In vitro and in vivo primed TH9 cells display key features of pathogenic TH2 cells
The transcriptomic signature of pTH2 cells has been previously identified by single-cell analysis of T cells extracted from multiple TH2-driven diseases2. To test whether human in vitro primed TH9 cells recapitulate the core pTH2 cell phenotype, we differentiated naive T cells into TH1 (IL-12), TH2 (IL-4), TH9 (IL-4+TGF-β), and iTREG (TGF-β) cells and performed transcriptomic profiling using RNA sequencing (RNA-seq) at day 7. Pairwise comparison to other subsets showed that 1492 genes were specifically upregulated in TH9 cells (Fig. 1a). We then compared our TH9 transcriptome with three pTH2-specific transcriptomes identified in EoE4, allergic asthma5, and allergen-specific TH2 cells7, respectively (Fig. 1b). PPARG, IL5, IL17RB, and IL9R, which are hallmarks of pTH2 cells, were upregulated in TH9 cells as well as in all three pTH2 datasets, while IL9 was upregulated in two of the three (Fig. 1c and Supplementary Fig. 1a). In contrast, SPI1, encoding the transcription factor PU.1, previously shown to be associated with IL-9 expression16,17, was neither expressed in pTH2-specific transcriptomes, nor in TH9 cells (Supplementary Fig. 1a). High levels of PPAR-γ were confirmed at the protein level in both in vitro primed TH9 cells and in vivo primed TH9 clones, which we generated from ex vivo isolated memory TH cells that were sorted based on chemokine receptor expression (Fig. 1d–f). In summary, these findings strongly support our hypothesis that in vitro and in vivo primed TH9 cells share key similarities with pTH2 cells.
As we had previously identified PPAR-γ+ TH9 cells in ACD9, we next considered validating the association between the TH9 and pTH2 phenotypes in acute allergic skin inflammation. Time course transcriptomics of untreated non-lesional (NL) skin and positive patch test reactions of lesional skin to nickel at 24 h, 48 h, and 120 h post allergen application showed a marked upregulation of the pTH2-associated genes in ACD (Fig. 1g and Supplementary Fig. 1b). Across individual samples, the expression of IL9 correlated with the expression of IL13 (R2 = 0.80; P < 0.0001), IL5 (R2 = 0.54; P < 0.0001), IL31 and IL19, but not IFNG (Fig. 1h and Supplementary Fig. 1c), consistent with the predominance of TH9 cells in the TH2 cell pool. Due to the fact that PPARG is expressed in various skin cell types, including keratinocytes9, the correlation analysis of PPARG with IL9 does not allow any conclusion with regard to TH9 cells.
Collectively, these data show that both in vitro and in vivo primed TH9 cells express the core features of pTH2 cells, including upregulated expression of PPAR-γ, IL-5, IL-9, and IL-9R. They can thus serve as model cells for studying the functional role of PPAR-γ in human TH cells. Further, human ACD appears to be a valid model for studying the functionality of TH9 cells ex vivo.
PPAR-γ mediates the high glycolytic activity of TH9 cells
To investigate the role of PPAR-γ in human TH cells, we first assessed the transcriptional response of activated TH9 clones to treatment with GW9662, a potent PPAR-γ antagonist. Pathway analysis of RNA-seq data revealed concerted downregulation of genes associated with T cell activation, glucose metabolism, and aerobic glycolysis (Supplementary Fig. 2a). At the single gene level, mRNA expression of all aerobic glycolysis enzymes was significantly downregulated by PPAR-γ inhibition (Fig. 2a, b). This prompted us to further analyze the role of PPAR-γ in aerobic glycolysis of TH9 cells. In vitro primed TH9 cells showed higher glycolytic activity than TH1- or TH2-primed T cells after activation with αCD3/CD2/CD28 (Fig. 2c and Supplementary Fig. 2b). To verify whether PPAR-γ was involved in this process, we starved in vitro primed TH9 cells in glucose-free medium in presence or absence of GW9662. We then performed measurements of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) in real-time before and after activation with αCD3/CD2/CD28 in either low or high-glucose environments. PPAR-γ inhibition by GW9662 or the alternative PPAR-γ antagonist, T0070907, hampered the glycolytic response in high- but not low-glucose environments, particularly following T cell activation (Fig. 2d, e and Supplementary Fig. 2c, d). These findings were corroborated by measurements of glucose uptake, in which in vitro primed TH9 cells showed higher glucose uptake compared to TH1, and TH2 cells (Fig. 2f). Importantly, glucose uptake in TH9 cells was reduced by PPAR-γ-inhibition or by siRNA-induced PPARG knockdown (Fig. 2g).
We next expanded our findings to in vivo primed memory TH cells, leveraging the ability to sort different subsets ex vivo based on their chemokine receptor profile, including CXCR3–/CCR4+/CCR6–/CCR8+ TH9 cells9,18. Consistent with our previous observations, in vivo primed TH9 cells showed higher glucose uptake compared to other TH cell subsets, and GW9662 significantly inhibited the glycolytic activity in TH9 cells but not in TH2 or TH1 cells (Fig. 2h).
Given the central role of aerobic glycolysis for T cell proliferation19, we finally tested whether PPAR-γ inhibition affected TCR stimulation-induced proliferation in low and high-glucose environments. Both GW9662 and T0070907 significantly reduced αCD3/CD2/CD28-induced proliferation in high-glucose environments (Fig. 2i and Supplementary Fig. 2e, f).
Since PPAR-γ has been implicated in mediating fatty acid (FA) uptake in activated TH cells20,21, we examined FA metabolism in response to PPAR-γ antagonism. In vitro and in vivo primed TH9 cells did not exhibit higher FA uptake compared with TH1 and TH2 cells (Supplementary Fig. 2g) and PPAR-γ antagonism had no effect on FA uptake (Supplementary Fig. 2h, i). In addition, glutamine uptake of TH9 was not affected by PPAR-γ inhibition, suggesting that glutaminolysis is not primarily regulated by PPAR-γ in these cells (Supplementary Fig. 2j).
PPAR-γ has previously been shown to regulate the expression of IL-33R in murine TH2 cells, thereby increasing their sensitivity to the tissue alarmin IL-3311,12 in allergic inflammation. While IL-33R (IL1RL1) is also associated with the pTH2 phenotype in humans4,6,7,8,22, PPAR-γ antagonism did not downregulate its expression in TH9 cells (Supplementary Fig. 2k).
Collectively, these data strongly suggest that both in vitro and in vivo primed human TH9 cells are characterized by high glycolytic capacity post-TCR-stimulation. Moreover, in the setting studied here, PPAR-γ signaling appears to be a crucial mediator of glycolysis, whereas FA oxidation and glutaminolysis remain unaffected.
High glycolytic activity in TH9 cells differentially regulates cytokine expression
PPAR-γ is required for the full effector function in TH9 and pTH2 cells, including the production of proinflammatory cytokines9,11,12. Based on our findings, we hypothesized that PPAR-γ might control cytokine production indirectly, namely by promoting glycolysis. To test this, we cultured in vitro primed TH9 cells in media containing different glucose concentrations and measured their cytokine profiles at day 7. Production of the pTH2 marker cytokine IL-9 showed a strong dependency on glucose availability, whereas production of IL-13, the conventional TH2 cytokine, did not (Fig. 3a). Further, in vivo primed TH9 clones cultured in different glucose concentrations downregulated the production of IL-9 but not IL-13 in low-glucose environments (Fig. 3b).
To demonstrate a direct relationship between high-glucose metabolism and cytokine production, we next sorted in vitro primed TH9 cells according to their glucose uptake level, measured by the uptake of 2-NBDG. Subsequently, we performed RT-qPCR for IL9, IL13, and PPARG. Glucose uptake correlated with the expression of IL9 and PPARG, but not IL13 (Fig. 3c and Supplementary Fig. 3c). Taken together, these data strongly suggests that glycolytic activity regulates the expression of IL-9. Similar regulation, albeit less pronounced, was observed for IL5 expression (Supplementary Fig. 3a–c).
In a next step, we hence inhibited glycolysis in TH9 cells using the glucose analog 2-deoxy-d-glucose (2-DG) and the aerobic glycolysis inhibitor lonidamine (LND) to investigate the effect on cytokine expression. In TH9 cells primed in vivo, 2-DG and LND inhibited the expression of IL-9 and IL-5 but not IL-13 (Fig. 3d and Supplementary Fig. 3d, e). Finally, PPAR-γ antagonism in high-glucose environments reduced the production of IL-9 to the levels seen in low-glucose environments, whereas IL-13 levels remained unaffected neither by PPAR-γ antagonism nor by low-glucose availability (Fig. 3e).
To investigate the contribution of FA metabolism to the regulation of cytokines in TH9 cells, we analyzed IL-9 expression in response to FA inhibition. Neither did culturing of TH9 clones in FA-free medium affect IL-9 or IL-13 expression (Supplementary Fig. 3f), nor did inhibition of FA metabolism with etomoxir, an inhibitor of carnitine palmitoyltransferase-1 (Supplementary Fig. 3g).
Taken together, these observations indicate a dichotomous role of glycolytic activity in regulating the production of IL-9, IL-5, and IL-13 by activated TH9 cells.
mTORC1 integrates glycolytic activity with the effector function in TH9 cells
We next investigated the mechanisms underlying the association between glycolysis and cytokine production in activated TH9 cells. Mammalian target of rapamycin complex 1 (mTORC1) is a central regulator of cellular metabolism and effector functions in T cells. Nutrients, such as glucose, are critical activators of mTORC123. Furthermore, the mTORC1-hypoxia-inducible factor-1α (HIF-1α) pathway is necessary for the expression of IL-9 in murine T cells, with HIF-1α binding directly to the Il9 promoter and activating its transcription24,25,26. Given the role of the established mTORC1-HIF-1α-IL-9 axis and our previous results, we hypothesized that mTORC1 might mediate the PPAR-γ-dependent expression of IL-9 in TH9 cells.
Phosphorylation of mTORC1 in activated TH9 cells measured by phosphorylated S6 (pS6) was glucose-dependent and reduced by PPAR-γ inhibition under high-glucose conditions (Fig. 4a, and Supplementary Fig. 4a, b). Indeed, IL-9+ T cells were strongly enriched in the pS6+ cell population, whereas the proportion of IL-13+ T cells was similar in pS6– and pS6+ populations (Fig. 4b, c). Moreover, inhibition of mTORC1 by either siRNA against RPTOR or by rapamycin decreased the production of IL-9 but not IL-13 in activated TH9 cells (Fig. 4d, e and Supplementary Fig. 4c, d).
Since the PPAR-γ agonist troglitazone (TGZ) unexpectedly reduced IL-9 expression in TH9 cells, we next investigated the mechanism by which this occurs. Interestingly, pS6 levels were reduced in the presence of TGZ (Supplementary Fig. 4e), suggesting that mTORC1 is inhibited. Previous studies have shown that PPAR-γ agonists, such as TGZ, activate AMP-activated protein kinase (AMPK)27. We, therefore, hypothesized that TGZ-mediated AMPK activation negatively regulates mTORC1, which in turn suppresses IL-9 expression. Indeed, Western blot analysis revealed that TGZ leads to phosphorylation of AMPK and mTORC1 inhibition (Supplementary Fig. 4f). In addition, the AMPK activator A-769662 also reduced IL-9, but not IL-13 levels (Supplementary Fig. 4g). Together, this data strongly supports our hypothesis that IL-9 expression is mTORC1-dependent.
To verify whether TH9 cells expressed activated mTORC1 in human skin inflammation, we performed immunofluorescence staining of normal skin (NS) and ACD skin samples and isolated T cells from such lesions. Double immunofluorescence revealed that CD3+ and CD4+ TH cells that express pS6 are significantly enriched in the infiltrate of ACD compared to NS (Fig. 4f and Supplementary Fig. 4h, i). Virtually all IL-9+ TH cells isolated from ACD show S6 phosphorylation, and thus have active mTORC1 signaling (Supplementary Fig. 4j). Incubation with GW9662 ex vivo showed reduced activation of mTORC1 and a significantly inhibited IL-9 production (Fig. 4g). As it is known that ACD skin is infiltrated by a substantial number of PPAR-γ+ TH cells9, we finally investigated whether those cells would show the activation of mTORC1. Indeed, PPAR-γ+pS6+ double-positive lymphocytes were significantly increased in the dermis of ACD compared to NS (Fig. 4h and Supplementary Fig. 4k).
Collectively, these findings strongly suggest that glucose- and PPAR-γ-dependent production of IL-9 in TH9 cells is regulated via mTORC1 in acute allergic skin inflammation.
Paracrine IL-9 promotes aerobic glycolysis in IL-9R+ TH cells by inducing the lactate transporter MCT1
After revealing the association between PPAR-γ-dependent glycolytic activity and IL-9 production in TH9 cells, we hypothesized that paracrine IL-9 might regulate glucose metabolism and downstream effector functions in IL-9R+ TH cells. Previous data suggested that IL9R is preferentially expressed in pTH2 and TH9 cells (Fig. 1 and ref. 4,5,7,8), but these findings had to be confirmed at the protein level and in human skin inflammation. Thus, we first confirmed the expression of IL-9R in TH9 clones primed in vivo (Fig. 5a), circulating CXCR3−/CCR4+/CCR8+ effector memory TH cells (Fig. 5b), and TH cells infiltrating lesions of ACD (Fig. 5c, d), showing that pTH2 and TH9 cells are important targets of IL-9 in human skin inflammation. Next, we performed RNA-seq of IL-9R+ TH clones and IL-9R+ TH cells isolated from ACD skin biopsies, incubated with or without IL-9. The pathway analysis of the 250 most IL-9-induced genes showed a coordinated induction of genes involved in aerobic glycolysis (Fig. 5e), most prominently SLC16A1 (Fig. 5f). SLC16A1 encodes the monocarboxylate transporter 1 (MCT1), enabling the rapid export of lactate across the plasma membrane and thereby exerting a glycolytic flux-controlling function28. IL-9-induced expression of MCT1 in TH9 clones was confirmed at the protein level (Fig. 5g). In contrast, inhibition of JAK3, which is central to IL-9R signal transduction29, by ritlecitinib suppressed IL-9-induced upregulation of SLC16A1 (Fig. 5h).
Next, we investigated the expression and regulation of SLC16A1 in different TH cell populations. In vitro TH9 differentiation induced higher levels of SLC16A1 than TH1, TH2, or iTREG differentiation (Fig. 5i). Time course transcriptomics of TCR-stimulated TH clones9 revealed a significantly higher expression of SLC16A1 in TH9 clones than in TH1 and TH2 clones (Fig. 5j), as well as a close correlation between the expression of IL9 and SLC16A1 (Fig. 5k). Finally, a Seahorse experiment confirmed that IL-9R+ TH clones stimulated by IL-9 exhibited strongly elevated ECAR, in line with IL-9-dependent induction of active aerobic glycolysis and efficient cellular export of lactate (Fig. 5l). Accordingly, IL-9 increased extracellular lactate levels in cultured IL-9R+ TH clones, and these levels were suppressed by the addition of BAY-8002, a potent MCT1 antagonist (MCT1-i) or by siRNA-mediated SLC16A1 knockdown (Fig. 5m). To link IL-9, glycolysis and MCT1 expression, we next investigated IL-9 levels in TH9 cells in presence of MCT1 inhibitor. We show that MCT1 inhibition reduces IL-9, but not IL-13 levels (Supplementary Fig. 5a). On the contrary, low-glucose environments, which in turn lead to reduced IL-9 levels, inhibited the induction of SLC16A1 (Supplementary Fig. 5b).
Taken together, these data indicate that paracrine IL-9 promotes aerobic glycolysis in IL-9R+ TH cells by inducing the lactate transporter MCT1, which controls glycolytic flux.
IL-9 promotes T cell proliferation in high-glucose environments
Considering the crucial role of aerobic glycolysis in supporting T cell proliferation30, we next investigated the functional effects of IL-9-induced MCT1 expression on T cell proliferation. IL-9 induced strong proliferative responses in IL-9R+ TH clones (Fig. 6a), and this proliferative boost was reversed by adding BAY-8002, the MCT1 inhibitor (Fig. 6b). Moreover, IL-9-induced proliferation was dependent on available glucose levels (Fig. 6c), which further supports the notion that IL-9 promotes glycolytic flux in IL-9R+ TH cells to facilitate proliferation.
Finally, we investigated whether tissue glucose levels are dynamically regulated in acute allergic skin inflammation, in which TH9 cells have been shown to highly express IL-99, 31. To this end, we determined glucose levels in the interstitial fluid of tissue homogenates from non-lesional and lesional ACD skin 48 h post allergen application, respectively. The interstitial fluid of the lesional skin contained higher glucose levels than the matched non-lesional skin samples (Fig. 6d), which suggests that IL-9-related human allergic skin inflammation is associated with changes in glucose availability in vivo.
Collectively, these observations suggest that paracrine IL-9 facilitates the proliferation of T cells by stimulating aerobic glycolysis through the induction of the lactate transporter MCT1, possibly contributing to the proliferation of IL-9R+ TH cells in the high-glucose environment of ACD and acute allergic tissue inflammation. In addition, we found that PPAR-γ is a positive regulator of aerobic glycolysis in activated human TH9 cells, which in turn, regulates the expression of IL-9 via mTORC1. Together, this suggests that PPAR-γ and IL-9 facilitate immunometabolic sensing of the tissue microenvironment (Fig. 6e).