Associations of Homocysteine, Folate, and Vitamin B12 with Osteoarthritis: A Mendelian Randomization Study

Homocysteine, inversely related to folate and vitamin B12, is an independent risk factor for several age-related disorders. However, little is known about the association of homocysteine and related vitamins with osteoarthritis (OA). This study aimed to elucidate the potential causal effects of homocysteine, folate, and vitamin B12 on site- and gender-specific OA by applying the two-sample Mendelian randomization (MR) approach. Genetically predicted homocysteine showed adverse effects on overall OA (95% confidence interval (CI): 1.044–1.155), knee OA (95% CI: 1.000–1.167), hip OA (95% CI: 1.057–1.297), and spine OA (95% CI: 1.017–1.216). Genetically predicted folate showed protective effects on overall OA (95% CI: 0.783–0.961) and spine OA (95% CI: 0.609–0.954). Folate (95% CI: 0.887–1.004) and vitamin B12 (95% CI: 0.886–1.009) showed a protective trend against knee OA. The patterns of associations were site and gender specific. In conclusion, homocysteine had adverse effects on OA, especially on OA at weight-bearing joints and in females. Folate and vitamin B12 had protective effects on OA. Homocysteine-lowering interventions may be a potential option in the treatment and prevention of OA.


Introduction
Osteoarthritis (OA) is a common joint disease characterized by cartilage damage and degradation of other tissues, such as subchondral bone within the joint [1]. OA is an ageand sex-related disorder with stiffness, pain, and impaired joint movement as the main symptoms. Studies have shown that the risk of OA increases rapidly between the ages of 50 years and 75 years and that it is more prevalent in women than in men [1,2]. OA can occur at all joints in the body. Notably, different patterns of OA have been demonstrated in different sites and genders in humans [2]. Although some symptomatic treatments have been recommended to reduce pain and improve joint movement, effective conservative treatments to halt or cure cartilage damage in OA are still currently unavailable, partly due to the limited understanding of the pathogenesis of OA [3].
Homocysteine, an intermediate metabolite within the methionine cycle, can be elevated in blood by deficiency of the related B vitamins, mainly folate and vitamin B12 [4]. A meta-analysis in China with 60,754 subjects aged 3-97 years demonstrated that the prevalence of hyperhomocysteinemia (HHcy, total homocysteine concentrations more than 15 µmol/L in blood) increased with age and was the highest in the group of people older than 65 years [4]. Higher blood concentrations of homocysteine have been identified as a risk factor for several age-related disorders, including cardiovascular disease (CVD) and osteoporosis [5,6]. Recently, in vitro studies revealed that homocysteine can stimulate osteoclast differentiation and cause chondrocyte dysfunction [7,8]. Considering the changes with aging for both OA and homocysteine and the potential relation between homocysteine and joint metabolism, a possible causal association of homocysteine with OA could be inferred. However, to date, no randomized controlled trials (RCTs) have been reported on the causal relationship between blood homocysteine and OA. Moreover, only a few observational studies have focused on the correlation between blood homocysteine levels and OA in humans, and the limited data provided inconsistent conclusions [9][10][11][12]. The Framingham OA study showed that homocysteine levels were not associated with the radiological incidence and progression of knee OA in women [10]. A cross-sectional study in Japan observed that homocysteine levels were related to the prevalence rate of spinal OA in postmenopausal women [9]. Another study in China observed a significantly elevated serum homocysteine level in patients with severe OA [12]. These inconsistent results are probably due to residue bias, confounding factors, and reverse causality. Thus, the exact causal relationship between blood homocysteine and OA is still unclear and deserves further investigation.
The Mendelian randomization (MR) approach is able to circumvent the disadvantages of the observational studies mentioned above by applying genetic variants as instrumental variables (IVs) and has been widely applied to evaluate the causal association of exposures with outcomes [13,14]. The present study aimed to examine the causal association of homocysteine and B vitamins with site-and gender-specific OA using the MR approach.

Study Design
MR analyses have 3 key assumptions: (1) IVs are strongly related to the exposure; (2) IVs should not be related to potential confounders; and (3) IVs should affect the outcome exclusively through the exposure. A sketch of the study design is illustrated in Figure 1.

Data Sources of Outcomes
Genetic data for OA were derived from the largest GWAS (genome-wide association

Data Sources of Outcomes
Genetic data for OA were derived from the largest GWAS (genome-wide association study) to date by Boer's group [15], including overall OA (177,517 cases and 649,173 controls) and 5 different sites of OA (knee OA, hip OA, spine OA, hand OA, and thumb OA) (Table S1). In addition, sex-specific analysis was conducted in this GWAS meta-analysis. Given that sex is related to both OA and homocysteine, genetic data for OA in each sex were also extracted from these GWAS.

Selection of Instrumental Variables
SNPs related to homocysteine, folate, and vitamin B12 were extracted from two previous GWASs, which included 44,147, 37,465, and 45,576 individuals of European ancestry [16,17].

MR Analyses
All analyses were performed in R software (version 4.1.3) using the R package "TwoSampleMR" [21]. The inverse-variance weighted (IVW) method based on the fixedeffects and random-effects model was utilized as the main analysis [22,23]. The weighted median and MR-Egger methods, which make diverse assumptions about horizontal pleiotropy, were performed as complementary methods to test the robustness of the main analysis [24,25]. Cochrane's Q value was used to assess the heterogeneity among estimates of SNPs [26]. The intercept in MR-Egger regression and MR pleiotropy residual sum and outlier (MR-PRESSO) test were applied to identify pleiotropy [27]. The leave-one-out method was implemented by sequentially excluding each SNP to determine whether the estimates were driven by any single SNP.

Discussion
In this MR study, we investigated the causal effects of homocysteine, folate, and vitamin B12 on the risk of overall and site-and gender-specific OA. The results revealed suggestive associations of higher genetically predicted homocysteine levels with increased risk of overall OA and OA at weight-bearing joints (knee, hip, and spine), especially in females. Folate and vitamin B12 had protective effects on knee OA in both males and females. Furthermore, folate was also associated with a lower risk of overall OA and spine OA in both males and females.
The present study shows that higher homocysteine levels could increase the risk of overall OA. The reason for this result is still unclear due to the lack of studies investigating the pathogenic mechanisms of homocysteine in OA. Based on the reported adverse effects of homocysteine in the literature, a possible association of homocysteine with OA could be inferred. A recent in vitro study reported that homocysteine could induce several detrimental changes in chondrocytes including mitochondrial dysfunction, apoptosis, and oxidative stress accumulation [8]. Moreover, accumulated evidence has suggested that homocysteine could affect bone metabolism and reduce bone strength [7]. Homocysteine in bone was not only bound to collagen of the extracellular matrix but also disturbed the osteoblastmediated calcification process and stimulated osteoclast differentiation. In diet-induced HHcy in rats, homocysteine accumulated in bone was accompanied by bone loss and bone strength reduction [28]. It is well known that subchondral bone and cartilage form a functional complex called the bone-cartilage unit, which is involved in the pathophysiology of OA biochemically and mechanically [1]. Subchondral bone is an important aspect of the mechanical environment for cartilage, and its microstructure has close relationship with cartilage metabolism [29,30]. Although there are different microstructural changes in subchondral bone at different stages of OA (enhanced subchondral bone turnover in early OA and subchondral bone sclerosis in the advanced stage), subchondral bone in OA undergoes an uncoupling process at different stages during OA progression [1]. The equilibrium between osteoclast-mediated bone resorption and osteoblast-mediated bone formation is crucial for bone homeostasis, and spatiotemporal uncoupling is responsible for the degradation of subchondral bone and progression of OA [30]. These findings suggest that homocysteine plays adverse roles in collagen metabolism, chondrocyte function, and bone formation and may be involved in OA pathophysiology.
Notably, the causal effects of homocysteine on OA were site specific. Homocysteine was associated with OA in weight-bearing joints (knee, hip and spine) but not with hand OA and thumb OA. OA itself is site specific with heterogeneity in terms of prevalence, epidemiological risk factors, genetics, and etiology [1,31,32]. Mechanical factors causing the wear of cartilage and affecting the integrity of subchondral bone have been considered one of the most important factors ascribed to the site specificity of OA [33]. There are three main patterns of hand OA: the symmetrical pattern (same joint involvement from left and right hands), the row pattern (several distal interphalangeal (DIP) or proximal interphalangeal (PIP) joint involvement), and the ray pattern (DIP/PIP joint involvement of the same finger). The progression of hand OA demonstrated in a symmetrical pattern, which mainly resulted from systemic factors rather than mechanical factors [33]. However, unlike the hand joint, which is thought to be influenced mostly by systemic factors, weightbearing joints (knee, hip, and spine) present a distinct biomechanical environment and are more susceptible to mechanical changes [34]. Mechanical factors may explain our MR results regarding homocysteine's prominent effects in weight-bearing joints. Among the periarticular tissues, subchondral bone contributes significantly to the mechanical environment and plays different roles in the onset and progression of OA at different sites. Thus, the differential effects of homocysteine on OA in weight-bearing and non-weightbearing joints suggest that blood homocysteine may be involved in OA pathogenesis not only through directly disturbing chondrocyte functions but also through affecting subchondral bone.
The sex-stratified MR analyses demonstrated that the causal effects of homocysteine on OA were gender specific. For females, homocysteine was associated with overall OA, knee OA, hip OA, and spine OA but not with hand OA and thumb OA. For males, homocysteine was only associated with overall OA and hip OA. Previous studies also reported sex specificity regarding the association between homocysteine and OA [9,10]. Both HHcy and OA are well-known sex-related disorders. Epidemiological studies have reported an increased prevalence of OA and HHcy in postmenopausal women. Estrogen deficiency could increase blood levels of homocysteine and induce or worsen OA [35][36][37]. Among the periarticular tissues, the subchondral bone responds most significantly to estrogen deficiency, exhibiting increased bone turnover accompanied by reduced stiffness and altered mechanical characteristics [37]. The sex-specific pattern of association seems to provide another clue that blood homocysteine levels influence OA development through bone structural alteration. Taking the site-and sex-specific patterns of homocysteine's effects on OA together, subchondral bone is nonnegligible during OA development stimulated by homocysteine. The exact mechanisms need to be further investigated.
Since folate and vitamin B12 supplementation are well-recognized methods to reduce the risk of diseases by lowering the blood level of homocysteine, the causal effects of folate and vitamin B12 on OA were evaluated in the present study. The protective and site-specific effects of folate and vitamin B12 on OA were manifested. Sex-stratified analyses showed similar patterns in males and females. In line with our findings, previous studies also demonstrated an inverse effect of folate and vitamin B12 on OA. Folate deficiency was reported to be associated with increased radiographic severity of knee OA [38]. Knee OA patients had lower intakes of dietary folate than controls [39]. Dietary folate and vitamin B12 supplementation can improve joint functions in OA patients [40]. Rationally, folate and vitamin B12 supplementation may be an optional intervention to prevent OA.
People remain unconvinced about whether and how homocysteine and human OA are related. To date, no related RCTs have been conducted. Traditional observational and cohort studies have demonstrated inconsistent results, probably due to some confounding factors. The Framingham osteoarthritis study concluded that homocysteine was not associated with the progression of radiographic knee OA [10], whereas cross-sectional studies in Japan and China observed the correlation of circulating homocysteine levels with spinal OA and knee OA, respectively [9,12]. A recent MR study indicated that hip OA is not correlated with higher homocysteine concentrations and the related B vitamins have no causal effects on OA [41]. These results seem different from ours, which might be attributed to the choice of outcome data sources and selection of SNPs. Our MR study used the largest GWAS for OA to date, which contains the data of site-and gender-specific OA, and selected instrumental variables complying with three key assumptions in MR design. The present design might provide more comprehensive results about the potential association of homocysteine and the related B vitamins with OA.
There are several limitations in the present study. First, sensitivity analyses of the association of folate with outcomes could not be performed since only two SNPs after screening were available for folate. Second, since the individuals included in our study were mainly European individuals, our findings may not be directly generalizable to other populations. Third, since the GWAS meta-analysis of exposures did not include sex-specific genetic data, only genetic data for OA in each sex were used to conduct sex-specific analysis in our MR analyses. The sex-specific patterns of association between exposures and OA provide interesting clues and may provoke more subsequent investigations. Furthermore, GWAS meta-analysis is warranted to conduct sex-stratified analyses, especially when sex plays a role in the disorder.

Conclusions
This MR study provided evidence that genetically predicted homocysteine was causally associated with an increased risk of OA, especially at weight-bearing joints and in females. Folate and vitamin B12 had suggestively protective effects on OA. Homocysteinelowering interventions, for example, through folate and vitamin B12 supplementation, may be a potential option for the treatment and prevention of OA, especially OA at weightbearing joints and in females.  Figure S5: Causal effect of homocysteine in OA in females in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Figure S6: Causal effect of folate in OA in females in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Figure S7: Causal effect of vitamin B12 in OA in females in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Figure S8: Causal effect of homocysteine in OA in males in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Figure S9: Causal effect of folate in OA in males in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Figure S10: Causal effect of vitamin B12 in OA in males in fixed-effects and random-effects IVW analyses. OR: odds ratio; CI: confidence interval; p value: p value of the causal estimate. Table S1 Detailed information on the genome-wide association study of OA. Table S2: Association (p < 5 × 10 −8 ) of SNPs used as candidate genetic instruments for OA at any site with confounders or OA.  Table S7: Weighted median and MR-Egger methods for genetic associations between exposures and outcomes.