Conservation of the Extended Substrate Specificity Profiles Among Homologous Granzymes Across Species

Granzymes are structurally related serine proteases involved in cell death and immunity. To date four out of five human granzymes have assigned orthologs in mice; however for granzyme H, no murine ortholog has been suggested and its role in cytotoxicity remains controversial. Here, we demonstrate that, as is the case for granzyme C, human granzyme H is an inefficient cytotoxin that together with their similar pattern of GrB divergence and functional similarity strongly hint to their orthologous relationship. Besides analyzing the substrate specificity profile of granzyme H by substrate phage display, substrate cleavage susceptibility of human granzyme H and mouse granzyme C was assessed on a proteome-wide level. The extended specificity profiles of granzymes C and H (i.e. beyond cleavage positions P4-P4′) match those previously observed for granzyme B. We demonstrate conservation of these extended specificity profiles among various granzymes as granzyme B cleavage susceptibility of an otherwise granzyme H/C specific cleavage site can simply be conferred by altering the P1-residue to aspartate, the preferred P1-residue of granzyme B. Our results thus indicate a conserved, but hitherto underappreciated specificity-determining role of extended protease-substrate contacts in steering cleavage susceptibility.Several molecular mechanisms are in place to combat transformed malignant cells and virally infected cells. Granzymes (Gr)¹, a family of structurally related serine proteases found in the granules of many immune cells, play crucial roles in such cellular defense mechanisms. The granzyme family consists of five human proteases (granzymes A, B, H, K, and M) and 10 murine members (granzymes A to G, K, M, and N). To date for four human granzymes (A, B, K, and M) clear murine orthologs have been assigned, while the most probable murine ortholog of human granzyme H (GrH) is granzyme C (GrC) based on their 70% sequence similarity, 61% sequence identity, and identical chromosomal location relative to granzyme B (GrB). Furthermore, both granzymes are expressed by NK and CD4+ T-cells (1, 2): GrH is constitutively expressed at high levels in NK cells and less in CD4+ T cells, whereas GrC expression can be observed after stimulation. Thus, their overlapping expression profiles further support possible functional similarities between mouse GrC and human GrH.The physiological role of granzymes was presumed to be the induction of death in target cells. Granzyme B is a highly efficient cytotoxin (3) and other granzymes such as GrA, GrC, GrF, and GrK can cause cell death at high concentrations (4–8). Recently, granzymes A, K, and M were shown to steer inflammatory processes when used at physiological levels (9–11). Two previous studies identified GrH as an alternative cytotoxic effector protease. Although both studies showed typical hallmarks of apoptosis, including mitochondrial depolarization, reactive oxygen species (ROS) production, DNA degradation as well as chromatin condensation, Fellows et al. (12) found that, in contrast to GrB mediated cell death, GrH induced cell death did not result in caspase activation, cytochrome c release, or cleavage of Bid and/or ICAD. In sharp contrast however, Hou et al. (13) demonstrated that GrH induced apoptosis depended on caspase activation and that GrH cleaved ICAD and Bid, the latter ultimately resulting in mitochondrial cytochrome c release. Such discrepancies have been documented in other granzyme studies and are usually linked to the use of different granzyme delivery systems, sources of recombinantly produced granzymes, differences in the granzyme concentration and species-specific differences in substrate specificities (14).GrC induces cell death reminiscent of GrH induced cell death as observed by Fellows et al. (12), as both exert their cytotoxic functions independent of caspase activation, Bid or ICAD cleavage, or by mitochondrial release of cytochrome c (5). GrC induced apoptosis was characterized by the rapid externalization of phosphatidylserine, nuclear condensation and collapse, and single-stranded DNA nicking. Supporting evidence implying a role for GrC in lymphocyte induced cytotoxicity was inferred from the fact that GrB cluster-deficient mice (mice that do not express GrB and show a five- to sixfold reduced expression of GrC and GrF respectively) display a more pronounced defect in the clearance of allogeneic tumor cells when compared with GrB-only knockout mice (4). This suggests that GrC and/or GrF may be important for correct functioning of cytotoxic lymphocytes. Besides, in GrB-only knockout mice, a likely compensatory mechanism occurs, given that during cytotoxic lymphocyte activation, peak expression of GrC occurs earlier, giving rise to overall higher GrC levels as compared with wild-type mice (1, 4, 15). Of note, despite its implication in cytotoxicity, GrC was shown to be an inefficient cytotoxin, with a 2900-fold greater EC₅₀ value as compared with hGrB when delivered into P815 cells with recombinant mouse perforin (16).Positional Scanning Synthetic Combinatorial Libraries (PS-SCL) revealed the chymotrypsin-like activity of GrH, which it shares with granzyme M (17–19), with an optimal P4-P1 peptide substrate sequence Pro-Thr-Ser-Tyr. Less stringent specificities were observed at positions P4, P3, and P2 (19) where GrH seems to tolerate multiple amino acids with different chemical characteristics (especially neutral and aliphatic amino acids). Although GrH and GrM (optimal peptide identified as Lys-Val-Pro-Leu) share a P1 chymotryptic activity, GrH prefers bulkier, aromatic amino acids (Tyr and Phe) at P1 whereas GrM prefers Leu. Both further recognize Leu and Met at P1, implying that some substrates could be cleaved by both granzymes. GrM also shows broader specificities at P3, but at P4 and P2 it prefers basic residues and Pro respectively. Besides, GrC chymase activity could be inferred from N-terminal COFRADIC and substrate phage display screens, which defined the P4-P3′ substrate specificity of GrC as [Ile/Val]-X-[Phe-Tyr]-[Phe-Leu-Tyr-Met]↓X-[Gly-Ser]-[Asp-Glu] (16).Recently, the crystal structure of the D₁₀₂>N GrH variant in complex with a decapeptide substrate (PTSYAGDDSG) or inhibitor (Ac-PTSY-chloromethylketone) was resolved (20). The electron density maps clearly showed the full length protease adopting a canonical structure of 2 α-helices and 13 β-strands assembled into two juxtaposed β-barrel domains bridged by the catalytic triad (composed of His₅₇, Asp₁₀₂ and Ser₁₉₅). The S1 specificity pocket is built up from residues of 2 loops; loop 189 (from residue 183–196) and loop 220 (from residue 215–226) with the determinants being Thr₁₈₉, Gly₂₁₆, and Gly₂₂₆. From these, Gly₂₂₆ was assigned as the most important determinant for the preference of bulky aromatic amino acids (such as Tyr and Phe) at P1. Where GrH contains a Gly at position 226, GrB, GrC, and GrM harbor an Arg, Gln, and Pro residue respectively (supplemental Fig. S1). Mutation of Gly₂₂₆ in GrH to Arg₂₂₆ compromised binding of bulky aromatic residues and enabled interaction with negatively charged amino acids. Hydrogen bond formation between a P1 Tyr and Asn₂₁₇ of GrH could furthermore strengthen the observed preference of Tyr over Phe at P1 (19). The presence of Pro at position 226 in GrM instead of Gly narrows the S1 pocket and might be indicative for the preference of Leu instead of Phe and Tyr at P1. In addition, the structure revealed that the S4′ pocket formed by the backbones of the Arg₃₉-Lys₄₀-Arg₄₁ motif resulted in a preference for acidic residues at P4′, in addition to influencing the P3′ preference for acidic residues via salt bridge formation with this Lys₄₀. Although the Arg₃₉-Lys₄₀-Arg₄₁ motif is a unique GrH feature, GrB possesses a partially degenerate Leu-Lys-Arg motif in which Lys₄₀ also enables interaction with acidic residues in P3′. Next to P3′, proteome-wide screening for GrB substrates led to the identification of a clear preference for acidic residues at P4′ caused by salt bridge formation with Arg₄₁ (21), a characteristic that was previously assigned to Lys₄₀ of GrB (22). To validate these structural observations, similar to the P4′ Asp mutation in the GrB substrate PI-9 (22), both Asp residues found at P3′ and P4′ in the nuclear phosphoprotein La, previously identified as a macromolecular GrH substrate (23), were mutated to Ala. These mutations completely abolished GrH mediated proteolysis, a first indication that the P4-P1 specificity profile is not a sole determinant for substrate recognition by GrH. Next to the La phosphoprotein, cleavage of the viral adenovirus DNA binding protein (DBP) and the 100K assembly protein (L4–100K), the latter resulting in the relieve of GrB inhibition by L4–100K, has been observed, implicating GrH in host antiviral defense and indicative for a functional synergism between GrH and GrB (24).Elucidation of the crystal structure, with an electron density map showing residues Ile₁₆-His₂₄₄ (i.e. 94% of full length GrC and 99.6% of active GrC), furthermore showed that wild-type GrC can be restrained in its proteolytic function despite the presence of the catalytic triad residues His₅₇-Asp₁₀₂-Ser₁₉₅ (16). Comparing the crystal structures of GrC and GrA showed an unusual conformation of the active site, which could explain the inactivity of GrC. Apparently, the 190-strand, preceding Ser₁₉₅ of the catalytic triad, has undergone a register shift on the structural level leading to Phe₁₉₁ filling the S1 pocket. This pocket is furthermore covered by Glu₁₉₂, which forms a salt bridge with Arg₉₉ and a hydrogen bond with the backbone amide of Ser₁₉₅. Because of this unusual conformation, the Glu₁₉₂-Glu₁₉₃ peptide bond points away from the substrate, as such leading to an improperly formed oxyanion hole, which normally stabilizes the negatively charged substrate oxygen atom during catalysis. Mutation of Glu₁₉₂-Glu₁₉₃ to the corresponding amino acids in its closest related homolog GrB (Arg₁₉₂-Gly₁₉₃) disrupted the Glu₁₉₂-Ser₁₉₅ hydrogen bond and caused a shift of the 190-strand, thereby clearing the S1 pocket and giving rise to an active GrC mutant. These results indicate allosteric control of wild-type GrC in which binding of a substrate or cofactor might stabilize the 190-strand, which becomes extremely mobile due to breaking the Glu₁₉₂-Ser₁₉₅ hydrogen bond and turning the region surrounding the active site more rigid.To elucidate a possible functional homology between GrH and GrC, we performed differential degradome analyses using N-terminal COFRADIC in the species-matching proteome backgrounds. For these analyses, we made use of the active E₁₉₂E₁₉₃>RG GrC mutant as described in (16) (further referred to as mut GrC). These analyses, further complemented with phage display data on granzyme H, clearly show that both granzymes display a highly similar substrate specificity profile, analogous to other orthologous granzymes. Besides, and in contrast to GrA, a general conservation of the extended substrate specificity profiles among the homologous granzymes B, C, H, and M could be observed across species, highlighting the importance of the extended substrate specificity in steering substrate cleavage susceptibility.