Entry and Trafficking of Granzyme B in Target Cells During Granzyme B-Perforin–Mediated Apoptosis

CYTOTOXIC T LYMPHOCYTES (CTL) are the effectors in cell-mediated immunity that destroy virus-infected and neoplastically transformed cells.1-4 After adhesion of CTL and target cell, the plasma membranes interdigitate and cytolytic granules migrate toward the site of contact, releasing their contents into the apposed space. This secretory process forms the basis of the lethal hit model of granule-mediated cytotoxicity in which it is postulated that the molecules contained within the cytoplasmic granules are the lytic effectors that act in concert to bring about target cell apoptosis.5 

A number of proteins have been isolated and characterized from cytolytic granules. The pore-forming protein, perforin,6 is a granule protein that is believed to insert into the membrane of the target cell, providing a channel that facilitates the entry of other granule components into the cytoplasm of the target cell.7Granzyme B, also known as cytotoxic cell proteinase 1 (CCP18) or fragmentin 2,9 is the prototypic member of the family of serine proteinases that reside in the cytolytic granules of CTL, natural killer (NK), and lymphokine-activated killer (LAK) cells,8,10 collectively termed granzymes. Although perforin alone causes necrotic cell death by osmolysis, apoptosis is mediated by the concerted actions of perforin and the granzymes.

Several laboratories have provided evidence for the synergistic requirement of granule proteins in stimulating apoptosis in target cells. Granzyme B and perforin have been shown to bring about rapid DNA fragmentation in target cells by a mechanism that is attributed to granzyme activity.9,11 Similarly, transfection of the rat basophilic cell line, RBL, with perforin and granzyme B renders the RBL capable of cytolytic activity when directed against a number of target cell lines.12,13 Conclusive evidence for a role of granzymes in target cell apoptosis has come from the generation of granzyme B-deficient mice. The CTL of granyzme B−/− mice display a severely reduced ability to induce rapid DNA fragmentation in target cells.14,15 

In support of the notion that the granzymes induce apoptosis, in part, by cleaving cytosolic proteins, granzyme B has been reported to activate certain ICE/CED-3 cysteine proteinases, or caspases,16 that are closely linked to apoptotic cell death (reviewed in Martin and Green17 and Nagata18), including caspases 3¹⁹ and 7.20,21 Caspase-3 is processed in murine targets treated with CTL,19 and targets exposed to granzyme B and perforin contain processed caspases 1,22 3, and 7.21 Thus, it is implied that granzyme B gains access to the interior of the target cell where it can cleave and activate these substrates. Activation of the caspases has been shown to be essential in DNA fragmentation,23 for externalization of phosphatidylserine in the plasma membrane of cells undergoing apoptosis,24 and for degradation of actin via gelsolin activation.25 In addition to the cytoplasmic substrates of granzyme B, there is an increasing body of evidence suggesting that granzyme B may act in the nucleus. In two independent studies, granzyme B was shown to bind to nuclear proteins.26,27 At least one of these potential substrates is associated with heterochromatin and the perinucleolar region of the nucleus.26 

Despite major advances in our understanding of the mechanisms of granule-mediated CTL killing, there has been no direct demonstration of transfer of molecules from cytolytic granules to target cells and no information on the resultant trafficking of CTL proteins once they have gained entry into the target cell. Previously, we have shown that granzyme B binds specifically to the surface of Jurkat cells, but apoptosis did not occur until the subsequent addition of perforin. The absolute requirement for the latter could be circumvented by the treatment of the cells with a replication-deficient adenovirus (Ad2). We hypothesized that granzyme B was taken into cells by endocytosis and that the Ad2 or perforin caused release of granzyme into the cytoplasm.28 

To test our hypothesis, we designed experiments to directly visualize granzyme B as it is internalized into the cell and elucidate its intracellular trafficking. Using confocal laser scanning microscopy (CLSM), we demonstrate that granzyme B does indeed enter the cell independently of perforin. It appears to be transiently associated with rab5⁺ vesicles, but ultimately accumulates in a novel compartment. After treatment with perforin or Ad2, the granzyme B is released into the cytoplasm, translocates rapidly to the nucleus, and apoptosis ensues. In contrast, direct injection of granzyme B into the cytoplasm induces death directly, without the requirement for perforin.

Antisera was raised against residues 9-16 of granzyme B (provided by Dorothy Hudig, University of Nevada, Reno) and used as previously described.26 Secondary antibodies goat antirabbit-fluorescein isothiocyanate (FITC), goat anti-rabbit–Texas Red, donkey anti-mouse–Texas Red, and mouse anti-rabbit rhodamine were purchased from Jackson Immunoresearch (West Grove, PA), rab5 (S-19) and rab4 (D-20) polyclonal antisera from Santa Cruz Biotechnology (Santa Cruz, CA), rab5 monoclonal antibody (clone #15) from Transduction Laboratories (Lexington, KY), cathepsin D antisera from Upstate Biotechnology, Inc (Lake Placid, NY), and cytochrome c monoclonal antibody from Pharmingen (San Diego, CA). All antisera were used at dilutions of 1:50 to 1:100 unless otherwise noted. Granzyme B was fluoresceinated according to the protocol described by Jans et al.29 

Human perforin and granzyme B were purified as previously described.28,30 Type 2 adenovirus was purified according to Seth et al.31 Granzyme B was added directly to Jurkat target cells at 1 μg/mL in RPMI supplemented with bovine serum albumin (BSA) (0.05% wt/vol). Activity of granzyme B was 120 U/μg where 1 U is the activity of enzyme required to hydrolyze 1 nmol/minute of the BAADT substrate. Sublytic doses of perforin were used at 90 U/mL, where 1 U is defined in the standard sheep red blood cell hemolytic assay. Cells were washed and resuspended four times in medium before further treatment.28 Volumes equivalent to the incubation volume were used for each wash. Infection with adenovirus (Type 2) was performed in RPMI with BSA (0.05%), as noted above for granzyme B at a multiplicity of infection of 10 plaque-forming units per cell. COS M5 cells were transfected to express granzyme B in the pAX142 cloning vector32 as previously described.33 Transiently transfected COS M5 cells were treated with adenovirus or perforin as described above without prior treatment of granzyme B. All cells were incubated at 37°C unless otherwise noted.

Apoptosis of target cells was assessed by terminal deoxytransferase nick end labeling (TUNEL) reaction to measure DNA fragmentation.34 Fluorescein-deoxyuridine triphosphate (dUTP) was incorporated at the sites of DNA nicks. The DNA-binding dye 4,6-diamidino-2-phenylindole dihydrochloride (DAPI; Sigma Chemical Corp, St Louis, MO) was added to saponin permeabilized cells at 1 μg/mL for at least 15 minutes before viewing by fluorescence microscopy. DAPI was used as a label for nuclear morphology.

Cell microinjection was performed on the stage of a Nikon Diaphot inverted microscope (Nikon, Melville, NY) using an Eppendorf pressure injector (Model 5246; Brinkmann Instruments, Inc, Westbury, NY) and micromanipulator (Model 5171). Microinjection needles (approximately 0.1 μm inner diameter) were pulled from glass capillaries using a horizontal electrode puller (Model P-97; Sutter Instrument Co, Novato, CA) and loaded using Eppendorf microloaders. Enzymatically active granzyme B and granzyme B inactivated with antigranzyme B were diluted in 0.15 mol/L NaCl. To identify injected cells, the injectate contained 0.3% Texas Red Lysine Fixable (Molecular Probes, Eugene, OR). The concentration of purified human granzyme B was 3.23 mg/mL, which corresponds to 119.4 U/μL. Solutions were injected into the cytoplasm of MCF-7 cells, which were plated on glass cellocate cover slips (Eppendorf) 18 hours before injection. Medium (RPMI 1640, 10% fetal bovine serum [FBS], 200 μg/mL G418, and 100 μg/mL hygromycin) was changed before and after injection. Minaschek et al,35 using similar equipment, demonstrated that the injection parameters used in the present study (pressure, 100 hecto Pascal; time 0.5 seconds) delivers approximately 0.05 pL into the cytosol of 3T3 cells. We estimate the volume of MCF-7 cells to be 5 pL, similar to that of 3T3 cells. Thus, intracellular concentrations of the injectate (Table1) are estimated to be 1% of the pipette concentration, reflecting a 100-fold dilution in the cell.

Jurkat cells were attached directly to glass microscope slides by centrifugation for 15 seconds at 600g. COS cells were grown directly onto 0.5-mm glass coverslips. All cells were fixed immediately in paraformaldehyde (2% wt/vol in phosphate-buffered saline [PBS]) and washed in PBS before permeabilization and immunolabeling. Cells were labeled as previously described26 and viewed with a Zeiss fluorescence microscope (Carl Zeiss, Inc, Thornwood, NY) or by confocal laser scanning microscopy and analyzed with the accompanying CLSM software (CLSM; Leica, Heidelberg, Germany). Images were acquired by 32 or 64 line scan averaging using 100×/1.32N.A. objective under oil immersion.

HeLa cells were incubated in medium containing granzyme B (1 μg/mL) for 60 minutes at 4°C, washed on ice, and placed at 20°C or 37°C for 45 minutes. Immediately after incubations, cells were fixed in 4% paraformaldehyde at 4°C for 5 minutes, then 60 minutes at room temperature. Cells were washed in PBS and dehydrated in increasing concentrations of ethanol. Dehydrated pellets were infiltrated with Lowicryl K4M:ethanol at −20°C followed by embedding at −20°C according to manufacturer's directions (Polysciences Inc, Warrington, PA).

Ultrathin sections (40 to 70 nm) were placed on carbon/parlodion-coated copper grids. Grids were blocked sequentially with double distilled water (1 minute), 0.01 mol/L glycine (5 minutes), and PBS containing 1% BSA and 0.1% gelatin (20 minutes). Labeling was performed at room temperature for 2 hours with antibody diluted in final blocking solution. After washing in blocking solution, grids were incubated in gold-conjugated secondary antisera (Jackson Immunoresearch) for 45 minutes at room temperature. After thorough washing in blocking solution, grids were rinsed in distilled water, counterstained in uranyl acetate and lead citrate, and viewed with a Philips 420 electron microscope (Philips Electron Optics, Eindhoven, The Netherlands).

Perforin has long been thought to be absolutely required to facilitate the entry of granzyme B into target cells. To test this hypothesis, we used direct and indirect fluorescence labeling techniques and confocal microscopy to visualize granzyme B when target cells were treated with the proteinase in the absence of perforin. Jurkat cells were incubated in medium containing soluble granzyme B at 37°C for up to 120 minutes. After washing and permeabilization, they were treated with antigranzyme antiserum, followed by a Texas Red–conjugated secondary antibody and analyzed by CLSM. Figure 1shows the accumulation of soluble granzyme B inside target cells. Within 60 minutes, there was a significant staining for granzyme B in a distinct intracellular pattern compared with immunolabeled control cells not exposed to granzyme B (Fig 1A).

We also studied the ability of granzyme B alone to trigger apoptosis in target cells. Jurkat, treated with granzyme B, were labeled by the TUNEL technique to assess DNA fragmentation and analyzed by CLSM. These experiments were also performed in YAC-1 cells with analysis of nuclear morphology, after DAPI staining, by fluorescence microscopy (Fig 1B and C). Cells labeled by either TUNEL or DAPI were also immunolabeled to detect granzyme B. No DNA fragmentation or nuclear condensation was observed under these conditions and the cells appeared entirely normal (Fig 1B and C). Despite the accumulation of granzyme B within target cells, we did not observe DNA fragmentation or nuclear condensation, two hallmarks of apoptosis, even upon prolonged incubation.

Jurkat and YAC-1 target cells were treated concurrently with granzyme B and sublytic doses of perforin. As shown by TUNEL label and DAPI stain, this resulted in rapid DNA fragmentation and nuclear condensation within 30 minutes. Targets were also incubated in medium containing granzyme B for 60 minutes and washed extensively before being treated with sublytic doses of perforin. Figure 1B (TUNEL label) and C (DAPI stain) show that cells treated in this two-step sequential manner also showed significant DNA fragmentation, severe nuclear condensation, and a drastic reduction in size. This was in stark contrast to the results seen with cells exposed to granzyme B alone. We conclude that although perforin is dispensable for the internalization of granzyme B, it is absolutely required to deliver a death signal to target cells containing granzyme B.

Internalized granzyme B showed a punctate cytoplasmic staining, but the cells were clearly not apoptotic. We therefore asked if either perforin or Ad2 influenced the intracellular distribution of granzyme. Jurkat were preloaded with granzyme B for 60 minutes at 37°C, and then perforin or Ad2 was added. The targets were fixed and immunolabeled for granzyme B. Simultaneous TUNEL label with dUTP-FITC was used to assay apoptosis and cells were viewed by CLSM.

As shown in Fig 2A (see page 1048), when perforin was added to granzyme-containing targets, we no longer observed a punctate granzyme B pattern. Rather, granzyme was detected in the nucleus concurrent with the onset of DNA fragmentation in TUNEL-positive cells. A similar result was obtained when granzyme B-containing Jurkat were subsequently treated with Ad2. In both of these systems, we observed localization of granzyme B label in the nucleus at the sites of TUNEL-positive DNA fragmentation (Fig 2A).

We have shown that granzyme B can enter the target cell unassisted by perforin or Ad2, but the possibility remained that sufficient residual granzyme B was bound to the outside of cells and that this could “piggyback” inside with perforin or Ad2 resulting in death. We took advantage of the fact that COS cells transfected with the granzyme B cDNA store the active enzyme in cytoplasmic vesicles. COS M5 cells expressing granzyme B were treated with perforin or Ad2, as described earlier for Jurkat, and analyzed by fluorescent TUNEL assay in conjunction with immunolabeling with antigranzyme antiserum to identify positively transfected cells. Perforin treatment of granzyme B positive cells exhibited rapid DNA fragmentation, while cells not expressing granzyme B were unaffected (Fig 2B). As seen with Jurkat cells, this system also resulted in a relocation of the granzyme B from the cytoplasm to the nucleus.

Despite extensive attempts, we have not been able to detect granzyme B activity in the supernatant or granzyme B protein on the surface of transfected COS cells. Thus, we believe that the ectopic protein is retained within a cytoplasmic vesicle and not secreted for reuptake in the presence of perforin or Ad2. In addition, COS cells appear to be resistant to lysis by treatment with externally added granzyme B. They can be infected with the virus, but do not appear to take up the granzyme.

Together the results obtained from visual assessment of granzyme B in Jurkat and COS M5 cells suggest that both perforin and Ad2 affect the intracellular compartment occupied by granzyme B. The enzyme is taken up independently of perforin, but is sequestered so that apoptosis does not occur. Upon exposure to perforin, or Ad2, the granzyme is released into the cytoplasm and then rapidly translocates to the nucleus. However, as soon as it is liberated, the granzyme B can interact with its substrates and hence initiate the cell death cascade.

Macromolecules can be internalized into cells by a variety of mechanisms, such as fluid phase engulfment and receptor-mediated endocytosis. Incubation of cells at 4°C prevents internalization of molecules due to a lack of fluidity of the plasma membrane, but does not prevent binding to surface receptors.36 Therefore, treatment at this temperature can be an effective means to differentiate between nonspecific fluid phase engulfment and receptor-mediated uptake. Accordingly, we treated Jurkat targets with soluble granzyme for 60 minutes at 4°C. The cells were then washed extensively with medium (pH 7.4) or citrate buffer (pH 3.0) and incubated for an additional 60 minutes at 37°C. Both treatment groups were then resuspended in medium (pH 7.4) and incubated in the presence of Ad2 for 60 minutes at 37°C. Aliquots of cells were labeled by TUNEL or annexin V-FITC and analyzed by fluorescence-activated cell sorting (FACS).

Cells treated at 4°C and washed in a neutral conditions exhibited similar levels of apoptosis to cells treated at 37°C (Fig 3). Those exposed to granzyme B at 4°C and washed under acidic conditions did not undergo apoptosis. Acid treatment of cells incubated at 37°C had no adverse effect on killing, suggesting that sufficient granzyme B had gained access to the cell before the acid wash.

If entry of granzyme B into the target cell is required to induce apoptosis, internalization must have occurred in order for killing to occur. Therefore, from these data we conclude that granzyme B interacts specifically with a surface receptor with sufficient affinity so as to withstand removal during washes, and that these interactions are susceptible to disruption by acid treatment. These data are consistent with the notion that internalization of granzyme B in the absence of perforin occurs via a receptor-mediated process and not passive engulfment of the fluid phase.

Granzyme B appeared to enter target cells autonomously, but required a second signal to induce apoptosis. This correlated with a redistribution of the granzyme within the target cell and suggested the possibility that perforin and Ad2 function to liberate granzyme B into the cytoplasm. To test this hypothesis, we used microinjection to introduce granzyme B directly into the cytoplasm of MCF-7 target cells. Granzyme B induced apoptosis in MCF-7 cells in a dose-dependent fashion (Table 1). Introduction of approximately 80 fg/cell (intracellular concentration of approximately 500 nmol/L) granzyme B led to apoptosis in 40% of cells in 2 hours. Staining of injected cells with Hoechst showed condensed nuclei, a hallmark of apoptotic cell death (data not shown). Apoptosis was due to the enzymatic activity of granzyme B, as an equivalent amount of granzyme that had been inactivated with antigranzyme B did not induce apoptosis.

Microinjection of granzyme B into the cytoplasm of target resulted in apoptosis, but granzyme internalized by the target did not. This confirms our hypothesis that the key step is release of granzyme B into the cytoplasm. The target cell appears to traffick granzyme B and sequester it away from its cellular substrates. Perforin treatment then effects the release of the enzyme where it can activate its substrates. Microinjection likely bypasses this trafficking control to cause apoptosis independently of the second signal.

Our results thus far suggest that internalized granzyme B is sequestered within a cytoplasmic vesicle. Because granzyme B appeared to be taken up by a receptor-mediated mechanism, we investigated the possibility of transport through the endocytic pathway by performing double label experiments with endosomal markers.

Small Ras-like proteins of the Rab family are known to be involved in vesicular trafficking.37,38 Rab5 is known to control early endocytic processes and is found in vesicles at the plasma membrane and in newly formed early endosomes and perinuclear recycling endosomes,39 and rab4 has been identified on early endosomes and endocytic vesicles that recycle plasma membrane constituents from early endosomes.40-43 To determine whether granzyme B was internalized into such a vesicle, colocalization studies were performed with antibodies to granzyme and rab4 or rab5.

Due to the difficulty in ascertaining distinct cellular morphology in Jurkat, we used HeLa for intracellular localization studies. We assessed the early trafficking of granzyme B by incubation of cells in the presence of granzyme B at 20°C, which prevents membrane fusion events required for endosomal maturation.36 HeLa, grown on glass coverslips, were incubated in medium containing 1 μg/mL FITC-conjugated granzyme B at 20°C. Cells were fixed, permeabilized, and immunolabeled with antisera against rab4 or antibody against rab5 and corresponding Texas Red-conjugated antisera. Immunolabeled cells treated with granzyme B-FITC were analyzed by CLSM (Fig 4A, see page 1048). We observed accumulation of granzyme B in early endosomes and perinuclear recycling vesicles characterized by colocalization with rab5. A significant, but lesser amount, of colocalization of granzyme B with rab4 was also seen, consistent with the indication that granzyme B is contained in an early endosome at 20°C.

To obtain a more detailed characterization of early trafficking of granzyme B, we performed immunoelectron microscopic analyses on targets treated with granzyme B at 20°C. HeLa were allowed to bind soluble granzyme at 4°C for 60 minutes, washed, then shifted to 20°C for 45 minutes to allow uptake. Ultrathin sections of fixed and embedded cells were labeled with antibodies against granzyme B and rab4 with corresponding gold-conjugated secondary antisera, 18 nm particle for granzyme B and 10 nm particle for rab4. As shown in Fig 4B, we observed colocalization of granzyme B with rab4 in targets treated at 20°C. Our observation of granzyme B in rab4⁺ early or recycling endosome is consistent with the colocalization of granzyme B with rab4 and rab5 by CLSM (Fig 4A).

Although granzyme B appeared to be internalized via receptor-mediated uptake and to follow an endocytic route, the further trafficking pathway remained to be established. To investigate the downstream events, we performed double label studies with granzyme B and markers of other vesicles and organelles after continuous granzyme B uptake at 37°C. HeLa were incubated for 60 minutes at 37°C in the presence of soluble granzyme B-FITC, fixed in paraformaldehyde, and immunolabeled for rab5 and rab4, mannose 6-phosphate receptor (MPR), a marker of late endosomes and the Golgi apparatus, cathepsin D and lgp120, both lysosomal markers, and cytochrome c, a mitochondrial protein. Figure 5 (see page 1048) shows representative cells labeled as described above, demonstrating that the site of granzyme B accumulation is not in vesicles or organelles characterized by any of these markers. Target cells incubated in medium containing granzyme B-FITC at 37°C for 60 minutes, or greater, no longer displayed dominant colocalization with either rab5 or rab4. This suggested that although granzyme B appeared to be transported through early endosomes, this was clearly not the site of granzyme B accumulation.

In an attempt to better define the site of granzyme B accumulation, we performed uptake experiments at 37°C for analysis by electron microscopy. HeLa were incubated at 4°C for 60 minutes with granzyme, washed, and shifted to 37°C to allow uptake and trafficking. Cells were fixed, embedded, and sectioned for immunolabeling. Figure6A shows the intracellular localization of granzyme B (10 nm gold) in HeLa treated at 37°C as a distinct compartment that is morphologically discernible from rab4⁺vesicles that contained granzyme B at 20°C (Fig 4B). Double label of granzyme B (10 nm gold) and rab5 (18 nm gold) showed some colocalization, but that the predominant granzyme label was distinct from rab5-positive vesicles (Fig 6B).

Taken together with colocalization of granzyme B with rab5 and rab4 at 20°C, these data suggest that granzyme B is transported through the early stages of the rab5/rab4 endosomal pathway, but that it accumulates in a cellular compartment that is not characterized by rab5, rab4, or any of the other vesicle markers tested.

As originally formulated, the granule exocytosis model of CTL-mediated cytotoxicity envisaged lethal lytic effector molecules being delivered to target cells by vectoral exocytosis of CTL granules. Initially it was believed that perforin was sufficient to induce target cell lysis, but then it became apparent that the granzymes (in particular granzyme B) were required to bring about physiologically relevant apoptotic death. The model was modified to include granzymes passing through perforin channels into the cytoplasm of the condemned cell. The formation of a complete macromolecular channel was not clearly demonstrated, but perforin damage to the membrane was a key factor in granzyme uptake. Despite the fact that many laboratories have been unable to replace perforin with a variety of channel forming and membranolytic agents, this model has become the widely accepted paradigm.

Although there is a wealth of indirect evidence to suggest that granzyme B enters the target, this key experimental fact has not been established. We therefore decided to look directly at whether granzyme B gains access to the target cell in the presence and absence of perforin. Using direct and indirect immunocytochemistry and CLSM, we found that granzyme became internalized into an early endosome and finally accumulated in a larger, uncharacterized vesicle. Importantly, we saw no evidence for the induction of apoptosis in these cells. In contrast, when granzyme B and perforin were added together, we observed rapid fragmentation of target cell DNA. Indeed, target cells could also be preloaded with granzyme B and then induced to die by subsequent addition of perforin. Thus, granzyme B alone can be taken up into a target cell but, for death to occur, perforin must also be present.

Perforin and Ad2 have been shown to stimulate granzyme-mediated apoptosis in targets loaded with granzyme B.28 Here we show that the granzyme is sequestered in cytoplasmic vesicles where it appears to be innocuous to the target. In addition, it appears that perforin and Ad2 caused a redistribution of the granzyme, which allowed the protease access to substrates and resulted in nuclear accumulation of granzyme B.

Ad2 is known to cause a disruption in endocytic vesicles,31,44,45 and perforin is also known to possess membrane-disrupting properties6 and also affects endocytic processes.46 The delivery of granzyme into the cytoplasm may be mediated directly by insertion of perforin channels in endosomal membranes or by stimulating a signal that ultimately results in release of granzyme within the target cell. At present, we cannot rule out that a signal triggered by perforin may be entirely independent of its pore-forming activity.

COS M5 cells transfected with granzyme B provided a system that allowed us to bypass the internalization process. Perforin and Ad2 both induced rapid DNA fragmentation in granzyme B-expressing COS cells and neither of these stimuli had adverse effects on mock transfected cells. Because the normal pathway targeting granzyme B to cytolytic granules of CTL^47,48is not present in COS cells, the granzyme may be targeted to a vesicle similar to that observed in target cells. It is unlikely that granzyme B is secreted by COS cells and reinternalized, as exhaustive efforts to detect granzyme B activity or protein in the supernatants of these cultures have been negative (data not shown). Thus, we believe that the perforin/Ad2 is acting not to increase uptake of granzyme B, but rather to stimulate the release of the granzyme into the cytoplasm. The induction of the apoptotic program could then be induced by the cleavage of key substrates, such as caspase-3. The importance of “free granzyme B” was confirmed by our demonstration that directly microinjected enzyme was able to induce death.

The granzyme does not seem to linger in the cytoplasm, as we observe a rapid translocation into the nucleus. The importance of this nuclear accumulation is unclear, but it has also been shown in an in vitro system described by Jans et al.29 It is intriguing that the site of granzyme buildup in the nucleus corresponds to the early regions detected by the TUNEL assay. Perhaps, as previously suggested,26 nuclear substrates for granzyme B exist and play a role in apoptotic events.

We have previously demonstrated that Jurkat target cells possess on their surface saturable binding sites for granzyme B (3 × 10⁴ sites per cell; kd ≈10 nmol/L) and suggested the likelihood of a granzyme B receptor. In this report, we expand the previous study by demonstrating that granzyme B binds to cells via high-affinity interactions at the cell surface. By incubating target cells in granzyme B at 4°C, we allowed binding, but inhibited its uptake. Washing the cells at pH 3.0 disrupts protein-protein interactions and dissociated bound granzyme B from the cell surface, thus preventing the induction of cell death after the addition of perforin or Ad2. Incubation of targets in granzyme B at 4°C followed by acid wash reduced killing to approximately 25% of that observed when targets were washed under neutral conditions. Residual killing of acid-washed cells at 4°C may be due to entry of granzyme during handling and processing.

These data lend further support to the notion of a specific granzyme B receptor. We have also ruled out the previous suggestion that granzyme B remained bound to its putative receptor and required perforin or Ad2 to stimulate entry,28 as acid treatment of cells incubated in granzyme B at 37°C did not show any reduction in apoptosis compared with cells washed under neutral conditions. These data strongly suggest that granzyme B binds to target cells as proposed by Froelich et al,28 but that the granzyme B is internalized by the target in the absence of additional exogenous factors, namely perforin or Ad2.

The likelihood of a granzyme B receptor and the involvement of endocytosis and trafficking of Ad2 led us to investigate the possibility that granzyme B follows a similar route. Using double label studies with markers of intracellular compartments, we found granzyme B label associated with early endosomal markers rab5 and rab4 in cells treated at 20°C to inhibit endosomal trafficking. Uptake of granzyme B by endocytosis is consistent with earlier reports that showed that the microtubule inhibitor, cytochalasin B, effectively inhibited granzyme B–induced killing.9 

Our results appear to be in direct contradiction to the recent report of Shi et al,49 who reported internalization of granzyme B directly into the cytoplasm. However, in their study, it was very difficult to discern the morphologic features represented in the electron micrographs. In addition, endocytic compartmentalization is generally believed to be the primary route of entry taken by the majority of molecules after receptor binding.50 In our study, we clearly show granzyme B in a distinct cytoplasmic compartment.

Our results, indicating the regulated uptake and trafficking of granzyme B, are supported by the observation that cytoplasmic microinjection of granzyme resulted in apoptosis. This result also appears to be contrary to a recent report in which microinjected granzyme B did not induce death.49 The apparent difference may be explained by the dose-dependent nature of the apoptotic response that we report in the present study. We observed apoptosis when granzyme B was injected at 54 fg/cell, which represents a ninefold higher cellular concentration than in the previous study. Our data support a model in which granzyme B enters the target via receptor-mediated endocytosis and is transported to a cytoplasmic compartment, where it is unable to access substrates. Transport to this compartment is not associated with cell death; however, release is vital for initiation of the apoptotic program.

Although granzyme B appears to be transported through early endosomes, these vesicles are clearly not the site of accumulation of granzyme. In target cells that were allowed to take up granzyme B at 37°C, the protease was found in a cellular compartment that is not characterized by the presence of rab5 or rab4 as seen at 20°C. Interestingly, Ad2-induced redistribution of granzyme B from its intracellular compartment did not occur at 20°C (data not shown). Additionally, we performed immunofluorescent labeling of other cellular markers on targets containing granzyme B-FITC and did not find granzyme B associated with markers of lysosomes, mitochondria, or structures containing MPR. Thus, it appears that granzyme B is internalized via a receptor-mediated process into an early endosome, but by further trafficking, it continues on to a compartment that is not characterized by any of our cellular markers.

Because of a lack of colocalization with the cellular markers we tested, we have not been able to better define the site to which granzyme B trafficks. However, it appears to be a homogeneous, translucent structure that resembles immature cytolytic granules of CTL.51 Although it remains to be conclusively established, it appears that trafficking of granzyme B to this novel compartment is necessary for apoptosis, as granzyme B does not reach this compartment at 20°C.

In this study, we have provided further evidence to support the existence of a granzyme B receptor and, more importantly, we have shown that trafficking of granzyme B by the target is required for apoptosis. This provides us with an additional level at which to direct drug design to enhance or abrogate CTL- or NK-based immune responses. Targeting the granzyme B receptor in transplants may provide a means to suppress rejection due to infiltrating cytotoxic lymphocytes. Our results suggest that a search for other granzyme B–binding molecules, in addition to substrates, may show proteins that play an essential role in the internalization and intracellular transport of this important cytotoxic effector.

The authors thank Dorothy Hudig and Ulrike Winkler for antigranzyme antiserum and Irene Shostak and Tracy Sawchuk for technical assistance.