Functional myeloma-reactive cells are present in the PD-1+CD8+ T cell subset
The immunogenic cancer antigens on 5T33 myeloma are unknown. Therefore, to identify T cells with myeloma antigen specificity, we used a 5T33 cell line expressing the model antigen SIINFEKL ovalbumin (OVA) peptide (5T33-GFP-OVA), along with GFP, to facilitate identification of the cells in vivo. To show that PD-1 is up-regulated on myeloma-reactive T cells, KaLwRij mice were inoculated with 2 × 106 5T33-GFP-OVA cells iv. Mice were euthanized, and spleens and bone marrow harvested 30–35 days after inoculation. CD8+ T cells that recognize SIINFEKL were detected by flow cytometry using fluorescently labeled H2Kb/SIINFEKL pentamers. Our results show that greater percentages and absolute numbers of both spleen and bone marrow PD-1+CD8+ T cells were SIINFEKL pentamer-positive as compared to PD-1−CD8+ cells (Fig. 1a). These data directly show that myeloma-specific CD8+ T cells are enriched in the PD-1+ population.
To examine whether PD-1+CD8+ T cells secrete cytokine in response to cancer antigen stimulation, IFN-γ ELISPOT assays were performed. For these assays, PD-1+CD8+ and PD-1−CD8+ T cells were sorted by flow cytometry and stimulated with 5T33 myeloma cells. While some PD-1+CD8+ T cells secreted IFN-γ in response to myeloma antigens (Fig. 1b), this number was significantly enhanced either by inclusion of anti-PD-L1 blocking antibody during the assay cell co-culture, or by antigen stimulation with 5T33 myeloma modified to express the co-stimulatory molecule CD80. These data clearly show that the PD-1+CD8+ T cell subset is enriched in myeloma-reactive T cells, but that many of the cells are relatively inactive in the absence of PD-1 blockade or additional co-stimulation.
To show that PD-1 expression identifies cancer antigen-reactivity in another hematologic malignancy model (C1498 acute myeloid leukemia), the percentages and absolute numbers of cancer-reactive cells were determined, and IFN-γ ELISPOT assays performed on T cells harvested from mice bearing C1498. PD-1+CD8+ and PD-1−CD8+ T cells were sorted from the spleens of mice that had been inoculated iv with C1498 cells engineered to express the model peptide antigen peptide SIY (SIYRYYGL; C1498-SIY). As with the 5T33 model, our results show that PD-1+CD8+ T cells are highly enriched in cancer antigen reactivity (Fig. 1c). Similar to the myeloma model, addition of anti-PD-L1 to the ELISPOT assays resulted in a significant increase in numbers of PD-1+CD8+ T cells secreting IFN-γ (Fig. 1d).
PD-1+ T cells from myeloma-bearing mice are phenotypically heterogeneous and secrete effector cytokines
In moribund myeloma-bearing (MB) mice, we previously showed that splenic PD-1+ T cells stimulated with anti-CD3 exhibit an altered cytokine profile (i.e., secreted less IL-2, IFN-γ and TNF-α) as compared to PD-1− T cells or T cells from non-MB mice [5]. This prompted us to determine if PD-1+ T cells co-express markers of T cell dysfunction or activation, or retain the ability to produce effector cytokines when analyzed prior to generation of advanced disease. The phenotype and function of PD-1+ T cells was determined 28 days after 5T33 inoculation. This time point is before mice become moribund, which is typically 35–45 days following 5T33 inoculation. At 28 days, myeloma comprises 1–4% of total spleen cells, unlike moribund mice, where approximately 5–20% of the spleen consists of myeloma (data not shown).
The percentage of spleen PD-1+CD4+ and CD8+ T cells in naïve non-myeloma bearing mice is relatively low (~4–7%), as compared to moribund 5T33 bearing mice where 20–60% are PD-1+. In naïve mice, only about 1% of PD-1+CD8+ spleen T cells co-express the checkpoint receptor TIM-3, whereas in moribund 5T33 mice approximately 10% of PD-1+CD8+ spleen T cells express TIM-3 [5]. For this study, we compared the phenotype of PD-1+ and PD-1− T cells from 5T33 bearing mice prior to advanced disease. To characterize PD-1+ T cells, spleens were harvested on day 28 and co-expression of PD-1 with various inhibitory and activation molecules was determined by flow cytometry. Figure 2a shows the percentage of total spleen cells co-expressing PD-1 and the other markers tested (upper right quadrant). The bracketed values in each upper right quadrant represent the percentages of PD-1+ T cells that co-expressed the marker of interest. Notably, 37% and 77% of PD-1+CD8+ T cells co-expressed the checkpoint receptors TIM-3 and LAG-3, respectively. However, 81% and 70% of PD-1+CD8+ T cells also co-expressed activation markers OX40 and CD103, respectively (Fig. 2a, top panel). 34% of CD8+PD-1+ T cells co-expressed CD137. For PD-1+CD4+ T cells, 51% and 79% expressed TIM-3 or LAG-3 checkpoint receptors, respectively (Fig. 2a, bottom panel). Of the PD-1+CD4+ T cells, 52% expressed Foxp3 as compared to approximately 12% of PD-1−CD4−T cells (Fig. 2b). These data show there are multiple subsets of PD-1+CD8+ and CD4+ T cells expressing both checkpoint receptors and activation markers. T cells that co-express multiple inhibitory receptors have been reported to be dysfunctional relative to cells that express PD-1 alone or no inhibitory receptors [11].
To compare how PD-1+ and PD-1− T cells respond functionally to activation signals, cells were sorted into PD-1+ and PD-1− T cell subsets and activated with plate-bound anti-CD3 and anti-CD28 for 6 h. This strong activation was used to optimize detection of cytokines produced by the cells. Functional status was assessed by examining presence of IFN-γ, TNF-α, granzyme B and Ki67 by intracellular flow cytometry. For CD8 T cells, there were no statistical differences in percentages of PD-1+ T cells expressing intracellular IFN-γ, TNF-α, granzyme B or Ki67 as compared to PD-1− T cells (Fig. 2c, top panel). However, there was a significant decrease in percentages of PD-1+CD8+ T cells that expressed both IFN-γ and TNF-α as compared to PD-1−CD8+ T cells. Similar to CD8+ T cells, significantly fewer PD-1+CD4+ T cells co-expressed IFN-γ and TNF-α as compared to PD-1−CD4+ T cells (Fig. 2c, bottom panel). Significantly lower percentages of PD-1+CD4+ T cells expressed TNF-α as compared to PD-1−CD4+ T cells. Surprisingly, the PD-1+CD4+ T cells had higher Ki67 expression as compared to PD-1−CD4+ T cells. Overall, these data suggest that in response to strong activation signals, PD-1+T cells may be proliferative and they produce similar IFN-γ but less TNF-α as compared to PD-1− T cells.
To further evaluate the ability of PD-1+ T cells to produce and secrete effector cytokines, PD-1+ and PD-1− T cells were stimulated with plate bound anti-CD3 for 48 h and culture supernatants collected. Supernatants were then analyzed for cytokine content using a multiplex platform. PD-1−CD8+ and CD4+ T cells produced significantly more IL-2 and GM-CSF than PD-1+ T cells (Fig. 2d). PD-1−CD4+ T cells produced significantly more TNF-α than PD-1+CD4+ T cells. However, the amount of IFN-γ in the PD-1+CD8+ T cell supernatant was not quantitatively different than that in the supernatant collected from PD-1−CD8+ T cells. In fact, there was significantly more IFN-γ in the supernatant of PD-1+CD4+ T cells as compared to PD-1−CD4+ T cells. Of particular note, both PD-1+CD4+ and CD8+ T cells produced increased amounts of IL-10 as compared to PD-1− T cells. Up-regulation of IL-10 production in IFN-γ-producing PD-1+ effector T cells may be a consequence of chronic antigen activation. Co-production of IFN-γ and IL-10 has been reported in Th1 T cells during chronic mouse infections [15, 16].
In summary, prior to advanced 5T33 myeloma burden, there are splenic PD-1+ T cells that appeared to be chronically activated, as demonstrated by expression of activation markers CD69, OX-40 and CD103, and inhibitory receptors LAG-3 and TIM-3. When activated, PD-1+ T cells expressed the Ki67 proliferation marker, and produced significantly less IL-2, similar or more IFN-γ and more IL-10 than PD-1− T cells.
PD-1+ T cells from myeloma-bearing mice expand in culture and maintain their reactivity
During chronic viral infection and cancer, up-regulation of PD-1 has been shown to be a marker of T cells with reduced ability to proliferate and secrete effector cytokines [17, 18]. In the 5T33 myeloma model we have shown that PD-1+ T cells harvested from non-moribund MB bearing mice can be activated to secrete cytokines. However, to use PD-1+ T cells for ACT, they must be able to undergo expansion ex vivo and retain effector function. To determine if these qualities persisted in T cells isolated from 5T33-bearing mice, flow cytometric sorted PD-1+ and PD-1− CD8 T cells were activated with anti-CD3 and anti-CD28 antibodies and expanded in culture for 7 days with 20 U/ml IL-2, 5 ng/ml IL-7 and 5 ng/ml IL-15. PD-1+CD8+ T cells expanded in vitro approximately 12-fold after 7 days in culture (Fig. 3a). Almost all expanded cells expressed the CD44 activation marker, and around 50% had a CD44+CD62L− effector phenotype (Fig. 3b). Interestingly, PD-1+CD4+ T cells lost expression of Foxp3 during the expansion (Fig. 3c versus Fig. 2b). To show that expanded T cells maintained effector function, IFN-γ ELISPOT assays were performed. Figure 3d shows that expanded PD-1+CD8+ T cells secreted IFN-γ in response to myeloma when checkpoint blockade or co-activation through CD80 was provided. The ELISPOT results show that when checkpoint blockade is provided, there are approximately 100 functional myeloma-reactive CD8+ T cells for every 105 PD-1+CD8+T cells. Significantly fewer PD-1−CD8+ T cells secreted IFN-γ under similar conditions. Together, these data show that within the population of ex vivo expanded PD-1+ T cells, around 50% have an activated effector phenotype, few of the cells are CD4+Foxp3+
, and 5T33-reactive PD-1+CD8+ T cells secrete IFN-γ.
ACT with cultured PD-1+ CD8+ and CD4+ T cells eliminates myeloma in vivo
To examine whether PD-1+ T cells could provide anti-myeloma immunity in vivo, cultured/expanded cells were infused into MB C57BL/6-Rag-1-deficient mice as ACT. Rag-1-deficient mice were chosen for these experiments to avoid the need for preconditioning (i.e., WBI), and to permit analysis of individual T cell subsets that were infused as ACT. Rag-1-deficient mice were inoculated with 106 5T33-GFP myeloma cells iv. Five days later, mice were given ACT with 3-4 × 106 PD-1+CD4+ and CD8+ T cells at a CD4:CD8 ratio of 1:1. Since our IFN-γ ELISPOT data demonstrated that myeloma-reactive PD-1+ T cells required PD-L1 blockade to enhance IFN-γ secretion, some mice also received anti-PD-L1 antibody intraperitoneally on days 7, 10, 14 and 17 (Fig. 4a). Mice were then followed for survival and euthanized when moribund. Mice given no treatment died within 40 days after 5T33 inoculation (Fig. 4b). There was a significant delay in cancer progression in mice that received ACT of expanded PD-1+ T cells, and about 30% of these mice survived beyond 100 days. Co-administration of expanded PD-1+ T cells and anti-PD-L1 further improved survival and eliminated myeloma in 100% of mice (Fig. 4b), demonstrating that ongoing PD-L1 blockade was needed to achieve optimal efficacy.
Next, we compared the anti-myeloma efficacy of different cultured/expanded T cell subsets given as ACT. Since PD-L1 blockade synergized with ACT to produce more effective cancer regression in Fig. 4b, all mice given ACT were treated with anti-PD-L1 for this study. Rag-deficient mice were treated as in Fig. 5a. Mice received the following T cell subsets: (1) combined 1:1 ratio of PD-1+ CD4+ and CD8+ T cells, (2) combined 1:1 ratio of PD-1− CD4+ and CD8+ cells, (3) PD-1+CD8+ T cells alone, or (4) PD-1+CD4+ T cells alone. For condition #3 (PD-1+CD8+ T cells alone), we were able to calculate from the ELISPOT data in Fig. 3d that there were approximately 20,000 functional myeloma-specific PD-1+CD8+ T cells infused. As observed in the previous experiment, mice that did not receive ACT died within 50 days after myeloma inoculation. Ninety percent of mice given the combination of PD-1+CD4+ and CD8+ T cells survived for 100 days (Fig. 4c). In contrast, none of the mice treated with PD-1−CD4+ and CD8+ T cells survived past day 50 after myeloma inoculation (Fig. 4c). These data provide compelling evidence that PD-1+ T cells provide anti-myeloma reactivity in vivo. Furthermore, while the PD-1+CD4+ and CD8+ T cell subsets each contained anti-myeloma reactivity, the combination of PD-1+CD4+ and CD8+ T cells provided the best anti-myeloma effect.
Adoptively transferred PD-1+ T cells persist in recipient mice and provide a long-term anti-myeloma response
The in vivo anti-myeloma immunity provided by the adoptively transferred PD-1+ T cells prompted us to test whether the cells persisted and were capable of providing memory. To test this, mice given PD-1+ T cells as ACT that had eliminated established 5T33 myeloma were re-challenged with 2 × 106 5T33 myeloma cells 120 days after the initial inoculation. Five days after myeloma re-challenge, spleens and bone marrow were harvested to analyze persisting T cells. Figure 5a shows the percentages of CD8+ (4.7%) and CD4+ (3.6%) T cells detected in the spleens by flow cytometry. Phenotypic analysis of surviving CD8+ T cells harvested from both the spleen and bone marrow is shown in Fig. 5b. Most of the transferred cells remained activated as indicated by CD44 expression (Fig. 5b). Importantly, both CD4+ and CD8+ T cells with a memory phenotype (CD44+CD62L+) were present in both the spleen and bone marrow. PD-1 was expressed on greater than 50% of splenic and 75% of bone marrow CD8+ T cells. The memory marker CD127 (IL-7Rα) was assessed on one cohort of pooled mice. Figure 5c shows expression of CD127 on CD8+ T cells harvested from both the spleen and bone marrow. IFN-γ ELISPOT assays were also performed on both spleen and bone marrow-derived T cells to assess anti-myeloma function. CD8+ T cells were isolated by immunomagnetic cell sorting and stimulated with wild-type 5T33 myeloma (5T33-WT) or 5T33-WT plus 10 μg/ml anti-PD-L1 in the assay wells (5T33-WT + anti-PD-L1). T cells from spleen and bone marrow produced IFN-γ in response to myeloma (Fig. 5d). As shown previously, IFN-γ production was increased when anti-PD-L1 was added to the assay wells. These data show that when PD-1+CD4+ and CD8+ T cells are adoptively transferred into Rag1-deficient mice, they remain activated long term with some cells expressing memory markers.