Peripheral blood and tissue eosinophilia is a prominent feature in allergic diseases and helminth infections. In cancer patients, tumor-associated tissue eosinophilia is frequently observed. Tumor-associated tissue eosinophilia can be associated with a favorable prognosis, notably in colorectal carcinoma. However, underlying mechanisms of eosinophil contribution to antitumor responses are poorly understood. We have in this study investigated the direct interactions of human eosinophils with Colo-205, a colorectal carcinoma cell line, and show that eosinophils induce apoptosis and directly kill tumor cells. Using blocking Abs, we found that CD11a/CD18 complex is involved in the tumoricidal activity. Coculture of eosinophils with Colo-205 led to the release of eosinophil cationic protein and eosinophil-derived neurotoxin as well as TNF-α secretion. Moreover, eosinophils expressed granzyme A, which was released upon interaction with Colo-205, whereas cytotoxicity was partially inhibited by FUT-175, an inhibitor of trypsin-like enzymatic activity. Our data present the first demonstration, to our knowledge, that granzyme A is a cytotoxic mediator of the eosinophil protein arsenal, exerting eosinophil tumoricidal activity toward Colo-205, and provide mechanistic evidence for innate responses of eosinophil against tumor cells.
Eosinophils are multifunctional granulocytes mainly involved in helminth infections and allergic diseases (1, 2). However, in many cancers, notably in carcinoma, eosinophils recruitment, commonly referred to as tumor-associated tissue eosinophilia (TATE), has been observed in some patients in tumor vicinity (3, 4). Several epidemiological studies have attempted to correlate TATE and cancer prognosis or to study the relationships between allergy-related disorders and cancer development (5, 6). Even though the beneficial effect of TATE remains controversial, the presence of eosinophils appears as an indicator of good prognosis indicator in some cancers, such as head and neck cancers (7–9) and gastrointestinal cancers (10), particularly in colorectal cancers, in which high eosinophil counts were associated to increased 5-y survival (11). Clinical observations that eosinophil accumulation in intestinal tumors was associated to a better prognosis have been recently supported by experimental data in murine models (12–14).
Eosinophils are potent effector cells that release several cytotoxic mediators upon activation. Specific granules contain high amounts of cytotoxic basic proteins, major basic protein (MBP), eosinophil cationic protein (ECP), eosinophil peroxidase (EPO), and eosinophil-derived neurotoxin (EDN). These cationic proteins have been described to exert cytotoxic properties for tumor cell lines with variable efficiency (15). Beside cytotoxic mediators, eosinophil granules contain multiple Th1, Th2, and immunoregulatory cytokines that are rapidly and selectively secreted (16). Among them, IFN, IFN-γ, and TNF, which are known to play a major role during antitumor responses, are released by eosinophils (17, 18). Some cellular interactions between eosinophils and tumor cells in situ have also been described (19, 20).
Although eosinophils can be recruited at tumor sites, their activation and direct involvement in antitumor responses have not been extensively investigated. Eosinophil granule proteins that are released upon activation are known to be highly cytotoxic at least in vitro (15), but receptors triggering eosinophil degranulation and mediator release upon contact with tumor cells are unknown. So far, a single study demonstrated the cytotoxic activity of eosinophils toward a B lymphoma cell line in vitro and the implication of 2B4 receptor, a member of the CD2 family, in antitumor responses of eosinophils (21). In a recent study, we showed that eosinophils express the CD3-γδTCR, a receptor shared with the γδ T cells (18). These data suggesting tumoricidal properties of eosinophils led us to further investigate the molecular interactions of eosinophils with tumor cells using the Colo-205 colorectal tumor cell line as target cells.
Materials and Methods
Cell line and reagents
The Colo-205 cell line (human colon carcinoma) was from American Type Culture Collection (Manassas, VA). Cells were grown at 37°C in 5% CO2 in culture medium RPMI 1640 without phenol red (supplemented with 10% FCS, 25 mM HEPES buffer, 2 mM l-glutamine, 10 mM sodium pyruvate, and 10 μg/ml gentamycin), hereafter referred to as complete medium (CM). Adherent tumor cells were harvested from the culture flasks posttreatment with 0.05% trypsin EDTA at 37°C, followed by one wash in CM. Cell viability was determined by trypan blue dye exclusion and was >98%. ECP and EDN purified cationic proteins were purchased from Diagnostics Development (A.P. & M. Venge, Uppsala, Sweden). Bromohydrin pyrophosphate (BrHPP) was a kind gift of Dr. J.J. Fournié (INSERM, Unité 563, Toulouse, France) (22).
Cell purification and culture
Peripheral venous blood was obtained from normal donors (NDs) or eosinophilic patients (allergic subjects [All], patients with skin diseases [Derm], or patients with hypereosinophilic syndrome [HES]) after informed consent. Human eosinophils were isolated as previously described (17) on a Percoll gradient followed by a negative immunomagnetic selection by anti-CD16–coated microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). Purity was determined on cytospin preparations after RAL555 staining and was >98%. Purified human eosinophils were cultured overnight at 37°C in 5% CO2 in CM. γδTCR lymphocytes were generated from PBMCs isolated from heparinized blood over Percoll gradient (23). Briefly, PBMCs were resuspended at a concentration of 0.5 × 106/ml in CM. After a single stimulation with 200 nM BrHPP, a major γ9δ2 TCR selective agonist peptide, γ9δ2 T lymphocytes are expanded for 10 d at 37°C in 5% CO2 in culture medium with recombinant human IL-2 added at day 0 and all 3 d at 20 ng/ml (22, 24). Cells were analyzed by flow cytometry for cell staining or cytotoxicity assays at day 8. At this time point, >60% of CD3+ cells were γ9δ2 T cells. γδT cells tested for cytotoxicity were sorted using an FACSAria cell sorter and Diva software (BD Biosciences, San Jose, CA).
Colo-205 cells were first stained with PKH26 (Sigma-Aldrich, St. Louis, MO; 10 μM) as recommended by the manufacturer. Eosinophil-mediated cytotoxicity against PKH26-labeled Colo-205 was measured in complete medium at 1.6 × 104 targets/well into U-bottom plates containing eosinophils (E:T ratio of 25:1). Apoptosis was measured after Annexin V-FITC (BD Pharmingen, San Diego, CA) staining for 15 min at room temperature. Analysis was performed on an FACScalibur flow cytometer with CellQuest software (BD Biosciences). In some experiments, discrimination between apoptosis and necrosis was achieved by additional staining with 5 μg/ml 7-aminoactinomycin D (7-AAD) (Sigma-Aldrich).
Determination of target necrosis was assessed by using a protocol adapted from the previously described fluorometric assessment of T lymphocyte Ag-specific lysis (FATAL) assay (25). Briefly, Colo-205 cells were first membrane stained with PKH26 and then intracellularly with CFSE. Lysis of Colo-205 was measured by the percentage of PKH26high-positive cells, which presented decreased CFSE fluorescence (in comparison with nonlysed PKH26highCFSEhigh Colo-205).
For inhibition experiments, eosinophils or Colo-205 were preincubated with blocking mouse mAbs (10 μg/ml) for 20 min, then washed, resuspended, and cocultured (200 μl/well, in CM). Blocking Abs were anti-CD11a (R7.1), anti-CD18 (R3.3), anti–ICAM-1 (RR1/1) mAbs (Bender MedSystems, Vienna, Austria), anti-CD11b (CBRM1/5), and anti-VLA4 (P1H4, Abcam, Cambridge, MA); isotype-matched controls were mouse IgG1 (Diaclone, Gen-Probe, San Diego, CA) or total mIgG isotype control (Jackson ImmunoResearch Laboratories, West Grove, PA). Caspase inhibitors (Z-VAD-fmk) (BD Pharmingen) were added to Colo-205 for 30 min at 37°C at 100 μM, preaddition of eosinophils, so that concentration of inhibitors was of 50 μM during coculture.
Cell adhesion assay
Colo-205 adhesion to purified eosinophils was measured by flow cytometry assay as previously described (26, 27). Briefly, eosinophils (1 × 106 5) were mixed in 200 μl CM with PKH26-stained Colo-205 (4 × 104) in a 12 × 75-mm polystyrene tube (BD Biosciences). Postcentrifugation at 400 rpm for 3 min, cells were cocultured for different times at 37°C. The cells were then placed on ice to stop the adhesion process, gently resuspended, and directly analyzed by flow cytometry with low settings. To determine the metabolic requirements for eosinophil binding, CFSE-labeled eosinophils or PKH26+ Colo-205 cells were separately fixed with 1% paraformaldehyde (PFA) for 20 min at 4°C and washed extensively with CM before mixing and coculture at 37°C. The percentage of Colo-205 cells binding to the purified eosinophils was calculated as follows: % binding = (number of Colo-205 bound to eosinophils [PKH26+CFSE+ cells]/total number of Colo-205 [PKH26+ cells]) × 100.
Purified eosinophils were cytocentrifuged on slides for 2 min at 300 rpm. In some experiments, effectors were previously cocultured with Colo-205 at E:T ratio of 5:1 for 3 h at 37°C. Cells were fixed in PBS-4% PFA for 10 min at room temperature. PBS washes were performed between each incubation step. Endogenous fluorescence was quenched by exposition to UV (2 min), followed by 0.1 M glycine for 4 min and 50 mM NH4Cl (pH 7.4). For intracellular staining, cells were permeabilized with PBS-1% BSA-0.2% Triton X-100 for 5 min on ice. Postincubation with PBS-3% BSA supplemented with 5% HSA for 30 min to prevent nonspecific binding, slides were directly incubated overnight at 4°C with primary Abs at 10 μg/ml in PBS-3% BSA-5% HSA. Anti-human eosinophil cationic proteins Abs are mouse mAbs: anti-EPO and -MBP (BD Pharmingen) and anti-ECP and -EDN (Diagnostics Development, Uppsala, Sweden). After washing in PBS-0.1% BSA, cells were incubated with FITC-secondary F(ab′)2
DNase-treated total RNA was prepared using the RNeasy kit (Qiagen, Valencia, CA) from 1072-microglobulin (5′-CAGCGTACTCCAAAGATTCAGGT-3′ and 5′-TGGAGACAGCACTCAAAGTAGAA-3′), 52°C (30 cycles).
For intracellular (IC) staining, purified eosinophils were first fixed with 2% PFA for 10 min and then permeabilized in IC buffer (0.01% saponin-PBS-1% BSA). Nonspecific binding was blocked with mouse serum for 10 min, then the cells were incubated for 30 min in IC buffer with FITC-conjugated mouse mAbs: anti-Perf (δC9) (BD Biosciences), anti-GzmA (CB9), anti-GzmB (GB11) (BD Pharmingen), or matched isotype control mouse IgG1 (MOPC-21). For indirect IC staining, mouse monoclonal IgG1 Abs were used: anti-GNLY (RB1, MBL International, Woburn, MA), anti-granzyme K, and anti-granzyme M (gift from J. Lohrmann and B. Bade, GENOVAC, Freiburg, Germany) (28). After washing and a 10-min blocking step with goat serum, cells were incubated with a goat FITC–anti-mouse IgG1 secondary Ab (Southern Biotechnology Associates) for 20 min. Cells were then washed and immediately analyzed on an FACSCalibur (BD Biosciences).
Eosinophils (2 × 106/ml) were incubated with Colo-205 (E:T ratios of 1:1, 1:2, and 1:5). After 18 h of coculture, supernatants were collected. ECP, EDN, TNF-α, and GzmA levels were measured by specific ELISA kits from MBL International, Diaclone, Hycult Biotechnology (Uden, The Netherlands), and Sanquin (Pelikine Compact, Amsterdam, The Netherlands), respectively. The lower detection limit was 0.125 ng/ml for ECP, 0.62 ng/ml for EDN, 8 pg/ml for TNF-α, and 4 pg/ml for GzmA.
All results were expressed as mean ± SEMs. Statistical analyses were performed using SPSS software (SPSS, Chicago, IL). Normality of data samples was first assessed with the normality test of Shapiro and Wilk, then the parametric Student t test for paired experiments was employed to compare variables. Independent Student t test was realized for experiment testing purified mediator on tumor cell lines. Differences were considered significant for a p value of <0.05; *p < 0.05, **p < 0.01, and ***p < 0.001 are presented in the figures.
Human eosinophils induce Colo-205 cell apoptosis and necrosis
We first investigated the antitumoral activity of purified human eosinophils toward various tumor cell lines (Supplemental Fig. 1). As the highest eosinophil cytotoxic response was obtained against the intestinal carcinoma cell line Colo-205, this cell line was used to further investigate the mechanisms underlying this effect. Tumor cells were cultured in the presence or absence of eosinophils with an E:T ratio of 25:1 for 0.25 to 6 h, and cell death was measured at each time point by Annexin V staining of PKH26-stained Colo-205. Eosinophils were able to rapidly induce target cell death, reaching 60% dead Colo-205 after 6 h as shown on representative histogram (Fig. 1A, 1B).
To discriminate between apoptosis and necrosis of Colo-205, we analyzed necrosis induction of target cells using the FATAL assay (25). Eosinophils induced 15.8 ± 3.3% specific lysis of Colo-205 after 3 h and 23.6 ± 6.0% after 6 h (Fig. 1C, 1D). As FATAL assay has never been tested with eosinophils as effectors, γδ T lymphocytes were used as a positive control. The cytotoxic effect of γδ T lymphocytes toward Colo-205 reached similar level of necrosis after 3 h (24.8 ± 2.7%; p < 0.01). Eosinophils and γδ T cells induced similar results using Annexin V–7-AAD double staining (Fig. 1E, 1F), respectively. Over the brief period of cytotoxicity assessment (0.25–6 h), no significant change in eosinophil viability was observed (Supplemental Fig. 2). Thus, eosinophils display significant cytotoxic functions against Colo-205 by inducing both apoptosis and necrosis.
Requirement of CD11a/CD18-dependent adhesion of Colo-205 to eosinophils for induction of cytotoxicity
Eosinophils might induce cell death through cell–cell contacts and release of soluble factors. A cell–cell adhesion assay was performed to evaluate the role of cellular contacts in Colo-205 cell death induction by eosinophils. Colo-205 adhered to purified eosinophils, reaching 26 ± 3.8% after 1 h of coincubation (n = 6) (Fig. 2A), which is higher than adhesion observed with γδ T cells, in the same conditions (10 ± 2%). To determine whether metabolic changes and membrane dynamics were required for cellular adhesion, Colo-205 or eosinophils were fixed with PFA prior to coculture. Colo-205 binding to eosinophils was strongly inhibited when granulocytes were prefixed with PFA, with inhibition values running from 70% to 53% at 30 min (data not shown). To study the role of cell–cell contact in eosinophil-induced Colo-205 cell death, we performed eosinophil fixation with PFA prior to incubation with Colo-205. This process significantly decreased Colo-205 cell death (Supplemental Fig. 3A). Moreover, separation of effector and targets by a semipermeable membrane prevented the development of cytotoxicity (Supplemental Fig. 3B). Cytotoxicity due to the passive release of cytotoxic mediators by dead/dying bystander eosinophils did not play a major role in the cytotoxicity mechanisms described. Indeed, neither supernatants from an overnight culture of resting or dead eosinophils nor soluble fraction of eosinophil sonicate were able to induce Colo-205 death, suggesting that eosinophil/Colo-205 contacts were essential to the cytotoxic process (Supplemental Fig. 3C, 3D). These data confirm that eosinophil adhesion is a prerequisite to eosinophil cytotoxicity.
Then we examined whether integrins, known to be expressed by eosinophils (1), were involved in such adhesion. Binding of Colo-205 to eosinophils was inhibited in the presence of specific blocking anti-CD11a and anti-CD18 Abs (Fig. 2B). This indicates that CD11a and CD18 may participate in the interaction of eosinophils with target cells. To demonstrate that eosinophil adhesion to Colo-205 was a major event to initiate cytolytic activities, we analyzed whether inhibiting cellular adhesion could reduce cytotoxicity. When blocking mAbs to CD11a or to CD18 were added to eosinophils precoincubation with Colo-205, eosinophil cytolytic activity was decreased by 58.3 ± 5.9% (p < 0.01) and 49.9 ± 27.1% (p < 0.05), respectively (Fig. 2C). These data strongly suggest that CD11a/CD18 play a crucial role in this cytolytic process.
ECP and EDN released by eosinophils are cytotoxic for Colo-205
We next intended to identify the cytotoxic mediators potentially involved in the target cell death observed. We focused on eosinophil-specific cationic proteins, EPO, ECP, and EDN, described to be highly cytotoxic for various cells, organisms, or tissues (1). EPO release could not be detected in culture supernatants by either chemiluminescence or colorimetric assays (data not shown). In contrast, ECP and EDN release was measured in the supernatants of eosinophils postincubation with different ratios of Colo-205 for 18 h (Fig. 3A, 3B). Increased levels of ECP and EDN were detected when eosinophils were cocultured with Colo-205, and dose-dependent EDN release was observed with increasing eosinophil/Colo-205 ratios. We performed intracellular immunofluorescence staining for each cationic eosinophil mediator in eosinophils and Colo-205 cells collected after 3 h of coculture (Fig. 3C). Staining indicates that EDN was detected inside the target cells in contrast to other cationic proteins including ECP or EPO and MBP (Supplemental Fig. 4). It is worth pointing out that Colo-205 cells were not stained with anti-EDN Abs in the absence of eosinophils. To determine whether this eosinophil protein could be internalized by tumor cells, during coculture with eosinophils, incubation of Colo-205 alone with purified EDN (10 μg/ml) was performed during 3 h, followed by immunofluorescence staining of EDN postremoval of extracellular protein by washes. Immunofluorescence staining obtained with anti-EDN suggests that this mediator could be internalized by the Colo-205 (data not shown). The direct effect of both cationic proteins on Colo-205 survival was then tested. Purified ECP and EDN display strong cytolytic activity against Colo-205 (Fig. 3D). Thus, eosinophil cationic proteins might therefore represent potent antitumoral cytotoxic effectors.
TNF released in the presence of eosinophils could be cytotoxic for Colo-205
Besides cationic proteins, other cytotoxic mediators potentially released by eosinophils are cytokines such as TNF (16, 29). We could indeed detect TNF release after overnight culture of eosinophils with Colo-205 (Fig. 4A). This secretion seemed to be dependent upon eosinophil-target adhesion because anti-CD11a and CD18-blocking Abs inhibited the amount of TNF released in supernatants (Fig. 4B). Colo-205 cells were sensitive to TNF, as concentrations of 5 and 10 ng/ml induced, respectively, 15% and 20% of death after overnight incubation. This cytotoxic effect on tumor cells was inhibited by blocking anti-TNF Ab used at 10 μg/ml (Fig. 4C). Taken together, these findings indicate that Colo-205 cells are sensitive to TNF, a cytokine released during coculture. This suggests that this mediator could play a role in Colo-205 cell death induced by eosinophils.
Eosinophils express and release GzmA in presence of Colo-205
Colo-205 cell death measured by Annexin V staining was not decreased in the presence of the caspase inhibitor Z-VAD-fmk, suggesting that Colo-205 cell death was caspase independent (Supplemental Fig. 5). This led us to evaluate whether cytotoxic mediators known to induce caspase-independent cell death could be implicated. Based on the expression of various cytotoxic molecules by NK cells or γδT cells, we focused on cytolytic mediators such as Perf, GNLY and serine protease family members, and granzymes, potent effector molecules in innate immunity against tumors and virus-infected cells (30). By RT-PCR, we solely detected expression of mRNA encoding GzmA but not Perf, GNLY, or other granzyme mRNA (Fig. 5A). CD8 mRNA was assessed to exclude some T lymphocyte contamination in eosinophil mRNA samples. GzmA was also detected in eosinophils by flow cytometry (Fig. 5B). Taken together, our results show a constitutive expression of GzmA in eosinophils, leading us to hypothesize that eosinophils may respond to Colo-205 by rapid release of GzmA. We detected GzmA release by eosinophils during coculture with Colo-205 (Fig. 5C). Eosinophil GzmA production was dependent upon increasing Colo-205 numbers. To determine the role of GzmA in the tumoricidal activity of eosinophils, we used a highly rapid, potent, and selective inhibitor of GzmA, FUT-175 (nafamostat mesilate) (31), which led to a partial but significant inhibition of Colo-205 cell death (Fig. 5D). No inhibition by FUT-175 was observed in EDN-, ECP-, or TNF-α–induced cell death of Colo-205 (Supplemental Fig. 6). Taken together, these findings indicate that GzmA is involved in eosinophil-mediated apoptosis of Colo-205.
Cytotoxicity of GzmA and ECP toward Colo-205
We studied the direct effect of purified GzmA on Colo-205 survival. GzmA alone, used at 10 μg/ml, exerted limited but significant proapoptotic activity against Colo-205 (Fig. 6A). However, GzmA could increase the proapoptotic effect of other tumoricidal mediators such as TNF, as shown in Fig. 6B. We also investigated whether GzmA could synergize with ECP, a potent cytotoxic mediator expressed by eosinophils (Fig. 6C). Additive effect between these two eosinophil compounds was detected when ECP was used at the concentration of 5 μg/ml. We then further studied this synergistic effect on Colo-205 cell death induction using Annexin V–7-AAD double staining (Fig. 6D). Addition of GzmA significantly increased the apoptotic and necrotic effect of ECP.
Besides helminth infections and allergic diseases, peripheral blood and tissue eosinophilia is associated with several tumors (3, 4), including intestinal tumors. However, as only very few studies have addressed the role of eosinophils during intestinal carcinoma, their impact on tumor cell development remains obscure, and functional data are particularly missing. We have investigated in this study the mechanism of interactions between eosinophil and colorectal carcinoma cells to provide some molecular basis for this observation.
Initial results indicate that purified eosinophils were able to induce tumor cell death of various cell lines with some selectivity in their tumoricidal properties (Supplemental Fig. 1). This study demonstrates that eosinophils could induce Colo-205 cell apoptosis and to a lesser extent necrosis. We observed that the tumoricidal effect of eosinophils was dependent on CD11a/CD18-mediated stable contacts with the target cells. As for T lymphocytes, eosinophil adhesion to target cells is a prerequisite for cytotoxicity by providing stable contacts with target cells. An interesting parallel could be drawn with the previously described cytotoxicity of human eosinophils toward Schistosoma mansoni parasitic larvae, which required adhesion molecules acting as coreceptors in this Ab-dependent cellular cytotoxicity mechanism (32). Being short-lived cells, eosinophils ultimately die upon activation and degranulation, as shown in antiparasite cytotoxicity reactions. Indeed, eosinophils are not classical cytotoxic cells such as CD8 or γδ T cells, which usually do recover and recycle following the cytotoxic process. In keeping with these differences, addition of IL-5 during overnight culture only marginally improved eosinophil survival (84 ± 5% versus 91 ± 2% upon IL-5 supplementation [n = 4, data not shown]). In addition, no major differences in the tumoricidal potential of eosinophils toward Colo-205 were observed in the presence or absence of IL-5 (Supplemental Fig. 7). Cytotoxicity due to the passive release of cytotoxic mediators by dead/dying bystander eosinophils did not play a major role in the cytotoxicity mechanisms described in the current study.
Concerning membrane interactions between the two cellular counterparts, several hypotheses have been explored. First, we demonstrated that LFA-1 was primordial for eosinophil adhesion to Colo-205 and subsequent activation of the granulocyte. LFA-1–ICAM-1 interaction was suggested to play a role in establishing the stable adhesion between γδT cells and tumor cells loaded with nonpeptide Ags (33). However, we could not detect inhibition with anti–ICAM-1 mAbs, suggesting that LFA-1 recognizes another counterreceptor expressed on Colo-205. Contradictory results have been reported regarding γδ T lymphocyte cytotoxic potential toward the Colo-205 cell line (34). Whereas we showed that γδ T cells induce cytotoxicity toward Colo-205, similar to eosinophils, negative results have been reported, suggesting that γδ anergy arises from a lack of ICAM-1 expression by Colo-205 (34). Indeed, in the current study, we could detect some weak membrane ICAM-1 expression on Colo-205. Supplementary evidence favoring the role of LFA1 role in the tumoricidal eosinophil activity relied on the observation that eosinophils purified from allergic donors, described to express an ncreased level of membrane CD11a (35), displayed exacerbated cytotoxicity toward Colo-205 compared with nonallergic donors (Supplemental Fig. 8). No expression of classic activation membrane markers CD69 and HLA-DR could be observed. In addition to these adhesion receptors, other membrane receptors participate in this cytotoxicity process. Indeed, we recently demonstrated that human eosinophils express γδ TCR and showed its involvement in Colo-205 cell death induction (18). Another receptor shared between T cells and eosinophils, 2B4, a member of the CD2 family, has also been involved in the cytotoxic activity of eosinophils toward a B lymphoma cell line (21). However, in our study concerning a different target cell, the colorectal carcinoma cell line, 2B4 only played a minor role in tumoricidal activity (data not shown).
We then demonstrated that Colo-205 could induce the release by eosinophils of different mediators potentially involved in tumor cell death: eosinophil cationic proteins, ECP, and EDN but also other lymphoid-associated tumoricidal soluble factors, TNF and GzmA. Concerning eosinophil-specific cationic proteins, ECP alone was found to be released and to induce Colo-205 cell death. ECP is a potent cytotoxic protein involved in the killing of different target cells including leukemia and carcinoma cells (reviewed in Ref. 33). This recent study reports that ECP increased TNF production by the bronchial epithelial cell line BEAS-2B, leading to apoptosis in a caspase-dependent manner, a mechanism thus distinct from the mechanism described in this study (36). Release of EDN by eosinophils was also observed as well as its subsequent internalization by tumor cells. In addition, we demonstrated that eosinophils released other cytotoxic mediators in the presence of Colo-205, including TNF and, interestingly, GzmA. As previously described (37), Colo-205 cells are sensitive to TNF, and this latter plays a role in Colo-205 cell death observed during incubation with eosinophils. A screening for Perf, GNLY, and granzyme families by RT-PCR and cytometry approaches revealed the expression by resting eosinophils of GzmA, a mediator never previously described in this latter cell type. This serine protease with trypsin-like specificity was classically found, together with saposin-like proteins, in the large granules of activated lymphoid cells: cytotoxic T lymphocytes, NK cells, and γδ T cells (38). Expression profiles and tumoricidal activation pathways have recently been questioned. Indeed, expression of GzmA could be found in the absence of Perf, such as in the majority of small intestinal intraepithelial lymphocytes, which do not express Perf, but where GzmA was found in 30% of cells. Besides eosinophils, neutrophils have also been reported to express GzmA (39, 40). GzmA is believed to act by entering tumor cells via pores formed by Perf, which participate in the apoptosis of abnormal cells (30, 38). We could not detect Perf expression by eosinophils. Recent results, however, suggest that, on one hand, Perf is not necessary for the entry of granzyme, raising doubt about the molecular mechanisms involved (41), and, on the other hand, that GzmA might be involved in vivo in enterocyte exfoliation through cleavage of extracellular matrix components (42, 43), stressing its putative toxic properties toward intestinal tissue. Purified GzmA induces detachment of the intestinal epithelial cell line or to induce rapid caspase-independent apoptosis of target cells in vitro (30, 44). A recent study also compared the cytolytic enzyme content of several lymphocyte populations (CD8+ or CD4+ αβ T cells and Vδ1, Vδ2 γδ T cell subsets) (45). Across these various T cell types, a characteristic pattern for Perf/granzyme expression has been observed. Notably, cells frequently expressed GzmA alone; many GzmA+GzmB− cells expressed low levels of Perf, and a small population of CD4+ αβ GzmA+GzmB− cells that did not express Perf at all was identified (45).
If Perf was indeed needed for proapoptotic properties of GzmA, we would speculate that in the case of eosinophils, the pore-forming effect of Perf could be substituted by other mediators, such as ECP, which are released during incubation of eosinophils with Colo-205. This eosinophil protein has been described for its tumoricidal effect (46–49) and pore membrane-forming properties (50, 51). A recent study showed that ECP was not internalized (52). Membrane immunofluorescence pathway of ECP released during coculture was consistent with a pore-forming effect.
In conclusion, our results present not only the description of CD11a/CD18-mediated tumoricidal activity of eosinophils toward a colorectal carcinoma cell line, but also the first evidence, to our knowledge, of GzmA expression and release by human eosinophils upon stimulation by the Colo-205. This study highlights a mechanism of direct recognition of tumor cells by human eosinophils, thereby broadening their functional importance as early direct sensors and effectors against tumors. Due to its functional plasticity and its strategic tissue localization, the eosinophil might play a previously unsuspected role in tumor immunosurveillance.
We thank Dr. B. Bade and J. Lohrman (Freiburg, Germany) for providing Abs against granzymes and Dr. J.J. Fournié for providing BrHPP. We also thank Gaëtane Woerly (Novartis Pharma, Basel, Switzerland) for technical assistance and Dr. J.E. Kahn and Prof. L. Prin of the French Hypereosinophilic Network for assistance with patient scheduling.
Disclosures The authors have no financial conflicts of interest.
This work was supported by grants from the Conseil Régional Nord-Pas-de-Calais and Agence Nationale pour la Recherche.
The online version of this article contains supplemental material.
Abbreviations used in this paper:
- 7-aminoactinomycin D
- allergic subject
- bromohydrin pyrophosphate
- complete medium
- patient with skin disease
- eosinophil cationic protein
- eosinophil-derived neurotoxin
- eosinophil peroxidase
- fluorometric assessment of T lymphocyte Ag-specific lysis
- granzyme A
- granzyme B
- patient with hypereosinophilic syndrome
- major basic protein
- normal donor
- tumor-associated tissue eosinophilia.
- Received February 9, 2010.
- Accepted October 9, 2010.