contribute to tumor immune surveillance by lysing tumor cells (Kishi et al., 2002; Sekiya et al., 2002; Stenger et al., 1998; Wang et al., 2000). The mechanism of granulysin action is unclear, but it is believed to mediate its effect by association with negatively charged lipids after granule exocytosis (Kaspar et al., 2001; Krensky, 2000). The crystal structure of granulysin shows a multihelical structure, and the association of granulysin with the plasma membrane predicts a scissoring motion, tunneling granulysin into the plasma membrane and resulting in membrane tearing (Anderson et al., 2003). Granulysin-induced membrane damage leads directly to the activation of apoptotic machinery within the cell through intracellular calcium flux and subsequent mitochondrial damage (Kaspar et al., 2001; Okada et al., 2003). Damaged mitochondria release cytochrome c and apoptosis-inducing factor, resulting in the activation of caspases and endonucleases (Kaspar et al., 2001; Pardo et al., 2001). It has been proposed that although granulysin can function independently on extracellular pathogens, it requires perforin to kill cells that harbor an intracellular pathogen (Dieli et al., 2001; Walch et al., 2007). In addition to its lytic role, granulysin can also function by recruiting immune cells to a site of inflammation (Deng et al., 2005).

3.3. Granzymes

Apart from perforin, granzymes represent the most abundant constituents of CL granules, and their role in cytotoxicity was proposed after certain protease inhibitors were shown to abrogate cytotoxicity in vitro (Chang and Eisen, 1980; Masson and Tschopp, 1987; Redelman and Hudig, 1980). Granzyme involvement in the apoptotic pathway of dying cells was subsequently confirmed when proteins corresponding to grA and grB were isolated from rat NK cells and displayed DNAfragmenting ability (Shi et al., 1992).

3.3.1. GrB-mediated apoptosis

In the early 1990s, caspases were emerging as proteases that orchestrated cell death by apoptosis. Caspases are present in the cytoplasm of cells; however, they must be activated by cleavage after specific aspartic acid residues. Because grB has aspase activity, it was postulated that grB may cleave and activate procaspases. Indeed, using cytosolic extracts, several groups demonstrated that grB efficiently cleaved caspase-3, -7, and -8 (Fernandes-Alnemri et al., 1996; Martin et al., 1996). Caspases also cleave their substrates after specific aspartic acid residues, permitting them to autoactivate.

GrB-mediated processing of caspase-3 was partly inhibited by the addition of the caspase inhibitor zVADfmk, suggesting that this event occurred by a two-step process in which grB was only required for the first step (Darmon et al., 1995; Sutton et al., 2000). Subsequently, it was shown that although grB can initiate caspase activation in intact cells, even high concentrations of grB could not fully process pro-caspases on its own (Barry et al., 2000; Sedelies et al., 2008; Sutton et al., 2000; Sutton et al., 2003; Waterhouse et al., 2006a).

Importantly, caspase inhibitors could block the nuclear damage associated with apoptosis, but they could not block grB-induced cell death (Sarin et al., 1998; Trapani et al., 1998a). In contrast, over-expression of Bcl-2 in the target cell did block cell death mediated, in particular, by human grB, leading to clonogenic survival (Davis et al., 2000; Heibein et al., 2000; Sutton et al., 2000; Sutton et al., 1997). The proapoptotic BH3-only Bcl-2 family member Bid is an excellent substrate for human grB, and the cleaved product, truncated Bid (tBid), translocates to the mitochondrial outer membrane, where it can interact with Bcl-2 to release its hold on the proapoptotic Bax and Bak proteins (Alimonti et al., 2001; Heibein et al., 2000; Sutton et al., 2000). Bid’s involvement in cell death mediated by human grB is crucial, as Bid-deficient cells were resistant to grB-mediated apoptosis and continued to proliferate in long-term survival assays (Waterhouse et al., 2005). After Bid cleavage, grB-mediated mitochondrial outer membrane permeabilization results in the release of proapoptotic proteins such as cytochrome c, Smac/DIABLO, and Htra2/omi (Alimonti et al., 2001; Barry et al., 2000; Sutton et al., 2000; Sutton et al., 2003). GrB has also been shown to cleave antiapoptotic Bcl2 family member Mcl-1, resulting in the release of proapoptotic Bim and subsequent cell cytotoxicity (Han et al., 2005) by a similar mechanism to that of Bid.

GrB has also been proposed to directly cleave various additional substrates that influence cell death, including inhibitor of caspase-activated DNAse (ICAD), and ICADdeficient murine embryonic fibroblasts were markedly resistant to grB-mediated DNA fragmentation (Cullen et al., 2007; Sharif-Askari et al., 2001; Thomas et al., 2000). Moreover, other nuclear substrates, such as poly(ADP-ribose) polymerase (PARP), DNA-dependent protein kinase, NuMA, and lamin B, were also directly cleaved by grB and may contribute to cell death (Andrade et al., 1998; Froelich et al., 1996a; Zhang et al., 2001a). Recent studies using human grB and primary NK cells have shown that over-expression of Bcl-2 and blocking caspase activity maintains the clonogenic survival

of target cells, suggesting that if human grB does cleave these other substrates, they do overtly result in cell death (Sedelies et al., 2008).

Many studies previously detailed strongly suggested that mitochondrial disruption precedes caspase activation during grB-mediated cell death; however, studies in mouse models suggested that mitochondria were not critical for grB-induced death (Metkar et al., 2003). Recently, the reason for these discrepancies has become clear: Although both require Asp at the P1 substrate position, human and mouse grB show differences in their substrate specificities, with human grB preferentially cleaving Bid and mouse grB being relatively more efficient at cleaving pro-caspase 3 directly (CasciolaRosen et al., 2007; Cullen et al., 2007; Kaiserman et al., 2006). Although these distinct differences exist at a biochemical level, similar to the human system (Sedelies et al., 2008), mouse grB delivered by CTL efficiently triggered both cytochrome c release and mitochondriaindependent activation of caspases (Pardo et al., 2008). In contrast to the human system, blocking mitochondrial damage and caspase activity did not protect target cells from death induced by mouse CTsL, suggesting that mouse grB may also target additional deathinducing substrates. The recent finding that the grB gene is highly polymorphic in wild mice may also provide a rationale for the different substrate specificities of human and mouse grB (Thia and Trapani, 2007). It has been proposed that viral pathogens of different species may have applied different selective pressures so that the grB gene in mice may have evolved to counter potential viral escape (Thia and Trapani, 2007). Thus the cell death pathways mediated by grB of different species may be significantly different.

3.3.2. GrA-mediated cell death

Granzyme A (grA)–induced cell death is entirely dependent on the contribution of the mitochondria and not on caspases (Beresford et al., 1999; Martinvalet et al., 2005). GrA induces loss of mitochondrial inner membrane potential and increased production of reactive oxygen species (ROS); however, the proapoptotic mitochondrial factors that are released in response to grB (cytochrome c, Smac/DIABLO, Htra2/omi) remain sequestered in the mitochondria (Martinvalet et al., 2005). ROS production is believed to then mediate the translocation of the SET complex to the nucleus, consistent with its involvement in DNA repair in response to oxidative stress (Chowdhury et al., 2006; Fan et al., 2003b). GrA has specificity for the DNA repair proteins of the SET complex, namely HMG2, Ape1, and SET (Beresford et al., 2001; Fan

et al., 2002; Fan et al., 2003b). GrA breaks the association between SET and HH23-H1, allowing HH23-H1 to act as the grA-activated DNase and nick DNA (Fan et al., 2003a). After its release into the cytosol, grA translocates to the nucleus, where it can target proteins involved in maintaining chromatin and nuclear envelop stability, such as histones, laminins, PARP, and Ku70 (Jans et al., 1998; Zhang et al., 2001a; Zhang et al., 2001b). Until recently, it was unclear how grA stimulated ROS production; however, a recent study proposes that grA enters the mitochondria and cleaves, NDUFS3, a subunit of complex I in the electron transport chain (Martinvalet et al., 2008). This disrupts complex I driven respiration, resulting in increased ROS production. This, however, leaves at least two unanswered questions: (1) How does grA enter the mitochondria? (2) How does the ROS produced result in caspase-independent cell death, especially because blocking complex I with rotenone triggers cell death by a caspase-dependent mechanism (Li et al., 2003)? It possible that ROS contributes to grA-induced death, but this is unlikely to be the whole story, and other substrates for grA remain to be uncovered. The pathways currently proposed for grAand grB-induced cell death are shown in Figure 10-2.

3.3.3. Orphan granzyme-mediated cell death

Granzymes C, D, E, F, G, H, K, M, and N have been termed the orphan granzymes because their specific substrates and functions are yet to be discovered (Grossman et al., 2003). The development of the grB gene-disrupted mice resulted in the accidental knock-down of additional genes within the locus, in particular grC and grF (Pham et al., 1996). These mice have been interpreted to represent a compound knockout for several granzymes; however, grC mRNA levels were only 10-fold less than wild-type mice, when CTLs from grA–/–grB–/– mice were activated in mixed lymphocyte reactions. It is not clear whether this reduction was sufficient to abrogate the function of grC; however, when grB alone was “knocked down,” the cytotoxic defects were not as pronounced, suggesting a role for these orphan granzymes in cytotoxicity (Revell et al., 2005). Individual knockout models have not yet been generated for all the granzymes but will be needed to understand the individual functions of the orphan granzymes.

Purified granzymes C, K, H, and M have all been shown to be capable of inducing cell death when delivered by perforin in vitro, but their cellular substrates are unclear (Fellows et al., 2007; Johnson et al., 2003; Kelly et al., 2004; MacDonald et al., 1999). Both granzyme C and K can induce death independently of caspases,

possibly through ROS production (Johnson et al., 2003; MacDonald et al., 1999; Zhao et al., 2007a; Zhao et al., 2007b). Granzyme C induces single-stranded DNA nicks, but the DNAse responsible is unknown (Johnson et al., 2003). Granzyme H and K can cleave Bid; however, the kinetics are slow. Granzyme M–induced death is rapid and occurs independently of caspases and mitochondrial perturbation by direct cleavage of nuclear substrates ICAD and PARP (Kelly et al., 2004; Lu et al., 2006). The putative cell death mechanisms activated by the various granzymes are described briefly (Table 10-4).

4. A ROLE FOR GRANULE PROTEINS IN VIRAL RESPONSE,

IMMUNE SURVEILLANCE, AND IMMUNE HOMEOSTASIS

It is imperative for the immune system to mount an attack on virus-infected or transformed cells, a process that is extremely dependent on cytotoxic granule proteins. It is also extremely important that the immune system diminishes lymphocyte populations afterwards to maintain cellular homeostasis. Although this latter process is thought to be primarily dependent on receptormediated death, cytotoxic granule proteins can also play a role, as indicated from studies with gene-disrupted mice and humans.

CL from perforin-deficient mice showed a marked decrease in their cytolytic function against lymphocytic choriomeningitis (LCMV) infected cells in vitro (Kagi et al., 1994). In addition to controlling the spread of viruses, there is evidence that perforin has a role in immune surveillance, that is, the elimination of transformed cells before their presentation as a clinical malignancy (Smyth et al., 2000; van den Broek et al., 1996). Considerable evidence suggests an additional role for perforin in immune regulation. Perforin-deficient mice show increased clonal expansion and persistence of virus-specific T cells and an inability to downregulate T-cell responses during chronic LCMV infection (Kagi et al., 1999; Matloubian et al., 1999). In LCMV infection, high levels of activated CTLs cannot be cleared, and the activated lymphocytes and macrophages infiltrate various organs, resulting in massive release of inflammatory cytokines, tissue necrosis, and organ failure, features very similar to those seen with human patients suffering from FHL (Arico, 1991). Among other causes, FHL can result from the absence of perforin expression within the granules and subsequent defective CL function (Stepp et al., 1999; Voskoboinik et al., 2006). Incomplete loss of perforin function has also been linked to hematological cancer, although these studies involved small numbers of patients (Clementi et al., 2005; Mehta et al., 2006; Voskoboinik et al., 2007). Perforin may also be a

critical mediator of tissue damage, controlling autoimmune beta-cell destruction that results in type 1 diabetes in nonobese diabetic mice (Kagi et al., 1997).

GrAand grB-deficient mice both show some increased mortality when infected with ectromelia virus. However, in contrast with CTLs from grA mice whose apoptotic response is unaltered, delayed nuclear apoptotic changes are evident in target cells when treated with grB-deficient CTL (Ebnet et al., 1995; Heusel et al., 1994). In contrast to mice deficient in a single granzyme, mice that are deficient for both grA and grB are remarkably (about a million-fold) more susceptible to fatal ectromelia infection (Mullbacher et al., 1999). Furthermore, allogeneic CTLs isolated from these mice induce an alternative form of cell death that largely resembles apoptosis morphologically but features the delayed expression of markers of phagocytosis (Waterhouse et al., 2006b).

5. CONCLUSIONS

Many unanswered questions still remain regarding the function of CL secretory granule proteins. As a crucial molecule in the granule exocytosis pathway, it is unclear how perforin functions at the molecular level. Perforin is critical for mediating granzyme entry into the target cell; however, the mechanism for this synergy still remains