Causal Relationship between Autoimmune Diseases and Risk of Osteonecrosis in European Population: A Two-Sample Mendelian Randomization Study

doi:10.21203/rs.3.rs-3981307/v1

Background: Prior epidemiological and observational studies have identified a series of autoimmune diseases (ADs) linked with osteonecrosis (ON), although potential confounding variables have not been fully accounted for. Furthermore, comprehensive studies elucidating the causal relationship between autoimmune diseases and osteonecrosis are lacking. Objective: This research endeavored to clarify the causal relationship between a myriad of autoimmune diseases and ON through the application of a Mendelian randomization (MR) methodology.

Methods: To explore the causal relationship between ADs and ON, Single Nucleotide Polymorphism (SNP) loci, firmly correlated with 12 autoimmune diseases including, but not limited to, type 1 diabetes mellitus (T1DM), rheumatoid arthritis (RA), ankylosing spondylitis (AS), and systemic lupus erythematosus (SLE), as well as osteonecrosis, were chosen as instrumental variables (IV). These were sourced from both the Finnish database (Finn-Gen) and the Genome-Wide Association Study (GWAS) public database. Various evaluation methods, such as inverse variance weighted (IVW), Simple Mode, MR-Egger, and weighted median and model, were employed to assess the causal relationship between 12 ADs and ON. Inverse variance weighted (IVW) is a widely utilized statistical method in genetic association studies for estimating and integrating genetic effects. It is particularly relevant for estimating single genetic effects across multiple genetic variant loci. Owing to its robust causal inference capabilities, IVW was employed as the primary analytical method in this MR study. For sensitivity analysis, MR-PRESSO, inverse variance weighted, MR-Egger, and a battery of other methods were utilized.

Results: Post-MR analysis, IVW findings indicated that T1DM (OR=1.062; P=0.007), RA (OR=1.248; P=7.19×10^-5), and AS (OR=8.50×10¹⁰; P=0.034) served as risk factors for osteonecrosis. Meanwhile, other autoimmune diseases such as myasthenia gravis (MG), Crohn's disease (CD), ulcerative colitis (UC), and psoriatic arthritis (PsA) did not exhibit a causal link with osteonecrosis. Additionally, the results of the reverse MR analysis failed to reveal any significant causal relationship between ON and ADs.

Conclusion: This research substantiated that T1DM, AS, and RA are prominent risk factors for osteonecrosis in European Population, utilizing a two-sample MR approach for both directions.

Type 1 diabetes

rheumatoid

ankylosing spondylitis

osteonecrosis

Mendelian randomization

Osteonecrosis (ON) is characterized by the necrosis of living tissue components within the human skeleton. This condition can be seen in various anatomical locations, such as the wrist bone, lunar bone, tibial tuberosity, metatarsal bone, foot navicular, heel bone, and femur. Clinically, femoral head necrosis is the most prevalent manifestation of ON. This challenging-to-treat disease can lead to osteoporosis, joint pain, and restricted movement in patients [1], significantly impacting their quality of life and imposing a substantial economic burden on patients' families. Given the high health-economic implications of ON, it is crucial from a societal perspective to investigate the risk factors associated with ON occurrence.

The etiology of osteonecrosis is multifaceted, encompassing both traumatic and non-traumatic factors. Trauma is the most common cause, contributing to over 20% of cases [2]. Non-traumatic factors include chronic or heavy use of glucocorticoids, alcoholism, decompression sickness, hypertension, atherosclerosis, and hematologic disorders [3–7]. In recent years, several studies have suggested a potential link between osteonecrosis and autoimmune diseases [8–9]; however, the potential influence of confounding factors cannot be dismissed. Understanding the causal relationship between osteonecrosis and autoimmune diseases can offer valuable insights into the pathogenesis of osteonecrosis, and importantly, guide preventive strategies for patients with autoimmune diseases. Therefore, there is a need for more reliable methods to assess the causal relationship between autoimmune diseases and osteonecrosis.

Mendelian randomization (MR) is a frequently utilized method in epidemiology to evaluate the relationship between exposure factors and disease risk [10]. MR is often considered the "most natural" randomized controlled trial (RCT) design that minimizes regression bias [11]. MR is impervious to external environmental factors (e.g., family income, parental occupation, community environment, etc.) present since birth and can effectively address endogeneity problems caused by omitted variables and reverse causality associations. Moreover, genetic data can be accurately measured with consistently lower costs [12–13].

Thus, in this study, we employed a bidirectional two-sample MR analysis to investigate the causal relationship between ON and various types of autoimmune diseases, including those of skin, connective tissue, hematologic, and endocrine types. The diseases analyzed include type 1 diabetes mellitus (T1DM), rheumatoid arthritis (RA), ankylosing spondylitis (AS), systemic lupus erythematosus (SLE), myasthenia gravis (MG), Crohn's disease (CD), ulcerative colitis (UC), psoriasis arthritic (PsA), juvenile idiopathic arthritis (JIA), primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), and pernicious anaemia (PA).

The current study employed a two-way two-sample Mendelian randomization (MR) analysis, considering both autoimmune disease and ON as exposure or outcome. We utilized Single Nucleotide Polymorphisms (SNPs) as instrumental variables (IVs) to assess the causal effect of exposure on outcome. The essential conditions to meet for the IVs are: (1) they are associated with exposure, (2) they are not associated with any confounders of the exposure-outcome association, and (3) they do not affect the outcome except possibly through association with exposure [14]. We selected SNPs with strong associations to both autoimmune diseases and ON as IVs according to our screening criteria and conducted a series of sensitivity analyses for these strongly associated SNPs. The study's flowchart is illustrated in Fig. 1.

3.1 Data Collection

We acquired data pertaining to 12 autoimmune diseases and osteonecrosis from the most extensive publicly available Genome-Wide Association Study (GWAS) suitable for bidirectional two-sample Mendelian Randomization (MR) analysis. From the Finn-Gen database, we extracted summary statistics for Type 1 Diabetes Mellitus (T1DM) which included 183,185 controls of European ancestry and 5,928 T1DM cases. Summary statistics for Systemic Lupus Erythematosus (SLE) were procured from the IEU Open GWAS database, including 1,311 SLE cases and 1,783 controls of European origin. Additionally, the GWAS database provided summary statistics for rheumatoid arthritis, which consisted of 6,329 cases and 212,463 controls of European origin. Table 1 presents the other disease-related data sources. This study did not necessitate ethical approval.

Table 1

The sample information of autoimmune diseases used in this article.
Autoimmune Disease	Number of SNPs	n_cases	n_controls	Population	Data sources
T1DM	16,380,008	5,928	183,185	European	FinnGen
RA	16,380,466	6,329	212,463	European	FinnGen
AS	9,851,867	1,296	461,637	European	UK Biobank
SLE	489,877	1,311	1,783	European	Hom G et al
MG	10,894,596	1,452	335,707	European	UK Biobank
CD	10,894,596	732	336,467	European	UK Biobank
UC	9,851,867	1,987	461,023	European	UK Biobank
PsA	16,380,462	1,637	212,242	European	FinnGen
JIA	103,767	2,816	13,056	European	Hinks A,et al
PBC	119,756	2,861	8,514	European	European Molecular Biology Laboratory
PSC	7,891,603	2,871	12,019	European	Ji SG,et al
PA	10,894,596	1,020	336,139	European	UK Biobank
Note: Type 1 diabetes mellitus (T1DM), rheumatoid arthritis (RA), ankylosing spondylitis (AS), systemic lupus erythematosus (SLE), myasthenia gravis (MG), Crohn's disease (CD), ulcerative colitis (UC), Psoriatic arthritis (PsA), juvenile idiopathic arthritis (JIA), primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), Pernicious anaemia (PA).

3.2 Screening of instrumental variables

We retrieved Single Nucleotide Polymorphisms (SNPs) associated with autoimmune diseases and ON. To meet the first hypothesis, we selected SNPs at a threshold of genome-wide significance (P < 5×10^− 5-5×10^− 8). We evaluated the strength of association of IVs by calculating the F statistic, excluding SNPs with F values less than 10 [15].

$$F={\beta }_{\text{e}\text{x}\text{p}\text{o}\text{s}\text{u}\text{r}\text{e}}^{2}/{SE}_{\text{e}\text{x}\text{p}\text{o}\text{s}\text{u}\text{r}\text{e}}^{2}$$

Here, β denotes the effect size of SNP on exposure, and SE denotes the standard error of β.

We set the chain imbalance threshold to R² = 0.05 and the clustering window size to 10M, with the echo sequences being removed [16]. In examining the relationship between ADs and osteonecrosis, factors such as hypertension, hyperlipidemia, and glucocorticoid use are likely major confounders. We utilized PhenoScanner (version V2), a comprehensive tool for exploring human genotype-phenotype associations, to identify SNPs and their surrogate traits. SNPs associated with these traits at thresholds of p < 1×10^− 5 and r² > 0.80 were excluded [17]. Using PhenoScanner, we excluded confounders and outcome-related SNPs linked with exposure-outcome associations to meet the conditions of hypotheses 2 and 3 for IVs. This strategy was similarly applied for reverse MR.

3.3 Statistical analysis

Our study used MR analysis to examine the causal relationship between autoimmune disease and ON. The analysis employed five tests: inverse variance weighted (IVW), Simple Mode, MR-Egger, Weighted Median (WM), and Weighted Model. The Inverse Variance Weighted (IVW) method, a prevalent statistical approach in genetic association studies for estimating and integrating genetic effects, is particularly suitable for assessing single genetic effects across multiple genetic variant loci. Due to its robust capability for causal inference, IVW was chosen as the main analytical method in this MR study. The IVW method is characterized by regression without an intercept term, and it is weighted with the inverse of the outcome variance (SE²) [18]. The p-value from the regression determines the result of the MR analysis (p < 0.05 is considered statistically significant). When applying the IVW method, it is essential to ensure that the SNPs are not multivariate to prevent significant bias in the results. The MR-Egger's hypothesis introduces an intercept term to evaluate the multiplicity of effects. If the intercept term approximates 0, the MR-Egger regression model closely aligns with IVW [19]. The weighted median test suggests that causal effects can be consistently estimated if a valid IV provides at least 50% of the information in the analysis [20].

3.4 Sensitivity Analysis

We used Cochrane's Q test to appraise the heterogeneity among SNPs associated with autoimmune diseases. If heterogeneity was present among the SNPs (P < 0.05), MR-PRESSO was utilized to discard outlier SNPs. The leave-one-out sensitivity test primarily computed the MR results of the remaining IVs after sequential elimination of each IV. If the MR results of the remaining IVs significantly deviated from the total results upon the removal of one IV, the MR results were deemed sensitive to that particular IV. We utilized the Pleiotropy test to examine the presence of horizontal pleiotropy across multiple IVs. If the intercept term of the IVW and MR-Egger significantly deviated from 0, the MR-Egger method indicated the existence of horizontal pleiotropy [21]. Besides, we frequently employed MR-PRESSO to test for horizontal pleiotropy. In addition, we visualized our data by creating scatter plots, funnel plots, and forest plots.

3.5 Co-localization Analysis

Co-localization Analysis is designed to assess the likelihood that two traits share causal genetic variations within a specific gene region. It represents an advanced fine-localization technique applicable to two or more traits [22]. In this study, we utilized a Bayesian inference-based Co-localization approach, known as the COLOC method [23], to quantify the probability of two traits sharing causal genetic variations. The COLOC method operates under the assumption that there is, at most, one causal genetic variant per trait in a given region. It adheres to the following five assumptions: H0, indicating that neither trait is genetically linked in the region; H1, where only the first trait is linked; H2, where only the second trait is linked; H3, suggesting both traits are linked but with distinct causal variants; and H4, proposing both traits are linked and share an identical causal variant. The COLOC method employs an approximate Bayesian factorial algorithm to estimate the posterior probability (PP) for each of these hypotheses, labeled as PP0, PP1, PP2, PP3, and PP4. The total of PP0 through PP4 equals 1. A high PP4 value (e.g., PP4 > 90%) implies that the first trait shares the same causal genetic variance with the second trait in the gene region. Additionally, this method identifies the SNP with the highest posterior probability, deeming it the most probable causal genetic variant.

In the course of performing the Mendelian Randomization (MR) analysis, with autoimmune diseases as the exposure and osteonecrosis as the outcome, a range of conditions were employed to screen instrumental variables (IVs) to derive Single Nucleotide Polymorphisms (SNPs) strongly associated with various autoimmune diseases. (For detailed SNPs, refer to Supplement Table 1). Post applying methods like IVW, MR-Egger, and others, the results from IVW indicated that Type 1 Diabetes Mellitus (T1DM), Rheumatoid Arthritis (RA), and Ankylosing Spondylitis (AS) acted as risk factors for ON.

SNPs associated with hypertension, hyperlipidemia, and glucocorticoids were discarded using the PhenoScanner database (T1DM-related SNPs rs149340530, rs3184504, and rs9275318 were removed; RA-related SNP rs7137828 was removed; AS-related SNPs rs2515919, rs3104415, and rs9266639 were removed). Following the exclusion of SNPs associated with confounders and subsequent causality tests, T1DM (Odds Ratio (OR) = 1.062; P = 0.007), RA (OR = 1.248; P = 7.19×10^− 5), and AS (OR = 8.50×10¹⁰; P = 0.034) continued to act as risk factors for osteonecrosis. However, no causal association with osteonecrosis was found for the remaining 9 autoimmune diseases (Fig. 2 depicts the forest plot). The derived F-statistic for IV was > 10, signifying that the outcome was not biased [24]. Table 2 presents detailed information. The scatter plot indicates that the five methods fit in the same direction, representing that the results of this Mendelian randomization experiment are relatively robust (Fig. 3 depicts the scatter plot).

Table 2

Results of MR analysis of autoimmune diseases and risk of osteonecrosis
Exposure		SNP	Inverse variance weighted			MR Egger				F-statistics
Exposure		SNP	Pval	OR	CI(95%)	Pval		OR	CI(95%)	F-statistics
T1DM	118		0.007	1.062	1.017–1.110		0.144	1.066	0.979–1.160	141.0
RA	24		7.19×10^− 5	1.248	1.119–1.393		0.040	1.232	1.022–1.487	116.6
AS	11		0.034	8.50×10¹⁰	6.65–1.09×10²¹		0.165	8.95×10¹⁷	4.82×10^− 6-1.66×10⁴⁰	216.8
SLE	6		0.519	0.911	0.687–1.209		0.112	0.276	0.080–0.957	36.3
MG	88		0.355	0.049	8.41×10^− 5-29.1		0.481	0.009	1.68×10^− 8-4.52×10³	166.9
CD	3		0.990	2.305	1.98×10^− 58-2.67×10⁵⁸		0.966	5.11×10⁶	4.89×10^− 244-5.63×10²⁵⁶	46.9
UC	8		0.317	6.83×10^− 11	2.93×10^− 35-5.72×10⁹		0.529	7.81×10³³	3.36×10^− 66-1.82×10¹³³	43.3
PsA	20		0.069	0.008	5.24×10^− 5-1.459		0.047	2.59×10^− 4	9.25×10^− 8-0.729	165.3
JIA	17		0.447	0.975	0.915–1.040		0.652	0.963	0.822–1.129	68.8
PBC	40		0.033	1.099	1.008–1.198		0.834	1.021	0.841–1.239	64.6
PSC	61		0.251	0.971	0.925–1.020		0.661	1.022	0.926–1.128	110.2
PA	51		0.693	0.017	3.67×10^− 11-8.61×10⁶		0.409	5.958×10⁴	3.67×10^− 7-6.917×10¹⁵	24.3
Note: OR stands for odds ratio, CI stands for confidence interval, MR stands for Mendelian randomization, and SNP stands for single nucleotide polymorphism

Furthermore, we performed sensitivity analyses to identify any pleiotropic effects in the IVs. The results from Cochrane's Q test indicated no heterogeneity in the IVs for T1DM (MR-Egger, Q_pval = 0.256; IVW, Q_pval = 0.278), RA (MR-Egger, Q_pval = 0.874; IVW, Q_pval = 0.903), and AS (MR-Egger, Q_pval = 0.301; IVW, Q_pval = 0.345). The Pleiotropy Test results demonstrated no horizontal pleiotropy for SNPs associated with T1DM (P = 0.934), RA (P = 0.873), and AS (P = 0.525). Similarly, MR-PRESSO calculations yielded T1DM (P = 0.285), RA (P = 0.933), and AS (P = 0.525), these results are displayed Supplement Table 2. Hence, the pleiotropy test results confirmed no violation of the MR core hypothesis for the SNPs associated with T1DM, RA, and AS. The funnel plot also shows that the distribution of causal effect estimates for IV is largely symmetric, with no overall heterogeneity (Fig. 4 depicts the funnel plot).

The results of the "Leave-one-out sensitivity test" showed the error line estimates positioned to the right of 0, signifying that no individual SNP exerted a substantial impact on the outcomes (Detailed information can be found in the supplementary document).

Additionally, the reverse MR analysis results displayed no detectable causal relationship between osteonecrosis and autoimmune diseases (Supplement Tables 3, 4 and 5). A reverse MR study was conducted with osteonecrosis as exposure and 12 ADs as endpoints, but no valid IVs could be extracted from the GWAS data related to JIA and PBC, so they were excluded, but to avoid the problem of endogeneity arising from reverse causation, we also supplemented with the Steiger Test, which was used to determine whether the unidirectional Mendelian randomization experiments were correctly oriented. The results of the Steiger Test showed that the p-value for both autoimmune diseases was less than 0.05, therefore the directionality of the unidirectional Mendelian randomization experiments for both autoimmune diseases was correct, which ensured that even without the reverse Mendelian experiments there was no impact on the results of this experiment. The results of the Steiger Test can be seen in Supplementary Table 6.

In response to the MR analysis that established a significant causal relationship between three autoimmune diseases, T1DM, RA and AS, and osteonecrosis, we conducted a Co-localization Analysis. This analysis aimed to investigate the possibility of pleiotropy between these conditions. The results indicated that the posterior probabilities, calculated using the COLOC method, were all below 90% for the gene regions in question. This finding does not align with hypothesis H4, thereby suggesting the absence of pleiotropy in the MR experiments involving T1DM, RA, AS, and osteonecrosis. Detailed information is available in Additional file 2.

Osteonecrosis, a detrimental degenerative disorder, leads to the death of bone and bone marrow components owing to insufficient blood supply [25]. This pathology, in turn, induces structural modifications in bone and joint functionality, severely impacting the patients' quality of life [26]. Predominantly afflicting young adults with an average onset age of 38, the disease witnesses 10,000 to 20,000 new cases annually in the United States [27] and 100,000 to 200,000 new cases in China [28]. A significant proportion of these cases constitutes non-traumatic osteonecrosis triggered by alcoholism and glucocorticoids. Nonetheless, emerging evidence suggests a heightened risk of osteonecrosis in individuals with autoimmune diseases. Inflammation plays an important role in autoimmune diseases, is identified as a crucial factor in osteonecrosis development [29]. Firstly, the inflammatory process in autoimmune diseases initiates various immune responses, including cytokine release and inflammatory mediator production. These immune components can directly or indirectly impair bone cell function, leading to conditions like osteoporosis and osteonecrosis. For instance, in rheumatoid arthritis, excessive TNF-α is linked to increased bone resorption and reduced bone formation [30]. Moreover, cytokines like IL-1 and IL-6 are believed to be instrumental in bone destruction associated with autoimmune diseases [31]. Secondly, inflammation can cause abnormal vascular endothelial function and hinder blood circulation, further exacerbating the risk of osteonecrosis. It leads to vascular endothelium damage and inflammatory cell infiltration in the vascular wall, disrupting normal blood vessel function. This vascular impairment results in insufficient bone tissue blood supply, accelerating osteoclastic necrosis [32]. Furthermore, the inflammatory process impacts bone cell differentiation and function. In autoimmune diseases, inflammatory cell infiltration and mediator release hamper osteoblast bone-forming activity while promoting bone resorption, creating an imbalance in bone metabolism that culminates in osteoporosis and osteonecrosis [33]. But in the early stages of osteonecrosis caused by autoimmune diseases, optimal treatment often gets delayed owing to the difficulty in diagnosing bone damage via standard X-rays. Consequently, recognizing autoimmune diseases linked with non-traumatic osteonecrosis and implementing early clinical interventions can significantly enhance the prognosis of non-traumatic osteonecrosis, eventually preserving the patient's femoral head or postponing the need for arthroplasty. Employing Mendelian randomization analysis, this study explored the relationship between 12 common autoimmune diseases and osteonecrosis, unearthing strong correlations with Type 1 Diabetes Mellitus (T1DM), Rheumatoid Arthritis (RA), and Ankylosing Spondylitis (AS), but not with the remaining 9 autoimmune diseases.

Type I Diabetes Mellitus (T1DM) is an autoimmune condition potentially triggered by genetic factors or disturbances in the autoimmune system. This disruption leads to a diminished population of insulin-producing pancreatic islet β-cells, necessitating patients to undergo insulin injections as a treatment strategy [34]. Individuals suffering from T1DM might exhibit irregular expression of autoimmune antibodies such as Glutamic Acid Decarboxylase antibodies (GAD), Insulin Cell antibodies (ICA), and Insulin Autoantibodies (IAA) at an abnormally high rate. These dysregulated autoantibodies can inflict damage on the insulin-secreting beta cells within human pancreatic islets, consequently impairing their insulin secretion capabilities.

In recent times, several studies have pointed towards T1DM as a potential contributor to osteonecrosis. Seah KT and associates proposed that microangiopathy, a condition prevalent in T1DM, could serve as a causative factor for osteonecrosis [35]. Additionally, Samarasinghe YP suggested a multifaceted involvement of factors in the pathogenesis, including diabetes-associated dyslipidemia, chronic renal failure (potentially leading to elevated PTH levels that cause cartilage subchondral bone resorption), and heightened coagulation capacity due to nephrotic syndrome [36]. Nevertheless, the precise causal association between T1DM and osteonecrosis calls for further exploration.

Moreover, researchers J Chen, Y Li, L Yang, M Taheri et al [37–40]. observed a correlation between the emergence of T1DM and the expression of circular RNA (circRNA). A foundational study by Kuang et al. revealed that circRNA ubiquitin-specific protease 45 could elevate PTEN expression by sequestering mir175-5p, thereby inhibiting the protein kinase B pathway and regulating bone volume in steroid-induced femoral head necrosis in rats [41]. These findings align with our results derived from Mendelian randomization analysis applied to T1DM and osteonecrosis, indicating a heightened risk of osteonecrosis in individuals with T1DM. Notwithstanding, these researchers only explored the co-morbidity between T1DM and osteonecrosis, without solidifying the causal link between the two. This present study offers a valuable extension to the prior investigations, presenting causative evidence for the observed associations.

In the past few years, international discourse has delved into the pathogenesis of osteonecrosis, especially with respect to glucocorticoid-induced osteonecrosis, resulting in a plethora of hypotheses. However, few studies have examined osteonecrosis of the femoral head stemming from Rheumatoid Arthritis (RA). Zabinski and colleagues conducted a comprehensive investigation into the disease process, treatment, and pathology of femoral head specimens from 545 patients undergoing hip replacement (of whom 507 had RA). They found that osteonecrosis afflicted 66 patients (12%), with the incidence amongst RA patients standing at 13%. Furthermore, it was observed that a significant majority of patients with ischemic (81%) and degenerative (68%) osteonecrosis had a history of glucocorticoid usage. This percentage was strikingly higher than those not diagnosed with femoral head necrosis (33%). The study concluded that low-dose glucocorticoids did not offer protection against femoral head necrosis. Moreover, the 3% of patients with femoral head necrosis without a history of glucocorticoid usage may be related to inflammatory changes inherent in RA [42].

Mylov et al. performed a broad analysis of 2392 rheumatological patients attending the Tula Regional Hospital in Russia between 2002 and 2003. They identified 71 patients afflicted with femoral head necrosis. By utilizing clinical data, laboratory tests, and X-ray results, the researchers undertook a correlational analysis to determine risk factors for the development of femoral head necrosis. They identified the presence of high inflammatory indexes combined with anemia, elevated blood lipid levels, and activation of the coagulation system as significant risk factors [43]. Hence, it becomes pertinent to explore if RA itself can have a substantial impact on osteonecrosis. Firestein et al. discovered an amplified inflammatory response in RA patients, characterized by elevated levels of interleukin-1 (IL-1) and tumor necrosis factor-alpha (TNF-α) [44]. These inflammatory factors stimulate inflammatory cells and the destructive proliferation of follicular fibroblasts (pannus), resulting in the degradation of articular cartilage and bone. Additionally, they enhance the activity of bone-resorbing cells (e.g., macrophages and osteoclasts), increasing bone resorption [45]. Concurrently, this inflammatory response inhibits the function of bone-forming cells (e.g., osteoblasts), reducing the repair and regeneration capacity of bone tissue [46]. Regrettably, research in this area is scant. In our study, we employed Mendelian randomization analysis to probe the causal link between RA and osteonecrosis within the GWAS database. The PhenoScanner database was used to exclude RA patients with a history of glucocorticoid usage. According to Fig. 3, a robust correlation persists between RA and osteonecrosis, providing a solid theoretical foundation for future biomolecular studies on these diseases.

International research on the intersection of ankylosing spondylitis and osteonecrosis remains inchoate. Zang et al., Chinese researchers, examined 11 patients diagnosed with both ankylosing spondylitis and osteonecrosis of the femoral head. Only three of these patients had a history of glucocorticoid usage, suggesting that glucocorticoids might not play a substantial role in the development of osteonecrosis in this disease type [47]. Jiang et al. identified SAA1, TUBA8, and monocyte dysregulation as key factors in ankylosing spondylitis-associated femoral head necrosis by analyzing ligaments and patient blood samples [48]. Patients with AS often experience increased spinal canal pressure, potentially leading to reduced blood supply and osteonecrosis [49]. Moreover, osteoarthritis and cartilage damage, prevalent in AS patients, may also promote osteonecrosis development. These conditions trigger apoptosis, extracellular matrix degradation, and bone marrow mesenchymal stromal cell dysfunction, culminating in bone destruction and osteonecrosis [50]. Glucocorticoid usage in patients with RA was also excluded in this study to mitigate bias. The derived results align with the observations of these scholars. Nevertheless, the exact mechanism underpinning osteonecrosis in ankylosing spondylitis remains enigmatic and necessitates future in-depth exploration.

Mendelian randomization, from a genetic standpoint, serves as a tool for epidemiological data analysis, aiming to assess causality between exposures of interest and outcomes. It uses genetic variation as an instrumental variable, capitalizing on the random nature of genetic variation to minimize confounding factors, thus providing an effective method for causal inference [51]. In this study, a two-way Mendelian randomization analysis, coupled with a Co-localization experiment, was conducted to investigate the causal relationship between autoimmune diseases and osteonecrosis in 12. Results from the two-way Mendelian randomization experiment suggest that autoimmune diseases elevate the risk of osteonecrosis, while osteonecrosis does not reciprocally increase the risk of autoimmune diseases. From a genetic perspective, autoimmune diseases are influenced by both genetic and environmental factors. Genetics play a crucial role in the onset and progression of autoimmune diseases [52]. Certain genetic variants are linked to an increased risk of autoimmune diseases and may also correlate with the development of osteonecrosis [53]. For instance, in conditions like ankylosing spondylitis and rheumatoid arthritis, The RANKL/RANK/OPG system, regulated by genes such as TNFSF11 (Tumor Necrosis Factor Superfamily Member 11), TNFRSF11A (Tumor Necrosis Factor Receptor Superfamily Member 11A), and TNFRSF11B (Tumor Necrosis Factor Receptor Superfamily Member 11B), plays a significant role [54]. In these diseases, inflammation and immune cell activation result in RANKL overexpression, leading to increased activity of bone-resorbing cells and enhanced bone resorption [55]. Concurrently, OPG expression may be reduced, resulting in overactive RANKL/RANK signaling. This imbalance can lead to excessive bone destruction compared to repair, culminating in osteoporosis and osteoarticular damage [56]. The development of osteonecrosis may be influenced by genetic factors. Several genetic variants have been linked to morphological, structural, and functional changes in bone tissue, thereby heightening the risk of osteonecrosis [53]. However, current osteonecrosis studies have not conclusively demonstrated that these genetic factors affect the development of autoimmune diseases. This finding also corroborates the reliability of our inverse Mendelian randomization experiment's results. Nonetheless, Mendelian randomization experiments may exhibit result multiplicity, indicating shared SNPs for exposure and outcome. Therefore, conducting Co-localization experiments is crucial. Co-localization Analysis, in contrast to MR's broader focus, concentrates on individual gene regions on chromosomes, not considering the overall association between two traits. This approach arguably offers a more in-depth examination of individual gene regions, providing further evidence of the association between traits [57]. At the protein level, pleiotropy is likely causal, where the SNP influences the protein and subsequently affects the downstream disease. In Mendelian randomization studies where the protein serves as the exposure and the disease as the endpoint, such Co-localization results would be reinforcing, suggesting a close relationship. However, in our study, we are examining disease-disease causality, necessitating the avoidance of this type of pleiotropy. From the results of our Co-localization experiment, it is apparent that the Mendelian randomization outcomes for T1DM, RA, and AS (three autoimmune diseases) and osteonecrosis did not display result multiplicity. This observation further validates the robustness of the current experiment's findings.

This experiment, like other Mendelian randomization (MR) studies, encounters several inherent limitations [58]. Firstly, two-sample MR methods do not address non-linear relationships between exposure factors and outcomes. Secondly, while this study's conclusions are based on MR results for causal inference, further experiments focusing on molecular mechanisms are necessary to validate these findings more comprehensively. Thirdly, factors such as the Beavis effect, canalization compensation, low statistical power, and the complexity of biological systems can influence the use of MR in causal inference. Lastly, the SNP data were derived from a European population, which may not represent all racial groups. Additionally, the lack of detailed data such as age and gender calls for a more refined study to reach more comprehensive conclusions.

In summary, this two-sample Mendelian Randomization (MR) study uncovers a causal association between rheumatoid arthritis, ankylosing spondylitis, and type 1 diabetes, with nontraumatic osteonecrosis in European Population. Despite these findings, there is a requirement for additional molecular biological investigations to elucidate the pathogenesis of osteonecrosis. Moreover, the randomization sample was drawn from a European population; therefore, the generalizability of these findings to other population groups remains uncertain.

7 Ethics approval and consent to participate

Not applicable.

8 Consent for publication

All the authors approved the manuscript for publication.

9 Availability of data and materials

All data generated or analysed during this study are included in this published article [and its supplementary information files].

10 Competing interests

The authors declare no conflict of interest, financial or otherwise.

11 Funding

Qingdao Science and Technology Benefit the People Demonstration Project (23-2-8-smjk-7-nsh)

12 Author Contributions

Conceptualization, Zheng.L and Fang.Z.; methodology, Zheng.L.; software, Fang.Z.; validation, Xuexiao.M and Kai.Z; formal analysis, Yongbin.W.; investigation, Zheng.L.; data curation, Wenzhe.B; writing—original draft preparation, Zheng.L.; writing—review and editing, Xuexiao.M. All authors have read and agreed to the published version of the manuscript.

13 Acknowledgments

This is my first paper during my PhD, thanks to the help of my fiancée, Dr. Qin.11

Yoon BH, Mont MA, Koo KH, et al. The 2019 revised version of association research circulation osseous staging system of osteonecrosis of the femoral head[J]. J Arthroplast. 2020;35(4):933–40.
Bergman J, Nordström A, Nordström P. Epidemiology of osteonecrosis among older adults in Sweden[J]. Osteoporos Int. 2019;30:965–73.
Powell C, Chang C, Naguwa SM, et al. Steroid induced osteonecrosis: an analysis of steroid dosing risk[J]. Autoimmun rev. 2010;9(11):721–43.
Kurtul S, Güngördü N. Dysbaric osteonecrosis in diving fisherman: a case report[J]. La Medicina del Lavoro, 2022, 113(4).
Cui Q, Jo WL, Koo KH et al. ARCO Consensus on the Pathogenesis of Non-traumatic Osteonecrosis of the Femoral Head[J]. J Korean Med Sci, 2021, 36(10).
Chang C, Greenspan A, Gershwin ME. The pathogenesis, diagnosis and clinical manifestations of steroid-induced osteonecrosis[J]. J Autoimmun. 2020;110:102460.
Svahn T, Elmstedt E, Lewander R et al. Avascular necrosis of the bone in 99 renal allograft recipients[J]. Scand J Urol Nephrol, 1975 (29 Suppl): 135–7.
Konarski W, Poboży T, Kotela A, et al. Does Diabetes Mellitus Increase the Risk of Avascular Osteonecrosis? A Systematic Review and Meta-Analysis[J]. Int J Environ Res Public Health. 2022;19(22):15219.
Borchers AT, Naguwa SM, Shoenfeld Y, et al. The geoepidemiology of systemic lupus erythematosus[J]. Autoimmun rev. 2010;9(5):A277–87.
Emdin CA, Khera AV, Kathiresan S. Mendelian randomization[J]. JAMA. 2017;318(19):1925–6.
Sheehan NA, Didelez V, Burton PR, et al. Mendelian randomisation and causal inference in observational epidemiology[J]. PLoS Med. 2008;5(8):e177.
Mills MC, Barban N, Tropf FC. An introduction to statistical genetic data analysis[M]. MIT Press; 2020.
Burgess S, Thompson SG. Mendelian randomization: methods for causal inference using genetic variants[M]. CRC; 2021.
Davies NM, Holmes MV, Smith GD. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians[J]. BMJ, 2018, 362.
Hemani G, Tilling K, Davey Smith G, Correction. Orienting the causal relationship between imprecisely measured traits using GWAS summary data[J]. PLoS Genet. 2017;13(12):e1007149.
Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports systematic causal inference across the human phenome[J]. elife. 2018;7:e34408.
Kamat MA, Blackshaw JA, Young R, et al. PhenoScanner V2: an expanded tool for searching human genotype–phenotype associations[J]. Bioinformatics. 2019;35(22):4851–3.
Baum CF, Schaffer ME, Stillman S. Instrumental variables and GMM: Estimation and testing[J]. Stata J. 2003;3(1):1–31.
Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression[J]. Int J Epidemiol. 2015;44(2):512–25.
Bowden J, Davey Smith G, Haycock PC, et al. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator[J]. Genet Epidemiol. 2016;40(4):304–14.
Verbanck M, Chen CY, Neale B, et al. Publisher Correction: Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases[J]. Nat Genet. 2018;50(8):1196–6.
Burgess S, Foley CN, Zuber V. Inferring causal relationships between risk factors and outcomes from genome-wide association study data[J]. Annu Rev Genom Hum Genet. 2018;19:303–27.
Giambartolomei C, Vukcevic D, Schadt EE, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics[J]. PLoS Genet. 2014;10(5):e1004383.
Burgess S, Smith GD, Davies NM et al. Guidelines for performing Mendelian randomization investigations[J]. Wellcome Open Res, 2019, 4.
Vulpiani MC, Vetrano M, Trischitta D, et al. Extracorporeal shock wave therapy in early osteonecrosis of the femoral head: prospective clinical study with long-term follow-up[J]. Arch Orthop Trauma Surg. 2012;132:499–508.
Zhao D, Liu Y, Ma C, et al. A Mini Review: Stem cell therapy for osteonecrosis of the femoral head and pharmacological aspects[J]. Curr Pharm Design. 2019;25(10):1099–104.
Chughtai M, Piuzzi NS, Khlopas A, et al. An evidence-based guide to the treatment of osteonecrosis of the femoral head[J]. bone joint J. 2017;99(10):1267–79.
Liu F, Wang W, Yang L et al. An epidemiological study of etiology and clinical characteristics in patients with nontraumatic osteonecrosis of the femoral head[J]. J Res Med sciences: official J Isfahan Univ Med Sci, 2017, 22.
Lubberts E. The IL-23–IL-17 axis in inflammatory arthritis[J]. Nat Rev Rheumatol. 2015;11(7):415–29.
Kotake S, Sato K, Kim KJ, et al. Interleukin-6 and soluble interleukin‐6 receptors in the synovial fluids from rheumatoid arthritis patients are responsible for osteoclast‐like cell formation[J]. J Bone Miner Res. 1996;11(1):88–95.
Tektonidou MG, Moutsopoulos HM. Immunologic factors in the pathogenesis of osteonecrosis[J]. Orthop Clin. 2004;35(3):259–63.
Wei Q, Zhou M, Liu J, et al. Relationship between osteonecrosis and antiphospholipid antibodies in patients with systemic lupus erythematosus: a systematic review protocol[J]. BMJ open. 2021;11(7):e046163.
Fernández-Codina A, Pope JE. Osteoporosis and osteonecrosis in systemic lupus erythematosus[J]. Revista Colombiana de Reumatología. 2021;28:3–11.
Atkinson MA, Eisenbarth GS, Michels AW. Type 1 diabetes[J]. Lancet. 2014;383(9911):69–82.
Matthew Seah KT, McEachan J, Davidson D. Bilateral Kienböck’s disease in association with type 1 diabetes[J]. J Hand Surg (European Volume). 2012;37(7):697–8.
Samarasinghe YP, Brunton SL, Feher MD. Avascular necrosis not Charcot's[J]. Diabet Med. 2001;18(10):846–8.
Chen J, Jia G, Lv X et al. Type 1 Diabetes Mellitus-Related circRNAs Regulate CD4 + T Cell Functions[J]. BioMed Research International, 2022, 2022.
Li Y, Zhou Y, Zhao M, et al. Differential profile of plasma circular RNAs in type 1 diabetes mellitus[J]. Diabetes metabolism J. 2020;44(6):854–65.
Yang L, Han X, Zhang C, et al. Hsa_circ_0060450 negatively regulates type I interferon-induced inflammation by serving as miR-199a-5p sponge in type 1 diabetes mellitus[J]. Front Immunol. 2020;11:576903.
Taheri M, Eghtedarian R, Dinger ME, et al. Emerging roles of non-coding RNAs in the pathogenesis of type 1 diabetes mellitus[J]. Volume 129. Biomedicine & Pharmacotherapy; 2020. p. 110509.
Zhou H, Kuang H. Circular RNAs: Novel target of diabetic retinopathy[J]. Reviews Endocr Metabolic Disorders. 2021;22:205–16.
Zabinski SJ, Sculco TP, Dicarlo EF, et al. Osteonecrosis in the rheumatoid femoral head[J]. J Rhuematol. 1998;25(9):1674–80.
Mylov NM, Polivanov AE. Aseptic necrosis of the head of the femur in patients with major rheumatic diseases of inflammatory and dystrophic genesis[J]. Vestn Rentgenol Radiol, 2005 (6): 37–43.
Firestein GS. Evolving concepts of rheumatoid arthritis[J]. Nature. 2003;423(6937):356–61.
McInnes IB, Schett G. The pathogenesis of rheumatoid arthritis[J]. N Engl J Med. 2011;365(23):2205–19.
de Molon RS, Hsu C, Bezouglaia O, et al. Rheumatoid arthritis exacerbates the severity of osteonecrosis of the jaws (ONJ) in mice. A randomized, prospective, controlled animal study[J]. J Bone Miner Res. 2016;31(8):1596–607.
Zang Yinsan. Clinical analysis of 11 cases of ankylosing spondylitis combined with femoral head necrosis[J]. Practical Clin Med. 2011;12(8):87–8.
Jiang J, Zhan X, Liang T, et al. Dysregulation of SAA1, TUBA8 and Monocytes Are Key Factors in Ankylosing Spondylitis with Femoral Head Necrosis[J]. Front Immunol. 2022;12:6037.
Haroon N, Inman RD, Learch TJ, et al. The impact of tumor necrosis factor α inhibitors on radiographic progression in ankylosing spondylitis[J]. Arthr Rhuem. 2013;65(10):2645–54.
Meng C, Bai R, Zhao Z et al. MMP-8 single-nucleotide polymorphisms are related to ankylosing spondylitis in Chinese Han population[J]. Medicine, 2018, 97(35).
Emdin CA, Khera AV, Kathiresan S. Mendelian randomization[J]. JAMA. 2017;318(19):1925–6.
Invernizzi P, Gershwin ME. The genetics of human autoimmune disease[J]. J Autoimmun. 2009;33(3–4):290–9.
Wang T, Azeddine B, Mah W, et al. Osteonecrosis of the femoral head: genetic basis[J]. Int Orthop. 2019;43:519–30.
Page G, Miossec P. RANK and RANKL expression as markers of dendritic cell–T cell interactions in paired samples of rheumatoid synovium and lymph nodes[J]. Arthritis Rheumatism: Official J Am Coll Rheumatol. 2005;52(8):2307–12.
Yasuda H. Discovery of the RANKL/RANK/OPG system[J]. J Bone Miner Metab. 2021;39(1):2–11.
Walsh MC, Choi Y. Biology of the RANKL–RANK–OPG system in immunity, bone, and beyond[J]. Front Immunol. 2014;5:511.
Lin J, Zhou J, Xu Y. Potential drug targets for multiple sclerosis identified through Mendelian randomization analysis[J]. Brain, 2023: awad070.
Haycock PC, Burgess S, Wade KH, et al. Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies[J]. Am J Clin Nutr. 2016;103(4):965–78.

No competing interests reported.

Additionalfile1.zip
additional file 1：Supplement Table 1. All instrumental variables used for MR analysis of the causal relationship between autoimmune diseases and osteonecrosis. Supplement Table 2. Pleiotropy and Heterogeneity Results from a Positive Mendelian Randomization Experiment. Supplementary Tables 3, 4, and 5.Results of all reverse Mendelian experiments. Supplementary Table 6. Steiger Test results.
Additionalfile2.zip
additional file 2: Results of Co-localization Analysis.

Causal Relationship between Autoimmune Diseases and Risk of Osteonecrosis in European Population: A Two-Sample Mendelian Randomization Study

Status:

Version 1

Abstract

Figures

Introduction

Methods

3.1 Data Collection

3.2 Screening of instrumental variables

3.3 Statistical analysis

3.4 Sensitivity Analysis

3.5 Co-localization Analysis

Results

Discussion

Conclusion

Declarations

11 Funding

13 Acknowledgments

References

Additional Declarations

Supplementary Files

Status:

Version 1