Peripheral blood mononuclear cells (PBMCs) of 7 ACR/EULAR classification criteria (2010)-defined RA patients and 10 healthy donors (HD) have been obtained by density centrifugation (FicoLite-H, Linaris, Wertheim, Germany) according to standardized laboratory procedures after written informed consent according to the principles of the declaration of Helsinki 2013 and ethical approval of the ethics committee at the University of Wuerzburg (protocol number 239/10). Exclusion criteria for RA and HD were malignoma, monogenetic syndromes, immunodeficiencies, administration of blood products in the past three months, severe infection or allergy requiring medical consultations in the past 8 weeks or vaccinations in the past 4 weeks. RA patients with biologicals or history of biological treatment were excluded. For demographics see Table 1 .

PBMCs were separated by two-step magnetic activated cell sorting (MACS) utilizing CD4+ T cell isolation kit and naive CD4+ T cell isolation kit (Miltenyi, Bergisch-Gladbach, Germany) according to the manufacturer’s protocol. Purity of CD4+CD45RA+ naive T cells was >95% as measured by flow cytometry (FACS Canto II, BD, New Jersey, USA) with FACSDiva V6 software (BD). Due to technical efforts to increase the purity of naive T cells for polarization experiments, the purity of non-naive T cells was less (about 50%).

Isolated naive CD4+CD45RA+ and non-naive CD4+CD45R0+ cells (2 x 10⁶ cells/well) were cultivated in RPMI1640 cell culture media (Lonza, Verviers, Belgium) supplemented with 10% fetal calf serum (FCS; Biochrom, Berlin, Germany) and 1% Pen/Strep (Biochrom) together with irradiated (30 Gy) autologous antigen-presenting cells (4 x 10⁴ cells/well). For co-stimulation, anti-CD3/anti-CD28 (1 µg/ml) (Biolegend, San Diego, USA) was added to each well in a 96-well plate. Cells were stimulated by a previously established Th17-inducing cytokine cocktail consisting of recombinant cytokines TGFβ (5 ng/ml) (Biolegend), IL-6 (20 ng/ml) (Immunotools, Friesoythe, Germany), IL-1β (10 ng/ml) (Immunotools) and IL-23 (100 ng/ml) (Biolegend) and by a previously established Th9-inducing cytokine cocktail containing IL-4 (20 ng/ml) (Immunotools), TGFβ (1 ng/ml) (Biolegend), IL-2 (5 ng/ml) (Immunotools) and anti-IFNγ (10 µg/ml) (Biolegend) for 3 days at 37°C, 5% CO2 and 95% humidity. The Th9-inducing cytokine cocktail was established in pilot experiments and modified after (1, 14–16).

For simulation of an IL-9 excess, IL-9 (20 ng/ml) (R&D Systems, Minneapolis, USA) was added to Th17- and Th9-inducing cytokine conditions and for inhibition of IL-9 anti-IL-9 (10 µg/ml) (R&D Systems) was added.

Resuspension of cells in FACS buffer containing phosphate-buffered saline (PBS) (pH 7.2), 0.5% bovine serum albumin (BSA) and 0.01% NaN₃ was performed to stain cells with monoclonal antibodies against various cell surface molecules. Monoclonal fluorochrome-labeled antibodies against CD3 were used to identify T cells, CD4 for helper T cells, CD45R0 and CD27 to define differentiation grade of T cells (naive CD45R0-CD27+, memory CD45R0+CD27+, effector CD45R0+CD27-, and effector II or terminally-differentiated-effector-memory T cell re-expressing CD45RA (TEMRA) by CD45R0-CD27-) (17). CCR6 was used as a chemokine receptor characteristic for Th17 cells. CD25+ and CD127- were used as additional markers to identify potential regulatory T cells (Treg) in addition to intracellular FoxP3 staining. Th9 cells were defined as IL-9+ CD4+ helper T cells. IL-9R was applied for identification of IL-9-sensitive cells. For intracellular staining, cells were permeabilized using permeabilization buffer and stained with monoclonal fluorochrome-labelled antibodies against FoxP3 (BD Bioscience) to identify Treg and activated T cells. Ki67 (BD Bioscience) was used to determine proliferating cells. Zombie aqua Dye (Biolegend) was used to exclude dead cells.

Intracellular cytokine production was analyzed following stimulation with phorbol 12-myristate 13-acetate (PMA) (0.03 µg/ml), ionomycin (1 µg/ml) and brefeldin A (10 µg/ml) (all purchased from Sigma-Aldrich, St. Louis, USA) for 4 h at 37°C. Cells were fixed for 20 min at room temperature using fixation buffer (BD, Franklin Lakes, USA) followed by a wash step with permeabilization buffer and stained using monoclonal fluorochrome-labelled antibodies against IL-9, IL-17A, IFNγ and IL-10 (all purchased from Biolegend), the latter two cytokines as key cytokines for Th1 and Treg, respectively. Surface and intracellular expression were assessed by flow cytometry (FACSCanto II, BD). Data analysis was performed using FACSDiva software V6 (BD).

In order to analyze IL-9 in the supernatants of cell cultures and in serum samples, an IL-9 ELISA was performed according to the manufacturer’s instructions (eBioscience, San Diego, USA).

Quantitative PCR for IL-9, IL-9R, IL-17, and PU.1 was performed using designed primers (Eurofins Genomics, Ebersberg, Germany) (for primer sequences see Supplementary Table 1 ). Total RNA was extracted from separated cells applying the RNA Nucleospin isolation kit (Machery Nagel, Düren, Germany) according to the manufacturer’s instructions. Complementary DNA (cDNA) was generated from 1µg RNA using oligodT₁₈ primers (Thermo Scientific, Waltham, MA, USA) for reverse transcription with Maxima reverse transcriptase (Thermo Scientific). Real-time PCR was performed using the applied Biosystems^® Real Time PCR 7500 (Applied Biosystems, Darmstadt, Germany) utilizing iTaq Universal SYBR green according to the manufacturer’s instructions (Biorad, Ismaning, Germany). Amplification was conducted for 40 cycles. Relative expressions of IL-9, IL-9R, IL-17, and PU.1 were determined by normalizing expression of each gene to β2-microglobulin.

Mann-Whitney U test for not normally distributed independent variables was applied after testing for distribution (Shapiro-Wilks test) with SPSS (Version 24, Chicago, IL). Wilcoxon Rank test was used for not normally distributed dependent variables. Bonferroni’s correction was performed to avoid bias by multiple testing. A p-value <0.05 was considered statistically significant.

In the naive CD4+ T cell fraction, stimulation by unspecific CD3/CD28 stimulation and Th9- and Th17-inducing cytokines, respectively, induced the switch to memory and effector T cell phenotypes as well as to Treg cells ( Supplementary Figure 1 and Table 2 ). Th9-inducing cytokines lead to significantly higher proportions of TEMRA cells compared to Th17-inducing cytokines in the naive CD4+ T cell fraction of HD. In the non-naive CD4+ T cell fraction, memory T cells did not change after Th9- or Th17-inducing cytokines. No significant difference was found between RA and HD.

Analysis of intracellular IL-9, IFNγ, IL-17 production, of CCR6 expression, IL-9R expression or of Ki67 in separated naive and non-naive CD4+ T cells before cell culture did not show any significant difference between RA and HD except for IFNγ-production which was higher in the naive fraction of HD ( Figure 1 and Supplementary Figure 2 ). After 3d cell culture of HD-derived naive CD4+ T cells, proportions of IL-9+ CD4+ T cells were elevated by Th9-stimulus and Th17-stimulus ( Figure 1A ). Despite limitation by moderate purity of separated non-naive T cells, increased proportions of IL-9+ CD4+ T cells were found in both Th9- and Th17-stimulated cultures of HD-derived non-naive CD4+ T cells. Intracellular IL-9 expression was highest in the effector CD4+ T cell subpopulation in switched previously naive CD4+ T cells ( Supplementary Figure 3 ).

In contrast to RA, in HD, Th17-stimulus significantly increased the proportion of IL-17+ ( Figure 1B ) and of IFNγ+ T cells in CD4+ ( Figure 1D ). This effect of higher IFNγ production in HD in differentiated cells of the previously naive T cell fraction was caused by the stimulation of effector CD4+ T cells producing IFNγ by Th17-inducing conditions ( Supplementary Figure 3 ).

A trend towards down-regulation of IL-9R cell surface expression could be found in the non-naive fraction by Th9- or Th17-stimulus ( Figure 1C ). However, both stimuli increased the Ki67+ proliferation of CD4+ T cells in a similar way in naive and non-naive CD4+ T cells in both HD and RA ( Figure 1E ). Although the CCR6 expression was higher in the non-naive fraction of CD4+ T cells in both HD and RA, CCR6 was neither induced by Th9-stimulus nor by Th17-stimulus ( Figure 1F ).

Proportions of Treg were similarly induced by Th9- and Th17-stimuli in naive CD4+ T cells of HD and of RA ( Table 2 ). Production of IL-10 by Treg was not different between Th9- and Th17-inducing stimuli ( Supplementary Figure 3 ). However, both stimuli were able to induce IL-17-production in Treg in HD in the naive T cell fraction after cell culture.

In order to analyze the overall IL-9 production and release into cell culture media, supernatants after 3d culture of separated naive and non-naive T cells were harvested and IL-9 determined by ELISA ( Figure 2 ). Th-9-inducing cytokine conditions lead to a significant increase of IL-9 concentrations in naive CD4+ T cells ( Figure 2A ), whereas in the non-naive T cell group both Th9- and Th17-stimuli drove IL-9 production ( Figure 2B ). In naive CD4+ T cells of RA, the Th17-stimulus significantly induced IL-9 production compared to unspecifically stimulated controls, whereas this difference was not significant in HD.

To adjust for disease activity, results were stratified into low, moderate and high disease activity according to DAS28 (<3.2, 3.2-5.2, >5.1, respectively). Highest IL-9 levels >6000 pg/ml were found in the two RA patients with highly active arthritis, whereas IL-9 concentrations of moderate and low DAS28 activity stayed in the range of concentrations determined in HD (data not shown). One HD had an extreme concentration of IL-9 of 9355 pg/ml without apparent explanation. No association between current medication and IL-9 concentrations was found in RA patients. Serum samples of HD and RA were negative for IL-9.

Molecular analysis of PU.1 transcription revealed similar values between RA and HD regardless which stimulus was applied ( Table 3 ). A significant change towards higher IL-9 mRNA expression could be shown for naive and non-naive CD4+ T cells after 3d culture under Th9- or Th17-stimulating conditions in RA, in HD only for non-naive CD4+ T cells.

In order to understand the possible involvement of IL-9-associated autocrine loops which may mitigate our results, an IL-9 excess was applied to Th9- and Th17-stimulated cultures as well as a blockade of IL-9 by a specific antibody ( Supplementary Tables 2 and 3 ). Inhibitory function of anti-IL-9 on IL-9 has been previously confirmed by determination of reduction of IL-9 in cell culture supernatants. A concentration of anti-IL-9 was applied leading to 50% inhibition of IL-9 in the Th0 control according to preliminary ELISA results. Neither IL-9 excess nor IL-9-blockade did change the differentiation of naive or non-naive T cell subpopulations in Th9- or Th17-stimulated cultures ( Supplementary Table 2 ). Intracellular cytokine production as well as IL-9R ( Figures 3A, B ) or CCR6 cell surface expression and Ki67 proliferation ( Figures 3E, F ) were not significantly influenced by this experimental approach in the naive CD4+ T cell fraction after 3d culture or in the non-naive T cell fraction ( Figures 3C, D, G, H and Supplementary Table 3 ). mRNA expression of PU.1, IL-9 and IL-9R was not significantly different regarding IL-9 excess or blockade. Also, IL-10 production in Treg was not influenced by IL-9 excess or IL-9 blockade in Th9- or Th17-stimulated cultures (data not shown). In the naive fraction, anti-IL-9 treatment of Th17-stimulated cultures caused a significant reduction of IL-9 concentrations in the supernatants of RA patients compared to HD ( Supplementary Figure 4 ). However, a trend to lower IL-9 concentrations secreted by naive or non-naive cells was seen in Th9- and Th17-driving cytokine conditions in both, HD and RA.