Putative causal inference for the relationship between obesity and sex hormones in males: a bidirectional Mendelian randomization study

Introduction

Obesity is a highly prevalent chronic disease (Burki, 2021) and seriously endangers public health (Kivimaki et al., 2022; Salah, Ghandour & Husseini, 2021). The worldwide prevalence of obesity has increased markedly during the past three decades (Ng et al., 2014; Pan, Wang & Pan, 2021; Withrow & Alter, 2011). According to the Global Burden of Disease Study, approximately 603.7 million adults were obese in 2015, double the number in 1980 (GBD 2015 Obesity Collaborators, 2017). The prevailing rate of obesity in adults was 39.8% in the United States in 2015–2016 (Hales et al., 2017). A demographic investigation indicated that the morbidity rate of obesity was approximately 16.4% for adults in China in 2015–2019 (Pan, Wang & Pan, 2021). Recent data from the WHO reported that the incidence was approximately 60% for overweight or obese individuals in the European population (Boutari & Mantzoros, 2022). Obesity is an established risk factor for many diseases, including cardiovascular diseases, diabetes, and cancer (Haslam & James, 2005; Kivimaki et al., 2022; Lingvay et al., 2022; Riaz et al., 2018), and body mass index (BMI) is a primary indicator for assessing obesity clinically (Smith & Smith, 2016). Although the prevalence rate of obesity varies by country or region, the overall trend is increasing yearly. Based on the harm of obesity to health, it is necessary to explore the potential mechanism of obesity to health to formulate public prevention policies.

Testosterone and estradiol play a crucial role in human growth and health (Kloner et al., 2016; McEwen & Milner, 2017; Mielke & Miller, 2021) and influence the development of the human nervous system by regulating many cellular and molecular processes (McEwen & Milner, 2017). They can affect human reproductive function by regulating spermatogenesis in males and follicular development in females (de Kretser et al., 1998; Roy, 1994). Imbalance in testosterone or estradiol level is associated with the risk of cardiovascular diseases (Kloner et al., 2016; Raparelli et al., 2022), diabetes (Saikia, Jabbar & Das, 2021; Vikan et al., 2010), and cancer (Ruth et al., 2020; Tin Tin, Reeves & Key, 2021). Sex hormone-binding globulin (SHBG) is a crucial protein that can bind to testosterone and estradiol in the blood and regulate the transport, tissue delivery, bioactivity, and metabolism of the hormones (Anderson, 1974; Bourebaba et al., 2022; Breckwoldt, Zahradnik & Wieacker, 1989; Winters, 2020). Obesity has a significant influence on male sex hormones and SHBG levels. Substantial differences in SHBG, testosterone, and estradiol levels have been reported among individuals with different BMIs (Grasa et al., 2017; Tchernof & Despres, 2000; Wei et al., 2009). Furthermore, obesity influences the specific binding of sex hormones to SHBG (Grasa et al., 2017). The effect of obesity on the risk of diseases can be mediated by sex hormones (Haffner et al., 1993). Therefore, understanding the potential causality between obesity and sex hormone level is essential to controlling the harmful effects of obesity on health.

Mendelian randomization (MR) is a powerful epidemiological method, and it mainly uses single-nucleotide polymorphisms (SNPs) from genome-wide association study (GWAS) data as instrumental variables to estimate the putative causal relationship between an exposure of interest and outcome of interest (Emdin, Khera & Kathiresan, 2017; Smith & Ebrahim, 2003). In this work, we investigated the potential causality among obesity (measured as BMI), sex hormones (bioavailable testosterone (BioT) and estradiol levels), and SHBG by using the MR method. The inverse-variance-weighted (IVW) method was utilized as a primary causal estimate algorithm. Given that obesity and sex hormone imbalance are hazardous to human health, further expounding the potential causality between obesity and sex hormone imbalance may facilitate the mapping of public health policy for promoting health.

Materials and Methods

Study design and data sources

The work is a bidirectional MR investigation. The study design is displayed in Fig. 1A. The GWAS summary-level data for SHBG, BioT, estradiol, and BMI were derived from the large sample studies of European ancestry. All GWAS data in this work are available in the IEU Open GWAS database (https://gwas.mrcieu.ac.uk/) and the Neale lab (http://www.nealelab.is/). BioT, estradiol, SHBG, and BMI were used as exposure data and outcomes in estimating the putative causal relationship between sex hormone-related traits and obesity. The IVW approach was used as the primary causal estimate algorithm in the bidirectional MR study.

Results

A summary of the analysis results is shown in Fig. 1B. For each MR analysis, the F statistics of the instrumental variables were all greater than 10. Detailed information on SNPs used as instrumental variables can be found in Tables S1–S10 in the Supplemental Materials.

Effect of BMI on BioT, estradiol, and SHBG

Genetically predicted, increase in BMI was associated with a reduction in BioT (IVW, β = −0.139, 95% confidence interval [CI] [−0.165 to −0.113], P = 9.54 × 10⁻²⁶) and SHBG (IVW, β = −0.103, 95% CI [−0.113 to −0.092], P = 1.50 × 10⁻⁷⁷) levels (Fig. 2A). A genetically predicted increase in BMI was associated with high estradiol levels with a β of 0.014 (IVW, 95% CI [0.009–0.019], P = 2.19 × 10⁻⁷; Fig. 2A).

The robustness and reliability of the above putative causal estimations were tested and verified. The results of four methods (MR–Egger, maximum likelihood, MR–PRESSO, and MR–RAPS) consistently supported the putative causal direction of BMI influence on BioT, estradiol, and SHBG levels (Figs. 2B–2G). The results of heterogeneity tests showed existing heterogeneity for the effect of the BMI-related SNPs on BioT and SHBG (all P < 0.05), except estradiol (all P > 0.05; Fig. 2A). The results from pleiotropy tests indicated no pleiotropy for BMI-related SNPs above each analysis (all P > 0.05; Fig. 2A). The leave-one-out analyses showed that no single SNP could reverse for above each IVW putative causal estimation (all P < 0.05; Tables S11–S13 in Supplemental Materials). The results of the MR–Steiger test supported that the effects of BMI on BioT, estradiol, and SHBG levels were all correct putative causal directions (all P < 0.05; Fig. 2A).

Effects of BioT, estradiol, and SHBG on BMI

The results from the IVW method showed that genetically predicted increased SHBG levels were associated with low BMI with a β of −0.051 (95% CI [−0.098 to −0.005], P = 0.030; Fig. 3A). However, the putative causality of the effect of BioT (IVW, β = −0.006, 95% CI [−0.035 to 0.023], P = 0.679) and estradiol (IVW, β = 0.091, 95% CI [−0.094 to 0.276], P = 0.336) on BMI was not statistically significant.

In sensitivity analyses, the four methods (MR–Egger, maximum likelihood, MR–PRESSO, and MR–RAPS) were consistent with IVW for estimating the effect of SHBG on BMI (Figs. 3B and 3C). The heterogeneity analysis results showed certain heterogeneity in the effects of the SHBG-related SNPs on BMI (P < 0.05; Fig. 3A). The results from the pleiotropy test indicated no pleiotropy for the putative causal estimation of the effect of the SHBG-related SNPs on BMI (P > 0.05; Fig. 3A). The leave-one-out analyses showed that no single SNP significantly altered the putative causal estimation of the effect of the SHBG-related SNPs on BMI (P < 0.05; Tables S14 in Supplemental Materials). The result of the MR–Steiger test indicated that the effect of SHBG on BMI was a correct putative causal assumption (P < 0.05; Fig. 3A).

Effect of SHBG on BioT and estradiol

Genetically predicted elevated SHBG levels were associated with decrease in BioT level (IVW, β = −0.126, 95% CI [−0.175 to −0.077], P = 5.97 × 10⁻⁷) and high estradiol level (IVW, β = 0.046, 95% CI [0.035–0.056], P = 6.51 × 10⁻¹⁷).

In sensitivity analyses, the results, including the putative causal estimations of the four methods (MR–Egger, maximum likelihood, MR–PRESSO, and MR–RAPS; Figs. 4B–4E), heterogeneity analyses (Fig. 4A), pleiotropy tests (all P > 0.05; Fig. 4A), leave-one-out analyses (all P < 0.05; Tables S15–S16 in Supplemental Materials), and MR Steiger tests (all P < 0.05; Fig. 4A), all supported the putative causal assumption of the effect of SHBG on BioT and estradiol.

Effects of BioT and estradiol on SHBG

Genetically predicted elevated BioT levels were correlated with low SHBG levels with a β of −0.044 (IVW, 95% CI [−0.061 to −0.026], P = 8.76 × 10⁻⁷; Fig. 5A). However, the genetically predicted effect of estradiol levels on SHBG levels (IVW, β = 1.035, 95% CI [−0.854 to 2.926], P = 0.283) was not significantly putative causal.

In sensitivity analyses, the results, including the putative causal estimations of four methods (MR–Egger, maximum likelihood, MR–PRESSO, and MR–RAPS; Figs. 5B and 5C), heterogeneity analyses (Fig. 5A), pleiotropy tests (P > 0.05; Fig. 5A), leave-one-out analysis (P < 0.05; Table S17), and MR Steiger tests (P < 0.05; Fig. 5A), all supported the putative causal hypothesis of the effect of BioT on SHBG.

Discussion

We comprehensively assessed the putative causal relationship among BMI, BioT, estradiol, and SHBG by using large-sample GWAS data. A summary of the findings is shown in Fig. 1B. We found that genetically predicted BMI was bidirectionally associated with SHBG levels and unidirectionally associated with BioT and estradiol levels. In addition, genetically predicted SHBG levels were bidirectionally correlated with BioT levels and unidirectionally correlated with estradiol levels.

Obesity, imbalance in sex hormones, and SHBG are all closely related to the risk of human diseases (Arnold et al., 2017; Boese et al., 2017; Guarner-Lans et al., 2011). The effects of obesity on human health may be partly mediated by sex hormones and SHBG levels (Grasa et al., 2017; Guarner-Lans et al., 2011; MacDonald et al., 2010). However, the correlation among BMI, sex hormones, and SHBG remains controversial. For example, some observational studies have reported that BMI is not associated with serum free testosterone and estradiol levels (Hofny et al., 2010; Winters et al., 2006). However, many observational data have indicated that BMI is negatively associated with serum free testosterone and SHBG levels (Foresta et al., 2009; Macdonald, Stewart & Farquhar, 2013; Yamacake et al., 2016) and positively associated with estradiol levels in males (Bieniek et al., 2016; Foresta et al., 2009; Jensen et al., 2004). Similarly, a recent meta-analysis including data from 28 articles revealed that increased BMI is associated with low serum testosterone and SHBG levels and high estradiol levels in males (Salas-Huetos et al., 2021). Additionally, MR investigation findings from Loh et al. (2022) showed that elevated BMI is causally associated with reduced serum BioT levels (β = −0.128) and high serum estradiol levels (β = 0.012). However, the putative causal effect of BioT and estradiol level imbalances on BMI was not statistically significant (Loh et al., 2022). Our results supported that the effect of BMI on serum BioT (β = −0.139) and estradiol levels (β = 0.014) is a unidirectional causality. Although the findings of Loh et al. (2022) were valuable, our study has more advantages. Importantly, our MR study used multiple methods to remove the pleiotropic and outlying SNPs that cause potential bias of causal estimation, thereby improving the accuracy of putative causal estimation. Moreover, our findings obtained more powerful support from statistical evidence of sensitivity analyses, including the putative causal estimations of the four methods (MR‒Egger, maximum likelihood, MR–PRESSO, and MR–RAPS), pleiotropy tests, leave-one-out analyses, and MR Steiger tests. Mechanistically, as previous research reported, obesity can increase the conversion of testosterone to estradiol in males, thereby causing reproductive axis suppression and reduced secretion of endogenous gonadotropin (Michalakis et al., 2013). Similarly, a recent study has shown that obesity reduces testosterone and elevated estradiol levels by mediating the hypothalamic–pituitary–gonadal axis in males (Leisegang et al., 2021). Additionally, in the present work, we analyzed the putative causal association between BMI and SHBG and found that they were a bidirectional putative causal relationship. Previous observational data revealed that individuals with a high BMI have lower SHBG levels than those with a low BMI (Allan & McLachlan, 2010). In addition, a recent meta-analysis revealed that BMI is negatively correlated with SHBG levels in males (Salas-Huetos et al., 2021). Our findings showed that genetically predicted elevated BMI is inversely associated with serum SHBG levels in males. We also observed that genetically determined increased SHBG levels are causally associated with low BMI. Similarly, observational studies have shown that high SHBG levels are associated with lowered risk of metabolic syndrome (Lewis et al., 2004; Salas-Huetos et al., 2021). Mechanistic findings from animal model studies have suggested that SHBG overexpression can reduce the risk of obesity by increasing insulin, leptin, and resistin levels in male rats (Chen et al., 2010; Saez-Lopez et al., 2020). According to the above evidence, the influence of genetically predicted high BMI on males’ health may be partly mediated by SHBG and sex hormones. Consequently, the healthy modulation of SHBG and sex hormones can attenuate the harmful effects of obesity on male health.

Given the importance of SHBG to sex hormone transport, tissue delivery, and bioactivity and metabolism regulation, we further examined the putative causal association between SHBG and the two sex hormones (BioT and estradiol). Our results suggested a bidirectional causal association between SHBG and BioT. However, a unidirectional causal association was found between SHBG and estradiol. Serum SHBG levels are associated with serum testosterone and estradiol concentrations in males (Wan et al., 2021; Yeap et al., 2021). High serum SHBG levels can cause a decline in androgen/estrogen ratio (Wan et al., 2021). In this work, we found that genetically determined raised SHBG levels are associated with high estradiol and low BioT levels in males. Additionally, we observed that elevated BioT levels were associated with decreased SHBG levels in males. A similar phenomenon was observed in previous observational studies in which patients with prostate cancer had lower SHBG levels and higher free testosterone levels than healthy individuals (Travis et al., 2007; Watts et al., 2021, 2022). Although our work preliminarily revealed the causal association between SHBG and sex hormones, the potential mechanism of interaction among them has been completely clarified. Therefore, we should cautiously comprehend the findings, and the underlying mechanism of SHBG and sex hormones interacting and affecting one another should be explored using molecular experiments. Given the putative causality of SHBG linked with BMI and sex hormones, SHBG could be a promising target for eliminating adverse effects caused by obesity and sex hormone imbalance on physical health.

Our work has some merits. First, all GWAS summary-level data were derived from recent extensive sample studies, ensuring our study’s statistical strength. Second, we used multiple methods to eliminate outlying and pleiotropic SNPs to guarantee instrumental variables’ validity. Third, multiple sensitivity analyses consistently supported our findings, suggesting the robustness and reliability of the results. Finally, our work solves the limitation of conventional observational research that does not explain the putative causality among obesity, SHBG, and sex hormones or handle unmeasured confounders well (Davies, Holmes & Davey Smith, 2018).

Inevitably, our work has some limitations. First, although our findings supported a putative causal association among BMI, sex hormones, and SHBG, the potential interaction mechanism remains unclear. Consequently, the underlying mechanism of their interactions and how they affect one another should be further investigated using molecular experiments. Next, whether and how much samples of the sex hormones-related GWAS overlap is unclear. Furthermore, given that all GWAS summary data were derived from European populations, generalizing to non-European populations may have some bias. Finally, all data were GWAS summary levels, and thus stratified analysis according to age was limited.

Conclusion

In summary, using genetic data from large sample studies, we conducted an MR study to expound the bidirectional causal association among BMI, sex hormones, and SHBG in men. We then found a bidirectional causality between BMI and SHBG and a unidirectional causality between BMI and sex hormones. Additionally, genetically predicted SHBG levels were bidirectionally associated with BioT levels and unidirectionally associated with estradiol levels. The findings provide novel insights into interventions based on BMI, sex hormones, and SHBG, specifically interventions that promote men’s health.

Supplemental Information