Structure of NADP+-dependent glutamate dehydrogenase from Escherichia coli – reflections on the basis of coenzyme specificity in the family of glutamate dehydrogenases


  • Michael A. Sharkey,

    1. School of Biomolecular and Biomedical Science, Conway Institute, University College Dublin, Ireland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Tânia F. Oliveira,

    1. School of Biochemistry and Immunology, Trinity College Dublin, Ireland
    2. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa (ITQB/UNL), Oeiras, Portugal
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Paul C. Engel,

    1. School of Biomolecular and Biomedical Science, Conway Institute, University College Dublin, Ireland
    Search for more papers by this author
  • Amir R. Khan

    Corresponding author
    1. School of Biochemistry and Immunology, Trinity College Dublin, Ireland
    • Correspondence

      A. R. Khan, School of Biochemistry and Immunology, Trinity College Dublin, Dublin 2, Ireland

      Fax: +353 1 677 2400

      Tel: +353 1 896 3868


    Search for more papers by this author


Glutamate dehydrogenases (GDHs; EC–4) catalyse the oxidative deamination of l-glutamate to α-ketoglutarate, using NAD+ and/or NADP+ as a cofactor. Subunits of homo-hexameric bacterial enzymes comprise a substrate-binding domain I followed by a nucleotide-binding domain II. The reaction occurs in a catalytic cleft between the two domains. Although conserved residues in the nucleotide-binding domains of various dehydrogenases have been linked to cofactor preferences, the structural basis for specificity in the GDH family remains poorly understood. Here, the refined crystal structure of Escherichia coli GDH in the absence of reactants is described at 2.5-Å resolution. Modelling of NADP+ in domain II reveals the potential contribution of positively charged residues from a neighbouring α-helical hairpin to phosphate recognition. In addition, a serine that follows the P7 aspartate is presumed to form a hydrogen bond with the 2′-phosphate. Mutagenesis and kinetic analysis confirms the importance of these residues in NADP+ recognition. Surprisingly, one of the positively charged residues is conserved in all sequences of NAD+-dependent enzymes, but the conformations adopted by the corresponding regions in proteins whose structure has been solved preclude their contribution to the coordination of the 2′-ribose phosphate of NADP+. These studies clarify the sequence–structure relationships in bacterial GDHs, revealing that identical residues may specify different coenzyme preferences, depending on the structural context. Primary sequence alone is therefore not a reliable guide for predicting coenzyme specificity. We also consider how it is possible for a single sequence to accommodate both coenzymes in the dual-specificity GDHs of animals.


bovine glutamate dehydrogenase


Clostridium symbiosum glutamate dehydrogenase


protomer A of Escherichia coli glutamate dehydrogenase


Escherichia coli glutamate dehydrogenase


protomer F of Escherichia coli glutamate dehydrogenase


glutamate dehydrogenase


Peptoniphilus asaccharolyticus glutamate dehydrogenase


Protein Data Bank


Pyrobaculum islandicum glutamate dehydrogenase


Glutamate dehydrogenases (GDHs; EC–4) play a fundamental role in nitrogen and carbon metabolism. GDH links amino acid metabolism to the tricarboxylic acid cycle through the conversion of l-glutamate to 2-oxoglutarate (α-ketoglutarate) via oxidative deamination. The inverse reductive amination of 2-oxoglutarate supplies nitrogen for several biosynthetic pathways [1]. GDH belongs to the amino acid dehydrogenase enzyme superfamily, whose members show different substrate specificities and have considerable potential in the production of novel nonproteinogenic amino acids for the pharmaceutical industry [2, 3]. Another application of amino acid dehydrogenases is in diagnostics: phenylalanine dehydrogenase, for example, has been widely exploited for the diagnosis of phenylketonuria [4].

Bacterial and mammalian GDHs are hexameric enzymes that assemble with ‘32’ point group symmetry. Each subunit consists of a largely N-terminal substrate-binding domain I that folds into an α/β structure with a central five-stranded β-sheet, and an NAD(P)+-binding domain II that adopts a modified Rossmann fold [5, 6]. Domain I mediates the majority of intersubunit contacts along the three-fold and two-fold axes of hexamers. The substrate-binding pocket is found in a deep groove at the juxtaposition of the two domains, and there is significant diversity in the relative orientation of domains I and II in known crystal structures [7]. Sequence identities among the bacterial enzymes so far described vary from 25% to > 90%. The mammalian hexameric enzymes are structurally similar to their bacterial counterparts, except that they have an insertion of an α-helical ‘antenna’ at the three-fold axis that regulates catalytic activity via binding to nucleotide effectors [8]. Several fungal species also contain tetrameric GDHs, but there are no known 3D structures of this class of dehydrogenases, although sequence homology indicates that their much larger (~ 115 kDa) subunits include the elements of a typical hexameric GDH subunit [9].

Despite their overall structural similarities, microbial GDHs have distinct cofactor preferences. The enzymes can be subdivided into three groups, depending on their preference for NAD+, NADP+, or both almost equally [10-12]. Analyses of sequences and structures of other nicotinamide cofactor-dependent enzymes have led to various explanations for the molecular basis of cofactor specificity. The presence of an acidic residue at a position near the 2′-OH and 3′-OH of adenosine has been reported to offer specific recognition of NAD+ and to discriminate against NADP+ by electrostatic repulsion [13]. This important site, termed the P7 position, has been emphasized, along with other residues adjacent to the glycine-rich P-loop (Gly-X-Gly-X-X-Gly/Ala), the so-called ‘fingerprint sequence’, for mediating cofactor specificity in the widespread Rossman fold of dehydrogenases [14-16]. The presence of glycine or alanine at P6 reputedly favours binding of NAD+ or NADP+, respectively.

Historically, the most widely studied bacterial GDH is the Clostridium symbiosum enzyme (CsGDH). However, CsGDH is an exception that violates the fingerprint rules [12, 17], differing in the residues at P6 (alanine) and P7 (glycine), despite rigorous (20 000-fold) NAD+ discrimination over NADP+ [18]. In marked contrast, our recent structure of another bacterial NAD+-dependent GDH, that of Peptoniphilus asaccharolyticus (PaGDH), reveals a pattern much more in keeping with the conventional view described above; a glutamate at the P7 position (Glu243) forms hydrogen bonds with the 2′-OH and 3′-OH of the adenosine ribose, facilitated by other residues in the vicinity that allow close approach to the sugar [19]. The comparison of these two structures thus supports the view that a single enzyme family may offer strikingly different ways of achieving the same specificity. Therefore, cofactor specificity is a complex phenomenon that cannot be extrapolated from sequence motifs alone.

The NADP+-dependent GDH from Escherichia coli has, like PaGDH, an acidic residue at P7 (Asp263), which would not seem to be ideal in the adenosine phosphate-binding pocket, although it is conserved in other NADP+-specific GDHs. Successful crystallization of E. coli GDH (EcGDH) was initially reported almost 20 years ago [20]. Also, in the course of completing our investigations reported here, a low-resolution structure of EcGDH (3.2 Å), determined with high-throughput proteomics techniques applied to the native source protein, was deposited in the Protein Data Bank (PDB) (PDB code 3sbo [21]). Here, we report the refined structure of the free enzyme at 2.5-Å resolution. In this structure, wild-type EcGDH has a closed catalytic cleft even in the absence of substrate and cofactor, which is unprecedented for the bacterial GDH family. Modelling of NADP+ in domain II reveals the presence of three lysine/arginine residues in the vicinity of the 2′-phosphate of the adenosine ribose, and mutagenesis of these residues has confirmed their role in coenzyme binding. Despite the apparent sequence conservation of one of these positively charged residues in both NAD+ and NADP+ enzymes, we observe that it is the context and spatial orientation of the α11–loop–α12 hairpin, where these residues are located, that dictate NADP+ specificity.


Overall architecture of EcGDH and the closed conformation

The free enzyme form of EcGDH is a hexamer with 32 symmetry, an assembly that is common to all mammalian and bacterial enzymes (Fig. 1). Domains I and II of each protomer fold with a central mixed β-sheet connected by α-helices, with domain II adopting a modified Rossmann nucleotide-binding fold. Domain I, which mediates the 32 symmetry contacts and is responsible for substrate binding, comprises residues 1–203 and 425–447. Domain II (residues 204–424) is responsible for cofactor binding, as recently confirmed by the construction of an active chimaeric enzyme comprising domain I of CsGDH and domain II of EcGDH, and showing NADP+ specificity [22]. EcGDH and CsGDH contain an extra α-helix (α1) at the N-terminus relative to other bacterial enzymes such as PaGDH (Fig. 2). In addition, helices α2/α3 can be considered as a single curved helix, rather than two distinct α-helices. However, in order to avoid confusion, the nomenclature used for the secondary structure assignment corresponds to the original CsGDH structure [5]. The final model (chains A–F) fits quite well in electron density maps, with only a small fraction (1.48%) lying outside favourable Ramachandran space. One exception is domain II (residues 210–360) in chain D, which contains poorly defined (residues 222–234 and 268–274) and disordered (residues 315–319) regions. The quality of the structures is shown with an electron density map in Fig. S1.

Figure 1.

Assembly of EcGDH into the biological hexamer. Left: view of the two-fold axis along a horizontal line. Right: view of the three-fold axis in the line of sight. The three final α-helices of the enzyme (α15–α17) are coloured red, and the pivot helix is identified by the ‘P’ mark (white label) on the right panel. The N-terminal segment of the pivot helix lies near the centre of the three-fold axis.

Figure 2.

Sequence alignment of bacterial GDHs. Secondary structure elements corresponding to EcGDH are shown above the sequence. The three triangles mark the trio of positively charged residues that are poised for recognizing the 2′-phosphate of NADP+. The P1, P6 and P7 residues are annotated below the sequences.

Analysis of the cleft between domains I and II reveals that five of six protomers (chains A–E) are in a relatively closed conformation. The degree of closure is roughly demonstrated by modelling the cofactor in domain II, and measuring its closest approach to domain I (Fig. 3). Protomer F (EcGDH-F) is the only molecule in the hexamer that adopts an open conformation that is sufficiently wide for diffusion of both substrate and cofactor. In contrast to EcGDH, the structures of all other wild-type GDHs in the substrate-free state have an open conformation [5, 7, 19, 23]. The distance between the basic residue (Lys136) in domain I and the 2′-phosphate in the model of EcGDH–NADP+ is only 7.6 Å (not shown). Although the conformation of EcGDH is clearly not fully closed, as compared with the bovine GDH (BvGDH)–NADP+ complex [24], the cleft is sterically inaccessible to both the cofactor and the substrate.

Figure 3.

Closed conformation of EcGDH. The N-terminal and C-terminal domains of EcGDH are coloured blue and grey, respectively. NADP+ is shown at the active site in space-filling models. The pivot helix (α15), which binds to substrate and mediates interdomain motion, is coloured yellow. Domain closure is exemplified by EcGDH-A (left), whereas EcGDH-F (right) is the only protomer in the open conformation. The measured distance corresponds to Asp168 (Cα) and the 2′-OH group of nicotinamide ribose in the two models of the EcGDH–NADP+ complex.

Previous work on the structure of the K89L mutant of CsGDH revealed domain closure even in the absence of substrate [7]. However, it was unclear whether domain closure is enabled by the K89L mutation, as it abrogates a positive charge at the interface of the two domains. The structure of wild-type EcGDH reveals that domain closure is a property of the free enzyme in the absence of substrate and cofactor. The functional consequences of a closed conformation of EcGDH prior to catalysis require further investigation. It would be speculative to correlate the crystalline state directly with the solution state of the enzyme. Furthermore, the effects on catalytic turnover would depend on the timescale of opening/closing, and the rate-limiting step in the reaction, neither of which have been well defined for the bacterial enzymes.

Cofactor specificity in EcGDH

The structure of NADP+-dependent EcGDH and its comparison with previously published NAD+-dependent enzymes provides insights into the molecular basis for cofactor specificity. Structural determinants can be extrapolated from molecular modelling, as domain II acts as a rigid body and appears to maintain most of the secondary and tertiary structural elements required for cofactor binding. Currently, there is no published structure of a bacterial GDH in complex with NADP+. Successful soaking of NAD+ into the active site of CsGDH has been reported previously [5]. Interpretable electron density was seen for one-third of molecules in the asymmetric unit, and binding was accompanied by subtle side chain rotations. The structure of an archaeal GDH from Pyrobaculum islandicum (PiGDH; PDB code 1v9l) in complex with NAD+ remains the only published structure of a nonmammalian GDH with its cofactor for which the coordinates are available in the PDB [25].

In the absence of a bacterial GDH–NADP+ complex, the cofactor was modelled at the active site by superimposing the bovine structure (1hwz) on domain II of EcGDH (see Experimental procedures). Docking of the cofactor onto protomer A of EcGDH (EcGDH-A) reveals complementarity to a substantial part of the cofactor pocket in domain II (Fig. 4). The specificity of EcGDH arises from a combination of sequence and conformational determinants that are evident from molecular modelling of NADP+ onto the free enzyme structure. Three positively charged amino acids – Lys286, Arg289, and Arg292 – are in the vicinity of the 2′-phosphate, and would be available for electrostatic interactions, with minor conformational adjustments (Fig. 4). Lys286 and Arg292 are within 2.0 Å and 2.8 Å of the phosphate, respectively, whereas the Cα atom of Arg289 is 10.2 Å away. Following superposition, no subsequent energy minimization or rigid-body docking was performed in our modelling. Therefore, small changes in the backbone/side chain conformations would be sufficient to optimize Lys286 contacts with the negatively charged phosphate in the EcGDH–NADP+ complex. With respect to the more distant Arg289, selection of an alternative side chain conformer results in a distance of only 4.2 Å between the guanidino group and the 2′-phosphate of adenine ribose. Enabling all of these potentially favourable electrostatic contacts is the position of α11–loop–α12, which contributes the basic residues. The side chain of Asp263 (the P7 residue of EcGDH) from βH points directly at the α-helical dipole of α12 (Fig. 4), which aids in the proper alignment of α11–loop–α12 for the recognition of NADP+ over NAD+. In fact, the side chain of Asp263 forms a hydrogen bond with the backbone NH of Val293 (3.0 Å). This aspartate is the equivalent of Glu243 in PaGDH (P7 position). In PaGDH, the glutamate forms hydrogen bonds with the 2′-OH and 3′-OH of adenine ribose, and thus points in a completely different direction [19]. The next residue, Ser264, which we propose to designate P8, forms a hydrogen bond with the 2′-phosphate of NADP+ (3.4 Å). A serine in this position is conserved in all bacterial NADP+-dependent enzymes. In EcGDH, the P-loop also contributes a hydrogen bond via Ser240(Oγ), which is within 2.1 Å of the 3′-OH of the ribose.

Figure 4.

Structural basis for NADP+ specificity of EcGDH. (A) Three basic residues and Ser264 are adjacent to the 2′-phosphate of NADP+, and contribute to cofactor specificity in EcGDH. (B) View of the equivalent region in PiGDH, which lacks the basic motif. The P7 acidic residue (Asp241) hydrogen bonds with 2′-OH and 3′-OH of adenine ribose. Adenine binding is dominated by hydrophobic residues in PiGDH (Ile242 and Ile300). The equivalent residues in EcGDH are more polar (Ser264 and Thr323). (C,D) BvGDH binding to NAD+ and NADP+, respectively. The orientation is similar, but the active site is zoomed out in the BvGDH–NADP+ complex to show Lys134 (domain I) interactions with the 2′-phosphate. PDB codes are 4bht [EcGDH (A)], 2yfq [PaGDH (B)], 1hwy [BvGDH (C)], and 1hwz [BvGDH (D)]. In (A) and (B), the cofactor was modelled into the unperturbed active sites of the enzymes, which were crystallized in the absence of reactants. No energy minimization was performed. (C) and (D) are crystal structures of enzyme–cofactor complexes.

Furthermore, the P8 serine, which we have shown is critical for NADP+ recognition in EcGDH, is a tryptophan in PaGDH, and would thus be incapable of partnering the proximal lysine of PaGDH, even if it were close enough to interact with the extra phosphate of NADP+. In the dual-specificity BvGDH, the side chain of the P8 serine (Ser276) interacts with Glu275 as part of a hydrogen bonding network involving the proximal lysine and the 2-OH groups of the ribose when NADH is bound (PDB code 3mw9). The side chain of Ser276 is rotated so that it can hydrogen bond with the extra phosphate group in the NADPH complex (PDB code 1hwz). Without this rotation, it would sterically interfere with the binding (being 1.0 Å from the position of the incoming phosphate). In this structure (PDB code 1hwz), Glu275 has moved away to allow room for the incoming phosphate, such that it can now only hydrogen bond with the 3′-OH. It also recruits the proximal lysine (Lys295), bringing it 0.5 Å closer to the ribose, thus enabling electrostatic interactions with the phosphate of NADPH.

In NAD+-dependent CsGDH, the P7 residue is Gly262, and the loop connecting βH/βI is much closer to the cofactor (relative to EcGDH), so that Pro263 stacks against the adenine ring (not shown). The P7 residue in BvGDH, which has dual specificity, plays two distinct roles in complexes with cofactor (Fig. 4). In the BvGDH–NADP+ complex, Glu275 forms a salt bridge with Lys295, which, in turn, binds to the 2′-phosphate of adenine ribose. In the BvGDH–NAD+ complex, alternative side chain conformations enable the formation of hydrogen bonds from Glu275 to the 2′-OH and 3′-OH of adenine ribose.

Mutagenesis and kinetic studies

Modelling of NADP+ at the active site of EcGDH suggested that a trio of positively charged residues may stabilize the negative charge of the 2′-phosphate. Therefore, these residues were subjected to single site and multisite mutagenesis to glutamines, and assayed for enzyme activity. With mutation to glutamine, the hydrogen bonding potential of the side chain is retained, allowing a critical assessment of the contribution of electrostatic complementarity to NADP+ recognition. The data reveal that the single site mutants are compromised in their catalytic ability (Table 1). As would be expected from disruption of charge complementarity, the apparent Km values are increased to varying extents in the single mutants, but the corresponding kcat is not adversely affected. The most severe effect is in mutant K286Q, which has a seven-fold increase in Km. This is consistent with the structural data, as Lys286 is proximal to the 2′-phosphate and preconfigured to bind the cofactor. The triple mutant (K286Q/R289Q/R292Q) shows a dramatic 50-fold increase in Km. The magnitude of the change highlights the critical role of these basic residues in charge complementarity with the cofactor.

Table 1. Apparent kinetic parameters for NAD+-dependent and NADP+-dependent catalysis by wild-type EcGDH and various mutant EcGDH proteins. Experiments were performed as described in Experimental procedures. Units of Km are μm, and units of of kcat are s−1. Dashes indicate that these parameters were not measured for these enzymes
 Wild typeK286QR289QR292QK286Q R289Q R292QK286Q R289Q R292Q S264LK286Q R289Q R292Q S264L S240A
  1. a

    These are S0.5 values, as these enzymes show an apparent mild negative cooperativity towards NADP+, with Hill coefficients varying from 0.89 to 0.94.

 K m 18.4 ± 0.8a129 ± 3.232.8 ± 1.7a81.9 ± 3.9a929 ± 4219 060 ± 147218 300 ± 1676
 k cat 37.3 ± 0.534.2 ± 0.337.6 ± 0.647.3 ± 0.824.0 ± 0.62.5 ± 0.26.9 ± 0.4
 kcat/Km2.032.65 × 10−11.155.78 × 10−12.58 × 10−21.31 × 10−43.77 × 10−4
 K m 21 680 ± 163324 190 ± 302419 480 ± 297413 740 ± 1826
 k cat 2.2 ± 0.13.9 ± 0.32.3 ± 0.23.4 ± 0.2
 kcat/Km1.02 × 10−41.61 × 10−41.18 × 10−42.48 × 10−4

In addition to the basic residues in the α11–loop–α12 region, the protein sequence alignment shown in Fig. 5 reveals a serine (Ser264 in EcGDH) that is conserved in all NADP+-dependent GDHs, but is absent from all NAD+-dependent GDHs. As previously mentioned, this P8 serine immediately follows the P7 aspartate, and participates in the recognition of the extra phosphate group of NADP+ in concert with the positively charged residue(s) present in the α11–loop–α12. Accordingly, we mutated this residue to leucine in order to evaluate its contribution to NADP+ binding. A nonpolar residue at P8 is observed in NAD+-utilizing enzymes, although the extent of hydrophobicity varies significantly (Fig. 5). None of these changes significantly improves or alters the activity of EcGDH with NAD+; Km remains high and kcat is unaffected with respect to the wild-type enzyme, resulting in catalytic efficiencies that are almost identical to those of the wild-type enzyme with NAD+ (Table 1). Strikingly, the Ser→Leu mutation has a major bearing on catalysis with NADP+, resulting in an almost 200-fold decrease in catalytic efficiency with this coenzyme, on top of the ~ 80-fold decrease effected by the removal of positive charges in the triple mutant. This confirms that the spatial disposition and composition of these residues are specifically adapted to NADP+. Further mutagenesis of Ser240 (P2 residue), which is generally conserved in NADP+-dependent GDHs (Fig. 5), has no additional effects on top of the quadruple mutation (Table 1). Ser240 forms a hydrogen bond with the 3′-OH group of adenine ribose, and would be expected to destabilize binding to both NAD+ and NADP+. Altogether, the kinetic analysis of these mutant enzymes reveals that the positively charged residues in the α11–loop–α12, together with P8 serine, account for the NADP+ specificity in EcGDH.

Figure 5.

Sequence alignment of cofactor-binding segments of bacterial GDHs. Sequences of NADP+-dependent enzymes are above, and sequences of NAD+-dependent enzymes are below. The P1, P6 and P7 sites are annotated below the alignment, and the P8 site is shown above the alignment. Filled triangles indicate the positively charged residues in EcGDH (α-helical hairpin) that contribute to NADP+ recognition.


Model for cofactor preferences

The structural, kinetic and modelling studies reveal a novel role for the acidic P7 residue in EcGDH. In PiGDH and PaGDH, an acidic P7 residue forms hydrogen bonds with the sugar moiety (adenine ribose), but in models of the EcGDH–NADP+ complex, the side chain of Asp263 points toward the α11–loop–α12 region, and forms a backbone hydrogen bond with Val293 (NH). The interaction aids in positioning of the α11–loop–α12 motif, thus orienting positively charged residues (Lys286, Arg289, and Arg292) towards the 2′-phosphate of NADP+. Surprisingly, three positively charged residues are also found in the α11–loop–α12 region of NAD+-dependent clostridial GDH (Fig. 5). The first positively charged residue is conserved in nearly all NAD+-dependent GDHs. In principle, these enzymes appear to have part of the 2′-phosphate recognition site, although they lack the P8 serine. However, the equivalent of Lys286 in EcGDH is Arg286 in CsGDH, which forms a salt bridge with the pyrophosphate group of NAD+. The other two basic residues in CsGDH (Arg290 and Lys292) face in completely the opposite direction relative to NAD+, and it is difficult to imagine a rearrangement that would guide them towards the adenine ribose (not shown). The role of positive charges in this region is therefore influenced by the 3D position and orientation of the α11–loop–α12 region (Fig. 6) relative to the adenine ribose.

Figure 6.

Conformational heterogeneity in the α11–loop–α12 motif. The BvGDH–NADP+ complex (PDB code 1hwz) is used as the reference structure to highlight the position of the cofactor. Complementarity of α9 and α10 in various structures indicates the high degree of convergence. However, the α11–loop–α12 motif is poorly conserved among the various NAD+-dependent and NADP+-dependent enzymes.

The equivalent helix–loop–helix motif in the structure of PiGDH has been identified previously as a possible determinant of nucleotide specificity [25]. In this study, a six-residue deletion in NADP+-dependent versus NAD+-dependent GDHs was correlated with spatial proximity to an adjacent α-helix (α9 in PiGDH; α12 in EcGDH), thus imparting specificity. However, the sequence alignments in this study were restricted to highly related enzymes from the same phylum (Euryarchaeota). Furthermore, one of the reportedly NADP+-dependent enzymes in their alignment, GDH from Thermotoga maritima, has the six-residue deletion and nevertheless prefers NAD+ as a cofactor over NADP+ [10]. In principle, our findings are consistent with the work of Bhuiya et al. in identifying a critical role for the nucleotide-proximal helix–loop–helix motif. However, the distinguishing feature of our model is that, in NAD+-specific enzymes, the α11–loop–α12 motif is improperly oriented relative to the adenine ribose to contribute positively charged side chains for 2′-phosphate recognition (Fig. 7). The known structures of NAD+-dependent enzymes (PiGDH, PaGDH, and CsGDH) reveal that cofactor recognition is dominated by hydrophobic interactions and hydrogen bonds with adenine ribose. The tight hydrophobic contacts from the α11–loop–α12 hairpin to the adenine in the NAD+-only enzymes may also serve to sterically preclude binding to NADP+.

Figure 7.

Model for the roles of the P7 residue in cofactor specificity. Left: cartoon depiction of the P7 aspartate in EcGDH, which orients the α11–loop–α12 hairpin toward the 2′-phosphate via a backbone hydrogen bond with Val293 (NH). The critical Ser264 (P8) forms a hydrogen bond with the 2′-phosphate. Right: the P7 aspartate in PiGDH forms hydrogen bonds with 2′-OH and 3′-OH of adenine ribose (NAD+). The P8 residue (Ile242) packs against the adenine ring of NAD+.

The P7 glutamate (Glu275) of BvGDH is bifunctional, conferring the dual specificity of the enzyme, as noted above. The composition of PaGDH (P7 glutamate) and the conformation of the α11–loop–α12 hairpin (Fig. 6, blue) resemble those of BvGDH (Fig. 6, green). However, in modelling of NADP+ in the active site, the proximal lysine (Lys268) is further away from the 2′-phosphate (4.5 Å) in PaGDH [19] than in EcGDH, in which the proximal lysine is significantly closer to the 2′-phosphate. PaGDH also lacks the basic residue in domain I (equivalent of Lys136 of EcGDH) that may contribute to NADP+ stabilization in the fully closed and catalytically competent conformation. Similarly, the three basic residues of NAD+-specific CsGDH in the α11–loop–α12 hairpin (Fig. 5) are improperly oriented for interactions with the 2′-phosphate.

In summary, cofactor specificity appears to be a function of the arrangement of secondary structural elements, rather than the amino acid sequence as such. Distinct structural roles for apparently conserved residues, such as an acidic residue (Asp/Glu) at the P7 position, suggest that caution should be used in interpreting the functional data from mutagenesis and kinetic studies without a structural context, and argue strongly against the tendency to infer coenzyme specificity from primary sequence data alone.

Experimental procedures

Crystallization of EcGDH

EcGDH was expressed and purified as previously described [26]. With a 10 mg ml−1 EcGDH solution in 20 mm Tris/HCl (pH 7.6), crystals were obtained by the hanging drop vapour diffusion method at 18 °C with 15–25% poly(ethylene glycol) 3350, Tris/Hepes (pH 7–8) and 0.2 m NaCl as a precipitant solution. Crystals (0.15 × 0.1 × 0.1 mm) grew to full size in 2 days. Prior to X-ray data collection, crystals were cryoprotected in a harvest solution with 25% glycerol, and immediately flash cooled in liquid nitrogen. X-ray diffraction data were collected on the NECAT 24ID-C beamline at the Advanced Photon Source, Argonne, USA. Despite numerous strategies to obtain EcGDH in complex with NADP+ and substrate–product complexes, including cocrystallization and soaking experiments, the only crystals obtained were of EcGDH in its free state.

Data collection and refinement

Diffraction data were integrated with hkl2000 [27] and scaled with scala [28] from the ccp4 program suite [29]. EcGDH crystals belong to the orthorhombic space group P212121, with the biological hexamer as the asymmetric unit. The structure was solved by molecular replacement with CsGDH (PDB code 1bgv; 53% sequence identity) as a search model. Molecular replacement involved division of the polypeptide into the substrate and cofactor domains, and a search for each independent domain against the diffraction data with phaser [30]. Model building and refinement were performed with coot [31] and phenix [32], making use of TLS groups selected on the TLSMD server [33]. Noncrystallographic symmetry restraints were used during all cycles of refinement, making use of the new torsion angle NCS algorithm implemented in phenix. Iterative cycles of model building and refinement were performed with coot and phenix, respectively, until convergence was reached. Structure analysis and model validation were performed with procheck from ccp4 and molprobity [29, 34]. Data collection and refinement statistics are shown in Tables 2 and 3. Coordinates and structure factors have been deposited in the PDB under the accession number 4bht.

Table 2. Crystallization and data collection. Values in parentheses correspond to the statistics for the highest-resolution shell
Protein10 mg·ml−1
20 mm Tris/HCl, pH 7.6
Reservoir15–25% poly(ethylene glycol) 3350
0.1 m Hepes/Tris, pH 7–8
0.2 m NaCl
Data collection
Beamline24ID-C (NECAT, APS)
DetectorADSC Q315
Space groupP212121
Unit cell lengths (Å)101.0, 152.9, 169.4
Asymmetric unitSix molecules
Wavelength (Å)0.9792
Resolution (Å)50.0–2.50 (2.59–2.50)
Total reflections542 372
Unique reflections91 251
Redundancy5.9 (4.3)
Completeness (%)99.8 (100.0)
Rmerge (%)9.7 (49.7)
I/σ all data18.1 (3.0)
Table 3. Refinement statistics
Model (chain/residues)
D6–315, 319–447
Ramachandran map (%)
Rwork/Rfree (%)15.9/22.7
High-resolution shell20.3/30.7
Nonhydrogen atoms
Protein20 371
Poly(ethylene glycol)1
Average isotropic B-factor (Å2)40.6
Average TLS B-factor22.5
Bond lengths (Å)0.015
Bond angles (°)1.592

Superpositions of structures and cofactor/substrate modelling

Crystallization of EcGDH in complex with NADP+ or NADPH was attempted in order to provide insights into determinants of cofactor specificity. However, all experiments (soaking and cocrystallization, with or without substrate–product) failed to yield a crystalline complex. As an alternative, superpositions and modelling of the cofactor were performed with the secondary structure matching algorithm in pdbset, as implemented in ccp4 [29]. The structure of BvGDH in complex with NADP+ (1hwz) was used to generate the approximate position of the cofactor in EcGDH. Domain II, comprising residues 204–368 of EcGDH, was superimposed on the equivalent region of the mammalian enzyme. The final three α-helices of the enzymes and the antenna of BvGDH were excluded from calculations. The rmsd values for the alignments were ~ 1Å between bacterial enzymes, and ~ 1.5–1.7Å upon alignment of bacterial and mammalian enzymes. As an example, equivalent regions of domain II in EcGDH-F superimpose on BvGDH with an rmsd of 1.72 Å for 145 aligned Cα atoms, and the rmsd for 185 equivalent Cα atoms of EcGDH-A (6–200) and glutamate-bound CsGDH (PDB code 1bgv; residues 6–203) is 0.95 Å. The value increases to 2.3 Å when both domains are superimposed simultaneously, and such an alignment would ignore the relative domain movements observed in the various subunits and between the different enzymes. Therefore, structural analysis was restricted to pairwise domain superpositions, and no further energy minimizations or rigid-body docking adjustments were performed.

Mutagenesis and kinetic analysis

Site-directed mutagenesis was applied to wild-type EcGDH in pTac-85 [26], with the primers listed in Table 4 (synthesized by Eurofins MWG Operon, Ebersberg, Germany), to generate the various mutant constructs. Sequences were verified by automated DNA sequencing (carried out by GATC Biotech, Konstanz, Germany). Expression and purification were performed as previously described [26], and protein was quantified with the BioRad dye-binding assay (BioRad Laboratories, Hercules, CA, USA), with BSA as the standard. EcGDH and mutant variants were diluted, when necessary, in 0.1 m potassium phosphate buffer (pH 8.0), containing 0.25 mg·ml−1 BSA to maintain enzyme stability. Typically, 0.2 μg of purified enzyme in a 10-μL volume was added to a reaction mixture consisting of 0.1 m potassium phosphate buffer (pH 8.0), containing l-glutamate and NADP+, bringing the final concentration of glutamate to 0.1 m. For reactions involving the unfavoured coenzyme, up to 6.3 μg of enzyme was used per reaction. Reduction of NAD(P)+ to NAD(P)H was monitored at 340 nm, and reaction rates were calculated with an extinction coefficient of 6.22 mm−1·cm−1. The apparent kinetic parameters (Table 1) were determined from initial rate measurements carried out over the NADP+ ranges 3–439, 22–439, 4–429, 6–429, 274–1499, 1867–14 938 and 2563–20 502 μm, for the wild type and the K286Q, R289Q, R292Q, triple, quadruple and quintuple mutants, respectively. The NAD+ range for the wild-type and the triple mutant was 3696–29 568 μm, and those for the quadruple and quintuple mutants were 4153–33 222 and 4062–32 493 μm, respectively. Nonlinear regression analysis was carried out with sigmaplot version 8.02.

Table 4. Oligonucleotide primers used for mutagenesis. Alterations to codons (bold) are underlined
Oligonucleotide nameSequence (5′- to 3′)


This work was supported by Science Foundation Ireland grant number 07/IN.1/B975 to A. R. Khan, and Fellowship grant 05/FE1/B857 to P. C. Engel. We would like to thank the staff of NE-CAT at the Advanced Photon Source, Argonne, Illinois for their help in collection of X-ray diffraction data. This work is based upon research conducted at the Advanced Photon Source on the Northeastern Collaborative Access Team beamlines, which are supported by grants from the National Center for Research Resources (5P41RR015301-10) and the National Institute of General Medical Sciences (8 P41 GM103403-10) from the National Institutes of Health. Use of the Advanced Photon Source, an Office of Science User Facility operated for the US Department of Energy (DOE) Office of Science by Argonne National Laboratory, was supported by the US DOE under Contract No. DE-AC02-06CH11357.