PPAR-Induced Fatty Acid Oxidation in T Cells Increases the Number of Tumor-Reactive CD8⁺ T Cells and Facilitates Anti–PD-1 Therapy

The checkpoint molecules CTLA-4 and PD-1 regulate the balance between immune surveillance by immune cells and immune escape by tumor cells (1–3). Inhibition of CTLA-4 and PD-1 with monoclonal antibodies (mAbs) suppresses tumor growth in animal models (4, 5). Clinical trials using these immune-checkpoint blockade mAbs led to FDA approval of ipilimumab (CTLA-4) and nivolumab (PD-1) for use in cancer treatment (6, 7). Nivolumab demonstrated durable clinical activity with less severe side effects than ipilimumab (8, 9). The FDA has approved additional antibodies such as pembrolizumab and durvalumab against PD-1 or its ligand PD-L1 to treat various human cancers, including melanoma, non–small cell lung carcinoma, kidney cancer, non-Hodgkin lymphoma, head and neck cancer, urothelial carcinoma, hepatocellular carcinoma, and gastric cancer (10).

Although PD-1 blockade therapy shows several advantages compared with conventional chemotherapies, a substantial number of patients do not respond to this therapy (10). Therefore, biomarkers for distinguishing responders from nonresponders before initiating PD-1 blockade treatment and combination therapy to improve PD-1 blockade efficacy must be developed. Despite attempts to identify biomarkers for cancer immunotherapy, only PD-L1 expression was informative for non–small cell lung carcinoma in the clinic (11, 12). To overcome the low response rate of PD-1 blockade therapy, various combinations involving chemotherapies, radiotherapy, cancer vaccines, other immune-checkpoint inhibitors, and several immune-stimulatory agents have been examined (10), but with disappointing results. The FDA has approved only two combinations for PD-1 blockade: CTLA-4 blockade or chemotherapy (13, 14).

Infiltration of lymphocytes to tumor sites is a prognostic factor for numerous cancers (15, 16). Additionally, the efficacy of PD-1 blockade is correlated with the infiltration of CD8⁺ T cells in tumors (17). Thus, it is necessary to understand the mechanism of tumor-reactive CTL activation and differentiation to control tumoricidal effects during PD-1 blockade. Naïve T cells are converted into effector T cells, which are responsible for tumor clearance, with an increased glycolytic rate during immune responses. After antigen clearance, a small fraction of T cells differentiate into long-surviving memory T cells, whose energy generation depends mainly on fatty acid oxidation/oxidative phosphorylation (FAO/OXPHOS) in the mitochondria (18).

PD-1 blockade enhances glycolysis in T cells and is responsible for recovery of effector function (19). The role of mitochondrial energy metabolism during development of T-cell antitumor immunity mediated by PD-1 blockade, however, remains unclear. Mitochondria are activated in tumor-reactive CTLs during PD-1 blockade therapy, indicating that T-cell activation and differentiation are associated with mitochondrial energy metabolism (20). Less mitochondrial activation was observed in CTLs derived from mice bearing PD-1 blockade–insensitive tumors, suggesting that mitochondrial activation may serve as a biomarker for the effectiveness of PD-1 blockade therapy (20). Others have shown that PGC-1α, a regulator of mitochondrial biogenesis, is important for development of T-cell–mediated antitumor immunity (21). We demonstrated previously that activation of PGC-1α/peroxisome proliferator–activated receptors (PPAR) by bezafibrate improves the efficacy of PD-1 blockade (20). However, the mechanism by which the PPAR pathway activates antitumor immunity remains unknown.

In this study, we investigated the effect of bezafibrate on the phenotype of effector CD8⁺ T cells and on mitochondria. We investigated the molecular mechanism by which bezafibrate modulates CTL differentiation and enhances T-cell–based antitumor immunity. We found that bezafibrate promotes differentiation of naïve to effector T cells, upregulates FAO, and inhibits apoptosis of effector T cells. These different effects of bezafibrate on the priming and effector phases resulted in increased numbers of functional effector T cells and improved effectiveness of PD-1 blockade. Therefore, combining the reprogramming of energy metabolism through PPAR signaling in T cells with PD-1 blockade may be a promising strategy for combination therapy.

C57BL/6N and BALB/c mice were purchased from The Charles River Laboratory. CD8^−/− mice were provided from The Jackson Laboratory (originally from Univ. Toronto, T.W. Mak). Mice were approximately 5 to 6 weeks old at the time of the experiment and maintained under specific pathogen-free conditions at the Institute of Laboratory Animals, Graduate School of Medicine, Kyoto University, under the direction of the institutional review board. The murine fibrosarcoma cell line (MethA) was obtained from the Cell Resource Center for Biomedical Research, and murine colon carcinoma cell line (MC38) was a gift from Dr. James P. Allison, Memorial Sloan-Kettering Cancer Center (New York, NY). Cell lines were cultured in RPMI medium (Gibco; 11875-093) with 10% (v/v) heat-inactivated fetal bovine serum and 1% (v/v) penicillin–streptomycin mixed solution (Nacalai Tesque; 26253-84). Cell lines were free of mycoplasma contamination. Cell lines were used within fifth passage and were not authenticated in the past year.

A total of 5 × 10⁵ cells of MC38 and MethA were intradermally (i.d.) injected into the right flank of C57BL/6N and BLAB/c mice, respectively (day 0). Combination therapy was started when the tumor size reached 60 to 70 mm³ (around day 7). Mice were intraperitoneally (i.p.) injected with 40 μg of anti–PD-L1 (clone 1-111A.4) and mAb injection was repeated every sixth day. Bezafibrate (Santa Cruz Biotechnology) was i.p. injected at 2.5 mg/kg every third day. For untreated mice, an isotype control for anti–PD-L1 (Rat IgG2a) was injected. Tumor measurement was performed on each alternate day and tumor volume was calculated using the formula for typical ellipsoid π × (length × breadth × height)/6.

CD8⁺ T cells were isolated from the lymph node and spleen of CD45.1 congenic mice using an autoMACS Pro Separator (Miltenyi Biotec). After washing with PBS, CD8⁺ T cells were labeled with CellTrace Violet (Thermo Fisher Scientific). For labeling, CD8⁺ T cells were incubated for 15 minutes with CellTrace diluted in PBS. After quenching with complete media and washing the cells twice with PBS, the labeled CD8⁺ CD45.1⁺ T cells were intravenously (i.v.) injected in to CD45.2 CD8^−/− mice. MC38 cells (5 × 10⁵ cells) were i.d. injected 2 days after CD8⁺ T-cell infusion. Anti–PD-L1 and bezafibrate were i.p. administered 5 days after tumor inoculation. Mice were sacrificed on day 9 for analysis.

Bezafibrate was used at the dose of 2.5 mg/kg for combination therapy. Bezafibrate was freshly prepared immediately before use in DMSO. Dissolved bezafibrate was diluted in PBS and 200 μL was i.p. injected per mouse. Bezafibrate was added at the concentration of 30 μmol/L for the entire culture period.

To isolate naïve CD8⁺ T cells from C57BL/6N wild-type mice, the spleen and three LNs (axillary, brachial, and inguinal LNs) from both the right and left sides were harvested. The spleen was minced, treated with ACK buffer for 2 minutes to lyse erythrocytes, and mixed with pooled and minced LN cells. Naïve (CD44⁻) CD8⁺ T cells were then purified from total pooled lymphocytes according to the manufacturer's instructions (Miltenyi Biotec; 130-096-543).

For draining lymph node (DLN) analysis, axillary, brachial, and inguinal LNs (one of each) were harvested from the right side of tumor-bearing mice. All LNs were minced and pooled. Average LN cell numbers (total pooled LN cells/3) were used as absolute cell numbers. For tumor-infiltrating lymphocyte (TIL) analysis, tumor tissue was harvested and minced into 1- to 2-mm pieces with scissors followed by digestion with collagenase type IV (Worthington Biochemical Corporation) using a gentleMACS Dissociator (Miltenyi Biotec). The numbers of TILs per mg of tumor tissue were used as the absolute numbers. For in vitro analysis, naïve CD8⁺ T cells were stimulated with beads coated with anti-CD3 and CD28 (Thermo Fisher Scientific) and with recombinant human IL2 (100 U/mL; PeproTech). The expanded T cells on days 2 and 13 were used for proliferation assays and apoptosis assays, respectively.

The following monoclonal antibodies (mAb) were used to detect the respective antigens: CD44 (IM7), CD45.2 (104), CD45.1 (A20), CD8 (53-6.7), CD62L (MEL-14), T-bet (4B10), IFNγ (XMG-1.2), and Bcl-2 (BCL/10C4) from BioLegend; EOMES (Dan11mag), and Ki67 (SolA15) from eBioscience: Cpt1a (8F6AE9) from Abcam. All flow cytometry experiments were performed on a FACSCanto II (BD Biosciences) and analyzed using FlowJo software (FlowJo, LLC). Details of intracellular staining were mentioned previously (20). Mitochondrial mass, mitochondrial potential, mitochondrial superoxide, and cellular reactive oxygen species (ROS) were determined by MitoTracker Green, MitoTracker Deep Red, MitoSOX Red, and CellROX Green reagents, respectively (all from Life Technologies). These dyes were added to the cells at final concentrations of 0.125, 0.125, 5.0, and 0.625 μmol/L and incubated at 37°C in a 5% CO₂ humidified incubator for 30 minutes, followed by surface staining.

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of CD8⁺ T cells isolated from treated mice were measured using an XF^e96 Extracellular Flux analyzer (Seahorse Biosciences). Cells (400,000 per well) were seeded in an XF^e96 plate, determined by titration, as previously described (20). Different parameters from the OCR graph were calculated. ATP turnover was defined as follows: (last rate measurement before oligomycin) − (minimum rate measurement after oligomycin injection). Maximal respiration was defined as follows: (maximum rate measurement after FCCP) − (nonmitochondrial respiration). Spare respiratory capacity (SRC) was calculated by subtracting basal respiration from maximal respiration. We measured the ECAR value in the same well, which contained an optimal glucose level so the basal ECAR (or glycolysis) value is the reading we obtained immediately before oligomycin injection. We prepared the assay medium as described in the XF cell Mito Stress Test Kit (Kit 103015-100). The glucose concentration in this medium is 10 mmol/L. In the classic glycolytic assay procedure (glucose-free media), the final concentration of glucose added to the port is 10 mmol/L while measuring flux. The basal ECAR value in this method is calculated by subtracting the last rate measurement before glucose injection from the maximum rate measurement before oligomycin injection, which gives the same value as calculated by our method. Glycolytic capacity was defined as the rate measured after oligomycin injection. Glycolytic reserve was defined as follows: (glycolytic capacity) – (basal ECAR value) (22).

We isolated RNA from purified CD8⁺ T cells with the RNeasy mini kit (QIAGEN) and synthesized cDNA by reverse transcription (Invitrogen). The primers used to perform real-time PCR are listed in Supplementary Table S1.

Total RNA was extracted using the RNeasy Micro kit (QIAGEN) according to the manufacturer's protocols. CD8⁺ T cells expanded in vitro until day 13 were lysed according to the protocol. Microarray analysis was performed by Macrogen using GeneChip Mouse Gene 2.0 ST Array. The data were deposited at GEO repository (http://www.ncbi.nlm.nih.gov/geo) and the accession number is GSE118659. Gene-enrichment and functional annotation analysis was performed using Gene Ontology (www.geneontology.org/) and KEGG (www.genome.jp/kegg/). All data analysis and visualization of differentially expressed genes was conducted using R 3.1.2 (www.r-project.org).

CD8⁺ T cells were isolated from DLNs using mouse CD8 MicroBeads (Miltenyi Biotec). After washing the cells with PBS twice, 2 × 10⁶ cells were used. Details were as described (20). Antibodies recognizing p-CREB (ab32096) and Cpt1a (ab128568) were obtained from Abcam.

Statistical analysis was performed using Prism 6 (GraphPad Software). One-way ANOVA analysis followed by Sidak multiple comparison test was utilized to analyze three or more variables. To compare two groups, Student t test was used. All statistical tests were two-sided assuming parametric data, and a P of <0.05 was considered significant. The variations of data were evaluated as the means ± standard error of the mean (SEM). Five or more samples were thought to be appropriate for the sample size estimate in this study. Samples and animals were randomly chosen from the pool and treated. Treatment of samples and animals was performed unblinded.

In our previous study, we demonstrated mitochondrial activation during PD-1 blockade therapy and developed several combination therapies using mitochondria-activating chemicals. We found that activation of the PGC-1α/PPARs axis by bezafibrate improved the efficacy of PD-1 blockade in murine colon carcinoma (MC38)-bearing C57BL/6 mice (20). We confirmed similar synergistic effects of bezafibrate with anti–PD-L1 on the growth of MethA tumors, a murine skin sarcoma line on a different genetic background, BALB/c (Supplementary Fig. S1A). Bezafibrate treatment alone did not exhibit any antitumor activity, indicating that enhanced antitumor activity is mediated through activated lymphocytes, but not directly through tumor cells (ref. 20; Supplementary Fig. S1B). These results indicate that the combination of bezafibrate with PD-1 blockade is applicable to multiple tumors on different genetic backgrounds.

Because the number of effector killer T cells determines the effect against cancer cells in vivo, we first investigated the effect of bezafibrate on the effector phenotype of CD8⁺ T cells (15–17). Both the frequency and number of effector/memory CD8⁺ T cells (CD62L⁻ CD44⁺ CD8⁺ T cells, P3) in DLNs were significantly increased by treatment combining anti–PD-L1 and bezafibrate compared with treatment with anti–PD-L1 alone (Fig. 1A). In contrast, the numbers of naïve (CD62L⁺ CD44⁻ CD8⁺ T cells, P1) and central memory T cells (CD62L⁺ CD44⁺ CD8⁺ T cells, P2) were not changed by bezafibrate addition, whereas injection of anti–PD-L1 alone was accompanied by increased P1 and P2 populations (Fig. 1A). Accordingly, CD8⁺ TILs, which mainly comprise the effector/memory T-cell population (P3), were expanded, as we previously described (ref. 20; Fig. 1B).

We attribute the antitumor effect not only to the number of effector T cells but also to their function. The T-bet/Eomes balance is a factor regulating cytokine synthesis and T-cell differentiation (22). We found that bezafibrate addition increased the amount of T-bet in DLN CD8⁺ T cells (Fig. 1C). In contrast, the frequency and number of Eomes⁺ CD8⁺ T cells decreased or were unchanged after bezafibrate addition (Fig. 1C). These data demonstrate that the bezafibrate enhanced the effector function of CTLs. We further investigated T-bet and Eomes expression in each P1-P3 population. Among subsets of T cells, the P3 population has the most T-bet and the least Eomes; T-bet increased after bezafibrate treatment only in P3 (Supplementary Fig. S2A). Although the naïve population (P1) is present most abundantly in CD8⁺ T cells, T-bet expression is lowest in this population and is unaffected by bezafibrate treatment (Supplementary Fig. S2A). Bezafibrate combination therapy increased the number of cells only in the P3 population (Fig. 1A). These results suggest that most of the changes in T-bet amounts are due to the P3 population in DLN CD8⁺ T cells (Fig. 1C). CD8⁺ TILs are T-bet⁺ and Eomes⁻ (Supplementary Fig. S2B) and exclusively comprised of P3 population (Fig. 1B). These data suggest that bezafibrate treatment enhances the function of effector/memory CD8⁺ T cells. Indeed, IFNγ in both CD8⁺ T cells of DLN and CD8⁺ TILs was upregulated (Supplementary Fig. S2C; Fig. 1D). Together, we showed the bezafibrate combination therapy increases the number and enhances the function of effector/memory CD8⁺ T cells in DLNs and at the tumor site.

To understand how bezafibrate enhances T-cell–mediated antitumor immunity under PD-1 blockade conditions, we investigated mitochondrial activities in CTLs isolated from DLNs of MC38 tumor-bearing mice treated with bezafibrate and anti–PD-L1. We found that the OCR, an indicator of mitochondrial respiration including basal respiration, as well as maximal respiration and ATP turnover were all significantly higher in CD8⁺ T cells isolated from DLNs of bezafibrate- and anti–PD-L1-treated mice (Fig. 2A, left; Supplementary Fig. S3A–S3B). SRC, which is calculated by subtracting basal respiration from maximal respiration, of CD8⁺ T cells isolated from DLNs was significantly higher in the combination-therapy group than after PD-1 blockade alone (Fig. 2A, right). Because SRC has been linked to cell survival, enhancement of SRC by bezafibrate and anti–PD-L1 combination therapy suggests that CD8⁺ T cells can survive longer (23). We also determined the effect of bezafibrate treatment on the ECAR to measure glycolysis. We observed that the ECAR values in the bezafibrate and anti–PD-L1 combination-therapy group were significantly higher than those in the group treated with anti–PD-L1 alone (Fig. 2B). The higher values for both OCR and ECAR indicated that CTLs were in a metabolically higher state in the bezafibrate combination group (Fig. 2B). The ratio of OCR and ECAR was higher in the bezafibrate combination group than in the group treated with PD-1 blockade alone (Fig. 2C), indicating that bezafibrate combination therapy raised CD8⁺ T cells to a higher energy state through mechanisms that relied more on mitochondrial metabolism than on glycolysis. Similar results were observed when the mice were sacrificed at a different time point (Supplementary Fig. S3C and S3D).

As the bezafibrate and PD-1 blockade combination enhanced the OCR values compared with PD-1 blockade alone, we investigated the effect of this combination on other mitochondrial activation parameters. The effector/memory population of CD8⁺ T cells (P3) in any treatment group showed larger mitochondrial areas, higher intensity of MitoTracker DeepRed, and more ROS than either the naïve (P1) or central memory (P2) CD8⁺ T cells (Fig. 2D and E). Cellular levels of the MitoTracker DeepRed and ROS increased when cells were treated with bezafibrate combined with anti–PD-L1 (Fig. 2E). Similar results were obtained from CD8⁺ TILs (Fig. 2F). PD-1 blockade significantly reduced various mitochondrial activation parameters in the P3 population as well as in TIL compared with untreated cells (Fig. 2E and F). This reduction may reflect the change of dependency on the energy metabolic pathways from OXPHOS to glycolysis by PD-1 blockade. Indeed, monotherapy enhanced glycolysis-dependent energy production more than OXPHOS (Fig. 2B and C) and (19, 24). In total, bezafibrate and PD-1 blockade combination treatment activated mitochondria in CTLs and increased mitochondrial SRC, contributing to the enhanced survival of CTLs in this group.

Because combined bezafibrate and anti–PD-L1 treatment enhanced mitochondrial activities in CD8⁺ T cells in MC38 tumor-bearing mice, we used qPCR analysis to ask whether the combination therapy affected transcription of genes involved in mitochondrial biogenesis. We detected increased transcription of PGC-1α and transcription factor A–mitochondrial (TFAM), both of which regulate mitochondrial biogenesis. Transcription of other mitochondria-associated genes such as Ubiquinol-Cytochrome C Reductase Core Protein I (Uqcrc1), NADH:ubiquinone oxidoreductase core subunit S8 (NDUSF8), and ATP synthase F1 subunit alpha (ATP5a1) also increased in CD8⁺ T cells from DLNs of mice treated with the combined bezafibrate and PD-1 blockade (ref. 25; Fig. 3A). Because PPAR signaling also activates the FAO pathway, we assessed transcription of enzymes involved in FAO (26). As shown in Fig. 3B, the expression of carnitine palmitoyl transferase 1B (Cpt1b), acyl-CoA dehydrogenase long chain (LCAD), and acyl-CoA dehydrogenase medium chain (MCAD) was significantly increased in CD8⁺ T cells isolated from DLNs after the combination treatment compared with cells treated with PD-1 blockade alone. We observed that expression of the Cpt1a protein, another FAO enzyme, was increased significantly by bezafibrate combination treatment in CD8⁺ T cells of both DLN and TIL (Supplementary Fig. S4A and B). Thus, the bezafibrate and PD-1 blockade combination activates mitochondrial biogenesis and FAO in CD8⁺ T cells in vivo.

To investigate the effect of combination therapy on tumor-reactive CTLs, we followed our previous strategy to identify the tumor-reactive CTLs in a mouse tumor model (20). Accordingly, CellTrace-labeled CD45.1⁺ CD8⁺ T cells were transferred into CD45.2⁺ CD8^−/− mice, and their proliferation in DLNs and tumor sites was examined (Fig. 4A). As tumors grow faster in CD8^−/− mice than in wild-type mice, tumor reached to the desired volume for start of the therapy earlier (on day 5) in CD8^−/− mice than in wild-type mice (on day 7). Among the transferred CD45.1⁺ CD8⁺ T cells, we identified the proliferating cell population as tumor-reactive cells in mice bearing MC38 tumor cells (Fig. 4B). The frequency and number of proliferating CD45.1⁺ CD8⁺ T cells were significantly increased in DLNs and at tumor sites of tumor-bearing mice treated with bezafibrate and anti–PD-L1 compared with mice injected with anti–PD-L1 alone (Fig. 4B). The bezafibrate and anti–PD-L1 combination therapy enhanced mitochondrial mass, MitoTracker DeepRed, mitochondrial ROS, and cellular ROS in tumor-reactive CTLs of DLN (Supplementary Fig. S5A).

An increased number of tumor-reactive CTLs enhances antitumor activity in PD-1 blockade therapy, as most PD-1–blocked CTLs undergo terminal differentiation and apoptosis (27). The increase in the number of tumor-reactive CTLs driven by bezafibrate treatment may be caused in two ways: (1) bezafibrate inhibited terminal differentiation and apoptosis of effector T cells or (ii) bezafibrate promoted proliferation associated with the transition from naïve to effector T cells. To test the first possibility, we analyzed apoptotic effector T cells in tumor-reactive CTLs by costaining with annexin V and propidium iodide (PI). We found that combination therapy significantly reduced the percentage of apoptotic cells (annexin V⁺ PI⁺ cells) in tumor-reactive CTLs compared with therapy with anti–PD-L1 alone, indicating that bezafibrate addition enhanced survival of tumor-reactive CTLs (Fig. 4C). We also used annexin V and PI staining to analyze the non–tumor-reactive cell population (CellTrace^highCD8⁺ T cells), which includes more live cells than found among the tumor-reactive population (Supplementary Fig. S5B). These observations are consistent with previous reports that showed that upon PD-1 blockade, dysfunctional effector cells regain effector function but die by terminal differentiation (27). We found that the bezafibrate combination treatment significantly increased expression of the antiapoptotic factor, Bcl2, in tumor-reactive CTLs at tumor sites (Fig. 4D).

Next, we investigated the second possibility by determining the expression of Ki67, a marker of cell proliferation, in CD8⁺ T cells of DLNs (Fig. 4E). We found that the number of Ki67⁺ CD8⁺ T cells increased significantly after bezafibrate combination therapy (Fig. 4E). The Ki67⁺ CD8⁺ T-cell number was increased in both P3 and P2 populations by the bezafibrate combination therapy (Fig. 4E). Together, these data suggest that the bezafibrate and PD-1 blockade in combination increased the number of tumor-reactive CTLs in DLN and at the tumor site by enhancing their survival capacity and proliferation.

To analyze the mechanism by which PPAR signaling inhibits apoptosis, we investigated the effect of bezafibrate on the survival capacity of CD8⁺ T cells in vitro. Naïve CD8⁺ T cells were stimulated as shown in Fig. 5A and expanded until day 13 in the presence of bezafibrate or solvent (DMSO) control. The effect of bezafibrate on T-cell longevity was tested in an overactivation-induced cell death system. After restimulation with anti-CD3 and anti-CD28 on day 13, bezafibrate treatment significantly reduced the number of apoptotic cells (Fig. 5B). Basal OCR and ECAR were not changed by bezafibrate treatment on day 13, but SRC and the OCR/ECAR ratio were significantly increased in the bezafibrate-treated group, again indicating that bezafibrate increased the survival capacity of CTLs (Fig. 5C). We confirmed that bezafibrate significantly increased the expression of Bcl2, Birc3, and API5 genes involved in the apoptosis inhibition pathway (Fig. 5D). We further performed GeneChip analysis to identify or differentiate gene-expression signatures between solvent control and bezafibrate-treated CD8⁺ T cells on day 13 (Supplementary Fig. S6A and S6B). KEGG pathway analysis demonstrated that bezafibrate-treated cells on day 13 displayed differential changes among genes involved in various pathways such as PPAR signaling, fatty acid metabolism, AMPK signaling, cytokine–cytokine receptor interaction, chemokine signaling pathway, complement and coagulation cascades, natural killer cell–mediated toxicity, and metabolism (Supplementary Fig. S6C and S6D).

We focused on upregulated genes involved in preventing apoptosis and analyzed the protein interactome using the STRING database. We found that Bcl2 may be stabilized by interacting with Cpt1, which is also induced by PPAR signaling (ref. 28; Fig. 5E). PPAR regulates cyclic AMP response element binding (CREB), which enhances the expression of PGC-1α in a feed-forward way (refs. 29–32; Fig. 5E). Indeed, both protein and mRNA of Cpt1a and CREB1 were upregulated in bezafibrate-treated CD8⁺ T cells on day 13 (Fig. 5F and G).

Our in vivo data suggest that enhanced proliferation of CTLs driven by the bezafibrate combination treatment contributes to increasing the number of effector killer T cells (Fig. 4E). To validate the effect of bezafibrate on T-cell proliferation, we stimulated naïve CD8⁺ T cells with anti-CD3 and anti-CD28 in the presence of bezafibrate or solvent (DMSO) control. As shown in Fig. 6A, in vitro–stimulated naïve CD8⁺ T cells in the presence of bezafibrate incorporated more ³H-thymidine than those without bezafibrate. The enhancement of proliferation by bezafibrate treatment was confirmed by the dye dilution experiment (Supplementary Fig. S7A). Because bezafibrate treatment improved mitochondrial activities in vivo, we also investigated whether enhanced CTL proliferation in vitro by bezafibrate is accompanied by mitochondrial activation. As shown in Fig. 6B, both OCR and ECAR were significantly increased, indicating that CTL reached a higher energy state with bezafibrate treatment. Although ATP turnover and glycolytic capacity were upregulated, SRC was decreased in the bezafibrate-treated group (Fig. 6C). As bezafibrate treatment enhanced OCR values, we investigated its effect on other mitochondrial activation parameters. We found that bezafibrate-treated cells possess larger mitochondrial areas, higher intensity of MitoTrackerDeepRed, and more ROS than DMSO-treated cells, proving that mitochondria are activated during T-cell priming (Supplementary Fig. S7B). These data suggest that the effect of PPAR signaling on T-cell priming (day 2) is more associated with proliferation (anabolic pathway) than with longevity (catabolic pathway). Thus, PPAR signaling enhances proliferation during the early (priming) phase and inhibits apoptosis during the effector phase of the T cells.

PD-1 blockade–based cancer immunotherapy has changed cancer treatment because it reaches more cancer targets and shows longer responses with fewer side effects than other cancer therapies (11, 33, 34). However, many patients do not respond to this therapy (12, 35). To enhance its efficacy and increase the range of patients who respond, combinatorial therapies involving PD-1 blockade have been designed and evaluated in clinical trials. One main cause of poor response is the absence or insufficiency of functional effector T cells at tumor sites, likely because of terminal differentiation and apoptosis mediated by PD-1 blockade (17, 27).

We found that bezafibrate treatment enhanced proliferation of CTLs and inhibited their apoptosis. Proliferation of primed CTLs generated more effector T cells. PPAR activation by bezafibrate enhanced both OCR and ECAR, explaining the upregulation of ATP production and glycolysis, both of which are necessary for induction of effector CTLs. These data are supported by studies showing that activation of both the mitochondria and PPAR/FAO pathways is necessary for proper activation of CTLs in the early stimulation phase (36, 37). The positive feed-forward loop in which PPAR signaling activates the mitochondria and enhances proliferation likely involves mTOR activation and promotes cellular proliferation through protein synthesis (38). mTOR signaling enhances cellular proliferation and AMPK activation improves cell survival (38). Therefore, the antiapoptotic and proproliferation effects of bezafibrate are indicated by our results showing that total CD8⁺ T cells from DLN isolated from mice treated with anti–PD-L1 and a mitochondria-activating chemical showed upregulation of both AMPK and mTOR activity, even though mTOR and AMPK compete with each other with in a single cell (20, 38).

Chronological dissection of the effects of PD-1 blockade and bezafibrate is schematically represented (Supplementary Fig. S8). Upon activation, naïve T cells proliferate and acquire effector function, a process that requires energy generation by glycolysis (Supplementary Fig. S8A). To avoid terminal differentiation and cell death induced by overactivation, T cells express immune checkpoints such as PD-1 and shift their energy metabolic pathway to FAO/OXPHOS (19, 27, 39). The PD-1 signal confers T-cell longevity by compromising proliferation and killing activity, producing dysfunctional cells that may allow tumor cells to escape immune surveillance. In contrast, PD-1 blockade recovers T-cell effector function and reactivates effective antitumor immunity, which in turn promotes terminal differentiation and apoptosis accompanied by increased glycolysis and decreased FAO (Supplementary Fig. S8B; ref. 27). Thus, PD-1 blockade alone may reduce the number of functional effector T cells, which explain the incomplete responsiveness to PD-1 blockade. Introduction of pan-PPAR signaling in T cells by bezafibrate led to enhanced FAO, even in the PD-1 blocked state, and prolonged survival of CD8⁺ T cells (Supplementary Fig. S8C). PPAR activation also boosts the proliferation of CD8⁺ T cells, which increases the number of functional effector T cells (Supplementary Fig. S8C).

Our transcriptome analysis showed that PPAR pathway activation rescues PD-1 blockade–induced apoptosis in T cells by upregulating the antiapoptotic proteins Bcl2, Birc3, and API5. Gene-expression data revealed that the FAO pathway is promoted during PPAR stimulation, through increased expression of Cpt1, an enzyme key to fatty acid metabolism in mitochondria. Cpt1 is stabilized by interaction with Bcl2, and inhibition of Cpt1 enhances cell death (28). Our results suggest that PPAR inhibits apoptosis of CTLs by upregulating Bcl2, Cpt1, and cross-talk between PPAR and Bcl2 (29). The effect of the PPAR pathway on T-cell longevity is indicated by the improvement in mitochondrial SRC defined as the potential ATP levels in OXPHOS that can be used when energy demand surges. Thus, SRC enhancement in CTLs by bezafibrate suggests that CTLs can survive longer and increase their number in a memory-like pool (23). We observed more depolarized mitochondria (∼55.2%) in the TIL under anti–PD-L1 plus bezafibrate therapy (Fig. 2F). This mitochondrial population characterized by low MitoTracker Deep Red and high MitoTracker Green supports longevity of T cells because cells with depolarized mitochondria would undergo mitophagy to eliminate damaged mitochondria and enhance cellular survival (40–42). Thus, bezafibrate treatment, which depolarizes mitochondria, would extend the life span of cells, as we observed.

PPAR signaling induces expression of a range of target genes (43). Our microarray data from CD8⁺ T cells showed that many of the genes identified belong to energy metabolism-related pathways, e.g., Cpt1, LCAD, MCAD, hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase (trifunctional protein), β-subunit (Hadhb), malic enzyme 1 (Me1), Perillipin 2 (Plin2), and stearoyl-CoA desaturase-1 (Scd1). Our data also revealed additional genes induced by PPAR signaling. For instance, genes in pathways related to immune function were induced, including cytokine–cytokine receptor interactions, the chemokine signaling pathway, complement and coagulation cascades, and natural killer cell–mediated toxicity. Therefore, our data suggest additional mechanisms for regulating the killing activity of CD8⁺ T cells by PPARs.

We found that each subset of the T-cell population (P1/P2/P3) exhibited different mitochondrial activities. However, the P3 population likely has a large contribution to the changes observed in the flux assay of total CD8⁺ T cells for the following reasons. We demonstrated that, after bezafibrate treatment, mitochondrial activity was increased in the P3 population but not in others. Although naïve and memory populations (P1 and P2) are represented in T cells, mitochondrial activities do not change in these populations after the bezafibrate combination treatment. Further, the bezafibrate combination therapy increased the number of cells only in the P3 population. Therefore, it is likely that upregulation in the flux assay in the total CD8⁺ T cells was mainly due to metabolic changes of the P3 population.

We previously showed that metabolic genes in tumor cells were unperturbed by treatment with Luperox or uncoupler alone (20). We expected that bezafibrate would not have large effect on tumor metabolism, as the dose used in this study was less than one-tenth the dose showing cytotoxicity toward tumor cells (44). If bezafibrate exerted its antiapoptotic effect or proproliferation effect on tumor cells, tumor growth would be faster compared with the untreated group. However, we observed no enhanced tumor growth in vivo when bezafibrate was administered alone, suggesting no direct effect of bezafibrate alone on tumor cells.

In summary, we showed that PPAR signaling reprograms CTL energy metabolism and overcomes the reduction in functional effector T-cell number associated with PD-1 blockade by decreasing apoptosis or increasing proliferation. Apoptosis was inhibited by upregulation of the antiapoptotic factor Bcl2 in CTLs. Enhanced proliferation can be explained by mitochondrial activation and upregulation of the mTOR pathway. Our results will enable development of improved combination therapies and allow for the treatment of patients who are less responsive to therapy blocking the PD-1/PD-L1 interaction alone.

No potential conflicts of interest were disclosed.

Conception and design: P.S. Chowdhury, K. Chamoto, T. Honjo

Development of methodology: P.S. Chowdhury, K. Chamoto

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.S. Chowdhury, K. Chamoto, A. Kumar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.S. Chowdhury, K. Chamoto, A. Kumar

Writing, review, and/or revision of the manuscript: P.S. Chowdhury, K. Chamoto, T. Honjo

Study supervision: P.S. Chowdhury, A. Kumar, T. Honjo

This work was supported by AMED under grant numbers JP17cm0106302, JP17gm0710012 (T. Honjo), and JP171k1403006 (K. Chamoto); Tang Prize Foundation (T. Honjo); JSPS KAKENHI grant numbers JP16H06149, 17K19593 (K. Chamoto), and 17F17119 (P.S. Chowdhury); the Cell Science Foundation (K. Chamoto); and Takeda Science Foundation (A. Kumar).

We thank M. Al-Habsi, M. Akrami, T. Oura, R. Hatae, Y. Nakajima, R.M. Menzes, and K. Yurimoto for assistance in sample preparation; Y. Kitawaki for helping with the western blotting; N.A. Begum for helping with the GeneChip data analysis; and S. Fagarasan for many stimulating discussions. We thank Bristol-Myers Squibb for the collaboration.

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.