JIA PBMCs from Vanderbilt Monroe Children’s Hospital and the Cincinnati Children’s Hospital Pediatric Rheumatology Tissue Repository and child healthy control (HC) PBMCs from the Cincinnati Children’s Hospital Medical Center genomic control cohort were obtained with IRB approved protocols. Isolated PBMC were stored in fetal bovine serum with 10% DMSO in liquid nitrogen.

PBMCs were cultured under T1, T2, and T17 polarizing conditions in a well-defined tissue culture model for T cell differentiation and cytokine secretion (14, 30–32). The T17 polarizing conditions include anti-IFNγ. Cultured cells were in RPMI 1640 (Gibco) supplemented with 10% fetal bovine serum (Atlanta Biologicals), L-glutamine (Gibco), and penicillin-streptomycin (Gibco) with 5% air CO₂. PBMCs at 100,000 cells/well in a 96-well plate or 500,000 cells/well in a 24-well plate were stimulated on anti-CD3 (10µg/mL) precoated plates with anti-CD28 (1µg/mL) and stimuli for T cell polarization. Stimuli were T1: IL-12 (10ng/mL), T2: IL-4 (10ng/mL), T17: IL-6 (50ng/mL), IL-1β (20ng/mL), TGF-β1 (2ng/mL), IL-23 (20ng/mL), IL-21 (100ng/mL), anti-IFNγ (10µg/mL). Antibodies were purchased from Biolegend and cytokines were purchased from BD Biosciences and R&D Systems. After 5 days for T1 and T2 cultures or 7 days for T17 cultures, media containing polarization factors was washed away and cells were re-stimulated on new anti-CD3 coated plates with media that did not contain polarization factors. After 2 days, culture fluids and cells were harvested for analysis. Culture fluids were analyzed for cytokines by ELISA to detect human IFNγ (BD Biosciences), TNFα (BD Biosciences), IL-17a (ThermoFisher), IL-5 (BD Biosciences), and IL-13 (ThermoFisher). T cell cultures were performed in triplicate and ELISA results averaged.

Naïve CD4⁺ cells were isolated from PBMCs using the EasySep™ Human CD4+ T Cell Isolation Kit (StemCell) using immunomagnetic negative selection according to the manufacturer’s instructions. Naïve CD4 T cell cultures had 200,000 cells/well in a 96-well plate that were stimulated on anti-CD3 (10µg/mL) precoated plates with anti-CD28 (1µg/mL) and IL-2 (100U/mL, PeproTech). In some samples T1 polarization factors were added.

PBMC were analyzed with LIVE/DEAD™ Fixable Violet Dead stain, anti-CD3 APC-H7, anti-CD8a PE, anti-CD4 perCP-cy5.5, anti-CD197 AF647, and anti-CD45RA FITC with standard flow cytometry protocols. Gating was performed with fluorescent minus one (FMO) control for CD197⁺ and CD45RA⁺ populations. Experiments were performed on a 3-Laser BD LSRFortessa instrument maintained by the Vanderbilt University Medical Center Flow Cytometry Shared Resource.

T1 cell cultures were restimulated on day 5 and then analyzed on day 7 after being treated with PMA (30ng/mL), ionomycin (1µg/mL), and GolgiStop (1µL/mL) for 6 hours. T1 cells were analyzed with Fixable Zombie NIR, anti-CD3 AF700, anti-CD4 BV510, anti-CD8 BV711, anti-IFNγ PEDazzle594, anti-IL-17 APC, anti-Tbet PE-Cy7, and anti-RORγT PE with standard flow cytometry protocols. Gating was performed with FMO control for IFNγ⁺ and IL-17⁺ populations. Experiments were performed on a Cytek Aurora instrument.

Naïve CD4⁺ cells were isolated and primary cells were analyzed or T1 polarized cells were analyzed. Primary cells were placed on anti-CD3 coated plates with added anti-CD28 and IL-2 for 16 hours, then treated with Cell Stimulation Cocktail (plus protein transport inhibitors) for 5 hours, and analyzed by flow cytometry. Naïve CD4⁺ cells were isolated and underwent T1 polarization on anti-CD3 coated plates with added anti-CD28, IL-2, and IL-12 for 5 days, then were treated with Cell Stimulation Cocktail (plus protein transport inhibitors) for 5 hours, and were analyzed by flow cytometry. Primary cells and T1 cells were analyzed with Fixable Zombie NIR, anti-CD3 AF700, anti-CD4 BV510, anti-CD8 BV711, anti-IFNγ PEDazzle594, and anti-IL-17 APC with standard flow cytometry protocols. Experiments were performed on a Cytek Aurora instrument.

Analysis was performed using FlowJo software. Reagents were purchased as antibodies from BD Pharmingen, Biolegend, and Life Technologies and reagents from BD Biosciences and ThermoFisher.

PBMCs were labeled with carboxyfluorescein succinimidyl ester (CFSE)(10µM) using the CellTrace CFSE Cell Proliferation Kit (ThermoFisher, C34554). Cells were then stimulated with immobilized anti-CD3 and anti-CD28 (1µg/mL). On day 3, harvested cells were analyzed by flow cytometry for live/dead cells, anti-CD3 APC-H7, anti-CD8a BV711, and anti-CD4 APC. Purchased reagents are from Life Technologies, BD Pharmingen, BD Biosciences, and Biolegend. Experiments were performed on a 3-Laser BD LSRFortessa instrument. Analysis was performed using FlowJo software Proliferation Modeling.

RNA was collected from T1 and T2 cell cultures polarized for 5 days using TRIreagent and complementary DNA synthesized using SuperScript First Strand Synthesis (ThermoFisher). Quantitative real time PCR measured expressed RNAs using SYBR green master mix (Applied Biosystems). RNA levels were standardized relative to the housekeeping gene GAPDH. Patient samples containing definite outliers (ROUT with Q=0.1% by GraphPad Prism software) were removed. Patient samples with both T1 and T2 available were included for analysis.

For RNA sequencing, RNA was collected by TRIreagent followed by DNase digestion. Library preparation was performed with the Illumina Tru-Seq Stranded RNA kit. RNA sequencing was performed with an Illumina HiSeq2500 instrument by the Vanderbilt Technologies for Advanced Genomics (VANTAGE) core facilities consecutively on all samples. 100bp paired end reads were generated. Average sequencing depth of all samples was 43 million mapped reads ± 13.5 million (standard deviation). FASTQ files were processed using DESeq2 to identify differentially expressed genes (33). Data are available in NCBI’s Gene Expression Omnibus (34) with GEO series accession number GSE185193. GO Enrichment analysis for overrepresented biological processes was performed using the Gene Ontogeny Resource (35–37).

Applied statistical tests were Welch’s t-test for comparison of 2 and Welch Anova for comparison of more than 2 samples. P < 0.05 was considered significant.

We performed studies in prepubescent children with JIA and age-, gender-, and race-matched controls (HC) ( Table 1 ). JIA patients were an average of 62 months old, ANA+, and mostly female, consistent with known clinical characteristics. JIA subtypes were defined using International League Against Rheumatism (ILAR) criteria (38). Clinical characteristics were collected at the time of PBMC collection ( Table 1 ). Overall, JIA patients had 0-1 active joints with arthritis and were on methotrexate and/or biologic medications including etanercept, adalimumab, infliximab, and abatacept. Five JIA PBMCs were collected within 1 month of diagnosis and before initiation of systemic medications other than non-steroidal anti-inflammatories (NSAIDs). This group had a higher number of active joints at time of biosample collection, with an average of 6 active joints.

Differences between JIA and HC peripheral blood mononuclear cells (PBMCs) were investigated by analyzing T cell subsets and naïve and memory phenotypes. JIA and HC PBMCs had the same frequencies of CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ cells ( Figures 1A-C ). On average, 60% were CD3⁺ cells, of which approximately 63% were CD4⁺ and 28% were CD8⁺. Memory and naïve cell phenotypes were identified by expression of CD197 (CCR7) and CD45RA with naïve cells: CD197⁺CD45RA⁺, central memory cells: CD197⁺CD45RA^-, and effector memory cells: CD197^-CD45RA^-. Pediatric (JIA and HC) CD3⁺CD4⁺ cells were an average of 69% naïve and 23% combined central and effector memory cells ( Figures 1D, E ). Four JIAs had naive CD3⁺CD4⁺ cells below the standard deviation and effector memory CD3⁺CD4⁺ cells above the standard deviation. These 4 JIA patients were all polyarticular RF- JIA with no common features shown by age in months (range 23-83), active joints (range 0->9), months since JIA diagnosis (range 0-54), and physician global assessment score (range 0-5). Pediatric CD3⁺CD8⁺ cells were an average of 70% naïve and 12% combined effector and memory cells ( Figures 1F, G ). In pediatric patients, ratios of naïve to memory CD3⁺CD4⁺ and CD3⁺CD8⁺ cells were 3 and 5.8 respectively. Importantly, no differences in memory and naïve cell phenotypes were identified between JIA and HC. Groups were color coded based on current use of methotrexate, biologic, methotrexate and biologic, prednisone, or none or NSAID therapy to assess the role of therapies on cell frequencies. There was no clustering of patient samples based on current therapy ( Figure 1 ).

JIA and HC T cell proliferation was assessed by labeling PBMCs with the fluorescent dye, CFSE. Labeled cells were stimulated with anti-CD3 and anti-CD28 and proliferation measured on day 3 ( Figure S1A ). The proliferation index (PI), which is the number of divisions per dividing cell, for JIA and HC CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ cells were not different ( Figure S1B ). On average, PIs for CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ cells were 1.4, 1.4, and 1.5 ( Figure S1B ). Importantly, no differences in PIs were identified between JIA and HC.

We determined if JIA T cells, which were largely naïve, developed inflammatory cytokine profiles with T cell polarization. We used short-term T cell cultures to model T cell polarization and cytokine production (30, 31). JIA and HC PBMCs underwent T1, T2, and T17 polarization, activation, and were then collected for analysis ( Figure 2A ). In T1 cultures, JIA exhibited increased production of IFNγ ( Figure 2B ). While in T2 cultures of HC IFNγ was completely suppressed, IFNγ was detectable in several JIA T2 cultures. JIA T1 and T17 cultures had increased production of IL-17 ( Figure 2C ). Notably, IL-17 production in JIA T1 cultures was similar or even greater than its production in HC T17 cultures. No differences were detected in TNFα levels ( Figure 2D ). Granulocyte macrophage colony-stimulating factor (GM-CSF) was analyzed in a subset of T1 polarized cultures and no differences were detected. In T2 cultures IL-17 and TNFα production were below the limits of detection. To ensure T2 cultures polarized appropriately, IL-13 and IL-5 were measured and showed that both cytokines were expressed and were not different between JIA and HC ( Figures S2A, B ). To determine if the 5 newly diagnosed JIA, who were receiving no therapy or NSAIDs, had a skewed pattern of inflammatory cytokine production, these patients were identified in our T cell culture analysis ( Figure 2 , red circles). The newly diagnosed JIA were not different than total JIA for T1, T2, and T17 culture production of IFNγ, IL-17, and TNFα. Groups were color coded based on current use of methotrexate, biologic, methotrexate and biologic, prednisone, or none or NSAID therapy to assess the role of therapies on cytokine production. There was no clustering of patient samples based on therapy ( Figures 2B-D ). A separate JIA subject had longitudinal biosamples at two timepoints that were separated by 5 months. Both biosamples were obtained while the patient was on methotrexate. The longitudinal JIA samples had T1 polarized cells with heightened production of IFNγ and IL-17 compared to HC ( Figure S3 ).

To determine if differences in cytokine production were due to T cell subset frequencies, HC and JIA T1, T2, and T17 CD3⁺CD4⁺ and CD3⁺CD8⁺ cells were analyzed by flow cytometry. We found that the subsets are distributed similar between JIA and HC cultures ( Figures S4A, B ).

T1, T2, and T17 polarization in Th and Tc cells involve similar STAT signaling, master transcription factor induction, and signature cytokine expression. We assessed gene expression levels in JIA and HC T1, T2, and T17 cells by quantitative real time qPCR. The cells used for this experiment were generated in parallel with the cytokine production data in Figure 2 . The inflammatory cytokine IFNγ was expressed higher in T1 than T2 and T17 cells ( Figure 3A ). The inflammatory cytokine IL-17 was expressed higher in T1 and T17 cells than T2 cells ( Figure 3A ). IFNγ expression was magnitudes higher than IL-17. Importantly, IL-17 was expressed at significantly higher levels in JIA T1 cells than in HC ( Figure 3A ). IFNγ exhibited a trend of higher expression in JIA T1 cells than in HC. IL-17 exhibited a trend of higher expression in JIA T17 cells than in HC. Master transcription factors Tbet, RORγT, and GATA3 are important in T1, T17, and T2 polarization, respectively. We found Tbet was more highly expressed in T1 cells and GATA3 was more highly expressed in T2 cells ( Figure 3B ). RORγT was more highly expressed in T1 cells and T17 cells. Importantly, both Tbet and RORγT were expressed significantly higher in JIA T1 cells than in HC ( Figure 3B ). In STAT signaling, STAT4 and STAT1 increase Tbet and are important in T1 polarization and STAT3 increases RORγT and is important in T17 polarization. STAT4, STAT1, and STAT3 gene expression levels were not different between JIA and HC in T1 and T2 cells ( Figure S5 ).

Biological pathway differences in JIA and HC T1 cells were determined by RNA sequencing and pathways analysis. RNA from 3 HC and 3 JIA T1 cells were analyzed for differential gene expression using DeSeq2. The 3 JIA patients were selected based on showing production of IFNγ and IL-17 in T1 polarized cultures, RNA availability, and having an age-matched HC with RNA available. The RNA was collected from a T1 polarized culture prepared in parallel with samples from Figure 2 . The demographic and clinical characteristics of the 3 HC and JIA samples were compared ( Table S1 ). Genes with p-value less than 0.05 were identified and found 17 genes with increased and 16 genes with decreased expression in JIA ( Figure 3C ) ( Table S2 ). The most overexpressed JIA T1 protein coding genes were IFNL1 and IL17A. The JIA genes with increased expression were analyzed for enriched biological processes using a PANTHER overrepresentation test. The identified biologic processes included important immune pathways and the IL-17 mediated signaling pathway ( Figure 3D ). The JIA genes with decreased expression were analyzed for enriched biologic processes and found enrichment in the B cell receptor signaling pathway ( Figure 3E ). RNA sequencing data identified pathways associated with IL-17 as highly dysregulated in JIA T1 cells.

We next asked if JIA T1 cell IFNγ and IL-17 were derived from single or dual cytokine producing cells, and if these cells were Th cells (CD3⁺CD4⁺) or Tc cells (CD3⁺CD8⁺). JIA T1 cultures from 8 patients were analyzed by flow cytometry for frequency of cells producing IFNγ, IL-17 or both IFNγ-IL-17. Representative flow cytometry diagrams of JIA T1 cells that are CD3⁺CD4⁺ and CD3⁺CD8⁺ showed cell production of IFNγ and IL-17 ( Figures 4A, B ). IL-17 was produced in CD3⁺CD4⁺ and not in CD3⁺CD8⁺ JIA T1 cells. IL-17 was produced in CD3⁺CD4⁺ cells that made only IL-17 and cells that made both IFNγ-IL-17. IFNγ was produced in both CD3⁺CD4⁺ and CD3⁺CD8⁺ JIA T1 cells. The frequency of IFNγ^-IL-17⁺, IFNγ⁺IL-17⁺, and IFNγ⁺IL-17^- cells were determined in the CD3⁺CD4⁺ and CD3⁺CD8⁺ JIA T1 cell subsets. A significantly higher frequency of IFNγ^-IL-17⁺ cells was present in CD3⁺CD4⁺ cells ( Figure 4C ). A trend of higher frequency of IFNγ⁺IL-17⁺ cells was present in CD3⁺CD4⁺ cells ( Figure 4C ). We next asked if the expression of the master transcription factors Tbet and RORγT were associated with IL-17 production in JIA T1 CD3⁺CD4⁺ cells. A representative flow cytometry diagram of JIA T1 CD3⁺CD4⁺ cells showed expression of IL-17 in cells that are Tbet^high and Tbet^low ( Figure 4D ). The frequency of IL-17 positive cells in JIA T1 CD3⁺CD4⁺Tbet^high cells was higher than in CD3⁺CD4⁺Tbet^low cells ( Figure 4E ). A representative flow cytometry diagram of JIA T1 CD3⁺CD4⁺ cells showed expression of IL-17 in cells that were RORγT^high and RORγT^low ( Figure 4F ). The frequency of IL-17 positive cells in JIA T1 CD3⁺CD4⁺RORγT^high cells was significantly higher than in CD3⁺CD4⁺RORγT^low cells ( Figure 4G ).

We then compared whether the production of cytokines in T1 polarized cultures by ELISA correlates with production of cytokines in T1 polarized cells measured in CD3⁺CD4⁺ cells by flow cytometry. We found that JIA T1 polarized culture measures were significantly correlated for IFNγ ( Figure S6A ) and trended towards significance for IL-17 ( Figure S6B ). We then analyzed RNA from T1 polarized cells generated during the flow cytometry experiment, finding that the JIA T1 polarized cells expressed high levels of the IFNγ, IL-17, Tbet, and RORγT genes ( Figure S7 ).

We asked if isolated naïve CD4⁺ JIA cells generated more IFNγ and IL-17 from primary cells and T1 polarized cells than naïve CD4⁺ HC cells. Naïve CD4⁺ cells from 6 JIA patients and 5 HCs (3 adult and 2 pediatric) were analyzed by flow cytometry for frequency of CD3⁺CD4⁺ cells producing IFNγ and IL-17. A representative flow cytometry diagram of JIA primary CD3⁺CD4⁺ cells showed low expression of IFNγ ( Figure 5A ). The frequency of IFNγ positive cells in JIA and HC primary CD3⁺CD4⁺ cells was not different ( Figure 5B ). A representative flow cytometry diagram of JIA primary CD3⁺CD4⁺ cells showed low expression of IL-17 ( Figure 5C ). The frequency of IL-17 positive cells in JIA and HC primary CD3⁺CD4⁺ cells was not different ( Figure 5D ). A representative flow cytometry diagram of JIA T1 CD3⁺CD4⁺ cells showed high expression of IFNγ ( Figure 5E ). The frequency of IFNγ positive cells in JIA T1 CD3⁺CD4⁺ cells was significantly higher than HC CD3⁺CD4⁺ cells ( Figure 5F ). A representative flow cytometry diagram of JIA T1 CD3⁺CD4⁺ cells showed higher expression of IL-17 ( Figure 5G ). The frequency of IL-17 positive cells in JIA T1 CD3⁺CD4⁺ cells was significantly higher than HC CD3⁺CD4⁺ cells ( Figure 5H ). Only JIA T1 CD3⁺CD4⁺ cells exhibited an increase in IL-17 frequency among all isolated naïve CD4⁺ cells studied. Importantly, the frequencies of IFNγ and IL-17 positive JIA CD4⁺ cells generated from PBMCs ( Figure 4C ) and naïve CD4⁺ cells ( Figures 5F-H ) by T1 polarization was similar.