Rare human Caspase-6-R65W and Caspase-6-G66R variants identify a novel regulatory region of Caspase-6 activity

The cysteine protease Caspase-6 (Casp6) is a potential therapeutic target of Alzheimer Disease (AD) and age-dependent cognitive impairment. To assess if Casp6 is essential to human health, we investigated the effect of CASP6 variants sequenced from healthy humans on Casp6 activity. Here, we report the effects of two rare Casp6 amino acid polymorphisms, R65W and G66R, on the catalytic function and structure of Casp6. The G66R substitution eliminated and R65W substitution significantly reduced Casp6 catalytic activity through impaired substrate binding. In contrast to wild-type Casp6, both Casp6 variants were unstable and inactive in transfected mammalian cells. In addition, Casp6-G66R acted as a dominant negative inhibitor of wild-type Casp6. The R65W and G66R substitutions caused perturbations in substrate recognition and active site organization as revealed by molecular dynamics simulations. Our results suggest that full Casp6 activity may not be essential for healthy humans and support the use of Casp6 inhibitors against Casp6-dependent neurodegeneration in age-dependent cognitive impairment and AD. Furthermore, this work illustrates that studying natural single amino acid polymorphisms of enzyme drug targets is a promising approach to uncover previously uncharacterized regulatory sites important for enzyme activity.


Results
Identification of rare missense variants of the human CASP6 gene with altered activity. Since Casp6 polymorphisms have not yet been associated with inherited diseases, we reasoned that natural polymorphisms of human Casp6 could inhibit its activity without revealing an obvious phenotype in humans. We screened the NCBI PubMed SNP database and identified several missense variants of CASP6. In this paper, we focus on two adjacent polymorphisms, Casp6-R65W (ExAC aggregated population allele frequency 0.0001159) and Casp6-G66R (no frequency data available) (Fig. 1a). Interestingly, despite being next to each other, R65 is found only in Casp6 compared to nine other human caspases, whereas G66 is entirely conserved in human caspases (Fig. 1b). Casp6-R65 is present in most mammals and some reptiles, birds and amphibians, but varies in fish, insects and molluscs (Fig. 1c). Despite these differences, no W65 was detected in the 19 species examined. By contrast, G66 is conserved in all species except for the Camelus ferus mammal and the Amazona aestiva bird. R65 and G66 are located in the extended helix B (amino acids [61][62][63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78][79][80], also known as the 60's helix of apo mature Casp6 (Fig. 1d). In the apo form, R65 and G66 are close to the active site but the two side-chain moieties are not interacting with other amino acid residues of Casp6 and the side chain of R65 is pointing away from the Casp6 active site catalytic dyad H121-C163. In a catalytically competent form of Casp6, represented by covalently bound Ac-VEID-CHO-Casp6, the 60's helix undergoes partial unfolding into the extension of L1 loop. G66 is retained in a position similar to that of the apo form, whereas R65 is solvent exposed and is observed in different conformations in two protomers of the dimeric structure (PDB ID: 3OD5): it either points away from the Casp6 active site (chain A) as in the apo form or is located close (2.7 Å) to the glutamate residue at the P3 position of Ac-VEID-CHO (chain B). Prokaryotically expressed recombinant Casp6-WT efficiently self-processes at TEVD 193, thus resulting in the generation of a higher amount of the large subunit with linker (LS-L) than without linker (LS) ( Fig. 1e and f). By contrast, recombinant Casp6-R65W generates more of the LS than LS-L, whereas Casp6-G66R remains unprocessed and migrates with the catalytically inactive proCasp6-C163A mutant (Fig. 1f). The identity of the subunits was confirmed by western blot analyses with neoepitope antiserum against the LS cleaved at DVVD 179, anti-p10 antibody that detects both full length (FL) Casp6 and its small subunit (SS) and anti-Casp6 antibody that detects LS, LS-L, or FL Casp6 (Fig. 1g). These results indicate that the auto-processing activity of these two rare human variants of Casp6 significantly differ from Casp6-WT Casp6-R65W and Casp6-G66R display reduced proteolytic activity on peptide and protein substrates through impaired active site binding. To assess the proteolytic activity of the variants, recombinant active site titrated ( Supplementary Fig. S1) enzymes were submitted to a fluorogenic assay with the Casp6-preferred Ac-VEID-AFC peptide substrate. Concentration-dependent cleavage of Ac-VEID-AFC by Casp6-R65W was achieved, but the Casp6-R65W activity was lower than that of Casp6-WT ( Fig. 2a and b). No measurable VEIDase activity was detected with Casp6-G66R at enzyme concentrations from 10 nM to 500 nM (Fig. 2a), indicating that Casp6-G66R is catalytically inactive on Ac-VEID-AFC. Nevertheless, purified recombinant Casp6-G66R generated by co-expression of Casp6 G66R-LS and WT-SS in E. coli ( Supplementary Fig. S2a), produced 2% VEIDase activity compared with Casp6-WT (Fig. 2c). These results show that R65W and G66R significantly suppress Casp6 activity on its preferred peptide substrate.
SCIeNtIfIC RepoRTS | (2018) 8:4428 | DOI:10.1038/s41598-018-22283-z While its k cat was unchanged, Casp6-R65W displayed approximately 2.5-fold lower k cat /K M and a 2.6-fold higher K M compared to Casp6-WT, indicating that it has a reduced binding affinity for Ac-VEID-AFC (Table 1). Thermal shift assays were performed in the absence of an active site ligand to assess the intrinsic stability of these variants. The results revealed that the intrinsic thermal stability of Casp6-R65W was essentially unchanged relative to Casp6-WT and the stability of Casp6-G66R was only decreased by 4 °C (Fig. 2d and e), suggesting that the changes in the activities of these variants was not due to global unfolding or misfolding of the proteins. However,  Data represent mean ± SD from three independent experiments. One-way ANOVA (p < 0.0001) followed by Bonferroni's multiple comparison tests compares Casp6-R65W with Casp6-WT, ***p < 0.001 (b) Relationship between initial reaction velocity (v) and substrate concentration of Casp6-WT and Casp6-R65W catalysed Ac-VEID-AFC cleavage. Data were fitted into Michaelis-Menten equation using nonlinear regression and represent the mean ± SD from three independent experiments. (c) Activity of co-expressed LS and SS from Casp6-G66R and Casp6-WT. Data represent mean ± SD from three independent experiments. in the presence of an excess amount of Ac-VEID-CHO substrate, a 7 °C stabilization of Casp6-WT was observed. The increase in stability upon incubation with Ac-VEID-CHO was lower for Casp6-R65W (4.6 °C) and absent for Casp6-G66R. This suggests that Casp6-R65W does bind substrate, albeit much less efficiently than Casp6-WT and that Casp6-G66R does not bind substrate to any appreciable extent. These observations are consistent with the observed catalytic parameters for Casp6-WT, -R65W and -G66R ( Table 1).
The observed difference in Casp6-R65W self-processing and VEIDase activity compared to Casp6-WT prompted us to investigate Casp6-R65W and Casp6-WT time-dependent activities on catalytically inactive proCasp6-C163A (Fig. 2f top panel and Supplementary Fig. S2b). As expected, Casp6-WT almost completely removed the pro-domain (Pro) of proCasp6-C163A within two hours, resulting in a rapid disappearance of the full-length (FL) proCasp6-C163A. By contrast, Casp6-R65W generated pro-domain lacking form (ΔCasp6) of Casp6-C163A more slowly than Casp6-WT. These results indicate that the R65W substitution decreases Casp6 self-cleavage efficiency at the TETD 23 site. As observed in self-processing, Casp6-R65W generated more LS cleaved at DEVD 179 (Fig. 2f, middle panel) and the corresponding SS with linker (L-SS) (Fig. 2f, bottom panel) from proCasp6-C163A, than Casp6-WT. No significant difference was observed in the levels of SS generated by Casp6-R65W compared to Casp6-WT. These results indicate that Casp6-R65W processes the DVVD 179 of pro-Casp6-C163A more efficiently than Casp6-WT. In order to test if R65W and G66R substitutions may alter Casp6 activity against natural protein substrates Lamin A/C, proteins extracted from Casp6 knock-out mice tissue were cleaved with increasing concentrations of Casp6-WT, Casp6-R65W and Casp6-G66R for two hours (Fig. 2g). Compared to Casp6-WT, Casp6-R65W generated 30-40% less cleaved Lamin A/C, whereas Casp6-G66R did not process Lamin A/C even at a concentration of 400 nM (Fig. 2h). Together, these results indicate that G66R completely abolishes and R65W significantly reduces Casp6 activity.
Furthermore, proCasp6-G66R was incubated with active Casp3 at 1:1, 2:1 and 3:1 ratios (Fig. 3c). After a 30 min incubation period, proCasp6-G66R was cleaved only at DVVD 179 generating a LS slightly larger than expected but detected with the neoepitope antiserum suggesting that it had retained the pro-domain (Pro-LS). By contrast, Casp3 efficiently cleaved Casp6-WT at DVVD 179 and TEVD 193 . The ability of Casp3 to cleave the pro-domain of Casp6-WT could not be determined in these experiments since the proCasp6-WT self-processed at the TETD 23 site during purification and naturally generated the Casp6-WT lacking its pro-domain (∆Casp6-WT). After a 16 h incubation period, Casp3 completely cleaved proCasp6-G66R at all three processing sites (Fig. 3d). However, Casp3 cleaved proCasp6-C163A mostly at TETD 23 , DVVD 179 and less efficiently at TEVD 193 since L-SS was clearly present (Fig. 3d). Together, these results indicate that the G66R substitution influences the conformation of Casp6 resulting in altered processing by Casp3 or Casp6.

Casp6-G66R acts as a dominant negative inhibitor of Casp6-WT.
To test if Casp6-G66R can act as a dominant negative inhibitor of Casp6-WT activity, VEIDase activity was measured in a mixture of ΔCasp6-WT and Casp3-processed Casp6-G66R or ΔCasp6-WT (Figs 3c and 4a). Unprocessed Casp6-G66R did not inhibit Casp6-WT (30 nM) VEIDase activity ( Supplementary Fig. S2d). As expected, active Casp3 showed a small amount of VEIDase activity whereas ∆Casp6-WT exhibited more VEIDase activity due to its ability to self-process. Casp3 cleavage of proCasp6-G66R did not increase VEIDase activity. However, Casp3 processing of ∆Casp6-WT increased VEIDase activity two-fold and this activity was significantly reduced by 30% with the addition of proCasp6-G66R, although not in a dose-dependent manner. Casp3 activity on its preferred Ac-DEVD-AFC substrate remained unaltered by the addition of proCasp6-G66R or ΔCasp6-WT (Fig. 4b).
Addition of fully processed Casp6-G66R, generated by co-expression and purification of LS and SS from E. coli, to fully processed Casp6-WT, resulted in full Casp6-WT VEIDase activity ( Supplementary Fig. S2e). These results support the previously proposed dominant negative inhibition mechanism for caspases where the LS of Casp6-WT and Casp6-G66R would associate to form an inactive enzyme.
(middle panel) and anti-p10 (bottom panel) western blots of time-dependent cleavage of Casp6-C163A by active Casp6-WT or Casp6-R65W. Quantitation of specific protein or peptide bands expressed as a % of time 0 for FL or Casp6-WT 16 h time point for LS and SS in bottom panel bar graphs. Data represent mean ± SD from three independent experiments. One-way ANOVA (p < 0.0001) followed by Bonferroni's multiple comparison tests, ***p < 0.001, **p < 0.01, *p < 0.05 compares Casp6-WT with Casp6-R65W. (g) Western blot of full-length Lamin A/C, cleaved Lamin A/C and anti-β-actin from Casp6 KO total protein lysates cleaved by 50-400 nM recombinant active site titrated Casp6-WT, Casp6-R65W and Casp6-G66R. (h) Quantification of data in (g) expressed as a % of cleaved Lamin A/C fragments (normalized to β-actin) relative to the 400 nM concentration. Data represent mean ± SD from three independent experiments. One-way ANOVA (p < 0.0001) followed by Bonferroni's multiple comparison tests, ***p < 0.001 and *p < 0.05 compares Casp6-R65W and Casp6-WT. Full-length images of gels and blots are provided in Supplementary Information. ity of Casp6 variants with Casp6-WT in mammalian cells, pro-domain-lacking ∆Casp6-R65W, -G66R, or -WT were expressed in the human embryonic kidney 293 T cell (HEK293T) line. As expected, the ∆Casp6 proteins migrated slightly ahead of the FL proCasp6-C163A (Fig. 5a). Compared to ∆Casp6-WT transfected cells, steady with anti-p10, anti-Casp6 and neoepitope anti-p20 antiserum using unprocessed proCasp6-C163A and fully processed Casp6-WT as molecular weight marker controls. Quantification of LS generated in (a and b) and expressed as % of LS with only Casp6-WT (bottom panels). One-way ANOVA followed by Bonferroni's (a) or Dunnett's multiple comparison (b) against Casp6-WT. Not significant (n.s.), **p < 0.01. Data represent mean ± SD from three independent experiments. Full-length images of blots are provided in Supplementary Information.

Molecular dynamics simulations of Casp6-G66R and Casp6-R65W indicate perturbation of active site and substrate-recognition components.
Ligand-free models of mature Casp6-WT 6 , Casp6-R65W and Casp6-G66R were subjected to all-atom explicit-solvent molecular dynamics (MD) simulations to reveal structural alterations that may account for observed differences in stabilities and kinetic parameters. Alignment-based comparison of Casp6-WT original crystal structure (Fig. 6a) and MD-optimized Casp6-WT (Fig. 6b) and Casp6-G66R ( Fig. 6c and e) models revealed that the backbone position of 61-66 fragment of the L1 loop deviates in Casp6-G66R from the position in Casp6-WT resulting in the perturbation of hydrogen-bonding between ε-N of catalytic His121 and carbonyl of the peptide bond between Pro62 and Glu63. Another notable topological difference is a significant deviation in the Arg220 hairpin position (Fig. 6e). Unlike G66 in Casp6-WT, the R66 in Casp6-G66R shifts away from the Arg220 hairpin since the bulky R66 points away towards the solvent to avoid steric hindrance with the Arg220 hairpin (Fig. 6e). In a Casp6-R65W model (Fig. 6d and f), a Trp65 indole is engaged in a cation-π interaction with Arg220, which is involved in the recognition of the P3 glutamate  of the substrate (Fig. 6a) and, therefore, Arg220 is made less available for this important ionic contact. These outcomes could translate into both the increases in the corresponding K M values and stability losses observed in the ligand-bound states of the Casp6 variants.

Discussion
This work shows that studying natural Casp6 variants is a promising approach to uncover novel areas important for enzyme's activity. Characterization of Casp6-R65W and Casp6-G66R reveals that R65W and G66R substitutions exhibit a destabilizing effect on the catalysis-enabling conformation of Casp6 and highlights the importance of previously uncharacterized R65-G66 region of L1 active site loop for Casp6 activity and substrate recognition.  Casp6 exists in a dynamic equilibrium between an inactive helix and substrate-binding induced active strand conformations 45 . During the transition from helix to strand forms, a region of 60's helix comprising R65-G66 unfolds into L1 loop extension which together with four additional loops flank the Casp6 active site 5 . The replacements of helix-breaking Gly66 by helix-indifferent Arg and helix-indifferent Arg65 by helix-favouring Trp 46 are both expected to stabilize the inactive conformation of Casp6. The strict conservation of G66 and the activity-abolishing effect of G66R substitution suggest that G66 may be crucial for providing L1 loop flexibility in the active sites of caspases. L1 loop conformation and mobility are known to control caspase activity. Mutations affecting Casp3 L1 loop dynamics reduce Casp3 activity 47 . Furthermore, calbindin-D28K anti-apoptotic protein inhibits Casp3 by locking the L1 loop in the position that occludes the Casp3 active site 48 . Similar restrained L1 loop position is observed in the inactive proCasp8 zymogen 49 . The G66R-induced pro-helical strain causes a slight repositioning of the 61-66 fragment of the L1 loop in Casp6-G66R, resulting in the loss of H-bond between the catalytic His121 and the carbonyl of the Pro62-Glu63 peptide bond. This contact, observed in crystal structures of both mature and proCasp6, has been implicated in serving a unique role of the third component in the catalytic triad-like arrangements of other caspases 36,50 . Importantly, this role is yet to be experimentally verified due to the participation of the backbone and this report provides the first instance of experimental evidence for supporting the hypothesis developed on the basis of static structure evaluation.
Our results suggest that R65W and G66R substitutions have a profound impact on Casp6 substrate recognition and strongly support the role of R65 in substrate binding. The R65W substitution significantly reduces Casp6 activity and affinity for VEID site, as observed with the Ac-VEID-AFC and Lamin A/C substrates. Human caspases prefer the glutamic acid at the P3 position of a peptide substrate 51 , since the strictly conserved R220 in the L3 loop binds the P3 side chain at the active site of Casp6 6 and most other caspases [29][30][31][32][33] . Interestingly, Casp6's L1 loop R65 can also mediate a water interaction with the P3 glutamate side chain 52 like the R258 in the L1 loop of Casp8 32 . However, this hypothesis remains tentative in the absence of consensus for Chain A or B of the VEID-bound Casp6 structure. Among the three Casp6 processing sites, TETD 23 , TEVD 193 and DVVD 179 , Casp6-WT intermolecularly cleaves the DVVD 179 site with the least efficiency because the P3 is occupied by valine and not the preferred glutamate. Yet, Casp6-R65W self-processes at DVVD 179 and cleaves Casp6-C163A-DVVD 179 more efficiently suggesting that R65W substitution improves recognition of DVVD 179 . The preference of Casp6-R65W for DVVD 179 over TETD 23 could be rationalized by the shift from favouring complementary charge-charge (R 65 -E 21 ) contact in the wild-type enzyme to hydrophobic-hydrophobic (W 65 -V 177 ) interaction in the mutant. In Casp6-R65W a relatively hydrophobic Trp replaces a charged Arg65 in a fully solvent-exposed position. A partial seclusion of the electron-rich indole could be accomplished via a cation-π contact with R220, responsible for the accommodation of both P1 and P3 substrate sites, which, when combined with the loss of R65, a secondary cationic contact for P3, offers a compelling account for the increased K M value with the peptide and native substrates for the Casp6-R65W. The replacement of fully buried and compact Gly66 in Casp6-G66R with the sterically demanding and charged Arg66 results in the reorganization of the R220-containing hairpin loop, thus negatively affecting substrate binding in the Casp6-G66R active site. This MD simulation-derived conclusion is strongly supported by the thermal shift assay findings. Furthermore, the observation that Casp6-G66R intersubunit linker TEVD 193 site is more accessible for intermolecular cleavage with active Casp6 or Casp3 compared to Casp6-C163A further suggests that G66R impairs linker binding at the active site of Casp6 6 .
Bioinformatics tools, like PolyPhen 53 and SIFT 54 , predicting effects of non-synonymous SNPs on protein function based on the analysis of multiple protein sequence alignments and protein three dimensional structures, suggest that R65W and G66R substitutions are deleterious to Casp6 function. Our experimental results support these predictions and strongly encourage the use of available computational tools for the initial estimation of mutation effect on protein function.
Casp6 is becoming an attractive therapeutic target for AD. Our study demonstrates that rare human CASP6 SNPs, encoding R65W and G66R substitutions in Casp6, suppress Casp6 activity and self-activation. Likewise, rare missense SNPs in human CASP1 gene associated with autoimmune conditions 55 cause a reduction or abrogate auto-processing and activity of Casp1 in vitro and in HEK293T 56 . In contrast to these Casp1 variants, rare Casp6 variants are not known to be associated with any disease. The fact that Casp6 variants with decreased or no activity are compatible with normal life suggests that Casp6 is not essential for healthy humans. Casp6 plays a role in axon pruning during development 17,18 and B cell activation and differentiation into plasmas cells 25 . Nevertheless, Casp6 knock-out mice develop normally 24 and cell death defects have not been reported for Casp6 mutants, supporting the idea that the role of Casp6 in apoptosis is either redundant or compensated in the Casp6 knock-out models 57 . In contrast to Casp6 physiological functions, increased active Casp6 in the human adult brains has been associated with age-dependent cognitive impairment and AD [11][12][13][14][15] . Casp6 causes axonal degeneration in NGF-deprived mouse sensory and wild type and mutant amyloid precursor protein-transfected human CNS neurons [16][17][18][19] . ProCasp6 is barely detectable in healthy foetal and adult brain 23 . Therefore, the absence of an association between these human CASP6 genetic variants and disease suggest that selective downregulation of Casp6 activity would unlikely cause serious side effects in humans.
Presently, there are 111 human CASP6 SNPs coding for missense Casp6 variants reported in the NCBI, 70% of which have been validated but have not yet been associated with severe clinical outcomes. Taking into consideration that active Casp6 is associated with AD and cognitive impairment, it is tempting to speculate that CASP6 SNPs described in this study could be associated with decreased risk of age-related Casp6-mediated neurodegeneration. In addition to finding SNPs that enhance the risk of AD, next-generation sequencing has allowed identification of rare SNPs linked to reduced risk of AD. The rare SNP coding for the Ala673Thr variant of amyloid precursor protein protects against AD and cognitive decline in the elderly 58 . Future systematic functional characterization of existing Casp6 polymorphic variants in combination with phenotypic information from healthy SCIeNtIfIC RepoRTS | (2018) 8:4428 | DOI:10.1038/s41598-018-22283-z and disease-affected human subjects would help to reveal any causative or protective effects of CASP6 SNPs on age-dependent cognitive impairment and AD.
Thermal shift assays. ΔN D179 CT 5 version of Casp6-WT, co-expressed Casp6-R65W large and small subunits and prodomain lacking ΔN Casp6-G66R described above were purified 45 . Casp6 (10 μM) was incubated in 20 mM Tris, pH 8.5 with 5 mM DTT with or without 50 μM Ac-VEID-CHO active-site inhibitor (Enzo Life Sciences) and 5 × SYPRO ® Orange dye (Thermo Fisher Scientific) in a 60 μL reaction. Fluorescence (ex/ em 490/575 nm) was measured in a 96-well plate using a CFX Connect Real-Time PCR instrument (BioRad). RFU recorded from 25 to 95 °C with 0.5 °C/3 second intervals were normalized to the highest observed intensity which was set to 1. The normalized fluorescence was fit to a Boltzmann sigmoidal curve using Prism (GraphPad) software. Melting temperature (T m ) was found to be the temperature at the midpoint of the denaturation curve.