Cytotoxic lymphocyte protease GrM (granzyme M) is a potent inducer of tumour cell death and a key regulator of inflammation. Although hGrM (human GrM) and mGrM (mouse GrM) display extensive sequence homology, the substrate specificity of mGrM remains unknown. In the present study, we show that hGrM and mGrM have diverged during evolution. Positional scanning libraries of tetrapeptide substrates revealed that mGrM is preferred to cleave after a methionine residue, whereas hGrM clearly favours a leucine residue at the P1 position. The kinetic optimal non-prime subsites of both granzymes were also distinct. Gel-based and complementary positional proteomics showed that hGrM and mGrM have a partially overlapping set of natural substrates and a diverged prime and non-prime consensus cleavage motif with leucine and methionine residues being major P1 determinants. Consistent with positional scanning libraries of tetrapeptide substrates, P1 methionine was more frequently used by mGrM as compared with hGrM. Both hGrM and mGrM cleaved α-tubulin with similar kinetics. Strikingly, neither hGrM nor mGrM hydrolysed mouse NPM (nucleophosmin), whereas human NPM was hydrolysed efficiently by GrM from both species. Replacement of the putative P1′–P2′ residues in mouse NPM with the corresponding residues of human NPM restored cleavage of mouse NPM by both granzymes. This further demonstrates the importance of prime sites as structural determinants for GrM substrate specificity. GrM from both species efficiently triggered apoptosis in human but not in mouse tumour cells. These results indicate that hGrM and mGrM not only exhibit divergent specificities but also trigger species-specific functions.
The first line of defence against tumour and virus-infected cells is formed by cytotoxic T-lymphocytes and NK cells (natural killer cells) [1,2]. Upon recognition of a target cell, these cytotoxic lymphocytes can initiate target cell death via either the death receptor  or the granule exocytosis pathway . In the granule exocytosis pathway, cytotoxic lymphocytes secrete a family of granule-associated serine proteases known as granzymes and the pore-forming protein perforin . Perforin facilitates the entry of granzymes into the target cell, enabling granzymes to induce cell death by cleaving intracellular substrates. In humans, five granzymes have been identified [GrA (granzyme A), GrB (granzyme B), GrH (granzyme H), GrK (granzyme K) and GrM (granzyme M)] [6–8]. The mechanisms via which GrA and GrB induce cytotoxicity have been studied extensively, whereas far less is known about the other human granzymes . Granzymes have also been postulated to play key roles in regulating inflammation [10,11].
hGrM (human GrM)  is unique in that it preferably cleaves after a leucine residue [13,14] and is highly expressed in NK cells and to a lesser extent in CD8+ T-cells [15,16]. hGrM is a potent and efficient inducer of tumour cell death in vitro and in vivo [17,18]. This cell death is characterized by rapid cell swelling, formation of large cytoplasmic vacuoles, chromatin condensation with only slight segmentation of the nuclei, and finally lysis of the cells . The molecular mechanisms by which hGrM initiates cell death remain controversial in the literature. hGrM has been demonstrated to trigger cell death in a caspase-independent fashion, with no fragmentation of DNA, no formation of ROS (reactive oxygen species) and no perturbation of mitochondria [17,19,20]. In these studies, hGrM has been demonstrated to cleave the microtubule network component α-tubulin, leading to disorganization of the microtubule network , and NPM (nucleophosmin)/B23, a multifunctional phosphoprotein essential for cell survival . In contrast, Fan and co-workers have reported that hGrM promotes cell death in a manner similar to GrB, including caspase 3 activation, DNA fragmentation, generation of ROS and cytochrome c release from the mitochondria, through cleavage of HSP-75 (heat-shock protein-75), iCAD (inhibitor of caspase-activated DNase), PARP [poly(ADP-ribose) polymerase] and survivin [21–23].
While there are five granzymes in humans, mice express at least ten granzymes (A, B, C, D, E, F, G, K, M and N) . hGrM has an orthologue in mice and mGrM (mouse GrM) functions in vivo have been studied in an mGrM knockout mouse model [10,18,24]. Like hGrM, mGrM has been implicated to participate in tumour clearance . More recently mGrM has also been proposed to play a key role in the regulation of inflammation in vivo via yet to be established mechanisms . Although hGrM and mGrM display extensive sequence homology, very little is known about the specificity of the mGrM protease. In the present study, we show that hGrM and mGrM exhibit divergent and species-specific substrate specificities. hGrM and mGrM display distinct P1 and subsite preferences, have narrow macromolecular substrate specificities that overlap only partially, cleave human but not mouse NPM, and trigger apoptosis in certain human but not mouse tumour cell lines. This stresses that caution is needed when using mouse models to elucidate GrM functions in humans.
Cell lines, antibodies and reagents
Human HeLa and murine C2C12 cells were grown in DMEM (Dulbecco's modified Eagle's medium), supplemented with 10% fetal bovine serum, 100 units/ml penicillin and 100 μg/ml streptomycin (Invitrogen). For the N- and C-terminal COFRADIC (combined fractional diagonal chromatography) analyses, human K-562 cells were grown in RPMI 1640 glutamax medium, supplemented with 10% dialysed fetal bovine serum, 100 units/ml penicillin and 100 μg/ml streptomycin containing 57.5 μM natural, 13C6 or 13C615N4L-arginine (Cambridge Isotope Laboratories). Cells were passaged for at least six population doublings for complete incorporation of the labelled arginine. Cell-free protein extracts were generated from exponentially growing HeLa, C2C12 and K-562 cells. Cells (108 cells/ml) were washed two times in a buffer containing either 50 mM Tris (pH 7.4) and 150 mM NaCl (HeLa and C2C12 cells) or 50 mM Tris (pH 8.0) and 100 mM NaCl (K-562 cells), and lysed in the same buffer by three cycles of freeze–thawing. Samples were centrifuged for 10 min at 20000 g at 4 °C, and cell-free protein extracts were stored at −80 °C. The protein concentration was quantified using the Bradford method. Antibodies used were anti-α-tubulin clone B-5-1-2 (Sigma), anti-β-tubulin clone TUB 2.1 (Sigma) and anti-NPM clone FC-61991 (Invitrogen). SLO (streptolysin O) was purchased from Aalto Bio Reagents. Homology modelling was performed using SWISS-MODEL .
The cDNA encoding mature mGrM (residues Ile27-Val264) was amplified from mouse thymus cDNA (MD-702, Zyagen) using the oligonucleotides 5′-CCGCTCGAGAAACGTATCATTGGGGGTCGAG-3′ and 5′-TAAAGCGGCCGCCTTAGACCAAAGATTGGGG-3′, and cloned into yeast expression vector pPIC9 (Invitrogen). Catalytically inactive mGrM-SA (mGrM with S195A mutation in catalytic centre), in which the Ser195 residue in the catalytic centre is replaced by alanine, was generated by site-directed mutagenesis (Stratagene). Plasmids were transformed into the GS115 (his4) strain of Pichia pastoris. hGrB, hGrM and mGrM and the catalytically inactive GrM-SA (GrM with S195A mutation in catalytic centre) mutants were produced in P. pastoris and purified using cation-exchange chromatography as described previously . GrM fractions were dialysed against 50 mM Tris (pH 7.4) and 150 mM NaCl and stored at −80 °C. GrM activity was determined using the synthetic chromogenic substrates Suc-Ala-Ala-Pro-Leu-pNA (p-nitroanalide) (Bachem) and Suc-Lys-Val-Pro-Leu-pNA (GL Biochem). To directly compare hGrM and mGrM substrate specificity, both enzymes were titrated using the synthetic chromogenic substrate Suc-Ala-Ala-Pro-Leu-pNA. hGrM was twice as efficient as its mouse counterpart in cleaving this substrate. The catalytically inactive GrM-SA mutants of both hGrM and mGrM did not show any activity (results not shown). Human and mouse granzyme concentrations in all experiments were matched based on Suc-Ala-Ala-Pro-Leu-pNA hydrolysis. Human α-tubulin, mouse α-tubulin, human NPM and mouse NPM cDNAs were amplified from IMAGE clones 3871729, 6306481, 5575414 and 30438901 respectively, and cloned into the bacterial expression vector pQE80L. N-terminally His-tagged α-tubulin and His-tagged NPM were expressed in Escherichia coli strain BL21 as recommended by the manufacturer (Roche). The NPM P1′, P2′ and P1′/P2′ mutants were generated by site-directed mutagenesis (Stratagene). Recombinant α-tubulin and NPM proteins were purified by metal-chelate chromatography (Clontech), dialysed against PBS, and stored at −80 °C.
Granzyme killing assays
HeLa and C2C12 cells were grown to confluence in a 96-well tissue-culture plate. Cells were washed twice in serum-free DMEM, after which they were incubated at 37 °C with a sublytic dose of SLO (Jurkat, 0.25 μg/ml SLO; HeLa and PC3, 0.5 μg/ml SLO; LR7, C26 and C2C12, 1 μg/ml SLO) and indicated concentrations of granzyme in a buffer containing 20 mM Tris and 150 mM NaCl for 30 min. The cells were washed twice with supplemented DMEM, after which the cells were incubated for another 20 h at 37 °C. Cell viability was assessed using flow cytometry. Cells were stained with Annexin V-fluos (Invitrogen) and PI (propidium iodide) for 15 min in a buffer containing 140 mM NaCl, 4 mM KCl, 0.75 mM MgCl2, 1.5 mM CaCl2 and 10 mM Hepes (pH 7.4). Flow cytometry was performed on a FACSCalibur™ instrument (BD Biosciences) and results were analysed using CellQuest Pro software (BD Biosciences). Cells that were negative for both Annexin V and PI were regarded as living cells. The percentage of viable cells after treatment with SLO only was set at 100% and the percentage of viable cells in other conditions were calculated accordingly.
Single substrate kinetics
Granzyme activity was monitored by the synthetic chromogenic substrates Suc-Ala-Ala-Pro-Leu-pNA (Bachem) and Suc-Lys-Val-Pro-Leu-pNA (GL Biochem Ltd) in 100 mM Hepes (pH 7.4) and 200 mM NaCl at 37 °C. The GrM concentration was 100 nM and the substrate concentration ranged from 0.5 to 3 mM. Hydrolysis of pNA substrates was monitored spectrophotometrically at 405 nm on an Anthos Zenyth 340 rt microtitre plate reader (Anthos). Kinetic parameters (kcat/KM) were determined using standard Michaelis–Menten kinetic equations.
Positional scanning synthetic combinatorial libraries
The preparation and characterization of the P1-diverse and P1-Met libraries of 7-ACC (amino-4-carbamoylmethylcoumarin) substrates used in the present study are described elsewhere [26,27]. Then 10−10 mol of each well of the P1-diverse stock library was added to 20 wells of a 96-well microfluor plate. The final concentration of each substrate in the assay was 250 μM. Assays were initiated by the addition of 5.9 μM GrM and were conducted at 25 °C in buffer containing 50 mM Tris (pH 7.4) and 150 mM NaCl for 2 h. Hydrolysis of substrates was monitored fluorimetrically with an λex of 380 nm and an λem of 460 nm on a Spectramax Gemini microtitre plate reader (Molecular Devices).
Fluorescent 2D-DIGE (two-dimensional difference gel electrophoresis)
For 2D-DIGE, 100 μg of HeLa or C2C12 cell-free protein extract was incubated for 1 h at 37 °C with 1 μM hGrM or mGrM or their catalytically inactive mutants. Samples were precipitated, solubilized, labelled, rehydrated and isoelectrically focused as we have described previously . The strips were reduced and overlaid on an SDS/12% PAGE gel. The images were acquired on a Typhoon 9410 scanner (GE Healthcare). Each condition was performed at least five times, and a dye swap was included to exclude preferentially labelled proteins from the analysis. The relative quantification of matched gel features was performed using Decyder DIA and BVA software (GE Healthcare). For inter-gel analyses, the internal standard method was used as described previously . Statistical analysis was performed using the Student's t test. P<0.05 was regarded as statistically significant.
N- and C-terminal COFRADIC: isolation of terminal peptides
For N- and C-terminal COFRADIC, 1250 μg of a K-562 cell-free protein extract was either left untreated (13C615N4L-arginine-labelled sample) or treated with 200 nM recombinant hGrM (12C6L-arginine-labelled sample) or recombinant mGrM (13C6L-arginine-labelled sample) for 1 h at 37 °C. Following protease incubation, guanidinium hydrochloride was added to the cell lysates to a final concentration of 4 M in order to denature and inactivate the proteases. The protein samples were reduced and S-alkylated, followed by trideutero-acetylation of primary amines and trypsin digestion as described previously [30,31]. N- and C- terminal COFRADIC analyses were performed as described in . In this setup, 12C4-butyrylation was used for the hGrM-treated sample, 13C4-butyrylation for the mGrM-treated sample and 13C2-butyrylation for the control sample.
LC (liquid chromatography)–MS/MS (tandem MS) analysis
LC–MS/MS analysis was performed using an Ultimate 3000 HPLC system (Dionex) in-line connected to an LTQ Orbitrap XL mass spectrometer (Thermo Electron) and, per LC–MS/MS analysis, 5 μl of sample (one-quarter of the total sample) was consumed. LC–MS/MS analysis and generation of MS/MS peak lists were performed as described in . These MS/MS peak lists were then searched with Mascot using the Mascot Daemon interface (version 2.2.0, Matrix Science). Searches were performed in the Swiss-Prot database with taxonomy set to human (UniProtKB/Swiss-Prot database version 2010_10 containing 20258 human protein sequences). Trideutero-acetylation at lysine residues, carbamidomethylation of cysteine residues and methionine oxidation to methionine-sulfoxide were set as fixed modifications. Variable modifications were trideutero-acetylation and acetylation of protein N-termini. Semi-ArgC was set as the used protease (no missed cleavages were allowed) and the mass tolerance on the precursor ion was set to 10 p.p.m. and on fragment ions to 0.5 Da. In addition, Mascot's C13 setting was set to 1. Only MS/MS spectra that exceeded the corresponding Mascot threshold score of identity (at 95% confidence level) were withheld. The estimated false discovery rate by searching decoy databases was typically found to lie between 2 and 4% on the spectrum level . All quantifications [SILAC (stable isotope labelling by amino acids in cell culture); 12C6L-arginine compared with 13C6L-arginine, 12C6L-arginine compared with 13C615N4L-arginine and 13C6L-arginine compared with 13C615N4L-arginine) and butyrylation (12C4-butyrylated compared with 13C4-butyrylated, 12C4-butyrylated compared with 13C2-butyrylated and 13C2-butyrylated compared with 13C4-butyrylated)] were carried out using the Mascot Distiller Quantitation Tool (version 2.2.1). The quantification method details were as follows: constrain search, yes; protein ratio type, average; report detail, yes; minimum peptides, 1; protocol, precursor; allow mass time match, yes; allow elution shift, no; all charge states, yes; fixed modifications, mass values. Ratios for the proteins were calculated by comparing the XIC (extracted ion chromatogram) peak areas of all matched light compared with medium, light compared with heavy and medium compared with heavy peptides. The calculated ratios that were reported as FALSE were all verified by visual inspection of all highest scoring MS spectra.
The primary and extended specificities of hGrM and mGrM are distinct
hGrM and its mouse orthologue display a considerable sequence homology of 69% (Figure 1A), whereas the serine, histidine and aspartic acid residues of the catalytic triad are completely conserved between hGrM and mGrM, several residues that are predicted to play key roles in the substrate recognition of hGrM  differ between both granzymes (Figure 1A). These differences result in an altered substrate-binding pocket in mGrM as can be visualized in a homology model of mGrM based on the known crystal structure of hGrM (Figure 1B) . hGrM prefers to cleave after a leucine residue at the P1 position (nomenclature for amino acid positions in substrates is Pn-P2-P1-P1′-P2′-Pn′, with amide bond hydrolysis occurring after P1, and the corresponding enzyme-binding sites denoted as Sn-S2-S1-S1′-S2′-Sn′) . The optimal P4–P1 tetrapeptide specificity of hGrM has previously been identified as Lys-Val-Pro-Leu [13,14]. To test whether the tetrapeptide specificity of mGrM is similar to that of hGrM, mGrM was incubated with the chromogenic optimal human tetrapeptide substrate Suc-Lys-Val-Pro-Leu -pNA (Figure 1C). While hGrM efficiently hydrolysed this substrate (kcat/Km=6.6×103 M−1·s−1), mGrM was approx. 4.5-fold less efficient (kcat/Km=1.5×103 M−1·s−1). This suggests that hGrM and mGrM are indeed structurally distinct proteases with distinct substrate-binding pockets. The primary specificity of mGrM was profiled using a PS-SCL (positional scanning synthetic combinatorial library) of tetrapeptide coumarin substrates known as the P1-diverse library [14,27]. For hGrM, this method has previously demonstrated a strong P1 preference for leucine (100%) over methionine (~16%) and non-physiological Nle (~46%), and no tolerance for other residues at this position [13,14]. In contrast with hGrM, mGrM displayed a preference for Nle (100%) and methionine (~95%) over leucine (~68%) (Figure 1D). The residues alanine, glutamine, histidine, lysine, phenylalanine, serinine, tryptophan and tyrosine were also tolerated at the P1 position, although to a lesser extent. The P4–P2 extended substrate specificity of mGrM was determined using a P1-fixed methionine PS-SCL library (Figure 1D). Strikingly, whereas hGrM shows an almost absolute requirement for proline at the P2 with some tolerance for alanine , mGrM clearly preferred alanine over proline and displayed a broad tolerance for other amino acids at P2. Although the preferred P3 residues of mGrM differed slightly from those of hGrM, both granzymes seemed to favour glutamine or valine at P3. At the P4 position, mGrM was highly tolerant to a large number of amino acids, whereas hGrM showed a distinct preference for a lysine, Nle or a histidine residue at P4 . Taken together, these results indicate that hGrM and mGrM differ in their primary and P4–P2 substrate specificity.
The primary and extended specificities of hGrM and mGrM are distinct
hGrM and mGrM display restricted macromolecular substrate specificities that overlap only partially
To compare the macromolecular substrate specificities of hGrM and mGrM, two different proteomics-based strategies were applied. First, a fluorescence 2D-DIGE proteomic approach was employed that scans the native proteome of tumour cells for macromolecular substrates of both granzymes and directly determines the efficiency of substrate cleavage (Figure 2). Cell lysates of human cervix carcinoma (HeLa) and mouse myoblast (C2C12) cells were incubated with hGrM, hGrM-SA, mGrM or mGrM-SA and were subsequently labelled with either a red fluorescent dye [Cy5 (indodicarbocyanine)] or a green fluorescent dye [Cy3 (indocarbocyanine)]. Two representative 2D-DIGE gels of cell lysates after hGrM (Figure 2A) or mGrM (Figure 2B) treatment are shown. GrM-treated lysates in these gels were labelled green, whereas the lysates that had been treated with GrM-SA were labelled red. Spots present in greater abundance in the control sample appear red and indicate possible intact GrM substrates, whereas spots present in greater abundance in the granzyme-treated sample appear green and reflect the appearance of specific cleavage products (the unaffected proteome appears yellow). Approx. 3000 protein spots were resolved from the human HeLa cell lysate, of which 44 spots clearly appeared (cleavage products) and 40 (~1.3%) decreased in intensity with different efficiencies (intact substrates) after hGrM treatment (Figure 2A) (log peak volume change >2-fold, P<0.05). For mouse C2C12 cell lysates, approx. 2300 proteins spots were resolved. After mGrM treatment, 39 protein spots clearly appeared (cleavage products) and 12 (~0.5%) decreased in intensity with different efficiencies (intact substrates) (Figure 2B) (log peak volume change >2-fold, P<0.05). This relatively low number of recognized substrates for hGrM and mGrM suggests that the macromolecular substrate specificities of both granzymes are highly restricted. Interestingly, 17 out of the 40 macromolecular substrates that could be detected were cleaved and shared by both granzymes in HeLa lysates (log peak volume change >2-fold, P<0.05), indicating a ~42% overlap in macromolecular substrate specificity (Figure 2C). For the cleavage fragments, 36 out of 51 spots with at least a 2-fold log peak volume change were shared, strongly suggesting that part of the detected cleavage events occurred at the same P1 cleavage sites, whereas, for the selection criteria applied, other events were considered unique for mGrM or hGrM. In the C2C12 lysate, 11 out of 32 proteins were cleaved by both granzymes, indicative of an ~34% overlap in macromolecular substrate specificity, while 19 out of 61 cleavage fragments overlapped (~31%) (Figure 2D).
hGrM and mGrM display highly restricted macromolecular substrate specificities that only partially overlap
To further probe the macromolecular differences in substrate selection among the human and mGrM orthologues, we made use of the COFRADIC-based complementary positional proteomics approach to study GrM-specific proteolysis in human K-562 cell lysates. This approach allows us to identify and to compare the (consensus) cleavage sites of hGrM and mGrM, but does not directly determine efficiencies of cleavage events. Analogous to previous GrB setups analysed [32,36], a SILAC  12C6-L-Arg- and 13C6-L-Arg-labelled cell lysate served as a hGrM and mGrM substrate pool respectively, whereas a 13C615N4-L-Arg-labelled proteome served as a control. This allowed quantification of (neo-)N-terminal peptides since all samples were subject to tryptic digestion and therefore end with an arginine residue. We further differentially tagged protein C-terminal peptides using NHS-esters of 12C4 (hGrM), 13C2 (control) or 13C4 butyric acid (mGrM). Following LC-MS/MS analysis, we here identified 720 unique hGrM- and/or mGrM-specific cleavage sites in 488 protein substrates based on their corresponding 577 neo-N and/or 155 neo-C-terminal peptide(s) (see Supplementary Table S1 at http://www.BiochemJ.org/bj/437/bj4370431add.htm). Of these, 411 hGrM- and/or mGrM-specific cleavage sites were generated by processing at leucine (57%), 75 by cleavage after methionine (10%) and the remaining 234 cleavage sites (33%) were, consistent with the positional scanning results, raised upon processing at alternative P1-specificities, including alanine, cysteine and glutamine. Besides monitoring of (neo-)termini generated by GrM, the differential labelling strategies applied allow for a direct comparison of differences in substrate specificity profiles for both granzymes. As such, single 12C6L-Arg neo-N- or 12C4-butyrylated neo-C-termini and 13C6L-Arg neo-N- or 13C4-butyrylated neo-C-termini indicate respectively unique hGrM and mGrM substrates, whereas couples spaced by six or four mass units respectively indicate neo-N- or neo-C-termini raised by both orthologues (Figure 3). The ratio of ion signal intensities of such couples is further indicative of the difference in substrate cleavage efficiency between hGrM and mGrM. Of all sites identified, 481 sites (67%) in 359 substrates were found to be cleaved by both proteases, whereas 230 (196 proteins) and nine cleavages (nine proteins) were uniquely introduced by the action of hGrM and mGrM respectively (see Supplementary Table S1). As deduced from the neo-termini-specific hGrM/mGrM ratios, on average, hGrM seemed to be 4- to 5-fold more efficient in cleaving its substrates as compared with mGrM.
Illustration of the different categories of isolated terminal peptides found in K-562 lysates
The COFRADIC-based complementary positional proteomics approach allows us to directly compare consensus cleavage sites of hGrM and mGrM. The general amino acid conservation in the set of GrM substrates with a leucine or methionine residue at P1 that were more efficiently cleaved by hGrM than by mGrM is shown in Figure 4(A). The general amino acid conservation in the set of more efficiently cleaved mGrM substrates with P1 leucine or methionine is shown in Figure 4(B). Overall, the identified P1–P4 specificity of mGrM is consistent with our PS-SCL data (Figure 1D). To further distinguish between differentially accommodated amino acids from P4 to P4′ in their respective subsites of hGrM and mGrM, a differential iceLogo  was created using the human and mouse leucine and methionine P1-specific data subsets for which the hGrM or mGrM cleavage efficiency (as deduced from the proteomics data) was at least 50% more efficient as compared with the mean hGrM compared with the mGrM cleavage ratio observed (Figure 4C). Statistically significant residues (P≤0.01) are plotted with the size of the amino acid proportional to the difference observed in cleavage efficiency between hGrM and mGrM on their respective substrates (Figure 4C). In line with the PS-SCL data, a P1 methionine residue seems to be better accommodated by mGrM as compared with hGrM. Other differentially accommodated residues include glycine or glutamine residues at P2, a P2 glycine residue being generally inhibitory for hGrM-specific cleavage (the increased prevalence of a glycine residue at P2 is also apparent when comparing the P2 hGrM and mGrM PS-SCL preferences), and a glutamine residue being inhibitory for mGrM-specific cleavage (see also representative examples Figures 3A–3F). In general, the presence of a positively charged P4 lysine or histidine residue is more stimulatory for hGrM-specific cleavage as compared with mGrM-specific cleavage (an arginine residue at P4). Beyond the P4–P1 motif, the amino acid occupancy of P2′ is also clearly discriminative with leucine, phenylalanine, tyrosine and aspartic acid residues being better accommodated by the mGrM S2′ pocket, whereas alanine, asparagine, lysine, serine, glycine and valine residues represent better accommodated hGrM P2′ residues. In addition, Supplementary Figure S1 (available at http://www.BiochemJ.org/bj/437/bj4370431add.htm) plots the different amino acid occurrences in percentages at the most discriminative P and P′-positions in more efficient hGrM compared with more efficient mGrM cleavage site motifs as histograms.
IceLogo representations of the cleavage sites identified by proteome-wide screening for GrM substrates
Collectively, these results indicate that hGrM and mGrM display narrow macromolecular substrate specificities that overlap only partially. Interestingly, the hGrM and mGrM substrates that were identified by this complementary positional proteomics approach include the already known hGrM substrates α-tubulin  and NPM  (Supplementary Table S1).
hGrM and mGrM cleave human and mouse α-tubulin with similar efficiency
We have previously demonstrated that hGrM cleaves the microtubule component α-tubulin, leading to a disorganization of the microtubule network that may contribute to cell death . The proteolysis of α-tubulin was further biochemically analysed in human (HeLa) and mouse (C2C12) cell lysates. Mouse α-tubulin was cleaved by mGrM in a time- and concentration-dependent manner similar to the cleavage of human α-tubulin by hGrM (Figure 5A). Semi-quantitative analysis of the protein bands indeed showed similar kinetics of α-tubulin cleavage by both hGrM and mGrM (Figure 5B). mGrM also cleaved human α-tubulin and vice versa with similar efficiency (Figure 5C). To exclude the possibility that GrM-induced cleavage of α-tubulin is indirect and to visualize GrM-induced cleavage products, these experiments were repeated using purified recombinant human and mouse His-tagged α-tubulin. Consistent with Figures 5(A)–5(C), both human and mouse His-tagged α-tubulin were cleaved by both hGrM and mGrM (Figure 5D). Despite the fact that α-tubulin is extremely well conserved between human and mouse (99.6% sequence homology) (Figure 5E), the cleavage fragments that appeared after GrM treatment of purified His-tagged α-tubulin differed between hGrM and mGrM (Figure 5D). These results indicate that although both hGrM and mGrM cleave α-tubulin with similar efficiency, distinct cleavage sites are preferred (see also Supplementary Table S1).
hGrM and mGrM cleave α-tubulin with similar efficiency
Species-specific substrate proteolysis: hGrM and mGrM cleave human but not mouse NPM
The nucleolar phosphoprotein NPM is a multifunctional tumour-suppressor protein that has been directly implicated in cancer pathogenesis  and apoptosis . Cleavage of NPM by GrM has been proposed to contribute to the mechanism by which GrM triggers tumour cell death . The proteolysis of NPM by hGrM and mGrM was tested in HeLa and C2C12 cell lysates (Figure 6A). As expected , cleavage of human NPM by hGrM was highly efficient and time- and concentration-dependent, with proteolysis almost going to completion at 15 min after the addition of 400 nM hGrM or at 1 h after the addition of 50 nM hGrM (Figure 6B). Cleavage of mouse NPM, however, was far less efficient, with virtually no reduction of the full-length protein band after 2 h incubation with 400 nM mGrM. Kinetic analysis revealed that mGrM was ~40-fold less efficient in cleaving its species-matched NPM as compared with hGrM (Figure 6B). To determine whether this marked variation in NPM cleavage was due to differences between hGrM and mGrM or due to differences between human and mouse NPM, NPM from both human and mouse cell lysates was exposed to both granzymes (Figure 6C). Interestingly, while human NPM was hydrolysed with similar efficiency by both hGrM and mGrM, mouse NPM was a poor substrate of both mGrM and hGrM. Similar results were obtained when the granzymes were incubated with purified recombinant human and mouse His-tagged NPM (Figure 6D), suggesting that differences in NPM between species rather than differences in GrM explain these findings. Human and mouse NPM share a sequence similarity of 93.9% and the known hGrM cleavage site  (at least P9–P1) in NPM is conserved between human and mouse. Cleavage of NPM at this site was also detected in a human K-562 lysate using complementary positional proteomics (observed hGrM/mGrM ratio of 1.6, see Supplementary Tables S1A and S1B). Notably, there are two alanine residues at the P1′ and P2′ positions of human NPM that are not present in mouse NPM. Instead, mouse NPM harbours the non-preferred (Figure 4A) negatively charged Asp159 and Glu160 residues at these positions (Figure 6E) respectively. To determine whether this difference is responsible for the impaired cleavage of mouse NPM by both mGrM and hGrM, the Asp159 and Glu160 residues of mouse NPM were ‘humanized’ using site-directed mutagenesis (Figure 6F). The efficiency of both hGrM and mGrM-mediated cleavage of mouse NPM slightly increased when the P1′ residue of mouse NPM was mutated into the corresponding residue of human NPM, and the efficiency was completely restored when only the P2′ residue was replaced with an alanine or when both residues were replaced simultaneously (Figure 6G). These results show that mouse NPM cannot be cleaved by both hGrM and mGrM because of species-specific differences in the prime site residues P1′ and predominantly P2′. Consistent with our positional proteomics analysis, this further stresses that prime sites in macromolecular substrates are important structural determinants for GrM substrate specificity.
hGrM and mGrM cleave human but not mouse NPM
GrM-induced cell death is species-dependent
The functional and structural species-dependent differences identified in the present study between hGrM and mGrM prompted us to investigate the tumour cell death potential of both granzymes. Human (HeLa) and mouse (C2C12) cells were incubated with purified GrM in the presence or absence of the pore-forming protein SLO, a perforin analogue. The cells were subsequently stained with apoptosis markers Annexin V and PI, and cell viability was assessed by means of flow cytometry. Both hGrM and mGrM efficiently induced cell death in human HeLa cells, with mGrM being slightly more potent than hGrM (Figure 7A). Additionally, hGrM and mGrM displayed clear cytotoxic potential towards the human prostate cancer PC3 and leukaemic Jurkat cell lines (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/437/bj4370431add.htm). Under the same conditions, however, neither hGrM nor mGrM was able to induce apoptosis in mouse C2C12 cells (Figure 7B). GrM from both species also failed to trigger apoptosis in several other mouse cell lines, including the colon carcinoma C26 cell line and the fibroblast LR7 cell line (Supplementary Figure S2). The catalytically inactive GrM-SA mutants and SLO alone displayed no cytotoxicity in these experiments, whereas human GrB efficiently induced apoptosis in all above-mentioned cell lines (Figure 7 and Supplementary Figure S2). These results indicate that, at least in the cell lines tested, both mGrM and hGrM have cytotoxic potential in human cells, but not in mouse cells.
hGrM and mGrM trigger cell death in human HeLa but not mouse C2C12 tumour cells
Granzymes are key players in the effector arm of the immune response against tumour and virus-infected cells [1,4]. Previously, it has been demonstrated that hGrB and mGrB are structurally and functionally different in that GrB from both species cleaves pro-caspase 3, whereas only hGrB is able to cleave BID [36,40–42]. In the present study, we have employed three distinct (proteomic) approaches to show that hGrM and mGrM also exhibit divergent and species-specific substrate specificities. mGrM and hGrM displayed distinct, but to some extent overlapping, tetrapeptide and macromolecular substrate specificities that were partially species-specific (Figures 1–4), whereas α-tubulin was a shared GrM substrate between species (Figure 5), only human but not mouse NPM was cleaved by both granzymes (Figure 6). Both hGrM and mGrM efficiently triggered apoptosis in human HeLa cells, but not in several mouse (tumour) cell lines (Figure 7). These results show that hGrM and mGrM are structurally and functionally divergent and cannot be used interchangeably.
Using a PS-SCL and a complementary positional proteomics approach, we showed that the P1 and extended specificities of hGrM and mGrM differ substantially. The PS-SCL results showed that, whereas hGrM prefers a leucine residue at P1 [13,14], mGrM favoured a methionine (or Nle) leucine residue at this position (Figure 1D). The results of the PS-SCL were confirmed by positional proteomics, which also revealed that leucine and methionine residues are major P1 determinants for hGrM and mGrM (Figure 4). Upon differential comparison between both granzymes (Figure 4C), mGrM was better able to accommodate a methionine residue at P1, whereas hGrM clearly preferred a leucine residue. These results are in agreement with the facts that leucine and methionine are large hydrophobic residues and that the S1 pocket of hGrM – which can only accommodate long, narrow hydrophobic amino acids [34,43] – is conserved in mGrM (Figure 1). The substrate specificity of hGrM and mGrM diverges further at the P2 position. PS-SCL and positional proteomics showed preferences for alanine and proline residues at P2 for hGrM [13,14] and mGrM (Figures 1D, 4A and 4B). Major differences between hGrM and mGrM at the P2 position include a better tolerance to glutamic acid, threonine and valine by hGrM and to glycine by mGrM (Figure 4C). Both hGrM and mGrM displayed a clear preference for a glutamic acid residue at the P3 position, which was demonstrated by both PS-SCL (Figure 1D) and positional proteomics (Figure 4). The PS-SCL results showed that mGrM had a broad preference at the P4 position (Figure 1D), whereas in natural substrates mGrM favoured a lysine or arginine residue at this position, the latter being discriminative for mGrM as compared with hGrM (Figure 4). In PS-SCL, hGrM shows a clear preference for the positively charged lysine and histidine residues at P4 , which can also be deduced from positional proteomics (Figure 4).
Apart from the non-prime subsite preferences, we show for the first time that prime site residues are also important structural determinants for GrM-substrate recognition. In our positional proteomics approach, the amino acid occupancy of the P2′ position was found to be clearly discriminative between hGrM and mGrM, with hGrM accommodating alanine, asparagine, lysine, serine, glycine and valine residues, and mGrM accommodating leucine, phenylalanine and tyrosine residues to a better extent at this position (Figure 4C). Furthermore, replacement of the putative P1′–P2′ residues in mouse NPM with the corresponding residues of human NPM fully restored cleavage of mouse NPM by both hGrM and mGrM (Figure 6). It seems unlikely that major structural differences around the GrM cleavage site between human and mouse NPM affect GrM proteolysis, since secondary structures and solvent accessibility of the GrM cleavage site regions in human and mouse NPM are predicted to be similar. The importance of prime site residues for GrM cleavage is compatible with the notion that relatively high concentrations of GrM were required for robust activity in PS-SCL (Figure 1D).
We employed two proteomic approaches to dissect the macromolecular substrate specificities of hGrM and mGrM (Figures 2–4 and Supplementary Table S1). Some substrates were shared, whereas others were specific for either hGrM or mGrM. Using the 2D-DIGE gel-based approach, only limited sets of hGrM (40, ~1.3% of total spots) and mGrM (12, ~0.5% of total spots) substrates were resolved in their species-matched cell lysates, indicating that both granzymes are highly specific. In the human HeLa cell lysate, 40 GrM substrates were identified, of which 17 (~42%) were shared by hGrM and mGrM. Using the gel-free complementary positional proteomics approach, we identified a total of 488 GrM substrates in the human K-562 cell lysate, of which 359 substrates (~74%) were shared between hGrM and mGrM. This indicates that the macromolecular substrate specificities of hGrM and mGrM only partially overlap. The discrepancy in the number of identified GrM substrates between the gel-free positional proteomics approach as compared with the gel-based 2D-DIGE approach is most likely due to the selection criteria and to the intrinsic higher sensitivity of gel-free proteomics . Both the 2D-DIGE and the COFRADIC approach showed that while hGrM and mGrM share some cleavage sites in macromolecular substrates, they also make use of unique sites, which is in agreement with the partially overlapping primary and extended substrate specificity data as determined by both the PS-SCL and the positional proteomics approach (Figures 1D and 4, and Supplementary Table S1). Interestingly, both approaches revealed only few unique mGrM substrates.
The ability of GrM to cleave specific macromolecular substrates partially depends on the substrate conservation between human and mouse, suggesting that GrM has species-specific functions (Figures 5 and 6). The multifunctional phosphoprotein NPM is essential for cell viability and plays a role in viral replication [20,45,46]. Recently, it has been demonstrated that cleavage of human NPM by hGrM abolishes NPM function, which has been suggested to contribute to GrM-induced cell death . NPM is a species-specific substrate of GrM: murine NPM was cleaved by neither mGrM nor hGrM, whereas human NPM was cleaved by both granzymes (Figure 6). Interestingly, hGRB and mGrB can also efficiently cleave human NPM , but fail to cleave mouse NPM . These results indicate that the cleavage and inactivation of NPM does not contribute to GrM and GrB function in mice. In contrast, the microtubule network component α-tubulin of both human and mouse origin was cleaved by GrM of both species (Figure 5). Although α-tubulin is highly conserved between human and mouse (>99%), it is cleaved by hGrM and mGrM at different sites and/or with different efficiencies at specific sites (Figure 5D and Supplementary Table S1). Apparently, targeting of α-tubulin by GrM and subsequent disorganization of the microtubule network are conserved between human and mouse and probably contributes to cytotoxic lymphocyte-induced cell death .
Although hGrM has previously been shown to efficiently induce cell death in multiple human cell lines [17,19,20,22,34], both hGrM and mGrM did not trigger cell death in several (tumour) cell lines of mouse origin (Figure 5). Whether or not GrM induces cell death in mice remains unclear and its lack of cytotoxicity in mouse cell lines may be due to several reasons. First, the cell death potential of GrM may depend on the type of target cell line that is used. mGrM efficiently induced cell death in human HeLa cells, indicating the apoptotic potential of this granzyme (Figure 5). Strikingly, mGrM did not trigger apoptosis in several mouse cell lines (i.e. C2C12, C26 and LR7) (Figure 5). In agreement with this, Pao et al.  have shown that GrM is not essential for NK cell-mediated cytotoxicity against tumour targets in mice. However, a more recent mouse study has demonstrated that adoptively transferred NK cells from wild-type mice, but not from GrM-deficient mice, effectively inhibit the growth of a subcutaneous tumour . Secondly, the ability of GrM to trigger apoptosis in mouse cell lines may be hampered by the presence of specific inhibitors of GrM-mediated apoptosis such as the murine serine protease inhibitor SPI-CI, which has previously been shown to directly inhibit GrM activity . Currently, no natural intracellular inhibitor of hGrM has been described. Finally, apart from α-tubulin and NPM, it remains unknown whether murine orthologues of other known hGrM death substrates such as HSP-75, iCAD, PARP and survivin [21–23] are cleaved by GrM in mice.
Mouse models are often employed to study the role of granzymes in vivo. GrM-knockout mice have demonstrated a role for GrM in the host response against tumours  and cytomegalovirus infection . Recently, GrM-deficient mice have allowed the identification of GrM as being a key regulator of inflammation, possibly by enhancing the inflammatory cascade downstream of LPS (lipopolysaccharide)-TLR4 (Toll-like receptor 4) signalling . Whether or not these in vivo data can be directly extrapolated to elucidate the functions of GrM in humans remains uncertain, since the species-specific substrate specificities of GrM signify that caution is essential when interpreting data from mouse studies.
combined fractional diagonal chromatography
two-dimensional difference gel electrophoresis
Dulbecco's modified Eagle's medium
GrM with S195A mutation in catalytic centre
inhibitor of caspase-activated DNase
mGrM with S195A mutation in catalytic centre
- NK cell
natural killer cell
positional scanning synthetic combinatorial library
reactive oxygen species
stable isotope labelling by amino acids in cell culture
Stefanie de Poot and Marijn Westgeest participated in the design of the study, performed experiments, analysed and interpreted the data, and wrote the paper. Daniel Hostetter and Petra Van Damme participated in the design of the study, performed experiments, analysed and interpreted the data. Kim Plasman, Kimberly Demeyer and Roel Broekhuizen performed experiments. Kris Gevaert and Charles Craik participated in the design of the study, and analysed and interpreted the data. Niels Bovenschen participated in the design of the study, analysed and interpreted the data, and wrote the paper.
We thank Dr C.E. Hack for critical reading of this manuscript before submission and Dr E.L. Schneider for helpful discussions. The recombinant granzymes used in the N- and C-terminal COFRADIC analyses were kindly provided by Dr P.I. Bird (Department of Biochemistry and Molecular Biology, Monash University, Melbourne, Victoria, Australia).
P. V. D. is a Postdoctoral Fellow of the Research Foundation-Flanders (FWO-Vlaanderen). K. P. is supported by a Ph.D. grant of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen). This work was supported by the Fund for Scientific Research-Flanders (Belgium) [grant numbers G.0077.06 and G.0042.07 (to K.G.)], the Concerted Research Actions [grant number BOF07/GOA/012 (to K. G.)] from Ghent University and the Interuniversity Attraction Poles [grant number IUAP06], the Dutch Cancer Society (KWF) [grant number UU-2009–4302 (to N. B.)] and the Netherlands Organization for Scientific Research (NWO) [grant number 916.66.044 (to N. B.)].
These authors contributed equally to this work.
Identified tandem MS (MS/MS) spectra are made publicly available in the Proteomics Identifications Database (PRIDE)  under the accession code 15475.