Using genetics, genomics, and transcriptomics to identify therapeutic targets in juvenile idiopathic arthritis

TL;DR

This study uses genetic and genomic data to find new treatment targets for juvenile idiopathic arthritis, focusing on a gene called HIPK1.

Contribution

The study identifies HIPK1 as a novel therapeutic target for JIA, linking it to disease risk and T cell regulation.

Findings

HIPK1 is a strong candidate target gene associated with JIA and is suppressed in CD4+ T cells of affected children.

HIPK1 is located in the PTPN22 locus and is connected to an enhancer element that influences its expression.

Downstream transcription factors of HIPK1 show enriched binding near promoters of JIA-related genes.

Abstract

Despite progress in improving outcomes for oligoarticular and polyarticular juvenile idiopathic arthritis (JIA), the field still faces considerable challenges. More than half of adults who have had JIA continue to have active disease and have developed functional limitations. Medication side effects are common and intrusive. Thus, the field continues to search for therapeutic agents that target specific aspects of disease pathobiology and will be accompanied by fewer and less intrusive side effects. We identified 28 candidate target genes that were associated with JIA according to Open Targets Genetics and were also differentially expressed in the CD4+ T cells of children with active JIA (when compared to healthy control subjects). Of the 28 candidates, the strongest new target to emerge was homeodomain-interacting protein kinase 1 (HIPK1), which suppresses T cell activation and is within the PTPN22 locus tagged by rs6679677. This locus includes an enhancer element that contacts the HIPK1 promoter, and HIPK1 shows decreased expression in JIA CD4+ T cells when compared to controls. Gene Ontology terms associated with HIPK1 were overrepresented among the differentially expressed genes between JIA and controls, and PML, a known coregulator of HIPK1, showed a similar suppressed gene expression profile. Two downstream transcription factors of HIPK1, TP53 and GATA4, showed enriched binding patterns near the promoters of JIA up-regulated genes. Taken together, these data suggest a pathogenic role for HIPK1 in JIA and make it a prime candidate for therapeutic modulation.

This study integrates genetic, genomic, and transcriptomic data to identify therapeutic targets in juvenile idiopathic arthritis (JIA). It highlights HIPK1 as a novel target, demonstrating its role in T cell regulation and its association with JIA risk variants. These findings offer insights into JIA pathobiology and potential drug development pathways.

Figures

Workflow to identify candidate drug targets for JIAFlowchart showing the steps to identify a candidate drug target for JIA, along with the number of genes identified at each step.Table 1JIA-associated gene targets with differential expressionJIA-associated SNP rsIDPutative gene targetLog2(fold change)FDR valueKnown clinical-stage drugsrs12430303AKAP110.3290.026nors6434390ASNSD1−0.4100.021nors4616297ATXN2L0.4310.001nors4616297BCL7C−0.4880.017nors2481065CKS1B−0.8510.010nors2481065DENND4B0.3450.014nors2481065DPM3−0.8450.015nors4766578ERP29−0.5550.001nors4616297FBRS0.4060.005nors4869314GLRX−0.9150

Genome browser snapshot of the rs6679677 JIA-associated SNPGenome browser snapshot shows data generated in T cells of patients with JIA (active disease under treatment [ADT] and clinical remission on medication [CRM]) and matched healthy control (HC) subjects. Pooled ATAC-seq, RNA-seq, CTCF ChIP-seq, and CTCF HiChIP data are shown, as generated in Tarbell et al.[10](#bib10) Genes in red indicate genes identified by Open Targets Genetics as target genes of the rs6679677 SNP.Figure 3HIPK1 and its gene network partners show disease-associated expression patterns(A) Violin plots of HIPK1 normalize

Regulatory potential analyses(A) Regulatory potential of TP53 ChIP-seq binding sites identified in A549 cells for differentially expressed genes in T cells of patients with JIA.(B) Regulatory potential of TP53 ChIP-seq binding sites identified in HepG2 cells for differentially expressed genes in T cells of patients with JIA.(C) Regulatory potential of GATA4 ChIP-seq binding sites identified in HepG2 cells for differentially expressed genes in T cells of patients with JIA.BETA generated the cumulative distribution function of the gene groups and the results of a one-tailed Kolmogorov-Smirnov te

Model of HIPK1’s role in the pathogenesis of JIAIn healthy CD4+ T cells, under activated conditions, *HIPK1* acts as a negative feedback, acting to suppress GATA4 and TP53 binding and reduce their gene expression programs. In JIA, genetic variants in the vicinity of the rs6679677 index SNP reduce *HIPK1* expression, lowering the effectiveness of its regulatory role on GATA4 and TP53 and prolonging a pro-inflammatory gene expression program that results in a hyper- and/or auto-inflammatory phenotype.

Full text

Introduction

Juvenile idiopathic arthritis (JIA) is a chronic condition characterized by chronic inflammation of the synovium and one of the most common chronic conditions affecting children in the USA.1^,^2 Considerable progress has been made in treating children with JIA over the past 20 years, particularly with the introduction of biological agents.3 However, the field still faces important challenges. Despite progress in improving functional outcomes, these outcomes have come at a cost, as most children experience significant and intrusive medication side effects from the initiation of treatment.4 Furthermore, we face additional serious challenges: (1) we have had limited success in getting children off medication,5 meaning that children with JIA have prolonged exposure to potent immunosuppressive medications, and (2) half of all adults diagnosed with JIA continue to have active disease and significant functional limitations, which results in exposure to these medications, and their toxicities/side effects, that continue into adulthood.6

In the past 5 years, investigators have assessed the potential of using genetic, genomic, and transcriptomic data to guide the development of new therapies or to repurpose known agents.7^,^8 The underlying premise of these investigations is that these datasets, used together, are likely to identify features of pathobiology that are not discernable from any single measurement/assessment. Furthermore, by elucidating genetic risk and the immunologic pathways influenced by risk-inducing variants, there is reason to believe that personalized approaches to therapy can be provided to individual patients.

While genetic risk is likely to be exerted on multiple cell types,9 it has long been known that CD4+ T cells are critical drivers of the immunopathological events in JIA.10^,^11 Distinct abnormalities in peripheral blood T cells have been described,12 and there is ongoing interest in these cells as therapeutic targets.13

In this paper, we demonstrate the feasibility of using the known JIA genetic risk loci, three-dimensional (3D) chromatin data, and RNA sequencing (RNA-seq) to identify candidate therapeutic targets. We identified twenty-eight candidate targets, one of which, homeodomain-interacting protein kinase 1 (HIPK1), is novel and represents a potentially druggable pathway.

Material and methods

Patients and control subjects

Patient data (RNA-seq, chromatin immunoprecipitation [ChIP]-seq, HiChIP, and ATAC-seq) used for these studies were previously reported in an earlier paper from our group.10 We repeat some details on the patient characteristics here. Patients fit criteria for polyarticular-onset, rheumatoid-factor-negative JIA as established by the International League Against Rheumatism (ILAR).14 We used the Wallace criteria15^,^16^,^17 to characterize children as having active disease on therapy (ADT) or clinical remission on medication (CRM). Patients with JIA were all being treated with combinations of methotrexate and the tumor necrosis factor (TNF) inhibitor etanercept. Samples were obtained from children classified as ADT and were obtained 6 weeks from the onset of therapy. Samples obtained from children with CRM were obtained when CRM status was confirmed. This was typically 12–15 months after the initial diagnosis.

Healthy control (HC) children were recruited from the Hodge General Pediatrics Clinic of the University at Buffalo Jacobs School of Medicine and Biomedical Sciences. Exclusion criteria included a fever ≥ 38°C within the previous 48 h, the presence of another autoimmune disease (e.g., type 1 diabetes), and treatment with systemic glucocorticoids or antibiotics.

Laboratory data

Data from RNA-seq and ChIP-seq/HiChIP for CTCF in purified CD4+ T cells are available on the Gene Expression Omnibus under accession number GEO: [GSE164215](GSE164215), uploaded in January 2021 and updated in April 2021. These data were used for the analyses described below.

The STRINGDB database18 for protein-protein interactions was utilized to identify the interacting partners to HIPK1. Each member of the interaction network was used as a search term for FactorBook,19 a repository of publicly available transcription factor binding data. Two of the interacting proteins were identified as transcription factors, TP53 and GATA4. FactorBook listed two TP53 ChIP-seq datasets (accessions ENCFF849LSO in HepG2 cells and ENCFF699UTZ in the A549 cell line) and one for GATA4 (accession ENCFF154VTV in HepG2 cells), which were downloaded.

Processing RNA-seq data

All RNA-seq data analyses were performed on the Galaxy bioinformatics online platform.20 Raw RNA-seq FASTQ files were downloaded from the SRA archive21 using the fastq-dump tool.22 The paired-end FASTQ files were aligned to the hg38 genome using RNA-STAR,23 using the GENCODE24 v.44 GTF file as the gene model for splice junctions and with the “GeneCounts” option set to create a table of read counts per gene. All other parameter settings were left on their Galaxy defaults. These count table files were used as the input to the edgeR differential expression analysis tool,25 which was run without low-count genes, defined as having a cpm value below 1.0 in less than 10 samples, and with all other settings at their default values, including the false discovery rate (FDR) threshold for defining differentially expressed genes (DEGs), which was set to 0.05. DEGs were used for GO term analysis using GOrilla.26 GOrilla was run using the two gene set option, with the up-regulated or down-regulated genes as the input and the total list of expressed genes as the background set.

Determining regulatory potential of transcription factors for DEGs

BETA27 was used to calculate the regulatory potential of the HIPK1-associated transcription factors for the DEGs in JIA T cells.

The regulatory potential, which is a gene’s likelihood of being regulated by a factor, is estimated for each gene. The regulatory potential is calculated as

$$[eqn]$$

All binding sites (k) near the transcription start site (TSS) of the gene (g) within a user-specified range (100 kb as default) are considered. Δ is the exact distance between a binding site and the TSS proportional to 100 kb (Δ = 0.1 means the exact distance = 10 kb). BETA then generates a cumulative distribution function of the gene groups and uses a one-tailed Kolmogorov-Smirnov test to determine whether the up-regulated and down-regulated groups differ significantly from the non-differential group.

Results

The following approach was employed in order to identify the pathogenic genes that are controlled by non-coding JIA-associated single-nucleotide polymorphisms (SNPs) and then determine which of those represented viable putative drug targets (Figure 1). First, the Open Targets Genetics (OTG) platform28 was used to link potential pathogenic genes to JIA-associated genetic variants. We find that there were a total of 267 predictions made by OTG, of which 30 were targets of known clinical-stage drugs (Table S1). Those genes whose link to JIA was predicted by OTG were then cross-referenced with genes that were determined to be differentially expressed in CD4+ T cells between patients with active JIA and HC subjects, originally published by Tarbell et al.10 A total of 29 genes met these criteria (Table 1). With the focus on novel drug targets, genes were then excluded if there were known clinical-stage drugs that targeted them. Four of the 29 genes initially identified met these criteria: ALDH2, which has been targeted for alcohol disorders and tested for some cancers (Table S2); CHRNB2, a target of nicotine; ITGAL (CD11a) was targeted for psoriasis, arthritis, and diabetes (Table S1); and ITGAM (CD11b), which is frequently, although unsuccessfully, targeted for cancers.29 We find that although our approach (OTG plus DEGs) has a modestly higher incidence of known drug targets (14.3%) as opposed to OTG only (11.2%), the difference is not statistically significant (Fisher’s exact test p = 0.5).Figure 1. Workflow to identify candidate drug targets for JIAFlowchart showing the steps to identify a candidate drug target for JIA, along with the number of genes identified at each step.Table 1JIA-associated gene targets with differential expressionJIA-associated SNP rsIDPutative gene targetLog2(fold change)FDR valueKnown clinical-stage drugsrs12430303AKAP110.3290.026nors6434390ASNSD1−0.4100.021nors4616297ATXN2L0.4310.001nors4616297BCL7C−0.4880.017nors2481065CKS1B−0.8510.010nors2481065DENND4B0.3450.014nors2481065DPM3−0.8450.015nors4766578ERP29−0.5550.001nors4616297FBRS0.4060.005nors4869314GLRX−0.9150.001nors2481065HAX1−0.6380.001nors4766578HECTD40.3540.020nors6679677HIPK10.3510.016nors4616297ITGAX1.4190.001nors2481065JTB−0.4810.005nors9960807LDLRAD40.4280.011nors79815064NBEAL20.8390.001nors6679677PPM1J1.8300.002nors9960807PSMG2−0.4960.010nors9960807PTPN2−0.3860.024nors4616297SETD1A0.3470.031nors4766578SH2B30.6470.001nors4616297SRCAP0.4630.001nors4616297TUFM−0.3190.036nors4766578VPS29−0.5770.001nors4766578ALDH21.0020.001yesrs2481065CHRNB2−1.6370.031yesrs4616297ITGAL0.4280.001yesrs4616297ITGAM1.0900.001yesList of genes that are associated with JIA, as determined by Open Targets Genetics, and differentially expressed in the CD4+ T cells of patients with JIA versus healthy control subjects. Listed are the fold change and corrected p value of the expression differences and whether there are any clinical-stage or approved drugs that target the gene.

The remaining 25 identified genes were then assessed for drug tractability using the Open Targets Platform, which includes tractability data that identify key details for small molecules, antibodies, and other modalities. For small molecules, assessments include whether the target has been co-crystallized with a small molecule, whether there is a high-quality ligand, whether there is a high-quality binding pocket, whether there is a med-quality pocket, or if the target is part of a druggable family, as per Finan et al.’s Druggable Genome pipeline.30 For antibodies, the assessments include whether there is high confidence that the subcellular location of the target is either plasma membrane, extracellular region/matrix, or secretion or medium confidence that the subcellular location of the target is either plasma membrane, extracellular region/matrix, or secretion; whether the target has a predicted signal peptide or transmembrane regions and not destined to organelles; or whether there is high confidence that the target is located in the plasma membrane. Overall, 6 of the 25 candidate genes/gene products met at least one of these criteria for either small-molecule or antibody tractability (Table 2). Of these top 6 candidates, the most promising one was HIPK1, a putative effector gene of the JIA-associated SNP rs6679677. The OTG platform was used to identify 10 putative effector genes of the rs6679677 SNP (Table 3). The highest-scoring target gene from OTG was PTPN22, followed by the nearest gene in linear genomic space, PHTF1, and HIPK1 rounding out the top 5. Of note is that HIPK1, although it had the fifth overall score assessed by OTG, had the third-highest rank in expression quantitative trait locus (eQTL) colocalization scores and chromatin interaction scores, components of the OTG model that use experimental data. To further prioritize the putative targets from OTG, publicly available transcriptomic and genomic data in the CD4+ T cells of patients with JIA and age-matched HC subjects, originally published by Tarbell et al.,10 were downloaded from the SRA data repository and processed (see material and methods). Transcriptomic data, in the form of RNA-seq, were pooled and averaged across the subjects in each disease group, as were genomic data in the form of ATAC-seq, CTCF ChIP-seq, and HiChIP, and graphically displayed in a genome browser, centered on the rs6679677 SNP (Figure 2). As shown in the genome browser screenshot, there are several regulatory regions in the vicinity of the SNP, denoted by spikes in the ATAC-seq signals. Those regulatory elements near the rs6679677 SNP are in close physical proximity to numerous distal regulatory elements, as shown with the HiChIP loops. Finally, many genes in the vicinity of the rs6679677 SNP and linked to it via chromatin loops show robust expression, as shown in RNA-seq signal tracks. Differential expression analysis between the patients and healthy subjects revealed three genes that both met the expression criteria and were physically linked to the rs6679677 SNP, only one of which, HIPK1, was also linked to the SNP via OTG. The normalized gene expression values between disease states for HIPK1, showing lower expression in individuals with active disease compared to HC subjects, are shown in Figure 3A. As shown in Tables 1 and 2, HIPK1 had a 1.28-fold decrease in T cell expression in individuals with active disease relative to HC subjects, with an adjusted p value of 0.016. To ensure that the expression differences we observed were disease associated rather than the result of treatment administration, we performed differential gene expression analysis on CD4+ T cells from patients in clinical remission with ongoing treatment versus individuals with active disease and HC subjects and found that HIPK1 had a 1.25-fold decrease in T cell expression in active disease relative to clinical remission, with an adjusted p value of 0.035. There was no statistical difference in HIPK1 expression between HC subjects and patients in clinical remission (edgeR DEG analysis, fold change = 1.03, adjusted p value = 1.0). Finally, we find that HIPK1 is predicted to be haploinsufficient and considered to be very intolerant to loss-of-function mutations, according to gnomAD,31 further demonstrating the deleterious effect of HIPK1 reduction. Overall, these data suggest that the suppression of HIPK1 in CD4+ T cells may be mediated by the rs6679677 SNP, which alters the regulatory activity of distal elements that influence HIPK1 expression via physical proximity.Table 2JIA-associated genes with strong drug tractabilityJIA-associated SNP rsIDPutative gene targetLog2(fold change)FDR valueSubcellular localizationrs12430303AKAP110.3290.026cytosol, nucleoli, plasma membraners2481065CKS1B−0.8510.010cytosol, mitochondriars4869314GLRX−0.9150.001cytosol, plasma membraners6679677HIPK10.3510.016cytosol, nucleoplasmrs9960807PTPN2−0.3860.024nucleoplasmrs4766578VPS29−0.5770.001cytosol, vesiclesList of genes that are associated with JIA, as determined by Open Targets Genetics, are differentially expressed in the CD4+ T cells of patients with JIA versus healthy control subjects, and had strong drug tractability for either small molecules or antibodies. Listed are the fold change and corrected p value of the expression differences and their subcellular location.Table 3. Predicted gene targets of the rs6679677 JIA-associated SNPGeneOverall L2G scorePartial L2G scoresVariant pathogenicityDistanceQTL colocChromatin interactionDistance to locusPTPN220.3520.7180.1980.2870.219110,567PHTF10.2360.0490.2660.0320.0341,697RSBN10.0920.0490.1940.0350.03851,290DCLRE1B0.0350.0640.0300.2770.155143,433HIPK10.0300.0640.0290.2430.155168,138BCL2L150.0270.0500.0350.0350.037126,395AP4B10.0180.0600.0200.0380.044144,015OLFML30.0170.0610.0180.1300.142218,205MAGI30.0120.0490.0130.0160.028370,671SYT60.0100.0490.0090.0320.034392,694Putative target genes of the rs6679677 SNP, as defined by Open Target Genetics. Listed is the overall lead to gene score from Open Targets for each gene associated with the SNP, along with the partial L2G scores from the Open Targets model.Figure 2. Genome browser snapshot of the rs6679677 JIA-associated SNPGenome browser snapshot shows data generated in T cells of patients with JIA (active disease under treatment [ADT] and clinical remission on medication [CRM]) and matched healthy control (HC) subjects. Pooled ATAC-seq, RNA-seq, CTCF ChIP-seq, and CTCF HiChIP data are shown, as generated in Tarbell et al.10 Genes in red indicate genes identified by Open Targets Genetics as target genes of the rs6679677 SNP.Figure 3HIPK1 and its gene network partners show disease-associated expression patterns(A) Violin plots of HIPK1 normalized gene expression, measured as log2(counts per million), for healthy control subjects and patients with active disease and under treatment. Black dots represent individual-level expression.(B) Protein binding partners of HIPK1 as determined by StringDB. Green lines represent associations gleaned by text mining, purple lines represent results of direct binding experiments, blue lines represent associations curated from other databases, and brown lines represent associations determined through coexpression.(C) Violin plots of PML normalized gene expression, measured as log2(counts per million), for healthy control subjects and patients with active disease and under treatment. Black dots represent individual-level expression.

To elucidate the downstream events of HIPK1 suppression in CD4 T cells and their role in the pathogenesis of JIA, additional analysis was performed. GO term analysis of the DEGs between active disease and healthy controls was conducted using a statistical overrepresentation test to identify enriched terms compared to a random expectation. From this analysis, we ranked and selected the top 15 most enriched GO terms, rather than all terms below a predefined threshold, to focus on the biological processes most relevant to the study. Among these top terms, several were directly associated with HIPK1, underscoring its involvement in key pathways implicated in JIA pathogenesis. By highlighting these specific terms, we aimed to illustrate HIPK1’s role in critical biological functions and emphasize its potential as a therapeutic target in JIA (Table 4).Table 4HIPK1 is involved in numerous critical processesGO termDescriptionFDR value0030155aregulation of cell adhesion1.63E−030090287regulation of cellular response to growth factor stimulus8.76E−040006928movement of cell or subcellular component2.24E−030090092regulation of transmembrane receptor protein serine/threonine kinase signaling pathway5.73E−031903053regulation of extracellular matrix organization5.58E−030050794aregulation of cellular process5.93E−030090288negative regulation of cellular response to growth factor stimulus6.16E−030043087regulation of GTPase activity7.04E−030023051aregulation of signaling1.06E−020090101negative regulation of transmembrane receptor protein serine/threonine kinase signaling pathway1.03E−020016477cell migration1.26E−020010646aregulation of cell communication1.20E−020051056regulation of small GTPase-mediated signal transduction1.17E−020007010cytoskeleton organization1.32E−021903391regulation of adherens junction organization1.26E−02Top 15 GO terms enriched among the JIA down-regulated genes.aTerms that are associated with HIPK1.

Understanding the role of different genes in the gene regulatory network of HIPK1 may shed light on the role HIPK1 plays in JIA. To that end, the gene regulatory network for HIPK1 was determined using StringDB, a repository of protein-binding data (Figure 3B). The PML nuclear body scaffold protein, a binding partner of and coregulator with HIPK1, was found to be differentially expressed between individuals with active disease and HC subjects in the same direction, but to a lesser degree, as HIPK1 (Figure 3C). The remaining binding partners of HIPK1 in the gene regulatory network were cross-referenced to the FactorBook transcription factor repository to determine which of the proteins acted as downstream transcription factors for HIPK1 signaling. Two factors were identified: p53 and GATA4. Transcription factor binding data, determined with ChIP-seq, were downloaded from FactorBook for both factors and run through BETA using the differential gene expression dataset. As is shown in Figures 4A–4C, both factors show a strong statistical enrichment of binding sites nearby up-regulated genes, as opposed to down-regulated or non-DEGs.Figure 4. Regulatory potential analyses(A) Regulatory potential of TP53 ChIP-seq binding sites identified in A549 cells for differentially expressed genes in T cells of patients with JIA.(B) Regulatory potential of TP53 ChIP-seq binding sites identified in HepG2 cells for differentially expressed genes in T cells of patients with JIA.(C) Regulatory potential of GATA4 ChIP-seq binding sites identified in HepG2 cells for differentially expressed genes in T cells of patients with JIA.BETA generated the cumulative distribution function of the gene groups and the results of a one-tailed Kolmogorov-Smirnov test to determine whether the up-regulated and down-regulated groups differ significantly from the non-differential group. The dotted line represents the background, the genes that are not differentially expressed, whereas the red and the blue lines represent the genes up-regulated and down-regulated, respectively. The y axis represents the proportion of genes in a category that are ranked at or better than the x axis value, which represents the rank on the basis of the regulatory potential score from high to low. The p value listed in the top left represents the significance of the UP or DOWN group relative to the NON group as determined by the Kolmogorov-Smirnov test.

Based on the findings presented here, we propose the following mechanism. In healthy CD4+ T cells, under activated conditions, HIPK1 plays a negative feedback role, acting to suppress GATA4 and TP53 and reduce their gene expression programs. In JIA, genetic variants in the vicinity of the rs6679677 index SNP result in the suppression of HIPK1 expression, reducing the effectiveness of its regulatory role on GATA4 and TP53, prolonging a pro-inflammatory gene expression program that results in a hyper- and/or auto-inflammatory phenotype (Figure 5). This mechanism represents an ideal target for therapeutic intervention. Indeed, TP53 is a target for several drug modalities, such as small molecules and oligonucleotides, for the treatment of different oncological indications, demonstrating the pathway’s potential.Figure 5. Model of HIPK1’s role in the pathogenesis of JIAIn healthy CD4+ T cells, under activated conditions, HIPK1 acts as a negative feedback, acting to suppress GATA4 and TP53 binding and reduce their gene expression programs. In JIA, genetic variants in the vicinity of the rs6679677 index SNP reduce HIPK1 expression, lowering the effectiveness of its regulatory role on GATA4 and TP53 and prolonging a pro-inflammatory gene expression program that results in a hyper- and/or auto-inflammatory phenotype.

Discussion

While the field of pediatric rheumatology has made significant advances in treating JIA in the past 20 years, significant challenges remain. For example, in a recent review using both Cincinnati Children’s Hospital Medical Center and Children’s Arthritis and Rheumatology Research Alliance data (from a total of 8,978 patients), Brunner et al.32 found that more than half of the patients were treated with one or more biological agents and that >10% of the patients continued to have active disease despite having been treated with ≥2 biological agents. Indeed, among those patients treated with ≥2 biological agents, more than half continued to experience active disease. Furthermore, the emergence of validated tools to capture patient-centered assessments of treatment outcomes33 is likely to reveal additional challenges not immediately obvious to physicians but unacceptable from a parent’s/patient’s point of view.

A major impediment to developing improved, more precisely targeted therapies in JIA is our poor understanding of either the pathobiology of the disease or the biology of therapeutic response. With respect to the latter, we know that clinically derived descriptions of therapeutic response (active disease, inactive disease, and CRM) as established in the Wallace criteria15 have objective correspondents in peripheral blood gene expression profiles.16 Therapeutic response is associated with complex re-organization of peripheral blood gene expression networks,17^,^34 but the most important elements of re-organization required to drive remission remain unknown. It is now recognized that remission is not typically associated with the normalization of peripheral blood gene expression profiles.35 However, while transcriptome studies have provided some useful information about the complex biology of treatment response, they have, by themselves, not led to the identification of novel or unrecognized therapeutic targets.

In the current study, we used a combined genetics-genomics-transcriptomics approach to identify new therapeutic target candidates. The validity of our approach is suggested by the fact that we agnostically identified known therapeutic targets, albeit not at a statistically enriched level relative to OTG alone, such as cell surface integrins CD11a and CD11b. Indeed, the anti-CD11a monoclonal antibody (mAb) efalizumab was used to treat autoimmune diseases, including psoriasis and psoriatic arthritis, and was tested in rheumatoid arthritis (Table S2) before being withdrawn for safety issues. More recently, Schittenhelm et al. showed that altered expression patterns of CD11a and CD11b in dendritic cells were associated with inflammation regulation in adult rheumatoid arthritis.36

In the current study, drug tractability was an important criterion for the selection process, and HIPK1 is notable for being a member of a druggable family. However, as a Ser/Thr kinase family member, the modality being referenced is a kinase inhibitor, which, if targeted against HIPK1 in the context of JIA, would be predicted to exasperate the problem. As such, HIPK1 itself may not represent a suitable drug candidate for JIA but rather an anchor point around which a search for a more suitable upstream regulator or downstream effector can take place.

The findings presented here regarding HIPK1’s role in the pathogenesis of JIA have interesting therapeutic implications for other indications. Bei et al. recently showed that the inhibition of HIPK1 in cardiomyocytes was a novel therapeutic approach to treating hypertrophic cardiomyopathy (HTC).37 Based on the data presented here, however, such a treatment would run the risk of inducing similar phenotypic consequences to JIA. Moreover, given that the rs6679677 SNP is associated with other indications beyond JIA, such as type 1 diabetes,38 rheumatoid arthritis,39 and systemic lupus erythematosus,40 it is possible that HIPK1 inhibition in the context of treating HTC may result in immune-modulated adverse reactions beyond synovial joints in other tissues, such as the pancreas. To avoid such adverse events, it may prove necessary to deploy the HIPK1 inhibitors using an active targeting approach by attaching it to a mAb to form an antibody-drug conjugate (ADC) or to enclose it in a liposome and attach mAbs to the liposome surface to create an antibody-conjugated liposome (ACL).

There are several limitations to this study. The first is that the wet lab and procedures here were performed exclusively with CD4+ T cells. Although these cells are known to be an important feature of the immunopathology of JIA,10 there is compelling evidence that both neutrophils41^,^42 and monocytes43^,^44 are involved in both genetic risk and disease biology. Thus, there are small but significant differences in 3D chromatin architecture across cells and cell lines that may highlight additional target candidates. A second limitation is the fact that our method, like that of Fang et al.,7 uses the genome-wide association study (GWAS)-identified SNPs as a starting point for surveying genetic architecture. Because of the phenomenon of linkage disequilibrium, the SNPs that tag risk loci are not necessarily those that drive disease risk. It is probable that knowing the actual risk-driving variants and using higher-resolution maps of 3D chromatin (e.g., using MicroC)45^,^46 will allow the identification of additional targets and more robust precision medicine approaches to treating JIA.

In conclusion, we demonstrate the use of genetic, genomic, and transcriptomic data in CD4+ T cells to identify new therapeutic targets in JIA. Our approach, which identified genes within or downstream of the GATA4-p53 pathways as potential targets, also provides new insights into the pathobiology of JIA. The role of genetic variants in attenuating HIPK1-mediated inhibition of CD4+ T cell activation deserves further investigation.

Data and code availability

Data from RNA-seq, ATAC-seq, and ChIP-seq/HiChIP for CTCF in purified CD4+ T cells is available on the Gene Expression Omnibus under accession number GEO: [GSE164215](GSE164215). FactorBook ChIP-seq datasets are available under accessions ENCFF849LSO, ENCFF699UTZ, and ENCFF154VTV. Code is available upon request.

Acknowledgments

This work was supported by R01 AR078785 from NIH/NIAMS (J.N.J.) and an Innovative Research Grant from the 10.13039/100006260Rheumatology Research Foundation.

Declaration of interests

The authors declare no competing interests.