Structure of Cryptosporidium IMP de­hydrogenase bound to an inhibitor with in vivo antiparasitic activity

Inosine 5′-monophosphate dehydrogenase (IMPDH) is a promising target for the treatment of Cryptosporidium infections. Here, the structure of C. parvum IMPDH (CpIMPDH) in complex with inosine 5′-monophosphate (IMP) and P131, an inhibitor with in vivo anticryptosporidial activity, is reported. P131 contains two aromatic groups, one of which interacts with the hypoxanthine ring of IMP, while the second interacts with the aromatic ring of a tyrosine in the adjacent subunit. In addition, the amine and NO₂ moieties bind in hydrated cavities, forming water-mediated hydrogen bonds to the protein. The design of compounds to replace these water molecules is a new strategy for the further optimization of C. parvum inhibitors for both antiparasitic and antibacterial applications.

1. Introduction

Cryptosporidium parasites are protozoan pathogens of increasingly recognized importance in both the developing and the developed world (Striepen, 2013). Cryptosporidiosis is a major cause of morbidity in infants and malnutrition in young children in Africa and Asia, and is a major cause of waterborne disease in North America and Europe. Infections are self-limiting in immunocompetent individuals, but can be chronic and fatal in immunocompromised patients (Checkley et al., 2014). In addition, Cryptosporidium oocysts are resistant to commonly used methods of water treatment and are readily available, making this organism a potential weapon for bioterrorism. The potential damage of such an incident was illustrated in the naturally occurring outbreak in Milwaukee in 1993, where 400 000 cases were reported and an estimated $96 million in damages were incurred.

Nitazoxanide is the only FDA-approved drug for the treatment of cryptosporidiosis (Checkley et al., 2014). It has limited efficacy in immunocompetent individuals and no efficacy in immunocompromised patients. High doses of paromomycin have anticryptosporidial activity in mice, where nitazoxanide is not effective (Tzipori et al., 1994; Baishanbo et al., 2006), but has not proved effective in humans (Hewitt et al., 2000). Thus, there is a critical need for new chemotherapy to treat Cryptosporidium infections.

Cryptosporidium parasites have a very streamlined purine-biosynthetic pathway that relies on the salvage of adenosine from the host (Striepen et al., 2004). Adenosine is converted into guanine nucleotides in a process that requires IMP dehydrogenase (IMPDH) activity. Curiously, the Cryptosporidium gene for IMPDH appears to have been obtained from bacteria via lateral transfer, so that the parasite enzyme is very different from the host ortholog. IMPDH is a clinically validated target for immunosuppressive, anticancer and antiviral therapy (Hedstrom, 2009). The IMPDH reaction involves two chemical transformations (Hedstrom, 2009). The first step is nucleophilic attack of the active Cys219 at the C2 position of IMP, followed by hydride transfer to form the covalent intermediate E-XMP* and NADH. The reduced cofactor dissociates from the enzyme and the mobile active-site flap (residues 302–330) folds into the cofactor site, positioning the catalytic Arg315 to activate water for the hydrolysis of E-XMP* to yield the product xanthosine 5′-monophosphate (XMP). IMPDH is a homotetramer with square-planar symmetry. The four active sites are located near the subunit interfaces. The IMP site is contained within a monomer, and the residues that contact IMP are strongly conserved among all IMPDHs (Hedstrom, 2009). The nicotinamide portion of the cofactor site also lies within the same monomer as IMP, and is also strongly conserved, as expected given that this is the site of chemical transformation. However, the position of the adenosine portion of the cofactor site varies widely among IMPDHs from different organisms. The adenosine site of eukaryotic IMPDHs is largely within the same monomer as the IMP site (Colby et al., 1999; Prosise & Luecke, 2003), but in prokaryotic IMPDHs the adenosine portion of NAD⁺ binds in a pocket in the adjacent subunit (Makowska-Grzyska et al., 2015).

We have been engaged in a program to design selective inhibitors of C. parvum IMPDH (CpIMPDH) as potential anticryptosporidial agents (Gorla et al., 2012, 2013, 2014; Johnson et al., 2013; Kirubakaran et al., 2012; MacPherson et al., 2010, Maurya et al., 2009; Sharling et al., 2010; Sun et al., 2014; Umejiego et al., 2004, 2008). Compound P131 is the most promising compound to arise from this effort (Fig. 1a), with superior activity to nitazoxanide and paromomycin in a mouse model of acute disease (Gorla et al., 2014).

Figure 1 (a) Structure of P131 and other cpIMPDH inhibitors. The numbering scheme is shown in italics. (b) 2mFo − DFc electron-density map for P131 contoured at the 1σ level. The inhibitor is shown in ball-and-stick representation. (c) Structure of the CpIMPDH tetramer. Subunit A is in tan, subunit B is in light blue, subunit C is in pink and subunit D is in light green. IMP and P131 are shown as spheres with the following color coding: carbon, cyan for IMP and magenta for P131; oxygen, red; nitrogen, blue; phosphorus, orange; chlorine, green. (d) Superposition of the individual monomers. Dashed lines indicate disordered regions. This figure was produced with UCSF Chimera (Pettersen et al., 2004).

The prokaryotic origin of CpIMPDH suggested that these inhibitors might also be effective against IMPDHs from pathogenic bacteria (Gollapalli et al., 2010). The CpIMPDH inhibitor collection does contain many potent inhibitors of Bacillus anthracis IMPDH, but only a few of these compounds possess antibacterial activity (Mandapati et al., 2014). The CpIMPDH inhibitors are generally more hydrophobic, with less polar surface area, than antibiotics (O'Shea & Moser, 2008), which suggests that they do not efficiently enter bacteria. Enzyme–inhibitor structures provide important information for the further modification of CpIMPDH inhibitors both for antiparasitic and antibacterial activities.

Here, we report the structure of the enzyme–IMP–P131 complex. This information will facilitate the further optimization of CpIMPDH inhibitors for the treatment of crypto­sporidiosis.

2. Materials and methods

Unless otherwise noted, all reagents and solvents were purchased from commercial sources and were used without further purification. Compound P131 was synthesized as described previously (Gorla et al., 2012).

2.1. Protein production and crystallization

The coding sequence for CpIMPDH was amplified by PCR and cloned into pMCSG7 vector (Stols et al., 2002) as described in Gorla et al. (2013). The recombinant construct produced a fusion protein with an N-terminal His₆ tag and a TEV protease recognition site (ENLYFQ↓S) (Table 1). The fusion protein was expressed in Escherichia coli strain BL21(DE3) harboring the pMAGIC plasmid encoding one rare E. coli Arg tRNA (covering the codons AGG/AGA) in the presence of 100 µg ml⁻¹ ampicillin and 30 µg ml⁻¹ kanamycin (Kim et al., 2011). The cells were grown in enriched M9 medium at 37°C followed by overnight induction with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 18°C. Cells were harvested, resuspended in lysis buffer (50 mM HEPES, pH 8.0, 500 mM KCl, 20 mM imidazole, 1.5 mM TCEP, 5% glycerol) and stored at −80°C. CpIMPDH protein was purified according to a standard protocol (Makowska-Grzyska et al., 2014). The protocol included cell lysis by sonication and nickel-affinity chromatography on an ÄKTAxpress system (GE Healthcare Life Sciences) followed by His₆-tag cleavage using recombinant TEV protease (Blommel & Fox, 2007) and additional Ni²⁺-affinity chromatography to remove the protease, the uncut protein and the affinity tag. In the final step, the protein was dialyzed against crystallization buffer consisting of 20 mM HEPES pH 8.0, 150 mM KCl, 1.5 mM TCEP, concentrated, flash-cryocooled and stored in liquid nitrogen. Crystallizations were set up with the help of a Mosquito liquid dispenser (TTP LabTech). Protein samples used for crystallization consisted of 13.4 mg ml⁻¹ protein in crystallization buffer (as above), 5 mM IMP, 1.5 mM P131. Diffraction-quality crystals appeared at 16°C in 0.1 M Bicine–NaOH pH 9.0, 20%(w/v) PEG 6000 (Table 2). The crystals were mounted on CryoLoops (Hampton Research) and flash-cooled in liquid nitrogen. The cryoprotectant consisted of 0.1 M Bicine–NaOH pH 9.0, 20%(w/v) PEG 6000, 15% glycerol.

Table 1
Protein-production information

Source organism	C. parvum Iowa II
CryptoDB ID	cgd6_20
CryptoDB sequence	MGTKNIGKGLTFEDILLVPNYSEVLPREVSLETKLTKNVSLKIPLISSAMDTVTEHLMAVGMARLGGIGIIHKNMDMESQVNEVLKVKNWISNLEKNESTPDQNLDKESTDGKDTKSNNNIDAYSNENLDNKGRLRVGAAIGVNEIERAKLLVEAGVDVIVLDSAHGHSLNIIRTLKEIKSKMNIDVIVGNVVTEEATKELIENGADGIKVGIGPGSICTTRIVAGVGVPQITAIEKCSSVASKFGIPIIADGGIRYSGDIGKALAVGASSVMIGSILAGTEESPGEKELIGDTVYKYYRGMGSVGAMKSGSGDRYFQEKRPENKMVPEGIEGRVKYKGEMEGVVYQLVGGLRSCMGYLGSASIEELWKKSSYVEITTSGLRESHVHDVEIVKEVMNYSK
CSGID Clone ID	IDP92622
Cloning/expression vector	pMCSG7
Expression host	E. coli BL21(DE3) pMAGIC†
Complete amino-acid sequence of the construct produced‡	HHHHHHENLYFQ^SNAMGTKNIGKGLTFEDILLVPNYSEVLPREVSLETKLTKNVSLKIPLISSAMDTVTEHLMAVGMARLGGIGIIHKNMDMESQVNEVLKVKNSGGLRVGAAIGVNEIERAKLLVEAGVDVIVLDSAHGHSLNIIRTLKEIKSKMNIDVIVGNVVTEEATKELIENGADGIKVGIGPGSICTTRIVAGVGVPQITAIEKCSSVASKFGIPIIADGGIRYSGDIGKALAVGASSVMIGSILAGTEESPGEKELIGDTVYKYYRGMGSVGAMKSGSGDRYFQEKRPENKMVPEGIEGRVKYKGEMEGVVYQLVGGLRSCMGYLGSASIEELWKKSSYVEITTSGLRESHVHDVEIVKEVMNYSK
Molecular weight (Da)	38459 (40077 with His6 tag)
Extinction coefficient§ (M^−1 cm^−1)/Abs 0.1% (= 1 g l^−1)	0.569 (0.583 with His6 tag)

†Kim et al. (2011).
‡The cloning tag is italicized, with the TEV protease site denoted `^'. This construct lacks residues 90–134 compared with the sequence deposited in CryptoDB. These residues were replaced with Ser-Gly-Gly (underlined).
§Calculated using ProtParam for the reduced protein sequence (Gasteiger et al., 2005).

Table 2
Crystallization

Method	Sitting-drop vapor diffusion
Plate type	Greiner Bio-One 96-well CrystalQuick standard profile, 3 round wells, flat bottom
Temperature (K)	289
Protein concentration (mg ml^−1)	13.4
Buffer composition of protein solution	20 mM HEPES pH 8.0, 150 mM KCl, 1.5 mM TCEP
Composition of reservoir solution	0.1 M Bicine–NaOH pH 9.0, 20% PEG 6000
Cryocondition	0.1 M Bicine–NaOH pH 9.0, 20% PEG 6000, 15% glycerol
Volume and ratio of drop	1:1 (0.4:0.4 µl)
Volume of reservoir (µl)	135

2.2. Data collection and structure determination

Diffraction data were collected at 100 K on the 19-ID beamline of the Structural Biology Center (SBC) at the Advanced Photon Source (APS), Argonne National Laboratory (Rosenbaum et al., 2006). Data were collected at a single wavelength of 0.97931 Å (12.6603 keV) to 2.05 Å resolution from a single crystal of CpIMPDH complexed with IMP and P131. The crystal was exposed for 3 s per 1.0° rotation in ω at a crystal-to-detector distance of 300 mm. The data were recorded on an ADSC Q315r CCD detector scanning 190°. The SBC-Collect program was used for data collection and visualization. Data collection, integration and scaling were performed with the HKL-3000 program package (Minor et al., 2006). A summary of the crystallographic data is presented in Table 3.

Table 3
Data-collection and processing statistics

Values in parentheses are for the highest resolution shell.

No. of crystals	1
Beamline	SBC 19-ID, APS
X-ray wavelength (Å)	0.97931
Temperature (K)	100
Detector	ADSC Quantum 315r
Crystal-to-detector distance (mm)	300
Total rotation range (°)	190
Rotation per image (°)	1.0
Exposure time per image (s)	3
Space group	P21
Unit-cell parameters (Å, °)	a = 89.19, b = 91.83, c = 92.02, α = 90, β = 103.75, γ = 90
Resolution range (Å)	40.57–2.05 (2.09–2.05)
Total No. of observations	349609
No. of unique reflections	89862 (4514)
〈I/σ(I)〉	12.66 (2.02)
Data completeness (%)	99.9 (99.6)
Rmerge† (%)	14.0 (69.5)
Rp.i.m.(I) (%)	7.9 (43.4)
Wilson B factor (Å^2)	27.5
Multiplicity	3.9 (3.3)
Mosaicity (°)	0.503–1.367

†R_merge = , where I_i(hkl) is the intensity of the ith measurement of an equivalent reflection with indices hkl.

The structure was determined by molecular replacement using the structure of CpIMPDH (PDB entry 3ffs ; MacPherson et al., 2010) as a search model with the HKL-3000 suite using data to 3.0 Å resolution (Bricogne et al., 2003; Minor et al., 2006). The initial model contained four protein chains of the search model, and there was extra electron density for additional protein residues that were not part of the search model. The presence of IMP and P131 in the active site was apparent from the initial electron-density map (F_o). Extensive manual model building with Coot (Emsley & Cowtan, 2004) and subsequent refinement using phenix.refine (Bricogne et al., 2003; Afonine et al., 2012) were performed against the full data set to 2.05 Å resolution until the structure converged to an R factor (R_work) of 0.187 and an R_free of 0.229 with an r.m.s.d. on bond distances of 0.004 Å and an r.m.s.d. on bond angles of 0.864°. The asymmetric unit contained four protein chains (A, B, C and D), with residues 1–392 for all chains and a Ser-Gly-Gly linker inserted in place of residues 90–134 (MacPherson et al., 2010). Several N- and C-terminal amino acids, including the N-terminal Ser-Asn-Ala introduced during cloning (Eschenfeldt et al., 2010), were missing because of disorder. In addition, several residues within the active-site flap were disordered and were not modeled. These included residues 311–324 in chains A, B and D and 310–325 in chain C. The final model also contained four IMP molecules, four P131 molecules, five ethylene glycol molecules, one glycerol molecule and 340 ordered water molecules. The stereochemistry of the structure was checked with PROCHECK (Laskowski et al., 1993) and a Ramachandran plot. Refinement statistics are shown in Table 4. Atomic coordinates and experimental structure factors of the structure have been deposited in the PDB as entry 4rv8 .

Table 4
Refinement statistics

Values in parentheses are for the highest resolution shell.

Resolution range (Å)	40.57–2.053 (2.076–2.053)
Reflections used (working/free)	85300/4499 (2661/139)
Data completeness (%)	99.6 (93.0)
R factor†/Rfree (%)	18.7/22.9 (26.7/31.8)
Total No. of non-H atoms in asymmetric unit	10414
No. of protein atoms	10312
No. of heteroatoms
P131	120
IMP	92
Glycerol	24
PEG	14
No. of water molecules	339
R.m.s.d. from ideal geometry
Bond lengths (Å)	0.004
Bond angles (°)	0.864
Average B factors (Å^2)
Overall	36.1
Protein atoms
All	36.0
Main chain	32.4
Side chain	38.4
Water molecules	40.4
Ligand atoms
P131	39.4
IMP	26.1
Solvent/cryoprotectant molecules
Glycerol	62.9
PEG	65.4
Ramachandran statistics (%)
Most favored	97.3
Outliers	0.0
MolProbity scores
Rotamer outliers	0.9
Clashscore	2.59
Overall score	1.15
PDB code	4rv8

†R = , where F_obs and F_calc are observed and calculated structure factors, respectively.

3. Results and discussion

The structure of CpIMPDH in complex with IMP and P131 was solved at 2.05 Å resolution using molecular replacement with the structure of apo CpIMPDH (PDB entry 3ffs ) as the search model (Table 4). Density for P131 is clearly visible in all four subunits (Fig. 1b). The unit cell contains a single tetramer with D4 symmetry (Fig. 1c). The four subunits are nearly identical, with r.m.s.d.s ranging between 0.15 Å (chain A versus chains B, C and D) and 0.16 Å (chain C versus chains B and D) (Fig. 1d). Residues 311–324, which are part of the active-site flap (see below), and residues 378–387 near the C-terminus are disordered in all four subunits. In addition, residues 310 and 325 are disordered in chain C.

The conformation and interactions of IMP in the P131 complex are essentially identical to those observed in the previously reported complexes of CpIMPDH with C64, Q21 and N109 (Fig. 2a; MacPherson et al., 2010; Gorla et al., 2013; Sun et al., 2014), as well as those in the complexes of Bacillus anthracis, Campylobacter jejuni and Clostridium perfringens IMPDHs with various CpIMPDH inhibitors (Makowska-Grzyska et al., 2015).

Figure 2 Substrate-binding and cofactor-binding sites. (a) The IMP site of CpIMPDH in complex with inhibitors. Residues in contact with IMP are shown. P131, PDB entry 4rv8 (pink); C64, PDB entry 3khj (green; MacPherson et al., 2010); Q21, PDB entry 4ixh (blue; Gorla et al., 2013); N109, PDB entry 4qj1 (tan; Center for Structural Genomics of Infectious Diseases, Sun et al., 2014). The B subunit is shown in all four cases. (b) Superposition of the CpIMPDH–IMP–P131 complex (AD dimer, pink) with the IMP–NAD^+ complex of V. cholerae IMPDH (AA dimer, tan; PDB entry 4qne ; Center for Structural Genomics of Infectious Diseases, Makowska-Grzyska et al., 2015). P131 is shown in stick representation with C atoms in magenta. NAD^+ is shown in ball-and-stick representation with C atoms in cyan. The residues that contact NAD^+ are shown, with the exceptions of IMP, Cys219 and Met302-Gly303, which were omitted for clarity (CpIMPDH numbering). This figure was produced with UCSF Chimera (Pettersen et al., 2004).

As observed with other CpIMPDH inhibitors, P131 binds in the cofactor site, spanning both the nicotinamide and adenosine subsites (Fig. 2b). The left-side aromatic ring (Fig. 1a) binds in the nicotinamide subsite, which is a pocket formed by IMP and residues Ser164-Ala165, Asn191, Gly212–Gly214, the catalytic Cys219, Thr221, Met302-Gly303, Met308, Glu329 and Tyr358′ (where a prime denotes a residue from the neighboring subunit). The electron density indicates that Cys219 has oxidized to the sulfenic acid, as often occurs with reactive cysteine residues. Similar oxidation of the catalytic Cys was observed in the structures of Tritrichomonas foetus IMPDH (Prosise & Luecke, 2003; Prosise et al., 2002). Despite this modification, the position of Cys219 is very similar to that observed in other CpIMPDH–inhibitor complexes (Fig. 3a). The oxime group of P131 lies across the hypoxanthine ring of IMP, while the N5 amine forms hydrogen bonds to the main-chain carbonyl O atom of Gly212 and N3 of IMP in all of the subunits (Fig. 3). The amine also interacts with water molecules to create a hydrogen-bonding network that extends to the 2′-OH of IMP, the sulfenic acid of Cys219, Ala165 N, Gly212 O, Gly214 N, Thr221 OG, Asp252 O and Tyr358 OH′. While the protein and inhibitor conformations are nearly identical in this region in all four subunits, the water occupancy varies slightly. WAT213, WAT282 and WAT316 (numbering in chains A, B and D, respectively) form hydrogen bonds to Tyr358 OH′ and Ala165 O in chains A, B and D, but this water is absent in chain C (Fig. 2b). Curiously, despite these many interactions, the presence of the alkylamine decreases the inhibitor potency by a factor of ∼30 versus the corresponding oxime (IC₅₀ = 0.66 nM for P96 versus 20 nM for P131; Gorla et al., 2012).

Figure 3 Interactions of P131 with CpIMPDH. (a) Interactions of the oxime and alkylamine. P131 is displayed in ball-and-stick representation and the protein residues are in ribbon representation with side chains depicted as sticks. Hydrogen bonds are shown in cyan. The dimer formed by subunits A and D is shown in tan and the corresponding waters are shown in brown. The BA dimer is shown in light blue and the corresponding waters are shown in dark blue. The CB dimer is shown in pink and the corresponding waters are shown in magenta. The DC dimer is shown in light green and the waters are shown in dark green. Residues within 5 Å of P131 are shown, with the exceptions of Asn191, residues 327–329 and the side chains of His166, Ile213 and Asn252, which have been omitted for clarity. (b) Interactions of the urea and the right side of P131 in the AD, CB and DC dimers. The color scheme is as above. (c) Interactions of the urea and the right side of P131 in the BC dimer. The color scheme is as above. (d) Interaction of the P131 Cl atom with the carbonyl O atom of Gly357′. The color scheme is the same as above. Hydrogen bonds are shown in cyan. The halogen bond is shown as a black dashed line. This figure was produced with UCSF Chimera (Pettersen et al., 2004).

Also as observed with other CpIMPDH inhibitors (MacPherson et al., 2010; Gorla et al., 2013; Sun et al., 2014), P131 bends in the linker region so that the plane of the right-side aromatic moiety is almost perpendicular to the left-side aromatic moiety (Figs. 1d, 3b and 3c). Both urea N atoms of P131 form hydrogen bonds to Glu329 OE2 in all four subunits (Fig. 3b and 3c). Analogous interactions with Glu329 are observed with other CpIMPDH inhibitors. In addition, water-mediated hydrogen bonds are found between the urea N1 atom and Val327 O as well as between the urea carbonyl O atom and Ala165 N.

The right-side aromatic ring binds in the adenosine subsite formed by residues Ser22′, Val24′-Leu25′-Pro26′, Ser164-Ala165-His166, Ser354′ and Gly357′-Tyr358′ (Makowska-Grzyska et al., 2015). As noted above, this subsite is highly divergent, accounting for the selectivity of the CpIMPDH inhibitors. The analogous residues in human IMPDH type 2 are Ile42′, Phe44′-Thr45′-Ala46′, Ser275-Ser276-Gln277, His466 and Gln469′-Asp470′. As observed with the primary amine, the NO₂ group binds in a hydrophilic pocket. In subunits A, C and D, the NO₂ group is roughly parallel to the aromatic ring and forms water-mediated hydrogen bonds to Ser22 O′, Val24 O′, Ser164 OG, His166 O, Ser169 O and Asn171 N and ND2. In contrast, the NO₂ group is perpendicular to the ring in subunit B and forms a hydrogen bond directly to Ser164 OG (Figs. 2c and 2d).

The P131 Cl atom interacts with the carbonyl O atom of Gly357′ (Fig. 3d; average distance 3.33 ± 0.12 Å), which suggests the presence of a halogen bond (expected distance 3.0–3.40 Å; Bissantz et al., 2010). The C—Cl⋯O and Cl⋯O—C angles are compatible with halogen-bond formation with the π system of the carbonyl acting as the donor (128 and 82°, respectively, versus C—Cl⋯O ≥ 120° and 180° ≥ Cl⋯O—C ≥ 70°; Bissantz et al., 2010; Voth et al., 2009; Vallejos et al., 2012). The Gly357′ carbonyl is further polarized by hydrogen bonds to Ser22 OG′ (average distance 2.85 ± 0.06 Å) and Gly360 N′ (average distance 3.19 ± 0.06 Å). The Gly357′–Ser22 OG′ hydrogen bond is at approximately 90° relative to the halogen bond, similar to the angle between other pairs of halogen and hydrogen bonds.

The water-mediated hydrogen-bonding networks of both the left-side alkylamine and the right-side NO₂ group offer many opportunities for further inhibitor optimization. In particular, the repurposing of CpIMPDH inhibitors as antibacterial agents requires the introduction of additional polarity (Mandapati et al., 2014), which might be attained by extending the inhibitor to replace water in these networks. Efforts in this area are ongoing.