Skip to main content
NCBI home page
Genetics logoLink to Genetics
. 2022 Sep 13;222(4):iyac141. doi: 10.1093/genetics/iyac141

3D chromatin structure in chondrocytes identifies putative osteoarthritis risk genes

Eliza Thulson 1,#, Eric S Davis 2,#, Susan D’Costa 3,#, Philip R Coryell 4, Nicole E Kramer 5, Karen L Mohlke 6, Richard F Loeser 7,8, Brian O Diekman 9,10,11,✉,4, Douglas H Phanstiel 12,13,14,15,16,✉,4
Editor: N Risch
PMCID: PMC9713432  PMID: 36099032

Abstract

Genome-wide association studies have identified over 100 loci associated with osteoarthritis risk, but the majority of osteoarthritis risk variants are noncoding, making it difficult to identify the impacted genes for further study and therapeutic development. To address this need, we used a multiomic approach and genome editing to identify and functionally characterize potential osteoarthritis risk genes. Computational analysis of genome-wide association studies and ChIP-seq data revealed that chondrocyte regulatory loci are enriched for osteoarthritis risk variants. We constructed a chondrocyte-specific regulatory network by mapping 3D chromatin structure and active enhancers in human chondrocytes. We then intersected these data with our previously collected RNA-seq dataset of chondrocytes responding to fibronectin fragment, a known osteoarthritis trigger. Integration of the 3 genomic datasets with recently reported osteoarthritis genome-wide association study variants revealed a refined set of putative causal osteoarthritis variants and their potential target genes. One of the putative target genes identified was SOCS2, which was connected to a putative causal variant by a 170-kb loop and is differentially regulated in response to fibronectin fragment. CRISPR-Cas9-mediated deletion of SOCS2 in primary human chondrocytes from 3 independent donors led to heightened expression of inflammatory markers after fibronectin fragment treatment. These data suggest that SOCS2 plays a role in resolving inflammation in response to cartilage matrix damage and provides a possible mechanistic explanation for its influence on osteoarthritis risk. In total, we identified 56 unique putative osteoarthritis risk genes for further research and potential therapeutic development.

Keywords: 3D chromatin structure, genome-wide association studies, osteoarthritis, genomics

Introduction

Osteoarthritis (OA) affects over 300 million people worldwide, yet treatment options are limited in large part because the mechanisms driving OA are not fully understood (Hunter and Bierma-Zeinstra 2019; Boer et al. 2021). Genome-wide association studies (GWAS) have identified over 100 loci associated with OA risk (Reynard and Barter 2020), but translating these broad loci into therapeutic targets has been challenging for several reasons. First, the effects of disease-associated variants are likely cell type and context specific (Umans et al. 2021); therefore, studying these variants in the correct system that mimics the OA phenotype is required. Second, linkage disequilibrium (LD) between nearby variants makes it difficult to identify the causal variant(s) at each locus. Finally, because the majority of OA risk variants occupy noncoding regions of the human genome and can regulate genes up to a million base pairs away, the genes impacted by most OA risk variants are unknown.

Several studies have successfully used genomic and bioinformatic techniques to identify the genes impacted by gene-distant noncoding GWAS variants for a variety of disease phenotypes (Claussnitzer et al. 2015; Won et al. 2016; Chesi et al. 2019; Laarman et al. 2019; Duan et al. 2021). Mapping regulatory loci using ChIP-seq, ATAC-seq, or CUT&RUN and intersecting the resulting data with disease-associated variants can identify a short list of putative causal variants. These variants can then be linked to potential target genes by quantifying 3D chromatin contacts using Hi-C or other chromatin conformation capture techniques. For example, chromatin interaction data were used to determine that an obesity-associated variant located in an intron of the FTO gene affects expression of the downstream genes IRX3 and IRX5, which are involved in obesity-related biological processes (Claussnitzer et al. 2015). Likewise, Hi-C in human cerebral cortex identified FOXG1 as a distal target of a schizophrenia GWAS variant, supporting its potential role as a schizophrenia risk gene (Won et al. 2016).

Because the effects of disease-associated variants are likely limited to particular biological states (Umans et al. 2021), studies of their impact must be conducted in the correct cellular and biological context. Several pieces of evidence suggest that chondrocytes—particularly those responding to cartilage matrix damage—are one of the most likely cell types to be affected by OA risk variants. Cartilage breakdown and loss is a primary feature of OA. Chondrocytes are the only cell type found in cartilage and are responsible for maintaining the cartilage matrix. Osteoarthritic cartilage harbors activated chondrocytes that exhibit a proinflammatory phenotype thought to contribute to progressive cartilage degradation, which includes production of bioactive matrix fragments (Loeser 2014; van den Bosch et al. 2020). We have developed an ex vivo system that simulates the OA chondrocyte phenotype by treating primary human articular chondrocytes with fibronectin fragment (FN-f) (Forsyth et al. 2002; Pulai et al. 2005; Wood et al. 2016; Reed et al. 2021). Fibronectin is a ubiquitous extracellular matrix protein, and high levels of FN-f are present in cartilage and synovial fluid of OA joints (Xie et al. 1992; Homandberg et al. 1998). Subsequently, FN-f has been shown to be an OA mediator that recapitulates gene expression changes associated with OA (Homandberg 1999; Forsyth et al. 2002; Pulai et al. 2005; Reed et al. 2021). We have leveraged this model of OA for use in clonal populations of genome-edited primary human chondrocytes, allowing us to quantify the phenotypic impact of putative target genes of genomic variants in an appropriate disease context.

In this study, we generated CUT&RUN data in primary human chondrocytes and Hi-C data in a human chondrocyte cell line and intersected them with publicly available RNA-seq data from our ex vivo OA model, ChIP-seq data from the Roadmap Epigenomics project, and OA GWAS variants from Boer et al. In doing so, we identified 56 putative OA risk genes, including SOCS2, whose promoter loops to an OA GWAS variant ∼174 kb away. Deletion of SOCS2 in primary human chondrocytes using CRISPR-Cas9 led to heightened expression of inflammatory markers in response to treatment with FN-f, providing a possible mechanism for influencing OA risk.

Methods

Experimental methods

Primary chondrocyte isolation and culture

Primary articular chondrocytes were isolated via enzymatic digestion from human talar cartilage obtained from tissue donors, without a history of arthritis, through the Gift of Hope Organ and Tissue Donor Network (Elmhurt, IL) as previously described (Loeser et al. 2003; Reed et al. 2021). For Cut and Run, 2 million primary articular chondrocytes from 2 male donors, aged 39 and 63 years, were plated onto four 6-cm plates in DMEM/F12 media supplemented with 10% fetal bovine serum, 1% penicillin streptomycin solution, 1% amphotericin B, and 0.04% gentamicin. For genome editing, primary chondrocytes from 3 male donors ages 56, 59, and 64 years were cultured in 6- or 10-cm dishes at a density of approximately 70,000 cells/cm2 in DMEM/F12 media supplemented with 10% FBS and antibiotics.

FN-f treatment

After serum starvation, cells were treated with either purified 42-kDa endotoxin-free recombinant FN-f (final concentration 1 µM in PBS), prepared as previously described, or PBS as control (Wood et al. 2016). Cells were harvested and crosslinked after 90 or 180 min and immediately subjected to Cut & Run, described below.

Hi-C

C-28/I2 cells were cultured in DMEM/F12 media with 10% fetal bovine serum, 1% penicillin streptomycin solution, 1% amphotericin B, and 0.04% gentamicin. Cells were treated with DMEM/F12 media with 1% ITS-Plus for 48 h prior to experiments to promote the chondrocyte phenotype. Cells were then washed with 1× PBS and treated with trypsin-EDTA (0.25%) for 3 min. Trypsin was quenched and cells were pelleted at 4°C for 5 min at 300 × g. Cells were resuspended in 1 mL DMEM/F12 per million cells and crosslinked in 1% formaldehyde for 10 min with rotation before quenching in a final concentration of 0.2 M glycine for 5 min with rotation. Cells were pelleted at 300 × g for 5 min at 4°C. Pellets were washed with cold PBS and aliquoted into ∼3 million cell aliquots. Pellets were flash frozen in liquid nitrogen and stored at −80°C. In situ Hi-C was performed as previously described (Rao et al. 2014). A full description of our methods is provided in the Supplementary Material.

Hi-C data processing

In situ Hi-C datasets were processed using a modified version of the Juicer Hi-C pipeline (https://github.com/EricSDavis/dietJuicer) with default parameters as previously described (Durand et al. 2016). Reads were aligned to the hg19 human reference genome with bwa (v0.7.17) and MboI was used as the restriction enzyme. Four biological replicates were aligned and merged for a total of 2,779,816 Hi-C read pairs in C-28/I2 cells yielding 2,373,892,594 valid Hi-C contacts (85.40%). For visualization, the merged Hi-C contact matrix was normalized with the “KR” matrix balancing algorithm as previously described (Knight and Ruiz 2013) to adjust for regional background differences in chromatin accessibility.

Looping interactions were called at 5-kb resolution with Significant Interaction Peak (SIP) caller (Rowley et al. 2020) (v1.6.2) and Juicer tools (v1.14.08) using the replicate-merged, mapq > 30 filtered hic file with the following parameters: “-norm KR -g 2.0 -min 2.0 -max 2.0 -mat 2000 -d 6 -res 5000 -sat 0.01 -t 2000 -nbZero 6 -factor 1 -fdr 0.05 -del true -cpu 1 -isDroso false.” Loop anchors were expanded to 20 kb and loops with overlapping anchors were filtered out (14 loops). This resulted in 9,271 loops after filtering.

Cut and Run

Primary chondrocytes were washed with 1× PBS and treated with trypsin-EDTA (0.25%) for 5 min. Trypsin was quenched and cells were pelleted at 4°C for 5 min at 1,000 × g. Cells were resuspended in 1 mL plain DMEM per million cells and crosslinked in 1% formaldehyde for 10 min with rotation before quenching in a final concentration of 125 mM glycine for 5 min with rotation. Cells were pelleted by spinning at 1,000 × g for 5 min at 4°C. Each 2 million cell pellet was washed in 1 mL cold PBS prior to flash freezing in liquid nitrogen. We performed Cut and Run following existing protocols (Skene and Henikoff 2017) but modified for crosslinked cells. A full description of our methods is provided in the Supplementary Material.

Cut and Run data processing and peak calling

Adaptors and low-quality reads were trimmed from paired-end reads using Trim Galore! (v0.4.3). Reads were aligned to the hg19 genome with BWA mem (v0.7.17) and sorted with Samtools (v1.9). Duplicates were removed with PicardTools (v2.10.3) and mitochondrial reads were removed with Samtools idxstats. Samtools was also used to merge donors, and index BAM files. Peaks were called from the merged alignments using MACS2 with the following settings: -f BAM -q 0.01 -g hs –nomodel –shift 100 –extsize 200 –keep-dup all -B –SPMR (v2.1.1.20160309). Peaks were then merged using bedtools (v2.26), and multicov was used to extract counts from each replicate BAM file. Signal tracks were made from alignments using deeptools (v3.0.1).

Genome editing of chondrocytes

Preparation of gRNA: Cas9 RNP complex

Two custom SOCS2 Alt-R crRNAs TGACAAGGGCCTATTCCCAC and TTACGCATTCCCAAGGACCC were synthesized by Integrated DNA technologies (IDT). Both sequences are written 5′ to 3′ and do not include PAM sequence. The first crRNA targets the plus strand and the second the minus strand. Ribonucleoprotein (RNP) complexes containing the Cas9 enzyme and sequence-targeting guide RNAs were prepared according to the manufacturer's recommendation. Briefly, Alt-R tracrRNA (1072533, IDT) and crRNA were resuspended in Tris-EDTA buffer to 100 µM concentration and equimolar concentration of crRNA and tracrRNA was combined, heated at 95°C for 5 min and cooled to room temperature to produce the gRNA. Separate RNP complex for each guide was prepared by combining the gRNA (50 µM) with Alt-R Cas9 Nuclease (61 µM) (1081058, IDT) and PBS at a ratio of 1:1.1:2 µL at room temperature for 15 min.

Transfection of primary human chondrocytes with RNP complex and single cell colony selection

Chondrocytes were trypsinized, washed with PBS, and transfected with the RNP complex as previously described with modifications; volumes were scaled up for transfection of more cells in larger cuvettes (D’Costa et al. 2020). Two million cells were resuspended in 100 µL of P3 Primary Cell Nucleofector solution (V4XP-3024, Lonza). The RNP complex and Alt-R Cas9 Electroporation Enhancer (1075916, IDT) was added to the cells. The mixture was gently pipetted up and down and transferred to 100 µL Nucleocuvette vessels (V4XP-3024, Lonza) and transfected using program ER-100 on a 4D-Nucleofector Core unit (Lonza). Cells were kept at room temperature for 8 min and then incubated in prewarmed antibiotic free media containing 20% FBS for recovery. An aliquot of the transfected cells was placed in a 96-well and used for DNA extraction and PCR. Following confirmation of editing, the transfected bulk cells were seeded at low cell density (200 cells per 6 cm2 dish) for generation of single-cell colonies. Individual colonies were picked under a microscope (EVOS FL, ThermoFisher), the colony was disrupted by pipetting and split into 96- and 24-well plates for genetic analysis and continued expansion, respectively.

PCR screening of genome-edited bulk and single-cell-derived colonies

DNA was extracted using QuickExtract DNA Extraction Solution (Lucigen), depending on confluency 25–100 µL of solution was added to the wells containing the cells for 15 min at 37°C, cell suspension was transferred to tubes and vortexed for a minute. Samples were then heated at 65°C for 6 min and 98°C for 2 min. The extracted DNA solution was stored at −20°C. PCR amplification was performed by adding 4 or 5 µL of template DNA, 1 µM forward (SOCS2_F1: accaagtttgtgtgggtgct) and reverse (SOCS2_R1: cttccagcgtgctaagaagc) primers, and EconoTaq PLUS GREEN 2× Master Mix (Lucigen) in a 25 µL reaction. PCR conditions included an initial denaturation at 94°C for 2 min, 35 cycles of denaturation at 94°C for 30 s, annealing at 63°C for 30 s, and extension at 72°C for 65 s, followed by a final extension at 72°C for 10 min. Following amplification of column purified genomic DNA, the PCR product was cleaned up and sequenced using the primers described above and the Bioedit software was used to visualize the chromatograms.

FN-f treatment and quantitative polymerase chain reaction analysis of genome-edited samples

Single-cell colonies in 24-well plates were passaged to 6-well plates for expansion. Following genotype confirmation, colonies with similar genotype were combined and seeded at 250,000 cells per well in a 12-well plate. Cultured cells were made serum free and treated with FN-f 1 µM or PBS. Following treatment, media was removed and cells were immediately lysed in the RLT buffer. RNA was isolated with RNeasy Plus columns (Qiagen) and reverse transcribed to cDNA using qScript XLT cDNA SuperMix (VWR) or iScript cDNA Synthesis Kit (1708891; Bio-Rad). DNase treatment was used for the second and third donors to confirm that detectable SOCS2 signal in knockout cells was due to the presence of genomic DNA. To evaluate the effect of SOCS2 editing on inflammatory gene response quantitative polymerase chain reaction (qPCR) was performed on a QuantStudio 6 Flex machine (Applied Biosystems) with TaqMan Universal Master Mix and TaqMan Gene Expression Assays for human CCL2 (Hs00234140_m1), IL6 (Hs00174131_m1), and housekeeping gene TBP (Hs00427620_m1). SOCS2 expression was assessed in pooled colonies with TaqMan Gene Expression Assay Hs00919620_m1.

Western blot analysis

Following genotype identification by PCR, cells from a wildtype, heterozygous and knockout colony were expanded in chondrocyte media supplemented with 5 ng/mL bFGF and 1 ng/mL TGF-β1 (Life technologies) for 11 days. Cells were lysed in standard cell lysis buffer (1×) (Cell Signaling Technology) containing phenylmethanesulfonyl fluoride (Sigma-Aldrich) and phosphatase inhibitor mix. Protein (15 µg) was separated by SDS-PAGE and transferred to nitrocellulose membrane. After blocking in 5% nonfat milk in TBST, the blot was incubated with SOCS2 antibody (PA5-17219; 1:1,000; Thermo Fisher) overnight at 4°C and secondary antibody solution for 1 h. The membrane was incubated in Radiance Plus Chemiluminescent Substrate (Azure Biosystems) and signal detected using the Azure c600 gel imaging system. The membrane was striped and incubated with the loading control beta tubulin antibody.

Computational methods

Cell type enrichment for OA risk variants

To identify the cell types that likely mediate genetic OA risk, we performed SNP enrichment analysis using GREGOR (Genomic Regulatory Elements and Gwas Overlap algoRithm) (Schmidt et al. 2015). Publicly available H3K27ac, H3K4me1, and H3K4me3 ChIP-seq narrowPeaks files from the NIH Roadmap Epigenomics Mapping Consortium were merged and sorted using bedtools (v2.29.2) (Quinlan and Hall 2010) to define regulatory loci for 98 cell types in hg19. GREGOR was used to determine each cell type’s enrichment for 104 OA lead SNPs (Boer et al. 2021) by comparing the observed overlap between regulatory loci and SNPs with their expected overlap and evaluating significance. Expected overlap is determined using a matched control set of ∼500 variants that control for the number of LD proxies, gene proximity and minor allele frequency. Reference data from 1000 Genomes Phase 1 version 2 EUR panel were used with GREGOR to control for LD proxies (1 Mb, r2 > 0.7) (1000 Genomes Project Consortium et al. 2010). Results were imported into R (v4.1.0) (R Core Team 2021) and visualized with ggplot2 (Wickham 2016) and plotgardener (Kramer et al. 2022).

Putative OA risk variants

LD proxies for 104 OA GWAS signals from Boer et al. were identified using the 1000 Genomes European reference panel since the GWAS data primarily analyzed individuals of European ancestry (11 of 13 cohorts are of European descent). r2 values were calculated with the –ld function in PLINK 1.9 (Purcell et al. 2007; 1000 Genomes Project Consortium et al. 2010) using a window of 1 Mb for LD calculation. Putative OA risk variants were defined as those in high LD (r2 > 0.8, n = 1,259) with lead variants.

Multiomic integration for assigning SNPs to putative OA risk genes

We took a multiomic approach to identify putative SNP–gene pairs implicated in OA. SNPs that (1) were predicted to affect coding regions of genes, (2) overlapped gene promoters, or (3) overlapped a regulatory peak looped to a gene’s promoter were assigned to the “Coding gene,” “Gene promoter,” or “Loops to gene promoter” categories, respectively. Genes in each category that change in response to FN-f (P ≤ 0.01 and LFC at any time point ≥1.25) were highlighted as putative OA risk genes.

Coding SNP–gene pairs were identified using ENSEMBL’s Variant Effect Predictor (VEP) tool. Putative OA risk variants (n = 1,259) were annotated with their predicted consequence on coding sequence using VEP run with the GRCh37.p13 human genome and default parameters. SNPs with a predicted consequence of “missense” or “synonymous” were paired with their affected genes assigned to the “Coding gene” category.

Promoter regions were defined as 2,000-bp upstream and 200-bp downstream of the TSS of transcripts obtained with the TxDb.Hsapiens.UCSC.hg19.knownGene Bioconductor package for a total of 82,960 transcripts. Gene symbols were linked to transcript ranges using the OrganismDbi and Homo.sapiens packages. Transcripts without gene symbols or those not present in the FN-f RNA-seq data were filtered out, leaving a total of 62,590 transcript promoters.

Chondrocyte regulatory regions were defined by combining Roadmap Epigenomics data with data from primary human articular chondrocytes. Specifically, H3K4me1, H3K4me3, H3K9ac, H3K9me3, H3K27ac, H3K27me3, and H3K36me3 peaks from mesenchymal stem cell derived chondrocyte cultured cells (E049) obtained through AnnotationHub (v3.1.7, snapshot date 20 October 2021) (Morgan et al. 2017) were combined with Donor-merged H3K27ac peaks from primary human articular chondrocytes. OA SNPs were overlapped with chondrocyte regulatory regions resulting in 507 SNPs.

SNPs overlapping chondrocyte regulatory regions that also overlapped a promoter region were assigned to their affected gene and added to the “Gene promoter” category. SNPs overlapping chondrocyte regulatory regions were intersected with loop calls from Hi-C in the C-28/I2 chondrocyte cell line (see Methods on Hi-C processing and loop calling). The linkOverlaps function from the InteractionSet package was used to identify chondrocyte regulatory SNPs that are connected to promoters by loops. These SNP–gene pairs were assigned to the “Loops to gene promoter” category.

Motif analysis

Tomtom (v5.4.1; release date: Saturday August 21 19:23:23 2021—0700) from the MEME suite was used to identify motif matches for sequences surrounding the rs7953280 variant (Gupta et al. 2007). All 7-mers surrounding rs7953280 (“GGCTTTG,” “GCTTTGA,” “CTTTGAG,” “TTTGAGG,” “TTGAGGC,” “TGAGGCA,” and “GAGGCAT”) and the entire 13-bp sequence (“GGCTTTGAGGCAT”) were used to identify motif matches. Sequences were input into the online motif comparison tool and queried against the JASPAR2022_CORE_vertebrates_nonredundant_v2 and HOCOMOCOv11_core_HUMAN_mono_meme_format motif databases. Pearson correlation coefficient was used as the motif column comparison function and the significance threshold was set to an E-value <10; no q-value threshold was set and reverse complementing of motifs was permitted. The following command summarizes the parameters used: “tomtom -no-ssc -oc . -verbosity 1 -min-overlap 5 -mi 1 -dist pearson -evalue -thresh 10.0 -time 300 query_motifs motif_databases.”

Transcription factor motif-binding propensity

We used SNP effect matrix (SEM) scores to predict the transcription factor (TF)-binding propensity between risk and nonrisk SNPs in OA. Precalculated SEMs for 211 TF motifs were obtained from SEMpl (https://github.com/Boyle-Lab/SEMpl) and used for scoring risk and nonrisk SNP sequences (Nishizaki et al. 2020). Binding propensity scores were determined by generating frame-shifted K-mers covering each TF motif position for both risk and nonrisk sequences. K-mers were scored against 211 TF SEMs using position-weight matrix scoring functions from the Biostrings Bioconductor package (Pagès et al. 2021). The best scoring K-mer frame for each TF motif was used to select the binding score for risk and nonrisk sequences. Scores were normalized by applying inverse-log transformation, subtracting the scrambled baseline provided with each SEM, and dividing the result by the absolute value of that baseline. TFs with positive scores are predicted to be bound while negative scores are predicted to be unbound.

Results

OA risk variants are enriched in chondrocyte regulatory loci

One of the first steps in decoding GWAS variant mechanisms is to determine the cell types that are likely mediating genetic OA risk. While different risk variants may impact distinct cell types, one approach to help direct research is to determine the cell types which harbor regulatory loci (e.g. enhancers) that are enriched for risk variants. To accomplish this, we performed SNP enrichment analysis using the Genomic Regulatory Elements and Gwas Overlap algoRithm (GREGOR) (Schmidt et al. 2015). Publicly available H3K27ac, H3K4me1, and H3K4me3 ChIP-seq peaks from the NIH Roadmap Epigenomics Mapping Consortium (Roadmap) were merged to define regulatory elements for 98 cell types. GREGOR was used to determine each cell type’s enrichment for 104 OA GWAS signals recently published in Boer et al. (2021).

The regulatory elements of “Chondrocytes from Bone Marrow Derived Mesenchymal Stem Cell Cultured Cells” (E049) exhibited a strong effect size and P-value of enrichment for OA risk variants (Fig. 1a), suggesting that many OA risk variants may impact regulatory events in chondrocytes. This is consistent with the known role of chondrocytes in maintaining joint homeostasis. Chondrocytes have been heavily implicated in OA, as activation of chondrocytes by mechanical and inflammatory stimuli triggers downstream inflammatory and catabolic response pathways in diseased tissue (Sandell and Aigner 2001; Pelletier et al. 2001; Loeser et al. 2012; Caron et al. 2015). An example of an OA risk variant that overlaps a chondrocyte-specific regulatory element (H3K27ac peak) is shown in Fig. 1b. For comparison, Fig. 1c shows a non-OA-associated variant that overlaps a noncell type specific, or ubiquitous, enhancer on chromosome 10 that is active in >90% of the 98 cell types evaluated. These examples underscore the importance of interpreting GWAS risk variants in light of the correct cellular context, as the variant-H3K27ac peak overlap shown in Fig. 1b would not have been detected in any of the other cell types investigated. In addition to E049, IMR90 fetal fibroblasts (E017) and HSMM cell derived Skeletal Muscle Myotubes (E121) were also enriched, suggesting that OA risk variants may also contribute to disease risk through altering the function of fibroblasts and muscle. However, given the strong enrichment in chondrocytes and their documented role in OA biology, we chose to focus our investigation of OA GWAS variants in human chondrocytes.

Fig. 1.

Fig. 1.

Open in a new tab

OA risk variants are enriched in chondrocyte regulatory elements. a) Enrichment analysis of 98 cell types from the NIH Roadmap Epigenomics Mapping Consortium reveals that OA GWAS variants are enriched in the regulatory regions (H3K27ac, H3K4me1, or H3K4me3 ChIP-seq peaks) of chondrocytes, skeletal muscle myotubes, and fibroblasts. b) Heatmap of H3K27ac signal from 98 cell types (bottom) highlights a chondrocyte-specific enhancer that overlaps Knee/Hip OA risk variant (Boer et al. 2021) (rs4760618, circled) that is in high LD (r2 > 0.8) with the lead variant (rs7967762, diamond) for this locus (top). c) Heatmap of H3K27ac signal from 98 cell types (bottom) highlights a ubiquitous enhancer (active in >90% of cell types) that does not overlap an OA GWAS variant (top).

Multiomic integration identifies putative variant–gene associations in OA

Due to high LD between variants and the fact that most risk variants reside in noncoding sequences, determining the causal variants and genes they impact remains a major challenge. To address these issues, we generated novel maps of epigenetic features in human chondrocytes and integrated them with GWAS results and publicly available genomic datasets to identify putative variant–gene associations for OA.

First, we identified OA risk variants that are predicted to directly affect protein sequences. We used ENSEMBL’s Variant Effect Predictor (VEP) tool to predict the consequences of 1,259 putative OA risk variants that were in high LD (r2 > 0.8) with 104 OA GWAS signals from Boer et al. (2021). VEP identified 29 variants at 19 loci predicted to affect the coding sequence of 24 unique genes (Fig. 2a, top). Eighteen of these variants encode a missense mutation impacting 17 genes, while 11 variants encode a synonymous mutation impacting 8 genes (Fig. 2a, top). Though synonymous variants do not impact the protein sequence directly, differences in transcription efficiency, tRNA availability, and mRNA stability introduced through these variants could contribute to the OA phenotype (Venetianer 2012; Zeng and Bromberg 2019). To identify genes most likely to impact OA risk, we incorporated our previously published RNA-seq FN-f time course data to find genes that change expression in an OA context. Of the 24 genes identified here, 6 exhibited differential expression in response to FN-f (Fig. 2a, bottom). Several of the genes identified have been previously implicated in OA, including Interleukin 11 (IL11), Solute Carrier Family 39 Member 8 (SLC39A8/ZIP8), and Serpin Family A Member 1 (SERPINA1). IL11 plays a role in bone turnover and is upregulated in subchondral bone and articular cartilage from OA tissue (Tuerlings et al. 2021). SLC39A8 is upregulated in OA chondrocytes and suppression of SLC39A8 in a mouse OA model significantly reduces cartilage degradation (Song et al. 2013). SERPINA1, a serine protease inhibitor with anti-inflammatory capabilities (Jain et al. 2011), is downregulated in OA (Boeuf et al. 2008; Wanner et al. 2013).

Fig. 2.

Fig. 2.

Open in a new tab

Multiomic integration for assigning SNPs to putative OA risk genes. a) ENSEMBL’s Variant Effect Predictor tool identified 29 unique OA risk SNPs (18 missense and 11 synonymous) overlapping coding regions of 24 unique genes (17 missense, 8 synonymous). b) Twenty-six unique OA risk SNPs overlapped both a chondrocyte regulatory region (H3K27ac, H3K4me1, or H3K4me3 ChIP-seq peaks) and a gene promoter for 21 unique genes. c) Forty-seven unique SNPs overlapped chondrocyte regulatory regions connected to 20 unique gene promoters via 14 C-28/I2 chromatin loops. RNA-seq data from our ex vivo OA model depicts how putative OA risk genes change in response to FN-f. Normalized expression of genes are shown below each category over an 18 h time-course of FN-f treatment. Differential genes (P ≤ 0.01, absolute log2 fold-change ≥1.25) are colored and labeled.

Next, we identified OA risk variants that could impart their phenotypic impact by altering promoter and/or enhancer activity. To define the most accurate regulatory regions in chondrocytes we used CUT&RUN to map histone H3K27 acetylation (H3K27ac)—a mark of active enhancers and promoters—in primary human chondrocytes isolated from the knees of 2 cadaveric donors. As expected, Roadmap chondrocyte (E049) H3K27ac peaks showed the highest degree of similarity by Jaccard distance to H3K27ac peaks called in our primary human chondrocyte data (Supplementary Fig. 1). Therefore, we merged our primary chondrocyte H3K27ac peaks with all available marks (including active and repressive marks) from the E049 chondrocyte cell line from Roadmap Epigenomics to define a comprehensive set of chondrocyte regulatory elements. Integration of public and newly generated sources of human chondrocyte features allowed us to identify 507 plausible regulatory variants from 1,259 OA risk variants.

Intersecting these 507 plausible regulatory variants with gene annotations (UCSC) identified 26 unique variants that overlapped the promoters of 21 genes (Fig. 2b). Two of these genes were differentially expressed in response to FN-f, both of which have been previously implicated in OA. Growth and Differentiation Factor 5 (GDF5), a member of the TGF-beta family, has roles in skeletal and joint development (Francis-West et al. 1999) and has been identified as a major risk locus for OA (Miyamoto et al. 2007; Southam et al. 2007). Specifically, variants in the GDF5 enhancers R4 and GROW1 have been associated with altered anatomical features of the knee and hip, which are thought to confer an increased risk of OA (Capellini et al. 2017; Richard et al. 2020; Muthuirulan et al. 2021). Solute Carrier Family 44 Member 2 (SLC44A2, aka choline transporter-like protein 2) is a mitochondrial choline transporter that has been identified as an expression quantitative trait locus in OA tissue (Steinberg et al. 2021) that colocalizes with the OA GWAS signal rs1560707 (Steinberg et al. 2020).

In addition to direct regulation of genes by their promoters, long-range regulation of genes also occurs via enhancer-promoter interactions mediated by chromatin loops (Maurano et al. 2012). To identify such connections, we conducted deeply sequenced (∼2.8 billion reads) in situ Hi-C in C-28/I2 chondrocyte cells and identified 9,271 chromatin loops with SIP caller at 5-kb resolution which we expanded to 20 kb for downstream analysis (Rowley et al. 2020). C-28/I2 cells were used because they could be expanded to easily provide the number of cells required for Hi-C analysis. To our knowledge, this is the first Hi-C map in a chondrocyte cell line, enabling us to discover novel OA-associated variant–gene connections. We performed 4 replicates which exhibited a high degree of reproducibility as measured by stratum-adjusted correlation coefficient (SCC > 0.98) with the HiCRep package (Supplementary Fig. 2) (Yang et al. 2017; Lin et al. 2021). Overlapping these data with OA risk variants identified 14 loops connecting 47 variants among 14 loci to 20 unique gene promoters (Fig. 2c). Four of these genes were differentially expressed in response to FN-f (P ≤ 0.01, absolute log2 fold-change ≥1.25) and are visualized in Fig. 2a and Supplementary Fig. 3. Several of these genes have interesting implications for OA, including FGFR3 (Fibroblast Growth Factor Receptor 3), which plays a role in skeletal development. FGFR3 may have an important function in the maintenance of articular cartilage (Zhou et al. 2016; Tang et al. 2016; Okura et al. 2018), possibly through the Indian hedgehog signaling pathway, which plays a role in regulating chondrocyte hypertrophy and the expression of cartilage matrix-degrading enzymes (Lin et al. 2009). FGFR3 is also downregulated in OA tissues, further implicating its potential role in limiting articular cartilage degeneration (Li et al. 2012; Shu et al. 2016).

Altogether we identified 24 genes impacted by a coding variant, 21 genes with at least 1 regulatory variant in their promoters, and 20 genes that were connected to a regulatory variant via a chromatin loop. Since genes can fall into multiple categories, the number of total distinct genes identified is 56. All putative variant–gene associations are reported in Supplementary Table 1. Boer et al. identified 637 putative effector genes and ranked them by the amount of evidence for association with OA signals (Boer et al. 2021). In general, genes with higher tiers of evidence as reported by Boer et al. were more likely to be supported by our analyses (Supplementary Fig. 4 and Supplementary Table 2). For example, 67% genes that were supported by 6 tiers of evidence were also detected in our study, whereas only 2.5% of tier 1 genes were supported by our work. Interestingly, 42% of the genes we identified were not previously implicated by Boer et al. Fifty-four percent of the genes unique to our study were supported by a chromatin loop compared to only 22% of genes implicated by both studies. This underscores the additional value our study provides by incorporating cell type-specific Hi-C data.

Chondrocyte chromatin features identify SOCS2 as a putative regulator of OA

Our multiomic analysis identified an association between rs7953280 and the promoter of Suppressor Of Cytokine Signaling 2 (SOCS2). rs7953280 is located in an intron of the CRADD gene, which is expressed at low levels in primary chondrocytes, does not change expression in response to FN-f, and lacks an obvious biological relevance to OA. However, rs7953280 overlaps a putative chondrocyte enhancer (i.e. histone H3K27ac peak), suggesting that it could alter the regulatory capacity of the enhancer and impact the expression of a proximal or distal gene. This enhancer is connected to the promoter of SOCS2 via a 174-kb chromatin loop (Fig. 3a). Unlike CRADD, SOCS2’s expression changes in response to FN-f, peaking at 3 h (Figs. 2c and 3a, orange signal tracks). Moreover, SOCS2 is known to play a role in resolving inflammatory response through NFKB and is downregulated in knee OA tissues (de Andrés et al. 2011; Paul et al. 2017), making it an intriguing candidate as an OA risk gene. No other SNPs from this locus can be assigned to genes using our integrated approach.

Fig. 3.

Fig. 3.

Open in a new tab

3D chromatin interactions identify SOCS2 as a putative regulator of OA. a) Hi-C performed in C-28/I2 cells reveals a chromatin loop connecting OA risk variant rs7953280 (right gray bar) to the promoter of SOCS2 (left gray bar). rs7953280 is located in an intronic region of CRADD and overlaps an H3K27ac peak in primary articular human chondrocytes from two donors (blue signal tracks). SOCS2 is differentially expressed in response to treatment with FN-f. Gene tracks are shown below with +/− indicating gene strand. b) Zoom-in on rs7953280 shows that the SNP is located within an H3K27ac peak in primary articular human chondrocytes. c) Motif analysis identifies a JUN-binding site at rs7953280. SEM data predicts decreased binding at the JUN motif (JASPAR ID: MA0099.2) with a G to C polymorphism in the second position. d) Motif analysis from 211 precomputed SEMs from SEMpl predicts that JUN/AP-1 motifs (red, upper left quadrant) bind to the nonrisk but not the risk allele.

To further understand how rs7953280 may confer risk for OA, we examined the sequence surrounding rs7953280 to see if it overlaps and alters any TF-binding motifs. Motif comparison with Tomtom from the MEME suite identified FOS and JUN as matching target motifs (Fig. 3c, Supplementary Table 3). FOS and JUN are members of the Activator Protein 1 (AP-1) complex, which is upregulated in response to FN-f (Reed et al. 2021), and the inhibition of which prevents cartilage degradation in a model of OA (Motomura et al. 2018; Fisch et al. 2018; Gao et al. 2019). We then used SEMs generated by the SNP effect matrix pipeline (SEMpl) (Nishizaki et al. 2020) to assess the predicted consequence of the G (nonrisk) to C (risk) variant on binding of JUN or any other of the 211 motifs included with SEMpl (Fig. 3d). Most TFs are predicted to be unbound at both alleles. However, multiple JUN/AP-1 motifs are predicted to bind to the nonrisk, but not the OA-risk sequence (Fig. 3d) providing further evidence that the G → C mutation in rs7953280 may disrupt JUN/AP-1 binding. Our analysis also showed that STAT-1 was predicted to bind only to the OA-risk sequence, although the SEM score was very close to the cutoff for predicted binding. Nevertheless, since STAT-1 is an important mediator for inflammatory signaling, rs7953280 could influence inflammation during OA progression by modulating STAT-1 binding.

SOCS2 deletion increases proinflammatory gene expression in response to FN-f

To assess the functional role of SOCS2, we used CRISPR-Cas9 to knock out SOCS2 in primary human chondrocytes isolated from 3 individual donors. After targeting the SOCS2 gene with 2 guide RNAs that flank exon 2 (a constitutive exon that contains the translational start site), we used our previously developed method that employs PCR to screen single-cell-derived colonies (D’Costa et al. 2020). The screening primers generated a 1,068-bp product if the region was intact and a novel 240-bp amplicon if the 2 guides successfully deleted the intended 828-bp region (Fig. 4a). We saw efficient deletion in each of the 3 donors, with 31% of the colonies showing no deletion, 49% of the colonies showing heterozygous deletion, and 20% showing homozygous deletion (Fig. 4b). Sanger sequencing was used to confirm deletions, while qPCR and western blotting confirmed partial (heterozygous) or complete (homozygous) loss of SOCS2 expression (Supplementary Fig. 5).

Fig. 4.

Fig. 4.

Open in a new tab

SOCS2 deletion increases proinflammatory gene expression in response to FN-f. a) PCR primers surrounding the intended SOCS2 deletion were used to screen single-cell-derived colonies from 3 independent donors. WT, wild type (+/+. purple); het, heterozygous deletion (+/−, blue); KO, homozygous knockout deletion (−/−, green). b) Efficiency of deleting the intended SOCS2 deletion in primary human chondrocytes from 3 independent donors. c) qPCR at 18 h after FN-f treatment revealed increased expression of the proinflammatory genes IL6 and CCL2 in SOCS2 deletion colonies from 3 independent donors.

Because SOCS2 is a known negative regulator of the inflammatory response in other settings (Paul et al. 2017; Monti-Rocha et al. 2018), we hypothesized that SOCS2 deletion would lead to an increased expression of inflammatory cytokines in chondrocytes during the response to FN-f. To test this hypothesis, we treated chondrocytes with defined genotypes (wild type, heterozygous, or homozygous knockout) with either FN-f or PBS for 18 h and quantified the change in proinflammatory cytokines C–C Motif Chemokine Ligand 2 (CCL2) and Interleukin 6 (IL6) using qPCR. IL6 and CCL2 have previously been shown to exhibit increased expression after 18 h of FN-f treatment and are also implicated in OA (Wojdasiewicz et al. 2014; Wang and He 2018; van den Bosch et al. 2020; Reed et al. 2021). Deletion of SOCS2 led to increased expression of both IL6 and CCL2 in response to FN-f treatment (Fig. 4c), and these increases were observed in a dose-dependent fashion, with greater increases observed in the homozygous compared to heterozygous genotypes. These results suggest that the loss of SOCS2 may promote a heightened inflammatory response to FN-f stimulation, which is consistent with a potential role in OA.

Discussion

We used a multiomic approach to identify putative causal SNPs and genes associated with OA risk. The efficacy of this approach was supported by the identification of previously known OA risk genes including GDF5, SLC44A2, and IL11. We generated the first maps of histone H3K27ac in primary human chondrocytes and integrated this dataset with publicly available genomic datasets to reduce thousands of OA risk GWAS variants to a small list of variants and genes for further study. By generating the first Hi-C contact map of human chondrocytes, we were able to uncover 73 previously unknown connections between OA risk variants and putative target genes. Most looped variant–gene pairs (71 of 73) skipped over the nearest gene, connecting variants to genes as far as 414 kb away. DNA looping revealed 20 unique genes, 13 of which were not identified by recent fine mapping approaches (Boer et al. 2021) and could provide new avenues for therapeutic interventions for OA.

Among the genes identified with Hi-C, 4 were found to be differentially expressed in our OA model. FGFR3 and SOCS2 have previously been implicated in OA, while Tropomoyosin 1 (TPM1) and Ral Guanine Nucleotide Dissociation Stimulator Like 1 (RGL1) have not. However, TPM1, an actin-binding protein involved in the contractile system of muscle cells and the cytoskeleton of nonmuscle cells has been shown to play roles in an inflammatory response in various cell types, such as human primary coronary artery smooth muscle cells (Li et al. 2022) and rod bipolar and horizontal cells in the retina (Gagat et al. 2021). RGL1, which functions as a RAS effector protein that activates GTPase by stimulating nucleotide exchange, has also been shown to modulate immune response in both vascular and immune cells (Kirkby et al. 2014) and, interestingly, is downregulated in human articular chondrocytes upon treatment with interleukin-1 and oncostatin-M (Yang et al. 2021). The functions of TPM1 and RGL1 in inflammatory responses may point to potentially undiscovered roles in OA.

One especially intriguing gene was SOCS2, whose promoter is looped to an OA risk SNP within a histone H3K27ac peak ∼170 kb away. SOCS2 is known to inhibit the JAK/STAT pathway and is induced by various pro-inflammatory cytokines such as interleukin-6, growth hormone, and tumor necrosis factor-alpha (Starr et al. 1997; Metcalf et al. 2000; Santangelo et al. 2005). CRISPR-mediated deletion of SOCS2 was associated with increased expression of IL6 and CCL2 in our ex vivo model of OA, suggesting that it may also play a role in mediating inflammation in response to cartilage matrix damage. These findings make SOCS2 a candidate for further studies and the activation of more robust SOCS2 expression could be a goal for future therapeutic development. The regulatory role of SOCS2 in chondrocytes is likely to be subtle, as Socs2 knockout mice did not show altered OA development (Samvelyan et al. 2022). Because that study used a global germline deletion, other members of the inflammatory cascade may have compensated for Socs2 loss, and it would be interesting to determine whether the inducible loss of Socs2 in adult chondrocytes would generate a different result.

Expression quantitative trait locus analysis (GTEx Project v8) provides evidence that variation at rs7953280 is associated with SOCS2 expression in fibroblasts. In that data set the C allele is associated with increased expression of SOCS2. This is contrary to what we predict for chondrocytes but could be explained by differences in cell type or condition (e.g. resting vs. stimulated with FN-f). Mapping of QTLs in chondrocytes responding to FN-f could shed light on these differences.

While further work is needed to clarify the role of rs7953280 and SOCS2 in mediating OA risk, our multiomic analysis suggests the following potential model. In cells harboring the nonrisk variant, proinflammatory cytokines such as IL-6 and matrix damage products such as FN-f may activate AP-1 via the JAK/STAT pathway. AP-1 may then bind the enhancer at the rs7953280 locus, increase enhancer activity, and upregulate transcription of SOCS2 via a chromatin loop between the enhancer and the SOCS2 promoter. In cells harboring the risk allele, AP-1 binding would be decreased, impeding enhancer activation and proper upregulation of SOCS2. As a result, JAK/STAT signaling would remain high, resulting in prolonged or heightened inflammation and further cartilage degradation.

This model, while compelling, will require further experimental investigation and validation. One such experiment would be to use genome editing of noncoding sequences to directly test the effect of rs7953280 on SOCS2. While implementing single base changes using CRISPR/Cas9 and homology-directed repair donor oligos in chondrocytes is technically challenging, engineering isogenic chondrocytes with the risk and protective alleles will help validate the association between the variant and SOCS2 expression and the observed inflammatory response. Moreover, future experiments are required to determine to the degree to which these findings translate from our ex vivo model into an in vivo system and/or if activation of SOCS2 could provide therapeutic avenue for OA treatment.

We generated the first maps of histone H3K27ac in primary human chondrocytes, provided the first maps of 3D chromatin contacts in chondrocytes of any type, and identified 56 putative OA risk genes using multiomic data integration. For each locus, we provide 0, 1 or multiple putative OA risk genes. While these analyses narrow the search space for the genes affected by OA risk variants and allow for the formation of new hypotheses, determining which genes are truly causal will require further experimental validation similar to the approach described here to investigate SOCS2. We chose to perform functional experiments on SOCS2 because it had the strongest genomic evidence for mediating OA risk of the genes looped to OA risk variants; however, many of the other genes implicated here (via looping or otherwise) may also influence disease risk and warrant further investigation. These putative risk genes and novel epigenetic datasets will provide a foundation for future studies to investigate the genetic variants responsible for OA risk and expedite our search for better prevention and treatment of OA.

Supplementary Material

iyac141_Supplementary_Figure_Legends
Click here for additional data file. (7.5KB, docx)
iyac141_Supplementary_Figure_S1
Click here for additional data file. (15.7KB, pdf)
iyac141_Supplementary_Figure_S2
Click here for additional data file. (14.4KB, pdf)
iyac141_Supplementary_Figure_S3
Click here for additional data file. (1.4MB, pdf)
iyac141_Supplementary_Figure_S4
Click here for additional data file. (4.7KB, pdf)
iyac141_Supplementary_Figure_S5
Click here for additional data file. (3MB, pdf)
iyac141_Supplementary_Tables_1_3
Click here for additional data file. (73.2KB, xlsx)

Acknowledgments

We thank Jesse Raab and Thomas Vierbuchen for helpful guidance on CUT & RUN protocols, Erika Deoudes for graphic design and typesetting, and Samantha Pattenden for use of the Covaris LE220 instrument, which was provided by the North Carolina Biotechnology Center Institute Development Program grant 2017-IDG-1005. We would also like to thank the Gift of Hope Organ and Tissue Donor Network, the donor families, and Dr. Susan Chubinskaya for providing normal donor tissue, as well as Dr. Pranav Mishra for donor tissue procurement and Mrs. Arnavaz Hakimiyan for technical assistance.

Funding

This work was supported by NIH grants (R35-GM128645 to DHP, R37-AR049003 to RFL, and R56-AG066911 to BOD) and multiple NIH training grants (T32-GM067553 for ESD and NEK and T32-GM007092 for ET). The project was also supported by the National Center for Advancing Translational Sciences (NCATS) through NIH Grant UL1TR002489 and by the UNC Thurston Arthritis Research Center through a pilot and feasibility grant. ET was supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-2040435. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Conflicts of interest

None declared.

Contributor Information

Eliza Thulson, Curriculum in Genetics and Molecular Biology, University of North Carolina, Chapel Hill, NC 27599, USA.

Eric S Davis, Curriculum in Bioinformatics and Computational Biology, University of North Carolina, Chapel Hill, NC 27599, USA.

Susan D’Costa, Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC 27599, USA.

Philip R Coryell, Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC 27599, USA.

Nicole E Kramer, Curriculum in Bioinformatics and Computational Biology, University of North Carolina, Chapel Hill, NC 27599, USA.

Karen L Mohlke, Department of Genetics, University of North Carolina, Chapel Hill, NC 27599, USA.

Richard F Loeser, Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC 27599, USA; Division of Rheumatology, Allergy and Immunology, University of North Carolina, Chapel Hill, NC 27599, USA.

Brian O Diekman, Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC 27599, USA; Joint Department of Biomedical Engineering, University of North Carolina and North Carolina State University, Raleigh, NC 27695, USA; Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC 27599, USA.

Douglas H Phanstiel, Curriculum in Genetics and Molecular Biology, University of North Carolina, Chapel Hill, NC 27599, USA; Curriculum in Bioinformatics and Computational Biology, University of North Carolina, Chapel Hill, NC 27599, USA; Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC 27599, USA; Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC 27599, USA; Department of Cell Biology and Physiology, University of North Carolina, Chapel Hill, NC 27599, USA.

Data Availability

Hi-C and CUT&RUN data can be accessed through GEO accession GSE200345 (see Datasets for publicly available datasets; OA GWAS, Epigenome Roadmap Data, and RNA-seq time course of FN-f treatment).

Supplemental material is available at GENETICS online.

Datasets

OA GWAS

Genome-wide association statistics for 11 OA phenotypes and lead variants identified in Boer et al. (2021) were obtained from the Musculoskeletal Knowledge Portal (Kiel et al. 2020).

Epigenome Roadmap Data

Consolidated reference human epigenomes for 98 cell/tissue types were obtained from the NIH Roadmap Epigenomics Project (Bernstein et al. 2010) and The Encyclopedia of DNA Elements (ENCODE) project (ENCODE Project Consortium 2012). Processed narrowPeak files for H3K27ac, H3K4me1, and H3K4me3 and BigWig files for H3K27ac were used for each cell/tissue type. Additional narrowPeak files for H3K9ac, H3K9me3, H3K27me3, and H3K36me3 were obtained for mesenchymal stem cell-derived chondrocyte-cultured cells (E049).

RNA-seq time course of FN-f treatment

RNA-seq data from a prior study of FN-f treated human chondrocytes was obtained from KSM Reed et al. (Reed et al. 2021) and vst-normalized, centered, and replicate-combined. The 0-h FN-f treatment time point was created by combining the 9 PBS-treated replicates. Genes were considered differential with a BH-adjusted P-value of 0.01 and a log2 fold-change threshold >1.25 across any time point.

Literature cited

  1. 1000 Genomes Project Consortium; Abecasis GR, Altshuler D, Auton A, Brooks LD, Durbin RM, Gibbs RA, Hurles ME, McVean GA.. A map of human genome variation from population-scale sequencing. Nature. 2010;467(7319):1061–1073. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. de Andrés MC, Imagawa K, Hashimoto K, Gonzalez A, Goldring MB, Roach HI, Oreffo ROC.. Suppressors of cytokine signalling (SOCS) are reduced in osteoarthritis. Biochem Biophys Res Commun. 2011;407(1):54–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bernstein BE, Stamatoyannopoulos JA, Costello JF, Ren B, Milosavljevic A, Meissner A, Kellis M, Marra MA, Beaudet AL, Ecker JR, et al. The NIH Roadmap Epigenomics Mapping Consortium. Nat Biotechnol. 2010;28(10):1045–1048. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Boer CG, Hatzikotoulas K, Southam L, Stefánsdóttir L, Zhang Y, Coutinho de Almeida R, Wu TT, Zheng J, Hartley A, Teder-Laving M, et al. ; Regeneron Genetics Center. Deciphering osteoarthritis genetics across 826,690 individuals from 9 populations. Cell. 2021;184(24):6003–6005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Boeuf S, Steck E, Pelttari K, Hennig T, Buneb A, Benz K, Witte D, Sültmann H, Poustka A, Richter W, et al. Subtractive gene expression profiling of articular cartilage and mesenchymal stem cells: serpins as cartilage-relevant differentiation markers. Osteoarthr Cartil. 2008;16(1):48–60. [DOI] [PubMed] [Google Scholar]
  6. van den Bosch MHJ, van Lent PLEM, van der Kraan PM.. Identifying effector molecules, cells, and cytokines of innate immunity in OA. Osteoarthr Cartil. 2020;28(5):532–543. [DOI] [PubMed] [Google Scholar]
  7. Capellini TD, Chen H, Cao J, Doxey AC, Kiapour AM, Schoor M, Kingsley DM.. Ancient selection for derived alleles at a GDF5 enhancer influencing human growth and osteoarthritis risk. Nat Genet. 2017;49(8):1202–1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Caron MMJ, Emans PJ, Surtel DAM, van der Kraan PM, van Rhijn LW, Welting TJM.. BAPX-1/NKX-3.2 acts as a chondrocyte hypertrophy molecular switch in osteoarthritis. Arthritis Rheumatol. 2015;67(11):2944–2956. [DOI] [PubMed] [Google Scholar]
  9. Chesi A, Wagley Y, Johnson ME, Manduchi E, Su C, Lu S, Leonard ME, Hodge KM, Pippin JA, Hankenson KD, et al. Genome-scale Capture C promoter interactions implicate effector genes at GWAS loci for bone mineral density. Nat Commun. 2019;10(1):1260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Claussnitzer M, Dankel SN, Kim K-H, Quon G, Meuleman W, Haugen C, Glunk V, Sousa IS, Beaudry JL, Puviindran V, et al. FTO obesity variant circuitry and adipocyte browning in humans. N Engl J Med. 2015;373(10):895–907. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. D’Costa S, Rich MJ, Diekman BO.. Engineered cartilage from human chondrocytes with homozygous knockout of cell cycle inhibitor p21. Tissue Eng Part A. 2020;26(7–8):441–449. [DOI] [PubMed] [Google Scholar]
  12. Duan A, Wang H, Zhu Y, Wang Q, Zhang J, Hou Q, Xing Y, Shi J, Hou J, Qin Z, et al. Chromatin architecture reveals cell type-specific target genes for kidney disease risk variants. BMC Biol. 2021;19(1):38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Durand NC, Shamim MS, Machol I, Rao SSP, Huntley MH, Lander ES, Aiden EL.. Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. Cell Syst. 2016;3(1):95–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature. 2012;489:57–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fisch KM, Gamini R, Alvarez-Garcia O, Akagi R, Saito M, Muramatsu Y, Sasho T, Koziol JA, Su AI, Lotz MK, et al. Identification of transcription factors responsible for dysregulated networks in human osteoarthritis cartilage by global gene expression analysis. Osteoarthr Cartil. 2018;26(11):1531–1538. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Forsyth CB, Pulai J, Loeser RF.. Fibronectin fragments and blocking antibodies to alpha2beta1 and alpha5beta1 integrins stimulate mitogen-activated protein kinase signaling and increase collagenase 3 (matrix metalloproteinase 13) production by human articular chondrocytes. Arthritis Rheumatol. 2002;46(9):2368–2376. [DOI] [PubMed] [Google Scholar]
  17. Francis-West PH, Parish J, Lee K, Archer CW.. BMP/GDF-signalling interactions during synovial joint development. Cell Tissue Res. 1999;296(1):111–119. [DOI] [PubMed] [Google Scholar]
  18. Gagat M, Zielińska W, Mikołajczyk K, Zabrzyński J, Krajewski A, Klimaszewska-Wiśniewska A, Grzanka D, Grzanka A.. CRISPR-based activation of endogenous expression of TPM1 inhibits inflammatory response of primary human coronary artery endothelial and smooth muscle cells induced by recombinant human tumor necrosis factor α. Front Cell Dev Biol. 2021;9:668032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gao X, Sun Y, Li X.. Identification of key gene modules and transcription factors for human osteoarthritis by weighted gene co-expression network analysis. Exp Ther Med. 2019;18(4):2479–2490. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gupta S, Stamatoyannopoulos JA, Bailey TL, Noble WS.. Quantifying similarity between motifs. Genome Biol. 2007;8(2):R24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Homandberg GA, Wen C, Hui F.. Cartilage damaging activities of fibronectin fragments derived from cartilage and synovial fluid. Osteoarthr Cartil. 1998;6(4):231–244. [DOI] [PubMed] [Google Scholar]
  22. Homandberg GA. Potential regulation of cartilage metabolism in osteoarthritis by fibronectin fragments. Front Biosci. 1999;4: d 713–30. [DOI] [PubMed] [Google Scholar]
  23. Hunter DJ, Bierma-Zeinstra S.. Osteoarthritis. Lancet. 2019;393(10182):1745–1759. [DOI] [PubMed] [Google Scholar]
  24. Jain S, Gautam V, Naseem S.. Acute-phase proteins: as diagnostic tool. J Pharm Bioallied Sci. 2011;3(1):118–127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kiel DP, Kemp JP, Rivadeneira F, Westendorf JJ, Karasik D, Duncan EL, Imai Y, Müller R, Flannick J, Bonewald L, et al. The musculoskeletal knowledge portal: making omics data useful to the broader scientific community. J Bone Miner Res. 2020;35(9):1626–1633. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kirkby NS, Lundberg MH, Wright WR, Warner TD, Paul-Clark MJ, Mitchell JA.. COX-2 protects against atherosclerosis independently of local vascular prostacyclin: identification of COX-2 associated pathways implicate Rgl1 and lymphocyte networks. PLoS One. 2014;9(6):e98165. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Knight PA, Ruiz D.. A fast algorithm for matrix balancing. IMA J Numer Anal. 2013;33(3):1029–1047. [Google Scholar]
  28. Kramer NE, Davis ES, Wenger CD, Deoudes EM, Parker SM, Love MI, Phanstiel DH.. Plotgardener: cultivating precise multi-panel figures in R. Bioinformatics. 2022;38(7):2042–2045. 10.1093/bioinformatics/btac057 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Laarman MD, Geeven G, Barnett P, Rinkel GJE, de Laat W, Ruigrok YM, Bakkers J; Netherlands Brain Bank. Chromatin conformation links putative enhancers in intracranial aneurysm–associated regions to potential candidate genes. J Am Heart Assoc. 2019;8(9):e011201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Li X, Ellman MB, Kroin JS, Chen D, Yan D, Mikecz K, Ranjan KC, Xiao G, Stein GS, Kim S-G, et al. Species-specific biological effects of FGF-2 in articular cartilage: implication for distinct roles within the FGF receptor family. J Cell Biochem. 2012;113(7):2532–2542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Li R, Liang Y, Lin B.. Accumulation of systematic TPM1 mediates inflammation and neuronal remodeling by phosphorylating PKA and regulating the FABP5/NF-κB signaling pathway in the retina of aged mice. Aging Cell. 2022;21(3):e13566. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lin AC, Seeto BL, Bartoszko JM, Khoury MA, Whetstone H, Ho L, Hsu C, Ali SA, Ali AS, Alman BA, et al. Modulating hedgehog signaling can attenuate the severity of osteoarthritis. Nat Med. 2009;15(12):1421–1425. [DOI] [PubMed] [Google Scholar]
  33. Lin D, Sanders J, Noble WS.. HiCRep.py: fast comparison of Hi-C contact matrices in Python. Bioinformatics. 2021;37(18):2996–2997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Loeser RF, Pacione CA, Chubinskaya S.. The combination of insulin-like growth factor 1 and osteogenic protein 1 promotes increased survival of and matrix synthesis by normal and osteoarthritic human articular chondrocytes. Arthritis Rheumatol. 2003;48(8):2188–2196. [DOI] [PubMed] [Google Scholar]
  35. Loeser RF, Goldring SR, Scanzello CR, Goldring MB.. Osteoarthritis: a disease of the joint as an organ. Arthritis Rheumatol. 2012;64(6):1697–1707. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Loeser RF. Integrins and chondrocyte-matrix interactions in articular cartilage. Matrix Biol. 2014;39:11–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Maurano MT, Humbert R, Rynes E, Thurman RE, Haugen E, Wang H, Reynolds AP, Sandstrom R, Qu H, Brody J, et al. Systematic localization of common disease-associated variation in regulatory DNA. Science. 2012;337(6099):1190–1195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Metcalf D, Greenhalgh CJ, Viney E, Willson TA, Starr R, Nicola NA, Hilton DJ, Alexander WS.. Gigantism in mice lacking suppressor of cytokine signalling-2. Nature. 2000;405(6790):1069–1073. [DOI] [PubMed] [Google Scholar]
  39. Miyamoto Y, Mabuchi A, Shi D, Kubo T, Takatori Y, Saito S, Fujioka M, Sudo A, Uchida A, Yamamoto S, et al. A functional polymorphism in the 5’ UTR of GDF5 is associated with susceptibility to osteoarthritis. Nat Genet. 2007;39(4):529–533. [DOI] [PubMed] [Google Scholar]
  40. Monti-Rocha R, Cramer A, Gaio Leite P, Antunes MM, Pereira RVS, Barroso A, Queiroz-Junior CM, David BA, Teixeira MM, Menezes GB, et al. SOCS2 is critical for the balancing of immune response and oxidate stress protecting against acetaminophen-induced acute liver injury. Front Immunol. 2018;9:3134. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Morgan M, Carlson M, Tenenbaum D, Arora S. AnnotationHub: Client to Access AnnotationHub Resources. R package version 2; 2017.
  42. Motomura H, Seki S, Shiozawa S, Aikawa Y, Nogami M, Kimura T.. A selective c-Fos/AP-1 inhibitor prevents cartilage destruction and subsequent osteophyte formation. Biochem Biophys Res Commun. 2018;497(2):756–761. [DOI] [PubMed] [Google Scholar]
  43. Muthuirulan P, Zhao D, Young M, Richard D, Liu Z, Emami A, Portilla G, Hosseinzadeh S, Cao J, Maridas D, et al. Joint disease-specificity at the regulatory base-pair level. Nat Commun. 2021;12(1):4161. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Nishizaki SS, Ng N, Dong S, Porter RS, Morterud C, Williams C, Asman C, Switzenberg JA, Boyle AP.. Predicting the effects of SNPs on transcription factor binding affinity. Bioinformatics. 2020;36(2):364–372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Okura T, Matsushita M, Mishima K, Esaki R, Seki T, Ishiguro N, Kitoh H.. Activated FGFR3 prevents subchondral bone sclerosis during the development of osteoarthritis in transgenic mice with achondroplasia. J Orthop Res. 2018;36(1):300–308. [DOI] [PubMed] [Google Scholar]
  46. Pagès H, Aboyoun P, Gentleman R, DebRoy S. Biostrings: Efficient Manipulation of Biological Strings; 2021. DOI: 10.18129/B9.bioc.Biostrings. [DOI]
  47. Paul I, Batth TS, Iglesias-Gato D, Al-Araimi A, Al-Haddabi I, Alkharusi A, Norstedt G, Olsen JV, Zadjali F, Flores-Morales A, et al. The ubiquitin ligase Cullin5SOCS2 regulates NDR1/STK38 stability and NF-κB transactivation. Sci Rep. 2017;7(42800):42800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Pelletier JP, Martel-Pelletier J, Abramson SB.. Osteoarthritis, an inflammatory disease: potential implication for the selection of new therapeutic targets. Arthritis Rheumatol. 2001;44(6):1237–1247. [DOI] [PubMed] [Google Scholar]
  49. Pulai JI, Chen H, Im H-J, Kumar S, Hanning C, Hegde PS, Loeser RF.. NF-κB mediates the stimulation of cytokine and chemokine expression by human articular chondrocytes in response to fibronectin fragments. J Immunol. 2005;174(9):5781–5788. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, Maller J, Sklar P, de Bakker PIW, Daly MJ, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81(3):559–575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Quinlan AR, Hall IM.. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841–842. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Rao SSP, Huntley MH, Durand NC, Stamenova EK, Bochkov ID, Robinson JT, Sanborn AL, Machol I, Omer AD, Lander ES, et al. A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell. 2014;159(7):1665–1680. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. R Core Team. R: A Language and Environment for Statistical Computing; 2021.
  54. Reed KSM, Ulici V, Kim C, Chubinskaya S, Loeser RF, Phanstiel DH.. Transcriptional response of human articular chondrocytes treated with fibronectin fragments: an in vitro model of the osteoarthritis phenotype. Osteoarthr Cartil. 2021;29(2):235–247. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Reynard LN, Barter MJ.. Osteoarthritis year in review 2019: genetics, genomics and epigenetics. Osteoarthritis Cartilage. 2020;28(3):275–284. [DOI] [PubMed] [Google Scholar]
  56. Richard D, Liu Z, Cao J, Kiapour AM, Willen J, Yarlagadda S, Jagoda E, Kolachalama VB, Sieker JT, Chang GH, et al. Evolutionary selection and constraint on human knee chondrocyte regulation impacts osteoarthritis risk. Cell. 2020;181(2):362–381.e28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Rowley MJ, Poulet A, Nichols MH, Bixler BJ, Sanborn AL, Brouhard EA, Hermetz K, Linsenbaum H, Csankovszki G, Lieberman Aiden E, et al. Analysis of Hi-C data using SIP effectively identifies loops in organisms from C. elegans to mammals. Genome Res. 2020;30(3):447–458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Samvelyan HJ, Huesa C, Cui L, Farquharson C, Staines KA.. The role of accelerated growth plate fusion in the absence of SOCS2 on osteoarthritis vulnerability. Bone Joint Res. 2022;11(3):162–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Sandell LJ, Aigner T.. Articular cartilage and changes in arthritis: cell biology of osteoarthritis. Arthritis Res. Ther. 2001;3:107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Santangelo C, Scipioni A, Marselli L, Marchetti P, Dotta F.. Suppressor of cytokine signaling gene expression in human pancreatic islets: modulation by cytokines. Eur J Endocrinol. 2005;152(3):485–489. [DOI] [PubMed] [Google Scholar]
  61. Schmidt EM, Zhang J, Zhou W, Chen J, Mohlke KL, Chen YE, Willer CJ.. GREGOR: evaluating global enrichment of trait-associated variants in epigenomic features using a systematic, data-driven approach. Bioinformatics. 2015;31(16):2601–2606. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Shu CC, Jackson MT, Smith MM, Smith SM, Penm S, Lord MS, Whitelock JM, Little CB, Melrose J.. Ablation of perlecan domain 1 heparan sulfate reduces progressive cartilage degradation, synovitis, and osteophyte size in a preclinical model of posttraumatic osteoarthritis. Arthritis Rheumatol. 2016;68(4):868–879. [DOI] [PubMed] [Google Scholar]
  63. Skene PJ, Henikoff S.. An efficient targeted nuclease strategy for high-resolution mapping of DNA binding sites. eLife. 2017;6:e21856. 10.7554/eLife.21856 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Song J, Kim D, Lee CH, Lee MS, Chun C-H, Jin E-J.. MicroRNA-488 regulates zinc transporter SLC39A8/ZIP8 during pathogenesis of osteoarthritis. J Biomed Sci. 2013;20(31):31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Southam L, Rodriguez-Lopez J, Wilkins JM, Pombo-Suarez M, Snelling S, Gomez-Reino JJ, Chapman K, Gonzalez A, Loughlin J.. An SNP in the 5’-UTR of GDF5 is associated with osteoarthritis susceptibility in Europeans and with in vivo differences in allelic expression in articular cartilage. Hum Mol Genet. 2007;16(18):2226–2232. [DOI] [PubMed] [Google Scholar]
  66. Starr R, Willson TA, Viney EM, Murray LJ, Rayner JR, Jenkins BJ, Gonda TJ, Alexander WS, Metcalf D, Nicola NA, et al. A family of cytokine-inducible inhibitors of signalling. Nature. 1997;387(6636):917–921. [DOI] [PubMed] [Google Scholar]
  67. Steinberg J, Southam L, Butterfield NC, Roumeliotis TI, Fontalis A, Clark MJ, Jayasuriya RL, Swift D, Shah KM, Curry KF, et al. Decoding the genomic basis of osteoarthritis. bioRxiv 835850, 2020.
  68. Steinberg J, Southam L, Roumeliotis TI, Clark MJ, Jayasuriya RL, Swift D, Shah KM, Butterfield NC, Brooks RA, McCaskie AW, et al. A molecular quantitative trait locus map for osteoarthritis. Nat Commun. 2021;12(1):1309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Tang J, Su N, Zhou S, Xie Y, Huang J, Wen X, Wang Z, Wang Q, Xu W, Du X, et al. Fibroblast growth factor receptor 3 inhibits osteoarthritis progression in the knee joints of adult mice. Arthritis Rheumatol. 2016;68(10):2432–2443. [DOI] [PubMed] [Google Scholar]
  70. Tuerlings M, van Hoolwerff M, Houtman E, Suchiman EHED, Lakenberg N, Mei H, van der Linden EHMJ, Nelissen RRGHH, Ramos YYFM, Coutinho de Almeida R, et al. RNA sequencing reveals interacting key determinants of osteoarthritis acting in subchondral bone and articular cartilage: identification of IL11 and CHADL as attractive treatment targets. Arthritis Rheumatol. 2021;73(5):789–799. [DOI] [PMC free article] [PubMed] [Google Scholar]
  71. Umans BD, Battle A, Gilad Y.. Where are the disease-associated eQTLs? Trends Genet. 2021;37(2):109–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Venetianer P. Are synonymous codons indeed synonymous? Biomol Concepts. 2012;3(1):21–28. [DOI] [PubMed] [Google Scholar]
  73. Wang T, He C.. Pro-inflammatory cytokines: the link between obesity and osteoarthritis. Cytokine Growth Factor Rev. 2018;44:38–50. [DOI] [PubMed] [Google Scholar]
  74. Wanner J, Subbaiah R, Skomorovska-Prokvolit Y, Shishani Y, Boilard E, Mohan S, Gillespie R, Miyagi M, Gobezie R.. Proteomic profiling and functional characterization of early and late shoulder osteoarthritis. Arthritis Res Ther. 2013;15(6):R180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Wickham H. ggplot2: Elegant Graphics for Data Analysis; 2016.
  76. Wojdasiewicz P, Poniatowski ŁA, Szukiewicz D.. The role of inflammatory and anti-inflammatory cytokines in the pathogenesis of osteoarthritis. Mediators Inflamm. 2014;2014:561459. [DOI] [PMC free article] [PubMed] [Google Scholar]
  77. Won H, de la Torre-Ubieta L, Stein JL, Parikshak NN, Huang J, Opland CK, Gandal MJ, Sutton GJ, Hormozdiari F, Lu D, et al. Chromosome conformation elucidates regulatory relationships in developing human brain. Nature. 2016;538(7626):523–527. [DOI] [PMC free article] [PubMed] [Google Scholar]
  78. Wood ST, Long DL, Reisz JA, Yammani RR, Burke EA, Klomsiri C, Poole LB, Furdui CM, Loeser RF.. Cysteine-mediated redox regulation of cell signaling in chondrocytes stimulated with fibronectin fragments. Arthritis Rheumatol. 2016;68(1):117–126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  79. Xie DL, Meyers R, Homandberg GA.. Fibronectin fragments in osteoarthritic synovial fluid. J. Rheumatol. 1992;19:1448–1452. [PubMed] [Google Scholar]
  80. Yang T, Zhang F, Yardımcı GG, Song F, Hardison RC, Noble WS, Yue F, Li Q.. HiCRep: assessing the reproducibility of Hi-C data using a stratum-adjusted correlation coefficient. Genome Res. 2017;27(11):1939–1949. [DOI] [PMC free article] [PubMed] [Google Scholar]
  81. Yang Y, Shen P, Yao T, Ma J, Chen Z, Zhu J, Gong Z, Shen S, Fang X.. Novel role of circRSU1 in the progression of osteoarthritis by adjusting oxidative stress. Theranostics. 2021;11(4):1877–1900. [DOI] [PMC free article] [PubMed] [Google Scholar]
  82. Zeng Z, Bromberg Y.. Predicting functional effects of synonymous variants: a systematic review and perspectives. Front Genet. 2019;10:914. [DOI] [PMC free article] [PubMed] [Google Scholar]
  83. Zhou S, Xie Y, Li W, Huang J, Wang Z, Tang J, Xu W, Sun X, Tan Q, Huang S, et al. Conditional deletion of Fgfr3 in chondrocytes leads to osteoarthritis-like defects in temporomandibular joint of adult mice. Sci Rep. 2016;6:24039. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

iyac141_Supplementary_Figure_Legends
Click here for additional data file. (7.5KB, docx)
iyac141_Supplementary_Figure_S1
Click here for additional data file. (15.7KB, pdf)
iyac141_Supplementary_Figure_S2
Click here for additional data file. (14.4KB, pdf)
iyac141_Supplementary_Figure_S3
Click here for additional data file. (1.4MB, pdf)
iyac141_Supplementary_Figure_S4
Click here for additional data file. (4.7KB, pdf)
iyac141_Supplementary_Figure_S5
Click here for additional data file. (3MB, pdf)
iyac141_Supplementary_Tables_1_3
Click here for additional data file. (73.2KB, xlsx)

Data Availability Statement

Hi-C and CUT&RUN data can be accessed through GEO accession GSE200345 (see Datasets for publicly available datasets; OA GWAS, Epigenome Roadmap Data, and RNA-seq time course of FN-f treatment).

Supplemental material is available at GENETICS online.

Datasets

OA GWAS

Genome-wide association statistics for 11 OA phenotypes and lead variants identified in Boer et al. (2021) were obtained from the Musculoskeletal Knowledge Portal (Kiel et al. 2020).

Epigenome Roadmap Data

Consolidated reference human epigenomes for 98 cell/tissue types were obtained from the NIH Roadmap Epigenomics Project (Bernstein et al. 2010) and The Encyclopedia of DNA Elements (ENCODE) project (ENCODE Project Consortium 2012). Processed narrowPeak files for H3K27ac, H3K4me1, and H3K4me3 and BigWig files for H3K27ac were used for each cell/tissue type. Additional narrowPeak files for H3K9ac, H3K9me3, H3K27me3, and H3K36me3 were obtained for mesenchymal stem cell-derived chondrocyte-cultured cells (E049).

RNA-seq time course of FN-f treatment

RNA-seq data from a prior study of FN-f treated human chondrocytes was obtained from KSM Reed et al. (Reed et al. 2021) and vst-normalized, centered, and replicate-combined. The 0-h FN-f treatment time point was created by combining the 9 PBS-treated replicates. Genes were considered differential with a BH-adjusted P-value of 0.01 and a log2 fold-change threshold >1.25 across any time point.

Articles from Genetics are provided here courtesy of Oxford University Press

RESOURCES