The Listeria monocytogenes protein listeriolysin O (LLO) is a pore-forming protein essential for virulence. Although the major role for LLO is to allow L. monocytogenes entry into the cytosol, it also induces apoptosis of activated lymphocytes, an obligatory cellular response that modulates the infection. Induction of apoptosis by LLO proceeds through a fast, caspase-dependent pathway and a slow, caspase-independent pathway. Polyclonal T cell lines were generated from either normal mice or mice deficient in granzyme and perforin proteins, and then treated with apoptogenic doses of LLO. In this study we show that apoptosis of lymphocytes induced by LLO was characterized by activation of caspases as quickly as 30 min that was dependent on the expression of granzymes. In the absence of granzymes, all parameters of apoptosis such as caspase activation, phosphatidylserine exposure, mitochondrial depolarization, and DNA fragmentation were dramatically reduced in magnitude. Removal of perforin inhibited the apoptotic effect of LLO on cells by ∼50%. Neutralization of intracellular acidification using chloroquine inhibited the rapid apoptotic death. In agreement with these findings granzyme-deficient mice harbored lower bacterial titers and decrease splenic pathology compared with normal mice following L. monocytogenes infection. Thus, LLO exploits apoptotic enzymes of the adaptive immune response to eliminate immune cells and increase its virulence.
Infection with the Gram-positive bacterium Listeria monocytogenes is a powerful model to examine bacterial virulence and immune regulation after infection. Infection of mice with L. monocytogenes causes marked apoptosis of lymphocytes, hepatocytes, and neurons (1, 2, 3). Indeed there is a growing list of bacterial pathogens that induce apoptosis making it important to understand the molecular mechanisms behind it (4, 5).
L. monocytogenes expresses a virulence cluster dedicated to invasiveness in mammalian species (6). One of the virulence factors in the cluster is the pore-forming toxin listeriolysin O (LLO),3 a member of the cholesterol-dependent cytolysin family (7, 8, 9, 10, 11, 12). Cholesterol-dependent cytolysins are expressed by a number of Gram-positive bacteria and have various functions, from delivery of toxins (streptolysin O) (13) to compromising phagosomes of infected cells (LLO). The main role attributed to LLO is to allow L. monocytogenes to escape from the phagosome into the permissive environment of the host cell cytosol (14, 15). Bacteria deficient in LLO are avirulent in vivo and in vitro. Treatment of mice with a mAb that neutralizes LLO also renders L. monocytogenes avirulent in vivo and in vitro (16, 17).
Lymphocyte apoptosis takes place in infective foci at the time of exponential growth of the microbe (1). Phagocytes also die after infection, but the mechanism of death is not understood. Mice deficient for the type I IFN receptor have decreased lymphocyte apoptosis and increased survival of a subset of macrophages (18, 19, 20). The apoptotic lesion is immunomodulatory, leading to decreased host resistance and increased bacterial proliferation (4, 21). We have postulated that during the exponential growth of L. monocytogenes, LLO is released to the extracellular fluids, leading to lymphocyte apoptosis. We favor this hypothesis based on the findings that 1) Abs to LLO blocked apoptosis in vivo (16); 2) free L. monocytogenes was found in the inflamed lesions surrounding apoptotic cells; 3) lymphocytes were never infected with L. monocytogenes; and 4) nanomolar doses of purified LLO induced apoptosis of dividing T cells with fast kinetics, and activation of caspase-3, noted as early as 30 min after treatment (22).
In examining the mechanism of action of LLO in causing lymphocyte apoptosis, we considered the role of granzymes. We reasoned that due to the rapid kinetics, a membrane proximal event should be the inductive event. Of all the apoptotic signals studied to date, granzyme-mediated induction of cellular death has kinetics most similar to LLO induced apoptosis. LLO has a pH optimum that enhances its activity in the phagosomal environment and could either lyse or permeabilize the acidic vesicles/granules that contain granzymes releasing them into the cytosol (23, 24). Hence, LLO may act as an endosomolytic agent. In fact, extracellular LLO has been used to deliver large amounts of purified recombinant granzyme B to target cells (25). Alternatively, LLO could be inducing signaling cascades inside cells that lead to granzyme-dependent apoptosis. To our knowledge this example is the only study of a protein that induces apoptosis through cell autonomous granzyme activity. We prove that granzyme is the major executor of the fast cellular death seen in activated T cells treated with LLO and, importantly, we also indicate an effect in the infection.
Materials and Methods
Mice and infections
129/SvJ mice were purchased from The Jackson Laboratory. Granzyme A × B−/−, granzyme B−/−, and granzyme A × B cluster−/− mice on a 129/SvJ background were a gift from Dr. T. J. Ley (Washington University School of Medicine, St. Louis, MO). A detailed description of how all the granzyme-deficient strains of mice generated can be found reported previously (26). In vivo infections were performed as previously described (21). H&E and TUNEL staining was performed as previously described (1). All mice were bred and maintained at our animal facility and used according to the protocols defined by the Division of Comparative Medicine, Washington University School of Medicine.
Cell culture and LLO treatment
T cell lines were generated by immunizing mice in the hind footpad with 10 nM OVA (Sigma-Aldrich) in CFA (Difco). Lymph nodes were isolated and lines were generated using conventional techniques. The initial T cell line was passaged as follows: 1 × 106 T cells and 2 × 107 dispersed irradiated (3000 rad) splenocytes were cultured in 20 ml of DMEM supplemented with 10% defined FCS, 50 U/ml IL-2, and 10 μM OVA. T cell lines were passaged every 7 days. At every passage and before any experiment, T cell lines were extensively purified by centrifugation over a Histopaque 1119 gradient (Sigma-Aldrich) to remove dead cells. This process also removes the irradiated splenocytes. T cells purified by this process were 100% CD4+, CD69+, and CD62L− by flow cytometric analysis (22). Wild-type 129/SvJ T cells were ∼30–40% positive for granzyme B expression by intracellular flow cytometry vs ∼0% for the granzyme A−/− × granzyme B−/−, granzyme B−/−, and granzyme A−/− × B cluster−/− T cell lines. T cell lines were treated with 250 ng/ml purified recombinant LLO as previously described (22). For chloroquine treatments, cells were incubated in the indicated concentrations of chloroquine (Sigma-Aldrich) for 10 min before being treated with LLO for the indicated times.
RNA was isolated from cultured 129/SvJ or granzyme A−/− × B cluster-deficient mice (gzmAC) T cells using the RNeasy Mini kit (Qiagen). First strand cDNA was made using the Superscript III First Strand Synthesis kit (Invitrogen). PCR was conducted with GoTaq Green Master mix (Promega). All protocols were performed using the manufacturer’s protocols.
Con A (Sigma-Aldrich) was used at 25 μg/ml. Whole splenocytes were incubated in Con A for 16 h, then purified by density gradient and treated with LLO. For experiments involving calcium-free medium, MEM base (Sigma-Aldrich) was supplemented with all the components of complete DMEM except calcium.
Conventional DMEM containing calcium was made following the same protocol. T cells were washed out of conventional DMEM and resuspended in DMEM with or without calcium and 1% FCS for the duration of the experiment.
Abs and immunoblotting
Whole protein was isolated from 1 to 6 million cells by standard techniques, resolved by SDS-PAGE and transferred to Hybond-P membranes (Amersham Biosciences). Caspase-3 (8G10), caspase-6 (9762), and caspase-9 (9504) Abs were obtained from Cell Signaling Technology and used at 1/4000, 1/2000, and 1/2000 dilutions, respectively. Goat anti-rabbit HRP was obtained from Cell Signaling Technology and used at 1/8000 (caspase-3) or 1/4000 (caspase-6 and caspase-9) dilution. Detection was performed with ECL Plus (Amersham Biosciences).
The cationic dye JC-1 was purchased from Molecular Probes. For JC-1 incorporation assays, cells were washed twice in PBS and resuspended in DMEM containing 10 μg/ml JC-1. Cells were incubated for 30 min at 37°C, washed twice in DMEM, and analyzed by flow cytometry. Anti-active caspase-3 Ab, annexin V-PE and 7-aminoactinomycin D (7-AAD) were obtained from BD Biosciences and used per the manufacturer’s protocol. All flow cytometry was conducted on a FACSCalibur (BD Biosciences), and data were collected using CellQuest software (BD Biosciences) and subsequently analyzed using FlowJo (Tree Star). CFDA-SE was used by the manufacturer’s instructions at a final concentration of 10 μM dye (Molecular Probes).
Granzymes are responsible for most of the rapid T cell apoptosis induced by LLO
CD4 T cells from normal mice and from mice deficient in granzyme expression were examined for LLO-induced apoptosis. We have chosen to mainly examine CD4 T cell lines because they are easily cultured, polyclonal, and are well propagated in culture. Limited experiments were made with CD8 T cells that are more fastidious in long-term culture.
Ag-specific T cell lines were generated against OVA by immunizing either 129/SvJ or granzyme A−/− × B cluster-deficient mice (gzmAC). The gzmAC mice are a cross between granzyme A−/− (gzmA−/−) and granzyme B−/− (gzmB−/−) cluster mice. The gzmB−/− cluster mouse is deficient for granzyme B and has reduced expression of granzymes C, D, F, and G due to a positional effect of neomycin insertion into the granzyme B gene locus (27, 28). The gzmAC T cell lines are completely deficient in granzymes A and B and have reduced expression of the granzyme B cluster granzymes C, D, F, and G. This observation was confirmed by RT-PCR. Our wild-type T cell lines expressed granzymes A and B at the RNA level (Fig. 1 A). We also detected granzymes C, D, and G. In the gzmAC T cells, we did not detect granzymes A, B, or C, but still detected RNA for granzymes D and G. The reduction in granzymes A, B, and C mRNA in the gzmAC T cells is consistent with previous data (27). There is little data on the function of granzymes D and G. Based on sequence analysis they appear to cleave hydrophobic residues, unlike granzymes A and B, which are a tryptase and aspase, respectively. We were unable to detect granzymes E, F, K, M, or N in either cellular preparation (29).
A time-course study examined whether the kinetics of LLO-induced apoptosis differed between cells derived from 129/SvJ and gzmAC mice. Fig. 1, B and C, are flow cytometry plots for annexin V and 7-AAD demonstrating a dramatic reduction of apoptosis following LLO treatment of gzmAC T cells. Fig. 1,D shows the compiled data of all the flow cytometry plots of annexin V and 7-AAD staining following LLO treatment. Apoptotic cells are annexin V+ (a marker of phosphatidylserine) and 7-AAD− (a marker of DNA), depicted in green in Fig. 1,D. Late apoptotic or necrotic cells are positive for 7-AAD (depicted in red in Fig. 1 D).
LLO caused ∼44.8%, 57.1%, and 76.7% T cell apoptosis at 2, 4, and 8 h posttreatment, respectively, in the 129/SvJ T cells with a background apoptosis in untreated cells of 9.61% (Fig. 1, B and D). In contrast, gzmAC T cells showed a reduction, with only 13.2%, 21.5%, and 28.6% apoptotic cells at 2, 4, and 8 h, respectively, over a background of 15.2% (Fig. 1, B and D).
Despite the dramatic reduction of apoptosis and total cell death within the first 8 h following LLO treatment, a slower death was found in the gzmAC T cells, at 24 h after treatment. By 24 h of treatment with LLO there was a total of 74.7% cell death in the 129/SvJ T cells, and 48.8% in the gzmAC T cells (annexin V-positive cells). The cell death seen in gzmAC cells at 24 h was distinct from the rapid granzyme-mediated death, the annexin V+/7-AAD− cells were not found but instead annexin V+/7-AAD dull to bright cells and annexin V−/7-AAD+ cells dominated the culture (Fig. 1, C and D).
CD4+ T lymphocytes are not conventionally thought to be cytotoxic cells that express granzyme and perforin. However, recent evidence has shown that human CD4+ T cells activated with anti-CD3 and anti-CD46 express granzyme B (30). Also, granzyme B is important in the regulation of activation induced cell death in different Th cell subsets (31). Granzyme B RNA was detected by quantitative RT-PCR in the wild-type but not the gzmAC T cell lines. Levels of message were >212 higher in the 129/SvJ T cells than the gzmAC T cells. This result was confirmed with a cross-reactive mouse anti-human granzyme B Ab. Intracellular staining for granzyme B protein, although weak, disclosed 29.6% of the T cells positive for granzyme B by flow cytometry in the wild-type and 0.61% granzyme-positive cells in the gzmAC T cells. Examination of lymphocytes by immunofluorescent microscopy revealed that granzyme B showed a punctate distribution compatible with compartmentalization in secretory granules.
Caspase activation, mitochondrial depolarization, and DNA fragmentation following LLO treatment was attenuated in gzmAC T cells
Caspase immunoblots performed on whole cell lysates of T cells treated with LLO detected both the pro- and active forms of the caspases (32). Activated caspase-3, caspase-6, and caspase-9 were detected as early as 30 min after treatment with LLO but were not found in gzmAC T cells (Fig. 2,A). Fig. 2 B shows representative flow cytometry plots for activated caspase-3 on either untreated or LLO-treated T cells. By 2 h after treatment with LLO, 41% of 129/SvJ T cells were caspase-3-positive over a background of 6.09%. In contrast, only 12.1% of the gzmAC T cells were caspase-3-positive after 2 h of LLO treatment compared with a background of 9.32%.
After 24 h of treatment there were 70.1% active caspase-3-positive T cells, but the background cell death had risen to 24.1% (Fig. 2 B). The background levels were approximately the same in gzmAC T cells after 24 h in culture, but after treatment with LLO only 46.1% active caspase-3-positive cells were found. This result indicates that most of the caspase activation seen within the first 2 h of treatment with LLO is granzyme-dependent, but that there is a slower pathway to caspase activation that operates in the absence of granzyme.
Fig. 3 A shows the data obtained after staining LLO-treated T cell lines with the mitochondrial dye JC-1, a cationic lipophilic dye that binds preferentially to mitochondrial membranes and undergoes a green to red color transition when the intermembrane space depolarizes due to apoptosis. A rapid increase in mitochondrial permeability was detected, with ∼66.1%, 72.5%, and 81.4% of the 129/SvJ T cells becoming JC-1 red after 2, 4, and 8 h compared with ∼21.9% background. This magnitude of mitochondrial depolarization was not seen in gzmAC T cells, which displayed 27.8%, 32.7%, and 42.4% JC-1 red at 2, 4, and 8 h after LLO treatment vs a background of ∼21.3%.
The final parameter examined was DNA fragmentation as measured by TUNEL reaction. As early as 2 h after treatment with LLO, 57.2% of the 129/SvJ T cells showed fragmented DNA, compared with 11% for the untreated (Fig. 3 B). The gzmAC T cells, showed 24.8% TUNEL-positive after LLO treatment compared with 13.8% for the control. By 24 h of treatment with LLO, 86.4% of the 129/SvJ T cells were TUNEL-positive over a 26.1% background. Even in the absence of granzyme we still detected DNA fragmentation, with 48.9% of the gzmAC T cells TUNEL-positive over a 17.1% untreated background.
Role of perforin, extracellular calcium, and mitogenic stimulation on granzyme-mediated LLO-induced apoptosis
Delivery of granzymes from cytotoxic lymphocytes to their targets is dependent on the expression of perforin. Perforin allows granzyme to translocate across membranes through an unknown mechanism. Both perforin and the cholesterol-dependent cytolysins contain a conserved membrane attack complex/perforin domain that participates in the oligomerization and pore formation of these proteins (33). To determine the extent of perforin’s contribution to the granzyme-dependent death seen after treatment with LLO, we treated perforin-deficient (prf−/−) T cells with LLO and tested annexin V or 7-ADD as previously described. Fig. 4,A shows representative flow cytometry plots of either untreated or LLO-treated T cell lines. We detected approximately the same level of background cell death in all three cell lines. After treatment with LLO, we detected a 22%, 16%, and 5% increase in the number of annexin V+/7-AAD− cells in 129/SvJ, gzmAC, and prf−/− cell lines, respectively. Analysis of three independent experiments revealed that prf−/− cells died early after LLO treatment but the incidence dropped ∼50% compared with the 129/SvJ T cells (Fig. 4 B). Perforin appears to enhance the ability of LLO to induce apoptosis of treated lymphocytes but the mechanism for this enhancement inside cells is unknown.
To examine a possible role of extracellular calcium, T cells were cultured in complete medium in the presence or absence of calcium and then treated with LLO for different times. We found equivalent caspase-3 activation after treatment with LLO regardless of the presence of extracellular calcium (Fig. 4 C). We also tested the involvement of the calcium-dependent protease calpain by treating T cells with the calpain inhibitor ALLM, but this treatment had no effect on LLO-induced apoptosis (data not shown). Perforin activity is calcium dependent (34). The fact that depletion of calcium had no effect on LLO activity in T cells suggests to us that there is no secretion of lytic granules that are causing neighboring T cells to die. This data would suggest that perforin potentiates the release of granzyme inside the LLO-treated cell.
One potential caveat to this work is that bulk Ag-specific T cells may reflect altered responses to stimuli after culture. We wanted to validate that the result obtained with bulk T cell lines in culture could be replicated with previously uncultured lymphocytes. Whole spleen cells were treated with Con A for 16 h to induce them to proliferate. The lymphocytes were then purified by density gradient centrifugation and treated with LLO for different times. Control splenocytes were treated in the same way but did not receive Con A. Fig. 4,C shows the result of immunoblot against active caspase-3 following LLO treatment. We observed that treatment with Con A was sufficient to induce up-regulation of pro-caspase-3 in the splenocytes. This observation is consistent with previous observations that showed up-regulation of caspase-3 in T cells following activation (35). Treatment with Con A also sensitized the cells to the killing effects of LLO. There was activation of caspase-3 only in Con A-treated, but not resting splenocytes. We fractionated the Con A-treated T cells to determine whether CD8+ T cells would also undergo apoptosis in response to LLO treatment. Fig. 4,D shows that activated CD8+ spleen T cells also activated caspase-3 in response to LLO exposure. Activation of caspase-3 in Con A activated CD8+ T cells was also dependent on expression of granzymes (Fig. 4 D).
Cellular acidification was required for LLO-induced rapid apoptosis
LLO is unique in the cholesterol-dependent cytolysin family of toxins in its pH dependence of its pore-forming activity (23, 24). At neutral pH and physiological temperatures, LLO becomes inactivated due to an irreversible conformational change in the domain responsible for penetrating lipid bilayers (36). This dependence on pH limits its activity in the cytosol of infected cells; at pH 6.5 LLO would only be very weakly active (24). Accordingly, we tested the pH dependence of apoptosis induced by LLO by using the phagosomal-neutralizing agent chloroquine (37).
Treatment of the T cell lines with chloroquine alone did not cause cell death as measured by annexin V or 7-AAD staining (Fig. 5,A). Chloroquine reduced the number of 129/SvJ T cells that became annexin V+/7-AAD− after 2 h of LLO treatment from 42.1% to 10.9%. The effect of chloroquine was titratable, as lowering the dose from 10 to 1 μM ablated its inhibitory ability on LLO (Fig. 5, B and C). Despite the effect of chloroquine on the wild-type cells, there was no effect on the 2-h LLO-treated gzmAC T cells. We found 9.72% annexin V+/7-AAD− gzmAC T cells after 2 h of LLO treatment and 8.38% if the cells were incubated in 10 μM chloroquine. Chloroquine reduced the number of annexin V+/7-AAD− 129/SvJ T cells found 24 h after treatment of LLO from 55.9% to 25.6% (30.3% reduction) (Fig. 5, A and C). However, this difference was the same found at 2 h (42.1% to 10.9% is 32.1% reduction). Interestingly, chloroquine reduced the number of dead gzmAC T cells at 24 h from 59.9% to 32.5%. This result suggests that part of the slow-death seen in the absence of granzymes also depends on phagolysosomal acidification. Neutralization of cellular acidity would have no effect on granzyme B activity in our assays because granzyme is more active at neutral than acidic pH (38). Also, perforin activity is also more active at neutral pH, therefore the entire perforin-granzyme pathway would be increased in activity by neutralization of acidity, arguing that the effect of choloroquine on LLO-induced apoptosis is at the level of LLO activity inside cells.
Granzyme B accounted for most of the rapid LLO-induced apoptosis
We wanted to determine the contribution of specific granzymes to the rapid LLO-induced apoptosis. By immunoblot, we did not find much activation of either caspase-3 or caspase-9 in either gzmA−/− × gzmB−/− or gzmB−/− T cell lines (Fig. 6,A). We detected a very small amount of caspase-3 and caspase-9 activation in gzmB−/− T cells by immunoblotting at 2 h after treatment. This finding would suggest that a minor component of the caspase activation seen after LLO treatment is dependent on granzyme A expression. The result was confirmed by intracellular staining for activated caspase-3; only the 129/SvJ T cells had activated caspase-3 over background after 2 h of treatment (Fig. 6,B). By 24 h after treatment with LLO, some level of activated caspase-3 was found in all granzyme-deficient T cell lines; however, none of them reached the level of the 129/SvJ (Fig. 6,C). Similar results were obtained with annexin V/7-AAD staining. The annexin V or 7-AAD data for the gzmB−/− and gzmA−/− × gzmB−/− T cells was similar to the data obtained with the gzmAC T cells. The granzyme-deficient T cells displayed cell death, but again not to the same level as the wild-type (Fig. 6, D and E). Removal of granzyme B alone eliminated most of the caspase-3 activation and diminished the rapid cell death seen after LLO treatment. In the absence of granzyme B there was still a small amount of caspase-3 activation that was dependent on granzyme A expression. It did not appear that this low level of caspase-3 activation was sufficient to induce extensive cell death, arguing that granzyme B was the dominant enzyme in the fast cell death induced by LLO.
It is possible that LLO is causing the release of lytic granules from the treated lymphocyte to adjacent lymphocytes. A variety of assays, including removal of calcium from the medium, treatment with mAbs against LLO, treatment with chloroquine, and using conditioned medium from LLO-treated cells suggests that this possibility is not the case. To directly examine this possibility, we labeled either 129/SvJ or gzmAC T cells with CFDA-SE and then treated these cells with LLO to see whether we could detect caspase-3 activation. We mixed CFDA-SE labeled 129/SvJ cells with unlabeled gzmAC cells or vice versa at an ∼1:1 ratio and then determined which population of cells was dying after treatment with LLO. In unlabeled wild-type cells treated with LLO, we detected 52.6% caspase-3-positive cells (Fig. 7). When we mixed labeled 129/SvJ cells with unlabeled gzmAC cells we found that only the wild-type 129/SvJ T cells activated caspase-3 following treatment with LLO. Approximately 38% of the CFDA-SE labeled cells were positive for active caspase-3 vs 6% positive for the unlabeled gzmAC T cells. In the reciprocal experiment, we found that 44% of the unlabeled 129/SvJ cells activated caspase-3 vs 7% for the labeled gzmAC cells. We also performed control experiments without mixing the cells and found ∼36% of the labeled 129/SvJ cells activated caspase-3 after treatment with LLO vs 3% for the labeled gzmAC T cells treated with LLO (data not shown). This demonstrates that granzyme is not being transferred to another cell after treatment with LLO.
Decreased apoptosis and increased resistance to infection in granzyme AC mice
129/SvJ and gzmAC mice were infected with ∼2.5 × 104 CFU i.p., and bacterial burden and histology were determined. At day 2 of the infection, gzmAC mice contained ∼15-fold fewer bacteria than the wild-type mice in both the spleens and livers (Fig. 8, A and B). At day 4 of the infection there was a ∼10-fold decrease in the spleens and livers of the gzmAC mice compared with 129/SvJ (Fig. 8, C and D). The difference in colony counts was significant, with a range from p = 0.0313 to 0.0015 when comparing the difference between the 129/SvJ and the gzmAC mice (Fig. 8). The results of one experiment showed a drop in L. monocytogenes burden in both strains after 6 and 8 days of infection, an indication that T cell immunity was taking place. Examination of histology at day 2 following infection, the peak time of lymphocyte apoptosis (1), revealed a decrease in apoptotic lesions in the spleens of gzmAC mice compared with the 129/SvJ (Fig. 9). The reduction in apoptosis found in the gzmAC mice is not due to absence of infection in the white pulp as staining for L. monocytogenes by immunofluorescence detected infectious foci (data not shown). Fig. 9, A and C, shows an apoptotic lesion found in the 129/SvJ mice at day 2 of infection. The 129/SvJ mice showed the typical lesions in the periarteriolar lymphoid sheath with depletion of the lymphocytes and TUNEL-positive cells. Approximately one-third of the white pulp profiles were TUNEL-positive. In contrast the profiles found in the gzmAC mice showed no apoptotic lesions as examined by TUNEL (Fig. 9, B and D). Similar results were obtained in two independent experiments each examining four and six mice. Thus, our data demonstrates that one of the potential mechanisms of induction of lymphocyte apoptosis by L. monocytogenes infection is dependent on granzyme. Experiments of other researchers have not shown such a protective effect with prf−/− mice. Perforin deficiency leads to impairment in sterilization of the splenic compartment, but the overall bacterial growth is unaffected (39). This result has been confirmed by our laboratory (data not shown). The major phenotype of perforin during L. monocytogenes infection is in the secondary response, at which it is required in the CD8+ T cell compartment to mount an effective recall response. Additionally, we have examined the splenic pathology in prf−/− mice following infection with L. monocytogenes, and did not see a major difference in the extent or presence of apoptotic lesion.
This study shows that granzymes are a major participant in lymphocyte apoptosis mediated by LLO, in vitro, as well as in vivo, during the first days of infection. The activation of caspases, kinetics of apoptosis, and the magnitude were all dependent on granzyme expression.
The results are best explained by an increase in permeability of vesicular compartments of the lymphocytes induced by LLO. When LLO is initially internalized it triggers a response by the cell that causes translocation of granzyme from the lytic granule into the cytosol. This effect is enhanced by the presence of perforin. Breakdown of the lytic granule requires that the phagolysosomal compartments acidify, suggesting that increased LLO activity in the endocytic compartment is required for inducing granzyme release into the cytosol. Indeed, chloroquine, which increases lysosomal pH, protected the lymphocyte from the rapid apoptotic death. We do not favor the interpretation that LLO allows transfer of granzyme from one cell to the next: neither cells treated with LLO nor the supernatant following LLO treatment induced apoptosis in fresh cells (4). Labeling with CFDA-SE demonstrated that only the 129/SvJ cells and not gzmAC cells could activate caspase-3 following LLO activation, further excluding the possibility of cell to cell transfer as the mechanism of LLO proapoptotic activities. Along this line, a previous report from the laboratory of Dr. Trapani (Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia) used LLO to deliver extracellular granzyme into target cells as a proof that LLO could act similarly to perforin in serving as a conduit to the cytosol. Thus the precedent is there that LLO can permit the passage of an enzyme into cytosol, and our studies place these findings into a pathophysiological context. Others have examined the mechanisms whereby granzyme induces apoptosis and envision a scenario in which granzyme induces maximal apoptosis through cleavage of bid to promote release of mitochondrial factors, apoptosome formation, and caspase-9 activation. Granzyme B can also directly cleave pro-caspase-3 into the active form, but this pathway is inefficient in the absence of bid activity (38).
The effects of LLO on T cells are akin to a model proposed for perforin in delivering granzyme to target cells (34). In the hybrid model proposed by Pipkin and Lieberman (34), perforin triggers coendocytosis of granzyme through calcium signaling and then disrupts the endocytic vesicle to allow entry of granzyme into the cytosol (34, 40, 41). It is believed that perforin and granzyme enter the target cells through calcium-dependent endocytosis and that the perforin allows granzyme egress from the endosome to the cytosol (25, 34, 42). In the case of LLO, its effects by way of granzyme were also enhanced by the presence of perforin.
The effects of LLO may vary depending on the cell type and its content of vesicular enzymes. In the case of L. monocytogenes-infected macrophages and dendritic cells, the cell death has slower kinetics than the death process seen in lymphocytes following LLO treatment (43, 44, 45). There is no immediate caspase activation or phosphatidylserine exposure in macrophages or dendritic cells either infected with L. monocytogenes or treated with LLO (J. A. Carrero and E. R. Unanue, unpublished observations) (45). We speculate that it would be disadvantageous to the microbe to kill its cellular host too quickly, but important to cause rapid death of adaptive immune cells. In fact, LLO is degraded very quickly in the cytosol of infected phagocytic cells and mutants of LLO that are not degraded or properly translationally regulated kill the host cell and decrease L. monocytogenes virulence (11, 12, 46). It remains to be determined whether other cell types that have high granzyme or protease content such as NK cells or neutrophils are also rapidly killed by LLO.
We still detected a slow cell death in limited cells, in the absence of granzymes and after blocking proteolytic activity. Examination of this slow death program is in progress but it appears not to depend on extracellular calcium signaling, cathepsins, granzymes, or autophagy. Some good candidates for slow death triggering of lymphocyte following LLO treatment are apoptosis-inducing factor, reactive oxygen intermediates, or reactive nitrogen moieties. Notalbly, pneumolysin triggers the release of intracellular calcium stores, which leads to release of apoptosis-inducing factor from the mitochondria and subsequent apoptosis in a peroxide-dependent manner (47).
In vivo, the peak of apoptosis was detected in wild-type mice at 48 h following infection. This coincides with the earliest detectable nonspecific activation of lymphocytes following infection with L. monocytogenes (19, 48). The apoptotic lesion resolves by 96 h following infection, presumably because of the initiation of a robust adaptive immune response that controls the infection (1). Removal of granzyme led to decreased apoptosis and bacterial titer. However, the difference was smaller compared with mice lacking lymphocytes (SCID or RAG−/−) or mice lacking the type I IFN receptor (18, 19, 20, 21). The difference in the former (i.e., normal vs granzyme AC) was 10- to 15-fold compared with the latter (i.e., wild-type vs SCID or RAG−/− mice or type I IFN receptor−/− mice), which was 100- to 1000-fold. We propose that there are multiple cell death pathways operating during L. monocytogenes infection. Some, like granzyme-dependent LLO cell death, are driven directly by the bacterial toxin. But in contrast, the perforin-deficient mice moreover, have a slight defect in clearance of the infection (39, 49). Other host molecules, like TRAIL enhance apoptosis after infection with L. monocytogenes, although the mechanism of induction of this pathway remains unresolved (50). In addition to a role for granzymes in the death of lymphocytes following infection, it is possible that other granzyme- or protease-expressing cell types, such as NK cells or neutrophils, may be dying by L. monocytogenes-induced apoptosis during the early phase of the infection. This possibility remains to be determined.
We thank Dr. Boris Calderon for assistance with staining of tissue sections. We thank Kathy Frederick for animal husbandry. We thank Dr. Timothy J. Ley for the gift of mice.
The authors have no financial conflict of interest.
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 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the National Institutes of Health.
Abbreviations used in this paper: LLO, listeriolysin O; 7-AAD, 7-aminoactinomycin D; gzmAC, granzyme A−/− × B cluster-deficient.