Poxviruses encode many proteins that enable them to evade host anti-viral defense mechanisms. Spi-2 proteins, including Cowpox virus CrmA, suppress anti-viral immune responses and contribute to poxviral pathogenesis and lethality. These proteins are ‘serpin’ protease inhibitors, which function via a pseudosubstrate mechanism involving initial interactions between the protease and a cleavage site within the serpin. A conformational change within the serpin interrupts the cleavage reaction, deforming the protease active site and preventing dissociation. Spi-2 proteins like CrmA potently inhibit caspases-1, -4 and -5, which produce proinflammatory cytokines, and caspase-8, which facilitates cytotoxic lymphocyte-mediated target cell death. It is not clear whether both of these functions are equally perilous for the virus, or whether only one must be suppressed for poxviral infectivity and spread but the other is coincidently inhibited merely because these caspases are biochemically similar. We compared the caspase specificity of CrmA to three orthologs from orthopoxviruses and four from more distant chordopoxviruses. All potently blocked caspases-1, -4, -5 and -8 activity but exhibited negligible inhibition of caspases-2, -3 and -6. The orthologs differed markedly in their propensity to inhibit non-mammalian caspases. We determined the specificity of CrmA mutants bearing various residues in positions P4, P3 and P2 of the cleavage site. Almost all variants retained the ability to inhibit caspase-1, but many lacked caspase-8 inhibitory activity. The retention of Spi-2 proteins’ caspase-8 specificity during chordopoxvirus evolution, despite this function being readily lost through cleavage site mutagenesis, suggests that caspase-8 inhibition is crucial for poxviral pathogenesis and spread.
Viral evolutionary success not only requires the virus to enter a host cell and hijack its cellular machinery to enable viral replication: the infected cell must also be kept alive until infective viral particles can be produced and released, to spread the infection to other cells within that host and to other individuals. Poxviruses employ numerous strategies to achieve this goal , including the expression of multiple proteins that inhibit various components of host cell death pathways [2,3]. The Spi-2 family of proteins constitute one weapon used by poxviruses to prevent the destruction of infected cells. Experiments using mutated viruses revealed that representatives of this family, Myxoma virus Serp2 and Ectromelia virus Spi-2, were required for pathological hallmarks of myxomatosis [4,5] and mousepox , respectively, and to minimize anti-viral immune responses. Deletion of the Spi-2 ortholog from Vaccinia virus (B13R) attenuated its virulence and prolonged the survival of infected CB6F1 mice , although other studies using different strains of mice and inoculation routes failed to discern a link between B13R and pathogenesis [8,9]. Spi-2 proteins belong to Clade N of the serpin superfamily . Most inhibitory serpins disable serine proteases, but the first described molecular target of a poxviral Spi-2 protein was the cysteine protease caspase-1 . That study, and most subsequent investigations into the biochemistry of Spi-2 proteins, focussed on the Cowpox virus Spi-2 protein CrmA.
In addition to efficiently inactivating caspase-1 and its paralogs caspases-4 and -5 [12–14], CrmA potently inhibited the extrinsic apoptotic protease caspase-8 [14,15]. In contrast, CrmA was a relatively weak inhibitor of executioner apoptotic caspases [12,14], possibly due to steric clashes between CrmA and a loop present on these proteases but absent from caspases-1 and -8 . The caspase-8 paralog caspase-10 was also relatively poorly targeted by CrmA [12,17]. In vitro evidence suggested that CrmA may also target the serine protease granzyme B, although much less efficiently than caspases-1 or -8 [18,19].
Spi-2 proteins primarily contribute to poxviral pathology by suppressing inflammatory and natural killer cell responses within infected hosts [6,20–22], and all three proteases reported to be targeted by CrmA play roles in these processes. Caspase-8 transmits apoptotic signals from death receptors like Fas, which could enable the destruction of virally infected cells by natural killer cells or cytotoxic T cells expressing Fas ligand [23,24]. Caspase-1 (assisted by caspases-4 and -5) processes precursors to generate the active inflammatory cytokines IL-1β and IL-18, promoting the recruitment of inflammatory cells to sites of viral infection . Caspase-8 has also been implicated in interleukin (IL)-1β and IL-18 production in some circumstances . Granzyme B is released via granules from natural killer cells and cytotoxic T cells into target cells, so could eliminate virally infected cells [24,27]. In vitro, CrmA inhibited granzyme B much less potently than caspases-1 and -8 [18,19]. Furthermore, granzyme B-deficient mice were substantially more susceptible to mousepox than wild-type mice , indicating that granzyme B could limit the pathogenesis of Ectromelia virus, despite its expression of a Spi-2 protein. This observation implies that ECTV Spi-2 did not neutralize granzyme B activity in vivo and reinforces the notion that this protease is not a biologically significant target of Spi-2 proteins.
Poxviruses probably acquired and retained Spi-2 proteins targeting caspase-1 and caspase-8 in response to selective pressure to withstand anti-viral immune responses involving caspase-1-driven inflammation and/or caspase-8-dependent destruction of infected cells by cytotoxic lymphocytes. However, it is not currently clear whether poxviruses benefit from expressing Spi-2 orthologs primarily because they inhibit either caspase-1 or caspase-8, or whether both proteases must be inhibited for the poxvirus to evade immune destruction and maximize poxviral propagation and spread.
CrmA inhibits its target proteases via a pseudosubstrate mechanism of action. The interaction between the protease and a cleavage site within the ‘reactive center loop’ (RCL) of CrmA is proposed to provoke a rapid and dramatic conformational change that converts this external loop to a strand within a β-sheet. Stable complexes have been detected between CrmA and its target caspases [19,29,30], consistent with a model in which the enzyme's catalytic dyad becomes distorted, leading to the thioacyl enzyme intermediate being trapped so the cleavage reaction is not completed. Caspase-1 and -8 heterotetramers were observed to dissociate into separate subunits during this process, with the larger subunit remaining covalently linked to CrmA . The only experimentally derived structures of Spi-2 family members were of CrmA cleaved by subtilisin  or an unidentified bacterial protease . These enzymes cleaved CrmA at sites distinct from, but near to, the caspase scissile bond that links residues denoted P1 and P1′. Those structures confirmed that, as in other cleaved serpins, the CrmA RCL sequence was inserted within a β-sheet in the core of the molecule. Unfortunately, structures have not been determined for the encounter complex, nor the final complex between the serpin and the trapped caspase. The Spi-2 residues responsible for disabling the caspase's active site after the conformational change may substantially influence specificity but have not been explored to date; specificity studies have instead focussed on exploring the initial interaction between caspases and the native (RCL-exposed) form of CrmA, extrapolating from knowledge of caspase/substrate interactions. As expected for a pseudosubstrate caspase inhibitor, mutating the P1 aspartate within CrmA abolished its caspase-inhibitory activity . Modifying the preceding P4-P2 sequence substantially affected CrmA's specificity: replacing its native P4-LVAD-P1 sequence with residues occupying the cleavage site of the baculoviral caspase inhibitor p35 (DQMD) conferred the ability to inhibit developmental cell death in nematodes  and intrinsic apoptosis in mammalian cells . However, as expected based on the multiple steps required for host serpins to inhibit their target proteases, the affinity of a CrmA's P4-P1 sequence for a caspase's active site was not the sole determinant of specificity. Substituting the native P4-LVAD-P1 sequence with a sequence efficiently cleaved by caspase-3 (DEVD) did not empower CrmA to inhibit intrinsic apoptotic stimuli  and only marginally enhanced in vitro caspase-3 inhibition . The disappearance of full-length wild-type CrmA or CrmADEVD (but not the CrmADQMD mutant) from cells undergoing intrinsic apoptosis  implied that those proteins were degraded in apoptotic cells. An alternative explanation has also been proposed to explain the relative resistance of caspase-3 to CrmA: molecular modeling suggested that a loop present on caspase-3 but not caspases-1 or -8 may block entry of the CrmA pseudosubstrate site into the active site of caspase-3 . Whether steric hindrance or proteolysis accounts for inefficient inhibition of caspase-3 by CrmA, these data reinforce the involvement of factors other than the affinity of P4-P1 sequences for protease active sites in determining Spi-2 proteins’ specificities.
Very little biochemical information has been published about Spi-2 proteins from poxviruses other than Cowpox virus. Yeast experiments suggested that Ectromelia virus Spi-2, a close ortholog of CrmA, shared its caspase-1/8 dual specificity . Myxoma virus, a species of the Leporipoxvirus genus that infects rabbits, encodes a distant relative of Spi-2 called Serp2. Serp2 inhibited caspases-1, -8 and granzyme B in vitro, although reports of its potency varied by up to 800-fold [19,36,37].
If it is important for poxviruses to inhibit either caspase-1 or -8, but not both, and a narrower specificity is biochemically feasible given the structural similarities between caspases-1 and -8, we reasoned that some poxviruses may encode serpins that could efficiently inhibit only the one critical caspase. Alternatively, if suppression of both caspases is important for poxviral propagation in vivo, we would expect that all poxviral Spi-2 proteins would exhibit a similar dual specificity to CrmA, efficiently inhibiting caspases-1 and -8. We explored these possibilities by comparing the specificities of Spi-2 proteins encoded by a panel of orthopoxviruses, and by attempting to create CrmA variants that only inhibited caspase-1 or -8.
The following plasmids have been previously described: pGALL-(LEU2)-caspase-1 , pGALL-(LEU2)-caspase-2, pGALL-(URA3)-caspase-2 , pGALL-(LEU2)-rev-caspase-3 , pGALL-(LEU2)-caspase-4, pGALL-(URA3)-caspase-4 , pGALL-(LEU2)-caspase-5 , pGALL-(LEU2)-caspase 8 , pGALL-(LEU2)-CED-3, pGALL-(URA3)-CED-3 , pGALL-(LEU2)-revDRICE , pGALL-(LEU2)-DCP-1 , pGALL-(LEU2)-DRONC , pGALL-(URA3)-DRONC , pGALL-(HIS3)-CrmA, MBP-His6-FLAG-pMAL-c2X , pEF-FLAG-CrmA . pcDNAT7-c IPAF1–577 was provided by Teresa and Emad Alnemri, pEF.myc.ER-E2-Crimson  was a gift from Benjamin Glick (Addgene plasmid # 38770) and Paul Ekert provided pEF-CrmAT291R .
DNA encoding caspase-1 was amplified from pGALL-(LEU2)-caspase-1 with primers 1657 and 1639, and ligated into pEF cut with BamHI and XbaI . DNA encoding pro-IL-1β was amplified using primers 1636 and 1637, cut with BamHI and XbaI and ligated into pEF-FLAG  cut with BamHI and XbaI. DNA encoding amino-terminal His6-tagged pro-domainless caspase-1 was amplified using pEF-caspase-1 as a template and oligonucleotide 1688 and reverse vector primer 741. The product was digested with EcoRI and NotI and ligated into pET23a-noT7 cut EcoRI/NotI.
pGALL-(TRP1)-caspase-4 was made by inserting the BamHI/SphI fragment from pGALL-(LEU2)-caspase 4 into BamHI/SphI-digested pGALL-(TRP1)-MCS . To make pGALL-(TRP1)-rev-caspase-6, DNA encoding the pro-domain and large subunit of caspase-6 was amplified with oligonucleotides 893 and 894 and the resultant PCR product was digested with XhoI and XbaI. DNA encoding the small subunit was amplified with oligonucleotides 895 and 896 and the PCR product was digested with BamHI and XhoI. The digested PCR products were ligated into pGALL-(LEU2) that had been cut with BamHI and XbaI as a three-way ligation. pGALL-(HIS3)-FLAG-Bcl-xL was made by amplifying the Bcl-xL open reading frame with primers 626 and 627, cutting the product with BglII and XbaI and ligating into pGALL-(HIS3)-FLAG cut with BamHI and XbaI. To incorporate a sequence encoding a FLAG tag upstream of the polylinker of pGALL-(HIS3) (generating pGALL-(HIS3)-FLAG), oligonucleotides 660 and 661 were annealed then ligated into pGALL-(HIS3) that had been cut with BamHI and XbaI. The CrmA coding region was excised from pGALL-(HIS3)-CrmA  by digestion with BglII and XbaI and ligated into pGALL-(HIS3)-FLAG which had been digested with BamHI and XbaI.
To make pGALL-(LEU2)KanR, a kanamycin resistance cassette was excised from pDORR221 (Invitrogen) using BspHI, and the sticky ends blunted by incubation with Klenow polymerase and dNTPs. This fragment was ligated into pRS315  cut with ScaI, to yield pRS314-KanR. An EcoRI/SalI fragment containing the promoter, polylinker and terminator was cut out of pGALL-(LEU2), blunted and inserted into pRS314-KanR cut with PvuII, to produce pGALL-(LEU2)-KanR. BamHI/XbaI fragments from pGALL-(LEU2)-caspase-1 and pGALL-(LEU2)-caspase-8 plasmids were then subcloned into BamHI/XbaI-digested pGALL-(LEU2)KanR.
Custom pUC57 plasmids bearing human codon-optimized open reading frames of COTV165 and YKV165 (based on the annotated start site ) were purchased from Genewiz (South Plainfield, NJ, U.S.A.). To produce pGALL-(HIS3)-COTV165 and pGALL-(HIS3)-FLAG-COTV165, the COTV165 coding region was excised with EcoRI and XhoI and cloned into pGALL-(HIS3) , then excised with EcoRI and SphI and cloned into pGAL-(HIS3)-FLAG. The pGALL-(HIS3) construct was cut with BamHI and religated to remove extraneous polylinker sites. The coding region of YKV165+15 (commencing from the annotated translational start site 15 codons before the probable native start site) was excised from the custom plasmid using BamHI and XbaI and inserted into pGALL-(HIS3)  and pEF-FLAG . Yeast expression constructs encoding YKV165 starting at the second ATG (encoding a protein starting MFID…) were created by amplifying this template with forward primer 1701 and a vector reverse primer 1639, cutting the product with EcoRI and XbaI and ligating into pGALL-(HIS3) . The resulting plasmid was then cut with BamHI and religated to remove extraneous polylinker sites. To express FLAG-tagged YKV165 in yeast, the same EcoRI/XbaI-digested PCR product was ligated into pGALL-(HIS3)-FLAG cut with EcoRI and XbaI.
For mammalian expression, was amplified from the pEF-FLAG-YKV165+15 template using with the forward primer 1725 and vector reverse primer 741, the product was cut with BamHI and XbaI and ligated into pEF-FLAG  cut BamHI/XbaI. The COTV coding region was amplified from the custom pUC57 plasmid using forward primer 1725 and an M13 vector reverse primer (1036), cut with BamHI/XbaI and ligated into pEF-FLAG  cut BamHI/XbaI.
Custom pET15b plasmids were purchased from Genscript bearing E. coli codon-optimized open reading frames of VACCB13R, MPXV165, CMLV191, TPV-149R and DPV-167. The coding regions were amplified using forward primers 1859 (TPV-149R), 1860 (VACCB13R, MPXV165, CMLV191) or 1861 (DPV-167) and T7 terminator reverse primer 1853. The products were cut with BamHI and XhoI and ligated into pGALL-(HIS3) which had also been digested with BamHI and XhoI. To express FLAG-tagged versions of these proteins, inserts were removed from those plasmids through BamHI/XbaI digestion and ligated into pGALL-(HIS3)-FLAG (for yeast expression) or pEF-FLAG (for mammalian expression) which had also been digested with BamHI and XbaI.
The CrmA coding region was amplified using primers 1692 and 1639 using the template pGALL(HIS3)-CrmA , digested with EcoRI and XhoI, then cloned into pET23a-noT7tag  cut EcoRI/XhoI. The COTV165 coding region was amplified from pGALL(HIS3)-COTV-165 with primers 1723 and 1639, cut with EcoRI and XhoI and ligated into pET23a-noT7tag cut with EcoRI and XhoI. The YKV165 coding region was amplified from pGALL-(HIS3)-YKV165 using primers 1639 and 1691 digested with EcoRI and XhoI then cloned into pET23a-noT7tag.
To create plasmids encoding CrmA P4-P1 variants, the 3′ portion of the coding region was initially subcloned, introducing a silent mutation to create EcoRI site just 3′ to the region encoding the RCL, through amplification from pGALL-(HIS3)-CrmA  using forward primer 1659 and reverse vector primer 1639. The product was cut EcoRI/XbaI and ligated into pGALL-(HIS3) to produce pGALL-(HIS3)-CrmACterm. Expression constructs encoding P4-P1 CrmA variants were created by performing a PCR with the common forward primer 1729 and specific mutagenic reverse primers (LVAA: 1707, WEHD: 1702, XXXD: 1764), cutting the fragments with BamHI and EcoRI and ligating into pGALL-(HIS3)-CrmACterm cut with BamHI and EcoRI. To generate the CrmAXXXD library, bacteria were transformed with the ligation of the cut PCR product then grown overnight in liquid media for plasmid purification via a Qiagen midiprep. The SacI/XhoI fragments from the CrmA variant yeast expression plasmids were subcloned into pET23a (Novagen) cut with SacI and XhoI for bacterial expression. For mammalian expression, DNAs encoding wild-type and variant CrmA proteins were amplified using the yeast expression constructs as templates with primers 1827 and 1639, cut with BglII and XbaI fragments then cloned into pEF-FLAG cut with BamHI and XbaI.
Yeast techniques, CrmA variant library screening
The Saccharomyces cerevisiae strain W303α was used in this study. Yeast culturing, transformations, death assays, plasmid extraction and immunoblotting were performed as previously described . Antibodies for immunoblotting were ‘THETM Anti-His’ clone 6G2A9 (Genscript; Piscataway, NJ, U.S.A.) followed by anti-mouse IgG-HRP (Sigma–Aldrich; St. Louis, MO, U.S.A.). To identify CrmA P4-P2 variants exhibiting altered specificity, yeast bearing a pGALL-(LEU2)KanR-caspase-1 or pGALL-(LEU2)KanR-caspase-8 plasmid were transformed with the pGALL-(HIS3)-CrmAXXXD library and plated onto minimal selective inducing (galactose) or repressing (glucose) media. Colonies from the glucose-containing plates were replica plated onto selective inducing and repressing plates to determine the frequency of transformants bearing caspase-inhibiting CrmA variants. DNA was extracted from some of these colonies, and from some that arose on the galactose-containing transformation plates. This DNA was transformed into MC1061 bacteria, which were plated onto media containing ampicillin to select those bearing the CrmA library plasmids rather than caspase expressing plasmids. Plasmids were isolated from those bacterial clones for sequencing and characterization.
Caspases-2, -3, and -8 were purified as described . Caspase-1 was purified from Arctic Express (DE3) E. coli (Agilent Technologies, Santa Clara, CA, U.S.A.) following induction with 2 mM IPTG for 45 min at 30°C. The bacteria were pelleted then resuspended in BugBuster Mastermix (Merck, Darmstadt, Germany) (5 ml/g) containing 10 mM imidazole and gently agitated at room temperature for 20 min. The lysate was centrifuged at 16 000×g for 20 min at 4°C. 1 ml Ni-NTA agarose slurry (Qiagen, Hilden, Germany) per liter of culture was washed in 10 column volumes of PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4) containing 10 mM imidazole. Cleared bacterial lysate was passed through a 0.45 µm sterile filter before washed pre-chilled Ni-NTA agarose was added to lysate and gently agitated at 4°C for 30 min. Ni-NTA agarose-lysate slurry was added to a gravity purification column (Bio-Rad, Berkeley, CA, U.S.A.) at 4°C and washed with 10 column volumes of PBS containing 25 mM imidazole prior to elution using one column volume of PBS containing 200 mM imidazole. Glycerol (15%) was added to eluates which were aliquoted and snap frozen in liquid N2 prior storage at −80°C. Caspases were active site-titrated using a published method .
AcP35 was purified as previously described . Serpins and MBP-FLAG-His6 were purified using a method based on a published protocol . BL21(DE3)pLysS (Merck) bacteria bearing expression plasmids were induced using 50 µM IPTG at 20°C for 16 h then pelleted and frozen. Induced bacterial pellets were resuspended in Ni-NTA resuspension buffer (50 mM Tris [pH 8.0], 500 mM NaCl, 5 mM imidazole, 10 mM MgCl2) supplemented with a protease inhibitor cocktail (Roche, Basel, Switzerland) and cells were lysed by a single pass through a cell disruptor (Constant Systems, Northants, U.K.) at 30 kpsi. DNase (New England Biolabs, Ipswich, MA, U.S.A.) was added to the lysate to 1 mM and incubated at 4°C for 15 min with gentle agitation before being centrifuged at 16 000×g for 20 min at 4°C. Ni-NTA slurry (Qiagen) was washed with Ni-NTA resuspension buffer. Lysate supernatant was filtered through a 0.45 µm filter before washed Ni-NTA was added and allowed to incubate at 4°C with gentle agitation for 1 h. Ni-NTA-lysate slurry was applied to a gravity purification column before being washed with two column volumes of Ni-NTA wash buffer (50 mM Tris [pH 8.0], 500 mM NaCl, 15 mM imidazole). Bound proteins were eluted with elution buffer (50 mM Tris [pH 8.0], 100 mM NaCl, 150 mM imidazole). Eluates were mixed to 10% glycerol, aliquoted and snap frozen in liquid N2 before long-term storage at −80°C. The concentrations of active serpins in each sample were determined via titration with caspase-1, using a published method .
Caspase activity in the presence or absence of serpins was quantified by calculating the gradient of fluorescence generation over time, using reactions containing 100 µM of fluorogenic substrates Ac-WEHD-AFC, Ac-DEVD-AFC, Ac-LEHD-AFC (Enzo Life Sciences, Farmingdale, NY, U.S.A.) or Ac-VDTTD-AFC (21st Century Biochemicals, Marlborough, MA, U.S.A.) in citrate buffer (10 mM HEPES [pH 7.0], 10% sucrose, 0.1% CHAPS, 10 mM freshly added DTT, 100 mM NaCl, 1 mM EDTA, 0.65 M Na-Citrate). The inverse of the gradient of (vo/vi) − 1 (where vo is the gradient of uninhibited caspase and viis the gradient of inhibited caspase) was plotted against the serpin concentration to calculate the inhibition constant in the presence of substrate ‘Ki(app)’ . The caspase's affinity for its substrate and the substrate concentration were taken into consideration using the formula Ki = Ki(app)/(1 + S/KM) where S is the substrate molar concentration and KM is the Michaelis constant, which was calculated as previously described .
Mammalian cell assays
LN18 and 293T cells were cultured in high glucose Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, U.S.A.) with 10% fetal calf serum (Invitrogen) at 37°C with 5% CO2. Transfections were performed using Lipofectamine 2000 or Lipofectamine Plus (Invitrogen) for 293T and LN18 transfections, respectively. Transfected LN18 cells were cultured in media containing 3 µg/ml puromycin until large colonies were visible. Using a dilute trypsin solution, isolated colonies were harvested and expanded for assessment of transgene expression, then phenotypic testing. Transiently transfected 293T cells were harvested 24 h post-transfection. A small proportion of the cells were analyzed by flow cytometry to determine the transfection frequency using a FACS-Canto (BD Biosciences, Franklin Lakes, NJ, U.S.A.). Lysis and immunoblotting of mammalian cells were performed as previously reported , using anti-IL1β clone 8516 (R&D Systems, Minneapolis, MN, U.S.A.), anti-FLAG M2 (Sigma–Aldrich) or anti-GAPDH clone 6C5 (Merck) followed by anti-mouse IgG-HRP (Sigma–Aldrich). LN18 cells were treated with agonistic α-DR5 antibody (MAB #631; R&D Systems) or ABT-737 (Selleck Chemicals, Houston, TX, U.S.A.) then propidium iodide uptake was measured by flow cytometry as previously published .
The NCBI BLAST search engine (https://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to identify genomes encoding proteins with homology to Cowpox virus CrmA (NP_619988.1), using the non-redundant protein database limited to poxviridae. Sequences of orthologs from Vaccinia virus (accession KC201194), Monkeypox virus (AY603973.1), Camelpox virus (AY009089.1), Cotia virus (HQ647181.2), Yoka virus (NC_015960.1), Tanapox virus (EF420157.1) and Deerpox virus (NC_006966.1) were aligned using Multialign . The P4-P1 residues in caspase-inhibitory CrmA variants were presented in a ‘Logo’ format .
Models of caspase-1 and -8 were generated using Coot , using published structures of caspase-1 (PDB 2HBQ)  and caspase-8 (PDB 4PRZ and 3H11) [63,64]. Residues in the originally bound peptides were substituted with the P4-P1 residues from wild-type or variant CrmA proteins, and the most stable conformations for the introduced residues that avoided steric clashes were selected. PyMOL (DeLano Scientific, San Carlos, CA, U.S.A.) was used to create the figures.
The frequencies of basic residues were compared at each position in the caspase-1-specific inhibitors versus the dual-specificity inhibitors using χ2 tests using Microsoft Excel. P values were subjected to Bonferroni corrections for multiple (3) comparisons. Differences in ABT-737 and anti-DR5 sensitivity of stable transfectant clones versus parental cells were analyzed using two-way ANOVAs with ad hoc Dunnett correction for multiple comparisons, using GraphPad Prism (San Diego, CA, U.S.A.).
To select Spi-2 orthologs for specificity characterization, translated poxviral genomes were searched for sequences predicted to encode proteins similar to Cowpox virus CrmA. Orthologs from seven chordopoxviruses were chosen for analysis. This panel included three orthopox viruses (Vaccinia virus, Camelpox virus and Monkeypox virus) and four chordopoxviruses from other genera: Tanapox virus (from the Yatapoxvirus genus), Deerpox virus (a Cervidpoxvirus), Yoka virus (a Centapoxvirus) and Cotia virus (which is currently unassigned) (Figure 1). The predicted start site for CrmA aligned with the experimentally established amino terminus of its Vaccinia virus counterpart B13R , and with the predicted translational start sites of all other orthologs except for YKV-165 from Yoka virus (Figure 1A). Bioinformatics analyses of the YKV-165 locus (Figure 1A) led us to suspect, however, that translation of YKV-165 actually commences 15 codons downstream of the predicted start site , at a position homologous to the validated translation initiation codon of VACV-B13R and the predicted start sites of the other relatives. Previous work identified a consensus promoter element required for transcription of CrmA  and other poxvirus early genes . An identical sequence was present in the Yoka virus genome, just downstream of the proposed translation initial site for YKV-165, implying that the proposed initiation codon would not be transcribed. Furthermore, the region surrounding the downstream ATG complied better with Kozak consensus initiation sequence  than the previously annotated upstream ATG. We, therefore, chose to express and analyze YKV165 bearing the amino-terminal sequence MDIF in this study. As expected from the phylogenetic relationships of these viruses, the sequences of the orthopoxvirus Spi-2 relatives were highly homologous to CrmA (89–90% identical), but the relatives from other poxvirus genera were less similar (37–55% identical) (Figure 1B). Key features required for the protease inhibitory activity of CrmA were conserved in each of the orthologs, including the P1 aspartate residue that interacts with the caspase active site , and a crucial threonine residue within an upstream hinge region  that facilitates the post-cleavage conformational change characteristic of serpins’ protease inhibitory mechanism . Interestingly, although all orthologs bore a P1 aspartate residue, only the Camelpox virus and Monkeypox virus orthologs shared the entire P4-P1 sequence with CrmA (LVAD). The identity of P4-P2 residues substantially influences the efficiency with which caspases cleave substrates , and altering these residues in CrmA modified its caspase-inhibitory profile [33,34], so we wondered whether the distinct sequences in other poxvirus Spi-2 proteins may impart distinct specificities to the orthologs, relative to CrmA.
Multiple sequence alignments of the Spi-2 proteins analyzed in this study.
We initially exploited a yeast-based system to evaluate the specificity of the CrmA orthologs. High-level expression of auto-activating caspases kills yeast, and co-expression of inhibitors with the appropriate specificity can prevent this toxicity [53,70]. We first checked tolerability of the Spi-2 proteins in yeast and their expression levels. Expression of neither untagged nor amino-terminal FLAG-tagged versions of each protein affected yeast growth appreciably (Figure 2A). Immunoblotting confirmed that each of the FLAG-tagged serpins was expressed in yeast, but their levels varied. Cowpox virus CrmA, Vaccinia virus B13R, Cotia virus COTV-165 and the negative control Bcl-xL were expressed at similar levels in yeast. CrmA was expressed at a lower level than the Camelpox virus ortholog CMLV-191 but at a higher level than Deerpox virus DPV-167, Tanapox virus TPV-149R and Monkeypox virus MPXV-165. Each inhibitor suppressed yeast death triggered by expression of the inflammatory caspases-1, -4 or -5 or the extrinsic apoptotic initiator protease caspase-8 (Figure 3, Supplementary Figure S1). All Spi-2 proteins conferred almost complete protection against death mediated by caspases-1 and 5 but provided less efficient protection against the toxicity provoked by caspase-4 or -8 expression. Those contexts revealed subtle differences between the inhibitors. Consistent with its relatively weak expression in yeast (Figure 2B), DPV-176 inhibited yeast death triggered by expression of caspases-4 or -8 less efficiently than the other orthologs. FLAG-tagged YKV-165 was slightly more active than untagged YKV-165, but epitope tagging did not substantially affect the activity of the other proteins. Our inability to evaluate the expression of untagged YKV-165 meant we could not determine whether the addition of the tag boosted the expression, stability and/or specific activity of the YKV165 protein.
Yeast tolerate the expression of poxviral Spi-2 proteins.
Spi-2 proteins from diverse poxviruses inhibit yeast death triggered by caspases-1, -4, -5 and -8 but not caspases-2, -3, or -6.
In yeast, as in vitro [12,14,16], CrmA was a weak inhibitor of apoptotic executioner caspases-3 and -6, and caspase-2 (whose function remains somewhat enigmatic)  (Figure 3, Supplementary Figure S1). In general, the orthologs shared this specificity, although COTV165 could impede caspase-6-mediated yeast death, albeit very inefficiently (Figure 3, Supplementary Figure S1). The viruses that encode these serpins infect mammalian cells, but we were also curious about the Spi-2 proteins’ abilities to inhibit caspases from other, non-host animals. We observed marked differences in the orthologs’ specificity with respect to insect and nematode caspases. Consistent with published data , CrmA had little impact on CED-3 activity in yeast (Figure 4, Supplementary Figure S2), and its counterparts from Vaccinia, Monkeypox, Camelpox and Tanapox viruses were similarly ineffective. In striking contrast, the Cotia virus Spi-2 protein, COTV-165, completely prevented CED-3-mediated yeast death, and less potent protection was also conferred by DPV-176 and tagged YKV165. The panel of Spi-2 proteins also differed in their inhibition of yeast death mediated by the initiator Drosophila caspase DRONC. We confirmed the previous observation that CrmA could inhibit DRONC  and most of the other serpins also strongly suppressed the lethality associated with expression of this caspase, but those from Vaccinia and Tanapox viruses were completely ineffective (Figure 4, Supplementary Figure S2). None of the Spi-2 proteins rescued yeast from death triggered by the Drosophila executioner caspases Dcp-1 or DrICE, although COTV165 and DPV-167 were very slightly protective (Figure 4, Supplementary Figure S2). The Spi-2 proteins would never encounter these caspases in nature, so these data do not provide any insight into the evolutionary forces that shaped viral serpin specificity, but the distinct specificity profiles of the different serpins may have useful applications in nematode or insect research and may inform structural biology studies of the caspase and Spi-2 families.
Poxviral Spi-2 proteins differ in their abilities to rescue yeast from death provoked by nematode or insect caspases.
Biochemical assays were performed using purified caspases and viral serpins (except DPV-167, which could not be purified with sufficient yield) to quantify the serpins’ specificities and potencies. None of the tested Spi-2 proteins prevented recombinant caspases-2 or -3 from cleaving peptide substrates (Figure 5A,B), although these caspases were sensitive to inhibition by the baculoviral pan-caspase inhibitor AcP35, as expected . The serpins all potently inhibited recombinant caspase-1. The Kis for these reactions clustered around 10 pM (Figure 5C–E, Supplementary Figure S3A), and the Ki calculated for CrmA's inhibition of caspase-1 (11 pM) closely matched previously reported figures [12,13]. All Spi-2 proteins also targeted caspase-8, but less efficiently and more heterogeneously than caspase-1, but even the weakest caspase-8 inhibitor (TPV-149R, Ki = 2 nM) possessed sufficient activity to suggest biological relevance [74,75]. The Ki for caspase-8 inhibition by CrmA (730 pM) fell between the previously documented values of 950 pM  and <340 pM .
Purified poxviral Spi-2 proteins inhibit the proteolytic activity of caspases-1 and -8 but not caspases-2 or -3.
The yeast and biochemical assays demonstrated that even distantly related CrmA relatives shared its ability to inhibit both caspase-1 and caspase-8. This dual specificity could be due to evolutionary pressure on diverse poxviruses to simultaneously suppress inflammatory cytokine production and death ligand-induced apoptosis. Alternatively, evolution may have selected for the ability to impede just one of these processes, via inhibition of either caspase-1 or -8, but the biochemical similarity of these proteases may have resulted in the other being coincidentally inhibited even in the absence of overt selective pressure. The latter hypothesis would predict that mutagenesis of Spi-2 genes would be unlikely to yield variants that only inhibited either caspase-1 or -8. Altering the P4-P2 residues of CrmA had been previously established to alter its specificity [33,34], so we investigated whether modification of the P4-P2 sequence could produce CrmA variants with specificities for either caspase-1 or 8 but not both. Using PCR with degenerate oligonucleotides and a template encoding a non-functional CrmA bearing a P1 alanine residue, we created a yeast expression library encoding CrmA variants possessing random residues in the P4-P2 positions and an invariant P1 aspartate. This library was transformed into yeast bearing a plasmid encoding caspase-8, and then the yeast were plated onto inducing selective media. The resulting colonies were presumed to have acquired library plasmids encoding CrmA variants capable of inhibiting caspase-8. These plasmids were isolated, then transformed into yeast bearing plasmids encoding either caspase-8, to verify its inhibition, or caspase-1, to check whether each caspase-8-inhibitory variant could also inhibit caspase-1. The opposite screen was also conducted, in which variants were selected for caspase-1-inhibitory activity, then tested for caspase-8 inhibition. All variants that inhibited caspase-8 also inhibited caspase-1, however only some of those that inhibited caspase-1 also inhibited caspase-8 (Figure 6A).
Some CrmA P4-P1 variants inhibit yeast death provoked by expression of caspase-1 but not caspases-2, -3 or -8.
We assessed the proportion of library plasmids that encoded caspase-1 or -8 inhibitors by plating library transformation mixtures onto selective repressing media, then replica plating co-transformants onto inducing and repressing plates. Approximately, two-thirds of co-transformants bearing library and caspase-1 plasmids survived transgene induction. Sequencing of those clones’ library plasmids revealed that almost all of the variants that failed to confer protection against caspase-1-mediated lethality contained the P4-P1 sequence LVAA. This was the P4-P1 sequence of the template used for the PCR reaction used to construct the library, which was expected and experimentally confirmed to be inactive (Figure 6B). Only two library plasmids that encoded CrmA variants with P1 aspartate residues failed to protect yeast from caspase-1-induced death. They bore the P4-P1 sequences PPPD and GRGD (Figure 6C). One-third of the library transformants survived co-expression with caspase-8. Upon purification of those library plasmids and retransformation, we ascertained that each of those variants could also inhibit caspase-1. Half of the caspase-8 non-inhibitors bore the LVAA P4-P1 sequence, indicating they were residual template plasmids from the original library construction; the other half harbored P1 aspartate residues, and all of these could also protect yeast from caspase-1-induced death. These analyses, therefore, confirmed the data from the screens for variants that could rescue yeast from caspase-1 or -8-dependent death: in the context of a CrmA variant bearing a P1 aspartate residue, the vast majority of P4-P2 sequences were compatible with caspase-1 inhibition, however, only a subset of those could also inhibit caspase-8. None of the CrmA P4-P2 variants we characterized could inhibit caspase-8 but not caspase-1.
The P4-P2 regions of all functional variants were sequenced (Figure 6A). The caspase-1-specific CrmA variants were more likely than the dual-specificity variants to contain basic residues in the pseudosubstrate site (P = 0.047 for P4 and P2), implying that positive charges in this region interfered with the ability of CrmA to either interact with and/or inhibit caspase-8 but did not impact on its ability to target caspase-1.
We also generated a CrmA variant that we hoped would be caspase-1-selective, by replacing its native P4-P1 residues (LVAD) with a sequence (WEHD) reported to be cleaved 376-fold more efficiently by caspase-1 than caspase-8 in a peptide context . We evaluated the ability of this rationally engineered mutant, and selected caspase-1-specific variants from the screens, to block yeast death provoked by caspases-1, -2, -3 or -8. Like wild-type CrmA, the variants all inhibited caspase-1 activity in yeast but failed to rescue yeast from caspase-3-dependent death (Figure 6B). The rationally designed mutant (CrmAWEHD) possessed the weak caspase-8-inhibitory activity and the selected screen-derived variants had a negligible impact on caspase-8-mediated lethality (Figure 6B). A P1 mutant, CrmALVAA, was non-functional as expected. Lysates from yeast expressing these functional variants and the inactive mutants (but no caspases) were immunoblotted. All variants were detected although most were expressed at somewhat lower levels than wild-type CrmA (Figure 6D). Intriguingly, the amino-terminally His6-tagged CrmALVAA and CrmAGRGD non-functional variants migrated slightly faster through SDS–PAGE than wild-type CrmA and the other mutants. This small apparent size difference would be consistent with proteolytic removal of the carboxyl terminus of these proteins, perhaps via cleavage by a yeast protease within this altered RCL. The expression and migration of the other inactive variant, CrmAPPPD, was similar to the wild-type protein, arguing that its inactivity was not due to proteolysis.
Selected CrmA variants were bacterially expressed and purified, and subjected to biochemical assays to quantitate their caspase-inhibitory potency. None inhibited caspases-2 or -3 (Figure 7A,B). All four variants possessed potent caspase-1-inhibitory activity, like wild-type CrmA (Figure 7C). Substitution of the P4-P1 residues of CrmA (LVAD) with WEHD decreased the efficiency of caspase-8 inhibition. Variants bearing REKD, KNKD or KGWD sequences in their RCPs were extremely poor caspase-8 inhibitors (Figure 7D,E).
Some purified CrmA P4-P1 variants inhibit the proteolytic activity of caspase-1 but not caspases-2, -3 or -8 in vitro.
Plasmids encoding FLAG-tagged versions each of these CrmA variants, Spi-2 orthologs or control proteins were stably transfected into LN18 human glioma cells. Clones were evaluated for transgene expression (Figure 8A) and sensitivity to intrinsic and extrinsic apoptosis (Figure 8B,C). None of the Spi-2 proteins conferred resistance to the BH3-mimetic ABT-737 (Figure 8B), which disables Bcl-2, Bcl-xL and Bcl-w to promote intrinsic apoptosis . This is consistent with our biochemical and yeast assay data showing that none of the CrmA orthologs or variants could inhibit caspase-3 and implies that they, like wild-type CrmA  were also unable to target cellular caspase-9. In contrast, responses of the stable transfectants to the extrinsic apoptotic stimulus anti-DR5 varied substantially. Reflecting their abilities to inhibit caspase-8 in yeast and in vitro, expression of the poxvirus Spi-2 proteins CrmA, YKV-165, COTV-165, DPV-167, or TPV-149R protected LN18 cells from death triggered by anti-DR5 (Figure 8C). In contrast, cells expressing CrmA mutants that failed to efficiently inhibit caspase-8 in yeast or in vitro were susceptible to extrinsic apoptotic signaling. CrmA mutants bearing REKD or WEHD residues in their P4-P1 sites were as sensitive to anti-DR5 toxicity as the parental cells, but both clones expressing CrmAKGWD were slightly less sensitive than the parental cells. A clone expressing an inactive hinge mutant of CrmA (CrmAT291R)  was somewhat more sensitive to anti-DR5-induced killing than the parental cells. This enhanced sensitivity could be a result of expression of the CrmA loss-of-function mutant, or may simply reflect clonal variability, as earlier experiments showed that this mutant did not render cells more sensitive to killing by anti-Fas or TNFα .
Spi-2 orthologs inhibit extrinsic but not intrinsic apoptosis; caspase-1-specific CrmA variants confer less protection from extrinsic apoptosis than wild-type CrmA.
The caspase-1-inhibitory activity of selected CrmA relatives and variants was also assessed in mammalian cells, using a transient transfection approach. Co-expression of caspase-1 and a constitutively active version of its activator Ipaf  led to cleavage of pro-IL-1β (Figure 9). This processing was prevented by co-expression of Spi-2 proteins CrmA, COTV-165 and YKV-165 or the CrmA variants bearing the sequences WEHD or REKD in their RCLs, confirming that each of these proteins could inhibit caspase-1 activity in a cellular context.
Spi-2 orthologs and caspase-1-specific CrmA variants block caspase-1-mediated IL-1β cleavage in 293T cells.
Earlier work had ascertained that Spi-2 proteins from Cowpox virus [12–14] and its close relatives Vaccinia virus  and Ectromelia virus  could potently inhibit caspases-1, -4, -5 and -8 but inhibited executioner apoptotic caspases and granzyme B too weakly to be biologically meaningful. Researchers have also explored the specificity of a more distant relative, Myxoma virus Serp2, whose protein sequence is 57% identical with CrmA . Although two groups reported Serp2-mediated inhibition of caspases-1, -8 and granzyme B in vitro, their calculations of the efficiency of this inhibition differed markedly [19,37]. This study encompassed the characterization of the specificity of seven Spi-2 orthologs, whose homologies to CrmA ranged from 37% to 90% identical. The P4-P1 sequences of the close orthologs from orthopox viruses were identical (LVAD) or similar (LVSD) to the corresponding sequence within CrmA, but the orthologs from more distantly related viruses were more divergent (LMCD, LITD, CVAD and CVTD). Despite this, all of the Spi-2 proteins shared with CrmA the ability to potently inhibit caspases-1 and -8. This suggested to us that either this dual specificity conferred an evolutionary benefit to these diverse poxviruses, or that the viruses had experienced selective pressure to inhibit one of these caspases and the other was merely coincidentally inhibited because of its biochemical similarity. We sought to distinguish between these alternatives by investigating whether experimental mutagenesis could yield a variant of CrmA that only inhibited one of these target proteases. By mutating the P4-P2 sequence of CrmA, we identified numerous caspase-1-specific derivatives. Around a third of the variants, we surveyed that contained the key P1 aspartate residue lacked the ability to inhibit caspase-8. In contrast, all but two of the variants retained the power to block caspase-1 activity. Thus, random mutation of the pseudosubstrate cleavage site of CrmA often abolished caspase-8 inhibitory potential, yet such mutations seem not to have arisen during poxviral evolution. We, therefore, suspect that the ability to inhibit caspase-8, which would protect against immune destruction mediated by death ligands, has been crucial for poxviral persistence throughout evolution.
The near-universal ability of CrmA P4-2 variants to inhibit caspase-1 was astonishing. One of the two variants that could not inhibit caspase-1 or -8 possessed proline residues in P4, P3 and P2 positions. Peptides bearing a proline residue in the P3 site were previously shown to be completely resistant to human caspase-1-mediated cleavage, and proline residues in P4 or P2 also hampered peptide cleavage [69,79]. Despite this, two of the variants which inhibited caspase-1 but not caspase-8 contained prolines at P3 (IPPD and PPSD), and other functional variants included proline at one or two of the three degenerate positions. Evidently, the structural distortion imposed by one or two proline residues did not prevent CrmA from interacting with and inhibiting caspases. There are many possible explanations for the inability of CrmA featuring three consecutive prolines in its pseudosubstrate site (which may form a helix structure ) to inhibit caspase-1. These residues may affect the stability or solubility of the protein, prevent the RCL from docking with the caspase active site, interfere with conversion of the cleaved loop to form a new β-strand, or prevent the deformation of the caspase following that conformational change, enabling completion of the cleavage reaction and release of the active enzyme. The CrmAPPPD variant was expressed at similar levels to wild-type CrmA in yeast, and migrated similarly, arguing that its inactivity was not due to degradation or cleavage. In contrast, the other non-functional variant identified during our screen, CrmAGRGD, migrated slightly faster than wild-type CrmA during SDS–PAGE, which may reflect proteolysis somewhere within this altered RCL by a yeast protease. Wild-type CrmA has been published to be sensitive to proteolysis within the RCL by other proteases at sites distinct from the position targeted by caspases: a bacterial protease cleaved between P4 and P3 (L↓VADCAST)  and subtilisin cleaved between P2′ and P3′ (LVADC↓AST) . It is there conceivable that mutagenesis from P4-LVAD-P1 to P4-GRGD-P1 rendered CrmA susceptible to cleavage and inactivation by a yeast protease, and therefore preventing it from targeting caspases we co-expressed in yeast. Curiously, the P1 mutant we designed, CrmALVAA, also exhibited an anomalously faster migration pattern. This mutant was expected to be non-functional because its P1 alanine could not productively interact with caspase active sites, however, these data suggest that proteolysis by a yeast protease may also have contributed to its inability to suppress caspase-induced yeast death.
All other CrmA P4-P2 variants we identified inhibited caspase-1-dependent yeast death. The caspase-1 active site features larger S4 and S2 pockets than other caspases , which may contribute to its promiscuity during its relatively short lifespan . We were nevertheless amazed that CrmA variants bearing almost any P4-P2 sequence could block caspase-1-mediated yeast lethality. It was particularly surprising that many caspase-1-specific CrmA variants featured lysine or arginine in P4-P2 residues, given that published work suggested that these amino acids would be expected to repel arginine residues in the S4 and S3 pockets of caspase-1  and their inclusion in peptides reduced caspase-1-mediated cleavage in a positional library scanning context . However, structural modeling revealed that the electropositive nature of the caspase-1 active site did not preclude it accommodating the P4-P1 residues of wild-type CrmA (LVAD) or the most caspase-1 selective variant we characterized (REKD) (Figure 10). The S2 and S4 subsites in caspase-1 are flanked by L2 and L4 loops that are mostly hydrophobic and allow interaction with a diverse set of amino acids in P2 and P4 positions.
Modeling of the binding mode of P4-P1 residues of wild-type CrmA (LVAD) and a caspase-1-selective variant (REKD) with caspase-1 and caspase-8.
Only a subset of the CrmA variants that inhibited caspase-1 could also inhibit caspase-8. The most distinctive difference between the dual-specificity CrmA variants, relative to those that inhibited caspase-1 but not -8, was the dearth of basic residues in the P4 and P2 positions within the CrmA mutants that could inhibit caspase-8. We reasoned that positively charged residues in P4 or P2 may have abolished caspase-8 targeting at either an early stage of the inhibitory mechanism, by blocking access of the P1 aspartate to its catalytic dyad residues, or later, by preventing the serpin from effectively trapping the covalently associated enzyme. Modeling led us to favor the former explanation: the caspase-8 active site was substantially more electropositive than that of caspase-1, particularly the S2 subsite (Figure 10). Basic residues such as K in P2 would experience charge repulsion, which may prevent the P1 aspartate from being positioned deeply enough into the S1 pocket of caspase-8 to interact with its catalytic cysteine.
Most CrmA variants were only examined in the context of caspase-dependent yeast death, but a subset of the caspase-1-specific derivatives was also evaluated biochemically and in mammalian cells. Those assays verified that CrmAREKD, CrmAKGWD and CrmAWEHD could inhibit caspase-1 but had severely impaired caspase-8 inhibitory activity. CrmAREKD was the most specific of the variants we studied: it efficiently targeted caspase-1 in vitro and in mammalian cells yet its expression failed to suppress extrinsic apoptosis in LN18 cells and the purified protein failed to significantly impede the proteolytic activity of caspase-8 in vitro.
Many endogenous serpins require exosite interactions to achieve maximal inhibitory activity , and exosites have been identified that influence caspases’ interactions with their substrates [84,85], but the only exosite interaction postulated to date between poxviral Spi-2 proteins and their targets was a clash between CrmA and a loop on caspase-3 that was speculated to account for that enzyme's resistance to CrmA-mediated inhibition . Our observation that CrmA variants whose P4-P2 residues seemingly lack strong affinity for the caspase-1 active site nevertheless potently inhibited the enzyme prompted us to wonder whether exosite interactions may help to orient the RCL to facilitate insertion of the P1 aspartate within the active site of enzyme. Further work, such as determination of an encounter complex structure, would be required to test this hypothesis and to define any additional sites of interaction between Spi-2 proteins and caspase-1 (or other caspases) that may encourage positioning of the Spi-2 cleavage site within the RCL into the caspase's active site. Exosites, either facilitating or impeding interactions between poxviral serpins and caspases, may also account for the strikingly differential specificity of Spi-2 orthologs for non-mammalian caspases. CED-3-mediated yeast death was strongly inhibited by COTV-165 and more weakly by DPV-165 and YKV-165, but not by the orthopoxvirus Spi-2 proteins nor TPV-165. DRONC-induced toxicity was suppressed by all Spi-2 orthologs except VACV-B13R and TPV-149R. VACV-B13R was completely ineffective in this context despite sharing 90–95% of its amino acid sequence with its DRONC-inhibitory relatives CrmA, MPXV-165 CMLX-191. The sequences of VACV-B13R and CrmA RCL surrounding the scissile bond are identical except for P2 (V versus S) and P5′ (V versus I) residues, so we suspect that the dramatic difference in their DRONC-inhibitory activity reflects differences elsewhere in their sequences. Presumably, structural elements of DRONC and VACV-B13R specifically prevented (or failed to facilitate) their interaction, but did not preclude DRONC inhibition by other Spi-2 proteins nor prevent VACV-B13R from efficiently targeting other caspases.
The caspase-8 inhibitory function of Spi-2 proteins was readily abolished via mutagenesis of the pseudosubstrate cleavage site, implying that selective pressure may have contributed to the retention of this easily lost activity during poxviral evolution. This hypothesis would predict that infection with Spi-2-deleted poxviruses would provoke less severe symptoms in wild-type mice than animals lacking caspase-8 (and RIPK3 or MLKL, to overcome necroptosis-mediated embryonic lethality [86,87]), and that recombinant poxviruses in which the native Spi-2 gene is substituted with a caspase-1-specific derivative (such as a P4-REKD-P1 variant) would be less pathogenic and infectious than wild-type viruses. It is possible that Spi-2 proteins are encoded by poxviruses solely or primarily to target caspase-8, with co-inhibition of caspase-1 merely being a biochemical by-product. Alternatively, poxviruses may need to inhibit both caspases to evade immune-mediated destruction to replicate and spread. Analysis of the pathogenesis of Spi-2-deficient viruses to wild-type mice versus caspase-1 null mutants may help resolve this issue.
The caspase-1-specific Spi-2 variants characterized in this study may assist researchers to delineate the involvement of caspase-1 versus caspase-8 in various biological pathways. The anti-inflammatory properties of CrmA and the Myxoma virus Spi-2 ortholog Serp2 have been investigated as approaches to control inflammation provoked by vascular injury or disease  or arthritis . Caspase-1-specific derivatives of these serpins, such as the P4-REKD-P1 variant, may provide the desired anti-inflammatory activities in these contexts while avoiding concomitant suppression of beneficial caspase-8-mediated functions such as elimination of infected or malignant cells by cytotoxic lymphocytes, and dampening of necroptotic signaling.
C.J.H. conceived the study; D.T.B. and C.J.H. designed experiments; D.T.B., T.K.-I., D.P.-E., Y.J. and M.A.M. performed experiments; D.T.B., B.H. and C.J.H. interpreted data; D.T.B., B.H. and C.J.H. generated figures; D.T.B., B.H. and C.J.H. wrote the text. All authors reviewed the manuscript and approved its submission.
This work was funded by a project grant from the Australian National Health and Medical Research Council (#602525).
We thank Teresa and Emad Alnemri, Benjamin Glick and Paul Ekert for providing plasmids, and Lakshmi Wijeyewickrema for valuable feedback on this manuscript.
The Authors declare that there are no competing interests associated with the manuscript.
Present address: Biotechnology Manufacturing Facility, CSL Behring, Broadmeadows, Victoria 3047, Australia