Introduction

The gut microbiota (GM) is considered to be a multifunctional “organ”, involved in immunity, metabolic regulation and other life activities. Studies have proved that there is an intimate relationship between the GM and human organs. The signal pathways between GM and different organs, can be divided into the gut–brain axis, gut–kidney axis, gut–liver axis, and the gut–bone axis1. Bone is a supporting organ composed of collagen and calcium phosphate, and it is also the main organ for storing minerals such as calcium and phosphorus. Homeostasis of the bone environment is maintained by osteoblasts and osteoclasts which are responsible for the dynamic process of bone remodeling2. When the balance is broken, it will result in low bone mass, destruction of bone tissue microstructure and eventually increased bone fragility, which is known as osteoporosis (OP) 3. The prevalence of OP in the Chinese population increased by 14.94% from 2003 to 2015, moreover, the current incidence of OP in the Chinese population aged over 50 is more than twice that in 20064. In view of the increasing prevalence of OP, prevention and control measures are even more important.

Osteocytes, the most abundant cells in bone tissue, are essential for the renewal and maintenance of the bone matrix. Osteocytes produce a variety of signals that induce osteoblast formation, such as prostaglandin E 2 (PGE 2), growth factors, and glycoproteins5. Recent advances have shown that, in addition to the respective signaling pathways and enhanced expression of genes in these osteocytes, further differentiation of osteoblasts and osteoclasts depends on metabolic reprogramming6. The GM and its metabolites, short–chain Fatty Acids (SCFAs), are intimately related to bone metabolism. A study has shown that SCFAs prevent bone loss in OP mouse model by metabolic reprogramming osteoclasts to inhibit osteoclast formation at an early stage7. SCFAs may also exert their effects by binding to free fatty acid receptor 2 (FFAR2) receptors expressed mainly by osteocytes8. These findings suggest that SCFAs have great potential to regulate bone health, and that GM metabolism produces large amounts of SCFAs. In a recent mouse model study, in vivo colonization of ovariectomized mice with Lactobacillus acidophilus enhanced trabecular and cortical bone microstructure, simultaneously increasing bone mineral density (BMD) and bone heterogeneity9. In a population study, the BMD and GM composition of 1776 people were analyzed using X–ray absorptiometry and 16S rRNA, and the results showed that BMD was related to the β diversity and species composition of the GM10. The diversity, composition and SCFAs of the GM were found to be closely related to the changes of bone mass and the development of bone metabolic diseases. Especially, the study of specific intestinal bacteria is very important for the prevention and treatment of OP and osteopenia.

Since the GM is affected by various factors such as diet, genetics, living environment and living habits, and some articles have pointed out that there are differences in the composition and structure of GM among different Chinese ethnic groups11. There were significant differences in the structure and composition of GM between the Plateau Tibetan and Plain Han populations with different BMD and T–score12. We only selected a cohort of Chinese Han youth of university students in Anhui Province, China, and which is located in the Yangtze River Delta region of China, one of the most Han populous regions. It is necessary to conduct research on population of different regions and nations, as this may supplement and validate previous research findings, especially for a nation with a large population base.

Materials and methods

Research participants

Non–sports majors of Han nationality were recruited as participants in Anqing Normal University, and the participants were all from Anhui Province, China. All participants were voluntary and gave informed consent to the content of the experiment. After excluding patients with diabetes/hypertension, osteoarthritis, fractures, gastroenteritis and peptic ulcer, 62 college students (28 men and 34 women) aged 19–25 years (21.60 ± 1.38 years) were finally recruited as the study participants. The basic information about the participants is shown (Table 1). This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Anqing Normal University (approval number: AQNU2018018H).

Table 1 Participants basic information.
Full size table

Standard for measurement and classification of BMD

The BMD at the lumbar spine (L1 to L4), total hip joint and femoral neck were measured by dual–energy X–ray absorptiometry (GE Lunar iDXA, GE Healthcare, Little Chalfont, UK) through a professional hospital. Before testing, participants were checked to ensure they had not had a barium meal, a radionuclide injection, or an injected or orally–administered contrast agent for CT or MRI within the previous 7 days. Using the World Health Organization (WHO) criteria, BMD is generally expressed as a T–score13. OP is diagnosed when the T–score of the lumbar spine, femoral neck and total hip joint is at least one T < −2.514. The specific criteria were that T < −2.5 was classified as osteoporosis\BDL (Hereinafter referred to as BDL), −2.5 < T < –1 was classified as osteopenia\BDM (BDM), and –1 < T < 1 was considered normal bone density\BDH (BDH) (Table 2).

Table 2 Baseline characteristics of the participants.
Full size table

Fecal collection and DNA extraction

We used disposable fecal collection boxes (GUHE Info Technology Co., Ltd, Hangzhou, China) to collect feces. The collection box consists of a sterile cotton swab and a fecal collection tube containing bacterial lysis preservation solution. Participants needed to use sterile cotton swabs to dip a small amount of feces into a fecal collection tube and kept in a portable Styrofoam box with ice packs (0–4 ℃). Fecal total DNA was extracted using the PowerMax (fecal/soil) DNA isolation kit (MoBio Laboratories, Carlsbad, CA). The quantity and quality of extracted DNA were measured using a NanoDrop ND–1000 photometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis respectively, to ensure that all samples met the sequencing requirements. Each faecal sample was frozen at −80 ℃ until sent to GUHE Info Technology Co., Ltd (Hangzhou, China) for sequencing analysis.

16S rRNA sequencing and quality control

In our study, the forward primer 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and the reverse primer 806R (5′-GGACTACHVGGGTWTCTAAT-3′) were used to amplify the V4 region of the bacterial 16S rRNA marker gene. PCR amplification was performed as follows: PCR reaction system (50 μL): Phusion High-Fidelity PCR Master Mix with HF Buffer 25 μL, 515 F and 806 R Primers (10 μmol/L) 3 μL each, DNA template 10 μL, DMSO 3 μL, DDH2O 6 μL. PCR reaction conditions: 98 ℃ 30 s, 98 ℃ 15 s, 58 ℃ 15 s, 72 ℃ 15 s, 25 cycles; 72 ℃ 1 min. The PCR amplification products were recovered and quantified, then purified using an Agencourt AMPure XP 60 mL Kit (Beckman Coulter, Brea, CA, USA) and quantified using the Qubit dsDNA HS Assay Kit (Vazyme, Nanjing, China) and finally two–terminal sequencing was performed using the Illlumina Hiseq 4000 Pair–End (Pair–End) 2 × 150 BP platform. According to Barcode sequence and primer sequence, the offline data were divided into the data of each sample. After truncating Barcode and primer sequences, the reads of each sample were spliced using VSEARCH v2.4.4 to obtain the Raw Tags. On this basis, the quality of the sequences was controlled and filtered, and the chimeric sequences were removed by UCHIME to obtain effective data, sequences with a similarity of 97% were clustered into operational taxonomic units (OTU’s) and the classification information for each sequence was annotated with the RDP classifier.

Bioinformatics analysis

Based on the clustering results, a Venn diagram was generated using the R package “Venn Diagram”. Moreover, QIIME and R Studio–based EasyAmplicon v1.14 software was used to compare the number of common and unique OTU’s among BDH, BDM and BDL groups. The R package “ggplot2” and its expansion package “ggtern” were used to generate histograms and ternary phase diagrams to analyze the composition of GM in BDH, BDM and BDL groups. The Kruskal–Wallis test was carried out with the function of “Kruskal. test” in R to analyze the taxa which differed significantly among BDH, BDM and BDL groups. The linear discriminant analysis (LDA) effect size (LEfSe) software v1.0 was used to do LEfSe analysis (LDA score threshold: 2) so as to find out the biomarkers. The co–occurrence network was analyzed by Cytoscape v3.9.1 software (https://cytoscape.org/), which reflects the interaction between genera and BMD groups. FastTree v2.1.3 (http://www.microbesonline.org/fast–tree/) was used to analyze the phylogeny and evolution of GM in a cohort of Chinese Han youth. The high–level phenotype prediction of GM in a cohort of Chinese Han youth was carried out by using BugBase (https://bugbase.cs.umn.edu/)15.

Statistical methods

The statistical tools R v3.5.3 and R Studio–based EasyAmplicon v1.14 software were used for statistical analysis and figure generation in this study. Data on height, weight, T–score, body mass index (BMI), and waist–to–hip ratio (WHR) are presented as the mean ± standard deviation (SD). The statistical differences were analyzed by the one–way ANOVA (conforming to a normal distribution), Chi–square test (categorical values) and Kruskal–Wallis (not conforming to a normal distribution). The Kruskal‒Wallis test was used for between-group comparisons of relative abundances at the phylum, genus and species levels and plotted by package “ggplot2” in R. R conducted alpha and beta diversity analysis with package “vegan”, which produced alpha diversity indexes (Chao1, ACE, Simpson and Shannon) and Beta diversity index (PCoA) at the OTU’s level. The significance of differences in the alpha diversity indexes was measured by the Kruskal–Wallis test between groups and was visualized as a box diagram using “phyloseq” package. Principal coordinate analysis (PCoA) was performed to evaluate microbial beta diversity based on Bray‒Curtis distance using R software (“phyloseq and vegan” package). Permutational multivariate analysis of variance (PERMANOVA) was performed to assess the between‒group differences in microbial communities using the “Adonis” function in the R package “vegan” with 9999 permutations. To explore the taxa differences in the three BMD groups, “Shapiro–Wilk. test” and “Levene’s. test” funtions in R were used for “normal distribution test” and “homogeneity test for variance”. According to the test results, the “Kruskal. test” function was used to compare the taxa differences in the three BMD groups, and the “pairwise–wilcox. test” function was used to calculate pairwise comparisons between groups. Post–hoc Benjamini–Hochberg Procedure correction was performed to determine pairwise differences. Spearman correlation analysis was applied to measure the correlations between microbiota abundance and T–score. Age, gender, height, weight, BMI and WHR as covariates, a regression model was constructed, and under the influence of the additive model (ADD, i.e. eliminating the influence of covariates). The results of regression were visually analyzed by GraphPad Prism v9.0.

Results

Basic information of participants in different BMD groups

There was a significant difference in WHR values. Not statistically significant differences were found in gender, height, weight, BMI, and T–score among the three groups (Table 2).

Comparisons of GM diversity between the three groups

The number of common and differential OTU’s among BDH, BDM and BDL groups was analyzed by sequence annotation for 62 samples. The total number of OTU’s obtained were 266 in the BDH group, 285 in the BDM group, 217 in the BDL group, with 197 in common (Fig. 1a).

Fig. 1
figure 1

Venn diagram and alpha\beta diversity analysis at OTU’s level. (a) Venn diagram. (b–e) Between-group comparisons of the alpha diversity of gut microbiota (ACE, Chao1, Shannon, Simpson). The results are presented in a box diagram, and P values were calculated by using the Kruskal–Wallis test. (f) Principal coordinate analysis (PCoA) of bacterial beta diversity based on the Bray‒Curtis distance. P values were calculated by Adonis. test with 9999 permutations.

Full size image

Analysis of alpha diversity indicated that the community richness and diversity in the BDH group were significantly higher than that in the BDL group (Fig. 1b–e), while beta diversity, the PCoA plot did not show an obvious separation between groups. PERMANOVA (R2 = 0.02, P = 0.03) has showed significant differences in gut microbial composition between groups (Fig. 1f).

Gut microbial community and taxonomic composition

Taxonomic composition of the GM was clearly different in the three BMD groups. The dominant phyla in a cohort of Chinese Han youth were Bacteroidota (50.6%) and Firmicutes (41.6%). The relative abundances of phyla Bacteroidota, Firmicutes and Proteobacteria in the BDH group were 51.7, 37.9 and 9.2%, While in the BDM group they were 53.9, 37.9 and 6.5%. The relative abundance of the dominant phyla in the BDL group were Bacteroidota (46.1%), Firmicutes (49.2%) and Proteobacteria (4.0%) (Fig. 2a).

Fig. 2
figure 2

Taxonomic composition of the gut microbial community. (a) Phylum–level abundance plot. (b) Ternary phase diagram. Each dot represents an OTU’s, and the size, color, and position of the OTU’s represent its relative abundance, genus–level name, and grouping, respectively. (c) Genus–level abundance plot. (d) Species–level abundance plot.

Full size image

Ternary phase diagram at the genus–level, g_Prevotella (31.2%), g_Bacteroides (16.9%), g_Faecalibacterium (13.2%) and g_Megamonas (4.6%) were the top 4 dominant genera in terms of average relative abundance for analysis (Fig. 2b). The relative abundance of g_Bacteroides (20.1%) in the BDH group was significantly higher than that in the BDM group (11.5%, P < 0.05) and higher than the BDL group (19.0%). The relative abundance of g_Prevotella (39.8%) in the BDM group was significantly higher than that in the BDH and BDL groups (28.1%, 25.6%, P < 0.01), and the relative abundance of g_Megamonas (7.8%) was significantly higher than that of the BDH and BDL groups (2.1%, 3.8%, P < 0.05). The relative abundance of g_Faecalibacterium (19.1%) in the BDL group was significantly higher than that of the BDH and BDM groups (11.8%, 8.6%, P < 0.01), and the relative abundance of g_Agathobacter (4.6%) was higher than that of the BDH and BDM groups (3.2%, 3.8%) (Fig. 2c).

Sample sequence annotation results at the species–level indicated that the dominant species were s_Prevotella copri (18.2%), s_unclassified_g_Faecalibacterium (12.7%), s_unclassified_g_Prevotella (9.2%), s_unclassified_g_Bacteroides (8.3%) and s_Bacteroides plebeius (7.2%). The relative abundance of s_unclassified_g_Bacteroides (11.1%) in the BDH group was significantly higher than that in the BDM group (4.1%, P < 0.01) and higher than the BDL group (10.0%). The relative abundance of s_unclassified_g_Prevotella (12.5%) in the BDM group was significantly higher than that in the BDH and BDL groups (8.2%, 6.9%, P < 0.01), while the relative abundance of s_Prevotella copri (22.3%) was also significantly higher than that of the BDH and BDL groups (16.8%, 15.5%, P < 0.01). The relative abundance of s_unclassified_g_Faecalibacterium (19.2%) in the BDL group was significantly higher than that of neither the BDH nor BDM groups (11.2%, 7.7%, P < 0.05), and the relative abundance of s_Bacteroides plebeius (7.6%) was higher than that of the BDH and BDM groups (7.2%, 6.9%) (Fig. 2d).

Identification of significant taxa differences and phenotype predication

The normal distribution and variance homogeneity test of genera and species in the three BMD groups were carried out, and the P values were all less than 0.05. Hence the data did not conform to the normal distribution and irregular variance that the Kruskal–Wallis test was used for the differences among the three groups. The pairwise-wilcox test was used to calculate pairwise comparisons between groups. The results showed that there were significant differences among the three groups in 9 genera and 1 specie: g_Faecalibacterium, g_Bacteroides, g_Prevotellaceae UCG–003, g_Lachnospiraceae UCG–010, g_Ruminococcaceae UCG–002, g_Ruminococcaceae NK4A214 group, g_Lachnospiraceae NK4A136 group, g_Lachnospira, s_Odoribacter splanchnicus and g_Ruminococcaceae UCG–005 (P < 0.05). In addition, 4 genera (g_Subdoligranulum, s_Ruminococcus bicirculans, g_Ruminococcus and g_Coprococcus) showed significant differences (P < 0.01) (Fig. 3a). The mean proportions of g_Faecalibacterium, g_Subdoligranulum and g_Ruminococcus were significantly higher in the BDL group than in the BDH group (P < 0.05), and the mean proportions of g_Faecalibacterium, g_Bacteroides, g_Ruminococcus and s_Ruminococcus bicirculans were significantly lower in the BDM group than in the BDH group (P < 0.05) (Fig. 3a).

Fig. 3
figure 3

Identification of significant taxa differences. (a) The Kruskal‒Wallis test was used for between-group comparisons of relative abundances at the genus-level and species-level. Only the taxa with a P value < 0.05 are shown. (b) LEfSe plot. The different colors represent different groups. The default LDA Score is 2, meaning that an LDA value greater than 2 indicates a statistical difference between groups. Note: * P < 0.05, ** P < 0.01.

Full size image

According to the LEfSe analysis, g_Bacteroides, s_unclassified_g_Bacteroides, s_gut_metagenome_g__Lachnospira and s_Lachnospiraceae bacterium TF01–11 were enriched in the BDH group. In the BDM group, g_Kocuria, s_Desulfovibrio and s_unclassified_g_Kocuria were enriched. s_unclassified_g_Butyricimonas, s_unclassified_bacterium_g__norank_f__Saccharimonadaceae, s_unclassified_organism_g__UCG–002 and s_unclassified_g_Ruminococcus were enriched in the BDL group (Fig. 3b).

We used BugBase phenotype prediction, which showed that g_Faecalibacterium accounted for the highest proportion of the anaerobic phenotype, and which in the BDL group was higher than in the BDH and BDM groups (Fig. 4a). The relative abundance of g_Megamonas and g_Prevotella of the anaerobic phenotype were higher in the BDM group than in the BDH and BDL groups (Fig. 4a). The prevailing gut microbial communities within the facultative anaerobic phenotype comprised g_Haemophilus, g_unclassified_f__Enterobacteriaceae, g_Erysipelotrichaceae UCG–003, g_Lactobacillus and g_Escherichia Shigella (Fig. 4b, Table S1). We ascertained that g_Sutterella, the predominant genus existing in the aerobic phenotype, was enriched in the BDH group. Meanwhile, in the BDM group of the aerobic phenotype, the relative abundance of g_Rothia was greater than that in the BDH and BDL groups (Fig. 4c).

Fig. 4
figure 4

Phenotype prediction and co–occurrence network analysis of bacterial taxa at genus-level. (a) Anaerobic. (b) Facultative Anaerobic. (c) Aerobic. (d) Co–occurrence network. The co–occurrence network explores the interaction between genera and BMD groups.

Full size image

Co–occurrence network analysis explores the correlation between genera and BMD groups, reflecting the range and strength of correlation between genera and BMD groups. The connecting lines between genera and BMD groups are considered as degrees, The number of degrees represents the influence of a genus in one or more BMD groups. The weighted degrees represent the strength of correlation the genera are with BMD groups. The higher the weighted degrees, the more strength of correlation the genera are with BMD groups, and the deeper the impact. The degrees of g_Megamonas, g_Faecalibacterium, g_Roseburia, g_Lachnospira, g_Prevotella, g_Bacteroides, g_Agathobacter, g_Phascolarctobacterium, g_Escherichia Shigella, and g_Dialister were three, which all had an impact in the three BMD groups. The weighted degrees of g_Prevotella, g_Bacteroides, g_Faecalibacterium, g_Megamonas, and g_Agathobacter were 45,800.62, 28,527.11, 20,643.70, 7525.19, 4998.01 respectively, with genera of Bacteroidota and Firmicutes phyla being highly intimately related to the three BMD groups. g_Alloprevotella, g_unclassified_f__Enterobacteriaceae, g_Ruminococcus, g_Sutterella and g_Haemophilus with one degree for the BDH group, while g_Eubacterium eligens group and g_Subdoligranulum with one degree for the BDL group, and which mean those genera only have an impact in specific BMD group. The BDH and BDL groups shared g_unclassified_f__Lachnospiraceae, while the BDM and BDL groups shared g_Lachnoclostridium (Fig. 4d, Table S2).

Associations of GM relative abundance and T–score at phylum–level and genus–level

No associations were found between gut microbiota abundance and T-score at phylum and genus levels by spearman correlation analysis (Table S3). After adjusting for age, BMI and gender in regression analysis, also no associations between gut microbiota abundance and T-score were found (Fig. 5, Table S3).

Fig. 5
figure 5

Regression analysis. (a): Phylum-level; (b): genus-level. Regression analysis reflects the associations between GM relative abundance and T-score.

Full size image

Analysis of the evolutionary relationship

At the genus–level, Fast–Tree software was used to analyze the evolutionary relationship of GM in the cohort of Chinese Han youth. The results showed that, with the exception of g_Prevotella and g_Bacteroides, the GM of the cohort of Chinese Han youth belonged to the same branch. The dominant genus was g_Faecalibacterium, which had the highest homology with the other genus, followed by g_Escherichia Shigella (Fig. 6).

Fig. 6
figure 6

Phylogenetic tree of bacterial taxa. Each branch of the phylogenetic tree represents a genus, and the branches are colored according to the high taxonomic level of the genera. The branch length is the phylogenetic distance between the two genera, that is, the degree of genera difference. The bar chart shows the proportion of reads of genera in different BMD groups.

Full size image

Discussion

The GM is closely related to bone health, and the diversity and richness of the GM are one of the important factors affecting bone health. Studies have shown that poor lifestyle, such as drinking alcohol, insufficient sleep, smoking, and irregular diet, are the critical factors leading to OP, and the proportion of contemporary youth with unhealthy lifestyles is increasing16,17. Therefore, our study analyzed the diversity, composition and structure, phenotypes and evolutionary relationship of GM community among different BMD levels in a cohort of Chinese Han youth, explored their differences and found some interesting phenomena.

Our findings performed that there were significant differences in the alpha diversity of GM between groups, which aligns with certain earlier studies18,19. PERMANOVA of beta diversity revealed that the microbial composition between groups was significantly different, and which is consistent previous study20. But there was also a similar study conducted contradict results21. The contradictory outcomes of alpha and beta diversity between different studies might be attributed to the vast heterogeneity of cohorts, also including age, sex, dietary habits, etc. On the other hand, this is an advantage of selecting a cohort of Chinese Han youth in Yangtze River Delta region as the participants, where they may have similar dietary habits, and to exclude the effects of age, dietary habits, locations.

We further analyzed the taxonomic differences. Compared with the BDH group, the relative abundance of Bacteroidota and Proteobacteria in the BDL group was decreased, while the relative abundance of Firmicutes was increased. A study of Japanese women showed that the incidence of fractures in the lower Bacteroides group was 5.6 times higher than that in the normal group22. Another study examined stool samples from 102 elderly Chinese patients with OP using 16S rRNA sequencing and showed that the abundance of Firmicutes was negatively correlated with BMD and T–score23. These findings suggested that changes in the proportion of gut dominant phyla were associated with BMD.

The g_Prevotella is one of the most abundant taxa in the oral and GM, and they are key participants in maintaining human metabolic homeostasis24. In the present study, we also annotated g_Prevotella as the dominant genus (with an average relative abundance of 31.2%). After groups annotating, it was found that the relative abundance of g_Prevotella in the BDL group (25.6%) was lower than that in the BDH group (28.1%), while the relative abundance of g_Prevotella in the BDM group (39.8%) was significantly higher than that in the BDH group (28.1%). Studies have shown that the dysbiosis of GM in OP mice induced by ovariectomized (OVX) was regulated by Prevotellaceae, which ameliorated bone loss in OP mice and has potential for preventing and treating OP25,26. Another OVX mice experiment showed that the abundance of g_Prevotella increased rapidly in the early and middle stages of OP and remained high for a certain period of time, but gradually decreased in the later stages of OP27, which is similar to our result. g_Prevotella may be a potential diagnostic biomarker.

Another dominant genus in our study is g_Faecalibacterium (with an average relative abundance of 13.2%), which is also the dominant genus in the BDL group. Literature shows that compared to the normal control group, Faecalibacterium prausnitzii is significantly enriched in patients with primary osteoporosis (PO) (0.331 vs. 0.132)28. We also found that the relative abundance of g_Megamonas in the BDM group (7.8%) was significantly higher than that in the BDH and BDL groups (2.1%, 3.8%). The relative abundance of Megamonas is positively correlated with femoral BMD29. The relative abundance of Megamonas in the gut of postmenopausal osteoporosis (PMOP) women is higher than that of the normal control group30. The reason for the difference may be a lack of estrogen.

We found that s_unclassified__g_Ruminococcus and s_unclassified__g_Butyrisimonas were enriched in the BDL group, but s_gut_metagenome__g_Lachnospira was enriched in the BDH group. Genus Butyricimonas, Lachnospira and Ruminococcus are the main members of the Clostridiales. This paradoxical result has also been reflected in previous studies. A study of the GM in an elderly Chinese population reported that g_Clostridium plays an important role in maintaining bone health, while another assessing study of analysis estimated heel bone mineral density (eBMD) genome–wide association research (GWAS) in 426,824 participants found a negative correlation between g_Clostridium abundance and BMD19,31. g_Lachnospira is closely related to heel BMD32. Probiotic metabolite Kefir fermentation peptide can regulate GM in OVX mice, leading to an increase in g_Ruminococcus33. SCFAs produced by g_Butyrisimonas participate in bone homeostasis regulation34. The results indicate that different genus of Clostridiales has different effects on regulating BMD, and their impact on BMD may be constrained by various factors such as region, population, and age.

In our study, the relative abundance of g_Kocuria in the BDM group was significantly higher than in the BDH and BDL groups, and g_Subdoligranulum was related to the BDL group. g_Kocuria infection is associated with low immunity, which has been shown to induce diseases such as peritonitis, myelodysplastic syndromes (MDS) and patients with short bowel syndrome are at increased risk of opportunistic infections such as with g_Kocuria, which may increase the likelihood of developing foot osteomyelitis35. g_Kocuria infection is most likely related to the expression of IL–6, IL–8, IFN–γ36. IFN–γ promotes osteoclast maturation in the late stage of osteoclast formation37. g_Subdoligranulum was found to stimulate TH17–cells proliferation, serum RA–relevant IgG autoantibodies, and joint swelling, while mice colonized with g_Subdoligranulum isolates developed OA, further analysis found that g_Subdoligranulum elicited CD4 T–cells response38. CD4 T–cells enter the thymus gland and differentiate to eventually form CD4 + T–cells, CD4 + T–cells can produce proinflammatory cytokines involved in inflammatory bone loss, increase osteoblast apoptosis, and promote osteoclastogenesis39. We can see that there is an intimate link among g_Kocuria and g_Subdoligranulum, the immune system, and BMD, and that g_Kocuria and g_Subdoligranulum may disrupt bone homeostasis by affecting the immune system. Interestingly, we annotated the s_Eubacterium eligens group (g_Eubacterium) was associated with the BDL group. g_Eubacterium is one of the core taxa of the human intestinal community, and its fermented glucose products typically contain large amounts of butyrate, acetate, or propionate40. Butyrate, acetate and propionate are the most common SCFAs, and SCFAs stimulate bone formation by increasing the production of bone Sialoprotein (BSP) and osteopontin (OPN) in osteoblasts, while also inhibiting osteoclast differentiation by inhibiting protein deacetylase and activating free fatty acid receptors (FFARs)41,42. Therefore, the appearance of g_Eubacterium in the gut tract of the BDL individuals is a very positive signal, indicating that g_Eubacterium has potential application value in protecting bone mass and regulating bone metabolism. High–fiber diets and Mediterranean diets have been shown to increase the relative abundance of g_Eubacterium43,44. This may provide a new idea for preventing and treating OP.

In the present study, Anaerobic bacteria were the dominant in the GM community, and a higher relative abundance of facultative anaerobic and aerobic bacteria in the BDH group compared to the BDM and BDL groups. The relative abundance of anaerobic bacteria was the highest between osteoarthritis (OA) and the control group, and the relative abundance of facultative anaerobic bacteria in OA was lower than that in the control group, but the relative abundance of aerobic bacteria in OA was higher than that in the control group45. Exercise, diet, and medication may all affect the gut environment by altering intestinal permeability, stimulating the production of immune factors46. A GWAS study investigated the composition of the GM in 181 elderly adults and found that some of the bacteria of Proteobacteria, such as g_Sutterella, f_Enterobacteriaceae and g_Citrobacter were highly enriched in the elder OP group21. Other studies have found that g_Sutterella secrete IgA protease, which can break down IgA in the intestinal mucosa, destroy the intestinal antibacterial immune system, and may cause intestinal inflammation. Intestinal inflammation increases intestinal permeability, decreases intestinal absorption of SCFAs and Ca2+, and increases the risk of osteoporosis47,48. In our study, we found that g_Sutterella was related to the BDH group, and the relative abundance of g_Sutterella was the highest in the aerobic phenotype. Some nutritional and psychiatric studies have pointed out that increased intake of dietary fatty acids leads to an increase in the abundance of g_Sutterella. g_Sutterella is a major component of the GM in more than half of children with autism, and altered intestinal permeability may be a cause of autism in children49,50. Therefore, we hypothesize that the high abundance of g_Sutterella was influenced by dietary and emotional changes, but this hypothesis needs further investigation. In addition, it is necessary to pay intimate attention to the psychological and mental state of the participants in the BDH group to avoid possible autistic tendencies.

In conclusion, our study demonstrates that the alpha diversity of GM in the BDL group are low. g_Prevotella and g_Sutterella play an important role in the three BMD groups, and these two taxa may be good candidates for monitoring and controlling OP.

However, our study has several limitations. First, the relatively small sample size might also have reduced the statistical power to detect subtle differences among the taxa, this may be one of the reasons why we did not found associations between GM and T-score, and differences in identification of dominant taxa at different perspectives. But the sample size of our study approached that of most similar studies20,51. Second, high–level BMI or WHR may lead to differences in alpha and beta diversity52,53, and some individuals of this study had high–level of BMI and there was a significant difference in WHR between groups. After regression analysis, not significant associations were found between GM abundance and T–score, and caution should be taken in interpreting these results. In addition, the study was not in depth enough, especially since it relied solely on 16S rRNA sequencing of stool samples to study the GM, a causal link between changes in GM and BMD cannot be determined. Therefore, it is necessary to consider the confounding factors of individuals in larger cohorts in future studies in order to ensure that the study is more rigorous.

Conclusion

The results of this study showed that the composition and structure of GM were changed. The richness and diversity of GM in the BDL were lower than in the BDH and BDM groups. Genus Prevotella, Eubacterium and Sutterella are potential candidate microbiota with diagnosis and therapy of OP.