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 CD4ϩ T-cells (1,2): GrH is constitutively expressed at high levels in NK cells and less in CD4ϩ 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 EC 50 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)(18)(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 de- Recently, the crystal structure of the D 102 ϾN 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 ␣-helices and 13 ␤-strands assembled into two juxtaposed ␤-barrel domains bridged by the catalytic triad (composed of His 57 , Asp 102 and Ser 195 ). 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 Thr 189 , Gly 216 , and Gly 226 . From these, Gly 226 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 Gly 226 in GrH to Arg 226 compromised binding of bulky aromatic residues and enabled interaction with negatively charged amino acids. Hydrogen bond formation between a P1 Tyr and Asn 217 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 S4Ј pocket formed by the backbones of the Arg 39 -Lys 40 -Arg 41 motif resulted in a preference for acidic residues at P4Ј, in addition to influencing the P3Ј preference for acidic residues via salt bridge formation with this Lys 40 . Although the Arg 39 -Lys 40 -Arg 41 motif is a unique GrH feature, GrB possesses a partially degenerate Leu-Lys-Arg motif in which Lys 40 also enables interaction with acidic residues in P3Ј. Next to P3Ј, proteome-wide screening for GrB substrates led to the identification of a clear preference for acidic residues at P4Ј caused by salt bridge formation with Arg 41 (21), a characteristic that was previously assigned to Lys 40 of GrB (22). To validate these structural observations, similar to the P4Ј Asp mutation in the GrB substrate PI-9 (22), both Asp residues found at P3Ј and P4Ј 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 Ile 16 -His 244 (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 His 57 -Asp 102 -Ser 195 (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 Ser 195 of the catalytic triad, has undergone a register shift on the structural level leading to Phe 191 filling the S1 pocket. This pocket is furthermore covered by Glu 192 , which forms a salt bridge with Arg 99 and a hydrogen bond with the backbone amide of Ser 195 . Because of this unusual conformation, the Glu 192 -Glu 193 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 Glu 192 -Glu 193 to the corresponding amino acids in its closest related homolog GrB (Arg 192 -Gly 193 ) disrupted the Glu 192 -Ser 195 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 Glu 192 -Ser 195 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 E 192 E 193 ϾRG 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.

EXPERIMENTAL PROCEDURES
Production of Recombinant Granzymes-Recombinant human granzyme B was expressed and purified from Pichia pastoris and its activity was determined as described previously (25,26). Recombinant, enterokinase activated wild-type GrC and the mutated form of GrC (E192R/E193G) were prepared as described previously (16). Recombinant human granzyme H was expressed in Pichia pastoris and purified to homogeneity. Correct folding and activity were assessed by serpin-mediated GrH complex formation.
Granzyme H Substrate Phage Display-Phage display was performed as described previously (16) using the random 9-mer library for 4 rounds of selection.
Preparation of Cell Lysates for N-Terminal COFRADIC Analysis-Stable isotope labeling by amino acids in cell culture (SILAC)-labeled K-562 and YAC-1 cells were washed in D-PBS and resuspended at 7 ϫ 10 6 cells per ml in 50 mM Tris-HCl (pH 8.0) and 100 mM NaCl. Cells were subjected to three rounds of freeze-thaw lysis and supernatants, cleared by centrifugation for 10 min at 16,000 ϫ g, were treated with 200 nM of recombinant human granzyme H (GrH), recombinant mutant mouse granzyme C (mut GrC) or left untreated (control). Solid guanidinium hydrochloride was added to a final concentration of 4 M to inactivate granzymes and denature all proteins. Before mixing the samples in equal amounts, proteins were reduced and alkylated using TCEP.HCl (1 mM f.c.) and IAA (2 mM f.c.) respectively, for 1 h at 30°C. Subsequent steps of the N-terminal COFRADIC protocol were performed as described previously (28). The proteome was digested overnight at 37°C with sequencing-grade, modified trypsin (Promega, Madison, WI) (enzyme/substrate of 1/100 w/w).

LC-MS/MS Analysis and Data Processing-Electrospray ionization liquid chromatography-tandem MS (ESI LC-MS/MS) analysis was
performed on a Q-TOF (Waters Corporation) mass spectrometer as described before (29). ESI-Q-TOF MS/MS peptide fragmentation spectra were converted to pkl files using the Masslynx® software (version 4.1, Waters Corporation). N-terminal peptides were identified using a locally installed version of the MASCOT database search engine version 2.2.1 and 2.2.04 (www.matrixscience.com) and the Swiss-Prot database (version 2012_09 of UniProtKB/Swiss-Prot protein database, containing in total 538,010 sequence entries from which 20,234 human and 16,563 mouse entries) was searched with restriction to human or mouse proteins. Truncated peptide databases made by DBToolkit (30), were searched in parallel to pick up protein processing events more efficiently (e.g. (31,32)). The following search parameters were used. Peptide mass tolerance was set at 0.2 Da and peptide fragment mass tolerance at 0.1 Da with the ESI-QUAD-TOF as selected instrument for peptide fragmentation rules for the Q-TOF Premier data. The protease setting semi-ArgC/P (i.e. no restriction toward arginine-proline cleavage) was used allowing one missed cleavage. Variable modifications were pyroglutamate formation of N-terminal glutamine, pyrocarbamidomethyl formation of N-terminal alkylated cysteine, deamidation of asparagine, acetylation or tri-deuteroacetylation of the alpha-N terminus. Fixed modifications were methionine oxidation (sulfoxide), carbamidomethyl for cysteine, trideuteroacetylation of lysine and for identifying heavy labeled peptides, [ 13 C 6 ] or [ 13 C 6 15 N 4 ]-arginine were additionally set as fixed modifications. Only MS/MS-spectra and corresponding identifications that exceeded the corresponding MASCOT's threshold score of identity (at 95% confidence level) and that were ranked one, were withheld. In addition, spectra that received a low MASCOT ion score (5 or less points above threshold for identity) were further interrogated and only spectra that contained b-and y-fragment ions covering a stretch of three consecutive amino acids were considered identified. The estimated false discovery rate by searching decoy databases was typically found to lie between 2 and 4% on the spectrum level (28). Whenever a peptide matched to multiple members of a protein family (redundancy), all protein entries were reported (column "Description," supplemental Tables S1 and S2) and ranked alphabetically. Corresponding orthologous mouse or human cleavage sites, as determined in (21), are reported in supplemental Tables S1 and S2. The peptide intensity ratios were manually calculated from the MS data in the Masslynx® software version 4.1 environment. To correct for varia-tions in initial protein concentrations, the 12 C 6 / 13 C 6 and 12 C 6 / 13 C 6 15 N 4 ratios of the 2670 human and 1866 mouse database-annotated protein N-terminal peptides (peptides starting at position 1 or 2) were first subjected to robust statistics on the base-2 logarithms of their ratios (33). All the mass spectrometry is deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) with the data set identifier PXD000152 and DOI 10.6019/PXD000152. Under the ProteomeXchange accession code; PXD000152 and PRIDE accession codes; PRIDE accessions: 28635-28636.
Cloning and Expression of (Truncated) FEN-1-The pOTB7-FEN-1 (RZPD Imagenes, cat # IRAUp969C102D, Germany) vector served as a template to generate PCR products of full-length and truncated FEN-1 suitable for use as substrates in a Gateway® BP recombination reaction with a donor vector (pDONR221, Invitrogen, cat # 12536 -017) thereby creating an entry clone. The following PCR primers were used to generate attB flanked PCR products (forward primer FEN-1: 5Ј-ggggacaagtttgtacaaaaaagcaggcttcaccatgggaattcaaggcctggcc-3Ј and truncated FEN-1: 5Ј-ggggacaagtttgtacaaaaaagcaggcttcaccatgtgtggtgaaaagcagttctctgagg-3Ј; reverse primer 5Ј-ggggaccactttgtacaagaaagctgggtcttttccccttttaaacttccctgc-3Ј). The reverse primer was designed to fuse the desired PCR products in frame with a C-terminal GFP tag.
K-562 cells were transiently transfected for 24 h with 0.4 g/ml (final concentration) of the eukaryotic expression vectors using Lipo-fectamine™ LTX and PLUS™ Reagents (Invitrogen) according to the manufacturer's instructions. Following transfection, cells were resuspended in 50 l 50 mM Tris-HCl (pH 8.0) and 100 mM NaCl and subjected to three rounds of freeze-thaw lysis and centrifuged for 10 min at 16,000 ϫ g. Whenever indicated, cleared cell lysates were incubated for 15 min with 5 mM iodoacetamide at 37°C in the dark before incubation with 200 nM mut GrC for 1 h at 37°C. Following mut GrC incubation, NuPAGE ® LDS Sample Buffer (Invitrogen) was added and samples were heated for 10 min at 70°C. Proteins were then separated in MOPS buffer using 4 -12% polyacrylamide Criterion XT-gels (1.0 mm thick gels) in MOPS buffer (Bio-Rad) at 150 V. Subsequently, proteins were transferred onto a PVDF membrane. Membranes were blocked for 30 min in a 50:50 Tris-buffered saline and 0.1% Tween-20 (TBS-T) Odyssey Blocking solution (LI-COR, cat n°927-40003) and probed for 1 h with primary anti-GFP antibody (Invitrogen, cat # A11122) in TBS-T/Odyssey blocking buffer. Following three 10 min washes in TBS-T, membranes were incubated with secondary anti-rabbit antibody (IRDye 800 CW goat anti-Rabbit antibody IgG, LI-COR, cat n°926 -32211) for 1 h in TBS-T/Odyssey blocking buffer followed by three washes in TBS-T or TBS (last wash step), bands were visualized using an Odyssey infrared imaging system (LI-COR).
Granzyme Cleavage of In Vitro Transcribed and Translated FEN-1-pOTB7-FEN-1 (RZPD Imagenes, cat # IRAUp969C102D, Germany) was used as template for in vitro coupled transcription/translation in a rabbit reticulocyte lysate system according to the manufacturer's instructions (IVTT; Promega) to generate [ 35 S]methionine-labeled FEN-1. The SP6 TNT Quick Master Mix (40 l) -containing TNT rabbit reticulocyte lysate, TNT reaction buffer, TNT SP6 RNA polymerase, an amino acid mixture without methionine and RNasin Ribonuclease Inhibitor -was mixed with 2 l [ 35 S]-methionine (20 Ci/ml), 1 l of plasmid DNA template (1 g/l) and 7 l nuclease-free water. Following 90 min incubation at 30°C, translates were incubated with iodoacetamide (f.c. 5 mM) for 20 min at 37°C as to avoid downstream caspase activity upon granzyme treatment (34). The translates (5 l) were subsequently incubated for 1.5 h at 37°C in granzyme assay buffer with the indicated concentrations of recombi-nant granzymes in a total volume of 30 l. Cleavage reactions were stopped by the addition of NuPAGE ® LDS Sample Buffer (Invitrogen) and heating the samples for 10 min at 70°C. Proteins were separated on 4 -12% NuPAGE ® Bis-Tris gradient gels and subsequently transferred onto a PVDF membrane as indicated above, these membranes were then air-dried and exposed to a film suitable for radiographic detection (ECL Hyperfilms, Amersham Biosciences, Buckinghamshire, UK). Radiolabeled proteins and their cleavage fragments were visualized by autoradiography.
Site Physical Property Motif Analysis-The positional amino acid frequency vectors of the ungapped substrate motif alignment calculated over sliding windows of 1 to 12 positions was correlated with an extensive collection of physical property scales and selected those that were significant with R 2 Ͼ 0.6. Properties that themselves closely correlated (R 2 Ͼ 0.9) with average database composition scales were excluded. If multiple, related properties (cross-correlation of R 2 Ͼ 0.4) were found for the same region, only those with the highest correlation were kept to eliminate redundant motif descriptions. To favor stretches of continuous significant correlations over high correlations on single positions, the values for the selection were normalized by the square root of the number of positions involved (21,35).
Structural Classification of Cleavage Sites-Structural classification of cleavage sites was conducted as follows. First, we compared the cleavage locations with the annotation of available structures for sequence entries in UniProtKB (13). For each overlap, we manually confirmed if the site is part of the resolved 3D structure or unresolved loops or termini without atomic coordinates of the PDB entries using YASARA (14). In the former case, the site would be classified as "known structure" and in the latter case as "known disorder." Furthermore, we extended the search for structure locations by homology searches of the respective region with BLASTP (15) against PDB (16), which ensures only highly similar hits and hence reliable structure prediction. As before, all structural locations of cleavage sites predicted by similarity were manually confirmed, giving rise to the classification of "predicted structure." Finally, cleavage sites in regions outside of known or predicted structure are divided into those with "unknown structure status" and those that are "predicted disorder" as determined by IUPred (17) using a threshold setting of 0.54, which was recently benchmarked (18) to give reliable disorder assignments with less or equal to 5% false positive predictions.

Granzyme H Displays a Low Cytotoxicity Potential Compared With
Granzyme B-Human granzyme H (GrH) was expressed in the yeast Pichia pastoris and purified to homogeneity. To confirm that GrH was correctly folded, an active site titration was performed by incubating recombinant granzyme H (GrH) with the serpin inhibitor alpha-1-antichymotrypsin (ACT). Serpin-mediated protease inhibition requires substrate recognition of an exposed reactive center loop (RCL), followed by cleavage at the reactive bond leading to a conformational change that distorts the protease active site, thereby trapping the protease in a 1:1 covalent, inhibitory complex with the serpin. ACT was selected as it contains a double Leu motif at the reactive site (supplemental Fig. S2), which PS-SCL predicts would be preferred as a substrate by GrH (19). Making use of excess ACT, gel-based analyses revealed formation of the SDS-stable complex, indicative of serpin-protease interaction, demonstrating that essentially all the GrH moves into complex (supplemental Fig. S2).
Having demonstrated the full catalytic potential of GrH, GrH cytotoxicity was studied by monitoring cell survival on streptolysin-O (SLO) mediated granzyme delivery in cell culture (Fig. 1). Quantification of granzyme cytotoxicity was performed using a dose-response analysis, in which the amount of granzyme required to kill 50% of the target cells (EC 50 ) was measured (25). The over 170-fold greater EC 50 value for human GrH (1.658 M Ϯ 0.457) as compared with human GrB (9.9 nM Ϯ 4.7) (Fig. 1), suggests that GrH is an inefficient cytotoxin similar to (unlocked) GrC (EC 50 values of 39.6 M and 11.2 M for locked and unlocked GrC respectively) (16), an observation supporting its role in viral clearance rather than cell death (24).
Substrate Phage Display-By employing a substrate phage display system, we examined the extended substrate specificity of GrH. Here, random 9-mer peptides linked to a His-tag were fused to the C terminus of the T7 phage coat protein (36). Phages were captured on a nickel-Sepharose support and eluted by GrH incubation. After several rounds of capture and release, sequencing of 96 phages yielded 77 unique sequences and a consensus recognition sequence Ile-[Ile/ Val]-[Phe/Tyr]-[Tyr/Phe]2Ser-[Gly/Ala]-X-[Glu/Asp] (Fig. 2). Incubation with the peptide substrate aminobenzoyl (Abz)-Arg-Ile-Met-Phe-Phe-Ser-Gly-Ala-Thr-Lys(dnp(dinitrophenyl))-Asp indicated cleavage after the double Phe (data not shown). Despite the identical P1 preference observed, a less restricted substrate specificity profile was identified by combinatorial peptide analysis using a completely diversified PS-SCL of ACC (7-amino-4-carbamoylmethylcoumarin) substrates yielding the consensus sequence Pro-Thr-Ser-Tyr (19). Especially, the broad preferences at P2 in contrast with the more restricted preference for Tyr and Phe observed in this study, potentially signifies the importance of a more extended substrate GrH specificity profile beyond P4-P1.
Differential N-Terminomics Analysis-N-terminomics in combination with SILAC (37) was performed to unravel the (differential) macromolecular substrate repertoires of GrH and mut GrC. Here the proteolytic activities of GrH and mut GrC were investigated simultaneously in either a human (K-562) or murine (YAC-1) proteome background. Overall, these analyses resulted in the identification of 6196 MS/MS spectra from 2310 peptides derived from 1029 human and 1244 mouse proteins (isoforms not taken into account). Sixty-nine cleavage sites in 67 human proteins were identified, of which 46 (67%) were cleaved by both granzymes, whereas 14 (20%) and 9 (13%) sites were unique to GrH or mut GrC respectively (supplemental Table S1). Of note, 12 cleavage sites (17%)including 9 P1 Leu specific cleavages were previously also generated on GrM cleavage (38). Besides, 35 of the 67 GrH and/or GrC substrates (52%) reported here were previously identified to be cleaved by GrM, indicative for a shared substrate repertoire among these homologous granzymes.
In the mouse proteome, 83 cleavage sites in 75 proteins were identified. GrH generated 23 unique sites (28%), whereas 31 sites (37%) were cleaved by mut GrC, leaving 29 sites (35%) cleaved by both proteases (supplemental Table  S2). Overall, 18 orthologous human and mouse substratepairs were identified, of which 17 were found to be cleaved at the same position(s) in both species.
Proteins playing a role in transcriptional regulation (e.g. PHD finger protein 6), RNA binding/processing (e.g. insulinlike growth factor 2 mRNA binding protein 1), microtubule dynamics (e.g. centrosomal protein of 170 kDa) and ribosome biogenesis (e.g. ribosome biogenesis regulatory protein homolog) were identified. Although cleavage of the phosphopro- tein La (previously identified as a physiological GrH substrate) cannot be identified via this approach (i.e. because of the size and the presence of two SCX nontolerated histidines at the GrH-introduced neo-N terminus of La (KTKF 364 2ASDDEH-DEHDENGATGPVKR)), the La-related protein-1 was identified in the human proteome as a sole mut GrC substrate, whereas La-related protein-7 was found in the YAC-1 background as a single GrH substrate.
Proteome-wide Substrate Specificity Profiles of GrH and Mutant GrC-The iceLogos (39) created for both proteases revealed an extended substrate specificity profile in line with previous data deduced from PS-SCL, phage display and Nterminal COFRADIC (16,19,20), with 86% of all human and mouse cleavage events after hydrophobic and/or aromatic amino acids (Fig. 3). For both granzymes we observed a pronounced preference for acidic residues from P3Ј up to P8Ј, an extended specificity profile shared with granzyme B (16,21) and which was recently shown for GrH, at least for P3Ј and P4Ј (20). Preliminary data obtained by proteome-wide screening of GrC and GrM substrate specificity also hinted to such an extended specificity profile (16,38). As is the case for many proteases, small amino acids like Ala and Ser were overrepresented at P1Ј and also at P2Ј, an observation in line with phage display data (16). The preference for small P2Ј residues was also apparent when comparing human and mouse GrB cleavage susceptibility; here efficiency of cleavage by mGrB increased significantly when Gly was present at P2Ј (7), whereas for hGrB, other amino acids at P2Ј were tolerated.
The major difference in cleavage efficiency between GrH and mut GrC was observed when a Phe was present at P1 (Fig. 4). In general mut GrC cleaved P1 Phe-containing substrates 4 to 5-fold more efficiently than GrH. Furthermore, 48% of all unique mut GrC cleavages held a Phe at P1 whereas this only occurred in 5% of the unique GrH cleavage sites identified.
Physical Property Analyses-To complement substrate sequence conservation based on amino acid identity we next analyzed aligned GrH and mut GrC cleavage motifs according to physical property constraints as described previously (21,40), thereby emphasizing conservation by merging amino acid type preferences sharing similar physical properties. Regions with significant physical property correlations as well as deviations from the database averages of selected properties are listed in supplemental Tables S3 and S4 and summarized in Fig. 5. As also apparent from the IceLogo visualizations (Fig. 3), both enzymes highly favor small and/or polar residues at positions P1Ј and P2Ј, whereas next to P1; P2 clearly favors the occupancy of bulky hydrophobic residues. Overall, P1 surrounding regions correlate well with hydrophilic and flexible properties, showing significant correlations encompassing stretches over 10 amino acids (Fig. 5). Classification of cleavage sites based on their structural location indicates that the

FIG. 2. Extended substrate specificity of granzyme H by substrate phage display.
A, A random 9-mer phage display library was subjected to four rounds of selection using 200 nM GrH. Variable regions of 96 random phage from two independent experiments were sequenced yielding 77 unique sequences. These were manually aligned and the occurrence of each amino acid at each subsite determined (43). Residues occurring with a frequency of at least four standard deviations (⌬) from random are shown. B, Consensus cleavage sequence for GrH derived from phage display. majority of the cleavage sites identified is located in disordered or unstructured regions, which is compatible with extended flexible conformations that would be accessible to the protease (supplemental Tables S1 and S2). More specifically, only 9 (or 11%) of the human and 13 (or 15%) of the mouse cleavage sites are located in known or predicted 3D structures. Furthermore, analyzing compliance of individual sites with the average characteristic physical properties described in Fig. 5, shows that cleavage events found in structured elements have a much lower score as compared with the majority of cleavage sites found in disordered or unstructured regions, which implies that they do not match the flexible, unfolded extended physical property motif in agreement with their observed structural constraints (supplemental Fig. S5).
Structural Analysis of the Substrate Specificity Profiles of GrH and Mutant GrC-To further elucidate the structural protease features underlying the observed substrate specificity profiles of GrH and mut GrC, next to the available decapep-

P1 Protease and Substrate Specificity and Beyond
followed by short simulated annealing molecular dynamics simulations of residues within 5 Ångströ m from the peptidic ligand for energy minimization using the AMBER03 force field in YASARA (41), resulted in a model for peptide ligand-bound unlocked version of mut GrC. As a consequence of energy minimization in the presence of ligand, the model shows opening of the initial loop causing the lock, thereby returning the specificity region back into register with the GrH conformation (supplemental Fig. S3). In both cases, the bound substrate confirms the observed size restrictions inferred by the binding pocket constraints at P1Ј and P2Ј. Besides, the observed preferences of acidic amino acids at P3Ј and P4Ј and the presence of larger residues at P2 are well supported by respectively the surrounding positive charges (blue residues in Fig. 6) and the fact that P2 is positioned on top of the open part of the binding pocket, not restricted in space.
Further, the crystal and modeled structure predicts a number of differences in the extended substrate binding cleft, which may explain the increased tolerance of positively charged lysines at P4-P6 by mut GrC as compared with GrH, viewing the exclusive availability of a negative surface charge in this region in mut GrC (red patch in lower right corner of mut GrC; Fig. 6) and the increased preference for acidic residues at P4Ј and P3 because of the complementary charge differences at the enzyme surface lining the binding pocket ( Fig. 3  and 6), all indicative for differences in the extended substrate binding cleft that could be causative for the (subtle) differences in substrate interactions observed.
The Extended Specificity Profile is a Major Determinant for Granzyme Substrate Recognition-Western blot analysis was performed to validate substrate cleavage of overexpressed Flap endonuclease 1 (FEN-1). Human and mouse FEN-1 were found to be cleaved by GrH and mut GrC in both differential degradomics analyses, generating an identical neo-N terminus ( 311 CGEKQFSEER 320 ) though more efficiently cleaved by mut GrC (K-562 GrH/mut GrC ratio of 0.38, YAC-1 GrH/mut GrC ratio of 0.22), and moreover harboring acidic residues at positions P3Ј, P8Ј and P9Ј (supplemental Table S1). Fig. 7A shows the MS-spectrum of 2 GIQGLAKLIADVAPSAIR 19 identified as the protein N terminus of human FEN-1 with overall about equal intensities in the three samples (1:0.9:0.8), whereas the neo-N terminus was exclusively present in the granzyme-treated samples with a GrH/mut GrC ratio of 0.38 and 0.22 in the human (Fig. 7B) and mouse proteome background respectively. C-terminally GFP tagged FEN-1 and its presumed GrH/GrC truncated form were overexpressed in K-562 cells. Cell lysates overexpressing full-length GFP-FEN1 and incubated with 200 nM mut GrC (Fig. 8) revealed a fragment of equal molecular weight as the truncated GFP-FEN-1; indicative for the fact that mut GrC cleavage occurred at IKFM 310 .
Next, we investigated the influence of P1 cleavage susceptibility on the efficiency of GrH/mut GrC cleavage by mutation of the FEN-1 P1 residue from Met to Phe or Tyr. Using site-directed mutagenesis PCR, the FEN-1 P1 methionine (M 310 ) was mutated to Tyr (M 310 ϾY) and Phe (M 310 ϾF) and the wild-type and mutant protein variants were in vitro transcribed and translated in the presence of 35 S-Met to investigate GrH/mut GrC cleavage profiles by autoradiography. Cleavage was assessed by incubating the radiolabeled protein with the different granzymes at various concentrations. Similar to the Western blot data (Fig. 8), cleavage of wt FEN-1 FIG. 6. Crystal structure of GrH (left) and model of mut GrC (right). Left side: crystal structure of GrH (PDB:3TJV) with bound substrate peptide PTSYAGDDS. Right side: model of "unlocked" mut GrC substrate recognition based on crystal structure of "locked" wt GrC (PDB:3FFZ) with double mutation of the EE motif to an "unlocked" RG configuration and superimposition of substrate ligand with same conformation as in GrH, followed by short simulated annealing molecular dynamics simulations of enzyme residues within 5 Ångströ m from the ligand for energy minimization using the AMBER03 force field in YASARA (41). Coloring: Asp, Glu… red; Arg, Lys ϭ blue, P1 residue… magenta).
(ϳ43 kDa) by GrH and mut GrC resulted in generation of a 35 kDa fragment (black arrow), again indicative for cleavage at IKFM 310 (Fig. 9A). The wild-type FEN-1 precursor was cleaved more efficiently by mut GrC as compared with GrH in line with the COFRADIC data. Mutating M 310 ϾY did not drastically affect cleavage susceptibility by GrH and mut GrC. However, for mut GrC a small increase in cleavage efficiency could be observed when mutating M 310 ϾF, and a decrease in GrH cleavage susceptibility was observed, overall correlating with the substrate specificity data obtained using N-terminal COFRADIC ( Fig. 4 and supplemental Table S1).
We further observed a shared extended specificity profile between GrB, GrM, GrH and mut GrC comprising a prevalence of acidic residues from P3Ј-P8Ј, which was not observed for GrA ( Fig. 3 and supplemental Fig. S4) (42). We therefore mutagenized FEN-1 to investigate whether the extended specificity profiles are sufficient to steer cleavage by specific granzymes based on an optimal P1 residue. To investigate whether GrB, processing wt FEN1 at distinct sites (Fig. 9B, dashed and gray arrows), can cleave in an otherwise GrH/mut GrC cleavage context, a P1 M 310 ϾD mutant was assayed for granzyme B cleavage susceptibility. Remarkably, GrB generated an additional fragment with an equal molecular weight as compared with the GrC/GrH M 310 cleavage product, accounting for about 33% of the precursor cleavage observed. This indicates that the presence of a P1 Asp residue in the extended substrate specificity profile of an otherwise GrH/mut GrC-specific cleavage event suffices to confer FIG. 7. Representative MS-spectra of FEN-1. The amino acid sequence of the Flap endonuclease 1 protein is shown at the bottom. A, The peptide corresponding to the protein N terminus of FEN-1 (underlined) was identified as a peptide ion triplet, implying its recovery from all setups analyzed with about equal intensities (corrected ratios of 1:0.9:0.8). B, A peptide ion couple separated by 6 Da, representative for their presence in the setups incubated with GrH ( 12 C 6 ) or mut GrC ( 13 C 6 ) and not in the control setup, was identified as 311 CGEKQFSEER 320 (bold in protein sequence) and assigned to FEN-1 cleavage after IKFM 310 (italic)with a GrH/mut GrC ratio of 0.38. GrB cleavage susceptibility. Unexpectedly, the P1 M 310 ϾD mutation did not abolish cleavage by GrH or mut GrC and still resulted in respectively 42 and 68% of the FEN-1 precursor being cleaved.
To investigate if cleavage at P2 (Phe 309 ) could account for this, an IKA 309 A 310 double mutant was created and shown to resist cleavage by either GrH or mut GrC. When mutating the P2-P1 residues FM 310 to DA to investigate if GrB cleavage can occur at position 309 similar to cleavage after Phe 309 by GrH and mut GrC, an IKAA 310 -like pattern could be observed (i.e. no cleavage at position 309 or 310), whereas analysis of the IKAD 310 mutant resulted in 41 and 71% of cleavage by GrB and mut GrC respectively. These results suggest that besides the P1 identity, the extended specificity profile is a key determinant for recognition by GrB, GrH, and mut GrC, and indicates that cleavages are predominantly confined to restricted protein areas. DISCUSSION In this study we investigated the postulated orthologous relationship between human granzyme H (GrH) and murine granzyme C (GrC) and whether they may exert similar functions. Therefore we complemented their high sequence sim-ilarities, sequence identities, highly similar chromosomal locations and overlapping expression profiles with information on their cytotoxicity and their substrate repertoires and specificities. By means of dose-response analyses, GrH (this study) and mutant unlocked GrC (7) were found to be very inefficient cytotoxins as compared with GrB because the amount of GrH and mut GrC needed to kill 50% of granzyme targeted cells is ϳ170-fold and up to 1000-fold higher as compared with GrB respectively. Next, substrate phage display with GrH was performed as described previously for assessing GrC subsite requirements (16). Whereas P4 Ile and P2Ј Gly preferences in GrH and mut GrC cleavage sites are similar, the biggest differences were observed at the P2 and P1 positions. Preferences for Tyr at P2 are quite similar whereas cleavage sites harboring P2 Phe residues are ϳ5fold more efficiently cleaved by GrC as compared with GrH. At P1 preferences for Tyr and Phe could be observed for both orthologous granzymes, with a Phe/Tyr ratio of 1/4 in GrH cleavage sites, whereas the inverse ratio for these amino acids was observed in GrC cleaved sites. N-terminal COFRADIC, allowing direct comparison of both granzymes, was applied as to examine whether the by phage display The degree of cleavage at the GrH/mut GrC cleavage site is indicated by "-" and "ϩ," with "ϩϩϩ" being cleaved most efficiently. The cleavage fragment generated upon GrH/mut GrC cleavage is indicated with a black arrow, whereas GrB cleavage fragments are indicated with dashed and gray arrows. determined subsite features also apply on the macromolecular level. N-terminal COFRADIC results obtained by comparing wild type GrC and an "unlocked" mut GrC previously resulted in the identification of 18 mut GrC substrates, with six of these found to be cleaved by both GrC variants although mut GrC cleaved these 10-fold more efficient (16). Among the substrates identified in this study, all 6 wt GrC substrates were identified. In accordance to phage display data, mut GrC cleaved P1 Phe-containing substrates four-to fivefold more efficiently than GrH. We furthermore observe that cleavage sites harboring either Phe or Tyr at P2 are cleaved more efficiently by mut GrC. Although these differences in cleavage efficiency are only subtle, substrate specificity profiles derived by phage display or positional proteomics gave highly similar, but none-identical patterns, most likely due to the process of phage display selection, specifically enriching for higher affinity substrate peptides through multiple rounds of selection and amplification. By contrast, the positional proteomics strategy used identifies cleavage events in their macromolecular context to provide a more all-encompassing pattern of potential substrates, though not restricted to the most efficiently cleaved substrates (5). Since the substrate specificity profiles and thus biochemical protease characterization were obtained using extensive in vitro degradome analyses, extrapolation to the functional implication and physiological importance of substrate cleavage requires caution. However, using the DAVID Functional Annotation Clustering Tool (7,8), two annotation clusters encompassing ribosomal proteins and helicases showed group enrichment scores of p Յ 0.01 when compared with the experimentally derived mouse and human proteome backgrounds respectively (i.e. respectively 826 and 654 unique mouse and human proteins based on the identification of their database annotated protein N termini). Interestingly, representatives of these categories of substrates have previously been found to be targeted by other granzymes under physiological conditions (9, 10), which might hint to the fact that similar to hnRNPK for instance, other hitherto unidentified substrates might represent pan-granzyme substrates.
In addition to these subtle subsite differences surrounding the P1 residue, a shared extended specificity profile for acidic residues was found, similar to former analyses on human GrB and GrM substrate motifs (21,38). We here showed that this extended protease-substrate interaction can steer protein cleavage because cleavage by GrB can be conferred by a sole mutation at P1 into GrB's preferred P1 aspartic acid residue. These observations suggest granzyme cleavage sites to reside in restricted substrate regions for the aforementioned granzymes, whereas for hGrA (and other proteases) such an extended specificity-profile has not been found (42). As such, the preferential targeting of acidic substrate regions may have exerted a selective force to conserve substrate recognition among specific granzymes and thus to target specific key substrates. On the other hand, investigation of cleavage site locations in macromolecular substrates reveals that cleavage events found in structures appear accessible at the protein surface but can sometimes be found within or at the border of secondary structure elements, as is the case for the FEN-1 IKFM 310 cleavage site. Although rare in our case, in line with the previously reported proteolytic cleavage sites residing in structural elements (i.e. ␣-helices) of proteins with resolved structures (11), we suggest an explanation through the dynamic nature of proteins, in which presumably stable structures exist in a thermodynamic equilibrium with a fraction of (partially) unfolded states, or by the structural rearrangement of the substrate and/or protease active site, as disorder/ order transitions have been demonstrated on protein complex formation (12).
Furthermore, the significant overlap of cleavage sites identified (50%) as compared with another pair of granzyme orthologs; i.e. hGrB and mGrB (21)(47%), are further indicative for a potential functional homology of GrH and GrC.
Of note is the fact that all these analyses were conducted making use of an unlocked, mutant GrC variant because GrC activation was suggested to require an allosteric control mechanism (16). Unlocking GrC by mutating those residues to their GrB counterparts, activates GrC and potentiates the killing potential of mut GrC fourfold.
As the proteomics data suggest similar substrate preferences between GrH and mut GrC, the original crystal conformation of the GrH substrate PTSYAGDDS superimposed into mut GrC as example ligand could be used to interpret relative positions of the surface charges to the substrate positions as well as binding pocket space constraints. Because based on the structural alignment no register shift for GrH can be observed (supplemental Fig. S3), it is difficult to predict whether because of evolution an allosteric control mechanism for only one of these proteases was preserved/has evolved (thus making GrC a strictly controlled protease for specialized functions), more so viewing the fact that granzyme paralogs could have diversified or taken over (parts) of GrH/C function.
In conclusion, we provide further evidence to assign GrH and mut GrC as orthologous granzymes exerting similar functions because they both were found to be poor cytotoxins and they target their substrates making use of similar specificity profiles. In addition, we showed that the extended specificity profile is a critical determinant among several granzymes for cleavage to occur because mutation of the P1 methionine into the GrB preferable aspartic acid infers GrB cleavage susceptibility and furthermore does not abolish GrH and mut GrC cleavage because a neighboring Phe residue together with the overlapping extended specificity profile of the substrate can still be recognized by mut GrC and GrH as such. □ S This article contains supplemental Figs. S1 to S5 and Tables S1 to S4. § § These authors contributed equally to this work. ‡ ‡ To whom correspondence should be addressed: Department of Medical Protein Research, Flanders Interuniversity Institute for Biotechnology, Ghent University, A. Baertsoenkaai 3, B9000 Ghent, Belgium. Tel.: ϩ32-92649279; Fax: ϩ32-92649496; E-mail: petra.vandamme@ vib-ugent.be.