Our human study was performed at the Friedrich-Alexander-University Erlangen-Nürnberg, Germany, after being approved by the ethics review board of the University of Erlangen (Re-No: 56-12B; DRKS-ID: DRKS00005376). To date more than one hundred and seventy (170) patients that suffered from primary NSCLC and three metastatic patients underwent surgery and gave their approval to be enrolled in this study in an informed written consent. The patient studies were conducted in accordance with the ethical guidelines of the Declaration of Helsinki. Patients enrolled in this study did not receive any therapy. The confidentiality of the patients was maintained.

Lung cancer diagnosis was based on pathological confirmation. The histological types of lung cancer were classified according to the World Health Organization (WHO) in 2004. Staging was based on the Cancer TNM Staging Manual formulated by the International Association for the Study of Lung Cancer (IASLC), issued in 2010. Clinical data including histological classification, TNM stage, age, gender and smoking status were provided by the Department of Thoracic Surgery and the Institute of Pathology and are summarized in Table S1 . Clinical data of the control cohort are shown in Tables S2–S4 .

Immediately after surgery, lung tissue samples were taken from three different regions: the tumoral region (TU: solid tumor tissue), peri-tumoral region (PT: 2-3 cm away from the solid tumor) and the tumor-free control region (CTR: > 5 cm away from the solid tumor). The post-surgery tissue samples were used for RNA and protein isolation as well as for total cell isolation followed by cell culture and FACS analysis. Paraffin-embedded lung tissue arrays were generated as previously described (25) and applied for immunohistochemistry (IHC).

Antibodies used in this study are listed in the Supplementary Material .

Immunohistochemistry was performed on paraffin-embedded histological sections. Before staining, paraffin was removed from the slides by incubation at 72°C for 30 min and treatment with Roti-Histol (Carl Roth) two times for 10 min. The tissue sections were then rehydrated by immersion in ethanol-series in descending concentrations (100%, 95%, 70%) for 3 min each and in deionized water for 1 min, followed by blocking endogeneous peroxidase in 3% H₂O₂ (in methanol) for 20 min. For heat-induced antigen retrieval, slides were placed into a rack containing 50 ml 1 mM Tris-EDTA buffer which was transferred into a pressure cooker followed by incubation at 120°C for 5 min. Slides were then cooled down for 30 min at room temperature (RT) followed by incubation for 1 min in deionized water. Tissue was surrounded with a hydrophobic barrier using a barrier pen. In a next step, slides were stained with the respective primary antibody against CD3, Foxp3, IL-9 or IL-9R ( Table S5 ) using the ZytoChem-Plus AP Polymer-Kit (Zytomed Systems GmbH) according to the manufactures instructions. For IHC single stainings, nuclei were stained with hematoxylin solution (Carl Roth) and slides were covered with coverslips using Aquatex (Merck). For IHC double staining, the second antibody against IL-9 ( Table S5 ) was applied and detected according to the manufacture instructions of the Dako EnVision Detection System Kit (Dako Deutschland GmbH). Negative controls were not treated with the primary antibody; the other steps remained the same. Stained slides were scanned using the digital slide scanner (Scan 150) at the Institute of Pathology. Whole slide images were visualized by the CaseViewer software (Version 2.0, 3D Histech Ltd). IL-9 single staining was quantified using the Definiens Tissue Studio 4.1 software (Definiens) while the IL-9R single and the IL-9/Foxp3 double staining have been evaluated using the ImageJ Cell Counter (Version 1.46).

Human tissue samples were cut into small pieces (1-3 mm²) using scalpels and digested with Collagenase (2700 U/ml, Sigma-Aldrich: Collagenase from Clostridium histolyticum, Cat# C98991-500MG) and 150µl DNase (10 mg/ml, Roche Diagnostics GmbH: DNase I, Cat#10104159001) diluted in 10 ml R10 medium (500 ml RPMI1640, anprotec, Cat# AC-LM-0060; 50 ml heat-inactivated fetal bovine serum (FCS), Sigma-Aldrich, Cat# S0615; 2mM L-Glutamine, anprotec, Cat# AC-AS-0001; 5 ml Penicillin-Streptomycin (Pen/Strep), anprotec, Cat# AC-AB-0024) in a shaker at 37°C for 1 h. After incubation, the samples were passed through a cell-strainer (100 µm, Greiner Bio-One GmbH: EASYstrainer; Cat# 542000), single-cell suspensions were washed with RPMI1640 and centrifuged at 300g for 7 min at 4°C. The supernatant was removed, and red blood cells (RBC) were lysed in ACK lysis buffer (0,15M NH4Cl, Carl Roth GmbH + Co. KG, Cat# P726.2; 0,01M KHCO3, Carl Roth GmbH + Co. KG, Cat#P748.1; 100M Na2EDTA, GERBU Biotechnik GmbH, Cat# 1034.1000; dissolved and steril filtered in deionized H2O; pH=7.2-7.4) with subsequent centrifugation at 1500rpm for 5 min at 4°C. The cell pellets were washed with Washing Buffer I (500 ml PBS EDTA pH 7.5, BioWhittaker, Cat# BE02-017F; 5 ml Pen/Strep), centrifuged again and resuspended in Washing Buffer II (500 ml PBS EDTA pH 7.5; 5 ml Pen/Strep; 25 ml FCS). Cells were centrifuged at 800rpm for 15 min at 4°C with a low deceleration mode 1. Supernatant was discarded and cells were taken up in Tumor tissue medium (500 ml RPMI1640; 50 ml FCS; 2mM L-Glutamine; 5 ml Pen/Strep; 1mM Sodium-Pyruvate, anprotec, Cat# 11360-070; 5 ml Non-Essential Amino Acids, Gibco, Cat# 11140-035). Cell numbers were determined using Trypan-blue staining in a Neubauer counting chamber. Cells were cultured in Tumor tissue medium at 37°C and 5% CO2.

PBMCs were freshly isolated using LeucoSep Tubes (Greiner Bio-One, Cat# 227290) according to the manufacturer´s protocol. After cell counting with Trypan-blue in a Neubauer counting chamber, PBMCs were cultured for 4-5 days with 5x10⁵ cells per condition as shown in Figure 2E .

Freshly isolated PBMCs from NSCLC patients and healthy control subjects were cultured in 1ml R10 medium at 5 x 10⁵ cells/well for 4-5 days with plate bound anti-CD3 (1µg/well) and soluble anti-CD28 antibodies (10µg/ml) in a 48 well cell culture plate (Greiner Bio-One, Cat# 677180) at 37°C and 5% CO₂ ( Figure 2E ). For skewing of IL-9 producing T cells, TGFβ (20ng/ml) and IL-4 (20ng/ml) were added, while the Treg skewing condition included TGFβ (20ng/ml) and IL-2 (2ng/ml). The respective cytokine information are listed in the Table below:

To prevent unspecific binding of antibodies to Fc receptors, single cell suspensions were incubated with unlabeled antibodies against IgG Fc receptors CD16/32 (human: 1:100, BD Biosciences Cat# 564220; murine: 1:100, BD Biosciences Cat# 553142) in FACS Buffer (500 ml PBS, anprotec Cat# AC-BS-0002; 10 ml FCS) for 5 min at 4°C. Cells were washed with FACS Buffer and centrifuged. Pellets were resuspended with FACS Buffer containing antigen-specific fluorochrome-conjugated antibodies in appropriate concentrations and incubated for 15 minutes at 4°C in the dark. After cells were washed with FACS Buffer and centrifuged, pellets were either resuspended in FACS Buffer to measure directly the surface stained cells or proceeded to intracellular staining. Therefore, the FoxP3 staining kit (Thermo Fisher Scientific Cat# 00-5523-00) was applied according to the manufacturer’s protocol. Permeabilized cells were stained for intracellular proteins in a volume of 50 µl PermWash Buffer supplemented with antibodies in appropriate concentrations. Cells were washed, taken up in FACS Buffer and analyzed immediately ( Table S6 ). Flow cytometric analyses were obtained with the FACS Canto II and the BD FACS Diva software (both BD Biosciences) for compensation. Data were analyzed with Flow-Jo v10.2 (FlowJo) software always gating for singlets using the forward scatter and for living cells using forward scatter vs. sideward scatter before specific analysis (also shown in Supplementary Material ).

Cells were first stained with Zombie (1:500, Biolegend, Cat#423101/423102) for 15 min at room temperature prior to Fc blocking and washed twice with FACS buffer. The staining of surface proteins was then continued as described above.

Data of applied FMOs (Fluorescence minus one) are shown in Supplementary Material .

Lung tissue samples were homogenized using Precellys Lysing Kits (Bertin Technologies, Cat# P000918-LYSK0-A) and the benchtop homogenizer Minilys (Bertin Technologies) as described in the manufacturer’s protocol. RNA of homogenized samples or single cell suspensions was isolated using peqGold RNA Pure (Peqlab, Cat# 30-1010) or Qiazol Lysis Reagent (Qiagen, Cat# 79306) according to the manufacturer´s instructions. RNA was reversely transcribed into cDNA using the RevertAid First Strand cDNA Synthesis Kit (ThermoFisher Scientific, Cat# K1622) according to the manufacturer’s protocol.

qPCR of synthesized cDNA was performed by using iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories, Cat# 1725124) in a total volume of 20 µl. Primer sequences for murine and human qPCR analyzes are depicted in Supplementary Table S7 which were purchased from Eurofins-Genomics Germany. qPCR Reactions (50 cycles, initial activation 98°C, 2 min, denaturation 95°C, 5 min, hybridization/elongation 60°C, 10 min) were performed using the CFX-96 Real-Time PCR Detection System and analyzed by the respective CFX Manager Software (both Bio-Rad Laboratories). Data were analyzed using the relative quantification 2^-ΔΔCT method by normalization to the housekeeping-gene hypoxanthine-guanine-phosphoribosyltransferase (Hprt).

The enzyme-linked immunosorbent assay (ELISA) technique was utilized to analyze the cytokine concentration in cell culture supernatants, BALF and Serum. ELISA was performed in accordance with the manufacturer’s instructions. Human ELISA Sets were purchased as followed: IL-9 ELISA (Human IL-9 DuoSet ELISA, R&D Systems, Cat# DY209-05), IL-21 ELISA (Human IL-21 DuoSet ELISA, R&D Systems, Cat# DY8879-05), IFNγ ELISA (BD OptEIA™ Human IFN-γ ELISA Set, BD Biosciences, Cat# 555142). Murine samples were analyzed by: IL-21 ELISA (Mouse IL-21 DuoSet ELISA, R&D Systems, Cat# DY594), IFNγ ELISA (BD OptEIA™ Mouse IFNγ ELISA Set, BD Biosciences, Cat# 555138), IL-10 ELISA (BD OptEIA™ Mouse IL-10 ELISA Set, BD Biosciences, Cat# 555252).

Lung tissue samples from patients were lysed with RIPA Buffer (Thermo Fisher Scientific, Cat# 89900) and inhibitor cocktail (complete Mini EDTA-free, Roche Diagnostics, Cat# 11836170 001) then homogenized with SpeedMill PLUS (Analytik Jena) in innuSPEED lysis Tubes P (Analytik Jena, Cat# 845-CS-1020250) and centrifuged (3000rpm, 5min, 4°C). The supernatant was incubated for 30 minutes on ice and centrifuged again (2000g, 5min, 4°C). Final extraction was done by another centrifugation step of the supernatant (45min, maximum speed, 4°C) and measured with Bradford Assay (Protein Assay Dye Reagent Concentrate, Bio-Rad, Cat# 5000006), followed by protein denaturation (95°C, 5 min) in a mix of reducing loading buffer (4xLDS Sample Buffer, Thermo Fisher Scientific, Cat# NP0007) and DTT (1M, Thermo Fisher Scientific, Cat# P2325). A mini-PROTEAN TGX Stain-Free Gel (Bio- Rad, Cat# 4568086) was used with 20µg of proteins per well and at 120V, 150mA for 1h. Proteins were then transferred to a 0,2µm nitrocellulose membrane (Trans-Blot Turbo Transfer Pack, Bio-Rad, Cat# 1704158) by using a western blot trans-blot turbo system (Bio-Rad). Transfer and protein load was assessed and recorded by using a ChemiDoc Imaging System (Bio-Rad). After washing and blocking (3% powdered milk (Carl Roth GmbH, Cat# T145.3) in Tween SDS buffer) the membrane for 1h at room temperature, the primary antibody was applied and incubated over night at 4°C. After that, the membrane was washed and incubated with the compatible secondary antibody in blocking buffer, for 1h at room temperature. Respective antibody information are listed in Table S8 . After final washing steps, detection was performed using SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Scientific, Cat# 34095). Visualization of the western blot results was done by using the ChemiDoc Imaging System. Protein bands and total protein levels were analysed with the ImageLab software (Bio-Rad).

The human A549 (ATCC^® CCL-185™) cell line was purchased authenticated from the ATCC bank (Manassas, Virginia, USA). Cells were cultured in F-12K Nut mix medium (Thermo Fisher Scientific, Cat# 21127-022) supplemented with 10% FCS, 1% Pen/Strep and 1% L-Glu at 37°C and 5% of CO₂.

The murine LL/2-luc-M38 (LL/2) cell line was purchased and authenticated from Caliper LifeScience (Bioware cell line, Caliper LifeScience, Waltham, Massachussets, USA). All Caliper Life Sciences cell lines are confirmed to be pathogen-free by the IMPACT profile I (PCR) at the University of Missouri Research Animal Diagnostic and Investigative Laboratory. LL/2 cells were cultured in D10 medium (anprotec, Cat# AC-LM-0012; 50 ml heat-inactivated fetal bovine serum (FCS), Sigma-Aldrich, Cat# S0615; 2mM L-Glutamine, anprotec, Cat# AC-AS-0001; 5 ml Penicillin-Streptomycin (Pen/Strep), anprotec, Cat# AC-AB-0024) at 37°C and 5% of CO₂.

The murine bronchoalveolar carcinoma cell line L1C2 (Dubinett et al. https://pubmed.ncbi.nlm.nih.gov/8382559/) was obtained from Prof R. Wiewrodt. Cells were cultured in R10 Medium at 37°C and 5% CO2.

3x10⁵ A549 were cultured overnight in 6-well plates (Greiner Bio-One, Cat#657160) in supplemented F-12K Nut mix or R10 medium with or without 50 ng/ml IL-9 (ImmunoTools, Cat# 11340093) for 72h. Other different experimental conditions were described in the respective figure. For the Fas agonistic challenge, we incubated A549 with 2 μg/mL anti-CD95 (Biolegend, Clone DX2, Cat# 305655) or IgG1, k (Biolegend, Clone MOPC-21, Cat# 400153).

1.25x10⁵ LL/2 cells were cultured in 24-well cell culture plates (Greiner Bio-One, Cat# 662160) for 24h at 37°C and 5% CO2 in D10 medium with 30 ng/ml or 100 ng/ml IL-9 or left untreated. Cells were harvested and counted for live and dead cells using trypan blue staining.

5x10⁵ LL/2 or L1C2 cells were cultured overnight in 6-well plates in D10 or R10 medium, respectively. After the first 24h the cells were stimulated with 30 ng/ml or 100 ng/ml IL-9 or left untreated for another 24h.

Determination of apoptotic cells was performed using Annexin V/PI staining according to the manufacturer’s protocol (BD Bioscience; Cat#550474 and #556463). Briefly, 1x10⁵ cells were stained with 5 μl AnnexinV-APC and 5 μl PI in 100 μl 1x AnnexinV binding buffer for a total volume of 110 μl and incubated for 15 min at RT in the dark. Reaction was stopped by adding 200 μl 1x AnnexinV binding buffer and samples were immediately analyzed with the FACSCanto II. Data sets were evaluated by Flow-Jo v10.2.

Determination of proliferating cells was performed using Ki67 staining according to the manufacturer’s protocol (BD Bioscience, Cat# 556026). For Ki67 analysis, 1x10⁶ LL/2 or L1C2 cells were fixed with 75% ethanol and incubated for at least 2 hours at -20°C. After washing the cells with FACS buffer the cells were stained with 20 µl Ki67-FITC diluted in FACS buffer for a total volume of 120 µl and incubated for 30 min at RT in the dark. Reaction was stopped by adding 1 ml PBS, centrifuged and resuspended in 250 µl PBS. Cells were immediately analyzed using the FACSCanto II. Data sets were evaluated by Flow-Jo v10.2.

All mice were bred and maintained under specific-pathogen-free (SPF) conditions in our animal facility (University Hospital Erlangen, Hartmannstraße 14, Erlangen). All experiments were performed in accordance with the German and European laws for animal protection and were approved by the local ethics committees of the Regierung Unterfranken (Az 55.2-2532-2-1286-20). IL-9^-/- mice were bred on Balb/c and C57BL/6 genetic backgrounds. And they were kindly presented by PD Dr. Benno Weigmann and PD Dr. med. Andreas Ramming from University Hospital Erlangen Medicine 1 and Medicine 3, respectively (26).

For lung tumor induction, 5x10⁵ cells resuspended in 200 µl DMEM medium (without supplements) were injected into the tail vein of 6-8 weeks old female mice. At the indicated time points, mice were weighed and injected intraperitoneally (i.p.) with luciferin (0.15 mg per 1 g body weight; Promega, Cat#P1043). Luciferase activity was measured after 20 min by the IVIS Spectrum In Vivo Imaging System (PerkinElmer) as previously described (5). Briefly, mice were anaesthetized using isoflurane and luciferase activity was measured by detecting luminescence intensity (photons per second). Analyses were performed in a logarithmic scale mode. Mice were sacrificed at day 14-23 after tumor cell injection. For the inhibition of IL-9 in vivo, we used two different protocols. In the first protocol, mice were treated i.p. with 200 µg of anti-IL9 antibody (BioXCell, Clone 9C1, Cat# BE0181) or IgG2a Isotype control (BioXCell, Clone C1.18.4, Cat# BE0085) dissolved in 100 µl PBS at day 1, 3, 6, 9, 13 and 16 after tumor induction with an experiment termination at day 21 to 23 depending on the lung tumor load. In the second protocol, mice were treated i.p. with 20 µg anti-IL9 antibody or IgG2a Isotype control resolved in 200 µl PBS at day 6, 9, 10 and 13 after tumor induction with an experiment termination at day 20.

For lung tumor induction by L1C2, 2x10⁵ cells resuspended in 200 µl DMEM or RPMI medium (without supplements) were injected into the tail vein of 6-8 weeks old female mice. Mice were sacrificed at day 14-23 after tumor cell injection, as indicated.

Representative parts of the sectioned mouse lung were fixed in 10% formalin-PBS solution, dehydrated, and embedded in paraffin. Five-micrometer-thick lung sections cut from paraffin blocks were stained with hematoxylin and eosin for visualization of lung tumors. Stained slides were scanned using a digital slide scanner (Scan 150, 3D Histech Ltd) and slide images were visualized using the CaseViewer software (Version 2.0, 3D Histech Ltd). Staining and scanning of slides was performed in cooperation with the Institute of Pathology in Erlangen

To analyze the tumor load in the lung for each mouse, the lung sections were histologically evaluated after Hematoxylin and Eosin staining. A pathologist analysed in a blinded codified manner the tumor load of the murine lungs. Moreover, each section was analyzed by means of area [µm²] and number of foci. The sections were investigated under the microscope with an object magnified 10 times (Carl Zeiss, AxioCam MRC) and live pictures were directly transmitted to the Zen 2 blue edition software (Carl Zeiss Microscopy GmbH). With the help of the graphic tool, the foci were localized, circled and automatically measured by means of area. The obtained data corresponding to each mouse was used for comparisons as well as correlation analyzes and transferred into graphs (GraphPad Prism 8).

Dissected lungs were cut into small pieces using scalpels and digested with Collagenase (2700 U/ml; Sigma-Aldrich: Collagenase from Clostridium histolyticum; Cat# C98991-500MG) and 1,5 mg DNase (10 mg/ml; Roche Diagnostics GmbH: DNase I; Cat#10104159001) diluted in 10 ml R10 medium in a shaker at 37°C for 1 h. After incubation the samples were passed through a cell-strainer (40 µm, Company, Cat#542040), single-cell suspensions were washed with RPMI1640 and centrifuged at 1500rpm for 10 min at 4°C. The supernatant was removed and RBC were lysed in 10 ml ACK lysis buffer with subsequent centrifugation at 1500rpm for 5 min at 4°C. The cell pellets were washed with 10 ml PBS and centrifuged at 1500rpm for 5 min at 4°C. After removal of the supernatant, the cells were taken up in 10 ml R10 medium and cell numbers were determined using Trypan-blue staining in a Neubauer counting chamber.

CD4+T cells were isolated from the lungs of tumor-bearing mice by magnetic cell separation using anti-CD4 MicroBeads (Miltenyi Biotec) according to the manufactures protocol. The purity of isolated CD4+ T cells was confirmed by FACS analysis. Isolated T cells were then cultured in RPMI medium supplemented with 10% FCS, 1% Pen/Strep and 1% L-Glu at 37°C and 5% CO₂, in the presence plate-bound anti-CD3 (5 µg/ml, BD Biosciences, Cat#) and soluble anti-CD28 (1 µg/ml, Biolegend, Cat#) antibodies. Supernatants and cells were harvested after 24h of cell culture for further analyses.

Naïve CD4 T cells were isolated from mouse spleens using the CD4⁺CD62L⁺ T cell isolation kit according to the manufacturer’s protocol (Miltenyi Biotec). Cells were cultured in R10 medium on anti-CD3 (2 µg/ml; BioXCell) coated cell culture plates with soluble anti-CD28 (2 µg/ml; BioXcell). Cells were cultured under Treg polarizing conditions including hTGF-β1 (2 ng/ml), hIL-2 (50 U/ml), anti-IFN-γ (XMG; 10 mg/ml) and anti-IL-4 (11B11; 10mg/ml). Th9 cells were cultured with hTGF- β1 (2 ng/ml), IL-4 (20 ng/ml), hIL-2 (50 U/ml) and anti-IFN-γ (10 mg/ml). Th0 cells were cultured with hIL-2 (50 U/ml), anti-IFN-γ (XMG; 10 mg/ml) and anti-IL-4 (11B11; 10mg/ml). On day 3, cells were expanded into fresh media containing the original concentrations of cytokines in the absence of co-stimulatory signals for additional 2 days. On day 5, mature T cell subsets were harvested for further analysis.

For transcription factor staining in Treg cells from different culture conditions were harvested on day 5 of differentiation whereas for cytokine staining, CD4+ T cells were stimulated with Phorbol 12-myristate 13-acetate (PMA, 5ng/ml, Sigma-Aldrich) and ionomycin (500ng/ml, Sigma-Aldrich) for 3 hours followed by monensin (2μM, Biolegend) for total 6 hours at 37°C. Cells were washed with FACS buffer (PBS with 0.5% BSA). CD4+ T cell subsets were then stained with a fixable viability dye (eBioscience) and surface markers (CD4, RM4-4, Biolegend; CD25, PC61, Biolegend) for 30 min at 4°C followed by washing and fixation with 4% formaldehyde for 10 min at room temperature. For transcription factor staining, after surface staining, cells were fixed with Fixation & Permeabilization Buffer (FoxP3 staining kit, Thermo Fisher Scientific, Cat# 00-5523-00) for 2 hours at 4°C, and then permeabilized with permeabilization buffer. Cells were stained with Foxp3 antibody (Foxp3, FJK-16s, eBioscience) for 30 min at 4°C followed by flow-cytometry. For cytokine staining, after fixation cells were then permeabilized with permeabilization buffer, and stained for cytokines (IL-9; RM9A4, Biolegend).

Total RNA was isolated from cells using TRIzol (Life Technologies). RNA was reverse transcribed according to manufactures directions (Quantabio, Beverly, MA). Quantitative Reverse Transcriptase (qRT-PCR) was performed with commercially available primers (Life Technologies) with a 7500 Fast-PCR machine (Life Technologies). Gene expression was normalized to housekeeping gene expression (β2-microglobulin).

For measurement of IL-9 secretion, on day 5, 1x10⁶ cells were re-stimulated with anti-CD3 coated on 12 well cell culture (Greiner Bio-One, Cat# 665180) plates for 12 hours. Supernatants were collected and IL-9 secretion was measured using IL-9 ELISA MAX™ kit provided by BioLegend (Cat#442704).

Naïve CD4⁺CD62L⁺ T cells were isolated from the spleens of naïve mice by magnetic cell sorting using the CD4⁺CD62L⁺ T Cell Isolation Kit II (Miltenyi Biotec). The purity of the cell isolation was confirmed by FACS analysis.

For T cell differentiation, CD4⁺CD62L⁺ T cell were cultured in RPMI medium supplemented with 10% FCS, 1% Pen/Strep and 1% L-Glu, at 37 °C and 5 % CO₂ in the presence of plate-bound anti-CD3 (5 µg/ml, BD Biosciences) and soluble anti-CD28 (1 µg/ml, Biolegend) antibodies.

For Th1 differentiation, cells were additionally treated with 5 µg/ml anti-IL-4 antibody (BioLegend) and 12 ng/ml recombinant IL-12 (PeproTech).

Afterwards ACK lysis buffer was applied as previously described (27).

All data analysis was done using Prims software version 8 (Graphpad Software) using a Student’s test, ordinary one-way ANOVA or two-way ANOVA. Post hoc Tukey test was used for multiple comparisons. Simple linear regression was used for the correlation. P<0.05 was considered statistically significant.

In this study, we aimed for characterizing the expression and function of IL-9, a cytokine with immune-regulatory functions (28, 29), in non-small cell lung cancer (NSCLC).

To investigate the IL-9 expression in the lung of patients with NSCLC, we did immunohistochemistry for IL-9 on post-surgery lung samples collected on tissue arrays and compared IL-9 production from the tumoral region (TU: solid tumor tissue), and the tumor-free control region (CTR: > 5cm away from the solid tumor) in NSCLC (see Materials and Methods and Figure 1A ). The clinical data of these patients as well as the healthy subjects whose post-surgery lung samples were analyzed in this study, are reported in Tables S1, S2 , respectively. This analysis showed induction of IL-9+ cells in the tumoral region of the lung as compared to the control region in these lung sections derived from patients with NSCLC ( Figure 1B and Supplementary Figure 1A ). Patients included in this study were not treated with any immunotherapy before undergoing surgery analysis. To further analyze the IL-9 induction in the tumoral region of these patients, we next analyzed interleukin 9 expression in proteins extracted from the control and tumoral regions of the lung of NSCLC patients. Furthermore, we analyzed the lung of 3 healthy controls (CN) and one subject with lung inflammation (Lu-infl.) by using Western blot analysis ( Figures 1C–E and Supplementary Figure 1B ). Here, we confirmed upregulation of IL-9 in the tumoral region of the lung of patients with NSCLC ( Figures 1C, D ).

We next started to investigate the cellular sources of IL-9 in NSCLC. Since IL-9 was induced in the tumoral region of the lung, we asked whether the tumor cells themselves would produce IL-9. To do so, we cultured the lung tumor adenocarcinoma cell line A549 and extracted proteins for performing Western blot analysis for IL-9. We found that A549 cells produce IL-9 although not as high as control healthy lungs when corrected for GAPDH expression ( Figure 1E ).

We further investigated IL-9 in epithelial cells isolated from the lung of one NSCLC patient and found decreased numbers of epithelial cells (EpCAM+) producing IL-9 (IL-9+ cells) in the tumoral region as compared to the control region by flow cytometric analysis ( Figure 1F ). Furthermore, we found few epithelial cells expressing IL-9 distributed in the different regions of the lung of the patients with NSCLC by immunohistochemistry (IHC) ( Supplementary Figure 1A ). We next further characterized the cellular source of IL-9 in lung tissue from NSCLC patients and focused our analysis on IL-9 production by tumor infiltrating T cells (TIL) and performed double immunohistochemistry with anti-IL-9 and anti-CD3 antibodies in lung tissue arrays as in Figure 1B . An example of this staining is shown in Figure 1G . Here, we found double positive cells, indicating the presence of lung CD3+ T cells producing IL-9, which infiltrated the tumoral regions of the lung in NSCLC. Accordingly, we analyzed IL-9 production of total lung cells from the three lung tumor regions after T cell stimulation with anti-CD3/CD28 antibodies in a cohort of NSCLC patients by ELISA ( Figure 1H ). Here, we noted significantly augmented IL-9 production by cultured lung cells stimulated with aCD3/28 antibodies from the TU region in NSCLC as compared to the CTR and PT regions of these patients. As IL-9 from T cells is mainly produced by Th9 cells which also produce IL-21 (20, 28), we next determined IL-21 concentration in the same cell supernatants where IL-9 was detected. We found a significant upregulation of IL-21 production in the supernatants of aCD3/aCD28 antibodies cultured total lung cells from the TU region of patients with NSCLC as compared to the CTR region ( Figure 1I ). This finding indicates higher T cell infiltration as the possible cause for the increased IL-9 levels found in the tumoral area of the lung of NSCLC patients (30). By contrast, the anti-tumoral cytokine IFN-ɣ was found, by trend, downregulated in these lung cells isolated from the tumoral region ( Figure 1J ). In confirmation of decreased anti-tumoral cells infiltrating the tumoral region of NSCLC patients, we found significantly less TNF-alpha mRNA levels in mRNA isolated from the lung tumoral regions as compared to the respective control region ( Figure 1K ).

In conclusion, we found that TIL in the tumoral region of NSCLC patients produced IL-9 while challenged with anti-CD3/CD28 antibodies. Moreover, in addition to tumor infiltrating lymphocytes, also lung tumoral cells have the capacity to produce IL-9.

We previously described induced immunosuppression in the tumoral region of the lung of patients with NSCLC (31, 32). To determine whether regulatory T (Treg) cells would produce IL-9 in the lung of patients with NSCLC, we performed double immunohistochemistry for IL-9 and Foxp3 in tissue array slides ( Figures 2A, B ). Here, we found an increased number of double positive cells for Foxp3 and IL-9 in the tumoral region compared to the control region of the lung of NSCLC patients suggesting that lung Treg cells may produce IL-9 in NSCLC. Consistently, we found a positive correlation between the Foxp3 mRNA levels and IL-9+ T cells in the tumoral area in NSCLC ( Figure 2C ).

Treg cells are known to be induced by low levels of IL-2, hereby competing with TIL for IL-2 induced survival (33, 34). By having a view on IL-2, we noted reduced IL-2 mRNA expression in the tumoral region compared to the control region ( Figure 2D ), indicating that low levels of IL-2 and the presence of other immunosuppressive cytokines like TGF-beta and IL-10 in the tumor microenvironment may contribute to augmented Treg numbers in NSCLC (31, 35).

To determine the capacity of Treg cells to produce IL-9, we isolated peripheral blood mononuclear cells (PBMCs) from five NSCLC patients and five healthy control subjects. Next, the PBMCs were cultured under T cell producing IL-9 and Treg skewing conditions with anti-CD3 and anti-CD28 antibodies. After 4-5 days of cell culture, we performed FACS analysis and measured IL-9 in the supernantant by ELISA ( Figures 2E, F and Figure S2A ). In the cell supernatants from PBMCs derived from both healthy controls as well as NSCLC patients, we observed an induction of IL-9 under conditions inducing IL-9 (TGF-beta+IL-4) as well as under immunosuppressive conditions (TGF-beta+ low IL-2) ( Figure 2F ). Moreover, we could detect induced CD3+CD4+CD25+Foxp3+ T regulatory cells both in Treg and IL-9 inducing conditions in PBMCs from both healthy control subjects and from NSCLC patients ( Figures 2G, H ). Thereby the NSCLC group showed an enhanced induction of Treg frequency. These data indicate that factors present in the tumor microenvironment like TGF-beta and low IL-2, can induce IL-9 in TILs from patients with NSCLC and simultaneously induce Foxp3+ T regulatory cells. In addition, our results from the peripheral blood demonstrate that immunosuppression can drive IL-9 both in healthy and in NSCLC subjects.

Lung cancer immunosuppression is mediated via effects of tumor cells on TILs resulting in T cells expressing Foxp3+ and IL-10 (35). To determine whether this concept is also relevant for tumor cells and TILs in NSCLC patients, we next assessed the distribution of IL-9R in patient samples. To this end, we performed immunohistochemistry for IL9R on the lung tissue arrays of our cohort of NSCLC patients and found an upregulation of IL-9R in the tumoral region as compared to the control region and healthy controls ( Figure 3A ).

Flow cytometric analysis of TILs from lungs of NSCLC patients revealed an upregulation of IL-9R expression in the CD3+ T cells localized in the peritumoral and tumoral region of the lung as compared to control regions ( Figure 3B , upper graph). We next addressed IL-9R expression in epithelial cells, as this is developed in lung tumors upon epithelial transformation. We found that IL-9R expression was increased by trend in Epcam+ epithelial cells in the lung tumoral region, as compared to the control and the peri-tumoral regions in patients with lung adenocarcinoma, indicating that normal epithelial cells as well as tumor cells express comparable amounts of IL-9R ( Figure 3B ).

Finally, we performed FACS analysis in the human adenocarcinoma lung cancer cell line A549 to understand if IL-9R is expressed in these tumor cells and detected IL9R expression ( Figures 4A, B ) in A549 cells. The gating strategy for FACS staining is shown in Supplementary Figure 3A . In summary, subsets of tumor infiltrating lymphocytes in the lung as well as lung adenocarcinoma cells express IL9R and may thus be targets for immunomodulatory and cancer promoting effects of IL-9.

We next asked if IFN-γ would regulate IL-9 signaling. Here we found that, IFN-ɣ, but not IL-21 and TNF-α treatment significantly downregulated IL-9R expression on A549 cells ( Figure 4B ). Moreover, independently from IL-9, we found that, TNF-alpha induced IFN-ɣ receptor (CD119) levels on A549 cells ( Figure 4C ), indicating an opposite role of IFN-ɣ and TNF-alpha as compared to IL-9 on lung tumor cells.

It has been reported that IL-9 induced A549 proliferation (36). Therefore, we investigated if IL-9 would influence the tumor cell survival and death. Here we found that, as opposed to the anti-tumoral cytokine IFN-ɣ, IL-9 suppressed lung tumor cell apoptosis ( Figure 4D ). Furthermore, pre-treatment of A549 cells with IFN-ɣ reversed the anti-apoptotic effect of IL-9 ( Figure 4D ). Taken together these data indicate a pro-tumoral role of IL-9 in lung cancer.

In search of the molecular mechanism of IFN-ɣ’s apoptotic effect on lung tumor cells, we discovered FAS-R (CD95) induction on A549 cells mediated by IFN-ɣ and an inhibition of FAS-R expression by IL-9 treatment ( Figure 5A ).

We next treated A549 cells either with IFN-ɣ or pretreated them with IL-9 and use agonistic antibodies to activate FAS-R (CD95). Here we could confirm that IFN-ɣ induced specifically FAS-R ( Figures 5B, C ). Moreover, treatment with agonistic anti-CD95 antibodies resulted in induction of IFN-ɣ mediated A549 cell death ( Figure 5D ).

By contrast, agonistic FAS-R treatment did not affect the effect of IL-9 on apoptosis demonstrating that IFN-ɣ-FAS mediated apoptosis is not influenced by IL-9. To prove that IFN-ɣ mediated A549 apoptosis, but not IL-9 induced tumor apoptosis via FAS-R.

To further explore the effects of IL-9 on Treg cells, we analyzed IL-9 production and IL-9R expression by T regulatory cells differentiated in vitro in naïve mice. In addition, we added IL-9 and anti-IL-9 antibodies to see their effects on Tregs. As comparison we also set up Th9 skewing conditions. It was found that IL-9 is produced by Treg skewed lymphocytes although at lower levels as compared to Th9 cells ( Figure S4A, B ). Moreover, Treg skewing conditions did not result in increased Th9 cells ( Figure S4C ). In addition, Th9 skewing conditions induced Foxp3+ Treg cells ( Figure S4D ) although at lower levels as compared to Treg skewing conditions. IL-9 did not further induce Treg development upon IL-2 and TGF-beta stimulation ( Figure S4D ). Finally, when naive T cells were differentiated into Treg cells, IL-9R expression mRNA was upregulated at day 5 as compared to day 3 ( Figure S4E ). These findings suggest that Treg cells produce lower levels of IL-9 as compared to Th9 cells but may have increased capability to respond to IL-9 signaling as they differentiate. We then analyzed levels of IFN-ɣ, an anti-tumor Th1 cytokine released by different immunocompetent cells such as CD8+ T cells that play a key role in anti-tumor mediated immune responses (27, 37, 38). To this aim, we first skewed naïve spleen CD4+CD62L+ T cells under Th1 skewing conditions. As IL-9 production of naive CD4+ T cells is inhibited by IFN-ɣ (9, 39), we induced Th1 development in naïve spleen T cells by incubating naïve cells with IL-12, a cytokine produced by dendritic cells and macrophages that induces IFN-ɣ production in naive T cells (40), and anti-IL-4 antibodies. In these studies, we found that Th1 skewing conditions inhibited IL-9 while inducing IFN-ɣ levels in the supernatants ( Figure S4F ).

In subsequent experiments, we performed studies in the syngenic LL/2 model and analyzed the direct anti-tumoral function of IL-9 on the tumor cell line LL/2 by using Ki67 and AnnexinV/PI as proliferation and apoptosis markers, respectively ( Figures 6A, B ). By using Ki67 staining we found that, IL-9 stimulation did not induce tumor cell proliferation of LL/2 cells ( Figure 6A ). In addition, IL-9 significantly reduced late apoptosis in LL/2 cells at the lower concentration analyzed, as determined by Annexin/PI staining, suggesting that IL-9 may directly regulate apoptosis of LL/2 lung tumor cells in a dose dependent manner ( Figure 6B ). The overall survival of LL/2 was induced by higher concentrations of IL-9 at 100ng/ml ( Figure 6C ).

We next induced lung tumor in a syngenic model of disease in C57Bl/6 mice. As the transcription factor PU1 has been implicated in controlling IL-9 production by T cells (16), we then asked if levels of PU1 (Spi1 mRNA) are upregulated in lung T cells of wild-type mice bearing tumors. In fact, Spi1 mRNA levels were significantly induced in tumor infiltrating T cells in WT mice as compared to lung T cells from mice lacking tumors ( Figure 6D ). In addition, we demonstrated that these lung T cells may respond to IL-9 stimulation because they express the IL-9R ( Figure 6E ). However, we could not find any difference in IL-9R expression between tumor infiltrating T cells and lung T cells from naïve mice.

In this syngenic model ( Figures 6F –J ), we found that, targeted deletion of IL-9, resulted in induced body weight as compared to the wild type littermates bearing tumor ( Figures 6F, G ). Moreover, by assessing the tumor load, we found that both the IL-9 deficient and the IL-9 heterozygous mice had a decreased (but not statistically significant) tumor load as compared to wild type littermates ( Figure 6H ). More importantly, 67% of the IL-9 heterozygous and 50% of the IL-9 knockout mice were tumor free as opposed to 40% in the wild type mice group, indicating an anti-tumoral effect of IL-9 deficient mice ( Figure 6I ). By looking into the mechanism we found that consistent with the presence of Th9 cells in the tumor of NSCLC patients, targeted deletion of IL-9 resulted in significant upregulation of IL-21, a cytokine produced by Th9 cells that may regulate anti-tumor effects of Th9 cells in lung cancer ( Figure 6J ).

We next asked if the presence of IL-9 affects tumor development in a second syngenic model of lung cancer after intravenous injection of L1C2 lung tumor cell line in a Balb/c genetic background ( Figure 7 ). Consistent with the other models of disease, we confirmed significant upregulation of body weight in the absence of IL-9 during tumor development as compared to wild type littermates ( Figures 7A, B ). We observed a reduction of the tumor load, by trend ( Figures 7C–E ). Similar to the LL/2 model, we observed in additional experiments that the number of mice with no tumor was induced in the absence of IL-9 ( Figure 7E ). Moreover, the immunosuppressive FoxP3+CD4+CD25+ T regulatory cells were reduced in the absence of IL-9 ( Figure 7F ), consistent with our finding that Treg cells express IL-9R. Finally, we directly analyzed the pro-tumoral function of IL-9 on the tumor cell line L1C2 by using cell counts ( Figure 7G ), Ki67 ( Figure 7H ) and AnnexinV/PI ( Figure 7I ) as proliferation and apoptosis markers, respectively. By using cell counts, we found that IL-9 did not influence the tumor cell count, while IFN-ɣ and TNF-alpha decreased the overall tumor cell count ( Figure 7G ). By Ki67 staining, we detected that IL-9 stimulation significantly induced L1C2 tumor cell proliferation in a dose dependent manner ( Figure 7H ). In addition, IL-9 significantly reduced early apoptosis in L1C2 cells at lower concentrations, as determined by Annexin/PI staining, suggesting that IL-9 may directly regulate both apoptosis and proliferation of L1C2 lung tumor cells ( Figure 7H ).

The above findings are consistent with a model in which IL-9 derived by TIL cells directly favours tumor cell survival and suppresses anti-tumor immune responses by effects back on TILs. Therefore, we asked if targeting of IL-9 function might be therapeutically useful in lung cancer. To explore this concept under in vivo conditions, we next studied the effects of neutralizing anti-IL-9 antibodies in our experimental lung cancer model induced by LL/2 cells in vivo. To this aim, we blocked IL-9 repeatedly during lung tumor development in WT mice via intraperitoneal injection of anti-IL-9 antibodies ( Figure 8A ). Consistent with the phenotype discovered for IL-9 deficient mice, intraperitoneal anti-IL-9 antibody treatment in WT mice resulted in increased body weight and an increased tumor rejection ( Figures 8B–D ) as compared to control IgG-treated WT animals. Moreover, we found that IgG control treatment in tumor model resulted in decreased number of CD4+ and CD8+ T-bet+TNF-α+ T cells which were induced by the anti-IL-9 antibody treatment ( Figures 8E, F ). Finally, we pretreated LL/2 cells with IL-9 for 48 hours and then challenge LL/2 cells with different cytokines. Shown is flow cytometric analysis of the percentage of necrotic cells (Annexin V-PI+ cells) after IL9 (100 ng/ml) pretreatment of LL/2 cells following 48h unstimulated cell culture condition or in conditions with IFNɣ, IL9, IL21, TNFα treatment, respectively. Under these conditions only TNF-α showed a pro-apoptotic function on LL/2 tumor cells ( Figure 8G ).

In subsequent studies on the effects of anti-IL-9 antibodies in experimental NSCLC, we explored the effects of these antibodies at later time points of tumor induction in the two models of disease in syngenic mice. In the LL/2 model we started the treatment at day 6 after tumor cell injection ( Figure 9A ). Here we observed that, blocking IL-9 at later time points during tumor development resulted in a significant tumor reduction as compared to IgG treatment ( Figures 9B, C ). pSTAT5 is activated downstream of IL-9 and thus we looked at pSTAT5 in lung CD4+CD25+ T cells and found that this cell population was down-regulated by anti-IL-9 antibody treatment ( Figure 9D ). By trend, we also found induction of IFN-ɣ after anti IL-9 treatment ( Figure 9E ).

We next analyzed a second syngenic model of disease using the cell line L1C2 in a BALB/c genetic background and started treatment at a later time point at day 9 when the tumor was already present in the lung ( Figure 9F ). In this model we found a reduction of the number of pre-tumoral lesions in the anti-IL-9 antibody treated mice ( Figure 9G ). Additionally, we studied the number of activated CD4+CD25 (IL2Ralpha)+ T cells and found them to be induced in the lung of anti-IL-9 antibody treated mice ( Figure 9I ). Moreover, the immunosuppressive cytokine IL-10 was downregulated upon anti-IL-9 treatment ( Figure 9J ), suggesting that IL-9 controls the balance between inflammatory and immunosuppressive cytokines in NSCLC.