X-ray crystal structure and specificity of the Toxoplasma gondii ME49 TgAPN2.

Toxoplasmosis is a parasitic disease caused by infection with Toxoplasma gondii that currently has few therapeutic options. The M1 aminopeptidase enzymes have been shown to be attractive targets for anti-parasitic agents and/or vaccine candidates, suggesting potential to re-purpose inhibitors between parasite M1 aminopeptidase targets. The M1 aminopeptidase TgAPN2 has been suggested to be a potential new drug target for toxoplasmosis. Here we investigate the structure and function of TgAPN2, a homologue of the antimalarial drug target PfA-M1, and evaluate the capacity to use inhibitors that target PfA-M1 against TgAPN2. The results show that despite a similar overall fold, the TgAPN2 has a unique substrate specificity and inhibition profile. Sequence and structure differences are investigated and show how comparative structure-activity relationships may provide a route to obtaining potent inhibitors of TgAPN2.

Introduction

In Plasmodium, Trypanosome, Eimeria and Cryptosporidium parasites, the clan MA family M1 aminopeptidases have been investigated as potential targets for new anti-parasitics and/or vaccine development [1]. M1 aminopeptidases, also known as aminopeptidase N, are generally membrane-anchored metallo-exopeptidases found throughout all kingdoms of life, and have a wide range of functions including cell maintenance, growth and development, and defence [1]. In the apicomplexan malaria parasite Plasmodium falciparum, there is a single M1 aminopeptidase, known as PfA-M1 [2]. PfA-M1 has been shown to be important for the final stages of haemoglobin digestion, an essential metabolic pathway required for blood-stage Plasmodium growth and development [3]. Inhibition of PfA-M1 activity in parasites has been shown to control both laboratory (P. falciparum) and murine (Plasmodium chaubaudi chaubaudi) models of malaria [4].

Toxoplasmosis is caused by infection of Toxoplasma gondii which, like Plasmodium, Eimeria and Cryptosporidum is an apicomplexan parasite. Toxoplasmosis in healthy individuals is generally asymptomatic, however, serious complications can arise in immunocompromised patients and pregnant women [5]. Congenital infection can lead to miscarriage, mental retardation, vision and hearing problems [5]. Preventing toxoplasmosis in immunocompromised patients relies on chemotherapy generally consisting of pyrimethamine with either sulfadiazine or clindamycin [6,7]. Agents to treat active and latent disease remain scarce and new molecular targets for therapeutics are desirable for future drug discovery.

The T. gondii ME49 strain contains three canonical M1 aminopeptidases, TgAPN1 (TGME49_221310), TgAPN2 (TGME49_224350) and TgAPN3 (TGME4_224460) [8], and their respective structural genes are located on different chromosomes [9]. TgAPN1 has previously been experimentally shown to be active as a metallo-aminopeptidase, as well as being immunogenic — suggesting an involvement in infection response [10]. TgAPN3 is also functional as a metallo-exopeptidase with a substrate preference for Ala > Tyr/Cys > Arg but interestingly has a signal peptide in place of the normal N-terminal transmembrane anchor. Recent studies into the localisation of TgAPN3 [9] showed that the protease was co-located with the GRA protein in the organelle and parasitophorous vacuole, and could be secreted with tachyzoites [9]. TgAPN2 is predicted to be localised within the parasite cytosol (https://toxodb.org [11]), and is expressed throughout the life cycle of the parasite, including during the tachyzoite stage where the parasite undergoes acute reproduction [11,12] (https://toxodb.org). The biological role(s) of TgAPN2 remain unclear but a genome wide CRISPR screen of the T. gondii reference strain, showed the loss of TgAPN2 resulted in a fitness cost of −1.21 in culture, meaning that it is important but probably not essential to tachyzoite growth [13]. The transcript is, however, up-regulated in bradyzoites, which is a more important stage for drug development [13].

Out of the three T. gondii M1 aminopeptidases, TgAPN2 shares the highest sequence similarity to that of PfA-M1 (38%) and as such, has been suggested to be target worthy of further investigation as a novel drug target [14]. Biochemical characterisation of recombinant TgAPN2, however, showed a substrate preference for a P1 Arg > Ala > Leu, suggesting significant changes in the substrate specificity pockets between TgAPN2 and PfA-M1 [14,15]. In this study, we aimed to produce structure activity relationship (SAR) data of a selection of M1 aminopeptidase inhibitors with the aim to map the similarities and differences in SAR between PfA-M1 and TgAPN2. The identification of a potent inhibitor of enzymatic activity would provide a valuable tool to probe the biological roles of TgAPN2 in T. gondii, as well as assess the capacity of the enzyme to act as a drug target. Our findings show that TgAPN2 displays a significantly different substrate specificity to that of PfA-M1. Additionally, inhibition studies indicate that there are significant changes in inhibitory kinetics likely due to differences within the S1 substrate specificity pocket of TgAPN2, as characterised in the crystal structure.

Experimental

Expression and purification of recombinant ecto-domain of TgAPN2

DNA coding for residues 520–1419 of TGME49_224350 from T. gondii ME49 (UniProt S8G5K8) with an in-frame C-terminal His6 tag was chemically synthesised by DNA2.0. Potential glycosylation sites at N840 and N1303 were altered to encode glutamine. The gene was provided in the pJ404 vector for expression, which also encodes for ampicillin resistance. The construct was transformed into BL21 (DE3) cells and protein expressed using an auto-induction method based on previously described protocols [16]. A 200 ml culture was grown for 24 h at 28°C followed by a further 24 h at 20°C. Cells were harvested and resuspended in PBS pH 8.0, 0.3 M NaCl, 5% (v/v) glycerol prior to lysis by sonication. Clarified lysates were bound to a Ni-NTA-agarose column in nickel-affinity buffer (PBS pH 8.0, 300 mM NaCl, 5% glycerol, 20 mM imidazole), and eluted in nickel-affinity buffer supplemented with 250 mM imidazole. The protein was further purified by size exclusion chromatography on a Superdex 200 16/60 using an AKTAxpress high throughput chromatography system in 50 mM Hepes pH 8.0, 0.3 M NaCl, 5% (v/v) glycerol. The major peak, containing TgAPN2, was pooled and concentrated to 5 mg/ml. The purity of the protein was assessed by SDS–PAGE and protein concentration determined using Pierce™ BCA protein assay.

Enzymatic analysis

Aminopeptidase activity was determined by measuring the hydrolysis of 7-amido-methylcoumarin (Mec) from the fluorogenic substrate L-leucine-7-amido-4-methylcoumarin hydrochloride (Leu-Mec) (Sigma L2145) or L-arginine-7-amido-4-methylcoumarin hydrochloride (Arg-Mec) (Sigma A2027). Reactions were carried out in 96-well microtitre plates (100 µl total volume, 60 min, 37°C) using a spectrofluorimeter (BMG LabTech Optima) with excitation at 355 nm and emission at 460 nm. For determination of kinetic parameters, enzyme was first added to 100 mM Tris–HCl, pH 8.0 for 10 min prior to the addition of substrate. Initial rates were obtained at 37°C over a range of substrate concentrations spanning KM (0.5–500 μM) and at fixed enzyme concentration (20 nM). Calculations of Michaelis–Menten constants (KM) and rates (kcat) were performed using GraphPad Prism. Observation of the rate of substrate hydrolysis under chelating and different pH conditions was performed by alteration of the enzyme buffer by either the addition of EDTA (0–200 mM) or alteration of pH (pH 5–9) of the buffer prior to enzyme addition. Assays were performed with fixed enzyme concentration (20 nM) and fixed substrate concentration (15 μM). Statistical significance was calculated using a Mann–Whitney test, comparing the mean of the slope velocity (fluorescence per sec, n = 6) between the different pH conditions. Ki(app) values were calculated as described for PfA-M1 [17], where inhibitor (0–1 mM) and enzyme were pre-incubated in 100 mM Tris–HCl, pH 8.0 for 20 min prior to the addition of substrate. Substrate concentration (15 μM) was selected to allow sensitive detection of enzyme activity while not exceeding the KM. Enzyme concentration was fixed (20 nM). The Ki(app) values were calculated by plotting the initial rates versus inhibitor concentration, and fitting to the Morrison equation in GraphPad Prism (non-linear regression method).

Substrate-specificity profiling was achieved by the use of a fluorogenic substrate library containing 63 amino acids [15]. For convenience in solid-phase synthesis, we employed a 7-amino-4-carbamoylmethylcoumarin (ACC) fluorogenic leaving group in this library. The final screening of the library was carried out at 10 µM substrate and 0.5 nM enzyme as described above and release of free ACC fluorophore was monitored. Each experiment was repeated at least three times and the average value with standard error was calculated. Concentration of DMSO in the assay was <1% (v/v).

Crystallisation and structure refinement

TgAPN2 was concentrated to 5, 10 and 13.5 mg ml−1 for initial sparse-matrix screening. The Monash Macromolecular Crystallization Facility was utilised to set up sitting-drop sparse matrix screens (INTELLI-PLATE 96-3, Art Robbins Instruments). Crystals were observed in seven different conditions. Two conditions were selected for optimisation/screening by the hanging drop method: (1) 15–25% PEG 8000, 0.1 M Hepes pH 6.5–8.0 and (2) 20–30% PEG 3350, 0.1 M Bis-Tris pH 5.0–7.0. Each tray was set with 6, 8 and 10 mg ml−1. Crystals were reproduced in both conditions and the best diffraction was observed from screening a crystal grown from 6 mg ml−1 in the PEG3350/Bis-Tris screen. Final crystals of TgAPN2 were obtained with a protein concentration of 5 mg ml−1 in 28% PEG 3350, 0.1 M Bis-Tris pH 5.0. Cryo-protection was achieved by briefly soaking single crystals in the reservoir containing 20% glycerol.

A single TgAPN2 crystal was harvested (CryoLoopTM, Hampton Research) and flash cooled in liquid nitrogen. X-ray diffraction data were collected at 100 K using synchrotron radiation on the MX2 beamline (3BM1) at the Australian Synchrotron. Data were indexed and integrated using XDS [18], and scaled and averaged in AIMLESS [19] from the CCP4 suite [20]. Initial phases were solved by molecular replacement using PHASER [21] and 3EBG as a search model [22]. Model building and refinement were performed using Phenix [23] and Coot [24]. Initial refinement made use of non-crystallographic symmetry until model was near completion. Final data collection and refinement statistics are shown in Table 1. The coordinates and structure factors are available from the Protein Data Bank (PDB ID: 6OIU). Analysis of the surface was performed using PDBePISA [25]. All crystal structure figures and analyses were prepared in the Pymol Molecular Graphics System, version 1.3r2 (Schrödinger, LLC). Structure and sequence comparisons were carried out using PDBeFold [26], PDBePISA [25] Stride [27] and Clustal Omega[28,29].

Table 1

Data collection and refinement

Data collection
Diffraction source	MX2 beamline, AS
Wavelength (Å)	0.9537
Temperature (K)	100
Detector	CCD Quantum 201r (ADSC)
Crystal-to-detector distance (mm)	250
Rotation range per image (°)	0.5
Total rotation range (°)	0–90
Exposure time per image (s)	1.0
Space group	P 1 21 1
a, b, c (Å)	92.93, 207.87, 102.87
	90, 94.34, 90
Mosaicity (°)	0.17
Resolution range (Å)	38.54–2.20 (2.28–2.20)
Total No. of reflections	741 396 (35 716)
No. of unique reflections	195 801(19 286)
Completeness (%)	99.6 (98.1)
Multiplicity	3.8 (3.8)
	5.3 (1.4)
R-merge (%)	0.205 (1.062)
R-pim (%)	0.122 (0.629)
CC1/2	0.979 (0.433)
Overall B factor from Wilson plot (Å2)	23.92
Refinement
R-work	0.1904 (0.2641)
R-free	0.2357 (0.3160)
non-hydrogen atoms	30 396
Macromolecules	27 586
Ligands	62
Water	2748
Protein residues	3575
RMS(bonds)	0.002
RMS(angles)	0.41
Ramachandran favoured (%)	97.47
Ramachandran outliers (%)	0.11
Clashscore	2.71/100th percentile[1]
Average B-factor	29.26
Macromolecules	28.99
Ligands	39.46
Solvent	31.73
PDB ID	6OIU

Values in parentheses are for the outer shell.

Molprobity N = 456, 2.200 Å ± 0.25 Å.

Results

Expression and purification of TgAPN2

M1 aminopeptidases are membrane anchored enzymes that can be anchored either intra- or extracellularly [1]. TgAPN2 is confidently predicted to be present in the cytosol (Probability MAP/MCMC 0.98/1, https://toxodb.org [11]) and has previously been shown to be a functional aminopeptidase [14]. In contrast with the previous biochemical study that — on the basis of amino acid sequence — suggested TgAPN2 is homohexameric [14], our analysis indicated that TgAPN2 is likely monomeric. The TgAPN2 structural gene encodes a 1419 amino acid protein (predicted molecular weight 156.7 kDa) that has an identifiable transmembrane region ∼300 amino acids from the N-terminus (Supplementary Figure S1) as is also observed in the other T. gondii M1 aminopeptidases, as well as to PfA-M1 (Supplementary Figure S1). This arrangement suggests an N-terminal domain that would be located on the opposite side of the membrane from the enzymatic domain. Sequence analysis of the first 319 amino acids shows significant regions of low-complexity, and analysis using the D2P2 server [30] confirms regions of intrinsic disorder (Supplementary Figure S2). PSI-Blast, and fold and function assignment failed to identify any homologues of this domain with known structure or function [31,32].

As we were interested in characterising the aminopeptidase function of TgAPN2, we chose to produce the soluble ecto-domain, excluding the N-terminal and transmembrane domains from our expression construct. The first residue of our truncated TgAPN2 (residues 520–1419) aligned with the beginning of truncated PfA-M1 (Supplementary Figure S1) [4]. Truncated TgAPN2 was produced using a bacterial expression system and purified from the soluble fraction using a two-step purification facilitated by the inclusion of an in-frame C-terminal hexa-histidine tag. Analytical size exclusion chromatography indicated that the purified, truncated TgAPN2 is a monomer with an expected molecular weight of 117 kDa (Figure 1).

Figure 1.

Expression and purification of TgAPN2.

(A) Analytical size exclusion chromatography of TgAPN2 using an S200 10/300 column and the standard curve showing the calculated molecular weight of purified TgAPN2 (red point). (B) An SDS–PAGE gel of purified TgAPN2.

Activity and substrate specificity of TgAPN2

To confirm if our purified recombinant TgAPN2 could function as an aminopeptidase, we used a fluorescence based assay with the fluorogenic substrate, L-Leucine-7-amido-4-methylcoumarin hydrochloride (Leu-Mec) [4]. The results show a slow enzyme possessing Michaelis–Menten kinetics with a KM of 31.8 ± 3.0 µM and a turnover rate of 1.32 s−1 (Figure 2A, and Table 2). Given the slow kinetics for Leu-Mec and previous study identifying Arg as the preferred P1 residue [14], we also determined the kinetic parameters toward L-arginine-7-amido-4-methylcoumarin hydrochloride (Arg-Mec) (Table 2 and Figure 2A). In comparison with a P1 Leu (Leu-Mec), the results suggest that TgAPN2 has a weaker substrate affinity for Arg (KM of 148 µM), but a significantly higher turnover rate (kcat of 1017 s−1). To investigate the metal-ion dependency of TgAPN2, we monitored aminopeptidase activity of TgAPN2 at increasing concentrations of EDTA and observed a reduction in enzyme velocity as EDTA concentration increased (Figure 2B). These results confirmed that, like other M1 aminopeptidases, TgAPN2 activity is metal-dependent. The pH dependency of the recombinant enzyme was investigated and found to be most active at pH 8.0, but retained majority of its activity at a neutral pH (Figure 2C). Changing the pH to either 6.0 or 9.0 resulted in a statistically significant decrease in the activity of TgAPN2 (P value of 0.0087 and 0.015, respectively) (Figure 2C).

Figure 2.

Aminopeptidase activity and P1 substrate specificity of TgAPN2.

(A) Michaelis–Menten plot for TgAPN2 against substrates Leu-Mec and Arg-Mec. (B) The effect of increasing concentration of EDTA (x-axis) on the velocity of TgAPN2 (y-axis). (C) Enzyme velocity of TgAPN2 at different pH conditions. Statistical significance was calculated using a Mann–Whitney test and comparing the mean (n = 6) of the velocity (fluorescence units per sec) at different pH conditions to pH 8.0. (D) Normalised activities of TgAPN2 against the extended substrate library where the x-axis represents the abbreviated amino acid names and y-axis represents the normalised relative activity relative to unnatural amino acid substrate hArg.

Table 2

Kinetic parameters of TgAPN2

Substrate	KM (µM)	kcat (s−1)	kcat/KM (s−1 µM−1)
L-Leu-Mec	31.8 ± 3.0	1.32 ± 0.03	0.0415
L-Arg-Mec	148 ± 5.1	1017 ± 15.2	6.87

To determine the P1 substrate specificity, we screened the activity of TgAPN2 against a fluorogenic peptide substrate library that incorporates 19 natural and 44 unnatural amino acids [15]. A total of 63 substrates were monitored for fluorescence post-cleavage, and the results confirmed that TgAPN2 strongly prefers a P1 Arg (Figure 2D). Homo-arginine (hArg), containing four carbons in the side chain, was the most active and the data was normalised by defining the activity against hArg as 100%. A natural Arg residue showed 81% activity, whilst activity toward Lys was 52% (Figure 2D). These results also correlate with the observed activity of TgAPN2 toward the P1 Leu as in our library screen the enzyme was only able to achieve 26% activity (Figure 2D). Alanine was able to be hydrolysed (∼20%) however other small hydrophobic residues such as valine, serine, proline or threonine showed no measurable activity (Figure 2D). Interestingly, P1 phenylalanine and tyrosine were digested poorly (<10% activity) but non-natural substitutes with longer aliphatic groups were effectively digested, (hTyr 54% and hPhe 18%), suggesting the extra carbon may allow the residues to occupy a more favourable position within the enzyme (Figure 2D).

X-ray crystal structure of TgAPN2 aminopeptidase domain

To investigate the structural basis for the substrate specificity, we solved the X-ray crystal structure of truncated TgAPN2 to 2.2 Å. A total of four TgAPN2 monomers were placed in the asymmetric unit, however, alignment of the four chains indicates that all are essentially identical (RMSD over 901 Cα atoms is 0.18–0.34 Å between chains A–D). Chain C in the asymmetric unit was the most complete, with only two residues not resolved in the density (1208–1209). Chain D was missing residues 520–521, and chains A and D also lacked continuous density for residues 1204–1210. The structural description and figures in this paper relate to chain C.

The final model confirmed that TgAPN2 is a monomer with a standard M1 aminopeptidase ecto-domain fold of four domains with a total of seven α-sheets and 24 α-helices (Figure 3A). The four domains consist of an N-terminal domain I (residues 520–731), the catalytic domain II (residues 732–980), a small Ig-like fold for domain III (residues 981–1080) and an entirely helical C-terminal domain IV (residues 1081–1419) (Figure 3A). The overall structure was in a closed and compact confirmation with the helical domain IV packed against the catalytic domain II (Figure 3A). Analysis of the structure shows that the tertiary structure of TgAPN2 is highly consistent with that of the clan MA, family M1 aminopeptidases (Supplementary Figure S3). The bacterial and parasite homologues showed the highest structural similarity, with an overall RMSD (Cα atoms) of 1.2 Å to the E. coli aminopeptidase N (PDB ID: 2DQ6 [33]) and 1.6 Å to PfA-M1 (PDB ID 3EBG [4]) (Supplementary Figure S3A). Comparison to the structures of human aminopeptidase N (APN, PDB ID: 5LHD [34]) and endoplasmic reticulum aminopeptidase 1 (ERAP1, PDB ID: 2YD0 [35]) when in a similar conformation to that of TgAPN2 share an overall RMSD (Cα atoms) of 3.5 Å and 2.3 Å, respectively (Supplementary Figure S3B). In particular, the N-terminal domain shows significant differences between TgAPN2 and human APN and ERAP1. The most noticeable difference is a small insertion that forms a helical loop in TgAPN2 domain I (residues 590–601) that is not present in the homologues (Supplementary Figure S3).

Figure 3.

The X-ray crystal structure of TgAPN2 (PDB ID: 6OIU).

(A) Cartoon representation of chain C of TgAPN2, showing each of the four domains, domain I (residues 520–731 in green), catalytic domain II (residues 732–980 in red), domain III (residues 981–1080 in blue) and domain IV (residues 1081–1419 in purple), and the zinc ions (grey spheres). (B) The electron density 2Fo–Fc maps contoured at one sigma showing density over the two zinc ions within the active site. Zinc coordinating residues (His835, His839 and Glu858) and Met801 and Glu802 of the conserved G799AMEN motif are shown as sticks with red carbon atoms.

The active site of TgAPN2 in domain II contains the two conserved M1 aminopeptidase motifs, the zinc-binding motif (H835EYFHX17KE858) and substrate binding G799AMEN motif (Figure 3B). Similar to other M1 aminopeptidases, a zinc ion is coordinated by a catalytic triad made up of the side-chains of a His835, His839, Glu858 (Figure 3B). Generally in unbound structures of M1 aminopeptidases, the tetrahedral zinc coordination state is completed by a water molecule in the active site cavity [36,37]. However, during the refinement of our structural model, it became evident that a water molecule would not satisfy the density observed in this position (Supplementary Figure S4). Individual refinement of a chloride, magnesium, nickel and zinc ion in this position indicated that the zinc and nickel were able to satisfy the density (Supplementary Figure S4). The second ion forms a bond with the active zinc ion (distance 2.3 Å, Supplementary Figure S4B) whilst the remaining coordination of the second ion results from longer bonds (2.9–3.7 Å) with residues Glu802, Glu836, Tyr919 and a water molecule (Supplementary Figure S4). A glycerol molecule is also present in this area (Supplementary Figure S4). Whilst the identification of a potential second metal site was unexpected, subsequent alignment of the coordinates of TgAPN2 with the unbound structure of PfA-M1 (PDB ID 3EBG [4]) showed that there is a Mg2+ ion in the same position of PfA-M1. The Mg2+ ion is suggested to be a result of high concentrations of Mg2+ in the crystallisation conditions of 3EBG [4]. For TgAPN2, there were no supplemental ions in the crystallisation buffers suggesting that the metal ion may have come from the nickel resin used to purify the protein or is a zinc ion that has likely been scavenged during purification. The final coordinates of TgAPN2 (6OIU.pdb) have been modelled with a zinc ion.

The substrate specificity pockets possess a similar architecture to that of PfA-M1 but sequence changes dramatically alter the local environment (Supplementary Figure S5A). The main wall of the S1 pocket is formed by the G799AMEN803 motif (Supplementary Figure S5A) which is highly conserved in M1 aminopeptidases, forming part of the S1 pocket and stabilising substrate prior to hydrolysis. Immediately preceding Gly799, TgAPN2 has a Met798 residue that provides the pocket with a more amphipathic nature than the equivalent hydrophobic valine in PfA-M1 (Supplementary Figure S5A). At the top of the S1 pocket, further sequence variation may account for substrate differences between PfA-M1 and TgAPN2, where TgAPN2 has a Gln1362 in place of the Met1034 in PfA-M1, and an Asp911 instead of the flexible Glu572 in PfA-M1 (Supplementary Figure S5B). The remaining residues that form the S1 pocket — Glu657 and Tyr914 — are conserved across the two proteins. The S1’ pocket is large and stretches into the internal cavity formed by the domain IV helices with the main contributing residues consisting of Arg828, Asn831, Val832 and Val862 (Supplementary Figure S5B).

Structure-activity relationships of P1 arginine mimetics

Previous studies have mapped the P1 substrate specificity for PfA-M1 and shown that this enzyme had a strong preference for a P1 Leu or Met [15]. Given that TgAPN2 prefers a P1 Arg and/or a longer substituent, we selected three arginine mimetics from a previous study [38] to probe the differences in the structure-activity relationships of the S1 pockets of PfA-M1 and TgAPN2. The compounds all contained a phosphonic acid zinc-binding group as well as various P1 arginine mimetics (Table 3). The inhibitory activity of the three selected compounds tested was moderate, but far greater than was observed for PfA-M1 (Table 3). The most potent inhibitor was compound 1 (Ki(app) 1.2 µM, Table 3) which resembles TgAPN2's preferred hArg substrate containing four carbons in the side chain. Compound 2 was shorter with three carbons in its side chain and had a slightly higher Ki(app) of 5.5 µM, matching our finding that longer, charged substrates may be preferred. The third compound tested (3) differs from 1 and 2 in that there is a phenyl ring within the side chain (Table 3). The addition of this phenyl ring was detrimental to inhibitory activity, increasing the Ki(app) to 11 µM.

Table 3

Inhibition of TgAPN2 by selected phosphonic (MD 1, 2, 3) and hydroxamic acid compounds 9b, 10o and 10q

Compound No[1]	R	Ki (µM)
		TgAPN2	PfA-M1[1]
1	1-amino-5-guanidinopentyl	1.2	11
2	1-amino-4-guanidinobutyl	5.5	193
3	1-amino-2-(4-guanidinophenyl)ethyl	11	>1000

We attempted to solve the co-crystal structure of TgAPN2 bound to 1 and/or 2, however, were unsuccessful at obtaining any bound structures. Therefore, we superimposed the coordinates of TgAPN2 onto the structures of PfA-M1 bound to 1 (PDB ID 4K5L) and 2 (PDB ID 4K5 M) [38], aligning the catalytic domains and assuming a displacement of the second, non-catalytic zinc ion from the active site (Figure 4). The phosphonic acid is a strong zinc-binding group and we assume that the coordination of the H3O3P group is similar in both enzymes. For PfA-M1-1, the guanidino moiety of the hArg forms a hydrogen bond with the backbone carbonyl of Glu572 (Figure 4A). A similar interaction would be possible for TgAPN1 with equivalent residue Asp911; however, it also seems likely that a second H-bond would be possible with the Oε atoms of Asp911 as well (Figure 4A). In addition, TgAPN2 has Gln1362 in H-bonding distance to the guanidino group, possibly providing a third coordinating bond (Figure 4A). This extra bonding capacity likely accounts for the increase in potency observed for TgAPN2 in comparison with PfA-M1, by locking the Arg mimetic into a more rigid position. The slightly shorter 2 was found to occupy two different conformations when bound to PfA-M1, the first was similar to that of 1 and the second saw the guanidino group form interactions with Val459, the residue that immediately precedes the GAMEN motif in PfA-M1 (Figure 4A). For TgAPN2, we suggest that this second conformation may not occur due to the ability of the compound to form hydrogen bonds with the side-chains of Asp911 and Gln1362. The carbonyl of Asp911 is beyond regular hydrogen bonding distance (4.5 Å) and this may account for the loss of potency observed for 2 against TgAPN2 compared with 1 — the loss of a hydrogen bond to the backbone may result in less stable binding due to the movement of the side-chains.

Figure 4.

PfA-M1 structures with bound inhibitors showing the superposition of TgAPNs.

The residues of PfA-M1 (denoted Pf) is shown with wheat carbon atoms and TgAPN2 (denoted Tg) with grey carbon atoms. Protein sidechain numbers are indicated according to each structure. The inhibitors are also shown in stick with compounds 1 (magenta carbon atoms) and 2 (purple carbon) in (A) and the hydroxamic acid scaffolds in (B). Compounds 9b (salmon carbon in stick) 10q (green carbon in sticks) 10o (yellow carbon in sticks) are shown.

The second conformation of 2 observed for PfA-M1 also highlights the known flexibility of the S1 pockets of M1 aminopeptidases. In both our superpositions, TgAPN2 Met798 is too close to the compound and would need to move to accommodate the inhibitor. Such a movement was observed in the crystal structure of PfA-M1-2 where the equivalent residue Val459 shifted to allow the compound to occupy the second site [38]. A similar movement was also observed with the equivalent EcAPN1 residue, with Met260 being labelled a gate-keeper residue that can move to allow substrate or inhibitor binding [33]. Therefore, it is possible that 2 could form interactions with this region, however, given the ∼17-fold increase in potency observed between PfA-M1 and TgAPN2, we feel it is likely that the compound is coordinated in one position in TgAPN2 (Figure 4a).

Compound 3, containing a 1-amino-2-(4-guanidinophenyl)ethyl core, was non-inhibitory toward PfA-M1 but did show moderate activity toward TgAPN2 (Table 3). The conclusions of the original study on the SAR of PfA-M1 showed that an amino(4-guanidinophenyl)methyl was capable of inhibition but the addition of the extra CH2 prior to the guanidinophenyl was what was detrimental to activity [38], placing the guanidino nitrogens too close to the end of the S1 pocket. However, TgAPN2 appears to like slightly longer charged P1 residues (Figure 2D), suggesting more flexibility of the residues that interact with the inhibitor.

Structure-activity relationships of hydroxamic acid containing PfA-M1 inhibitors

Previous work by our group has generated a library of inhibitors that combine a hydroxamic acid zinc-binding group linked various bi-aryl moieties that show inhibitory activity toward PfA-M1 [17,39,40]. To assess the capacity to re-purpose this scaffold to target TgAPN2, we selected three compounds of varying functional groups where R1 is the S1’ anchor position and R2 explores the S1 pocket (Table 3). Compound 9b is the least complex of the selected compounds, providing a Boc group in the S1′ of PfA-M1 and bromo-substituent in R2. The compound showed a near 100-fold loss in activity toward TgAPN2 in comparison with PfA-M1 (Table 3). The other two compounds, each had a slightly longer S1′ anchor, probed the capacity of TgAPN2 S1 pocket to accommodate either 3,4,5-trifluorophenyl (10o) or a thiophen-3-yl substituent (10q). The 3,4,5-trifluorophenyl analogue was a poor inhibitor of TgAPN2, whereas the thiophen-3-yl analogue had a similar potency to that of 9b. Interrogation of the superpositions of the structures onto PDB ID 4ZX3 provide some structural explanation as to the reason for the loss of potency (Figure 4B). Assuming a movement of the Met798, 9b can easily fit into the active site but appears to form very few interactions beyond the hydroxamic acid core. A similar situation can be seen with 10q, however, the thiophenyl group is within bonding distance of Asp911 and Gln1362 (Figure 4B). The similar potency of 9b and 10o suggests there may be a water network that assists the positioning of 9b that cannot be captured in a superposition, hence stabilising the inhibitor to provide an equivalent potency. The trifluorophenyl of 10o has a likely clash with Asp911 that may not have the flexibility that is the equivalent Glu572 has in PfA-M1 [41]. This likely accounts for the extremely poor inhibition by this compound (Figure 4B).

Discussion

Toxoplasmosis is a zoonotic infection caused by the apicomplexan parasite T. gondii. Infections are often asymptomatic and only present a risk to health for immunocompromised individuals and through congenital infection of a foetus. Current therapeutics require prolonged courses of drugs from weeks to more than a year and commonly have toxic side effects [42]. New medicines with enhanced efficacy as well as features that address the unique aspects of toxoplasmosis (e.g. eliminating tissue cysts, achieving therapeutic concentrations in the eye and brain) would greatly improve therapy. Conservation amongst apicomplexan protozoan parasites makes drug re-purposing an attractive option for screening for new anti-toxoplasmosis agents. In this study, we have investigated TgAPN2 — homologue of the malarial drug target, PfA-M1 — to assess the capacity of current aminopeptidase inhibitors to target T. gondii homologues.

TgAPN2 is a member of the M1 aminopeptidase family of metallo-exopeptidases that has a membrane anchor, as well as large and potentially disordered N-terminal domain that precedes both the transmembrane anchor and the enzymatic domain. It appears from bioinformatic analysis available on ToxoDB, that TgAPN2 may be the only typical intra-cellular M1 aminopeptidase and given its sequence identity to PfA-M1 (38% overall, 55% in the catalytic domain) may be a potential target for inhibition. A previous study has shown that recombinant TgAPN2 behaves as a typical metallo-exopeptidase with a substrate preference for a P1 Ala and Arg, and that it can be inhibited by generic aminopeptidases inhibitors bestatin and amastatin [14]. The authors of the biochemical study stated that TgAPN2 would share a similar homohexameric arrangement to that of other M1 aminopeptidases [14], which is in contradiction to the field as the M1 aminopeptidases or aminopeptidase N family are generally monomeric or dimeric in structure and function [1]. In this study, we produced the TgAPN2 ecto-enzyme and characterised its structure and function, producing comparative structure-activity relationships of PfA-M1 purposed inhibitors.

We produced the recombinant ecto-domain of TgAPN2 and confirmed that it acted as a typical M1 aminopeptidase in both structure and function. It exhibited metal-dependent aminopeptidase activity with optimal activity at near neutral pH and formed the typical bacterial M1 aminopeptidase fold. One unexpected finding in the resolution of the structure was the presence of unexplained density coordinated to the catalytic zinc. We have modelled this as a second zinc ion coordinated to the active site zinc. The position of a possible second metal ion is similar to that of magnesium ions in the unbound structures of PfA-M1, but the lack of significant concentration of metal ions in the crystallisation buffers suggests that the ion was scavenged. We attribute this finding to a crystallisation artefact and therefore do not assign a functional role however do suggest that the ability to coordinate a second metal ion to the active site ion is a possible reason that some M1 aminopeptidases are inhibited in the presence of excess zinc. Interrogating the role of the second metal-binding site with molecular biology techniques is challenging as the main bond positioning the ion is to the active site zinc, and residues that contribute to its stabilisation would be required for the normal architecture of the active site. Enzymatic analysis of natively purified TgAPN2 would help answer the question as to the number of zinc ions found in the catalytic domain of this protease.

The M1 aminopeptidase family is large and diverse, and often shows a variety of substrate preferences. Comparison of the extended substrate specificity of TgAPN2 with PfA-M1 shows a significant difference in what P1 residue the enzymes prefer. PfA-M1 has a much broader substrate specificity but a clear preference for hydrophobic natural amino acids (Met > Leu > Ala) [15]. This P1 preference is similar to the Eimeria tenalla APN1 [8] and human aminopeptidase N [43]. Similarly, extensive characterisation of the human ERAP1 family [44] has shown exclusive activity toward a P1 Leu and Met, but also differential activity depending on the length of the peptide substrate[45-48]. Longer peptide substrates can interact with an allosteric site that affects both specificity and enzymatic rates [47-49]. TgAPN2 prefers basic branched residues (Arg > Lys) and has a reduced activity toward Leu and Ala. The kinetic parameters (KM, kcat) confirm this preference but interestingly show that a P1 Arg has a reduced affinity for binding but a significantly increased turn-over rate. The bacterial homologues from E. coli (Arg > Ala > Lys [33]) and Neisseria meningitidis (Arg > Ala > Lys/Leu/Met [50]) share a preference for a P1 Arg, as does the human insulin receptor aminopeptidase (IRAP, Arg >>> Lys/Met/Leu, [44]) and leukotriene A4 hydrolase (LTA4H, Arg >>> Ala/Lys > Phe/Pro/Leu, [51]). Therefore, it can be hard to understand the capacity to re-purpose inhibitors from substrate profiling alone. The addition of comprehensive structure-activity relationship data to correlate the specificity profiles adds to our ability to predict cross-species homologue inhibitors, and potential off-target effects.

The structure of TgAPN2 provides clear insight to explain why compounds that potently inhibit PfA-M1 (9b, 10o and 10q) are only weak inhibitors of TgAPN2. The differences in the S1 pocket that produce a preference for a longer, charged P1 substrate also affect inhibitor interactions, with Asp911 and Gln1362 changing the bonding capacity of the pocket. Future design could use the hydroxamic acid zinc-binding group but should focus on capturing interactions as predicted for compounds 1 and 2 tested in this study, locking the inhibitor into hydrogen bonds with multiple residues within the S1 pocket. The structure produced in this study should contribute to the design of potent inhibitors of TgAPN2, as well as TgAPN3, that would be valuable tools to assess the suitability of these enzymes as drug targets.

Conclusions

TgAPN2 has been proposed as a potential new drug target for future anti-toxoplasmosis agents. In this study, we have produced and analysed the recombinant ecto-domain of TgAPN2 and shown that it has significant differences to the apicomplexan homologue PfA-M1. The value of these comparative structure-activity relationships should inform the design of potent inhibitors of the T. gondii M1 aminopeptidases. Furthermore, the TgAPN2 appears to be the main intracellular M1 aminopeptidase in T. gondii but does have an unusual N-terminal domain that was not characterised in this study. Investigation into the biological role(s) of this enzyme would also be of great benefit to assess its capacity to be a drug target.

Compound No[2]	R1	R2	Ki (µM)
			TgAPN2	PfA-M12
9b	Boc	Br	2.9	0.027
10o	C(O)tBu	3,4,5-trifluorophenyl	44.9	0.078
10q	C(O)tBu	thiophen-3-yl	2.4	0.64

Ki values are the mean of three independent experiments.

Sivaraman et al. [38].

Drinkwater et al. [40].