Perforin/Fas-Ligand Double Deficiency Is Associated with Macrophage Expansion and Severe Pancreatitis¹

Cytotoxic effector mechanisms utilized by CTL and NK cells consist of rapid pathways, mediated by perforin/granzyme (1, 2) and Fas-L³/Fas (3), and slow pathways mediated by death receptors such as TNF and lymphotoxin (4, 5). Fas and Fas-L deficiency (6, 7) is characterized by autoimmune disease, lymphadenopathy, and lymphoproliferation, and expansion of CD4, CD8 double-negative T cells expressing B220 (8), suggesting a critical role for Fas and Fas-L in lymphocyte homeostasis (9). Perforin deficiency is accompanied by a decreased ability to clear viral infections (10, 11, 12) and to reject tumors, and by diminished tumor surveillance (13).

Fas and Fas-L have been shown to be important molecular effectors for the homeostatic control of the lymphoid system, including B cells (for review, see 14 , T cells (for reviews, see Refs. 15–17), and dendritic cells, and may also function as regulators of APCs (18). CD4⁺Th1 cells lyse normal but not Fas-deficient macrophages in an Ag-specific MHC-restricted reaction (19). A role for Fas and Fas-L in the control of tissue macrophages has been observed in the pathogenesis of arteritis in the MRL mouse (20). APCs are critical elements for T cell activation and for the generation of Ag-specific, MHC-restricted effector T cells of both the CD4⁺ and CD8⁺ phenotype. Activated Ag-specific CD4⁺ and CD8⁺ T cells can express Fas-L and other death receptors in addition to cytolytic granules containing perforin and granzymes (21, 22, 23, 24, 25). APCs expressing the appropriate Ag and MHC molecule are potential target cells for activated T cells and may be lysed by previously activated T cells through the perforin/granzyme or Fas-L/Fas pathway. Effector T cells therefore are capable of turning off the original antigenic stimulus by lysing APCs. This negative feedback control of Ag presentation may be necessary to limit T cell activation and avoid tissue destruction.

In this study, we ablated the two rapidly acting cytotoxic pathways, perforin and Fas-L, by cross-breeding of perforin-deficient with Fas-L-deficient mice. Double-deficient mice develop a spontaneous syndrome accompanied by tissue destruction due to infiltrating T cells and monocytes/macrophages, weight loss, and early death. We present evidence that the failure of activated T cells to lyse Ag-presenting monocytes/macrophages may be responsible for this syndrome. Cytotoxicity mediated by perforin thus may contribute to limiting T cell activation, in addition to controlling intracellular pathogens.

P815 mastocytoma (H-2^d), EL-4 (H-2^b), L1210 (H-2^d), and YAC-1 lymphomas and WEHI-164 fibrosarcoma (H-2^d) were passaged in Iscove’s modified Dulbecco’s medium containing 10% heat-inactivated FCS.

Hamster anti-murine CD3 Ab 2C11, a gift of Dr J. A. Bluestone, University of Chicago (Chicago, IL), was purified by passage over protein G-Sepharose. For staining of membrane-associated TNF, polyclonal rabbit anti-mouse TNF-α (IP-400) obtained from Genzyme (Cambridge, MA) was used at a final concentration of 1:20, and detected by fluoresceinated goat anti-rabbit IgG at 1:50 (Organon Technica, West Chester, PA). All other Abs were from PharMingen (San Diego, CA).

Female C57BL/6 perforin-deficient (PKO) mice, generated as described (10), were bred to C3H/HeJ-Fasl^gld and to C57BL/6^gld males (The Jackson Laboratory, Bar Harbor, ME) under viral Ag-free conditions in the animal facilities at the University of Miami School of Medicine (Miami, FL). Heterozygous first generation (F₁) mice were then brother-sister mated to obtain second generation (F₂) mice that were screened for perforin and Fas-L deficiency by PCR (10) and ligase chain reaction (26), respectively, as described (27). bm1 mice were obtained from The Jackson Laboratory.

One million cells were incubated for 30 min in 50 μl hybridoma 2.4G2 anti-FcRIII Ab supernatant to block FcR. Cells were then incubated for 30 min with FITC- or phycoerythrin-conjugated Abs diluted to 5 μg/ml in PBS containing 0.5% BSA and 0.01% sodium azide. Cells were washed three times and analyzed in a FACScan flow cytometer (Becton Dickinson, Mountain View, CA).

Pancreas and uterus were minced into small cubes and agitated for 1 h on ice in PBS, pH 7.2, containing 10 mM EDTA. Cells were then pipetted and filtered through wiremesh to obtain single cell suspensions. Single cell suspensions from the pancreas were also prepared by cannulating the main duct and injecting collagenase V (2 mg/ml) (Sigma, St. Louis, MO) in tissue culture medium. Cells were dispersed by 10-min incubation at 37° in medium containing 2 mg/ml collagenase V with constant agitation. Cells were washed several times, and infiltrating white cell elements were separated from other cells by gradient centrifugation using Lympholyte M (Accurate Biochemicals, Westbury, NY).

If cdd mice were to be used for breeding, carrageenan injection was started on days 35 to 40 with weekly i.p. injections of 100 μg carrageenan (Sigma) dissolved at 1 mg/ml in PBS, pH 7.2, until pregnancy was noted. In experiments measuring the effect of carrageenan on pancreatic infiltrating cells, injection was begun on day 70 and repeated once or twice at weekly intervals before analysis of infiltrating cells.

CD8⁺CTL lines from C57BL/6, C57BL/6 gld/gld, C57BL/6 P0/0, or C57BL/6 cdd mice were prepared by culturing 10 × 10⁶ spleen cells with 10 × 10⁶ irradiated (6000 rad) DBA-2 or bm1 stimulator spleen cells in 4 ml Iscove’s modified Dulbecco’s medium containing 10% heat-inactivated FCS and 50 μM 2-ME (basic medium). Spleen cells were restimulated weekly with 1 × 10⁶ irradiated stimulator cells. After 3 wk, cells were transferred to CTL medium (above medium plus 25 U/ml murine rIL-2, 2.5% rat spleen Con A supernatant, 20 mM methyl-α-d-mannopyranoside) and maintained by weekly restimulation with irradiated spleen cells. For short-term cultures, CTL were tested for cytotoxicity 5 days after primary stimulation or one restimulation.

Staphylococcal enterotoxin B (SEB)-specific CD4 CTL were generated by culturing 2 × 10⁶ spleen cells/ml (C57BL/6) with 5 μg/ml SEB (Sigma) (28) for 5 days. Cytotoxicity assays were conducted against syngeneic cells pulsed with 5 μg/ml SEB.

Peritoneal macrophages used as targets for SEB-specific CD4⁺CTL or bm1-specific CD8⁺CTL were elicited in C57BL/6 or bm1 mice with thioglycolate, as described (29), and labeled with ⁵¹Cr for cytotoxicity assays.

⁵¹Cr release assays were performed according to standard protocols for 4 or 16 h. All assays were performed in triplicate. For redirected lysis, anti-CD3 Ab at 1 μg/ml was present throughout the assay. Inhibition of TNF activity was accomplished by addition of neutralizing anti-TNF-α antiserum (IP-400; Genzyme) at the start of each assay at the indicated concentrations.

To compare the influence of the MHC background on perforin/Fas-L double deficiency, C3H (H-2^a) and C57BL/6 (H-2^b) gld/gld mice were interbred with perforin-deficient C57BL/6 (PKO) mice under viral pathogen-free conditions in appropriate facilities. Double heterozygous F₁ mice were interbred, and tail biopsies of the F₂ generation were screened by PCR for homozygous perforin deficiency, detectable through the insertion of the neomycin resistance gene into the perforin gene (10). The homozygous gld point mutation (6) was screened by a ligase chain reaction assay (26), as described previously (27). Animals homozygous for perforin and Fas-L cdd were identified and set aside for further breeding. Surprisingly, however, it soon became evident that cdd mice, after thriving for about 8 wk of age, did not breed, became ill, lost weight, and died usually by 15 wk (Fig. 1, inset in A). Death was unrelated to any detectable viral or bacterial infection and occurred uniformly within a similar time period, regardless of differences in the MHC background, pure H-2^b, or mixed H-2^b × H-2^a, suggesting that death was caused by the cytotoxic double deficiency and unrelated to genetic background. Male and female cdd mice are equally susceptible to disease and early death. Female cdd mice were infertile, and no pregnancy or delivery was observed even after mating with normal, not immunodeficient C57BL/6 males. Male cdd mice, in contrast, were fertile, but short lived. Mice homozygous for the gld defect, but heterozygous at the perforin locus (P+/0, gld) and expressing one functional perforin allele, were protected from early death (Fig. 1). They were therefore used for breeding to generate sufficient numbers of cdd mice for further study. Heterozygous P+/0, gld mice reproduced normally; as expected, the frequency of cdd offspring was 25%, while 50% of the litter remained perforin heterozygous. P+/0, gld mice were used as controls for experiments with cdd animals.

At autopsy, 11- to 14-wk-old cdd mice showed pronounced splenomegaly, but only moderate to minimal lymph node enlargement in peripheral and mesenteric nodes, which was similar to heterozygous P+/0, gld controls. The size of the pancreas was considerably reduced in cdd compared with P+/0, gld mice. On microscopic examination, mild levels of cellular infiltration were apparent in the liver, lung, and kidneys of both cdd and P+/0, gld controls. However, a striking difference was apparent in the pancreas, which in 20 of 20 cdd mice exhibited severe pancreatitis (Fig. 1,A, right), whereas all P+/0, gld control pancreata appeared normal (Fig. 1,A, left). In cdd mice, the pancreas contained extensive mononuclear infiltrates, which were accompanied by a loss of exocrine acinar cells. Despite the severity of the inflammatory response, secretory ducts were unaffected and islets of Langerhans remained intact, as evidenced by immunostaining for insulin and the absence of hyperglycemia (not shown), even in terminally ill mice. Ovaries from cdd mice were infiltrated by mononuclear cells, with a corresponding reduction in the number of maturing follicles (Fig. 1,B), and infiltration of the uterus with mononuclear cells was severe when compared with P+/0 gld controls (Fig. 1 C).

The gld defect is associated with the expansion of a B220-positive T cell subset in lymph node and spleen (8, 30) with increased Fas expression (31). While gld mice show severe lymphadenopathy usually after 6 mo, they do not develop weight loss or pancreatitis later in life. To determine the nature of infiltrating cells that may be responsible for pancreatitis and hysterosalpingitis in cdd mice (P0/0, gld/gld), we compared them to infiltrating pancreatic cells in perforin heterozygous, gld/gld control mice (P+/0, gld) differing by one functional perforin allele (Fig. 2). Since perforin heterozygous gld mice had only few resident white cells in the pancreas, it was necessary to pool the infiltrating pancreatic white cells from three P+/0, gld mice controls and compare them with each cdd mouse. P+/0, gld mice had normal body weight without detectable lymphadenopathy, while the cdd cells came from 70–100-day-old mice of about 20 g body weight, decreased from 26 g within the previous 2 wk. The analysis was repeated 12 times over a period of more than 1 yr, with similar results using mice of pure H-2^b (C57BL/6) or mixed H-2^b × H-2^a (C57BL/6 × C3H) background. In comparison with pancreatic infiltrating cells, data are also presented for cdd spleen cells (n = 6), cdd peripheral (n = 5) and mesenteric lymph nodes (n = 3) together with P+/0, gld controls (Fig. 2). The frequency of Mac-1-positive cells infiltrating the pancreas depended on the stage of disease (see Fig. 5 below) and increased with increasing age of the mice. Data of surface markers (mean and SD) showing significant differences between P+/0, gld, and cdd mice are shown.

Infiltrating mononuclear cells in the pancreas were separated from glandular pancreatic parenchyma cells by density-gradient centrifugation following collagenase digestion and mechanical dispersion of the pancreas. Approximately 30 million infiltrating mononuclear cells could be recovered from each cdd pancreas, while P+/0 gld pancreata yielded less than 5 million mononuclear cells per pancreas. In addition to the quantity of cellular infiltrates in pancreata from cdd mice, infiltrating mononuclear cells were phenotypically different from those in P+/0 gld pancreata. Infiltrating cells in cdd mice differed primarily by an increased frequency of the expression of the Mac-1 (CD11b) surface marker (Fig. 2). Both Mac-1/CD3-positive T cells and Mac-1/F4.80-positive macrophages are increased in cdd pancreata when compared with the perforin heterozygous gld state (Fig. 2). B220/CD3-positive cdd T cells, on the other hand, are increased to a lesser extent in the cdd pancreas than the P+/0 gld pancreas (Fig. 2). The increase of Mac-1/CD3-positive cdd T cells, relative to P+/0, gld, is found to a lesser extent in cdd peripheral lymph nodes (Fig. 3, A and B) and spleens (Fig. 3 C). In contrast, Mac-1/F4.80-positive macrophages are not increased in spleen and lymph nodes of cdd mice. Mesenteric and peripheral lymph nodes and spleens of both cdd and P+/0 gld mice show increased CD3/B220-positive T cells. cdd spleens are characterized by an increase of T cell cellularity that may be responsible for the enlarged spleens in cdd mice noted above.

The cdd state is characterized by the presence in the pancreas of Mac-1/F4.80-positive monocytes/macrophages in addition to Mac-1-positive T cells and by tissue destruction (Fig. 3, A and B). Mac-1-positive T cells, but not monocytes/macrophages, are also found in cdd lymphoid organs. A small number of Mac-1/CD3-positive cells is normally present in healthy spleens, probably representing a subset of activated T cells (Fig. 3,C). This population is increased in the perforin heterozygous gld state and quadruples in the spleen of cdd mice (Fig. 3 C).

Infiltrating mononuclear cells in the uterus of two cdd mice analyzed likewise are characterized by Mac-1-positive T cells and monocytes/macrophages (data not shown), suggesting a similar pathogenetic mechanism for pancreatitis and hysterosalpingitis. Moreover, the CD4 to CD8 ratio of infiltrating cells in both organs undergoing tissue destruction in cdd disease was found to be inverted, suggesting expansion of the CD8 subset over and above the CD4 subset (Fig. 4). The inversion of the CD4/CD8 ratio was not found in cdd lymph nodes and spleen cells of the same cdd animal (Fig. 4). Infiltrating T cells are TCR αβ positive and express high levels of Fas, and a fraction is CD4/CD8 double negative (data not shown).

The presence of monocytes/macrophages in pancreas and uterus of cdd mice suggests their participation in destruction of these organs either directly or through the activation and expansion of infiltrating CD8⁺ T cells. Thus, destruction of the pancreas and infertility of female cdd mice may be causally related to the presence of increased numbers of monocytes/macrophages. To test this hypothesis, monocytes/macrophages were inactivated in cdd mice by weekly i.p. carrageenan injection (100 μg in PBS), beginning on day 35 to 40 of their life.

Carrageenan injection improved the fertility of female cdd mice to the extent that they became pregnant and delivered healthy cdd litters (Fig. 5), while untreated female cdd mice never became pregnant. About 70% of treated mice became pregnant and remained healthy for up to 130 days, indicating that carrageenan treatment, possibly together with the ensuing pregnancy, delayed onset of cdd disease. Male mice, even when treated continuously by weekly carrageenan injections, showed only a comparatively small benefit when compared with female mice in disease onset and survival (Fig. 5, upper panel).

Weekly carrageenan injection also delayed pancreatic disease. The extent of pancreatic infiltration by Mac-1-positive cells, including both T cells and monocytes/macrophages, over time increases steadily from 20% on day 60 to 80% by day 100 and remains constant thereafter in the few mice surviving beyond this time. If a regimen of weekly carrageenan injection is implemented on day 70 at a time when the mice are beginning to lose weight, the infiltration of Mac-1-positive cells in the pancreas is halted and appears to be reversed to levels of less than 10% in some animals (Fig. 5, lower panel). Mac-1-positive cells include both T cells and monocytes/macrophages, indicating that infiltration of both cell types is reversed by carrageenan. Since carrageenan affects only macrophages, this finding suggests that expansion of Mac-1-positive T cells is secondary to monocyte/macrophage presence in the tissues undergoing destruction.

Allospecific CD8⁺CTL lines (H-2^b, anti-H-2^d) were generated from spleens of cdd mice in parallel with control CTL lines from C57BL/6, PKO, B6 gld/gld mice and compared for their ability to lyse allospecific (H-2^d) tumor targets in a standard 4-h ⁵¹Cr release assay (Fig. 6,A). Wild-type B6 CTL and gld CTL exhibited a high level of cytotoxicity against P815 targets, while killing by PKO CTL was significantly reduced. cdd CTL, in contrast, were completely unable to lyse P815 targets in 4 h. Similar results were obtained using primary allospecific CTL generated in 5-day MLC from cdd and control mice (not shown). cdd CTL remained incapable of P815 lysis even when redirected by anti-CD3 (2C11) Ab (Fig. 6,B). Similarly, Fas transfection of P815 was also incapable of rendering targets susceptible to 4-h lysis by cdd CTL (Fig. 6 C), further demonstrating loss of the Fas-L- and perforin-killing pathways.

cdd CTL are unable to mediate acute lysis of P815 or P815-Fas by perforin, Fas-L, or any other mechanism under these conditions. To determine whether other, slow lytic pathways are present, cdd CTL were tested against TNF-sensitive WEHI-164 (H-2^d) targets in a 16-h ⁵¹Cr release assay. As reported by us previously for cdd lymphokine-activated killer cells (27), cdd CTL effectively killed WEHI-164 cells at very low E:T ratios within 16 h (Fig. 6 D). Lysis was completely inhibited by neutralizing anti-TNF-α antiserum.

The inability of cdd CTL to lyse target cells in 4-h assays using the perforin or the Fas-L effector pathway suggested a mechanism by which expansion of monocytes/macrophages in the pancreas and uterus of cdd mice could take place. Assuming that resident tissue macrophages present Ag and activate cognate T cells, ensuing T cell activation coupled to the expression of Fas-L and perforin will result in the generation of cytotoxic T cells that are capable of eliminating APCs through one or both of the major cytolytic pathways. The lytic action of CTL is expected to limit monocyte/macrophage action, thereby down-regulating further T cell activation. To test this hypothesis, CD4⁺ cdd CTL were generated by addition of SEB to B6 gld or B6 cdd spleen cells and cultured for 5 days. Thioglycolate-elicited macrophages from normal B6 mice were pulsed with SEB and used as targets (Fig. 6,F). CD8⁺cdd and gld CTL were generated in MLC against bm1 targets (H-2^b with a mutation in K^b) and tested against bm1 peritoneal macrophages (Fig. 6 E). While gld CTL of the CD4 or CD8 phenotype were able to lyse the appropriate targets in 4 h, cdd CTL were completely unable to lyse peritoneal macrophages, indicating that perforin is essential for lytic activity in the gld state in this system and that cdd T cells are unable to eliminate cognate Ag-presenting macrophages.

The unexpected occurrence and surprising severity of autoimmune disease in perforin/Fas-L double-deficient mice point to an essential function of these two rapidly acting cytolytic effector pathways in regulating cell-mediated tissue destruction. Since neither perforin deficiency alone nor Fas-L deficiency alone produces this particular syndrome, characterized by pancreatitis and hysterosalpingitis, it is clear that each effector pathway alone is capable of controlling this form of tissue destruction and that the combined absence of the two effector pathways is necessary for disease expression. Therefore, in regard to preventing this syndrome and maintaining homeostasis, perforin and Fas-L function are redundant. This reveals a novel role for perforin in maintaining homeostasis of the immune system, which is apparent only in the absence of functional Fas-L. This conclusion is supported by earlier observations in mice with autoimmune syndromes. MRL mice develop spontaneous autoimmunity characterized by glomerular disease and mononuclear infiltration of liver and salivary glands. In Fas-deficient MRL (MRL/lpr) mice, the autoimmunity and early mortality are significantly more severe. Autoimmune disease and mortality are further aggravated by the additional deficiency of perforin (MRL/lpr/P0/0) (32). In contrast to MRL mice, normal C57BL/6 mice do not develop spontaneous autoimmune disease. Fas-L deficiency causes primarily lymphoproliferation and lymphadenopathy in C57BL/6 mice without an initial component of autoimmunity. Eventually, however, massively proliferating lymphocytes infiltrate virtually all organs and cause mortality toward the end of the first year of life, presumably through defects in peripheral tolerance induction in T cells. Additional perforin deficiency in C57BL/6/gld mice, as shown in this study, not only causes early death by 8–10 wk of age, but overt organ-specific tissue destruction, accompanied by only moderate lymphoproliferation. The tissue specificity of cdd disease suggests the contribution of an autoimmune component aggravating the genetic defects. This is further supported by the inversion of the CD4/CD8 ratio in affected organs. Cell transfer studies and culture of pancreas-specific CD8 cells will address this question. Alternatively, pancreas and uterus of cdd mice may lack redundant mechanisms of restraining T cell expansion present in other tissues, resulting in the organ-specific T cell expansion.

Early death and generalized lymphoproliferation are also seen in CTLA-4-deficient mice with a predilection for severe pancreatitis and myocarditis (33, 34). While mechanistically clearly different from CTLA4 deficiency, our study supports the concept of organ-specific protection against lymphocyte damage.

Mechanistically, we propose that perforin and Fas-L down-regulate Ag presentation by killing APCs via either the perforin- or the Fas-L-mediated mechanism. Continued survival of APCs in the face of activated T cells may result in continuous restimulation and expansion of both CTL and APCs, ultimately leading to organ destruction by perforin- and Fas-L-independent mechanisms. This explanation of the pathogenetic mechanism of cdd disease implies that activated cytotoxic T cells control their own down-regulation by a negative feedback loop, eliminating the original antigenic stimulus through lysis of the APCs (Fig. 7). If this negative feedback loop is inactivated through the ablation of perforin and Fas-L gene expression, autoimmunity ensues. This disease mechanism also implies that cdd T cells are able to mediate tissue destruction.

The nature of the molecular effector of tissue destruction has not been explored, but several candidates may be considered. Pancreas and uterus are the sites of tissue destruction in cdd disease and harbor both expanded monocyte/macrophage and activated T cell populations. Monocytes/macrophages may produce nitric oxide and activated oxygen as effectors of tissue destruction, in addition to unbridled TNF production. Infiltrating cdd T cells, in addition to producing macrophage-activating factors such as IFN-γ and TNF, may exert cytotoxic functions via TRAIL (35, 36, 37, 38, 39, 40), the ligand for DR4 to DR7, or DR3-ligand (41, 42, 43, 44, 45). cdd CTL, in addition to producing TNF (Fig. 6 (27)), express high levels of TRAIL mRNA (data not shown).

In addition to the demonstration of expanded monocyte/macrophage populations in affected tissues, the postulated pathogenetic mechanism of cdd disease is supported by the dramatic effect of carrageenan treatment on cdd mice. Carrageenan, a monocyte/macrophage toxic agent, delays and reverses tissue expansion of both monocytes/macrophages and T cells, restores fertility in female mice, and significantly delays autoimmune disease onset and death. Apparently, the elimination of functional monocytes/macrophages in vivo by carrageenan can partially replace the defect in cytotoxicity normally responsible for control of APCs. cdd mice generated by Braun et al. (46) also died relatively early, but survived and bred for 120 to 150 days without treatment in facilities that, unlike ours, were not reported to be viral Ag free (46). It is possible that pathogen exposure is beneficial in cdd disease and delays onset of disease, reminiscent of the effect of carrageenan. Our results show that cdd T cells are capable of mediating tissue damage directly or through activation of other cells. In ongoing studies, we have found that cdd T cells can cause lethal graft versus host disease in allogeneic recipients if precautions are taken against rejection of the graft. The effector mechanisms of cdd graft versus host disease and tissue destruction are under analysis.

We acknowledge the excellent technical assistance of Joseph F. Rhoderick.