We assessed the causal effects of the gut microbiome on BMDs at different skeletal sites by using a two-sample MR. Of note, the results suggested that higher number Eisenbergiella significantly caused lower H-BMD while Oscillibacter had an opposite effect on H-BMD. The use of Mendel's second law allowed us to reduce the confounding factors when using the MR design. A variety of techniques were adopted to investigate heterogeneity and correct pleiotropy, which led to results of remarkable consistency. Furthermore, the reverse MR study showed no sign of any causal effect of the H-BMD on the Eisenbergiella and Oscillibacter. Throughout our study, PhenoScanner played a role in finding probable pleiotropic SNPs and exploring the disease mechanisms as well as biology links to the selected SNPs.
The gut microbiome is crucial to human health(48). Study of the gut-bone axis, one of the recent popular study subjects, suggested that the gut might indirectly influence bone health. Consequently, focusing on the gut as an alternative treatment for bone disease rather than using bone-targeting medicine such as bisphosphonates is a more appealing method. The immune system, the gut-brain axis, alterations in the intestinal mucosal barrier permeability, and gut microbial excretion products are a few of the ways that the gut microbiome impacts bone(49). Studies found that when compared to mice raised normally, germfree female mice had higher BMD, which could be attributed to the absence of gut microbiome(50, 51). However, these findings are contradictory, possibly due to the different strains of mice used, or the differences in age and sex(52). Despite an increasing number of animal experiments demonstrating an effect of the gut microbiome on bone development, the evidence of the association between the gut microbiome and skeletal health in humans is still sparse. One research employed a polygenetic risk scoring based MR analysis to show an effect of the gut microbiota on pelvis BMD (P = 0.0437) (25). Wang et al. reported that patients with osteoporosis and osteopenia had a different bacterial composition and diversity than the healthy controls, which, however, was carried out with a small sample size(20). Orwoll et al. used a large cohort study of community-dwelling older men (n = 831) to demonstrate that the abundances of four bacterial genera (Anaerofilum, Methanomassilicoccus, Ruminiclostridium 9, and Tyzzerella) were weakly associated with bone density, strength, or structure(53). The role of a single microbiota in pathogenesis may be difficult to convince considering the intactness of the ecosystem and the dynamic response of the gut microbiome to environmental factors. Thus, it poses a challenge for gut microbiota MR analysis to completely exclude other microbiome functions or their interactions. Furthermore, the two genus we found were significantly associated with H-BMD alone but not with other skeletal sites, which calls for more caution causality studies. Association replication studies and/or functional validation of supporting mechanistic principles will be important for our identified bacterial genera in the future.
Most of the previous research on the gut microbiome-BMD relationship has failed to reveal a convincing causal relationship, while our findings may provide strong evidence. The study used observational and linear regression techniques to identify the genus Escherichia Shigella as negatively related to L1-L4 BMD. Although significant association was not detected, our study showed that higher number of Escherichia Shigella was negatively associated with LS-BMD with a β of IVW is 0.2(54). Established causal links deepen our understanding of low BMD related diseases and the gut microbiome while also having the ability to direct the selection of interventions.
When it comes to measuring BMD, DXA is the gold standard technique, which could be applied to different sites and provide variable information about the fracture risk and osteoporosis diagnosis(55). Notably, the BMD for heel bone is estimated by QUS, which is an alternative to DXA that represents a combination of broadband ultrasound attenuation and speed of sound(31). According to the AACE’s guidelines in 2020, the best sites for diagnosing osteoporosis in postmenopausal women are the lumbar spine BMD, hip BMD and 1/3 radius BMD(56). Several studies have identified different ideal measurement sites for osteoporosis diagnosis, such as the proximal femur and the femoral neck(57, 58). It is worth noting that the forearm and proximal femur are the most typical choices for BMD measurements of adults, whereas for most children and adolescents total body less head and the posterior-anterior spine are the preferable sites, which is related to the transition from cartilage to bone during growth and the closure of growth plates(59). The measured rate at which BMD increases fracture risk varies significantly with different techniques and bone sites(1). Multiple studies have indicated that the best method for assessing the likelihood of fracture at a given place is a site-specific measurement of BMD, which, however, has some limitations(60–62). Considering the choice of which site to measure BMD relies on characteristics like the patient's age and sex, we used BMD data from five different skeletal sites as the outcome data in our article. Since QUS and DXA are two independent procedures for measuring BMD, a low result with either approach can indicate a great risk of fracture(63, 64). In view of the fact that DXA-derived BMD is more relevant to the clinical diagnosis of osteoporosis and that the direction of Eisenbergiella and Oscillibacter effects on other skeletal sites is not the same at all, the causal effects in our MR analysis must be interpreted with caution. Differences in procedures and sample size may explain why Eisenbergiella and Oscillibacter have an exclusive significant association with H-BMD.
In the Genome Taxonomy Database website, the genus Eisenbergiella is classified as a member of the family Lachnospiraceae, order Eubacteriales, class Clostridia in the phylum Firmicutes (http://gtdb.ecogenomic.org/). According to one study that used observational and GWEI analysis, Eisenbergiella has a significant effect on TB-BMD and plays a potential role in the aetiology of osteoporosis(54). Likewise, Wei M et al. discovered that Eisenbergiella is more prevalent in individuals with osteoporosis than in the control group (44 patients with osteoporosis and 64 controls), suggesting that Eisenbergiella is associated with BMD and osteoporosis risks(65). Furthermore, the Lachnospiraceae family, to which the genus Eisenbergiella belongs, has been proved to demonstrate a reverse regulation pattern with heel BMD(25). More importantly, sphingolipids and xanthurenic acid both have a negative correlation with the number of Eisenbergiella(66). Sphingolipids are associated with skeletal development, and disturbed sphingolipids metabolism can lead to early-onset osteoporosis(67, 68). In addition, xanthurenic acid has been positively correlated with BMD(69). In brief, we suggest the reason why Eisenbergiella causes low BMD is that high Eisenbergiella level is associated with low sphingolipids and xanthurenic acid, which are two essential acids for BMD. However, more research into the pertinent mechanism is required due to the absence of biological evidence. We hypothesized that the detrimental effect of Eisenbergiella on BMD may be genus-specific by combining data from observational studies, MR analysis, 16S rRNA, and animal experimentation.
The genus Oscillibacter is listed as a member of the family Oscillospiraceae, order Eubacteriales, class Clostridia, and phylum Firmicutes on the Genome Taxonomy Database website (http://gtdb.ecogenomic.org/). A significant bacterial linked to inflammatory bowel disease is Oscillibacter(70). However, the association between Oscillibacter and BMD remains unclear. Our results showed a positive correlation, and we speculate that Oscillibacter are related to the gut microbial excretion products. According to some studies, Oscillibacter can create short-chain fatty acids, which increases bone density effectively by influencing osteoclast metabolism and bone mass(71–74). Furthermore, Oscillibacter is a butyrate-producing bacteria(75). Butyrate stimulates Tregs while suppressing osteoclasts, promoting bone formation by modulating WNT10B expression via regulatory T cells(76–78). More importantly, regulatory T cells inhibit bone resorption while also encouraging bone formation via parathyroid hormone and osteoblast development(79–81). In some studies, higher relative abundance of fecal Oscillibacter has been causally linked to lower triglyceride concentrations, and lower triglyceride levels have been linked to lower risks of osteoporosis(82, 83). Further research is needed to decipher the latent potential mechanism.
We then searched for the 17 selected SNPs by using the PhenoScanner v2 curated database (http://www.phenoscanner.medschl.cam.ac.uk/) to detect the possible causal mechanistic associations between the gut microbiome and BMD (P < 5 × 10− 6) (Supporting Information Table S6)(45). Out of the six SNP variants genetically associated with Eisenbergiella, two were found in the intron region of the genes ROBO4 and SNTG1; also, four SNPs were found in the intergenic region of the genes RP11-747E23.1, RP11-209E8.1 and LOC105370823 and LOC107983981, DPH6-AS1, and AP000472.3. Out of the 11 SNP variants genetically linked to Oscillibacter, five were found in the intron region of the genes RP11-362A1.1 and LINC02426, FAM129A, NUBPL, GFRA2, and TENM4; three SNPs were found in the intergenic region of NDRG2, LINC02171, RP11-443B9.1, and LOC105375951; one SNP was found in the upstream of LINC01656 gene; also, one SNP was found in the 3_prime_UTR of ADNP gene; and one SNP was found in the missense region of TTN. A few SNPs had no known connection with diseases. It should be noted that the disease and trait in the PhenoScanner curated database for the rs12278566, which is the Eisenbergiella SNP, is H-BMD, suggesting a linkage between the genus Eisenbergiella and BMD biology. The Oscillibacter SNP rs11627628 was associated with the NDRG2 gene, which encourages osteoblast differentiation while inhibiting osteoclast differentiation(84, 85). What’s more, through the control of FAM129, which is Oscillibacter SNP rs234108 associated gene, could affect a crucial regulator of autophagy in osteoclasts proliferation and differentiation(86). According to one study, ADNP mutations affect the regulation of ossification and osteoblast differentiation, and ADNP is the Oscillibacter SNP rs761240 associated gene(87). As we all know, in vivo, two competing processes—osteoblasts produce bone while osteoclasts resorb bone—maintain a balanced bone remodeling process. Loss and decreased BMD can be caused by unbalanced bone remodeling, like inadequate osteoblastic bone formation and/or excessive osteoclastic bone resorption. TTN, which is the Oscillibacter SNP rs16866406 associated gene, shows the function in suppressing the expression of TNF-α, and TNF inhibitors can maintain femoral neck BMD while increasing lumbar spine BMD and total hip BMD(88, 89). As a consequence, we believe the Oscillibacter may have a protective effect on BMD, which could be caused by the gene as mentioned above. It is necessary to investigate the mechanistic pathway underlying this impact using experimental laboratory data.
This study has benefits in several ways. Firstly, to boost the statistical power to identify causal relationships, our work used a number of variants that have been summarized from large-scale GWAS studies on the gut microbiome and BMDs. Secondly, we studied up to five taxonomic ranks (phylum, class, order, family and genus) and BMDs at five different skeletal sites. Thirdly, to fully satisfy the fundamental MR assumptions, we used independent SNPs as IVs and performed various sensitivity studies, such as heterogeneity evaluation and reverse MR analysis. To account for multiple testing, the Bonferroni approach was utilized, which produced compelling evidence of the relationships.
In previous studies, Ni et al. performed MR analyses to investigate the effects of gut microbiota on Heel BMD(25). The difference in results could be explained in the following three aspects: First, Ni et al. utilized summary statistics of gut microbiota from a sample size that was relatively small. The researchers utilized 1,126 twin pairs form the TwinsUK cohort and 984 Dutch participants from the LifeLines-DEEP cohort. But the data we analyzed was from the MiBioGen study, which include 22 additional cohorts in addition to TwinsUK and LIFELINES cohort for a total sample size of 18,340. Second, because linkage disequilibrium SNPs can bias the results, we set the threshold at r2 < 0.01, distant = 10000 kb; however, their threshold was set to be r2 < 0.1 and the distant = 500 kb. Third, the quality control approach for selecting IVs was stricter in our study. We selected IVs with a P value < 5×10− 6 and performed a series of sensitivity analyses such as Cochrane’s Q test, to maximally satisfy MR fundamental assumptions. In contrast, the previous study utilized a rather loose P-value (P < 1 × 10− 5) to select eligible IVs.
Our article contains some limitations that should be noted. Firstly, the two-sample MR analyses in this study could not ascertain the exposure and outcome GWAS overlapping degree. Although the maximum estimate overlapping rate is 15%, bias caused by overlapping samples can be reduced by making use of powerful instruments like F-statistic(90). Secondly, despite the fact that most of the gut microbiome data included in our analysis came from European populations, a small proportion was acquired from groups of other ancestries, which could have skewed our results. Finally, we advise that future GWASs for BMD should take sex specificity into account because several disorders that cause lower BMD, such as osteoporosis, are more common in postmenopausal women. Nevertheless, the greater sample size for BMD might lessen the bias.
In conclusion, this MR study lends support to the hypothesis that the gut microbiome influences BMD in a causative manner. Given that people are living longer lives as a result of modern medicine, a better understanding of maintenance for bone health and treatment for diseases associated with low BMD is critical. Unlike randomized controlled trials tell whether an intervention will be effective, MR studies reveal the role of a definite causal pathway(91). Furthermore, MR studies often identify long-term exposures, whereas randomized controlled trials only describe the effects of transient exposures. It is plausible that our findings may contribute to a better understanding of the relationship between the gut microbiome and bone health. And, theoretically, direct the design of interventional studies to promote bone health with a higher probability of success.