Resistance of cytolytic lymphocytes to perforin-mediated killing. Murine cytotoxic T lymphocytes and human natural killer cells do not contain functional soluble homologous restriction factor or other specific soluble protective factors

CTL and NK cells produce a cytolytic pore-forming protein (perforin, cytolysin) localized in their cytoplasmic granules. These cytotoxic cells are resistant to killing mediated by other lymphocytes and by purified perforin. A membrane factor, known as homologous restriction factor (HRF), has been suggested to confer protection to different cell types against both C- and perforin-mediated lysis. The granules of human large granular lymphocytes have been reported to contain, in addition to perforin, a soluble HRF activity that can be eluted from anion-exchange columns at 115 mM NaCl. Here, we report that a soluble HRF activity is absent in the granules or the cytosol of murine CTL and human NK cells. Our data indicate that the inhibition attributed to HRF could be explained by exogenous EDTA added during granule fractionation. EDTA was shown to bind to Mono Q and to elute at 90 to 120 mM NaCl. A second perforin-inhibitory activity was also eluted from such a column. However, it was present in preparations obtained not only from CTL and NK cells, but also from some perforin-susceptible tumor cell lines, indicating that it has nonrestricted distribution and suggesting that it is probably irrelevant to the perforin-protection mechanism. Our results argue against a role for soluble granule HRF or other soluble factors in mediating resistance of cytotoxic lymphocytes against perforin-mediated lysis.     

Localization of Fas ligand in cytoplasmic granules of CD8+ cytotoxic T lymphocytes and natural killer cells: participation of Fas ligand in granule exocytosis model of cytotoxicity

Fas ligand (FasL) has been implicated in cytotoxic T lymphocyte (CTL)- and natural killer (NK) cell-mediated cytotoxicity. In the present study, we investigated the localization of FasL in murine CTL and NK cells. Immunocytochemical staining showed that FasL was stored in cytoplasmic granules of CD8+ CTL clones and in vivo activated CTL and NK cells, where perforin and granzyme A also resided. Immunoelectron microscopy revealed that FasL was localized on outer membrane of the cytoplasmic granules, while perforin was localized in internal vesicles. Western blot analysis showed that the membrane-type FasL of 40 kDa was stored in CD8+ CTL clones but not in CD4+ CTL clones. By utilizing a granule exocytosis inhibitor (TN16), we demonstrated that FasL translocated onto cell surface upon degranulation of anti-CD3-stimulated CD8+ CTL clones. Moreover, TN16 markedly inhibited the FasL-mediated cytotoxicity by CD8+ T cell clones and NK cells. These results suggested a substantial contribution of FasL to granule exocytosis-mediated target cell lysis by CD8+ CTL and NK cells.     

Granulysin delivered by cytotoxic cells damages endoplasmic reticulum and activates caspase-7 in target cells

Granulysin is a human cytolytic molecule present in cytotoxic granules with perforin and granzymes. Recombinant 9-kDa granulysin kills a variety of microbes, including bacteria, yeast, fungi, and parasites, and induces apoptosis in tumor cells by causing intracellular calcium overload, mitochondrial damage, and activation of downstream caspases. Reasoning that granulysin delivered by cytotoxic cells may work in concert with other molecules, we crossed granulysin transgenic (GNLY(+/-)) mice onto perforin (perf)- or granzyme B (gzmb)-deficient mice to examine granulysin-mediated killing in a more physiologic whole-cell system. Splenocytes from these animals were activated in vitro with IL-15 to generate cytolytic T cells and NK cells. Cytotoxic cells expressing granulysin require perforin, but not granzyme B, to cause apoptosis of targets. Whereas granzyme B induces mitochondrial damage and activates caspases-3 and -9 in targets, cytotoxic cell-delivered granulysin induces endoplasmic reticulum stress and activates caspase-7 with no effect on mitochondria or caspases-3 and -9. In addition, recombinant granulysin and cell-delivered granulysin activate distinct apoptotic pathways in target cells. These findings suggest that cytotoxic cells have evolved multiple nonredundant cell death pathways, enabling host defense to counteract escape mechanisms employed by pathogens or tumor cells.     

The inflammasome: in memory of Dr. Jurg Tschopp

A decade ago, Jurg Tschopp introduced the concept of the inflammasome. This exciting discovery of a macromolecular complex that senses 'danger' and initiates the inflammatory response contributed to a renaissance in the fields of innate immunity and cell death. Jurg led the biochemical characterization of the inflammasome complex and demonstrated that spontaneous hyperactivation of this interleukin (IL)-1β processing machinery is the molecular basis of a spectrum of hereditary periodic fever syndromes, caused by mutated forms of the inflammasome scaffolding receptor, NLRP3. The identification of the underlying mechanism in these disorders has led to their now successful therapy, with the use of the IL-1 receptor antagonist in the clinic. Jurg's pioneering work has subsequently defined a number of inflammasome agonists ranging from microbial molecules expressed during infection, to triggers of sterile inflammation, most notably gout-associated uric acid crystals, asbestos, silica and nanoparticles. More recently, Jurg introduced the critical new concept of the metabolic inflammasome, which senses metabolic stress and contributes to the onset of the metabolic syndrome associated with obesity and type 2 diabetes. Jurg was an outstanding and skillful biochemist, an elegant and rigorous researcher often far ahead of his peers. He was a truly amiable person, fair, generous and inspiring, and will be most remembered for his infectious enthusiasm. We write this review article on the inflammasome in his honor and dedicate it to his memory.     

Perforin-dependent cytolytic responses in beta2-microglobulin-deficient mice.

The ability of beta2-microglobulin-deficient mice (B6.beta2micro(o)) mice to reject syngeneic and major histocompatability (MHC) class I-deficient tumor grafts was examined with a view to determining residual cytotoxic activities that exist in these mice. In particular, the cytotoxic activities of NK cells and CD8(+) cytotoxic T lymphocytes (CTL) reactive against self-MHC class I were assessed using a variety of gene-targeted mice. The creation of mice doubly deficient for perforin and beta2micro (B6.P(o).beta2micro(o)) enabled the determination that perforin was responsible for the cytotoxic activity of NK cells and CD8(+) CTL reactive against self-MHC class I. Dependence on perforin function was demonstrated for the cytotoxicity of these effectors in vitro and for the ability of these effectors to reject a variety of tumors in vivo.     

Inhibition of the death receptor pathway by cFLIP confers partial engraftment of MHC class I-deficient stem cells and reduces tumor clearance in perforin-deficient mice

NK cells mediate acute rejection of MHC class I-deficient bone marrow cell (BMC) grafts. However, the exact cytotoxic mechanisms of NK cells during acute BMC graft rejection are not well defined. Although the granule exocytosis pathway plays a major role in NK cell-mediated rejection, alternative perforin-independent mechanisms also exist. By analyzing the anti-apoptotic effects of cellular Fas-associated death domain-like IL-1-converting enzyme-inhibitory protein (cFLIP) overexpression, we investigated the possible role of death receptor-induced apoptosis in NK cell-mediated cytotoxicity. In the absence of perforin, we found that cFLIP overexpression reduces lysis of tumor cells by NK cells in vitro and in vivo. In addition, perforin-deficient NK cells were impaired in their ability to acutely reject cFLIP-overexpressing TAP-1 knockout stem cells. These results emphasize the importance of NK cell death receptor-mediated killing during BMC grafts in the absence of perforin.     

Veto activity of activated bone marrow does not require perforin and Fas ligand

Veto cells suppress generation of CD8(+) T cell immune responses in an antigen-specific manner, with specificity dictated by antigens on the veto cell surface. Activated bone marrow (ABM) veto cells belong to the NK cell type lineage and veto by clonally deleting antigen-specific precursor cytotoxic T cell lymphocyte (CTL). In vitro cytotoxicity of ABM depends largely on the perforin/granzyme and Fas/Fas ligand pathways. Utilizing perforin-deficient and functional Fas ligand-deficient gld mice as a source of ABM and functional Fas-deficient lpr mice as a source of precursor CTL, we demonstrate in this study that ABM cells utilize a perforin- and Fas-independent pathway to veto allogeneic cell-mediated cytotoxic responses. We also show that ABM cells mediate perforin- and Fas-independent veto activity even in an 8-h clonal deletion assay. We conclude that ABM veto activity does not require the two primary pathways of cell-mediated death.     

Functional defects of NK cells treated with chloroquine mimic the lytic defects observed in perforin-deficient mice

NK cells are the primary effectors mediating acute rejection of incompatible bone marrow cell grafts. To reduce rejection, we evaluated the ability of chloroquine (CHQ) to prevent perforin-dependent NK cell activity. Perforin is a key cytotoxic component released from the lytic granules of activated NK cells. Generation of functional perforin requires an acidic protease activity that occurs in the secretory, lytic lysosomes. Our hypothesis was that CHQ, a lysosomotropic reagent, would raise the pH of the acidic compartment in which perforin is processed and thereby block perforin maturation and cytotoxicity. We have measured NK cytotoxicity in vivo by clearance of YAC-1 tumor cells from the lungs and by rejection of incompatible bone marrow transplants and in vitro by cytolysis of YAC-1 and Jurkat cells. The engraftment of bone marrow cells was monitored by recolonization of the spleen with hemopoietic cells from transplants of MHC class I-deficient bone marrow cells into lethally irradiated recipient mice. Transplant rejection was compared in two inbred strains of mice: 129, which apparently use perforin-dependent cytotoxicity, and C57BL/6, in which rejection can be perforin-independent. CHQ treatment reduced NK cell activity in 129 mice in which perforin is important for mediating rejection. CHQ affected the fraction of NK cell cytolysis that was Fas independent. In addition, we found that CHQ prevents perforin processing by LAK cells in vitro. These data indicate that CHQ may impair rejection of incompatible bone marrow transplants and other functions mediated by NK and cytotoxic T cells.     

Mouse granzyme A induces a novel death with writhing morphology that is mechanistically distinct from granzyme B-induced apoptosis

Human and mouse granzyme (Gzm)B both induce target cell apoptosis in concert with pore-forming perforin (Pfp); however the mechanisms by which other Gzms induce non-apoptotic death remain controversial and poorly characterised. We used timelapse microscopy to document, quantitatively and in real time, the death of target cells exposed to primary natural killer (NK) cells from mice deficient in key Gzms. We found that in the vast majority of cases, NK cells from wild-type mice induced classic apoptosis. However, NK cells from syngeneic Gzm B-deficient mice induced a novel form of cell death characterised by slower kinetics and a pronounced, writhing, ‘worm-like'' morphology. Dying cells initially contracted but did not undergo membrane blebbing, and annexin-V staining was delayed until the onset of secondary necrosis. As it is different from any cell death process previously reported, we tentatively termed this cell death ‘athetosis''. Two independent lines of evidence showed this alternate form of death was due to Gzm A: first, cell death was revealed in the absence of Gzm B, but was completely lost when the NK cells were deficient in both Gzm A and B; second, the athetotic morphology was precisely reproduced when recombinant mouse Gzm A was delivered by an otherwise innocuous dose of recombinant Pfp. Gzm A-mediated athetosis did not require caspase activation, early mitochondrial disruption or generation of reactive oxygen species, but did require an intact actin cytoskeleton and was abolished by latrunculin B and mycalolide B. This work defines an authentic role for mouse Gzm A in granule-induced cell death by cytotoxic lymphocytes.     

Reduced release and binding of perforin at the immunological synapse underlies the age-related decline in natural killer cell cytotoxicity

Physiological aging is accompanied by a marked reduction in natural killer (NK) cell cytotoxicity (NKCC) at the single cell level, but the underlying mechanisms are unknown. To address this issue, we isolated NK cells from healthy young (≤ 35 years) and old (≤ 60 years) subjects and examined the effect of age on events fundamental to the process of NKCC. Simultaneous assessment of NKCC and NK cell–target cell conjugate formation revealed a marked age‐associated decline in NK cell killing but comparable conjugate formation, indicating a post‐target cell binding defect was responsible for impaired NKCC. Despite a reduction in the proportion of NK cells expressing the activatory receptor NKp46, NK cells from old donors were not hyporesponsive to stimulation, as no age‐associated difference was observed in the expression of the early activation marker CD69 following target cell coculture. Furthermore, intracellular levels of the key cytotoxic effector molecules perforin and granzyme B, and the fusion of secretory lysosomes with the NK cell membrane were also similar between the two groups. However, when we examined the binding of the pore‐forming protein perforin to the surface of its target cell, an event that correlated strongly with target cell lysis, we found the percentage of perforin positive target cells was lower following coculture with NK cells from old subjects. Underlying this reduction in binding was an age‐associated impairment in perforin secretion, which was associated with defective polarization of lytic granules towards the immunological synapse. We propose that reduced perforin secretion underlies the reduction in NKCC that accompanies physiological aging.     

Heterogeneity of channel catfish CTL with respect to target recognition and cytotoxic mechanisms employed

Two types of catfish alloantigen-dependent cytotoxic T cells were cloned from PBL from a fish immunized in vivo and stimulated in vitro with the allogeneic B cell line 3B11. Because these are the first clonal cytotoxic T cell lines derived from an ectothermic vertebrate, studies were undertaken to characterize their recognition and cytotoxic mechanisms. The first type of CTL (group I) shows strict alloantigen specificity, i.e., they specifically kill and proliferate only in response to 3B11 cells. The second type (group II) shows broad allogeneic specificity, i.e., they kill and proliferate in response to several different allogeneic cells in addition to 3B11. "Cold" target-inhibition studies suggest that group II CTL recognize their targets via a single receptor, because the killing of one allotarget can be inhibited by a different allotarget. Both types of catfish CTL form conjugates with and kill targets by apoptosis. Killing by Ag-specific cytotoxic T cells (group I) was completely inhibited by treatment with EGTA or concanamycin A, and this killing is sensitive to PMSF inhibition, suggesting that killing was mediated exclusively by the secretory perforin/granzyme mechanism. In contrast, killing by the broadly specific T cytotoxic cells (group II) was only partially inhibited by either EGTA or concanamycin A, suggesting that these cells use a cytotoxic mechanism in addition to that involving perforin/granzyme. Consistent with the presumed use of a secretory pathway, both groups of CTL possess putative lytic granules. These results suggest that catfish CTL show heterogeneity with respect to target recognition and cytotoxic mechanisms.     

Treatment of melanoma with 5-fluorouracil or dacarbazine in vitro sensitizes cells to antigen-specific CTL lysis through perforin/granzyme- and Fas-mediated pathways

Several factors may influence sensitivity of melanoma cells to CTL lysis. One is the avidity of the CTL TCR. A second is that certain cytotoxic drugs have been reported to sensitize cancer cells to CTL lysis through Fas-mediated apoptosis. In this study, we examined whether antineoplastic agents 5-fluorouracil (5-FU) and dacarbazine (DTIC) sensitize melanoma cells to lysis of G209 peptide-specific CTL. Our results show that CTL generated from PBMC are HLA-A2 restricted and gp100 specific. Treatment with 5-FU or DTIC sensitized melanoma cells to lysis of G209-specific CTL. Most importantly, 5-FU- or DTIC-treated melanoma cells also became sensitive to low-avidity CTL, which per se are less cytolytic to melanomas. We sought to identify apoptotic pathways mediating this effect. The enhanced cytolysis was mediated through the perforin/granzyme pathway. Although 5-FU up-regulated FasR expression on melanoma cells, sensitization was not blocked by anti-Fas Ab, and the G209-specific CTL was Fas ligand (FasL) negative. However, when G209-specific CTL were stimulated to express FasL, FasL signaling also contributed to enhanced cytolysis. DTIC treatment, which did not increase FasR expression, also sensitized FasL-mediated killing induced by neutralizing anti-Fas Ab. For CD95L-positive G209-specific CTL, the sensitization was primarily mediated through the perforin/granzyme pathway regardless of up-regulation of FasR. The findings demonstrate that cytotoxic drug-mediated sensitization primes both perforin/granzyme and Fas-mediated killing by melanoma-specific CTL. Considering that most of autoreactive antitumor CTL are low avidity, the findings provide experimental basis for understanding cytotoxic and immunologic therapeutic synergy in melanoma.     

In vivo elimination of MHC-I-deficient lymphocytes by activated natural killer cells is independent of granzymes A and B

NK cells kill target cells mainly via exocytosis of granules containing perforin (perf) and granzymes (gzm). In vitro, gzm delivery into the target cell cytosol results in apoptosis, and induction of apoptosis is severely impaired in the absence of gzm A and B. However, their importance for in vivo cytotoxicity by cytotoxic T cells has been questioned. We used an in vivo NK cytotoxicity assay, in which splenocytes from wild-type and β(2)microglobulin-deficient (MHC-I(neg)) mice are co-injected into recipients whose NK cells were activated by virus infection or synthetic Toll-like receptor ligands. Elimination of adoptively transferred MHC-I(neg) splenocytes was unimpaired in the absence of gzmA and gzmB, but dependent on perforin. This target cell rejection was NK cell dependent, since NK cell depletion abrogated it. Furthermore, target cell elimination in vivo was equally rapid in both wild-type and gzmAxB-deficient recipients, with the majority of specific target cells lost from lymphoid tissue within less than one to two hours after transfer. Thus, similar to T cell cytotoxicity, the contribution of gzmA and B to in vivo target cell elimination remains unresolved.     

The cytotoxic lymphocyte protease, granzyme B, targets the cytoskeleton and perturbs microtubule polymerization dynamics

Granzyme B, a serine protease derived from cytotoxic T lymphocyte (CTL) and Natural Killer (NK) cell granules, plays an important role in coordinating apoptosis of CTL and NK target cells. Here, we report that granzyme B targets the cytoskeleton by cleaving and removing the acidic C-terminal tail of alpha-tubulin. Consistent with this, Granzyme B markedly enhanced rates of microtubule polymerization in vitro, most likely by removal of an autoinhibitory domain within the tubulin C terminus. Moreover, delivery of Granzyme B into HeLa target cells promoted dramatic reorganization of the microtubule network in a caspase-independent manner. These data reveal that granzyme B directly attacks a major component of the cell cytoskeleton, which may contribute to the incapacitation of target cells during CTL/NK-mediated killing.     

Differences in target cell DNA fragmentation induced by mouse cytotoxic T lymphocytes and natural killer cells

Fragmentation of YAC-1 target cell DNA during cytolysis mediated by mouse natural killer (NK) cells and cytotoxic T lymphocytes (CTL) was compared. Cleavage of nuclear chromatin was always an extensive and early event in CTL-mediated cytolysis, whereas with NK cell-mediated killing the degree of DNA fragmentation showed an unexpected relationship to the effector:target (E:T) ratio. At low NK:YAC-1 ratios, DNA fragmentation and 51Cr release were equivalent and increased proportionately until a ratio of about 50:1 was reached; at higher ratios, 51Cr release increased as expected but DNA fragmentation decreased dramatically. Comparison of time course data at E:T ratios producing similar rates of 51Cr release showed that the target cell DNA fragmentation observed in NK killing was not nearly as rapid nor as extensive as that observed with CTL effectors. These results suggest that NK cells induce target cell injury via two different mechanisms. One mechanism would involve lysis mediated by cell-to-cell contact, while the other may induce DNA fragmentation via a soluble mediator. In support of this notion, cell-free culture supernatants containing NK cytotoxic factor (NKCF) induced DNA fragmentation in YAC-1 cells. The DNA fragments induced by NK cells and NKCF-containing supernatants consisted of oligonucleosomes indistinguishable from those induced by CTL. The results presented here show distinct differences in target cell DNA fragmentation induced by CTL and NK cells, and suggest that these two effectors use different mechanisms to achieve the same end. CTL seem to induce DNA fragmentation in their targets by direct signaling, whereas NK cells may do so by means of a soluble factor.     

Perforin is activated by a proteolytic cleavage during biosynthesis which reveals a phospholipid-binding C2 domain.

Perforin is a secreted protein synthesized by activated cytotoxic T lymphocytes (CTL) and natural killer (NK) cells. It is a key component of the lytic machinery of these cells, being able to insert into the plasma membrane of targeted cells, forming a pore which leads to their destruction. Here we analyse the synthesis, processing and intracellular transport of perforin in the NK cell line YT. Perforin is synthesized as a 70 kDa inactive precursor which is cleaved at the C-terminus to yield a 60 kDa active form. This proteolytic cleavage occurs in an acidic compartment and can be inhibited by incubation of the cells in ammonium chloride, concanamycin A, leupeptin and E-64. The increased lytic activity of the cleaved form can be demonstrated by killing assays in which cleavage of the pro-piece is inhibited. Epitope mapping reveals that cleavage of the pro-piece occurs at the boundary of a C2 domain, which we show is able to bind phospholipid membranes in a calcium-dependent manner. We propose that removal of the pro-piece, which contains a bulky glycan, allows the C2 domain to interact with phospholipid membranes and initiate perforin pore formation.     

Impaired cytolytic activity in calreticulin-deficient CTLs

Calreticulin is an endoplasmic reticulum-resident chaperone that is stored in the cytotoxic granules of CTLs and NK cells and is released with granzymes and perforin upon recognition of target cells. To investigate the role of calreticulin in CTL-mediated killing, we generated CTL lines from crt(+/+) and crt(-/-) mice expressing a constitutively active form of calcineurin in the heart. Crt(-/-) CTLs showed reduced cytotoxic activity toward allogeneic target cells despite normal production, intracellular localization, and activity of granzymes and despite perforin overexpression. Comparable or higher amounts of granzymes were degranulated by crt(-/-) cells in response to immobilized anti-CD3 Abs, indicating that calreticulin is dispensable for the signal transduction that leads to granule exocytosis. The ability to form conjugates with target cells was affected in the crt(-/-) CTLs, explaining the observed reduction in cytotoxicity. Conjugate formation and cytotoxicity were completely restored by treatments that facilitate recognition and contact with target cells, a prerequisite for degranulation and killing. Therefore, we conclude that calreticulin is dispensable for the cytolytic activity of granzymes and perforin, but it is required for efficient CTL-target cell interaction and for the formation of the death synapse.     

Perforin and Fas act together in the induction of apoptosis, and both are critical in the clearance of lymphocytic choriomeningitis virus infection

In this report we questioned the current view that the two principal cytotoxic pathways, the exocytosis and the Fas ligand (FasL)/Fas-mediated pathway, have largely nonoverlapping biological roles. For this purpose we have analyzed the response of mice that lack Fas as well as granzyme A (gzmA) and gzmB (FasxgzmAxB(-/-)) to infection with lymphocytic choriomeningitis virus (LCMV). We show that FasxgzmAxB(-/-) mice, in contrast to B6, Fas(-/-), and gzmAxB(-/-) mice, do not recover from a primary infection with LCMV, in spite of the expression of comparable numbers of LCMV-immune and gamma interferon-producing cytotoxic T lymphocytes (CTL) in all mouse strains tested. Ex vivo-derived FasxgzmAxB(-/-) CTL lacked nucleolytic activity and expressed reduced cytolytic activity compared to B6 and Fas(-/-) CTL. Furthermore, virus-immune CTL with functional FasL and perforin (gzmAxB(-/-)) are more potent in causing target cell apoptosis in vitro than those expressing FasL alone (perfxgzmAxB(-/-)). This synergistic effect of perforin on Fas-mediated nucleolysis of target cells is indicated by the fact that, compared to perfxgzmAxB(-/-) CTL, gzmAxB(-/-) CTL induced (i) an accelerated decrease in mitochondrial transmembrane potential, (ii) increased generation of reactive oxygen species, and (iii) accelerated phosphatidylserine exposure on plasma membranes. We conclude that perforin does not mediate recovery from LCMV by itself but plays a vital role in both gzmA/B and FasL/Fas-mediated CTL activities, including apoptosis and control of viral infections.     

Subcellular localization of perforin and serine esterase in lymphokine-activated killer cells and cytotoxic T cells by immunogold labeling

CTL, NK cells, and lymphokine-activated killer (LAK) cells are cytolytic lymphocytes known to produce a pore-forming protein, named perforin or cytolysin, that lyses target cells by forming large pores on the plasma membrane of the target cell. Other proteins besides perforin are found in the cytoplasmic granules of effector lymphocytes, and these include a family of serine esterases. Ultrastructural immunogold labeling studies with antibodies against perforin and a serine esterase (MTSP-1, also known as granzyme A and SE-1) show that all the granules of LAK cells and a CTL cell line contain perforin and serine esterase. For both LAK cells and CTL, perforin has been located mostly in the fine granular matrix of the granules, whereas gold particles corresponding to serine esterase have been found in both the matrix and the cap regions of the granules. Results from double immunogold labeling indicate that perforin and serine esterase colocalize to the same granules.     

In vivo expression of perforin by CD8+ lymphocytes in autoimmune disease. Studies on spontaneous and adoptively transferred diabetes in nonobese diabetic mice

A potent cytolytic pore-forming protein (perforin or cytolysin) has previously been found to be associated with the cytoplasmic granules of CTL and NK cells. Inasmuch as all previous studies on perforin have been conducted with cultured CTL and NK cell lines, it is not clear whether perforin may play a role in the cytotoxicity mediated by CTL that have been primed in vivo. In this study, we investigated the presence of perforin in pancreata from nonobese diabetic (NOD) mice, which have been studied as a model of autoimmune, insulin-dependent (type I) diabetes mellitus. Whereas adult NOD mice spontaneously develop diabetes, it is possible to induce diabetes in young, irradiated NOD mice by adoptive transfer of splenocytes obtained from diabetic donors. By means of immunohistochemical analysis, we were able to detect perforin Ag in a small subpopulation of CD8+/Thy-1+/asialo GM1-/CD4- lymphocytes in the pancreatic islets of animals undergoing both spontaneous and adoptive transfer-mediated insulitis. Perforin+/CD8+ lymphocytes were found in small clusters and were observed to display the morphology of large granular lymphocytes. These observations show for the first time the presence of perforin-containing CD8+ lymphocytes in tissues of animals undergoing autoimmune disease.