Mechanism-based inhibition of HsaD: a C-C bond hydrolase essential for survival of Mycobacterium tuberculosis in macrophage

Authors


Abstract

Mycobacterium tuberculosis remains the leading cause of death by a bacterial pathogen worldwide. Increasing prevalence of multidrug-resistant organisms means prioritizing identification of targets for antituberculars. 2-hydroxy-6-oxo-6-phenylhexa-2,4-dienoate hydrolase (HsaD), part of the cholesterol metabolism operon, is vital for survival within macrophage. The C-C bond hydrolase, HsaD, has a serine protease-like catalytic triad. We tested a range of serine protease and esterase inhibitors for their effects on HsaD activity. As well as providing a potential starting point for drug development, the data provides evidence for the mechanism of C-C bond hydrolysis. This screen also provides a route to initiate development of fragment-based inhibitors.

Introduction

Although Mycobacterium tuberculosis has been almost eradicated in the developed world, around 1.4 million people died from the disease in 2011 (WHO, 2012) (95% were in developing countries) and 8.7 million people became infected. Around 3.4% of all cases were multidrug-resistant (MDR-TB) tuberculosis (defined as those with resistance to rifampicin and isoniazid), while there were around 25 000 cases of extremely drug-resistant tuberculosis (defined as those MDR-TB which are also resistant to fluoroquinolone and a second-line antitubercular e.g. amikacin).

The vital role of cholesterol in the infection cycle of M. tuberculosis is becoming increasingly apparent (Ouellet et al., 2011). Cholesterol is vital for phagocytosis of M. tuberculosis by macrophage (Peyron et al., 2000) and also plays an important role as an energy source during bacterial survival within macrophage (Van der Geize et al., 2007). The cholesterol metabolism operon of M. tuberculosis has been identified and includes the genes HsaA-D (Van der Geize et al., 2007). Gene deletion mutants of HsaC and HsaD have shown that these enzymes are required for survival inside macrophage (Rengarajan et al., 2005). As HsaD is an essential gene for survival inside macrophage, it is a promising target for antitubercular therapy.

HsaD is a member of the meta-cleavage product (MCP) hydrolase class of enzymes which are a subfamily of the α/β hydrolases (Lack et al., 2008). HsaD catalyses the cleavage of 4,9-DHSA within the cholesterol metabolism pathway (Van der Geize et al., 2007). HsaD cleaves carbon-carbon bonds via a serine protease-like catalytic triad (Lack et al., 2008, 2010).

Three classes of inhibitors were tested for activity against HsaD (Supporting Information, Fig. S1). The largest group was serine protease inhibitors. A number of covalent inhibitors, for example phenylmethylsulphonyl fluoride (PMSF), were tested alongside noncovalent inhibitors, for example benzamidine. Acetylcholinesterases are also members of the α/β hydrolase family and catalyse their reactions via a serine protease-like catalytic triad (Shafferman et al., 1992). A range of acetylcholinesterase-specific inhibitors were also tested, for example neostigmine. Humans have a structural homologue of HsaD called monoglyceride lipase [MGL (Bertrand et al., 2010)]. Like acetylcholinesterases, it shares the same overall fold as HsaD and also acts via a serine protease-like catalytic triad. A number of MGL-specific inhibitors have been tested, for example cholesterol-like pristimerin (King et al., 2009). Due to the hydrophobic nature of the physiological substrates of MGL and HsaD, it is proposed that similar inhibitors would be favoured by both enzymes. We describe the effect of these inhibitors on HsaD and the results shed light on the mechanism of action of HsaD and provide a basis for future approaches to inhibitor design.

Materials and methods

All reagents were obtained from Sigma-Aldrich unless specified. The structures of all inhibitors are provided in Fig. S1. 3,4-dichloroisocoumarin (DCI) was obtained from Calbiochem, JLK6 from Tocris Biosciences, and N-arachidonyl maleimide (NAM) was from Cayman Chemicals. All inhibitors were dissolved in DMSO except NAM, which was obtained dissolved in ethanol. 2-hydroxy-6-oxo-6-phenylhexa-2,4-dienoic acid (HOPDA) was synthesized as described previously (Lack et al., 2009) by Almac Sciences and dissolved in ethanol as it proved to be more stable in ethanol than DMSO (Fig. S2).

HsaD was expressed in Pseudomonas putida KT2442 and purified as described previously (Lack et al., 2009). Enzymatic activity was measured via monitoring OD450 nm on a Sunrise plate reader (Tecan). ε450 nm of HOPDA was measured as 13 200 M−1 cm−1. All inhibition studies were carried out with the following reaction mixture: 16 μg mL−1 HsaD, 100 μM HOPDA in 100 mM phosphate buffer pH 7.5, 20 mM NaCl, 5% (v/v) DMSO, 1% (v/v) ethanol. All inhibitors were incubated at 21 °C with HsaD for 20 min unless otherwise stated.

Mass spectroscopy was carried out via electrospray ionization time-of-flight mass spectroscopy (ESI-TOF) using an LCT mass spectrometer (Micromass).

Results

Serine protease inhibitors

PMSF is a broad spectrum serine protease inhibitor that forms a covalent adduct with the catalytic serine and has previously been shown to inhibit members of the MCP hydrolase family (Ahmad et al., 1995; Khajamohiddin et al., 2006). PMSF showed relatively weak inhibition of HsaD, with an IC50 of 630 μM after 20 min incubation (Fig. 1a). As well as PMSF a range of other serine protease inhibitors have also been tested including 4-amidinophenylmethanesulphonyl fluoride (APMSF), 4-(2-aminoethyl)benzenesulphonyl fluoride (AEBSF), benzamidine, DCI and JLK6 (Fig. 1b). APMSF, a close relative of PMSF, showed significantly poorer inhibition than PMSF (Fig. 1a and b), indicating that addition of the positively charged amidino group has a detrimental effect on binding. A third sulphonyl fluoride-based inhibitor, AEBSF, also poorly inhibited HsaD (Fig. 1b).

Figure 1.

Inhibition of HsaD by serine protease inhibitors. (a) Inhibition of HsaD by PMSF and APMSF. (b) Comparison of serine protease inhibitors. All compounds in red were tested at 1 mM, while all compounds in blue were tested at 100 μM. (c) Inhibition of HsaD by DCI and JLK-6. (d) Inhibition of HsaD by DCI and PMSF. In (a), (c) and (d), logarithms were taken of the molar concentration of the inhibitors. All inhibitors were incubated for 20 min with HsaD prior to addition of HOPDA. All tests were carried out in triplicate, and error bars show the standard deviation of the readings. Curve fitting was carried out in graphpad Prism version 6 using nonlinear fit and the log(inhibitor) – response variable slope equation.

Like PMSF, DCI is known to be a broad spectrum covalent inhibitor of serine proteases (Hedstrom, 2002), although their chemical structures are unrelated. DCI showed the strongest inhibition of all compounds tested with an IC50 of 17 μM (Fig. 1c). Covalent modification of HsaD by DCI was shown via mass spectrometry to increase the molecular weight of HsaD by an amount consistent with a single covalent modification by DCI (Table 1). The structural homologues of DCI, JLK-6 and S920428 showed significantly poorer inhibition (Fig. 1b and c). The noncovalent inhibitors including benzamidine, leupeptin and nafamostat mesylate also showed weak inhibition of HsaD (Fig. 1b) compared to PMSF and DCI.

Table 1. Molecular weight of HsaD determined by mass spectroscopy before and after modification with DCI
ProteinMW predicted (Da)MW actual (Da)
  1. Molecular weight was predicted based upon amino acid sequence using protparam from expasy (Wilkins et al., 1999). Actual molecular weights were obtained via ESI-TOF mass spectroscopy.

HsaD34 03934 045 ± 7
HsaD – DCI34 23534 231 ± 12

MGL and acetylcholinesterase inhibitors

MGL like HsaD catalyses the turnover of highly hydrophobic substrates: as such the inhibitors that have been identified tend to be insoluble (e.g. pristimerin and NAM). Although pristimerin is the most active noncovalent inhibitor tested (35% inhibition at 50 μM – Fig. 2a), further investigation was hampered by its poor aqueous solubility under conditions that are required for HsaD to remain active.

Figure 2.

Effect of inhibitors of other MCP hydrolases. (a) Inhibition of HsaD by MGL inhibitors. (b) Inhibition of HsaD by acetylcholinesterase inhibitors. All compounds in red were tested at 1 mM, those in blue at 100 μM and those in green at 50 μM. All experiments were carried out in triplicate, and error bars show the standard deviation of the readings.

NAM and JZL184 are covalent inhibitors: JZL184 like DCI and PMSF modifies the catalytic serine of MGL (Long et al., 2009), while NAM modifies a cysteine in the active site of MGL (Saario et al., 2005). Consistent with the lack of a cysteine residue in the active site of HsaD, NAM does not significantly inhibit HsaD (Fig. 2a). JZL 184 proved a better inhibitor (Fig. 2a) but was difficult to work with due to its hydrophobic nature and hence poor solubility.

A series of specific acetylcholinesterase inhibitors were tested for inhibition of HsaD (Fig. 2b). These included eserine, edrophonium, tacrine, neostigmine, pyridostigmine and trichlorfon. After incubation with HsaD, trichlorfon inhibited poorly. Eserine and neostigmine show better inhibition, but still not as strong as was observed with DCI (c. 30% inhibition at 1 mM). The other acetylcholinesterase inhibitors did not significantly inhibit HsaD.

Discussion

Two mechanisms have been proposed for the hydrolysis of substrates by MCP hydrolases. The first is based on the mechanism known to occur in serine proteases and proceeds via an acyl enzyme and tetrahedral intermediate (Ruzzini et al., 2012). The second requires a keto-enol tautomerization resulting in a gem-diol intermediate (Horsman et al., 2007). Recent mutagenesis experiments combined with structural studies resulted in trapping of the acyl enzyme intermediate of HOPDA hydrolysis, by another member of the C-C bond hydrolase family, BphD (Ruzzini et al., 2012) strongly supporting the first mechanism. Inhibition by PMSF and DCI is also consistent with this mechanism as PMSF and DCI act as tetrahedral and acyl enzyme intermediate analogues, respectively, when they modify the active site serine.

The most successful inhibitors were those that covalently modify HsaD (e.g. DCI). The primary issue with DCI and other covalent inhibitors tends to be their broad specificity profile making them poor starting points for inhibitor design. To help understand the specificity observed among the covalent inhibitors, the structure of HsaD modified with PMSF was solved (Fig. 3). Although density was observed for the sulphonate group covalently linked to Ser114, there was insufficient density to accommodate the phenylmethyl group of PMSF. A lack of electron density for PMSF in the structure with HsaD might suggest that PMSF acts reversibly. To test whether PMSF is a reversible inhibitor of HsaD, the protein used for crystallization was diluted 80-fold to measure enzyme activity. The concentration of PMSF following dilution was 10 μM which is noninhibitory, however, the enzyme activity was reduced to only 20% of a control that had been treated identically apart from preincubation with PMSF. As a result, PMSF is likely to act irreversibly. The structure of another α/β hydrolase fold protein (RsbQ) has been solved when modified with PMSF (Kaneko et al., 2005). A comparison of the active sites of RsbQ and HsaD is shown in Fig. 4. In contrast to the small hydrophobic active site of RsbQ (Fig. 4a), HsaD has a large open active site (Fig. 4b). The RsbQ active site is perfect for binding the hydrophobic phenylmethyl group of PMSF as it is bordered by three phenylalanine residues. The more open site of HsaD means that PMSF is more mobile, explaining the lack of density for the phenylmethyl group. The hydrophobic nature of the active site close to the catalytic serine (Fig. 4b) makes binding of the positively charged amidino group of APMSF unfavourable and explains its relatively poor inhibition compared with PMSF (Fig. 1a).

Figure 3.

Structure of HsaD complexed with PMSF. Structure was solved at 1.8 Å and refined to give Rwork and Rfree values of 17.3%, 20.3%, respectively. The catalytic triad of HsaD is shown and labelled. Green density represents an FoFc map at 3σ, while the blue density represents the 2FoFc map at 1σ.

Figure 4.

Comparing the sizes of the binding sites of various enzymes. (a) Binding of PMSF to the catalytic site of RsbQ. (b and d) Binding of PMSF to HsaD and MGL, respectively, modelled on its binding to RsbQ. (c) Binding of PMSF to trypsin. The structure of HsaD was taken from PDB 2VFC (Lack et al., 2008), while RsbQ is from PDB 1WPR (Kaneko et al., 2005), trypsin from PDB 1PQA (Schmidt et al., 2003) and MGL from PDB 3JW8 (Lack et al., 2010). RMSD of overlay between RsbQ and HsaD was 2 and to MGL was 2.1 via secondary structure matching. PMSF is shown with green carbon atoms, and in (a), the three interacting phenylalanine residues are labelled and are shown in purple. The surfaces shown in (b)–(d) are coloured by contact potential. Structures were visualized and aligned in CCP4MG (McNicholas et al., 2011).

The Hill slope of the DCI and JLK-6 dose–response curves are very similar (Fig 1c – fitted as 0.88 and 0.9, respectively). Dose–response curves that have similar Hill slopes indicate that the inhibitors work via the same mechanism which reflects the similar chemical structures of DCI and JLK6 (Fig. S1). PMSF is a member of a different family of inhibitors (sulphonylfluoride rather than isocoumarin) and consistent with this has a different Hill slope to that of DCI (Fig. 1d – fitted as 1.9).

Those inhibitors with the broadest specificity against serine proteases and acetylcholinesterases are also the inhibitors which show the best inhibition against HsaD. PMSF and DCI inhibit a wide range of serine proteases, for example thrombin, elastase and trypsin (Turni et al., 1969; Hedstrom, 2002); both also inhibit acetylcholinesterase (Turni et al., 1969; Hedstrom, 2002), and PMSF inhibits MGL (Muccioli et al., 2008). Thus, it is unsurprising that they also inhibit HsaD. More selective serine protease inhibitors such as APMSF [does not inhibit either chymotrypsin or acetylcholinesterase (Laura et al., 1980)] do not inhibit HsaD. The acetylcholinesterase inhibitors, for example eserine, are drug molecules and designed to show very good specificity for acetylcholinesterase, which is consistent with their poor inhibition of HsaD.

The majority of the noncovalent inhibitors were not very effective inhibitors of HsaD: as the main anchor for covalent inhibitors is the active site serine, whereas the noncovalent inhibitors are dependent upon the shape/charge distribution of the active site. Poor inhibition by the majority of noncovalent inhibitors (e.g. benzamidine) can be linked to their relatively small size. HsaD has a large open active site (Fig. 4b) which is considerably larger and more hydrophobic than the active site of either serine proteases or acetylcholinesterases (e.g. trypsin – Fig. 4c). The active site of MGL is more comparable in size to that of HsaD (Fig. 4d). Noncovalent inhibitors of MGL are thus significantly larger than those of serine proteases (e.g. compare pristimerin and benzamidine –Fig. S1) and fill more of the HsaD active site and thus have lower IC50 values. The lipophilicity of the inhibitors also has a direct effect with the more hydrophobic inhibitors, for example pristimerin, being favoured over charged ones, for example neostigmine, due to the apolar nature of the HsaD active site.

The aim of this work was to identify leads for fragment-based drug design (Scott et al., 2012). DCI has emerged as a good covalent inhibitor with a low IC50 value (Fig. 1a), it is however limited in its usefulness due to its ability to inhibit a broad range of enzymes (Hedstrom, 2002). Structural studies are ongoing to determine the mode of binding of DCI within the active site to improve specificity.

Acknowledgements

We would like to thank Dr David Staunton (Biochemistry, Oxford University) for carrying out the mass spectroscopy for this manuscript. We would also like to thank Dr Edward Lowe (Biochemistry, Oxford University) for his help with the data collection and structure solution.

Ancillary