|Home | About | Journals | Submit | Contact Us | Français|
The development of highly selective small molecule inhibitors for individual caspases, a class of cysteine-dependent aspartate-specific proteases, has been challenging due to conservation of the active site. Previously we discovered an allosteric site at the dimer interface of caspases-3, -7 and -1 using disulfide trapping. Here we show this approach can generate selective tethered ligands and inhibitors for caspase-5 which is remarkable considering its high sequence similarity to caspase-1. Among the 62 hit out of a screen of ~15,000 thiol-containing fragments, a naphthyl-thiazole containing molecule was identified that selectively inhibited and labeled the allosteric cysteine in the p10 subunit of caspase-5, but caused very little inhibition or labeling of caspase-1. Interestingly, some of allosteric tethered compounds to caspase-5 did not inhibit its enzymatic activity, suggesting that thiol-labeling itself is not sufficient to drive inhibition. These studies validate an allosteric site on caspase-5 and provide a useful starting point to develop selective compounds to probe the role of caspase-5 separate from caspase-1 in the innate immune response.
Caspases comprise a family of aspartate-specific thiol proteases with dimeric structures. Based on sequence homology and functionality, they can be broadly divided into two groups: the apoptotic and the inflammatory caspases. Caspase-5 is part of the latter group which also includes caspase-1 and caspase-4 in humans, and caspase-11 and caspase-12 in mice (1). The human inflammatory caspases are believed to be involved in driving innate immune response (for reviews, see (2–8)). However, unlike caspase-1 which is primarily responsible for inflammatory cytokine processing, the specific role of caspase-5 remains unclear. Selective inhibitors to caspase-5 would help further dissect its specific role in innate immune response.
It has been challenging to find selective active site inhibitors for caspase-5 due to virtually identical substrate specificities among the inflammatory caspase family (preferred cleavage after Trp-Glu-Xaa-Asp). However, the discovery of a novel allosteric site located at the dimer interface some 15 Å away from the active site on caspase-1 (Figure 1A), caspases-3 and -7 opened up new opportunities for developing selective chemical tools for individual caspases (9–11). Even though the structure of caspase-5 is currently unknown, sequence alignment to its closest homolog, caspase-1, indicates that the allosteric site is likely to be preserved.
To test the hypothesis that caspase-5 activity can be regulated through compounds binding in the allosteric cavity, we employed a site-directed fragment discovery approach called disulfide trapping or tethering to target the site (12, 13). This powerful approach has been proven effective for fragment-based drug discovery and developing inhibitors for a variety of targets including proteases, kinases, and protein-protein interaction targets (10, 12, 14, 15). Here, we present the results of screening and characterization of a novel allosteric inhibitor of caspase-5 from a ~15,000 member fragment library (12, 13). These studies further validate the allosteric cavity as an alternative for developing selective inhibitors for the individual caspases and provide a potentially useful starting point to develop chemical probes for caspase-5 function.
The p20 subunit (residues 122-311) and p10 subunit (residues 331-418) of wild-type human caspase-5 were separately expressed in E. coli BL21 (DE3) as inclusion bodies from a pRSET expression vector (Invitrogen, CA). The preparation of inclusion bodies was performed as previously described (16) with the following modifications. Cells were lysed with a microfluidizer and inclusion-body pellets were collected by centrifuging at 4°C for 30 min. The pellets were washed twice with 50 mM Tris–HCl, pH 8.0, 100 mM NaCl, 0.25 M guanidine, and 0.5% Triton X-100, followed by two washes using the same buffer without the detergent. Washed pellets were re-suspended in 6 M guanidine–HCl, 20 mM DTT, 0.1 M Tris-HCl, pH 8.0 and frozen at −80 °C. The refolding and purification was carried out using the same procedure as previously described (17) without using malonate. After purification, the protein fractions were pooled, concentrated, and analyzed by SDS–PAGE.
The screening construct caspase-5 contained five cysteine to alanine mutations denoted C5A (Cys333Ala, Cys370Ala, Cys376Ala, Cys377Ala, Cys378Ala). The mutant was generated by site-directed mutagenesis using the QuikChange Site-Directed Mutagenesis kit (Stratagene, CA). Two sets of primers were included in a single QuikChange reaction to simultaneously introduce all mutations (extension time of 18 min at 68 °C, 18 cycles). This procedure produced 4 correct clones out of 6 clones sequenced.
Disulfide trapping screen was performed following published procedures (10) with a few modifications. Briefly, purified caspase-5 C5A was freshly diluted to 10 μM in the screening buffer (50 mM Hepes, pH 7.5, 50 mM NaCl, 100 μM β-ME) and was incubated at room temperature for 1 h. with pools of disulfide-containing compounds in 96-well plates. Following the equilibration period, reaction mixtures were analyzed by high-throughput mass spectrometry (LCT Premier, Waters, MA). Hits were identified by comparing the molecular mass of compounds covalently bound to the p10 subunit to the molecular masses of compounds in the pool.
The following two-step procedure was used for parallel re-synthesis of hits. 1) Disulfide dimer formation: in a 4-mL glass vial add EDC (0.11 mmol), the free acid coupling partner (0.10 mmol), a solution of cystamine.2HCl (0.05mmol), HOBt (0.01mmol), triethylamine (0.10 mmol), dH2O (25 μL), and DMF (300 μL). The resulting reaction mixture was stirred overnight. 2) Disulfide exchange: a solution of bis[2-(N,N-dimethylamino)ethyl]disulfide dihydrochloride (0.25 mmol), cysteamine hydrochloride (0.01–0.02 mmol) in water (100 μL) and DMSO (100 μL) was added to the above reaction mixture. Triethylamine (0.7 mmol) is then added and stirred overnight. After reaction, the mixture was diluted with 2:1 DMSO:dH2O to a final volume of 1 mL and injected onto a Waters Xterra 19×50mm Prep MS OBD HPLC column and eluted with a acetonitrile/water (0.05% TFA) gradient (0% to 40% acetonitrile in 8 mins, 40% to 100% in 2 mins, hold at 100% for 2 mins, and decrease to 0% in 1 min).
To determine the DR50, the testing compound was serially diluted by 2-fold starting at 100 μM before pre-incubated with 2 μM caspase-5 in presence of 100 μM β-ME. For measuring β-ME50, the concentration of the reducing agent was increased by adding freshly prepared β-ME to the reaction mixture containing 2 μM caspase-5 and 50 μM of compound. After 1 h of incubation, the samples were analyzed on LC-MS and the percentage of labeling was calculated based on the ratio of compound-conjugated p10 vs. unconjugated p10. Nonlinear regression was used to calculate DR50 and β-ME5o.
Caspase-5 or its variants was diluted in assay buffer (50 mM Hepes, pH 7.5, 50 mM KCl, 200 mM NaCl, 100 μM β-ME, 0.1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate) to 250 nM and incubated with or without compounds at room temperature for 1 h before assaying with fluorescent substrate Ac-WEHD-fmk. The change in relative fluorescence units (RFU) over time was monitored for 10 min using a Spectromax M5 fluorescence plate reader (Molecular Devices, CA) with excitation at 365 nm and emission at 495 nm. Enzyme activity was reported as the rate of change in RFU. All kinetic parameters were determined by fitting with nonlinear regression using the Michaelis-Menten model.
Sequence analysis showed that the proposed allosteric Cys341 is preserved in caspase-5 (Figure 1B). To reduce the complexity of mass spectrum we mutated the 5 other Cys residues (Cys333, Cys370, Cys376, Cys377, Cys378) on the p10 subunit to Ala. All Cys residues on the p20 subunit were left intact to serve as the internal control for eliminating non-specific labeling by thiol-containing compounds. The composition of the screening construct, named caspase-5 C5A, was confirmed by DNA sequencing as well as mass spectrometry analysis.
To prepare a large quantity of recombinant caspase-5 for the screening, we employed a protocol used for caspase-1 (16) that expresses the large and the small subunits separately as inclusion bodies and obtained the active enzyme through refolding and co-purifying the subunits together (see Materials and Methods for details). This method delivered an average yield of ~5 mg of caspase-5 C5A per litter of shake flask E. coli culture. The purity was > 99% as judged by SDS-PAGE (Figure 2A) and mass spectrometry (data not shown). Enzyme kinetics analysis using fluorescent substrate Ac-WEHD-fmk indicated that the recombinant caspase-5 is enzymatically active and in line with literature values (Figure 2B). The wild-type and C5A mutant had virtually the same kinetic constants showing the Cys to Ala mutations had minimal effect on enzyme activity.
The disulfide trapping screen and triage was carried out as outlined in Figure 3. Before the full-scale run, we tested and optimized the assay condition using a pilot library containing ~400 compounds from the library. The β-ME concentration was set to 100 μM in the assay buffer which is sufficient to allow disulfide conjugation to the target allosteric Cys while minimizing non-specific modification of other Cys residues on the protein. This yielded a typical hit rate that is less than 0.5%.
Having optimized the conditions we proceeded to screen ~15,000 thiol compounds. We identified 62 compounds (hit rate of ~0.3%) that showed greater than 50% labeling, defined as the percentage of compound-bound form among all forms of p10 subunit. Based on structural attractiveness and the least similarity to hits generated in previous caspase-1 campaign, 10 first-tier hits (Figure 3) were chosen to be re-synthesized. Five of the top compounds were re-synthesized and purified to greater than 90% and were tested on wild-type caspase-5 for inhibition potency (Figure 4).
Compound 8, a naphthyl-thiazole containing molecule, showed the most potent inhibition of caspase-5 activity among all hits. To confirm that compound 8 specifically bound to the p10 subunit of capase-5, we tested the labeling on wild-type caspase-5 by mass spectrometry. As shown in Figure 5B, there was a single peak with a M.W. shift of + 327 Da which corresponded to the mass of the p10/compound 8 complex. No labeling of the p20 subunit or other labeling of p10 was observed.
To assess the reversibility of compound 8 labeling of caspase-5, we determined the β-ME50 value by measuring the percentage of labeling in the presence of various concentration of reducing reagent. This has been used to effectively rank compounds for their relative potencies as the more potent compounds generally require higher β-ME concentration to reverse the labeling (for review see (12)). As seen in Figure 5C, the labeling percentage of caspase-5 by compound 8 showed a dose dependent decrease with increasing concentration of β-ME and virtually abolished at the highest β-ME concentration. The observed β-ME50 is 2.3 mM indicated a very high conjugation strength (23-fold above the concentration of β-ME used in the primary screen and about 50-fold in excess of the compound concentration). We also estimated the relative strength of labeling in presence of 100 μM β-ME by titrating caspase-5 with serial diluted compound 8, called DR50. The titration curve depicted in Figure 5D showed an excellent correlation between compound dose and percentage of labeling. The calculated DR50 is 0.19 μM, indicating that compound 8 can label 50% of the enzyme subunits when dosed at 0.19 μM in the presence of 100 μM β-ME (500-fold molar excess).
We further tested the functional effects of compound 8 binding to caspase-5 activity. As shown in Figure 6A, compound 8 inhibited capase-5 activity in a dose-dependent manner. The presence of inhibitor significantly reduced the maximum velocity of the reaction (Vmax) and also affected the binding affinity of substrate (KM) (Table 1). This observation argues that although compound 8 binds to an allosteric pocket away from the active site, it may still interfere the substrate binding by inducing unfavorable conformational change. To further assess the specificity of compound 8, we tested its inhibition of recombinant caspase-1. No inhibitory effect on caspase-1 was observed in presence of compound 8 (Figure 6B).
Highly selective inhibitors provide powerful tools to probe the function of enzymes in complex mixtures. One general approach is to design inhibitors based on known substrate structure. A number of “pseudo-substrate” competitive inhibitors (e.g. YVAD-fmk) have been developed with a wide range of potency and selectivity against caspases (18). However, due to the fact that caspase-5 has nearly identical substrate specificities as caspase-1, it has been difficult to find a selective active-site inhibitor for this enzyme. Using a fragment-based targeted screen, we were able to identify a novel non-peptic inhibitor which binds allosterically to capase-5 and completely inhibits its enzyme activity. This success validates the allosteric cavity located at the dimer interface of most caspases as a possible site from which the activity can be selectively modulated. Because of the relatively higher sequence diversity at the allosteric site among caspases family (Figure 1B), it provides a better chance for developing selective small molecule probes for each one of them.
The site-directed nature of disulfide trapping allowed us to test if binding at this putative allosteric site had a functional impact. To ensure selective labeling we mutated all of the Cys residues to Ala except the allosteric Cys residue on the p10 subunit while all Cys residues on the p20 subunits remained intact. By doing this we improved the chance of observing compounds that specifically tether to the allosteric site while maintaining the ability to triage non-specific disulfide bond forming compounds based on reaction with Cys residues on the p20 subunit.
Out of the five compounds we re-synthesized, only compound 8 had inhibitory effect on caspase-5 activity, despite the fact that all of them can tether to the enzyme. This shows that simple labeling of the allosteric cysteine is insufficient for inhibition therefore additional interactions with the compound are required due to the highly dynamic nature of the allosteric pocket. As it has been shown before for caspase-1 and caspase-7, the binding of substrates and allosteric inhibitors are mutually exclusive (10, 11). Thus, even though the compounds bind distally from the active site they are effectively competitive inhibitors. Therefore, only allosteric compounds with reasonably high affinity can inhibit caspase-5 in the presence of high concentration of substrates.
Even though conjugation of compound 8 relies on the thiol-reactive group, we believe it offers a reasonable starting point for developing highly potent and selective, stand-alone inhibitors for caspase-5. The scaffold of compound 8 is chemically simple and offers good opportunities for further improvement by SAR. In addition, extended tethering (19, 20) or other fragment growth strategies have been applied before to advance such compounds to potent non-covalent analogs suitable to seed drug discovery efforts. Our studies reveal a viable site and early starting point for this process. Moreover, it extends the generality of this site for drug discovery in other caspase family members, even those that share very close sequence homology.
We thank J. Zorn and D. Wolan for advice and assistance. Special thanks to M. Burlingame at the Small Molecule Discovery Center of UCSF for help with the chemical synthesis. This work was supported by NIH grant 5R01AI070292-02 and the Sandler Family Foundation gift to J.A.W.