Article Text
Abstract
Introduction This study aimed to assess the causal relationship between diabetes and frozen shoulder by investigating the target proteins associated with diabetes and frozen shoulder in the human plasma proteome through Mendelian randomization (MR) and to reveal the corresponding pathological mechanisms.
Research design and methods We employed the MR approach for the purposes of establishing: (1) the causal link between diabetes and frozen shoulder; (2) the plasma causal proteins associated with frozen shoulder; (3) the plasma target proteins associated with diabetes; and (4) the causal relationship between diabetes target proteins and frozen shoulder causal proteins. The MR results were validated and consolidated through colocalization analysis and protein–protein interaction network.
Results Our MR analysis demonstrated a significant causal relationship between diabetes and frozen shoulder. We found that the plasma levels of four proteins were correlated with frozen shoulder at the Bonferroni significance level (p<3.03E-5). According to colocalization analysis, parathyroid hormone-related protein (PTHLH) was moderately correlated with the genetic variance of frozen shoulder (posterior probability=0.68), while secreted frizzled-related protein 4 was highly correlated with the genetic variance of frozen shoulder (posterior probability=0.97). Additionally, nine plasma proteins were activated during diabetes-associated pathologies. Subsequent MR analysis of nine diabetic target proteins with four frozen shoulder causal proteins indicated that insulin receptor subunit alpha, interleukin-6 receptor subunit alpha, interleukin-1 receptor accessory protein, glutathione peroxidase 7, and PTHLH might contribute to the onset and progression of frozen shoulder induced by diabetes.
Conclusions Our study identified a causal relationship between diabetes and frozen shoulder, highlighting the pathological pathways through which diabetes influences frozen shoulder.
- Diabetes Mellitus, Type 1
- Diabetes Mellitus, Type 2
- Muscles
- Muscle, Skeletal
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
In recent years, there has been increasing evidence supporting that diabetes may be an important factor in the development of frozen shoulder. Several epidemiological studies reported a marked rise in the incidence of frozen shoulder in individuals with diabetes versus those without.
WHAT THIS STUDY ADDS
This study found that insulin receptor subunit alpha, interleukin-6 receptor subunit alpha, interleukin-1 receptor accessory protein, glutathione peroxidase 7, and parathyroid hormone-related protein participate in the pathological process of diabetes and impact the pathogenesis of frozen shoulder.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY
Our findings preliminarily elucidate the pathological mechanisms whereby diabetes affects the development of frozen shoulder, offer guidance on preventing the condition in diabetic patients, and shed light on the development of therapeutic agents for frozen shoulder.
Frozen shoulder, also known as adhesive capsulitis of the shoulder, is a common condition causing shoulder pain and limited motion and is predominantly prevalent among women who are aged between 40 and 60 years.1 This condition is typically characterized by fibroplasia and tissue fibrosis with inflammation, neo-angiogenesis and neo-innervation, which can ultimately result in fibrotic contracture of the shoulder capsule, thereby leading to shoulder pain.1 The incidence of frozen shoulder may be influenced by a range of factors, including age, gender, endocrine disorders, and shoulder injuries.1 Furthermore, female patients who undergo breast cancer surgery are at a higher risk of developing frozen shoulder.2
In recent years, there has been increasing evidence supporting that diabetes may be an important factor in the development of frozen shoulder. Several epidemiological studies reported a marked rise in the incidence of frozen shoulder in individuals with diabetes versus those without.3–5 However, the specific pathological mechanisms by which diabetes is implicated in the development of frozen shoulder remain unclear. A possible explanation is that diabetes can cause a series of pathological changes such as microangiopathy, inflammation and oxidative stress,6 7 which can further lead to elevation of a number of cytokines in soft tissues such as muscles, tendons, and joint capsules,8 therefore inducing inflammatory and fibrotic responses in the shoulder joint. Moreover, the formation of advanced glycation end products (AGEs) can be enhanced by both inflammation and oxidative stress, resulting in their accumulation around the shoulder joint.9 10 This phenomenon may result in neo-innervation and neo-angiogenesis, leading to limited mobility and shoulder pain.10
Mendelian randomization (MR) is an epidemiological research technique that utilizes genetic variation to ascertain the causal effect between exposure and outcome, while gamete formation is the process of acquiring genetic variation from the parent, abiding by the laws of MR, in order to obtain the genetic variation largely unconstrained by confounding factors.11 Therefore, by examining the impact of bearing a particular genetic variation on the outcome, a causal link between the specific exposure and outcome can be established. MR analysis is not plagued by issues of reverse causal associations and confounders that are commonly seen in traditional observational studies.12 For a randomized controlled study, MR analysis can not only circumvent conditions, difficulties, and ethical constraints, but also help generate design ideas. With the development of proteomic genome-wide association studies (GWAS) in recent years, MR analysis has been widely used to explore the relationship between plasma proteins and diseases.13 Given the crucial role of human proteins in disease pathology, the pathological processes by which diseases occur may be determined by protein identification.13 However, there is still a lack of MR analyses on the pathological processes of diseases based on plasma proteins. As for MR analysis examining the causal proteins of frozen shoulder based on plasma proteins, no report has been published at all.
This study is likely the first MR analysis of disease pathology by plasma proteins. Specifically, we aimed to examine how diabetes affects the initial onset and progress of frozen shoulder based on the plasma proteome and to pinpoint the precise pathological mechanisms involved.
Research design and methods
Study design
Figure 1 illustrates the overall design of this MR study, which consists of four steps: (1) Establish the causal relationship between diabetes and frozen shoulder; (2) Identify the causal proteins for frozen shoulder in the human plasma proteome; (3) Identify the target proteins for diabetes in the human plasma proteome; (4) Establish the causal relationship between the diabetes target proteins and the frozen shoulder causal proteins.
Data on diabetes
Six categories of statistics data related to diabetes were examined: fasting glucose (281 416 samples), type 1 diabetes (520 580 samples), type 2 diabetes (179 000 samples), fasting insulin (151 013 samples), glycated hemoglobin (146 806 samples), and 2-hour postprandial glucose (63 396 samples) (online supplemental table 1).14–16 The genetic variants linked to diabetes and fasting glucose that meet the criteria of p<5E-08 and low linkage disequilibrium (R2<0.001) in the related GWASs were selected for MR analysis.
Supplemental material
Data on frozen shoulder
The statistics data concerning frozen shoulder were obtained from a recent GWAS containing 10 104 subjects.4 To consolidate and validate the findings, summary statistics from the UK Biobank (1198 cases and 359 996 controls) and the FinnGen study (2942 cases and 167 641 controls) were also utilized (online supplemental table 1).
Source and selection of plasma protein data
The plasma protein quantitative trait loci (pQTL) from several large-scale human proteomic GWASs17–22 were retrieved, and only cis-pQTLs satisfying the following criteria were included: (1) indicating a genome-wide significant association (p<5E-08); (2) cis-pQTLs were defined as pQTLs within 1 Mb from the gene encoding the protein; (3) linkage disequilibrium clumping R2<0.001; and (4) situated outside the major histocompatibility complex region (chr6, 26‒34 Mb). Eventually, 2419 cis-pQTLs were identified for 1651 proteins. In addition, the plasma cis-pQTLs data obtained from a recently published study by Zhang et al23 were used for external validation, including 2004 plasma proteins from 7213 European Americans and 1618 plasma proteins from 1871 African Americans.
MR analysis
The inverse variance weighted (IVW) method was adopted as the primary MR method, while the weighted median and MR-Egger methods were used as a complement to the IVW method. In MR analysis, the Wald ratio method was employed if the exposure was a protein and only one cis-pQTL was available. The strength of instrumental variables was evaluated using the F-statistic, with a critical value >10 indicating good strength of the selected variables. To enhance the accuracy of the MR analysis proteome for frozen shoulder causal proteins, a Bonferroni correction was applied to account for multiple tests and the results were prioritized using a critical p<3.03E-5 (0.05/1651 proteins). All MR analyses were performed using the R software package ‘TwoSampleMR’ (https://github.com/MRCIEU/TwoSampleMR).
Sensitivity analysis
Following MR analysis, a reverse MR was performed to detect potential reverse causality. In the MR analysis of frozen shoulder causal proteins, we employed the Steiger filtering method to ensure the directionality of the association between the concerned proteins and frozen shoulder, where p<0.05 indicated the absence of reverse causality. The MR-Egger regression was used to detect horizontal pleiotropy by intercept test, which was considered to be absent when p>0.05. The Cochran’s Q test was used to estimate whether there was a high degree of heterogeneity among the included instrumental variables, which was considered to be absent when p>0.05.
Colocalization analysis
Colocalization analysis provides posterior probabilities for five hypotheses regarding whether two traits share a common variant.24 In this study, we tested the posterior probability of hypothesis 4 (PPH4). If the posterior probability of sharing a causal variant (PPH4) was >0.8, the two traits were considered to be of high support for colocalization; if the posterior probability was >0.5 and <0.8, the strength of colocalization for the two traits was considered to be medium. The colocalization analysis was performed using the R software package ‘coloc’ (https://github.com/chr1swallace/coloc).
External validation and PPI network
In the MR analysis of frozen shoulder causal proteins, we externally validated the findings with different datasets. The same cis-pQTLs and the most significant cis-pQTLs from external datasets were used for the exposed proteins. The external validation data for frozen shoulder were obtained from the UK Biobank and FinnGen. We then explored the protein–protein interaction (PPI) network to confirm whether there were any known interactions between the diabetes target proteins and frozen shoulder causal proteins. The PPI analysis was performed using the Search Tool for Retrieving Interacting Gene Databases V.12.0 (https://string-db.org/).
Data and resource availability
The summary statistics for diabetes and frozen shoulder were obtained from the MRC Integrated Epidemiology Unit (https://gwas.mrcieu.ac.uk/) (online supplemental table 1). The cis-pQTL datasets for plasma proteins were retrieved from the original studies.17–23
Results
Causal relationship between diabetes and frozen shoulder
Online supplemental table 2 summarizes the results of MR analyses on diabetic confounding variables and frozen shoulder. The IVW-based MR analysis did not find a significant causal relationship between frozen shoulder and fasting insulin, glycosylated hemoglobin, or 2-hour postprandial glucose. Online supplemental figure 1 and online supplemental table 3 summarize the results of MR analyses on diabetes and frozen shoulder. IVW-based MR analysis revealed that fasting glucose (OR 1.37 (95% CI 1.11, 1.69), p=0.003), type 1 diabetes (OR 1.03 (95% CI 1.02, 1.04), p=1.04E-05), and type 2 diabetes (OR 1.05 (95% CI 1.01, 1.10), p=0.014) were causally and positively associated with frozen shoulder, which was further supported by the weighted median and MR-Egger methods. Our results are consistent with those from the UK Biobank and FinnGen data. In reverse MR, no evidence of causality was found in the reverse direction between diabetes and frozen shoulder (online supplemental table 3). Pleiotropy was detected in the analysis of type 1 diabetes (p=0.029) (online supplemental table 3). No heterogeneity was observed except for type 1 diabetes in FinnGen and type 2 diabetes in both the UK Biobank and the FinnGen data (online supplemental table 3).
Frozen shoulder causal proteins
The genetically predicted levels of four proteins were significantly associated with the risk of frozen shoulder after Bonferroni correction (p<3.03E-5), including dipeptidase 1 (DPEP1) (OR 0.91 (95% CI 0.87, 0.95), p=1.19e-05), glutathione peroxidase 7 (GPX7) (OR 0.92 (95% CI 0.89, 0.95), p=7.60e-06), parathyroid hormone-related protein (PTHLH) (OR 1.33 (95% CI 1.17, 1.51), p=9.63e-06), and secreted frizzled-related protein 4 (SFRP4) (OR 1.42 (95% CI 1.25, 1.61), p=4.06e-08) (figure 2A and online supplemental table 4). In reverse MR, no evidence of causality was found in the reverse direction between these four proteins and frozen shoulder (p>0.05), which was further supported by Steiger filtering (online supplemental figure 2 and online supplemental table 4). In colocalization analysis, SFRP4 showed high colocalization support with frozen shoulder (PPH4=0.97), PTHLH showed medium colocalization support with frozen shoulder (PPH4=0.68), while DPEP1 (PPH4=0.27) and GPX7 (PPH4=0.07) showed no significant colocalization support with frozen shoulder (figure 2B and online supplemental table 4). In addition, DPEP1 and SFRP4 were causally associated with frozen shoulder in external validation using the FinnGen outcome dataset, and GPX7 was associated with frozen shoulder in external validation using the UK Biobank outcome dataset (online supplemental figure 3 and online supplemental figure 4). However, PTHLH was only considered to be associated with frozen shoulder when the exposed protein in external validation was derived from the most significant cis-pQTL in African Americans (online supplemental figure 4).
Diabetes target proteins
In order to verify the specific mechanisms by which diabetes affects the onset and progression of frozen shoulder, we searched for a number of possible target proteins across human plasma proteins. First, we verified the causal relationship between diabetes and four frozen shoulder causal proteins (SFRP4, GPX7, DPEP1, and PTHLH). The IVW-based MR results revealed that fasting glucose, type 1 diabetes, and type 2 diabetes had no effects on frozen shoulder causal proteins (online supplemental figure 5A and online supplemental table 5); consistently, the results of colocalization analysis were not significant (figure 2B and online supplemental table 5). Second, the main pathological processes of diabetes include the occurrence of inflammation and oxidative stress.6 7 Therefore, we selected sixteen proteins related to these two processes in the human plasma proteome, among which nine proteins were identified as being involved in the pathology of diabetes (p<0.05), namely interleukin-6 receptor subunit alpha (IL6R), intercellular adhesion molecule 1 (ICAM1), interleukin-17B (IL17B), insulin receptor subunit alpha (INSR), interleukin-6 receptor subunit beta (IL6ST), C-C motif chemokine 5 (CCL5), interleukin-1 receptor type 1 (IL1R1), interleukin-1 receptor type 2 (IL1R2), and interleukin-1 receptor accessory protein (IL1RAP) (online supplemental figure 3B and online supplemental table 6). In reverse MR analysis, IL6R (p=0.003) was found to be associated with type 1 diabetes, while both ICAM1 (p=0.006) and IL1RAP (p=0.006) were found to be associated with type 2 diabetes (online supplemental table 6). Except for INSR and fasting glucose, no horizontal pleiotropy was detected during the MR-Egger intercept analysis (online supplemental table 6). The heterogeneity among diabetes target proteins is shown in online supplemental table 6. No significant colocalization of genetic variants was found between these proteins and diabetes (figure 2B and online supplemental table 6). The results of MR analysis for the remaining seven proteins are shown in online supplemental table 7.
Causal relationship between diabetes target proteins and frozen shoulder causal proteins
Further, we analyzed the causal relationship between the nine diabetes target proteins and the four frozen shoulder causal proteins. MR analysis revealed that IL6R was causally and negatively correlated with both PTHLH (OR 0.99 (95% CI 0.97, 1.00), p=0.064) and GPX7 (OR 0.98 (95% CI 0.97, 1.00), p=0.028); INSR was causally and positively correlated with GPX7 (OR 1.23 (95% CI 1.00, 1.52), p=0.049); and IL1RAP was causally and negatively correlated with PTHLH (OR 0.99 (95% CI 0.98, 1.00), p=0.057) (figure 3A and online supplemental table 8). Reverse MR analysis revealed that SFRP4 was causally correlated with INSR (p=0.025) and IL6R (p=0.042); GPX7 was causally correlated with ICAM1 (p=0.045) and IL1R2 (p=0.056); DPEP1 was causally correlated with IL6ST (p=0.001), INSR (p=5.69E-06), IL1R1 (p=0.020), and IL6R (p=0.035); and PTHLH was causally correlated with IL1R2 (p=0.007) (figure 3B and online supplemental table 8). In PPI network analysis, IL6R-PTHLH and ICAM1-GPX7 were shown to have a known interaction, with a composite score of 0.420 and 0.501, respectively (figure 4A).
Conclusions
Overall, this MR study mainly consisted of four steps. First, three diabetic confounders that could contribute to the development of frozen shoulder were excluded: fasting insulin, glycosylated hemoglobin, and 2-hour postprandial glucose. We then determined that fasting glucose, type 1 diabetes, and type 2 diabetes were all positively associated with frozen shoulder. Although pleiotropy was detected in one of the included datasets for type 1 diabetes, it had no impact on the overall findings. Besides, the inclusion of a large-sized sample in our GWAS data resulted in unavoidable heterogeneity. Nevertheless, our results aligned well with previous epidemiological studies.3–5 Second, we screened four frozen shoulder causal proteins, all of which passed reverse MR analysis and further validation using external datasets. The colocalization analysis results of SFRP4 and PTHLH suggested a strong correlation with frozen shoulder, as they shared common genetic variants. Notably, though low PPH4 values were observed in the colocalization analysis of GPX7 and DPEP1, their correlation with frozen shoulder could not be ruled out. We speculated that the lower PPH4 values might be attributed to limited power24 or due to the fact that these proteins might not act directly on frozen shoulder, but indirectly through other proteins. Third, we identified nine diabetes target proteins, which were supported by reverse MR analysis and heterogeneity analysis. In the pleiotropy test, only fasting glucose was found to be in pleiotropy with INSR. The PPH4 values derived from colocalization analysis did not reject the results of MR analysis, as the majority of pleiotropy tests failed. Finally, we found that INSR-GPX7, IL6R-GPX7, IL6R-PTHLH, and IL1RAP-PTHLH might be pathological pathways contributing to the development of frozen shoulder in individuals suffering from diabetes. The PPI network analysis revealed a known interaction between IL6R-PTHLH and ICAM1-GPX7, but no interactions have been reported between the remaining proteins. Furthermore, the results of reverse MR analysis were consistent with the PPI network interaction. In conclusion, INSR, IL6R, IL1RAP, GPX7, and PTHLH are involved in the pathological processes of diabetes and can influence the pathogenesis of frozen shoulder (figure 4B).
Several epidemiological studies have supported a correlation between diabetes, especially type 1 diabetes, and frozen shoulder.3–5 Patients with type 1 diabetes have a significantly higher incidence of frozen shoulder, with a lifetime prevalence up to 76%.10 Consistent with previous research, our MR analysis also indicated a robust causal positive correlation between type 1 diabetes and frozen shoulder, with three separate frozen shoulder datasets generating comparable results. Type 1 diabetes has a longer disease duration than type 2 diabetes, resulting in more severe inflammatory responses and higher accumulation of AGEs in the shoulder joint.9 Despite extensive epidemiological literature on the association between diabetes and frozen shoulder, the underlying pathology of this relationship has not been thoroughly clarified. At present, it is commonly argued that individuals with diabetes suffer from diminished insulin secretion due to β-cell dysfunction, which impairs the body’s glucose regulation.25 When the blood glucose level increases excessively, oxidative stress will take place, leading to the production of inflammatory factors.6 7 The joint effects of oxidative stress and inflammatory factors can generate a combined impact on the body, resulting in mitochondrial dysfunction, endoplasmic reticulum stress, and consequently, accumulation of AGEs.9 10 25 The aggregation of AGEs, immune cells, inflammatory factors, and matrix metalloproteinases in the shoulder joint will further induce fibroblast activation, dysregulation of collagen synthesis, and neo-angiogenesis, ultimately leading to inflammatory responses and scarring contracture of the shoulder joint.10 25 To clarify the impact of diabetes on the pathogenesis of frozen shoulder, we conducted MR analysis at the protein level, and the results suggested that INSR, IL6R, IL1RAP, GPX7, and PTHLH might play a role in this pathological process. Nevertheless, the diabetes-induced pathogenesis of frozen shoulder is multifaceted, requiring continuous scrutiny of various proteins and molecular pathways.
In diabetic individuals, the elevation of the blood glucose level will elicit the generation of inflammatory mediators that aggregate around the shoulder joint, initiating an inflammatory reaction; this process mainly involves interleukins such as IL1, IL6, and IL17.6 7 10 25 IL1 consists of 21 different molecular systems, among which IL1R1 and IL1RAP are responsible for generating transmembrane signaling complexes that initiate IL1-dependent intracellular signaling.26 IL1RAP is closely associated with diabetes, and previous studies have shown that IL1RAP expression is identified in gene expression changes in patients with impaired glucose regulation, type 1 diabetes and type 2 diabetes.27 28 IL6 expression was reported to significantly increase in synovial fibroblasts of the shoulder joint in frozen shoulder patients versus healthy controls.29 IL6R acts as a receptor for IL6 and has been demonstrated to contribute to the progression of diabetic nephropathy.30 PTHLH is a neuroendocrine peptide that plays a critical role in regulating the growth, development, migration, differentiation, and survival of cells and organs, as well as epithelial calcium ion transport. It has been demonstrated to control the endochondral bone formation and epithelial–mesenchymal interactions during mammary gland and tooth formation.31 An in vitro study revealed that elevated levels of inflammatory mediators IL1 and IL6 would impede PTH secretion and mRNA expression in parathyroid chief cells.32 Our study provided reliable evidence that IL1RAP and IL6R were negatively correlated with PTHLH (figure 4B and online supplemental table 8). However, an increased production of PTHLH was observed in the synovial fibroblasts from patients with rheumatoid arthritis cultured in vitro and treated with IL1α.33 The discrepancy above may stem from differences in species or diseases. Both IL1R1 and IL1R2 were found to be positively correlated with PTHLH in our results (online supplemental table 8). GPPX7 is a crucial antioxidant enzyme present in human beings, and its function is to restrain oxidative stress in diverse cells and to prevent oxidative DNA damage and double-strand breakage.34 Our research revealed a negative correlation between GPX7 and frozen shoulder, implying that GPX7 can be potentially used to prevent and treat frozen shoulder by reducing diabetes-induced oxidative stress (figure 4B). A prior case–control study confirmed that there was a negative correlation between the IL6 level and glutathione peroxidase.35 In our study, IL6R was found to be negatively correlated with GPX7 (figure 4B). ICAM1 is a transmembrane protein that can be glycosylated. It is essential for efficient leukocyte function and can promptly respond to proinflammatory cytokines.36 Studies have shown that ICAM1 plays a significant role in diabetes, and its knockout can prevent diabetes in NOD mice.36 It was reported that ICAM1 expression was notably elevated in the synovial fluid and serum of patients with frozen shoulder versus healthy controls, similar to the situation in patients with diabetes.37 38 Based on our findings, there is a strong relationship between ICAM1 and GPX7 (figure 3B and figure 4A). We speculated that ICAM1 might be indirectly involved in the process by which diabetes affects the pathogenesis of frozen shoulder through GPX7. Nevertheless, the precise mechanism can be very complex and requires further research. The SFRP4 protein is an extracellular regulator of the Wnt pathway.39 Studies have shown that SFRP4 can reduce insulin secretion and serves as a biomarker of islet dysfunction in type 2 diabetes.39 40 Similarly, our reverse MR analysis demonstrated an association between SFRP4 and type 2 diabetes (online supplemental table 5). In addition, SFRP4 has been suggested as a potential causal gene for frozen shoulder,4 which was confirmed by our MR analysis (figure 2A and online supplemental table 4). Moreover, our colocalization analysis revealed a strong correlation between SFRP4 and frozen shoulder, suggesting the presence of a common genetic variant (figure 2B). Based on these findings, it is possible that SFRP4 is a shared causal factor for both diabetes and frozen shoulder.
To our best knowledge, the application of MR in analyzing and exploring the pathological processes between diseases has not been previously reported, making our study the first attempt of its kind. The MR analysis method used in our study minimized the bias due to confounding and reverse causation to the best extent, thereby strengthening causal inference. Furthermore, we utilized colocalization analysis to further authenticate the protein disease links. However, our study is also subjected to several limitations. First, the plasma proteins we selected as exposures included only proteins with cis-pQTLs that met our selection criteria. Despite the possibility that the selection of cis-pQTLs can help avoid pleiotropy, the small number of genetic instruments restricted the use of other sensitivity analysis techniques. However, we have selected alternative methods such as reverse MR, external validation, and colocalization analysis to ensure the reliability of the results. Second, our screening for diabetes target proteins only included proteins such as oxidative stress and inflammatory factors but not the entire plasma proteome, making further screening of possible target proteins necessary in future research. Third, our MR analysis was performed based on human plasma proteins, but proteins in the brain tissue and cerebrospinal fluid were not included due to the presence of blood–brain barrier. Therefore, further investigations on proteins in the brain tissue and cerebrospinal fluid are needed following the development of proteomics. Finally, most of our selected GWAS datasets were derived from European populations, which may lack population representativeness. If the GWAS datasets set for non-European populations become sufficient in the future, further verification of the applicability of our findings to different populations will be necessary.
In conclusion, our study has preliminarily elucidated the pathological mechanisms by which diabetes affects the development of frozen shoulder, providing an important reference for preventing the condition in patients with diabetes.
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
Not applicable.
Acknowledgments
The authors acknowledge the participants and investigators of the IEU OpenGWAS project, UK Biobank, and FinnGen. The authors thank all of the investigators who provided these data to support this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors KC, TT, and LL conceived and designed the study. KC, TT, PG, and XF collected data. KC, TT, WJ, and ZL analyzed data. KC, TT, KT, and PO wrote the first draft of the manuscript. All authors revised the manuscript and gave final approval to the submitted versions. LL is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding This research was funded in part by the Key project of Hunan Provincial Health Commission number 20201902.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.