STAT5 Represses a STAT3-Independent Th17-like Program during Th9 Cell Differentiation

This article is featured in Top Reads, p.1223

CD4⁺ Th cells differentiate in response to inflammatory cytokines and downstream signaling through receptor-associated signal transducer and activator of transcription (STAT) factors. For example, IL-4–producing Th2 cells differentiate in response to IL-4 and signaling through STAT6 (1). However, in the presence of TGF-β, IL-4–driven Th2 cell development is subverted and results in the differentiation of IL-9–producing Th9 cells that play detrimental roles in allergic disease and exhibit enhanced antitumor immunity after adoptive cell therapy (2–4). During the differentiation process, IL-4–induced STAT6 binds regulatory elements within the Il9 locus and additionally suppresses expression of transcription factors (TFs) (i.e., T-bet, Foxp3) that inhibit Th9 differentiation (2, 5). Similarly, TGF-β induces expression of PU.1, which diverts cells from a Th2 to a Th9 phenotype (6). Despite the importance of these cytokines and TFs in Th9 differentiation, they are insufficient to drive Th9 differentiation in the absence of IL-2/STAT5 signaling.

IL-2 is a pleiotropic cytokine that acts to drive the proliferation, survival, and differentiation of Th cells by binding the IL-2R complex and inducing the phosphorylation of STAT5 (7). Upon phosphorylation, STAT5 forms homodimer or tetrameric structures that translocate to the nucleus, where they either activate or suppress gene expression based on the molecular context of each bound loci. Although IL-2/STAT5 signaling is detrimental for the differentiation of Th17 and T follicular helper cells, it enhances Th1 and Th2 lineage programs (8–11). In previous work, we and others have shown that IL-2 and STAT5 signaling also promotes the differentiation and cytokine-producing capacity of Th9 cells (12–15). In the differentiation process, STAT5 represses expression of BCL6, which competes for binding with STAT5 at the Il9 locus, and likely other Th9-associated gene loci (12, 13). Additionally, STAT5 plays a “pioneering” role in Th9 cell differentiation by facilitating the binding of other TFs (e.g., BATF and IFN regulatory factor [IRF] 4) to the Il9 locus (16–18). IL-2 and STAT5 also enhance the cytokine-producing capacity of Th9 cells postdifferentiation. In this role, IL-2 production by Th9 cells feeds back in a paracrine fashion and amplifies production of IL-9 and several other STAT5-induced cytokines after TCR activation (14). Despite the known roles of IL-2 and STAT5 signaling in the in vitro differentiation and function of Th9 cells, much less is known about how these signals impact the differentiation of Th9 cells during disease. Furthermore, how IL-2/STAT5 signaling impacts Th9 differentiation outside of repressing BCL6 and modulating the Il9 locus has not been well established.

We demonstrate in this study that STAT5 signaling is associated with IL-9–producing Th cells in human allergic disease. Although a number of known STAT5 target genes were induced in human Th9 cells, we also observed a strong signature of STAT5-mediated gene repression in IL-9–producing cells. STAT5-mediated gene repression correlated with the suppression of a Th17-like signature, indicating that STAT5 may regulate the balance of Th9/Th17 differentiation during disease. In a murine Th cell differentiation model, we further demonstrated that IL-2–deprived Th9 cells produced high levels of IL-17 and the Th17-associated factors Rorγt and BATF. Despite their capacity to produce IL-17, IL-2–deprived Th9 cells did not exhibit a classical Th17 transcriptional profile and, surprisingly, maintained IL-17 production and Rorγt and BATF expression in the absence of STAT3. STAT3-independent IL-17 production, however, was dependent on both Rorγt and BATF. Despite the potential for IL-17–producing cells to differentiate independently of STAT3 in these conditions, STAT3 was required for the prolonged survival of IL-2–deprived Th9 cells. These data may explain why STAT3-independent IL-17–producing Th cells are not readily apparent in vivo. Together, our data identify an important role of STAT5-mediated gene repression in human and mouse Th9 cell development and define a novel STAT3-independent pathway involved in the differentiation of IL-17–producing T cells. Insight into this STAT3-independent pathway may be useful for overcoming STAT3 loss-of-function immunodeficiencies that feature a lack of IL-17–producing Th cells and susceptibility to multiple opportunistic pathogens.

C57BL/6 mice were bred in-house at Purdue University. Stat3^fl/fl mice expressing Cd4- Cre recombinase, Stat3^fl/fl Cd4^cre, were originally obtained from Dr. D. Levy (New York University, New York, NY). Batf ^−/− mice were kindly provided by Dr. Elizabeth Taparowsky (Purdue University, West Lafayette, IN). All mice experiments were in compliance with the protocol approved by the Purdue Animal Care and Use Committee.

Naive CD4⁺ T cells were obtained from the spleen and lymph nodes using the Naive CD4⁺ T Cell Isolation Kit (Miltenyi Biotec, Auburn, CA). Purity was >90% checked through flow cytometry. A total of 1 × 10⁶ enriched cells/ml were cultured in a 1:1 mixture of RPMI or DMEM (Life Technologies) supplemented with 1% penicillin–streptomycin (Life Technologies), 10% FBS, and 50 µM 2-ME (Life Technologies). Cells were cultured in anti-CD3ε (2 µg/ml; 2C11; Bio X Cell) -coated plates and soluble anti-CD28 (4 µg/ml; 37.51; Bio X Cell) under the T cell–polarization conditions described below. The Th9 conditions included recombinant murine IL-4 (10 ng/ml; PeproTech), anti-IFN-γ (12.5 µg/ml; XMG1.2; Bio X Cell), recombinant human TGF-β1 (4ng/ml; PeproTech), and anti-IL-10R (12.5 µg/ml; 1B1.3A; Bio X Cell). The Th17 conditions included anti-IL-4 (12.5 µg/ml; 11B11; Bio X Cell), anti-IFN-γ, recombinant murine IL-6 (100 ng/ml; PeproTech), TGF-β1, and anti-IL-10R. Additionally, in some conditions we neutralized IL-2 in Th9 and Th17 cultures with anti-IL-2 (5 µg/ml; S4B6-1; Bio X Cell) and anti-CD25 (10 µg/ml; 3C7; BioLegend). The cells were incubated at 37°C and harvested on day 4 of culture. In some experiments, either the RORγt-specific inhibitors digoxin (10 µM; Cayman Chemical) or TMP778 (2.5 µM; AOBIOUS) or the corresponding dilution of DMSO (Fisher Bioreagents) were added to the conditions mentioned above. In long-term cell culture experiments, cells from each condition were harvested at day 4 and rested for 24 h in the absence of anti-CD3 with half concentrations of the original cytokines. After this rest period, cells were reseeded at 1 × 10⁶ cells/ml into fresh anti-CD3–coated wells and cultured for an additional 4 d under the initial conditions stated, at which time cells were assessed for viability and expression of Th lineage-specific cytokines and TFs.

At day 4 of culture, cells were harvested, and 2 × 10⁵ cells at 1 × 10⁶ cells/ml were used for either intracellular cytokine staining or TF staining. For intracellular cytokine staining, the differentiated Th cells were incubated at 37°C with PMA (0.5 µg/ml) and ionomycin (0.5 µg/ml; Sigma-Aldrich) for 2.5 h, and monensin (2 µM) (BioLegend) was added for additional 2.5 h. The cells were subsequently stained with fixable viability dye (eFlour 780; Thermo Fisher Scientific) and mouse anti-CD4 Ab, (RM4-5; 1 µg/ml; BioLegend) and then were fixed with 3.7% formaldehyde. After fixation, cells were permeabilized with Intracellular Staining Permeabilization Wash Buffer (BioLegend) and stained in the same buffer with a cytokine Ab mixture that includes anti–IL-17A (TC11-18H10.1), anti–IL-9 (RM9A4), anti–IL-2 (JES6-5H4), anti–GM-CSF (MP1-22E9), and anti-TNF (MP6-XT22) (all purchased from BioLegend and used at 1 µg/ml). For TF staining, the cells were fixed instead with True-Nuclear Fixation Buffer (BioLegend). After fixation, the cells were permeabilized with True-Nuclear Permeabilization Buffer (BioLegend) and were stained with Abs to BATF (S39-1060; 1:100 dilution of stock; BD Biosciences), Rorγt (Q31-378; 2 µg/ml; BD Biosciences), and IRF4 (IRF4.3E4; 0.5 µg/ml; BioLegend). Stained samples were run on an Attune NxT Flow Cytometer (Thermo Fisher Scientific), and the data were analyzed using FlowJo software (v.10.0).

Retroviral particles were generated using the bicistronic vectors MSCV-IRES-eGFP and MSCV-Thy1.1 for RORγt and BATF (gifts from Dr. Mark Kaplan, Indiana School of Medicine and Dr. Elizabeth Taparowsky, Purdue University, respectively). Plasmids (10 µg) containing each gene of interest were cotransfected with 5 µg of packaging vector pCL-Eco (a gift from Dr. Mark Kaplan, Indiana School of Medicine) into HEK-293T cells using Lipofectamine 2000 according to manufacturer’s instructions. At 48 h posttransfection, virus containing cell supernatant was collected and used for Th cell transduction. Briefly, naive CD4 T cells were cultured under Th17 conditions as per above and on day 2 of culture, cells were “spinfected” at 2,000 rpm for 1.5 h at 37°C in the presence of polybrene (8 µg/ml; Sigma-Aldrich). Cells were harvested 2 d after transduction, and the efficiency of the transduction as well as intracellular IL-17 production was assessed via flow cytometry.

Data plots and statistics were generated using the software GraphPad Prism, v8. The statistics tests used are indicated in the figure legends and were considered significant when p < 0.05.

The fastq files were downloaded from GSE41317. RNA expression levels in each library were estimated by “rsem-calculate-expression” from RSEM v1.3.0 (19) using default parameters, except “–bowtie-n 1 –bowtie-m 100 –seed-length 28.” The bowtie index (GRCm38) required by RSEM software was generated by “rsem-prepare-reference” on all Reference Sequence genes and obtained from the University of California Santa Cruz Table Browser in April, 2017. The edgeR v3.24.3 (20) package was used to identify differentially expressed genes between IL-2 knockout (KO) and wild-type (WT) samples. Gene Set Enrichment Analysis (GSEA) (21) was performed by GSEA version 3.0 using default settings, except “permutation type = gene_set” and “Collapse = No_Collapse.”

The raw single-cell RNA-sequencing (RNA-seq) data from allergen-specific T cells in allergy and asthma patients were sourced from GSE146170 (22). All data were reanalyzed according to what has been described in the original manuscript. Briefly, cellranger count (v3.1.0) was used to align sequencing reads to the human reference genome (GRCh38) to extract barcodes and unique molecular identifier counts.

Seurat v3 standard integration workflow was followed to enable comparative analysis among samples (23). Genes expressed in <3 cells, cells with <200 expressed genes and/or >5% mitochondrial genes were removed from downstream analyses in each dataset. The filtered count matrices were then normalized by total unique molecular identifier counts, multiplied by 10,000, and transformed to natural log space. The top 2000 variable features were determined based on the variance stabilizing transformation function (FindVariableFeatures) by Seurat with default parameters. All samples were integrated using canonical correlation analysis function with default parameters. Potential artifacts arising from library size and percentage of mitochondrial genes were regressed out by the ScaleData function. To perform clustering, principal component analysis was first performed, and the top 50 principal components were included in a uniform manifold approximation and projection dimensionality reduction. Clusters were then identified on a shared nearest-neighbor modularity graph using the top 50 principal components and the original Louvain algorithm. Cluster annotations were derived from the same marker genes used in the original paper (22). Seurat function AddModuleScore (24) was used to calculate gene list (module) scores. This function calculates the average scaled expression levels of each gene list, subtracted by the expression of control sets. The expression of individual genes across selected cell types were extracted, plotted by GraphPad Prism v8, and compared using the Wilcoxon test. Additional statistical tests used are indicated in the figure legends and were considered significant when p < 0.05.

IL-2–mediated STAT5 signaling plays a critical role in the in vitro differentiation of Th9 cells by acting as both a transcriptional activator and repressor (12, 13, 15, 16). However, less is known about the roles of IL-2/STAT5 signaling in human Th9 cell development in vivo. To address this, we reanalyzed recent single-cell transcriptomics data from house dust mite (HDM)–allergic patients, in which IL9–producing (IL9^hi) and low IL9-producing (IL9^lo) “Th2” cells populations were readily apparent (Fig. 1A, Supplemental Fig. 1A) (22). Consistent with the original study, both IL9^hi and IL9^lo cells were enriched in HDM-allergic but not in asthmatic patients with no HDM sensitivity as compared with healthy individuals. Furthermore, these subsets of cells were distinct from other Th lineages (e.g., Th1, type I IFNR⁺ cells or activated Th cells) within these patients, whose presence was less affected by disease status (Supplemental Fig. 1A–E). Collectively, these data suggest that IL9^hi and IL9^lo Th cells are associated with allergic disease.

Based on these data, we reasoned that a direct comparison of IL9^hi and IL9^lo cells may help identify how STAT5 signaling shapes pathogenic Th cells during the allergic response. To address this point, we initially examined the expression of a number of known STAT5-regulated genes in IL9^hi and IL9^lo cells. Although the levels of IL4 were similar in IL9^hi and IL9^lo subsets (Fig. 1A), other STAT5 target genes, such as IL5, IL13, IL1RL1, and GATA3 were more highly expressed in IL9^hi versus IL9^lo Th cells (Fig. 1A, Supplemental Fig. 1D, 1E). To test whether other STAT5 target genes were differentially affected, we sourced a list of IL-2–induced or –repressed genes in Th9 cells (13) with known STAT5 binding (25) as a surrogate for direct in vivo–induced and –repressed targets of STAT5 (see Supplemental Table I). The expression of STAT5-induced genes was only moderately increased in IL9^hi versus IL9^lo Th cells in HDM-allergic patients (Fig. 1B). Surprisingly, however, the expression of STAT5-repressed genes were markedly lower in IL9^hi versus IL9^lo cells (Fig. 1C), suggesting that STAT5-mediated gene repression is enhanced in IL-9–producing cells and thus may have a role in their differentiation.

STAT5 has previously been shown to inhibit Th17 cell differentiation (8); thus, we speculated that IL9^lo cells might exhibit a Th17-like signature as compared with IL9^hi cells. To test this, we sourced a list of Th17 signature genes (26) (see Supplemental Table I) and compared the expression of these genes in IL9^lo versus IL9^hi cells. Although the expression of IL17 and RORC were below the level of detection within these single-cell data, IL9^lo cells did exhibit higher expression of Th17-associated genes as compared with IL9^hi cells (Fig. 1D). Together, these data indicate that human Th9 cell differentiation is associated with enhanced STAT5 activity and that STAT5 potentially represses a Th17-like phenotype in IL-9–producing cells.

As STAT5 repressive activity inversely correlated with the emergence of a Th17-like signature, we questioned whether STAT5 signaling repressed a Th17-like differentiation program during Th9 cell development. To this point, we used a well-characterized in vitro murine model of Th9 cell differentiation, in which naive CD4 T cells were cultured under standard Th9-polarizing conditions in the presence or absence of IL-2– and CD25-blocking Abs. After 4 d of culture, we examined the expression of IL-9 and several Th17-associated proteins (e.g., IL-17A, IL-17F, Rorγt, IRF4, BATF) by flow cytometry. As previously reported, inhibition of STAT5 signaling through IL-2 blockade resulted in an almost complete loss of IL-9 production by Th9 cells and a reduction in the expression of the STAT5 target IRF4 (Fig. 2A–C) (13, 15, 27). As predicted by our analysis of human IL9^hi and IL9^lo cells, we observed a striking increase in the protein expression of IL-17A, IL-17F and Rorγt that was similar or equivalent to that of bona fide Th17 cells cultured with anti-IL2/CD25 Abs (Fig. 2A–C). To further confirm these findings, we reanalyzed existing RNA-seq data (13) comparing WT and IL-2–deficient Th9 cells. IL-2–deficient Th9 cells exhibited dramatic reductions in the STAT5-induced genes Il9, Irf4, and Batf3, as originally described (13). Similar to our protein data, we observed a dramatic induction of Th17-associated genes Il17a, IL17f, and Rorc, as well as Maf and Ccr6, in the absence of IL-2 in these Th9 cells (Fig. 2D). Together, these data indicate that IL-2/STAT5 suppresses a Th17-like phenotype in Th9 cells and defines a transcriptional signature of Th9-derived IL-17–producing cells.

To examine the functional features of IL-2–deprived Th9 cells, we compared the transcriptomes of IL-2–deficient (KO) and –sufficient (WT) Th9 cells against Th17-related programs using GSEA. Despite the increased expression of Il17 and Rorc in IL-2–deficient cells, there was not a significant correlation with the transcriptomes of these cells and that of the core Th17 signature (Fig. 2E). However, genes that were more highly expressed in IL-2–sufficient WT cells as compared with IL-2–deficient Th9 cells, were enriched in genes that antagonize Th17 cell differentiation (Fig. 2E). Furthermore, IL-2–deprived Th9 cells had an enhanced capacity to produce the neuroinflammatory cytokine GM-CSF and IL-2 and a reduced capacity to produce IL-17F as compared with bona fide Th17 cells (Fig. 2F). These data indicate that while IL-2–deprived Th9 and Th17 cells share a number of similarities, they differ in their capacity to produce proinflammatory cytokines and likely require unique cytokine signaling and TF networks for their differentiation.

As IL-2–deprived Th9 cells did not resemble canonical Th17 cells at the transcriptional level, we questioned whether these cells also differed in other key aspects of the Th17 differentiation pathway. Differentiation of Th17 cells requires IL-6– and IL-23–induced STAT3 signaling (8, 28–31). However, IL-2–deprived Th9 cells, as cultured in this study, lack exogenous STAT3 activating cytokines (i.e., IL-6, IL-10, and IL-21 (27),) and did not exhibit transcriptomic similarities with genes induced by STAT3 but, instead, were modestly correlated with STAT3-suppressed genes (Fig. 3A). Based on these data, we hypothesized that IL-17–producing IL-2–deprived Th9 cells may develop in a STAT3-independent fashion. To test this hypothesis, we isolated naive CD4 T cells from Stat3^fl/fl Cd4-Cre⁻ (i.e., WT STAT3 expression) or Stat3^fl/fl Cd4-Cre⁺ (i.e., T cell–specific deletion of STAT3) and cultured these cells under Th9 or Th17 conditions in the presence or absence of IL-2 and CD25 blocking Abs. In agreement with previous studies (31), IL-17A and IL-17F production in canonical Th17 conditions was completely dependent on STAT3 (Fig. 3B, 3C). However, in IL-2–deprived Th9-polarizing conditions, production of IL-17A and -17F were STAT3-independent (Fig. 3B, 3C), indicating that STAT3 is not required for IL-17 production under these conditions.

In Th17 cells, STAT3 is also required for the optimal expression of the Th17 master TF Rorγt and the “accessory” TFs BATF and IRF4. As STAT3 was not required for IL-17 production under IL-2–deprived Th9 cell–polarizing conditions, we questioned whether expression of Rorγt and BATF were also STAT3-independent. Similar to IL-17, we observed virtually identical levels of Rorγt, BATF, and IRF4 in WT and STAT3-deficient IL-2–deprived Th9 cells, whereas all of these factors were decreased in STAT3-deficient Th17 cells (Fig. 3D, 3E). Taken together, these data indicate that STAT3 is dispensable for induction of IL-17 and several Th17-associated TFs under IL-2–deprived Th9-polarizing conditions.

In our previous work, we showed that the prolonged culture of Th9 cells resulted in reduced IL-2 bioavailability and the subsequent emergence of cells with the capacity to produce IL-17 (32). Moreover, increased IL-17 production in these settings correlated with enhanced levels of activated STAT3. Despite this appealing correlation, it was unclear whether STAT3 was required for IL-17 production in this setting. To address this, we repeated our initial prolonged culture experiments (32) with both WT and STAT3-deficient naive CD4 T cells. Similar to our previous report, prolonged culture of WT Th9 cells resulted in a reduced frequency of IL-9–producing cells and an increased proportion of IL-17–producing cells as compared with day 4 of culture (Figs. 4A, 2A). Likewise, STAT3-deficient cells exhibited a similar capacity to produce IL-17 and had Rorγt and BATF expression levels that were not significantly different (p > 0.05) than WT Th9 cells at this time point (Fig. 4A–C). These data support our initial findings and indicate that IL-2/STAT5 signaling suppresses a STAT3-independent Th17-like phenotype in differentiating Th9 cells.

Previous work has shown that STAT3 is required for the generation of IL-17–producing Th cells in vivo (33, 34). However, this lack of IL-17 production is potentially complicated by the poor survival and persistence of STAT3-deficient Th cells in these settings (30, 33). To test whether STAT3 is required for the persistence of IL-2–deprived Th9 cells, we used a prolonged culture approach in which naive WT or STAT3-deficient CD4 T cells were cultured under Th9-polarizing conditions with and without IL-2 blockade for two rounds of culture, as indicated in (Fig 4. After the second round of culture, live cell counts and the frequency of viable cells (Supplemental Fig. 2A) from each condition were assessed by flow cytometry. Under standard Th9 conditions, there was little difference in the total number or frequency of viable cells after two rounds of culture. However, when Th9 cells were deprived of IL-2, STAT3-deficient cells exhibited significantly (p < 0.05) poorer survival as compared with their WT counterparts (Supplemental Fig. 2B). These data indicate that STAT3 is required for the prolonged survival of IL-17–producing cells when IL-2 is limiting.

In Th17 cells, Rorγt directly binds the Il17 locus and is required for IL-17 expression (28). Above, we demonstrated that IL-2–deprived Th9 cells expressed Rorγt in a STAT3-independent manner, indicating that IL-17 production by IL-2–deprived Th9 cells may require Rorγt. To initially test this possibility, we compared the transcriptomes of IL-2–deficient and –sufficient Th9 cells against Rorγt-induced and -repressed genes [sourced from (28)]. Interestingly, genes that were more highly expressed in IL-2–deficient compared with IL-2–sufficient Th9 cells were highly enriched in Rorγt-induced but not Rorγt-repressed genes (Fig. 5A). These data suggest a role for Rorγt in STAT3-independent IL-17 production.

To determine whether Rorγt activity was required for STAT3-independent IL-17 production, WT and STAT3-deficient naive CD4 T cells were isolated and cultured under Th9 and Th17 conditions as above. In some conditions, we also included the Rorγt-specific inhibitor digoxin, which interferes with Rorγt ligand binding and induction of transcription (35). Although Th9 cells express high levels of Rorγt, digoxin treatment did not significantly inhibit the production of IL-9 in these experiments or the expression of BATF or IRF4 in either Th9 or IL-2–deprived Th9 cell conditions (Fig. 5B, 5C). Of note, addition of the vehicle control alone (DMSO) had an effect on the ability of STAT3-deficient IL-2–deprived Th9 cells to produce IL-17, likely due to the mild cytotoxic effects of DMSO on T cells (36). Nonetheless, between 8–16% of STAT3-deficient cells were capable of producing IL-17 in this setting (compared with ∼0% in conventional STAT3-deficient Th17 conditions; see (Fig. 3A). Importantly, addition of digoxin or TMP778 [another Rorγt inhibitor (37)], to both WT and STAT3-deficient IL-2–deprived Th9 cultures virtually abolished IL-17 production in these settings (Fig. 5B, 5C). As previously described (35, 37), neither digoxin nor TMP778 altered Rorγt protein expression, and we further demonstrate, in this study, that abrogation of Rorγt activity with digoxin or TMP778 did not alter the expression of the Th17-associated accessory factors BATF or IRF4 (Fig. 5C). Together, these data indicate that Rorγt activity was required for STAT3-independent IL-17 production by Th cells, and this was not associated with a loss of BATF or IRF4 expression.

BATF is required for both Th9 and Th17 differentiation and also binds enhancer elements near the Il9 locus (16, 17) and in the Il17 promoter (28). Similar to Rorγt, BATF expression in IL-2–deprived Th9 cells, but not Th17 cells, was also maintained in the absence of STAT3 (see (Fig. 3C–E), suggesting that BATF may also play a role in STAT3-independent IL-17 production. We demonstrate here that BATF-deficient cells exhibit dramatic reductions of IL-9 and IL-17A production under both IL-2–deprived Th9- and Th17-skewing conditions (Fig. 6A, 6B). While BATF was required for optimal IL-17F production in Th17 cells, it was not required in IL-2–deprived Th9-polarizing conditions (Fig. 6B), suggesting that Rorγt may promote IL-17F production in the absence of BATF. Indeed, Rorγt protein expression was similar in WT and BATF-deficient IL-2–deprived Th9 cells, whereas conventional Th17 cells required BATF for optimal Rorγt expression under IL-2–neutralizing and –nonneutralizing conditions (Fig. 6C, 6D). Together, these data suggest that Rorγt and BATF play nonredundant roles in STAT3-independent IL-17 production by Th cells.

Our data above indicate that expression of Rorγt and BATF are STAT3-independent in IL-2–deprived Th9 cells and are required for IL-17 production in this setting. In contrast, Rorγt and BATF expression are dramatically reduced in STAT3-deficient Th17 culture conditions (see (Fig. 3). Based on these data, we hypothesized that restoration of Rorγt and BATF expression could rescue IL-17 production in STAT3-deficient Th17 cells. To test this hypothesis, we transduced WT and STAT3-deficient Th cells cultured under conventional Th17 conditions with retroviruses expressing Rorγt-GFP and BATF-Thy1.1 (Supplemental Fig. 3A) and examined IL-17 production after 4 d of culture. In WT cells, single transduction with Rorγt- or BATF-expressing retroviruses enhanced IL-17 production, with ectopic Rorγt expression having the dominant effect. Transduction with both factors resulted in an additive increase in IL-17A and IL-17F production (Supplemental Fig. 3A–C). In STAT3-deficient cells, however, transduction with Rorγt or BATF alone had minimal effect on IL-17A or IL-17F production. In contrast, STAT3-deficient cells with ectopic expression of both factors exhibited enhanced IL-17A/F production to levels seen in control-transduced WT cells but ∼4- to 5-fold less than WT cells transduced with both Rorγt and BATF. This difference in IL-17 production was not due to differential expression of Rorγt and BATF in WT and STAT3-deficient cells after transduction, as Rorγt and BATF protein (as measured by the associated hemagglutinin [HA] tag) levels were virtually identical (Supplemental Fig. 3C). Taken together, these data indicate that at least part of STAT3's role in promoting the IL-17 production is through the induction of Rorγt and BATF.

Although STAT5-mediated gene transactivation clearly plays an important role in the differentiation of Th9 cells, its role as a transcriptional repressor in this lineage is also critical. As a transcriptional repressor, STAT5 has been previously shown to directly bind and limit the expression of the T follicular helper cell–associated factor BCL6 that competes for STAT5 binding sites within the Il9 promoter (12, 13). However, our recent work indicated that WT and BCL6-deficient Th9 cells exhibited very similar capacities to differentiate into IL-9–producing cells under optimal conditions (27), indicating that STAT5 may have other repressive activities outside of regulating BCL6 expression.

In this study, we demonstrate that STAT5 activity is also associated with the development of IL-9–producing Th cells in human allergic disease. These data are consistent with previous reports indicating that CD4⁺ CCR4⁺ cells, known IL-9 producers in human allergic airway disease, exhibit enhanced STAT5 phosphorylation in allergic versus nonallergic patients. Furthermore, STAT5 activity in these cells correlated with increased chromatin accessibility at IL9 promoter and enhancer regions (16). Although a number of STAT5 target genes (e.g., IL5, IL9, IL13, GATA3, IL1RL1) were induced in human Th9 cells during an HDM-driven allergic response (22) (Fig. 1, Supplemental Fig. 1D, 1E), a STAT5-mediated suppressive signature was also prevalent in human IL9^hi cells as compared with IL9^lo Th2 cells (Fig. 1), suggesting that STAT5-mediated gene repression may also contribute to human Th9 cell development. Interestingly, IL9^lo cells exhibited a lesser STAT5-repressed gene signature and an elevated Th17-like signature, which suggests that STAT5 may regulate the balance of a Th9- versus Th17-like phenotype cells during human allergic disease.

In our studies, we observed increased IL-17 mRNA and protein expression by Th9 cells cultured in an IL-2–limiting environment (Fig. 2). Although a number of Th17-associated transcripts (e.g., Rora, Rorc, Ccr6) were increased in the absence of IL-2, these cells did not adopt a canonical Th17- or STAT3-associated transcriptional profile (Figs. 2, 3). Instead, these data suggest an alternative pathway of IL-17–producing cell differentiation. In support of this, IL-17–producing cells from IL-2–deprived Th9 conditions developed in the absence of the key Th17-associated factor STAT3 (Fig. 3). Interestingly, innate-like T cells (i.e., γδ TCR⁺, invariant NKT cells, and innate-like αβ T cells) also exhibit a similar STAT3-independent IL-17 pathway (38–40), whereas innate lymphoid cells (i.e., non–TCR-expressing cells) maintain a requirement for STAT3 for the production of IL-17 (41, 42). As innate lymphoid cells lack the capacity to receive signals through the TCR, this may suggest that TCR signaling compensates for lack of STAT3 in IL-17 production. Indeed, innate-like αβ T cells produced IL-17 only in response to TCR engagement in the presence of IL-1 (38). Likewise, STAT3-independent IL-17 production by IL-2–deprived Th9 cells in our studies was optimal in cells continuously cultured on anti-CD3–coated plates and was diminished in cells that were rested prior to restimulation (data not shown). Thus, TCR signaling is likely a critical factor that drives STAT3-independent production of IL-17 by all TCR⁺ cells.

In Th17 cells, STAT3 signaling induces the expression of Rorγt and BATF, which in turn, is required for Th17 cell differentiation (28, 31). In our studies, Th9 cells also expressed surprisingly high levels of Rorγt, and this expression was further enhanced upon IL-2 blockade (Figs. 2, 5). In contrast to canonical Th17 cells, Rorγt and BATF expression was STAT3-independent in IL-2–deprived Th9 cells, and ablation of these factors dramatically reduced IL-17A production under IL-2–deprived Th9-polarizing conditions. These data suggest that while IL-17 production under these conditions is STAT3-independent, it maintains a requirement for Rorγt and BATF. Interestingly, STAT3-independent production of IL-17 by essentially all innate-like T cells is associated with or requires expression of Rorγt (38, 43). However, the role of BATF in IL-17 production by innate-like T cells is more controversial. BATF and its transcriptional cofactor IRF4 are dispensable in for IL-17 production by certain subsets of γδTCR⁺ cells (43, 44). In contrast, BATF but not IRF4 is required for development of IL-17–producing NKT cells (45, 46). Given that IL-17 is critical for the antimicrobial response at anatomical barrier sites (i.e., skin, intestines, eyes, lungs), the evolution of multiple IL-17–inducing pathways likely ensures protection of these susceptible organs.

The data above indicate that multiple subsets of innate-like T cells can produce IL-17 in the absence of STAT3, and our data demonstrate that adaptive Th cells have similarly retained this capacity. Despite this, STAT3-independent production of IL-17 by Th cells is not commonly observed in vivo (31). One potential explanation for this may be the multiple roles of STAT3 in promoting Th cell survival/proliferation in vivo. STAT3 has been long known for its ability to promote Th cell survival by directly binding and inducing antiapoptotic and proliferative factors (i.e., Bcl2, Ier3, Fos, Fosl2) (33). Furthermore, recent data in Th17 cells demonstrates that STAT3 suppresses the antiproliferative functions of STAT1 and maintains mitochondrial membrane potential in established Th17 cells (30). STAT1 hyperactivation in cells with a STAT3 loss-of-function mutation also leads to induction of PD-L1, which inhibits IL-17 production in vivo (47), and may also influence Th cell survival. We demonstrated in this study that while STAT3 was not required for the initial differentiation of IL-17–producing cells in IL-2–deprived Th9 cultures, it was required for their prolonged survival after multiple rounds of differentiation (Supplemental Fig. 2). These data may explain why STAT3-independent IL-17–producing Th cells are not observed in the steady-state intestines, mouse models of disease, or in patients with STAT3 loss-of-function mutations (34, 48).

Work by our group and others has noted the relative instability of Th9 cells in culture (32) after adoptive cell therapy (3, 49) and in murine disease models (50). In culture, we demonstrated that Th9 cells rapidly lose their capacity to produce IL-9 upon secondary culture and gain the capacity to produce IL-10 and IL-17 (32). Although the loss of IL-9 was associated with increased STAT3 activation, blockade of IL-10–induced STAT3 in these cultures only moderately rescued IL-9 production and did not impair the transition to an IL-17–producing phenotype (32). Of note, this loss of IL-9 and gain in IL-17 production correlated with loss of responsiveness to IL-2 (32). We extended these findings in this study to show that STAT3 was also not required for the generation of IL-17–producing cells after long-term culture of Th9 cells (Fig. 4). These data indicate that when IL-2 is limiting, through either IL-2 blockade or natural decay of bioavailable IL-2, this favors the differentiation of STAT3-independent IL-17–producing Th cells. As a whole, these data indicate a role for the transcriptional inhibition by IL-2/STAT5 for maintenance of the Th9 phenotype and that this occurs at least partially through limiting the emergence of a Th17-like differentiation program. Furthermore, our data suggest that STAT5 transcriptional inhibition is a key player in Th9-driven immunopathology in human allergic disease.

The authors thank Sungtae Park and Nicole Anderson for helpful comments on this manuscript and Tripti Bera for technical assistance that made this work possible.

The authors have no financial conflicts of interest.