Variant to gene mapping for carpal tunnel syndrome risk loci implicates skeletal muscle regulatory elements

Summary

Background

Carpal tunnel syndrome (CTS) is a common disorder caused by compression of the median nerve in the wrist, resulting in pain and numbness throughout the hand and forearm. While multiple behavioural and physiological factors influence CTS risk, a growing body of evidence supports a strong genetic contribution. Recent genome-wide association study (GWAS) efforts have reported 53 independent signals associated with CTS. While GWAS can identify genetic loci conferring risk, it does not determine which cell types drive the genetic aetiology of the trait, which variants are “causal” at a given signal, and which effector genes correspond to these non-coding variants. These obstacles limit interpretation of potential disease mechanisms.

Methods

We analysed CTS GWAS findings in the context of chromatin conformation between gene promoters and accessible chromatin regions across cellular models of bone, skeletal muscle, adipocytes and neurons. We identified proxy variants in high LD with the lead CTS sentinel SNPs residing in promoter connected open chromatin in the skeletal muscle and bone contexts.

Findings

We detected significant enrichment for heritability in skeletal muscle myotubes, as well as a weaker correlation in human mesenchymal stem cell-derived osteoblasts. In myotubes, our approach implicated 117 genes contacting 60 proxy variants corresponding to 20 of the 53 GWAS signals. In the osteoblast context we implicated 30 genes contacting 24 proxy variants coinciding with 12 signals, of which 19 genes shared. We subsequently prioritized BZW2 as a candidate effector gene in CTS and implicated it as novel gene that perturbs myocyte differentiation in vitro.

Interpretation

Taken together our results suggest that the CTS genetic component influences the size, integrity, and organization of multiple tissues surrounding the carpal tunnel, in particular muscle and bone, to predispose the nerve to being compressed in this disease setting.

Funding

This work was supported by NIH Grant UM1 DK126194 (SFAG and WY), R01AG072705 (SFAG & KDH) and the Center for Spatial and Functional Genomics at CHOP (SFAG & ADW). SFAG is supported by the Daniel B. Burke Endowed Chair for Diabetes Research. WY is supported by the Perelman School of Medicine of the University of Pennsylvania.

Research in context.

Evidence before this study

We searched PubMed for the terms Carpal Tunnel GWAS to identify the most recent genome wide association study. The GWAS identified 53 genetic loci associated with risk for Carpal Tunnel Syndrome. However, a limitation of GWAS is that the method does not directly identify the precise genetic variants or effector genes responsible for increased disease risk. These loci often comprise large blocks of linked variants, with only a subset likely causally relevant to disease risk. In addition to this, many of these signals are in non-coding regions of the genome, which limits identification of the relevant effector genes for these genetic variants. Multiple approaches to nominate effector genes, such as associating expression at the gene expression or protein level with disease risk variants, have been employed.

Added value of this Study

Our study utilizes genome-wide epigenetic and chromatin conformational state data for a set of relevant cell types. We observed enrichment of carpal tunnel associated heritability in the promoter-connected regulatory elements of skeletal muscle myotubes and osteoblasts, suggesting these cells are relevant. We intersected the putative cis-regulatory elements with the variants associated with Carpal Tunnel Syndrome, to identify the subset that most likely impact gene regulatory activity and the genes they are likely to regulate. From this approach, we identified putative effector genes for 20 of the 53 GWAS signals using a combination of epigenetic and chromatin conformation datasets. Additionally, we identified a novel function for one gene in influencing muscle cell differentiation in vitro.

Implications of all the available evidence

Our data identified a potentially underappreciated role for skeletal muscle in the risk for carpal tunnel syndrome. Characterizing the effector genes associated with Carpal Tunnel Syndrome risk can provide valuable information for developing therapeutic interventions, whether through drug repurposing or target prioritization, aimed at treating patients or preventing development of the disease in at-risk individuals.

Introduction

Carpal tunnel syndrome (CTS) is a common chronic disorder caused by compression of the median nerve in the wrist, resulting in pain and numbness throughout the hand.¹ In the United States, carpal tunnel affects 3–6% of adults, and risk factors include a known heritable component.² The median nerve is located within the synovial tendinous sheath, which surrounds several tendons and muscles. Current treatments for CTS include surgical intervention and behaviour modification with pharmaceuticals for reducing symptoms.² While treatments such as wrist stretches and anti-inflammatory drugs can reduce mild symptoms, severe CTS often requires surgical intervention, with approximately 600,000 carpal tunnel releases per year in the United States, which is achieved by cutting the carpal ligament to relieve pressure on the medial nerve.³ Identifying genes associated with this disorder should aid understanding of CTS aetiology and provide additional avenues for treatment and prevention to reduce the need for surgical intervention.

A recent large-scale GWAS identified 53 signals associated with CTS from a large European cohort with 48,843 cases and 1,190,837 controls derived from Iceland, the United Kingdom, Denmark, and Finland.⁴ Estimated CTS polygenetic signal has been reported to be highly correlated with traits such as BMI, height, osteoarthritis, and restlessness,^4,5 suggesting a role for anthropomorphic traits as contributors to CTS genetic risk.

Multiple cell and tissue types are relevant to CTS: the medial neuron, along with the cartilage, tendons, bone, and muscle surrounding it becomes compressed leading to ischemia and axonal degeneration. In addition, as obesity is a risk factor, tissues relevant to metabolism also likely contribute to disease pathology. Identification of the underlying cells/tissues of action for at least a proportion of the lead variants is an important step in resolving their molecular mechanisms underlying diseased risk.

The vast majority of GWAS variants reside in non-coding regions of the genome. These variants can occupy cis-regulatory elements (cREs) that regulate their target genes from up to megabases away by long-distance chromatin folding. One approach to identify such ‘effector’ genes is use expression quantitative trait loci (eQTLs). However, large tissue collections can lack data from relevant cell or tissue types. Another approach is to assess chromatin conformation data generated from methods such as Hi-C performed in specific cellular settings. Chromatin conformation assays represent an adaptable approach to implicate target effector genes at GWAS loci. For example, the BMI associated FTO locus was shown to contact IRX3 and IRX5 rather than the closest gene for which the locus was annotated.^6,7 These approaches can therefore implicate candidate effector genes in specific cellular contexts.

Building on previous experience with such ‘variant-to-gene mapping’ campaigns for other common complex traits, we sought to use epigenetic and chromatin conformation data to connect non-coding CTS associated variants with their potential effector genes by integrating GWAS data with a combination of ATAC-seq, chromatin conformation capture and RNA-seq. We identified skeletal muscle myotubes and osteoblasts as cell types of interest through via partitioned LD score regression.⁸ Motivated by these results, we implicated effector genes by identifying promoters in physical contact with cREs harbouring CTS-associated putative causal variants in these given cellular contexts.

Methods

Partitioned LD score regression enrichment analyses

Partitioned heritability LD score regression (v1.0.0)⁸ was used to identify enrichment of GWAS summary statistics among open accessible regions identified in each cell type. The analysis was performed using the baseline LD score data⁹ (https://data.broadinstitute.org/alkesgroup/LDSCORE) and summary statistics from carpal tunnel syndrome.⁴ The celltypes we considered were skeletal muscle myotubes, human mesenchymal stem cell derived adipocytes and osteoblasts, iPSC neuronal cells, and human fetal osteoblast cell line (hFOB) either undifferentiated or differentiated. As the minor allele frequencies for the total population were not provided, we estimated the minor allele frequency across the by taking the weighted mean of each allele of each cohort (the sample size in specific population/sample size in the entire cohort). We generated annotations for each cell type using the coordinates of the promoter-interacting open chromatin regions.

Variant to gene mapping analyses

We queried for proxy SNPs in high LD (R2 > 0.8) from the European ancestry sample using LDLinkR,¹⁰ with the default ± 500,000 bp window from the sentinel variant. We next intersected the carpal tunnel sentinel and proxy variants with the set of OCRs annotated to promoter regions (−1500/+500 bp of TSS) and OCR overlapping promoter interacting regions identified by Capture C/HiC. Genomic coordinate overlaps were identified using the R package GenomicRanges (ver 1.42)⁹ using the hg19 reference genome. These implicated variants were intersected with ENCODE regions called for enriched histone mark and TF binding. The resulting number of proxies were also intersected with skeletal muscle eQTLs identified from the FUSION skeletal muscle tissue collection.¹¹ Putative causal variants were intersected with 95% credible sets identified using ` with the 1000G Phase3 as the LD reference with the parameter outlier.switch = TRUE.¹²

GO term enrichment

GO biological process (c5.bp) datasets annotated in MSigDB (v7.0) were used for gene set enrichment analyses. Prior to enrichment testing, the background was limited to genes within a 5 megabase window around the CTS GWAS sentinels. Statistical significance of gene set enrichment was determined using the hypergeometric test, implemented in the R phyper function with lower.tail = FALSE. P values were adjusted using the R function p.adjust with the parameter, method = “BH”.

GTEX expression comparison

Using the GTeX v8 median expression values (https://gtexportal.org/home/datasets), we calculated the specificity of the implicated target genes by taking the scaled (mean centered at 0) transform of the ratio of skeletal muscle TPM value over the sum of the remaining tissues.

Transcription factor motif analysis

Transcription factor binding site motifs overlapping with proxy variants implicated in by variant to gene mapping analysis were identified using the R package motifbreakR (v2.0.0) using the HOCOMOCO v11 database as our reference set of position weight matrices.¹³

Cell culture

Human skeletal myoblast cell line LHCN-M2 was purchased from Evercyte (Austria). They were maintained and differentiated to myotubes according to Evercyte’s instructions. Briefly, cells were maintained in tissue culture flasks coated with 0.1% gelatin in Maintenance Media consisting of 4:1 DME/M199, 20 mM Hepes, 3 μg/ml ZnSO4, and 1.4 μg/ml Vitamin B12 with 15% fetal bovine serum (FBS), 55 ng/ml dexamethasone, 2.5 ng/ml hepatocyte growth factor (Millipore, Cat# GF116), 10 ng/ml bFGF (R&D Systems, Cat# 3718-FB-100), 50 U/ml penicillin, 50 μg/ml streptomycin, and 625 ng/ml fungizone. Maintenance Media were changed every two days. For differentiation, cells grown to 90% confluence were washed with phosphate-buffered saline and re-fed with low glucose DMEM supplemented with 2% horse-serum. Media were refreshed every other day until day 6.

Plasmid construction

Guide RNAs (gRNAs) were designed using the online tool (https://www.benchling.com/crispr) and two oligonucleotides, 5′-CACCG-gRNA sequence-3′ and 5′-CAAA- (reverse compliment of gRNA)-C-3′ were synthesized for cloning. Cloning primer sequences are listed in Supplemental Table S6. Oligonucleotides were synthesized by Integrated DNA Technologies (IDT) and were annealed and ligated into lentiCRISPR v2 (Addgene Cat #52961) plasmid after BsmBI digestion. Lentiviral plasmid lentiCRISPR v2-sgControl (Cat #125836) used as sgCtrl1 and a mixture of LentiCRISPRv2_sgCtrl1 (Cat #174148) and LentiCRISPRv2_sgCtrl2 (Cat #174149) used as sgCtrl2 were purchased from Addgene. Two gRNA sequences targeting BZW2 were used to construct separate lentiviral plasmids and the plasmids were mixed together before the mixture was used for viral packaging. Primers rs2240855-Δ-sg1, rs2240855-Δ-sg2 and rs2240855-Δ-sg3, rs2240855-Δ-sg4 were used to create Δrs2240855line1 and rs2240855line2 respectively.

Lentivirus generation and CRISPR/Cas9-based targeting of LHCN-M2 cells

Lentiviruses were packaged with the third-generation lentivirus system. Lentiviral packaging helper plasmids pHDM-G (Addgene, cat #164441), pHDM-Hgmp2 (Addgene, cat #164441), pHDM-tat1b (Addgene, cat #164441), and RC/CMV-rev1b (Addgene, cat #164441) were used. HEK293T cells were seed in a 6 cm dish at a density of 5 × 10⁶ cells the day before transfection. Five μg of the lentiviral backbone plasmid and 2.6 μg RC/CMV-rev1b, 0.5 μg pHDM-tat1b, 1 μg pHDM-tat1b, and 2 μg pHDM-Hgmp2 were co-transfected into HEK293T cells the Lipofectamine™ 3000 Transfection Reagent (Invitrogen, Cat#L3000015). The media were changed after 12 h and then were collected after 48 h. Following centrifugation at 400 g for 4 min, the virus-containing supernatant was passed through a 0.45 μm filter The viruses were further precipitated with PEG-it Virus Precipitation Solution (System Biosciences, Cat# LV810A-1) and dissolved in 200 μl DMEM. Myoblasts were infected with 20 μl resuspended viruses in the presence of 8 mg/ml of polybrene and selected with 2 mg/ml puromycin (Sigma Aldrich, cat P8833) for three days to generate stable Cas9-sgRNA-expressing LHCN-M2 myoblast cell lines.

Immunostaining of myotubes

Immunostaining of human muscle myotubes was performed as described previously.¹⁴ Myoblasts were differentiated into myotubes for 6 days, and then were fixed with 4% paraformaldehyde (PFA). Myotubes were treated with 0.1% Triton X-100 in PBS for 10 min. After blocking with 2% BSA in PBS for 30 min, the expression of pan myosin heavy chain (MHC) in myotubes was detected with the MF-20 antibody (Developmental Studies Hybridoma Bank) followed by secondary antibody (Alexa Fluor 561 goat anti–mouse IgG1 1:1000; Invitrogen). Nuclei were counterstained with DAPI (Thermo Fisher Scientific). Representative images of cells were taken on a Keyence inverted fluorescence microscope (BZX-710).

RNA analyses

Total RNA was extracted from myoblasts and myotubes using TRIzol™ Reagent (Thermo Fisher Scientific). 1 μg total RNA samples were then reverse transcribed with the high-capacity cDNA reverse transcription kit with RNase inhibitor (Applied Biosystems™) using random hexamer primers according to the manufacturer’s instructions. Real-time quantitative PCR was performed using the SYBR™ Green PCR Master Mix (Applied Biosystems). Specific oligonucleotide primers for target gene sequences are listed in Supplemental Table S6. Gene expression was normalized to TBP.

Statistics

All cell studies were analysed by Student’s t test (2-tailed). No statistical methods were used to pre-determine sample sizes. Data represent the mean ± SEM, with a statistically significant difference defined as a value of P < 0.05.

Ethics

Research was conducted in accordance with the Children’s Hospital of Philadelphia’s ethics and compliance policies. The researchers had access to deidentified biospecimens and datasets.

Role of funding source

The funding source had no role in data collection, analysis, interpretation, writing or decision to submit the manuscript. No authors were paid to write this article by any pharmaceutical company or other entity. Authors were not precluded from accessing data in the study, and they accept responsibility to submit for publication.

Results

Identification of cell types enriched for carpal tunnel syndrome heritability

We sought to investigate a potential role for neural, skeletal, and metabolic tissue types in the pathogenesis of CTS in the context of cREs integrated with GWAS findings. Our group has generated profiles for chromatin accessibility and chromatin conformation in disease relevant cell types, which we employed for this variant to gene mapping approach15, 16, 17, 18, 19 (Fig. 1a).

Fig. 1.

Integration of GWAS empowers identifying of key tissue types and effector genes. (a) Schematic of the variant to gene mapping strategy. After generation of chromatin contact and accessibility atlases, we intersected putative regulatory elements with proxy variants in high LD (R² > 0.8, ± 500,000 bp from the sentinel variant) with significant GWAS signals. (b) Results from LD score regression of hFOB bone cell line (prediff and postdiff), hMSC osteoblasts, hMSC Adipocyte, primary skeletal muscle myotubes. Circle size indicates the heritability enrichment of each cell type’s cRE annotation (proportion heritability/proportion of SNPs in the annotation). The color indicates the degree of significance zscore calculated from P-value. Yellow indicates more significant enrichment; Dots below are the scale.

To explore cell types with common GWAS heritability associated with CTS risk, we performed partitioned LD score regression using the set of cREs coinciding with chromatin contacts to gene promoters in skeletal muscle myotubes (Myotubes), induced pluripotent stem cell (iPSC) derived neurons,¹⁸ pre and post differentiated fetal osteoblastic cells (hFOBs) human mesenchymal stem cell (hMSC) derived osteoblasts¹⁶ and hMSC derived adipocytes¹⁷ (Supplemental Table S1). We observed significant enrichment for GWAS heritability in the skeletal muscle myotubes (FDR = 0.0027) and hMSC osteoblasts (FDR = 0.046) (Fig. 1b), suggesting that hMSC-derived osteoblasts and skeletal muscle myotubes are cell types relevant for further exploration of putative effector genes with our chromatin epigenetic and chromatin accessibility data. Our observation is consistent with prior observations that identified enriched CTS heritability in osteoblasts and smooth muscle myocytes using regions near specifically expressed genes across a range of cell types.⁵

Variant to gene mapping for CTS risk loci implicates effector genes in myotubes and osteoblasts

Motivated by our partitioned LD score enrichment results, we elected to query the CTS GWAS-implicated signals for mapping effector genes using our chromatin accessibility and chromatin conformation data (Fig. 2a). First, we identified proxy SNPs as the variants in high LD (1000G EUR, r² > 0.8) for the 53 CTS sentinel variants representing the independent GWAS signals.^10,20 Next, we intersected the resulting 1501 SNP list with open chromatin regions, identified using ATAC-seq, that contacted gene promoters, identified with Hi-C or promoter-focused Capture C, in myotubes and hMSC-derived osteoblasts. In the skeletal muscle myotubes, we identified at least one CTS-associated variant contacting at least one gene for 20 of the 53 CTS signals, with a total of 60 variants in contact with 115 genes.⁴ In the hMSC-derived osteoblast dataset we identified at least one gene implicated for 12 of the signals, with a total of 24 variants contacting 30 gene promoters (Fig. 2a and b and Supplemental Table S2). We observed a median of 4 genes implicated per signal. Only 19 genes corresponding to 9 signals were consistently implicated by both the myotubes and hMSC-derived osteoblast datasets, with 96 genes specifically called in myotubes and 11 genes specific to osteoblasts (Fig. 2c and Supplemental Table S3). We also compared our dataset to a collection of RNA-seq data derived from CTS cases,⁵ 85 of our implicated genes (69.1%) were expressed (TPM > 1), with 49 of these (39.8%) of these being highly expressed (TPM > 10) in diseased carpal tunnel tissue; however, control tissue was not available to test whether genes expression is specifically altered in CTS (Supplemental Fig. S1). We compared the variants implicated to those within the calculated 95% credible set and observed that 66.18% of the implicated proxy variants were included within a credible set (Supplemental Fig. S2). In addition, we compared our list of implicated genes with eQTLs from bulk RNA-seq from skeletal muscle tissue.¹¹ We observed limited overlap with only 11 genes in agreement between the two approaches (Supplemental Fig. S3).

Fig. 2.

Variant to gene mapping of CTS loci in skeletal muscle and osteoblast models. (a) We identified coding variants based on their predicted annotation. Additionally, we identified non-coding proxy variants that were in open chromatin (putative cREs) that were in contact with gene promoters in either skeletal muscle myotubes or osteoblasts. The number of proxy variants are the number that meet our criteria to be putatively causal (located in a putative cRE that interacts with a promoter), and genes are the number of genes implicated by our approach. (b) Manhattan plot showing the signals with a least one implicated gene from our dataset (highlighted in green). The x-axis depicts genomic position indicated by chromosomes (alternating colors for clarity), the y axis depicts −log10 P value from the GWAS.⁴ (c) Overlap in the number of genes implicated in both skeletal muscle myotubes and osteoblasts (d) Enrichment of GO biological processes, the top 30 terms with more than one gene annotated are shown (the full list of significant terms is available in Supplemental Table S4). (e) The number of implicated proxy variants predicted to disrupt transcription factor (TF) motifs (x-axis) vs the expression (log2 TPM) of the TF (HOCOMOCO v11). (f) Ranked prediction of skeletal muscle expression in GTEX for the implicated genes. BZW2 which was selected for further follow up is highlighted in red. TPM = Transcripts per million mapped reads.

We performed enrichment analyses for Gene Ontology (GO) terms on the lists of myotube and osteoblast implicated genes to test if any functionally related group of genes were overrepresented. We observed enrichment for pathways relevant to metabolic processes and development (Fig. 2d and Supplemental Table S4): Glucose metabolic processes (FDR = 0.0063958 Osteoblast), hyaluronan metabolic processes (FDR = 1.01 × 10⁻⁴ Myotube; 4.66 × 10⁻⁴ Osteoblasts), Response to Laminar Fluid Shear Stress (FDR = 0.0293 Myotubes), and Segment Specification (FDR = 0. Myotubes) [one-sided hypergeometric test](Supplemental Table S4). We also observed many more general terms (Supplemental Table S4).

As non-coding variants can affect gene expression by disrupting cREs and altering affinity of transcription factor (TF) binding sites,²¹ we predicted the effect of our implicated “causal variants” on known TF motifs (Supplemental Table S5). Relatively few TFs were predicted to be impacted by multiple motifs (Fig. 2e). The notable exception to this pattern was the androgen receptor, which on average exhibited higher affinity motifs associated with CTS variants. We observed that the odds ratio for carpal tunnel of the putative causal variants were weakly correlated with loss of transcription factor affinity (r (1151) −0.1189808, P = 5.119 × 10⁻⁵)[Pearson’s correlation coefficient], suggesting a weak tendency for CTS variants preferentially disrupt, rather than stabilize, these transcription factor binding sites.

Identifying muscle-specific candidate genes

Given that we observed strongest heritability enrichment in skeletal muscle myotubes, we sought to identify muscle-relevant candidate genes for CTS. For this purpose, we used GTEX expression data²² and calculated a ratio of the median gene expression in skeletal muscle compared to all other tissues (Fig. 2f and Supplemental Fig. S4). From this analysis, we implicated several genes with strong enrichment for expression in skeletal muscle including: TNNC1 (Troponin C1), a component of the complex that interfaces between calcium signalling and the actin cytoskeleton during muscle contraction,²³ MKNK2 (MAPK interacting serine/threonine kinase 2) a member of the calmodulin-dependent protein kinases that regulates the eukaryotic initiation factor complex during translation,²⁴ BZW2 (basic leucine zipper and W2 domains 2) a factor important for cadherin binding and an inhibitor of translation that competes with catalytic subunit of the eukaryotic initiation factor,25, 26, 27 MUSTN1 (Musculoskeletal, Embryonic Nuclear Protein 1) a small nuclear protein that is broadly expressed in the musculoskeletal system (osteoblasts, chondroblasts, myoblasts, myocytes, etc) and its expression drives both myogenesis and regeneration of bone fractures,²⁸ WNT9A (WNT Family Member 9A) is known to regulate myogenesis by inhibiting differentiation of satellite cells,²⁹ and ZNF438 (Zinc Finger Protein 438), encoding a transcriptional repressor that responds to androgen and glucocorticoid signalling.^30,31 Bone is not represented in GTeX and thus we were unable to carry out similar analyses for target genes in the hMSC-derived osteoblasts.²² From this approach we therefore prioritized both novel and well-characterized genes likely to confer their effect in muscle.

Loss of BZW2 leads to precocious muscle differentiation in vitro

From the CTS gene list implicated by our variant to gene mapping approach that also showed enriched expression in skeletal muscle, we focused on BZW2 (Fig. 3a). This gene has been previously shown to be expressed in skeletal muscle but has not been functionally characterized in the context of skeletal muscle biology to date. Given its role in regulating translation and cadherin junctions, we hypothesized that variants in BZW2 may contribute to skeletal muscle size and organization in order to increase risk for compression of the carpal tunnel.

Fig. 3.

Analysis of putative effector gene BZW2. (a) Genomic track of BZW2, showing HiC Matrix (top), ATAC-seq (green), from skeletal muscle myocytes. Yellow shows several Chip-seq tracks from different ChIP-seq experiments from ENCODE (H3K4me3; H3K27ac, enhancer mark; H3K27me3, repressive mark; CTCF, TF known to anchor chromatin contacts/topical associated domain boundaries). Regions called as open are shown as OCRs. The location of the sentinel rs1239945 variant (blue) and implicated proxy variant rs2240855 that physically interacts with BZW2 promoter. (b) Phase contrast images and myosin heavy chain staining of day 6 differentiated M2 myotubes. (c–e) qPCR results showing relative expression of BZW2 and skeletal muscle differentiation marker genes in control (sgCtrl1 and sgCtrl2) and BZW2 mutant (sgBZW2) differentiating M2 cells at day 0, 2 and 6.

To assess the functional impact of BZW2 on muscle cell differentiation, we mutated BZW2 in the myoblast cell line LHCN-LHCN-M2 (M2) line using lentivirus mediated CRISPR/Cas9 gene-editing. The expression of BZW2 was reduced by 90% in BZW2 mutated M2 cells (sgBZW2) compared to that in control cells expressing non-targeting guide RNA or guide RNA targeting a safe harbour locus (sgCtrl). We then differentiated the cells into myotubes for 6 days. Morphologically, we did not observe any significant differences between D6 BZW2 mutant and control myotubes (Fig. 3b). Consistently, there are no significant changes in gene expression of markers of skeletal muscle differentiation that we tested, except for 2 contractile myosin protein genes (MYH7 and MYH8). However, when we examined earlier differentiation time points, we observed that myogenic transcription factor genes were upregulated before the initiation of differentiation (D0) and several myosins as well as myotube fusion genes are all increased in BZW2 mutant cells at D2 (Fig. 3c–e).

The intronic regions adjacent to the rs2240855 proxy SNP associated with sentinel variant rs12539945 displayed enrichment of open chromatin by ATAC-Seq and activated transcriptional epigenetic modifications including H3K27ac and H3K4me3. The region also lacks repressive transcriptional epigenetic modification H3K27me3, suggesting rs2240855 is located in a enhancer active in muscle cells. Furthermore, there is strong DNA interaction between this proxy SNP and BZW2 promoter indicated by Hi-C of human myotubes. This led us to hypothesize that the putative causal variant, rs2240855 has a regulatory role in BZW2 gene expression in human muscle cells (Fig. 3a). To study the function of this proxy SNP, we used two independent strategies with lentiviral delivery of CRISPR-Cas9 system to delete a ∼400 bps region spanning the putative causal variant, rs2240855 in human M2 muscle cell line. PCR amplification of the genomic region surrounding rs2240855 in control and CRISPR targeted cells indicates a near complete deletion of the targeted region. However, deletion of this region led to no change in gene expression of BZW2 in either undifferentiated myoblasts or myotubes (Fig. 4a). We speculate that the immortalized M2 cell line may not be a suitable system to recapitulate BZW2 gene regulation observed in primary myotubes. It is also possible that the effect of perturbing the immediate region harbouring this susceptibility variant was below our detection range.

Fig. 4.

Analysis of deletion of region containing putative causal SNP. (a) qPCR results showing BZW2 expression in M2 cells harbouring deletions spanning SNP rs2240855 at day 0 and 6 of differentiation. NS: Not Significant.

Discussion

Enrichment for CTS GWAS heritability implicates skeletal muscle as a potentially important tissue setting for CTS pathogenesis (Fig. 1b). As CTS is known to be associated with other anthropomorphic measures, such as height and BMI,⁴ it is possible that factors influencing tissue, and specifically muscle size, may be relevant. Repetitive motion results in damage of musculoskeletal tissue. Similarly, we observed weaker enrichment in osteoblast cells. This finding highlights the importance of considering factors beyond traditional risk factors, such as age and gender. Further studies are needed to elucidate the molecular mechanisms underlying this association. While our variant to gene mapping approach identified more genes in myotubes, which is consistent with the observed higher enrichment GWAS heritability. However, the difference in the number of implicated genes may reflect technical differences as the myotube chromatin contacts were determined with Hi-C and was sequenced comparably deeper (∼6 billion reads passing quality control).³² To confirm the enrichment of muscle cells, we also performed similar enrichment analyses using the 66 cell types with cell specificity scores available from the Descartes single cell ATAC-seq atlas.³³ Using the 25,000 most specific peaks per cell type, we similarly detected enrichment of skeletal muscle cells, as well as multiple vascular relevant cells and Schwann Cells, peripheral neuronal support cells, suggesting important contribution from celltypes that were not present in our dataset.

Our approach implicated genes involved in pathways including cholesterol, glucose, and hyaluronan metabolism. Diabetes is a known risk factor for CTS, where high sugar levels could lead to the accumulation of advanced glycation end products in carpal tunnel tissues, which may in turn cause reduced elasticity and stiffness, leading to compression of the tissue.^34,35 Similarly, genes implicated by cholesterol metabolism have been associated with fibrosis and median neuron diameter, reducing blood flow, and causing nerve damage.³⁶ Indeed, idiopathic CTS has been associated with hypercholesterolemia and the cross-sectional area of the median nerve associates with low-density lipoprotein levels in middle-aged patients.³⁷ Hyaluronan is a major component of the extracellular matrix that influences hydration and lubrication of tissues; indeed, a recent clinical trial found that hyaluronan injections could provide a degree of short-term relief from CTS but was not beneficial to relieve chronic neuropathy.³⁸

In addition, our chromatin conformation data nominated genes involved in developmental processes, which points to the possible contribution of growth and patterning of the surrounding tissue. One gene of interest implicated by our approach was ZEB1, a transcription factor important for restraining myogenesis during.³⁹ ZEB1 competes with MyoD, the master regulator of myogenesis, to retrain myoblast differentiation. Loss of ZEB1 results in precocious expression of myotube and differentiation in animal models.⁴⁰ These observations, in addition to the known genetic correlations with anthropomorphic traits such as height, led us to investigate whether genes potentially involved in tissue size and organization could be relevant for CTS pathogenesis.

We identified a novel role in muscle differentiation for BZW2 in the M2 muscle cell line. The BZW2 promoter contacted a candidate causal variant located in a potential muscle enhancer and was subsequently prioritized due to its expression being enriched in muscle. Our observations; loss of BZW2 results in precocious muscle cell differentiation. Using the M2 cell line, we leveraged CRISPR to disrupt BZW2 and the putatively causal variant, rs2240855. We observed that loss of BZW2 resulted in accelerated differentiation, suggesting that BZW2 acts to repress muscle cell differentiation. Changes in muscle size could lead to increased compression on the carpal tunnel. However, we did not observe changes when we introduced ∼400 bp deletions in the regulatory region, including the putative causal variant. We speculate that the immortalized M2 cell line may not be a suitable model to recapitulate BZW2 gene regulation observed in primary myotubes. It is also possible that the effect of perturbing the immediate region harbouring this susceptibility variant was below our detection range. Moreover, it is likely that regulation by the implicated long-distance enhancer is only functionally and temporally important under specific physiological contexts that could be missed in the in vitro differentiated M2 cell model. As we did not assess if the contact was also present in M2 cells, future work characterizing the degree of differentiation of these cells mirrors in vivo muscle differentiation would be valuable to interpret differences between these models. These observations highlight the importance of considering context for interpreting the result of functional assays to study GWAS variants.

Taken together, these observations suggest that BZW2 acts to repress muscle cell differentiation; loss of BZW2 results in precocious muscle cell differentiation. We hypothesize that variation in BZW2 expression may lead to alterations in muscle cell size or organization that could impact compression of the medial nerve. Further work in cell and animal models will be necessary to elucidate the precise mechanism through which BZW2 acts in muscle cell differentiation and CTS pathogenesis.

There are several limitations to this study. For instance, the set of cell types we used are not fully comprehensive. Other tissue including cartilage and peripheral neuronal cells may play a role, as well as endocrine tissues such as the thyroid. While we do see enrichment for skeletal muscle cREs, it is possible that these other cell types may harbour distinct cREs that contact other target genes.

Chromatin conformation approaches offer the ability to help prioritize non-coding variants at scale by identifying disease associated variants in cREs that contact gene promoters. A strength of this approach is that it does not require large cohorts, in contrast to needs of molecular QTL studies, to implicate specific genes, allowing for different cell types or cell states to be efficiently assayed. Resolving these challenges warrant the further increase in the number of cell types and states assayed, which should in turn increase the number of signals explained beyond the <50% detected in this study.

With this work bringing a focus on the potential role of skeletal muscle in conferring the risk for CTS highlights the importance of understanding the contribution of different cell types in the pathogenesis of this common disorder. While this study provides valuable insight into potential effector genes driving muscle dysfunction, it highlights the need for additional studies in other cell types hypothesized to be involved in the disease. Furthermore, additional studies are necessary to uncover mechanistically how these genes contribute to CTS pathology. Characterizing the effector genes associated with CTS risk can better inform strategies for developing therapeutic interventions, through informing either drug repurposing or target prioritization, to potentially treat patients or prevent development of disease in at-risk individuals.

Contributors

MCP: Conceptualization; Data Curation; Formal Analysis; Visualization; Writing-Original Draft; Writing-Review & Editing.

LL: Data Curation; Formal Analysis; Investigation; Visualization; Writing-Original Draft; Writing-Review & Editing.

JAP: Data Curation; Investigation; Writing-Review & Editing.

YW: Writing-Review & Editing.

KB: Investigation; Investigation.

KDH: Funding Acquisition, Writing-Review & Editing.

ADW: Supervision; Writing-Review & Editing.

WY: Data curation, Formal Analysis, Investigation, Funding Acquisition, Writing-Review & Editing, Supervision.

SFAG: Conceptualization, Investigation, Funding Acquisition, Writing-original draft, Writing-Review & Editing.

All authors have read and approve the final version of the manuscript. All underlying data has been verified by more than one author MP and SFAG for computational analyses; LL and WY for functional experiments.

Equal contributions: MP and LL; WY and SFAG.

Data sharing statement

Sequencing data is available from public repositories ArrayExpress and GEO. IPSC Neurons: ArrayExpress, E-MTAB-9159; E-MTAB-9087; hMSC Osteoblasts: ArrayExpress, E-MTAB-6862; E-MTAB-6834; hMSC Adipocytes: GSE164745, GSE164911. Skeletal muscle myotube data will be made available in GEO upon acceptance of a sister manuscript investigating Type II Diabetes GWAS loci.

Declaration of interests

KDH is the president of the Orthopaedic Research Society. All other authors report no actual or potential conflicts.

Appendix ASupplementary data

Supplemental Fig. S1

Supplemental Fig. S1: Expression of effector genes in resected tenosynovium from carpal tunnel patients.⁵ Boxplots depict the median (central line), 25–75% percentiles (box edges), the 1.5 interquartile range (whiskers), and outliers (points).

Supplemental Fig. S2

Supplemental Fig. S2: Results from fine-mapping of CTS variants using the 1000G Phase EUR LD reference and CARMA.¹² (a) Credible set size (x-axis) vs maximum individual posterior inclusion probabilities (PIP) values (y-axis). (b) Overlap of the implicated genes in each celltype with the predicted 95% credible set.

Supplemental Fig. S3

Supplemental Fig. S3: Venn diagram of the implicated proxies and genes implicated by myotube chromatin looping of cREs to promoters (brown) and bulk muscle eQTL (blue). Shared genes are highlighted.

Supplemental Fig. S4

Supplemental Fig. S4: Heatmap showing the standardized median TPM values for each gene implicated in our variant to gene mapping approach across GTEX tissues. Red/orange indicates a higher degree of expression for a gene in the indicated tissue while blue indicates lower expression.

Supplemental Table S1

Supplemental Table S1: Results from partitioned LD score regression in the cREs of the indicated cells using the latest CTS GWAS study.⁴

Supplemental Table S2

Supplemental Table S2: Results of variant to gene mapping for CTS loci, for GWAS signals where we implicated at least one gene.

Supplemental Table S3

Supplemental Table S3: Variant to gene mapping pairs either shared or specific across cell types.

Supplemental Table S4

Supplemental Table S4: Enriched GO terms from MSigDB GO Biological Processes annotation.

Supplemental Table S5

Supplemental Table S5: Predicted changes in motif affinities for implicated variants.

Supplemental Table S6

Supplemental Table S6: RT-PCR, PCR and cloning primers.