Impaired lipid biosynthesis hinders anti-tumor efficacy of intratumoral iNKT cells

Dysfunction of invariant natural killer T (iNKT) cells in tumor microenvironment hinders their anti-tumor efficacy, and the underlying mechanisms remain unclear. Here we report that iNKT cells increase lipid biosynthesis after activation, and that is promoted by PPARγ and PLZF synergically through enhancing transcription of Srebf1. Among those lipids, cholesterol is required for the optimal IFN-γ production from iNKT cells. Lactic acid in tumor microenvironment reduces expression of PPARγ in intratumoral iNKT cells and consequently diminishes their cholesterol synthesis and IFN-γ production. Importantly, PPARγ agonist pioglitazone, a thiazolidinedione drug for type 2 diabetes, successfully restores IFN-γ production in tumor-infiltrating iNKT cells from both human patients and mouse models. Combination of pioglitazone and alpha-galactosylceramide treatments significantly enhances iNKT cell-mediated anti-tumor immune responses and prolongs survival of tumor-bearing mice. Our studies provide a strategy to augment the anti-tumor efficacy of iNKT cell-based immunotherapies via promoting their lipid biosynthesis.

I mmunotherapies have been suggested as the most specific and effective treatments against tumors. Except for the cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells, invariant natural killer T (iNKT) cells play important roles in anti-tumor immune responses as well. They could either kill tumor cells directly through antigen recognition or augment anti-tumor responses by depleting tumor associated macrophages (TAMs) and promoting activation of NK and CTLs [1][2][3][4] . Therefore, iNKT cells are promising candidates for anti-tumor immunotherapies. Administration of alpha-galactosylceramide (αGC), a ligand for iNKT cells, or αGC-pulsed DCs inhibits tumor growth in both mouse models and human patients [5][6][7][8] . Moreover, recent studies report a notable anti-tumor efficacy of CAR iNKT cells 9,10 . However, not all patients respond to these iNKT cell-based immunotherapies 7,8 . Those failures are linked to impaired iNKT cell function. Although previous studies suggest that  , myeloid derived suppressor cells 14 , regulatory T cells 15,16 , exosomes 17 , metabolic products 18,19 , and suppressive cytokines 20 in tumors repress anti-tumor immunity and cause the failure of immunotherapies, the mechanisms underlying dysfunction of intratumoral iNKT cells still remain unclear. Elucidating the underlying mechanisms would help to design new immunotherapies augmenting the efficacy of iNKT cell-based immunotherapies.
Recent studies demonstrate a link between cellular metabolism and immune cell functions. Impaired metabolism has been reported as a main reason causing the dysfunction of anti-tumor immune cells in tumor microenvironment 21,22 . Whether metabolism regulates function of iNKT cells in tumors remains unclear. Here, we show that activation of iNKT cells increases their lipid biosynthesis, and that is promoted by PPARγ in cooperation with PLZF and is required for the optimal IFN-γ production. Lactic acid in tumor microenvironment reduces PPARγ expression in intratumoral iNKT cells and thereby diminishes their lipid synthesis and IFN-γ production. Importantly, activation of PPARγ via pioglitazone (PIO), a drug used to treat type 2 diabetes, restores lipid synthesis and IFN-γ production in those intratumoral iNKT cells, and significantly enhances anti-tumor efficacy of iNKT cell-based immunotherapy. Our results demonstrate the mechanisms underlying dysfunction of iNKT cells in tumor microenvironment, and provide important insights into interventions enhancing iNKT cell-mediated antitumor immune responses.

Results
PPARγ promotes lipid biosynthesis in iNKT cells. As the major component of cells, lipids have been reported to regulate the differentiation and function of immune cells. However, their roles in iNKT cells remain unclear. We detected higher level of PPARγ (Fig. 1a), a well-known regulator of adipocyte differentiation and lipid metabolism, in sorted mouse iNKT cells ( Supplementary  Fig. 1) in comparison to conventional CD8 + and CD4 + T cells. The higher expression of PPARγ in iNKT cells was further confirmed at protein level by flow cytometry and at mRNA level by qPCR (Fig. 1b, c). Notably, anti-CD3 plus anti-CD28 stimulation further increased PPARγ expression (Fig. 1d), as well as the lipid amount in iNKT cells (Fig. 1e) as indicated by the staining of 4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene (BODIPY). Next, we analyzed fatty acid composition in iNKT cells, and found that anti-CD3 plus anti-CD28 stimulation significantly increased amount of intracellular fatty acids, including C14:0, C16:0, C16:1, C18:1n9t, and C18:2n6t (Fig. 1f). PIO, the agonist of PPARγ, further enhanced the amount of these intracellular fatty acids in activated iNKT cells, whereas T007, the antagonist of PPARγ, significantly reduced the amount of these fatty acids (Fig. 1f). These results suggested that PPARγ activity controlled the amount of fatty acids in iNKT cells. In addition, we found that mRNA of genes regulating fatty acid biosynthesis were increased in iNKT cells after activation, and these mRNA levels were further enhanced by PIO and reduced by T007 (Fig. 1g). All these data demonstrate that PPARγ promotes fatty acid biosynthesis in iNKT cells after activation.
PPARγ-controlled lipid synthesis promotes IFN-γ production. Next, we investigated the functional impacts of PPARγ-controlled lipid synthesis in iNKT cells. We activated iNKT cells with either plate-coated anti-CD3 plus anti-CD28 antibodies or with platecoated mCD1d-PBS57 tetramers, and inhibited lipid synthesis with antagonist of PPARγ, T007, or with inhibitor of fatty acid synthesis, Tofa. After activation, iNKT cells proliferated and increased expression of activation markers CD69 and CD25. T007 and Tofa significantly reduced expression of CD69 and CD25 (Fig. 2a, b), and inhibited cell proliferation (Fig. 2c). Interestingly, T007 and Tofa dramatically inhibited IFN-γ production from sorted iNKT cells in a dose dependent manner, whereas showed only minor effect on their IL-4 production (Fig. 2d, e). These results indicated polarization of iNKT cell functions toward Th2 response. In addition to IFN-γ, reduced Th1 cytokines IL-2 and TNF-α were detected in iNKT cells treated with GW9662, another antagonist of PPARγ (Supplementary Fig. 2a, b). In agreement with its in vitro effects, GW9662 also inhibited αGC induced IFN-γ production in vivo. Frequency of IFN-γ + hepatic iNKT cells and mean fluorescence intensity of intracellular anti-IFN-γ were reduced by GW9662, whereas no effect of GW9662 on IL-4 production was detected ( Supplementary Fig. 2c, d). Moreover, we found that Ifng mRNA was dramatically reduced in iNKT cells treated with PPARγ antagonists, including GW9662 and T007, or with inhibitors of fatty acids synthesis, including Tofa and C75 (Fig. 2f, g). Again, these inhibitors showed minor effects on Il4 mRNA level (Fig. 2f, g). These results indicated that PPARγ-controlled lipid synthesis promoted IFN-γ production in iNKT cells at the transcriptional level. To further confirm the role of PPARγ in iNKT cells, we used shRNA to knock down its expression (Fig. 2h). Knockdown of PPARγ significantly reduced IFN-γ production (Fig. 2i). In addition, by crossing Pparg fl/fl mice with PLZF-cre mice, we deleted PPARγ in iNKT cells but not in conventional T cells (Fig. 2j). PPARγ deficiency reduced iNKT cell frequencies in thymuses but not in spleens or livers from PLZF-cre Pparg fl/fl mice ( Supplementary Fig. 3). In line with the knockdown of PPARγ, deletion of PPARγ in iNKT cells reduced their IFN-γ production when cells were activated in vitro (Fig. 2k). Moreover, we showed that PIO increased IFN-γ production and T007 reduced IFN-γ production in wide type iNKT cells but not in PPARγ deficient iNKT cells (Fig. 2k). These results further confirmed that PIO and T007 regulated IFN-γ production in iNKT cells by targeting PPARγ. Taken together, our results demonstrate that PPARγ promotes activation and IFN-γ production in iNKT cells via enhancing lipid synthesis.
Although fatty acid β-oxidation has been reported to regulate T cell fate 23,24 , transcription of Cpt1a, a gene controlling fatty acid β-oxidation, was reduced after activating iNKT cells. Inhibitor of CPT1α, Etomoxir, showed no influence on IFN-γ or IL-4 production from iNKT cells ( Supplementary Fig. 4). Therefore, fatty acid β-oxidation is not required for iNKT cell effector functions.
Cholesterol is required for IFN-γ production in iNKT cells. Previous findings have shown that certain of saturated fatty acids and unsaturated fatty acids play important roles in early activation of T cells 25,26 . It is possible that diminished IFN-γ production caused by those inhibitors results from shortage of certain lipid. To identify the lipid controlling IFN-γ production in iNKT cells, distinct fatty acids were, respectively, added to iNKT cells treated with Tofa. Both saturated and unsaturated fatty acids, including octanoic acid, decanoic acid, palmitic acid, oleic acid, linolenic acid, and linoleic acid, failed to restore IFN-γ production ( Fig. 3a). High concentration of palmitic acid, oleic acid, and linoleic acid even reduced IFN-γ production (Fig. 3a), possibly due to the increased cell apoptosis 27 . In contrast, supplementation of water-soluble cholesterol successfully restored IFN-γ production (Fig. 3b). In addition, cholesterol also restored IFN-γ production in iNKT cells treated with T007 (Fig. 3c). The recovery of IFN-γ by cholesterol was in line with the findings that T007 and Tofa reduced total cholesterol, as well as plasma membrane cholesterol in iNKT cells (Fig. 3d). These results indicated that inhibition of lipid synthesis diminished IFN-γ production by reducing cholesterol. Moreover, we showed that simvastatin (Sim), an inhibitor for cholesterol synthesis, significantly inhibited IFN-γ production in iNKT cells (Fig. 3e), whereas addition of cholesterol elevated their IFN-γ production (Fig. 3f). Again, the amount of cholesterol only slightly influenced IL-4 production (Fig. 3e, f). In line with the role of lipid synthesis in promoting iNKT cell activation and proliferation, we found that Sim inhibited upregulation of CD69 and iNKT cell proliferation ( Supplementary Fig. 5a, b). Taken together, these results demonstrate that reduction of cholesterol is responsible for the suppressive effect of impaired lipid synthesis. Cholesterol has been previously shown to regulate the TCR signaling [28][29][30] . In our studies, when iNKT cells were activated by PMA plus ionomycin bypassing TCR signaling, exogenous cholesterol, C75, and Tofa showed no influence on IFN-γ production (Fig. 3g). Therefore, cholesterol regulated IFN-γ production in iNKT cells likely by modulating proximal TCR signaling. Clustering and accumulation of TCR at immunological synapse were inhibited by T007, Tofa, and Sim, as indicated by decreased fluorescence intensity of TCR (Fig. 3h, i). The reduced area of immune synapses indicated impaired formation of synapse (Fig. 3j). To further investigate the activation of TCR signaling pathways by western blot, we used expanded iNKT cells which responded similarly to T007 as fresh iNKT cells did ( Supplementary Fig. 6). When expanded iNKT cells were activated and treated with T007, Tofa, and Sim, respectively, decreased phosphorylation of Lck and LAT were detected, confirming impaired proximal TCR signaling (Fig. 3k). Again, supplementation of water-soluble cholesterol successfully restored four independent experiments (f, g), or six mice (k), pooled from three to four independent experiments. Data were analyzed by unpaired Student's t-test (a-e) or Mann-Whitney test (f-i, k). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.
synaptic TCR accumulation, immune synapse formation, and phosphorylation of Lck and LAT ( Fig. 3h-k). These results demonstrate that cholesterol enhances TCR signaling at immunological synapse of iNKT cells.
PPARγ2 and PLZF synergically promotes Srebf1 transcription. PPARγ has been previously shown to promote fatty acid uptake in CD4 + T cells 25 . However, antagonists of PPARγ reduced genes controlling cholesterol synthesis, including Srebf1, Hmgcr, and Sqle (Fig. 1g), but showed no influence on genes controlling cholesterol efflux or uptake, including Abca1, Abcg1, Ldlr, and Ldol ( Supplementary Fig. 7). Among those genes controlled by PPARγ, Srebf1 encodes sterol regulatory element-binding protein 1 (SREBP1), a major transcription factor regulating the biosynthesis of lipids 31 . In agreement with the amount of Srebf1 mRNA (Fig. 1g), SREBP1 protein level was increased after cell activation and was reduced by T007, in both mature and immature forms (Fig. 4a). The reduction of total SREBP1 protein in presence of T007 was further confirmed by flow cytometry analysis (Fig. 4b). On the other hand, both Srebf2 mRNA level (Fig. 1g) and SREBP2 protein level were not influenced by T007 (Fig. 4a). Knockdown of PPARγ significantly reduced expression of SREBP1 in activated iNKT cells (Fig. 4c), confirming that PPARγ promoted expression of SREBP1 in iNKT cells after activation. To demonstrate that reduction of SREBP1 could inhibit cholesterol synthesis in iNKT cells, we knocked down SREBP1 by shRNA (Fig. 4c) and measured the cholesterol level by Filipin III staining. Knockdown of SREBP1 decreased cholesterol level as efficiently as knockdown of PPARγ (Fig. 4d). In agreement with the roles of cholesterol in promoting IFN-γγ production, knockdown of SREBP1 inhibited IFN-γ production as well (Fig. 4e). Consistently, we found reduced SREBP1 and cholesterol level in PPARγ deficient iNKT cells (Fig. 4f, g). Together, these results suggest that PPARγ enhances cholesterol synthesis and IFN-γ production in iNKT cells via promoting expression of SREBP1.
To investigate whether PPARγ could directly regulate transcription of Srebf1, PPARγ-binding DNA fragments from iNKT cells were collected by chromatin immunoprecipitation, and different sets of primers covering the −2000 bp to +1 bp region of Srebf1 promoter were used to scan the potential binding sites of PPARγ (Fig. 4h). A pair of primers targeting the known PPAR response element (PPRE) at Nr1h3 promoter region 32 was used as the positive control. Increased binding of PPARγ to fragment P3 (−359 to −477 bp), P4 (−613 to −730 bp), and P7 (−1338 to −1445 bp) were detected in activated iNKT cells (Fig. 4i). Moreover, we found that T007 completely inhibited binding of PPARγ to these fragments (Fig. 4i). These results indicate direct binding of PPARγ at promoter region of Srebf1 in activated iNKT cells.
Notably, antagonists of PPARγ did not influence the amount of mature or immature forms of SREBP1 or total SREBP1 protein in CD4 + T and CD8 + T cells after activation (Supplementary Fig. 8a-d). Consistently, binding of PPARγ to Srebf1 promoter was not influenced by the activation status or by PPARγ antagonist in conventional T cells ( Supplementary Fig. 8e, f). As a result, T007 and GW9662 showed no influence on cholesterol synthesis in these conventional T cells (Supplementary Fig. 8g, h). These results indicate that PPARγ regulates expression of SREBP1 in a cell type specific manner. PLZF is a cell type-specific master regulator that controls development and effector functions of iNKT cells, and is not expressed by conventional T cells 33 . We detected PLZF among the proteins co-immunoprecipitated with PPARγ. Activation of iNKT cells further promoted the binding of PPARγ to PLZF, though antagonist and agonist of PPARγ did not influence the association of these two proteins (Fig. 4j). Moreover, increased co-localization between PLZF and PPARγ was observed after activating iNKT cells (Fig. 4k, l). Again, PIO and T007 showed no influence on the co-localization coefficient. These results indicate that interaction between PPARγ and PLZF is controlled by activation signal other than ligands of PPARγ. Importantly, the interaction between PPARγ and PLZF implied a possible explanation for the cell type-specific regulatory role of PPARγ. Next, we investigated whether PLZF was required for PPARγ to promote Srebf1 transcription. There are two isoforms of PPARγ, PPARγ1, and PPARγ2. We detected both Pparg1 and Pparg2 mRNA in iNKT cells (Fig. 4m). The antibody against PPARγ used in our studies was unable to distinguish PPARγ1 from PPARγ2. PPARγ and PLZF are conserved between mouse and human (96% and 97% homology). Therefore, we overexpressed mouse PPARγ1 and PPARγ2, respectively, in human cell line 293T cells, with or without co-overexpressing mouse PLZF. Overexpressing PLZF, PPARγ1, or PPARγ2 alone failed to promote the transcription and expression of SREBP1 in 293T cells. Co-overexpression of PPARγ1 and PLZF also failed to do so. Interestingly, co-overexpressing PPARγ2 and PLZF significantly increased amount of mature SREBP1, total SREBP1 protein, and Srebf1 mRNA (Fig. 4n-p). These results demonstrated that PPARγ2 cooperated with PLZF to promote the transcription and expression of SREBP1. Consistently, co-overexpressing PPARγ2 and PLZF in iNKT cells increased SREBP1 expression as well, whereas co-overexpressing PPARγ1 and PLZF did not (Fig. 4q). Comparing to PPARγ1, PPARγ2 has additional 30 amino acids at N-terminal. Coimmunoprecipitation experiments demonstrated that overexpressed PPARγ2 rather than PPARγ1 bound PLZF in 293T cells (Fig. 4r). Therefore, the additional 30 amino acids in PPARγ2 were important for its interaction with PLZF. These results also explained the findings that some PPARγ protein, likely PPARγ1, was not co-localized with PLZF in iNKT cells (Fig. 4k), and PPARγ2 but not PPARγ1 cooperated with PLZF in promoting Srebf1 transcription (Fig. 4n-q). Furthermore, we found that co-expression of PLZF significantly enhanced binding of PPARγ2 to the Srebf1 promoter region in 293T cells (Fig. 4s). Together, our results demonstrate that PLZF is required for PPARγ2 to bind at the promoter region of Srebf1 and to promote its transcription.
In line with our conclusion, iNKT2 and iNKT17 subsets that expressed higher amount of PLZF and PPARγ showed higher amount of SREBP1 and cholesterol in comparison to iNKT1 subset ( Supplementary Fig. 9b, c). Similar to conventional T cells, adipose tissue iNKT cells, which expressed PPARγ but negligible PLZF (Supplementary Fig. 9d) 34 , did not change their amount of SREBP1 or cholesterol in response to PIO or T007 treatment ( Supplementary Fig. 9e).  3 Cholesterol restores IFNγ production in iNKT cells lacking lipid synthesis. a Influences of indicated fatty acids on IFN-γ production in iNKT cells activated by plate-coated mCD1d-PBS57 tetramer in the presence of Tofa. b Influences of water-soluble cholesterol on IFN-γ and IL-4 production from iNKT cells activated by plate-coated mCD1d-PBS57 tetramer in the presence of Tofa. c Influences of water-soluble cholesterol on IFN-γ and IL-4 production in iNKT cells activated by plate-coated mCD1d-PBS57 tetramer in the presence of T007. d Total cholesterol or plasma membrane cholesterol in iNKT cells activated by anti-CD3 plus anti-CD28 for 24 h with or without T007, or Tofa. e Influences of Sim on IFN-γ and IL-4 production in iNKT cells activated by mCD1d-PBS57 tetramer. f Influences of water-soluble cholesterol on IFN-γ and IL-4 production in iNKT cells activated by mCD1d-PBS57 tetramer. g Influences of Tofa, C75, and cholesterol on IFN-γ and IL-4 production from iNKT cells activated by PMA plus ionomycin. h-j Distribution (h) and fluorescence intensity (i) of surface TCR, and area of immunological synapse (j) in iNKT cells activated by coverslip-coated mCD1d-PBS57 tetramer for 30 min in the presence or absence of T007, Tofa, Sim, or water-soluble cholesterol (n = 100 cells per group). Arbitary unit, A.U. Bar, 5 μm. k P-LCK (Y394), LCK, P-LAT (Y191), LAT in iNKT cells activated by plate-coated anti-CD3 plus anti-CD28 for 20 min in the presence or absence of T007, Tofa, Sim, or water-soluble cholesterol. Data are means ± SEM of three independent experiments (a, d, g) or nine biological replicates (b, c, e, f), pooled from three independent experiments, or are representative of three independent experiments (h, k). Data were analyzed by Mann-Whitney test (a, d, g) or unpaired Student's t-test (b, c, e, f, i, j). *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.

Reduced PPARγ and IFN-γ in human intratumoral iNKT cells.
Previous findings indicate that altered cellular metabolism influences immune cell functions in tumor microenvironment. The metabolic state and effector functions of iNKT cells in tumor microenvironment remain unclear. In hepatocellular carcinoma (HCC) patients, reduced PPARγ was detected in human tumor-infiltrating iNKT cells, in comparison to cells in para-carcinoma tissues (Fig. 5a). In agreement with our findings that PPARγ promoted expression of SREBP1 and synthesis of cholesterol in mouse iNKT cells (Fig. 4a-d, f, g), reduced SREBP1 (Fig. 5b) and cholesterol levels (Fig. 5c) were detected in human tumorinfiltrating iNKT cells, and the level of SREBP1 and cholesterol γ . V .  were positively correlated with the level of PPARγ in cells from tumors and para-carcinoma tissues (Fig. 5d, e). Furthermore, in line with the important roles of PPARγ and cholesterol in promoting mouse iNKT cell function (Fig. 2d, i, j, 3e, f), intratumoral human iNKT cells significantly reduced IFN-γ production, as indicated by frequencies of IFN-γ + iNKT cells and mean fluorescence intensity of intracellular anti-IFN-γ (Fig. 5f). To demonstrate that the dysfunction of intratumoral iNKT cells was due to impaired cholesterol synthesis, we added water-soluble cholesterol to tumor-infiltrating iNKT cells and found that it restored IFN-γ production successfully (Fig. 5g). Consistently, PIO, the agonist of PPARγ, successfully restored SREBP1 expression, cholesterol synthesis, and IFN-γ production in human tumor-infiltrating iNKT cells (Fig. 5h-j). These results demonstrate that reduced PPARγ causes impaired cholesterol synthesis and dysfunction of human tumor-infiltrating iNKT cells, and that could be restored by PIO.

PIO restores function of mouse intratumoral iNKT cells.
To further elucidate the role of PPARγ in regulating iNKT cell function in tumors, mice subcutaneously inoculated with B16F10 melanoma cells were used in our studies. In comparison with splenic iNKT cells, tumor infiltrating-iNKT cells reduced IFN-γ (Fig. 6a), PPARγ (Fig. 6b), SREBP1 (Fig. 6c), and cholesterol (Fig. 6d). These results were in line with those findings in human intratumoral iNKT cells (Fig. 5a-f). Tumor microenvironment is enriched with lactic acid, which has been shown previously to inhibit IFN-γ production from iNKT cells 35 . Here, we found that lactic acid significantly reduced PPARγ, and hence inhibited expression of SREBP1 and synthesis of cholesterol in iNKT cells ( Supplementary Fig. 10a). In agreement with the previous findings 35 , reduced mTOR protein and impaired activation of mTORC1 were detected in iNKT cells treated with lactic acid (Supplementary Fig. 10b). mTORC1 pathway has been reported to promote transcription of Pparg in CD4 T cells 25 . Using rapamycin to inhibit mTORC1 pathway, we showed that activation of mTORC1 was required for expression of PPARγ and hence for SREBP1 expression and cholesterol synthesis in iNKT cells (Supplementary Fig. 10c). These results indicate that reduction of PPARγ in tumor-infiltrating iNKT cells is caused by lactic acid in tumor microenvironment. To test whether PIO could restore intratumoral iNKT cell function in vivo, B16F10 melanoma-bearing mice received PIO daily for 7 days via oral gavage, and then the mice were administered with iNKT cell-specific antigen αGC intraperitoneally (Fig. 6e). PIO elevated SREBP1 expression and cholesterol synthesis in tumor infiltrating-iNKT cells after activation (Fig. 6f, g). Consequently, PIO treatment significantly increased αGC-induced IFN-γ production in those intratumoral iNKT cells (Fig. 6h). Moreover, we confirmed that PIO treatment promoted synaptic TCR accumulation and synapse formation of tumor-infiltrating iNKT cells in vitro (Fig. 6i-k).
PIO promotes iNKT cell-mediated anti-tumor immune responses. It is well known that activation of iNKT cells by αGC or αGC-pulsed DCs inhibits tumor growth and provides survival benefit in mouse and in human patients [5][6][7][8] . The anti-tumor efficacy of iNKT cells is correlated to their IFN-γ production 3,36,37 . Although PIO did not completely restore the lipid synthesis and IFN-γ production in intratumoral iNKT cells (Fig. 6h), augmentation of IFN-γ in tumors indicated a better anti-tumor response. In agreement with the restoration of IFN-γ production in tumor-infiltrating iNKT cells (Fig. 6h), PIO treatment significantly promoted efficacy of iNKT cell-based immunotherapy against B16F10 melanoma cells (Fig. 7a-d). PIO or αGC alone only modestly reduced tumor size and prolonged survival of tumor bearing mice, whereas combination of PIO and αGC did so strongly (Fig. 7b-d). As important immune regulatory cells, iNKT cells promote downstream activation of NK and CD8 + T cells indirectly 5,38,39 . Consistently, PIO and αGC in combination rather than PIO or αGC alone caused higher frequencies of intratumoral NK and CD8 + T cells (Fig. 7e), and induced more IFN-γ production from those cells (Fig. 7f). These results indicated that PIO plus αGC enhanced the activities of anti-tumor networks downstream iNKT cells. Conversely, in iNKT cell deficient Jα18 −/− mice and in Ifng −/− mice, combination of PIO and αGC showed no influence on the tumor size, frequencies of intratumoral NK cells and CD8 + T cells, or their IFN-γ production (Fig. 7g-k). These results further confirmed that PIO plus αGC-augmented anti-tumor immune responses were dependent on iNKT cells and IFN-γ production. To demonstrate that PIO enhanced iNKT cell-mediated anti-tumor immune responses in vivo by targeting at PPARγ in iNKT cells, we generated mixed bone marrow chimeric mice to delete PPARγ specifically in iNKT cells. Mixed bone marrow cells from Jα18 −/− mice and pLck-cre Pparg fl/fl mice (1:1 ratio) were injected intravenously into irradiated mice to create iNKT cell-specific deletion of PPARγ (Fig. 7l, m). As wide type controls, irradiated mice received mixed bone marrow cells from Jα18 −/− mice and Pparg fl/fl mice (1:1 ratio). In line with previous findings that PLZF-cre Pparg fl/fl mice had normal numbers of peripheral iNKT cells, deficiency of PPARγ specifically in iNKT cells did not influence the intratumoral iNKT cell numbers in chimeras (Fig. 7n). However, in chimeras, deficiency of PPARγ in intratumoral iNKT cells reduced their IFN-γ production after PIO plus αGC treatment (Fig. 7o). Notably, chimeric mice with iNKT cell-specific deletion of PPARγ exhibited larger tumor size and lower survival than their wide type controls in response to PIO plus αGC treatments (Fig. 7p, q). These results further confirmed that PPARγ in iNKT cells was essential for the increased antitumor efficacy of PIO plus αGC treatment. In addition, combination of PIO and αGC enhanced efficacy of iNKT cell-based immunotherapy in MC38 colorectal cancer mouse model as well. PIO and αGC in combination but not alone significantly reduced tumor size and prolonged survival of MC38-bearing mice (Supplementary Fig. 11a, b). Taken together, PIO restores lipid synthesis and IFN-γ production in tumor-infiltrating iNKT cells and enhances the iNKT cell-mediated anti-tumor immune responses.

Discussion
Lipid metabolism has been shown to regulate differentiation and functions of immune cells. Inhibition of ACC1, a key enzyme for de novo synthesis of fatty acids, favors development of regulatory T cells but inhibits Th17 differentiation 40 . Although expansion of CD8 + T cells depends on lipid synthesis, development of memory CD8 + T cells requires lipolysis to support their fatty acid β-oxidation 23,41 . Here, we report an important role of lipid synthesis, especially cholesterol synthesis, in promoting activation, proliferation, and effector function of iNKT cells. On the other hand, inhibition of fatty acid β-oxidation showed no influence on iNKT cell function (Supplementary Fig. 4). Although iNKT cells release both Th1 and Th2 cytokines upon activation, their IFN-γ production requires longer TCR signaling than IL-4 production 42,43 . Lack of cholesterol inhibited TCR signaling transduction (Fig. 3h-k), and thereby significantly diminished IFN-γ production whereas only showed minor effect on IL-4 (Fig. 3e). The polarization of iNKT cell function toward Th2 favors tumor growth rather than inhibits tumor growth. Therefore, alteration of lipid metabolism in iNKT cells would influence the progression of diseases. Notably, the roles of cholesterol in regulating T cell functions are not consistent 44,45 . Our results are in agreement with previous findings that cholesterol in cell membrane promotes TCR signaling in CD8 + T cells 30 .
A recent study indicates that PPARγ promotes early T cell activation via enhancing fatty acid uptake 25 . However, PPARγ regulated biosynthesis rather than uptake of cholesterol in iNKT cells (Supplementary Fig. 7). It is well known that synthesis of cholesterol is controlled by SREBP1 [46][47][48] . Interestingly, transcription of Srebf1 was regulated by PPARγ only in the presence of PLZF (Fig. 4n-p). In conventional T cells and adipose tissue iNKT cells which lack of PLZF expression, PPARγ failed to promote SREBP1 expression ( Supplementary Fig. 8a-d, 9e). These results are in line with previous findings that TCR-mTOR signaling upregulates PPARγ but shows no influence on SREBP1 in CD4 + T cells 25 . The cell type-specific role of PPARγ is mediated by its cooperation with PLZF, a master transcriptional regulator in iNKT cells. Similar cell type-specific role of PPARγ has been  reported previously in adipose tissue Tregs via cooperation with Foxp3 49 . Notably, transcription of Srebf1 was promoted by PPARγ2 rather than by PPARγ1 (Fig. 4n-q). These results were further explained by the findings that PPARγ2 but not PPARγ1 interacted with PLZF (Fig. 4r). Although it is still unclear whether PPARγ2 binds PLZF directly or indirectly, the extra 30 amino acids at N terminus of PPARγ2 are required for its interaction with PLZF. Moreover, we showed that PLZF promoted binding of PPARγ2 to Srebf1 promoter region (Fig. 4s). Further studies are required to characterize the PPREs at Srebf1 promoter region and to demonstrate the mechanisms by which PLZF interacts with PPARγ2 and promotes its binding to those PPREs. Our data Although PPARγ does not regulate lipid synthesis in CD4 + T cells, it has been reported to promote lipids uptake and cell proliferation in CD4 + T cells 25 . In addition, it has been shown to promote differentiation and effector function of Th2 cells, whereas shows no influence on Th1 cells 50 . Conversely, suppressive effects of PPARγ on Th1 cytokine production 51,52 , Th2 cytokine production, and T cell proliferation 53,54 are reported as well. The mechanisms underlying these discrepancies remain unclear, and cell typespecific regulators might be involved. iNKT cells play important roles in tumor clearance. However, tumor microenvironment influences their cellular metabolism and hinders their anti-tumor functions. Reduction of PPARγ and IFN-γ was observed in tumor-infiltrating iNKT cells in both human patients and mouse models (Figs. 5a-f, 6a-d). Notably, impaired lipid synthesis is not the only reason dampening antitumor effects of iNKT cells. Low pH in microenvironment and impaired glycolysis would contribute to dysfunction of intratumoral iNKT cells as well 35,43 . Here, we reported that lactic acid downregulated PPARγ and inhibited cholesterol synthesis in iNKT cells (Supplementary Fig. 10a). Cholesterol only modulated proximal TCR signaling (Fig. 3g-k), whereas lactic acid inhibited IFN-γ production even TCR signaling was bypassed 35 . Therefore, PPARγ-cholesterol pathway is not the only pathway mediating suppressive effect of lactic acid. For these reasons, PIO only restored IFN-γ production in intratumoral iNKT cells to some extent in vivo (Fig. 6h). Nevertheless, it significantly promoted anti-tumor efficacy of iNKT cell-based immunotherapy in animal models (Fig. 7b-d). Here, PIO alone only slightly increased frequencies of intratumoral NK and CD8 + T cells and showed no effect on their IFN-γ production (Fig. 7e, f). Significantly elevated NK and CD8 + T cell numbers and their IFN-γ production were only observed in mice receiving both PIO and αGC (Fig. 7e, f). Moreover, in the absence of iNKT cells, combination of PIO and αGC failed to promote accumulation and activation of NK and CD8 + T cells in tumors (Fig. 7g-i). Therefore, NK and CD8 + T cell-mediated anti-tumor immune responses were promoted by iNKT cells. It has been reported that activation of iNKT cells indirectly promotes downstream activation of NK and CD8 + T cell. After injecting αGC, iNKT cells release IFN-γ, activate DCs reciprocally, and induce IL-12 production from DCs. Sequentially, matured DCs activate NK and CD8 + T cells, and IFN-γ and IL-12 further promote the activation of NK and CD8 + T cells 2,38,55,56 . Therefore, activation of iNKT cells efficiently enhances multiple anti-tumor immune responses. These direct and indirect tumor-killing manners of iNKT cells make them ideal candidates for anti-tumor immunotherapies. In Ifng −/− mice, PIO and αGC in combination did not reduce tumor size or increase the intratumoral NK and CD8 + T cells (Fig. 7j, k). These results further confirmed that IFN-γ was the key factor mediating iNKT cell-based immunotherapy against tumors. Our results could not exclude the possibility that PIO may also directly influence the function of NK and CD8 + T cells when they are transactivated downstream iNKT cells. In addition, previous studies demonstrate a role of PPARγ in promoting CD1d and Cathepsin D expression in DCs 57,58 . Here, the contribution of PIO on lipid antigen presentation could not be excluded as well. However, with mixed bone marrow chimeric mice in which PPARγ was specifically deleted in iNKT cells, we confirmed that PPARγ expression in iNKT cells was required for the anti-tumor efficacy of PIO and αGC treatment (Fig. 7o-q). These results demonstrate that PIO targets PPARγ in iNKT cells and enhances their anti-tumor efficacy.
Together, we demonstrate that impaired lipid synthesis hinders anti-tumor effect of tumor-infiltrating iNKT cells. Restoring lipid synthesis via activating PPARγ recovers αGC-induced IFN-γ production and significantly enhances the efficacy of iNKT cellbased immunotherapy against tumors. Our results exhibit a promising prospect for combinational therapies targeting iNKT cells and their cellular metabolism. Importantly, PIO has been already used to treat type 2 diabetes, and that warrants its potential application for anti-tumor treatment in clinic.

Methods
Mice. Vα14 Tg.cxcr6 gfp/+ mice were previously described 59 and provided by Dr. Albert Bendelac. All mice used were on the C57BL/6J background and were 8 to 12 weeks old. Cohoused littermates were used in our experiments. Mice were housed under specific pathogen-free conditions. All animal procedures were approved by the Animal Care and Use Committee of University of Science and Technology of China, and all experiments were performed in accordance with the approved guidelines.
Clinical biospecimens. Clinical tumor biospecimens were obtained from HCC patients in accordance with protocols approved by the Biomedical Ethics Committee of Anhui Medical University. Informed consent was obtained from all subjects.
respectively. To study the influences of cholesterol on TCR signaling and iNKT cell function, water-soluble cholesterol was added to cells at 1 μM. To investigate the effects of GW9662 on iNKT cells in vivo, mice were administered intraperitoneally with 2 mg kg −1 GW9662 or PBS buffer 20 h before injecting 2 μg αGC. Mice were sacrificed 4 h later, and intracellular IL-4 and IFN-γ in hepatic iNKT cells were measured by flow cytometry.
To isolate tumor-infiltrating lymphocytes from tumor-bearing mice or HCC patients, tumors were cut into small pieces and digested with collagenase IV (1 mg ml −1 ) for 60 min at 37°C, and followed by centrifugation in Percoll density gradient. Tumor-infiltrating iNKT cells from tumor-bearing mice were enriched with mCD1d-PBS57 tetramer, and were stimulated with plate-coated mCD1d-PBS57 tetramer for cytokine measurement. Tumor-infiltrating iNKT cells isolated from HCC patients were stimulated with plate-coated hCD1d-PBS57 tetramer (0.5 μg ml −1 ) for 5 h in vitro, and intracellular IFNγ was measured by flow cytometry. Quantification of long-chain fatty acids. iNKT cells (1 × 10 7 ) were sorted and stimulated with plate-coated anti-CD3 plus anti-CD28 for 24 h with or without T007, or PIO. Lipids were extracted to analyze fatty acid composition as described 26 . In brief, C-17 heptadecanoic acid was added to cell pellets as an internal standard for fatty acids. Cells were dried and lipids extracted using chloroform: methanol. The lipid extract was saponified and fatty acids were derivatized to methyl ester forms with methanolic boron trifluoride, extracted into hexane, and then injected into an Agilent 7890B/5977 gas chromatography-mass spectrometry (GC-MS) fitted with an Agilent DB-225 column.
RNA isolation and qPCR. Total RNA was isolated with the ReliaPrepTM RNA Cell Miniprep System (Promega Z6010) and was reverse transcribed with Reverse Transcription System (Promega A3500). Real-time qPCR was performed with PikoReal 96 (Thermo Scientific). In each experiment, samples were run in triplicate and were normalized to β-actin to determine relative expression levels. Primer sequences were listed in Supplementary Methods.
Measurement of the cholesterol. Filipin III was dissolved in DMSO to reach the final concentration of 25 mg ml −1 . Cells were fixed with 4% paraformaldehyde (PFA) and stained with 50 μg ml −1 filipin III for 60 min at 4°C.
In some experiments, the total cellular cholesterol level was quantified using the Amplex Red cholesterol assay kit (Invitrogen). To quantify the intracellular cholesterol, cells were fixed with 0.1% glutaraldehyde and then were treated with 8 U ml −1 cholesterol oxidase for 30 min to oxidize the plasma membrane cholesterol. The intracellular cholesterol was then quantified using the Amplex Red cholesterol assay kit. The value of the plasma membrane cholesterol was obtained by subtracting the intracellular cholesterol from the total cellular cholesterol.
Total internal reflection fluorescence microscopy. Eighteen millimeter round coverslips (Thermo) were cleaned and coated with poly-L-lysine for 15 min. mCD1d-PBS57 tetramer (0.5 μg ml −1 ) or anti-CD3 (5 μg ml −1 ) plus anti-CD28 (5 μg ml −1 ) were used to coat the coverslips for 60 min at 37°C. iNKT cells were seeded onto coverslips for 30 min with or without T007, Tofa, or simvastatin. To detect surface TCR distribution, cells were stained with Alexa-647-labeled anti-TCRβ (5 μg ml −1 ) for 60 min on ice, and then were fixed in 4% paraformaldehyde for 15 min on ice. To view the location of PPARγ and PLZF, iNKT cells were fixed with 4% paraformaldehyde and permeabilized with buffer containing 5% FBS and 0.1% Triton X-100 for 30 min on ice, and then cells were stained with PE labeled anti-PLZF and FITC labeled anti-PPARγ for 60 min. Samples were visualized with a Leica SR GSD microscope at 100 nm penetration depth. Quantitative analyses were performed using ImageJ software.
Chromatin immunoprecipitation (ChIP)-qPCR. Expanded iNKT cells were activated by immobilized anti-CD3 plus anti-CD28 overnight in the presence or absence of T007. Cells were harvested and cross-linked with 1% formaldehyde at room temperature for 15 min, followed by neutralization with 125 mM glycine. Digestion and isolation of protein-bound DNA was performed using a magnetic ChIP kit (Thermo scientific). Antibody against PPARγ (Abcam ab41928, 1:100 dilution) and IgG control antibody (Santa Cruz sc-2025, 1:100 dilution) were used for ChIP. To detect the binding sites of PPARγ at srebp1 promoter region, different sets of primers covering the −2000 bp to +1 bp region of Srebf1 promoter were used for qPCR experiments, and were listed in Supplementary Methods. The fold enrichment is normalized to IgG control.
Mouse tumor models. In B16F10 melanoma models, B16F10 cells (1 × 10 5 ) were subcutaneously injected into the oxter part of mice. From day 6, tumor size was measured every 2 days, and animal survival rate was recorded every day. In MC38 models, MC38 cells (1 × 10 6 ) were subcutaneously injected into the oxter part of mice. From day 8, tumor size was measured every 2 days, and animal survival rate was recorded every day. Tumor size was calculated as length × width. Mice with tumor size larger than 20 mm at the longest axis were euthanized for ethical consideration. From day 6 or 8, tumor-bearing mice were randomly divided into four groups and were intragastrically administered with PBS buffer or with PIO daily at the dose of 30 mg kg −1 . On day 14, 2 μg αGC was injected intraperitoneally. To analyze the functions of intratumoral iNKT cells, NK cells, and CD8 + T cells, mice were euthanized 5 h after αGC injection, and IFNγ production were measured by flow cytometry using cytokine secretion assay kit (Miltenyi Biotec). To visualize the melanoma, mice were euthanized on day 20.
Statistical analysis. Statistical analyses were performed with paired or unpaired Student's t-test, Mann-Whitney test, Pearson's correlation test, two-way analysis of variance (ANOVA), or log-rank test. Particularly, the paired t-test is used for normally distributed parameters in two paired samples, unpaired Student's t-test is used for normally distributed parameters in two unpaired samples, Mann-Whitney test is used for non-normally distributed parameters in two unpaired samples, and log-rank test is used for survival analysis. All statistical analyses were performed using GraphPad Prism software. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 were considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.