Granzyme M Mediates a Novel Form of Perforin-dependent Cell Death*

Cell death is mediated by cytotoxic lymphocytes through various granule serine proteases released with perforin. The unique protease activity, restricted expression, and distinct gene locus of granzyme M suggested this enzyme might have a novel biological function or trigger a novel form of cell death. Herein, we demonstrate that in the presence of perforin, the protease activity of granzyme M rapidly and effectively induces target cell death. In contrast to granzyme B, cell death induced by granzyme M does not feature obvious DNA fragmentation, occurs independently of caspases, caspase activation, and perturbation of mitochondria and is not inhibited by overexpression of Bcl-2. These data raise the likelihood that granzyme M represents a third major and specialized perforin-dependent cell death pathway that plays a significant role in death mediated by NK cells.

Cytotoxic lymphocytes kill tumor or virus-infected target cells by two major pathways (1)(2)(3)(4)(5)(6) that are mediated either by directed release of cytotoxic granules into the synapse between cytotoxic lymphocytes and their targets or by ligation of cell death receptors on target cells. Cytotoxic granules contain several components including perforin and a family of unique serine proteases called granzymes. Perforin released from cytotoxic granules facilitates the entry of granzymes into the target cell, probably through a process of endosomal disruption (7)(8)(9).
GrB has been shown to play a critical role in triggering apoptotic cell death either directly (19) or via the activation of cellular caspases (proteases that orchestrate apoptosis) (20). There is considerable evidence from in vitro studies and GrBdeficient (GrB Ϫ/Ϫ ) mice to support the concept that lymphocyte-mediated oligonucleosomal fragmentation of DNA in tumor cells requires perforin and GrB (21)(22)(23)(24). GrB Ϫ/Ϫ mice also do not express Gr C, D, F, and G that additionally exist in mice (25); however, despite the lack of all of these Gr, the defect in inducing nuclear damage in vitro is kinetic and not absolute (26). GrA and GrB trigger nuclear damage in the presence of perforin (23) and can synergize in order to mediate target cell death (27). Recently, GrA has been demonstrated to induce caspase-independent cell death, cleave lamins (28), and generate single-stranded DNA nicks rather than causing oligonucleosomal fragmentation of DNA ("laddering") (29 -31) by a process that appears to be mediated by a known tumor suppressor, NM23-H1 (32). A recent report has also described the ability of mouse GrC (most closely related to human GrH) to cause cell death characterized by the rapid externalization of phosphatidylserine, nuclear condensation and collapse, and single-stranded DNA nicking (33). The protease specificity of this Gr has not been defined. Collectively, these data suggest that at least three Gr may induce cell death by distinct pathways. Intriguingly, lymphocytes from mice deficient in GrA, GrB, and GrC still kill tumor targets (34), thus suggesting that additional Gr or granule components may contribute to cell death.
GrM is highly expressed in NK cells but is not expressed in activated primary human, rat, or mouse T cells (17,35). Given the unique cellular expression and protease specificity of GrM, we undertook to study the cytotoxic potential of GrM in the presence of perforin. Surprisingly, GrM induced a novel form of perforin-dependent cell death that was independent of caspases and mitochondrial disruption and occurred in the absence of discernable DNA fragmentation.

Cell Lines
Human Jurkat, Jurkat.pgk, and K562 cells were grown in RPMI 1640 supplemented with 10% fetal calf serum and 2 mM glutamine. HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 2 mM glutamine. Jurkat cells expressing human Bcl-2 (Jurkat.Bcl-2) were derived and cultured as described (36).

Granzyme M and Perforin
Recombinant human GrM and the enzymatically inactive mutant, GrM-SA, were produced in baculovirus Sf21 cells and purified as pre-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: Gr, granzyme(s); pfp, perforin; Z-VADfmk, benzyloxycarbonyl-VAD-fluoromethyl ketone; PBS, phosphatebuffered saline; PI, propidium iodide; ROS, reactive oxygen species; DAPI, 4Ј,6-diamidine-2Ј-phenylindole dihydrochloride; TMRE, tetramethyl rhodamine methyl ester.

Cell Death Assays
Cell lines were first titrated with recombinant human perforin to define a sublytic dose (Ͻ15% lysis in a 4-h 51 Cr release assay). That sublytic dose of perforin was then used in combination with one of recombinant GrM, GrM-SA, or GrB (generally to a final concentration of 10 -83 nM). The assays were performed in triplicate with 2 ϫ 10 4 cells (unless otherwise stated) in a 96-well plate. The cells were re-suspended in Hanks' buffered salt solution (JRH Biosciences) supplemented with 20 mM Hepes buffer, pH 7.2, 2 mM CaCl 2 , and 0.04% bovine serum albumin. The Gr and pfp were added in 10 mM Hepes buffer, pH 7.2, 150 mM NaCl (HE buffer).
Clonogenic Assays-These were performed to ultimately determine that target cells were undergoing cell death. After exposure to Gr and pfp, cells were pelleted and resuspended in media. 1200 cells were added to 1.5 ml of molten soft agar (20% fetal calf serum, 0.3% Bacto agar (Difco) in RPMI), and 0.5 ml were plated in triplicate in a 24-well plate. The set agar was overlaid with 1 ml of complete RPMI, and the plate was incubated for 7 days. The number of viable colonies was counted in triplicate. Trypan blue exclusion counts were also performed on a sample of cells from the same assay. The number of dead cells (blue) were expressed as a percentage of the total cell number. 51 Cr and 125 I Release Assays-Assays measuring the release of 51 Crand 125 I-labeled DNA from apoptotic cells were performed as described (40,41). To measure membrane perturbation, target cells were radiolabeled with 51 Cr for 1 h, then washed thoroughly to remove free label and incubated with Gr and pfp in 96-well round bottom plates for various periods of time as indicated. The supernatants were harvested using a Skatron Cell Harvester (Lier, Norway), and the specific release of 51 Cr into the supernatant was measured using a Wallac Wizard 1470 automatic gamma counter (PerkinElmer). To additionally measure 125 I release, target cells were co-labeled with 51 Cr-and 125 I-labeled 2Ј deoxyuridine. After harvesting manually for 51 Cr release, cells were treated with 1% Triton X-100 for 10 min as previously described (41), and cells were centrifuged. For caspase inhibition experiments, radiolabeling and assay was performed in the presence or absence of final concentration of 200 M Z-VAD-fmk or Z-FA-fmk inhibitor. The spontaneous release was determined by incubating target cells with buffer alone. The maximum release was determined by adding SDS to a final concentration of 5%. The percent specific lysis was calculated as follows: 100 ϫ [(experimental release Ϫ spontaneous release)/(maximum release Ϫ spontaneous release)].
Apoptosis-Apoptosis was measured by DNA content and cell surface phosphatidylserine exposure. Target cells were treated with Gr and pfp for 2 h after which cells were washed once in PBS, re-suspended in 200 l of PBS, and fixed with 200 l of ice-cold 100% ethanol and left for 30 min on ice. The cells were re-suspended in propidium iodide (PI) (46 g/ml) with 50 g of RNase. The cells were incubated for 15 min at 37°C and immediately analyzed by flow cytometry. For the analysis of phosphatidylserine on the outer surface of apoptotic cells, target cells were double-stained with annexin V-Fluos (Roche Diagnostics) and PI. Cells were pelleted after incubating with Gr and pfp for 4 h at 37°C, washed, and treated with Annexin V-Fluos on ice for 30 min. After washing, PI was added to a final concentration of 1 g/ml, and the cells were analyzed by flow cytometry using a FACSCalibur (BD Biosciences). Live cells were considered to be PI Ϫ and annexin V-Fluos Ϫ .
Mitochondrial Function-Changes in mitochondrial membrane potential were measured by flow cytometry after treatment with the mitochondrion-specific fluorescent dye, tetramethyl rhodamine methyl ester (Molecular Probes), at a final concentration of 100 nM. A control sample was stained with carbamoyl cyanide n-chlorophenylhydrazone at 10 M to determine the level of fluorescence in target cells with depolarized mitochondria. Determination of reactive oxygen species (ROS) was performed with H 2 DCFDA dye (C-400, Molecular Probes) as described (42).
Morphology-To examine the morphological aspects of cell death, target cells undergoing GrM-and pfp-mediated death were assessed by light and electron microscopy. Jurkat target cells were treated with Gr and pfp for 1 and 4 h. To examine the morphological changes in the nucleus, the nuclear DNA was stained with 4Ј,6-diamidine-2Ј-phenylindole dihydrochloride (DAPI; Roche Diagnostics). Jurkat cells (5 ϫ 10 4 ) were treated with Gr and pfp for 4 h. The cells were washed twice with Hanks' buffered salt solution and once with methanol, and finally the cells were re-suspended in 200 l of 1 g/ml DAPI (in methanol) and incubated at 37°C for 15 min. After spinning, the supernatant was removed, and the cells were washed in methanol and re-suspended in 100 l of PBS plus 100 l paraformaldehyde and left for 20 min in the dark. After washing in PBS the cells were re-suspended in 20 l of PBS, pipetted onto a microscope slide, left to dry, and a drop of mounting solution (glycerol pyrogallol) was added. The DAPI staining was observed on a Zeiss microscope (Model Axioskop 2) under a 40ϫ lens. For transmission electron microscopy cells were fixed in a modified Karnovsky's fixative consisting of 2% paraformaldehyde, 2% glutaraldehyde in 0.08 M Sorensen's phosphate buffer. Post-fixation was in 2% buffered osmium tetroxide and processed for routine electron microscopy. Cells were embedded in Spurrs (ProScitech) resin. Ultrathin sections of 70 -90-nm thickness were cut on a Leica Ultracut S ultramicrotome (Leica Instruments, North Ryde, Australia), stained with uranyl acetate and lead citrate, and examined in a Hitachi H600 transmission electron microscope.
Cytostasis-Jurkat cells (2 ϫ 10 5 ) were treated with Gr and pfp (125 l) for 4 h at 37°C. The cells were spun and re-suspended in 1 ml of RPMI media and incubated in a 24-well plate for 24 h at 37°C in a 5% CO 2 incubator. The cells were treated with PI as described in the apoptosis method and analyzed by flow cytometry to determine whether the cells showed any evidence of cell cycle arrest.

Cytochrome c Release
Cells (6 ϫ 10 4 ) were resuspended in 100 l of plasma membrane lysis buffer (50 g/ml digitonin, 120 mM KCl in PBS) and incubated on ice for 5 min (43). Samples were tested by trypan blue to ensure that Ͼ95% of cells were permeabilized. Cells were fixed by adding 100 l of paraformaldehyde (4% in PBS), pelleted (800 ϫ g for 5 min in V-bottom 96-well plates), and incubated in paraformaldehyde (4% in PBS) for 20 min. The cells were washed three times in blocking buffer (3% bovine serum albumin, 0.05% saponin in PBS) and incubated overnight at 4°C with anti-cytochrome c (clone 6H2.B4 BD, PharMingen) diluted 1:200 in blocking buffer. Cells were washed three times in blocking buffer and incubated for 1 h at room temperature in phycoerythrin (PE) antimouse secondary antibody (Silenus, Melbourne, Australia) diluted 1:200 in blocking buffer. The cells were washed 3ϫ in PBS and analyzed by flow cytometry. PE fluorescence was detected in FL2.

Sheep Red Blood Cell Lysis
Recombinant pfp was added to a 96-well plate and serially diluted in Hepes/saline buffer (20 mM Hepes/150 mM NaCl). Twenty million sheep red blood cells were added to each well followed by GrM (final 60 nM), and lysis was initiated by adding CaCl 2 to a final concentration of 1 mM.
The plate was incubated at 37°C for 30 min. After centrifugation at 1400 ϫ g for 4 min, the supernatant was transferred to a fresh plate, and hemoglobin release was evaluated as the absorbance measured at 405 nm. Maximal hemoglobin release was defined as that occurring after re-suspending the sheep red blood cell in double distilled water.

Western Blot Analysis
Western blotting was performed essentially as described (44). Jurkat.pgk cells (1 ϫ 10 6 ) were loaded with pfp and Gr (60 nM) for 2 h at 37°C. The cells were washed twice with PBS and lysed on ice for 30 min in Nonidet P-40 lysis buffer (0.5% Nonidet P-40, 5 mM MgCl 2 , 25 mM KCL, 10 mM Tris-HCl, pH 8.0) containing complete-mini protease inhibitor mixture (Roche Applied Science). Cell debris was removed by centrifugation at 13,000 rpm for 5 min, and the supernatant was retained. Equal amounts of protein (determined by means of a Bradford reaction) were separated on 12.5 or 15% SDS-polyacrylamide gels and electroblotted on to nitrocellulose transfer membranes. Immunoblots were probed with mouse monoclonal antibody C31720 to human caspase-3 (Transduction Laboratories, Lexington, KY), mouse monoclonal antibody clone 10-1-87 to human caspase-9 (a kind gift from Dr. Y. Lazebnik, Cold Spring Harbor Laboratory, NY), mouse monoclonal antibody to human Bid (Junying Yuan, Harvard Medical School, Boston, MA), or mouse monoclonal antibody to human ␣-tubulin (Sigma) and visualized by enhanced chemiluminescence (Amersham Biosciences).

Caspase Assay
Caspase-3 activity in cell lysates was measured using a caspase assay. Jurkat.pgk cells (2 ϫ 10 6 ), treated with pfp and Gr for 2 h at 37°C, were harvested, washed twice with PBS, and lysed in ice-cold Nonidet P-40 lysis buffer without protease inhibitors. An equal amount of protein for each sample was incubated with or without 1.25 mM Ac-Asp-Glu-Val-Asp-p-nitroanilide substrate (Bachem, Bubendorf, Switzerland) in buffer, pH 7.3, containing 0.1 M Hepes, 0.05 M CaCl 2 at 37°C. The absorbance at 405 nm was read at 15-min intervals for 2 h, and the substrate cleavage was measured by the change in absorbance in the presence or absence of the caspase-3 substrate.

GrM Induces Cytolysis and Death of Target Cells-Cell-free
assays have been extremely useful in the analysis of granule serine proteases that induce cell death in the presence of pfp (23,45). Thus far GrA, GrB, GrC, and GrK have been shown by a variety of assays to trigger cell death in the presence of pfp (7,23,33). Initially, we compared the cytolytic activity of human GrM and GrB in the presence of sublytic (Ͻ15% lysis) concentrations of pfp against Jurkat target cells in a 4-h 51 Cr release assay (Fig. 1A). Clearly, increasing concentrations of GrM (10 -83 nM) caused 51 Cr release in the presence of pfp. This level of cytolytic activity was comparable with similar concentrations of GrB in the presence of pfp (Fig. 1A). By contrast, the highest concentration of GrM alone (83 nM) displayed only background levels of cell death (Fig. 1A) as did the highest concentration of active site serine-mutated GrM (GrM-SA, 83 nM) in combination with pfp (Fig. 1A). Similar results were obtained when loss of viability was assessed by trypan blue uptake (data not shown). Given the rapid cell death mediated by NK cells, we next assessed the kinetics of cytolysis by incubating Gr and pfp for various periods of time before harvest (Fig. 1B). Surprisingly, Jurkat target cells exposed to GrM and pfp displayed measurable 51 Cr release within 20 min, and by 1 h cell death had approximated maximal levels. By contrast, GrB and pfp did not induce detectable cell death until after 1 h of exposure, and death had reached a maximum after ϳ4 h (Fig. 1B). Similar levels and kinetics of cell death were also observed for other human target cells (HeLa, K562) exposed to GrM and pfp ( Fig. 1C and data not shown). Because clonogenic assays provide a reliable and definitive method of measuring cell death (4, 46), we exposed Jurkat target cells to GrM or GrB and pfp for 4 h and assessed the number of colonies grown that arose in soft agar 7 days later (Fig. 1D). Both GrM and GrB reduced colony number by more than 50% in the presence of pfp, whereas pfp alone or in the presence of GrM-SA was without effect.  51 Cr and 125 I release was measured as described. Results were recorded as the mean Ϯ S.E. of triplicate samples and were representative of two experiments performed. B, whole cell lysates treated as above for 2 h in the absence of Z-VAD-fmk were assessed for cleavage of bid, caspase-3, and caspase-9 as indicated. ␣-Tubulin expression is included as a loading control. C, capsase-3 activity in the lysates of Jurkat cells exposed to pfp and GrM or GrB for 4 h was measured by cleavage of the chromogenic peptide substrate Ac-Asp-Glu-Val-Asp-p-nitroanilide.

GrM Does Not Enhance pfp Lytic Function and Mediates
suggest that serine proteases may be required for optimal pfp function (49). Therefore, initially we examined the ability of GrM to simply potentiate the lytic function of pfp against red blood cell membranes. Over a broad range of pfp concentrations we demonstrated that neither GrM nor GrB was able to simply enhance the lytic activity of pfp ( Fig. 2A).
To further characterize the death of nucleated Jurkat cells exposed to pfp and GrM we next analyzed their scatter properties and DNA content. Consistent with previous findings (45), Jurkat cells exposed to both pfp and GrB for 2 h developed apoptotic changes, including DNA fragmentation (87% sub-G 1 ), reduced size (reduced forward scatter), and increased granularity (increased side scatter) (Fig. 2B). By contrast, no such changes in propidium iodide staining, forward scatter, or side scatter were observed in target cells incubated in the presence of pfp or pfp and GrM in combination. GrM/pfp also failed to cause any cytostatic effect after 4 h of exposure as determined by assessing the cell cycle after 24 h by PI staining (data not shown). Staining with the vital dye PI and Fluos-coupled annexin V, which detects externalized phosphatidylserine, is a further hallmark of apoptotic cells. It has been recognized that cells undergoing apoptosis transit from an annexin V ϩ /PI Ϫ stage to terminate as annexin V ϩ /PI ϩ . After 2 h of exposure we detected an equivalent number of live cells (annexin V Ϫ PI Ϫ ) in cultures incubated with GrB/pfp and GrM/pfp (46 and 53%, respectively) compared with control cultures (Fig. 2C). Strikingly, however, Jurkat cells exposed to GrB/pfp were 26% single annexin V ϩ and 27% annexin V ϩ PI ϩ , compared with 8% single annexin V ϩ and 36% annexin V ϩ PI ϩ for GrM/pfp-exposed cultures (Fig. 2C). These data indicated that GrM/pfp killed Jurkat target cells but did not induce a detectable single annexin V ϩ stage of cell death. A similar lack of transition of Jurkat cells through a single positive stage was also observed in cells exposed to GrM/pfp for shorter periods (15 min to 1 h) of time (data not shown).
GrM Does Not Activate the Mitochondria-dependent Apoptotic Pathway-Mitochondria have been shown to play a role in GrB-induced cell death. In this instance Bcl-2 family proteins regulate the release of proteins from the mitochondrial intermembrane space (e.g. cytochrome c and smac), which then participate in the activation of caspases (50). Therefore, we investigated whether Bcl-2 could inhibit GrM/pfp-mediated lysis in Jurkat cells. Consistent with previous reports (40), Bcl-2 inhibited GrB/pfp-mediated DNA fragmentation and lysis (Fig.  3A). By contrast, GrM/pfp was equally effective in killing Jurkat-Bcl-2 and parental Jurkat target cells (Fig. 3A). To inves-tigate whether mitochondrial intermembrane space proteins were released during GrM-induced cell death, we followed cytochrome c release by immunocytochemistry. For immunocytochemistry, the pellets of digitonin-treated samples were stained for the presence of cytochrome c and analyzed by flow cytometry (43). In this assay the percentage of cells that have maintained cytochrome c within the mitochondria can be quantitated. We found that after 2 h, 31% of GrB/pfp-treated cells had released cytochrome c (Fig. 3B). At this point, when GrM/ pfp had caused significant 51 Cr release, there was no evidence of cytochrome c release (Figs. 1B and 3B). To determine whether GrM/pfp induced changes in mitochondrial membrane potential, release of the mitochondrion-specific fluorescent dye, TMRE, was measured by flow cytometry (Fig. 3C). Mitochondrial transmembrane potential was maintained in the majority of untreated Jurkat cells or those exposed to pfp alone, GrM/ pfp, or GrM-SA/pfp. By contrast, about half the Jurkat cells (55%) exposed to GrB/pfp demonstrated loss of mitochondrial transmembrane potential. Totally depolarized mitochondria were represented by Jurkat cells treated with carbamoyl cyanide n-chlorophenylhydrazone (99%). Because the production of ROS has also been implicated in cell death, we used H 2 DCFDA (C-400) dye to determine whether ROS were generated in Jurkat cells exposed to GrM/pfp. A detectable level of ROS was only observed in cells exposed to GrB/pfp (85%) but not GrM/pfp (data not shown).
GrM-induced Cell Death Is Caspase-independent-We next determined whether caspase activity was necessary for GrMmediated cell death by comparing cell death (at 1 or 4 h) in the presence of the polycaspase inhibitor Z-VAD-fmk or a Z-FAfmk control. Despite the ability of Z-VAD-fmk to specifically inhibit GrB/pfp-mediated DNA fragmentation (by Ͼ85%) and lysis (by Ͼ50% in most experiments) of Jurkat cells, Z-VADfmk was without effect upon GrM/pfp-mediated cell death (Fig.  4A). Clearly, GrM/pfp did not induce discernable 125 I-labeled 2Ј deoxyuridine release (DNA fragmentation) at either time point (Fig. 4A). These data were supported by the inability of GrM/ pfp to induce the specific cleavage of Bid, pro-caspase 3, and procaspase-9 (Fig. 4B), all hallmark features of apoptosis induced by GrB/pfp (Fig. 4B) (45). Furthermore, caspase-3 activity was detected in lysates of Jurkat cells exposed to GrB/pfp but not GrM/pfp (Fig. 4C).
Morphological Characteristics of Cells Undergoing GrM-induced Cell Death-To understand the nature of GrM-induced death, we investigated nuclear morphology of the dying cells using the DNA-specific dye DAPI (Fig. 5, A-C). The nuclei of cells treated with pfp alone (Fig. 5A) were uniformly stained. In contrast, the nuclei of GrB/pfp-treated cells (Fig. 5B) were segmented, a feature that is typical of the chromatin condensation observed during apoptosis. Staining indicative of chromatin condensation was also observed in cells treated with GrM/pfp. In the panel shown (Fig. 5C) three distinct and representative patterns of staining are shown. In cell 1, the chromatin appears only slightly fragmented. In cell 2, the chromatin is more typical of apoptosis, and in cell 3 the nucleus appears extremely shrunken but not typically segmented. To investigate the morphology further, we performed electron microscopy analysis (Fig. 5, D-F). The morphology of pfp-treated cells (Fig. 5D) displayed no distinct features of cell death. In GrB/pfp-treated cells (Fig. 5E), the chromatin was condensed against the nuclear membrane. In GrM/pfp-treated cells the chromatin was generally condensed but was not as strikingly associated with the nuclear membrane (Fig. 5F). Evidence of chromatin condensation is also shown in micrographs of GrM/ pfp-treated cells (Fig. 6, A-C). In Fig. 6B, the chromatin appeared more condensed than in other GrM/pfp-treated cells  Fig. 5C). The appearance of the condensed chromatin in the micrographs of GrM/pfp-treated cells was, therefore, consistent with our observations in the DAPI-stained samples. The cytoplasm of GrM/pfp-treated cells (Fig. 6, A-C) also frequently appeared more electron-dense (Fig. 6, A-C) and had dilated endoplasmic reticula (Fig. 6, A and C, arrows). Large vacuoles were prevalent in GrM/pfptreated cells (Fig. 6, A and B). There was no evidence of swelling of mitochondria in GrM/pfp-treated cells; however, some appeared rounded (Fig. 6E) compared with cells treated with GrB/pfp (Fig. 6F). It is unclear from the micrographs whether these features are early or late features of death; however, the appearance of cells that were quite obviously lysed suggests that in culture these cells will eventually succumb to a secondary necrosis (Fig. 6D). DISCUSSION The unique protease specificity of GrM among granzyme family members suggested this NK cell-associated protease might have a novel biological function or trigger a novel form of cell death. To our knowledge this is the first report of the ability of GrM to cause cell death. In the presence of pfp, the protease activity of GrM was required to display an effective capability of inducing cell death in target cells derived from tissues of different origins (T, erythromyeloid, and epithelial). Equimolar amounts of GrB and GrM were seen to induce an equivalent amount of cell death (measured by 51 Cr release within 4 h). The kinetics of lysis induced by GrM appeared rapid if not faster than that observed for GrB. However, most importantly, the hallmarks of death induced by GrM were quite distinct from GrB and that described for other Gr. In contrast to GrA (11,29,51) and GrC (33), cell death induced by GrM did not feature early mitochondrial perturbation or early cytochrome c release and was not inhibited by overexpression of Bcl-2. In addition, unlike GrB (36) (shown herein), cell death induced by GrM did not trigger obvious DNA fragmentation and occurred independently of caspase activation. Cell death induced by intact cyto-toxic lymphocytes is not blocked by caspase inhibitors and overexpression of the anti-apoptotic protein, Bcl-2 (19,36). Our studies add weight to recent studies which have demonstrated that granule components such as GrA (29), GrC (33), and granulysin (52) can all trigger caspase-independent cell death pathways in their own right. In addition to inducing a caspaseindependent death pathway, GrM also bypasses inhibition of Bcl-2 and appears to fail to trigger hallmark mitochondrial changes (membrane depolarization, increased ROS, cytochrome c release, etc.) that are observed in response to other Gr and granulysin. Collectively, these data suggest that GrM triggers a pathway to cell death that is molecularly distinct to that induced by other Gr family members and raise the likelihood that GrM represents a third major and specialized perforin-dependent cell death pathway that plays a different role in death mediated by NK cells. Clearly, NK cell cytotoxic granules may co-express Gr that can trigger caspase-dependent (GrB) and caspase-independent (GrA and GrM) pathways. However, the diversity of Gr specificities may be an essential mechanism by which cytotoxic lymphocytes can overcome inhibitors of caspases and other mediators of apoptosis that various intracellular pathogens synthesize. It remains to be determined what target proteins are cleaved by GrM.
A functional role for GrM in cytotoxic lymphocyte-mediated cell death has not yet been established. Indeed, although GrA and GrB have been unequivocally shown to be involved in cytotoxic T lymphocyte-mediated apoptosis, the importance of these pathways remains to be illustrated in a biological context. Cytotoxic lymphocytes from mice deficient in GrA and GrB and associated cluster Gr (25) still retain cytolytic function and can reject many experimental tumors (34, 53); however, it remains possible that other Gr, like GrM, may compensate, and therefore, analysis of killer cells from mice additionally lacking GrM will be of interest. Thus far, expression of GrM protein has been found to be restricted to NK cells, ␥␦ ϩ T cells, and some CD56 ϩ T cells, and thus, this expression pattern suggests a role for GrM in innate immunity. In concert with GrM being evolutionarily conserved for a particular role in innate immunity, a GrM-like gene was recently identified in pathogen-challenged fish (54). The morphology of cells undergoing GrM-induced cell death (electron dense cytoplasm, large cytosolic vacuoles, nongeometric chromatin condensation) was quite distinct from those dying in response to GrB. Indeed, other caspase-independent forms of programmed cell death described include autophagy, characterized by the formation of large lysosomalderived cytosolic vacuoles (55,56), and dark cell death (57). There are many forms of programmed cell death in which the chromatin condenses into less geometric shapes; however, phagocytosis markers on the plasma membrane are usually displayed before cell lysis (58). At no stage were GrM/pfptreated cells annexin V ϩ PI Ϫ , suggesting that target cells exposed to GrM/pfp may be releasing some cytosolic contents but not displaying an "eat me" indicator to neighboring phagocytic cells. This raises the provocative possibility that GrM-induced cell death plays a role in stimulating an early inflammatory response. Future efforts to characterize the molecular targets and biological relevance of the GrM-mediated cell death pathway will be of great interest.