Bone remodeling is a continuous physiologic process that preserves skeletal mechanical competence and calcium homeostasis; disturbance of this coupled turnover precipitated metabolic bone disease (Kim and Koh, 2019). Osteoporosis is defined by a reduction in bone mineral density (BMD) and increased skeletal fragility, thereby elevating fracture risk (Aslan, 2022). According to the World Health Organization guidelines, a dual-energy X-ray absorptiometry (DXA) T-score ≥−1.0 defines normal BMD; −1.0 to −2.5 indicates osteopenia; <−2.5 defines osteoporosis; and <−2.5 with a fragility fracture constitutes severe osteoporosis (Al-Hourani et al., 2021).

Bone turnover markers are categorized into two major groups. The initial category encompasses key bone formation markers, including serum alkaline phosphatase, osteocalcin, bone-specific alkaline phosphatase, procollagen type I N-terminal propeptide, procollagen type I C-terminal propeptide, and bone sialoprotein. Urinary hydroxyproline, urinary and serum type I collagen cross-linked N-telopeptide, serum type I collagen cross-linked C-telopeptide, and tartrate-resistant acid phosphatase (TRAP) fall into the second category of bone resorption markers, among others. These fracture-related biomarkers are valuable for diagnosing postmenopausal osteoporosis and evaluating therapeutic efficacy (Kanis et al., 2019; Vasikaran et al., 2011).

Data from a 2018 survey by China’s National Health Commission showed that the prevalence of osteopenia among women aged 40–49 years was 31.4%, increasing to 67.6% in women aged ≥50 years. In Japan, approximately 60% of postmenopausal women are classified as high-risk for osteoporosis (Iki, 2012). Similarly, a Korean study reported that osteoporosis prevalence among postmenopausal women reached 35–40%, approaching 50% in women aged ≥60 years (Lee et al., 2013). In India, prevalence rates vary widely, ranging from 8 to 62% (Agrawal and Garg, 2023). By 2050, it is projected that over 50% of global osteoporotic fractures will occur in Asia, emphasizing the urgent need for region-specific prevention strategies (Vijayakumar et al., 2025).

Currently, non-pharmacological interventions for maintaining BMD primarily involve calcium and vitamin D supplementation. However, accumulating evidence suggests that the gut microbiota is implicated in the regulation of bone metabolism through the gut-bone link (Hernandez et al., 2016; Bedree et al., 2023). Therefore, investigating the relationships between postmenopausal bone loss (T-score <−1.0), gut microbiota, serum biochemical markers, and serum metabolites offers remarkable potential for developing multi-pathway prevention and treatment strategies (Song et al., 2022). Nevertheless, the specific characteristics of gut microbiota and serum metabolites in postmenopausal females with osteopenia, and their correlations with bone quality, is poorly understood.

This study aimed to establish a scientific foundation for developing targeted prevention and treatment strategies for postmenopausal bone loss. Compared with prior studies that typically examined either gut microbiota or circulating metabolites alone, our study integrates 16S rRNA profiling and untargeted serum metabolomics in the same cohort and links multi-omic features to site-specific bone outcomes (L1–L4 T-scores). This integrative approach highlights a coherent association pattern involving NAA and the taxa Prevotella/Dorea, providing new clues to microbiota-metabolite-bone co-variation in postmenopausal bone loss.

The key new insight from this work is the integrative multi-omics correlation map linking gut taxa and serum metabolites with lumbar bone status. In particular, NAA showed a positive association with L1–L4 T-scores and co-varied with Prevotella and Dorea, suggesting a new potential microbiota–metabolite model of bone metabolic regulation that warrants further mechanistic validation.

Decreased estrogen levels in postmenopausal women disrupted bone remodeling, leading to an imbalance in which bone resorptive processes exceed osteogenic activity and induces persistent bone loss. In this study, the “normal group” and “bone loss group” represent different stages of this process. The lower T-scores observed in the bone loss group reflected the cumulative effect of long-term remodeling imbalance, indicating that bone mass has entered the osteopenia or osteoporosis range. Some researchers have argued that osteoporosis should not be strictly defined as a disease since rigid diagnostic thresholds may prevent treatment for individuals who do not meet the criteria but are at high fracture risk. Instead, bone loss can be regarded as a common process affecting most individuals from around 50–60 years of age onward (Reid and Billington, 2022). Therefore, the osteopenia and osteoporosis groups were unified as a single bone loss group, reflecting a continuous and progressive pathological process. By comparing differences in gut microbiota, bone turnover biomarkers, blood biochemical indices, and serum metabolomic profiles between the osteopenia and normal groups—and analyzing their correlations with bone mass—this study identified potential diagnostic, preventive, and therapeutic strategies for postmenopausal bone loss.

Compared with the normal group, the bone loss group exhibited a markedly longer menopause duration, independent of age. Estrogen deficiency following prolonged menopause accelerates bone breakdown and inhibits osteogenesis, ultimately culminating in bone depletion (Melis et al., 2025). Bone status assessments revealed that the T-scores and BMD values for the total left femur, neck of the right femur, and overall right femoral region proved significantly reduced among the bone loss group than in the normal bone mass group, indicating that bone loss substantially reduces BMD. A significant difference in Z-score was only found at the L1–L4 vertebrae; all other sites showed no significant variations. This may be because the Z-score reflects bone density relative to individuals of the same age, sex, and ethnicity; no considerable discrepancies were noted due to the sample in this study being highly homogeneous (Muhan et al., 2022).

Blood biochemical analysis revealed that 25(OH)D and 25(OH)D3 concentrations were higher in the bone loss group than in the normal group; however, absolute levels in both groups remained below commonly used reference ranges for sufficiency. Correlation analysis further showed significant inverse associations between circulating 25(OH)D/25(OH)D3 and the L1–L4 T-score (25(OH)D3: r = −0.23, p = 0.016; 25(OH)D: r = −0.24, p = 0.011). These cross-sectional findings may reflect altered vitamin D metabolism, supplementation or health-seeking behaviors, and/or reduced receptor sensitivity (e.g., lower VDR expression), rather than a direct pathogenic effect of vitamin D (Anderson et al., 2017). In principle, activation of the PTH axis can increase renal 1α-hydroxylase activity and promote the conversion of 25(OH)D to 1,25(OH)2D3, thereby enhancing intestinal calcium absorption; however, in our cohort PTH levels were similar between groups. Importantly, because the study is cross-sectional and vitamin D concentrations were subnormal overall, causal interpretation is not warranted. Longitudinal investigations are required to clarify the relationships among vitamin D metabolism, receptor sensitivity, and bone loss rates.

Overall, the serum metabolomic profile of women with normal bone status differed from that of the bone-loss group, characterized by coordinated changes across multiple metabolite classes relevant to bone remodeling. Specifically, flavonoids such as taxifolin and isoliquiritigenin were relatively higher in the normal group, a pattern that is directionally consistent with prior experimental evidence linking these compounds to reduced osteoclastogenic signaling and bone-resorptive activity (Zhang et al., 2019; Liu et al., 2016). In addition, several amino acids and related metabolites—including L-arginine and the polyamine spermidine—as well as aromatic amino acid derivatives, were more abundant in the normal group; these observations are broadly consistent with reported links between amino acid metabolic axes and osteoblast–osteoclast balance, although such mechanistic interpretations remain indirect in an observational setting (Riancho et al., 1995; Lee et al., 2017; Izawa et al., 2016; Voronov et al., 2008). Phospholipid alterations were also evident: phosphatidylethanolamine (PE 18:0_18:1) was higher in the normal group, whereas multiple LPC species were lower. Given that membrane phospholipid composition may influence cellular functions relevant to mineralization and signaling, these findings may reflect a metabolic milieu more favorable to bone homeostasis (Fisk and Kano-Sueoka, 1992; Agrawal et al., 2011).

Among the differential metabolites, NAA emerged as a key correlate. NAA was higher in the normal group and positively associated with the L1–L4 T-score, while also co-varying with the relative abundances of Prevotella and Dorea. Notably, although both genera were associated with NAA, their relative abundances were not significantly correlated, suggesting that they might contribute to the metabolite pool through independent or complementary pathways rather than synchronized expansion. Prior evidence suggests that NAA may act as an endogenous ligand involved in aryl hydrocarbon receptor-related pathways that are relevant to osteoclast/osteoblast regulation (Park et al., 2020). Taken together, these results highlight a microbiota-metabolite signature linked to lumbar bone status in our cohort and provide a testable hypothesis for microbiota–metabolite–bone crosstalk. However, given the cross-sectional design and genus-level microbiota resolution, the directionality and mechanistic basis of these associations require confirmation in longitudinal cohorts and functional studies.

Pathway enrichment analysis suggested several nominally enriched pathways (e.g., arginine/proline metabolism and aromatic amino acid-related pathways), but these signals did not remain significant after FDR correction (all q > 0.1). Therefore, pathway-level interpretations should be regarded as exploratory and primarily hypothesis-generating, and the lack of FDR significance likely reflects limited statistical power in a moderate-sized cohort. Notably, GKK (Gly-Lys-Lys), a small peptide detected in the untargeted metabolomic analysis, showed one of the stronger contributions to group discrimination based on fold-change patterns; however, its biological relevance to bone metabolism remains unclear at present. Thus, GKK is more appropriately interpreted as a marker of altered peptide/amino acid metabolism in the bone-loss state, warranting further validation in larger cohorts and targeted functional studies.

The gut microbiota-bone link is known to regulate bone metabolism (Tu et al., 2023; Lyu et al., 2023). Our analysis revealed no significant disparities in α- and β-diversity across the groups, genus-level analysis revealed higher Prevotella abundance in the normal group. As Prevotella may protect against estrogen-deficiency–induced bone loss (Wang et al., 2021; Zhang et al., 2023) by influencing carbon, branched-chain amino acids, as well as bone metabolism through the citric acid cycle (Lu et al., 2024). Additionally, the normal group exhibited higher Dorea abundance than that of the bone loss group, whereas Limosilactobacillus, significantly enriched in the bone loss group, exhibited an inverse association with the L1–L4 T-score (r = −0.27). Although Prevotella (r = 0.183, p = 0.065) and Dorea (r = 0.188, p = 0.058) tended to correlate positively with the L1–L4 T-score, these associations were not statistically significant and therefore warrant cautious interpretation. Studies of spaceflight and simulated microgravity have reported increased Dorea and Limosilactobacillus abundance in rodent intestines, accompanied by shifts in circulating metabolites relevant to bone remodeling (e.g., short-chain fatty acids) and evidence of trabecular bone loss and increased bone marrow adipogenesis (Bedree et al., 2023). However, other studies found that certain Lactobacillus spp. have a bone-improving effect. Lactobacillus reuteri administration increased femoral and lumbar trabeculae bone volume fraction and BMD in male mice, and prevented bone loss in ovariectomized mice (Villa et al., 2017). This contradiction may stem from the functional heterogeneity of bacterial species/strains and environmental influences. Since this study analyzed only genus-level data, metagenomic sequencing is required to clarify the species-grade specificity.

We detected a significant positive correlation between Olsenella (enriched in the bone loss group) and Limosilactobacillus. Both genera showed negative correlations with the L1–L4 T-score, suggesting the possible symbiotic regulation of bone metabolism. The negative correlation between Staphylococcus and the L1–L4 T-score may reflect an opportunistic expansion of inflammation-associated taxa, potentially contributing to a pro-inflammatory milieu. Staphylococcus may indirectly influence bone metabolism through inflammation resulting from gut microbiota alterations (Waldbaum et al., 2023). Overall, because genus-level differences did not remain significant after BH correction (all q > 0.05), the taxa discussed above should be regarded as exploratory candidates.

The differential serum metabolite NAA was significantly positively correlated with the L1–L4 T-score for BMD and protective genera Prevotella, Dorea, Defluviitaleaceae_UCG-011, and N-acetyl-L-phenylalanine (Figure 4). Therefore, gut microbiota may be involved in NAA production and function via a synergistic metabolic network, influencing bone homeostasis (Figure 5). Mechanistically, Prevotella degraded dietary fiber (e.g., arabinoxylan) to produce short-chain fatty acids such as propionate (Bartsch et al., 2025). Propionate provided energy for host colonic cells and may optimize the growth environment for Dorea by reducing intestinal pH (Brinck et al., 2025); however, in our cohort Prevotella and Dorea were not significantly correlated in relative abundance (r = 0.015, p = 0.880), This lack of direct correlation suggests that their contributions to the NAA-associated bone protection likely reflect functional complementarity within the gut-bone axis, rather than a synchronized expansion of these two genera. Certain fungi expressed anthranilate synthase, a precursor of NAA, which converts tryptophan to anthranilic acid (Zhao et al., 2025). Genes encoding enzymes involved in aromatic amino acid metabolism (including phenylalanine deamination) have been reported in several gut bacterial taxa, such as Enterococcus and Clostridium (Shen et al., 2022; Alili et al., 2022). However, it is also possible that Dorea harbors similar functional capacity. This possibility cannot be confirmed from genus-level 16S data alone and should be validated using shotgun metagenomics and/or isolate-based functional assays. Although the specific metabolic pathways of Defluviitaleaceae_UCG-011 in the context of bone health are not yet fully elucidated, its consistent co-variation with bone phenotypes and protective metabolites suggests that it may be a key member of the microbial network contributing to the NAA metabolic hub. As an acetylated tryptophan derivative (Paul and Ratledge, 1971), NAA participates in tryptophan metabolism, thereby potentially modulating osteochondrocyte activity through multiple signaling pathways, including the receptor mediation, intercellular coupling, gut-bone link crosstalk (kynurenine), serotonin (5-HT), and indole pathways (Xiang et al., 2025). Cross-regulation between phenylalanine and tryptophan metabolism may represent a core mechanism linking NAA to bone homeostasis. The positive correlation between NAA and N-acetyl-L-phenylalanine suggests that phenylalanine and tryptophan metabolism may form a synergistic regulatory network by sharing intermediate products, such as anthranilic acid (Figure 5), by working together to maintain the dynamic equilibrium between osteogenesis and resorption.

In this cross-sectional cohort, postmenopausal bone loss was associated with specific gut microbial and serum metabolic signatures. The co-variation of NAA with Prevotella/Dorea and L1–L4 T-scores provides a testable hypothesis and potential candidate biomarkers for future longitudinal and interventional studies.