Fecal samples were collected from postmenopausal women, and all experimental procedures were ethically approved by the Medical Ethics Committee of Shanghai Ninth People’s Hospital (SH9H-2019-T101-2). Participants were recruited from the Osteoporosis Clinic of Shanghai Ninth People’s Hospital. Informed written consent was obtained from all participants, adhering to the Declaration of Helsinki.

All the subjects in the current work were strictly enrolled. They were from a relatively concentrated environment, the differences in the diets were relatively small. Inclusion criteria for postmenopausal women were as follows: age older than 50 years and at least 12 months since their last menstrual period. Exclusion criteria encompassed cancer, chronic liver disease, heart disease, renal disease, diabetes, and secondary osteoporosis-related conditions (hyperthyroidism, steroid abuse, Cushing’s syndrome, hyperparathyroidism, etc.). The participants using antibiotics or affected bone metabolism drugs (such as estrogen and glucocorticoids) over the past three months, or are currently were excluded. Unclear primary osteoporosis diagnosis, untreated osteoporosis, or a history of brittle fracture of individuals are excluded. Besides, factors that might affect the gut microbiome were taken into account. Participants taking probiotics were excluded. After the age and weight matching. Finally, five postmenopausal osteoporosis women and five healthy controls were included.

Dual-energy X-ray absorptiometry (DXA, Hologic Discovery A, United States) was then utilized to measure bone mineral density (BMD) and T-score in the postmenopausal women donors. The two groups were defined based on the T-score threshold corresponding to BMD. Following the World Health Organization’s definition of osteoporosis, a T score of ≤ -2.5 indicates osteoporosis, while a T score of ≥ -1 indicates normal bone mass (Kanis, 1994; Compston et al., 2019). Subjects were categorized into the postmenopausal osteoporosis group (n=5) and the healthy control group with normal bone mass (n=5) as donors for fecal microbiota transplantation.

Fresh stool samples were collected from each participant within 15 minutes of excretion, then promptly frozen and preserved in liquid nitrogen for subsequent fecal microbiota transplantation.

Fourteen 7-week-old female C57BL/6 mice were procured from Shanghai Jiesijie Laboratory Animal Co., LTD. The mice were housed in a specific pathogen-free (SPF) environment with a 12-hour light-dark cycle, and they had ad libitum access to pelleted rodent feed and autoclaved water. After one week of acclimation, the mice were randomly assigned to two groups: the Vehicle group (n=7) received intestinal bacteria transplantation from the healthy control group, and the FMT group (n=7) received intestinal bacteria transplantation from postmenopausal osteoporosis patients. All animal experiments were ethically approved by the Ethical Committee of the Ninth People’s Hospital of Shanghai Jiao Tong University School of Medicine and adhered to the recommendations of the Guide for the Care and Use of Laboratory Animals by the Chinese Association for Laboratory Animal Science.

Based on previous studies, both groups of mice were treated with an antibiotic cocktail (ABX) to establish a pseudo-sterile mouse model (Borody et al., 2013; Liu et al., 2020; Wang et al., 2020). C57BL/6 mice received a mixture of antibiotics (Vancomycin, 100 mg/kg; Neomycin sulfate 200 mg/kg; Metronidazole 200 mg/kg; Ampicillin 200 mg/kg) via gavage for 14 days, with the drinking solution changed every other day to deplete the gut microbiota. Fresh fecal samples were collected on day 15.

Fecal microbiota transplantation (FMT) was conducted following established methods (Hamilton et al., 2012; Borody et al., 2013). The feces (1.5 g) of five humans collected from each group were mixed, steeped, and shaken in sterile PBS (pH 7.4, 10 mL) and then filtered through a 100 μm pore mesh. Centrifugation at 800G for 3 minutes was performed to obtain two sets of supernatants, which were subsequently frozen at −80°C. At 8 weeks of age in the pseudo-sterile mice, the diluted supernatant (200μL per mouse) was administered to recipient mice daily via gavage for 8 weeks. The body weight of the two groups of mice was monitored weekly. One day before euthanasia, the mice were fasted for 12 hours, and fecal samples were collected and stored at −80°C. Then, the mice were weighed after anesthesia. And femur, serum, and colon tissues were collected from each mouse for subsequent analysis. The operation flow chart was created using Adobe Illustrator 2020 (V24.0.1.341).

At the end of the experiment, mice were fasted for 12 h and then given sodium pentobarbital (100 mg/kg). Mice were sacrificed (Euthanasia by CO₂ inhalation) after blood collection from the abdominal aorta. Serum was obtained by centrifuging the blood for 10 min at 4000 rpm/min (4°C). Bilateral femoris were dissected, removed, and fixed in 4% paraformaldehyde. Approximately 2 cm of small intestine was removed and washed with PBS. A portion of the small intestine was frozen at − 80 °C and the remainder was fixed in 4% paraformaldehyde, for subsequent analyses.

Ipsilateral femur samples from each group were collected and fixed in 4% paraformaldehyde. The bone structure of mouse femurs was evaluated using a high-resolution CT81 system (Scanco Medical AG, Bruettiselien, Switzerland) (Fu et al., 2020). The structural parameters of the femur were further analyzed using program CTAn (Bruker, Karlsruhe, Germany). 3D reconstruction of the tibial metaphyseal was performed using the built-in Scanco software.

Mice serums were separated by centrifugation (2500 × g, 10 min) and stored at – 80°C. CTX-1 elisa assay kits (LSBio cat. no. LS-F21349), Soluble Cluster of Differentiation 14 (sCD14) elisa assay kits (Cell Sciences cat. #: CKM034), TNF-α elisa assay kits (Beyotime, China. cat. no.PT512), IL-17A elisa assay kits (Beyotime, China. cat. no. PI545) were used to ELISA measured. All procedures were performed according to the manufacturer’s instructions.

Fasting blood glucose from mice was measured with the use of a handheld glucose meter, FreeStyle Lite Glucose strip (Abbott Laboratories), from serum samples obtained by centrifugation after a 12-hours fast.

After fasting for 4 h, mice were given 200 μl FITC-dextran (FD4) (440 mg/kg body weight) by gavage, and blood samples were collected 4 h later. The concentration of FITC-dextran was determined using a fluorometer with an excitation wavelength of 490nm and an emission wavelength of 530 nm. Fitc-dextran in serum was serially diluted. It was used to establish a standard curve.

Microbial DNA was extracted from the stool samples using the QIAamp DNA Stool Mini Kit (QIAGEN, USA). The DNA was subsequently fragmented into 400 bp readings and processed with the Biomek^® FXp (Bio Scientific, USA) compatible NEXTflex™ DNA sequencing kit to construct a paired-end genomic library. Paired-end reads were generated using an Illumina HiseqTM2500 platform. Raw FASTQ files were filtered using the FASTX-Toolkit. The high-quality reads were then subjected to splicing and assembly using Mothur software (V1.44.1) with the main splicing parameter Kmer values set between 55 and 85. Scaffolds exceeding 500 bp were selected for bioinformatics analysis in preparation for shallow shotgun sequencing.

Shallow shotgun sequencing was conducted using the Illumina NovaSeq sequencing platform. Predicted gene sequences in the stool samples were clustered using CD-HIT (version 4.6.1) with parameters set to 90% identity and 90% coverage to construct non-redundant gene sets. The high-quality readings of each sample were then aligned with the non-redundant gene set (95% identity) using SOAPaligner software (version 2.21), and the abundance of genes in each sample was calculated. Sequences of the non-redundant gene sets were compared with the database RefSeq version 82 using Diamond (version 0.8.35) with an expected value of 10^-5 for blastp alignment. Species annotations were obtained through the corresponding taxonomic information database of the database RefSeq, and species abundances were determined by summing the abundances of the corresponding genes for each species (Wang et al., 2022b).

Alpha diversity analysis (ACE, Chao, Shannon, and Simpson indices) were conducted using Mothur software (V1.44.1). Principal component analysis (PCOA) was employed to visualize the similarities or differences between the two groups by calculating ecological distance. LEfSe (linear discriminant analysis effect size) was utilized to compare the two groups, and the Kruskal-Wallis test and paired Wilcoxon rank-sum test and linear discriminant analysis (LDA) were employed to identify biomarkers with both statistical differences and biological significance, aiming to identify features with different abundances and related categories. The enzyme functions were classified based on the NC-IUBMB classification and nomenclature of enzymes (commonly known as the Enzyme List, an officially recognized functional classification system). Enzyme data can be found in IntEnz (http://www.ebi.ac.uk/intenz), supported by NC-IUBMB (Fleischmann et al., 2004). Enzyme abundances between groups were compared by Benjamini-Hochberg FDR. Corrected p < 0.05 was considered significant in predicting pathway status across the community (Hillmann et al., 2018).

The α diversity visualization, columnar stacking maps at the phylum, genus, and species levels, principal coordinate analysis, and correlation heat map were generated using R software (V4.2.2). Enzyme enrichment analysis visualization was performed using stamp (V2.1.1.0). Welch’s t-test and White’s non-parametric t-test were used to compare profiles between the two groups, and the effect size was measured using the difference in mean proportion along with Welch’s confidence intervals. Principal component analysis (PCA) plots and extended error bar plots were created based on corrected p-values (Parks et al., 2014). LEfse analysis and linear discriminant analysis visualization by Galaxy cloud Platform implementation (galaxy Platform, http://huttenhower.sph.harvard.edu/galaxy) (Boekel et al., 2015).

Bilateral femoral samples were collected and fixed in 4% paraformaldehyde. Decalcification was performed with a decalcification solution containing EDTA for 2 weeks at room temperature. After decalcification, the femoral tissue was embedded in paraffin and cut longitudinally into 5 μm thick sections. The sections were subjected to H&E (Hematoxylin-eosin) and TRAP (Tartrate resistant acid phosphatase) staining using the TRAP staining kit (Servicebio, Wuhan, China). After washing off the staining solution, the HE-stained and TRAP-stained femur sections were observed under an optical inverted microscope (Olympus, Japan), and the number of TRAP-positive osteoclasts was quantified using ImageJ software (NIH, Bethesda, MD, USA).

After fixation with 4% (v/v) paraformaldehyde, the distal ileum of mice was embedded in paraffin and cut into 4 µm thick sections. The sections were stained with H&E to observe the morphological features of the two groups under an Olympus light microscope (Japan).

After blocking sections with 0.3% triton X-100, in Tris-hcl buffer (0.1 M) containing 5% fetal bovine serum, Incubation with anti-ZO-1 (1:100, NOVUS), anti-claudin (1:100, SAB signaling pathway antibody), and anti-occludin (1:100, Proteinbtech) primary antibodies was performed overnight at 4 °C. Jackson Immunology Research Laboratories LTD), secondary antibody was FITC-labeled goat anti-rabbit and Fluor 488-conjugated goat anti-mouse (GeneTex, Irvine, CA, USA). Nuclei were counterstained with 40, 6-diamidino-2-phenylindole (DAPI) (1:30 000 Roche Biochemistry, Monza, IT). A laser scanning confocal microscope (Olympus, Japan) was used for image acquisition. ImageJ software was used for quantification (NIH, Bethesda, Maryland, USA).

For immunofluorescence of femora, the primary antibodies were anti-TNF-α (Servicebio, China) and IL-17A (Servicebio, China), the secondary antibodies were Cy3-labeled goat anti-mouse IgG (Servicebio, China), Fluorescence intensity quantification was done using Image J. The remaining procedures were the same as above.

For OCN staining, femur sections were antigen repaired and then blocked with 100%BSA and Triton 3. The sections were then treated with primary anti-osteocalcin antibodies (catalog M173; Takara) at 4°C overnight. Immunoreactivity was determined using HRP-DAB cells in the tissue staining kit according to the manufacturer’s instructions (Yang et al., 2019).

GraphPad Prism 9.0 statistical software was used to visualize body weight gain, all skeletal parameters, and count data of osteoblasts and osteoclasts in mice. SPSS (SPSS^®Statistics v26) was used for independent-samples T test to analyze statistical differences. The overall quantified data were presented as mean ± standard deviation (SE). P-value < 0.05 was considered statistically significant.

Ten female subjects provided fecal samples, including 5 postmenopausal women without osteoporosis as healthy controls (Con group) and 5 postmenopausal women diagnosed with osteoporosis (Op group). The inclusion and exclusion criteria were detailed in the Methods section. The age and body weight of the two groups were comparable (P > 0.05). The mean bone mineral density (BMD) was significantly lower in the Op group (P < 0.05) ( Table 1 ).

The FMT process was depicted in Figure 1A . Colony counts before and after ABX treatment indicated a significant impact of antibiotic administration before FMT, successfully establishing a pseudo-sterile mouse model ( Figure 1B ). The body weight of the mice during FMT was recorded. The FMT group showed faster weight gain after receiving the gut microbiota of Op patients, showing a statistically significant difference from 4 weeks after receiving gut microbiota transplantation (p < 0.05, Figure 1C ). After 8 weeks, compared with the weight before fecal microbiota transplantation, the weight gain of the mice in the FMT group was significantly higher than that in the Vehicle group (p < 0.01, Figure 1D ).

We first analyzed mouse bone tissue to verify the transfer of osteoporosis representations. Micro-CT analysis was conducted to assess changes in the trabecular bone structure of the distal femur in mice inoculated with the intestinal flora of postmenopausal osteoporosis patients. Three-dimensional images ( Figure 2A ) revealed evident bone loss in the FMT group, with an enlarged medullary cavity and reduced bone trabeculae in the distal femur. Furthermore, the common bone parameters showed that the FMT group had significantly lower bone mineral density (BMD, p < 0.05), bone volume (BV, p < 0.01), and bone volume fraction (BV/TV, p < 0.05). ( Figure 2B ). Then, we analyzed bone microstructural parameters to assess bone condition. FMT group had less trabecular number (Tb. N, p > 0.05), smaller trabecular thickness (Tb. Th, p < 0.05), and higher trabecular separation (Tb. Sp, p < 0.05) ( Figure 2C ). To obtain a more comprehensive understanding of the bone condition during fecal microbiota transplantation, the related parameters of cortical bone were analyzed. The results showed that cortical bone thickness (Ct. Th, p < 0.05) was significantly decreased in the FMT group. There was no significant difference in cortical bone area (Ct. Ar, p > 0.05) between the two groups. However, the ratio of cortical bone area and total cortical bone area (Ct. Ar/Tt. Ar, p < 0.05) was significantly higher in the FMT group ( Figure 2D ). These results indicate that the bone loss and bone microstructural changes in the FMT group of mice, which were similar to those in osteoporosis patients.

The differences in bone mass could be directly reflected by the staining of bone tissue sections. H&E staining of the femur showed a decrease in the area of trabecular bone and trabecular bone itself in the FMT group, indicating obvious bone mass loss ( Figure 3A ). We subsequently examined the concentration of serum CTX-1 in mice and found a significant increase in the FMT group ( Figure 3B ). To further clarify the effect of enterobacteria transplantation on bone mass, Trap staining was performed on femoral samples of mice. Trap staining revealed significant osteoclast differentiation in the FMT group, and the quantitative results of Trap staining indicated a notable increase in the number of osteoclasts along the trabecular bone ( Figure 3C ). At the same time, we analyzed the performance of bone formation in mice. In the analysis of immunohistochemical results, OCN staining of the femur in the FMT group showed a reduction in the number of osteoblasts and attenuated bone formation ( Figure 3D ). These results demonstrated that the bone metabolic status of the mice was changed, as reflected by the enhanced bone resorption and decreased bone formation in the FMT group.

Previous studies demonstrated the important influence of gut microbiota on osteoporosis. We asked about the role of the gut microbiota in altered bone metabolism. To clarify the changes of bacterial flora during FMT. We first used shallow shotgun sequencing to collect data on the donor gut microbiota. Among postmenopausal women, Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and Verrucomicrobes were dominant phyla ( Figure 4A ). In addition, the relative abundance of the microbiota at genus level was characterized in the two groups of donors, Bacteroides (0.40 vs 0.079, p < 0.01) were significantly enriched in the Con group, while Lachnospiraceae (0.0080 vs 0.12, p < 0.05) and Eubacterium (0.019 vs 0.062, p < 0.05) were significantly enriched in the Op group ( Figure 4B ). To further analyze the differences and similarities of the microbiota between the two groups, Bray-Curtis based principal coordinate analysis (PCOA) was used, and the results showed that the distribution of the microbiota was significantly different between the two groups ( Figure 4C ). Alpha diversity showed that there was no significant difference in ACE index and Chao 1 index between the two groups. The Shannon index and Simpson index of intestinal bacteria in Op group were significantly increased ( Figures 4D–G ). These results showed the gut microbiota increased richness and the significant changes in gut microbiota composition and structure of the Op group.

To elucidate the colonization of intestinal flora in mice after transplantation, Venn diagram was used to visualize intestinal bacterial colonization. Among the bacterial genera in the FMT group, 257 (60.05%) were consistent with those in the Op group (The total number of genus-level bacteria was 428), and 252 (60.58%) were consistent with those in the Vehicle group (The total number of genus-level bacteria was 416), which showed that the two groups of mice had comparable colonization effect ( Figure 5A ). Similarly, we further analyzed the differences in bacterial abundance at the genus level in mice. We found that among the top 30 bacterial genera in relative abundance, Lachnospiraceae (0.38 vs 0.17, p < 0.05), Oscillibacter (0.036 vs 0.019, p < 0.05), and Firmicutes (0.0093 vs 0.0046, p < 0.05) were significantly increased in the Vehicle group, while Halobiforma (0.096 vs 0.16, p < 0.05), Enterorhabdus (0.021 vs 0.10, p < 0.05), Alistipes (0.010 vs 0.026, p < 0.01), Butyricimonas (0.0026 vs 0.020, p < 0.01), Natronorubrum (0.0065 vs 0.011, p < 0.05), and Streptomyces (0.0020 vs 0.0027, p < 0.05) were significantly increased in the FMT group ( Figure 5B ). To further clarify whether the changes in microbiota richness in human subjects were successfully replicated in mice, we analyzed the alpha diversity of the two groups of mice. The results showed that ACE index, Chao 1 index, Shannon index, and Simpson index were significantly increased in the FMT group ( Figures 5C–F ). PCOA analysis also showed a significant difference in the microbiota composition between the two mice groups ( Figure 5D ). These results indicated that the composition and structure of gut microbiota in FMT mice were altered similarly to those in Op mice, as reflected by the increased richness of microbiota and the significant increase in the relative abundance of specific bacterial groups.

To further identify the intestinal flora affecting bone mass metabolism. LEfSe analysis was employed to identify statistically and biologically significant species. The cladogram revealed significantly enriched genera, seventeen bacterial genera were enriched in the FMT group and 3 bacterial genera were enriched in the Vehicle group ( Figure 6A ). To screen out the final differential bacteria (biomarker), we set the LDA value to >3.5. The enriched bacterial genera were visualized as LDA histograms. Finally, we identified 4 dominant genera that were significantly enriched in the FMT group, including Halobiforma (0.096 vs 0.16, p < 0.05), Enterorhabdus (0.021 vs 0.10, p < 0.05), Alistipes (0.010 vs 0.026, p < 0.01), Butyricimonas (0.0026 vs 0.020, p < 0.01). Besides, two dominant genera in the Vehicle group, including Lachnospiraceae (0.38 vs 0.17, p < 0.05), Oscillibacter (0.036 vs 0.019, p < 0.05) ( Figures 6B–H ).

The underlying mechanisms of bone loss induced by these dominant species were further explored. Our previous work had demonstrated the critical role of gut barrier impairment in the gut microbiota-osteoporosis axis (Wang et al., 2022a). We therefore first asked whether gut microbiota derived from human donors could contribute to gut barrier damage in mice. H&E staining showed obvious intestinal mucosa damage in the FMT group, with aggravated swelling and disordered villus structure ( Figure 7A ). In line with these observations, FMT group mice showed increased small intestine permeability for FITC-dextran MW 4000 (FD4) ( Figure 7B ). To add more evidence of intestinal barrier dysfunction, we analyzed serum sCD14 levels, which has been described as a marker of microbial translocation. ELISA results showed that the serum sCD14 content of mice in the FMT group was significantly increased ( Figure 7C ). To obtain further direct evidence of intestinal permeability impairment, immunohistochemical staining was used to detect the content of Claudin, Occludin, and ZO-1 related to intestinal permeability. Immunohistochemical staining revealed significantly reduced expression of claudin, occludin, and ZO-1, which were related to the integrity of the intestinal epithelial barrier, compared with the Vehicle group. Staining quantification demonstrated statistically significant differences in the expression of claudin and ZO-1 between the two groups ( Figures 7D-F ). These results demonstrated significantly damage to the intestinal barrier in mice transplanted with gut bacteria from osteoporosis patients.

To clarify the role of gut microbiota disorder in bone loss and intestinal barrier damage, the correlation of bone loss and impaired intestinal barrier with gut microbiota was visualized as heatmaps based on Spearman analysis. We found a significant positive correlation between Lachnospiraceae and Ct.Th, and a significant negative correlation between Lachnospiraceae and CTX-1. Oscillibacter was significantly positively correlated with BV and BV/TV, and negatively correlated with Tb.Sp, Ct.ar/Tt.ar. Enterorhabdus was negatively correlated with BMD, Tb.Th, and Ct.Th, and positively correlated with Ct.ar/Tt.ar, sCD14, and CTX-1. Alistipes was significantly negatively correlated with Ct.Th. Butyricimonas was negatively correlated with Ct.Th and positively correlated with FD-4 ( Figure 8A ). These results suggested that bone loss and gut barrier impairment are significantly associated with dysbiosis.

To investigate the effect of gut microbiota on the expression function of specific genes, we utilized PICRUST to predict the function of sequencing data based on the integrated relational enzyme database IntEnz. This approach helped identify biologically relevant differences between the two groups. Principal coordinate analysis revealed significant differences in the enzyme expression functions of the two groups under the three principal components ( Figure 8B ). Expanded error bar plots were used to characterize the differentially expressed enzymes between the two groups, and 10 enzymes were significantly up-regulated. In contrast, 13 enzymes were significantly down-regulated in the FMT group ( Figure 8C ). Furthermore, Spearman correlation analysis based on heatmap was performed to determine the correlation between the differentially expressed enzymes and the gut microbiota ( Figure 8D ). Overall, Lachnospiraceae was strongly associated with the detected enzymes whose expression were significantly upregulated in the FMT group. Enterorhabdus and Butyricimonas also played important roles in regulating the characteristic enzyme profile. In addition, Oscillibacter interacts with the enzyme with a disulfide as acceptor (1.2.4), and there was a significant negative correlation between the expression of enzymes with NAD+ or NADP+ as acceptor (1.5.1) and enzymes transferring alkyl or aryl groups (2.5.1). Halobiforma was positively correlated with Endodeoxyribonucleases producing 5-phosphomonoesters (3.1.21) and the enzymes in cyclic amidines (3.5.4). Alistipes were positively correlated with the enzymes with an iron-sulfur protein as acceptor (1.17.7). These results suggest that the transplanted gut microbiota altered the expressed enzyme profile of the mice, which has important implications for the regulation of metabolic reactions in which the enzymes are involved.

To explore whether the bacterial translocation caused inflammation after the change of intestinal barrier after transplantation, we used immunofluorescence to detect the expression of TNF-α and IL-17A in the proximal femur of mice. The immunofluorescence results showed that the inflammatory markers (TNF-α and IL-17A) in the femur of mice in the FMT group were significantly up-regulated. Quantitative results showed that expression levels were significantly different between the two groups ( Figures 9A, B ).

Bacterial translocation and femoral inflammation are usually accompanied by inflammatory changes in the blood of mice, which may be the connection mechanism of intestinal barrier and bone mass loss. Therefore, we detected the changes of TNF-α and IL-17A in the serum of mice, and the results showed that the serum TNF-α and IL-17A in the FMT group were significantly increased ( Figures 9C, D ).

In addition, the fasting blood glucose content of the two groups of mice was examined, and the results showed no statistically significant difference in blood glucose between the two groups of mice ( Figure 9E ), which ruled out impaired glucose tolerance after fecal microbiota transplantation.