The molecular signals that allow primed CD8 T cells to persist and be effective are particularly important during cancer growth. With response to tumor-expressed Ag following adoptive T cell transfer, we show that CD8 effector cells deficient in OX40, a TNFR family member, could not mediate short-term tumor suppression. OX40 was required at two critical stages. The first was during CD8 priming in vitro, in which APC-transmitted OX40 signals endowed the ability to survive when adoptively transferred in vivo before tumor Ag encounter. The second was during the in vivo recall response of primed CD8 T cells, the stage in which OX40 contributed to the further survival and accumulation of T cells at the tumor site. The lack of OX40 costimulation was associated with reduced levels of Bcl-xL, and retroviral expression of Bcl-xL in tumor-reactive CD8 T cells conferred greatly enhanced tumor protection following adoptive transfer. These data demonstrate that OX40 and Bcl-xL can control survival of primed CD8 T cells and provide new insights into both regulation of CD8 immunity and control of tumors.
Costimulatory molecules expressed on T cells and APC are thought to impact many aspects of a T cell response, including the phases of clonal expansion and functional differentiation, as well as clonal contraction and memory reactivity. These cells are also very promising targets for modifying disease processes provided their functional activities are well characterized. Although the importance of a number of costimulatory interactions for CD4 T cell responses has been well documented (1), the rules defining the costimulatory requirements for CD8 T cells are not as clear. Based on studies of CD8 responses with acute viral or parasite infection models, the need for costimulatory signals from either Ig or TNFR family members varies from total dependence to complete independence (2, 3, 4, 5, 6, 7). In regards to CD8 memory generation after acute viral infection, members of the TNFR family, but not CD28, can be important in determining the size of the memory pool and thus impact the magnitude of some secondary responses (5, 8).
CD8 T cells are also critical in the battle against tumor cells. Antitumor responses appear to be regulated more by the availability of costimulation than antiviral responses. There are a number of studies showing that forcible expression of costimulatory molecules such as B7 in tumor cells, or injection of agonist Abs such as anti-4-1BB, can enhance CTL generation and tumor reactivity (1). Although these types of reports have shown that costimulation can dramatically affect CD8 responses, in most cases reports have not demonstrated a direct requirement for signaling to the CD8 cell, and many studies might have indirectly amplified CTL responses through actions on CD4 or NK cells. Thus, there is still a profound lack of knowledge of which costimulatory molecules might be directly relevant to CD8 cells, how and at what stage they function, and the types of CD8 response that are dependent on costimulation.
OX40 (CD134), an inducible member of the TNFR superfamily, has been described as a principal costimulatory receptor for CD4 T cells. Without OX40 signals, CD4 T cells are impaired in clonal expanding and a low number of memory T cells develop in vivo (9). We previously demonstrated that at least part of the action of OX40 on CD4 cells could be explained through sustaining the expression of antiapoptotic members of the Bcl-2 family that suppress overt cell death (10). Although OX40 has been visualized on some CD8 T cells, there is little data on whether it plays a central role in many types of CD8 response. In fact, initial reports using OX40-deficient animals showed that this molecule was dispensable in the generation of CTL to several viruses (6, 11). A number of studies have since demonstrated that agonist Abs to OX40, or OX40 ligand (OX40L)3 transfected tumor cells, can result in tumor suppression, but such effects were dependent on the presence of CD4 T cells, which puts into question whether the action was directed to the CD8 cell itself (12, 13, 14, 15, 16, 17). Although a few studies have shown that agonist Abs to OX40 are able to enhance CD8 T cell response without CD4 T cell help (18, 19), the question still remains whether OX40 signals influence CD8 T cells directly. More recently, we provided the first demonstration that OX40 could participate in regulating CD8 T cell priming directly, showing that OX40-deficient CD8 T cells were impaired in expanding and accumulating in a primary response to Ag in adjuvant (20).
In the present study, through adoptive T cell transfer, we addressed whether OX40-OX40L interactions critically regulate tumor-specific CD8 T cells and whether OX40 signals are crucial to the CD8 T cell that is thought to orchestrate killing at the site of tumor growth. Using in vitro generated effector or primed tumor-specific CD8 T cells from wild-type and OX40−/− mice, we show that OX40 is indispensable at two stages of the CD8 response. The first is during CD8 priming in vitro, at which OX40 endows the ability to survive following adoptive transfer in vivo. The second is during the recall response when primed CD8 T cells encounter tumor-expressed Ag, at which OX40 contributes to the further persistence and accumulation of T cells at the tumor site. Moreover, the action of OX40 can be replaced through forced expression of the antiapoptotic molecule, Bcl-xL, in CD8 cells, resulting in substantially improved therapeutic efficacy of adoptive transfer.
Materials and Methods
Mice and cell lines
OT-I TCR transgenic mice bred on the C57BL/6 background (21) were used as a source of Vα2Vβ5 CD8 T cells reactive to SIINFEKL peptide derived from OVA (OVA257–264). OX40−/− mice were bred with OT-I mice to generate OT-I/OX40−/− mice (20). C57BL/6 and B6.PLThy1.1 mice were purchased from The Jackson Laboratory. All types of mice were housed under specific pathogen-free conditions. The OVA-expressing EG.7 thymoma cell line was obtained from American Type Culture Collection and cultured in complete RPMI 1640 (10% FCS, 0.1% penicillin/streptomycin, 0.1% HEPES, and 0.05% 2-ME) containing 400 μg/ml G418 (Invitrogen Life Technologies) for 7–10 days before injection into mice.
Vα2Vβ5 CD8 T cells from OT-I or OT-I/OX40−/− mice were purified with CD8 T cell isolation kits (Miltenyi Biotec) following the manufacturer’s instructions. The purity was >90% with >95% of resulting cells expressing the Vα2Vβ5 transgene. Purified CD8 T cells (0.5 × 106) and irradiated spleen cells (2 × 106) from C57BL/6 were cultured in IMDM (10% FCS, 0.1% penicillin/streptomycin, 0.1% HEPES, and 0.05% 2-ME) with 0.5 μM SIINFEKL peptides in 24-well plates. In some experiments, purified CD8 T cells (3 × 106) were stimulated with plate-coated anti-CD3 (4 μg/ml) and soluble anti-CD28 (3 μg/ml). Cultured CD8 T cells were either used as CD8 effector cells at day 4 or rested in IL-15 (10 ng/ml; PeproTech) containing IMDM at day 3 for 2–3 days and then used as primed CD8 T cells.
Adoptive transfer of transgenic CD8 cells and tumor challenge
Activated CD8 T cells (1–5 × 106) from OT-I or OT-I/OX40−/− were washed twice in PBS, and injected i.v. into 7- to 10-wk-old unirradiated female B6.PLThy1.1 congenic or C57BL/6 mice. Injected cells were allowed to rest in recipient mice for 1 day, and then the recipients were challenged i.p. with E.G7 cells (4 × 106) in 0.5 ml of PBS. In some experiments, activated CD8 T cells were labeled with CFSE (5 μM; Molecular Probes) following the manufacturer’s instructions before injection. All experimental groups consisted of more than three recipient mice.
Flow cytometry analysis
Recipient mice were sacrificed at indicated times, and the draining lymph node (LN; inguinal, mesenteric, and paraaortic) and spleen were harvested, homogenized, and treated with RBC lysing buffer (Sigma-Aldrich). The peritoneal cavity was flushed with 10 ml of ice-cold PBS twice to extract tumor cells and recruited lymphocytes. The total number of cells obtained from each site and the number of tumor cells from the peritoneal cavity were counted by trypan blue exclusion. An aliquot of cells were stained with CyChrome-conjugated anti-CD8, FITC-conjugated anti-Vβ5, PE-conjugated Vα2, and allophycocyanin-conjugated Thy1.2 (BD Pharmingen) for 20 min on ice. Immunostained cells were analyzed on a FACSCalibur flow cytometer (BD Biosciences) with FlowJo software (Tree Star). The numbers of OVA257–264-specific T cells were calculated by multiplying total cell numbers by percentages of Vα2/Vβ5/CD8 cells or Thy1.2/CD8 cells in the draining LN, spleen, or peritoneal cavity lavage.
Activation status-associated phenotypes of OT-I cells after in vitro stimulation or in vivo tumor challenge were characterized by immunostaining with specific Abs: CyChrome-conjugated anti-CD8, FITC-conjugated anti-CD44, and PE-conjugated anti-CD25, or PE-conjugated anti-CD62L, or biotin-conjugated anti-OX40 followed by PE-conjugated streptavidin. Early apoptotic cells were detected with annexin V (BD Pharmingen) following the manufacturer’s instructions. All Abs were purchased from BD Pharmingen.
For intracellular cytokine staining, 2 × 106 cells were cultured with 0.5 μM SIINFEKL peptide in 24-well plates in the presence of GolgiPlug (BD Pharmingen) for 5 h. Cells were then stained with CyChrome-conjugated anti-CD8, FITC-conjugated anti-CD44, and allophycocyanin-conjugated anti-Thy1.2 as described, followed by intracellular staining using PE-conjugated anti-IL-2 or anti-IFN-γ.
Murine stem cell virus vectors containing GFP alone or GFP and Bcl-xL were used for transfection as previously described (10). Supernatants of transfected Phoenix E cells were collected after 48 h, and used to transduce T cells, which had been stimulated either with APC and peptides or anti-CD3 and anti-CD28 for 2 days. For transduction, supernatants of cultured T cells in 24-well plates were replaced with 1 ml/well viral supernatants containing 5 μg/ml polybrene (Sigma-Aldrich), and cells were spun for 2 h at 32°C. T cells were then incubated at 32°C for 8 h before continuing cultures at 37°C. GFP+ cells were sorted by flow cytometry at day 5 or 6.
In vitro recall responses and cytokine production
Draining LN and spleen cells (5 × 105) were stimulated with varying concentrations of SIINFEKL peptide in 96-well flat-bottom plates. Proliferation was measured in triplicate after 72 h by the incorporation of [3H]thymidine (1 μCi/well; ICN Biochemicals) during the last 12–16 h of culture. IL-2 and IFN-γ levels in culture supernatants were measured by ELISA at 36–40 h with Abs to IL-2 (JES6-1A12 and biotin-JES6-5H4) and IFN-γ (R46A2 and biotin-XMG1.2; all from BD Pharmingen).
Numeric data were compared by unpaired t test, p values were two-tailed, and significance was set at 5%. Mouse survival was compared by log-rank test (life table). All statistics were calculated using Prism4.
OX40-deficient CD8 effector cells fail to suppress tumor growth following adoptive transfer
We first analyzed the inhibitory effects of OX40-deficient CD8 effector cells on tumor growth after adoptive transfer. The model involved stimulating wild-type and OX40−/− OT-I TCR transgenic CD8 T cells in vitro with OVA257–264 peptide and APCs for 4 days, at which time equal numbers of activated CD8 cells were injected i.v. into naive mice. This was followed 1 day later with an i.p. challenge of OVA-expressing E.G7 thymoma cells.
At day 10 and 15, when low numbers (1 × 106) of effector cells were transferred, tumor growth in mice receiving OX40−/− CD8 T cells was similar to growth in control mice that did not receive T cells, and tumor growth was ∼2- to 3-fold more than that in mice receiving wild-type CD8 T cells (day 10, p = 0.0009; day 15, p = 0.006) (Fig. 1,A). When higher numbers of effector CD8 cells (5 × 106) were used for adoptive transfer, mice receiving wild-type T cells had dramatically lower tumor growth at day 15 compared with mice receiving OX40−/− CD8 T cells (p = 0.004), although growth at day 10 was suppressed substantially by the OX40-deficient cells (Fig. 1,B). Given the stringent conditions we used in this tumor model in which high doses (4 × 106) of tumor cells were injected and relatively low numbers (3 × 106) of primed CD8 T cells were transferred, wild-type T cells did not significantly alter overall mouse survival. Nevertheless, wild-type T cells markedly delayed mortality between day 35 and day 50. In contrast, OX40-deficient T cells failed to have any effect on survival over time (Fig. 1,C). In vitro recall responses showed strikingly weaker OVA peptide-specific proliferative responses from mice receiving OX40−/− T cells (Fig. 1,D). Consistent with this, lower levels of IL-2 (Fig. 1,E) and IFN-γ (Fig. 1 F) production were detected.
OX40-deficient CD8 effector cells exhibit a survival defect initiated during in vitro priming
The difference in functionality could have reflected reduced expansion and/or survival of OX40−/− CD8 T cells after adoptive transfer. To initially investigate this possibility, the response of OX40-deficient CD8 T cells was analyzed in vitro over a more extended time. OX40−/− CD8 T cells expanded dramatically after 2–3 days, resembling the response of wild-type CD8 T cells (data not shown), but started to show a defective phenotype from day 4 onward, which was the time we chose to perform the adoptive transfer. OX40−/− CD8 T cells expressed comparable levels of Bcl-xL (and Bcl-2, data not shown) to wild-type cells at day 3, but considerably fewer OX40−/− cells maintained high levels of these proteins at days 4 and 5 (Bcl-2 data not shown) (Bcl-xL; Fig. 2,A). To investigate whether this affected the survival of transferred CD8 effectors, we quantified the number in recipients that were or were not challenged with tumor. Without tumor injection, there were ∼2-fold more wild-type T cells as early as 24 h after the transfer (equivalent to day 5 in vitro) (p = 0.007; Fig. 2,B). Similar results were seen with tumor challenge, and wild-type T cells expanded at the tumor site in the peritoneal cavity, whereas OX40−/− T cells did not accumulate there (p = 0.0003; Fig. 2,C). Correspondingly, annexin staining demonstrated a dramatic increase in apoptosis of adoptively transferred OX40−/− CD8 T cells (Fig. 2 D).
These results showed that OX40 was required for the survival of activated CD8 T cells when used during adoptive transfer, and at least part of the defect was directly related to an inability to receive appropriate signals during the in vitro priming period.
OX40 regulates the ability of primed CD8 T cells to suppress tumor growth after Ag re-encounter in vivo
We then asked whether OX40 was also required for the persistence and functional effectiveness of already primed CD8 T cells when tumor-expressed Ag was re-encountered in vivo. To address this question, wild-type and OX40-deficient OT-I CD8 T cells were stimulated with anti-CD3 and anti-CD28 for 3 days and rested in IL-15-containing medium for 3 days before the adoptive transfer. The rationale in this model was that by not stimulating with Ag plus 5 APCs and supplementing cultures with cytokines, we bypassed the requirement for OX40-OX40L interactions during effector generation. Wild-type CD8 T cells expressed OX40, CD25, and high levels of CD44, whereas these cells down-regulated CD62L 2–3 days after in vitro stimulation (Fig. 3,A). However, further culture of these cells in the presence of IL-15 resulted in a decrease in cell size and phenotypic changes associated with transition from effector to memory state (Fig. 3 A).
Upon transfer into mice, CFSE analysis showed that the majority of primed CD8 cells either did not divide or only divided one time over the next 4–7 days without tumor challenge (day 7 data not shown) (Fig. 3,B, top). OX40-deficient CD8 T cells behaved similarly to wild-type T cells with equivalent numbers found after 4 or more days (day 7 data not shown) (Fig. 3,B, bottom), and there was no difference between wild-type and OX40-deficient CD8 T cells in retaining the ability to produce cytokines (Fig. 3, C and D).
To determine whether OX40 was then critical during the recall CD8 T cell response when tumor-expressed Ag was encountered in vivo, the effectiveness of these primed CD8 cells was compared following tumor injection. Although OX40−/− CD8 T cells were able to suppress initial tumor growth nearly as efficiently as wild-type cells over 10 days (Fig. 4,A), they failed to provide continued protection through to day 20 (p = 0.0004; Fig. 4,B) and did not alter the kinetics of mortality (Fig. 4,C). Correlating with the idea that this was due to an inefficient recall response of the primed OX40-deficient CD8 T cells, recovered splenocytes and draining LN cells (data not shown) displayed a weak proliferative response to in vitro stimulation with OVA (Fig. 4,D) and a profound defect in production of IL-2 (Fig. 4,E) and IFN-γ (Fig. 4 F).
OX40 regulates the survival of primed CD8 T cells in recall responses to tumor-expressed Ag in vivo
To determine whether primed OX40−/− CD8 T cells were impaired in their ability to divide when encountering recall tumor-expressed Ag, CFSE-labeled cells were tracked. Wild-type CD8 T cells divided up to 8 times at the site of tumor growth in the peritoneal cavity by day 3, and divided 2–3 times in draining LN (Fig. 5,A) and spleen (data not shown). Significantly, no difference was observed with OX40-deficient cells. However, by day 6, the OX40−/− CD8 T cells that were detectable had undergone over 10 divisions in the peritoneal cavity (Fig. 5,B), but far fewer were present compared with wild-type T cells (p = 0.01; Fig. 5, B and C). This dramatic decrease in CD8 T cell recovery was clearly seen in LN and spleen, as well as the peritoneal cavity at day 10 (LN, p < 0.0001; spleen, p = 0.0003; peritoneal, p = 0.001) (Fig. 5,D). As some division of T cells was visualized in the secondary lymphoid organs, this suggested that tumor Ag was presented to the T cells outside, as well as inside, the tumor site, and that OX40 was required for allowing the survival of these Ag-activated CD8 T cells. This was substantiated by finding increased numbers of apoptotic OX40−/− CD8 T cells at day 5 (Fig. 5 E).
Normal responsiveness but decreased survival of primed CD8 T cells in the absence of OX40
In vitro assays demonstrated weak proliferative and cytokine responses (Fig. 4, C–E) from mice receiving primed OX40−/− CD8 T cells. To determine whether a lack of OX40 during the secondary in vivo response resulted in a defect in functional differentiation as well as survival, we assessed cytokine production by intracellular staining. At 3 days after tumor injection, OX40−/− CD8 T cells retained the ability to make IL-2 (Fig. 6,A) and IFN-γ (Fig. 6,B) and the numbers of IL-2- or IFN-γ-positive cells in the draining LN and peritoneal cavity were comparable with wild-type populations (Fig. 6,C). However, by day 10, far fewer IL-2-positive (Fig. 6,D) and IFN-γ-positive (Fig. 6,E) CD8 T cells were detected in mice receiving OX40−/− cells (IFN-γ, p < 0.0001; IL-2, p = 0.0002) (Fig. 6 F), although those that did produce cytokines appeared to display similar levels to wild-type T cells. Collectively, these data demonstrated that primed CD8 T cells could not survive and persist after re-encountering tumor Ag in vivo, and therefore failed to provide sustained protection.
Expression of Bcl-xL restores the ability of OX40-deficient CD8 T cells to suppress tumor growth during adoptive immunotherapy
The finding that a lack of OX40 signaling correlated with reduced expression of Bcl-2 antiapoptotic molecules, including Bcl-xL, suggested that maintenance of these proteins over time might be key to the effectiveness of the transferred T cells. We therefore asked whether forced expression of Bcl-xL in OX40−/− CD8 T cells could rescue the phenotype and restore their ability to inhibit tumor growth. OX40−/− CD8 T cells were stimulated in vitro with either anti-CD3 and anti-CD28 or peptide plus APC and retrovirally transduced with a GFP-expressing vector encoding Bcl-xL, similar to previous experiments with CD4 cells (10). GFP+ cells were sorted and equal numbers (2.5 × 106) adoptively transferred into naive mice followed by challenge with tumor. Six days later, draining LN (p < 0.03; Fig. 7,A), spleen (p = 0.0003; Fig. 7,B), and peritoneal lavage (p < 0.008; Fig. 7,C) from mice receiving Bcl-xL-transduced OX40−/− CD8 T cells contained ∼4- to 7-fold more GFP+ cells than mice receiving control vector-transduced cells. Although the number of GFP+ cells in mice receiving Bcl-xL-transduced T cells markedly decreased at days 10 and 20 (data not shown) possibly due to tumor suppression, the number was still significantly higher than in mice receiving vector-transduced cells. Importantly, adoptive transfer of Bcl-xL-transduced OX40−/− CD8 T cells effectively suppressed tumor growth, with far less tumor burden in the peritoneal cavity at day 20 (p = 0.009; Fig. 7,D). More significantly, Bcl-xL-transduced OX40−/− CD8 T cells reduced mouse mortality between day 35 and day 60 (p < 0.006; Fig. 7 E), although most of the mice were not tumor free and eventually died.
Bcl-xL augments antitumor activity following adoptive CD8 T cell immunotherapy
Although we showed that OX40 was essential for control of short-term tumor growth over 20 days (Figs. 1 and 4), the adoptive transfer of wild-type CD8 T cells that still received OX40 signals only delayed the kinetics of growth. As such, mouse survival was prolonged by a matter of 10–20 days, but ultimately the majority of mice succumbed to the tumor (Fig. 4). This correlated with the fact that the number of CD8 T cells at the tumor site and in secondary lymphoid tissues constantly diminished over time, even when they received OX40 signals in their first week of response in vivo. This result suggested that survival signals received by the primed CD8 cells in response to tumor Ag were not sufficient or optimal.
We then asked whether forced expression of Bcl-xL in primed wild-type CD8 T cells could sustain their antitumor activity. Wild-type CD8 T cells were stimulated in vitro with anti-CD3 and anti-CD28 and retrovirally transduced with Bcl-xL. Mice receiving Bcl-xL-transduced cells contained distinctly higher numbers of GFP+ T cells in the draining LN, spleen, and peritoneal cavity at day 6 (data not shown), and strikingly lower numbers, or in many mice undetectable numbers, of tumor cells were found in the peritoneal cavity at day 20 (p < 0.04; Fig. 8,A). More importantly, the majority of mice receiving Bcl-xL-transduced CD8 T cells survived up to 90 days after the first tumor challenge (p < 0.02; Fig. 8,B), and remained tumor free even after a second tumor challenge (Fig. 8 B). In summary, the data show that both OX40 and Bcl-xL can critically contribute to, and augment, the persistence of CD8 T cells following encounter with tumor Ag in vivo.
We show in this study that OX40 signals maintain the expression of antiapoptotic molecules in CD8 T cells, dictating the duration, and thus the effectiveness, of short-term activity in a tumor setting. Moreover, we demonstrate that CD8 T cells modified with the antiapoptotic gene Bcl-xL, a target of OX40, accumulate in high numbers following in vivo expansion to tumor-expressed Ag, and CD8 T cells persist for an extended period of time at both the tumor site and secondary lymphoid organs. This finding consequently led to augmented long-term tumor protection.
Interactions between costimulatory receptors and ligands occur only transiently, yet the signals transmitted can influence the fate of a T cell during, as well as after, receptor/ligand engagement (9, 10, 20). A goal in cancer research is to generate and sustain tumor reactive CD8 T cells in tumor-bearing subjects. Cancer vaccines with synthetic peptides, peptide-pulsed dendritic cells, and recombinant viruses have achieved only limited clinical effects, probably due to the failure in generating sufficient numbers of high avidity tumor reactive CD8 T cells (22). Recently adoptive T cell transfer has emerged as a viable treatment for cancer patients and has been shown to mediate a degree of successful tumor regression in both mice and humans (22, 23, 24, 25, 26, 27). The results from human clinical trials demonstrate a significant correlation between tumor regression and the persistence of adoptively transferred T cells, implying that a major obstacle to success is the failure to sustain survival of transferred cells (28). We then first asked whether OX40-deficient CD8 effector cells generated in vitro could provide antitumor activity. Although CD8 T cells initially proliferated in vitro in the absence of OX40, they failed to maintain Bcl-xL and Bcl-2 expression shortly after transitioning into effector cells. This resulted in a survival defect reflected in CD8 T cell impaired persistence in vivo 24 h or more after the adoptive transfer, regardless of tumor challenge. These data therefore indicate that OX40 signals, provided during priming in vitro, regulate the survival capacity of CD8 T cells in vivo. They dramatically illustrate that although it can appear that viable effector T cells are being introduced into a tumor-bearing host, if these cells do not receive appropriate costimulatory signals, they will not survive even over a short period of time, regardless of whether tumor Ag is encountered.
As relevant to CD8 immunity and control of tumor growth is what happens to already primed CD8 T cells when they re-encounter Ag in recall responses. More specifically, are costimulatory signals essential for the continued survival and functionality of tumor-reactive T cells over an extended period of tumor growth? CD8 memory T cells specific to tumor Ags can successfully develop in vivo (24) after the initial eradication of tumor cells, and they can provide enhanced protection in response to tumor rechallenge in some murine models. However, in most situations in which tumor cells are not completely eliminated and regrow, tumors are not efficiently suppressed by CD8 memory T cells. To address the requirement for costimulation and for provision of further survival signals, we established an in vitro culture system to activate then rest CD8 T cells, generating cells that phenotypically resemble central memory cells (29). Adoptively transferred wild-type and OX40−/− CD8 primed “memory-like” cells were maintained at equivalent levels in the absence of tumor challenge, allowing conclusions to be gained regarding the requirements for CD8 persistence after tumor Ag encounter. Quite conclusively, we found that CD8 T cells that could not receive OX40 signals exhibited an impaired ability to survive after responding to tumor-expressed Ag, with greatly reduced numbers detected at the tumor site and in secondary lymphoid organs.
Our data show that inadequate costimulation to an already primed and fully functional CD8 cell is a limitation for the persistence of these cells during the time at which tumor Ag is presented. This finding then translates into being vital to future tumor control and regression. Interestingly, although we demonstrated a major role for OX40 in allowing CD8 T cells to survive over a week or more following in vivo tumor Ag encounter, even wild-type T cells that initially received survival signals were still only partially effective in controlling the tumor, and simply delayed its outgrowth. Suboptimal costimulation, or a lack of sustained costimulation over time, might account for the incomplete tumor protection from wild-type CD8 T cells. We demonstrated that reconstituting Bcl-xL expression in OX40−/− CD8 T cells rescued their survival defect, and restored their ability to suppress tumor growth over time. More significantly, sustained expression of Bcl-xL in wild-type CD8 T cells conferred a dramatically enhanced capacity to suppress tumor growth, and converted a moderate effectiveness into complete and long-lasting protection in a majority of recipients. This might then illustrate a general principle that sustaining the balance of antiapoptotic vs proapoptotic members of the Bcl-2 family in favor of the antiapoptotic molecules can fundamentally alter the effectiveness of T cells over time. It is noted that Bcl-xL-transduced CD8 T cells did not provide absolute tumor free protection, suggesting that this antiapoptotic molecule is not the only molecule that is relevant, and in fact, Bcl-2 expression in OX40−/− CD8 T cells was also not sustained. In line with this possibility, forced expression of Bcl-2 was recently demonstrated to have a similar effect in another tumor model (30) similar to the action of Bcl-xL we show in this study. More recently, mice with a conditional Bcl-xL deletion were shown to generate normal CD8 T cell responses to Listeria monocytogenes, suggesting that there might be a Bcl-xL-independent mechanism for T cell survival (31).
These data highlight the need for survival or antiapoptotic signals, at least at the time of initial tumor Ag encounter, to allow primed CD8 T cells to effectively combat tumor growth. We feel, however, that it is unlikely that the long-term antitumor effect of retroviral expression of Bcl-xL was due to the continued persistence of those T cells over months rather than weeks. We were unable to detect transferred transgenic T cells (based on GFP or Thy1.2 expression) 30 days after initial tumor injection. It is possible a low level still persisted, but we favor the notion that they may have eventually died. This result would correlate with previous data on CD8 cells obtained from Bcl-xL transgenic mice in which Bcl-xL antagonized apoptosis of effector cells, but only delayed the contraction phase of an Ag-induced CD8 T cell response rather than preventing it (32). Recent results have suggested that recruitment of host cells is essential to the therapeutic action of adoptively transferred CD8 cells, rather than these cells directly targeting the tumor (27, 33). If so, long-term protection will then be mediated by host cells and will not require adoptive CD8 cells. Together with our data, this would then imply that there is a critical time period over which transferred primed CD8 cells need to survive to recruit this host response.
In summary, we demonstrate that OX40-delivered survival signals can contribute to priming of anti-tumor CD8 effector T cells and to the recall CD8 T cell response to tumor-derived Ag. Genetic modification of CD8 T cells in vitro with the antiapoptotic molecule Bcl-xL, a target of OX40, can compensate for the defective short and longer term survival signals, increase the persistence of tumor-reactive CD8 T cells in vivo, and engender enhanced antitumor activity.
This is manuscript no. 694 from the La Jolla Institute for Allergy and Immunology, San Diego, CA.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was funded by an American Cancer Society grant and by National Institutes of Health Grant CA91837 (to M.C.).
Abbreviations used in this paper: OX40L, OX40 ligand; LN, lymph node.