Extended subsite profiling of the pyroptosis effector protein gasdermin D reveals a region recognized by inflammatory caspase-11

Pyroptosis is the caspase-dependent inflammatory cell death mechanism that underpins the innate dysregulated in inflammatory disorders. Pyroptosis occurs via two pathways: the canonical pathway signaled by caspase-1 and the noncanonical pathway regulated by mouse caspase-11 and human caspases-4/5. All inflammatory caspases activate the pyroptosis effector protein gasdermin D, but caspase-1 mostly activates the inflammatory cytokine precursors pro-interleukin-18 and pro-interleukin-1β (pro-IL18/pro-IL1β). Here, in vitro cleavage assays with recombinant proteins confirmed that caspase-11 prefers cleaving gasdermin D over the pro-ILs. However, we found that caspase-11 recognizes protein substrates through a mechanism that is different from that of most caspases. Results of kinetics analysis with synthetic fluorogenic peptides indicated that P1’– P4’—the C-terminal gasdermin D region adjacent to the cleavage site—influences gasdermin D recognition by caspase-11. Furthermore, introducing the gasdermin D P1’–P4’ region into pro-IL18 enhanced catalysis by caspase-11 to levels comparable to that of gasdermin D cleavage. Pro-IL1β cleavage was only moderately enhanced by similar substitutions. We conclude that caspase-11 specificity is mediated by the P1’–P4’ region in its substrate gasdermin D, and similar experiments confirmed that the substrate specificities of the human orthologs of caspase-11—i.e. caspases-4 and -5—are ruled by the same mechanism. We propose that P1’–P4’-based inhibitors could be exploited to specifically target inflammatory caspases.


Introduction
Caspases constitute a family of cysteine proteases that are obligate dimers in their active conformations (1,2). They have a stringent requirement for aspartic acid at the primary specificity position in their substrates (P1, Figure 1) (3). A great number of studies have aimed to understand the specificity mechanisms of these central proteases of regulated cell death (3,4). Early works pointed out the requirement for interaction with four amino acid residues to the N-terminal side of a cleavage site (5) thus caspases can be distinguished by their favored P1-P4 sequences (6).
The main objective of learning the specificity requirements of a caspase is to use this information to develop specific and selective chemical biology tools for their study in cellular biology and disease models. However, success is limited by the overlap in the preferred peptide sequences, which prevents the generation of selective reagents for individual caspases (7). Cleavage sequences have also been explored aiming to predict natural protein substrates, however, many endogenous substrates are cleaved at suboptimal sites (8). Alternative substrate specificity mechanisms like exositesenzyme regions distant from the catalytic site that participate in substrate binding-are emerging as mechanisms explaining the discrepancies between consensus sequences and actual cleaved proteins by caspases (9)(10)(11). Investigating these mechanisms is important because they may reveal novel targeting sites for caspase inhibition.
Inflammatory caspases regulate pyroptosis, a form of regulated cell death central to the innate immune response and clearance of pathogens (12). Pyroptosis is divided into canonical and noncanonical pathways. Caspase-1 is involved in signaling canonical pyroptosis, whereas mouse caspase-11 and human caspases-4 and -5 regulate the non-canonical pathway. Caspase-1 is activated in protein complexes known as inflammasomes, which are assembled upon recognition of pathogen or damage associated molecular patterns (13,14). In contrast, lipopolysaccharide from gram negative bacteria is the only known activator the noncanonical pathway and is thought to directly induce activation of caspases (15)(16)(17).
Inflammatory caspases cleave the pyroptosis effector protein gasdermin D to release its Nterminal domain (18,19). The N-terminal gasdermin D domain oligomerizes in the plasma membrane causing the distinguishing lytic event of pyroptotic cell death (20,21). A recognized function of gasdermin D is to allow the release of the proinflammatory cytokines interleukin-1β and -18 (22,23). Genetic and biochemical evidence indicates that the precursors of these cytokines (pro-IL1β and pro-IL18) are proteolytically activated by caspase-1 (15,24). Caspase-11 does not activate the interleukins directly, but caspase-1 is activated downstream of caspase-11 in the noncanonical pathway (15). The detrimental effects of autoinflammatory diseases can be halted in part by anti-IL1 therapy, signifying the potent inflammatory nature of this cytokine (25)(26)(27), making pyroptosis an attractive target for therapeutics of inflammation disorders. Development of inhibitors and probes to selectively target each individual inflammatory caspase would be beneficial to investigate their mechanisms in cell death signaling and disease and provide a path towards anti-inflammatory therapeutics.
One of the main conundrums in the field is the discrepancy of the number of inflammatory caspases in different mammals and the substrates they cleave. Extensive substrate specificity analysis using peptide combinatorial libraries revealed that using traditional active site-directed strategies is insufficient to selectively target individual inflammatory caspases (6,24,28). In contrast, protein substrates can be readily distinguished by caspase-11, leading to the prediction that its specificity is mediated by exosite interactions (24). Indeed, an interaction between gasdermin D and a region of caspase-11 may contribute towards catalysis (11), but cannot explain the specificity differences between inflammatory caspases. Therefore, we hypothesize that there are other mechanisms besides the exosite that determine substrate specificity. By using recombinant proteins and peptide substrates we describe the role of alternative substrate recognition in driving specificity of inflammatory caspases.

In vitro cleavage assays of protein substrates suggest that inflammatory caspases have specificity determinants outside of the P1-P4 region
Caspase specificity is generally dictated by the S1-S4 pockets that interact with the P1-P4 region of substrates ( Figure 1) (3,4). We postulated that extended substrate recognition elements may account for the restricted specificity of caspase-11 and its human orthologues.
Expression of full-length caspase-3 in E. coli results in a fully functional enzyme (29). We found that expression of the CARD-deleted version of caspase-1, caspase-4 and caspase-5 resulted in fully functional enzymes, as previously demonstrated for caspase-11 (24).
Recombinant inflammatory caspases and apoptotic caspase-3 were used for in vitro cleavage assays of gasdermin D, pro-IL18 and pro-IL1β. Cleavage products were visualized on SDS-PAGE ( Figure  2A). The inflammatory caspases cleaved the protein substrates at the expected positions previously reported ( Figure 2B) (18,19,(30)(31)(32)(33)(34) to generate characteristic products that were quantitated by densitometry. Band quantitation allowed calculation of the cleavage efficiency (kcat/KM) ( Figure 2C). Caspase-1 cleaved all three substrates at high efficiency. Gasdermin D was the preferred substrate of caspase-11. Pro-IL18 was also cleaved by caspase-11 but 50 times slower than gasdermin D. Only minimum cleavage pro-IL1β was observed with caspase-11.
To control for folding of the substrates we employed the apoptotic caspase-3 which has been shown to generate loss-of-function alternative cleavages to those of the inflammatory caspases in gasdermin D and the pro-interleukins ( Figure 2B) (31,35,36). As previously reported, we observed cleavage by caspase-3 (Figure 2A and 2C). These results suggest that caspase-3 may inactivate the interleukins as part of a mechanism that blunts pyroptosis (36).
Human caspase-4 and caspase-5 are often considered synonymous enzymes and orthologues of murine caspase-11. However, these enzymes differ in their tissue expression and are upregulated differently (37)(38)(39)(40)(41). We hypothesized that these enzymes are not interchangeable. Biochemical analysis showed that caspase-4 and caspase-5 have very similar specificity in P1-P4 (6). However, analysis of the human protein substrates revealed a very different picture. Overall, gasdermin D was preferred to the pro-interleukins by caspase-4 and caspase-5. However, whereas caspase-4 was more like caspase-1 in its cleavage of gasdermin D and pro-IL18, caspase-5 was more like caspase-11 ( Figure 3). However, caspase-4 does not cleave pro-IL1β and caspase-5 cleaves it very poorly ( Figure 3). This data revealed that caspases signaling noncanonical pyroptosis prefer gasdermin D as a substrate. Based on the cleavage site sequence of the protein substrates, previous peptidyl substrate library screens have shown that P1-P4 specificity do not determine specificity for protein substrates in these caspases [caspases-4/-5/-11] (6,24).
Extended synthetic substrates confirm the importance of the prime side of the gasdermin D sequence in enhancing activity of caspase-11 Extended interactions at the active site level beyond the P1-P4 sites are a possible mechanism for enhancing activity. An example of this is caspase-2 which has been demonstrated to cleave pentapeptides at rates 10-40 times higher than tetrapeptides (42). Given our previous observations that such interactions occur outside P1-P4 area and the faster processing of gasdermin D by caspase-11 compared to the pro-ILs, we hypothesized that interactions may occur at positions outside but proximal to the P1-P4 sites. To investigate such potential interactions, we designed and synthesized extended IQF substrates containing different portions of the mouse gasdermin D sequence around the cleavage site within positions P7-P5' ( Table 1). Caspase-1 was far more efficient at cleaving these substrates than caspase-11 ( Figure  4). Like most caspases, caspase-1 is most influenced by P1-P4. In contrast, caspase-11 is more highly influenced by the prime side amino acids P1'-P4' (Figure 4). These observations confirm that the prime side contains specificity determinants in substrate recognition by inflammatory caspases.

Mutational analysis reveals that cleavage of protein substrates by caspase-11 is strongly influenced by the prime side residues
Given the influence of prime side region of peptidyl substrates on the activity of caspase-11, we asked if these constitute specificity determinants for interaction with gasdermin D. Consequently, we predicted that introducing gasdermin D sequences from this region into pro-IL18 would make this a better substrate for caspase-11. The hypothesis was supported for caspase-11 where we observed a 40fold increase in cleavage of the mutant containing P1'-P4' substitution ( Figure 5). In contrast, we observed minimal enhancement in caspase-1 catalysis.
We sought to pinpoint what particular subsite has the most influence on caspase-11 and we individually substituted the amino acids in the P1'-P4' region of pro-IL18 for those of gasdermin D. P1' and P4' mutations in pro-IL18 demonstrated the largest effect on the activity of caspase-11, both leading to a 5-fold efficiency increase ( Figure 5). In contrast, similar mutations moderately enhanced the cleavage of pro-IL1β by caspase-11 ( Figure 6). We conclude that pro-IL18 is a better protein substrate scaffold than pro-IL1β for caspase-11.
Caspase-3 was used as control for cleavage of the mutant proteins, it cleaved all pro-IL18 mutant proteins at alternative sites producing the same product sizes as for WT pro-IL18. Caspase-3 had the same activity in each of the introduced mutations, suggesting that these mutations did not result in generally more accessible or better sites for any caspase.

Human caspase-4 and -5 are similarly influenced by prime side residues from gasdermin D
To investigate if caspase-4 and caspase-5 specificity is determined by the prime side we used the mouse protein substrates and their mutants. Gasdermin D was a good substrate for both caspase-4 and caspase-5. Pro-IL1β was not cleaved by caspase-4, was a poor substrate for caspase-5, being cleaved at an alternative upstream site. Cleavage of pro-IL1β by caspase-5 was enhanced by introduction of the gasdermin prime region, which also resulted in a shift of cleavage site to generate the active cytokine. Pro-IL18 was not cleaved by caspase-4 but was cleaved at moderate rate by caspase-5. Introducing the gasdermin D prime side region in pro-IL18 enhanced catalysis by both caspases to levels comparable to those of gasdermin D (Figure 7). The prime side residues of gasdermin D differ between human and mouse ( Figure 7C). Nevertheless, both substrates are cleaved by caspase-4 and caspase-5. These data revealed that caspase-4 and caspase-5 share with caspase-11 a requirement for prime side residues to enhance specificity and catalysis.

Discussion
Inflammatory caspases are central regulators of pyroptosis because they activate the pyroptosis effector protein gasdermin D by limited proteolysis (18,19). A function of caspase-1, which is not shared with caspase-11, is processing of IL18 and IL1β precursors to produce the active forms of these inflammatory cytokines (15,24). This differential substrate recognition is centered in caspase-11 and likely involves sites on the protease that are distinct from the conventional active site cleft (S1-S4).
Caspase-11 has two orthologues in human, caspase-4 and caspase-5, but their individual functions are not yet defined. Caspase-4 and caspase-5 may be influenced by different activators (43). There are also differences in in their expression profiles that indicate cell type-dependent diversification (40,44). Consistent with their function in signaling the pyroptotic non-canonical pathway we observed that, similarly to caspase-11, caspase-4 and caspase-5 favor gasdermin D processing over the pro-ILs. A point of divergence of caspases-4 and caspase-5 is the distinct preference of the former for pro-IL18. The preference of caspase-4 over caspase-5 for pro-IL18 is consistent with the observation that epithelial cell lines respond to gram negative bacteria by activating pro-IL18 in a caspase-4 dependent manner (39). These observations suggest that caution should be exercised in assuming equivalence between caspase-4 and caspase-11.
Preference of caspase-11 for gasdermin D is in part explained by the recently described exosite interaction between inflammatory caspases and gasdermin D (11). Caspase-11 is intrinsically less efficient than caspase-1 (lower catalytic rates on substrates) (24). The exosite implicates interaction between inflammatory caspases and the C-terminal gasdermin D domain that stabilizes the caspase active dimeric conformation thereby increasing the activity towards gasdermin D (11). Cleavage of fluorogenic peptidyl substrates based on the gasdermin D cleavage site motif sequence led us to hypothesize that the C-terminal region adjacent to the cleavage site (P1'-P4') contains amino acids that enhance caspase-11 activity beyond the established P1-P4 specificity determining region.
A limitation of the peptidyl fluorogenic substrates containing the gasdermin D cleavage site motif sequence is that it does not take into consideration the overall protein substrate scaffold. Therefore, we used protein substrates to test our hypothesis. If the gasdermin D P1'-P4' region contains specificity determinants for caspase-11, then substitution of these sites in pro-ILs would enhance caspase-11 catalysis. Indeed, introducing the gasdermin D P1'-P4' region in pro-IL18 enhanced catalysis by caspase-11 to levels comparable to gasdermin D. Individual mutations in the P1'-P4' region in pro-IL18 had moderate effects, signifying that amino acids in this region have additive effects on caspase-11 catalysis. The most influential positions were P1'-Gly and P4'-Glu. Glycine is generally preferred in P1' by caspases and this was previously corroborated in inflammatory caspases (42,45,46). P1'-Gly is strongly conserved in gasdermin D (Figure 8), hence P1' steric hindrance is a limiting factor for caspase-11 activity. Pro-IL18 contains Lys or Arg in the P4' position, while gasdermin D most often contains a negatively charged or a neutral amino acid residue (Figure 8). The positive charge in P4' possibly disfavors interaction with pro-IL18.
In contrast to pro-IL18, we had only limited success on improving IL1β as a substrate for caspase-4, caspase-5 and caspase-11 with P4-P4' substitutions. This implies that IL18 is a better substrate protein scaffold than pro-IL1β for these caspases and underscores the importance of caspase-1 as the most efficient IL1β converting enzyme. Our data indicated that the dominant specificity sites for caspase-1 are S1-S4, consistent with pro-IL1β recognition (5,47). In contrast, dominant sites for caspase-11 are S1'-S4', consistent with gasdermin D recognition. This is supported by P1'-P4' deletion mutations of gasdermin D. Although initially interpreted to shorten the distance between the exosite and the scissile bond (11), we interpret this to reflect sequence alterations of the prime side that interfere with caspase-11 catalysis.
Cleavage of gasdermin D, pro-IL18 and pro-IL1β by inflammatory caspases induces gain-of function of these proteins (18)(19)(20)30). Upon cleavage, gasdermin D N-terminal domain induces pyroptosis and mature ILs are able to exert their proinflammatory properties. In contrast, caspase-3, an apoptosis executioner, cleaves these proteins at alternative sites inducing loss-of-function (31,36). Caspase-3 activity on the pyroptotic substrates is thought to be part of the interplay mechanism between apoptosis and pyroptosis to ensure that apoptotic cells remain inflammatory silent (36,48). Caspase-3 cleavage occurred at the expected sites. Mutations did not alter caspase-3 cleavage. This signifies that P1'-P4' recognition is specific to inflammatory caspases.
We have showed that prime side interactions (the C-terminal region adjacent to the cleavage site) are determinant for substrate recognition by inflammatory caspases. We propose that in addition to the previously described exosite interaction (11), the P1'-P4' region also influences caspase-11 specificity for gasdermin D (this also applies to its human orthologues, caspase-4 and caspase-5, with the caveat that in our study the protein scaffolds were mouse, not human). Ours is a mechanistic study that seeks to identify the specificity determinants of substrate recognition. Therefore, it should not matter whether the sequences of our test substrates match the natural species-specific sequences ( Figure 7C). Naturally, these conclusions are independent of any hypothetical regulatory proteins that may exist in cells undergoing pyroptosis.
To date, there are inhibitors preferentially targeting inflammatory caspases over apoptotic caspases (49,50). In contrast, targeting individual inflammatory caspases has been less successful. Inflammatory caspases differ in their set of protein substrates and learning their specificity determinants will be conducive for inhibitor design. In fact, prime side exploring inhibitors have been investigated to target the activity of caspases (51). Future studies will expand the use inhibitor libraries to explore the prime side in inflammatory caspases.

Plasmids and cloning
Constructs encoding pro-IL18, pro-IL1β, as well as CARD deleted modified versions of caspase-1 and caspase-11 in pET29b(+) containing a C-terminal 6×His tag were previously described in (24). Deletion of the N-terminal caspase recruitment and activation domain (CARD) of inflammatory caspases results in a form that encompasses only the catalytic domain, enabling direct comparisons between caspase activity and specificity. DNA encoding the CARD deleted version of caspase-4 and caspase-5 containing a C-terminal 6×His tag were purchased from Integrated DNA technologies (IDT, San Diego, CA) and cloned into pET29b(+) by using Nde I and Xho I. The expression construct for caspase-3 containing a C-terminal 6×His tag was in pET23b as previously described (52). Cloning of pro-IL18 and pro-IL1β mutants was performed from DNA gene block purchased from IDT or by mutagenesis using PCR-driven overlap extension and cloned into pET29b(+) containing a C-terminal 6×His tag. Gasdermin D sequence was amplified from a pET29b(+) construct using primers to add an N-term 8xHis and cloned into pET15b using Nde I and Xho I restriction enzymes.

Protein expression and purification
All constructs were transformed into BL21(DE3) competent E. coli and cultured in 2xYT media containing appropriate antibiotics, expressed and purified as described previously (24). Protein concentration was calculated by absorbance at 280 nm and concentration of active enzyme was calculated by active site titration against benzoxycarbonyl-VAD-fluoromethyl ketone in 20 mM PIPES, 10% sucrose, 100 mM NaCl, 0.1% CHAPS, 1 mM EDTA, 10 mM DTT and was supplemented with 0.75 M sodium citrate to enhance caspase activity using Ac-WEHD-7amino-4-methylcoumarin as substrate (53).

Enzymatic assays with fluorogenic substrates
Enzymatic assays of recombinant caspases consisted on 100-μl final volume and were done in 96-well opaque plates (Costar, Corning). Assay buffer was 20 mM PIPES, 10% sucrose, 100 mM NaCl, 0.1% CHAPS, 1 mM EDTA, 10 mM DTT and was supplemented with 0.75 M sodium citrate to enhance caspase activity. Caspases were incubated in caspase buffer for 10 min at 37 °C before the assay and activity assay was initiated by addition of substrate solution. Activity assays consisted on kinetic measurement of fluorescence by using a CLARIOstar plate reader (BMG LabTech). For ACC fluorophore was detected at excitation/emission 355/460 nm, and AFC at 400/505nm. Reaction velocity was calculated as the linear portion of kinetic curves using MARS data analysis software (BMG LabTech).

Synthesis of peptidyl fluorogenic substrates
Internally quenched fluorescent (IQF) substrates with the ACC-Lys(dnp) donor-acceptor pair were synthesized as previously described (54). In brief, IQF substrates were synthesized on Rink Amide AM polystyrene resin using Fmoc solid phase peptide synthesis. Fmoc-protected amino acids were subsequently coupled to the resin using HATU/2,4,6-collidine reagents dissolved in DMF. Fmoc-protecting groups were removed after each coupling cycle using 20% piperidine in DMF. Finally, the Fmoc-ACC-OH was coupled twice to the N-terminus of peptides using HOBt/DICI reagents dissolved in DMF (24 hours), followed by removal of the Fmoc-protecting group. Next, the resin was dried and the substrates were cleaved from the resin using TFA. All substrates were purified using RP-HPLC, dissolved in DMSO to the concentration of 10mM and stored at -20 °C until use.

Enzyme kinetics of peptidyl substrates
We determined kcat/KM parameters for fluorogenic substrates toward caspases. For this, serial dilutions of substrates were prepared in eight wells of 96well plates and activity assays were initiated by addition of enzyme. Reactions were monitored for 30 min and the reaction velocity was plotted against substrate concentration. Kinetic parameters Vmax and KM were calculated with Prism 7 (GraphPad) using the allosteric sigmoidal equation, kcat was calculated by using equation 1.

In vitro cleavage of recombinant protein substrates
Recombinant inflammatory caspases were subjected to 2-fold dilution series and incubated for 30 min at 37 °C with 4 μM gasdermin D, pro-IL18, pro-IL1β or mutant proteins. Reaction controls consisted on incubation of substrate or enzyme only reactions. Reactions consisted on 60 μl final volume and were performed using assay buffer without supplementation of sodium citrate. After incubation, reactions were terminated by the addition of 30 μl of 3× SDS loading buffer and incubated at 95 °C for 5 min, reaction tubes were centrifuged for 1 min at 16,000 × g. Reaction products were separated on 4-12% Bis-Tris SDS-PAGE and stained with Instant Blue (Expedeon). The gels were scanned by using an ODYSSEY CLx imaging equipment (LI-COR) and images were exported to Image Studio software for band intensity quantification corresponding to protein substrate remaining after assay. Band intensity values were normalized relative to those of noncleaved substrate and values were plotted against log [E] estimated as active site titration. E1/2 values were calculated with Prism 7 (GraphPad) using the log(inhibitor) vs response (three parameters) equation. E1/2 values were used to calculate the catalytic efficiency of caspases for protein substrates according to equation 2 (53). Were kcat/KM is the second order rate constant for substrate hydrolysis, E1/2 is the concentration of caspase at which 50% of substrate is converted and t is the reaction time in seconds.

kcat/KM=ln2/(E1/2 × t) (Equation 2)
Data availability: All data are contained within the article.  Figure 1. The convention for naming substrate residues interacting with a protease active site (55). The residue to the N-terminal side of the cleaved peptide bond is named P1 and the one to the C-terminal side is P1', then residues are numbered consecutively from this origin. Many substrate residues interact with specificity pockets in the protease which are named with an S and numbered following the same logic used for the substrate. The specificity pockets are defined by several enzyme residues.       . Cleavage site motifs of inflammatory caspase protein endogenous substrates. Sequences were retrieved by Blast search in the Uniprot database, cleavage sites were predicted by multiple sequence alignment using Clustal Omega and consensus sequences were generated iceLogo (n= number of sequences, p value =0.05, % difference describes the frequency for an amino acid compared to the human proteome) (56).