Cytosolic Delivery of Granzyme B by Bacterial Toxins: Evidence that Endosomal Disruption, in Addition to Transmembrane Pore Formation, Is an Important Function of Perforin

Abstract

Granule-mediated cell killing by cytotoxic lymphocytes requires the combined actions of a membranolytic protein, perforin, and granule-associated granzymes, but the mechanism by which they jointly kill cells is poorly understood. We have tested a series of membrane-disruptive agents including bacterial pore-forming toxins and hemolytic complement for their ability to replace perforin in facilitating granzyme B-mediated cell death. As with perforin, low concentrations of streptolysin O and pneumolysin (causing <10% ⁵¹Cr release) permitted granzyme B-dependent apoptosis of Jurkat and Yac-1 cells, but staphylococcal alpha-toxin and complement were ineffective, regardless of concentration. The ensuing nuclear apoptotic damage was caspase dependent and included cleavage of poly(ADP-ribose) polymerase, suggesting a mode of action similar to that of perforin. The plasma membrane lesions formed at low dose by perforin, pneumolysin, and streptolysin did not permit diffusion of fluorescein-labeled proteins as small as 8 kDa into the cell, indicating that large membrane defects are not necessary for granzymes (32 to 65 kDa) to enter the cytosol and induce apoptosis. The endosomolytic toxin, listeriolysin O, also effected granzyme B-mediated cell death at concentrations which produced no appreciable cell membrane damage. Cells pretreated with inhibitors of endosomal trafficking such as brefeldin A took up granzyme B normally but demonstrated seriously impaired nuclear targeting of granzyme B when perforin was also added, indicating that an important role of perforin is to disrupt vesicular protein trafficking. Surprisingly, cells exposed to granzyme B with perforin concentrations that produced nearly maximal ⁵¹Cr release (1,600 U/ml) also underwent apoptosis despite excluding a 8-kDa fluorescein-labeled protein marker. Only at concentrations of >4,000 U/ml were perforin pores demonstrably large enough to account for transmembrane diffusion of granzyme B. We conclude that pore formation may allow granzyme B direct cytosolic access only when perforin is delivered at very high concentrations, while perforin’s ability to disrupt endosomal trafficking may be crucial when it is present at lower concentrations or in killing cells that efficiently repair perforin pores.

Cytotoxic T cells (CTL) and natural killer (NK) cells kill target cells by either of two mechanisms, both of which depend on direct effector-target cell contact. The first mechanism requires cross-linking of Fas (CD95) molecules on the target cell by its ligand on the effector cell (27). Fas-mediated apoptosis is vital for lymphoid homeostasis, and defects of this mechanism result in abnormal lymphoproliferation, failure to eliminate self-reactive T and B cells, high levels of circulating immunoglobulin, and a strong predisposition to autoimmune diseases (23). The second mechanism requires the exocytic release of cytotoxins from cytosolic granules of the effector cell toward the target cell (38). Genetically engineered deficiencies of proteins involved in this mechanism suggest that its major physiological role is to protect the host against virus infection and cellular transformation (18).

Cytolytic granules contain two principal proapoptotic mediators that jointly induce cell death. Perforin, first identified because of its membranolytic properties (22), contributes to the apoptotic response by mediating transport into the target cell of a family of serine proteases (granzymes) with which it is coreleased from the killer cell (30, 31, 36). The granzymes are then responsible for initiating the molecular events that culminate in cell death. Granzyme B (grB), which cleaves target proteins after aspartate residues (Asp-ase activity), is the most efficient mediator (30), and the CTL of mice that lack this enzyme require prolonged times to induce nuclear disintegration in their targets (12). Recently, the pathways responsible for transducing the death signal through grB and perforin have come under investigation. It is now evident that many of the lethal effects of grB are mediated through its cleavage of proapoptotic cysteine proteases (caspases) constitutively expressed as proenzymes in most cells (35, 37). GrB can activate both proximal (adapter) and distal (effector) caspases in vitro (11, 33, 37) but probably initiates the process in vivo by directly cleaving pro-caspase-3 in the first instance (48). Caspase-7 is then activated through a novel mechanism in which caspase-3 first cleaves the propeptide of caspase-7, making it susceptible to further cleavage by grB (48). While grB is clearly able to activate caspases, it can also directly and rapidly target nuclear substrates (14, 15) and cleave them at signature sites different from those used by caspases (1). Evidence also exists for caspase-independent pathways to cell death, as grB can kill cells in which caspases are irreversibly inactivated, through a mechanism that does not produce significant nuclear damage (28, 39). The key substrates responsible for caspase-independent cell death have not been defined.

Despite advances in delineating some of the downstream events induced by grB, how it gains access to its substrates in the cytosol and nucleus remains obscure. Although it has been assumed that grB enters a cell’s cytosol through transmembrane channels formed by polyperforin, evidence for this mechanism in a physiological setting is lacking, leading us to formulate an alternative model (7). We have reported that grB undergoes receptor-mediated endocytosis, but the granzyme remains innocuously confined to intracellular vesicles unless the target is also treated with sublytic perforin (7, 8, 14, 15, 32). More recent studies pointing out the ability of grB to kill cells infected with replication-defective adenovirus (without the need for perforin or any other lytic stimulus) is consistent with perforin’s role being to enable grB to escape from the endosomal compartment into the cytosol and nucleus (8), a form of viral mimicry. This suggestion was supported by our recent observations that perforin causes the very rapid redistribution of grB from the cytoplasmic vesicles into the nucleus, where it becomes highly concentrated well before the onset of nuclear apoptotic changes such as DNA fragmentation and loss of nuclear membrane integrity (14, 15, 26, 40).

In this study, we examined the potential biological function of perforin by replacing it with a number of agents that have documented membranolytic and/or endosome-disrupting (endosomolytic) properties. We demonstrate that formation of large transmembrane channels by streptolysin O (SLO) and pneumolysin (PLO) is not necessary for delivery of grB from vesicles. Furthermore, listeriolysin O (LLO), which is predominantly endosomolytic, can faithfully mimic perforin’s proapoptotic activities without inducing appreciable membrane damage. Finally, access of grB to the cytosol occurs predominantly through endosomal disruption rather than by transmembrane diffusion, even at concentrations of perforin which caused 100% Cr release from the target cell. Only at extremely high perforin concentrations (>4,000 U/ml) was it possible to identify transmembrane pores that could account for transmembrane diffusion of grB. We thus postulate that grB reaches intracellular substrates by two mechanisms, the relative importance of which may depend on the amount of perforin secreted by the effector cell and the ability of different types of target cells to adapt to perforin-mediated osmotic stress.

MATERIALS AND METHODS

Cell culture.

Jurkat human T-lymphoma cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum. FDC-P1 mouse myeloid cells and Yac-1 mouse lymphoma cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum. Interleukin-3 was additionally added to FDC-P1 cultures, as previously described (41).

Membranolytic agents.

Human perforin was purified as described previously (9). Recombinant SLO and staphylococcal alpha-toxin (SAT) were kind gifts from S. Bhakdi (4). SLO was diluted in buffer containing 1 mM dithiothreitol prior to use. Pneumococcal PLO, a kind gift from James Paton, Children’s and Women’s Hospital, Adelaide, South Australia, Australia, was activated in phosphate-buffered saline containing 2-mercaptoethanol for 30 min at room temperature. Freshly prepared serum from naive rabbits was used as the source of hemolytic complement (RC). Small aliquots of serum were stored at −70°C and discarded after each use. Heat-inactivation at 56°C for 30 min caused total inhibition of complement activity of the serum (data not shown). Recombinant LLO was a gift from D. Portnoy. A sublytic dose of the membranolytic agents was defined as that producing <10% specific release of ⁵¹Cr in a 4-h assay at 37°C. None of the membranolytic agents were inhibitory for the Asp-ase activity of grB, nor was membrane perforation by any of the agents negatively affected by caspase inhibition (data not shown).

Chemicals and reagents.

Immunoaffinity purification of human grB from nuclear lysates of YT cells was performed as described previously (42). The grB was free of grA and grM activities and perforin, as routinely demonstrated by Western blotting and peptide cleavage functional assays (42). Fluorescein isothiocyanate (FITC) labeling of grB was performed as previously described (26) and resulted in the loss of <30% of grB esterolytic activity. The oligopeptide caspase inhibitors z-Val-Ala-Asp-fluoromethylketone (z-VAD-fmk), z-Leu-Phe-fluoromethylketone (z-LF-fmk), and Phe-Ala-fluoromethylketone (z-FA-fmk) were purchased from Enzyme Systems Products, San Diego, Calif., dissolved in dimethyl sulfoxide, and stored in aliquots at −20°C. Final concentrations of Me₂SO did not exceed 0.5% in any of the assays. FITC-labeled 4-kDa dextran was purchased from Sigma (St. Louis, Mo.). FITC-labeled protein markers used in the study were bacterial azurin (8 kDa) and HS1 (9 kDa) (kindly provided by E. Williams, University of Western Australia), p13^Suc (13-kDa subunit of cyclin-dependent kinase [40]), green fluorescence protein (GFP; intrinsically fluorescent, His-tagged 27-kDa protein expressed in and purified from bacteria by Ni affinity chromatography), a 46-kDa glutathione S-transferase (GST) fusion protein expressed in pGEX3 (Pharmacia), and bovine serum albumin (BSA; 67 kDa, purchased from Sigma). In some experiments, Jurkat cells were preincubated with medium containing 10 μM brefeldin A (BFA; Sigma) for 15 min at 37°C prior to exposure to proapoptotic agents.

Assays of apoptosis.

Assays measuring the release of ⁵¹Cr and ¹²⁵I-DNA from apoptotic cells were performed as described previously, as were terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assays (39). FDC-P1 cells used in apoptosis assays were incubated in medium supplemented with interleukin-3 for the length of the experiment. Nuclear morphology was assessed by staining with Hoechst 33342 as described previously (29).

Confocal microscopy.

Cells undergoing apoptosis in response to perforin and FITC-grB were analyzed by confocal laser scanning microscopy as previously described in detail (13, 38). In some experiments, apoptosis was scored according to morphological criteria using light microscopy as previously described (14, 15, 40). Nuclear transport was quantitated as described previously (14, 15, 39), and image analysis was performed with the Macintosh NIH Image 1.49 public domain software (15).

Western blot analysis.

Antibodies to caspase-3 (Santa Cruz Biotechnology Inc.) and poly(ADP-ribose) polymerase (PARP) (Boehringer Mannheim) were purchased for use in protein blot analysis.

RESULTS

The purpose of this study was to ascertain how perforin facilitates the access of grB to its substrates and to determine whether this function could be replaced by other stimuli. On the basis that one of the best-characterized actions of perforin in vitro is its ability to induce transmembrane pores approximately 5 nm in diameter (31), we sought to replace the pore-forming function of perforin with other membrane-disruptive agents. Previous studies had shown that membrane solubilization with various detergents was inefficient at inducing apoptosis in combination with grB (32), so we studied protein toxins that form discrete transmembrane pores (SLO, SAT, PLO, and RC). In addition we used LLO, which has a potent ability to disrupt endosomes but a lesser ability to disrupt the cell membrane when applied to cells at neutral pH (3). We initially titrated the membrane-damaging effects of hemolytic complement (RC), perforin, and the bacterial toxins on Jurkat cells (Table 1) and Yac-1 cells (data not shown) and selected concentrations of each toxin that reproducibly caused minimal (0 to 10%, referred to as sublytic) or high (80 to 100%) levels of specific ⁵¹Cr release. We then used the toxins in various assays of apoptosis in the presence or absence of grB.

TABLE 1.

Membranolytic/endosomolytic agents used in this study^a

Agent	Concn producing >80% specific 51Cr releaseb	Concn producing 10% specific 51Cr releasec	Concn used in apoptosis assays
Perforin	1,000–1,500 U/ml	300–500 U/ml	130–350 U/ml
PLO	160 U/ml	10–20 U/ml	5–20 U/ml
SLO	6–12 μg/ml	0.5–1.0 μg/ml	0.02–0.5 μg/ml
SAT	3–6 μg/ml	0.1–0.2 μg/ml	0.02–5.0 μg/mld
RC	25% (vol/vol)	1.5–3.0%	1.5–25%d
LLO	100–200 ng/ml	15–30 ng/ml	1.2–1.5 ng/ml

^aThe concentrations refer to specific release of ⁵¹Cr from Jurkat cells over 4 h at 37°C (see Materials and Methods). The ranges shown were determined from three or four independent experiments performed with different batches of cells and different dilutions of the toxins. 

^bHigh concentration of toxins used in apoptotic and other assays. 

^cSublytic concentration of toxins used in apoptotic and other assays. 

^dNot permissive for apoptosis. 

Complement cannot replace the proapoptotic function of perforin.

The membrane lesions formed by complement are known to have a topology similar to that of perforin, and C9, in particular, has structural and antigenic characteristics that are reflected in significant amino acid homology to perforin, especially in a 270-amino-acid region about the center of both molecules, where the proteins share 22% amino acid identity (33). Unlike perforin, however, neither sublytic (Fig. 1) nor high (data not shown) doses of RC were able to induce DNA fragmentation in Jurkat cells coexposed to grB. These results were confirmed in TUNEL assays performed on Yac-1 cells (Fig. 2) and FDC-P1 cells (data not shown), and by Hoechst staining, which showed an absence of both nuclear and cell membrane features of apoptotic death (see below). The absence of cell death was not due to inhibition of grB by serum, as grB-mediated cleavage of an Asp-containing synthetic substrate was unaffected by coincubation with preparations of RC (data not shown).

FIG. 1.

Hemolytic complement cannot replace perforin in inducing apoptosis with grB. Release of ¹²⁵I-DNA from Jurkat cells exposed to purified human grB (0 to 1.5 μg/ml) with purified human perforin (130 U/ml; solid bars), RC (3.0% [vol/vol]; hatched bars), or no lytic agent (gray bars). The values shown are for triplicate estimates ± standard error of the mean. The experiment shown is representative of three such experiments.

FIG. 2.

Some but not all bacterial toxins are permissive for grB-mediated apoptosis. Cell death of Jurkat and Yac-1 cells was measured by positive TUNEL staining following exposure to sublytic concentrations of various pore-forming agents (Table 1) alone or in combination with purified human grB (4 μg/ml) for 90 min at 37°C. The numerals indicate the percentage of fluorescent (TUNEL-positive) cells in each panel. For each combination of lysin and cell line, the experiment depicted is representative of three to eight similar experiments. Pfp, perforin; nd, not determined.

Some but not all bacterial toxins can synergize with grB to induce apoptosis.

Many gram-positive bacteria such as staphylococci and streptococci synthesize exotoxins that kill host cells by causing membrane perforation and osmotic stress. Others, such as Listeria monocytogenes, produce membrane-disruptive agents designed to act primarily inside the cell. LLO is an important virulence factor, as it disrupts macrophage phagosomes, allowing phagocytosed bacteria access to the cytosol where efficient bacterial replication can occur (3). The optimal pH for LLO is 5.5 (approximately the pH of endophagosomes), and alkalinization of this compartment reduces bacterial proliferation by reducing bacterial escape from phagosomes (3). Despite its pH optimum, purified LLO still caused a dose-related release of ⁵¹Cr from cells exposed to it at pH 7.4 (Table 1). When we exposed Jurkat and Yac-1 cells to the combined effects of the various pore-forming agents and grB, we found that LLO, SLO, and PLO were permissive for apoptosis whereas SAT was not, whether applied at sublytic (Fig. 2) or high (data not shown) doses. Although some of these bacterial toxins may be capable of inducing apoptosis when applied alone to certain cell types (17), we saw neither evidence of DNA fragmentation (Fig. 2) nor any morphological consequences of apoptosis by Hoechst staining (Fig. 3) in the absence of grB. However, when grB was present with SLO, PLO, or LLO, typical morphological changes of apoptosis including intense cell membrane blebbing, chromatin condensation, and nuclear degradation were obvious in many cells (Fig. 3).

FIG. 3.

Apoptosis induced by pore-forming bacterial toxins in combination with grB, determined by Hoechst staining of Jurkat cells exposed to sublytic quantities of various lytic agents (Table 1) in the presence or absence of grB (4 μg/ml) for 60 min. (A) GrB alone; (B) perforin alone; (B′) perforin with grB; (C) SLO alone; (C′) SLO with grB; (D) PLO alone; (D′) PLO with grB; (E) LLO alone; (E′) LLO with grB; (F) SAT alone; (F′) SAT with grB. The arrows indicate typical morphological changes of apoptosis, including nuclear collapse induced by perforin plus grB (B′) or by SLO plus grB (C′), and intense membrane blebbing induced by PLO plus grB (D′). Original magnification, ×400.

It has previously been demonstrated that nuclear apoptotic changes induced by grB and perforin depend on caspase activation, as these parameters can be blocked by viral serpins (28) and oligopeptide caspase inhibitors such as z-VAD-fmk (28, 29, 37). We therefore preincubated target cells with caspase inhibitors prior to exposure to the permissive toxins and grB. Asp-Glu-Val-Asp-fluoromethylketone (DEVD-fmk; an inhibitor of caspase-3 and related caspases) and z-VAD-fmk inhibited the cleavage of 32-kDa pro-caspase-3 to its active p19-p12 heterodimeric form induced by either PLO or perforin in combination with grB, whereas the chymase inhibitor z-FA-fmk had no effect (Fig. 4). Since the fmk inhibitors do not react with grB, cleavage of pro-caspase-3 by grB at Asp¹⁷⁵ was still apparent (35, 48). The nuclear damage induced by perforin or bacterial toxins and grB was accompanied by cleavage of PARP, which was also caspase dependent (Fig. 4). We also found that DNA fragmentation in response to all the permissive toxins and grB was totally inhibited by blockage of the caspase cascade, as TUNEL-positive cells were almost totally abolished in the presence of z-VAD-fmk or DEVD-fmk (Fig. 5). However, the caspase inhibitors failed to protect Jurkat and Yac-1 cells from the grB-dependent nonnuclear form of apoptosis when the granzyme was delivered by either perforin or bacterial toxins. The cells still became markedly shrunken (reduced forward scatter) and much more granular (increased side scatter) despite preincubation with the caspase inhibitors (Fig. 6) and were not rescued from cell death when put back into cell culture after treatment (data not shown). Thus, as has been shown for perforin- and grB-induced apoptosis (39), caspase inhibition had no effect on overall cell survival in response to the permissive toxins but did largely abolish the nuclear changes of apoptosis. Nuclear accumulation of grB in the presence of the permissive toxins was also inhibited by Bcl-2 expression (data not shown), as has been demonstrated for perforin-mediated cell death (16). Overall, our data strongly suggest that the same or closely related pathways to nuclear and nonnuclear apoptosis were being activated by perforin and the permissive bacterial toxins alike.

FIG. 4.

Cleavage of PARP and pro-caspase-3 in response to grB and lytic toxins is caspase dependent. Shown is Western blot analysis of whole-cell extracts of Jurkat cells exposed to perforin or LLO in the presence (+) or absence (−) of grB (4 μg/ml) for 60 min at 37°C. Cells were incubated with the fmk inhibitors indicated (20 μM) at 37°C for 30 min or with no inhibitor (control) prior to exposure to pfp or LLO ± grB. The Western blots were probed with antiserum detecting PARP or caspase-3 (see Materials and Methods). Cleavage products, including the 19-kDa caspase-3 chain indicative of active enzyme, are indicated with arrowheads. Numerals at the left indicate the migration of molecular size markers in kilodaltons.

FIG. 5.

GrB-mediated DNA fragmentation induced in the presence of perforin or permissive bacterial toxins is caspase dependent. The percentage of Jurkat cells showing positive TUNEL staining is indicated in the absence (−) or presence (+) of grB (4 μg/ml) and the toxin indicated (sublytic concentrations [Table 1]) for 60 min. Cells were preincubated in medium containing fmk inhibitors at 37°C for 90 min prior to assay. The assay shown is representative of three such experiments. ∗ not determined.

FIG. 6.

DNA fragmentation of Jurkat cells in response to grB in combination with sublytic concentrations of perforin or SLO is caspase dependent, but other morphological indicators of cell death are not. Cells were exposed to either the lysin alone (A) or the lysin with grB (B) following preincubation in medium containing the indicated fmk inhibitor (20 μM) for 90 min at 37°C. For each determination, the panel at the left shows TUNEL staining (FL, fluorescence), while the panel at the right is a plot of forward scatter (FS, indicative of cell size) against side scatter (SS, indicative of cell granularity). In each case, cells positive for TUNEL staining appear in the scatter profile as a collection of light gray dots.

Endosomolysis by LLO mimics perforin-mediated nuclear targeting of grB.

The ability of LLO to induce leakage of proteins from the endosomal compartment has been utilized to target the delivery of polypeptides into the cytosol for immunological detection and other purposes (13, 20). When LLO was added externally to Jurkat cells at neutral pH, concentrations of ∼15 ng/ml were necessary to induce discernible release of ⁵¹Cr (Table 1). However, far lower concentrations (0.75 to 1.5 ng/ml) readily disrupted endosomes containing FITC-grB, enabling it to access the nucleus (Fig. 7), without demonstrable release of ⁵¹Cr. When the endosomolytic effect of LLO was inhibited with the weak base ammonium chloride (10 mM), grB remained in a vesicular distribution (Fig. 7A). The absence of significant cell membrane pore formation under these conditions was verified by showing that LLO-treated cells excluded 4-kDa FITC-dextran, even at concentrations of toxin reaching 10 ng/ml (Fig. 7B and C). Despite the absence of transmembrane pores, grB accumulated in the nucleus (ratio of nuclear to cytoplasmic fluorescence [Fn/c] of ∼1.5) in the presence of LLO by 25 min (Fig. 7C). However, when the cells were incubated with ammonium chloride, grB remained principally within the cytoplasm (Fn/c of ∼0.75; [Fig. 7A]). Although LLO facilitated delivery of grB to the nucleus, the cytoplasmic levels of grB also rose by 50 min (ratio of fluorescence in the cytoplasm to fluorescence in the medium [Fc/Fmed] of ∼1.2), indicating that further uptake of grB was also stimulated by the toxin (Fig. 7B).

FIG. 7.

Perforin can be replaced by LLO to mediate grB cellular uptake and nuclear translocation, both of which are inhibitable by ammonium chloride. (A) Representative confocal microscope images of Jurkat cells exposed to FITC-labeled grB (4 μg/ml) or 4-kDa FITC-dextran in the presence or absence of LLO (10 ng/ml) at 37°C for 15 min. Cells in the lower right panel were preincubated with medium containing 10 mM ammonium chloride prior to assay. (B) Fc/Fmed ratio of Jurkat cells exposed to LLO (10 ng/ml) in the presence of 4-kDa FITC-dextran or FITC-grB at 37°C for the times indicated. In these experiments, an Fc/Fmed of <0.6 indicates no significant uptake of fluorescent molecules into the cytoplasm. (C) Fn/c ratio of Jurkat cells exposed to FITC-grB (4 μg/ml) and LLO (10 ng/ml) at 37°C for the times indicated. Each bar represents the mean of six to eight separate measurements for each of Fc, Fmed, and Fn after subtraction of autofluorescence. The standard error of the mean in each case was no more than 10% of the mean.

We have previously shown that addition of perforin with grB to Jurkat (and other) cells results in rapid redistribution of grB from the cytoplasm to the nucleus, together with further rapid endocytic uptake of grB (7, 8, 14, 15, 40); cells showing this pattern of grB redistribution are destined for rapid apoptotic death (39, 40, 43). Further evidence for the importance of vesicular trafficking in the redistribution of grB was gained from experiments in which target cells were pretreated with BFA before exposure to grB in the presence or absence of perforin. BFA is known to interfere with the redistribution of proteins out of the endosomal system (19), probably by inhibiting the recruitment of cytoplasmic coat complexes including members of the COP-1 and ARF families (2, 46). This causes cisternal elements of the Golgi stack to fuse with the endoplasmic reticulum, while the trans-Golgi network fuses with endosomes (47). Although anterograde protein transport out of the Golgi and vesicular secretion from the cell are blocked, BFA does not interfere substantially with targeting of proteins to lysosomes or with receptor-mediated recycling (19); thus, transferrin uptake and intracellular iron delivery are unaffected by BFA treatment (21). Accordingly, Jurkat cells exposed to BFA (10 μM) for 20 min prior to incubation with FITC-grB alone showed no difference from mock-treated cells in overall grB uptake over 90 min (as indicated by a similar Fc/Fmed ratio at 15, 45, and 90 min) (Fig. 8A, left). These results indicated that as with transferrin uptake, endocytic uptake of grB across the plasma membrane was not influenced by BFA treatment. As expected, no nuclear accumulation of grB was seen in the absence of perforin, as indicated by the static Fn/c ratio (Fig. 8C, left). In contrast, the coaddition of perforin resulted in rapid accumulation of grB in the nucleus of mock-treated cells (Fn/c of ∼1.3 and 1.6 at 15 and 45 min, respectively); BFA pretreatment, however, virtually abolished the redistribution of grB from the cytoplasm to the nucleus (Fn/c of ∼0.9 at 90 min [Fig. 8B, right]) and prevented additional perforin-induced uptake of grB into the cytoplasm (Fig. 8A, right). Preincubation of cells with agents that increase endosomal pH (ammonium chloride and bafilomycin A) also slowed perforin-mediated grB redistribution to the nucleus (data not shown), again implying that perforin functions by disrupting endosomal trafficking. Treatment with these agents had no significant effect on pore formation by perforin (data not shown).

FIG. 8.

BFA inhibits perforin-mediated enhancement of grB cellular uptake and nuclear translocation. (A) Fc/Fmed ratio of Jurkat cells exposed to perforin (400 U/ml; right) or to no perforin (left) in the presence of FITC-grB (4 μg/ml) at 37°C for the times indicated. (B) Fn/c ratio of Jurkat cells exposed to FITC-grB (4 μg/ml) with or without perforin (400 U/ml) at 37°C for the times indicated. In each case, cells were either pretreated in medium containing BFA (10 ng/nl) for 20 min at 37°C or incubated without BFA (mock). Each bar represents the mean of six to eight separate measurements for each Fc, Fmed, and Fn after subtraction of autofluorescence. The standard error of the mean in each case is shown.

Lytic concentrations of perforin and the permissive toxins facilitate intracellular delivery of grB despite the absence of large transmembrane defects.

A large body of evidence indicates that efficient apoptosis in response to purified perforin and grB can be readily achieved with a quantity of perforin that causes minimal lysis (30–32, 38, 40). Indeed, we have previously shown that with these sublytic quantities of perforin, the membrane perforations formed are too small to admit a 13-kDa protein, p13^Suc (40), despite the observation that much larger granzymes (32-kDa grB and 65-kDa grA) are transported to the nucleus in as little as 2 min and induce rapid apoptotic death (14, 15, 40, 41). The data for grB- and LLO-mediated apoptosis also strongly support the notion that diffusion of grB through cell membrane defects is not a prerequisite for apoptosis. However, these studies did not address the possibility that perforin might be delivered in much higher than sublytic concentrations in a physiological situation where an effector-target cell conjugate forms. This raises the question whether endosomolysis is physiologically relevant when perforin is delivered in quantities that cause greater osmotic stress. The amount of perforin delivered to the surface of a target cell by a CTL has not yet been determined either in vitro or in vivo and could conceivably vary depending on the number of times the cell has previously degranulated. In addition, some cell types might compensate for the osmotic stresses of even large quantities of perforin by membrane budding or by stimulating vigorous endocytic repair of the cell membrane.

To address this issue, we examined whether grB presented to cells with high concentrations of perforin achieved intracellular trafficking by endosomal disruption or by direct entry into a cell via membrane perforations. To do this, we determined the functional pore size (FPS; defined as the molecular mass of the largest fluorescent protein able to enter cells over the course of the experiment) (Table 2) of the cell membrane at escalating concentrations of perforin. Cells exposed to perforin at 1,600 U/ml (sufficient to cause 100% ⁵¹Cr release [Table 1]) were able to exclude 8-kDa azurin, 9-kDa HS1, and 27-kDa GFP for longer than 70 min (Fig. 9 and 10), indicating an FPS of less than 8 kDa. In previous experiments using sublytic quantities of perforin with FDC-P1 cells, we had observed the relatively weak uptake of only 8-kDa azurin and complete exclusion of p13^Suc (40), indicating that the membrane lesions formed with 1,600 U per ml of perforin were functionally similar. When we examined the ability of FITC-grB to target the nucleus in the presence of 1,600 U per ml of perforin, nuclear accumulation of FITC-grB was still clearly evident [Fn/c of >1.0 by 10 min and ∼1.9 by 70 min [Fig. 10]). Therefore, we were able to conclude that (i) the endosomolytic function of perforin can operate even when it is present at high concentrations and (ii) endosomolysis can predominate over direct transmembrane access by grB, despite considerable osmotic stress being applied to the target cell. Using the same approach, we also determined the FPS in Jurkat cells following exposure to various concentrations of PLO (Table 2) and SLO (data not shown). Although the membrane perforations formed by high concentrations of the toxins excluded the 8- and 9-kDa proteins, both toxins induced rapid nuclear targeting of 32-kDa FITC-grB (and 65-kDa FITC-grA [data not shown]) when used at the same concentrations. The results of these experiments with high-dose pore-forming agents confirmed that diffusion of grB across the cell membrane need not contribute significantly to the cytosolic pool of grB, even when the cell membrane experiences a major osmotic stress.

TABLE 2.

FPS induced by various concentrations of perforin and PLO in Jurkat cells compared with concentrations producing minimal or near-maximal Cr release

Lysin	Concn (U/ml)	FPSa (kDa)	Minimum concn producing >80% specific 51Cr release2 (U/ml)	Minimum proapoptotic concn (U/ml)
Perforin				130
	200	<8
	400	<8
			1,000
	800	<8
	1,600	<8
	4,000	>27
	5,000	>67
PLO	2	<8
				5
	6	<8
			160
	160	<9
	800	>67

^aSee Results for definition. 

^bAs indicated in Table 1, and interpolated into the FPS data above, for purposes of comparison. 

FIG. 9.

Determination of FPS in the cell membrane of Jurkat cells. Cells were exposed to various concentrations of perforin for the times indicated in the presence of FITC-labeled proteins of the sizes indicated. In the presence of 800 or 1,600 U per ml of perforin, the 8-kDa azurin marker was excluded from cells for the length of the experiment, whereas both the 46-kDa GST fusion protein and 67-kDa BSA size markers were able to equilibrate across the cell membrane at the highest perforin concentrations (5,000 U/ml).

FIG. 10.

High concentrations of perforin enhance grB uptake and nuclear translocation but do not permit uptake of small proteins. Shown are Fc/Fmed and Fn/c ratios of Jurkat cells exposed to FITC-grB (4 μg/ml) or to similar concentrations of GFP, FITC-HS1, or FITC-azurin with perforin (1,600 U/ml) for the times indicated. Each bar represents the mean of six to eight separate measurements for each of Fc, Fmed, and Fn after subtraction of autofluorescence. The standard error of the mean in each case was no more than 10% of the mean.

To determine the dose of perforin required to allow the free diffusion of granzymes into the cell, we exposed Jurkat cells to even higher concentrations of perforin. At 4,000 U/ml, we saw cytoplasmic uptake of 8-, 9-, and 27-kDa proteins within 25 min but continuing exclusion of 46-kDa GST for over 50 min, indicating an FPS of between 27 and 46 kDa at this perforin concentration (Table 2). When the perforin concentration was raised to 5,000 U/ml, both GST and BSA (67 kDa) also entered the cytosol (Fig. 9; Table 2), indicating an FPS of >67 kDa. We concluded that perforin concentrations greater than ∼4,000 U/ml produced pores that could directly admit a protein the size of grB into the cytosol of Jurkat cells.

DISCUSSION

This present study was intended to provide insight into the proapoptotic functions of perforin. Gene knockout studies indicate a pivotal role for perforin in cell-mediated immunity, as congenital perforin deficiency cannot be compensated for and perforin-deficient mice have a reduced ability to survive certain viral infections, do not reject some major histocompatibility complex-mismatched tissue and tumor allografts, and lack NK cytolytic function (5). In contrast, deficiency of either or both granzymes A (6, 34) and B (12) results in a far milder immune deficit. Collectively, the data imply that perforin contributes to granule-mediated apoptosis by delivering a family of granule proteases which possess considerable functional redundancy despite their different proteolytic specificities. Two principal hypotheses have been proposed to account for perforin’s role in providing access into the cytosol for grB and other granule-bound toxins. The first proposed that perforin provides large channels for passive diffusion of grB into the cytosol; alternatively, perforin has been postulated to disrupt endosomes containing grB (the “facilitated access” hypothesis) (38, 43). While the former mechanism may operate at perforin concentrations of >4,000 U/ml, our data here showing that BFA, an inhibitor of endosomal trafficking, abrogates perforin-mediated enhancement of grB cellular uptake and nuclear translocation provide strong evidence for the second hypothesis, when perforin is present at concentrations at least as high as 1,600 U/ml. A third possibility recently canvassed (43; see also reference 16) is that perforin does not directly affect the localization of grB but rather generates a membrane signal that exerts an effect downstream of grB entering the cytosol. There is no direct experimental evidence to support this likelihood, and we have observed that perforin insertion into membranes does not appear to be accompanied by significant, reproducible phosphorylation events (6a). Furthermore, the fact that the permissive toxins SLO, LLO, and PLO all utilize cholesterol to attach to cells (3), while perforin requires phosphorylcholine (44), makes it less likely that docking of perforin to a specific signalling moiety is of functional importance.

As no previous study has systematically examined the requirements for perforin-mediated pore formation and/or endosomal disruption in concert with grB, we sought to examine the relative importance of these two mechanisms under conditions where the amount of osmotic damage could be regulated by varying the dose of perforin. The findings we present here are significant because they clearly indicate that transmembrane pores large enough to permit passive diffusion of grB into the cytosol do not have to be present for apoptosis, either when perforin is present in limiting (sublytic) concentrations (defined as Cr release of <10%) or, surprisingly, when it is present in abundance (Cr release of >80%). Even at high concentrations of perforin and PLO, we found that the cell membrane lesions induced were not sufficiently large (<8 kDa for PLO and perforin [40]) to account for the passage of grB through the plasma membrane. Moreover, we have recently demonstrated that grB is secreted from cytotoxic cells as a macromolecular complex with the carrier proteoglycan, serglycin, and that targets pulsed with the complexes undergo apoptosis after the addition of perforin (10). On the basis of the FPS generated by perforin in vitro and the large complexes that require entry, the transmembrane channel model for granzyme delivery becomes even less attractive. Supporting the feasibility of endosomal disruption, we showed that LLO, which has well-characterized endosomolytic properties, can substitute faithfully for perforin to facilitate grB-mediated apoptosis without measurable transmembrane pore formation. The fact that BFA, ammonium chloride, and bafilomycin, agents inhibiting endosomal trafficking, all inhibited grB subcellular redistribution without blocking grB uptake or perforin-mediated pore formation (data not shown) lends further credence to the argument that disruption of vesicular trafficking is a key function of perforin. Only when extremely high concentrations of perforin (>∼4,000 U/ml) were present was it possible to identify functional pores that were large enough to account for transmembrane entry of grB into the cell. Overall, our findings imply that endosomal disruption by perforin can predominate over cell membrane pore formation under most circumstances. However, it is possible that pore formation becomes relevant when perforin is present in extremely high concentrations.

We speculate that the ability of granzymes to gain access to their substrates through two alternative pathways may confer advantages to the host under specific circumstances (Fig. 11). For example, it is possible that a cell’s stores of perforin become depleted following several consecutive degranulations, and it may become progressively more difficult to inflict osmotic damage on successive targets. Under such circumstances, an ability to achieve grB delivery through other than a purely osmotic mechanism would provide the CTL with an alternative mechanism for inducing apoptotic death. An alternative scenario arises with target cells that may be very adept at limiting membrane damage by activating repair mechanisms to restore their osmotic balance. In addition to shedding damaged membrane buds or vesicles, an important part of this process may involve stimulating endocytosis, to limit exposure of “leaky” areas of the membrane to the external milieu. Again, the CTL might capitalize on this process, using it to augment granzyme-mediated delivery to the target cell within endosomes. Unfortunately, no study has yet estimated what range of perforin concentrations is delivered into the cleft between given effector and target cells and to what extent its lytic effects can be neutralized by a metabolically active target cell. Accurate measurements of these phenomena would be required to place our findings in a fuller pathophysiological context.

FIG. 11.

Two hypothetical but possibly complementary models of perforin-granzyme synergy. (A) Passive diffusion. In this model, perforin monomers (large ellipses) elongate, attach to phosphorylcholine headgroups in the plasma membrane, coalesce with like molecules, and intercalate the membrane to form transmembrane pores for the passage of apoptotic mediators such as grB (small ellipses) into the target cell cytosol. Free grB can then process proapoptotic substrates, leading to apoptotic cell death. We have found evidence for this mechanism only when perforin is present in very high concentrations, i.e., >4,000 U/ml. (B) Facilitated access. GrB is taken up within an endocytic vesicle after interaction with a specific receptor in the plasma membrane (small rectangle). Perforin destabilizes grB-containing endosomes by an unknown mechanism, perhaps by entering the cell in separate vesicles which then fuse with grB-carrying endosomes (as shown, hypothetically), again facilitating grB access to substrates associated with apoptosis induction. We have found evidence for this mechanism at perforin concentrations as high as 1,600 U/ml (see text for further details).

Given that perforin is indispensable for both the caspase-dependent and caspase-independent pathways to cell death, it is remarkable that perforin can be replaced by pore-forming agents with such dissimilar structures and mechanisms of polymerization. The three permissive molecules LLO, PLO, and SLO are all members of the family of thiol-activated toxins, whose pore-forming ability depends on the highly conserved stretch of 11 amino acids at their C termini, containing Trp residues crucial for pore formation and a Cys residue that renders them intolerant of oxidizing conditions (24). Based on the knowledge that the integral function of LLO is to permit phagocytosed bacteria to exit from endophagosomes and that this process occurs most efficiently at acidic (endosomal) pH, it might have been predicted that LLO could also effectively deliver grB to the cytosol in a pH-dependent manner. However, a surprising result from this study was that PLO and SLO, which are most active at neutral pH, also delivered grB despite forming membrane perforations too small to enable direct grB access into the cytosol. These results suggest that the mechanism through which SLO and PLO deliver grB as well as other macromolecules such as immunoglobulin G may not necessarily be reliant on plasma membrane perforation, raising the need to reevaluate some of the biologic effects of these toxins on eukaryotic cells. Interestingly, SAT, which could not deliver grB, is completely different in structure and mode of action from the thiol-activated toxins and forms very small pores (typically 1.5 nm in diameter) (5, 45). Although this might suggest a requirement for pores above a critical size, this interpretation is not consistent with our observation that complement, which is very similar to perforin with respect to diameter of pores formed (>5 nm) and overall topology (29), failed to deliver the granzyme. Overall, our data indicate that neither pore diameter nor overall shape and size of membrane pores per se have a significant bearing on the ability of toxins to facilitate intracellular delivery of grB.

How then, might these diverse agents permit grB access to the cytosol? We propose that following insertion in the plasma membrane, the pores are subjected to clearance by endocytosis. It is not yet clear whether the pores are removed by classical clathrin-dependent endocytosis or by alternative pathways, but there are at least two ways in which the granzyme might reach the cytosol. The internalized porin might be incorporated into the same coated vesicle as grB, rendering the vesicle unstable and releasing the granzyme, but since the pores formed either by perforin (at concentrations of <1,600 U/ml) or the bacterial toxins exclude proteins significantly smaller than grB, such a hypothesis becomes less tenable. Alternatively, the internalized pore-forming protein could disrupt the fidelity of vesicular fusion, leading to the release of the granzyme when the vesicle containing the porin and grB coalesces with another intracellular vesicle. Regardless of the mechanism of vesicle release, an obvious prediction of postulating an endosomolytic function for perforin is that it should be possible to identify perforin in the cytoplasm of the target cell and not just within the intracellular cleft and at the target cell membrane as previously demonstrated (25). Immunohistochemical evidence for the presence of perforin within cells is lacking, however, possibly due to the use of antibodies that were not capable of detecting polyperforin. If indeed perforin does not enter cells, it must be concluded that plasma membrane-inserted perforin can modulate endosomal stability or function, possibly through ion fluxes or by inducing changes in membrane lipid turnover or composition (perhaps prior to the grB endocytosis step), such that endosomal exit by internalized grB is facilitated. A high priority of our laboratories is to develop experimental approaches to distinguish these possibilities.