Causal relationships between basal metabolic rate and epigenetic age acceleration: bidirectional Mendelian randomization and mediation analysis

doi:10.21203/rs.3.rs-4871047/v1

Download PDF

Research Article

Causal relationships between basal metabolic rate and epigenetic age acceleration: bidirectional Mendelian randomization and mediation analysis

https://doi.org/10.21203/rs.3.rs-4871047/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Basal metabolic rate (BMR) reflecting the body's energy expenditure at rest, may be a crucial indicator of accelerated aging. However, there is a lack of evidence for the causal effect of BMR on accelerated aging. We aimed to apply Mendelian randomization (MR) and mediation analysis to explore the causal relationships between BMR and epigenetic age acceleration.

Method

Genetic variants linked to BMR and epigenetic age (GrimAge, PhenoAge, HannumAge and HorvathAge) acceleration were extracted from genome-wide association study datasets. MR analysis primarily utilizes the Inverse variance weighted (IVW), followed by a range of sensitivity analyses to validate the robustness of our findings.

Results

The univariable MR demonstrated a significant causal effect of BMR on GrimAge (IVW: Beta 0.38 [95% CI 0.12 to 0.64]), and and PhenoAge accelerations (IVW: Beta 0.91 [95%CI 0.67 to 1.45]), but not in the reverse direction. In the multivariate MR analysis adjusting for body mass index (BMI), hypertension, type 2 diabetes mellitus (T2D), physical activity, smoking, alcohol intake frequency, total and bioavailable testosterone levels, we found that the causal effects of BMR on PhenoAge and GrimAge accelerations were both dependent on BMI, and also dependent on T2D, smoking, and bioavailable testosterone levels for GrimAge acceleration. Mediation analysis indicated that 7.24%, 16.29%, 6.95%, and 5.36% of effect from BMR on GrimAge acceleration was mediated by hypertension, T2D, lack of physical activity and alcohol intake.

Conclusion

A positive and unidirectional causal effect of BMR on epigenetic age acceleration, which was dependent of BMI.

Basal metabolic rate

Epigenetic age acceleration

Mendelian randomization

Mediation

Aging is a multifaceted process characterized by a gradual loss of metabolic homeostasis and plasticity, which leads to the onset of a wide range of age-related diseases and ultimately death[1]. Energy metabolism is a critical hallmark of aging[1, 2], and understanding the causal relationship between energy metabolism and aging could provide valuable insights into the mechanisms underlying aging and age-related diseases. Basal metabolic rate (BMR) reflects the body's energy expenditure at rest and plays a crucial role in energy balance, accounting for 60%-70% of total daily energy expenditure of human[3]. Therefore, it is important to clarify the association between BMR and aging.

High BMR casually correlated with impaired kidney function[4], and elevated risks of venous thromboembolism[5], cardiovascular disease[6], osteoporosis[7], and cancer[8, 9]. Despite the casual association of BMR with aging-related conditions has been widely explored, the direct association of BMR with aging is less reported. Previous studies have identified mass-specific BMR or BMR as a reliable biomarker or modified risk of human aging as its relation with mortality rate, survival curve, or age gap, respectively[10, 11]. A cross-section research has also reported that telomere length was negatively correlated with BMR in women[12]. However, no study explored the bidirectional causal relationship between BMR and aging.

Central to the study of aging is the concept of biological age, which can differ from chronological age due to the impact of physiological changes at the molecular and cellular level[13]. The epigenetic age has emerged as a promising tool for assessing biological age and predicting age-related health outcomes[13–16]. To date, several algorithms of epigenetic age have been developed, including HannumAge, Intrinsic HorvathAge, PhenoAge and GrimAge[17, 18]. Their difference with chronological age, that is, the epigenetic age acceleration, reflects the rate at which an individual's biological age is increasing over time and provides a more comprehensive view of aging. Therefore, understanding bidirectional causal relationships between BMR and epigenetic age acceleration could provide valuable insights into the mechanisms underlying aging.

Mendelian randomization (MR) is widely used for causal inference, which can reduce the potential confounding biases inherent in conventional epidemiological studies[19]. To elucidate these relationships, here, by leveraging data from this recent large-scale genome wide association study (GWAS) datasets on BMR and epigenetic age acceleration of HannumAge, Intrinsic HorvathAge, PhenoAge, and GrimAge, we implemented two-sample MR: (i) bidirectional two-sample MR analysis was conducted to explore the associations of genetic predisposition to BMR with epigenetic age acceleration and vice versa; (ii) mediation analysis using multivariable MR (MVMR) and two-step MR was further conducted to elucidate the mechanistic pathways through which BMR impacts (or is impacted by) epigenetic age acceleration; (iii) validate these findings using another GWAS datasets.

Study design

To investigate the genetic relationship between BMR and epigenetic age acceleration (PhenoAge, HannumAge, Intrinsic HorvathAge, and GrimAge accelerations), as well as the influence of potential mediators, we conducted bidirectional two-sample MVMR and two-step MR using large publicly available GWAS databases. The total effect was split into the direct effect and the indirect effect. The indirect effects mediated by mediating factors was also called mediated effect. We explored the mediated effect of BMR on epigenetic age acceleration by two-sample MR analysis. First, we explored BMR casual effect on potential mediating factors. Next, we explored the casual effect of potential mediating factors on epigenetic age acceleration. Finally, a MVMR was conducted to assess the direct effect of exposure on the outcome, with adjustments made for potential mediating factors (hypertension, T2D, BMI, heavy physical activity, light physical activity, lack of physical activity, packing years of smoking, alcohol intake frequency, total testosterone levels, bioavailable testosterone levels). Figure 1 represents the specific flow chart for MR designs.

Data resources for MR analysis

Summary statistics of single nucleotide polymorphisms (SNPs) associated with BMR were obtained from GWAS databases (https://gwas.mrcieu.ac.uk/). BMR SNPs were selected from two different GWAS databases, with GWAS ID ukb-b-16446 as the main analysis (the dataset was constructed by MRC Integrated Epidemiology Unit consortium using UK Biobank) and ukb-a-268 as subsequent replication analysis (the dataset was derived from the Neale Lab Consortium’s summary data from the UK Biobank). These data were drawn from 454874 and 331307 subjects of European ancestry, respectively.

Four types of epigenetic age acceleration, including PhenoAge (n = 107460), GrimAge (n = 34710), HannumAge (n = 34710), and HorvathAge (n = 34710) acceleration, were derived from a recent GWAS meta-analysis of biological aging[20, 21]. SNPs related to the acceleration of epigenetic age acceleration can be found in Supplementary data 1–4.

The SNPs of the 10 potential adjusted factors were all obtained from GWAS database (https://gwas.mrcieu.ac.uk/). These data were drawn from Europeans, which include hypertension (n = 463010), T2D (n = 655666), BMI (n = 681275), heavy physical activity (n = 460376), light physical activity (n = 460376) lack of physical activity (n = 460376), pack years of smoking (n = 124387), alcohol intake frequency (n = 462346), total testosterone levels (n = 425097) and bioavailable testosterone levels (n = 382988). Supplementary data 5–10 exhibit SNPs associated with 10 potential mediators factors.

Since the data used in our study were collected from publicly available sources such as published articles and GWAS databases, ethical approval is therefore waved. The summary information of all data are provided in Table 1.

Table 1

The summary of exposure, mediator indicators and biomarkers of aging.
Trait	Sample Size	Population	p-Value	LD	No. of SNPs	GWASID/PMID
Basal metabolic rate	454874	European	5×10^− 8	0.001	521	ukb-b-16446
Basal metabolic rate	331307	European	5×10^− 8	0.001	378	ukb-a-268
Hypertension	463010	European	5×10^− 8	0.001	70	ukb-b-12493
Body mass index	681275	European	5×10^− 8	0.001	490	Ieu-b-40
Type 2 diabetes	655666	European	5×10^− 8	0.001	118	ebi-a-GCST006867
Heavy physical activity	460376	European	5×10^− 8	0.001	18	ukb-b-13184
Light physical activity	460376	European	5×10^− 8	0.001	13	ukb-b-11495
Lack of physical activity	460376	European	5×10^− 8	0.001	5	ukb-b-15869
Pack years of smoking	124387	European	5×10^− 8	0.001	11	ukb-b-10831
Alcohol intake frequency	462346	European	5×10^− 8	0.001	95	ukb-b-5779
Total testosterone levels	425097	European	5×10^− 8	0.001	163	ebi-a-GCST90012114
Bioavailable testosterone levels	382988	European	5×10^− 8	0.001	107	ebi-a-GCST90012104
PhenoAge*	107460	European	5×10^− 8	0.001	16446666	34038024
GrimAge*	34710	European	5×10^− 8	0.001	7568149	34187551
HannumAge*	34710	European	5×10^− 8	0.001	7565493	34187551
HorvathAge*	34710	European	5×10^− 8	0.001	7567980	34187551
LD, linkage disequilibrium.
*, indicates the SNPs for PhenoAge, GrimAge, HannumAge, and HorvathAge accelerations, respectively.

Selection of instrumental variables

We chose to utilize SNPs that had a significant association with exposure (P < 5×10^− 8) from GWAS databases as our instrumental variables (IVs). Then, in order to ensure independence among each SNPs, we conducted linkage disequilibrium (LD) analysis (kb = 10000, r² = 0.001) on the candidate IVs. The F-statistic (F) can be employed to evaluate the robustness of genetic instrumental variables. When F < 10, it is considered a weak instrumental variable. The presence of weak instrumental variables in the analysis give rise to weak instrument bias. Therefore, instrumental variables with F > 10 are excluded. The formula for calculating the F-statistic is as follows:

F=\(\:\frac{{Beta\:\left(\beta\:\right)}^{2}}{{Standard\:error\:\left(SE\right)}^{2}}\).

Additionally, we carried out SNP harmonization to rectify the orientation of alleles.

Statistical analysis

The bidirectional two-sample MR analysis was performed to assess the causal relationship between BMR and four types of epigenetic age acceleration. We primarily employed the IVW random effect to assess the causality. It is possible to obtain unbiased estimates from IVW as long as all genetic variants are valid. We also carried out secondary analyses using the weighted median, weighted mode and MR Egger method. Using the weighted median, valid estimates are obtained when at least 50% of the weight comes from valid variants. MR-Egger can capture horizontal pleiotropy when the instrument strength is independent of the direct effect. However, due to potential limitations of MR-Egger in statistical power, our emphasis was on the direction and magnitude of the effect rather than on its statistical significance.

MVMR and two-step MR analyses were performed to assess potential mediators if there is a causal relationship. The first step was to estimate the causal effect of BMR on mediating factors using 2SMR, we can get a β1. In the second step, we employed 2SMR approach to evaluate the impact of the mediating factors on outcomes, we can get a β2, and MVMR was utilized to examine the association between BMR and outcomes while adjusting for the mediators, yielding β3. The total effect of BMR was decomposed into a direct effect (the effect of BMR on four types of epigenetic age independent of the mediator) and an indirect effect (the effect of BMR on four types of epigenetic age). The total effect of an exposure on the outcome is estimated by univariable MR. The following formula was used to estimate the proportions of mediating effect (E%):

E (%) = \(\:\frac{{\sum\:}_{k=1}^{k}\beta\:1\times\:\beta\:2k}{{\sum\:}_{k=1}^{k}\beta\:3+\beta\:1\times\:\beta\:2k}\)

For sensitivity analysis, we applied MR-Egger global test to assess pleiotropy, and inferred the presence of pleiotropy when the MR-Egger intercept value significantly deviates from zero[22]. MR-Egger P < 0.05 reveals a pleiotropic effect, suggesting the invalidity of the IVW estimate. We employed Cochran’s Q test assess the presence of heterogeneity among the chosen IVs[23]. We defined the presence of heterogeneity if the Cochran’s Q P ≤ 0.05 and consequently, we used a random-effect IVW method in MR analysis. To ensure the reliability of our findings, we conducted MR pleiotropy residual sum and outlier test (MR-PRESSO) and “leave-one-out” analyses to identify outliers, and, if necessary, MR-PRESSO to correct for horizontal pleiotropic outliers. Furthermore, to confirm the robustness of our results, we carried out replication analysis.

Analyses were performed by the R software (version 4.3.2), employing TwoSampleMR (version 0.5.8), MRPRESSO (version 1.0), as well as the ggplot2 (version 3.4.4) R package and associated functions.

Overview of the instruments

Table 1 presents the summary of genetic instruments. After rigorous screening (P < 5×10^− 8, r² = 0.001), we respectively included 521 and 392 SNPs as instruments for BMR, 70 SNPs for hypertension, 490 SNPs for BMI, 118 SNPs for T2D, 18 SNPs for heavy physical activity, 13 SNPs for light physical activity, 5 SNPs for lack of physical activity, 11 SNPs for pack years of smoking, 95 SNPs for alcohol intake frequency, 163 SNPs for total testosterone levels, 107 SNPs for bioavailable testosterone levels, 16446666 SNPs for PhenoAge acceleration, 7568149 SNPs for GrimAge acceleration, 7565493 SNPs for HannumAge acceleration, and 7567980 SNPs for HorvathAge acceleration Supplementary data 1–16 provided descriptive information on the instruments used. The specific characteristics of these instrumental SNPs encompassed the effect, standard error (SE), effect allele, effect allele frequency (Eaf), other allele, F-statistic (F), etc. The F for each instrumental variable were observed to be higher than 10, indicating the satisfactory instrumental strength.

Total effect of BMR on four types of epigenetic age acceleration

The two-sample MR analysis was first performed to explore the total effect of BMR on four types of epigenetic age acceleration (Fig. 2). The main IVW method indicated a positive association between genetically predicted BMR and GrimAge acceleration (IVW: Beta 0.38 [95% CI 0.12 to 0.64]). Additionally, there was a positive relationship between genetically predicted BMR and PhenoAge acceleration (IVW: Beta 0.91 [95%CI 0.67 to 1.45]). However, genetically predicted BMR was not significantly correlated with HannumAge acceleration and (IVW: Beta 0.06 [95%CI -0.21 to 0.33]) HorvathAge acceleration (IVW: Beta 0.19 [95%CI -0.09 to 0.47]). Furthermore, scatter plots demonstrated the above impact of BMR on the four epigenetic age acceleration (Figure S1). All the above results were validated by MR-Egger, weighted median, and weighted mode methods (Fig. 2 and Table S1).

In sensitivity analyses, MR-Egger intercept test found no evidence of pleiotropy between BMR and any of the four epigenetic age acceleration (each P for intercept > 0.05, Table S1). Cochran’s Q test suggested evidence of heterogeneity (each P ≤ 0.05), indicating that random effect IVW model was applied in this study (Table S1). We also utilized MR-PRESSO to identify and exclude any potential IVs outliers. After removing the outliers through MR-PRESSO, our results remained robust (Table S1). Moreover, funnel plots were symmetrical without sign of bias and leave-one-out analysis further revealed that no SNP drove the findings (Figures S2-S6). Overall, the sensitivity analysis demonstrates the stability of the results obtained from the BMR and epigenetic age acceleration.

In addition, we conducted a reverse causality analysis, and the findings found the absence of the causal effect of genetically predicted GrimAge, PhenoAge, HannumAge and HorvathAge accelerations and BMR (Figure S7).

Causal Effect of BMR on potential adjusted factors

Figure 3 presents the causal effect of BMR on potential mediating factors. We observed a significant causal relationship between genetic prediction of BMR and hypertension (IVW: Beta 0.024 [95% CI 0.016 to 0.033]), BMI (IVW: Beta 0.676 [95% CI 0.605 to 0.747]), T2D (IVW: Beta 0.543 [95% CI 0.402 to 0.684]), lack of physical activity (IVW: Beta 0.006 [95% CI 0.001 to 0.011]), pack years of smoking (IVW: Beta 0.102[95% CI 0.068 to 0.136]), and alcohol intake frequency (IVW: Beta 0.068 [95% CI 0.025 to 0.111]). Conversely, no causal relationship was detect between genetic prediction of BMR and heavy physical activity (IVW: Beta − 0.007 [95% CI -0.017 to 0.002]), light physical activity (IVW: Beta 0.001 [95% CI -0.010 to 0.012]), total testosterone levels (IVW: Beta − 0.023 [95% CI -0.048 to 0.001]), and bioavaible testosterone levels (IVW: Beta 0.007 [95% CI -0.014 to 0.028]). Similar results were marginally found by MR-Egger, weighted median, and weighted mode.

In sensitivity analyses, Cochran’s Q test suggested the presence of heterogeneity (each P ≤ 0.05, Table S2). MR-Egger intercept test found no evidence of pleiotropy between BMR and any of the ten potential mediating factors (each P for intercept > 0.05, Table S2). After removing the outliers through MR-PRESSO, our results remained unchanged except for total testosterone levels (IVW: Beta − 0.04 [95% CI -0.05 to -0.02]) (Table S1).

Causal effect of potential adjusted factors on GrimAge and PhenoAge accelerations

As BMR was only significantly associated with GrimAge and PhenoAge accelerations, only the causal effect of potential mediating factors on GrimAge and PhenoAge accelerations were further analyzed (Fig. 4). A significant causal relationship was found between genetically predicted BMI (IVW: Beta 0.604 [95% CI 0.407 to 0.801]), T2D (IVW: Beta 0.119 [95% CI 0.019 to 0.218]) and packing years of smoking (IVW: Beta 0.975 [95% CI 0.078 to 1.871]) and GrimAge. However, there was no significant correlation between gene prediction of hypertension, heavy physical activity, light physical activity, lack of physical activity, alcohol intake frequency, total testosterone levels, and bioavaible testosterone levels and GrimAge. Meanwhile, there was a positive correlation observed between genetically predicted hypertension (IVW: Beta 2.865 [95% CI 1.151 to 4.578]), BMI (IVW: Beta 1.303 [95% CI 1.135 to 1.472]), T2D (IVW: Beta 0.299 [95% CI 0.190 to 0.408]) and lack of physical activity (IVW: Beta 12.571 [95% CI 6.252 to 18.884]), packing years of smoking (IVW: Beta 0.912 [95% CI 0.075 to 1.748]), and alcohol intake frequency (IVW: Beta 0.858 [95% CI 0.527 to 1.190]). In contrast, PhenoAge showed no significant correlation with genetically prediction of heavy physical activity, light physical activity, total testosterone levels, and bioavaible testosterone levels. Likewise, these results were marginally supported by MR-Egger, weighted median, and weighted mode.

Cochran’s Q test suggested the presence of heterogeneity for some of potential mediating factors (Table S3). MR-Egger intercept test found no evidence of pleiotropy except for the effect of hypertension on GrimAge (P for intercept = 0.03, Table S3).

MVMR analysis and mediated effect by mediating factors

MVMR analysis and mediated effect were calculated (Table 2). In the MVMR of PhenoAge acceleration results for BMR exposure, after individually adjusting for hypertension, T2D, heavy physical activity, lack of physical activity, light physical activity, pack years of smoking, alcohol intake frequency, total testosterone levels, and bioavailable testosterone levels, we still observed a significant causal relationship. However, the causal effect became nonsignificant after adjusting BMI. Overall, hypertension, T2D, lack of physical exercise and alcohol intake frequency were considered as the mediating factors for the causal effect of BMR on PhenoAge acceleration. The mediation effects were 7.24%, 16.29%, 6.95% and 5.36%, respectively. In the MVMR of GrimAge acceleration results for BMR exposure, after adjusting for above potential mediation factors, except for BMI or T2D, all results showed significant causal relationships. Overall, smoking was considered as the mediating factor for the causal effect of BMR on GrimAge acceleration. The mediated effect adjusted was 22.95% for packing years of smoking.

Table 2

Multivariate MR analysis of the direct effect of adjusted factors on PhenoAge and GrimAge.
Exposure/Outcome	Adjusted Factors	Multivariate MR Analysis			Mediation Effect (%) (95%CI)
Exposure/Outcome	Adjusted Factors	nSNP	p-Value	β (95%CI)	Mediation Effect (%) (95%CI)
BMR/PhenoAge	None	516	1.55×10^− 13	0.91 (0.67 to 1.15)
BMR/PhenoAge	Hypertension	433	6.63×10^− 13	0.94 (0.68 to 1.20)	7.24 (2.65 to 11.07)
BMR/PhenoAge	BMI	261	0.80	0.06 (-0.40 to 0.67)	/
BMR/PhenoAge	T2D	314	2.63×10^− 8	0.83 (0.54 to 1.13)	16.29 (12.39 to 19.83)
BMR/PhenoAge	Heavy physical activity	449	4.81×10^− 16	1.02 (0.78 to 1.27)	/
BMR/PhenoAge	Lack of physical activity	446	1.89×10^− 15	1.00 (0.75 to 1.25)	6.95 (0.81 to 9.03)
BMR/PhenoAge	Light of physical activity	448	3.10×10^− 17	1.06 (0.81 to 1.30)	/
BMR/PhenoAge	Pack years of smoking	443	2.01×10^− 13	0.94 (0.69 to 1.19)	/
BMR/PhenoAge	Alcohol intake frequency	441	1.52×10^− 15	1.03 (0.78 to 1.29)	5.36 (1.67 to 9.33)
BMR/PhenoAge	Total testosterone levels	383	1.25×10^− 14	1.06 (0.79 to 1.33)	/
BMR/PhenoAge	Bioavailable testosterone levels	403	6.50×10^− 17	1.14 (0.87 to 1.40)	/
BMR/PhenoAge	Hypertension, T2D	304	1.55×10^− 13	0.76 (0.46 to 1.06)	18.71 (17.13 to 28.80)
BMR/PhenoAge	T2D, hypertension, lack of physical activity	304	6.79×10^− 7	0.74 (0.45 to 1.57)	33.52 (22.69 to 60.83)
BMR/PhenoAge	T2D, hypertension, lack of physical activity, alcohol intake frequency	296	8.28×10^− 7	0.73 (0.44 to 1.03)	42.01 (26.30 to 74.31)
BMR/PhenoAge	Hypertension, T2D, heavy physical activity, lack of physical activity, light physical activity, pack years of smoking, alcohol intake frequency, total testosterone levels, and bioavailable testosterone levels	252	9.80×10^− 7	0.79 (2.20 to 0.47)	/
BMR/GrimAge	None	444	0.0048	0.38 (0.12 to 0.64)
BMR/GrimAge	Hypertension	427	0.00303	0.14 (0.14 to 0.68)	/
BMR/GrimAge	BMI	261	0.39	-0.19 (-0.62 to 0.24)	/
BMR/GrimAge	T2D	315	0.13	0.24 (-0.07 to 0.56)	/
BMR/GrimAge	Heavy physical activity	443	0.004	0.38 (0.12 to 0.64)	/
BMR/GrimAge	Lack of physical activity	441	0.0076	0.36 (0.10 to 0.62)	/
BMR/GrimAge	Light physical activity	442	0.004	0.38 (0.12 to 0.64)	/
BMR/GrimAge	Pack years of smoking	437	0.03	0.29 (0.03 to 0.56)	22.95 (16.78 to 31.07)
BMR/GrimAge	Alcohol intake frequency	435	0.008	0.36 (0.09 to 0.62)	/
BMR/GrimAge	Total testosterone levels	380	0.01	0.36 (0.09 to 0.64)	/
BMR/GrimAge	Bioavailable testosterone levels	398	0.0009	0.46 (0.19 to 0.73)	/
BMR/GrimAge	Hypertension, heavy physical activity, lack of physical activity, light physical activity, pack years of smoking, alcohol intake frequency, total testosterone, and bioavailable testosterone levels	252	0.046	0.34 (0.006 to 0.68)	/
BMR GWAS id: ukb-b-16446; BMR, basal metabolic rate; BMI, body mass index; T2D, type 2 diabetes; /, the results of the two-step Mendelian analysis are nonsensical, hence it is impossible to calculate mediating effects.

Replication analysis

To ensure the credibility of our findings, we carefully selected BMR SNPs from an alternative GWAS database to conduct replication analyses. The causal effects of BMR on PhenoAge and GrimAge accelerations and potential mediating factors were align with our previous findings (Figures S8 and S9). The MVMR and mediation results between BMR and PhenoAge acceleration did not materially change in the replicate analysis (Figure S2 and Table S4). however, The MVMR and mediation results revealed that the causal effect of BMR on GrimAge acceleration was dependent on not only BMI and T2D, but also pack years of smoking and bioavailable testosterone levels in the replicate analysis (Figure S2 and Table S4).

To our knowledge, this study is first to analyze the causal association of BMR with epigenetic age (GrimAge, PhenoAge, HannumAge and HorvathAge) acceleration. We found that, in univariable MR analyses, genetically predicted higher BMR was causally associated with an increased risk of GrimAge and PhenoAge accelerations, but not HannumAge and HorvathAge accelerations. However, in the reverse MR analysis, genetic liability to these four epigenetic age acceleration measures did not show significant associations with BMR. Multivariable MR analyses found that the causal effect of BMR on PhenoAge acceleration was independent on hypertension, T2D, heavy physical activity, lack of physical activity, light physical activity, pack years of smoking, alcohol intake frequency, total testosterone levels, and bioavailable testosterone levels, but dependent on BMI., The causal effect of BMR on GrimAge acceleration was independent on most of variables mentioned above except for BMI, hypertension, pack years of smoking, and bioavailable testosterone levels. Mediation analysis revealed that hypertension, T2D, lack of physical exercise and alcohol intake frequency mediated the causal effect of BMR on PhenoAge but not GrimAge acceleration. Above results were supported by sensitivity and validation analyses.

Consistent with previous findings in the cross-sectional analysis[12], we found that elevated BMR was causally associated with an increased risk of epigenetic age acceleration, as reflected by GrimAge and PhenoAge accelerations; however, causal associations were not detected for HannumAge and HorvathAge accelerations. We speculated that the inconsistent result may be in part due to the distinct compositions included in each measure. HannumAge and Intrinsic HorvathAge were trained using DNAm levels at specific CpG sites associated closely with chronological age[24, 25]. PhenoAge and GrimAge incorporated data from additional CpG sites and clinical biomarkers, providing improved prediction of age-related health outcomes[26, 27]. These clinical biomarkers include systolic blood pressure, cholesterol, glycated hemoglobin, C-reactive protein, or serum glucose, which are strongly related to energy metabolism, thus further affects BMR. Supported by our speculation, a Mendelian randomization analysis has found that systolic blood pressure and cholesterol were causally associated with longevity[28]. However, the specific reasons for the difference in causal associations can vary, and further research is needed to elucidate the underlying mechanisms. Further, we did not find a significant causal effect of four types of epigenetic age acceleration on BMR in the reverse MR analysis, which indicated that high BMR is the driver, but not the result, of epigenetic aging. Therefore, targeting BMR may be a effective strategy to promote healthy aging.

A positive causal effect was observed between BMR and hypertension, BMI, T2D, lack of physical activity, pack years of smoking, and alcohol intake frequency. Previous epidemiology study further supported our findings and has showed a positive association between BMR and blood pressure[29]. It is surprised that genetically determined higher BMR was causally associated with an increased BMI, in line with previous study[30]. Genetic factors interact with environmental influences, such as diet and physical activity, to determine overall health and weight outcomes. Individuals with higher BMR may have faster metabolic rates, but if they lack physical activity or are rich in food, they may also be more susceptible to store energy as fat, leading to weight gain. Additional research is needed to validate these findings and delve into the biological mechanisms driving this phenomenon.

Abundant researches have suggested preexisting metabolic dysfunction (e.g., BMI, hypertension, and T2D) may aggravate the aging progression[11, 31–33]. Aging is also well known to related with unhealthy lifestyle (e.g., physical activity, smoking, and alcohol intake) and showed a sex-specific[34–37]. For instance, higher BMI, childhood obesity, and T2D risk were causally associated with GrimAge and PhenoAge accelerations[32]. Genetically determined T2D and BMI were associated with a lower odds of longevity[38], or increased risk of frailty[28]. BMI, diabetes and physical activity showed statistically significant relationships with epigenetic age acceleration[39]. Consistent with our findings, we found a significant causal effect of genetically predicted hypertension, BMI, T2D and lack of physical activity, packing years of smoking, or alcohol intake frequency on PhenoAge or GrimAge accelerations. However, no causal association was found for testosterone level, a measure to reflect sex-hormone level, indicating that sex may not impact PhenoAge or GrimAge accelerations.

The MVMR analysis found that, independent of most factors, the causal effects of BMR on PhenoAge and GrimAge accelerations were dependent on BMI, and also dependent on T2D, smoking, and bioavailable testosterone levels for GrimAge acceleration. Mediation MR further suggested that increased risk of hypertension, T2D, lack of physical activity and alcohol intake mediated the causal association of BMR on PhenoAge but not GrimAge acceleration. Notably, the causal effects of BMR on both PhenoAge and GrimAge accelerations became non-significant after adjusting for BMI. BMI was associated with increased risk of T2D, smoking, and bioavailable testosterone levels[40–42]. It may be strongly suggested that the causal relationship between the ratio of BMR to body weight and epigenetic age acceleration merited to be investigated. If so, the causal effects of BMR/BMI on GrimAge acceleration may be independent on T2D, smoking, and bioavailable testosterone levels. Thus, we speculated that the ratio of BMR to BMI may be a more stable measure to independently reflect aging than BMR alone. Measure from animal experiments supported our speculation showing that the ratio of BMR to body weight rather than BMR alone was applied[43, 44]. However, the GWAS study based on the ratio of BMR to body weight is lacking. Further research was required to explore the SNP associated with the ratio of BMR to BMI.

This study had several advantages. Although observation study has suggested a link between BMR and epigenetic age acceleration, the causal relationship is unknown. Although MR Analysis is less indicative of causality compared to high-quality RCTs, the implementation of an RCT based on BMR and epigenetic age acceleration presents challenges. Therefore, MR may be a more feasible option for conducting the study. This study is first to analyze the causal association of BMR with epigenetic age acceleration by MR analysis. Second, we applied a bidirectional MR analysis to investigate the bidirectional causal association of BMR and epigenetic age acceleration, which was less influenced by reverse causation and potential confounding bias. Second, multiple sensitivity analyses, including MR-Egger, weighted median, and MR-PRESSO, were carried out to bolster the reliability of the results. Moreover, two independent BMR-associated GWAS datasets were employed, which increased external validity of the results. Last, all GWAS datasets were derived from individuals of European descent, which could minimize the potential bias of population stratification.

Our findings also have some limitations. This study was predominantly limited to European population, and whether our findings can be generalized to other populations are unclear. In addition, the causal effect of BMR on epigenetic age acceleration was dependent of BMI. Therefore, we speculated that the ratio of BMR to BMI may be a better predictor for aging than BMR alone However, the SNP associated with the ratio of BMR to BMI is lacking. Further GWAS study should focus on screening SNP associated with the ratio of BMR to BMI. We were also unable to detect non-linear associations due to the lack of individual data. Finally, we may still have unrecognized confounding factors between our exposure and outcome variables even though no pleiotropy was detected in this study.

Our MR analysis provides evidence of a causal effect of BMR on epigenetic age acceleration, which was independent on hypertension, T2D, physical activity, smoking, alcohol intake frequency, or sex hormone levels, but dependent of BMI. However, no evidence was found for a reverse causal effect of epigenetic age acceleration on BMR. Our study suggests that a high BMR could accelerate the aging process and the direction is critical to informing prevention strategies.

Acknowledgements

Jibin. Li and D. Shi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: D. Shi. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Mingyi. Wang. Critical revision of the manuscript for important intellectual content: Mingyi. Wang and D, Shi. Statistical analysis: Mingyi. Wang. Obtained funding: D. Shi. Administrative, technical, or material support: All authors. Supervision: Jibin. Li and D. Shi.

Conflict of interest disclosures

The authors declare no conflict of interest.

Funding support

This study was funded by Dan Shi from Start-up grant for young innovative talents in scientific research of Chongqing medical university (R4028). The funders had no role in the design and conduct of the study, collection, analysis, and interpretation of the data.

Acknowledgement

We would like to thank the participants and staff of UK Biobank for their valuable contributions.

Data sharing

No additional data available

Smith HJ, Sharma A, Mair WB. Metabolic Communication and Healthy Aging: Where Should We Focus Our Energy? Dev Cell. 2020;54:196-211.
Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G. Hallmarks of aging: An expanding universe. Cell. 2023;186:243-278.
Ravussin E, Lillioja S, Anderson TE, Christin L, Bogardus C. Determinants of 24-hour energy expenditure in man. Methods and results using a respiratory chamber. J Clin Invest. 1986;78:1568-1578.
Zhou C, Peng Y, Zhan L, Zha Y. Causal relationship between basal metabolic rate and kidney function: a bidirectional two-sample mendelian randomization study. Front Endocrinol (Lausanne). 2024;15:1319753.
Huang J, Xie Y. Genetically predicted basal metabolic rate and venous thromboembolism risk: a Mendelian randomization study. Front Nutr. 2023;10:1263804.
Zhao P, Han F, Liang X, et al. Causal Effects of Basal Metabolic Rate on Cardiovascular Disease: A Bidirectional Mendelian Randomization Study. J Am Heart Assoc. 2024;13:e031447.
Zhou J, Ye Z, Wei P, et al. Effect of basal metabolic rate on osteoporosis: A Mendelian randomization study. Front Public Health. 2023;11:1096519.
Kliemann N, Murphy N, Viallon V, et al. Predicted basal metabolic rate and cancer risk in the European Prospective Investigation into Cancer and Nutrition. Int J Cancer. 2020;147:648-661.
Ng JCM, Schooling CM. Effect of Basal Metabolic Rate on Cancer: A Mendelian Randomization Study. Front Genet. 2021;12:735541.
Kitazoe Y, Kishino H, Tanisawa K, Udaka K, Tanaka M. Renormalized basal metabolic rate describes the human aging process and longevity. Aging Cell. 2019;18:e12968.
Liu WS, You J, Ge YJ, et al. Association of biological age with health outcomes and its modifiable factors. Aging Cell. 2023;22:e13995.
Nonsa-Ard R, Aneknan P, Tong-Un T, Honsawek S, Leelayuwat C, Leelayuwat N. Telomere Length Is Correlated with Resting Metabolic Rate and Aerobic Capacity in Women: A Cross-Sectional Study. Int J Mol Sci. 2022;23.
Moqri M, Herzog C, Poganik JR, et al. Biomarkers of aging for the identification and evaluation of longevity interventions. Cell. 2023;186:3758-3775.
Moqri M, Herzog C, Poganik JR, et al. Validation of biomarkers of aging. Nat Med. 2024;30:360-372.
Hillary RF, Stevenson AJ, McCartney DL, et al. Epigenetic measures of ageing predict the prevalence and incidence of leading causes of death and disease burden. Clin Epigenetics. 2020;12:115.
Horvath S, Raj K. DNA methylation-based biomarkers and the epigenetic clock theory of ageing. Nat Rev Genet. 2018;19:371-384.
Duan R, Fu Q, Sun Y, Li Q. Epigenetic clock: A promising biomarker and practical tool in aging. Ageing Res Rev. 2022;81:101743.
Li A, Koch Z, Ideker T. Epigenetic aging: Biological age prediction and informing a mechanistic theory of aging. J Intern Med. 2022;292:733-744.
Sekula P, Del Greco MF, Pattaro C, Kottgen A. Mendelian Randomization as an Approach to Assess Causality Using Observational Data. J Am Soc Nephrol. 2016;27:3253-3265.
Kuo CL, Pilling LC, Liu Z, Atkins JL, Levine ME. Genetic associations for two biological age measures point to distinct aging phenotypes. Aging Cell. 2021;20:e13376.
McCartney DL, Min JL, Richmond RC, et al. Genome-wide association studies identify 137 genetic loci for DNA methylation biomarkers of aging. Genome Biol. 2021;22:194.
Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44:512-525.
Bowden J, Spiller W, Del Greco MF, et al. Improving the visualization, interpretation and analysis of two-sample summary data Mendelian randomization via the Radial plot and Radial regression. Int J Epidemiol. 2018;47:1264-1278.
Hannum G, Guinney J, Zhao L, et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell. 2013;49:359-367.
Horvath S. DNA methylation age of human tissues and cell types. Genome Biol. 2013;14:R115.
Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10:573-591.
Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY). 2019;11:303-327.
Ni X, Su H, Lv Y, et al. Modifiable pathways for longevity: A Mendelian randomization analysis. Clin Nutr. 2023;42:1041-1047.
Ali N, Mahmood S, Manirujjaman M, et al. Hypertension prevalence and influence of basal metabolic rate on blood pressure among adult students in Bangladesh. BMC Public Health. 2017;18:58.
Ning L, He C, Lu C, Huang W, Zeng T, Su Q. Association between basal metabolic rate and cardio-metabolic risk factors: Evidence from a Mendelian Randomization study. Heliyon. 2024;10:e28154.
Chen Z, Chen Z, Jin X. Mendelian randomization supports causality between overweight status and accelerated aging. Aging Cell. 2023;22:e13899.
Kong L, Ye C, Wang Y, et al. Genetic Evidence for Causal Effects of Socioeconomic, Lifestyle, and Cardiometabolic Factors on Epigenetic-Age Acceleration. J Gerontol A Biol Sci Med Sci. 2023;78:1083-1091.
Chen M, Wang Z, Xu H, Teng P, Li W, Ma L. Association between modifiable lifestyle factors and telomere length: a univariable and multivariable Mendelian randomization study. J Transl Med. 2024;22:160.
Cui F, Tang L, Li D, et al. Early-life exposure to tobacco, genetic susceptibility, and accelerated biological aging in adulthood. Sci Adv. 2024;10:eadl3747.
Ortola R, Garcia-Esquinas E, Carballo-Casla A, Sotos-Prieto M, Banegas JR, Rodriguez-Artalejo F. Alcohol consumption patterns and unhealthy aging among older lifetime drinkers from Spain. Drug Alcohol Depend. 2022;235:109444.
Hagg S, Jylhava J. Sex differences in biological aging with a focus on human studies. Elife. 2021;10.
Thomas A, Belsky DW, Gu Y. Healthy Lifestyle Behaviors and Biological Aging in the U.S. National Health and Nutrition Examination Surveys 1999-2018. J Gerontol A Biol Sci Med Sci. 2023;78:1535-1542.
van Oort S, Beulens JWJ, van Ballegooijen AJ, Burgess S, Larsson SC. Cardiovascular risk factors and lifestyle behaviours in relation to longevity: a Mendelian randomization study. J Intern Med. 2021;289:232-243.
Oblak L, van der Zaag J, Higgins-Chen AT, Levine ME, Boks MP. A systematic review of biological, social and environmental factors associated with epigenetic clock acceleration. Ageing Res Rev. 2021;69:101348.
Sahle BW, Chen W, Rawal LB, Renzaho AMN. Weight Gain After Smoking Cessation and Risk of Major Chronic Diseases and Mortality. JAMA Netw Open. 2021;4:e217044.
Zornitzki T, Tshori S, Shefer G, Mingelgrin S, Levy C, Knobler H. Seasonal Variation of Testosterone Levels in a Large Cohort of Men. Int J Endocrinol. 2022;2022:6093092.
Shen QM, Li HL, Li ZY, et al. Joint impact of BMI, physical activity and diet on type 2 diabetes: Findings from two population-based cohorts in China. Diabet Med. 2022;39:e14762.
Shi D, Fang G, Chen Q, Li J, Ruan X, Lian X. Six-hour time-restricted feeding inhibits lung cancer progression and reshapes circadian metabolism. BMC Med. 2023;21:417.
Shi D, Han T, Chu X, et al. An isocaloric moderately high-fat diet extends lifespan in male rats and Drosophila. Cell Metab. 2021;33:581-597 e589.

No competing interests reported.

Download PDF

Reviewers agreed at journal
03 Sep, 2024
Reviewers invited by journal
30 Aug, 2024
Editor assigned by journal
08 Aug, 2024
Submission checks completed at journal
08 Aug, 2024
First submitted to journal
06 Aug, 2024

You are reading this latest preprint version

Causal relationships between basal metabolic rate and epigenetic age acceleration: bidirectional Mendelian randomization and mediation analysis

Status:

Version 1

Abstract

Background

Method

Results

Conclusion

Figures

Introduction

Methods

Study design

Data resources for MR analysis

Selection of instrumental variables

Statistical analysis

E (%) = \(\:\frac{{\sum\:}_{k=1}^{k}\beta\:1\times\:\beta\:2k}{{\sum\:}_{k=1}^{k}\beta\:3+\beta\:1\times\:\beta\:2k}\)

Results

Overview of the instruments

Total effect of BMR on four types of epigenetic age acceleration

Causal Effect of BMR on potential adjusted factors

Causal effect of potential adjusted factors on GrimAge and PhenoAge accelerations

MVMR analysis and mediated effect by mediating factors

Replication analysis

Discussion

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1