Research article

Structure and kinetic characterization of human sperm-specific glyceraldehyde-3-phosphate dehydrogenase, GAPDS

Apirat Chaikuad, Naeem Shafqat, Ruby Al-Mokhtar, Gus Cameron, Anthony R. Clarke, R. Leo Brady, Udo Oppermann, Jan Frayne, Wyatt W. Yue


hGAPDS (human sperm-specific glyceraldehyde-3-phosphate dehydrogenase) is a glycolytic enzyme essential for the survival of spermatozoa, and constitutes a potential target for non-hormonal contraception. However, enzyme characterization of GAPDS has been hampered by the difficulty in producing soluble recombinant protein. In the present study, we have overexpressed in Escherichia coli a highly soluble form of hGAPDS truncated at the N-terminus (hGAPDSΔN), and crystallized the homotetrameric enzyme in two ligand complexes. The hGAPDSΔN–NAD+–phosphate structure maps the two anion-recognition sites within the catalytic pocket that correspond to the conserved Ps site and the newly recognized Pi site identified in other organisms. The hGAPDSΔN–NAD+–glycerol structure shows serendipitous binding of glycerol at the Ps and new Pi sites, demonstrating the propensity of these anion-recognition sites to bind non-physiologically relevant ligands. A comparison of kinetic profiles between hGAPDSΔN and its somatic equivalent reveals a 3-fold increase in catalytic efficiency for hGAPDSΔN. This may be attributable to subtle amino acid substitutions peripheral to the active centre that influence the charge properties and protonation states of catalytic residues. Our data therefore elucidate structural and kinetic features of hGAPDS that might provide insightful information towards inhibitor development.

  • glyceraldehyde-3-phosphate dehydrogenase (GAPDH)
  • sperm-specific glyceraldehyde-3-phosphate dehydrogenase (GAPDS)


The glycolytic enzyme GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (EC catalyses the NAD+-dependent oxidative phosphorylation of D-G3P (D-glyceraldehyde 3-phosphate) to form DPG (1,3-diphosphoglycerate). GAPDH is widely recognized as a drug target due to its central role in glycolysis, and in non-glycolytic processes such as nuclear RNA transport, DNA replication and repair, membrane fusion and apoptosis [1]. For example, GAPDHs from trypanosomatid species, which depend heavily on glycolysis for ATP production, are promising drug targets to combat tropical parasitic diseases [24]. In mammals, accumulation of GAPDH in the nucleus is linked to cell death [5], hence its inhibition might provide an anti-apoptotic application. In several neurodegenerative diseases associated with the expression of mutant polyglutamine proteins, e.g. Huntington's disease, GAPDH has been shown to bind the mutant proteins [6]. GAPDH also reportedly binds APP (amyloid precursor protein), implying a potential role in the development of Alzheimer's disease [7].

Mammals possess not only the somatic GAPDH isoenzyme present in all tissues except for spermatozoa, but also a sperm-specific isoenzyme [GAPDS (sperm-specific GAPDH)] expressed at late stages of spermatogenesis [8,9]. GAPDS shows significant sequence divergence from its somatic counterpart. For example, hGAPDS (human GAPDS) shares only 56% sequence identity with its somatic counterpart [hGAPDH (human GAPDH)] (see Supplementary Figure S1A at, and contains a unique N-terminal 72-residue polyproline extension [8] necessary for tight association with the cytoskeletal fibrous sheath of the spermatozoa flagellum [10]. The co-ordinated movements of sperm require substantial amounts of ATP provided by glycolysis, as is evident from the compartmentalization of glycolytic enzymes in spermatogenic cells [11]. A Gapds−/− mouse suffered defects in sperm motility, a crucial determinant for male fertility [12], revealing the potential of targeting GAPDS as a novel contraceptive method alternative to the hormone-based approaches. Chlorinated anti-fertility compounds have been shown to inhibit sperm motility by selectively targeting the sperm-specific and not the somatic enzyme [1315], albeit with side effects from in vivo trials [16,17].

To date, the wealth of structural data from archaeal [18], bacterial [19,20] and eukaryotic GAPDHs [2123] has revealed a common two-domain architecture and homotetramer assembly. Uniform cofactor and D-G3P substrate-binding pockets featuring two anion-recognition sites (traditionally known as Ps and Pi sites, for the binding of substrate and inorganic phosphate respectively) have also been well documented [20,24]. Combined with extensive kinetic studies [25,26], the structures are consistent with a well-accepted flip–flop reaction mechanism. This begins with a nucleophilic attack by a cysteine thiol on the carbonyl of D-G3P to form a thiohemiacetal intermediate followed by hydride transfer from the thiohemiacetal to NAD+. The resulting thioacyl-enzyme is then attacked by an inorganic phosphate to form DPG. In this model, the C-3 phosphate of substrate D-G3P is expected to bind to the Pi site during the formation of the hemiacetal intermediate, and flips to the Ps site before hydride transfer, to vacate the Pi site for an incoming inorganic phosphate [20,27].

Compared with somatic GAPDHs which have been characterized structurally and kinetically (reviewed in [28]), GAPDS has proven intractable to study largely due to the insolubility of native and recombinant proteins. Recently, the structure of a heterotetrameric complex of rat GAPDS–Escherichia coli GAPDH has been reported to moderate resolution [29], and provided the first view of a mammalian GAPDS, albeit within an unnatural oligomer. In the present study, we have overexpressed hGAPDS truncated at the N-terminus (hGAPDSΔN) to gain insights into its structural and kinetic properties. The hGAPDSΔN homotetrameric crystal structures, determined in complexes with NAD+ and phosphate (hGAPDSΔN–NAD+–PO4) and with NAD+ and glycerol (hGAPDSΔN–NAD+–Gol), provide the first high-resolution view of the two anion-recognition sites in a human enzyme. Combined with a comparative kinetic analysis, these structures provide a rigorous description of GAPDS that may assist inhibitor design.


Protein expression and purification

A DNA fragment encoding hGAPDS (hGAPDSΔN, amino acids 69–398; GenBank® accession number NP_055179.1) was subcloned into the pNIC-CTHF vector (GenBank® accession number EF199844.1) incorporating a C-terminal TEV (tobacco etch virus)-cleavable His6 tag. The corresponding plasmid was transformed into E. coli BL21(DE3)-R3 cells, cultured in 1 litre of Terrific broth at 37 °C, and induced with 0.5 mM IPTG (isopropyl β-D-thiogalactopyranoside) overnight at 18 °C. Cells were homogenized in lysis buffer [50 mM Hepes (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol and 20 mM imidazole] and insoluble material was removed by centrifugation at 43000 gfor 60 min. The supernatant was purified by affinity (Ni-Sepharose) and size-exclusion (Superdex S75) chromatography. Purified protein was treated with His6-tagged TEV protease overnight at 4 °C, and then passed over 1 ml of Ni-Sepharose resin. The TEV-treated protein, containing the remnant vector-encoded sequence AENLYFQ at the C-terminus, was concentrated to 12 mg/ml and stored in storage buffer {10 mM Hepes (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 0.5 mM TCEP [tris-(2-carboxyethyl)phosphine]} at −80 °C.

Crystallization and data collection

hGAPDSΔN was crystallized at 4 °C in 150-nl drops using the sitting-drop vapour-diffusion method. Viable crystals of the hGAPDSΔN–NAD+–PO4 complex were obtained by mixing protein pre-incubated with 5 mM NAD+(Sigma) with the reservoir solution {20% PEG [poly(ethylene glycol)] 3350, 0.2 M sodium/potassium phosphate and 10% (v/v) ethylene glycol} in a 1:2 volume ratio. Crystals of the hGAPDSΔN–NAD+–Gol complex were fortuitously obtained from mixing protein alone with reservoir solution [20% PEG 3350, 0.2 M Na2SO4, 10% (v/v) ethylene glycol and 0.1 M Bis-Tris propane (pH 6.5)] in a 1:1 volume ratio. Attempts to crystallize hGAPDSΔN in the apo form have not been successful to date. All crystals were cryoprotected with reservoir solution supplemented with 25% (v/v) glycerol, and flash-cooled in liquid nitrogen. Diffraction data were collected on beamline I03 at the Diamond Light Source (hGAPDSΔN–NAD+–PO4) and in-house on a Rigaku FR-E SuperBright™ source (hGAPDSΔN–NAD+–Gol). Data were processed with MOSFLM and SCALA from the CCP4 Suite [30].

Structure determination

Both crystal forms belong to the C-centred monoclinic space group C2 (Table 1). Structures were solved by molecular replacement with PHASER [31] using the rat GAPDS co-ordinates [29] as a search model. Density modification and NCS averaging were performed with DM [32], and the improved phases were used for automated model building with Buccaneer [33]. Iterative cycles of manual model building using COOT [34] alternated with refinement in REFMAC5 [35] were performed. At the last refinement step, TLS (Translation Libration Screw-motion)-restrained refinement was applied using TLS tensor groups determined by the TLSMD server [36]. Data collection and refinement statistics are summarized in Table 1.

View this table:
Table 1 Data collection and refinement statistics for hGAPDSΔN structures

Values in parentheses are for the highest resolution shells. P/L/O denotes protein, active-site ligands and other solvent molecules respectively.

Steady-state enzyme kinetics

Recombinant hGAPDSΔN was dialysed against storage buffer containing activated charcoal to remove bound cofactor, enabling accurate measurement of protein concentration. Examination of the protein spectra showed no NAD+ absorbance at 260 nm, showing that the NAD+ had been removed successfully. Kinetic assays were based on the forward reaction, in which phosphorylation of D-G3P was catalysed in the presence of NAD+ and inorganic phosphate [37]. All reactions were performed in 10 mM sodium pyrophosphate, 20 mM sodium phosphate (pH 8.5), 3 μM dithiothreitol and 10 mM sodium arsenate, using 500 ng/ml enzyme. The Km for D-G3P was measured using a concentration range 0.3–4.0 mM at a fixed NAD+ concentration (0.5 mM). The Km for NAD+ was determined using a concentration range 0.02–1 mM at a fixed D-G3P concentration (1.5 mM). All reagents apart from hGAPDSΔN were dispensed in a 96-well plate, and pre-incubated at 25 °C for 2 min. The reaction was initiated by the addition of hGAPDSΔN. Steady-state reactions were recorded at 340 nm for 3 min with readings taken every 2 s, using a VersaMax® microplate reader (Molecular Devices). The tangent at zero time was well defined and accurately defined the initial velocity. Values for Km, Vmax and kcat were calculated using GraphPad Prism by non-linear regression analysis. The same process and parallel kinetic assays were also performed with hGAPDH from erythrocytes (Sigma).


Recombinant production and structure determination of hGAPDSΔN

To identify the optimal domain boundary of hGAPDS for structural studies, we expressed a series of full-length and truncated constructs in E. coli as N/C-terminally His6-tagged proteins (see Supplementary Figure S1B). We observed that the full-length protein was predominantly insoluble when recombinantly expressed. High levels of soluble expression were found in several N-terminally truncated constructs harbouring the His6 tag at the C-terminus, whereas fusing the affinity tag to the N-terminus of the equivalent constructs render them less soluble. We subsequently purified the C-terminal His6-tagged construct encompassing amino acids 69–407 (referred hereafter as hGAPDSΔN) to homogeneity (see Supplementary Figure S1C) and its crystal structure was determined in two complexes: one bound with NAD+ and inorganic phosphate (hGAPDSΔN–NAD+–PO4) at 1.72 Å (1 Å=0.1 nm) resolution, and the other bound with NAD+ and glycerol (hGAPDSΔN–NAD+–Gol) at 2.15 Å resolution (Table 1).

In both hGAPDSΔN complexes, the asymmetric unit comprises two subunits O and P [Cα-RMSD (root mean square deviation) between subunits ~0.1–0.2 Å] (Figure 1A), which generate another subunit pair Q and R by a two-fold symmetry operation (lying approximately on the RQ plane, using previous nomenclature [38]), resulting in a homotetramer with 222 point symmetry (Figure 1B). Similar to other GAPDHs, each subunit consists of an N-terminal nucleotide-binding domain (N-domain; amino acids 69–220) and a C-terminal catalytic domain (C-domain; amino acids 223–407). The N-domain consists of an eight-stranded β-core flanked by helices on both sides in a classical Rossmann fold. The C-domain is built from a five-stranded β-sheet core with helices on one side, and harbours the active-site catalytic residues Cys224 and His251, as well as a conserved S-shaped loop (S-loop; amino acids 252–278) that forms the tetrameric core. The S-loop constitutes the main site of intersubunit contacts and has previously been suggested to determine NAD/NADP cofactor specificity [39]. The hGAPDSΔN structures are similar to other GAPDHs, e.g. human liver/placental (hGAPDH) [22], Geobacillus stearothermophilus (formerly Bacillus stearothermophilus) (GsGAPDH) [20,27,40], E. coli (EcGAPDH) [19] and Leishmania mexicana (LmGAPDH) [24] enzymes, as is evident from Cα-RMSD values ranging from 0.7 to 2.3 Å, albeit sharing only 20–67% sequence identity.

Figure 1 Structural overview, surface electrostatic potentials and selectivity cleft of hGAPDSΔN

(A) The asymmetric unit of hGAPDSΔN–NAD+–PO4 complex comprises subunits O and P, each composed of the N- (green) and C-domains (cyan) (inset). (B) The hGAPDSΔN homotetramer is assembled from the OP dimer and the QR dimer generated by a two-fold crystallographic symmetry operator (grey plane) lying on the RQ plane. (C) Surface electrostatic potentials of hGAPDSΔN calculated using the APBS program [46] reveal three prominently charged regions: the negatively charged S-loop (blue), positively charged 350–372 region (red) and mixed-charged 110–130 region (orange). The inset shows calculated electrostatic potentials for somatic hGAPDH (PDB code 1U8F). (D) Superposition of the hGAPDSΔN selectivity cleft in closed conformation (yellow backbone) with two structures of somatic hGAPDH showing the closed (blue, PDB code 1U8F) and open (orange, PDB code 1ZNQ) conformations. The closed conformation in hGAPDS is stabilized by water-mediated hydrogen bonds (blue broken lines, water in yellow sphere), similar to that seen in hGAPDH (green broken lines, water in blue spheres). (E) Space-filling representation of the hGAPDSΔN selectivity cleft (yellow) reveals a much reduced cavity than LmGAPDH (magenta, PDB code 1I32) due to the bulky Phe108, Pro263 and Arg265. The trypanosomal GAPDH inhibitor NMDBA (orange) is modelled into the hGAPDS structure based on the binding position of the adenine ring of the NAD+ cofactor and illustrates several clashes between the ligand and the protein.

The hGAPDSΔN homotetramer

The quaternary arrangement and intersubunit interfaces of hGAPDSΔN are consistent with the functional oligomeric state proposed for other GAPDHs, with subtle differences in two features, namely the electrostatic surface potential and intersubunit selectivity cleft. The hGAPDSΔN structures reveal three prominently charged patches at the dimeric and tetrameric interfaces, contributed by the positively charged S-loop (amino acids 252–278), the negatively charged 350–372 region and the mixed-charged 106–130 region (Figure 1C). The widespread charge distribution in hGAPDSΔN is more prominent than somatic hGAPDH (PDB codes 1U8F and 1ZNQ) owing to several non-conserved charged/polar residues (e.g. Glu112, Lys119, Tyr127, Lys128, Ser252, Tyr253, Arg265, His275, Asp351, Glu352 and Lys370) (see Supplementary Figure S1A), and suggests that hGAPDSΔN tetramerization is facilitated via intersubunit electrostatic complementarity. For example, at the dimeric interface, the positively charged S-loop of subunit P protrudes into a negatively charged groove from the 350–372 region of subunit O, creating a number of salt bridges (see Supplementary Figure S2 at At the tetrameric interface, close to the cofactor-binding site, the mixed-charged 106–130 region of subunit O packs against the positively charged S-loop of subunit R and the negatively charged 350–372 region of subunit Q.

A second differentiating feature between hGAPDH and hGAPDSΔN is the selectivity cleft adjacent to the adenosine pocket of the NAD+-binding site (Figure 1D). The selectivity cleft exhibits cross-species variability in size, and has been exploited for small-molecule inhibitor development as exemplified by the selective binding of an adenosine derivative NMDBA [N6-(1-naphthalenemethyl)-2′-deoxy-2′-(3,5-dimethoxybenzamido)adenosine] to the selectivity cleft of the trypanosomatid LmGAPDH (PDB code 1I32) [24]. In our hGAPDSΔN structures, loop β2–α3 residues (Phe108, Ile109 and Asp110) from one subunit (e.g. O) form one side of the selectivity cleft, whereas S-loop residues (Pro263, Ser264, Arg265 and Lys266) of another subunit (e.g. R) form the other side (Figure 1D). These residues are conserved between hGAPDH and hGAPDS, with the exception of Arg265 in hGAPDS (glycine in hGAPDH) (see Supplementary Figure S1A). This arginine residue confers reduced conformational flexibility on to the S-loop compared with the more adaptable glycine residue in hGAPDH. As a result, while the hGAPDH S-loop adopts two flexible conformations (resulting in ‘open’ and ‘closed’ states of the cleft; PDB codes 1U8F and 1ZNQ) proposed to be inherent to the somatic isoenzyme [22,23], only the ‘closed’ state is observed in our hGAPDSΔN selectivity cleft. Furthermore, the Arg265 side chain forms electrostatic interactions and water-mediated hydrogen bonds with Asp110 on the opposite side of the cleft, thereby constricting the cleft cavity to ~4 Å in width, compared with ~5 Å in hGAPDH [23] and 7–8 Å in LmGAPDH [2,3]. Modelling the NMDBA ligand on to the NAD+ molecule in our structure reveals severe steric clashes with Pro263 from the S-loop and Phe108 from the β2–α3 loop (Figure 1E). This, together with a much reduced selectivity cleft cavity compared with that of LmGAPDH, might prevent the binding of NMDBA to hGAPDSΔN.

Conservation of anion-binding sites in the hGAPDSΔN–NAD+–PO4 complex

Our hGAPDSΔN complexes reveal an NAD+ molecule bound in each subunit (Figure 2A), adopting essentially identical conformation and interactions as the rat GAPDS and somatic hGAPDH structures. Additional electron density was observed in the substrate-binding pocket adjacent to the cofactor-binding site (Figure 2A) and was refined as two bound phosphate ions per subunit, presumably originating from the crystallization buffer. The positions of the two phosphate ions are identical in both subunits, and map two anion-binding sites in hGAPDS that are commonly observed in other GAPDHs (Figure 2B). The first phosphate site, known previously as the Ps site, is adjacent to the nicotinamide and ribose moieties of NAD+. The bound phosphate interacts with residues enclosing the Ps site (Thr254, Thr256 and Arg306) as well as the nicotinamide ribose hydroxy group of NAD+ (Figure 2C, left-hand panel). The second phosphate site is close to the catalytic residue Cys224 (closest phosphate oxygen is ~3.3 Å from the cysteine thiol) (Figure 2C, right-hand panel) and corresponds to the ‘new Pi site’ observed previously in the LmGAPDH (PDB code 1GYP) and EcGAPDH structures (PDB code 1DC4) [19,24]. This is at a distance of 2.5–3.0 Å from the ‘original Pi site’ reported for the GsGAPDH structure (PDB code 1GD1) [20] which is further away from Cys224.

Figure 2 Phosphate-binding sites in hGAPDSΔN

(A) Refined 2FoFc map (contoured at 1.0 σ) for NAD+ and phosphate ions in the hGAPDSΔN–NAD+–PO4 complex. (B) Superposition of phosphate-bound LmGAPDH (semi-transparent ‘lm’, PDB code 1GYP) and sulfate-bound GsGAPDH (semi-transparent ‘bs’, PDB code 1GD1) on to hGAPDSΔN–NAD+–PO4 subunit O (grey) maps the two bound phosphates in hGAPDSΔN (grey sticks) at the Ps site, and at the new Pi site. The positions of the original and new Pi sites identified from various GAPDH structures are shown. (C) Interactions of phosphates at the Ps (left) and new Pi (right) sites. Note that at the new Pi site, the closest phosphate oxygen atom is 3.3 Å away from the Cys244 thiol group. (D) Two conformations of the active-site segment in hGAPDSΔN–NAD+–PO4 complex: an ‘in’ conformation in subunit O (left), and mixed ‘in’ (semi-transparent)/‘out’ (grey) conformations in subunit P (right).

The two disparate Pi locations observed in various GAPDH structures are proposed to be commensurate with the two different conformations of a nearby β-loop–α-segment (referred to hereafter as the active-site segment), namely an ‘out’ conformation for the ‘original Pi site’ and an ‘in’ conformation for the ‘new Pi site’ (Figure 2B) [27]. In the hGAPDSΔN–NAD+–PO4 structure, the active-site segment (amino acids 280–292) adopts the ‘in’ conformation for subunit O, and a mixed ‘in’/‘out’ conformation (50% occupancy each) for subunit P (Figure 2D). The two conformers differ structurally in the orientation of the conserved residues Thr283 and Gly284, which are displaced from the ‘in’ (closer to the second phosphate) to the ‘out’ (away from the second phosphate) conformations by ~2.0–3.8 Å (Cα-RMSD). Interestingly, the second phosphate from both subunits occupy the ‘new Pi site’ (despite the presence of the ‘out’-conformer in subunit P), and are anchored by Ser223, Thr225 and His251 (Figure 2C, right-hand panel). The ‘in’-conformer of subunits O and P further contributes hydrogen bonds to the second phosphate ion via the main-chain Thr283 and Gly284 atoms (~2.5–2.7 Å), which are not provided in the subunit P ‘out’ conformer as the two residues are more distant from the second phosphate (3.7 and 5.7 Å respectively) (Figure 2D, compare left- and right-hand panels).

Binding of glycerol in the anion-recognition sites

Structural characterization of an hGAPDSΔN–NAD+ crystal grown in a sulfate-containing condition reveals electron-density peaks at the two anion-binding sites (Figure 3A) that do not fit the shape of sulfate ions often observed to bind in various GAPDH structures. Instead, this density is a good match with glycerol present in our purification and cryoprotectant buffers. As a result, two glycerol molecules were refined into the Ps and ‘new Pi’ sites of each subunit, and we refer to this structure as the hGAPDSΔN–NAD+–Gol complex. Neither of the two glycerol molecules overlaps with the expected location of the substrate D-G3P glyceraldehyde moiety between the two anion-recognition sites (Figure 3A). The interactions between the two glycerol molecules and the enzyme are similar to the phosphate ions from the hGAPDSΔN–NAD+–PO4 structure, albeit less extensive than the phosphate ions as reflected by a degree of structural degeneracy and flexibility in the binding conformation of glycerol. Although in this structure, the active-site segment adopts fully the ‘out’ conformation, the glycerol at the new Pi site is further away from the catalytic cysteine residue (compared with the phosphate in the hGAPDSΔN–NAD+–PO4 complex), enabling it to form direct and water-mediated polar interactions with residues from the active-site segment (Ser282, Thr283 and Gly284) (Figure 3B). The water-mediated hydrogen-bond network bridging the two glycerol molecules is also conserved in the hGAPDSΔN–NAD+–PO4 structure.

Figure 3 Glycerol binding in the hGAPDSΔN–NAD+–Gol complex

(A) Left-hand panel: FoFc difference density in the active site of both subunits (contoured at 3σ) was modelled as glycerol. Right-hand panel: phosphate anions from the hGAPDSΔN–NAD+–PO4 structure (semi-transparent), and two D-G3P molecules (dotted transparent) one bound at the Ps site (‘rs’, from rat GAPDS–NAD+–G3P complex, PDB code 2VYV) and the other at the Pi site (‘gs’, from GsGAPDH–NAD+–G3P complex, PDB code 3CMC) are overlaid. (B) Interactions of the two glycerol molecules (Gol) in subunits O (top) and P (bottom) of hGAPDSΔN. Water molecules are represented as blue spheres, and hydrogen-bond networks are shown as broken lines. Note that the water-mediated hydrogen-bond network bridging the two glycerol molecules is highly conserved among various binding conformations of glycerol.

Comparative kinetic analysis of hGAPDSΔN

The steady-state kinetic properties of hGAPDSΔN were measured and compared with those of somatic hGAPDH (Table 2). The two isoenzymes share similar Km values for D-G3P substrate and rate constants. However, hGAPDSΔN has a 3-fold lower Km for NAD+ (35 μM) than does hGAPDH (100 μM), hence resulting in an approx. 3-fold increased catalytic efficiency (kcat/Km) with regard to NAD+. The steady-state kinetics for hGAPDH in the present study are comparable with those reported previously for hGAPDH purified from brain tissue [28], showing a kcat of ~100 s−1 and Km of 0.07 mM and 0.16 mM for NAD+ and D-G3P respectively.

View this table:
Table 2 Steady-state kinetic properties of hGAPDSΔN and hGAPDH

Calculations of Km for D-G3P were based either on *an assumption of an equal concentration of D- and L-G3P in the racemic DL-G3P, or **an approx. 4% D-G3P as the active aldehyde in the racemic substrate according to [45]; however, the latter is not considered in other GAPDH kinetic analyses.


hGAPDS shares conserved structural features with other GAPDHs

In the present paper, we report the first detailed structural and kinetic characterization of hGAPDS. By fusing a C-terminal affinity tag and truncating the N-terminal polyproline region that hinders correct protein folding as shown in both the present study and a recent study [41], we have obtained high levels of recombinant soluble hGAPDSΔN. The present study significantly extends a recent structural study on rat GAPDS [29], where very low recombinant expression levels and solubility required complexation with native EcGAPDH for crystallization. Cross-species heterotetramers in various stoichiometries have been reported previously between GAPDH and GAPDS subunits [29,42]. Nevertheless, given the proximity of the active-site cleft to the subunit interfaces, it is unlikely that studies of GAPDS in these unnatural heterotetrameric forms provide a realistic assessment of GAPDS structure and activity. Our high-resolution hGAPDSΔN structures confirm homotetrameric formation and show that electrostatic interactions and charge complementarity appear to play a substantial role in tetrameric assembly, more so than in hGAPDH, to facilitate the correct orientation of subunits. Our structures also reveal an effectively closed selectivity cleft in hGAPDS as compared with other GAPDHs, particularly the trypanosomal enzymes, and therefore lend further support to the strategy of designing species-specific inhibitors that fill the trypanosomal selectivity cleft without inhibiting the human enzymes [2,3].

Mapping of anion-recognition sites in hGAPDS

The binding positions of phosphate in our hGAPDSΔN–NAD+–PO4 complex map the two anion-recognition sites in hGAPDS, which are occupied by glycerol in our hGAPDSΔN–NAD+–Gol structure. In addition, the trapping of active-site glycerol was not unprecedented for GAPDHs, since glycerol binding to the Ps site was previously reported in an inhibitor-bound Trypanosoma cruzi GAPDH structure (PDB code 3DMT). It remains to be determined how glycerol, presumably derived from our purification or cryoprotection buffers, out-competes the binding of sulfate ions (closely similar to the physiological phosphates) that are present in high concentration in the crystallization solution. Our findings nevertheless reveal the flexible nature and capacity of the anion-recognition sites to accommodate uncharged, polar and non-physiologically relevant ligands, and provide important information on their mode of interactions.

Previous structural characterization of GAPDH structures has led to the widely discussed Ps and Pi nomenclature and flip–flop mechanism [19,27,40]. In contrast with the Ps site, the true position of the Pi site remains unclear since the identification of a ‘new Pi’ site (at a distance from the ‘original’ Pi site) in a number of GAPDH structures. The divergence in the location and number of Pi site(s) is postulated to correlate with the conformations of the neighbouring ‘active-site segment’ alternating during catalysis, to maintain interactions between the phosphate ion and the conserved threonine and glycine residues from the active-site segment (Thr283 and Gly284 in hGAPDS) [27]. The hGAPDSΔN–NAD+–PO4 structure clearly indicates that a phosphate ion is bound at the new Pi site when the active site segment adopts the fully ‘in’ conformation (presumably mimicking a ternary state of enzyme poised for catalysis) and the partially ‘out’ conformation (presumably allowing flipping of the hemiacetal intermediate), whereas we did not observe phosphate binding at the ‘original’ Pi site. This contrasts with the GsGAPDH ternary complex, known to have both new and original Pi sites competent and partially occupied by two phosphate moieties, with the active-site segment adopting two conformations [27]. It is not clear whether coexistence of two alternative Pi positions portrays a common catalytic property or a species-specific feature [27]. Since this is not the case in hGAPDS, a different mechanism for the human enzymes can be envisaged where the original Pi site may not be present.

Kinetic differences between hGAPDS and hGAPDH may arise from electrostatic properties

Disregarding the unique polyproline N-terminus in hGAPDS, the amino acid differences between hGAPDS and hGAPDH are mostly confined to the protein surface, contributing to differences in their surface electrostatic potentials. The active-site architecture, ligand-binding mode and catalytic residues, however, are highly conserved among the isoenzymes, in part reflected by their kinetic parameters being of the same order of magnitude. Nevertheless, subtle differences above S.D. values are notable, particularly the 3-fold higher catalytic efficiency for hGAPDS compared with hGAPDH. This may be attributable to the effect of amino acid substitutions peripheral to the catalytic centre, often known as ‘second sphere’ residues for enzymes known to have identical active sites, but different enzymatic properties [43,44]. A number of amino acid positions lining the adenine and nicotinamide pockets of the NAD+-binding site (Figure 4) involve point charge changes between hGAPDS and hGAPDH. For example, Tyr173 and Asp311 in hGAPDS are replaced by phenylalanine and asparagine respectively in hGAPDH (Figure 4). These ‘second sphere’ residues, while not perturbing the structural integrity of the active site, could influence the catalytic efficiency of the enzyme by possibly removing a hydrogen bond with NAD+ (Y173F substitution) or by altering the pKa and consequently the protonation state (D311N substitution) of nearby catalytic residues.

Figure 4 Amino acid substitutions peripheral to the active sites between hGAPDSΔN and hGAPDH

(A) Amino acid differences between hGAPDSΔN and hGAPDH are clustered in two regions of the active site: one adjacent to the adenine-binding pocket and the other at the nicotinamide-binding pocket. Residue labels are coloured blue (basic), red (acidic), green (uncharged polar) or grey (non-polar). (B) Moderate differences in electrostatic potentials of the cofactor pocket and catalytic pocket between hGAPDSΔN and hGAPDH are circled.

To conclude, isoform-specific inhibitor development for hGAPDS has been implemented for several decades as a strategy for non-hormonal contraception, although it suffers from a lack of ligand-bound structural information for the human GAPDH isoenzymes. Our data highlight subtle yet significant differences in the structural and kinetic features between hGAPDS and other GAPDH enzymes, and therefore need to be considered for the future design of inhibitors targeting the sperm-specific isoenzyme.


Apirat Chaikuad and Naeem Shafqat performed the structural determination. Ruby Al-Mokhtar, Jan Frayne, Gus Cameron, Anthony Clarke and Leo Brady performed the kinetic experiments and analysed the data. Wyatt Yue, Udo Oppermann, Jan Frayne and Leo Brady supervised the project. Apirat Chaikuad and Wyatt Yue designed the experiments and wrote the paper. All authors read and approved the final paper.


The Structural Genomics Consortium is a registered charity (number 1097737) that receives funds from the Canadian Institutes for Health Research, the Canadian Foundation for Innovation, Genome Canada through the Ontario Genomics Institute, GlaxoSmithKline, Karolinska Institutet, the Knut and Alice Wallenberg Foundation, the Ontario Innovation Trust, the Ontario Ministry for Research and Innovation, Merck & Co., Inc., the Novartis Research Foundation, the Swedish Agency for Innovation Systems, the Swedish Foundation for Strategic Research and the Wellcome Trust. The work was also supported by the NIHR Biomedical Research Unit (to U.O.), and by the Wellcome Trust [grant number 07746] and Biotechnology and Biological Sciences Research Council [grant number BB/F007256] (to R.L.B.).


We thank Georgina Berridge and Rod Chalk for assistance in MS analyses, and staff at the Diamond Light Source for access to synchrotron data collection facilities.


  • Atomic co-ordinates and structural factors have been deposited in the PDB under accession numbers 3H9E (hGAPDSΔN–NAD+–PO4) and 3PFW (hGAPDSΔN–NAD+ –Gol).

Abbreviations: DPG, 1,3-diphosphoglycerate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GAPDS, sperm-specific GAPDH; EcGAPDH, Escherichia coli GAPDH; hGAPDH, human GAPDH; GsGAPDH, Geobacillus stearothermophilus GAPDH; LmGAPDH, Leishmania mexicana GAPDH; G3P, glyceraldehyde 3-phosphate; Gol, glycerol; hGAPDS, human GAPDS; NMDBA, N6-(1-naphthalenemethyl)-2′-deoxy-2′-(3,5-dimethoxybenzamido)adenosine; PEG, poly(ethylene glycol); RMSD, root mean square deviation; TEV, tobacco etch virus; TLS, Translation Libration Screw-motion


View Abstract