Gut Microbiota and SCFA Changes After Alendronate–Calcitriol Therapy in Postmenopausal Osteoporosis

Abstract

Background

Postmenopausal osteoporosis (PMO) is driven primarily by estrogen deficiency, and gut microbiota-derived metabolites may participate in bone remodeling. The intestinal effects of conventional anti-osteoporosis therapy remain unclear. We investigated gut microbial and short-chain fatty acid (SCFA) changes associated with alendronate–calcitriol therapy in PMO.

Methods

Forty-one postmenopausal women were enrolled, including 21 healthy controls and 20 patients with PMO. PMO patients received oral alendronate plus calcitriol for six months, with paired fecal samples collected before and after treatment. Shotgun metagenomic sequencing was used to assess microbial taxonomy, KEGG pathways, and CAZy profiles. Fecal SCFAs were quantified by gas chromatography–mass spectrometry.

Results

After treatment, lumbar spine BMD T-score improved from −2.98 to −2.65, and femoral neck BMD T-score increased from −2.40 to −2.12, and all measured bone turnover markers (BALP, TRACP-5b, PINP and osteocalcin) decreased significantly (P < 0.05). Global microbial diversity did not differ significantly among groups, but treatment was accompanied by shifts in specific taxa. Roseburia, Megamonas, Dialister, and unclassified Lactobacillaceae increased, whereas Enterobacter, Salmonella, Raoultella, Serratia, and Rahnella decreased. Functional profiles showed treatment-associated shifts in propanoate metabolism, pyruvate metabolism, the TCA cycle, and selected carbohydrate-active enzyme patterns. Overall fecal SCFA composition remained relatively stable; however, propionic acid concentration increased significantly after treatment (FDR-adjusted P = 0.021). Correlation analyses linked selected commensal taxa and SCFAs with BMD and bone turnover markers.

Conclusion

Alendronate–calcitriol therapy was associated with improved bone metabolism and treatment-associated shifts in gut microbial composition, microbial functional potential, and propionate-related SCFA profiles. These findings provide preliminary evidence for microecological changes accompanying conventional PMO therapy.

Introduction

Postmenopausal osteoporosis (PMO) is a common age-related skeletal disorder primarily driven by estrogen deficiency. This hormonal change disrupts the balance between bone resorption and formation, leading to progressive bone loss and an increased risk of fragility fractures.^1–3 PMO affects a substantial proportion of postmenopausal women⁴ and imposes a considerable burden in terms of morbidity and healthcare costs.⁵ In current clinical practice, alendronate sodium and calcitriol are commonly used drugs for the treatment of PMO. Alendronate sodium can inhibit the activity of osteoclasts, thereby reducing bone resorption. Calcitriol can enhance the intestinal absorption of calcium and improve calcium homeostasis. Previous clinical and experimental studies have shown that combined treatment with alendronate and calcitriol improves bone mineral density or bone strength in postmenopausal osteoporosis and ovariectomy-induced osteopenia models.^6,7 However, whether this standard anti-osteoporotic regimen also reshapes gut microbial dysbiosis and thereby modulates the gut–bone axis remains unclear.

The gut microbiota regulates skeletal homeostasis through the “gut-bone axis”,^8,9 and clinical evidence indicates that PMO is accompanied by gut microbial dysbiosis and altered microbial metabolic profiles.¹⁰ In cohorts, Firmicutes and Clostridiales are positively correlated with bone mineral density (BMD) and depleted in PMO patients.^11,12 Estrogen deficiency drives microbial compositional shifts, impairs intestinal barrier function, and promotes systemic inflammation, which in turn enhances osteoclast activity and accelerates bone loss.^13–15 Short-chain fatty acids (SCFAs), key microbial metabolites, mediate host-microbe interaction and regulate bone metabolism by modulating immune responses, enhancing mineral absorption, and suppressing osteoclastogenesis.^16–18 Consistently, SCFA-producing bacteria such as Faecalibacterium prausnitzii and Roseburia are associated with better bone health,^19–23 and reduced fecal propionic acid has been reported in PMO patients.² Orally administered drugs can interact bidirectionally with the gut microbiota, and microbial composition may influence therapeutic efficacy.^19,24 Nevertheless, the effects of conventional anti-osteoporotic treatments on gut microbial structure and SCFA metabolism remain poorly defined.^24,25 We therefore hypothesized that PMO is characterized by gut microbial dysbiosis with reduced SCFA-producing bacteria and disrupted SCFA profiles, and that alendronate-calcitriol therapy is accompanied by partial normalization of these microbial and metabolic features.

High-throughput sequencing enables high-resolution characterization of gut microbiota composition and function, allowing more precise relationships between microbial communities and human health or disease states.²⁵ Although 16S rRNA gene sequencing has been widely used to describe overall microbial community structure, its resolution is generally limited to the genus level and it provides only indirect inference of functional potential. In contrast, shotgun metagenomic sequencing offers species-level taxonomic resolution and simultaneously captures gene-level functional profiles, including Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and carbohydrate-active enzyme (CAZy) annotations. In addition, gas chromatography‑mass spectrometry (GC-MS)-based metabolomics allows systematic identification and quantification of microbiota-derived metabolites, thereby elucidating molecular pathways through which the microbiota influences host physiology.²⁶ Integrating high-throughput sequencing with metabolomics therefore enhances our understanding of how the microbiome dynamically shapes the organ environment, host metabolic status, and the microbiota–host co-metabolic network.²⁷

In this study, we integrated shotgun metagenomics and GC-MS-based targeted SCFA metabolomics to profile fecal samples from 41 postmenopausal women, including healthy controls and PMO patients assessed before and after six months of oral alendronate plus calcitriol therapy. We aim to characterize PMO-associated microbiota and SCFA alterations, and to determine whether therapeutic bone improvements are accompanied by concurrent shifts in gut microbial and metabolic homeostasis. These findings may provide new insight into the microecological changes associated with standard PMO treatment.

Materials and Methods

Study Population and Grouping

Postmenopausal women were recruited from Karamay Central Hospital between May 2024 and August 2025. The study was approved by the hospital’s Ethics Committee (YL-2024-005), and all participants provided written informed consent. Inclusion criteria were as follows: (1) postmenopausal status, defined as women aged 50–70 years with at least 12 consecutive months of amenorrhea or ≥1 year after bilateral oophorectomy; (2) availability of bone mineral density (BMD) measurements at the lumbar spine (L1-L4) and femoral neck assessed by dual-energy X-ray absorptiometry (DXA); and (3) willingness to participate with signed informed consent. Based on BMD, participants were stratified into three groups: (i) PMO-pre, defined as T-score ≤ −2.5 at either the lumbar spine or femoral neck; (ii) healthy controls (HC), defined as T-score ≥ −1.0; and (iii) PMO-aft, comprising patients from the PMO group after six months of treatment with alendronate plus calcitriol. Exclusion criteria included: (1) use of medications affecting bone metabolism (eg, bisphosphonates, denosumab, glucocorticoids, thyroid hormones, or estrogen therapy) within the previous six months; (2) use of agents known to alter gut microbiota, including antibiotics, antifungals, probiotics, prebiotics, synbiotics, or proton pump inhibitors, within one month before enrollment or during the follow-up period; (3) history of secondary osteoporosis, other metabolic bone diseases, malignancy, thyroid disorders, or gastrointestinal diseases; and (4) recent gastrointestinal surgery (eg, bariatric surgery, bowel resection, or gastrectomy) within the past three months. Patients in the PMO group received combination therapy consisting of oral alendronate sodium (10 mg once daily) and calcitriol (0.25 μg twice daily) for six months. All PMO participants were additionally prescribed calcium supplementation. Treatment adherence was assessed using different approaches for inpatients and outpatients. For inpatients, compliance was evaluated based on medication administration records and pill counts. For outpatients, adherence was assessed through prescription record verification, pharmacist follow-up, and pill counts. Only patients with ≥80% adherence during the treatment period were included in the final analysis. Based on the manufacturer’s nominal precision of Hologic Discovery - W (lumbar CV 1.0%, femoral neck CV 1.7%), combined with the average BMD of the reference population in this region, the lumbar LSC was calculated to be approximately 0.03 g/cm², and the femoral neck LSC was approximately 0.05 g/cm².

Sample Collection and Processing

Participants were provided with sterile collection tubes one day prior to sampling and instructed to collect fresh fecal samples the following morning. Samples were placed on ice immediately after collection. Approximately 0.5–1.0 g of feces was allocated for metagenomic sequencing, and 0.2–0.5 g for targeted SCFA analysis. Samples were transferred into 2 mL sterile cryovials, transported on ice, and stored at −80°C until analysis. For serum collection, participants fasted for 8–12 hours prior to blood sampling. Venous blood was drawn from the antecubital vein into vacuum tubes containing a clot activator and allowed to clot at room temperature for 30–60 minutes. Serum was then separated by centrifugation at 3000 × g for 15 minutes at 4°C. Aliquoted serum samples were stored at −80°C within 90 minutes of collection for subsequent analyses.

Metagenomic DNA Extraction and Sequencing

Genomic DNA was extracted from fecal samples using the OMEGA Soil DNA Kit (D5635-02; Omega Bio-Tek, Norcross, GA, USA) according to the manufacturer’s instructions. DNA concentration and integrity were assessed using the Agilent 5400 system (Agilent Technologies, USA). Extracted DNA was stored at −20 °C until further processing. For library construction, 0.2 μg of DNA per sample was used as input. Genomic DNA was fragmented to an average size of approximately 350 bp using a Covaris LE220R-plus system (Covaris, USA). The fragmented DNA underwent end repair, A-tailing, and ligation to full-length Illumina sequencing adapters, followed by PCR amplification. Amplified products were purified using the AMPure XP system (Beckman Coulter, USA). Library quality was evaluated using the Agilent 5400 system and quantified by real-time PCR. Qualified libraries were pooled based on effective concentration and target sequencing output, and sequencing was performed on the Illumina NovaSeq X Plus platform to generate 150 bp paired-end reads.

Quality Control, Assembly, and Taxonomic Annotation

Raw image data generated by the sequencing platform were converted into FASTQ format. Quality filtering was performed using Fastp (v0.23.4). Reads were trimmed using a 4-base sliding window with a mean quality threshold of Q20, and bases with quality scores below Q15 were removed. Reads shorter than 15 bp were discarded.

To remove host contamination, reads were aligned to the host reference genome using Bowtie2 (v2.5.1), and only unmapped reads were retained for downstream analysis. High-quality reads were assembled de novo using MEGAHIT (v1.2.9) with a minimum contig length of 500 bp. Assembly quality was assessed using QUAST (v5.2.0).

Coding sequences (CDSs) were predicted from assembled contigs using MetaGeneMark (v3.38), and CDSs shorter than 100 nucleotides were excluded. A non-redundant gene catalog was constructed using MMseqs2 (easy-linclust module) with a sequence identity threshold of 90% and coverage of 90%. Gene abundance was estimated by mapping reads back to the gene catalog using Bowtie2, and normalized as transcripts per million (TPM).

Taxonomic profiling was performed using an assembly-based gene annotation strategy. Specifically, predicted protein sequences from the non-redundant gene catalog were aligned against the NCBI non-redundant NR meta database using DIAMOND (v2.1.8), with an e-value cutoff of 1×10⁻⁵. Taxonomic lineages from domain to species level were assigned based on the DIAMOND-NR annotation results. The relative abundance of each taxon at different taxonomic ranks was calculated by summing the TPM values of all unigenes assigned to the corresponding taxon. MetaPhlAn was not used in the primary taxonomic profiling analysis.

Functional Annotation and Analysis

Predicted protein sequences were functionally annotated against the KEGG and the Carbohydrate-Active enZYmes (CAZy) databases using DIAMOND or database-recommended tools. Annotations with e-values ≤ 1e−5 were retained. Functional abundance profiles were generated by integrating annotation results with gene abundance data.

Differential KEGG pathway abundance was analyzed using covariate-adjusted linear regression with age, BMI, dietary fiber intake, carbohydrate intake, menopausal duration, and serum 25(OH)D included as applicable covariates. CAZy enzyme family differences were assessed using Wilcoxon tests, with rank-sum tests for independent comparisons and signed-rank tests for paired PMO-pre versus PMO-aft comparisons. P values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method.

Bioinformatics and Statistical Analysis

Alpha diversity was calculated from the Unigene count table using QIIME2 with the qiime diversity alpha-rarefaction command. The maximum rarefaction depth was set to the minimum sequencing depth across all samples, with 30 steps and 10 iterations. Two indices were computed: Chao1 (richness) and Shannon (diversity). Rarefaction curves were generated to assess sequencing depth sufficiency, and violin plots were constructed using diversity estimates at the maximum rarefaction depth.

Beta diversity was assessed using Bray–Curtis and Jaccard distance metrics within the QIIME2 framework. Principal coordinate analysis (PCoA) was performed based on the corresponding distance matrices to visualize differences in microbial community composition among groups. Statistical significance of group differences was evaluated using analysis of similarities (ANOSIM) implemented in QIIME2 (qiime diversity beta-group-significance), with permutation-based P values calculated from the Bray–Curtis and Jaccard distance matrices.

Differential taxonomic features at the species level were identified using linear discriminant analysis effect size (LEfSe, v1.1.2), with thresholds set at LDA score > 2.5 and P < 0.05 for biomarker discovery in pairwise and multi-group comparisons.

Functional annotation was performed by aligning predicted protein sequences from the non-redundant Unigene catalog against functional databases. KEGG annotation was conducted using KofamKOALA, and Unigene abundance was normalized as TPM. Functional abundance at each KEGG pathway level was calculated by summing the TPM values of all Unigenes assigned to the corresponding pathway.

Differentially abundant genes were identified from the non-redundant Unigene TPM abundance matrix. For pairwise group comparisons, differences in gene abundance were assessed using Wilcoxon rank-sum tests for independent comparisons and paired Wilcoxon signed-rank tests for paired PMO-pre and PMO-aft comparisons, as appropriate. P values were adjusted using the Benjamini–Hochberg false discovery rate method. Genes with FDR-adjusted P < 0.05 were defined as differentially abundant genes.

KEGG functional abundance profiles were generated by integrating KEGG annotation results with Unigene abundance data. Briefly, Unigene abundance was normalized as TPM, and the abundance of each KEGG Level 3 pathway was calculated by summing the TPM values of all Unigenes assigned to that pathway. Differential KEGG pathway abundance between groups was assessed using covariate-adjusted linear regression, with age, BMI, dietary intake, menopausal duration, and serum 25-hydroxyvitamin D included as covariates where appropriate.

Targeted Metabolomic Analysis of Short-Chain Fatty Acids

Samples were transferred into 2 mL microcentrifuge tubes and extracted with 1.0 mL of ultrapure water, followed by vortex mixing for 10s. The samples were homogenized using a ball mill at 40 Hz for 4 min and subsequently sonicated for 5 min in an ice-water bath. The homogenization and sonication steps were repeated three times. The homogenates were centrifuged at 5000 rpm for 20 min at 4 °C, and 0.8 mL of the supernatant was transferred to a fresh 2 mL microcentrifuge tube. Subsequently, 0.1 mL of 50% sulfuric acid (H₂SO₄) and 0.8 mL of extraction solution containing the internal standard (25 mg/L in methyl tert-butyl ether) were added. The mixture was vortexed for 10s, shaken for 10 min, and sonicated for an additional 10 min in an ice-water bath. After centrifugation at 10,000 rpm for 15 min at 4 °C, the samples were stored at −20 °C for 30 min. The resulting supernatant was transferred into a 2 mL glass vial for GC-MS analysis.

GC-MS analysis was performed using an Agilent 7890B gas chromatograph coupled to a 5977B mass spectrometer (Agilent Technologies, Santa Clara, CA, USA) equipped with an HP-FFAP capillary column. A 1 μL aliquot of each sample was injected in split mode (split ratio, 5:1). Helium was used as the carrier gas at a constant flow rate of 1.2 mL/min, with a front inlet purge flow of 3.0 mL/min. The oven temperature program was as follows: 50 °C for 1 min; increased to 150 °C at 50 °C/min and held for 1 min; increased to 170 °C at 10 °C/min; increased to 225 °C at 25 °C/min and held for 1 min; and finally increased to 240 °C at 40 °C/min and held for 1 min. The injector, transfer line, quadrupole, and ion source temperatures were maintained at 220, 240, 150, and 240 °C, respectively. Mass spectrometric detection was performed in electron impact (EI) ionization mode at 70 eV.

For quantitative analysis, calibration curves were established for all target SCFAs using authentic standards. The calibration curves showed good linearity, with correlation coefficients (R²) ranging from 0.9904 to 0.9950 across the 11 analytes. The calibration ranges were 0.02–50 mg/L for acetic acid, propionic acid, and isobutyric acid; 0.02–20 mg/L for butyric acid, hexanoic acid, isovaleric acid, and valeric acid; and 0.02–0.5 mg/L for heptanoic acid, octanoic acid, nonanoic acid, and decanoic acid. 2-Methylvaleric acid was used as the internal standard at a concentration of 25 mg/L in methyl tert-butyl ether and was added during sample extraction to monitor analytical variability. Quality control (QC) standard samples were inserted after every 10 study samples to evaluate calibration accuracy and instrument stability. The QC accuracy ranged from 89.9% to 109.3%, and the relative standard deviation (RSD) ranged from 7.6% to 10.7%, indicating acceptable analytical precision and system stability. The detailed calibration and QC parameters are provided in Table S1.

Measurement of Serum Bone Metabolism Markers

Serum markers of bone metabolism were quantified using enzyme-linked immunosorbent assay (ELISA). Frozen serum samples stored at −80 °C were thawed at 4 °C and centrifuged at 10,000 × g for 10 minutes at 4 °C. The resulting supernatant was collected for analysis.

The analytes measured included tartrate-resistant acid phosphatase 5b (TRACP-5b) and bone-specific alkaline phosphatase (BALP). Commercial ELISA kits were obtained from Jiangsu Jingmei Biotechnology Co., Ltd. (Jiangsu, China) (TRACP-5b: JM-03987H1; BALP: JM-03982H1). All assays were performed strictly in accordance with the manufacturers’ instructions. Each serum sample was measured in duplicate wells.

Multi-Omics Integrative Analysis (DIABLO)

Integrative analysis of gut microbiota and SCFA datasets was performed with the Data Integration Analysis for Biomarker discovery using Latent variable approaches for Omics (DIABLO) framework, implemented in the mixOmics R package. This approach enables simultaneous modeling of multiple omics layers to identify key variables driving group separation and to characterize their association patterns.

Microbial abundance and SCFA concentration matrices were combined into a multi-block dataset, with group assignment (HC, PMO-pre, PMO-aft) specified as the response variable. A block sparse partial least squares discriminant analysis (block sPLS-DA) model was constructed with two components (ncomp = 2). All variables were scaled (scale = TRUE) to ensure comparability across datasets.

Sample distribution and group separation in the integrated component space were visualized using sample plots. Correlation circle plots were used to identify key microbial taxa and metabolites contributing to group discrimination and to illustrate the direction and strength of their associations. In addition, a correlation network was constructed to depict interactions between core microbiota and SCFAs, using a correlation threshold of |r| ≥ 0.5. Edge color intensity reflected the magnitude and direction of correlations. All analyses were conducted using mixOmics in R (version 4.3.1).

To evaluate the internal performance of the supervised DIABLO model, repeated 5-fold cross-validation was performed using the perf function in the mixOmics package with 50 repeats. Classification error rates were calculated across one and two components using three prediction distances: maximum distance, centroid distance, and Mahalanobis distance. Lower classification error rates indicated better internal discrimination. Because no independent external validation cohort was available, the DIABLO analysis was interpreted as an exploratory supervised multi-omics integration approach rather than as an externally validated predictive classifier.

Sample Size and Power Consideration

No formal a priori sample-size calculation was performed before recruitment because this was an exploratory longitudinal clinical metagenomic study with limited prior data for estimating treatment-associated microbiota and SCFA effect sizes in PMO. To provide quantitative justification, we performed sample-size and post hoc sensitivity power analyses using the pwr package in R. For the main paired PMO-pre versus PMO-aft comparison, assuming a two-sided α = 0.05, 80% power, and an expected within-subject effect size of dz = 0.70, 19 paired PMO patients were required; therefore, the 20 patients who completed baseline and 6-month follow-up assessments met this requirement. For the baseline HC versus PMO-pre comparison, assuming a large effect size of d = 0.90, approximately 21 participants per group were required, and the actual sample size of 21 HC and 20 PMO patients provided approximately 80.2% power. Sensitivity analysis indicated that the actual sample size could detect minimum effect sizes of dz = 0.66 and d = 0.90, respectively (Table S2).

Because shotgun metagenomic analyses involve high-dimensional microbial taxonomic and functional features with multiple testing correction, these calculations indicate that the study was mainly powered to detect moderate-to-large feature-level differences, whereas smaller changes in low-abundance taxa, KEGG/CAZy pathways, SCFA profiles, or multi-omics signatures may have remained undetected.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 10 and R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria). Data normality was assessed using the Shapiro–Wilk test.

PMO-pre and PMO-aft samples were paired longitudinal samples from the same PMO participants. Paired tests or repeated-measures models were used where applicable.

Normally distributed data are presented as mean ± standard deviation (mean ± SD), whereas non-normally distributed data are expressed as median and interquartile range (IQR). For comparisons between the HC and PMO-pre groups, one-way analysis of variance (ANOVA) was applied. Paired comparisons within the PMO group (PMO-pre vs PMO-aft) were conducted using a paired t-test for normally distributed data or the Wilcoxon signed-rank test for non-normally distributed data.

To account for multiple comparisons, P values were adjusted using the Benjamini–Hochberg FDR method for analyses involving multiple testing, including differential taxonomic features, KEGG pathway abundance, CAZy abundance, SCFA concentrations, and correlation analyses between microbial taxa, SCFAs, and clinical parameters. FDR-adjusted P< 0.05 were considered statistically significant. For significantly different taxonomic features, effect estimates with 95% confidence intervals were calculated where applicable. For baseline clinical characteristics and global diversity analyses, unadjusted P values were reported unless otherwise specified.

Intergroup differences in alpha diversity (Shannon and Chao1 indices) were assessed using the Kruskal–Wallis rank-sum test. Beta diversity was evaluated based on the Bray–Curtis distance matrix and visualized using principal coordinate analysis (PCoA). Analysis of similarities (ANOSIM) with 999 permutations was used to test for differences in microbial community structure among groups.

Differences in taxonomic composition at both genus and species levels were initially analyzed using the Wilcoxon rank-sum test. Multiple testing was controlled using the Benjamini–Hochberg FDR method. LEfSe analysis was performed with a linear discriminant analysis (LDA) threshold > 2.5 and P < 0.05 to identify discriminative microbial features. To further validate the robustness of the differential taxonomic findings, MaAsLin2 was applied as a sensitivity analysis at both genus and species levels. Group status was included as the primary fixed effect, and BMI, dietary fiber intake, carbohydrate intake, menopausal duration, and serum 25-hydroxyvitamin D levels were included as covariates. P values were adjusted using the Benjamini–Hochberg method.

To evaluate the independent effect of treatment on gut microbiota composition while accounting for repeated measures, linear mixed-effects models (LMMs) were fitted using the lmer function in the lme4 R package (version 4.1.0). Treatment group (HC, PMO-pre, PMO-aft) was included as a fixed effect, with participant-specific random intercepts to account for within-subject correlation. Covariates, including body mass index (BMI), dietary fiber intake, years since menopause, serum 25-hydroxyvitamin D levels, and carbohydrate intake, were included as fixed effects. Dietary intake was assessed using a structured food frequency questionnaire (FFQ) administered by trained study personnel. The questionnaire was designed with reference to previously validated FFQs used in Chinese adult populations and was used to estimate dietary fiber, carbohydrate, and calcium intake.²⁸ Dietary intake was assessed using a structured FFQ at the initial admission/enrollment visit for all participants and repeated at the 6-month follow-up admission/reassessment visit for PMO patients. The FFQ was administered within the same clinical assessment window as fecal sample collection, fasting blood sampling for bone turnover markers, and DXA examination. Calcium supplementation was not standardized across study groups and was prescribed according to routine clinical practice. Treatment adherence was assessed through follow-up visits and medication monitoring. Significance of fixed effects was evaluated using Satterthwaite’s approximation for degrees of freedom. Post-hoc pairwise comparisons were conducted using the emmeans package with Bonferroni correction. A two-sided P < 0.05 was considered statistically significant.

Spearman rank correlation analysis was used to examine associations between gut microbial taxa and bone-related parameters, including BMD at the lumbar spine and femoral neck, and bone turnover markers (BALP, TRACP-5b, PINP, and β-CTX). To avoid pseudo-replication from repeated observations, primary correlation analyses were restricted to independent baseline samples from HC and PMO-pre participants, excluding paired PMO-aft samples. Correlations were also assessed between differential microbial taxa and four SCFAs (butyric, propionic, acetic, and valeric acids). To avoid pseudo-replication from repeated observations, microbiota–SCFA correlation analyses were restricted to independent baseline samples from HC and PMO-pre participants, excluding paired PMO-aft samples. FDR correction was applied using the Benjamini-Hochberg method, with FDR < 0.05 considered significant. Correlation heatmaps were generated using the pheatmap R package, with color intensity representing Spearman correlation coefficient and asterisks indicating significance after FDR correction (P < 0.05, *P < 0.01).

Differential KEGG pathway abundance was assessed using covariate-adjusted linear regression, whereas CAZy enzyme family profiles were analyzed using Wilcoxon tests, with rank-sum tests for independent comparisons and signed-rank tests for paired PMO-pre versus PMO-aft comparisons. P values were corrected using the Benjamini–Hochberg FDR method, and FDR-adjusted P < 0.05 was considered statistically significant. SCFA concentrations were analyzed using linear mixed-effects models, with participant-specific random intercepts used for paired PMO-pre and PMO-aft comparisons. PCA of metabolomic data was conducted using Euclidean distance following Pareto scaling. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Results

Participant Characteristics and Clinical Efficacy

A total of 41 postmenopausal women were included in the study, comprising 21 healthy controls (HC) and 20 patients with postmenopausal osteoporosis (PMO). All 20 PMO patients completed the six-month follow-up, and post-treatment samples (PMO-aft) were paired with their corresponding baseline samples (PMO-pre) from the same individuals. Baseline demographic and lifestyle characteristics, including age, years since menopause, BMI, carbohydrate intake, dietary fiber intake, daily calcium intake, and smoking and alcohol use, were comparable between the HC and PMO-pre groups (P > 0.05) (Table 1).

Table 1.

Baseline Characteristics of the Research Subjects and Changes in Bone Metabolism Indicators After Treatment

Parameter	HC (n=21)	PMO-Pre (n=20)	PMO-Aft (n=20)
Demographic characteristics
Age (years)	60.5 ± 10.0	54.0 ± 9.5	-
Years since menopause (years)	6.53±0.27	6.25 ± 0.46	-
BMI (kg/m2)	25.56 ± 2.84	25.87 ± 3.12	26.18±2.79
Dietary intake
Carbohydrates (g/d)	223.8±16.2	224.8±26.4	214.2±18.4
Dietary fibre (g/d)	18.9±4.4	17.0±3.0	17.7±3.1
Dietary calcium intake (mg/d)	835.00±2.10	828.00±1.90	832.90 ± 2.20
Smoking, n (%)
Never smoker	20 (95.2)	18 (90.0)	18 (90.0)
Current smoker	1 (4.8)	2 (10.0)	2 (10.0)
Drinking, n (%)
Never drinker	21 (100)	20 (100)	20 (100)
Bone mineral density
Lumbar spine BMD (T-score)	−0.17 ± 0.20	−2.98±0.15****	−2.65±0.16#
Femoral neck BMD (T-score)	−0.33 ± 0.19	−2.40 ± 0.09****	−2.12 ± 0.10#
Serum biochemistry
Calcium (mmol/L)	2.29 ± 0.10	2.30 ± 0.17	2.24 ± 0.09
Phosphorus (mmol/L)	1.22 ± 0.12	1.20 ± 0.14	1.17 ± 0.20
25-hydroxyvitamin D (ng/mL)	14.92 ± 7.96	17.80 ± 6.11	17.87 ± 7.02
Parathyroid hormone (ng/mL)	41.10 ± 22.70	33.15 ± 12.70	37.13 ± 8.90
Bone turnover markers
BALP (U/L)	184.04± 102.85	253.44 ± 71.76*	179.31±81.83#
TRACP-5b (U/L)	298.80 ± 55.07	430.40±151.68*	316.25±164.31#
PINP (ng/mL)	37.15 ± 24.65	69.10 ± 24.87*	47.30 ± 17.37#
β-CTX (ng/mL)	0.31 ± 0.49	0.47 ± 0.27	0.34 ± 0.15
Osteocalcin (ng/mL)	13.45 ± 8.26	19.82 ± 7.85*	13.71 ± 5.11#

Notes: Data are presented as mean ± SD. Statistical analysis: *P < 0.05, ****P < 0.0001 vs HC group; ^#P < 0.05 vs PMO-pre group. Demographic comparisons between HC and PMO-aft groups showed no significant differences in age, years since menopause, and BMI (P > 0.05), indicating comparability of baseline characteristics.

Abbreviations: HC, healthy control; PMO-pre, postmenopausal osteoporosis patients before treatment; PMO-aft, postmenopausal osteoporosis patients after 6-month treatment with alendronate plus calcitriol; BMI, body mass index; BMD, bone mineral density; BALP, bone alkaline phosphatase; TRACP-5b, tartrate-resistant acid phosphatase 5b; PINP, procollagen type I N-terminal propeptide; β-CTX, beta-isomerized C-terminal telopeptide of type I collagen.

At baseline, the PMO-pre group exhibited a high-turnover bone metabolism phenotype. Compared with healthy controls, patients with PMO had significantly lower BMD at both the lumbar spine and femoral neck (P < 0.05), accompanied by markedly elevated levels of bone turnover markers, including TRACP-5b, BALP, PINP, and osteocalcin (P < 0.05). Following six months of alendronate-calcitriol therapy, significant clinical improvements were observed. BMD at the lumbar spine and femoral neck increased significantly relative to baseline (P < 0.05), the mean change in lumbar spine BMD exceeded the LSC threshold of 0.03 g/cm², whereas the mean change in femoral neck BMD did not reach its LSC threshold of 0.05 g/cm², suggesting that the observed femoral neck increase may still be within measurement error, while all measured bone turnover markers decreased significantly (P < 0.05), consistent with suppression of bone turnover. Importantly, dietary fiber and calcium intake remained stable throughout the intervention period (P > 0.05), suggesting that subsequent microbiome changes were unlikely to be driven by dietary variation.

Global Alterations in Gut Microbiota Diversity

Alpha diversity analysis showed no statistically significant differences in Shannon or Chao1 indices among the three groups (P > 0.05) (Figure 1A and B). Nevertheless, a consistent trend was observed. The PMO-pre group displayed reduced microbial diversity compared with the HC group (Shannon: 7.149 ± 0.773 vs 6.660 ± 0.634; Chao1: 13,236.83 ± 2,556.36 vs 12,386.53 ± 2,543.30). After treatment, both indices in the PMO-aft group increased toward control levels (Shannon: 7.000 ± 0.975; Chao1: 12,575.09 ± 2,657.01), indicating partial restoration of microbial richness and diversity.

Figure 1.

Alpha and Beta Diversity Indices. (A and B) Intergroup comparison of intestinal microbiota alpha diversity (Chao1, Shannon). P values were calculated with the Kruskal–Wallis test, which showed no significant differences among groups. (C) Principal coordinate analysis (PCoA) of bacterial beta diversity based on Bray-Curtis distance. The first two principal coordinates explain 8.59% and 7.13% of the total variance, respectively. (D) ANOSIM based on Bray-Curtis distance. R > 0 indicates that between-group differences exceed within-group differences, and P > 0.05 indicates no statistically significant difference between groups.

Beta diversity analysis based on Bray-Curtis distances revealed partial separation of microbial community structure across groups, as visualized by PCoA (Figure 1C). Samples from the HC group clustered near the center of the ordination space, whereas PMO-pre samples were shifted along the PC1-PC2 axes. Samples from the PMO-aft group occupied an intermediate position, suggesting a treatment-associated shift toward the healthy microbiota profile.

The first two principal coordinates explained 8.59% (PCo1) and 7.13% (PCo2) of the total variance, respectively (cumulative 15.72%). However, overall differences in community structure did not reach statistical significance based on ANOSIM (R = 0.037, P = 0.072) (Figure 1D). The direction of microbiome changes following treatment showed partial convergence toward the healthy control profile, suggesting that the observed alterations may reflect both disease-associated recovery and potential pharmacologic effects of anti-osteoporosis therapy.

Taxonomic Composition and Key Microbial Biomarkers

At the phylum level, gut microbial communities across all groups were dominated by Firmicutes, Bacteroidota, and Pseudomonadota, together accounting for more than 90% of total relative abundance (Figure 2A and B). Compared with healthy controls, PMO patients at baseline exhibited a significant expansion of Pseudomonadota (FDR-adjusted P = 0.0474), indicating a state of microbial dysbiosis.

Figure 2.

Taxonomic composition and differential microbial taxa among HC, PMO-pre, and PMO-aft groups. (A) Stacked bar plot showing the relative abundance of dominant bacterial phyla across the HC, PMO-pre, and PMO-aft groups. (B) Box plot showing the relative abundance of Pseudomonadota among the three groups. (C and D) Bar plots showing significantly altered bacterial genera among groups. (E) Bar plot showing significantly altered bacterial species among groups. (F) LEfSe analysis identifying discriminative bacterial species among the HC, PMO-pre, and PMO-aft groups. The length of each bar represents the linear discriminant analysis (LDA) score. Red, blue, and green indicate taxa enriched in the HC, PMO-pre, and PMO-aft groups, respectively. HC, healthy control group; PMO-pre, postmenopausal osteoporosis group before treatment; PMO-aft, postmenopausal osteoporosis group after treatment with alendronate sodium combined with calcitriol. Data are presented as relative abundance. *FDR-adjusted P < 0.05 for PMO-pre vs HC; #FDR-adjusted P < 0.05 for PMO-aft vs PMO-pre.

At the genus level, Wilcoxon rank-sum tests with FDR correction identified a reduction in several SCFA-associated and potentially beneficial taxa in PMO-pre, including Anaerostipes, Limosilactobacillus, Bombilactobacillus, Petrotoga, Duganella, Oscillibacter, Faecalibacterium sp., and Anaerobutyricum hallii. In contrast, opportunistic or inflammation-associated genera, including Desulfovibrio, Cedecea, Kosakonia, and Parabacteroides, were significantly enriched in PMO-pre compared with healthy controls (Figure 2C).

Following six months of alendronate–calcitriol therapy, a marked shift in microbial composition was observed. SCFA-associated taxa, including Roseburia, Megamonas, Dialister, and unclassified Lactobacillaceae, were significantly increased, whereas multiple opportunistic taxa, including Enterobacter, Salmonella, Raoultella, Serratia, Copranaerobaculum, and Rahnella, were significantly reduced (Figure 2D and Table S3).

At the species level, the butyrate-producing bacterium Anaerobutyricum hallii and Faecalibacterium sp. were significantly depleted in PMO-pre compared with healthy controls and increased after treatment, indicating a directional shift toward normalization of key SCFA-producing members of the gut ecosystem (Figure 2E). In addition, species-level opportunistic pathogens, including Salmonella enterica and Enterobacter hormaechei, were reduced following treatment.

Longitudinal modeling using linear mixed-effects models (Table S3) confirmed that these microbial changes were consistent after adjustment for BMI, dietary fiber intake, carbohydrate intake, years since menopause, and serum 25-hydroxyvitamin D levels. In particular, taxa depleted in PMO-pre relative to healthy controls tended to increase after treatment, whereas opportunistic taxa showed the opposite pattern, consistent with a gradual shift toward the healthy control profile.

MaAsLin2 sensitivity analysis further validated the robustness of these findings under multivariable adjustment (Table S4), demonstrating consistent directional shifts in both SCFA-associated and opportunistic taxa. LEfSe analysis further identified Faecalibacterium prausnitzii, Agathobacter rectalis, and unclassified Blautia as characteristic taxa in the HC group. In contrast, Raoultella ornithinolytica and Lactobacillus kitasatonis were enriched in the PMO-pre and PMO-aft groups, respectively (Figure 2F).

Correlation of Microbiota with Bone Health Indicators

Spearman correlation analysis based on independent baseline observations from HC and PMO-pre participants revealed significant associations between gut microbial taxa and bone-related parameters (Figure 3A).

Figure 3.

Associations between gut microbiota, bone metabolism, and functional alterations of the gut microbiome. (A) Spearman correlation analysis was performed using independent baseline observations from healthy controls and PMO-pre participants. PMO-aft samples were excluded from this primary correlation analysis to avoid pseudo-replication caused by repeated observations from the same PMO patients. The heatmap shows Spearman correlation coefficients between differential microbial taxa and lumbar spine T-score, femoral neck T-score, PINP, β-CTX, BALP, and TRACP-5b. Red indicates positive correlations and blue indicates negative correlations. *FDR-adjusted P < 0.05. (B) Volcano plot showing differentially abundant CAZy enzyme families between the HC and PMO-pre groups. Red dots represent significantly upregulated features, and gray dots represent features without significant differences. Representative altered CAZy families, including PL13, are labeled. (C) Volcano plot showing differentially abundant CAZy enzyme families between the PMO-pre and PMO-aft groups. Red dots indicate significantly upregulated features, blue dots indicate significantly downregulated features, and gray dots indicate nonsignificant features. Representative altered CAZy families, including CBM17, GH90, and GT56, are labeled. (D) Differential KEGG pathway analysis among the HC, PMO-pre, and PMO-aft groups. The left panel shows the mean relative abundance of selected KEGG Level 3 pathways across groups, and the right panel shows differences in mean proportions with 95% confidence intervals for pairwise comparisons. HC, healthy control group; PMO-pre, postmenopausal osteoporosis group before treatment; PMO-aft, postmenopausal osteoporosis group after treatment with alendronate sodium combined with calcitriol. *FDR-adjusted P < 0.05.

PMO-pre-enriched or potentially opportunistic taxa, including Cedecea, Parabacteroides, Desulfovibrio, Copranaerobaculum, Pseudomonadota, and Kosakonia, were negatively correlated with lumbar spine and femoral neck T-scores and were positively correlated with bone turnover markers.

In contrast, taxa reduced in PMO-pre but enriched after treatment, including Dialister, Anaerostipes, Limosilactobacillus, Bombilactobacillus, Petrotoga, Duganella, Oscillibacter, Roseburia, Megamonas, Lactobacillaceae_unclassified, Faecalibacterium sp., and Anaerobutyricum hallii, showed the opposite pattern, with positive correlations with T-scores and negative correlations with bone turnover markers. Several of these associations remained significant after FDR correction, suggesting that baseline microbial dysbiosis was associated with reduced bone mass and increased bone turnover. Complete correlation results are provided in Table S5.

Functional Potential of the Gut Microbiome

To investigate the functional consequences of postmenopausal osteoporosis and its treatment on the gut microbiome, KEGG pathway enrichment analysis at Level 3 was performed using the covariate-adjusted linear regression (Figure 3B). Compared with the HC group, the PMO-pre group exhibited significant enrichment of pathways associated with cell motility and lipoarabinomannan (LAM) biosynthesis. These pathways showed directional shifts toward control levels following treatment with alendronate sodium combined with calcitriol. In contrast, several pathways involved in energy metabolism and SCFA precursor metabolism, including motor proteins, propanoate metabolism, pyruvate metabolism, and taurine and hypotaurine metabolism, were significantly depleted in the PMO-pre group and showed directional increases toward control levels after treatment. Similarly, pathways related to the citrate cycle (TCA cycle), ether lipid metabolism, other glycan degradation, and riboflavin metabolism demonstrated a consistent “decrease- shift toward control levels” trend across the three groups.

Functional annotation based on the CAZy database further revealed disease-associated alterations in carbohydrate metabolic capacity. In the PMO-pre group, the polysaccharide lyase families PL13 were significantly enriched compared with the HC group (Figure 3C), suggesting an increased capacity for degradation of host-derived mucus glycans and plant polysaccharides during osteoporosis progression. Following treatment, the cellulose-binding module family CBM17 was significantly enriched in the PMO-aft group (Figure 3D). In contrast, glycoside hydrolase GH90 and glycosyltransferase GT56 were significantly reduced after treatment, indicating decreased potential for pro-inflammatory surface polysaccharide biosynthesis and specific virus-associated enzymatic functions. Detailed results are provided in Table S6.

Targeted Metabolomic Profiling of SCFAs

Differences in fecal SCFA profiles among groups were evaluated using PCA based on Euclidean distance. PC1 and PC2 accounted for 58.3% and 19.0% of the total variance, respectively (Figure 4A). The HC and PMO-pre groups showed partial separation in multivariate space, whereas the PMO-aft group occupied an intermediate position and clustered more closely with the PMO-pre group. However, PERMANOVA analysis demonstrated that the overall SCFA composition did not differ significantly among the three groups (R² = 0.026, P = 0.528).

Figure 4.

Short-chain fatty acid profiles and their associations with gut microbiota and bone metabolism among HC, PMO-pre, and PMO-aft groups. (A) Principal component analysis (PCA) of fecal short-chain fatty acid (SCFA) profiles among the HC, PMO-pre, and PMO-aft groups. Each point represents one sample, and the ellipse indicates the 95% confidence interval. PC1 and PC2 explained 58.3% and 19.0% of the total variance, respectively. (B) Spearman correlation heatmap showing the correlation of baseline data of differential microbial taxa and major short-chain fatty acids in HC and PMO-pre participants. PMO-aft samples were excluded to avoid pseudo-replication from repeated observations of the same PMO patients. (C) Bar plots showing the concentrations of individual SCFAs across the three groups. Data are presented as mean ± SEM. (D) Spearman correlation heatmap showing associations between major SCFAs and bone-related parameters, including bone mineral density and bone turnover markers. PMO-aft samples were excluded to avoid pseudo-replication from repeated observations of the same PMO patients. Red indicates positive correlations, whereas blue indicates negative correlations. Asterisks indicate statistically significant correlations. HC, healthy control group; PMO-pre, postmenopausal osteoporosis group before treatment; PMO-aft, postmenopausal osteoporosis group after treatment with alendronate sodium combined with calcitriol. *FDR-adjusted P < 0.05, **FDR-adjusted P < 0.01 for PMO-pre vs HC; #FDR-adjusted P < 0.05 for PMO-aft vs PMO-pre.

Given the marked alterations observed in pyruvate and propanoate metabolism pathways, individual SCFAs were further quantified and analyzed using linear mixed-effects models with covariate-adjusted pairwise comparisons. After adjustment for relevant clinical and dietary covariates, no significant differences in SCFA concentrations were observed between the HC and PMO-pre groups or between the HC and PMO-aft groups (Table S7).

In the paired comparison within PMO patients, propionic acid was the only SCFA that showed a significant treatment-associated change, with concentrations increasing significantly after six months of therapy compared with baseline levels (PMO-aft vs PMO-pre: estimate = 0.214, SE = 0.076, t = 2.808, FDR-adjusted P = 0.021; Figure 4C and Table S7). In contrast, acetic acid, butyric acid, valeric acid, isobutyric acid, isovaleric acid, hexanoic acid, heptanoic acid, octanoic acid, nonanoic acid, and decanoic acid showed no significant treatment-related changes.

Notably, the increase in propionic acid was consistent with the enrichment of propanoate metabolism pathways identified in the functional prediction analysis, suggesting a potential treatment-associated enhancement of propionate-related microbial metabolism. Correlation analyses further revealed significant associations between major SCFAs, differential microbial taxa, and bone-related parameters (Figure 4B and D), supporting a potential link between gut microbial metabolism and skeletal health.

Taken together, these findings indicate that the metabolic response to treatment was characterized by a selective increase in propionic acid rather than a broad restoration of the overall fecal SCFA profile.

Inter-Relationships Among the Gut Microbiota, SCFAs, and Bone Health

Spearman correlation analysis with FDR correction was performed to investigate the associations between differentially abundant bacterial genera and four major SCFAs, including butyric, acetic, propionic, and valeric acid (Figure 4B and Table S8). After correction by the FDR, although the results showed that there was no statistically significant correlation between different bacterial genera and the four major short-chain fatty acids, it was found that Dialister was positively correlated with Acetic acid; Limosilactobacillus, Parabacteroides, Faecalibacterium_sp was positively correlated with Butyric acid; Limosilactobacillus and Bombilactobacillus were positively correlated with Propionic acid, Desulfovibrio and Copranaerobaculum were negatively correlated with Propionic acid; Desulfovibrio, Oscillibacter, Faecalibacterium_sp. was positively correlated with Valeric acid.

Associations between SCFAs and clinical bone parameters were further evaluated (Figure 4D and Table S9). After FDR correction, propionic acid was significantly positively correlated with LS-BMD and FN-BMD (r = 0.499, FDR-adjusted P = 0.007; r = 0.410, FDR-adjusted P = 0.046), and was significantly negatively correlated with PINP, β-CTX, and TRACP-5b (r = −0.562, FDR-adjusted P = 0.002; r = −0.569, FDR-adjusted P = 0.002; r = −0.559, FDR-adjusted P = 0.002), while acetic acid, butyric acid and valeric acid had weak correlations with most bone-related indicators and were not statistically significant. It indicates that propionic acid is closely related to higher bone mineral density of the lumbar vertebrae and femoral neck as well as lower bone turnover levels.

Multi-Omics Integration via DIABLO

To comprehensively characterize the interactions between the gut microbiota and SCFA metabolism, supervised multi-omics integration was performed using the DIABLO framework. Sample projection plots showed partial separation among groups in the microbiota block, with PMO-aft samples distributed between the HC and PMO-pre clusters, indicating partial restoration of microbial community structure following treatment (Figure 5A). In the SCFA module, group discrimination was less pronounced; however, the PMO-aft group exhibited a directional shift toward the HC distribution, suggesting a gradual recovery of metabolic profiles after intervention.

Figure 5.

Multi-omics integration analysis of gut microbiota and short-chain fatty acids using DIABLO. (A) Sample projection plots derived from DIABLO analysis showing the distribution of HC, PMO-pre, and PMO-aft groups based on the microbiota block and SCFA block. Each point represents one sample, and ellipses indicate the distribution range of each group. (B) Correlation circle plot showing key variables contributing to group discrimination in the integrated microbiota-SCFA model. Microbial genera are shown in blue, and SCFAs are shown in red. Variables located farther from the center contribute more strongly to group separation, while variables with similar directions indicate positive associations. (C) Correlation network showing associations between key gut microbial genera and SCFAs. Blue nodes represent bacterial genera, and green nodes represent SCFAs. Edges indicate correlations between microbial taxa and SCFAs, with edge color reflecting the direction and strength of correlation. Correlation network analysis was performed using a correlation coefficient threshold of |r| ≥ 0.5. HC, healthy control group; PMO-pre, postmenopausal osteoporosis group before treatment; PMO-aft, postmenopausal osteoporosis group after treatment with alendronate sodium combined with calcitriol. (D) Internal validation of the DIABLO multi-omics model. Repeated 5-fold cross-validation with 50 repeats was performed to evaluate classification performance across one and two components. Classification error rates were calculated using maximum distance, centroid distance, and Mahalanobis distance. Lower error rates indicate better classification performance.

The correlation circle plot identified several variables contributing to group discrimination, including major SCFAs and bacterial genera such as Oscillibacter, Dialister, Anaerostipes, and Desulfovibrio (Figure 5B). These variables displayed coordinated loading patterns, suggesting that shifts in microbial composition were associated with changes in fecal SCFA profiles. Correlation network analysis further revealed co-occurrence patterns between selected bacterial genera and SCFAs, including propionic, butyric, and valeric acids (Figure 5C). These associations support a potential link between microbial community remodeling and SCFA-related metabolic changes during PMO treatment.

Repeated 5-fold cross-validation showed that the microbiota block provided moderate internal discrimination among HC, PMO-pre, and PMO-aft groups (Figure 5D and Table S10). Classification error rates decreased from 0.462–0.476 with one component to 0.303–0.328 with two components, with the lowest error rate observed using maximum distance classification. In contrast, the SCFA block showed higher error rates of 0.633–0.647, indicating limited stand-alone discriminative performance. Therefore, the DIABLO results were interpreted as exploratory and mainly used to identify coordinated microbiota–SCFA patterns rather than to establish a definitive predictive classifier.

Discussion

PMO is characterized by estrogen deficiency-driven acceleration of bone turnover and progressive deterioration of bone microarchitecture.^1,2 In recent years, increasing attention has been directed toward the gut-bone axis, in which the gut microbiota and its metabolites contribute to the regulation of bone metabolism.^8,9 For instance, SCFAs may contribute to PMO-related bone metabolism through pathways such as regulating immunity, activating G protein-coupled receptors, and promoting intestinal calcium absorption.¹⁷ Bile acids intervene in bone metabolism by activating key pathways such as farnesoid X receptor (FXR) and G protein-coupled bile acid receptor 1 (TGR5).²⁹ Tryptophan metabolites derived from the gut microbiota regulate bone metabolic balance through the AhR signaling pathway.³⁰ Among them, SCFAs are among the most widely studied microbial metabolites of intestinal microbiota. In the present study, six months of treatment with alendronate sodium combined with calcitriol significantly improved bone mineral density and reduced biochemical markers of bone turnover. These clinical improvements are primarily consistent with the established pharmacological actions of alendronate in inhibiting osteoclast-mediated bone resorption and calcitriol in regulating calcium–phosphate and vitamin D-dependent bone metabolism. The accompanying partial restoration of gut microbial composition and SCFA-related metabolic functions suggests that gut microecological changes may occur alongside therapeutic improvement and may potentially complement the direct skeletal effects of alendronate–calcitriol therapy.^24,25

At the community level, PMO was not associated with marked alterations in overall alpha or beta diversity, which is consistent with previous studies indicating that osteoporosis is more closely associated with shifts in specific functional taxa than with large-scale disruption of the entire microbial ecosystem.^3,31 Instead, PMO patients showed enrichment of Pseudomonadota and several opportunistic taxa belonging to the Enterobacteriaceae family. Expansion of Pseudomonadota is widely regarded as a hallmark of gut dysbiosis and has been associated with impaired intestinal barrier integrity and chronic low-grade inflammation.^12,32 This result is consistent with the research findings of Xu Q et al, who proposed that gut microbiota may participate in PMO through immune-inflammatory regulation and intestinal permeability.³³ Meanwhile, we observed depletion of several SCFA-producing genera, including Roseburia, Faecalibacterium, Anaerostipes, and Oscillibacter. Collectively, these findings suggest that PMO is characterized by a transition toward a pro-inflammatory, low-SCFA microbial environment rather than by complete loss of microbial diversity.^34,35 Following treatment, this pattern was partially reversed, with recovery of beneficial commensal taxa and reduction of opportunistic pathogens, despite relatively stable global diversity indices.

The observed taxonomic alterations were further supported by functional changes in microbial metabolic pathways. Pathways associated with energy metabolism and SCFA precursor metabolism, including pyruvate metabolism and propanoate metabolism, were reduced in PMO and partially restored after treatment. Correlation analyses further demonstrated that SCFA-producing bacteria were positively associated with BMD and negatively associated with bone turnover markers, whereas Enterobacteriaceae-related taxa displayed the opposite trend. In particular, Anaerostipes, Roseburia, and Faecalibacterium were associated with more favorable bone parameters, consistent with their established roles in butyrate production, maintenance of intestinal barrier function, and regulation of host immune responses.^17,23 By contrast, Pseudomonadota showed a significant negative association with lumbar spine BMD, further supporting its potential correlation to systemic inflammation and bone loss.^12,35

At the metabolite level, the overall SCFA profile remained relatively stable among groups. Although propionic acid was significantly increased after treatment in the paired PMO comparison, no significant differences were observed between HC and PMO groups after covariate adjustment. Although global differences in SCFA composition did not reach statistical significance, several biologically relevant associations were identified. Butyrate and propionate were positively correlated with SCFA-producing bacteria and inversely associated with markers of bone resorption, particularly TRACP-5b and β-CTX.^18,34 These findings suggest that microbial metabolic activity may respond more slowly than microbial composition during treatment. Nevertheless, the coordinated relationships among SCFA-producing taxa, SCFAs, and bone turnover markers support the presence of functionally relevant microecological remodeling during PMO therapy.

Multi-omics integration analysis further demonstrated coordinated interactions between the gut microbiota and SCFA metabolism in distinguishing disease and treatment states. Key bacterial genera, including Oscillibacter, Anaerostipes, and Dialister, clustered closely with major SCFAs such as acetate, propionate, and butyrate, highlighting their functional interdependence. The research by Seely et al indicates that the gut microbiota may influence osteoporosis and osteogenesis through interactions with osteoclastogenesis, osteoblast activity, RANKL-related signaling, immune regulation, and microbial metabolites.³⁶ Among these taxa, Oscillibacter has been reported to produce valeric acid, a metabolite capable of suppressing osteoclast differentiation and promoting osteogenesis through inhibition of the pro-inflammatory NF-κB signaling pathway.^37,38 In contrast, opportunistic pathogens such as Desulfovibrio displayed opposing associations, further supporting their potential involvement in disruption of the gut-bone axis. Importantly, treatment shifted both microbial and metabolic profiles toward those observed in healthy controls, although recovery at the metabolite level appeared less pronounced than changes in microbial composition. This delayed metabolic recovery may reflect the complex determinants of fecal SCFA concentrations. In addition to the abundance of SCFA-producing bacteria, SCFA levels are influenced by dietary substrate availability, microbial cross-feeding interactions, intestinal absorption, and host metabolic utilization. As a result, restoration of microbial community structure may precede measurable changes in fecal SCFA concentrations during a relatively short intervention period.

Functional metagenomic analyses further supported these observations. A systematic review and meta-analysis based on 16S rRNA gene sequencing by Huang et al reported that osteoporosis is associated with altered gut microbial composition, although the direction of specific taxonomic changes may vary across populations and geographic regions.³⁹ Notably, reduced abundance of several commensal taxa, including Anaerostipes and Megamonas, was reported in Chinese studies included in that meta-analysis, which partially overlaps with the disease-associated microbial pattern observed in our cohort. In the present study, shotgun metagenomic analysis further suggested that these compositional changes were accompanied by alterations in microbial functional potential. Following treatment, pathways associated with mucosal degradation and lipopolysaccharide-related biosynthesis were reduced, whereas functions related to fiber utilization and energy metabolism showed treatment-associated shifts toward levels observed in healthy controls. These microbial changes occurred alongside improvements in bone metabolism; however, this study design cannot establish whether microbiota alterations contribute to therapeutic bone effects, as the two may be independent outcomes of pharmacologic treatment. Although alendronate is primarily recognized as an inhibitor of osteoclast-mediated bone resorption, accumulating evidence indicates that it can also induce shifts in gut microbial composition and promote the growth of beneficial bacterial taxa.^40,41 Similarly, calcitriol has been shown to enhance the abundance of SCFA-producing bacteria and strengthen microbial interaction networks within the gut ecosystem.^42,43

Taken together, our findings demonstrate that PMO is associated with reduced SCFA-producing bacterial taxa and enrichment of opportunistic bacteria, alongside impaired bone metabolism and increased bone turnover. Treatment with alendronate sodium combined with calcitriol improved bone mass and suppressed excessive bone resorption, and was accompanied by concurrent shifts in gut microbial composition and SCFA-related metabolic function toward healthy control levels. This observational study cannot confirm that modulation of the gut-bone axis contributes to the therapeutic effects of the regimen; this hypothesis requires further testing in controlled interventional and mechanistic studies.

Limitations

Several limitations of this study should be acknowledged. First, no untreated parallel PMO control group was included in this before–after observational design, so observed microbial and metabolic shifts cannot be fully disentangled from temporal variation, regression to the mean, or other unmeasured confounding factors over the six-month follow-up, which restricts causal interpretation of treatment effects. In addition, because PMO patients received alendronate, calcitriol, and calcium supplementation as part of routine clinical care, the observed microbiome and SCFA changes cannot be attributed to any single component of the treatment regimen. Second, although the actual sample size was sufficient to detect moderate-to-large paired longitudinal effects, the relatively small sample size (21 healthy controls and 20 paired PMO samples) limited the detection of small effects in high-dimensional metagenomic, functional, SCFA, and multi-omics analyses involving multiple comparisons. Third, the identified microbial biomarkers lack independent external validation in separate cohorts, and no mechanistic experiments were performed to establish causal links between microbiota changes and bone outcomes. Fourth, although multiple SCFA-producing taxa increased significantly after treatment, only propionic acid showed a statistically significant elevation in feces. This discrepancy likely arises because fecal SCFA concentrations reflect the net balance of microbial production, cross-feeding utilization, intestinal epithelial absorption, host metabolism, and colonic transit, rather than microbial production capacity alone. Butyrate and other SCFAs are rapidly metabolized by colonocytes, so production changes may not be fully captured in fecal levels; the limited sample size and six-month intervention window may also have reduced power to detect subtler shifts in other SCFA species. Finally, the six-month intervention, while sufficient to improve BMD and bone turnover markers, may be too short to drive comprehensive remodeling of the gut microbial community and global SCFA profile. Future large-scale, multicenter randomized controlled trials coupled with mechanistic investigations are needed to clarify the causal role of the gut-bone axis in PMO.

Conclusions

This study demonstrates that six months of combined alendronate sodium and calcitriol therapy is associated with partial treatment-associated shifts in gut microbial composition and functional potential in patients with postmenopausal osteoporosis. Treatment selectively enriched several SCFA-producing taxa, including Roseburia, Megamonas, Dialister, and unclassified members of the Lactobacillaceae family, while reducing the abundance of opportunistic pathogens within the Enterobacteriaceae family, such as Enterobacter, Salmonella, and Raoultella. Functional analyses further revealed a characteristic “decrease-shift toward control levels” pattern in pathways related to propanoate metabolism, the TCA cycle, and taurine metabolism. In addition, multi-omics integration identified coordinated microbiota-metabolite interaction modules associated with the therapeutic state. Collectively, these findings provide preliminary evidence of microecological changes accompanying conventional anti-osteoporotic therapy, and support the potential value of microbiota-targeted strategies as potential adjunctive interventions for PMO.

Funding Statement

This study was funded by the Karamay Central Hospital, “High-throughput sequencing was conducted to study the effect of alendronate sodium combined with calcitriol on the intestinal flora structure of postmenopausal osteoporosis patients (No. YXH202423)” and the Xinjiang Uygur Autonomous Region Pharmaceutical Society, “A study on the mechanism of alendronate sodium combined with calcitriol in improving postmenopausal osteoporosis based on intestinal flora” (No. 20240105); Tianshan Talents’ Program 2023TSYCQNTJ0029 (this fund belongs to the corresponding author Shengzhao Zhang).

Data Sharing Statement

The data generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Ethics Approval and Consent to Participate

This study was approved by the Ethics Committee of the Karamay Central Hospital (grant number: YL-2024-005, date of approval: April 22, 2024) in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants included in this study.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

Mrs Fang Shen reports Support for the manuscript from Karamay Central Hospital, Xinjiang Uygur Autonomous Region Pharmaceutical Society, during the conduct of the study; Grants or contracts from Xinjiang Uygur Autonomous Region Pharmaceutical Society, Karamay Central Hospital, outside the submitted work. Mr Shengzhao Zhang reports Support for the manuscript from Xinjiang Uygur Autonomous Region Pharmaceutical Society, Karamay Central Hospital, during the conduct of the study. The authors declare that there is no conflict of interest.