INTRODUCTION

The oxidative deamination of amino acids to their corresponding keto acids is catalyzed by the family of amino acid dehydrogenases and provides a route for the incorporation of ammonia into organic compounds and links the metabolism of carbohydrates and amino acids. Sequence homology between glutamate (GluDH),¹ leucine (LeuDH), valine (ValDH), and phenylalanine (PheDH) dehydrogenases (1-5) clearly indicates the existence of an enzyme superfamily related by divergent evolution (1). This enzyme family has considerable commercial potential for the production of novel non-proteogenic amino acids for the pharmaceutical industry (6, 7) and for the diagnosis of genetic diseases of amino acid metabolism including phenylketonuria (8), maple syrup urine disease (9, 10), and homocystinuria (11). A thorough understanding of the way in which members of this family achieve differential substrate specificity might not only enhance our understanding of the relationship between molecular structure and biological function, but may also have important industrial applications.

ValDH (EC 1.4.1.8) catalyzes the reversible oxidative deamination of L-valine to 2-ketoisovalerate, with the corresponding reduction of the cofactor NAD⁺ (Scheme 1). This enzyme has been characterized in species of Streptomyces, where it functions in the catabolism of branched chain amino acids and also plays an important role in providing precursors for the biosynthesis of polyketide antibiotics (12). The ValDH genes from Streptomyces cinnamonensis (4), Streptomyces coelicolor (13) and Streptomyces fradiae (14) have been cloned and sequenced and contain 358, 364, and 371 amino acids, respectively, with subunit M_r of approximately 38,000. Biochemical studies have established that the ValDHs from S. cinnamonensis and S. coelicolor are homodimers (15, 16). In contrast, the common quaternary structure for GluDH is based on a hexamer and for LeuDH and PheDH an octamer.

There are small differences in substrate specificity between ValDH and LeuDH. Members of the ValDH family (15-19) exhibit preferential substrate specificity toward valine; however, other hydrophobic branched chain amino acids are also accepted by these enzymes but with lower activity (Table I). For example, the k_cat/K_m values for S. cinnamonensis ValDH (15) when valine and leucine are used as substrate are 21.8 and 3.0 mM¹ s¹, respectively. In contrast, LeuDH (20-24) favors the 1-carbon longer branched chain amino acid leucine, with, for example, the k_cat/K_m values for Corynebacterium pseudodiptheriticum LeuDH (20) with valine and leucine as substrate being 73.3 and 101.7 mM¹ s¹, respectively (Table II). Using a structure-based sequence alignment, we have analyzed the similarities and differences between ValDH, LeuDH, and GluDH and present our findings below, arranged into sections that describe the similarity in amino acid sequence and its consequences for the tertiary structure, quaternary structure, catalytic mechanism, and substrate specificity of ValDH.

Table I. Substrate specificities in the oxidative deamination direction for the ValDHs from S. cinnamonensis (15), S. coelicolor (16), S. aureofaciens (17), S. fradiae (18), and A. faecalis (19) Substrate S. cinnamonensis S. coelicolor S. aureofaciens S. fradiae A. faecalis Km kcat kcat/Km Relative activitya Km Relative activitya Km Relative activitya Km Relative activitya Km mM s1 mM1s1 % mM % mM % mM % mM L-Valine 1.3 28 21.8 100 10.0 100 2.5 100 1.0 100 2.3 L-Norvaline 3.2 18 5.6 43 5.7 98 1.3 44 L--Aminobutyrate 19.6 104 5.3 130 17 14.8 69 3.1 33 L-Leucine 4.0 11.9 3.0 8 36 6.3 25 0.7 73 L-Isoleucine 10.8 5.8 0.5 28 47 5.0 29 1.2 72 L-Norleucine 30.0 17.9 0.6 11 15.6 52 6.7 16 a Relative activity is the specific activity of the particular amino acid substrate relative to L-valine, expressed as a percentage.

Table II. Substrate specificities in the oxidative deamination direction for the LeuDHs from C. pseudodiptheriticum (20), B. sphaericus (21), T. intermedius (22), B. licheniformis (23), and B. cereus (24) Substrate C. pseudodiptheriticum B. sphaericus T. intermedius B. licheniformis B. cereus Km kcat kcat/Km Relative activitya Km Relative activitya Km Relative activitya Km Relative activitya Km mM s1 mM1s1 % mM % mM % mM % mM L-Valine 0.6 41.8 73.3 74 1.7 87 2.4 59 12.5 61 2.5 L-Norvaline 0.5 15.5 28.7 41 3.5 27 28 2.9 L--Aminobutyrate 2.7 7.5 2.8 14 10.0 8 32 24 22.0 L-Leucine 0.3 30.5 101.7 100 1.0 100 2.0 100 2.1 100 1.5 L-Isoleucine 0.8 12.5 15.4 58 1.8 89 0.4 72 3.3 61 1.0 L-Norleucine 2.8 4.2 1.5 10 6.3 7 6 1.5 a Relative activity is the specific activity of the particular amino acid substrate relative to L-leucine, expressed as a percentage.

RESULTS AND DISCUSSION

The close relationship between the sequences of ValDH and LeuDH and the more remote relationship with GluDH are illustrated in the DIAGON plots presented in Fig. 1. S. cinnamonensis ValDH exhibits 49% and 20% identities with Thermoactionomyces intermedius LeuDH and Clostridium symbiosum GluDH over the 346 and 349 residues that can be aligned, respectively. Nevertheless, of the 68 residues that are strongly conserved in the GluDH family (25), 33 are also conserved in the ValDH family, indicating that all these enzymes are closely related.

The crystal structures of C. symbiosum GluDH (26, 27) and B. sphaericus LeuDH (28) have been solved and can be seen to be closely related despite having a low sequence identity of 19% (28). In both cases, the subunit structure is based on two domains separated by a deep cleft. Domain I is constructed primarily from residues in the N-terminal half of the polypeptide chain and is exclusively involved in subunit assembly. In contrast, Domain II comprises residues from the C-terminal half of the polypeptide chain and is responsible for nucleotide binding. The structure-based sequence alignment of representative members of the ValDH, LeuDH, and GluDH families is presented in Fig. 2, and, unless otherwise stated, the sequence numbering and identification of the secondary structural elements in B. sphaericus LeuDH are used to identify equivalent residues in the ValDH sequences throughout this paper. With the exception of residues close to the N and C termini and a region of the polypeptide chain that includes 8, i, and 9 (all of which lie on the periphery of the molecule), ValDH and LeuDH show considerable sequence similarity over the entire length of the polypeptide chain, implying that their tertiary structures will be almost identical (Fig. 3a). Three sites of insertion and deletion are found between LeuDH and all ValDHs, including approximately 11 additional residues at the N terminus, 3 additional residues in the loop connecting 8 to i, and the deletion of approximately 15 residues at the C terminus. Additionally, there are 5 extra residues in the d-3 loop and a 1-residue deletion in the 9-j loop in S. fradiae ValDH. Previous studies have noted that insertions and deletions at these positions occur commonly within the wider enzyme superfamily (1).

a, a schematic diagram of a B. sphaericus LeuDH subunit, produced using MOLSCRIPT (37), showing the helices and strands conserved in ValDH colored yellow and portrayed as labeled helical ribbons and arrows, respectively. The helices 8 and 9 and the strand i in LeuDH, which show low sequence similarity with ValDH, are highlighted in green. Regions of insertions and deletions between LeuDH and ValDH are highlighted in red. b, a view from the 42 symmetry point down the two-fold axis, which relates the dimer, AB, with the four-fold axis vertical. c, close-up of view b. The secondary structural elements implicated in interactions across the two-fold axis relating dimers are labeled. d, a view from the 42 symmetry point down the two-fold axis, which relates the pairs of dimers, AB and AB, with the four-fold axis vertical. e, close-up of view d. The interaction of 14 and the C-terminal loop of subunit A (red) with the complementary U-shaped pocket formed by the symmetry-related protruding arm and the loops connecting b to c and 3 to e of subunit B (yellow) and 3, 4, and the loop connecting a to b of subunit A (blue) about the two-fold axis can clearly be seen. The C-terminal residues that form the last turn of 14 and the loop to the C terminus in LeuDH and which are deleted in ValDH are highlighted by a dashed outline for subunits A and B, respectively.

Crystallographic studies have shown that C. symbiosum GluDH assembles into a hexamer with 32 symmetry (26), this being the most common quaternary structure within this enzyme family. In contrast, the structure determination of B. sphaericus LeuDH has shown it to assemble into an octamer with 42 symmetry (Refs. 28 and 29; Fig. 3, b-e). Unfortunately, the complete sequence of the latter is not yet available and the structure is currently based on a partial amino acid sequence and a combination of a sequence determined by inspection of the electron density map and the aligned sequences of other species variants of the enzyme. In LeuDH, the dimer interface consists of three interacting complementary areas (Fig. 3b). Centrally, strand a from Domain I interacts with its symmetry-related mate across the two-fold axis to form an antiparallel sheet of 12  strands spanning both subunits (Fig. 3c). Another secondary structural element involved in interactions across the two-fold interface is the C-terminal end of 3, which packs against its two-fold related counterpart. The final interaction involves residues from one face of 1, which pack against a, b, d, and 2 in the symmetry-related subunit. The sequence alignment strongly indicates that, with the exception of 1 for which the sequence similarity is low, each of the elements of secondary structure that are involved in these interactions are conserved in ValDH. Furthermore, of the 33 residues that take part in interactions across the two-fold axis in LeuDH, 15 residues are identical in at least six of the eight aligned LeuDH and ValDH sequences, implying that the character of this interface is maintained.

Around the two-fold axis relating pairs of dimers (Fig. 3d), an important feature of the four-fold interface in B. sphaericus LeuDH involves the interaction of a protruding arm, formed by residues at the end of 14 (residues 347-350) and the loop to the C terminus (residues 351-364) from subunit A of an AB dimer, with residues in a complementary U-shaped pocket constructed by the neighboring dimer, AB (Fig. 3e). One face of this pocket is built from the symmetry-related protruding arm of subunit B, the base of the pocket is constructed from the loops connecting b to c and 3 to e of the same subunit, and the other face is formed by the loops connecting a to b, the N-terminal end of 3, and the C-terminal end of 4 from subunit A. Calculations of the accessible surface area using the algorithm of Lee and Richards (30) have shown that the LeuDH monomer possesses an accessible surface area of 16,100 Å² and on octamer formation 3,400 Å² (21%) of this surface is buried, of which 1,250 Å² is buried on dimer formation and the remainder on assembly of dimers to form the octamer (Fig. 4). Analysis of the structure-based sequence alignment in Fig. 2 illustrates that the final 15 C-terminal residues in LeuDH are deleted in the ValDH sequences. In LeuDH, these residues contribute 45% (950 Å² out of 2,150 Å²) of the surface that becomes buried on the assembly of dimers to form the octamer. Therefore in ValDH, the deletion of these residues implies that an octameric arrangement for this enzyme is unlikely, which is consistent with the proposed dimeric quaternary structures of the ValDHs from S. cinnamonensis and S. coelicolor (15, 16).

Proposals for the molecular basis of the catalytic mechanism of members of the amino acid dehydrogenase superfamily have been suggested following the structure determination of the binary complex of the glutamate dehydrogenase from C. symbiosum with glutamate (27), coupled to earlier kinetic studies on bovine GluDH (31, 32). The structural studies have highlighted a number of potentially important residues, which are fully conserved in GluDH, LeuDH, and PheDH. In LeuDH, these key residues include 5 glycines (residues 41, 42, 77, 78, and 290) important in the design of the active site; Lys⁶⁸, which recognizes the 1-carboxyl group of the amino acid substrate; Asp¹¹⁵, which is believed to be involved in proton transfer to and from the amino acid substrate during catalysis; and Lys⁸⁰, which has an unusually low pK_a and is thought to enhance the nucleophilicity of an essential water molecule involved in the proposed reaction mechanism (Fig. 6a). Comparison of the sequences for LeuDH and ValDH in Fig. 2 reveals that all of these residues are conserved, indicating that they share an identical catalytic mechanism.

LeuDH shows a considerable structural resemblance to other NAD(P)⁺-linked dehydrogenases in its nucleotide binding domain (Fig. 5) and recognizes the nucleotide in a similar manner (28, 33). Specific interactions between LeuDH and the NAD⁺ include hydrogen bonds between the adenine ribose hydroxyl groups and the side chain carboxyl group of Asp²⁰³. Additionally, the side chain of Asp²⁰³ forms a hydrogen bond to the main chain peptide NH group of Leu¹⁸¹, which lies in a glycine-rich region (residues 180-185) between g and the dinucleotide-binding helix, 7 (34). Furthermore, in LeuDH the pyrophosphate moiety of the NAD⁺ forms hydrogen bonds to the main chain peptide NH groups of Asn¹⁸³ and Val¹⁸⁴ within this region. In ValDH, the presence of an aspartate equivalent to Asp²⁰³ and the glycine-rich region can be inferred from the sequence alignment, implying that the interactions with this part of the NAD⁺ cofactor in ValDH and LeuDH are similar. In B. sphaericus LeuDH, the 2 and 3 hydroxyl groups of the nicotinamide ribose make hydrogen bonds to the side chain carboxyl group and main chain peptide NH group of Asp²⁶¹, respectively, a residue that is also conserved in B. stearothermophilus LeuDH. Although this residue is an asparagine in the six remaining LeuDH and ValDH sequences, it is anticipated that it will also be involved in cofactor recognition.

In LeuDH, the nicotinamide ring lies in the cleft between the two domains, above 12 which forms one face of the active site pocket, and close to 6. Considerable homology between LeuDH and ValDH can be detected in this area. This includes the conservation of Thr¹⁵⁰, which forms a hydrogen bond to the carboxyamide moiety of the nicotinamide ring and which is responsible for determining the conformation of the glycosidic bond that leads to the presentation of the 4-pro-S hydrogen of the NADH toward the active site in LeuDH. The conservation of residues in this region, and in particular Thr¹⁵⁰, provides a molecular explanation for the identical stereospecificity of the hydride transfer step observed in S. cinnamonensis ValDH (15).

A comparison of the x-ray structures of C. symbiosum GluDH and B. sphaericus LeuDH (26, 28) has revealed that the discrimination of the amino acid substrate between these two enzymes is achieved by a combination of point mutations at the base of the substrate side chain binding pocket (Lys⁸⁹ and Ser³⁸⁰ in GluDH, equivalent to Leu⁴⁰ and Val²⁹⁴ in LeuDH, respectively) and changes in the shape of this pocket resulting from movements in the main chain.

In LeuDH, there are 13 residues with a side chain atom lying within 6 Å of the expected position of the leucine substrate determined by molecular modeling using the binding site of glutamate in GluDH as a guide (Table III). These include Leu⁴⁰ and Val²⁹⁴, which are involved in crucial interactions with the side chain of the amino acid substrate and are thought to be principally responsible for controlling the amino acid specificity. All of these residues are identical in the seven aligned ValDH and LeuDH sequences and in the x-ray LeuDH sequence (Fig. 6a). Given the difference in substrate specificity between ValDH and LeuDH, this is somewhat surprising. However, a number of differences occur in the residues that lie more remote from the substrate binding site. Thus, of the 40 residues with either a side chain or main chain atom between 6 and 10 Å from the leucine substrate, 28 are identical in all the aligned ValDH and LeuDH sequences, including the B. sphaericus LeuDH x-ray sequence (Table IV). Although the sequences at three positions (residues 43, 62, and 261 in LeuDH) differ between some ValDHs and LeuDHs, similarities can also be seen between some members of the two families. Thus, we assume that these positions are less important in the observed differential substrate specificity. The remaining nine positions provide substitutions that are identical within, but not between, members of the LeuDH and ValDH families and include T66S, L76H, T81A, I111V, E114C, V133T, F140N, P146S, and N293Q. The sequence differences are highlighted in Fig. 6b, where it can be seen that they form the second shell of residues surrounding those on the surface of the substrate binding site. Furthermore, examination of Fig. 6b also indicates that the C-terminal tail in LeuDH lies directly behind the elements of the secondary structure that form the substrate binding pocket and therefore its absence in ValDH may lead to further subtle changes in the detailed conformation within this region. Taken together, the similarities in the residues which lie directly in the amino acid specificity pocket coupled with the changes in the second shell suggest that the modulation of substrate specificity between these enzymes arises in part through changes in the shape of the amino acid specificity pocket caused by differences in the relative position of residues which line the pocket, rather than their mutation. The values for the relative rates of LeuDH and ValDH, highlighted in Tables I and II, indicate that the energetics of discrimination between the substrates leucine and valine are small and thus minor movements of the side chains within the active site can be sufficient to produce this discrimination. Currently it is impossible to predict the nature of any such subtle changes by homology-based modeling, and structural data are required to understand how these differences provide fine control over the substrate specificity. In the longer term, information gleaned from the comparison of closely related but distinct structures may improve our ability to model these types of changes and advance our understanding of the relationship between structure and function.

Table III. Residues in B. sphaericus LeuDH with at least one side chain atom (or C in the case of glycine) residing within 6 Å of the modeled leucine substrate These include Leu40 and Val294, critical for the amino acid specificity of LeuDH. All 13 residues are identical in the ValDH sequences, implying that the character of the amino acid binding pocket is preserved. Residue (LeuDH x-ray sequence) No. of side chain atoms <6 ÅA Side chain atom nearest substrate Closest substrate atom Separation Å Leu40 4 C2 C2 4.0 Gly41 C C1 4.3 Gly42 C N 4.2 Arg44 1 NH2 O1 5.6 Met65 4 C C1 2.8 Lys68 4 C O2 4.5 Asn69 2 O1 C1 5.3 Lys80 3 N O1 3.6 Ala113 1 C C2 3.5 Asp115 4 O2 N 2.7 Thr134 1 C2 C2 4.9 Val291 3 C2 O2 2.8 Val294 3 C2 C2 3.7

Table IV. Patterns of sequence substitution between LeuDH and ValDH close to the active site Of the 40 residues in LeuDH with either a side chain or main chain atom lying between 6 and 10 Å from the leucine substrate, 28 residues are identical in all the aligned ValDH and LeuDH sequences: Ala39, Leu61, Gly64, Tyr67, Ala72, Leu74, Gly77, Gly78, Gly79, Tyr110, Thr112, Val116, Gly117, Thr118, Met123, Gly135, Thr150, Asn262, Tyr284, Asn287, Ala288, Gly289, Gly290, Ile292, Ala295, Asp296, Gly297, and Leu298. The sequences at three positions (43, 62, and 261) differ between some ValDHs and LeuDHs, but similarities can be seen between some members of the two families. The remaining nine positions provide substitutions that are identical within, but not between, members of the LeuDH and ValDH families (T66S, L76H, T81A, I111V, E114C, V133T, F140N, P146S, and N293Q). Residue no. (B. sphaericus x-ray sequence numbering) Residue type in LeuDH (x-ray/Bc/Bl/Bst/Ti) Residue type in ValDH (Sci/Sco/Sfr) 43 Ala/Thr/Thr/Thr/Met Thr 62 Ala/Ala/Ala/Ala/Gly Ser/Ala/Ala 66 Thr Ser 76 Leu His 81 Thr Ala 111 Ile Val 114 Glu Cys 133 Val Thr 140 Phe Asn 146 Pro Ser 261 Asp/Asn/Asn/Asp/Asn Asn 293 Asn Gln