A genome-wide association study identifies 5 loci associated with frozen shoulder and implicates diabetes as a causal risk factor.

Frozen shoulder is a painful condition that often requires surgery and affects up to 5% of individuals aged 40-60 years. Little is known about the causes of the condition, but diabetes is a strong risk factor. To begin to understand the biological mechanisms involved, we aimed to identify genetic variants associated with frozen shoulder and to use Mendelian randomization to test the causal role of diabetes. We performed a genome-wide association study (GWAS) of frozen shoulder in the UK Biobank using data from 10,104 cases identified from inpatient, surgical and primary care codes. We used data from FinnGen for replication and meta-analysis. We used one-sample and two-sample Mendelian randomization approaches to test for a causal association of diabetes with frozen shoulder. We identified five genome-wide significant loci. The most significant locus (lead SNP rs28971325; OR = 1.20, [95% CI: 1.16-1.24], p = 5x10-29) contained WNT7B. This variant was also associated with Dupuytren's disease (OR = 2.31 [2.24, 2.39], p<1x10-300) as were a further two of the frozen shoulder associated variants. The Mendelian randomization results provided evidence that type 1 diabetes is a causal risk factor for frozen shoulder (OR = 1.03 [1.02-1.05], p = 3x10-6). There was no evidence that obesity was causally associated with frozen shoulder, suggesting that diabetes influences risk of the condition through glycemic rather than mechanical effects. We have identified genetic loci associated with frozen shoulder. There is a large overlap with Dupuytren's disease associated loci. Diabetes is a likely causal risk factor. Our results provide evidence of biological mechanisms involved in this common painful condition.

Introduction

Frozen shoulder, also known as adhesive capsulitis, affects 2–5% of the population at some point in their lives [1,2]. It is characterised by initial shoulder pain followed by a gradual reduction in range of movement or “freezing”. Pain subsides but joint stiffness can continue for years, causing significant disability [3]. The onset of frozen shoulder is usually between 40 and 60 years [4].

Little is known about the causes of frozen shoulder. Genome-wide association studies can provide new insights into underlying biological mechanism and possible drug targets. For example, a recent GWAS of osteoarthritis has been used to identify new therapeutic targets for the condition [5]. To date, there have been no published genome wide association studies (GWAS) of frozen shoulder.

Diabetes is the strongest known risk factor for frozen shoulder. Individuals with diabetes have a greatly increased lifetime risk, with a hazard ratio of 1.33 [6]. Whether diabetes causes frozen shoulder is unclear because the association may reflect residual confounding by other risk factors such as age, obesity [7] and Dupuytren’s disease [8].

Mendelian randomization is a statistical method that can be used to infer causal relationships between an exposure and an outcome by using genetic variants associated with the exposure [9]. The exposure associated variants can be used as an unconfounded proxy for the exposure, as their inheritance is random at conception. This method is now extensively used to infer causal associations and proof of principle examples include evidence that increased BMI causes diabetes [10] and increased LDL cholesterol causes coronary artery disease [11].

Using the UK Biobank, we performed a genome-wide association study (GWAS) to identify genetic variants associated with frozen shoulder. We used publicly available summary statistics from 176,899 individuals from the FinnGen study for replication [12]. We then used Mendelian randomization to test for causal associations of diabetes and obesity with frozen shoulder.

Methods

Ethics statement

Ethics approval for the UK Biobank study was obtained from the North West Centre for Research Ethics Committee (11/NW/0382)[13]. Written informed consent was obtained from all participants.

Frozen shoulder cases in the UK Biobank

The primary GWAS was performed on cases of frozen shoulder identified from the ICD-10 code M750, OPCS4 code W871 and read codes (N210, XE1FL and XE1Hm) from the primary care data in UK Biobank. Controls were defined as individuals without a record of one of these codes.

Other phenotypes

ICD-10 codes were used for related disorders (Dupuytren’s disease (M720), rotator cuff (M751), and calcific tendinitis of shoulder (M753)). Further, OPCS in Supplementary Table 1 of [14] (all sections except anatomical codes) and primary care read codes 7H320, 7H340, N236., Xa8RB, XE07l, XE1Fj, XM0ug, and XM1HO were used for Dupuytren’s disease. Type 1 diabetes was classified favouring specificity over sensitivity, defined by diagnosis aged ≤ 20, on insulin within 1 year of diagnosis and at time of recruitment, not using oral antihyperglycemic agents, and not self-reporting type 2 diabetes, as described in [15]. Type 2 diabetes was defined by participants answering yes to the question ‘Has a doctor ever told you that you have diabetes?’ excluding those who reported using insulin within one year of diagnosis, were diagnosed under the age of 35, or were diagnosed within the past year [16]. We used UK Biobank variable 21001 for BMI measurement. HbA1c in mmol/mol is defined using variable n_30750_0_0 in UK Biobank. We defined diabetic retinopathy using self-report code 1276, diabetic neuropathy / ulcers using 1468 and diabetic nephropathy using 1607. Diabetes duration was defined by the difference between age at baseline and data-field 2976 (age diabetes diagnosed).

Genome-wide association study

We performed two case-control genome-wide association studies using BOLT-LMM [17], which applies a linear mixed model to test for an association between each SNP and the outcome trait, including population structure as a part of the model. Age, sex, study centre and genotyping chip were also included as covariates. We performed our primary GWAS using ICD10, OPCS and primary care codes, and a sensitivity analysis including only cases identified in ICD10 and OPCS. Briefly, principal component analysis was performed using individuals from the 1000 Genomes Project prior to projection of UK Biobank individuals into the principal component space. K-means clustering was subsequently applied to classify individuals as European, with centres initiated to the mean principal component values of each 1000 Genomes sub-population. The first 4 principal components were used in this analysis. We used LocusZoom to plot the resulting significant locus [18] and GTEx V8 to explore the possibility of functional variants [19]. P-values < 5x10-8 were considered significant for the GWAS. Betas and standard errors from BOLT-LMM were converted to log-odds ratios using log(OR) = β/(μ(1−μ)), where μ is the case-control ratio, as per the BOLT-LMM user manual [17]. Standard errors were also divided by μ(1−μ).

Replication and meta-analysis

We used summary statistics from freeze 4 of the FinnGen publicly available resource [12] to replicate the genome-wide significant findings in the discovery GWAS. Cases of frozen shoulder were defined in FinnGen using the ICD-10 code M750. METAL was used to perform a genome-wide meta-analysis of UKBB and FinnGen results, based on log-OR and standard errors [20]. We performed two meta-analyses using UK Biobank and FinnGen, the first using all cases from UK Biobank and the second using ICD-10 and OPCS codes.

Observational analyses

Observational analyses were performed on a subset of 379,708 unrelated individuals, 2,132 cases and 377,576 controls. We used a KING Kinship[21] to exclude those third-degree relatives or closer. An optimal list of unrelated individuals was generated by preferentially removing individuals with the maximum number of relatives to allow maximum numbers of individuals to be included.

We tested for associations with demographic and clinical features using logistic regression models, firstly using a univariable logistic regression model on 7 variables of interest (sex, age, TDI, type 1 diabetes, type 2 diabetes, BMI and WHR) and secondly using all variables in a multivariable logistic regression model to identify independent associations. We analysed Type 1 and Type 2 diabetes separately. p<0.007 (0.05/7) was considered significant. We performed a separate analysis specifically of diabetes-related variables (type, duration, retinopathy, neuropathy and nephropathy) for which p<0.01 (0.05/5) was considered significant.

Mendelian randomization

We applied two different methods. The one sample Mendelian randomization (MR) results were performed in two stages: first, the association between each exposure and a genetic risk score was used to derive a genetically predicted exposure value, and second, these predicted values were used in a logistic regression model of frozen shoulder, adjusting for ancestral principal components, assessment centre, genotyping platform, age and sex.

Second, we used two-sample methods, which involve regressing the effect sizes of variant–outcome associations against the effect sizes of the variant–risk factor associations for a set of SNPs associated with the exposure. We performed inverse variance weighted (IVW) instrumental variable analysis and two further methods that are more robust to potential violations of the standard MR assumptions (MR-Egger [22] and weighted-median MR [23]). We used effect sizes and 30 SNPs for type 1 diabetes reported in [24] using the Wellcome Trust Case Control Consortium, and a type 2 diabetes genetic risk score of 88 variants identified in [25,26]. The T1D-GRS uses 2 SNPs, rs2187668 and rs7454108, to tag and categorize the high risk DR3/DR4 haplotypes [24], this requires access to individual level data which we did not have for the FinnGen study. We therefore performed two analyses: one where we used the 2 SNPs to impute DR3/DR4 status and another where we used the SNPs as standard risk SNPs. We performed the above two-sample Mendelian randomisation in UK Biobank cases identified by ICD10 and OPCS codes, all UK Biobank cases and the FinnGen cohort. We also performed Mendelian-Randomisation using log-odds ratios and standard errors from the two genome-wide meta-analyses.

P-values < 0.01 were considered significant for Mendelian randomisation results. Mendelian randomisation results using a diabetes status as an exposure should be interpreted in terms of liability, which may cause frozen shoulder via changes in diabetes status and / or by independent pathways such as glycaemia[27].

Results

Diabetes is associated with frozen shoulder in the UK Biobank

To identify associated risk factors for frozen shoulder we tested several measures of diabetes and its related traits to assess their independent contributions. There were 2,540 cases of frozen shoulder in the UK Biobank based on ICD-10 and OPCS codes, and 10,104 including GP records. A Venn diagram showing overlap can be found in . Demographic and clinical associations with frozen shoulder defined by ICD10 and OPCS can be found in . Those with frozen shoulder were more likely female, had lower Townsend deprivation index, more likely obese (higher BMI and waist-hip-ratio) and more likely to have either type 1 or type 2 diabetes, with type 1 giving the strongest association. In a multivariable logistic regression model, only sex, types 1 and type 2 diabetes showed association.

Diabetes duration was nominally associated with frozen shoulder adjusted for diabetes type (OR per additional year with diabetes 1.02 [1.00–1.04], p = 0.02). HbA1c associated with frozen shoulder independently of type or duration of diabetes (OR = 1.09 [95% CI: 1.02–1.17], p = 0.009). Diabetic retinopathy (OR = 1.99 [95% CI, 1.19–3.33], p = 0.009) associated with frozen shoulder independently of type or duration of diabetes. This suggests that individuals with longer duration and less well controlled diabetes have higher risk of frozen shoulder.

Three loci are associated with frozen shoulder from discovery GWAS

presents the results of the frozen shoulder genome-wide association study using all frozen shoulder cases. We identify a genome-wide significant peak on chromosome 22. The A allele at the lead SNP, rs28971325, has a frequency of 26.4% in cases and 23.2% in controls (OR = 1.20, [95% CI: 1.16–1.24], p = 5x10-29). We also observe two additional genome-wide significant signals: rs5777216 on chromosome 1 (OR = 0.92 [95%CI 0.89–0.94], p = 1x10-9) and rs1042704 on chromosome 14 (OR = 1.11 [95%CI 1.07–1.15], p = 1x10-9).

Manhattan plot of discovery GWAS for frozen shoulder in UK Biobank.

The plot shows–log10(p) values for each single nucleotide polymorphism [SNP] in the HRC Imputation Panel and their association with frozen shoulder defined by ICD10, OPCS codes and GP records with p < 0.01, computed using BOLT-LMM and plotted using an in-house MATLAB script which we have made publicly available on the MATLAB File Exchange [31]. The horizontal dashed line is the genome-wide significance threshold at p = 5×10−8. Positions are based on the hg19 reference human genome.

When including only ICD10 and OPCS codes in the analysis, rs28971325 associated with OR = 1.32 (95%CI 1.22–1.41), p = 7x10-17. rs5777216 (OR = 0.91 [95%CI 0.86–0.96], p = 1x10-3) and rs1042704 (OR = 1.12 [95%CI 1.05–1.20], p = 8x10-4) were not genome-wide significant. A Manhattan Plot showing these results can be found in

Replication and meta-analyses of genome-wide association studies

We sought replication of the association signals in the FinnGen study. FinnGen have recently released GWAS summary statistics for 1801 diseases and traits, including frozen shoulder [12]. The lead SNP from the UK Biobank, rs28971325, was associated with frozen shoulder in FinnGen with a similar effect size to the UK Biobank results using ICD10 and OPCS codes (OR = 1.30 [95%CI 1.20–1.41], p = 6x10-10). The frequency of the A allele was 20.8% in cases compared to 17.6% in controls. rs1042704 was nominally associated with frozen shoulder in the FinnGen cohort (OR = 1.15 [95%CI: 1.07–1.24] p = 2x10-4), while rs5777216 (OR = 0.94 [95%CI: 0.89–1.00] p = 6x10-2) was not.

We note five genome-wide significant loci when performing a meta-analysis using all UK Biobank cases and FinnGen, including the three previously associated loci. Odds ratios for these SNPs were similar across all three definitions, except for rs28971325, which was weaker in UKBB’s GP records. The summary statistics can be found in . Manhattan Plots of the meta-analyses can be found in .

WNT7B, MMP14 and SFRP4 are potential causal genes

S5–S9 Figs provide LocusZoom plots [18] for the 5 associated loci. For the strongest associated locus, rs28971325 (), there were no coding variants with an r2 > 0.8. There are no gene expression associations in GTEX for rs28971325. However, a recent study found WNT7B was the second most differentially expressed transcript genome-wide with a log fold change of 7 (p = 1x10-16) in anterior capsule tissue from 22 patients undergoing arthroscopic capsulotomy surgery for frozen shoulder compared to 26 controls [28]. The lead SNP at the next strongest associated locus, rs1042704, is a missense variant in the MMP14 gene (). This variant has previously been associated with Dupuytren’s disease and has been shown to have a 99% posterior probability of being the causal variant [29]. The chromosome 7 locus is also a known Dupuytren’s locus, and previous work has suggested SFRP4 is the causal gene [29].

Frozen shoulder associated SNPs have been associated with Dupuytren’s contracture and bone mineral density

A previous GWAS study of Dupuytren’s disease found an association in the same genomic region of WNT7B [29]. The lead SNP for Dupuytren’s disease, rs7291412, was not associated with frozen shoulder in UK Biobank (OR = 0.98 [95%CI: 0.95–1.01], p = 0.17) or FinnGen (OR = 1.00 [95%: 0.94, 1.06], p = 0.92). Of the five genome-wide significant loci in the meta-analysis, rs28971325, rs1042704 and rs2472660 associated with Dupuytren’s disease at the genome-wide significant level, rs17570529 was nominally significant (p = 1x10-5), and rs2472660 did not associate (p = 0.21) (). None of these five loci associated with either rotator cuff or calcific tendinitis of shoulder at p<0.01.

Dupuytren’s disease and other fibroblastic diseases

There was limited overlap between Frozen Shoulder and Dupuytren’s disease observationally: 61 (2.4%) of frozen shoulder cases in the ICD10+OPCS definition and 198 (2.0%) including GP record cases had an ICD10 code for Dupuytren’s disease. However, many of the previously reported Dupuytren’s disease SNPs were associated with frozen shoulder with a correlation in effect sizes (R2 = 0.72, p = 1x10-7, ). The association of the 5 SNPs with frozen shoulder did not change when we excluded individuals with Dupuytren’s disease from the cases.

Mendelian randomization provides evidence that diabetes is causal to frozen shoulder

We used Mendelian randomization to explore whether the associations with Type 1 and Type 2 diabetes reported in were causal. Using 1-sample Mendelian randomization methods, genetic data provided evidence that type 1 diabetes causes frozen shoulder: OR 1.05 [95% CI: 1.02–1.09], p = 0.002 using ICD-10 and OPCS codes and OR 1.04 [95% CI: 1.02–1.06], p = 2x10-6 including primary care records. Evidence of a causal role of type 2 diabetes was weaker: OR 1.10 [95% CI: 0.99–1.22], p = 0.07 using ICD-10 codes and OR 1.07 [95% CI: 1.02–1.13], p = 0.006 including cases identified by UK Biobank primary care records. These results were consistent when we performed a sensitivity analysis adjusting both stages of the regression for significant variables in (BMI, WHR and TDI) to correct for potential pleiotropy (.

The association with type 1 diabetes was replicated using more robust, two-sample MR methods IVW (p = 3x10-6) and MR-Egger (p = 2x10-6). There was limited evidence from the two-sample methods that type 2 diabetes causes frozen shoulder (IVW: p = 0.06, MR-Egger: p = 0.96). The supplementary tables contain the full results from our Mendelian randomization analyses, including T1D-GRS results that exclude HLA variants which have greater potential for pleiotropy, for which the effect sizes were consistent ( shows the two stage regression, shows the IVW results and heterogeneity p values, and shows the MR-Egger results). Figs show plots showing the lines of best fit for MR-Egger, IVW, Median IV and Penalised Median IV methods for type 1 and type 2 diabetes defined by ICD-10 codes, demonstrating in for type 1 diabetes SNPs, a strong correlation between the betas for type 1 diabetes and frozen shoulder and in no association for type 2 diabetes SNPs.

Discussion

We have identified robust associations of common genetic variants with frozen shoulder. Frozen shoulder is a condition which affects up to 5% of the population around between the ages of 40 to 60, but the causes of the disease, and particularly the transient nature of the condition are unknown. The genome-wide association study and Mendelian randomization analyses we report here provide new insights into the underlying causes of the condition.

WNT7B is a candidate causal gene at the most strongly associated locus. As with most GWAS studies, further work is needed to identify the causal variant at the locus. GTEX analyses did not identify any strong eQTL’s with the lead SNP at the associated locus. However, a recent study performed RNA-seq on anterior capsule tissue from 22 patients undergoing arthroscopic capsulotomy surgery for adhesive capsulitis and compared to 26 undergoing arthroscopic stabilization surgery for a different condition [28]. WNT7B was the second most differentially expressed transcript genome-wide with a log fold change of 7 (p = 1x10-16), although it was noted that the expression levels of WNT7B was relatively low. The WNT signalling pathway has been highlighted by previous GWAS of related fibroblastic diseases [30]. Other potential causal genes at associated loci included MMP14, a missense variant previously finemapped for Dupuytren’s disease, and SFRP4.

Three of the identified loci for frozen shoulder also associate with Dupuytren’s disease. Dupuytren’s disease is a common condition which is characterised by a hand deformity where there is contracture of connective tissue within the palm and digits preventing full finger extension. A previous GWAS of Dupuytren’s identified 26 associated loci with the condition [29]. One of the more strongly associated loci also contained the WNT7B locus. However, the lead SNP at that locus has not been associated with frozen shoulder and does not associate with frozen shoulder in the UK Biobank or FinnGen. The odds ratio for the lead SNP in our study has a significantly stronger odds ratio for Dupuytren’s than the strongest association signal from the Dupuytren’s GWAS. The explanation may be that older versions of SNP chips or imputation panels were used in the previous study which did not capture the lead SNP from our current study of frozen shoulder. Overall there is a strong overlap in signals between Dupuytren’s disease and frozen shoulder. Some overlap might be expected, as both conditions are a result of contracture of connective tissue planes. WNT is a known regulator of planar cell polarity and causes tissue planes to contract via the non-canonical pathway. Dupuytren’s disease differs from frozen Shoulder in that frozen shoulder is most commonly a transient condition.

Diabetes is known to be an observational risk factor for frozen shoulder, but the causal nature of the association was unclear. Our Mendelian randomization analyses in UK Biobank provide evidence that Type 1 diabetes is causal for the condition. The weaker association with Type 2 diabetes is likely due to differences in duration of diabetes or earlier diabetes onset, as Type 1 diabetes is generally diagnosed earlier (half before the age of 30 years) whereas the diagnosis of Type 2 diabetes generally occurs later in life (after 50 years). It is likely all individuals with diabetes would have an increased risk of frozen shoulder, with those with longer duration (like those individuals with Type 1 diabetes) and worse glycaemic control at increased risk.

There are limitations to our study. The UK Biobank ICD-10 data is reliant on patients having a coded diagnosis at hospital, which is reliant on accuracy of hospital coding, and also reliant on the condition being severe enough for the patient to go to hospital. This may result in less serious cases being classified as controls. As such, they may be overly specific, while the primary care records are potentially overly sensitive and include cases that are not true frozen shoulder cases because it can be difficult to accurately diagnose in primary care. Consistent with this the effect size for frozen shoulder was smaller when cases were based on presence in primary care records rather than inpatient ICD-10 codes for the most associated loci. This study was performed using only white Europeans, and further study is needed to determine if results replicate for other ethnic groups. The UK Biobank only includes patients between the ages of 40 and 69 at recruitment, although this covers the most common incidence age range for frozen shoulder.

We have identified genetic loci associated with frozen shoulder. There is a large overlap with Dupuytren’s disease genetic loci. Diabetes is a likely causal risk factor. Our results provide evidence of biological mechanisms involved in this common painful condition.

One sample Mendelian randomisation results.

A table showing results from a two-stage regression where the GRS for the exposure was regressed against the exposure, and the genetically predicted exposure was used to predict frozen shoulder. Adjusted refers to a sensitivity analysis in which both stages of the regression were adjusted for the significant variables in Table 1.

Table 1

Demographic and clinical associations with frozen shoulder in the UK Biobank.

The OR and p value columns were calculated using univariable logistic regression. The adjusted OR and adjusted p value columns were calculated using a multivariable logistic regression model. Adjusted results for diabetes types were calculated from a model excluding the other type. These analyses used only unrelated individuals in the UK Biobank.

Variable	Cases (2,132)	Controls (377,576)	OR [95% CI]	p value	Adjusted OR [95% CI]	Adjusted p value
Male sex	174,125 (46%)	820 (38%)	0.73 [0.67–0.80]	2x10-12	0.64 [0.58–0.71]	1x10-18
Age [SD]	57.2 [8.0]	57.2 [7.4]	1.00 [0.99–1.01]	0.86	1 [1.00–1.00]	0.59
Townsend deprivation index [SD]	-1.26 [3.1]	-1.48 [3.0]	1.02 [1.01–1.04]	7x10-4	1.01 [0.99–1.03]	0.10
Type 1 diabetes	23 (1.2%)	356 (0.09%)	12.23 [8.00–18.69]	9x10-31	13.09 [8.55–20.03]	2x10-32
Type 2 diabetes	179 (9.0%)	11,887 (3.2%)	2.99 [2.56–3.49]	1x10-43	3.00 [2.54–3.54]	6x10-39
BMI [SD]	28.0 [4.9]	27.4 [4.8]	1.12 [1.07–1.18]	1x10-10	1.03 [0.97–1.09]	0.12
WHR [SD]	0.873 [0.09]	0.871 [0.09]	1.16 [1.11–1.21]	1x10-17	1.03 [0.97–1.09]	0.09

(DOCX)

Click here for additional data file.

IVW Mendelian randomisation results.

A table showing results the IVW analysis. Meta Analysis 1 refers to using the betas and standard errors from the meta-analysis GWAS with FinnGen using ICD10 + OPCS from UKBB. Meta Analysis 2 refers to the same using ICD10 + OPCS + GP records from UKBB. The P het column contains heterogeneity.

(DOCX)

Click here for additional data file.

MR-Egger results.

A table showing results the MR-Egger results. Meta Analysis 1 refers to using the betas and standard errors from the meta-analysis GWAS with FinnGen using ICD10 + OPCS from UKBB. Meta Analysis 2 refers to the same using ICD10 + OPCS + GP records from UKBB.

(DOCX)

Click here for additional data file.

Venn diagram of UK Biobank cases.

This Venn diagram shows overlap between the different case definitions of frozen shoulder (ICD10, OPCS, and GP record codes) in the UK Biobank.

(TIF)

Click here for additional data file.

Manhattan plot of sensitivity analysis GWAS for frozen shoulder in UK Biobank.

The plot shows–log10(p) values for the association of each single nucleotide polymorphism [SNP] in the HRC Imputation Panel and their association with UK Biobank frozen shoulder cases defined by ICD10 and OPCS codes. The horizontal dashed line is the genome-wide significance threshold at p = 5×10–8.

(TIF)

Click here for additional data file.

Manhattan plot of primary meta-analysis.

The plot shows–log10(p) values for the association of each single nucleotide polymorphism [SNP] in the HRC Imputation Panel and their association in the meta-analysis using UK Biobank ICD10 + OPCS and FinnGen. The horizontal dashed line is the genome-wide significance threshold at p = 5×10–8.

(TIF)

Click here for additional data file.

Manhattan plot of secondary meta-analysis.

The plot shows–log10(p) values for the association of each single nucleotide polymorphism [SNP] in the HRC Imputation Panel and their association in the meta-analysis using UK Biobank ICD10 + OPCS + GP Records and FinnGen. The horizontal dashed line is the genome-wide significance threshold at p = 5×10–8.

(TIF)

Click here for additional data file.

LocusZoom plot of rs28971325.

(TIF)

Click here for additional data file.

LocusZoom plot of rs1042704.

(TIF)

Click here for additional data file.

LocusZoom plot of rs5777216.

(TIF)

Click here for additional data file.

LocusZoom plot of rs17570529.

(TIF)

Click here for additional data file.

LocusZoom plot of rs2472660.

(TIF)

Click here for additional data file.

Scatterplot of Frozen Shoulder betas against Dupuytren betas for Dupuytrens SNPs.

Betas and 95% CIs for association between Frozen Shoulder and between Dupuytren’s Disease for all Dupuytren’s SNPs in Ng. et. al.

(TIF)

Click here for additional data file.

25 Jan 2021

Dear Dr Weedon,

Thank you very much for submitting your Research Article entitled 'A genome-wide association study of frozen shoulder identifies a common variant of WNT7B and diabetes as causal risk factors' to PLOS Genetics.

The manuscript was fully evaluated at the editorial level and by two peer reviewers. As you will see, both reviewers are generally positive but identify several major concerns that would be necessary to address for the manuscript to move forward. The comments of reviewer #2 should be straightforward to address, but note that the comments form reviewer #1 will require substantial effort. If the reviewers' comments can be addressed satisfactorily, we would be interested to evaluate a revised manuscript. We cannot, of course, promise publication at that time.

Should you decide to revise the manuscript for further consideration here, your revisions should address the specific points made by each reviewer. We will also require a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript.

If you decide to revise the manuscript for further consideration at PLOS Genetics, please aim to resubmit within the next 60 days, unless it will take extra time to address the concerns of the reviewers, in which case we would appreciate an expected resubmission date by email to plosgenetics@plos.org.

If present, accompanying reviewer attachments are included with this email; please notify the journal office if any appear to be missing. They will also be available for download from the link below. You can use this link to log into the system when you are ready to submit a revised version, having first consulted our Submission Checklist.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see our guidelines.

Please be aware that our data availability policy requires that all numerical data underlying graphs or summary statistics are included with the submission, and you will need to provide this upon resubmission if not already present. In addition, we do not permit the inclusion of phrases such as "data not shown" or "unpublished results" in manuscripts. All points should be backed up by data provided with the submission.

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool.  PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

PLOS has incorporated Similarity Check, powered by iThenticate, into its journal-wide submission system in order to screen submitted content for originality before publication. Each PLOS journal undertakes screening on a proportion of submitted articles. You will be contacted if needed following the screening process.

To resubmit, use the link below and 'Revise Submission' in the 'Submissions Needing Revision' folder.

[LINK]

We are sorry that we cannot be more positive about your manuscript at this stage. Please do not hesitate to contact us if you have any concerns or questions.

Yours sincerely,

Gregory S. Barsh

Editor-in-Chief

PLOS Genetics

Gregory Copenhaver

Editor-in-Chief

PLOS Genetics

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: PLOS Genetics peer review report

A genome-wide association study of frozen shoulder identifies a common variant of

WNT7B and diabetes as causal risk factors

Green et al.

This article describes a GWAS of frozen shoulder in UK Biobank with replication in FinnGen. The authors discovered 1 locus, and used MR to show that type 1 DM is a causal risk factor for frozen shoulder.

I enjoyed reading the paper. The work is novel, and will be of interest to the scientific community. The paper overall is well-written. I have suggested major changes that must be performed before acceptance, and would favour the authors being given a chance to perform these analyses and address my comments in a re-submission.

Major points

1. The authors have used only diagnostic codes to define disease cases for both FS and DD. This seriously underestimates the number of cases, as many are only identified by surgical (OPCS) codes. For example, for FS, a quick browse of the UK biobank data showcase reveals the following numbers of cases:

Code Cases

ICD10 – M750 2314

OPCS – W781 922

This misclassification may have led to a critical loss of power to detect association (I note several sub-threshold loci in figure 1). This is a missed opportunity, and should be rectified before publication.

Surgical codes for DD can be found in supplementary data file of https://doi.org/10.1038/s41598-020-73595-y

2. In a similar vein, GP data has been recently released, and the authors identify 7,913 cases of FS in the GP data. These cases are also misclassified in the GWAS analysis, and should be included as cases (note only 323 overlapped).

The authors looked only at the associated variant in this data I think, though it is not clear as there is no methodology reported for this part of the genetic analysis. The authors should combine the cases from both hospital records above and GP records into a proper case-control cohort to avoid misclassification and maximise power. They should then do a single proper GWAS. This again is a major missed opportunity to more accurately define the genetic architecture of FS, and should be performed before publication.

3. The reporting of the analysis in FinnGen is inadequate. Did the authors just test the single associated variant? What about suggestive variants that were below the genomewide significance threshold? Did they do a GWAS and then meta-analysis with the summary statistics? They should perform a full GWAS (or use the pre-prepared GWAS data) and meta-analysis as it is possible that this will reveal further associated loci, and further delineate the genetic architecture of FS. Again, a missed opportunity that should be rectified before publication.

4. Supplementary tables 2 and 3: I am unsure as to the precise definition of cases and controls here, and what the methodology for “meta-analysis” was (this is not described in the methods, and definitely needs to be), but I have a serious concern about performing meta-analysis when cases and (especially) controls are likely to be the same in two analyses. Specifically here in the “UK Biobank ICD” and “UK Biobank GP” cohorts I think it is likely that there is very large sample overlap. How is this dealt with in such a “meta-analysis”.

Minor points

1. Introduction Page 3 “Diabetes is the strongest known risk factor for Diabetes” – I think you mean for FS.

2. Remove “a deeply phenotyped…..from the UK” – PLOS genetics readers know the UK Biobank.

3. Results: “…had a lower Townsend Deprivation Index…” Not really. This is very marginally significant (if you were to Bonferroni correct your seven univariable tests), and is not significant in the multivariable model. Please remove.

4. “Diabetic eye disease….” – how many variables related to diabetes did you test? Did you do a Bonferroni correction? This is very marginally “significant” (p=0.04) and is likely a false positive.

5. Locuszoom should be LocusZoom

6. Page 7 – what is the r2 between rs62228062 and rs7291412? Was rs62228062 in the 95% credible set for that locus in Ng et al? And therefore is the predisposition to FS and DD in this region of Chr. 22 likely driven by the same underlying causal variant?

7. Page 7 – how many cases overlapped between FS and DD?

8. Page 7 – I think the misclassification bias discussed above extends to the MR analysis, but the methodology of which cases and controls were used in this analysis is unclear - the methods section needs to be very clearly written.

9. Page 9 – replace elongated with extended.

10. Page 9 – “…chronic condition like DD…” – please read up a bit on DD. It is a fibrotic condition just like FS, it has a similar age of onset (mean 63 years). Many patients with DD also get FS. It is not at all unexpected that both conditions might share some genetic predisposition. For example:

https://doi.org/10.1302/0301-620X.77B5.7559688

https://doi.org/10.1067/mse.2001.112883

11. Page 9, final paragraph, remove the first sentence – it is repetitive, and also misleading, as only Type1 DM was really shown to be causative.

12. Please remove all claims to identifying the “first” variant associated with FS. This should be determined in retrospect, and such claims of primacy are not fitting of a scientific article.

Tables

1. Table 1 should have numbers of participants, not just percentages. I don’t know if all of the cases in UK Biobank were included in this analysis, or just the ones identified by ICD10 coding. It should be the former – if not, please re-run the analyses.

Reviewer #2: The authors conducted a GWAS on frozen shoulder in the UK Biobank sample, and replicated the finding using available GWAS results of a well-powered independent cohort. Using the GWAS results, 1-sample and 2-sample Mendelian randomization analyses were conducted revealing type 1 diabetes but not type 2 diabetes as causal risk factor for frozen shoulder. The analyses were generally sound, the findings robust, and the paper is well written. However, more details regarding the analysis methods need to be provided, and several points listed below should be addressed or clarified.

Although the conducted discovery-replication approach is valid and provides a robust result, it would be of interest to perform additionally a combined UK Biobank and FinnGen GWAS meta-analysis (without replication) to see, if there are any additional genome-wide significant associations on frozen shoulder by maximizing the sample size in the discovery stage.

Methods: please provide how significance was defined in the observational analysis, the MR analysis, the GWAS, and provide the replication criteria applied for the GWAS results.

How was required independence of instruments (SNPs) for the MR assessed?

Results page 6, first paragraph: It’s not quite clear how to “strong” associations were defined and compared, given that the effects presented in Table 1 are based on a mixture of continuous variables on different scales, and binary traits with different prevalence used as exposure. In addition, the sentence “ In a multivariable logistic regression model only the type 1 and type 2 diabetes showed strong association” is imprecise, as also sex had an OR below 1 and a very low adjusted p-value. Please clarify.

Observational results were provided for the duration of diabetes, but the methods are missing describing how diabetes duration (and of which type of diabetes) was assessed. This information needs to be added.

Regarding the 1-sample MR, please provide more details how and on which scale the exposure (i.e. diabetes) was estimated using the genetic factors and included in the second stage, and explain how the causal OR on frozen shoulder can be interpreted with respect to the units of the exposure.

As a sensitivity analysis, I suggest to adjust the first stage association in the 1-sample MR for the traits listed in Table 1 (except diabetes) to correct for potential pleiotropic effects of the SNPs that could lead to invalid instruments.

Please provide the number of instruments included in the respective MR analyses (e.g. in the Supplementary Tables), and provide heterogeneity measures of the 2-sample MR results.

Regarding the Dupuytren’s contracture, are the two WNT7B SNPs rs7291412 and rs62228062 in linkage disequilibrium? Please provide their R².

Although WNT7B is a strong candidate gene as stated in the Discussion, I do not see the conclusion that WNT7B is the likely causal gene at the locus supported by any analysis conducted in this project – the missing eQTL associations and coding variants rather suggest the opposite. Please re-phrase or remove this conclusion.

Please rephrase the “strong” eQTL in the Discussion because the effect size cannot be solely quantified by the association p-value. Furthermore, I encourage to re-assess the use of “strong” in combination with “OR” and “association” throughout the manuscript, and to use more appropriate wordings like higher/lower OR or association p-values.

Discussion end of page 9: the weaker effect of type 2 diabetes on frozen might be because of earlier onset of type 1 diabetes, not because of time point of diagnosis. Please clarify or correct this statement. In the next sentence, it should be written “would have an increased risk” instead of “would be”.

It should be stated were the UK Biobank frozen shoulder GWAS results will be available and accessible.

minor issues:

- page 4: please provide the version of GTEx used

- please add the number of cases and controls to Table 1

- please provide a meaningful p-value for the “T1D no DR3/DR4 haplotyping” meta-analysis results in Supplementary Tables 2 and 3

**********

Have all data underlying the figures and results presented in the manuscript been provided?

Large-scale datasets should be made available via a public repository as described in the PLOS Genetics data availability policy, and numerical data that underlies graphs or summary statistics should be provided in spreadsheet form as supporting information.

Reviewer #1: Yes

Reviewer #2: No: It is not stated where the UK Biobank frozen shoulder GWAS results are available.

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Dominic Furniss

Reviewer #2: No

29 Mar 2021

Submitted filename: Review_Response.docx

Click here for additional data file.

13 Apr 2021

Dear Dr Weedon,

Thank you very much for submitting your Research Article entitled 'A genome-wide association study of frozen shoulder identifies a common variant of WNT7B and diabetes as causal risk factors' to PLOS Genetics.

The revised manuscript was seen by the original reviewers. As you will see, they are both positive. There are some remaining concerns from reviewer #2 that we ask you address in a hopefully final round of minor revision that will not necessarily need additional external review.

We therefore ask you to modify the manuscript according to the review recommendations. Your revisions should address the specific points made by each reviewer.

In addition we ask that you:

1) Provide a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript.

2) Upload a Striking Image with a corresponding caption to accompany your manuscript if one is available (either a new image or an existing one from within your manuscript). If this image is judged to be suitable, it may be featured on our website. Images should ideally be high resolution, eye-catching, single panel square images. For examples, please browse our archive. If your image is from someone other than yourself, please ensure that the artist has read and agreed to the terms and conditions of the Creative Commons Attribution License. Note: we cannot publish copyrighted images.

We hope to receive your revised manuscript within the next 30 days. If you anticipate any delay in its return, we would ask you to let us know the expected resubmission date by email to plosgenetics@plos.org.

If present, accompanying reviewer attachments should be included with this email; please notify the journal office if any appear to be missing. They will also be available for download from the link below. You can use this link to log into the system when you are ready to submit a revised version, having first consulted our Submission Checklist.

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

Please be aware that our data availability policy requires that all numerical data underlying graphs or summary statistics are included with the submission, and you will need to provide this upon resubmission if not already present. In addition, we do not permit the inclusion of phrases such as "data not shown" or "unpublished results" in manuscripts. All points should be backed up by data provided with the submission.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

PLOS has incorporated Similarity Check, powered by iThenticate, into its journal-wide submission system in order to screen submitted content for originality before publication. Each PLOS journal undertakes screening on a proportion of submitted articles. You will be contacted if needed following the screening process.

To resubmit, you will need to go to the link below and 'Revise Submission' in the 'Submissions Needing Revision' folder.

[LINK]

Please let us know if you have any questions while making these revisions.

Yours sincerely,

Gregory S. Barsh

Editor-in-Chief

PLOS Genetics

Gregory Copenhaver

Editor-in-Chief

PLOS Genetics

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: Thank-you for responding so thoroughly to all of the comments from myself and the other reviewer. The paper is now vastly improved, and I strongly recommend publication without any further changes.

Reviewer #2: I thank the authors for addressing all my questions satisfactorily. However, a few additional information need to be added to the revised manuscript:

Please add the effect alleles and their allele frequencies to new Table 3.

As far as I understand, the linear mixed model of BOLD-LMM does not generate estimates that can be directly transformed into odds ratios (compared to e.g. logistic regression). Please add to the methods how the odds ratios (and CI) were approximated from the betas and SEs obtained from the linear mixed model. Are there any known limitations of this approximation e.g. with respect to allele frequency or case-control ratio? Apologies for missing to ask this question after the initial submission already, but addressing this point would be quite important for the reproducibility and interpretation of possible uncertainties of the GWAS results.

On page 6 (MR methods), please specify what the newly added sentence “We also performed Mendelian-Randomisation using betas and odds ratios from the two genome-wide meta-analyses.” exactly means. Did you use for the MR the betas estimated from log(odds ratio) (i.e. taking my comment on the BOLD-LMM results for FS into account)? Given that both exposures and outcome are binary traits, the simple statement “using betas and odds ratios” in the two-sample MR is somehow misleading.

**********

Have all data underlying the figures and results presented in the manuscript been provided?

Large-scale datasets should be made available via a public repository as described in the PLOS Genetics data availability policy, and numerical data that underlies graphs or summary statistics should be provided in spreadsheet form as supporting information.

Reviewer #1: Yes

Reviewer #2: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Dominic Furniss

Reviewer #2: No

28 Apr 2021

Submitted filename: Summary of changes.docx

Click here for additional data file.

4 May 2021

Dear Dr Weedon,

We are pleased to inform you that your manuscript entitled "A genome-wide association study identifies 5 loci associated with frozen shoulder and implicates diabetes as a causal risk factor" has been editorially accepted for publication in PLOS Genetics. Congratulations!

Before your submission can be formally accepted and sent to production you will need to complete our formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Please note: the accept date on your published article will reflect the date of this provisional acceptance, but your manuscript will not be scheduled for publication until the required changes have been made.

Once your paper is formally accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you’ve already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosgenetics@plos.org.

In the meantime, please log into Editorial Manager at https://www.editorialmanager.com/pgenetics/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production and billing process. Note that PLOS requires an ORCID iD for all corresponding authors. Therefore, please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field.  This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager.

If you have a press-related query, or would like to know about making your underlying data available (as you will be aware, this is required for publication), please see the end of this email. If your institution or institutions have a press office, please notify them about your upcoming article at this point, to enable them to help maximise its impact. Inform journal staff as soon as possible if you are preparing a press release for your article and need a publication date.

Thank you again for supporting open-access publishing; we are looking forward to publishing your work in PLOS Genetics!

Yours sincerely,

Gregory Barsh

Editor-in-Chief

PLOS Genetics

Gregory Copenhaver

Editor-in-Chief

PLOS Genetics

www.plosgenetics.org

Twitter: @PLOSGenetics

----------------------------------------------------

Comments from the reviewers (if applicable):

----------------------------------------------------

Data Deposition

If you have submitted a Research Article or Front Matter that has associated data that are not suitable for deposition in a subject-specific public repository (such as GenBank or ArrayExpress), one way to make that data available is to deposit it in the Dryad Digital Repository. As you may recall, we ask all authors to agree to make data available; this is one way to achieve that. A full list of recommended repositories can be found on our website.

The following link will take you to the Dryad record for your article, so you won't have to re‐enter its bibliographic information, and can upload your files directly:

http://datadryad.org/submit?journalID=pgenetics&manu=PGENETICS-D-20-01823R2

More information about depositing data in Dryad is available at http://www.datadryad.org/depositing. If you experience any difficulties in submitting your data, please contact help@datadryad.org for support.

Additionally, please be aware that our data availability policy requires that all numerical data underlying display items are included with the submission, and you will need to provide this before we can formally accept your manuscript, if not already present.

----------------------------------------------------

Press Queries

If you or your institution will be preparing press materials for this manuscript, or if you need to know your paper's publication date for media purposes, please inform the journal staff as soon as possible so that your submission can be scheduled accordingly. Your manuscript will remain under a strict press embargo until the publication date and time. This means an early version of your manuscript will not be published ahead of your final version. PLOS Genetics may also choose to issue a press release for your article. If there's anything the journal should know or you'd like more information, please get in touch via plosgenetics@plos.org.

21 May 2021

PGENETICS-D-20-01823R2

A genome-wide association study identifies 5 loci associated with frozen shoulder and implicates diabetes as a causal risk factor

Dear Dr Weedon,

We are pleased to inform you that your manuscript entitled "A genome-wide association study identifies 5 loci associated with frozen shoulder and implicates diabetes as a causal risk factor" has been formally accepted for publication in PLOS Genetics! Your manuscript is now with our production department and you will be notified of the publication date in due course.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript.

Soon after your final files are uploaded, unless you have opted out or your manuscript is a front-matter piece, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting PLOS Genetics and open-access publishing. We are looking forward to publishing your work!

With kind regards,

Katalin Szabo

PLOS Genetics

On behalf of:

The PLOS Genetics Team

Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom

plosgenetics@plos.org | +44 (0) 1223-442823

plosgenetics.org | Twitter: @PLOSGenetics

Table 2

Association of lead SNPs in the full meta analysis in the two input GWAS results (ICD+OPCS+GP in UKBB and FinnGen) and the UKBB result excluding GP records.

The effect allele frequency is from the UK Biobank cohort.

SNP	Effect Allele (Freq)	Meta Analysis	ICD+OPCS	ICD+OPCS+GP	FinnGen
22:46367144 rs28971325	A (0.233)	1.21 (1.25–1.18) 8E-37	1.32 (1.24–1.41) p = 7E-17	1.20 (1.16–1.24) p = 5E-29	1.30 (1.20–1.41) p = 6E-10
14:23312594 rs1042704	A (0.215)	1.12 (1.15–1.08) 1E-12	1.12 (1.05–1.20) p = 8E-04	1.11 (1.07–1.15) p = 1E-09	1.15 (1.07–1.24) p = 2E-04
1:115681565 rs5777216	GT (0.403)	1.09 (1.11–1.06) 3E-10	0.91 (0.86–0.96) p = 1E-03	0.92 (0.89–0.94) p = 1E-09	0.94 (0.89–1.00) p = 6E-02
15:86125607 rs17570529	T (0.266)	0.92 (0.94–0.89) 3E-09	1.10 (1.03–1.17) p = 4E-03	1.09 (1.05–1.12) p = 1E-07	1.10 (1.03–1.18) p = 6E-03
7:37981961 rs2472660	C (0.355)	0.93 (0.95–0.90) 2E-08	1.06 (1.00–1.12) p = 4E-02	1.07 (1.04–1.10) p = 6E-06	1.13 (1.06–1.20) p = 3E-04

Table 3

Association of frozen shoulder genetic loci with Dupuytren’s Disease.

CHR	SNP	BP	OR	95%CI	P
22	rs28971325	46367144	2.31	2.24–2.39	<1x10-300
14	rs1042704	23312594	1.18	1.14–1.22	3x10-21
1	rs5777216	115681565	1.02	0.99–1.05	0.21
15	rs17570529	86125607	1.08	1.04–1.11	1x10-5
7	rs2472660	37981961	1.32	1.28–1.36	7x10-73