CD8⁺ T Cells Induce Complete Regression of Advanced Ovarian Cancers by an Interleukin (IL)-2/IL-15–Dependent Mechanism

Purpose:In vitro studies suggest that ovarian cancer evades immune rejection by fostering an immunosuppressive environment within the peritoneum; however, the functional responses of ovarian cancer–specific T cells have not been directly investigated in vivo. Therefore, we developed a new murine model to enable tracking of tumor-specific CD8⁺ T-cell responses to advanced ovarian tumors.

Experimental Design: The ovarian tumor cell line ID8 was transfected to stably express an epitope-tagged version of HER-2/neu (designated Neu^OT-I/OT-II). After i.p. injection into C57BL/6 mice, ID8 cells expressing Neu^OT-I/OT-II gave rise to disseminated serous adenocarcinomas with extensive ascites. CD8⁺ T cells expressing a transgenic T-cell receptor specific for the OT-I epitope of Neu^OT-I/OT-II were adoptively transferred into tumor-bearing mice, and functional responses were monitored. Cytokine signaling requirements were evaluated by comparing the responses of wild-type donor T cells with those with genetic deletion of the interleukin (IL)-2/IL-15 receptor β subunit (CD122) or the IL-2 receptor α subunit (CD25).

Results: On adoptive transfer into tumor-bearing hosts, wild-type OT-I T cells underwent a striking proliferative response, reaching peak densities of ∼40% and ∼90% of CD8⁺ T cells in peripheral blood and ascites, respectively. OT-I cells infiltrated and destroyed tumor tissue, and ascites completely resolved within 10 days. By contrast, CD122^−/− OT-I cells and CD25^−/− OT-I cells proliferated in blood but failed to accumulate in ascites or tumor tissue or induce tumor regression.

Conclusions: Contrary to expectation, advanced ovarian cancers can support extraordinary CD8⁺ T-cell proliferation and antitumor activity through an IL-2/IL-15–dependent mechanism.

Epithelial ovarian cancers disseminate in a peritoneal environment that is rich in potentially tumor-reactive lymphocytes. Moreover, ascites contains many factors that could potentially promote T-cell responses, such as interleukin (IL)-6 (1), IL-15 (2), CD40 (3), and LETAL (4). Consistent with this, human ovarian tumors frequently trigger cellular and/or humoral immune responses to a variety of tumor-associated proteins, including HER-2/neu (5). Despite immune recognition, however, ovarian tumors routinely undergo metastatic progression in immunocompetent individuals; therefore, they must deploy one or more immune evasion mechanisms to prevent immune rejection. Indeed, many immunosuppressive factors have been identified in advanced ovarian cancer, including IL-10 (6), Fas ligand (7), transforming growth factor-β1 (8), B7-H1 (9), B7-H4 (10), SDF-1 (11), EBAG9/RCAS1 (12), soluble IL-2 receptor (13), and PD-L1 (14). Immunosuppressive cell populations have also been found in ovarian cancer, including macrophages (10), plasmacytoid dendritic cells (11), IL-10–producing CD8⁺ cells (15), and CD4⁺CD25⁺FoxP3⁺ regulatory T cells, the latter of which correlate to decreased survival (16–18).

Despite these potentially immunosuppressive factors, several recent studies have shown that the presence of intratumoral T cells in ovarian cancer correlates to a favorable clinical outcome after standard treatments (19, 20). Specifically, tumors that contain a high ratio of CD8⁺ T cells to FoxP3⁺ regulatory T cells are associated with prolonged overall survival (14, 18). These findings agree with earlier studies showing a positive correlation between survival and expression of IFN-γ, a cytokine released by activated CD8⁺ T cells (21). Thus, it seems that, in many patients undergoing standard treatments, endogenous T cells are able to overcome immunologic barriers to mediate an antitumor effect. With better understanding of the signals or conditions that promote T-cell responses against ovarian cancer, it may be possible to improve clinical outcomes further.

The survival and proliferation of naive, activated, and memory T cells is largely controlled by cytokines. IL-2 was the first T-cell growth factor to be molecularly defined and is used to promote T-cell proliferation in vitro and, in the case of certain cancers, in vivo. However, IL-2–deficient mice are largely immunocompetent, which underscores the role of other T-cell growth factors in vivo. Indeed, IL-7, IL-15, and IL-21 have been shown to promote T-cell proliferation in vitro and in vivo, with greater potency than IL-2 in many circumstances (22–24). A general view has emerged that cytokines play specific physiologic roles according to T-cell subset, stage of differentiation, and anatomic location. For example, in wild-type (WT) hosts undergoing conventional immunization, neither IL-2 nor IL-15 is required for the primary expansion of CD8⁺ T cells in blood and lymph node (25) nor to mount antiviral responses or reject tissue grafts (26). However, other studies have shown an essential role for IL-2 in the expansion and survival of regulatory T cells in the periphery (27–29) and in suppressing the differentiation of Th17 T cells (30). IL-2 is also essential for maximal expansion of CD8⁺ effector T cells in nonlymphoid compartments, such as gut and lung epithelium (31, 32), and for programming memory CD8⁺ T-cell responses (33). For its part, IL-15 is required for the long-term maintenance of CD8⁺ memory T cells (34, 35). Thus, the physiologic roles of cytokines, such as IL-2 and IL-15, are complex and highly context dependent.

Much less is known about the signals that govern the expansion, differentiation, and persistence of CD8⁺ T cells in different tumor environments, which can be highly complex mixtures of epithelial, stromal, and hematopoietic cell types. To better understand these issues in the setting of advanced ovarian cancer, we developed a novel mouse model that allows precise tracking of tumor-reactive CD8⁺ T-cell responses. We show for the first time that, contrary to expectation, ascites can be a favorable environment for the clonal expansion and function of CD8⁺ T cells. Furthermore, we show that IL-2/IL-15 signaling, although dispensable for the primary expansion of CD8⁺ T cells in tumor-free hosts, is absolutely essential for CD8⁺ T-cell responses in the ovarian tumor environment.

ID8-G7 tumor model. All mice were on the C57BL/6 background and, except where indicated, were obtained from The Jackson Laboratory. Mice were maintained at the Animal Care Unit of the University of Victoria under standard conditions. All protocols followed the guidelines of the Canadian Council for Animal Care and were approved by the Animal Care Advisory Committee of the University of Victoria. Ovarian tumors were established in mouse mammary tumor virus (MMTV)/neu^OT-I/OT-II transgenic mice (36) with the ovarian cancer cell line ID8-G7, which was derived from the ID8 cell line (Supplementary Data; ref. 37). Adoptive transfer and analysis of CD8⁺ OT-I and P14 T cells was done essentially as described (36) with some modifications (Supplementary Data).

Bromodeoxyuridine incorporation assay. Bromodeoxyuridine (Invitrogen) was injected at 1 mg/0.4 mL PBS into the peritoneal cavity. Two hours later, ascites and solid tumor were harvested. Cells were stained for CD8 and Thy1.1 at 4°C. Subsequent steps were done in the dark. Cells were fixed with 1% paraformaldehyde plus 1% Tween 20 for 15 min at 37°C and permeabilized with 2% Tween 20/10% DMSO solution for 15 min at room temperature. After fixing a second time with 4% paraformaldehyde for 15 min, cells were treated with DNase I to expose incorporated bromodeoxyuridine, blocked with 6% goat serum for 15 min, and stained with Alexa Fluor 488–conjugated anti-bromodeoxyuridine antibody (Molecular Probes). Cells were washed once between each step with PBS + 1% FCS.

Generation and characterization of the ID8-G7 ovarian tumor cell line. To study T-cell responses to ovarian tumors in immunocompetent mice, we used a previously described strategy, in which CD4⁺ and CD8⁺ T-cell epitopes from the model antigen ovalbumin were attached to the COOH terminus of rat neu to generate a fusion protein designated Neu^OT-I/OT-II (36). An expression vector containing the neu^OT-I/OT-II cDNA was transfected into the ovarian tumor cell line ID8 (37). Because ID8 cells are already transformed, they do not require neu^OT-I/OT-II for tumorigenesis. Hence, the WT allele of neu was used instead of the activated allele described previously (36). Although not essential for tumor formation, Neu nevertheless served as a convenient, traceable, and physiologically relevant carrier protein for the ovalbumin epitopes.

We selected one transfected subclone, designated ID8-G7, for all subsequent experiments (Supplementary Data). Flow cytometric analysis revealed that ID8-G7 cells expressed Neu^OT-I/OT-II, MHC class I, and the OT-I epitope (SIINFEKL) complexed with MHC class I (Fig. 1A). These cells also weakly expressed Fas but failed to express MHC class II, the costimulatory molecules B7-1 and B7-2, or Fas ligand. After treatment with 200 units/mL IFN-γ for 20 h, cells showed increased expression of MHC class I, the SIINFEKL/MHC class I complex, and Fas, and MHC class II became expressed. However, cells remained negative for B7-1, B7-2, and Fas ligand (Fig. 1A; data not shown).

ID8-G7 cells give rise to disseminated ovarian cancer in mice. Initially, we injected ID8-G7 cells (1 × 10⁶ per mouse) i.p. into syngeneic WT C57BL/6 mice in an attempt to form ovarian tumors. However, these cells were spontaneously rejected by WT mice, whereas parental ID8 cells or ID8 cells expressing WT Neu readily formed tumors (data not shown). We concluded that the OT-I and OT-II epitopes on Neu^OT-I/OT-II were triggering endogenous CD8⁺ and/or CD4⁺ T cells to reject ID8-G7 cells.

We have previously described transgenic mice that express Neu^OT-I/OT-II in the mammary epithelium under the control of the MMTV promoter for use as a model of spontaneous breast cancer (36). Because MMTV/neu^OT-I/OT-II mice are tolerant to the Neu^OT-I/OT-II protein and do not develop mammary tumors until >12 months of age, we reasoned that at a younger age they could serve as hosts for Neu^OT-I/OT-II-expressing ID8-G7 cells. Indeed, when MMTV/neu^OT-I/OT-II mice were injected i.p. with the ID8-G7 cell line, disseminated ovarian tumors formed in >90% of cases, with an average latency of 32 days. Peritoneal tumor nodules were generally palpable and were associated with extensive ascites, similar to tumors induced by parental ID8 cells (37). The tumor nodules comprised poorly differentiated epithelial carcinoma cells without obvious glandular differentiation. The tumor cells were organized in clusters separated by collagenous stroma and generally had large vesicular nuclei, prominent multiple nucleoli, and frequent mitotic figures (Fig. 1B and C). Tumor nodules were restricted to the peritoneal cavity and implanted on multiple organs, including pancreas, intestine, kidney, and peritoneal wall. Mice typically became moribund and required euthanasia within 40 days of cell line injection. Thus, MMTV/neu^OT-I/OT-II transgenic mice readily accept ID8-G7 tumor cells, likely owing to the fact that Neu^OT-I/OT-II represents a self-protein in this model. This resembles many human ovarian cancers in which HER-2/neu, a self-protein, is overexpressed by tumor cells and hence can potentially be targeted by immunotherapy (5).

Antigen-specific CD8⁺ T cells proliferate vigorously in response to established ovarian tumors. To analyze the response of antigen-specific CD8⁺ T cells to ovarian tumors in vivo, 3 × 10⁶ donor OT-I cells were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) and adoptively transferred into mice bearing established ID8-G7 ovarian tumors. At the time of adoptive transfer (typically 30-40 days after implantation of ID8-G7 cells), tumors were palpable and/or had visible ascites. Control tumor-bearing mice received a similar dose of CD8⁺ T cells expressing an irrelevant transgenic T-cell receptor obtained from P14 donor mice. OT-I T cells (but not P14 T cells) began to proliferate in tumor-bearing mice within 3 days of adoptive transfer, as evidenced by dilution of the CFSE label (Fig. 2A; data not shown). By day 9, they constituted from 5% to 40% of the CD8⁺ T-cell population in blood (Fig. 2B) and from 45% to 96% of CD8⁺ T cells in ascites. By contrast, OT-I cells showed little or no proliferation when adoptively transferred into tumor-free MMTV/neu^OT-I/OT-II mice (data not shown). Moreover, when OT-I cells were transferred into WT mice and stimulated by s.c. injection of whole ovalbumin protein (1 mg in PBS), they proliferated transiently but never reached >10% of CD8⁺ T cells in peripheral blood (Fig. 2C). Thus, in tumor-bearing mice, OT-I cells underwent an exceptionally large, sustained expansion despite the presence of extensive ascites.

Antigen-specific CD8⁺ T cells infiltrate ovarian tumors and induce regression. ID8-G7 tumors from untreated mice showed very sparse CD3⁺ T-cell infiltrates, as did tumors from mice infused with P14 CD8⁺ T cells expressing an irrelevant transgenic T-cell receptor (Fig. 3A; data not shown). Consistent with this, such tumors lacked FoxP3⁺ cells and had only small numbers of granzyme B⁺ cells, which were largely confined to stroma (Fig. 3B and C). In contrast, mice that received donor OT-I cells showed marked infiltration of CD3⁺ lymphocytes in tumor nodules by day 7 after adoptive transfer (Fig. 3D). These infiltrates contained granzyme B⁺ cells (Fig. 3E) as well as a smaller number of FoxP3⁺ cells, suggesting they contained predominantly effector T cells with a minor regulatory T-cell component (Fig. 3F). T-cell infiltration was accompanied by extensive cell death and disintegration of the tumor tissue, consistent with a cytolytic T-cell response (Supplementary Fig. S1A). Flow cytometric analysis of freshly excised tumor tissue revealed that OT-I was the dominant clone in these infiltrates, ranging from 60% to 90% of total CD8⁺ T cells (Supplementary Fig. S1B).

Mice were also monitored for changes in ascites over time. All mice receiving P14 T cells had to be euthanized by day 10 due to progressive growth of ascites and tumor (Table 1). By contrast, 75% of mice receiving OT-I T cells achieved complete resolution of ascites and loss of all palpable tumors by day 10 (Table 1). Necropsy of a subset of these mice on day 10 confirmed that most tumor nodules had disappeared and only a few small fibrotic nodules (≤3 mm) remained. Mice remained healthy for ∼3 weeks, and OT-I T cells remained abundant in peripheral blood, constituting 2% to 16% of CD8⁺ T cells. However, despite the continued presence of OT-I cells, tumors recurred in all cases within 3 to 4 weeks of adoptive transfer.

We investigated antigen loss as a potential mechanism of immune evasion. Those tumors that failed to regress initially (∼25%) were found to still express Neu^OT-I/OT-II, MHC class I, and MHC class II (Supplementary Fig. S2); therefore, antigen loss did not account for the failed T-cell response in these cases. In contrast, most recurrent tumors (seven of nine cases tested) no longer expressed Neu^OT-I/OT-II, suggesting they evaded immune rejection through antigen loss. Nevertheless, a small subset of recurrent tumors remained positive for Neu^OT-I/OT-II, MHC class I, and MHC class II (Supplementary Fig. S2), indicating that other evasion mechanisms were also operant. Recurrent tumors generally showed moderate-to-dense CD3⁺ T-cell infiltrates, with high numbers of granzyme B⁺ cells and low numbers of FoxP3⁺ cells (Fig. 3J-L). Thus, many tumors recurred despite containing a seemingly favorable ratio of effector to regulatory T cells.

The antitumor effect of OT-I cells is dose dependent. Given the ability of OT-I T cells to undergo major expansion in response to ID8-G7 tumors, we reasoned that lower starting doses of OT-I T cells might be able to expand to the same levels and mediate an antitumor effect. To test this, we reduced the transferred cell dose by an order of magnitude (i.e., from 3 × 10⁶ to 3 × 10⁵ OT-I cells per mouse). At the lower dose, the vast majority of OT-I cells still proliferated after adoptive transfer, as measured by a loss of CFSE labeling by day 3 in peripheral blood (data not shown). Moreover, the number of peripheral OT-I cells increased from a starting point of <0.1% of CD8⁺ T cells to a peak of 1% to 2% of CD8⁺ T cells on days 6 to 9, which represented at least a 10-fold expansion (Fig. 4A). Nevertheless, in no case did the donor T cells ever exceed the 2% mark in peripheral blood, indicating that their proliferative response was constrained.

We next evaluated donor OT-I cells in ascites and solid tumor. Irrespective of the starting dose (i.e., 3 × 10⁵ or 3 × 10⁶ cells), OT-I cells represented the dominant clone within the CD8⁺ infiltrate in ascites (Fig. 4B). Furthermore, in vivo bromodeoxyuridine labeling revealed that OT-I cells were actively cycling in ascites and solid tumor, again irrespective of starting dose (Fig. 4C). However, the absolute number of T cells in ascites and solid tumor was less in mice that received a lower starting dose of OT-I cells, as evidenced by a decreased number of flow cytometric events per unit of ascites (data not shown) and reduced infiltration of tumor tissue by CD3⁺ cells (Fig. 3G). The number of tumor-infiltrating granzyme B⁺ and FoxP3⁺ cells was also reduced, suggesting that the proportion of effector to regulatory T cells remained roughly constant despite a lower starting dose of OT-I cells (Fig. 3H and I). No tumor regressions were observed at the lower OT-I cell dose (Table 1). Indeed, an intermediate dose of OT-I cells (15 × 10⁵ per mouse) also failed to induce tumor regression, although this represented only a 2-fold reduction from the curative dose (Table 1). Collectively, these results suggest that the clonal expansion of OT-I cells in response to ovarian tumors is programmed, with no evidence of compensatory proliferation when lower doses of OT-I cells are given. We therefore asked what factors control the clonal expansion of antigen-specific CD8⁺ T cells in the ovarian cancer environment.

IL-2/IL-15 signaling is required for CD8⁺ T-cell responses in ovarian cancer. As discussed above, IL-2 signaling is dispensable for the primary expansion of CD8⁺ T cells in peripheral blood and lymph node but essential for sustained T-cell proliferation in epithelial tissues, such as gut and lung (25, 31, 32). Given that the ovarian tumor environment contains both fluid (i.e., ascites) and epithelial components, we asked whether IL-2 signaling was required for the expansion and antitumor activity of OT-I cells in this setting.

To address this initially, we evaluated expression of the IL-2 receptor α subunit (CD25) on adoptively transferred OT-I cells as they responded to ID8-G7 tumors. OT-I cells isolated from lymph node on day 7 after adoptive transfer failed to express CD25 (Supplementary Fig. S3), consistent with our prior studies in tumor-free mice (25). By contrast, the majority of OT-I cells isolated from ascites and solid tumor on day 7 expressed CD25 (Supplementary Fig. S3), raising the possibility that IL-2 could indeed be involved in proliferative signaling once T cells enter the ovarian tumor environment.

To test this hypothesis, we did adoptive transfer experiments with donor OT-I cells that lacked expression of CD122, an essential signaling subunit of both the IL-2 and IL-15 receptors (38). It is technically difficult to obtain naive donor T cells from CD122-deficient mice, as these mice develop severe lymphadenopathy within 3 weeks of birth (39). However, this problem can be prevented by thymic expression of a CD122 transgene, which rescues development of regulatory T cells (40). Thus, CD122^−/− mice that express a thymic CD122 transgene (designated CD122^−/−Tg⁺ mice) produce peripheral T cells that have a naive phenotype yet are functionally CD122 deficient and refractory to IL-2 or IL-15 signaling (40).

We generated CD122^−/−Tg⁺ donor OT-I cells and confirmed that they lacked expression of CD122 yet retained a naive phenotype (CD62L^high, CD44^low, CD69^low; Supplementary Fig. S4). After adoptive transfer into mice bearing ID8-G7 tumors, both WT and CD122^−/−Tg⁺ OT-I cells proliferated in peripheral blood, as evidenced by loss of CFSE florescence by day 7 (data not shown). However, the number of CD122^−/−Tg⁺ OT-I cells was generally lower than WT OT-I cells (Fig. 5A). This difference was even more striking in ascites and solid tumor, where CD122^−/−Tg⁺ OT-I cells represented <3% of the CD8⁺ infiltrate compared with approximately 60% to 80% for WT OT-I cells (Fig. 5B). Consistent with this, CD122^−/−Tg⁺ OT-I cells failed to induce tumor regression (Table 1). This suggests that IL-2/IL-15 signaling is dispensable for the initial proliferation of OT-I cells but absolutely required for accumulation of OT-I cells in the ascites and tumor environment.

The above results could be attributed to a loss of either IL-2 or IL-15 signaling, or both, in OT-I cells. To specifically elucidate the contribution of IL-2 signaling, we did adoptive transfer experiments with OT-I cells deficient for CD25. As with CD122^−/− mice, T cells from CD25^−/− mice undergo spontaneous activation in vivo, which makes it difficult to obtain naive donor T cells. This problem was largely ameliorated in the present experiments by expression of the OT-I T-cell receptor transgene as well as the use of young donor mice (41). Indeed, before adoptive transfer, the vast majority of donor CD25^−/− OT-I cells showed a naive phenotype (CD44^low, CD62L^high, CD69^low), and only a small percentage showed up-regulated expression of CD44 (Supplementary Fig. S5).

In control experiments, we first transferred CD25^−/− OT-I cells into tumor-free mice followed by immunization with ovalbumin. Consistent with prior studies (25), CD25^−/− OT-I cells showed a similar proliferative response in peripheral blood as WT OT-I cells (Fig. 5C). In contrast, when CD25^−/− OT-I cells were transferred into mice bearing ID8-G7 tumors, they still proliferated in peripheral blood but to a lesser extent than WT OT-I cells (Fig. 5D). An even more striking difference was seen in ascites and solid tumor, where the number of CD25^−/− OT-I cells ranged from 2% to 7% of total CD8⁺ T cells compared with 25% to 65% for WT OT-I cells (Fig. 5E). Accordingly, tumor regressions were never observed in mice receiving CD25^−/− OT-I cells (Table 1). Thus, IL-2 signaling through CD25 is required for effective CD8⁺ T-cell responses against advanced ovarian tumors.

Ovarian cancer has a complex immunobiology owing to the extensive ascites that typically accompanies advanced disease. Numerous immunosuppressive components have been identified in the ovarian tumor environment, including molecules such as IL-10, Fas ligand, transforming growth factor-β1, and EBAG9/RCAS1 as well as suppressive cell populations such as regulatory T cells, macrophages, plasmacytoid dendritic cells, and IL-10–producing CD8⁺ cells (6–8, 11, 42). Despite these potential barriers, several recent studies have shown a correlation between the presence of intratumoral CD8⁺ T cells in ovarian cancer and increased progression-free survival, suggesting that ovarian cancer may be an immunologically responsive disease (18–20). Indeed, we show here in a new mouse model of ovarian cancer that adoptively transferred CD8⁺ T cells can mount effective antitumor responses that are sufficient to cause regression of highly advanced disease. This was associated with extensive T-cell proliferation within ascites and solid tumors through an IL-2/IL-15–dependent pathway. Our results support the concept that, contrary to prior assumptions, ovarian cancer may represent a highly favorable setting for T-cell–based immunotherapy.

A major barrier to the successful adoptive immunotherapy of cancer is the poor persistence and expansion of T cells after transfer. Recent progress has been made in the setting of advanced melanoma through pharmacologic depletion of the lymphoid compartment before transfer. Such protocols can increase the frequency of circulating tumor-reactive T cells to the point that they represent >90% of the CD8⁺ T-cell compartment, likely due to the removal of suppressive cell populations and the creation of a favorable cytokine milieu (43, 44). Unfortunately, this comes at the expense of considerable morbidity due to the lymphodepletion regimen, the use of high-dose IL-2, and the impairment of host antiviral responses. Moreover, the persistence of tumor-reactive T cells remains a limitation in many patients, apparently due in part to telomere shortening during T-cell proliferation (45). Ultimately, one would hope to refine such adoptive immunotherapy protocols to the point where adequate T-cell expansion can be achieved with less intensive regimens.

In this study, we observed striking T-cell expansion without the need for lymphodepletion, exogenous cytokines, donor CD4⁺ T cells, or vaccination. This shows that, under the right conditions, a tumor-reactive T-cell clone can overtake the CD8⁺ T-cell compartment for extended periods, resulting in an effective antitumor response. Our findings may be attributable to several factors. First, tumor cells were made to express a foreign, high-affinity CD8⁺ T-cell epitope. Importantly though, we used host mice that also expressed this epitope in mammary epithelium; therefore, from an immunologic perspective, the SIINFEKL epitope represented a self-protein that was overexpressed by the tumor, similar to the situation with HER-2/neu and many other human tumor antigens.

Second, we used a T-cell clone with high affinity for the SIINFEKL epitope. Indeed, OT-I T cells have shown remarkable efficacy in several other tumor and infectious disease models (46, 47). Although CD8⁺ T cells with similarly high avidity for tumor antigens are relatively rare in human cancer patients, they can be more readily isolated using tetramer technology (48). Moreover, advances in T-cell receptor engineering may ultimately bypass the need to identify and isolate naturally occurring T cells with such properties (49, 50).

Third, T cells were in a naive state at the time of adoptive transfer and hence possessed a greater proliferative capacity than the activated or exhausted T cells that are typically obtained from cancer patients (51, 52). This important advantage is more difficult to translate into the clinic. Approaches may include the use of CD8⁺ T cells with a central memory phenotype, which also have high proliferative capacity (53). More speculatively, recent progress with in vitro methods to promote the differentiation of T cells from hematopoietic stem cells may in future provide a source of naive human T cells from patients (54).

Fourth, the starting dose of OT-I cells was a critical determinant of the antitumor response. At lower doses, OT-I cells expanded proportionately but failed to achieve a density sufficient for tumor infiltration and destruction. Studies in other models have shown that naive CD8⁺ T cells undergo both an expansion and contraction phase in response to antigen, and the initial precursor frequency has a major influence on the magnitude and duration of these two phases (55–57). When given in high doses, CD8⁺ T cells undergo proportionately smaller proliferative responses but show increased expression of CD127, CD62L, IL-2, and tumor necrosis factor relative to physiologic T-cell populations (57). Based on these studies, one could speculate that the high dose of OT-I cells used in the present study may result in a phenotype that is functionally distinct from that of endogenous T cells yet favorable for tumor eradication.

Fifth, tumors in this model showed low baseline numbers of FoxP3⁺ regulatory T cells, which might otherwise oppose the CD8⁺ T-cell response. A significant proportion of human ovarian cancers also have sparse regulatory T-cell infiltrates (16, 18). Intriguingly, after adoptive transfer, the density of FoxP3⁺ cells increased in proportion to the CD3⁺ and granzyme B⁺ infiltrate and hence was highest under conditions associated with tumor regression. Likewise, recurrent tumors were generally devoid of FoxP3⁺ cells but positive for CD3⁺ and granzyme B⁺ cells. Thus, although FoxP3⁺ cells could potentially dampen the OT-I response, their presence showed no obvious association with either nonregression or recurrence. Instead, FoxP3⁺ cells seemed to be a relatively constant factor within the overall T-cell infiltrate. We are currently investigating other mechanisms that may have limited the OT-I response in cases where immune rejection did not occur.

Finally, it seems the environment provided by ascites might be especially conducive to CD8⁺ T-cell expansion and function. This phenomenon is not unique to OT-I cells, as activated P14 T cells also preferentially accumulate in ascites.³

Given that OT-I cell proliferation required CD25 and CD122 expression, we considered the possibility that ascites may contain IL-2 or IL-15 (2). However, in unpublished experiments, ascites from ID8-G7 tumors (harvested either before or after adoptive transfer of OT-I cells) failed to support the proliferation or viability of the T-cell line CTLL-2, which is highly responsive to IL-2 and IL-15 (data not shown). Thus, the CD25/CD122 dependence of the OT-I proliferative response may reflect autocrine IL-2 signaling, implying that the ovarian tumor environment resembles other nonlymphoid tissues and tumors in this regard (28, 31, 32). Paracrine IL-2 production by host CD4⁺ T cells could also potentially contribute. In addition, ascites may contain other immunostimulatory factors, such as IL-6 and IFN-γ (58–60), which collaborate with IL-2 to induce the extraordinary T-cell proliferation we observed. Lastly, by providing an additional fluid compartment in the host, ascites might relieve physical constraints on T-cell expansion.

The success of the OT-I response was underscored by the finding that most recurrent tumors represented antigen loss variants, implying all antigen-positive tumor cells had been eliminated. This is a common outcome of adoptive immunotherapy and highlights the need to target additional antigens. Adoptive immunotherapy has been shown to induce antigen spreading in other models (61, 62), and we are currently investigating whether this process can be promoted by the provision of OT-II T-cell help. Thus, our model provides a unique system to define the essential variables for the successful immunotherapy of ovarian cancer.

We thank Dr. Paul Terranova for providing the ID8 cell line, Tom Malek for providing CD122^−/−Tg⁺ mice, and Julie Stewart for technical assistance.