Structural analyses to identify selective inhibitors of glyceraldehyde 3-phosphate dehydrogenase-S, a sperm-specific glycolytic enzyme

Abstract

STUDY HYPOTHESIS

Detailed structural comparisons of sperm-specific glyceraldehyde 3-phosphate dehydrogenase, spermatogenic (GAPDHS) and the somatic glyceraldehyde 3-phosphate dehydrogenase (GAPDH) isozyme should facilitate the identification of selective GAPDHS inhibitors for contraceptive development.

STUDY FINDING

This study identified a small-molecule GAPDHS inhibitor with micromolar potency and >10-fold selectivity that exerts the expected inhibitory effects on sperm glycolysis and motility.

WHAT IS KNOWN ALREADY

Glycolytic ATP production is required for sperm motility and male fertility in many mammalian species. Selective inhibition of GAPDHS, one of the glycolytic isozymes with restricted expression during spermatogenesis, is a potential strategy for the development of a non-hormonal contraceptive that directly blocks sperm function.

STUDY DESIGN, SAMPLES/MATERIALS, METHODS

Homology modeling and x-ray crystallography were used to identify structural features that are conserved in GAPDHS orthologs in mouse and human sperm, but distinct from the GAPDH orthologs present in somatic tissues. We identified three binding pockets surrounding the substrate and cofactor in these isozymes and conducted a virtual screen to identify small-molecule compounds predicted to bind more tightly to GAPDHS than to GAPDH. Following the production of recombinant human and mouse GAPDHS, candidate compounds were tested in dose–response enzyme assays to identify inhibitors that blocked the activity of GAPDHS more effectively than GAPDH. The effects of a selective inhibitor on the motility of mouse and human sperm were monitored by computer-assisted sperm analysis, and sperm lactate production was measured to assess inhibition of glycolysis in the target cell.

MAIN RESULTS AND THE ROLE OF CHANCE

Our studies produced the first apoenzyme crystal structures for human and mouse GAPDHS and a 1.73 Å crystal structure for NAD⁺-bound human GAPDHS, facilitating the identification of unique structural features of this sperm isozyme. In dose–response assays T0501_7749 inhibited human GAPDHS with an IC₅₀ of 1.2 μM compared with an IC₅₀ of 38.5 μM for the somatic isozyme. This compound caused significant reductions in mouse sperm lactate production (P= 0.017 for 100 μM T0501_7749 versus control) and in the percentage of motile mouse and human sperm (P values from <0.05 to <0.0001, depending on incubation conditions).

LIMITATIONS, REASONS FOR CAUTION

The chemical properties of T0501_7749, including limited solubility and nonspecific protein binding, are not optimal for drug development.

WIDER IMPLICATIONS OF THE FINDINGS

This study provides proof-of-principle evidence that GAPDHS can be selectively inhibited, causing significant reductions in sperm glycolysis and motility. These results highlight the utility of structure-based drug design and support further exploration of GAPDHS, and perhaps other sperm-specific isozymes in the glycolytic pathway, as contraceptive targets.

LARGE SCALE DATA

None. Coordinates and data files for three GAPDHS crystal structures were deposited in the RCSB Protein Data Bank (http://www.rcsb.org).

STUDY FUNDING AND COMPETING INTEREST(S)

This work was supported by grants from the National Institutes of Health (NIH), USA, including U01 HD060481 and cooperative agreement U54 HD35041 as part of the Specialized Cooperative Centers Program in Reproduction and Infertility Research from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, and TW/HD00627 from the NIH Fogarty International Center. Additional support was provided by subproject CIG-05-109 from CICCR, a program of CONRAD, Eastern Virginia Medical School, USA. There are no conflicts of interest.

Introduction

Extensive tissue-specific gene expression is a hallmark of spermatogenesis, providing a large number of sperm-specific proteins that are potential targets for male contraception (Schultz et al., 2003; Matzuk and Lamb, 2008). The central metabolic pathway of glycolysis presents an extraordinary example of tissue specificity within the male germ line. Mammalian sperm have distinct isozymes at most steps of this pathway, resulting from germ cell-specific expression of at least three genes and from novel alternative splicing events (Eddy et al., 2003; Vemuganti et al., 2007; Ijiri et al., 2013). Enzymes are typically considered good candidates for pharmaceutical intervention and account for >25% of the molecular targets for all known drugs (Nass and Strauss, 2004).

Glyceraldehyde 3-phosphate dehydrogenase, spermatogenic (GAPDHS), in particular, is a promising contraceptive target. This sperm-specific glycolytic isozyme is expressed only in the post-meiotic period of spermatogenesis, replacing the somatic isozyme (GAPDH) in mammalian sperm (Welch et al., 1992, 2000; Bunch et al., 1998). In multiple mammalian species the sperm isozyme shares ∼70% amino acid identity with the somatic isozyme and has a novel proline-rich extension at the N-terminus. This N-terminal extension plays a role in anchoring GAPDHS to the fibrous sheath in the principal piece of the sperm flagellum (Bunch et al., 1998; Krisfalusi et al., 2006), but is not required for enzymatic activity (Elkina et al., 2010; Sexton et al., 2011). Gene targeting studies in mice established that GAPDHS (Miki et al., 2004) and other sperm-specific isozymes in the glycolytic pathway (Odet et al., 2008; Danshina et al., 2010; Nakamura et al., 2013) are essential for sperm motility and male fertility. These studies demonstrate the importance of glycolysis for sperm energy production. Spermatogenesis and sperm production appear normal in mice lacking GAPDHS, providing evidence that inhibition of this isozyme should not impair testicular function (Miki et al., 2004).

Like the somatic isozyme, homotetrameric GAPDHS catalyzes the oxidation and phosphorylation of glyceraldehyde 3-phosphate (GAP) to form 1,3-bisphosphoglycerate, a reaction that requires NAD⁺ and inorganic phosphate. This reaction occurs at an important transition point in glycolysis between the enzymatic steps that consume and generate ATP. Consequently, ATP levels in mouse sperm lacking GAPDHS are ∼10% of wild-type levels immediately after collection in glucose-containing M16 medium, which fuels the ATP-consuming steps of glycolysis (Miki et al., 2004). The requirement for glycolytic ATP production has also been demonstrated for human sperm (Williams and Ford, 2001; Nascimento et al., 2008) and sperm from several other mammalian species (Storey, 2008; Mukai and Travis, 2012).

As a contraceptive target, GAPDHS must be subject to selective inhibition that does not disrupt glycolysis in other tissues. Initial evidence for selective inhibition was provided by early studies of α-chlorohydrin (3-chloro-1,2-propanediol) conducted before GAPDHS was identified as a distinct sperm-specific isozyme (reviewed in Jones, 1978; Jones and Cooper, 1999). S-3-chlorolactaldehyde, the major metabolite of α-chlorohydrin and related compounds, inhibited GAPDHS, sperm glycolysis and motility in a dose-dependent manner. These effects on sperm occurred at concentrations that did not inhibit GAPDH activity in other tissues (Brown-Woodman and White, 1975; Brown-Woodman et al., 1978; Ford and Harrison, 1983). Efforts to develop these agents as contraceptives were abandoned due to toxic effects at high doses, although impurities and racemic mixtures of reactive compounds may have contributed to the observed toxicity (Jones and Cooper, 1999).

To assess the potential for selective inhibition of GAPDHS, we used structural analyses to highlight differences between the sperm and somatic isozymes that are conserved between species. We initially constructed homology models of human and mouse GAPDHS and GAPDH templated on the 2.0 Å crystal structure of Palinurus versicolor GAPDH (PDB 1CRW; Shen et al., 2000). Active site structural features of our models were later confirmed in high resolution crystal structures of human GAPDHS (PDB 3H9E, Chaikuad et al., 2011; PDB 5C7L and PDB 5C7O, this study), mouse GAPDHS (PDB 5C7I, this study) and human GAPDH (PDB 1U8F, Jenkins and Tanner, 2006). We conducted virtual screening to identify small-molecule compounds that are predicted to bind more tightly to GAPDHS than GAPDH. These compounds were then tested in enzymatic assays to compare inhibition of the sperm and somatic enzymes, and in functional assays to determine their effects on sperm metabolism and motility.

Materials and Methods

Reagents

Standard reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) with the exception of glucose, magnesium sulfate heptahydrate, sodium bicarbonate, sodium chloride, sodium isopropyl-β-d-thiogalactopyranoside (IPTG), sodium pyruvate, tris(2-carboxyethyl)phosphine (TCEP) (ThermoFisher Scientific, Waltham, MA, USA); potassium chloride, potassium phosphate (Mallinckrodt Baker, Phillipsburg, NJ, USA); penicillin/streptomycin 100X stock solution containing 10 000 U/ml of penicillin G and 10 mg/ml of streptomycin (Gemini Bio-Products, West Sacramento, CA, USA); cOmplete protease inhibitor cocktail (Roche Life Science, Indianapolis, IN, USA); and PEG 3350 (Emerald Biosystems, Bainbridge Island, WA, USA).

Homology modeling and virtual screening

The 2.0 Å crystal structure of Palinurus versicolor GAPDH (PDB 1CRW; Shen et al., 2000) was used as a template for homology modeling of human and mouse GAPDH and GAPDHS isozymes, and binding sites for GAP and NAD⁺ were determined by reference to structures PDB 1DC4 and 1DC6 (Yun et al., 2000). The proline-rich N-terminal extensions that are specific to the sperm isozymes were not included in these models. Therefore, mouse GAPDHS was modeled without the first 105 amino acids of its sequence, and human GAPDHS without the first 75 amino acids. Homology models were constructed using the modeler module of the Insight II molecular modeling program (Accelrys, San Diego, CA, USA). Final structures were tested for sequence-structure compatibility using the Verify function in the Profiles-3D module. Results were displayed graphically using the PyMOL Molecular Graphics System (Schrödinger, LLC, New York, NY, USA).

To assess the potential for selective inhibition of the sperm isozymes, we conducted virtual screening with the SYBYL 6.9 molecular modeling program (Tripos, St. Louis, MO, USA). The SiteID module was used to identify three solvent-accessible binding pockets surrounding the active site in our homology models. After pre-filtering by size, compounds from the Maybridge small molecule database were docked into each of the binding pockets using the SYBYL FlexX algorithm for the flexible docking of small ligands (Rarey et al., 1995). Protein-ligand interaction scores were computed for each isozyme, and the predicted discrimination between sperm and somatic isozymes was calculated as the difference between interaction scores. Docking was repeated thirty times for the compounds with the largest discrimination scores for the human isozymes to confirm that these compounds consistently docked in the predicted binding pocket.

Expression and purification of recombinant proteins

As in previous studies (Frayne et al., 2009; Elkina et al., 2010; Sexton et al., 2011), we expressed only trace amounts of full-length mouse or human GAPDHS even though we purchased DNA sequences with codon usage optimized for expression in Escherichia coli (GeneArt, Regensburg, Germany). Therefore, we subcloned the optimized sequences and expressed truncated forms of the sperm isozymes lacking their proline-rich N-terminal extensions (tGAPDHS). Several fusion constructs were tested to optimize expression of the sperm isozymes.

The DNA fragment encoding human tGAPDHS (amino acids 76–408) was cloned into the pGEX-4T-1 vector (GE Healthcare Life Sciences, Piscataway, NJ, USA) for expression as a glutathione S-transferase (GST) fusion protein. Recombinant protein was expressed in gapA-deficient E. coli DS112, strain K-12 (Yale Coli Genetic Stock Center, New Haven, CT, USA) to avoid the formation of mixed tetramers that contain bacterial GAPDH (Frayne et al., 2009; Sexton et al., 2011). Cells were grown at 37°C in M9 minimal media supplemented with 1% glycerol, 0.1% casamino acids, 0.4% sodium succinate, 0.001% thiamine, 0.004% methionine, 0.004% tryptophan and 0.002% uracil until absorbance at 600 nm was 0.5–0.6. Expression was induced by 0.1 mM IPTG during overnight culture at 18°C. Cells were harvested by centrifugation and resuspended in phosphate-buffered saline (PBS: 136 mM NaCl, 2.7 mM KCl, 10 mM NaH₂PO₄·7H₂O, 1.7 mM KH₂PO₄; pH 7.4) supplemented with 2 mM dithiothreitol (DTT), 5 mg/ml lysozyme, 5 μg/ml DNase I and cOmplete protease inhibitor cocktail. After incubation for 30 min on ice, the suspension was sonicated with short pulses for 2 min and centrifuged at 30 000g for 1 h at 4°C. The resulting supernatant was loaded onto a glutathione Sepharose 4B (GE Healthcare Life Sciences) column prepared according to the manufacturer's instructions and washed with PBS containing 2 mM DTT. To remove the GST tag, the column was incubated overnight at room temperature with ∼40 units of bovine thrombin/ml bed volume. Cleaved tGAPDHS was eluted, frozen in liquid nitrogen and stored at −70°C.

The DNA fragment encoding mouse tGAPDHS (amino acids 106–438) was subcloned into the pMal vector (New England Biolabs, Ipswich, MA, USA), which incorporates a thrombin-cleavable maltose-binding protein (MBP) tag. Recombinant mouse tGAPDHS was expressed using the same procedure described for human tGAPDHS, except that buffer A (20 mM Tris–HCl, 200 mM NaCl, 10 mM EDTA, pH 7.4) replaced PBS in the cell lysis solution. The clarified supernatant was loaded onto an amylose column (New England Biolabs) equilibrated with buffer A and 5% glycerol, followed by overnight incubation at room temperature with ∼40 units of bovine thrombin/ml bed volume. The eluate from this column, containing recombinant protein and a fraction of the cleaved MBP tag, was dialyzed against 2000 volumes of 20 mM Tris–HCl, 25 mM NaCl, 2 mM β-mercaptoethanol, pH 8.0. Following dialysis, the protein solution was loaded onto a diethylaminoethyl (DEAE)-Sepharose column equilibrated with 20 mM Tris–HCl, 10 mM NaCl, 2 mM DTT, pH 8.0, which retained the MBP tag. Mouse tGAPDHS was eluted, frozen in liquid nitrogen and stored at −70°C.

Mouse somatic GAPDH was expressed as a GST-fusion protein and purified according to the same procedures used for human tGAPDHS. The purity of all recombinant enzyme preparations was confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (Sexton et al., 2011). Human somatic GAPDH, purified from erythrocytes, was purchased from Sigma-Aldrich (Cat. No. G6019).

Crystallization of tGAPDHS

Crystal structures were obtained using resources in the Macromolecular X-Ray Crystallography Core Facility at the University of North Carolina School of Medicine (Chapel Hill, NC, USA) as well as the SER-CAT beamline at the Advanced Photon Source of the Argonne National Laboratory (Lemont, IL, USA). The PACT Suite screening kit (Qiagen, Valencia, CA, USA) was used to assess multiple conditions for crystallization with the hanging-drop vapor-diffusion method. Apoenzyme crystals of mouse tGAPDHS were obtained when purified protein (3 mg/ml in PBS) was mixed in a 1:1 volume ratio with a reservoir solution containing 0.2 M potassium thiocyanate, 0.1 M Bis-Tris propane, pH 6.5, 20% PEG 3350. These crystals, which appeared in the hanging drop after 6 days at 20°C, were cryoprotected with 25% glycerol in crystal growth solution and then flash-frozen in liquid nitrogen.

Human tGAPDHS was concentrated to 10 mg/ml in 10 mM HEPES (pH 7.5), 500 mM NaCl, 5% glycerol, 0.5 mM TCEP, 0.01% of sodium azide and stored at −80°C. Purified protein was mixed in a 1:1 volume ratio with each of the selected crystallization conditions from the screening kit. Crystals appeared within two days at 20°C. Apoenzyme crystals were obtained with F8 PACT Suite reservoir conditions (0.2 M Na₂SO₄, 0.1 M Bis Tris propane, pH 6.5, 20% PEG 3350) and holoenzyme crystals of tGAPDHS complexed with NAD⁺ were obtained with E5 PACT Suite reservoir conditions (0.2 NaNO₃, 20% PEG3350). Crystals were cryoprotected and flash-frozen as described for mouse tGAPDHS.

Diffraction data were collected from flash-frozen crystals at 100 K either on a Rigaku rotating anode generator using the Saturn944 CCD detector (mouse apoenzyme), or the Advanced Photon Source SER-CAT beamline (human apo- and holoenzymes). Crystal structures were solved by molecular replacement with PHASER (McCoy et al., 2007) in the CCP4 program suite (Winn et al., 2011), using human placenta GAPDH (PDB 1U8F, Jenkins and Tanner, 2006) and human sperm GAPDHS (PDB 3H9E, Chaikuad et al., 2011) structures as search models for the mouse and human sperm enzymes, respectively. The 2.01 Å mouse tGAPDHS apoenzyme structure was refined initially using CCP4 RefMac5 (Vagin et al., 2004), with final refinements using PHENIX (Adams et al., 2010) and keeping the R_free constant. Human tGAPDHS structures were refined to 1.86 Å (apoenzyme) and 1.73 Å (holoenzyme) using PHENIX (Adams et al., 2010). Data collection and refinement statistics are shown in Table I.

Table I.

Data collection and refinement statistics for human and mouse truncated glyceraldehyde 3-phosphate dehydrogenase, spermatogenic (tGAPDHS) structures.

Parameter	Human tGAPDHS-NAD+	Human tGAPDHS-apo	Mouse tGAPDHS-apo
PDB code	5C7O	5C7L	5C7I
Data collection
Beamline	Advanced Photon Source, 22-ID	Advanced Photon Source, 22-ID	Rigaku 007-HF/Saturn944 CCD
Wavelength (Å)	1.0746	1.0746	1.5418
Space group	C2	P 3121	P 3121
Monomers/asymmetric unit	2	2	2
Unit cell dimensions	a = 144.4, b = 71.8, c = 80.8; α = γ = 90.0°, β = 123.01°	a = 86.9, b = 86.9, c = 159.1; α = β=90.0°, γ = 120.0°	a = 86.7, b = 86.7, c = 158.4; α = β = 90.0°, γ = 120.0°
Resolution range (Å)	31.71–1.73 (1.79–1.73)	38.2–1.86 (1.93–1.86)	30.6–2.01 (2.08–2.01)
# of unique reflections	72 313 (1819)	59 104 (1462)	44 704 (4260)
Completeness	93.9 (55.7)	99.8 (99.9)	95.9 (92.9)
I/σI	19.2 (2.0)	17.5 (2.1)	23.7 (4.6)
Rmeas	0.091	0.169	0.071
Rpim	0.044	0.052	Not calculated
CC1/2	0.922	0.895	Not calculated
CC*	0.981	0.952	Not calculated
Redundancy	4.2 (4.0)	8.7 (6.3)	3.5 (3.1)
Refinement
# of non-hydrogen atoms (protein/ligand/water O)	5218/98/629	5210/384	5107/447
Number of reflections and cutoff	68 115/F > 0σF	54 421/F > 0σF	44 704/F > 0σF
Rwork (%)	13.7 (18.9)	18.7 (22.1)	19.2 (23.2)
Rfree (%)	16.6 (21.2)	22.7 (26.0)	21.3 (27.8)
Average B-factor (Å2) (protein/ligand/water O)	23.8/26.5/35.2	29.5/NA/33.9	29.2/NA/35.7
Wilson B-factor	19.7	23.1	24.6
RMSD bond length (Å)	0.005	0.013	0.004
RMSD bond angles (o)	1.03	1.4	0.907
Ramachandran favored/allowed (%)	95.9/3.8	96.3/3.4	95.2/4.5
Ramachandran outliers (%)	0.3	0.3	0.3

Values in parentheses are for highest resolution shells.

Standard crystallography statistics are shown for three structures determined for tGAPDHS. X-ray diffraction provides the intensity (I) of n measurements of reflections, which yield structure factor (F) values with defined indices (h, k, l). R factor statistics measure agreement between the experimental X-ray diffraction data and the crystallographic model. R_meas and R_pim are indicators of data quality and R_work and R_free assess the quality of the model. RMSD refers to root-mean-square deviation. Formulas for calculating the statistics shown are:

I/σI = average intensity/average standard deviation of intensity.

R_meas= (∑_hkl(n/n − 1)^1/2∑_i=1,n |I_i(hkl) − <I> (hkl)|)/∑_hkl∑_i=1,nI_i(hkl)), where <I> = average I of the n measurements.

R_pim= (∑_hkl(1/n − 1)^1/2∑_i=1,n |I_i(hkl) − <I> (hkl)|)/∑_hkl∑_i=1,nI_i(hkl)), where <I> = average I of the n measurements.

R_work= (∑_hkl |F_obs(hkl) − F_calc(hkl)|/∑_hkl F_obs(hkl)), ALL reflections, obs = observed, calc = calculated.

R_free= (∑_hkl |F_obs(hkl) − F_calc(hkl)|/∑_hkl F_obs(hkl)), 5% randomly chosen reflections NOT included in refinement.

CC_1/2 = Pearson's correlation coefficient between randomly chosen ‘half’ of each data set.

CC* = (2 CC_1/2/1+ CC_1/2) ^½.

Enzyme inhibition and kinetics

The dehydrogenase activity of GAPDH or GAPDHS was monitored in a kinetic assay that measures NADH accumulation at 340 nm (Schmalhausen et al., 1997). The assay was adapted for 96-well plates and the final reaction mixture in each well (200 μl) contained 100 mM glycine, 100 mM potassium phosphate, 5 mM EDTA, 0.5 mM NAD⁺, 0.5 mM d-glyceraldehyde-3-phosphate (Sigma-Aldrich, 39705) and 0.5 μg of protein (GAPDH or tGAPDHS). One enzyme unit is defined as the amount of enzyme necessary for the formation of 1 μmole 1,3-diphosphoglycerate/min. Twenty-five compounds identified as potential selective inhibitors of GAPDHS in our virtual screen were purchased (Ryan Scientific, Mount Pleasant, SC, USA) and tested for inhibition of both sperm and somatic isozymes. Stock solutions (10 mM) of each compound were prepared in dimethyl sulfoxide (DMSO), aliquoted and frozen at −20°C. Following pre-incubation of the test compounds with enzyme and NAD⁺ for 30 min at 37°C, enzymatic reactions were initiated by the addition of GAP. Triplicate samples were included for each concentration tested and DMSO controls were included in each assay. IC₅₀ values for each compound were calculated in Prism (GraphPad Software, La Jolla, CA, USA) using the four parameter nonlinear regression model. Results are expressed as the mean ± SEM for at least three different experiments with different batches of recombinant isozymes.

Kinetic assays of inhibition were conducted for each substrate (GAP or NAD⁺) at several inhibitor concentrations (1–100 μM) and the results were analyzed with Enzyme Kinetics Pro (ChemSW, Fairfield, CA, USA). Duplicate samples were measured in all assays, which were repeated three times with different enzyme preparations. Data were fit using the Lineweaver–Burk linearization method to identify the mode of inhibition and estimates K_i values.

Inhibitor effects on sperm motility and metabolism

Mouse sperm were collected from the cauda epididymides of adult CD1 males (>8 weeks old; Charles River Laboratories, Raleigh, NC, USA) in human tubal fluid (HTF) medium as described previously (Goodson et al., 2011). All procedures involving mice were performed according to the Guide for the Care and Use of Laboratory Animals with prior approval by the Institutional Animal Care and Use Committee within the AAALAC accredited program at the University of North Carolina at Chapel Hill (UNC-CH) (Animal Welfare Assurance Number: A-3410-01).

De-identified human semen samples were obtained from excess stocks collected by the UNC-CH Fertility Clinic, Department of Obstetrics and Gynecology. All human samples were from healthy donors with normal semen parameters and screened for HIV (human immunodeficiency virus), RPR (rapid plasma reagin: syphilis) and hepatitis. Criteria for selection of donor sperm included normal morphology using strict criteria (World Health Organization, 2010) and a post-thaw count of >80 × 10⁶ motile sperm. Semen samples were diluted 1:1 with TEST-yolk buffer (Prins and Weidel, 1986) following collection, divided into aliquots with ≥20 × 10⁶ sperm, and stored in liquid nitrogen. After thawing, sperm were isolated from seminal plasma and cryoprotectant by low speed centrifugation (365g) for 10 min through a 90% colloidal silica suspension (ISolate, Irvine Scientific, Santa Ana, CA, USA) at a 1:1 volume ratio. The upper layer was discarded and the remaining sperm pellet was resuspended into 2 ml HEPES buffered Sperm Washing Medium (Irvine Scientific) and centrifuged for 5 min at 365g. The washing step was repeated twice and the final pellet was resuspended in 1–2 ml of HTF medium without human serum albumin.

Sperm were incubated at 37°C under 5% CO₂ in air for 2 h in HTF medium containing 50 or 100 μM T0501_7749. Effects on motility were monitored by computer-assisted sperm analysis (CASA) with the Hamilton Thorne CEROS imaging system, version 12.3H IVOS software (Goodson et al., 2011). Sperm tracks (1.5 s) were captured at a frame acquisition rate of 60 Hz. Aliquots were removed at 30 min intervals to compare sperm motility in the presence of inhibitors with motility in the DMSO vehicle control. Motility analyses were repeated in modified HTF medium, omitting bovine serum albumin (BSA) or human serum albumin or replacing BSA with 0.4 mM methyl-β-cyclodextrin and 0.01% polyvinyl alcohol (Hasegawa et al., 2012). For each condition, experiments were repeated at least three times with sperm from different mice or human donors. Sperm viability was monitored by the uptake of propidium iodide by cells with damaged plasma membranes, as described previously (Goodson et al., 2012).

Lactate production and accumulation in the medium was measured in mouse sperm to monitor inhibitor effects on sperm glycolysis. Sperm were incubated for 2 h ± T0501_7749 under the same conditions used for motility analyses in HTF medium with 0.4 mM methyl-β-cyclodextrin and 0.01% polyvinyl alcohol, except that lactate and pyruvate were omitted from the medium. Osmolality of the medium was adjusted to ∼315 mOsm/kg with 5 M NaCl using a Model 3300 micro-osmometer (Advanced Instruments, Norwood, MA, USA). Duplicate aliquots were removed at time 0 and 2 h, centrifuged to remove sperm and assayed to measure lactate accumulation in the medium. The spectrophotometric assay monitors the conversion of lactate to pyruvate by lactate dehydrogenase in the presence of NAD⁺ and hydrazine (Pesce et al., 1975; Danshina et al., 2001). In this assay the concentration of lactate in the sample is proportional to the increase in absorbance at 340 nm as NAD⁺ is reduced to NADH.

Statistical analyses were performed using Prism (GraphPad Software, La Jolla, CA, USA). Data are shown as mean ± SEM and statistical significance (P< 0.05) was calculated using either two-tailed unpaired t-tests or two-way analysis of variance followed by Tukey's multiple-comparison test.

Results

Structural differences between sperm and somatic isozymes

Both mouse and human sperm require glycolysis (Miki et al., 2004; Nascimento et al., 2008) and have GAPDHS isozymes with 83% amino acid identity, suggesting that the mouse may be useful as a model system for testing GAPDHS inhibitors as potential contraceptives. Alignment of the amino acid sequences for human and mouse GAPDHS and GAPDH shows that 68 amino acids are identical in the sperm isozymes (highlighted in red in Fig. 1), but distinct from the corresponding amino acids in the somatic isozymes. The proline-rich N-terminal regions of human (amino acids 1–75) and mouse (amino acids 1–105) GAPDHS are not included in the alignment since homologous regions do not exist in their somatic counterparts. Unless noted otherwise, we will refer to amino acid numbering based on the sequences of human GAPDHS (top row in Fig. 1) and human GAPDH (bottom row in Fig. 1). Each subunit of the tetrameric enzyme has an NAD⁺-binding domain in the N-terminal segment of the protein preceding the active site C224 and a catalytic domain in the C-terminal half. Eight amino acids in the C-terminal domain that have crucial roles in catalysis (denoted by dots above the sequences in Fig. 1; Seidler, 2013) are conserved in both sperm and somatic isozymes. Forty-six of the sperm-specific residues are in the NAD⁺-binding domain and the remaining 22 are in the catalytic domain.

Figure 1.

Alignment of human (h) and mouse (m) glyceraldehyde 3-phosphate dehydrogenase, spermatogenic (GAPDHS) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) amino acid sequences. The human GAPDHS sequence (NP_055179) is shown on the top row, without the proline-rich N-terminal region (amino acids 1–75) that is characteristic of the sperm isozyme. Similarly, mouse GAPDHS (NP_032111) in the second row does not include the N-terminal 105 residues. Highlighted in red are amino acids that are conserved in these sperm isozymes, but distinct from the corresponding amino acids in the somatic GAPDH isozymes shown in the third and fourth rows (mouse, AAH83149 and human, NP_002037). The sperm isozymes have 83% amino acid identity, compared with the ∼70% identity between the sperm and somatic isozymes in each species. Dots above the sequences denote amino acids required for catalysis. Colored boxes indicate amino acids that form pockets 1 (blue), 2 (green) and 3 (yellow) surrounding the active site, as identified by SiteID in the SYBYL molecular modeling program. Arrows denote sperm-specific amino acids in pockets 1 and 2. Amino acids forming the S-loop are labeled.

We constructed homology models for mouse and human GAPDHS and GAPDH based on the 2.0 Å crystal structure of Palinurus versicolor GAPDH (PDB 1CRW; Shen et al., 2000) to assess the three-dimensional location of the sperm-specific amino acids. Models of the sperm isozymes represent tGAPDHS without the proline-rich N-terminal extensions. Most of the 68 sperm-specific amino acids noted in Fig. 1 are located on the surface of the protein subunit (pink in Supplementary Video S1). However, two regions with sperm-specific amino acids (red in Supplementary Video S1) are closer to the active site. One region near the center of the model is adjacent to the catalytic site where both GAP (orange) and the nicotinamide ring of NAD⁺ (gray) bind. The second region near the top of the model is at the opposite end of the NAD⁺ binding site surrounding the adenine moiety of the cofactor.

To evaluate the potential for differential binding to the sperm and somatic isozymes, we used SiteID in the SYBYL molecular modeling program to identify three binding pockets surrounding the active site in our tGAPDHS homology models (Fig. 2A). Subsequent comparisons with high resolution crystal structures for human tGAPDHS (PDB 3H9E, Chaikuad et al., 2011 and crystal structures solved in this study) and human GAPDH (PDB 1U8F, Jenkins and Tanner, 2006) confirmed the conformations of all three pockets. Pockets 1 and 2 (blue and green, respectively, in Fig. 2A) contain the two regions with sperm-specific amino acids noted in Supplementary Video S1. These amino acid differences are highlighted in Fig. 2B, where structures of the sperm (pink) and somatic (gray) isozymes are aligned and superimposed.

Figure 2.

Binding pockets and sperm-specific amino acids near the active site. (A) To facilitate virtual screening, three binding pockets were identified in our homology model of truncated GAPDHS (tGAPDHS) using SiteID in the SYBYL molecular modeling program. Surfaces are shown for pockets 1 (blue), 2 (green) and 3 (yellow) which surround the glyceraldehyde 3-phosphate (GAP) substrate (orange) and NAD⁺ cofactor (black). (B) Superimposed structures of human tGAPDHS (pink, PDB 3H9E) and GAPDH (gray, PDB 1U8F) highlighting the position of sperm-specific residues in pockets 1 (lower box) and 2 (upper box).

Pocket 1 (blue in Fig. 2A) surrounds the substrate and the nicotinamide moiety of NAD⁺. It is formed by the side-chains of R85, I86, L89, S193, S223, C224, H251, S252, Y253, T254, A255, P310, N388, E389, Y392 and S393 of human GAPDHS (blue boxes in Fig. 1). Pocket 1 includes two polar residues, S252 and Y253 (arrows in Fig. 1 and in the lower box of Fig. 2B), that are specific to the sperm isozymes and are located within 6 Å of the active site where they are flanked by two of the amino acids required for catalysis (Fig. 1). Unlike the smaller, hydrophobic residues (A180 and I181) present at corresponding positions in the somatic enzymes, S252 and Y253 have the potential for hydrogen bond formation. Pocket 1 of human GAPDHS has a third amino acid (P310) that is distinct from the corresponding residue in human GAPDH (A238).

Pocket 2 (green in Fig. 2A) surrounds the adenine moiety of NAD⁺ and is formed by the side-chains of N81, G82, F83, G84, N105, D106, P107, F108, C150, K151, E152, P153, E168, S169, T170, V172 and Y173 of human GAPDHS (green boxes in Fig. 1). There are three sperm-specific amino acids in pocket 2, C150, K151 and Y173 (arrows in Fig. 1 and in the upper box of Fig. 2B) replacing E79, R80 and F102 in the somatic isozymes. Both K151 and Y173 are within 2.5 Å of the adenine portion of NAD⁺. C150 and K151 in GAPDHS are smaller than the corresponding residues in GAPDH, and the C150 residue alters charge and provides a nucleophilic thiol group. The polar Y173 residue replaces a smaller, nonpolar residue in GAPDH (F102), shown previously to interact directly with NAD⁺ through hydrophobic contacts (Moras et al., 1975). It is interesting to note that the five sperm-specific residues identified in pockets 1 and 2 are all highly conserved in mammalian GAPDHS orthologs identified in Ensembl (Supplementary Table SI). C150 and Y173 in pocket 2 are also conserved in reptile and fish orthologs.

Pocket 3 (yellow in Fig. 2A) surrounds the inorganic phosphate binding site and is formed by the side-chains of T256, Q257, K258, S264, A267, R269, D270, G271, I279, P280, A281, S282, A304 and R306 of human GAPDHS (yellow boxes in Fig. 1). This pocket does not contain sperm-specific amino acids that are conserved between human and mouse GAPDHS, although L195 in the somatic isozyme is replaced by A267 in human GAPDHS and D297 in mouse GAPDHS. Two additional residues (G300 and S311) are distinct only in mouse GAPDHS, replacing D198 and A209 in the somatic isozyme.

To explore GAPDHS structures in greater detail, we expressed recombinant human and mouse tGAPDHS without their proline-rich N-terminal regions and compared multiple crystal structures for these sperm isozymes. We solved three distinct crystal structures (Table I), including the first apoenzyme crystal structures for human tGAPDHS (PDB 5C7L, resolution 1.86 Å) and mouse tGAPDHS (PDB 5C7I, 2.01 Å). We also obtained a 1.73 Å holoenzyme structure for human tGAPDHS complexed with NAD⁺ (PDB 5C7O). The coordinates and data files for these crystal structures were deposited in the RCSB Protein Data Bank (http://www.rcsb.org). Using subunit nomenclature previously established for GAPDH (Buehner et al., 1974), the asymmetric unit of the human sperm NAD⁺-bound holoenzyme is a dimer of O and P subunits, while the asymmetric unit in the human and mouse apoenzyme structures is an O-Q dimer. In both cases, the consensus tetramer of the GAPDH family is formed by crystallographic symmetry that recapitulates the biological tetramer. As reported for multiple GAPDH and GAPDHS structures (Cowan-Jacob et al., 2003; Ismail and Park, 2005; Frayne et al., 2009), the active site cysteines (C224 in human GAPDHS, C256 in mouse GAPDHS) in our structures were oxidized to the sulfinic acid form, with clear electron density for two oxygens attached to the cysteine sulfur.

The C2 space group of our NAD⁺-bound structure for human tGAPDHS (PDB 5C7O) is the same as the published NAD⁺-bound structures for this isozyme (PDB 3H9E and 3PWF, Chaikuad et al., 2011), but distinct from the P3₁21 space group of both apoenzyme structures. A novel feature of our NAD-bound tGAPDHS structure is the presence of disulfide bonds between tetramers, linking the C150 residue of each monomer with the C150 of the symmetry-related monomer in the adjacent tetramer. C150 is one of the highly conserved sperm-specific amino acids in GAPDHS (pocket 2 in our homology model), replacing E79 in the somatic isozyme. Figure 3A shows the tetramers stacked and translated as they are in the holoenzyme crystal, with the C150 residues shown as black spheres. Disulfide bonds between these residues facilitate stacking of the tetramers parallel to the R-axis of the C2 space group, which would continue along the direction of this axis. The C150 residue is within 5 Å of the adenine ring in the NAD⁺ bound within the crystal. As shown in the electron density plot of this tetramer interface (Fig. 3B), the disulfide bond between C150 residues is the major conformation observed in this crystal structure (occupancy = 0.88). The minor non-disulfide-bonded conformation of C150 has an occupancy of only 0.12.

Figure 3.

Features of the holo- (PDB 5C7O) and apoenzyme (PDB 5C7L) crystal structures of human tGAPDHS. (A) Two holoenzyme tetramers stacked and translated as they are in the C2 crystal form. Subunits and axes are labeled as described by Buehner et al. (1974). The bound NAD⁺ cofactors are shown in each subunit as black ball-and-stick structures. In this holoenzyme structure the sperm-specific C150 residues (highlighted as black spheres) form disulfide bonds between tetramers, facilitating stacking parallel to the R-axis of the C2 space group. The S-loops of each subunit are shown as thicker ribbons. (B) Electron density (2Fo-Fc contoured at 1σ, where Fo is the observed structure factor and Fc is the calculated structure factor) around the inter-tetramer disulfide bond seen in the NAD⁺-bound holoenzyme. The figure shows that there is a major disulfide-bonded conformation and a minor non-disulfide-bonded conformation of C150. The distance between sulfur atoms (2.3 Å) is slightly longer than the consensus disulfide bond, but there is continuous electron density between them. (C) Apoenzyme tetramers are more open, packing in the P3₁21 crystal form without disulfides between the C150 residues (black spheres).

Inter-subunit disulfide bonds are not found in either the human or mouse tGAPDHS crystal structures without NAD⁺ (PDB 5C7L and PDB 5C7I), where the loop containing C150 (residues 148–158) has a slightly different conformation. Figure 3C shows two tetramers of the human apoenzyme that are rotated and translated, illustrating that they cannot stack as they do in the C2 holoenzyme (Fig. 3A). When the human apo- and holoenzyme structures are superimposed (CCP4 Superpose, Krissinel and Henrick, 2004), the root-mean-square deviation (RMSD) for all atoms of the C150 loop (including hydrogens and side-chains) is 0.7 Å, larger than the 0.38 RMSD over the entire structure. In the apoenzyme, the crystal packing changes since the relationship between monomers in the tetramer is slightly more open. For this reason, the apposition of C150 residues and consequent disulfide bonds that are facilitated in the holo space group are not allowed in the trigonal space group that the apoenzyme forms.

Previous comparisons of GAPDH crystal structures indicate that differences between the apo- and holoenzyme occur primarily in the N-terminal NAD⁺-binding domain (Skarzynski and Wonacott, 1988; Shen et al., 2000; Yun et al., 2000), reflecting the conformational changes that occur during cofactor binding. To assess conformational changes in the sperm isozyme, we superimposed our human apo- and holo-tGAPDHS structures with CCP4 Superpose (Krissinel and Henrick, 2004) and plotted the RMSD between equivalent Cα backbone atoms along the polypeptide chain for both subunits of the crystallographic unit (Fig. 4). Comparable to the somatic isozyme, the greatest displacement for tGAPDHS was seen in the NAD⁺-binding domain (residues 76–223). Only minor variations (<0.5 Å) were apparent in most of the C-terminal catalytic domain. However, there was a larger conformational difference between the apo- and holo-tGAPDHS structures in the S-loop region (residues 253–276, thicker ribbon regions in Fig. 3, Banas et al., 1987), with maximum displacement (0.88 Å) centered at R265 and K266. Several S-loop amino acids are included in binding pockets 1 and 3 identified with SiteID, including the sperm-specific Y253 residue in pocket 1. Structural analyses of GAPDH indicate that the S-loops form the core of the tetrameric enzyme, making contacts with neighboring subunits across the R and P axes (Biesecker et al., 1977). Supplementary Video S2 illustrates the conformational change in the GAPDHS S-loop that occurs when NAD⁺ binds.

Figure 4.

Root-mean-square deviation (RMSD, Å) between equivalent Cα atoms along the polypeptide chains of the O and P subunits when the human apo- and holo-tGAPDHS crystal structures (PDB 5C7L and PDB 5C7O) are superimposed. Displacement is greatest within the N-terminal NAD⁺-binding domain and in the S-loop (residues 253–276) within the catalytic domain.

We also compared the mouse and human tGAPDHS apoenzyme structures (Fig. 5) by superimposing the Cα backbone atoms using CCP4 Superpose (Krissinel and Henrick, 2004). The overall arrangements of helices and strands in these structures were in very good agreement, with an average RMSD of 0.356 Å for all 1320 main chain atoms in the O subunit and 0.373 for the P subunits, not including side-chains since there are sequence differences between the orthologs. The most significant relative displacement between the apoenzyme structures was in the surface loop formed by residues W158-P164 of human tGAPDHS (black arrow in Fig. 5). Four of the seven amino acids in this loop are different in the human (WRAVGSP) and mouse (WSSIGNP) isozymes. The RMSD of the 28 main chain atoms for this loop is 1.08 Å. The position of this loop is unlikely to alter the novel structural features as described for human tGAPDHS, including potential disulfide bonds and S-loop conformational changes (which occur in the regions highlighted in red in Fig. 5).

Figure 5.

Mouse (green, PDB 5C7I) and human (blue, PDB 5C7L) tGAPDHS apoenzyme structures superimposed along their Cα backbones. The largest displacement between the structures was seen in a surface loop (black arrow) that is not close to the S-loops or the C150 residues that form disulfides in the holoenzyme (highlighted in red).

Virtual screening to identify compounds predicted to inhibit GAPDHS selectively

With the goal of identifying selective inhibitors of GAPDHS, we used the SYBYL FlexX algorithm to compare docking of compounds from the Maybridge small molecule database into the binding pockets of the sperm and somatic isozymes. The Maybridge collection used for this screen included over 300 000 organic compounds designed to exhibit diverse ring structures with drug-like characteristics. We first prescreened the Maybridge compounds on the basis of size to eliminate compounds that did not fit into the three pockets identified by SiteID. The remaining compounds (up to 8000) were docked with FlexX into each of the pockets, comparing predicted binding in both mouse and human tGAPDHS with the somatic isozyme in each species. Protein-ligand interaction scores were determined for each isozyme, with more negative scores reflecting tighter binding. Discrimination scores were calculated as the difference between the interaction scores for the sperm and somatic isozymes. For compounds with the largest discrimination scores for the human isozymes, docking was repeated thirty times to identify compounds that docked consistently within the predicted binding pocket. Following cluster analysis and visual inspection for drug-like properties and distinct scaffolds, we selected 25 representative compounds for further testing in enzyme assays. These compounds had discrimination scores between −2.2 and −17.6 and were predicted to form hydrogen bonds within one of the pockets of the sperm isozyme (eleven pocket 1, eleven pocket 2 and three pocket 3 compounds).

Testing compounds to compare inhibition of GAPDHS and GAPDH

We tested predicted inhibitors in a spectrophotometric assay of enzyme activity that measures NADH accumulation at 340 nm (Schmalhausen et al., 1997), initially comparing inhibition of mouse tGAPDHS and GAPDH. Two compounds inhibited ∼50% mouse tGAPDHS activity at concentrations that did not alter activity of the somatic enzyme, but complete inhibition was not achieved because these compounds exhibited limited solubility at higher concentrations. To estimate IC₅₀ values, detailed dose–response assays were conducted comparing inhibition of human tGAPDHS and GAPDH.

T0501_7749 (IUPAC Name: 2-[2-amino-3-(4-methylphenyl)sulfonylpyrrolo[3,2-b]quinoxalin-1-yl]-1-(4-nitrophenyl)ethanol; structure shown in Fig. 6A) inhibited >60% human tGAPDHS activity at a concentration of 5 μM (Fig. 6B). IC₅₀ values, calculated using the four parameter nonlinear model (Prism, Graphpad Software), were 1.2 μM for human tGAPDHS and 38.5 μM for human GAPDH, indicating that T0501_7749 inhibited the sperm isozyme with >10-fold selectivity. FlexX docking predicted that T0501_7749 binds in pocket 1 of human tGAPDHS (Fig. 6C), forming hydrogen bonds with Y253 and P310, two residues that are distinct from the somatic isozyme. T0501_7749 did not form bonds with the corresponding amino acids (I181 and A238) when docked in human GAPDH, interacting instead with pocket 1 residues (R13, T182, A183) that are shared in the sperm and somatic isozymes (Fig. 6D).

Figure 6.

Properties of T0501_7749. (A) Chemical structure of this small-molecule inhibitor. (B) Dose–response curves monitoring inhibition of human tGAPDHS (red, n = 3–6 for all concentrations except first and last, where n = 2) and GAPDH (black, n = 4–16) in the presence of 2 mM GAP and 2 mM NAD⁺. (C) FlexX predicted binding pose of T0501_7749 within pocket 1 (blue) of human tGAPDHS, forming hydrogen bonds (dotted lines) with Y253 and P310, which are distinct from the corresponding residues (I181 and A238) in GAPDH. The sperm-specific residues (S252 and Y253) in pocket 1 are highlighted in red. (D) In the somatic isozyme, FlexX predicted that T0501_7749 forms hydrogen bonds with three pocket 1 residues that are shared with the sperm isozyme (R13, T182, A183). GAPDH amino acids (A180, I181) corresponding to the sperm-specific residues are highlighted in gray.

T0506_9350 (IUPAC Name: 1-cyclohexyl-3-[4-[(4-methoxyphenyl)sulfamoyl]-2-nitroanilino]urea; structure shown in Fig. 7A) inhibited ∼60% human tGAPDHS activity at a concentration of 200 μM (Fig. 7B). Consistent inhibition was observed with this compound only when the temperature of the 30 min pre-incubation period with enzyme was reduced to 4°C. Although T0506_9350 is not a very potent inhibitor, it did exhibit partial selectivity for the sperm isozyme, with estimated IC₅₀ values of 95.8 μM for GAPDHS compared with 860 μM for GAPDH. FlexX docking predicted that T0506_9350 binds in pocket 2 of tGAPDHS forming five hydrogen bonds with two sperm-specific residues, Y173 and K151 (Fig. 7C). The corresponding residues in GAPDH (F102 and E79) were not predicted to bind to T0506_9350. Instead, docking predicted hydrogen-bond interactions with GAPDH residues (F11, D35 and T99) that are shared with the sperm isozyme (Fig. 7D).

Figure 7.

Properties of T0506_9350. (A) Chemical structure of this small-molecule inhibitor. (B) Dose–response curves monitoring inhibition of human tGAPDHS (red, n = 4–6) and GAPDH (black, n = 5–6) in the presence of 2 mM GAP and 2 mM NAD⁺. (C) FlexX predicted binding pose of T0506_9350 within pocket 2 (green) of human tGAPDHS, forming hydrogen bonds (dotted lines) with Y173 and K151. These residues flank the adenine ring of NAD⁺ (see Fig. 2B) and are two of the three sperm-specific residues in pocket 2 (highlighted in red). (D) FlexX predicted that T0506_9350 binding to GAPDH is removed from the corresponding amino acids (highlighted in gray), forming hydrogen bonds with pocket 2 residues that are shared with the sperm isozyme (F11, D35, T99).

Kinetic analyses were conducted to determine the modes of inhibition and K_i values for T0501_7749 and T0506_9350, and the resulting Lineweaver–Burk plots for human tGAPDHS are shown in Fig. 8. Each double reciprocal line (1/V_o versus 1/[S]) shows the effects of a single inhibitor concentration with variable concentrations of either GAP (Fig. 8A and C) or NAD⁺ (Fig. 8B and D). When GAP concentrations vary, lines representing higher concentrations of T0501_7749 have steeper slopes (K_m/V_max) and smaller x intercepts (−1/K_m) with little variation in the y intercept (1/V_max) (Fig. 8A). This pattern of increasing K_m and unchanging V_max indicates that T0501_7749 inhibition is competitive with the GAP substrate. In contrast, the double reciprocal lines for T0501_7749 are approximately parallel when NAD⁺ concentrations vary (Fig. 8B), reflecting decreases in both V_max and K_m. This pattern reflects uncompetitive inhibition, indicating that T0501_7749 binds following formation of the NAD⁺-tGAPDHS complex. Mean K_i values for T0501_7749 were 10.9 μM for GAP and 7.1 μM for NAD⁺. Identical modes of inhibition were observed in kinetic analyses of mouse tGAPDHS (Supplementary Fig. S1A and B), with mean K_i values of 19.9 μM for GAP and 31.6 μM for NAD⁺. Kinetic analyses indicate that T0506_9350 inhibition of human and mouse tGAPDHS is competitive with both GAP (Fig. 8C and Supplementary Fig. S1C) and NAD⁺ (Fig. 8D and Supplementary Fig. S1D). Mean K_i values were 63.3 μM for GAP and 60.2 μM for NAD⁺ for human tGAPDHS, and 71.0 μM for GAP and 56.9 μM for NAD⁺ for mouse tGAPDHS.

Figure 8.

Lineweaver–Burk plots for the kinetic analysis of T0501_7749 (A and B) and T0506_9350 (C and D) inhibition of human tGAPDHS. Each line (1/initial velocity V_o versus 1/substrate concentration [S]) shows the effects of a single inhibitor concentration with varying concentrations (0.05–2 mM) of either GAP (A and C) or NAD⁺ (B and D). Inhibitors concentrations were: no inhibitor (▪), 1 μM (♦), 3 μM (▴),10 μM (○), 20 μM (□), 30 μM (◊), 80 μM (Δ), and 100 μM (x).

GAPDHS inhibitor effects on sperm motility and metabolism

Based on the phenotype of GAPDHS knockout mice (Miki et al., 2004), we expect inhibitors of GAPDHS to impair sperm motility by blocking the glycolytic pathway. Our studies of inhibitor effects on sperm focused on T0501_7749, since T0506_9350 had a very high IC₅₀ and limited stability above 4°C. Using CASA, we first monitored effects on mouse sperm motility in HTF medium containing 5 mg/ml BSA throughout a 2 h incubation at 37°C, conditions which support capacitation (Goodson et al., 2011). Under these conditions, T0501_7749 did not reduce the percentage of motile sperm. In other preliminary studies, however, we found that T0501_7749 inhibition of GAPDHS enzymatic activity was substantially reduced by the addition of BSA or sperm from GAPDHS knockout mice (not shown). Since these results suggested that nonspecific protein binding to T0501_7749 reduced its effectiveness, we assayed sperm motility in modified HTF medium. T0501_7749 caused significant reductions in the percentage of motile sperm during in vitro capacitation when BSA was replaced by 0.4 mM methyl-β-cyclodextrin and 0.01% polyvinyl alcohol (Fig. 9A). More pronounced reductions to <15% motile sperm occurred within 30 min when T0501_7749 incubations were conducted in HTF medium without BSA or replacement compounds that facilitate capacitation (Fig. 9B). After human samples were washed to remove seminal plasma and cryoprotectant, T0501_7749 also significantly reduced the percentage of motile sperm when incubated in HTF without human serum albumin (Fig. 9C). Mean sperm viabilities at 120 min were comparable for control samples (48.0%) and samples treated with 100 µM T0501_7749 (56.7%).

Figure 9.

Inhibition of sperm motility by T0501_7749 monitored by computer-assisted sperm analysis. The percentages of motile sperm were determined at 30 min intervals, plotted as mean values ± SEM, with treatments compared by two-way analysis of variance followed by Tukey's multiple-comparison test. Mouse sperm were incubated in (A) human tubal fluid (HTF) medium with 0.4 mM methyl-β-cyclodextrin and 0.01% polyvinyl alcohol replacing bovine serum albumin (BSA) or (B) HTF with neither these constituents nor BSA. In both media T0501_7749 caused significant reductions in the percentage of motile sperm compared with DMSO vehicle controls. For (A), n = 3 and P≤ 0.03 for 50 or 100 μM at 60, 90 and 120 min. For (B), n = 3 to 7; P ≤ 0.003 for 50 μM at 30, 60 and 120 min; P < 0.0001 for 100 μM at these time points. (C) T0501_7749 also significantly reduced the motility of human sperm incubated in HTF without human serum albumin (n = 8; P for 100 μM compared with control < 0.05 at 60 and 90 min, <0.01 at 120 min).

Sperm lactate production and accumulation in the medium during in vitro incubations has long been used as an indicator of glycolytic activity (Mann and Lutwak-Mann, 1981). T0501_7749 reduced lactate production by ≥50% (P= 0.017 for 100 μM T0501_7749 versus control) when mouse sperm were incubated for 2 h in HTF containing 0.4 mM methyl-β-cyclodextrin, 0.01% polyvinyl alcohol and glucose as the sole substrate (Fig. 10), providing evidence that this compound inhibits glycolysis in the target cell.

Figure 10.

Inhibition of mouse sperm lactate production by T0501_7749. Mouse sperm were incubated for 2 h under the same conditions as in Fig. 9A, except that lactate and pyruvate were omitted from the HTF medium. Data are shown as mean values ± SEM for sperm treated with 50 μM (n = 1) or 100 μM T0501_7749 (n = 3) compared with control sperm incubated with equal amounts of the DMSO vehicle (n = 3). The reduction was statistically significant when the 100 μM and control treatments were compared using a two-tailed unpaired t-test (*P= 0.017).

Discussion

Our structural analyses of GAPDHS focused on differences between the sperm and somatic isozymes that are conserved between species, with the goal of identifying regions uniquely important for GAPDHS function that are suitable for selective targeting to achieve contraception. We solved the first apoenzyme structures for human and mouse GAPDHS and demonstrated very close alignment between these orthologs. Comparison of our human apo- and holoenzyme structures also identified features not previously observed in GAPDHS or GAPDH crystal structures. Our virtual screen compared the predicted binding of compounds to the sperm and somatic isozymes and identified two small-molecule inhibitors with partial selectivity for the sperm isozyme. Furthermore, T0501_7749, which inhibited GAPDHS with micromolar potency and >10-fold selectivity, exhibited the expected inhibitory effects on sperm motility and lactate production.

In addition to its distinctive N-terminus, GAPDHS has 68 amino acids that are conserved in the human and mouse sperm isozyme, but are different from the corresponding amino acid in the somatic isozymes of both species. Although most of these sperm-specific residues are on the surface, five are clustered in pockets 1 and 2 surrounding the binding sites for GAP and NAD⁺. Chaikuad et al. (2011) noted that differences in electrostatic properties in these two regions may contribute to kinetic differences between the sperm and somatic enzymes, notably a lower K_m and increased catalytic efficiency for tGAPDHS with respect to NAD⁺. Two sperm-specific residues, S252 and Y253, are in pocket 1 surrounding the substrate and the nicotinamide portion of NAD⁺. Y253 is predicted to bind to the T0501_7749 selective inhibitor. This residue also provides a potential phosphorylation site that is not present in the somatic isozyme. The sperm-specific K151 and Y173 residues are in binding pocket 2 where they are localized within 2.5 Å of the adenine moiety of NAD⁺. In multiple GAPDH orthologs this region is important for cofactor binding and induction of conformational changes that precede catalysis (Skarzynski and Wonacott, 1988; Shen et al., 2000; Yun et al., 2000). Another sperm-specific residue, C150, is located in the same pocket between K151 and Y173, but is oriented in the opposite direction away from NAD⁺ (Fig. 2).

Highly conserved in mammalian, reptile and fish orthologs of GAPDHS, C150 is an interesting feature of the sperm isozyme since it mediates disulfide bond formation between adjacent tetramers in our holoenzyme crystal structure (PDB 5C7O). Inter-tetramer disulfide bonds were not observed in the previously published human sperm holoenzyme structure (PDB 3H9E, Chaikuad et al., 2011), with the same space group and very similar unit cell constants. Perhaps differences in protein production and crystallization conditions may have contributed to a slightly more ‘oxidized’ tGAPDHS in this study, resulting in packing that allows disulfides and the oxidation of the active site cysteine to sulfinic acid. Recombinant tGAPDHS has enhanced stability compared with the somatic isozyme, exhibiting greater resistance to inactivation by heat or guanidine hydrochloride (Elkina et al., 2010; Kuravsky et al., 2014). Furthermore, the native protein with its N-terminal extension is tightly bound to the fibrous sheath in mature sperm (Westhoff and Kamp, 1997; Bunch et al., 1998), persisting when this cytoskeletal structure is isolated by sequential extraction with 1% Triton X-100, 0.6 M potassium thiocyanate and 6 M urea (Krisfalusi et al., 2006). Disulfide bond formation between GAPDHS tetramers may contribute to the stability of this protein and the protein complexes that form the fibrous sheath. Further studies are needed to assess disulfide bond formation between the highly conserved C150 residues of GAPDHS during flagellar formation and during maturation in the epididymis, where disulfide linkages increase in both head and tail proteins of mammalian sperm (Calvin and Bedford, 1971; Balhorn, 2007; Baker et al., 2015).

Although we were unable to obtain crystals of tGAPDHS complexed with T0501_7749, our crystal structures should prove useful in the design and analysis of selective inhibitors. The monoclinic C2 space group of tGAPDHS holoenzyme structures (PDB 5C7O and PDB 3H9E, Chaikuad et al., 2011) is distinct from our apoenzyme structures (PDB 5C7L and 5C7I), which belong to the trigonal P3₁21 space group and cannot form disulfide bonds between tetramers. Comparison of our apo- and holoenzyme structures confirms that GAPDHS, like GAPDH from multiple species (Skarzynski and Wonacott, 1988; Shen et al., 2000; Yun et al., 2000), undergoes significant conformational change in the N-terminal NAD⁺-binding domain when the cofactor binds. Unlike GAPDH in these earlier structural comparisons, GAPDHS also exhibits distinct conformational change in the S-loop region (residues 253–276) of the catalytic domain upon cofactor binding. The maximum displacement in this region occurs at basic residues R265 and K266. The basic residue at position 265 in GAPDHS (R in human or K in mouse) replaces a nonpolar glycine in the somatic isozyme. There are also sperm-specific amino acids (Y253 and H275) at both ends of the S-loop. Previous studies suggest that the S-loops, which extend across the R-axis (see Fig. 3) and interact with NAD⁺ in the adjacent subunit, may contribute to cooperativity between subunits for cofactor binding (Moras et al., 1975; Biesecker et al., 1977). A recent study found that human GAPDHS exhibits positive cooperativity, opposite to the negative cooperativity observed in mammalian GAPDH (Kuravsky et al., 2015). Although this property was eliminated by mutagenesis to disrupt the D311-H124 salt bridge (Kuravsky et al., 2015), the role of S-loop amino acids in GAPDHS cooperativity has not been explored to our knowledge.

S-loop residues also form one side of the selectivity cleft, which is wider in trypanosomatid GAPDH structures and has been targeted for the design of selective inhibitors to combat illnesses caused by these parasitic protozoa (Verlinde et al., 1994; Suresh et al., 2001). This is an intersubunit cleft across the R-axis that is adjacent to the adenosine ribose of NAD⁺ in the neighboring subunit. ‘Open’ (PDB 1ZNQ; Ismail and Park, 2005) and ‘closed’ (PDB 1U8F; Jenkins and Tanner, 2006) conformations of the S-loop have been reported in human GAPDH structures, although the distance between residues closest to NAD⁺ (S-loop P191 and F37 in the neighboring subunit) is similar in these two structures. These residues are conserved in the sperm isozyme (P263 and F108). Only the ‘closed’ S-loop conformation has been observed in both apoenzyme (human PDB 5C7L and mouse 5C7I) and holoenzyme (human PDB 5C7O and PDB 3H9E, Chaikuad et al., 2011) structures of GAPDHS. The S-loop P263 in our holoenzyme structure (PDB 5C7O) is 3.8 Å from the 2′-hydroxyl of the NAD⁺ in the neighboring subunit. Consequently, adenosine analogs designed to fit in the larger trypanosomatid selectivity cleft are unlikely to inhibit the sperm isozyme.

Two compounds predicted to bind more tightly to GAPDHS than to GAPDH in our virtual screen inhibited GAPDHS in enzymatic assays and exhibited partial selectivity for the sperm isozyme. FlexX docking predicted that both of these inhibitors interact with amino acids that are conserved in human and mouse GAPDHS, but are distinct from the corresponding amino acids in the somatic isozymes in both species. Although both compounds exhibited limited solubility, we found that T0501_7749 inhibited human GAPDHS with an estimated IC₅₀ of 1.2 μM and >10-fold selectivity. Kinetic analyses with both human and mouse GAPDHS indicate that T0501_7749 competes with GAP substrate binding following formation of the cofactor-enzyme complex.

T0501_7749 appears to bind nonspecifically to BSA and other proteins, hampering full evaluation of its effects on sperm function. Nevertheless, we demonstrated that this compound inhibited both glycolysis and motility in mouse sperm when albumin in the medium was replaced with methyl-β-cyclodextrin and polyvinyl alcohol. The percentage of motile sperm was further reduced to <15% when these additives and BSA were omitted completely. T0501_7749 also caused significant inhibition of human sperm motility, although the effect was not as pronounced. It should be noted that the human sperm used in this study were exposed to proteins from seminal fluid and egg yolk in the cryopreservation medium, potentially modulating the effectiveness of this compound.

This study provides proof-of-principle evidence that GAPDHS can be inhibited with at least partial selectivity and that small-molecule inhibitors of this enzyme significantly reduce sperm glycolysis and motility. We are currently using medicinal chemistry to optimize structure-activity relationships within the T0501_7749 scaffold, with the goal of improving potency, selectivity and drug-like chemical properties. In addition, we will continue to use structure-based cheminformatics approaches to search for additional GAPDHS inhibitors, efforts that are informed by our crystal structures.

Supplementary data

Supplementary data are available at http://molehr.oxfordjournals.org/.

Authors' roles

P.V.D. had primary responsibility for all aspects of data acquisition and analysis, and prepared the initial draft of the manuscript. WD.Q. developed the initial constructs and procedures for the expression of recombinant proteins. B.R.T. directed our homology modeling and virtual screening and R.J.R. assisted with the virtual screen. M.J.M., M.M. and L.B. provided guidance for the preparation of tGAPDHS crystals, collecting diffraction data and analyzed the crystal structures. D.A.O. is responsible for the experimental design and supervision of this project, including data analysis and manuscript preparation.

Funding

This work was supported by National Institutes of Health grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development: U01 HD060481 and cooperative agreement U54 HD035041 as part of the Specialized Cooperative Centers Program in Reproduction and Infertility Research. Also supported by National Institutes of Health Fogarty International Center grant D43 TW/HD00627 and subproject CIG-05-109 from CICCR, a program of CONRAD, Eastern Virginia Medical School, USA.

Conflict of interest

None declared.