The Intracellular Granzyme B Inhibitor, Proteinase Inhibitor 9, Is Up-Regulated During Accessory Cell Maturation and Effector Cell Degranulation, and Its Overexpression Enhances CTL Potency¹ Free

Natural killer cells and CTLs kill virus-infected and tumor cells by inducing apoptosis through exocytosis of cytotoxic granule contents or ligation of death receptors (reviewed in Ref. 1). Granzyme B (grB)³ is a key cytotoxic granule proteinase that is endocytosed by the target cell following cytotoxic lymphocyte (CL) degranulation, probably by binding to the mannose-6-phosphate receptor (M6PR) (2). The cytotoxic pore-forming protein, perforin, is also internalized and mediates the release of grB into the cytoplasm of the target cell. GrB then rapidly induces apoptosis by cleaving one or more cytoplasmic or nuclear proteins. These include Bid (3, 4, 5, 6), caspases (7, 8), components of the DNA repair machinery (7, 9), and inhibitor of caspase-activated DNase (10, 11).

Although CL granule contents are efficiently directed into the target cell via the immunological synapse (12), some grB may escape from granules or the synaptic zone into the cytoplasm of the CL or into the extracellular milieu. For example, free grB is evident in the sera of patients with elevated CTL responses (13) or severe Gram-negative bacterial infections (14). Given its cytotoxic potency and ability to degrade extracellular proteins (13, 15, 16, 17, 18), it is likely that protective mechanisms exist to counter misdirected or ectopic grB.

GrB is efficiently inhibited by the nucleocytoplasmic serpin, proteinase inhibitor 9 (PI-9) (19, 20), with transfection studies demonstrating that PI-9 protects cells from grB-mediated apoptosis (21). PI-9 is present in cells at immune-privileged sites such as testis and placenta (22, 23), and the expression of PI-9 in endothelial and mesothelial cells suggests that it protects bystander cells from grB released during an immune response (24). Here we show that PI-9 is up-regulated in CTLs in response to grB production and degranulation, that it associates with grB-containing granules in activated CTLs and NK cells, and that overexpression increases CTL potency. We also show that PI-9 is present in several dendritic cell (DC) types. The localization and regulated expression of PI-9 in these leukocyte subsets strongly support the hypothesis that PI-9 protects effector, accessory, and bystander cells from ectopic grB during an immune response.

PI-9 was detected with the specific mAb 7D8 (22) or with rabbit 12 or rabbit 15 polyclonal antisera raised against recombinant PI-9 (19). PI-6 was detected with mAb 3A (25). GrB was detected with 2C5 (26) or rabbit anti-grB/grH polyclonal antiserum (27), both provided by J. Trapani (The Peter MacCallum Cancer Institute, East Melbourne, Australia), or GrB-7 (Chemicon, Temecula, CA). CD markers were detected with hybridoma supernatants from 3C10 (CD14; provided by R. Steinmann, Rockefeller Institute, New York), FMC63 (CD19; provided by H. Zola, Child Health Research Institute, Adelaide, Australia), and OKT3 (CD3) and OKT8 (CD8; both from American Type Culture Collection (Manassas, VA). Anti-CD4 and anti-CD56 were obtained from BD Biosciences (Franklin Lakes, NJ). Actin was detected with anti-actin (I-19) from Santa Cruz Biotechnology (Santa Cruz, CA).

Intracellular flow cytometry was performed using peripheral blood drawn from healthy volunteers. Erythrocytes were removed from whole blood by lysis in 167 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA for 5 min at room temperature, centrifugation at 100 × g for 5 min, then washing twice in PBS containing 1% FCS and 0.02% sodium azide (repeated for all subsequent washes). Approximately 2.5 × 10⁶ PBLs were fixed in 4% formaldehyde in PBS for 10 min at room temperature, centrifuged, resuspended in FACS Permeabilizing Solution (BD Biosciences) for 10 min, centrifuged, washed, and incubated with 7D8 or isotype control Ab (IgG1; BD Biosciences) for 30 min. Cells were washed twice, then incubated with FITC-conjugated anti-mouse Ig (Chemicon). After two more washes the cells were analyzed using a FACSCalibur and CellQuest software (BD Biosciences).

Highly purified B, CD4⁺, CD8⁺, monocyte, and NK cell populations were obtained from whole blood using RosetteSep Enrichment cocktails (StemCell Technologies, Vancouver, Canada) according to the manufacturer’s instructions. Cells were plated on poly-l-lysine-coated slides, fixed in 4% formaldehyde for 20 min, then permeabilized in 0.5% Triton X-100 for 5 min. PI-9 and PI-6 were detected with 7D8 hybridoma supernatant or 3A hybridoma supernatant, respectively. Bound Ab was detected with anti-mouse Ig conjugated to FITC. Cells were counterstained with propidium iodide (1 μg/ml), and cross-sections through the nucleus were obtained using laser scanning confocal microscopy (TCS-NT; Leica, Wetzlar, Germany). The purity of each cell population and the presence of any contaminating cell types were assessed using Abs to CD3, CD4, CD8, CD14, CD19, and CD56.

A cDNA encoding PI-9 (19) was subcloned into the MIGR1 retroviral vector in the forward or reverse orientation. PI-9 mRNA was transcribed as a bicistronic message with green fluorescence protein (GFP) (28). Viral supernatant for transduction was obtained by transient transfection of the 293GP packaging cell line (29) using Lipofectamine Plus (Invitrogen, Carlsbad, CA). Supernatant was tested for transduction efficiency on Jurkat cells. PBLs were obtained from normal healthy donors and were cultured in for 2 days with PHA and human IL-2 (300 U/ml), then transduced with retrovirus (29). After transduction, PBLs underwent one round of rapid expansion with irradiated feeder cells (BRT Laboratories, Baltimore, MD), and the brightest GFP-expressing PBLs were purified by FACS. Sorted PBLs (72% GFP⁺, 75–80% CD8⁺) were then kept in culture by rapid expansion every 3 days. PI-9 mRNA was quantitated by real-time PCR on cDNA (30) using primers and probes specific for PI-9 and GAPDH, respectively (MegaBases, Evanston, IL). Retrovirally transduced PBLs were analyzed in ⁵¹Cr release assays (in triplicate wells) with hybridoma targets that express the anti-human CD3 mAb OKT3 (31). The percent specific lysis of JY cells, a human B lymphoblastoid cell line, was determined in parallel (<6% specific lysis) and subtracted from the lysis of the OKT3 hybridoma to report CTL killing only.

YT cells (32) were cultured as previously described (20). PBMCs were isolated from whole blood using Ficoll-Paque Plus. PBMCs (depleted of monocytes by adherence to plastic) were cultured in IL-2 (Sigma-Aldrich, St. Louis, MO) or a combination of Con A and PMA (both from Sigma-Aldrich) as previously described (33). Culture-generated quiescent NK cells were cultured and activated as described previously (34, 35). At the indicated time points, activated cells were harvested, and cell lysates were prepared by lysis in Nonidet P-40 (NP40) lysis buffer (1% NP-40 in 50 mM Tris-HCl (pH 8.0), and 10 mM EDTA (pH 8.0)). To prevent postlysis association of PI-9 and grB, cells were lysed on ice in the presence of protease inhibitors (1 mM PMSF, 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin). Lysates from 0.5 × 10⁶ cells were resolved by reducing 12.5% SDS-PAGE, transferred to nitrocellulose, immunoblotted for PI-9 with 7D8, and detected with HRP-conjugated anti-mouse Ig (Chemicon) using ECL (NEN, PerkinElmer, Boston, MA). The membranes were stripped (62.5 mM Tris-HCl (pH 6.8), 2% SDS, and 0.1 M 2-ME for 30 min at 50°C), then reprobed for grB (2C5) and actin (diluted 1/1000). Densitometric analysis of PI-9 levels was determined using MCID Image Analysis software (BD Biosciences).

Lysate from COS-1 cells transfected with PI-9 (20) was incubated with or without 100 ng of recombinant grB (36) at 37°C for 10 min. Samples were resolved by reducing 12.5% SDS-PAGE, transferred to nitrocellulose, and immunoblotted with rabbit 12 and rabbit 15 antisera (PI-9) or 2C5 (grB). YT cells (1 × 10⁶) were cultured for 20 h with or without 25 μM calpain inhibitor I (N-acetyl-Leu-Leu-Norleu; Sigma-Aldrich). Cells were lysed in NP-40 lysis buffer or modified Laemmli sample buffer (60 mM Tris-HCl (pH 6.8), 2% SDS, and 10% glycerol) (37) and passed through a 26-gauge needle to shear the DNA. Twenty micrograms of cell lysate was resolved by reducing 12.5% SDS-PAGE, transferred to nitrocellulose, immunoblotted with rabbit 12 antiserum, then stripped and reprobed with rabbit 15 antiserum.

Activated CTLs were attached to poly-l-lysine-treated slides, fixed and permeabilized in 1/1 acetone/methanol for 2 min at room temperature, and stained for PI-9 (7D8), grB (rabbit anti-grB antiserum, diluted 1/200), or cathepsin A (rabbit antiserum, diluted 1/500, obtained from A. Pshezhetsky, University of Montreal, Montreal, Canada). Activated NK cells were cultured on irradiated MM-170 cells for 4 h, fixed in 4% formaldehyde, then permeabilized in 100% methanol and stained for PI-9 (7D8) and grB (rabbit anti-grB antiserum, diluted 1/200). YT cells were attached to poly-l-lysine-treated slides, fixed and permeabilized in acetone/methanol, and stained for PI-9 (rabbit anti-PI-9 antiserum, diluted 1/2000) and grB (2C5, diluted 1/200). BeWo cells were cultured and handled for microscopy as previously described (20). Primary Abs were detected with the appropriate secondary Ab conjugated to FITC or rhodamine B isothiocyanate (Chemicon). Two-color images were obtained using laser scanning confocal microscopy.

Human tissues fixed in neutral buffered formalin were obtained from the archive of the Pathology Department of Box Hill Hospital (Melbourne, Australia). Immunohistochemistry was performed as previously described (22). Cryopreserved normal human thymus was provided by R. Boyd (Department of Pathology and Immunology, Monash University, Prahran, Australia). Sections (6 μm) of tissue were air-dried onto silanized slides. Sections were washed and blocked in normal goat serum before incubation with 7D8 or an isotype-matched control Ab (IgG1). Bound Ab was detected with anti-mouse Ig conjugated to FITC. Sections were then stained with anti-CD3, -CD4, or -CD45 mAbs directly conjugated to PE (Diatec, Oslo, Norway). Two-color images were obtained using laser scanning confocal microscopy.

PBMCs were isolated from normal donors over Ficoll-Paque gradients (Red Cross Blood Transfusion Service, Melbourne, Australia). Monocytes were prepared by elutriation at 2100 rpm using a Beckman J6 M (Palo Alto, CA) with a standard chamber. Monocytes were further purified by sorting using a MoFlo CLS cell sorter (Cytomation, Fort Collins, CO) based on forward and side light scatter and subpopulations sorting for CD16⁺ and CD14⁺ expression. Lineage-negative DCs were purified from the monocyte elutriation fractions by negative selection with Abs to CD3, CD14, CD11b, CD19, and CD16. Lineage-negative DCs were sorted into CD1b/c⁺ (B-B5; BioSource International, Camarillo, CA) Langerhans cell precursors (38) or CD123⁺ (CD123, BD Bioscience) plasmacytoid DCs (39). MDDCs were generated from CD14⁺ monocytes positively selected using MACS columns after labeling for CD14 and goat anti-mouse MACS beads (Miltenyi Biotec, Bergisch Gladbach, Germany). Cells were selected on a MACS column as recommended by the manufacturers. CD14⁺ monocytes were cultured in IL-4 (20 ng/ml) and GM-CSF (40 ng/ml) for 5–7 days as previously described (40). Immature DCs were matured in the presence of TNF-α (10 ng/ml; R&D Systems, Minneapolis, MN) for 2 days.

To extend our previous observations that PI-9 is expressed in lymphoid (T, B, and NK-like) cell lines, but not myeloid cell lines (19), we examined PI-9 in PBLs by intracellular flow cytometry using a specific mAb (22). PI-9 was evident in the majority of PBLs in both mononuclear cells (R1) and granulocytes (R2; Fig. 1,A). Most mononuclear cells were clearly positive (>90%). Dual-color analysis showed that >95% of CD3⁺ cells are PI-9 positive, 100% of CD4⁺ and CD8⁺ cells are PI-9 positive, and ∼75% of CD19⁺ cells are PI-9 positive (data not shown). Many granulocytes (>70%) were also positive, but appeared to have lower amounts of PI-9, with only a slight positive shift in staining apparent over the isotype control (Fig. 1 A). Comparison of the mean fluorescence intensities indicated that mononuclear cells express at least 5-fold more PI-9 than granulocytes.

To further identify the populations of mononuclear cells that produce PI-9, B cells, CD4⁺ and CD8⁺ T cells, NK cells, monocytes, and granulocytes were enriched from PBLs using established immunodepletion procedures. These were analyzed for PI-9 by indirect immunofluorescence so that its intracellular distribution could be simultaneously assessed (Fig. 1,B). The efficiency of immunodepletion was monitored by staining for CD3, CD4, CD8, CD14, CD19, or CD56 (Fig. 1 C). PI-9 was evident in all populations, with the highest levels in NK and T cells (CD4⁺ and CD8⁺ T cells expressed equivalent levels). B cells, monocytes, and granulocytes had much lower levels. In all cell types PI-9 exhibited a nucleocytoplasmic localization, consistent with previous reports (19, 20). The expression pattern of PI-9 differed from that of the closely related serpin, PI-6, which is a cathepsin G inhibitor present in myeloid cells (41).

The observation of high PI-9 expression in NK and CTLs suggests that it protects grB-producing cells against endogenous grB. Using indirect immunofluorescence and laser scanning confocal microscopy we examined the intracellular localization of PI-9 in primary NK cells and the NK-like cell line, YT. This showed that in addition to its previously reported cytoplasmic and nuclear distribution (Fig. 2,A) (20), PI-9 is associated with vesicles in the CL cytoplasm that also contain grB (Fig. 2, B and C). These vesicles were positive for the lysosomal/granule marker lysosome-associated membrane protein-1 and not with markers for the endoplasmic reticulum or Golgi apparatus (data not shown), indicating that they are secretory lysosomes (cytotoxic granules). To determine whether the association of PI-9 with lysosomes is unique to CLs, the distribution of endogenous PI-9 in the epithelial cell line BeWo was also examined (Fig. 2 D). Here PI-9 was evident only in the cytoplasm and nucleus, and no colocalization between PI-9 and lysosomal markers was observed.

In a previous study we separated membrane and cytosolic fractions of CLs and showed that PI-9 is present in the cytosol and does not copurify with grB in the granule fractions (19). Given that PI-9 is not present within granules, the results shown in Fig. 2 suggested that it is associated with the cytoplasmic surface of the granule and is poised to rapidly inactivate grB entering the cytoplasm from leaking granules. To test this idea we examined YT cells for evidence of extragranular grB in complex with PI-9. We took advantage of the fact that grB-PI-9 complexes are stable under SDS-PAGE and immunoblotting conditions (19, 21), and that we have two high affinity PI-9 Abs that detect PI-9 in complex with grB very efficiently. As shown in Fig. 3,A, these Abs (rabbit 12 and 15) detect the grB-PI-9 complex better than they detect free PI-9, and they also detect the degradation products commonly observed following serpin-proteinase complex formation. Degradation of the complex occurs because distortion of the protease (in particular) renders it highly susceptible to proteolysis (42, 43). The Abs do not bind grB (Fig. 3,A), and their corresponding preimmune sera do not recognize PI-9, complex, or degradation products (data not shown). None of the four grB Abs we have tested detected the complex or degradation products with similar sensitivity to the PI-9 Abs. An immunoblot using the most sensitive of these grB Abs (2C5) is shown in Fig. 3 A.

When YT cells are lysed in buffers containing NP-40, active grB is released from granules and binds cytoplasmic PI-9 postlysis (20, 21). By contrast, lysis of CTLs in buffers containing SDS prevents postlysis interactions, presumably by rapidly denaturing cellular proteins (37). Thus, lysis of YT cells in SDS will denature free PI-9 and free grB, and a complex evident in SDS-treated YT cell extracts will have formed before lysis. As shown in Fig. 3,B, lysis of YT cells in NP40 generated a postlysis complex of ∼70 kDa as well as smaller products resulting from complex degradation that were detected by both PI-9 antisera, but more efficiently by rabbit 15. When cells were lysed in SDS buffer, small, but reproducible, amounts of complex and degradation products were detected by rabbit 15 (Fig. 3 B). This suggests that there is a pool of extragranular grB in the cytoplasm of YT cells that is bound to PI-9 and undergoing proteolysis.

Because serpin/proteinase complexes are irreversible and involve distortion of both serpin and proteinase (42, 43), we predicted that cytoplasmic PI-9/grB complexes would be recognized and degraded by the ubiquitin-proteosomal machinery. Cells cultured in the presence of a proteosome inhibitor should therefore accumulate complexes. Indeed, greater amounts of complexes were detected by both PI-9 antisera when cells were incubated with the proteosomal inhibitor, calpain inhibitor I, and lysed in SDS (Fig. 3 B). This also supports the idea that grB enters the cytoplasm of CLs during normal cellular function and is rapidly inactivated by PI-9.

Since PI-9 potentially protects CLs from death induced by misdirected grB, we wanted to determine the effect on CTL potency of overexpressing PI-9 in activated PBLs. Normal donor PBLs were activated with PHA and transduced with retroviral vectors containing the PI-9 gene in either the forward (coding) or the reverse (noncoding) orientation. These vectors allow transcription of the inserted cDNA as a bicistronic mRNA that also encodes GFP (28, 44). Cultures transduced with either the coding or noncoding vectors were enriched to equivalent amounts of GFP-expressing cells (∼72%) by FACS (data not shown). Those transduced with the PI-9 gene in the coding orientation expressed 4-fold higher levels of PI-9 mRNA (p < 0.02) compared with those in which the PI-9 gene was in the noncoding orientation (Fig. 4 A). Both cultures contained an equal proportion of CD8 cells, which was 75–80% of the total PBLs (data not shown).

The ability of uncloned populations of transduced CTLs to lyse target cells expressing an anti-CD3 mAb (OKT3) (31) was evaluated. In parallel, we also determined the lysis of JY target cells, which do not express anti-CD3 mAbs, so we were able to account for NK cell activity in the cultures. In each experiment the specific lysis of JY targets (NK cell killing) was subtracted from that of OKT3 targets (total killing) to give the level of CTL activity. At every E:T cell ratio tested we observed increased potency of CTLs that expressed elevated levels of PI-9 (Fig. 4 B). For example, at an E:T cell ratio of 0.5:1 we observed a 3-fold increase in potency of killing by CTLs that overexpressed PI-9 (p < 0.02). Since PI-9 can protect cells from apoptosis caused by grB, we infer that overexpression of PI-9 increases CTL potency by improving viability.

CL stimulation, which increases the level of grB, should also result in a corresponding increase in PI-9. We therefore examined the regulation of PI-9 in T and NK cells under conditions known to induce grB synthesis and release. PBLs were stimulated with IL-2, which expands and activates T cells, or a combination of Con A and PMA, which results in T cell activation and granule exocytosis. As shown by immunoblotting, both these stimuli induced grB expression, peaking on days 6 and 3, respectively (Fig. 5, A and B). No significant increase in PI-9 was observed in T cells stimulated with IL-2 alone (Fig. 5,A). However, stimulation of T cells with Con A/PMA resulted in a 3-fold induction of PI-9 over endogenous levels (Fig. 5 B). This treatment also generated the higher M_r form of grB (35 kDa) that is only observed following T cell degranulation (45). Thus, in vitro, PI-9 up-regulation is associated with the release of grB from CLs.

The regulation of PI-9 was also investigated in NK cells. Culture-generated quiescent NK cells were activated by incubation with targets (irradiated MM-170 cells) in the presence of IL-2. These NK cells constitutively express both PI-9 and grB, as evident on day 1 of stimulation (Fig. 5 C). While activated NK cells killed MM-170 cells (data not shown) and up-regulated grB, which peaked on day 3, no significant increase in PI-9 was detected over time. Taken together, these results suggest that the level of PI-9 in proliferating, nondegranulating T cells and activated NK cells is sufficient to protect against grB. However, the increase in PI-9 expression during T cell degranulation suggests that granule exocytosis is associated with increased entry of grB into the effector cell cytoplasm, and that degranulating cells consequently require higher levels of PI-9.

Since PI-9 is up-regulated in response to T cell degranulation in vitro, its expression was also investigated in activated T cells in vivo by immunohistochemical analysis of normal and inflamed human tissues. When normal spleen was examined using the specific PI-9 mAb, 7D8, in standard immunohistochemical procedures, no PI-9-positive lymphocytes were detected (Fig. 6, b and c). However, when lymphocytes were extracted from freshly isolated spleen and stained for PI-9 using the more sensitive technique of indirect immunofluorescence, PI-9-positive cells were apparent (Fig. 6,a). This indicated that the level of PI-9 expressed in quiescent lymphocytes is below the level of immunohistochemical detection using this Ab. By contrast, when we examined inflamed tissue samples by immunohistochemistry, PI-9-positive cells were clearly evident within lymphocytic infiltrates. For example, PI-9-positive lymphocytes were observed in sections of a ductal carcinoma of the breast (Fig. 6,d) that were also positive for grB (Fig. 6 e). Thus, PI-9 is up-regulated in grB-expressing effector cells in vivo.

In addition to protecting CLs from endogenous grB, PI-9 should protect accessory and bystander cells from exogenous grB released during the immune response. Supporting this idea, PI-9-positive DCs were evident in tonsillar germinal centers using the immunohistochemical procedures described above (Fig. 6, g and h). PI-9-positive DCs were also detected by two-color immunofluorescence in the thymus. These thymic PI-9-positive cells were CD3⁻ CD4⁺ CD45⁺ (Fig. 6, j–l) and CD8⁻ (data not shown) with DC morphology and were located in the keratin-negative medulla. The morphology and phenotype of these PI-9-positive DCs are consistent with several populations of human thymic DCs recently described (46). It is important to note that no PI-9-positive lymphocytes were detected in either tonsil or thymus, suggesting that DCs constitutively express higher levels of PI-9 compared with resting or immature lymphocytes.

Since immunohistochemistry is qualitative, highly purified DC populations were isolated from whole blood to assess the relative levels of PI-9. Consistent with the FACS data, low levels of PI-9 were present in monocytes purified from PBLs. IL-4- and GM-CSF-induced differentiation of these cells into immature MDDCs resulted in a significant increase in PI-9 expression (Fig. 7,A). Maturation of MDDCs by exposure to TNF-α resulted in a further increase in the expression of PI-9 (Fig. 7 B). By densitometric analysis, this increase in PI-9 in MDDCs was ∼4-fold over monocytes and increased a further 3.5-fold upon maturation with TNF-α.

To identify specific DC types that express PI-9, subpopulations were purified from whole blood and examined for PI-9 by immunoblotting. The CD16⁺ monocytic precursors of tissue DCs (47, 48) expressed higher levels of PI-9 than CD16⁻ monocytes (Fig. 7,B). Lineage-negative DCs purified from peripheral blood expressed PI-9 at an intermediate level between monocytes and MDDCs (Fig. 7,A). These lineage-negative DCs contain two major populations: CD1b/c⁺ Langerhans cell precursors (38) and CD123⁺ plasmacytoid DCs (39). The plasmacytoid DCs expressed an intermediate level of PI-9, while Langerhans cell precursors expressed virtually no PI-9 (Fig. 7 B).

The differential expression of PI-9 in DC populations from both the myeloid and lymphoid lineages is illustrated in Fig. 7 C. The distribution of PI-9 in DC subsets suggests that expression is not limited by ancestry, and that differing requirements for protection from grB during Ag presentation to naive CD8⁺ T cells during induction of effector functions determines PI-9 expression.

Cells have evolved a number of mechanisms to control apoptosis. For example, inhibitor of apoptosis proteins inhibit caspases directly (49), cellular FLIP blocks Fas-mediated apoptosis (50), and Bcl-2 family members regulate apoptosis at the mitochondria (reviewed in Refs. 51 and 52). We and others believe that PI-9 and other intracellular serpins are part of the anti-apoptotic machinery of cells involved in or exposed to the cellular immune response (23, 41, 53, 54). In particular, we have proposed that PI-9 protects cells against misdirected grB (19, 21, 22, 24).

A role for PI-9 in protecting grB-expressing cells is supported by its unusual intracellular distribution, described in the first part of this study. In addition to its cytoplasmic and nuclear localization (20), PI-9 in CLs is clearly associated with grB-containing cytotoxic granules. Fractionation studies indicate that PI-9 is not present within granules (19). Other studies show that, like PI-6 (55), when PI-9 is provided with an efficient signal peptide, it is retained in the endoplasmic reticulum and cannot move through the secretory pathway (A. Gillard and P. Bird, unpublished observations). Therefore, PI-9 must associate with the external, cytoplasmic face of the granule through interaction with a specific granule component. Our demonstration that PI-9 does not associate with lysosomes in epithelial cells suggests that it interacts with a protein or other component not found on nonsecretory lysosomes or on other membrane-bound organelles. For example, it may bind a specific lipid, as the lipid content of the granule membrane differs from that of the plasma membrane and other organelles (56).

GrB is stored in granules in an active form (57); thus, leakage from granules or from the synapse would allow it to access nucleocytoplasmic substrates and kill the CL. The positioning of PI-9 at the granule surface and of the PI-9/grB complexes evident within CLs suggests that granules do leak and that PI-9 is present to rapidly inactivate extragranular grB. Once granule exocytosis is triggered, the granules are refilled with newly synthesized grB; however, some of the newly synthesized grB is constitutively secreted as a 35-kDa form (45). The 35-kDa form of grB is specifically associated with degranulating cells, so the correlation we report with a 3-fold increase in PI-9 strongly suggests that grB misdirection occurs during T cell degranulation. While this in vitro system probably does not fully recapitulate the situation in vivo, our demonstration of an increase in CTL killing efficiency on 3- to 5-fold overexpression of PI-9 suggests that the similar fold up-regulation seen in Con A/PMA-treated cells is physiologically relevant. We therefore conclude that CTL viability is enhanced by a PI-9-mediated reduction in suicide or fratricide induced by misdirected grB.

What might cause granule leakage? Granules are secretory lysosomes containing lysosomal hydrolases in addition to granzymes and perforin (58, 59, 60). While there is no direct evidence that granules leak, it is clear that lysosomal rupture can be induced by stressors such as oxidation and UV irradiation, leading to apoptosis and necrosis (61, 62). Increased levels of reactive oxygen species have been observed in CLs, which may lead to apoptosis (63). This suggests that granule leakage occurs during effector cell activation and/or function.

Release of grB into the effector cell cytoplasm need not only occur from leaking granules. It is possible that secreted grB is endocytosed by effectors or bystanders. The uptake of grB into target cells is thought to be mediated by the 300-kDa M6PR (2). The M6PR is expressed in all nucleated cells, with up to 20% of the receptor present at the cell surface (64, 65, 66, 67, 68). Interestingly, the M6PR is up-regulated on activated T cells (69), which might increase their susceptibility to secreted grB.

The cellular immune response involves a complex interplay between many cell types. APC or accessory cells (DCs, macrophages, and B cells) induce differentiation of naive T cells into cytotoxic or Th lymphocytes by secretion of cytokines and expression of costimulatory molecules. These accessory cells are closely associated with CLs and are likely to be exposed to collateral damage mediated by grB and perforin during the immune response. The presence of PI-9 would provide protection against inadvertent killing of these cells.

How likely is such inadvertent killing? DCs have established roles in presenting Ag to CD4⁺ Th cells and eliciting Th1 (CTL) or Th2 (B cell) responses. However, distinct DC subpopulations also directly interact with B cells or CD8⁺ CTLs. For example, follicular DCs in the germinal center directly contribute to B cell proliferation and differentiation (70), while virally infected DCs and DCs purified from blood can directly present to CD8⁺ T cells in the absence of CD4⁺ Th cells (71, 72, 73, 74).

This close association of activated CTLs and DCs may result in elimination of DCs by effector CTLs (33, 75). Elimination of DCs in normal mice is unusual, presumably due to protective mechanisms, but elimination of Ag-specific DCs by cognate CTLs has been observed in transgenic mouse models (76, 77). Although this demonstrates that DCs are potentially susceptible to CTLs, the mechanism of CTL-dependent clearance of DCs is unclear. One study found DC elimination to be independent of Fas and perforin (77), while another reports that it is mediated partly by the perforin pathway (78). Consistent with the latter results, we and others have suggested that expression of PI-9 in DCs prevents grB-mediated apoptosis during Ag presentation to CTLs (23, 53). This is further supported by the recent report that SPI6, one of seven murine PI-9 homologues (79), protects murine DCs from CTL-induced apoptosis (54).

Our results show that PI-9 is expressed in specific DC subsets. Thymic medullary DCs comprise three different subsets: a major CD11b⁻ subset of lymphoid origin, a minor CD11b⁺ subset of myeloid origin (46), and a population of plasmacytoid DCs (80). The CD11b⁺ thymic DCs resemble tonsillar germinal center DCs (81) and are thought to be phenotypically and morphologically related (46). Considering this relationship, it is likely that the PI-9-positive DCs observed in the thymus are related to the PI-9-positive DCs located in the tonsil. Both these DC populations are probably derived from CD16⁺ myeloid precursors (47, 48, 82). PI-9 is highly expressed in CD16⁺ monocytes, which is consistent with PI-9 expression in the CD11b⁺ thymic DCs and tonsillar germinal center DCs.

The differential expression of PI-9 in DCs suggests differing requirements for protection from grB-mediated apoptosis in DC subsets. Thymic medullary DCs are essential in the positive and negative selection of thymocytes. GrB-positive cells are present in the thymic medulla, with grB transcripts detected in double-positive (CD4⁺ CD8⁺) thymocytes (83) and in both CD4⁺ or CD8⁺ single-positive thymocytes (84). This suggests that thymocytes undergoing selection express grB and have cytotoxic potential. Thus, the presence of PI-9 in thymic medulla DCs is consistent with a role in protecting these DCs from grB-mediated apoptosis.

Some subsets of germinal center DCs are involved in the presentation and activation of T cells directly (81), while plasmacytoid DCs are IFN-producing cells (85, 86, 87) that initiate potent Th1 (CTL) responses (88). The expression of PI-9 in these cells suggests that DC populations that present to and activate CD8⁺ precursor T cells are at risk from the effector functions of the T cells they activate. This is also supported by the up-regulation of PI-9 in MDDCs upon TNF-α-induced maturation, suggesting that mature APC require protection from inadvertent apoptosis.

It would be advantageous for other cell types to express PI-9 to protect against misdirected grB during the immune response. For example, PI-9 is expressed in endothelial and mesothelial cells likely to be exposed to CLs and is up-regulated by inflammatory stimuli (24, 89). The expression of grB and perforin has also been demonstrated in human CD4⁺ CTLs (90, 91), suggesting that PI-9 has a cytoprotective role in some CD4⁺ T cells. The level of PI-9 expressed in B cells, monocytes, and granulocytes is lower than in CLs or DCs, and it is unlikely that they would be protected from direct CTL attack. Hence, these cells have sufficient PI-9 to cope with low levels of misdirected grB, but could still be cleared by CTLs if they become infected or tumorigenic.

Two recent papers have indicated a role for PI-9 in tumor evasion by providing neoplastic cells with an advantage against grB-mediated cytotoxicity (92, 93). However, both studies failed to establish the baseline expression of PI-9 in tissues and cells or to compare the level of PI-9 expression in normal and tumor samples. The expression of PI-9 in epithelial cells has been reported (24), and PI-9 is in the ductal tissue of the breast, the columnar epithelia of the colon, and the ciliated columnar epithelia lining the female reproductive tract (M. Buzza, unpublished observations). The presence of PI-9 in carcinomas of breast, colon, and cervix (92) is therefore not surprising.

Additionally, the work reported here shows that PI-9 is widely distributed in normal leukocytes, which may account for the presence of PI-9 in T, B, and Hodgkin lymphomas (93). Analysis of the latter data indicates that only 28% of 224 biopsies were positive for PI-9. Furthermore, the number of PI-9-positive cells within each biopsy varied, with only 17% containing a majority of PI-9-positive tumor cells. The lack of PI-9 in most of these lymphomas suggests it is unlikely that PI-9 up-regulation is a common mechanism by which lymphomas resist immune destruction.

We thank J. Sun for recombinant grB, J. Trapani (Peter MacCallum Cancer Institute, Melbourne, Australia) for grB Abs, A. Pshezhetsky (University of Montreal, Montreal, Canada) for cathepsin A Abs, P. Hosking (Department of Anatomical Pathology, Box Hill Hospital, Melbourne, Australia) for archival tissues, and R. Boyd and M. Malin (Department of Pathology and Immunology, Monash University, Clayton, Melbourne, Australia) for thymus tissue.