Activation of Peroxisome Proliferator-Activated Receptors α and δ Synergizes with Inflammatory Signals to Enhance Adoptive Cell Therapy

Cellular therapies, including adoptive cell therapy (ACT) with autologous tumor infiltrating lymphocytes (TIL), have demonstrated promise in the treatment of advanced malignancy. CD8⁺ memory T (T_mem) cells, compared with fully differentiated effector (T_eff) cells, are superior mediators of ACT due to an enhanced ability to expand and persist in vivo. Multiple methodologies have been explored to allow the ex vivo culture and expansion of CD8⁺ T cells with memory cell characteristics. Altering cellular metabolism has emerged as a promising strategy for the ex vivo generation of memory-like T cells as metabolic programming underlies many of the phenotypic differences between CD8⁺ T_mem and T_eff cells (1).

Aerobic glycolysis, involving an increased flux of glucose to pyruvate down the glycolytic pathway and the subsequent conversion of pyruvate to lactate, is the hallmark of the metabolic program engaged by T_eff cells (2). This highly glycolytic phenotype of T_eff cells is required for both rapid cellular division and maximal cytokine expression (3, 4). Conversely, CD8⁺ T_mem cells adopt an oxidative metabolic program that relies upon mitochondrial fatty acid oxidation (FAO; refs. 5, 6). CD8⁺ memory cells have been demonstrated to convert glucose to fats, particularly triglycerides (TAG), in order to fuel FAO (7, 8). The utilization of FAO by CD8⁺ T_mem cells supports their in vivo longevity (6–8). These data established a dichotomous model in which short-lived effectors rely upon glycolysis to engage their effector function, whereas CD8⁺ T_mem cells depend upon fatty acid metabolism to fuel oxidative phosphorylation (OXPHOS) and sustain their persistence in vivo.

Recent findings, however, have altered this paradigm in the context of cancer immunotherapy. T_eff cells were discovered to require mitochondrial metabolism and FAO for optimal effector function activity in the tumor microenvironment (9, 10). T_eff cells genetically or pharmacologically altered to have increased usage of OXPHOS displayed enhanced antitumor immunity (9–11). Thus, enhanced mitochondrial metabolism is not mutually exclusive with an effector phenotype in CD8⁺ T cells. Furthermore, agents that boost mitochondrial metabolism and FAO in effector cells represent a potential avenue to improve the efficacy of ACT.

The peroxisome proliferator-activated receptors (PPAR) are members of the nuclear receptor superfamily. Three isoforms, PPARα, PPARδ (also called PPARβ), and PPARγ have been identified. The PPARs have been implicated in the regulation of multiple metabolic pathways, particularly lipid metabolism. Activation of PPARα and PPARδ increases FAO in multiple tissues, whereas activation of PPARγ increases the synthesis and storage of fatty acids. In CD4⁺ T cells, PPARγ has been linked with fatty acid uptake and metabolism (12) as well as the differentiation of adipose-resident T-regulatory cells (13). In CD8⁺ T cells, PPARα has been demonstrated to have role in promoting fatty acid catabolism (10). The role of PPARδ in the regulation of T-cell metabolism and function, particularly in CD8⁺ T cells, has not been well defined. GW50156 (GW), a PPARδ-specific agonist, has recently been demonstrated to enhance FAO in hematopoietic stem cells and thereby promote their maintenance in the bone marrow (14). Therefore, we examined whether activation of PPARδ by GW in activated CD8⁺ T cells would shift their metabolism to increase FAO and mitochondrial metabolism and enhance their efficacy for ACT.

C57BL/6 mice and C57BL/6 CD45.2 mice were purchased from Taconic. PPARδ^Flox/Flox, PPARα^−/−, and CD4-Cre mice were purchased from the Jackson Laboratory. Generation of P14 mice, which express a transgenic TCR specific for the H2-D^b gp33 peptide of the lymphocytic choriomeningitis virus (LCMV) were made in the lab of Dr. Hanspeter Pircher. All mice were maintained at the Ontario Cancer Institute animal facility according to institutional guidelines and with approval of the Ontario Cancer Institute Animal Ethics Committee. The B16F10-gp33 (B16-gp33) tumor line was obtained from Dr. Rolf M. Zinkernagel (University of Zurich) and was not authenticated since acquisition and was not tested for Mycoplasma since being acquired.

All experiments were conducted on CD8⁺ T cells, which were magnetically purified (Miltenyi Biotec) from spleens and lymph nodes of mice of the indicated genotypes. T cells were activated either by cocultured with mature, gp33-pulsed bone marrow dendritic cells (BMDC) for 3 days in IMDM supplemented with 10% FCS, l-glutamine, β-mercaptoethanol, penicillin, and streptomycin or by stimulation with platebound anti-CD3 and anti-CD28 antibodies (1 μg/mL; eBioscience) in the presence of IL12 (5 ng/mL; Biolegend) and/or IL2 (20 IU/mL; Biolegend) for 72 hours as indicated. After 72 hours of culture, the cultures were >98% pure CD8⁺ T cells using either method of T-cell activation (data no shown). The BMDCs were generated from the pooled bone marrow collected from mouse femurs and tibiae. Briefly, bone marrow was cultured in complete RPMI containing 40 ng/mL GM-CSF (Peprotech) with medium changes on days 3, 6, and 8 of culture. On day 8–10, nonadherent dendritic cells (DC) were collected and cultured with 100 ng/mL of lipopolysaccharide (Invitrogen) for 16 hours to induce activation. Activated DCs were then pulsed with gp33 peptide for 2 to 3 hours before being co-cultured with P14 CD8⁺ T cells. GW501516 (Santa Cruz Biotechnology) was dissolved in DMSO vehicle. Antibodies used for Western blotting were CPT1a (Abcam) and actin (Sigma).

For real-time PCR, RNA was extracted from cells using the RNeasy Plus Mini Kit (Qiagen) according to manufacturer's instructions. RNA was reverse transcribed into cDNA using qScript cDNA Fast Mix (QuantaBio) according to manufacturer's protocol. Real-time PCR was performed on cDNA using PerfeCTa SYBR Green FastMix (QuantaBio) on the Applied Biosystems 7900HT using the two-step fast cycling protocol as per manufacturer's recommendations. Gene expression for all experiments were normalized to the house keeping gene β-actin and expressed as relative quantity (RQ) compared with vehicle-treated controls. For IFNγ production, cells were restimulated with cell stimulation cocktail (eBioscience) and stained using the Intracellular Staining Kit (BD Biosciences) as per the manufacturer's instructions. For intracellular flow cytometry analysis of T-bet and PPAR protein expression, primary antibodies conjugated to PE were obtained from eBioscience (T-bet) and Santa Cruz Biotechnology (PPARs). For the assessment of lipid uptake, cells were stained with BODIPY FL C₁₆ (Molecular Probes/ThermoFisher Scientific) as per the manufacturer's instructions. Data were then acquired on a FACS CANTO II (BD Biosciences) and analyzed using Flowjo software.

Metabolomic and statistical analyses were conducted at Metabolon. Briefly, cell pellets (n = 5 biological replicates per group) were disrupted using a GenoGrinder (675 strokes/minute, 2 minutes) and subjected to methanol extraction. Extracts were split into four aliquots and processed for analysis by ultra-high performance liquid chromatography/mass spectrometry (UHPLC/MS) in the positive (two methods), negative, or polar ion mode. Metabolites were identified by automated comparison of ion features to a reference library of chemical standards followed by visual inspection for quality control. For statistical analyses and data display, any missing values were assumed to be below the limits of detection; these values were imputed with the compound minimum (minimum value imputation). To determine statistical significance, Welsh's two-factor t test was performed on protein-normalized data in ArrayStudio (Omicsoft) or “R” to compare data between experimental groups; P < 0.05 was considered significant. An estimate of the false discovery rate (Q-value) was calculated to take into account the multiple comparisons that normally occur in metabolomic-based studies, with Q < 0.05 used as an indication of high confidence in a result. ArrayStudio was also used to generate high-level overview display items (PCA, heatmaps). Pathway enrichment scores for GW-treated cells were calculated relative to vehicle treated control effector CD8⁺ T cells using the formula: [no. of significant metabolites in pathway (k)/total no. of detected metabolites in pathway (m)]/[total no. of significant metabolites (n)/total no. of detected metabolites (N)] (k/m)/(n/N). A P value cut off of P < 0.01 was used. Only pathways in which at least three metabolites were examined were included.

Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measure on an XFe96 extracellular analyzer (Agilent Technologies) as per standard protocols (15). Briefly, T cells were plated in Seahorse XF assay media (unbuffered DMEM, pH 7.4, 25 mmol/L glucose, 2 mmol/L l-glutamine, and 1 mmol/L sodium pyruvate) at a density of 4 × 10⁵ cells per wells on Seahorse cell plates coated with poly-d-lysine (50 μg/mL). Cells were then incubated in a CO₂-free incubator at 37°C for 30 minutes before OCAR and ECAR were assessed using the extracellular analyzer. Five technical replicates were acquired for each reading. To assess the contribution of FAO to the OCR, etomoxir (4 μmol/L; Sigma) was injected during the Seahorse assay. For the glycolytic rate assay, XF base media without phenol red (Agilent Technologies) with 10 mmol/L glucose, 2 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, and 5 mmol/L HEPES was used as the base media. Rotenone with antimycin A (both 0.5 μmol/L) and 2-deoxy-d-glucose (2-DG; Agilent Technologies) were injected in sequence. Seahorse data were analyzed as per the manufacturer's instruction using Wave software with the appropriate assay template (Agilent Technologies).

For cell transfer experiments, equal numbers of activated P14 Thy1.2⁺ CD45.2⁺ T cells and P14 Thy1.1⁺ CD45.2⁺ T cells were mixed at 1:1 ratio ex vivo and then 1 × 10⁶ cells were injected intravenously into CD45.1⁺ hosts. Spleens were harvested at the times indicated and persistence of transferred cells was assessed by flow cytometry. For ACT, female C57LB/6 mice were inoculated subcutaneously with 4 × 10⁵ B16F10-gp33 melanoma cells. Mice then received 1 × 10⁶ activated CD8⁺ P14 T cells intravenously 10 days after tumor inoculation. Tumor size was assessed every few days using calipers until mice reached experimental endpoint (tumor volume ≥ 1,500 mm³ or severe ulceration/necrosis). Tumor volume was calculated using the formula volume = 0.5 × (length × width²). For assessment of cytokine production of TILs, tumors were harvested 4 days after the transfer of T cells and mechanically separated. IFNγ production in the Thy1.1⁺ CD8⁺ population of TILS was then assessed by flow cytometry as described above.

First, a dose titration was performed to assess the effect of increasing GW concentrations on the expression of CPT1a in P14 CD8⁺ T cells activated by gp33 peptide-pulsed, mature DCs (Supplementary Fig. S1A). We found GW treatment increased CPT1a expression at all doses tested. We selected the 5 μmol/L dose for further experiments, as it appeared to induce maximal CPT1a expression. We also confirmed that this dose also increased CPT1a expression in purified CD8⁺ T cells activated with antibodies to CD3 and CD28 (Supplementary Fig. S1B). These data indicated that GW acts directly upon the CD8⁺ T cells to induce CPT1a expression.

Despite being designed as a PPARδ-specific ligand, GW has been suggested to activate PPARα directly when used in micromolar concentrations (16). Alternatively, activation of PPARα downstream of GW-mediated PPARδ activation has also been suggested (17). We observed that CD8⁺ T cells expressed all three PPAR isoforms and that treatment GW did not alter the expression of these proteins (Supplementary Fig. S1C). Therefore, to determine which PPAR was required for increasing the expression of CPT1a in CD8⁺ T cells, we bred the P14 transgenic TCR into various knockout and floxed mice to generate P14 PPARα^−/− mice, P14 PPARδ^Fl/FL CD4-CRE (PPARδ^ΔT), and P14 double knockout mice (DKO, PPARα^−/− PPARδ^Fl/FL CD4-CRE). We observed that P14 CD8⁺ T cells lacking PPARδ, but not PPARα, had reduced expression of CPT1a in response to GW treatment (Fig. 1A). The induction of CPT1a upon GW treatment was not observed in the DKO CD8⁺ P14 T cells (Fig. 1A). These data demonstrate that GW engages both PPARδ and PPARα, but primarily PPARδ, in CD8⁺ T lymphocytes to increase the expression of CPT1a.

CPT1a protein expression is decreased upon T-cell activation compared to naïve T cells (Supplementary Fig. S1B; ref. 18). To exclude the possibility that the augmented CPT1a expression induced by GW was due to impaired T-cell activation, we measured the expression of activation markers on naïve and activated cells. Regardless of treatment with GW, the P14 T cells increased CD44 expression and decreased 62L expression to similar levels after 72 hours of activation. These data demonstrated that treatment with GW did not impair the cells from acquiring an activated phenotype despite the high expression of CPT1a (Fig. 1A and B).

In addition to CPT1a, multiple other proteins associated with lipid metabolism have been implicated in the in vivo longevity of T_mem cells. These include the glycerol channel aquaporin 9 (AQP9) and DAG O-acyltransferase 1 (DGAT1; ref. 8), the enzyme that catalyzes the ultimate step in TAG synthesis. The liposomal acid lipase (19), encoded by the LIPA gene, has also been found to be required for the persistence of CD8⁺ T_mem cells (7). We assessed the expression of the genes encoding for these proteins, as well as CPT1a, in P14 cells activated in the presence or absence of GW (Fig. 1C). Compared with control cells, we found that GW significantly induced the expression of transcripts of CPT1a, AQP9, and DGAT1 but not LIPA, suggesting that GW induced a subset of the metabolic genes associated with T_mem differentiation and persistence but did not completely recapitulate the metabolic program of memory T cells. Moreover, consistent with studies in other cell types (20), we found that GW treatment also increased the RNA expression of other known PPARδ target genes associated with FAO including pyruvate dehydrogenase kinase 4 (PDK4) and uncoupling protein 2 (UCP2; Supplementary Fig. S1D). Therefore, we found that GW treatment induced the transcription of multiple genes involved in FAO and lipid metabolism in activated CD8⁺ T cells.

To further characterize the metabolic phenotype of the GW-treated cells, we utilized the Seahorse extracellular flux analyzer. This allows for simultaneous, real-time assessment of OCR, a marker of OXPHOS, as well as ECAR, a marker of glycolysis and lactate production (15). Treatment with GW increased the basal OCR whereas the ECAR was comparable to control cells (Fig. 1D). This resulted in a higher OCR to ECAR ratio for GW-treated cells, indicating an increased reliance on OXPHOS.

Previous studies have shown that carbon dioxide (C0₂) produced in the mitochondria can result in acidification of media and thus contribute to the measured ECAR (21). Accordingly, to further characterize the relative utilization of glycolysis and OXPHOS in GW-treated cells, we performed the Glycolytic Rate Assay. This assay allows the calculation of the contribution of mitochondrial acidification to the measured ECAR. Consistent with the higher basal OCR of the GW-treated cells (Fig. 1D), we found that GW-treated cells had a reduced proton efflux rate (PER) due to glycolytic activity (glycoPER) and an increased rate of acidification due to mitochondrial metabolism (mitoOCR/glycoPER; Supplementary Figs. S2A and S2B). Interestingly, we did not detect an increase in cellular reactive oxygen species (ROS) despite this enhanced mitochondrial activity (Supplementary Fig. S2C). Collectively, these data indicate that treatment with GW shifts the bioenergetics profile of active CD8⁺ away from aerobic glycolysis and towards oxidative metabolism.

To elucidate the metabolic changes underpinning the oxidative phenotype of GW-treated cells, we utilized an ultrahigh performance liquid chromatography/mass spectrometry (UHPLC/MS) based approach to assess the quantity of over 400 intracellular metabolites. Using both principle component analysis and unsupervised hierarchical cluster analysis, it was evident that GW induced a unique metabolic signature in activated CD8⁺ T cells (Fig. 2A). Pathway enrichment analysis further demonstrated significant alterations in multiple metabolic pathways in GW-treated cells, with pathways associated with glycerol metabolism, fatty acid metabolism, oxidative phosphorylation, and nucleotide metabolism being the most enriched (Fig. 2B). Among the metabolites whose abundance were most affected by GW treatment were various acylcarnitines, obligate intermediates of FAO, as well as multiple monoacylglyerides and diacylglycerides (Fig. 2C). Also consistent with increased FAO in the GW-treated cells was the observed effect on oxygen consumption with the addition of the CPT1a inhibitor etomoxir at a dose at which etomoxir maintains CPT1a-specific effects (22). Injection of 4 μmol/L etomoxir during analysis on the Seahorse demonstrated a significant decline in OCR in the GW-treated cells that was not observed in the controls (Fig. 2D). The observed CPT1a-dependent OCR in the GW-treated cells is consistent with increased utilization of fatty acid as a substrate for oxidative metabolism. Additionally, an increased ability for the uptake of fatty acids was also observed in the GW-treated cells as assessed by the uptake of Bodipy-FL C16 (Fig. 2E). These findings, combined with the increased expression of CPT1a (Fig. 1A), indicated that treatment with GW shifted the cellular metabolism of activated CD8⁺ T cells to enhance the uptake and oxidation of fatty acids.

Because aerobic glycolysis enhances the production of IFNγ at a posttranscriptional level (4), we investigated whether the change in cellular metabolism induced by GW had an impact on IFNγ production. Surprisingly, GW treatment increased the percentage of cells that were IFNγ⁺ as well as the mean fluorescent intensity of IFNγ staining (Fig. 3A) without an observed increase in glycolytic metabolism, as assessed by the measurement of ECAR (Fig. 1D). Given these findings, we examined whether GW was affecting IFNγ production at the transcriptional level and found increased IFNγ mRNA in the cells treated with GW (Supplementary Fig. S3A). In CD4⁺ T cells, PPARα negatively regulates the expression of T-bet and ligation relieves this repression (23). Given that treatment with GW appeared to activate both PPARα and PPARδ (Fig. 1A), we measured expression of T-bet and found GW increased T-bet protein expression (Fig. 3B). To exclude the possibility that GW was acting via the DC in our culture system to enhance T-bet and IFNγ expression, we activated T cells with antibodies and used IL12 as the proinflammatory signal in lieu of DCs. GW alone modestly boosted T-bet expression (Supplementary Fig. S3B). In the presence of IL12, however, GW augmented T-bet expression to a level higher than treatment with IL12 alone. Collectively, these data indicated that GW synergizes with inflammatory signals to drive the expression of T-bet and subsequently IFNγ at the transcriptional level in activated CD8⁺ T cells.

Elevated expression of T-bet has been associated with terminal differentiation of CD8⁺ T cells and diminished persistence (24). We next examined the ability of GW-treated cells to persist in vivo. Congenically marked P14 cells were activated in the presence or absence of GW and then mixed in a 1:1 ratio. The mixture was then adoptively transferred to mice and the survival of the cells was assessed 14 and 40 days after infusion. At both time points, an increased survival of the GW-treated P14 Thy1.1⁺ cells was observed relative to the P14 Thy1.1⁻ control cells (Fig. 3C and D). We then tested if this combination of increased in vivo persistence and increased T-bet expression translated to superior activity of GW-treated P14 cells in the B16-GP33 model of ACT. Mice treated with the cells activated in the presence of GW displayed delayed tumor growth (Fig. 4A) as well as a statistically significant improvement in overall survival (Fig. 4B). When we examined the IFNγ production of the transferred T cells that were present in the tumor microenvironment, we found a higher percentage of cells expressing IFNγ among the cells that had been activated in the presence of GW ex vivo (Fig. 4C and D). These data indicate that the GW-treated cells transferred for ACT retain their ability to produce more IFNγ in vivo. Therefore, the combination of metabolic reprogramming to FAO and increased T-bet and IFNγ expression resulted in superior antitumor function in GW-treated cells.

This study demonstrates that GW, and potentially other combined PPARα/δ dual agonists, are attractive candidates for further investigation and rapid translation into clinical trials of ACT. GW displayed a novel synergy with inflammatory signals to promote both in vivo persistence (Fig. 3C and D) as well as effector function via the modification of both cellular metabolism and T-bet expression respectively. Previously, it was found that T cells engineered to express IL12 did not result in improved tumor control in patients with melanoma, due to lack of persistence of the transferred T cells (25). Combining GW, or genetic approaches to activate PPARα and PPARδ, may overcome this survival problem without diminishing the effector function of the transferred T cells, thereby resulting in a survival benefit for patients.

Russell G. Jones reports receiving a commercial research grant from Agios Pharmaceuticals, has ownership interest (including stock, patents, etc.) in Immunomet Therapeutics, and is a consultant/advisory board member of Agios Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.D. Saibil, A.R. Elford, L.T. Nguyen

Development of methodology: S.D. Saibil, C. Robert-Tissot

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.D. Saibil, M. St. Paul, R.C. Laister, C.R. Garcia-Batres, K. Israni-Winger, A.R. Elford, N. Grimshaw, D.G. Roy, R.G. Jones

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.D. Saibil, M. St. Paul, R.C. Laister, C.R. Garcia-Batres, A.R. Elford, R.G. Jones, P.S. Ohashi

Writing, review, and/or revision of the manuscript: S.D. Saibil, M. St. Paul, P.S. Ohashi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.D. Saibil, A.R. Elford, C. Robert-Tissot, P.S. Ohashi

Study supervision: S.D. Saibil, P.S. Ohashi

This work was supported by a Canadian Institute for Health Research Grant (Foundation Scheme: FDN 143220) to P.S. Ohashi and a Natural Science and Engineering Research Council Scholarship to M. St. Paul. P.S. Ohashi holds a Canada Research Chair in Tumor Immunity and Immunotherapy (950-230660). The authors would like to thank Leanne Wybenga-Groot and Justin Barham of the SPARC Biocentre, Toronto, Canada, for assistance with Seahorse Assays.