Acta nate dehydrogenase refined at 2.5 A Crystallog. sect. B, 47, 817–820.

917 19. Adams, M. J., Ellis, G. H., Gover, S., Naylor, C. E. & Phillips, C. (1994). Crystallographic study of coenzyme, coenzyme analogue and substrate binding in 6-phosphogluconate dehydrogenase: implications for NADP specificity and the enzyme mechanism. Structure (Camb), 2, 651–668. 20. Hawes, J. W., Harper, E. T., Crabb, D. W. & Harris, R. A. (1996). Structural and mechanistic similarities of 6-phosphogluconate and 3-hydroxyisobutyrate dehydrogenase reveal a new family, the 3-hydroxyacid dehydrogenases. FEBS Letters, 389, 263–267. 21. Roosild, T. P., Miller, S., Booth, I. R. & Choe, S. (2002). A mechanism of regulating transmembrane potassium flux through a ligand-mediated conformational switch. Cell, 109, 781–791. 22. Rossmann, M. G., Moras, D. & Olsen, K. W. (1974). Chemical and biological evolution of nucleotidebinding protein. Nature, 250, 194–199. 23. Yokoyama, S., Hirota, H., Kigawa, T., Yabuki, T., Shirouzu, M., Terada, T. et al. (2000). Structural genomics projects in Japan. Nature Struct. Biol. 7, 943–945. 24. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. 25. Terwilliger, T. C. & Berendzen, J. (1999). Automated MAD and MIR structure solution. Acta Crystallog. sect. D, 55, 849–861. 26. Terwilliger, T. C. & Berendzen, J. (2001). Maximumlikelihood density modification with pattern recognition of structural motifs. Acta Crystallog. sect. D, 57, 1755–1762. 27. Collaborative Computational Project, Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760–763. 28. Bru¨nger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallographic & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905–921. 29. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946–950. 30. Esnouf, R. M. (1999). Further additions to Molscript version 1.4, including reading and contouring of electron-density maps. Acta Crystallog. sect. D, 55, 938–940. 31. Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505–524. 32. Gouet, P., Robert, X. & Courcelle, E. (2003). ESPript/ ENDscript: extracting and rendering sequence and 3D information from atomic structures of proteins. Nucl. Acids Res. 31, 3320–3323. 33. Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic diagrams of protein–ligand interactions. Protein Eng. 8, 127–134.

Edited by R. Huber (Received 14 June 2005; received in revised form 21 July 2005; accepted 27 July 2005) Available online 11 August 2005