Abstract
CD40, 4-1BB, and OX40 are costimulatory molecules belonging to the TNF/nerve growth factor superfamily of receptors. We examined whether simultaneous costimulation affected the responses of T cells using several different in vivo tracking models in mice. We show that enforced dual costimulation through 4-1BB and OX40, but not through CD40, induced profound specific CD8 T cell clonal expansion. In contrast, the response of specific CD4 T cells to dual costimulation was additive rather than synergistic. The synergistic response of the specific CD8 T cells persevered for several weeks, and the expanded effector cells resided throughout lymphoid and nonlymphoid tissue. Dual costimulation through 4-1BB and OX40 did not increase BrdU incorporation nor an increase in the number of rounds of T cell division in comparison to single costimulators, but rather enhanced accumulation in a cell-intrinsic manner. Mechanistically speaking, we show that CD8 T cell clonal expansion and effector function did not require T help, but accumulation in (non)lymphoid tissue was predominantly CD4 T cell dependent. To determine whether this approach would be useful in a physiological setting, we demonstrated that dual costimulation mediated rejection of an established murine sarcoma. Importantly, effector function directed toward established tumors was CD8 T cell dependent while being entirely CD4 T cell independent, and the timing of enforced dual costimulation was exquisitely regulated. Collectively, these data suggest that simultaneous dual costimulation through 4-1BB and OX40 induces a massive burst of CD8 T cell effector function sufficient to therapeutically treat established tumors even under immunocompromising conditions.
Productive T cell activation is mediated by at least two different signals (1). The first is based on the specificity of the TCR for MHC/peptide complexes on APCs, and the second is stimulation of a costimulatory molecule. Although activation of both signals do not guarantee productive immunity (2), the absence of either signal leads to unproductive immune responses or sometimes even tolerance induction (3). Nevertheless, signal 2 provides a practical and therapeutic means to manipulate lymphocyte responses in vivo. This is based on the ability of antagonistic vs agonistic reagents that either block or enforce T cell costimulation (4).
The TNF superfamily of receptors contains a number of costimulatory molecules that are expressed pervasively throughout the immune system (5). Some of these molecules typically enhance activation resulting in protective immune responses (6, 7). For example, OX40 (CD134) has been shown to potently costimulate T cell effector function as measured by proliferation and cytokine production (8, 9, 10), while being capable of delivering a bidirectional signal into APCs via OX40 ligand (11). Moreover, OX40 was shown to reverse unresponsiveness in anergized T cells (12). Another example is CD40 stimulation, which directly induces APC activation, leading to numerous effects including enhanced proliferation and cytokine production (13, 14), as well as class switching of Abs (15, 16, 17). Recently, CD40 expression was shown to occur on T cells, which when ligated led to enhanced immunity (18), although this does not occur in all systems (19).
4-1BB (CD137) is another TNF family member with wide-spanning surface expression patterns and is also capable of costimulating T cells and APCs (20). A unique feature of 4-1BB, compared with OX40 or CD40, is its penchant for enhancing CD8 T cell responses over other cell types (21, 22). This has been particularly useful in controlling viral infections as well as enhancing antitumor immunity (23, 24, 25). Therefore, we investigated whether simultaneous costimulation through 4-1BB and either OX40 or CD40 would result in even better immune responses compared with activating one costimulator alone. Nevertheless, this was unlikely, because simultaneous stimulation of OX40 and CD40 does not significantly enhance T cell responses in vivo (26), suggesting that signaling through these TNF family members is at least partially redundant.
However, our results show that simultaneous dual costimulation through 4-1BB and OX40, but not with CD40, synergistically induces specific CD8 T cell clonal expansion in several in vivo models. The expanding T cells produced large amounts of effector cytokines like IFN-γ and TNF when primed with dual costimulation, and effector function did not require T help. These functional outcomes were shown to be a result of enhanced effector T cell accumulation as opposed to increased entry into the cell cycle or increased rounds of cell division. Ultimately, we tested whether this approach could be used therapeutically to eliminate a murine sarcoma. Our data show that established tumors were eliminated by dual costimulation. Furthermore, the mechanism is CD8 T cell dependent and functions in the absence of T help. Therefore, enforced T cell dual costimulation by 4-1BB and OX40 may be an efficacious approach in fighting human cancer even under immunocompromising conditions.
Materials and Methods
Mice, reagents, and injection schedule
B10.A and C57BL/6 mice were purchased from the National Cancer Institute (Frederick, MD) and The Jackson Laboratory (Bar Harbor, ME). BALB/c mice were purchased from Charles River (Frederick, MD). The OT-I transgenic RAG2-deficient (RAG−/−) mice were kindly provided by Dr. L. Lefrançois (University of Connecticut Health Center). All of the animals were maintained in accordance with federal guidelines at the University of Connecticut Health Center.
The staphylococcal enterotoxin A (SEA)321) is specific for murine 4-1BB. Anti-OX40 is an agonist rat mAb specific for murine OX40 (OX86 hybridoma; Ref. 27); and anti-CD40 (FGK45.5) is a rat mAb specific for murine CD40 which was a kind gift from T. Rolink (University of Basel, Basel, Switzerland) (28).
In all experiments, Ag, peptide, or superantigen (SAg) injection, as well as mAbs, were given by the i.p. route, and adoptive transfer of cells was by the i.v. route. To study endogenous T cell responses, 0.3 μg of SEA or 1 mg of OVA was injected into B10.A or C57BL/6 mice, respectively. For in vivo stimulation by enforced costimulation, mice were given anti-4-1BB, anti-OX40, and anti-CD40 alone or in various combinations at the same time as signal 1. For different experiments, the Ab dose changed due to batch-to-batch variation of each Ab preparation. This is based on careful titration experiments examining the in vivo effects of these mAbs on T cells. For the adoptive transfer model, OT-I RAG−/− splenocytes were transferred into C57BL/6 mice, and the day after, either 100 μg of SIINFEKL peptide or OVA diluted into balanced salt solution (BSS) was injected. The number of cells transferred is given in the figure, and on occasion, we found that establishment of OT-I cells was impaired, which was probably due to low transfer number or a poor take by the host.
Cell processing, staining, and flow cytometry
Spleen, peripheral lymph nodes (PLN; inguinal, axillary, brachial), and mesenteric lymph nodes (MLN) were crushed through nylon mesh cell strainers (Falcon; BD Biosciences, San Diego, CA), and then RBC in spleen were lysed with ammonium chloride. After several washes, the cells were counted using a Z1 particle counter (Beckman Coulter, Miami, FL). For nonlymphoid tissue, we followed published procedures (29
For flow cytometry, cells were stained with primary Abs in the presence of a blocking solution containing 5% normal mouse serum (Sigma-Aldrich), 10 μg/ml human γ-globulin (Sigma-Aldrich), and 0.1% sodium azide in culture supernatant from the 2.4.G.2 hybridoma (anti-FcR; Ref. 30) for 30 min on ice and then washed in wash buffer (3% FBS and 0.1% sodium azide in BSS). For intracellular cytokine staining, 1 × 106
For intracellular propidium iodide staining, the cells were stained with anti-CD45.1-FITC (BD Pharmingen) on ice for 30 min. After several washes, the cells were fixed, and the pelleted cells were resuspended in a solution containing 5 mM EDTA, 5 μg/ml propidium iodide, 50 μg/ml RNase A, and 0.3% saponin in PBS. All stained cells were assayed on a FACSCalibur (BD Biosciences), and the data were analyzed with CellQuest (BD Biosciences) or FlowJo (Tree Star, San Carlos, CA) software.
Cell purification, depletion, and culturing
To study the survival of purified OT-I T cells, 2 × 106 splenocytes from OT-I RAG−/− mice were adoptively transferred into C57BL/6 mice, and 1 day later, OVA protein was injected with either anti-4-1BB alone, anti-OX40 alone, or the combination of anti-4-1BB and anti-OX40. After 60 h, spleens were crushed through a cell strainer and run through a nylon wool column (PerkinElmer Life and Analytical Sciences, Boston, MA) for bulk removal of APC. Cells were stained with anti-CD45.1-PE (BD Pharmingen), and then labeled with anti-PE microbead (Miltenyi Biotec, Auburn, CA) according to the manufacturer’s recommendation. CD45.1+ cells were positively selected using MACS columns (Miltenyi Biotec). After washing, purity of CD45.1 OT-I was measured by flow cytometry and was typically 95–98% pure as determined by SIINFEKL tetramer staining. The OT-I T cells were resuspended in CTM (MEM containing amino acids, salts, antibiotics, and FBS), and 100,000 cells were placed into a well of a 96-well plate and cultured at 37°C.
For CD4 T cell depletion in vivo, B10.A mice were treated with 200 μg of purified anti-mouse CD4 (L3/T4) mAb (Cedarlane, Ontario, Canada) on 6 and 2 days before immunization. This technique induced profound depletion of the CD4 T cells as determined by flow cytometry.
CFSE and BrdU analysis
For in vivo CFSE profile studies (31), spleen cells from OT-I RAG−/− mice were isolated, and pelleted cells were resuspended in TC solution (MEM, amino acids, and antibiotics). The cells were labeled with 15 μM CFSE (Molecular Probes, Eugene, OR) and incubated at 37°C for 10 min. The labeling procedure was stopped with cold CTM, and the cells were transferred into recipient mice. For in vitro CFSE profile studies, OT-I cells were transferred into C57BL/6 mice and immunized, and after 60 h, spleen cells were labeled with 1.5 μM CFSE at 37°C for 10 min, and the labeling reaction was halted with cold CTM. The washed cells were resuspended in CTM, and 1 × 106 cells were placed into a well of a 96-well plate. After 20 or 45 h, the cells were stained with anti-CD45.1 PE Ab.
For BrdU staining, we followed a previous report (32). Briefly, 10 μM BrdU (Sigma-Aldrich) was added at time 0 or 30 h of culture. After 15 h of BrdU, the cells were stained with anti-CD45.1-PE on ice for 30 min, and then dehydrated and fixed in ice-cold 95% ethanol. The cells were permeabilized with 1% paraformaldehyde, 0.01% Tween 20 in PBS, and then treated with 50 Kunitz U of DNase I (Sigma-Aldrich) at RT for 10 min. After several washes, the cells were stained with anti-BrdU-FITC (BD Pharmingen) at RT for 30 min, and then analyzed by flow cytometry.
Fibrosarcoma tumor model
MethA fibrosarcoma cells were maintained by weekly i.p. passage in BALB/c mice as done previously (33). Briefly, the effect of dual costimulation on the fibrosarcoma MethA tumor was studied by using an intradermal injection of 0.5–2.0 × 106 MethA tumor cells in the right flank next to the midline of 6- to 8-wk-old female BALB/c mice. Growth was monitored twice a week using calipers to measure tumor diameter. For CD4, CD8 depletion experiments, anti-CD4, anti-CD8, or rat IgG control Ab was injected into BALB/c mice 3 days before and 6 days after injection of 1 × 106 MethA cells. Lymphocyte depletion was confirmed by flow cytometry.
Results
Effects of dual costimulation on T cell clonal expansion
SEA stimulates T cells in an oligoclonal fashion by activating TCR Vβ3-bearing CD4 or CD8 T cells in an MHC class II-restricted manner (34). SEA was previously shown to induce specific T cell clonal expansion followed by profound deletion, and sometimes anergy induction (35, 36). To test the effects of simultaneous dual costimulation, mice were injected with SEA plus control rat IgG, anti-4-1BB, -CD40, -OX40, or dual combinations of various costimulators (Fig. 1⇓). On day 10 after injection, PLN, MLN, and spleen cells were analyzed ex vivo by flow cytometry for the presence of CD8 (Fig. 1⇓a), or CD4 (b) T cells expressing TCR Vβ3. Deletion of SEA-responsive T cells was observed in all tissues when SEA was given with control IgG. Consistent with previous data, 4-1BB costimulation rescued CD8, but not CD4, T cells from deletion (22). CD40 or OX40 costimulation did not block CD8 T cell deletion and had minimal effects on CD4 T cells at this time point. CD40 and 4-1BB costimulation induced additive effects on the SEA-responsive T cells. In contrast, 4-1BB and OX40 synergistically induced CD8 Vβ3 responses in all tissues examined. This was not the case for CD4 T cells where the responses were additive at best. To stringently test the notion that 4-1BB and OX40 mAbs synergistically stimulate T cell responses, equivalent amounts of mAb were given as shown in Table I⇓. The data show that the bulk amount of mAb did not determine the level of T cell response, because 75 μg of each individual mAb was not equivalent to 25 plus 50 μg of 4-1BB and OX40 mAbs, respectively.
Dual costimulation by 4-1BB and OX40, but not CD40, synergistically stimulates CD8 T cells. B10.A mice were separated into seven groups and immunized as follows: no injection; SEA plus 275 μg of rat IgG; SEA plus 25 μg of anti-4-1BB plus 250 μg of rat IgG; SEA plus 250 μg of anti-CD40 plus 25 μg of rat IgG; SEA plus 50 μg of anti-OX40 plus 25 μg of rat IgG; SEA plus 25 μg of anti-4-1BB plus 250 μg of anti-CD40; and SEA plus 25 μg of anti-4-1BB plus 50 μg of anti-OX40. On day 10 postimmunization, PLN, MLN, and spleen were isolated, and resident CD8 and CD4 T cells were stained for specific cells. The mean percentage ± SEM of CD8 (left) or CD4 (right) T cells expressing TCR Vβ3 in each tissue was measured by flow cytometry. The filled boxes indicate the type of treatment given, and the data come from the combination of two separate experiments for a total of five mice per group.
Dual costimulation by 4-1BB and OX40 is purely synergistic for CD8 T cell stimulationa
We used the SEA model to examine a detailed time course of 4-1BB and OX40 dual costimulation in comparison with either stimulus alone. The data show that clonal expansion (days 4 and 7) and inhibition of deletion (days 12 and 14) were enhanced with dual costimulation compared with the other groups (Fig. 2⇓a). To fully evaluate this response, the number of SEA-specific CD8 T cells was determined (Table II⇓). The data on days 4–14 show large increases of specific CD8 T cells in the dual-costimulation group compared with the control groups. For example, during the effector phase on day 4, there was a 60-fold increase in the number of CD8 Vβ3 T cells in the dual-costimulation group compared with the control IgG. Even in comparison to the 4-1BB and OX40 groups alone, the dual-costimulation group induced a >4-fold increase on days 4 and 7. Collectively, these data show that there is a massive increase in proportion and number of specific CD8 T cells during the effector phase after in vivo stimulation with dual costimulation by anti-4-1BB and -OX40.
Dual costimulation synergistically stimulates CD8 T cell clonal expansion in multiple in vivo models. a, B10.A mice received SEA and 125 μg of rat IgG (□); SEA plus 25 μg of anti-4-1BB (○); SEA plus 100 μg anti-OX40 (⋄); and SEA plus anti-4-1BB and anti-OX40 (▴). Resident T cells in spleen were analyzed on days 2, 4, 7, 9, 11, and 14. The data show mean percentage ± SEM of CD8 T cells expressing Vβ3 from three mice combined from three separate experiments. b, C57BL/6 mice were immunized with OVA plus 25 μg, or 100 μg in a separate experiment, of anti-4-1BB mAb (○); OVA plus 100 μg or 50 μg of anti-OX40 mAb alone (⋄); or OVA plus both Abs (dual costimulation) (▴). Control rat IgG was injected into the two former groups to equalize the amount of Ab between the groups, and this was true for the remaining data. On days 5–9, CD8 peripheral blood T cells were gated, and the percentage of CD11ahigh cells binding SIINFEKL tetramer was determined by flow cytometry. Shown are the mean percentage ± SEM combined from both experiments for a total of eight mice per group. c, C57BL/6 mice received 50,000 (top) or 1,000 (bottom) OT-I RAG−/− splenocytes. The following day, mice were challenged with OVA plus 100 μg of anti-4-1BB, OVA plus 50 μg of anti-OX40, or OVA plus dual costimulation. On day 6, peripheral blood was analyzed for the presence of OT-I CD8 T cells by staining for the congenic marker CD45.1 expressed on CD8 CD11a T cells. The mean percentage ± SEM of CD8 CD11ahigh OT-I T cells is shown. Each group contained three mice, and comparable data were seen with transfer of 105 or even 2 × 106 OT-I cells (data not shown).
Dual costimulation massively enhances CD8 T cell expansion in vivoa
Perhaps the most stringent test of in vivo T cell immunity is stimulation of endogenous peptide-specific T cells with whole protein in the absence of any apparent adjuvant. SIINFEKL-specific CD8 T cells were stimulated in vivo on day 0 with whole OVA protein in the presence of either anti-4-1BB, -OX40, or dual costimulation. The percentage of peripheral blood CD8 CD11ahigh T cells with specificity for SIINFEKL tetramer is shown for days 5–9 (Fig. 2⇑b). Day 6 after immunization was the peak point where dual costimulation induced the largest amount of clonal expansion compared with the other treatments. OX40 costimulation was the weakest signal, and 4-1BB alone yielded a response similar to what we have seen in the past (37). Nevertheless, tracking the specific cells before immunization is extremely difficult, and thus we used a more workable experimental model as a third test for dual costimulation.
The transfer of OT-I T cells into recipient mice has proven to be a useful model to study basic CD8 T cell activation (38), antitumor immunity, and other immunological responses (39, 40, 41). We used the CD45.1 congenic marker to track RAG−/− OT-I T cells after transfer into C57BL/6 mice. Recipient mice were immunized with OVA protein with anti-4-1BB, -OX40, or dual costimulation followed by monitoring the presence of peripheral blood CD8 T cells that are CD11ahigh, CD45.1 double-positive using flow cytometry (Fig. 2⇑c). Dual costimulation, in comparison with the other treatments, was more effective at stimulating OT-I T cells in vivo. This was true for recipient mice, which received 50,000 OT-I cells (Fig. 2⇑c, upper panel) where there was a >3-fold increase in the frequency of specific cells in blood. Even the 1,000 OT-I cell (Fig. 2⇑c, lower panel) transfer showed an increase; however, these results were not synergistic and were more variable, potentially due to competition from endogenous SIINFEKL-specific cells. Thus, 4-1BB and OX40 dual costimulation, compared with treatment with either costimulator alone, synergistically enhances in vivo specific-CD8 T cell clonal expansion.
Characterizing effector function and fitness after dual costimulation
Our next goal was to determine whether the effects of dual costimulation were confined to lymphoid tissue. Sixty thousand OT-I cells were transferred into recipient mice, and the day after, they were treated with OVA protein and control rat IgG, anti-4-1BB, -OX40, or dual costimulation. On day 12, peripheral blood, PLN, MLN, spleen, and lung were analyzed for the percentage of CD8 T cells expressing high levels of CD11a and CD45.1 (Fig. 3⇓a). In a separate experiment, 500,000 cells were transferred, and day 20 was examined (Fig. 3⇓a, see bottom panel for liver data). In both experiments, dual costimulation led to a dramatic increase of OT-I cells not only in spleen and LNs but also in lung and liver tissue. Interestingly, the frequency of lung OT-I cells in the OX40 alone group was much higher than the 4-1BB alone group, but the opposite was true for day 20 liver; this may also be related to the difference in the number of OT-I cells transferred between the two experiments. Nevertheless, in both cases, the dual-costimulated mice contained the largest OT-I population. These data show that dual costimulation mediates spread of effector CD8 T cells throughout the body.
Dual costimulation increases the frequency of effector OT-I T cells at late time points in lymphoid and nonlymphoid tissue. Two separate experiments, one transferring 60,000 and a second transferring 500,000 spleen OT-I RAG−/− cells into C57BL/6 mice, were conducted. The following day recipient mice received OVA plus 100 μg (or 150 μg in experiment 2) of control rat IgG; OVA plus 50 μg (or 100 μg) of anti-4-1BB; OVA plus 50 μg of anti-OX40; or OVA plus dual costimulation. On day 12, or day 20 in the second experiment, after immunization, blood, PLN, MLN spleen, lung, and liver were isolated, and the presence of OT-I cells evaluated. a, Shown are dot plots gated on CD8 T cells examining CD11a vs CD45.1 expression. The number indicates the percentage of CD11ahigh cells that are CD45.1+. The bottom panel is from the second experiment analyzing day 20 liver. b, Contour plots show intracellular cytokine staining for day 12 T cells from dual costimulation, which were restimulated in vitro with or without 1 μg of SIINFEKL peptide. The number shows the percentage of CD45.1 cells staining positive for IFN-γ, TNF, or isotype control IgG. Four other similar experiments show data demonstrating that dual costimulation enhances stimulation of CD8 T cells (not shown).
As a test for potency of effector function, spleen cells from both time points (day 20; not shown) were restimulated in vitro with or without SIINFEKL peptide and then assayed for intracellular cytokine production (Fig. 3⇑b). The data show that dual costimulation induced nearly two-thirds of the transferred OT-I cells to produce IFN-γ after restimulation and about one-third made TNF. Although the other individual costimulatory groups generated OT-I IFN-γ producers at both time points (data not shown), the amount of IFN-γ produced per OT-I cell was more when dual costimulation was used (data not shown). Therefore, the fact that many more OT-I T effector cells were present and produced greater amounts of IFN-γ, demonstrates that dual costimulation generates a robust boost to effector responses.
Because of the increase in effector cells observed (Fig. 3⇑a), it was possible that dual costimulation induced more proliferation or ability to undergo an increase in the number of cycles of division. To test this idea, we isolated OT-I cells, stained them with CFSE, and then transferred them into recipient mice. At 50 h after immunization with OVA and costimulation, dilution of CFSE in the OT-I population was analyzed (Fig. 4⇓a). The data show that there was no difference in any of the groups examined. Nevertheless, it was possible that earlier time points may have revealed differences between the groups; thus, 28 or 32 h were examined but showed comparable dilution of CFSE between the groups (data not shown).
Dual costimulation does not increase cell cycle entry or more rounds of division over single costimulators, but does increase OT-1 T cell number. a, OT-I RAG−/− splenocytes were labeled with CFSE and transferred into C57BL/6 mice. After 50 h of in vivo priming with OVA and either anti-4-1BB, anti-OX40, or dual-costimulation treatment, splenocytes were analyzed for CFSE dilution on CD45.1 T cells. Results from 28 or 32 h of in vivo priming were similar (data not shown). b–d, C57BL/6 mice received splenocytes from OT-I RAG−/− mice, and the next day, recipient mice were immunized as in Fig. 3⇑. After 60 h, 1 × 106 spleen cells from each group was cultured separately in a single well without added peptide or any known stimulus. At 20 or 45 h after culture, the cells were analyzed. In b, the data show mean ± SEM of the fold increase of the percentage of OT-I cells from time 0 h of culture: OVA plus rat IgG (□); OVA plus anti-4-1BB (○); OVA plus anti-OX40 (⋄); and OVA plus dual costimulation (▴). The data represent six combined but separate experiments, with one mouse assayed in each group per experiment. c, In this study, 10 μM BrdU was added at time 0 or 30 h, and the cells were removed from culture at 20 h (0↔20 h) or at 45 h (30↔45 h) and analyzed by flow cytometry. Shown are the mean percentage ± SEM of CD45.1 cells staining positive for BrdU. The black bar represents data from the rat IgG group; dotted bar, from anti-4-1BB alone; striped bar, from anti-OX40 alone; and open bar, from the dual-costimulation group. The data are from four combined but separate experiments for a total of four mice per group. d, Splenocytes were labeled with CFSE, and 1 × 106 cells were placed in culture. After 20 (gray line) or 45 h (thick black line), CD45.1+ cells were analyzed for CFSE dilution. These data are from one experiment and are comparable to two other experiments.
Because CFSE dilutes to a point at which the number of cell cycles are no longer discernable, we performed several experiments to study the mechanism of OT-I clonal expansion. The day after OT-I transfer, mice were immunized, and 60 h later, whole spleen populations were placed in vitro without any stimulus except for culture medium. After 20 and 45 h in culture, the percentage of OT-I cells was determined in proportion to time 0 h of in vitro culture. There was a ∼5-fold increase in the percentage of OT-I cells from the dual-costimulation group at 45 h; the other treatments generated a <2-fold response (Fig. 4⇑b). The fold increase in the number of OT-I cells in these cultures from 0 to 45 h was 0.34 for the IgG control group, 2.64 for 4-1BB alone, 1.38 for OX40 alone, and 5.41 for dual costimulation. Thus, dual costimulation enhanced OT-I cell accumulation in vitro without added peptide, although it is possible that peptide was still being presented from the initial immunization.
One possibility was that a greater proportion of OT-I T cells were initiating division. To test this idea, an identical culture was set up as in Fig. 4⇑b, and BrdU was added either at the beginning or halfway through incubation. BrdU incorporation is a sensitive measurement of DNA synthesis, and under our test conditions, the percentage of OT-I cells incorporating BrdU would be attributable to their ability to initiate division. BrdU incorporation during the first 20 h of culture was higher for OT-I cells that received enforced in vivo costimulation compared with the IgG control group (Fig. 4⇑c). Nevertheless, the type of costimulation did not affect the OT-I response, and all groups showed very similar levels of BrdU incorporation. When BrdU was added during the last 15 h of culture, all OT-I populations incorporated equivalent levels of BrdU. Therefore, dual costimulation did not induce a greater proportion of OT-I cells to incorporate BrdU as would be expected if these cells possessed an enhanced ability to initiate cell division.
To further test this idea, cells were taken from recipient mice as above, labeled with CFSE, and assayed for dye dilution at 20 and 45 h after culture. In all groups, the extent of CFSE dilution was similar, but cell counts between the groups were vastly different (Fig. 4⇑d). At both time points, cells primed in dual-costimulation-treated mice induced a massive increase in cell counts compared with the other groups even though the extent of dye dilution was similar. Taken together, the data in Fig. 4⇑ suggest that the OT-I CD8 T cells primed with dual costimulation are far more likely to remain viable after, and during, cell cycle traverse as opposed to treatment with either costimulator alone.
Perhaps the most stringent test of this idea is to assay pure OT-I T cells in the absence of contaminating populations that may provide survival signals in vitro. Thus, OT-I T cells were stimulated in vivo as in Fig. 4⇑ and, at 60 h, purified using MACS columns with anti-CD45.1 mAb. Purity was ∼98% (CD8+CD45R.1+), and cells were placed in vitro without added APCs, other lymphocytes, or peptide. At 0 h of culture, the cells were analyzed for cell cycle status, and all groups possessed a substantial population of cells cycling through S phase, as measured by intracellular DNA content using propidium iodide (data not shown). After 18 h, the cells were analyzed for viability (Fig. 5⇓, top panel) and cell cycle status (bottom panel) by flow cytometry. In vivo priming with OX40 mAb generated 439 viable OT-I cells/50,000 events (viable cells = cells from top panel multiplied by the percentage of OT-I cells from bottom panel), followed by 4-1BB with 2239 OT-I cells. Dual costimulation yielded the greatest number of viable OT-I cells at 4284. Collectively, these data show that dual costimulation does not induce more rounds of division (Fig. 4⇑, a and d), nor a greater ability to initiate DNA synthesis (Fig. 4⇑c), but does enhance accumulation (Fig. 4⇑b), through efficient cell cycle traverse and survival (Fig. 5⇓).
Peptide-specific CD8 T cells possess an intrinsic ability to resist apoptosis when primed with dual costimulation. C57BL/6 mice received 2 × 106 splenocytes from OT-I RAG−/− mice and were immunized the next day with OVA plus anti-4-1BB; OVA plus anti-OX40; or OVA plus dual costimulation. After 60 h, splenocytes were isolated from each mouse, and CD45.1+ T cells were isolated using a MACS column. One hundred thousand cells were placed into a well of a 96-well plate and cultured in medium. After 18 h, cells were stained for CD45.1 expression and then with propidium iodide as described in Materials and Methods. The top panel shows dot plots representing forward and side scatter of each sample, and the number indicates viable cells per 5 × 104 events as determined by flow cytometry. In the bottom panel, dot plots show intracellular propidium iodide profile of OT-I cells gated from the top. Each group had one mouse, and this experiment was repeated three times.
Clonal expansion and effector function of specific CD8 T cells is not dependent on CD4 help
Using multiple in vivo models, we showed for the first time that dual costimulation by 4-1BB and OX40 synergistically induce specific CD8 T cell clonal expansion. We tested whether the responding CD8 T cells required T cell help for this effect. Thus, one group of mice were treated with a CD4-depleting mAb and the other given a control IgG, and depletion was confirmed throughout the experiment. Both groups were injected with SEA plus dual costimulation, and on days 2–5, peripheral blood lymphocytes were analyzed for the presence of responding CD8 Vβ3 T cells (Fig. 6⇓a). The data show that there was a profound increase in specific CD8 T cell clonal expansion whether or not CD4 T cells were present; however, in the presence of CD4 T cells, expansion was enhanced by ∼20%. Therefore, the vast majority of clonal expansion was CD4 independent. Multiple tissues were examined on day 11, and the data show that effector CD8 T cells from CD4-depleted mice did not accumulate to the same degree as cells in control mice (Fig. 6⇓b). It is unlikely that migration can explain this difference, because accumulation was inhibited in nonlymphoid as well as lymphoid tissue. Collectively, this suggests that CD4 help may contribute to CD8 T cell survival in vivo.
CD4 help is dispensable for CD8 effector function but not for survival of effector cells. a–c, B10.A mice were treated with anti-CD4 mAb or rat IgG on day −6 and −2. On day 0, all mice were treated with SEA plus dual costimulation, and peripheral blood T cells were analyzed on days 2, 3, 4, 5, and 11. Depletion of CD4 T cells was confirmed by flow cytometry. a, Shown are the mean percentage ± SEM of peripheral blood CD8 T cells expressing TCR Vβ3 from CD4 depleted (▵) or nondepleted (▴) mice. b, Cells from spleen, PLN, MLN, and liver were analyzed on day 11. Shown are the mean percentage ± SEM of CD8 Vβ3+ T cells, and these data represent combined data from two separate experiments for a total of at least nine mice in each group. c, Spleen cells from days 5 and 12 mice were restimulated with SEA for 5 h and then stained for intracellular cytokine levels. The number represents the percentage of CD8 T cells staining positive for IFN-γ. The data represent one of three mice in each group, and this is one experiment of two completed. d and e, Three separate experiments were conducted in which C57BL/6 mice received either 0.5 × 106, 60,000, or 50,000 splenocytes from OT-I RAG−/− mice. The next day, mice were immunized with SIINFEKL peptide plus anti-4-1BB, peptide plus anti-OX40, or peptide plus dual costimulation. On day 5, spleen and PLN cells were isolated from each group and analyzed as described in Materials and Methods. d, Scatter graphs show percentage (top) and number (bottom) of OT-I cells from 0.5 × 106 (•; mean is black bar), 60,000, or 50,000 (○; mean is gray bar) recipient mice in spleen (left) and PLN (right). e, Shows intracellular cytokine staining of day 5 T cells in response to SIINFEKL peptide. The number represents the percentage of CD45.1 T cells staining positive for IFN-γ. Dot plots from 500,000 (top), 60,000 (middle), and 50,000 (bottom) cell transfer groups are shown.
In a second experiment, we determined whether CD8 effector function was influenced in the absence of CD4 help. Spleen cells were taken from CD4-depleted or nondepleted mice treated 5 and 12 days earlier with SEA and dual costimulation. Cells werestimulated in vitro with SEA for ∼5 h, and IFN-γ synthesis was evaluated by intracellular cytokine staining. Because TCR is down-regulated in this assay, we were only able to examine the percentage of CD8 T cells possessing IFN-γ (Fig. 6⇑c). Thus, on day 5, there was little difference in the percentage of CD8 T cells producing IFN-γ. On day 12, there appeared to be weakened effector function, but this was not likely the case, because the proportion of Vβ3 T cells in the CD4-depleted mice was lower on day 12 compared with CD4-intact mice (data not shown; but for a comparison, see Fig. 6⇑b). Therefore, effector function is intact despite the fact that survival of the effector CD8 T cells is impaired in the absence of help.
To test this idea without depleting cells, we stimulated OT-I recipient mice with SIINFEKL peptide instead of OVA. It is unlikely that CD4 T cells respond to SIINFEKL/MHC I complexes, and therefore, they have little opportunity for TCR ligation, especially because no adjuvant is used. Three groups of OT-I recipient C57BL/6 mice were injected with soluble SIINFEKL peptide plus either anti-4-1BB, -OX40, or dual costimulation. On day 5 after injection, the spleen (Fig. 6⇑d, left panels) and PLN cells (right panels) were analyzed for the percentage of OT-I cells (top panels) and absolute OT-I number (bottom panels) by flow cytometry and carefully counting cell numbers. Differences in the percentages or numbers of OT-I cells between the 4-1BB and OX40 groups were very small regardless of how many cells were transferred. The dual-costimulation group generated massive amounts of OT-I T cells in spleen and PLNs as measured by absolute number or percentage. Although not as dramatic for 50–60,000 cell transfer groups, this trend was still very evident. To measure whether effector function was intact, cells from the same experiment were restimulated in vitro with SIINFEKL peptide or without (data not shown), and the data show that the dual-costimulation group yielded cells that either responded as well as the other groups or even better (Fig. 6⇑e). Therefore, by factoring the proportion, number, and responsiveness of the effector cells, we conclude that dual costimulation provides a massive boost to effector function.
We tested whether dual costimulation would be useful in the rejection of established tumors. The MethA fibrosarcoma (MethA) is a stringent tumor model based on the fact that MethA is a progressive fibrosarcoma induced by methylcoanthrene in immunocompetent mice. A single injection of as few as 100,000 cells intradermally will lead to complete establishment of tumors and uniform mortality within 4 wk in normal BALB/c mice (42). The rejection of MethA induced by heat shock proteins was shown to be dependent on both CD4+ and CD8+ T cells (43, 44). Therefore, in the first experiment, we tested whether dual costimulation would hamper or enhance an established vaccine strategy in this model. In the first experiment, we demonstrated that 500,000 MethA cells expressing membrane-bound gp96 were rejected faster if they received dual costimulation (data not shown); however, dual-costimulation mAbs administered on days 3 and 6 completely protected mice from the control neo-transfected MethA tumor cell line (Fig. 7⇓a). The control rat IgG did not protect the tumor-bearing mice.
Dual costimulation can inhibit tumor growth independent of CD4 help. a, Five hundred thousand MethA sarcoma cells were injected into two groups of BALB/c mice, and 3 and 6 days later, one group received control IgG (□) and the other dual costimulation (▴). b, Four groups of BALB/c mice received 2 × 106 MethA cells. Three days later, one group received control IgG (□), and the second group received dual costimulation (▵); 7 days later, a third group received control IgG (▪), and the last received dual costimulation (▴). c, Either anti-CD4 (▵), anti-CD8 (△), or rat IgG control Ab (▴) was injected into BALB/c mice on days −3 and 6 for in vivo cell depletions. On day 0, 1 × 106 MethA cells were injected into all the mice. On day 7, mice received dual costimulation.∗, Indicates that mice were sacrificed before the end of the experiment due to rapid tumor growth. For all experiments, tumor growth was monitored twice a week using calipers to measure diameters, and b and c use the same x-axis. Mean ± SEM of tumor size from at least five mice per group is shown.
To vigorously test this system, we found that a single injection of dual-costimulatory mAbs was very effective in retarding tumor cell growth. A lethal dose of 2 × 106 MethA tumor cells were injected on day 0, and on days 3 or 7, either control rat IgG or dual costimulation was administered (Fig. 7⇑b). Day 3 dual-costimulation mAbs were not effective, but day 7 injection completely blocked tumor growth compared with the IgG control. Based on this data and previous figures, we hypothesized that dual costimulation may impair tumor cell growth in a CD8-dependent manner. Therefore, three different groups of mice were given IgG control, or CD4- or CD8-depleting mAbs, and inoculated with 1 × 106 tumor cells. All groups were treated with dual-costimulation mAbs on day 7 (Fig. 7⇑c). We found that CD8 T cells were necessary for inhibition of tumor cell growth, but perhaps surprisingly, we found that CD4 T cells were completely dispensable. Collectively, these data show that dual costimulation can drive vigorous antitumor effector responses even under immunocompromising conditions such as those in the absence of T cell help.
Discussion
It has been hypothesized that triggering multiple costimulators may enhance T cell immune responses in vivo. In the past, it has been shown that OX40- and CD40-combined costimulation do not significantly enhance T cell responses in vivo (26), and therefore, the data showing that dual costimulation by OX40 and 4-1BB synergistically induce CD8 responses were surprising. This was not the case for 4-1BB and CD40, which induced only an additive effect on both CD4 and CD8 T cells (Fig. 1⇑).
To rigorously study dual costimulation by OX40 and 4-1BB, we conducted experiments in three different murine models where specific T cells were tracked in vivo. The data convincingly show that simultaneous dual costimulation by OX40 and 4-1BB mAbs synergistically stimulate SAg- or peptide-specific CD8 T cells to clonally expand. This included endogenous CD8 T cells like those specific to SAg or peptides, as well as cells that have been adoptively transferred into recipient mice and then immunized with whole protein (Fig. 2⇑). Perhaps the most impressive data testing dual costimulation was that precursor frequency of the specific CD8 T cells was not restrictive. Throughout this study, it was demonstrated that high or low precursor frequency was not a limiting factor. Even when transferring as much as 500,000 or as little as 50,000 OT-I spleen cells, the synergistic effect was clearly evident, and at 1,000 OT-I cells, an enhancement was also observed, although this was complicated by the fact that competition by endogenous SIINFEKL-specific CD8 T cells was likely. This notion is consistent with a previous study demonstrating that endogenous cells can compete with transferred cells (45). Nonetheless, the endogenous cells responded synergistically by day 6 after immunization, demonstrating that precursor frequency did not limit the effects of dual costimulation (Fig. 2⇑, b and c).
Recently, data from infectious disease models (29, 46) have demonstrated that responding primary and memory specific-CD8 T cells can migrate to nonlymphoid tissue, and likely take up residence (47). Perhaps this is an important mechanism to control future infections (48) but nevertheless is a trademark of a robust effector response. On day 12, OT-I cells primed with dual costimulation migrated to (non)lymphoid tissues with the highest frequency of OT-I cells detected in lung, which was the same for OX40 or 4-1BB alone (Fig. 3⇑a). Interestingly, we detected a substantial population of activated CD11ahighCD45.1− endogenous cells. Presumably, these cells were responsive to OVA or possibly the agonist mAb itself. However, upon restimulation with SIINFEKL peptide, only the OT-I T cells responded by producing effector cytokines like TNF or IFN-γ (Fig. 3⇑b). Therefore, dual costimulation did not alter migration, but maintained survival of the effector T cells and their function, although the mechanism of expansion was not clear.
To address this issue, we tested whether dual costimulation accelerated or increased rounds of cellular division, thereby leading to robust clonal expansion. The evidence suggested that dual costimulation did not accelerate division in vivo (Fig. 4⇑a), although it was impossible to tell whether increased rounds of division occurred, because as the cells divided, they retained an undetectable level of CFSE. To address this issue, cells from dual-costimulated mice vs individual-costimulated mice were placed in vitro without any known stimulus and then enumerated 2 days after culture (Fig. 4⇑b). Based on proportion or absolute number, there was a ∼5-fold increase in the dual-costimulatory group and a small increase in the others. This is even greater than the effect that LPS has on rescuing T cells from deletion (49). Therefore, dual costimulation increased clonal expansion by enhancing accumulation, but not by accelerating division.
Accumulation of activated T cells after immunization is a complicated biological process. This phenomenon is based on at least the ratio of T cell survival to death, vs proliferation. In one example of many, T cells may proliferate and die at the same rate and thereby do not accumulate beyond their initial numbers; however, populations that proliferate at the same rate but die less frequently will accumulate. Nonetheless, the half-life of a circulating T cell is very difficult to accurately measure, because techniques like BrdU incorporation are fraught with underappreciated toxicity issues (50). Thus, incorporation of BrdU can be minimally or reliably interpreted as an attempt to initiate cell cycle traverse. This is based on a short exposure to BrdU in a time span that permits BrdU incorporation. Our experiments show that costimulation priming provided an advantage in the uptake of BrdU in specific T cells at the beginning of culture, but not at the end (Fig. 4⇑c). Nevertheless, T cells from dual-costimulated mice did not possess more or less of an ability to initiate cell division than the other costimulation groups. Therefore, even if peptide was being presented in this in vitro culture as a result of the initial immunization, these data show that it did not preferentially influence cells from the dual-costimulation group compared with the others.
To visualize survival and proliferation at the same time, the cells were labeled with CFSE and monitored in vitro as before (Fig. 4⇑d). These data suggest that, although the ability to initiate division is the same between all costimulation groups, there is a greater propensity of OT-I cells to survive through the cell cycle. One possibility to explain how dual costimulation mediates survival is to examine whether the OT-I cells possess an intrinsic program allowing them to preferentially survive. Alternatively, nonspecific cells like dendritic cells, macrophages, or other lymphocytes may provide the OT-I cells with survival signals. Pure OT-I T cells were clearly more capable of surviving when they were derived from mice treated with dual costimulation through OX40 and 4-1BB as opposed to either alone (Fig. 5⇑). The molecules involved in this process have not been identified. One possibility is a role for Bcl-xL, Bcl-2, and/or protein kinase B, which are reported to be essential for OX40 signaling in CD4 T cells (51, 52), although our data show little survival of CD8 T cells when costimulated through OX40 alone. From a cellular perspective, it is also possible that CD8 T cell survival is enhanced when CD4 T cells are present during priming of the CD8 T cells. Recently, CD4 T cells were shown to play a minimal role in helping the primary response of CD8 T cells, but in contrast influenced the secondary response of specific CD8 T cells (53, 54, 55). Therefore, we formulated the hypothesis that CD4 T cells may not play a role influencing CD8 T cells in our dual-costimulation model.
Depletion of CD4 T cells minimally affected CD8 T cell clonal expansion, but in contrast, survival of the effector cells was substantially impaired in all (non)lymphoid tissues examined (Fig. 6⇑, a and b). These data complement and extend those reported in the pathogen models discussed above. However, perhaps the most important result was that effector function was maintained even at the later stages when specific T cell accumulation was impaired (Fig. 6⇑c). These data demonstrate that survival and effector function are not always linked or dependent upon each other. In a second test of this idea, mice were primed with a CD8-specific peptide instead of whole protein to eliminate helper function, and we detected massive clonal expansion coupled to a robust effector response (Fig. 6⇑, d and e). One caveat is that the mAbs themselves may have provided a CD4 TCR signal; however, the response of these potential CD4 cells would theoretically lag behind the CD8 response, because SIINFEKL peptide was already processed in contrast to the putative CD4 epitopes in the mAbs. Taken together, whether the life span of the effector CD8 T cell was impaired or not, or importantly whether CD4 T cells were present or present but not stimulated, CD8 effector cytokine production was robust. Therefore, we hypothesized that this strategy may be effective under recalcitrant conditions of immunity.
We tested whether dual costimulation could help fight immunologically resistant tumors. Initially, we showed that the combination of 4-1BB and OX40 efficiently induced rejection of an established fibrosarcoma in mice (Fig. 7⇑a). To investigate the mechanism, we demonstrated that a single injection of mAb 7 days, as opposed to 3 days, after tumor establishment was sufficient to completely retard tumor growth (Fig. 7⇑b). Two major aspects with these experiments require mention: first, total tumor rejection was observed with only two injections of mAb; and second, we used mAbs without any immunizing tumor Ag. In contrast, CTLA4 blockade without a tumor Ag vaccine targeting p53 was not very effective in MethA tumor rejection (56). It remains to be seen whether the combination of dual costimulation and a tumor Ag vaccine is most effective.
Ultimately, we examined antitumor effectiveness of dual costimulation under immunocompromising conditions such as in the absence of CD4 T cells. In this model, there is a substantial amount of evidence suggesting that CD4 T cells are essential for tumor rejection under vaccinating conditions or even during immunotherapy (42, 57). Dual costimulation was very effective in the presence or absence of CD4 T cells, but completely dependent upon CD8 T cells (Fig. 7⇑c). Taken together, these data suggest that dual costimulation triggers CD8 effector function to eliminate tumor cells even in the absence of T cell help. Therefore, we conclude that dual costimulation may provide a fertile avenue for immunotherapy in humans suffering from cancer.
Acknowledgments
We thank Drs. Lefrancois and McSorley for helpful comments regarding data in the manuscript. We also thank the University of Connecticut Health Center (UCHC) flow cytometry core for help with flow cytometry, and the UCHC animal facility.
Footnotes
-
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.
-
↵1 This work was supported by in part by National Institutes of Health Grants AI42858A and AI52108 (to A.T.V.). L.M. is supported by National Institutes of Health Immunology Training Grant T32-AI07080.
-
↵2 Address correspondence and reprint requests to Dr. Anthony T. Vella, Division of Immunology, MC1319, University of Connecticut Heath Center, 263 Farmington Avenue, Farmington, CT 06032. E-mail address: vella{at}uchc.edu
-
↵3 Abbreviations used in this paper: SEA, staphylococcal enterotoxin A; SAg, superantigen; BSS, balanced salt solution; PLN, peripheral lymph node; MLN, mesenteric lymph node; RT, room temperature.
- Received April 7, 2004.
- Accepted June 25, 2004.
- Copyright © 2004 by The American Association of Immunologists
References
In this issue
