Reprogramming lipid metabolism prevents effector T cell senescence and enhances tumor immunotherapy

Abstract

The functional state of T cells is a key determinant for effective antitumor immunity and immunotherapy. Cellular metabolism, including lipid metabolism, controls T cell differentiation, survival, and effector functions. Here, we report that development of T cell senescence driven by both malignant tumor cells and regulatory T cells is a general feature in cancers. Senescent T cells have active glucose metabolism but exhibit unbalanced lipid metabolism. This unbalanced lipid metabolism results in changes of expression of lipid metabolic enzymes, which, in turn, alters lipid species and accumulation of lipid droplets in T cells. Tumor cells and T_reg cells drove elevated expression of group IVA phospholipase A₂, which, in turn, was responsible for the altered lipid metabolism and senescence induction observed in T cells. Mitogen-activated protein kinase signaling and signal transducer and activator of transcription signaling coordinately control lipid metabolism and group IVA phospholipase A₂ activity in responder T cells during T cell senescence. Inhibition of group IVA phospholipase A₂ reprogrammed effector T cell lipid metabolism, prevented T cell senescence in vitro, and enhanced antitumor immunity and immunotherapy efficacy in mouse models of melanoma and breast cancer in vivo. Together, these findings identify mechanistic links between T cell senescence and regulation of lipid metabolism in the tumor microenvironment and provide a new target for tumor immunotherapy.

INTRODUCTION

Immunotherapies, including immune checkpoint blockade therapy and adoptive T cell therapy, have resulted in promising outcomes for patients with certain types of cancer, but the overall effective rates are still varied across the tumor types (1–3). One of the key determinants for the therapeutic efficacy and immune responses is functional state of the transferred or preexisting T cells in the suppressive tumor microenvironment (4, 5). It is now well recognized that T cells are exhausted, as shown by expression of inhibitory receptors and loss of effector functions and proliferation in the tumor microenvironment in patients (4, 6, 7). However, current checkpoint blockade therapies using antibodies to target programmed death-1 (PD-1) and PD ligand 1 or cytotoxic T lymphocyte–associated protein 4 (CTLA-4) only have limited success rates, further suggesting that there are other mechanisms or checkpoints involved in T cell dysfunction mediated by tumors (1, 2, 5). A better understanding of the distinct mechanisms responsible for T cell dysfunctional states within the suppressive tumor microenvironment should provide avenues for tumor immunotherapy.

Cellular metabolism directs T cell survival, proliferation, and effector functions (8, 9). Aerobic glycolysis is the main metabolic pathway for activated T cells and is specifically required for T cell effector function upon activation (8, 10–12). Furthermore, different T cell subsets have different metabolic profiles (9). Activated CD4⁺ T cells increase both glycolysis and fatty acid (FA) metabolism, whereas CD8⁺ T cells dominantly shift metabolism to glycolysis to rapidly produce adenosine triphosphate (13, 14). Increasing evidence suggests that malignant tumors rewrite T cell metabolic programs and functions as a means to sustain a suppressive tumor microenvironment (4, 15, 16). Tumor cells and tumor-infiltrating T cells (TILs) directly compete for key nutrients, such as glucose and glutamine, within the suppressive tumor microenvironment. This competition impairs T cell metabolism and effector functions and drives tumor progression (12, 17–20). Furthermore, tumor-derived regulatory T (T_reg) cells also drive glucose consumption and trigger cell senescence and DNA damage in responder T cells during their interaction (21, 22). In addition, tumor cells highly expressing checkpoint molecules such as PD-1 and CTLA-4 alter T cell metabolic program by inhibiting glucose transporter 1 (Glut1) and glycolysis, as well as by enhancing lipid oxidation (23, 24). The tumor microenvironment also decreases peroxisome proliferator–activated receptor γ coactivator 1α and subsequent mitochondrial biogenesis and function in TILs (25). Although these more recent studies have improved our understanding of how reprogramming of glucose metabolism affects T cells within the tumor microenvironment, the active metabolic pathways and regulations in lipid metabolism in TILs modulated by malignant tumors are still unclear. Lipids not only are structural molecules but also play important roles in regulating fundamental cellular processes, including cell death and division (26, 27). Recent studies have shown that increased cholesterol in the tumor microenvironment induces exhaustion in TILs (28). Furthermore, T_reg cells also modulate lipid metabolism in tumor-associated macrophages to promote tumor suppression (29). In addition, tumor-derived dendritic cells (DCs) with tolerogenic functions exhibit inhibited glycolysis but enhanced lipid droplet (LD) accumulation, leading to impaired antigen-presenting functions and T cell priming (30, 31). However, the molecular processes responsible for alterations in lipid metabolism in tumor-specific T cells are still under investigation. Comprehensively exploring the lipid metabolic profiles of T cells and the causative molecular interactions within the suppressive tumor microenvironment will facilitate the development of strategies for cancer therapy via metabolic reprogramming of cell fate and functions.

Immunosenescence is another state of T cell dysfunction within the tumor microenvironment, which is a key strategy used by tumors to evade immune surveillance (22, 32–36). Senescent T cells exhibit dysfunctional antitumor activity but are not functionally exhausted nor anergic (21, 35). Accumulation of senescent CD8⁺ T cells has been found in human TILs isolated from various types of cancers, including lung, colorectal, ovarian, and breast cancers (37–39). Furthermore, multiple types of tumor cells can directly induce T cell senescence via the tumor-derived metabolite cyclic adenosine monophosphate (22, 34). In addition, tumor-derived T_reg cells can induce cell senescence in effector T cells (21, 32, 33). These studies indicate mechanisms responsible for the accumulation of senescent T cells in patients with cancer. Although the metabolic signature of senescent T cells in the tumor microenvironment is unknown, studies from other types of senescent cells have demonstrated that senescent cells exhibit permanent cell cycle arrest but active metabolism with elevated glycolysis (40–42). Thus, the mechanisms by which the tumor microenvironment reprograms metabolism in TILs during senescence should be investigated to develop effective immunotherapies.

To further explore the mechanisms and functional states of TILs in the tumor microenvironment, we demonstrated that development of T cell senescence induced by malignant tumor cells or T_reg cells is a general feature in the suppressive tumor microenvironment. We found that unbalanced lipid metabolism is involved in T cell senescence and that elevated group IVA phospholipase A₂ (cPLA₂α), which is molecularly controlled by mitogen-activated protein kinase (MAPK) and signal transducer and activator of transcription (STAT) signaling, is mechanistically involved in altered lipid metabolism. Our in vivo studies further demonstrate that inhibition of cPLA₂α can prevent effector T cell senescence and enhance antitumor immunity in mouse models of melanoma and breast cancer treated with adoptive T cell transfer therapy. These studies identify molecular processes responsible for T cell dysfunction in the tumor microenvironment and provide proof of concept for therapeutic reprogramming of T cell lipid metabolism for enhanced antitumor immunity.

RESULTS

Development of T cell senescence is a general feature of the tumor microenvironment

Tumor-reactive T cells are known to be suppressed and dysfunctional in the tumor microenvironment, which presents a major obstacle for successful tumor immunotherapy (4, 5). We found that human T_reg cells can induce both CD4⁺ and CD8⁺ T cells to become senescent as measured by senescence-associated β-galactosidase (SA-β-gal) staining (Fig. 1A and fig. S1A) (21, 32, 33). In addition to T_reg cells, different types of human cancer cells, including breast cancer MCF-7 and melanoma M586 cells, can induce senescence in cocultured human T cells (Fig. 1B and fig. S1B) (22, 34). Senescent T cells were characterized by decreased surface expression of the costimulatory molecules CD27 and CD28 (fig. S1C) and increased expression of genes encoding proinflammatory cytokines such as interleukin-1β (IL-1β), IL-6, IL-8, tumor necrosis factor–α (TNFα), and interferon-γ (IFN-γ) (Fig. 1, C and D, and fig. S1D) (21, 22, 32, 33). In addition, senescent T cells potently inhibited the proliferation of other responding CD4⁺ T cells (fig. S1E).

Fig. 1. Regulatory T cells and tumor cells induce T cell senescence.

(A) Naïve CD4⁺ T cells were incubated alone or cocultured with nT_reg cells or control effector CD4⁺CD25⁻ T cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for 3 days and then stained for SA-β-gal. Arrows indicate SA-β-gal⁺ T cells. Scale bars, 20 μm. Data are presented as means ± SD from three independent experiments. n = 6 different representative CD4⁺ T cells. **P < 0.01. (B) Anti-CD3–activated naïve CD4⁺ T cells were cocultured with MCF-7 or M586 cells at a ratio of 1:1 for 1 day, then purified, and stained for SA-β-gal after culture for an additional 3 days. Arrows indicate SA-β-gal⁺ T cells. Scale bars, 20 μm. Data presented as means ± SD from three independent experiments. n = 3 different representative CD4⁺ T cells. *P < 0.05 and **P < 0.01. (C and D) CD4⁺ T cells were cocultured with nT_reg cells (C) or MCF-7 cells (D) as in (A) and (B), and changes in cytokine mRNA expression were measured relative to untreated CD4⁺ T cells 24 hours later. Expression was normalized to β-actin expression as a control. Data are presented as means ± SD from three independent experiments. n = 3 to 6 different representative CD4⁺ T cells. *P < 0.05 and **P < 0.01. (E to H) CD4⁺ and CD8⁺ T cells were isolated from blood, lymph nodes (LNs), spleens (SPs), and tumor tissues of B16F0 subcutaneous tumor (E and F)– or E0771 subcutaneous tumor (G and H)–bearing mice and tumor-free littermates when primary tumors reached 10 to 15 mm in diameters. SA-β-gal staining (E and G) and cytokine mRNA expression (F and H) were measured. Data shown are means ± SD from seven mice in each group. *P < 0.05 and **P < 0.01. (I) CD3⁺ TILs were isolated from freshly digested human melanoma (MTIL) and breast cancer (BTIL) tissues and stained for SA-β-gal⁺. Arrows indicate SA-β-gal⁺ T cells. Scale bars, 20 μm. Naïve CD4⁺ and CD8⁺ T cells from healthy donor peripheral blood were included as controls. Data presented as means ± SD. Each dot represents an individual donor. **P < 0.01. (J) Proinflammatory cytokine expression was measured in purified CD3⁺ TILs isolated from human melanoma or breast cancer tumors. Purified naïve CD4⁺ or CD8⁺ T cells from healthy donor peripheral blood were used as controls. Data shown are means ± SD. *P < 0.05. One-way analysis of variance (ANOVA) was performed in (A), (B), (C), (F), (H), (I), and (J). Paired Student’s t test was performed in (D), and unpaired Student’s t test was performed in (E) and (G).

We next investigated the existence of senescent T cell populations in the tumor microenvironment in vivo. We used the murine mammary cancer cell line E0771 and the melanoma cell line B16F0 to establish breast and melanoma tumors, respectively. T cells from different organs and from tumors were isolated and analyzed after tumor diameters reached 10 to 15 mm. We observed elevated SA-β-gal⁺ CD4⁺ and CD8⁺ T cells in blood, lymph nodes, spleens, and tumors derived from B16F0 tumor–bearing mice, but not in control tumor-free mice (Fig. 1E). Furthermore, CD4⁺ and CD8⁺ T cells from blood and tumor tissues in B16F0-bearing mice displayed increased mRNA expression of genes encoding IL-1β, IL-6, TNFα, and IFN-γ relative to tumor-free mice (Fig. 1F). We observed similar results in the E0771 mouse model (Fig. 1, G and H).

We next generated TILs from tumor tissues obtained from patients with either breast cancer or melanoma. Elevated senescent T cell populations were observed among the TILs from both melanoma and breast tumor tissues (P < 0.001, TILs versus naïve T cells; Fig. 1I) (22). In addition, these TILs had increased expression of genes encoding proinflammatory cytokines such as IL-6 and TNFα (P = 0.044 and P = 0.047, breast cancer tumor-infiltrating T cell (BTIL) versus naïve CD4 or CD8 in IL-6, respectively; P = 0.013 and P = 0.04, BTIL versus naïve CD4 or CD8 in TNFα, respectively; P = 0.037 and P = 0.041, MTIL versus naïve CD4 or CD8 in IL-6, respectively; P = 0.013 and P = 0.022, melanoma tumor-infiltrating T cell (MTIL) versus naïve CD4 or CD8 in TNFα, respectively; Fig. 1J). Our results collectively indicate that induction of T cell senescence is an important strategy used by tumors to mediate T cell dysfunction and evade immune surveillance.

Senescent T cells have active glucose metabolism–associated enzymes

We have shown that senescent T cells induced by T_reg cells have increased phosphorylation of adenosine monophosphate-activated protein kinase, an important nutrient and energy sensor (21). Therefore, we reasoned that development of senescent T cells might involve metabolic regulation induced by T_reg cells and tumor cells. Glucose metabolism is the main metabolic pathway required for effector T cell function upon activation (43). We conducted transcriptome analyses of senescent T cells induced by human naturally occurring T_reg (nT_reg) cells at early (4 to 8 hours), middle (24 to 48 hours), and late (72 hours) stages during senescence development using Illumina whole-genome HumanHT-12 BeadChips. Our transcriptome analyses demonstrated that T_reg treatment increased gene expression of enzymes involved in glycolysis and also slightly increased gene expression of enzymes involved in tricarboxylic acid cycle pathways in responder T cells in the middle and late stages of senescence development (Fig. 2A).

Fig. 2. Senescent T cells have increased expression of enzymes associated with glucose metabolism.

(A) Human naïve CD8⁺ T cells and nT_reg cells were isolated from PBMCs of healthy donors and cocultured at a ratio of 5:1 for indicated time points. Transcriptional analysis of nT_reg-treated CD8⁺ T cells was performed, and differentially expressed genes involved in glycolysis were identified. (B) Naïve CD8⁺ T cells were cocultured with nT_reg or control CD4⁺CD25⁻ T cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for 3 days, and glucose metabolism–related gene expression was evaluated by real-time quantitative PCR (qPCR). Expression of each gene was normalized to β-actin expression and adjusted to the expression in CD8⁺ T cells alone. Data shown are means ± SD from eight to nine different independent donors. Each dot represents one individual donor. *P < 0.05 and **P < 0.01. (C) Anti-CD3–activated naïve CD8⁺ T cells were cocultured with MCF-7 cells at a ratio of 1:1 for 2 days, and glucose metabolism–related gene expression was evaluated by real-time qPCR as in (B). Data shown are means ± SD from six different independent donors. Each dot represents one individual donor. *P < 0.05 and **P < 0.01. (D and E) CD8⁺ T cells were purified from blood and tumor tissues in melanoma B16F0-bearing (D) and breast cancer E0771-bearing (E) mice when primary tumors reached 10 to 15 mm in diameters, and mRNA expression of glucose metabolism–related genes was evaluated by real-time qPCR. CD8⁺ T cells purified from blood in tumor-free mice served as controls. Data shown are means ± SD from seven mice in each group. *P < 0.05 and **P < 0.01. (F) CD3⁺ cells were purified from melanoma (MTIL) and breast cancer (BTIL) tissues from patients and gene expression of glucose transporters, and key enzymes involved in glycolysis was evaluated by real-time qPCR. Naïve CD4⁺ or CD8⁺ T cells were purified from healthy donor blood and activated with plate-coated anti-CD3 (2 μg/ml) for 3 days for controls. Data shown are means ± SD from three to eight different independent donors. Each dot represents an individual donor or patient. *P < 0.05 and **P < 0.01. (G) Naïve CD4⁺ or CD8⁺ T cells were cocultured with nT_reg or control CD4⁺CD25⁻ T cells in anti-CD3–coated (2 μg/ml) plates in the presence of different concentrations of glucose and then stained for SA-β-gal after culture for 3 days. Data shown are means ± SD from three independent experiments. **P < 0.01. (H) Anti-CD3–activated naïve CD4⁺ or CD8⁺ T cells were cocultured with MCF-7 cells at a ratio of 1:1 in the presence of different concentrations of glucose for 1 day. The treated T cells were then separated and stained for SA-β-gal after culture for an additional 3 days. Data shown are means ± SD from three independent experiments. **P < 0.01. (I) Senescent CD8⁺ T cells induced by nT_reg or MCF-7 cells were cultured in the presence of normal (11 mM) or high (25 mM) dosages of glucose for 2 days and then stained for SA-β-gal. Data shown are means ± SD from three independent experiments. (J and K) Senescent CD8⁺ T cells induced by nT_reg cells (J) or MCF-7 cells (K) were cultured in the presence of glycolysis inhibitors 3-BrPA (30 μM) or 2-DG (2 mM), and mRNA expression of each cytokine was determined relative to untreated senescent CD8⁺ T cells 24 hours later by real-time qPCR. Expression was normalized to β-actin as a control. Data are means ± SD from three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001. One-way ANOVA was performed in (B), (D), (E), and (F). Paired Student’s t test was performed in (C), and unpaired Student’s t test was performed in (G), (H), (J), and (K).

We then characterized the key metabolic genes involved in glucose metabolism in senescent T cells induced by human T_reg cells and tumor cells to validate our transcriptional analysis results using real-time quantitative polymerase chain reaction (PCR) (44). Those molecules include Glut1 and Glut3, as well as glycolysis-related enzymes hexokinase 2 (HK2), glucose-6-phosphate isomerase (GPI), phosphofructokinase 1 (PFK1), triosephosphate isomerase 1 (TPI1), enolase 1 (ENO1), pyruvate kinase muscle 2 (PKM2), and lactate dehydrogenase A (LDHα). We found that the mRNA expression of the majority of these metabolic genes was increased or unchanged in senescent CD8⁺ T cells induced by both nT_reg cells and MCF-7 cells, except down-regulation of LDHα in senescent CD8⁺ T cells induced by nT_reg cells (Fig. 2, B and C).

We next determined the expression of glucose metabolism–related genes in T cells in vivo in melanoma tumor– and breast cancer tumor–bearing mice. We found that mRNA expression of the majority of these molecules was increased in CD8⁺ TILs obtained from B16F0 tumor– and E0771 tumor–bearing mice (Fig. 2, D and E). In addition, we purified TILs from fresh tumor tissues obtained from patients with melanoma or breast cancer and observed that these molecules were similarly increased or unchanged in TILs from tumor tissue (Fig. 2F).

To dissect the role of glucose metabolism in T cell senescence, we found that addition of high concentrations of glucose (25 mM) to T cell cocultures markedly prevented responder T cell senescence mediated by nT_reg cells (P = 0.004 and P = 0.005, CD4 or CD8 T cells treated with T_reg cells in 11 mM glucose medium versus those in 25 mM glucose medium, respectively; Fig. 2G). Addition of high concentration of glucose also prevented induction of senescence in both CD4⁺ and CD8⁺ T cells mediated by breast cancer cells (P = 0.002 and P = 0.004, CD4 or CD8 T cells treated with MCF-7 cells in 11 mM glucose medium versus those in 25 mM glucose medium, respectively; Fig. 2H). These results indicate the importance of glucose competition between responder T cells and T_reg or tumor cells for the development of T cell senescence. We next explored the critical role of glucose metabolism in the function of established senescent T cells. Our results showed that high concentration of glucose (25 mM) cannot reverse senescence in T cells after coculture with T_reg cells or tumor cells (Fig. 2I). However, blockade of glycolysis in senescent T cells using the inhibitors of 2-deoxy-d-glucose (2-DG) or 3-bromopyruvate (3-BrPA) decreased their secretion of proinflammatory cytokines (P = 0.042, P = 0.006, and P = 0.008, nT_reg-induced senescent CD8 T cells treated with 3-BrPA versus medium in IL-1β, IL-6, or IL-8, respectively; P = 0.075, P = 0.003, and P = 0.0002, MCF-7–induced senescent CD8 T cells treated with 2-DG versus medium in IL-6, IL-8, or IFN-γ, respectively; Fig. 2, J and K) (21). Our results indicate that senescent T cells have active glucose metabolism that may be required for T cells to execute their biological functions.

Senescent T cells have unbalanced lipid metabolism

Lipid metabolism is critical for proper functioning of certain T cell populations, including T_reg, T helper 17, and memory CD8⁺ T cells (10, 45–48). Furthermore, FA metabolism is linked to cellular senescence (49). Therefore, we determined whether lipid metabolic regulation is involved in the development of T cell senescence induced in vitro. We performed transcriptome analyses of senescent T cells induced by human nT_reg cells at early (4 to 8 hours), middle (24 to 48 hours), and late (72 hours) stages of senescence development and further identified the altered genes involved in lipid metabolism. We found that T_reg treatment markedly promoted the expression of genes involved in cholesterol biosynthesis and transport and FA biosynthesis in responder T cells during the early and middle stages of senescence development, but those genes were decreased in senescent T cells at the late stage of T cell senescence. Furthermore, genes involved in regulation of triacylglycerol (TAG) metabolism were also up-regulated in senescent T cells. In addition, the genes involved in the catabolism of cholesterol, FA, and TAG were markedly down-regulated in senescent T cells (Fig. 3A). These results suggest that lipid metabolism might be critical for T cell senescence.

Fig. 3. Senescent T cells have unbalanced lipid metabolism.

(A) Transcriptional analysis of nT_reg-treated CD8⁺ T cells was performed as in Fig. 2A. Alterations of genes involved in lipid metabolism were identified and normalized to log₂ expression at different time points. (B) Naïve CD8⁺ T cells were cocultured with nT_reg at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for indicated time points. Lipid metabolism–related enzyme gene expression was evaluated by real-time qPCR. Expression of each gene was normalized to β-actin expression and normalized relative to expression in CD8⁺ T cells at 8 hours. Data shown are means ± SD from four to six different independent donors. Paired Student’s t test was performed. *P < 0.05 and **P < 0.01. (C) Anti-CD3–activated CD8⁺ T cells were cocultured with MCF-7 cells at a ratio of 1:1 for different times. The treated CD8⁺ T cells were then separated, and lipid metabolism–related gene expression was evaluated by real-time qPCR. Expression of each gene was normalized to β-actin expression and adjusted to expression in CD8⁺ T cells at 8 hours. Data shown are means ± SD from six different independent donors. Paired Student’s t test was performed. *P < 0.05 and **P < 0.01. (D) CD3⁺ TILs were purified from melanoma (MTIL) and breast cancer (BTIL) tissues, and gene expression of key enzymes involved in lipid metabolism was evaluated by real-time qPCR. CD4⁺ or CD8⁺ T cells from healthy donors activated with plate-coated anti-CD3 (2 μg/ml) for 3 days served as controls. Data shown are means ± SD from three to eight different independent donors. Each dot represents an individual donor or patient. One-way ANOVA and unpaired Student’s t test were performed. *P < 0.05. (E) Naïve CD4⁺ or CD8⁺ T cells were cocultured with nT_reg or control CD4⁺CD25⁻ T cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for 3 days. Treated T cells were purified, stained with Fillipin III, and then analyzed for cholesterol content by fluorescence microscopy. Scale bars, 20 μm. (F) CD3⁺ TILs were purified from breast cancer and melanoma tissues isolated from patients and stained with Fillipin III as in (E). Anti-CD3 (2 μg/ml) preactivated CD4⁺ and CD8⁺ T cells purified from healthy donors served as controls. Scale bars, 20 μm.

We next determined gene expression of the key enzymes related to both cholesterol synthesis and FA oxidation and synthesis in senescent CD4⁺ and CD8⁺ T cells induced by human nT_reg cells and MCF-7 cancer cells, including 3-hydroxy-3-methyl-glutaryl–coenzyme A (CoA) reductase (HMGCR), 3-hydroxy-3-methylglutaryl-CoA synthase 1 (HMGCS1), squalene monooxygenase (SQLE), isopentenyl-diphosphate delta isomerase 1 (IDI1), carnitine palmitoyltransferase I (CPT-1), FA synthase (FASN), and acetyl-CoA carboxylases 1 (ACC1) (44). Consistent with transcriptional analysis, we found markedly up-regulated gene expression of the key enzymes related to cholesterol and FA metabolism in senescent CD8⁺ T cells induced by human nT_reg and MCF-7 cells (Fig. 3, B and C). Furthermore, these enzymes were dynamically changed during senescence development in T cells induced by nT_reg and tumor cells (Fig. 3, B and C). We then determined the alterations of these lipid-associated enzymes in T cells isolated from B16F0 melanoma tumor– or E0771 breast cancer tumor–bearing mice. We observed markedly up-regulated expression of most key enzymes related to cholesterol and FA metabolism in blood-derived T cells and TILs, except down-regulation of CPT-1 and IDI1 in TILs from both B16F0 tumor– and E0771 tumor–bearing mice (fig. S2, A and B). However, the mRNA expression of those lipid enzymes was not obviously altered in CD4⁺ and CD8⁺ T cells purified from spleens and lymph nodes of B16F0 tumor– and E0771 tumor–bearing mice compared with those in T cells from tumor-free mice (fig. S2, C and D). In addition, we found up-regulated expression of genes encoding all the key enzymes measured except SQLE in TILs from patients with breast cancer, as well as up-regulation of genes encoding FASN, HMGCR, and IDI1 in TILs from patients with melanoma (Fig. 3D). These results demonstrate dynamic changes and unbalanced lipid metabolism in senescent cells in vitro and both mice and humans in vivo.

Given that gene expression of enzymes important for cholesterol synthesis was increased in senescent T cells, we determined the cholesterol concentrations in T_reg-induced senescent T cells using a fluorometric assay (28) . We found cholesterol markedly accumulated in responder CD4⁺ and CD8⁺ T cells after coculture with nT_reg cells, but not control effector T cells (Fig. 3E). Furthermore, increased amounts of cholesterol were also observed in TILs from both melanoma and breast cancer tissues isolated from patients, as well as from B16 tumor– and E0771 tumor–bearing mice (Fig. 3F and fig. S2, E and F).

We then identified what lipid species were changed in senescent T cells induced by T_reg and tumor cells and whether the altered lipid components are directly related to T cell senescence using a mass spectrometry–based lipidomic analysis. We were particularly interested in glycerophospholipids and sphingolipids (figs. S3 and S4A) (50, 51). nT_reg-induced senescent T cells showed higher concentrations of total and majority of subfractions of free FAs (FFAs) and cholesteryl esters (CEs) but lower concentrations of ceramide (Cer), sphingomyelins (SMs), phosphatidylcholine (PC), and phosphatidylethanolamine (PE) than those in control T cells (Fig. 4, A and B). Senescent T cells induced by nT_reg cells did not show changes of lysophosphatidylcholine (LPC) compared with control T cells. Furthermore, senescent T cells induced by MCF-7 breast cancer cells also showed markedly increased total and subfractions of CEs and decreased Cer and PE compared with control T cells (Fig. 4, C and D). MCF-7–induced senescent T cells did not markedly alter the concentrations of PC, phosphatidylserine (PS), SM, and LPC (Fig. 4D). We then determined whether the reduced lipid species were critical for T cell senescence. We observed that addition of different concentrations of Cer, PE, PS, LPC, 16:0 alkyl-lysophosphatidylcholine (aLPC), and 16:0 lysoplasmenylcholine (pLPC) markedly prevented senescence induction in T cells cocultured with nT_reg cells (Fig. 4E), although these lipids themselves, aside from Cer, did not induce T cell senescence (fig. S4B). Collectively, these data strongly indicate that unbalanced lipid metabolism plays a crucial role in regulation of senescence development in T cells mediated by human T_reg and tumor cells in vitro.

Fig. 4. Alterations of lipid species in senescent T cells induced by T_reg cells and tumor cells.

(A and B) Naïve CD8⁺ T cells were cocultured with nT_reg or control CD4⁺CD25⁻ T cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for 3 days. Treated CD8⁺ T cells were purified, and total and multiple molecular species in free fatty acid (FFA) (A) and cholesteryl ester (CE), phosphatidylcholine (PC), phosphatidylethanolamine (PE), and Cer (B) from different treatment groups were subjected to mass spectrometry. Data shown are means ± SD from T cells purified from two to three independent donors. *P < 0.05, **P < 0.01, and ***P < 0.001. (C) Anti-CD3–activated naïve CD4⁺ T cells were cocultured with MCF-7 cells at a ratio of 1:1 for 1 day. The treated CD4⁺ T cells were then separated, and lipid extracts of T cells from different treatment groups were analyzed by mass spectrometry after an additional 3-day culture. Data shown are means ± SD from T cells purified from four different independent donors. *P < 0.05 and **P < 0.01. (D) Summary of lipidomic analysis results in senescent T cells induced by T_reg and tumor cells as described in (A) to (C). (E) Naïve CD4⁺ T cells were cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plates in the presence or absence of the indicated concentrations of lipid fractions for 3 days and then stained for SA-β-gal. Data shown are means ± SD from three independent experiments. n = 4 to 6 different representative CD4⁺ T cells. **P < 0.01. Unpaired Student’s t test was performed in (A) and (C). One-way ANOVA was performed in (B) and (E).

Accumulated LDs contribute to the development of T cell senescence

We next explored the role of the increased lipid species in senescent T cell development mediated by T_reg cells and tumor cells. Our studies have shown that CE is markedly increased in senescent T cells, as are increased FFA and cholesterol (Figs. 3E and 4, A to C). More recent work has demonstrated that T cells with increased cholesterol esterification have impaired proliferation and anti-tumor effector function (52). These are key components for LDs, which are lipid-rich cytoplasmic organelles that directly regulate inflammation and immune responses to cancer (fig. S4C) (52–54). Therefore, we determined whether LD formation was induced in the context of T cell senescence using Oil Red O staining. We observed that Oil Red O⁺ T cell populations were markedly increased in responder CD4⁺ and CD8⁺ T cells after coculture with nT_reg cells, as well as with MCF-7 and M586 tumor cells (P < 0.001, naïve CD4 and CD8 T cells treated with nT_reg, MCF-7, or M586 cells versus naïve T cells in the culture medium only; Fig. 5, A and B). We also used the lipophilic fluorescent dye BODIPY 493/503 to evaluate the amount of lipids in senescent T cells (55). Consistent with the Oil Red O staining results, T_reg-induced senescent T cells displayed more fluorescence intensity than that of CD4⁺ T cells cultured in medium only or with CD4⁺CD25⁻ effector T cells (fig. S4, D and E). To identify the relationship between the accumulated LDs and cell senescence in T cells, we isolated BODIPY^high, BODIPY^medium, and BODIPY^low populations in senescent T cells induced by nT_reg cells with fluorescence-activated cell sorting (FACS) and found that almost all the SA-β-gal⁺ T cells were in the BODIPY^high populations (P < 0.001, BODIPY^high versus BODIPY^medium or BODIPY^low groups; Fig. 5C). In addition, LD formation was colocalized with increased cell cycle regulatory molecules P21 and P53 in nT_reg-induced senescent T cells (fig. S4F). We next determined whether reduced lipid species were correlated with accumulated LDs in senescent T cells. Decreased lipid species, except Cer and PS, did not promote LD formation in T cells (fig. S4G). However, supplementation of these lipids could prevent LD formation in senescent T cells induced by T_reg cells, as indicated by decreased Oil Red O⁺ T cell populations in senescent T cells (Fig. 5D). These results suggest that the interactions among lipid species and LDs are critical for the development of T cell senescence.

Fig. 5. Accumulated LDs contribute to the development of T cell senescence mediated by T_reg and tumor cells.

(A and B) Naïve CD4⁺ and CD8⁺ T cells were cocultured with nT_reg cells (A) or with MCF-7 or M586 tumor cells (B). The cocultured CD4⁺ and CD8⁺T cells were purified and stained for Oil Red O to measure lipid droplet (LD) accumulation. Oil Red O ⁺ T cells are indicated by arrows. Scale bars, 20 μm. Data shown are means ± SD from three independent experiments. n = 3 to 5 different representative T cells. ***P < 0.001. (C) T_reg-treated senescent T cells were stained with BODIPY 493/503, sorted on the basis of BODIPY 493/503 fluorescence by FACS, and measured for SA-β-gal expression. Scale bars, 20 μm. Data shown are means ± SD from three different independent donor T cells. **P < 0.01 and ***P < 0.001. (D) Naïve CD4⁺ T cells were cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plate in the presence or absence of the indicated concentrations of lipid species for 3 days and then purified and stained for Oil Red O. Data shown are means ± SD from three independent experiments. n = 3 to 6 different representative CD4⁺ T cells. *P < 0.05 and **P < 0.01. (E and F) Naïve CD4⁺ and CD8⁺ T cells were cocultured with nT_reg cells (E) or MCF-7 cells (F), and mRNA expression of ACAT1 and ACAT2 was determined by the real-time qPCR. Expression was normalized to β-actin expression and adjusted to expression in T cells cultured alone. Data are presented as means ± SD from three to four independent experiments. *P < 0.05 and **P < 0.01. (G and H) Naïve T cells were pretreated with ACAT inhibitor avasimibe (Ava) for 24 hours and then cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plates for 3 days. The treated CD4⁺ and CD8⁺ T cells were purified and stained for SA-β-gal (G) and Oil Red O (H), respectively. Data shown are means ± SD from three independent experiments. n = 5 to 10 different representative T cells. **P < 0.01. (I) Naïve CD8⁺ T cells were cocultured with nT_reg cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for indicated time points, and mRNA expression of lipase A (LIPA) was evaluated by real-time qPCR. Expression was normalized to β-actin expression and adjusted to expression in T cells at 8 hours. Data are means ± SD from four independent experiments. **P < 0.01. (J and K) Naïve CD8⁺ T cells were cultured in anti-CD3–coated (2 μg/ml) plates in the presence or absence of lipase inhibitor orlistat (5 μM) for 3 days. The treated CD8⁺ T cells were stained for SA-β-gal (J) and Oil Red O (K). Data shown are means ± SD from five different independent donor T cells. **P < 0.01. One-way ANOVA was performed in (A) to (E), (G), and (H). Paired Student’s t test was performed in (F) and (I) to (K).

We next dissected the molecular mechanisms responsible for increased CEs and LDs during T cell senescence. CE synthesis is induced by the cholesterol acyltransferase 1 (ACAT1) and ACAT2 (52–54). However, the expression of ACAT1 and ACAT2 was down-regulated rather than increased in senescent CD4⁺ and CD8⁺ T cells induced by nT_reg and MCF-7 cancer cells (Fig. 5, E and F). We further determined the kinetic changes of these two genes during T cell senescence. ACAT1 and ACAT2 mRNA concentrations were not changed at the early time points but decreased at the late time point in both T_reg and tumor-induced senescent T cells (fig. S4, H to J). However, we observed increased ACAT2 in CD8⁺ TILs isolated from both B16 tumor– and E0771 tumor–bearing mice (fig. S4K). To identify the importance of increased CEs for T cell senescence, we pharmacologically inhibited ACAT with avasimibe and determined whether blockage of CE synthesis can prevent cell senescence and LD formation in T cells. We found that treatment with avasimibe markedly reduced the proportion of senescent T cells and Oil Red O⁺ T cell populations in CD4⁺ and CD8⁺ T cells cocultured with T_reg cells (Fig. 5, G and H). We ruled out the possibility that avasimibe treatment alone does not induce cell senescence or LD formation in naïve T cells (fig. S4, L and M). In addition to ACAT, the lysosomal acid lipase (LAL), encoded by LIPA, is another important enzyme involved in down-regulation of CE and LD formation (fig. 4C) (52–54). We therefore determined LIPA expression in senescent T cells and found that LIPA was down-regulated in senescent T cells during the senescence development mediated by T_reg cells (Fig. 5I). Furthermore, inhibition of LAL with the specific inhibitor orlistat markedly increased cell senescence and promoted LD formation in naïve T cells, suggesting that LAL might also be involved in the induction of T cell senescence and increases in CEs and LDs (Fig. 5, J and K). Collectively, these results indicate that accumulation of LDs in T cells induced by T_reg and tumor cells is a critical process for the development of T cell senescence.

Elevated cPLA₂α expression is required for LD accumulation and senescence induction in T cells

Recent studies demonstrated that cPLA₂α is critical for regulation of phospholipid metabolism and LD formation (56–58). We therefore hypothesized that cPLA₂α might be involved in the LD formation in T cells, resulting in T cell senescence and dysfunction during interactions with T_reg cells or tumor cells. We first determined the kinetic alterations of expression of the gene encoding cPLA₂α in senescent T cells during coculture with T_reg cells or tumor cells. We observed an increase in expression of the gene encoding cPLA₂α in T cells during the progression to senescence and in already senescent T cells mediated by T_reg cells or MCF-7 breast tumor cells (Fig. 6, A and B). cPLA₂α protein was also elevated in senescent T cells induced by T_reg cells using both Western blot and flow cytometry analyses (Fig. 6C and fig. S5A). In addition, we purified T cells from blood and tumor tissues from B16 tumor– and E0771 tumor–bearing mice and evaluated expression of the gene encoding cPLA₂α in T cells. We observed that mRNA expression in the tumor-infiltrating CD4⁺ and CD8⁺ T cells from both melanoma tumor– and breast cancer tumor–bearing mice was higher than that of T cells purified from blood in control tumor-free mice (Fig. 6, D and E). These results demonstrate that the expression of cPLA₂α is elevated in senescent T cells in the tumor microenvironment.

Fig. 6. Elevated cPLA₂α promotes LD accumulation and induction of senescence in T cells.

(A and B) Naïve CD8⁺ and CD4⁺ T cells were cocultured with nT_reg cells (A) or MCF-7 cells (B) for indicated time points, and mRNA expression of cPLA₂α was determined by real-time qPCR. Expression was normalized to β-actin expression and adjusted to expression in T cells cultured alone at 4 hours. Data shown are means ± SD from three independent experiments. *P < 0.05 and **P < 0.01. (C) Naïve CD4⁺ and CD8⁺ T cells were cocultured with or without T_reg cells or control CD4⁺CD25⁻ T cells at a ratio of 4:1 in the presence of plate-bound anti-CD3 (2 μg/ml) for 3 days. cPLA₂α protein in treated T cells was analyzed by Western blot. nT_reg-1 and nT_reg-2 cells are nT_reg cells purified from two individual healthy donors. (D and E) T cells were purified from blood and tumors of melanoma B16F0 tumor–bearing (D) and breast cancer E0771 tumor–bearing (E) mice, as described in Fig. 1. mRNA expression of cPLA₂α was evaluated by real-time qPCR. Data shown are means ± SD from seven mice in each group. *P < 0.05 and **P < 0.01. (F) Immunofluorescence after staining with anti-cPLA₂α antibody and BODIPY 493/503 was evaluated in responder CD4⁺ T cells treated with nT_reg or control CD4⁺CD25⁻ effector T cells. Scale bars, 20 μm. DAPI was used to stain cell nuclei. (G) Naïve CD4⁺ and CD8⁺ T cells were pretreated with the cPLA₂α inhibitor MAFP for 24 hours and then cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plates for 3 days. The treated CD4⁺ and CD8⁺ T cells were purified and stained for Oil Red O. Data are means ± SD from three independent experiments. **P < 0.01. (H) Anti-CD3 (2 μg/ml)–preactivated naïve CD4⁺ and CD8⁺ T cells were transfected with siRNAs specific for PLA2G4A or control (Ctr) nontargeting siRNA for 2 days. mRNA expression of cPLA₂α was determined by real-time qPCR. Expression was normalized to β-actin expression and adjusted to expression in T cells cultured in medium. Data shown are means ± SD from three independent experiments. **P < 0.01 and ***P < 0.001. (I) Anti-CD3 (2 μg/ml)–preactivated naïve CD4⁺ and CD8⁺ T cells were transfected with siRNAs specific for PLA2G4A or Ctr-siRNA and cultured for 2 days. The transfected T cells were further cocultured with nT_reg cells at a ratio of 4:1 for 2 days and were subsequently purified and stained for Oil Red O. Data shown are means ± SD from three independent experiments. **P < 0.01. (J) Cells were treated and cultured as in (G). Protein concentrations of cPLA₂α, P53, and P21 in treated CD4⁺ T cells were analyzed by Western blot and further quantitatively analyzed and compared against GAPDH expression using densitometry. Results shown are means ± SD from three independent experiments. **P < 0.01. (K and L) CD4⁺ and CD8⁺ T cells were pretreated with cPLA₂α inhibitor MAFP (K) or transfected with siRNAs specific for PLA2G4A (L) and cocultured with nT_reg cells. The treated CD4⁺ and CD8⁺ T cells were purified and stained for SA-β-gal. Data shown are means ± SD from three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001. (M) Flow cytometry analysis showed that inhibition of cPLA₂α with MAFP restored costimulatory molecules CD27 and CD28 in responder T cells. The gray-shaded histogram indicates T cells with the isotype control antibody staining. Unpaired Student’s t test was performed in (A) and (B). One-way ANOVA was performed in (D), (E), (G), and (H) to (L).

We next investigated the relationship between cPLA₂α expression and LD formation in senescent T cells induced by T_reg cells using immunofluorescence in situ analysis. We observed the expression of cPLA₂α in responder T cells treated with T_reg cells, but not with control T cells (Fig. 6F). Furthermore, cPLA₂α expression was paralleled with the accumulated LDs in T_reg-induced senescent T cells as evaluated by BODIPY 493/503 staining (Fig. 6F). We next determined whether blocking cPLA₂α in T cells with the cPLA₂α-specific pharmacological inhibitor methyl arachidonyl fluorophosphonate (MAFP) or knockdown of cPLA₂α expression with specific small interfering RNA (siRNA) could prevent LD formation in senescent T cells induced by T_reg cells (56). Pretreatment of CD4⁺ and CD8⁺ T cells with MAFP markedly decreased LD formation in responder T cells cocultured with T_reg cells (Fig. 6G). However, MAFP itself did not induce cell senescence and LD formation in naïve T cells (fig. S5, B and C). Furthermore, transfection of cPLA₂α-specific siRNA down-regulated cPLA₂α expression and prevented LD formation in responder T cells cocultured with T_reg cells (Fig. 6, H and I). These results were also confirmed by studies using another cPLA₂α-specific pharmacological inhibitor arachidonyl trifluoromethyl ketone (ATK; AACOCF3) (fig. S5D) (59). Together, our studies identify a causal relationship between increased cPLA₂α and alterations to lipid metabolism during T cell senescence in vitro and in vivo.

We next determined whether elevated cPLA₂α is responsible for senescence induction in responder T cells induced by T_reg cells. Our previous studies have shown that senescent T cells induced by T_reg cells and tumor cells have increased expression of cell cycle regulatory molecules P16, P21, and P53 and production of inflammatory cytokines and down-regulation of costimulatory molecules CD27 and CD28 (21, 22, 32, 33). We found that inhibition of cPLA₂α with MAFP down-regulated cPLA₂α protein expression in responder T cells induced by T_reg cells (Fig. 6J). MAFP treatment not only inhibited the T_reg-induced increases in P21 and P53 expression but also prevented responder T cells from becoming senescent (Fig. 6, J and K). In addition to MAFP, knockdown of cPLA₂α with siRNA or blockade with the other inhibitor, ATK, reduced senescence induction in responder T cells during coculture with T_reg cells (Fig. 6L and fig. S5E). Furthermore, cPLA₂α inhibition with MAFP or siRNA led to restoration of the expression of costimulatory molecules CD27 and CD28 and suppression of inflammatory cytokine secretion in T_reg-induced senescent T cells (Fig. 6M and fig. S5, F and G). To further identify a causative role of cPLA₂α in the promotion of LD formation and senescence induction in T cells, we also used a gain-of-function strategy with cPLA₂α activators phospholipase A₂–activating protein (PLAP) and mastoparan (60, 61). Our results demonstrated that PLAP or mastoparan treatment not only induced cell senescence but also promoted LD accumulation, in both naïve CD4⁺ and CD8⁺ T cells (fig. S5, H and I). Collectively, these studies suggest that cPLA₂α is a key molecule that controls both lipid metabolism and cell fate in responder T cells in vitro.

We also validated our findings of changes in lipid metabolism in senescent T cells in aged mice and irradiation-induced senescent human T cells. T cells purified from blood and spleens of aged mice (greater than 10 months old) had increased SA-β-gal⁺ cell populations compared with those from adult mice (fig. S6A). Furthermore, T cells from aged mice also had decreased Lamin B1 gene expression, another marker of T cell senescence (fig. S6B) (62). These studies indicated that T cells derived from aged mice are senescent. In addition, T cells purified from aged mice showed increased LD formation and cPLA₂α expression as compared with T cells from adult mice (fig. S6, C and D). We also observed that x-ray irradiation at 5 or 10 grays (Gy) could induce senescence in human T cells, showing increased SA-β-gal expression and down-regulation of CD27 and CD28 (fig. S6, E and F) (32, 63). In addition, senescent T cells induced by irradiation exhibited accumulated LDs and increased cPLA₂α mRNA expression (fig. S6, G and H). These studies further confirm that senescent T cells accumulate LDs and that cPLA₂α is important for increased lipid metabolism.

MAPK and STAT signaling control lipid metabolism and cPLA₂α activity in responder T cells during T cell senescence

Our recent studies have shown that initiation of ataxia-telangiectasia mutated (ATM)–associated DNA damage is the cause for T cell senescence and dysfunction induced by both human T_reg cells and tumor cells (21, 22). nT_reg treatment induced activation and phosphorylation of DNA damage molecules in responder CD4⁺ T cells, including ATM, H2A histone family member X (H2AX), tumor suppressor p53 binding protein 1 (53BP1), and checkpoint kinase 2 (CHK2) (fig. S7A). We therefore determined whether ATM-associated DNA damage is also causally related to altered lipid metabolism in senescent T cells. Blockade of ATM activation with the ATM-specific inhibitor KU55933 markedly decreased both mRNA and protein expression of cPLA₂α in T_reg-induced senescent T cells (Fig. 7, A and B). In addition, blockage of ATM-associated DNA damage inhibited the accumulation and formation of LDs and prevented senescence induction in responder CD4⁺ and CD8⁺ T cells induced by T_reg cells (Fig. 7, C and D). These results further support mechanistic relationships among DNA damage initiation, altered lipid metabolism, and cell senescence in T cells.

Fig. 7. MAPK and STAT signaling regulate cPLA₂α and lipid metabolism in senescent T cells.

(A and B) Anti-CD3–activated CD8+ T cells were pretreated with or without the ATM inhibitor KU55933 (10 μM) for 1 day and then cocultured with nT_reg cells at a ratio of 4:1 for 24 hours (A) or 3 days (B). The treated CD8⁺ T cells were purified and cPLA₂α mRNA and protein expression were evaluated using real-time qPCR (A) and flow cytometry (B), respectively. Data shown in (A) were normalized to β-actin expression, adjusted to expression in CD8⁺ T cells without KU55933 treatment, and are presented as means ± SD from four different independent donor T cells. **P < 0.01. The gray shaded histogram (B) indicates T cell control with the secondary antibody staining. (C and D) Naïve CD4⁺ and CD8⁺ T cells were pretreated with KU55933 (10 μM) for 24 hours and then cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3 coated (2 μg/ml) plate for 3 days. The treated CD4⁺ and CD8⁺ T cells were purified and stained for SA-β-gal (C) and Oil Red O (D), respectively. Data shown are means ± SD from three independent experiments. *P < 0.05 and **P < 0.01. (E and F) Immunofluorescence staining showing colocalization of MAPK (E) or STAT (F) molecules with LDs in responder T cells treated with T_reg or control CD4⁺CD25⁻ effector T cells using BODIPY 493/503 and antibodies against p-ERK, p-P38, p-STAT1, or p-STAT3. Scale bars, 20 μm. DAPI was used to stain cell nuclei. (G) Naïve CD4⁺ T cells were pretreated with indicated inhibitors for 24 hours and then cocultured with nT_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plates for 3 days. cPLA₂α protein concentration in treated CD4⁺ T cells were determined by Western blot, quantitatively analyzed, and compared against GAPDH expression with a densitometer. Data shown are means ± SD from three independent experiments. **P < 0.01. (H and I) Responder CD4⁺ and CD8⁺ T cells were treated with indicated inhibitors and nT_reg cells, subsequently purified, and stained for Oil Red O (H) and SA-β-gal (I), respectively. Data shown are means ± SD from three independent experiments. **P < 0.01. One-way ANOVA was performed in (A), (C), (D), and (G) to (I).

We have previously showed that both MAPK and STAT1/3 signaling pathways are involved in the regulation of T_reg-induced T cell senescence (fig. S7, B and C) (21, 32). We thus investigated how these two signaling pathways interact to control lipid metabolism in senescent T cells. We performed immunofluorescence analysis to visualize the interaction between MAPK signaling, as measured by phosphorylated P38 and extracellular signal–regulated kinase (ERK) or STAT signaling, as measured by phosphorylated STAT1 and STAT3, and LD accumulation in T_reg-induced senescent T cells. We found increased phosphorylated activation of P38, ERK, STAT1, and STAT3 in responder T cells cultured with T_reg cells but not with control T cells (Fig. 7, E and F). Furthermore, senescent T cells also displayed accumulation of LDs as measured by BODIPY 493/503 staining. In addition, phosphorylated P38, ERK, STAT1, and STAT3 molecules were colocalized with LDs in T_reg-induced senescent T cells (Fig. 7, E and F).

We next functionally blocked the activities of MAPKs and STAT1/3 signaling in T cells and explored alterations of T cell lipid metabolism and senescence during the senescence development induced by T_reg cells. Pharmacologic inhibition of P38 and ERK signaling with respective inhibitors SB203580 and U0126, as well as STAT1/STAT3 signaling with respective inhibitors methylthioadenosine (MTA) and S3I-201, suppressed cPLA₂α expression in T_reg-induced senescent T cells (Fig. 7G). However, blockage of cPLA₂α activity with MAFP did not inhibit the phosphorylation of ERK and P38 or STAT1 and STAT3 in T_reg-treated CD4⁺ T cells, further suggesting that MAPK and STAT signaling directly control cPLA₂α expression in T cells during T cell senescence (fig. S7D). In addition, inhibition of MAPK and STAT signaling decreased Oil Red O⁺ T cell populations and prevented senescence induction in responder T cells induced by T_reg cells (Fig. 7, H and I). Furthermore, inhibition of ATM, MAPK, and STAT signaling with these inhibitors also prevented the T_reg-induced up-regulation of CE and down-regulation of PC in senescent T cells (fig. S8). These studies clearly indicate that ATM-associated DNA damage initiation and MAPK signaling and STAT signaling activation control lipid metabolism in senescent T cells in vitro.

Reversal of tumor-specific effector T cell senescence by inhibition of cPLA₂α activation enhances antitumor immunity in vivo

Our studies suggest that preventing the generation of senescence in tumor-specific T cells is critical for antitumor immunity (21, 22, 32–34). Our current in vitro studies have indicated that accumulation of LDs induced by cPLA₂α is a checkpoint for control of T cell senescence and function mediated by both tumor cells and T_reg cells. Therefore, we next explored whether we can manipulate cPLA₂α activation and LD formation in effector T cells as a strategy for tumor immunotherapy. We first used the well-established B16 melanoma and Pmel (gp100-specific) T cell receptor (TCR) transgenic mouse models to test our hypothesis (64). Preactivated gp100-specific CD8⁺ T cells were adoptively transferred into B16F10 tumor–bearing mice through intravenous injection at day 6 after tumor cell inoculation. MAFP (7.5 mg/kg per mouse) was injected intraperitoneally into the mice every 3 days for a total of four injections after the adoptive transfer of T cells. Tumor growth and mouse survival were evaluated. Furthermore, adoptively transferred gp100-specific CD8⁺ T cells were purified, recovered from indicated organs at the end of experiments, and further analyzed for the effects of the MAFP treatment on LD formation and senescence induction.

B16F10 tumor cells grew quickly in untreated mice and in mice receiving adoptive transfer of Pmel-1 T cells. MAFP treatment promoted the inhibition of tumor growth mediated by gp100-specific Pmel-1 CD8⁺ T cells (Fig. 8A). Furthermore, Kaplan-Meier survival analysis showed that mice treated with Pmel-1 T cells combined with MAFP survived longer than the other two groups (Fig. 8B). MAFP treatment alone did not have effects on tumor growth (fig. S9A). We confirmed the molecular changes of recovered tumor-specific T cells in vivo from different organs and tumors after MAFP treatment. We observed that administration of MAFP inhibited gene expression of cPLA₂α in recovered gp100-specific Pmel-1 CD8⁺ T cells in vivo (fig. S9B). Furthermore, MAFP treatment markedly increased CD8⁺ T cell fractions in different organs and increased granzyme B⁺, perforin⁺, and IFN-γ⁺ cell populations in the adoptively transferred Pmel-1 T cells (fig. S9, C and D). In addition, MAFP treatment markedly decreased LD formation and prevented cell senescence in the adoptively transferred T cells purified from blood, lymph nodes, spleens, and tumor tissues (Fig. 8, C and D). These data suggest that MAFP treatment enhances antitumor activity of tumor-specific effector T cells rather than directly suppressing tumor growth.

Fig. 8. Reversal of T cell senescence via reprogramming of T cell lipid metabolism enhances antitumor immunity in vivo.

Mouse B16F10 tumor cells (2 × 10⁵ per mouse) were subcutaneously injected into C57BL/6 mice. Activated Pmel-1 T cells (1.7 × 10⁶ per mouse) were adoptively transferred through intravenous (i.v.) injection into B16F10 tumor–bearing mice at day 6 after tumor inoculation. MAFP (7.5 mg/kg per mouse) was injected intraperitoneally into the mice at days 1, 4, 7, and 10 after T cell transfer to inhibit cPLA₂α. (A) Tumor volumes were measured and presented as means ± SD (n = 6 mice per group). (B) Survival was evaluated with Kaplan-Meier analysis (n = 11 to 12 mice per group). (C and D) Transferred Pmel-1 T cells were isolated from blood, LNs, SPs, and tumors (TILs) in B16F10 tumor–bearing mice at day 28 after tumor injection and then stained for Oil Red O (C) and SA-β-gal (D), respectively. Data shown are means ± SD from four to five mice each group. **P < 0.01. (E) Mouse E0771 tumor cells transduced with retroviral vector encoding melanoma tumor antigen gp100 were subcutaneously injected into NSG mice at day 0. Activated Pmel-1 T cells were adoptively transferred into tumor-bearing mice at day 6. MAFP were administered intraperitoneally at days 1, 4, 7, and 10 after adoptive transfer of Pmel-1 T cells. Tumor sizes were measured and presented as means ± SD (n = 5 mice per group). (F and G) The transferred Pmel-1 T cells were isolated from blood, SPs, and tumors (TIL) of E0771-bearing mice at day 24 after tumor injection and then stained for Oil Red O (F) and SA-β-gal (G), respectively. Data shown are means ± SD (n = 4 to 5 mice per group). *P < 0.05 and **P < 0.01. (H) Mouse B16F10 tumor cells (2 × 10⁵ per mouse) were subcutaneously injected into C57BL/6 mice. Activated Pmel-1 T cells (2 × 10⁶ per mouse) were adoptively transferred into B16F10 tumor–bearing mice at day 11 after tumor inoculation by intravenous injection. MAFP (15 mg/kg per mouse) was administered intraperitoneally at days 1, 4, 7, and 10 after T cell transfer. Tumor volumes were measured and presented as means ± SD (n = 5 to 6 mice per group). (I) Human 586mel tumor cells (5 × 10⁶ per mouse) were subcutaneously injected into NSG mice. Tumor-specific CD8⁺ TIL586 cells (5 × 10⁶ per mouse) were intravenously injected on day 5 after tumor injection. MAFP (7.5 mg/kg per mouse) was administered intraperitoneally at days 1, 4, 7, and 10 day after adoptive transfer of CD8⁺ TIL586 T cells. Tumor volumes were measured and presented as means ± SD (n = 5 to 6 mice per group). (J and K) The transferred human TIL586 T cells were isolated from blood, SPs, and tumors (TIL) of 586mel tumor-bearing mice at day 39 after tumor injection and then stained for Oil Red O (J) and SA-β-gal (K), respectively. **P < 0.01. One-way ANOVA and unpaired Student’s t test were performed in (A), (E), (H), and (I). Unpaired Student’s t test was performed in (C), (D), (F), (G), (J), and (K).

To further validate the ability of cPLA₂α inhibition to enhance effector T cell functions, we used the cPLA₂α inhibitor, ATK. Consistent with the results shown in MAFP studies, we found that administration of ATK enhanced antitumor efficacy and prolonged mouse survival mediated by tumor-specific Pmel-1 T cells in the B16F10 melanoma model (fig. S10, A and B). Furthermore, ATK treatment similarly decreased LD accumulation and prevented senescence induction, as well as promoted effector T cell populations in tumor-specific Pmel-1 T cells, suggesting that AKT treatment enhanced antitumor immunity (fig. S10, C to E).

We then extended our studies to the E0771 breast cancer model. E0771 tumor cells were transduced with a retroviral vector encoding tumor antigen gp100 such that mice could be treated with adoptive transfer of Pmel-1 T cells. We obtained similar results as shown in the B16F10 melanoma model. The adoptively transferred tumor-specific Pmel-1 T cells suppressed E0771/gp-100 tumor growth, which was enhanced by MAFP administration (Fig. 8E). In addition, MAFP treatment decreased the SA-β-gal⁺ and Oil Red O⁺ T cell populations in the recovered gp-100 CD8⁺ T cells from blood, spleen, and tumors in E0771 tumor–bearing mice (Fig. 8, F and G).

We next determined whether this strategy could enhance T cell–mediated antitumor immunity in a late stage of cancer development to explore its translational potential. We adoptively transferred Pmel-1 T cells after B16F10 tumor grew to a larger size (greater than 8 mm by 8 mm) and then administered MAFP as before. We found that inhibition of cPLA₂α activity with MAFP could promote control of tumor burden by adoptively transferred T cells in larger B16F10 melanoma tumors (Fig. 8H). In addition to the murine tumor models, we investigated whether reprogramming of effector T cell lipid metabolism via cPLA₂α inhibition could enhance antitumor immunity in a human melanoma model in humanized non-obese diabetic–severe combined immunodeficient IL-2Rγ^null (NSG) mice (22, 65, 66). Human 586mel tumor cells were subcutaneously injected into NSG mice on day 0. Tumor-specific CD8⁺ TIL586 T cells (which recognize and kill autologous 586mel tumor cells) were adoptively transferred through intravenous injection on day 5, followed by intraperitoneal injection of MAFP. Consistent with our previous studies, we found that M586 melanomas grew progressively in NSG mice. Adoptive transfer of tumor-specific CD8⁺ TIL586 T cells inhibited tumor growth, which was further enhanced with MAFP treatment (Fig. 8I). In addition, MAFP injection decreased LD formation and prevented cell senescence in tumor-specific TIL586 T cells (Fig. 8, J and K). Collectively, these in vivo studies suggest that reprogramming of lipid metabolism in tumor-specific T cells via cPLA₂α inhibition can prevent T cell senescence and enhance antitumor immune responses, which is an effective strategy for tumor immunotherapy.

DISCUSSION

The suppressive tumor microenvironment maintains tumor-specific T cells in a dysfunctional state, which is a major obstacle for successful tumor immunotherapy (4). The tumor microenvironment can reprogram T cell metabolism, which directs T cell survival, proliferation, and function (4, 15, 16). Our current and previous studies have clearly demonstrated that T cell senescence induced by both tumor cells and T_reg cells is a general feature in different types of cancers (21, 22, 32–34). In this study, we further demonstrate that tumor and T_reg cells alter lipid metabolism in T cells, which induces T cell senescence and dysfunction. Molecularly, MAPK or STAT1 and STAT3 signaling pathways cooperate to promote cPLA₂α expression in responder T cells, resulting in alterations of lipid metabolism, LD accumulation, and development of a senescent phenotype. Using T cell adoptive transfer models, we further establish the concept that reprogramming lipid metabolism in T cells within the tumor microenvironment is a potential strategy for cancer immunotherapy.

Tumors can use various strategies to modify T cell glucose metabolism to inhibit antitumor immunity, including direct competition for glucose consumption, metabolite regulation, inhibition of Glut1 or glycolysis, and decrease in mitochondrial biogenesis and function (4, 12, 15–19, 23–25). However, the metabolic changes that tumor cells and T_reg cells induce in TILs have not been fully characterized. In the current study, we characterized the key metabolic genes involving glucose metabolism in TILs isolated from patients with cancer or from tumor-bearing mice, as well as responder T cells cultured with tumor cells and T_reg cells, using transcriptome and real-time quantitative PCR analyses. Our studies suggest that TILs from both human cancer tissues and from tumor-bearing mice have active glucose metabolism to maintain their biological and functional demands.

In addition to modulating glucose metabolism, we found that tumor cells and T_reg cells altered lipid metabolism in T cells. Recent studies have demonstrated that T cells with increased cholesterol esterification have impaired proliferation and antitumor effector function (52). Furthermore, increased cholesterol also induces exhaustion in TILs (28). Our current studies collectively indicate that tumor cells and T_reg cells can modify T cell lipid metabolism through different molecular processes. First, both tumor cells and T_reg cells induce changes in expression of the key enzymes important for cholesterol synthesis and FA oxidation and synthesis in senescent CD4⁺ and CD8⁺ T cells, as well as in TILs. Second, both tumor and T_reg cells alter the concentration of lipid species in responder T cells or TILs by increasing cholesterol, CE, and FFA and decreasing concentration of key phospholipids. These altered lipid species are critical for the development of T cell senescence. Our studies further demonstrate that the altered lipid species promote formation of LDs in responder T cells or TILs, which was strongly associated with the development of T cell senescence.

LDs are lipid-rich cytoplasmic organelles and are important for the regulation of inflammation and cancer (52–54). Recent studies have shown that LDs accumulate in tumor-derived DCs, which results in tolerogenic functions (30, 31). However, little information is known about whether LDs regulate T cell function in antitumor immunity. In this study, we identified accumulation of LDs in both TILs from human patients with cancer and tumor-bearing mice, as well as in senescent T cells induced by tumor and T_reg cells in vitro. We further identified the molecular mechanisms responsible for the increase in LDs in senescent T cells in the tumor microenvironment. Our studies suggest that ACAT does not regulate LD accumulation in senescent T cells. Rather, down-regulated LAL and increased cPLA₂α appear to control both LD formation and cell fate. Whereas cholesterol, CE, and TAG are all key components of LDs (53, 67), our current studies did not investigate the effects of TAG changes in inducing T cell senescence. In addition, we have not determined how loss of phospholipids affects in the development of T cell senescence. Future studies are warranted to further characterize regulators of lipid metabolism in the context of antitumor immunity.

Our studies provide a proof of concept that reprogramming T cell lipid metabolism to prevent senescence in effector T cells induced by T_reg cells and tumor cells may lead to effective strategies for tumor immunotherapy. We have identified that cPLA₂α is a key target regulating lipid metabolism, cell senescence, and function in T cells. Our in vivo studies demonstrate that reprogramming of T cell lipid metabolism via cPLA₂α inhibition can prevent effector T cell senescence and dysfunction, resulting in enhanced antitumor immunity and therapeutic efficacy of adoptively transferred T cells in melanoma and breast cancer models at both early and late stages. Our hypothesis and concept are supported by a recent study showing that modulation of CD8⁺ T cell cholesterol metabolism can enhance cancer immunotherapy (52). Our previous studies have demonstrated that activation of human Toll-like receptor 8 signaling in T_reg cells and multiple types of tumor cells can prevent effector T cell senescence and reverse the suppressive activity mediated by both T_reg cells and senescent T cells (22, 32, 33, 44, 65, 66). These strategies manipulating T cell metabolism and senescence, combined with the immune checkpoint therapies such as CTLA-4 or PD-1 blockade therapy, may yield more promising results for improved cancer immunotherapy (1, 2, 5).

Although our studies provide compelling evidence that activated lipid metabolism in T cells induced by T_reg cells and tumor cells is responsible for induction of senescence and dysfunction of T cells, there are several limitations. First, our studies have shown that senescent T cells induced by T_reg cells and tumor cells have increased gene expression of enzymes involved in glycolysis. However, we still do not understand how active glucose metabolism contributes to development of senescence. Furthermore, the relative contributions of lipid metabolism and glucose metabolism in inducing senescence in T cells are unclear. These questions are important for us to understand the functional state of T cells in patients with cancer, which will, in turn, provide insights for effective cancer immunotherapy. The other restriction of this study is the limited availability of tissue samples from patients with cancer used to purify TILs for metabolic analysis. Our current studies investigated gene expression of key enzymes involved in glycolysis and lipid metabolism in human TILs from patients with melanoma or breast cancer. However, the variability is high due to small sample sizes. More studies with human TILs directly purified from tumor tissues from different types of human cancers should be included in future studies to explore the generality of our findings. Although our studies have identified that MAPK P38/ERK and STAT signaling pathways are involved in regulation of accumulated lipids and the ATM-associated DNA damage response in senescent T cells, we did not perform in vivo studies to explore the translational potential of blocking these signaling pathways. Future studies should also focus on strategies targeting these signaling pathways in combination with immune checkpoint therapies for cancer treatment.

In summary, our current study provides evidence that development of senescence is an important dysfunctional state for T cells within the suppressive tumor microenvironment. T_reg cells and tumor cells reprogram lipid metabolism in responder T cells to promote this senescent phenotype, and this process is driven by elevated cPLA₂α. Further, altered lipid metabolism and senescence induction in T cells is molecularly controlled by both MAPK and STAT signaling (fig. S11). Thus, our studies identify potential targets and therapeutic strategies to reprogram lipid metabolism in T cells to enhance antitumor immunity.

MATERIALS AND METHODS

Study design

In this study, we used both in vitro studies with a cell coculture system and purified T cells from patients with cancer, as well as in vivo studies in different tumor models, to identify molecular mechanisms responsible for T cell senescence and to develop strategies to manipulate tumor-specific T cell metabolism and function for cancer immunotherapy. Transcriptome analyses of senescent CD8⁺ T cells induced by human T_reg cells were performed to identify potential molecular and metabolic signaling pathways involved in T cell senescence using the Illumina whole-genome HumanHT-12 BeadChips. Real-time quantitative PCR, Western blot, and flow cytometry analyses were used to determine gene and protein expression of enzymes involved in glucose and lipid metabolism. For T cell senescence analysis, SA-β-gal activity, CD27/CD28 expression, and suppressive activity were evaluated using SA-β-gal staining, flow cytometry analysis, and [³H]-thymidine incorporation assay, respectively. Lipidomic analysis was used to identify the changes of lipid species and metabolites in senescent T cells induced by T_reg and tumor cells. We also used the loss-of-function strategy with pharmacological inhibitors or gene knockdown with siRNA and gain-of-function strategy with signaling activators to dissect the functional role of the identified molecules in induction of senescence in T cells mediated by T_reg and tumor cells. Furthermore, both syngeneic mouse tumor models and humanized NSG mouse models with human T cells and tumor cells were used to test whether reprogramming lipid metabolism in T cells enhanced antitumor immunity. Five to 12 mice were included in each group for in vivo studies. In vitro coculture experiments had at least three biological replicates for each group. T cell lipidomic data were derived from T cells purified from two to four individual donors, except for inhibitor treatments, which were derived from T cells purified from two individual donors. Transcriptome analysis data were derived from T cells purified from two individual donors with multiple time points and technical replicates.

Human samples and cell lines

Tumor samples of melanoma and breast cancers were obtained from patients treated in the Department of Surgery at St. Louis University from 2004 to 2017 who have given informed consent for enrollment in a prospective tumor procurement protocol approved by the Saint Louis University Institutional Review Board (IRB no. 15283). Buffy coats from healthy donors were obtained from the Gulf Coast Regional Blood Center at Houston. Peripheral blood mononuclear cells (PBMCs) were purified from buffy coats using Ficoll-Paque. Human naïve CD4⁺ and CD8⁺ T cells were purified by EasySep enrichment kits (STEMCELL Technologies). Naturally occurring human CD4⁺CD25^hi T_reg cells (nT_reg cells) were purified from CD4⁺ T cells by FACS after staining with anti–CD25-PE (BD Biosciences) or isolated from PBMCs by negative selection with the human CD4⁺CD127^lowCD49d⁻ T_reg cell enrichment kit (STEMCELL Technologies) (32). Melanoma (B16F0 and B16F10) and breast cancer (MCF-7 and E0771) cell lines were purchased from the American Type Culture Collection. Human melanoma 586mel and paired TIL586 cells were obtained from the Surgery Branch of the National Cancer Institute. Human T cells were maintained in T cell medium containing 10% human AB serum, l-glutamine, 2-mercaptethanol, and recombinant human IL-2 (50 U/ml). Tumor cell lines were cultured in RPMI 1640 containing 10% fetal bovine serum.

SA-β-gal staining

SA-β-gal activity in senescent T cells was detected as previously described (21, 32, 33). Naive CD4⁺ or CD8⁺ T cells were labeled with carboxyfluorescein succinimidyl ester (CFSE; 4.5 μM) and cocultured with T_reg cells at a ratio of 4:1 in anti-CD3–coated 24-well plates for 3 days. Cocultured T cells were then separated using FACS gated on CFSE⁺ populations and then stained with SA-β-gal staining reagent (Sigma-Aldrich). For tumor cell–induced T cell senescence, anti-CD3–activated naïve CD4⁺ T cells or CD8⁺ T cells were cocultured with or without tumor cells at ratio of 1:1 for 1 day and then separated and cultured for additional 3 days. For irradiation-induced human T cell senescence studies, purified human CD4⁺ and CD8⁺ T cells from healthy donors were irradiated with 5- or 10-Gy doses using X-RAD 320 irradiator (Precision X-Ray) and then cultured for different times, as described previously (63). For indicated experiments, naïve T cells were pretreated with the following inhibitors and then cocultured with T_reg cells: ACAT inhibitor avasimibe (5 and 10 μM; Cayman Chemical), lipase inhibitor orlistat (10 μM; Cayman Chemical), cPLA₂α inhibitors MAFP (1 and 5 μM; Cayman Chemical) and ATK (1 and 5 μM; Cayman Chemical), ATM inhibitor KU55933 (10 μM; Tocris Bioscience), STAT1 inhibitor MTA (5 μM; Sigma-Aldrich), STAT3 inhibitor S3I-201 (10 μM; Sigma-Aldrich), and MAPK inhibitors U0126 (10 μM; Calbiochem) and SB203580 (10 μM; Calbiochem).

Flow cytometry analysis

T cell protein expression was determined by flow cytometry analysis after surface staining or intracellular staining with anti-human specific antibodies conjugated with PE or Alexa Fluor 488. For surface staining, T cells were stained with PE-conjugated mouse anti-human CD27 (clone M-T271, BD Biosciences; dilution, 1:100) or anti-human CD28 (clone CD28.2, BD Biosciences; dilution, 1:100) on ice for 30 min. For intracellular staining, T cells were fixed with 4% formaldehyde at room temperature for 15 min and followed to incubate in 90% methanol on ice for 30 min after washing with phosphate-buffered saline. The fixed T cells were incubated with indicated diluted primary antibodies at room temperature for 1 hour and then stained with goat anti-rabbit immunoglobulin G (IgG) conjugated with Alexa Fluor 488 (Cell Signaling Technology; dilution, 1:1000) at room temperature for 30 min. The primary human antibodies included the following: anti–phospho-ERK (Thr²⁰²/Tyr²⁰⁴; dilution, 1:200), anti–phospho-P38 (Thr¹⁸⁰/Tyr¹⁸², clone D3F9; dilution, 1:200), anti–phospho-STAT1 (Tyr⁷⁰¹, clone D4A7; dilution, 1:200), anti–phospho-STAT3 (Tyr⁷⁰⁵, clone D3A7; dilution, 1:200), anti–phospho-ATM (Ser¹⁹⁸¹, clone D25E5; dilution, 1:200), anti–phospho-H2AX (Ser¹³⁹/Tyr¹⁴²; dilution, 1:200), anti–phospho-C HK2 (Thr⁶⁸; dilution, 1:200), anti–phospho-53BP1 (Ser¹⁷⁷⁸; dilution, 1:100), and anti-cPLA₂ (clone D49A7; dilution, 1:100), which were purchased from Cell Signaling Technology. For BODIPY 493/503 staining, T cells were incubated with BODIPY 493/503 (1 μg/ml; Thermo Fisher Scientific) at room temperature for 15 min. All stained cells were analyzed on an LSR II cytometer (BD Biosciences), and data were analyzed with FlowJo software (BD Biosciences).

Western blotting analysis

CFSE-labeled naïve CD4⁺ or CD8⁺ T cells were cocultured with T_reg or control T cells at a ratio of 4:1 in anti-CD3 antibody-precoated plates (2 μg/ml) for 3 days. In indicated experiments, T cells were pretreated with the cPLA₂α inhibitor MAFP (5 μM) for 24 hours and then cocultured with T_reg cells. Treated T cells were then sorted by FACS gating on CFSE⁺ T cells and lysed in the buffer [50 mM tris-Cl (pH 8.0), 150 mM NaCl, 0.2% SDS, 1% NP-40, 1 mM NaF, 1 mM sodium orthovanadate, 0.5% sodium deoxycholate, phosphatase inhibitor PhosSTOP (Roche), and proteinase inhibitor (Sigma-Aldrich)]. T cell lysates were briefly sonicated for 30 s, followed by boiling for 10 min. Samples were loaded on 10% gradient tris-glycine SDS–polyacrylamide gels and separated by electrophoresis. Proteins were electroblotted onto polyvinylidene difluoride membranes and blocked in 5% milk at room temperature for 2 hours. The membranes with transferred proteins were incubated with primary antibodies at 4°C overnight, followed by incubation with a secondary antibody of horseradish peroxidase–conjugated anti-rabbit IgG (dilution, 1:10,000) at room temperature for 1 hour. Western blots were developed with chemiluminescent substrate (KPL). The rabbit polyclonal antibodies that used to stain Western blots include anti-cPLA2 (clone D49A7; dilution, 1:500), anti-P53 (clone 7F5; dilution, 1:1000), anti-P21 (clone 12D1; dilution, 1:1000), and anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH; dilution, 1:10,000). All antibodies were purchased from the Cell Signaling Technology.

Reverse transcription quantitative PCR analysis

Total RNA was extracted from T cells using TRIzol reagent (Invitrogen), and complementary DNA was transcribed using a SuperScript II Reverse Transcriptase kit (Invitrogen), both according to the manufacturers’ instructions. Expression of each gene was determined by reverse transcription PCR using specific primers, and the mRNA concentration in each sample was normalized to the expression of the housekeeping β-actin gene. All experiments were performed in triplicate. The specific primers used for T cells are listed in table S1.

Functional proliferation assay

Proliferation assays were performed using a [³H]-thymidine incorporation assay as previously described (21, 22, 32, 65, 66). Naïve CD4⁺ T cells (1 × 10⁵ per well) purified from healthy donors were cocultured with T_reg-treated T cells at a ratio of 10:1 in 200 μl of T cell assay medium containing 2% human AB and recombinant human IL-2 (20 U/ml). After 56 hours of culture, [³H]-thymidine was added at a final concentration of 1 μCi per well, followed by an additional 16 hours of culture. The incorporation of [³H]-thymidine was measured with a liquid scintillation counter.

Lipidomic analysis

Lipids from T cells were extracted by the method of Bligh and Dyer in the presence of internal standards including eicosanoic acid, 1–0-heptadecanoyl-LPC, dieicosanoyl-PC, ditetradecanoyl-PE, ditetradecanoyl-PS, N-heptadecanoyl-Cer, and cholesteryl heptadecanoate (68). Extracted lipids were resuspended in a methanol-chloroform solution (2:1, by volume) and analyzed by electrospray ionization mass spectrometry in direct infusion mode at a flow rate of 3 μl/min using a Thermo Fisher Scientific TSQ Quantum Ultra instrument. Samples were analyzed in both the positive and negative ion mode using a shotgun lipidomic approach (50). For LPC, neutral loss (NL) scanning of 59.1 was monitored in positive ion mode for sodiated molecular ions. NL scanning of 368.5 was performed for sodiated CE molecular species in positive ion mode (51). PE was derivatized to 9-fluorenyl methoxycarbonyl–PE species and monitored in negative ion mode using NL scanning of 222.2. NL scanning for Cer (NL, 256.2), PC (NL, 50), and PS (NL, 87) was performed in negative ion mode (50). FAs were converted to pentafluorobenzyl esters and quantified using negative ion chemical ionization detection and gas chromatography (69). Spectra were averaged over 3 to 5 min and processed using Xcalibur software (Thermo Fisher Scientific). Individual molecular species were quantified by comparing the ion intensity of individual molecular species to that of the appropriate internal standards after corrections for type I and type II 13C isotope effects (50). Additional corrections were made from response curves for CE molecular species (51). Each sample was normalized to cell number, and values are expressed per million cells.

Lipid supplement assay

LPC (16:0 LPC, 16:0 αLPC, and 16:0 pLPC), PE (16:0 lyso PE), and PS (16:0 lyso PS) were purchased from the Avanti Polar Lipids Inc. C2-Cer was purchased from Sigma-Aldrich. All lipids were dissolved in chloroform, evaporated under a gentle stream of nitrogen, and then immediately dissolved in an ethanol vehicle (0.1%). Naïve CD4⁺ T cells were first treated with or without the lipids (5 or 10 μM) for 1 hour and then cocultured with T_reg cells at a ratio of 4:1 in anti-CD3–coated (2 μg/ml) plates for 3 days. The treated naïve CD4⁺ T cells were then stained for SA-β-gal and Oil Red O.

Cholesterol content detection

Cellular cholesterol accumulation in different types of T cells was determined by a simple fluorometric method staining for Filipin III according to the manufacturer’s instructions (10009779, Cayman Chemical) (28). Briefly, T cells from mice or from in vitro cocultures were stained with Filipin III and then analyzed by fluorescence microscopy or flow cytometry.

Oil Red O staining

Treated T cells were washed in phosphate-buffered saline (pH 7.2), fixed with 4% formaldehyde for 30 min, and treated with 60% isopropanol for 2 to 5 min. Cells were further stained with freshly prepared Oil Red O staining solution (Sigma-Aldrich) in isopropanol (60%) for 5 min, followed by a rinse with isopropanol (60%). The stained cells were then washed with water thoroughly, and LDs in T cells were evaluated using a light microscope.

Indirect immunofluorescence staining

Naïve CD4⁺ T cells were cocultured with T_reg cells or control T cells in the presence of plate-bound anti-CD3 antibody (2 μg/ml) for 3 days. The cocultured T cells were incubated with primary rabbit anti-human antibodies at room temperature for 1 hour, including anti–phospho-ERK (Thr²⁰²/Tyr²⁰⁴; dilution, 1:100), anti–phospho-P38 (Thr¹⁸⁰/Tyr¹⁸², clone D3F9; dilution, 1:100), anti–phospho-STAT1 (Tyr⁷⁰¹, clone D4A7; dilution, 1:100), anti–phospho-STAT3 (Tyr⁷⁰⁵, clone D3A7; dilution, 1:100), anti-P53 (clone 7F5; dilution, 1:100), anti-P21 (clone 12D1; dilution, 1:100), and anti-cPLA2 (clone D49A7; dilution, 1:100) antibodies. All primary antibodies were purchased from Cell Signaling Technology. The treated CD4⁺ T cells were stained with BODIPY 493/503 (1 μg/ml) at room temperature for 15 min, then incubated with Alexa Fluor 594–conjugated anti-rabbit secondary antibody (Cell Signaling Technology; dilution, of 1:200) for 30 min, and further counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen).

Transcriptome analyses of senescent T cells

Anti-CD3–activated CD8⁺T cells were cocultured with medium only or CD4⁺CD25^hiFoxP3⁺ T_reg cells at a 5:1 ratio for indicated time points, including early (4 to 8 hours), middle (24 to 48 hours), and late (72 hours) stages. Total RNA was purified from the T_reg-treated and untreated naïve CD8⁺ T cells using an RNeasy kit (QIAGEN) per the manufacturer’s instructions. Transcriptome analyses of senescent CD8⁺ T cells induced by human T_reg cells were performed using Illumina whole-genome HumanHT-12 BeadChips as previously described (21, 32). Gene ontology terms associated with each gene were used to characterize the functionally related genes and identify processes associated with networks of differentially expressed genes. The normalized log₂ expression of each gene was calculated.

In vivo studies

C57BL/6 mice, NSG (stock no. 005557) mice, and Pmel-1 TCR/Thy1.1 transgenic mice on a C57BL/6 background (6- to 8-week-old female) were purchased from the Jackson Laboratory and maintained in the institutional animal facility. The facility is fully accredited by Association for Assessment and Accreditation of Laboratory Animal Care International and treated in accordance with principles outlined in the Guide for the Care and Use of Laboratory Animals. Animals were grouped and housed (five per cage or less) under specific pathogen–free conditions on a 12-hour light/12-hour dark cycle with free access to autoclave water and standard rodent diet. The caging and bedding are sterilized by autoclave, and food is irradiated. Mice were kept in a temperature-controlled (72°F) and humidity-controlled (30 to 70%) environment with air changes (10 to 13 per hour). All animal studies have been approved by the Institutional Animal Care Committee at Saint Louis University (protocol no. 2411).

To evaluate senescence in aged mice, 10- to 12-month-old C57BL/6 mice were used. Eight-week-old C57BL/6 mice were used as healthy adult mouse controls. CD4⁺ and CD8⁺ T cells were purified from the blood and spleens of aged and adult mice for subsequent SA-β-gal and Oil Red O staining and for related gene expression analyses.

To analyze T cell responses in the tumor microenvironment, mouse E0771 (2 × 10⁵ per mouse) breast tumor cells and melanoma cell line B16F0 (2 × 10⁵ per mouse) in a total volume of 100 μl of buffered saline were subcutaneously injected into the mammary fat pad and back of C57BL/6 mice, respectively. Tumor volumes were assessed every 3 days by external measurement of the length and width of the tumors in two dimensions with caliper measurement. When the tumor volumes reached the end-point criteria (diameter, 10 to 15 mm), mice were euthanized. Blood, lymph nodes, spleens, and tumor tissues were harvested, and CD4⁺ and CD8⁺ T cells were purified for subsequent SA-β-gal staining, Oil Red O staining, and metabolic gene profile analyses. In addition, CD4⁺ and CD8⁺ T cells from indicated organs of normal littermates were harvested and used as negative controls.

For adoptive transfer experiments, splenocytes from Pmel-1 TCR/Thy1.1 transgenic mice were prepared and activated in the presence of plate-coated anti-mouse CD3 (2 μg/ml) and anti-mouse CD28 (1 μg/ml) antibodies for 6 to 9 days. For the in vivo melanoma model, activated Pmel-1 T cells (1.7 × 10⁶ per mouse) in a total volume of 200 μl of buffered saline were adoptively transferred via tail vein into B16F10 tumor–bearing mice at day 6 (early-stage treatment) or day 11 (late-stage treatment) after tumor inoculation (2 × 10⁵ per mouse). For the E0771 breast cancer model, E0771 tumor cells were transduced with retroviral vector encoding tumor antigen gp100 and mIL-4R genes as previously described (70). E0771/gp-100 tumor cells were obtained after FACS based on mIL-4R expression. E0771/gp100 (2 × 10⁵ per mouse) cells were subcutaneously injected into the mammary fat pad of NSG mice. On day 6 after E0771/gp-100 tumor cell injection, activated Pmel-1 T cells (2 × 10⁶ per mouse) in a total volume of 200 μl of buffered saline were adoptively transferred via tail vein into E0771 tumor–bearing mice. In the B16F10 model, the tumor-bearing mice were irradiated with a non-myeloablative dose (500 cGy) to induce lymphopenia 1 day before T cell adoptive transfer. In parallel experiments, MAFP (7.5 or 15 mg/kg per mouse) or ATK (10 mg/kg per mouse) was injected intraperitoneally into the mice at 1 day after T cell transfer and then injected every 3 days for a total of four injections. In addition, recombinant human IL-2 (1 × 10⁵ IU daily) was also intraperitoneally administered for a total of 3 days after T cell transfer for all T cell adoptive transfer experiments. Five to 12 mice were included in each group. Tumor size was measured every 2 days with caliper, and tumor volume was calculated on the basis of two-dimensional measurements. Mouse survival was determined on the basis of the tumor sizes calculated according to a standard formula (length × width² × 0.52). The experiments were terminated for ethical considerations at the end point (tumor volume, >2000 mm³), and mouse survival was then determined. Blood, spleens, and tumors were harvested at the end of each experiment. The transferred CD8⁺ T cells from different organs, and tumor tissues were isolated by antibody-coated micro beads (STEMCELL Technologies) for subsequent SA-β-gal staining, Oil Red O staining, and gene expression profile analyses.

For human T cell adoptive transfer experiments, 586mel tumor cells (5 × 10⁶ per mouse) were subcutaneously injected in 100 μl of buffered saline into NSG mice. Tumor-specific CD8⁺ TIL586 cells (5 × 10⁶ per mouse) were intravenously injected on day 5 into tumor-bearing mice. In a parallel experiment, MAFP (7.5 mg/kg per mouse) was injected intraperitoneally into the mice at 1 day after adoptive transfer of CD8⁺ TIL586 cells and followed an identical treatment procedure and dosing as the above Pmel-1 T cell experiments. Five to six mice were included in each group. Tumor size was measured as the above B16F10 melanoma experiment, and the transferred human TIL586 CD8⁺ T cells were recovered for subsequent phenotypic and functional analyses in vitro as described above.

Statistical analysis

Statistical analysis was performed with GraphPad Prism5 software. Unless indicated otherwise, data are expressed as means ± SD. For multiple group comparison for in vivo studies, a one-way analysis of variance (ANOVA) was used, followed by the Dunnett’s test for comparing experimental groups against a single control. For single comparison between two groups, paired Student’s t tests were used. Nonparametric t test was chosen if the sample size was too small and did not fit Gaussian distribution. Differences in survival were determined on the basis of Kaplan-Meier survival analysis.

Data and materials availability:

The authors declare that the data supporting the findings of this study are available within the article and its supplementary information files. Microarray data that support the findings of this study have been deposited in Gene Expression Omnibus with the primary accession code GSE38765.