- Research
- Open access
- Published:
Causal association of epigenetic aging and osteoporosis: a bidirectional Mendelian randomization study
BMC Medical Genomics volume 16, Article number: 275 (2023)
Abstract
Background
The relationship between aging and osteoporosis is well established. However, the relationship between the body's physiological age, i.e. epigenetic age, and osteoporosis is not known. Our goal is to analyze the bidirectional causal relationship between epigenetic clocks and osteoporosis using a bidirectional Mendelian randomization study.
Methods
We used SNPs closely associated with GrimAge, Hannum, PhenoAge, and HorvathAge in epigenetic age and SNPs closely associated with femoral neck bone mineral density, lumbar spine bone mineral density, and forearm bone mineral density as instrumental variables, respectively, using the inverse variance weighting method and several other MR methods to assess the bidirectional causal relationship between epigenetic age and osteoporosis.
Result
There was no evidence of a clear causal relationship of epigenetic age (GrimAge, Hannum, PhenoAge, and HorvathAge) on femoral neck bone mineral density, lumbar spine bone mineral density, and forearm bone mineral density. In reverse Mendelian randomization analysis showed a significant causal effect of lumbar spine bone mineral density on GrimAge: odds ratio (OR) = 0.692, 95% confidence interval (CI) = (0.538–0.890), p = 0.004. The results suggest that a decrease in lumbar spine bone mineral density promotes an acceleration of GrimAge.
Conclusion
There was no significant bidirectional causal relationship between epigenetic age and osteoporosis A decrease in lumbar spine bone density may lead to an acceleration of the epigenetic clock "GrimAge". Our study provides partial evidence for a bidirectional causal effect between epigenetic age and Osteoporosis.
Introduction
Osteoporosis (OP) is the most common metabolic bone disease, characterized by reduced bone mineral density (BMD) and deterioration of trabecular architecture, and is capable of causing many skeletal-related diseases, with a serious negative impact on the health of the elderly in particular [1, 2]. With more than 9 million OP-related fractures occurring each year, OP has become a major global public health problem, which imposes a significant clinical and economic burden on society and an increased health cost and disease burden on patients [3]. Age is a very important factor in the reduction of bone mineral density and because of lifestyle habits, physical conditions, etc., older adults are more prone to osteoporotic fractures [4, 5]. In addition to age, BMD in osteoporosis and its prognosis involve many other factors such as genetic factors, gender, obesity, calcium and milk intake, caffeine intake, smoking and physical activity [6, 7]. However, both the current diagnostic technique for OP, dual-energy X-ray absorptiometry (DXA), and pharmacological treatment of OP have limitations, and patients with osteoporosis have poor compliance with pharmacological treatment due to adverse reactions associated with drug side effects, so it is crucial to find other risk factors for OP to optimize the diagnosis, prevention, and treatment of OP [8,9,10,11].
Epigenetic age is a heritable indicator of biological aging derived from DNA methylation (DNAm) data, which differs from actual age and is a very promising biomarker of aging that has emerged recently [12]. Accelerated epigenetic age means that the predicted biological age is greater than the actual age, and such populations have an increased risk of death from various causes [13, 14]. "First generation" epigenetic clocks, such as HannumAge and Intrinsic HorvathAge, are derived based on DNAm levels at CpG loci that are closely related to actual age [15, 16]. Hannum et al. analyzed whole blood samples from 656 individuals and identified 71 age-associated CpGs, and HannumAge results were derived by training on these loci [16]. Intrinsic HorvathAge is based on training on 353 age-related CpG loci found in human tissues and cells, and further adjusting blood cell counts [15]. "Second generation" epigenetic clocks, such as PhenoAge and GrimAge, have been developed in recent years to predict age-related morbidity and mortality [17, 18]. PhenoAge integrates data from 513 CpG and nine clinical biomarkers (e.g., albumin, creatinine, serum glucose, C-reactive protein, lymphocyte percentage, etc.) associated with mortality [17]. GrimAge was derived from seven plasma proteins (i.e. cystatin C, leptin, tissue inhibitor metalloproteinases 1, adrenomedullin, beta-2-microglobulin, growth differentiation factor 15, and plasminogen activation inhibitor 1 (PAI-1)) and a DNAm-based estimator of smoking pack-years [18].
Since OP is inextricably linked to aging at real age, is there an equally strong relationship between OP and aging at biological age? Differences in DNA methylation signatures of bone tissue samples and mesenchymal stem cells (MSCs) have been found between patients with fragility fractures and osteoarthritis controls, and accelerated epigenetic aging of chondrocytes has been identified [19, 20]. However, no study has been able to prove whether there is a causal relationship between epigenetic age and osteoporosis. In this study, we used MR analysis methods to analyze the causal relationship between epigenetic age and osteoporosis.
Mendelian randomization (MR) analysis is a method of causal inference using genetic variation as an instrumental variable (IV), thereby testing the causal relationship between exposure factors and disease [21,22,23]. By using GWAS data for Mendelian randomization, we can understand the relationship between a factor and individual diseases [24, 25]. A systematic review and meta-analysis of Mendelian Randomization studies collated Mendelian randomization studies of BMI and multiple diseases and performed a meta-analysis for each disease to systematically look at the causal relationship between exposure measures and disease risk [26]. Standard MR studies must satisfy three core assumptions: (1) instrumental variables are strongly associated with exposure factors; (2) genetic variables are not associated with any confounding factors affecting the causal relationship between exposure and outcome; (3) instrumental variables can only affect outcome through exposure and are not directly associated with outcome [22, 27]. Therefore, we performed a two-way MR analysis to investigate the association between epigenetic age and OP (measured as BMD).
Methods
Research design
In this study, we first used epigenetic age (PhenoAge, GrimAge, Hannum, HorvathAge) as the "exposure" and femoral neck BMD (FN BMD), lumbar spine BMD (LS BMD) and forearm BMD (FA BMD) as the "outcome" for Mendelian randomization analysis. In contrast, we used FN BMD, LS BMD and FA BMD as "exposures" and epigenetic age as "outcomes". MR studies consistently meet their three key assumptions: (1) instrumental variables are strongly associated with exposure factors; (2) genetic variables are not associated with any confounding factors affecting the causal relationship between exposure and outcome; and (3) instrumental variables can only influence outcomes through exposure and are not directly associated with outcomes [22, 27]. Figure 1 shows an overview of the three hypotheses.
Data sources
We obtained summary genetic association estimates for epigenetic age acceleration measures from the recent GWAS biological aging meta-analysis: Intrinsic HorvathAge [15], HannumAge [16], PhenoAge [17] and GrimAge [18]. The analysis was a meta-analysis based on the European ancestry of 34,710 participants from 28 cohorts, 57.3% of whom were women, and it identified 137 loci for DNA biomarkers associated with aging. For more information and a detailed description of the methodology, please see the latest GWAS meta-analysis of biological aging [28].
The diagnosis of OP in clinical practice is often made by BMD at three common skeletal sites: femoral neck BMD (FN BMD), lumbar spine BMD (LS BMD), and forearm BMD (FA BMD). Therefore, we selected the GWAS summary statistics for FA BMD, FN BMD and LS BMD from the GWAS meta-analysis of 53,236 individuals of European ancestry conducted by the Genetic Factors for Osteoporosis (GEFOS) consortium [29]. Relevant GWAS summary statistics are available from the GEFOS website (http://www.gefos.org/?q=content/data-release-2015). Table 1 shows the details of the studies and datasets used for the analysis.
Selection of instrumental variables
Selection of appropriate instrumental variables (IV) for MR analysis from the different GWAS summary results. Independent single nucleotide polymorphisms (SNPs) with significant genome -wide significance (p < 5 × 10–8) associated with the exposure were selected as IV. However, only three SNPs were selected for FA BMD when screening instrumental variables from GWAS data for FA BMD, so we adjusted the threshold (p < 5 × 10–6) in order to obtain a relatively appropriate number of instrumental variables. Set the parameter r2 threshold to 0.001 and kilobase pairs (kb) to 10,000, and exclude the linkage disequilibrium (LD) by the clump_data function in the "TwoSampleMR" R package. The last valid SNPs obtained that were significantly associated with exposure were used as IV. In this study, we calculated the F values by which the intensity of the screened IVs was assessed [30]. F value of a single SNP = beta2/SE2 [31]. IVs with F values > 10 have good instrumental strength and can mitigate the potential bias present in MR analysis. The proportions of trait variance explained by genetic instruments (R2) is calculated by the following equation: R2 = 2 × MAF × (1-MAF) × beta (MAF = minor allele frequency, beta = effect size, SE = standard error, N = sample size, k = number of IVs) [32]. In order to verify whether the selected instrumental variables satisfied the independence hypothesis, PhenoScanner (http://www.phenoscanner.medschl.cam.ac.uk/) was used to detect whether the remaining SNPs were correlated with other phenotypes, such as basal metabolic rate and rheumatoid arthritis, and the results showed that no SNPs was significantly associated with these factors [33, 34].
Statistical analysis
The inverse variance weighting (IVW) method uses meta-analysis methods to combine the Wald estimates for each SNP to assess the effect of exposure on outcome. The IVW method was used as the primary statistical method to assess the bidirectional relationship between epigenetic age and OP. The IVW method provides unbiased effect estimates when the selected IVs are all valid IVs (when there is no horizontal pleiotropy or heterogeneity). We also performed MR-Egger method, weighted median method (WME), simple mode and weighted mode to complement the MR results. The WME method produces valid causal estimates when the valid IVs of all selected IVs are greater than 50% [35]. We also calculated causal effect estimates (equivalent to beta coefficients) and converted them into dominance ratios (OR).
A series of sensitivity analyses were conducted for the accuracy and robustness of the results. Heterogeneity between IVs was assessed by Cochran Q statistic [36]. If there was heterogeneity in the data, we removed the SNPs causing heterogeneity by MR-PRESSO analysis(with P value less than the threshold in the MR-PRESSO outlier test). If the heterogeneity was still significant after removing the abnormal values by MR-PRESSO, we would perform MR analysis again under the condition of removing all abnormal SNPs with P values less than 1 in MR-PRESSO [37]. We assessed horizontal pleiotropy by analyzing MR Egger intercept values [27]. Then, we analyzed each SNP by using the "leave-one-out" method, and checked whether there was a high effect on the effect estimates by eliminating each SNP one by one.
The above statistical treatments were implemented by software R4.2.2 and TwoSampleMR package.
Results
Instrumental variable
We screened 4, 9, 24, and 11 IVs from the GWAS data of GrimAge, HannumAge, Intrinsic HorvathAge, and PhenoAge, respectively, and the F statistics of all selected SNPs were greater than 10 (GrimAge range 31–45, HannumAge range 31–99, Intrinsic HorvathAge range 31–240, PhenoAge range 32–89) (Supplementary Table 1). 20, 16, 22 IVs were screened from the GWAS data of FN BMD,FA BMD, LS BMD. All selected SNPs had F statistics greater than 10 (FN BMD range 32–111, FA BMD range 22–114, LS BMD range 32–89) (Supplementary Table 2). The instrument strength in this study is high and the results of MR analysis are not affected by weak instrumental variable bias.
Bidirectional Mendelian randomization results
We observed no causal relationship between epigenetic age and BMD in Mendelian randomization analysis. In the causal relationship analysis of the epigenetic clock on FN BMD, we calculated the OR and 95% CI: GrimAge and FN BMD: 1.023 (95%CI 0.980–1.067),P = 0.297;HannumAge and FN BMD: 1.002(95%CI 0.973–1.031), P = 0.915; Intrinsic HorvathAge and FN BMD: 0.988(95%CI 0.973–1.002), P = 0.100; PhenoAge and FN BMD: 1.007(95%CI 0.991–1.023), P = 0.401. Therefore, we found that there was no statistically significant (Table 2). For detailed information on the MR analysis of the causal relationship of the epigenetic clock on BMD refer to Supplementary Table 3. The scatter plot is shown in Fig. 2. When performing reverse MR (exposure: FN BMD, outcome: epigenetic age), we found that the IVW results of FN BMD and PhenoAge showed a causal relationship between the two. (beta = 0.427, SE = 0.181,p = 0.019). But in the MR Egger method the result is: beta = -0.325,SE = 0.929,p = 0.730 (Table 3). It can be found that the beta values of the two methods are not in the same direction, indicating that the results obtained are not robust, so a clear causal relationship between the two cannot be established yet. It can be found by other methods that only the results of IVW are more significant, while all other methods are not significant, and on the whole, it cannot be considered that there is a causal relationship between them. Additional analysis methods can be found in Supplementary Table 4. The scatter plot is shown in Fig. 3. The pleiotropy analysis showed no pleiotropy in the results. Heterogeneity analysis revealed a heterogeneity between FN and GrimAge, so we performed MR-PRESSO analysis. After removing the abnormal IV, the heterogeneity disappeared, and MR analysis was performed again, which resulted in negative results (Table 4).
In the causal relationship analysis of the epigenetic clock on FA BMD, we calculated the OR and 95% CI: GrimAge and FA BMD: 0.975 (95%CI 0.893–1.064),P = 0.569;HannumAge and FA BMD: 1.029 (95%CI 0.979–1.081), P = 0.259;Intrinsic HorvathAge and FA BMD:0.991(95%CI 0.958–1.024), P = 0.579; PhenoAge and FAN BMD:1.011(95%CI 0.977–1.045), P = 0.534. We found no statistical significance (Table 4) (Supplementary Table 3). The scatter plot is shown in Fig. 4. In reverse MR analysis (exposure: FA BMD, outcome: epigenetic age), the result is: FA BMD and GrimAge: 0.870(95%CI 0.697–1.085), P = 0.217; FA BMD and HannumAge: 0.943(95%CI 0.774–1.148), P = 0.557; FA BMD and Intrinsic HorvathAge: 1.114 (95%CI 0.928–1.337), P = 0.248; FAN BMD and PhenoAge: 1.006(95%CI 0.799–1.266), P = 0.960. The results are not statistically significant (Table 3) (Supplementary Table 4). The results are not heterogeneous or pleiotropic (Table 4). The scatter plot is shown in Fig. 5. Therefore, the results of the bidirectional Mendelian analysis between epigenetic clock and FA BMD concluded that there is no causal relationship between them.
In the causal analysis of the epigenetic clock on LS BMD, we did not find a significant causal relationship between the two. Heterogeneity was found between Hannum and LS BMD when heterogeneity analysis was performed. The heterogeneity disappeared after removal of the abnormal IV (Table 4). The results showed no causal relationship between the two (Table 2) (SupplementaryTable 3). The scatter plot is shown in Fig. 6. When reverse MR analysis was performed (exposure: LS BMD, outcome: epigenetic age), due to the heterogeneity of the results of the two groups, LS BMD and GrimAge, and LS BMD and Hannum, we removed the abnormal IVs in both groups after MR-PRESSO analysis, and the heterogeneity disappeared after removal, and the information after removal of heterogeneity is presented in Table 4. We found a significant causal relationship of LS BMD on GrimAge, with the rest being negative results (Table 3) (Supplementary Table 4). There was no horizontal pleiotropy in the results (Table 4). The scatter plot is shown in Fig. 7.
Sensitivity analysis
Sensitivity analysis of the IVW method results was performed using the leave-one-out method. The results of “leave-one-out” analysis are shown in Supplementary Figure. Removing one SNP in turn and analyzing the remaining SNPs, we found that the absence of SNPs had a significant impact on the results and the results had significant confidence.
Discussion
Aging at real age is currently considered to be an important factor in OP, but biological aging is still a potential area for further study. If there is a potential causal relationship between accelerated biological aging and OP, then slowing down biological aging has become a new research direction for preventing OP. In recent years, research between epigenetic mechanisms and osteoporosis is gradually developing, and researchers believe that there is an inextricable relationship between epigenetics and osteoporosis [38]. In our observation of the causal relationship between epigenetic clock and OP, based on our MR results, we found no clear causal relationship of epigenetic clock on osteoporosis. And in the reverse MR analysis, we also did not find an association between the two.We found that the epigenetic clock did not show a clear effect on osteoporosis. In another study of the association between epigenetic age and osteoporosis, investigators analyzed genome-wide DNA profiles of peripheral blood from patients with apparent OP and non-OP controls, but did not find CpG sites with significantly abnormal DNAm in OP, and did not find epigenetic age acceleration in the blood of patients with OP [39].
In this study, we tried to investigate whether osteoporosis could accelerate or delay epigenetic age by reverse Mendelian randomization analysis. We found that LS-BMD has a negative causal relationship to the GrimAge clock of epigenetic age, i.e., a reduction in lumbar spine bone mineral density promotes acceleration of epigenetic age (GrimAge). According to our results, among the four epigenetic clocks, it was found that LS-BMD only showed a significant causal relationship with the GrimAge clock, but this result was not meaningless. Different clocks capture different processes of biological aging, so different biological mechanisms of aging can be understood [12].
GrimAge is better able to predict age-related clinical phenotypes and all-cause mortality than the other three clocks [40]. Our results support that reduced lumbar spine bone mineral density promotes accelerated GrimAge, and perhaps lumbar spine bone mineral density can be a factor affecting Physiological aging. Our findings may provide a promising target for intervention in aging. However, there is a lack of research on the effect of lumbar spine bone mineral density on epigenetic age, so it is hoped that our research will help this new field.
Our study is the first to use Mendelian randomization analysis to explore the causal relationship between epigenetic age and osteoporosis. Mendelian randomization has the advantage that any association between risk factors and disease outcome can be examined by using genetic variants as instrumental variables (IV), which can minimize confusion and avoid reverse causal bias [22, 41]. We selected SNPs with genome-wide associations and independent inheritance but no LD as IV to assess the causal relationship between epigenetic age and osteoporosis. In our analysis, the F-statistic for each SNP was much larger than 10, suggesting that the possibility of weak instrumental variable bias is small. In addition, to ensure the reliability of our results, we identified and removed abnormal IVs by MR-PRESSO outlier testing. We also performed sensitivity analysis to observe whether the results were pleiotropic and heterogeneous.
Our study still has some limitations. Because the pooled GWAS data came from populations of European ancestry, our conclusions may not apply to other ethnic groups. So we use our conclusions with caution in racially and ethnically diverse populations [42]. In this study, GWAS aggregate data was used, and due to the lack of individual data, we could not conduct a stratified analysis of factors such as gender and age. Although every effort was made to exclude confounding in our study, the extent of overlap in exposure and outcome data used in the two-sample MR analysis could not be estimated, and we could only minimize bias in sample overlap by using strong instruments (e.g., F statistic much larger than 10). While we minimized the confounding bias of SNPS, it is still possible that some SNPS are associated with undetected factors that may influence the association between epigenetic age and OP. The level of pleiotropy introduced by this SNP we cannot eliminate its effect on the results. Therefore, the results of MR Analysis should be interpreted with caution.The epigenetic biological GWAS we used are calculated based on blood counts, clinical markers, etc., and may have different results as more samples and GWAS emerge in the future.
Conclusion
Our findings suggest that there may not be a causal relationship between epigenetic age and osteoporosis, nor is there an association between osteoporosis and changes in epigenetic age. Decreased bone density in the lumbar spine may contribute to the acceleration of GrimAge. Our results are based on the Mendelian randomization analysis we performed. More studies are needed to explore the relationship and potential mechanisms between epigenetic clocks and osteoporosis.
Availability of data and materials
There is no use of raw, unprocessed data in this study. The datasets mentioned in this study can be found in online repositories.
The epigenetic age (Intrinsic HorvathAge, HannumAge, PhenoAge and GrimAge) GWAS data source: https://datashare.ed.ac.uk/handle/10283/3645
GWAS summary statistics for bone mineral density (FA BMD, FN BMD and LS BMD): http://www.gefos.org/?q=content/data-release-2015.
Abbreviations
- BMD:
-
Bone mineral density
- FN BMD:
-
Femoral neck bone mineral density
- LS BMD:
-
Lumbar spine bone mineral density
- FA BMD:
-
Forearm bone mineral density
- DXA:
-
Dual-energy X-ray absorptiometry
- CpG:
-
Cytosine-phosphate-guanine
- DNAm:
-
DNA methylation
- PAI-1:
-
Plasminogen activation inhibitor 1
- MSCs:
-
Mesenchymal stem cells
- MR:
-
Mendelian randomization
- IV:
-
Instrumental variable
- GEFOS:
-
Genetic Factors for Osteoporosis
- SNPs:
-
Single nucleotide polymorphisms
- kb:
-
Kilobase pairs
- LD:
-
Linkage disequilibrium
- IVW:
-
Inverse variance weighting
- WME:
-
Weighted median method
References
Compston JE, William MRM. Osteoporosis. Lancet (London, England). 2019;393:364–76.
Armas LA, Recker RR. Pathophysiology of osteoporosis: new mechanistic insights. Endocrinol Metab Clin North Am. 2012;41:475–86.
Burge R, Dawson-Hughes B, Solomon DH, Wong JB, King A, Tosteson A. Incidence and economic burden of osteoporosis-related fractures in the United States, 2005–2025. J Bone Miner Res. 2007;22:465–75.
Ettinger MP. Aging bone and osteoporosis strategies for preventing fractures in the elderly. Arch Intern Med. 2003;163(18):2237–46.
Monemi Amiri A, Hosseini SR, Rahmaninia F, Nooreddini H, Bijani A. slamic Azad University Guilan science and research branch. relationship between bone mineral density and physical activity level in the elderly. Ann Appl Sport Sci. 2015;3:23–32.
Alghadir AH, Gabr SA, Al-Eisa E. Physical activity and lifestyle effects on bone mineral density among young adults: sociodemographic and biochemical analysis. J Phys Ther Sci. 2015;27:2261–70.
Korkmaz N, Tutoglu A, Korkmaz I, Boyaci A. The relationships among Vitamin D Level, balance, muscle strength, and quality of life in postmenopausal patients with osteoporosis. J Phys Ther Sci. 2014;26:1521–6.
Aibar-Almazán A, Voltes-Martínez A, Castellote-Caballero Y, Afanador-Restrepo DF, Carcelén-Fraile MD, López-Ruiz E. Current status of the diagnosis and management of osteoporosis. Int J Mol Sci. 2022;23:9465.
Chandran T, Venkatachalam I. Efficacy and safety of denosumab compared to bisphosphonates in improving bone strength in postmenopausal osteoporosis: a systematic review. Singapore Med J. 2019;60:364–78.
Schultz K, Wolf JM. Emerging technologies in osteoporosis diagnosis. J Hand Surg Am. 2019;44:240–3.
Cummings SR, Lorentzon M. Osteoporosis: the evolution of a diagnosis. J Intern Med. 2015;277(6):650–61.
Liu Z, Leung D, Thrush K, Zhao W, Ratliff S, Tanaka T. Underlying features of epigenetic aging clocks in vivo and in vitro. Aging Cell. 2020;19(10):e13229.
Christiansen L, Lenart A, Tan Q, Vaupel JW, Aviv A, McGue M. DNA methylation age is associated with mortality in a longitudinal Danish twin study. Aging Cell. 2016;15:149–54.
Chen BH, Marioni RE, Colicino E, Peters MJ, Ward-Caviness CK, Tsai PC. DNA methylation-based measures of biological age: meta-analysis predicting time to death. Aging (Albany, NY). 2016;8:1844–65.
Horvath S. DNA methylation age of human tissues and cell types. Genome Biol. 2013;14(10):1–20.
Hannum G, Guinney J, Zhao L, Zhang Li, Hughes G, Sadda SriniVas. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell. 2013;49:359–67.
Levine ME, Lu AT, Quach A, Chen BH, Assimes TL, Bandinelli S. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10:573–91.
Lu AT, Quach A, Wilson JG, Reiner AP, Aviv A, Raj K. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany, NY). 2019;11:303–27.
Del Real A, Pérez-Campo FM, Fernández AF, Sañudo C, Ibarbia CG, Pérez-Núñez MI. Differential analysis of genome-wide methylation and gene expression in mesenchymal stem cells of patients with fractures and osteoarthritis. Epigenetics. 2017;12:113–22.
Delgado-Calle J, Fernández AF, Sainz J, Zarrabeitia MT, Sañudo C, García-Renedo R. Genome-wide profiling of bone reveals differentially methylated regions in osteoporosis and osteoarthritis. Arthritis Rheum. 2013;65:197–205.
Boef AGC, Dekkers OM, le Cessie S. Mendelian randomization studies: a review of the approaches used and the quality of reporting. Int J Epidemiol. 2015;44:496–511.
Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. 2014;23:R89–98.
Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008;27:1133–63.
Jones G, Trajanoska K, Santanasto AJ, Stringa N, Kuo CL, Atkins JL. Genome-wide meta-analysis of muscle weakness identifies 15 susceptibility loci in older men and women. Nat Commun. 2021;12(1):654.
Gergei I, Zheng J, Andlauer TF, Brandenburg V, Mirza-Schreiber N, Müller-Myhsok B. GWAS meta-analysis followed by Mendelian randomization revealed potential control mechanisms for circulating α-Klotho levels. Hum Mol Genet. 2022;31:792–802.
Larsson SC, Burgess S. Causal role of high body mass index in multiple chronic diseases: a systematic review and meta-analysis of Mendelian randomization studies. BMC Med. 2021;19:
Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44:512–25.
McCartney DL, Min JL, Richmond RC, Lu AT, Sobczyk MK, Davies G. Genome-wide association studies identify 137 genetic loci for DNA methylation biomarkers of aging. Genome Biol. 2021;22(1):194.
Zheng HF, Forgetta V, Hsu YH, Estrada K, Rosello-Diez A, Leo PJ. Whole-genome sequencing identifies EN1 as a determinant of bone density and fracture. Nature. 2015;526:112–7.
Palmer TM, Lawlor DA, Harbord RM, Sheehan NA, Tobias JH, Timpson NJ. Using multiple genetic variants as instrumental variables for modifiable risk factors. Stat Methods Med Res. 2012;21:223–42.
Thompson SG, Burgess S. Mendelian randomization: methods for causal inference using genetic variants. CRC Press (2021)
Pierce BL, Ahsan H, VanderWeele TJ. Power and instrument strength requirements for Mendelian randomization studies using multiple genetic variants. Int J Epidemiol. 2011;40:740–52.
Deng Y, Wong MCS. Association between rheumatoid arthritis and osteoporosis in Japanese populations: a mendelian randomization study. Arthritis Rheumatol. 2023;75(8):1334–43.
Zhou J, Ye Z, Wei P, Yi F, Ouyang M. Shilang Xiong. Effect of basal metabolic rate on osteoporosis: A Mendelian randomization study. Frontiers in Public Health; 2023. p. 11.
Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol. 2016;40:304–14.
Burgess S, Dudbridge F, Thompson SG. Combining information on multiple instrumental variables in Mendelian randomization: comparison of allele score and summarized data methods. Stat Med. 2016;35:1880–906.
Wu F, Huang Y, Hu J, Shao Z. Mendelian randomization study of inflammatory bowel disease and bone mineral density. BMC Med. 2020;18(1):1–9.
Visconti VV, Cariati I, Fittipaldi S, Iundusi R, Gasbarra E, Tarantino U. DNA methylation signatures of bone metabolism in osteoporosis and osteoarthritis aging-related diseases: an updated review. Int J Mol Sci. 2021;22:4244.
Fernandez-Rebollo E, Eipel M, Seefried L, Hoffmann P, Strathmann K, Jakob F. primary osteoporosis is not reflected by disease-specific DNA methylation or accelerated epigenetic age in blood. J Bone Miner Res. 2018;33:356–61.
McCrory C, Fiorito G, Hernandez B, Polidoro S, O’Halloran AM, Hever A. GrimAge outperforms other epigenetic clocks in the prediction of age-related clinical phenotypes and all-cause mortality. J Gerontol Series A. 2021;76:741–9.
Ebrahim S, Davey Smith G. Mendelian randomization: can genetic epidemiology help redress the failures of observational epidemiology? Hum Genet. 2008;123:15–33.
Wu F, Huang Y, Hu J, Shao Z. Mendelian randomization study of telomere length and bone mineral density. Aging (Albany, NY). 2020;13:2015–30.
Acknowledgements
None.
Funding
There is no funding to report.
Author information
Authors and Affiliations
Contributions
Huafeng Zhang, Zhijun Li and Xinyu Liang conducted the design of this study. Xinglong Zhang, Ran Pang, Zhaohui Jiang, Zhengxuan Liu and Chunlei Xu carried out data collection and collation. Xinyu Liang, Wei Shi and Kai Zhang performed data analysis and wrote the manuscript. Hui Li, Qian Xu, Zhi Jun Li, Wenhao Cui, Dong Li and Xin Wan contributed suggestions for manuscript revision and revised the manuscript.All authors have read and approved the submitted version.Xinyu Liang and Wei Shi contributed equally to this work.
Corresponding authors
Ethics declarations
Ethics approval and consent to participate
Not applicable. The studies for which we have used publicly available GWAS data have received ethical approval from their respective institutional review boards.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Additional file 1:
Supplementary Table 1. Information on epigenetic age instrumental variables. Supplementary Table 2. Information on BMD instrumental variables. Supplementary Table 3. Specific information for MR analysis of epigenetic age as exposure and BMD as outcome. Supplementary Table 4. Specific information for MR analysis of BMD as exposure and epigenetic age as outcome.
Additional file 2:
Supplementary Figure 1. “leave-one-out” analyses for MR analysis of epigenetic age as exposure and FN BMD as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge. Supplementary Figure 2. “leave-one-out” analyses for MR analysis of FN BMD as exposure and epigenetic age as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge. Supplementary Figure 3. “leave-one-out” analyses for MR analysis of epigenetic age as exposure and FA BMD as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge. Supplementary Figure 4. “leave-one-out” analyses for MR analysis of FA BMD as exposure and epigenetic age as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge. Supplementary Figure 5. “leave-one-out” analyses for MR analysis of epigenetic age as exposure and LS BMD as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge. Supplementary Figure 6. “leave-one-out” analyses for MR analysis of LS BMD as exposure and epigenetic age as outcome. (A) GrimAge (B) Hannum (C) HorvathAge (D) PhenoAge.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Liang, X., Shi, W., Zhang, X. et al. Causal association of epigenetic aging and osteoporosis: a bidirectional Mendelian randomization study. BMC Med Genomics 16, 275 (2023). https://doi.org/10.1186/s12920-023-01708-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12920-023-01708-3