The strain LC86 was provided by Wecare Probiotics Co., Ltd. (Suzhou, China). The strain was cultured at 37°C in De Man, Rogosa, and Sharpe (MRS) medium for 18 hours (20). Bacterial cells were collected by centrifugation at 6000 × g for 8 min and resuspended in sterile water to achieve a final concentration of 1×10⁹ CFU/mL.

Alendronate sodium, used as the positive control for osteoporosis treatment, was obtained in white tablet form (Batch No. X002979, Savio Industrial S.r.l.) and dissolved in ultrapure water. Chondroitin sulfate, used as the positive control drug for cartilage injury protection, was obtained in white powder form (Batch No. I2003201, Shanghai Aladdin Biochemical Technology Co., Ltd.) and dissolved in standard dilution water.

Key equipment used in this study included a dissection microscope (OLYMPUS SZX7), fluorescence microscope (Nikon AZ100), CCD camera, precision electronic balance (OHAUS CP214), ultrasonic cleaner, automatic sample grinder, automatic nucleic acid extractor (Auto-Pure32A), conventional PCR thermal cycler (BIO-RAD T100), real-time quantitative PCR instrument (BIO-RAD CFX Connect), high-speed refrigerated centrifuge, UV-Vis spectrophotometer (NanoDrop 2000), and standard 6-well plates. Major reagents included dexamethasone (DXMS), dimethyl sulfoxide (DMSO), methylcellulose, alizarin red, Universal RNA Extraction Kit, FastKing cDNA First Strand Synthesis Kit, and ChamQ Universal SYBR qPCR Master Mix. All reagents were obtained from commercial suppliers and used according to the manufacturers’ protocols.

This study utilized two zebrafish strains: the wild-type AB strain and the transgenic cartilage green fluorescent (Tg (OlaSp7:nlsGFP) cy25) zebrafish. All zebrafish were maintained in custom aquaculture water at 28°C, which was prepared by adding 200 mg of instant sea salt per liter of reverse osmosis water. The conductivity was maintained between 450-550 μS/cm, with a pH range of 6.5-8.5, and water hardness between 50-100 mg/L CaCO₃. The zebrafish were bred and provided by Hunter Biotechnology Inc. (Hangzhou, China) aquaculture center. This experiment received an animal use permit (Permit Number: SYXK (Zhe) 2022-0004) and strictly adhered to the standards of the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC) (Certification Number: 001458). All experimental procedures were ethically reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) (Review Number: IACUC-2024-8511-01).

DXMS-induced cranial bone injury model in wild-type zebrafish (21, 22): To investigate the impact of dexamethasone (DXMS) on cranial bone mineralization, a cranial bone injury model was established using 3 days post-fertilization (dpf) wild-type AB strain zebrafish. Zebrafish were randomly assigned to four groups: control (CTL), model control (MC), positive control (ALN), and LC86 treatment group. Thirty fish were maintained in 20 mL of culture water per beaker. Except for the CTL group, zebrafish in the MC, ALN, and LC86 groups were exposed to 1.5 μM DXMS for 4 consecutive days to induce cranial bone injury. The ALN group received 5.00 μg/mL of alendronate sodium, while the LC86 group was treated with probiotic LC86 at a final concentration of 3×10⁷ CFU/mL. All groups were maintained at 28°C with daily renewal of treatment solutions. At the end of the experiment, cranial bones were evaluated via fluorescence microscopy and quantitative analysis of fluorescence intensity ( Figure 1A ).

DXMS-induced osteogenic injury model in transgenic zebrafish (21, 22): A separate osteogenic injury model was developed using 3 dpf transgenic zebrafish expressing green fluorescence in osteoblasts (Tg (OlaSp7:nlsGFP) cy25) to evaluate the effect of LC86 on osteogenesis. Zebrafish were randomly divided into four groups (CTL, MC, ALN, and LC86), with 30 fish per well in 3 mL of water in 6-well plates. Except for the CTL group, the other groups were treated with 10 μM DXMS for 4 days to induce osteogenic impairment. The ALN group received alendronate sodium (5.00 μg/mL), and the LC86 group received probiotic LC86 at 3×10⁷ CFU/mL. Water and treatments were refreshed daily. At the conclusion of the experiment, zebrafish were imaged using fluorescence microscopy to assess hard bone fluorescence, and pixel intensity values were quantified to evaluate mineralization levels ( Figure 1B ).

DXMS-induced cartilage injury model in transgenic zebrafish (23): To explore the potential protective effects of LC86 against cartilage damage, a cartilage injury model was established using 2 dpf transgenic zebrafish expressing green fluorescence in cartilage tissue. Fish were assigned to four groups: CTL, model control (MC), positive control (CS), and LC86 treatment group. Thirty fish were placed in each well of a 6-well plate containing 3 mL of water. DXMS was added to all groups except CTL at a concentration of 25 μM to induce cartilage injury. The CS group received 1000 μg/mL of chondroitin sulfate as a positive control, and the LC86 group received 3×10⁷ CFU/mL of LC86. All fish were maintained at 28°C with daily solution changes. After 3 days of treatment, zebrafish were collected for imaging, and cartilage integrity was assessed by fluorescence intensity analysis using microscopy ( Figure 2A ).

Alizarin red staining and fluorescence microscopy imaging are critical techniques for assessing bone mineralization. Alizarin red dye specifically binds to the calcium salts in bone tissue, making it used for quantifying the degree of bone mineralization (24). At the end of the experiment, 10 zebrafish were randomly selected from each experimental group, placed under a fluorescence microscope for imaging. Images were analyzed using the NIS-Elements D3.20 advanced image processing software to collect data on the fluorescence intensity of the skull, hard bone, and cartilage.

At the end of the experiment, three independent biological samples were randomly selected from each group, with each replicate consisting of 30 pooled zebrafish larvae. Total RNA was separately from each biological replicate using an automatic nucleic acid extractor, and its concentration and purity were measured using a UV-Visible spectrophotometer (A₂₆₀/A₂₈₀ ratio). From each sample, 2.00 μg of total RNA was used to synthesize 20.0 μL of cDNA according to the guidelines provided by the cDNA first strand synthesis kit. Subsequently, real-time quantitative PCR (q-PCR) was employed to measure the expression levels of genes associated with bone formation, including β-actin, sp7, runx2a, bmp2a, bmp4, and col2a1a. Throughout this process, β-actin served as the internal reference gene, facilitating the calculation of relative RNA expression levels for sp7, runx2a, bmp2a, bmp4, and col2a1a. The related primer sequences are detailed in Supplementary Table S1 .

Each treatment group consisted of 18 biological replicates, with 30 zebrafish larvae per replicate (totaling 540 larvae per group). After a 4-day intervention, six biological replicates were randomly selected for untargeted metabolomics. For each replicate, 90 zebrafish larvae (pooled from three wells) were washed with ultrapure water, transferred to 1.5 mL Eppendorf tubes, flash-frozen in liquid nitrogen for 3 min, and stored at −80°C. For metabolite extraction, 25 mg of homogenized zebrafish tissue was mixed with 10 μL of internal standard and 800 μL of cold methanol: acetonitrile: water (2:2:1, v/v/v), followed by homogenization with magnetic beads. Samples were precipitated at −20°C for 2 h, centrifuged at 25,000 g at 4°C for 15 min, and 600 μL of supernatant was freeze-dried. The dried extracts were reconstituted in 600 μL of 50% methanol, vortexed, centrifuged again, and the final supernatant was used for analysis. Chromatographic separation was performed using a Waters BEH C18 column (1.7 μm, 2.1 × 100 mm) with a 15-minute gradient at 0.35 mL/min and 45°C. Mass spectrometry was conducted on a Q Exactive HF system in both positive and negative ion modes (scan range m/z 70–1050), using a Top3 DDA strategy. Data were processed with Compound Discoverer 3.3 and annotated against BMDB, mzCloud, and ChemSpider databases. Peak detection, normalization, and inter-group differential analysis were conducted for metabolic profiling.

The intestinal microbiota of zebrafish was analyzed using 6 biological replicates per group (n = 6). Four days post-treatment, the zebrafish samples were washed with ultrapure water at 28°C to remove any residual culture, then quickly transferred to 1.5 mL Eppendorf tubes (30 zebrafish per tube, per biological replicate). After removing all liquid, the samples were flash-frozen in liquid nitrogen for 3 min and stored at -80°C for subsequent 16S rRNA gene sequencing analysis. Genomic DNA extraction was performed using the QIAamp DNA Microbiome Kit (Qiagen, Germany) according to the manufacturer’s protocol, with additional mechanical lysis using bead beating (5.5 m/s for 60 s, MP FastPrep-24). DNA quality and concentration were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific) to measure the A₂₆₀/A₂₈₀ ratio (acceptable range: 1.8-2.0) and a Qubit 3.0 Fluorometer (Thermo Fisher Scientific) for DNA quantification (minimum 30 ng DNA required per sample). Post-PCR, the amplification products were purified using the Agencourt AMPure XP magnetic bead system (Beckman Coulter, USA) and dissolved in Elution Buffer to complete the library construction. The library fragment size and concentration were then assessed using an Agilent 2100 Bioanalyzer. Only libraries that passed quality control proceeded to high-throughput sequencing based on the size of the insert fragments. Sequencing data processing involved adapter sequence trimming with cutadapt v2.6 software and removal of low-quality sequences through a sliding window method to obtain high-quality clean data.

Raw sequencing reads were demultiplexed and quality-filtered using cutadapt v2.6 to remove adapter sequences. Low-quality bases were removed using QIIME 2 (DADA2 plugin, default parameters). Reads shorter than 400 bp or containing ambiguous bases (N > 2) were discarded. Paired-end reads were assembled into tags using overlapping relationships, forming the full-length V3–V4 region sequences. The merging criteria were as follows (1): minimum overlap length of 15 bp and (2) a maximum mismatch rate of 0.1 in the overlapping region. The merged tags were clustered into OTUs using USEARCH (v7.0.1090_i86linux32). The process included: (1) OTU clustering at 97% similarity using UPARSE software (version 7.1, http://drive5.com/uparse/) (25), generating representative OTU sequences; (2) chimera removal from OTU representatives using UCHIME (v4.2.40); and (3) OTU abundance profiling by aligning all tags back to representative sequences using the usearch_global method to generate the OTU table for each sample. Taxonomic classification was performed using the RDP Classifier (http://rdp.cme.msu.edu/) against the SILVA 138 database, with a confidence threshold of 70%. Alpha and beta diversity analyses were conducted in QIIME 2 (26), including Shannon and Chao1 indices for species diversity, principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarity, and linear discriminant analysis effect size (LEfSe) for biomarker identification. Based on the OTU results, additional analyses such as intersample species complexity, interspecies differences, correlation analysis, and predictive functional profiling using PICRUSt2 were performed.

The non-parametric Kruskal-Wallis test was used for multi-group comparisons. Differences between two groups were assessed using the non-parametric Mann-Whitney U test. A p-value of less than 0.05 was considered statistically significant. Continuous data were represented as mean ± standard deviation. PCoA based on unweighted UniFrac distances was constructed to visualize gut microbiome community parameters (27). LEfSe was conducted using the microeco package, PICRUSt data were evaluated using the Statistical Analysis of Metagenomic Profiles (STAMP, v 2.1.3), and genus-level abundance differences among the three groups were visualized with the heatmap package (28, 29). Statistical charts and graphs of gut microbiome community parameters were generated using the ggplot2 package in R (30). All statistical analyses were performed using R version 4.2.2.

Figure 1C illustrated the fluorescence intensity of zebrafish cranial bones under different treatment conditions and their quantification to assess the impact of various interventions on bone mineralization. Qualitative analysis showed that the CTL group displayed normal baseline fluorescence intensity with no apparent abnormalities. Compared to the CTL group, the MC group exhibited slightly reduced fluorescence intensity, suggesting a potential osteoporotic condition. Relative to the MC group, both the ALN group and LC86 group showed significantly enhanced fluorescence intensity. Further quantitative analysis revealed that, compared to the CTL group, the MC group’s fluorescence pixel values significantly decreased (p < 0.001), reflecting reduced levels of bone mineralization consistent with an osteoporotic phenotype. In contrast, cranial bone fluorescence intensity pixel values in both the probiotic LC86 and ALN-treated groups were significantly higher (p < 0.01 and p < 0.001, respectively) than those in the MC group, consistent with the increase in fluorescence intensity. The effect of LC86 intervention was similar to that of the known bone mineralization promoter, sodium alendronate, effectively enhancing bone mineralization.

Figure 1D demonstrated the fluorescence microscopy images and their quantitative analysis of zebrafish hard bones under different treatment conditions, assessing LC86’s osteogenic impact. Qualitatively, the CTL group displayed the natural state of untreated hard bones, showing baseline green fluorescence. In contrast, the MC group showed significantly diminished fluorescence, indicating reduced mineralization typical of osteoporosis. Conversely, the LC86 and ALN groups showed markedly enhanced fluorescence, suggesting successful promotion of bone mineralization. Quantitative assessments supported these findings. Compared to CTL group, the fluorescence pixel values in the MC group significantly decreased (p < 0.001), reflecting lowered bone mineralization levels. However, both LC86 and ALN treatments significantly increased fluorescence in hard bones (both p < 0.01) compared to the MC group, reversing mineralization losses and potentially boosting structural integrity.

The qualitative analysis ( Figure 2B ) revealed that CTL group exhibited baseline fluorescence, indicating healthy cartilage, whereas the MC group’s reduced fluorescence in specific areas suggested potential cartilage damage or decreased mineralization due to the model conditions. Both the ALN and LC86 groups exhibited significant enhancements in fluorescence intensity, especially the ALN group. Quantitative analysis ( Figure 2C ) corroborated the qualitative findings, showing significantly lower fluorescence pixel values (p < 0.05) in MC group compared to CTL group, confirming reduced cartilage mineralization. Conversely, the ALN and LC86 groups showed significantly higher fluorescence pixel values (both p < 0.05), indicating that these treatments reversed the mineralization loss and potentially enhanced the cartilage’s structural integrity. Although ALN effects were slightly superior, LC86 also demonstrated significant protective benefits.

At the end of study, total RNA was extracted from zebrafish, with UV-visible spectrophotometry confirming high quality suitable for q-PCR (A260/A280 ratios between 1.8-2.2, Supplementary Table S2 ). Figure 3 analyzed bone health-related gene expression in the zebrafish model to evaluate the effects of LC86. Compared to CTL group, sp7 gene expression in MC group decreased significantly (p < 0.01), but notably recovered in the LC86 and CS groups (both p < 0.01), with LC86 showing greater enhancement. RUNX family transcription factor 2a (runx2a), essential for bone development, was reduced in MC group (p < 0.01), indicating suppressed bone formation. Post-intervention, both LC86 and CS significantly boosted runx2a expression, with LC86 performing slightly better. Expression levels of bmp2a and bmp4, crucial for bone formation and repair, dropped in MC group but were significantly elevated after LC86 and CS treatment, partially reversing osteoporosis effects. Type II collagen (col2a1a) expression also increased substantially post-intervention, improving connective tissue integrity, though slightly less effectively with LC86 than CS. Overall, LC86 markedly mitigated gene expression downregulation in the osteoporotic model.

As illustrated in Figure 4A , PCA analysis revealed distinct distribution trends between groups, with clear separation between CTL and MC groups indicating significant metabolic differences. Overlaps between LC86 and MC suggest some shared metabolites. Figure 4B , through univariate and multivariate analyses, identified 528 differential metabolites between CTL and MC groups, with 290 upregulated and 238 downregulated. In LC86 group compared to MC, 366 differential metabolites were identified, including 150 upregulated and 216 downregulated, suggesting that LC86 intervention partially restored or regulated the metabolic status disrupted by the disease model. Further analysis of differential metabolites between (CTL vs. MC) and (LC86 vs. MC) groups revealed 144 shared metabolites with significant expression trends. Specifically, the top 15 upregulated and downregulated metabolites post-LC86 intervention ( Figure 4C , Supplementary Table S3 ) reflected significant metabolic activity changes. Upregulated metabolites like Metconazole and Daprodustat, associated with antifungal activity and erythropoiesis, suggest an enhancement or protective effect on metabolism by LC86. Downregulated metabolites like Dimethenamide OXA and Dexamethasone 21-sulfate indicate a potential reduction in metabolites associated with inflammation or chronic stress. Log₂FC values further quantified these changes, with positive values indicating upregulation and negative values indicating downregulation. By modulating these metabolic pathways, LC86 may exhibit anti-inflammatory, immunomodulatory, and antioxidant effects. Metabolic pathway enrichment analysis based on the KEGG database helped elucidate significant changes in the metabolic pathways, aiding in the interpretation of the biological phenotype. Metabolic pathways with a p-value below 0.05 were defined as significantly enriched, and the top ten pathways with the smallest p-values were plotted as bar graphs. Results in Figures 4D, E show that the top three enriched pathways with the most differential metabolites between CTL and MC were purine metabolism, nucleotide metabolism, and alanine, aspartate, and glutamate metabolism, while in LC86 versus MC, the top pathways were nucleotide metabolism, beta-alanine metabolism, and purine metabolism.

Microbial diversity analysis: Species accumulation curves ( Figure 5A ) indicated that sequencing depth was sufficient, with species richness reaching a saturation plateau across all groups. Alpha diversity indices (ACE, Chao1, Simpson, and Shannon; Figures 5B–E ) showed a trend of reduced microbial diversity in the MC group compared to CTL group, although the differences were not statistically significant. LC86 intervention did not significantly alter these diversity indices. Beta diversity analysis suggested some overlap in species composition between CTL and MC groups, but no distinct clustering was observed post-LC86 intervention, indicating greater sample dispersion. These results suggest that while LC86 did not significantly impact alpha diversity, it modulated the overall structure of the gut microbiome.

Taxonomic composition changes: Using LEfSe and STAMP, significant differences were identified at the genus level between groups. LEfSe analysis ( Figure 5G ) identified significant taxonomic shifts. LEfSe analysis revealed enrichment of pathogenic genera like Mycobacterium and Klebsiella in MC group, while beneficial genera such as Ligilactobacillus, Limosilactobacillus, Lacticaseibacillus, and Mesorhizobium increased in LC86 group, highlighting the beneficial impact of LC86. STAMP analysis further detailed these variations. Vibrio, Achromobacter, and Brucella decreased in the MC group, whereas Burkholderia, Mycobacterium, and Aquariibacter increased ( Figure 5H ). Notably, LC86 reversed these changes, significantly reducing harmful bacteria like Achromobacter and Aquariibacter and enhancing beneficial ones including Lacticaseibacillus and Ligilactobacillus ( Figure 5H ).

Functional pathway prediction: MetaCyc Level 2 pathway analysis ( Figures 6A, B ) evaluated the intervention’s effects on metabolic pathways. The glycolysis pathway was significantly more active in the MC group (p = 0.045), indicating increased energy metabolism demands in osteoporosis. While not statistically significant, slight increases in nucleoside and nucleotide biosynthesis and co-factor biosynthesis pathways were observed in the MC group, reflecting cellular proliferation, repair, and functional demands in the osteoporosis model. Compared to MC group, LC86 group exhibited significantly greater abundance of antibiotic resistance pathways (p = 0.005), suggesting that LC86 may boost resistance gene expression in zebrafish gut microbiome, enhancing antibiotic resistance. Additionally, a higher relative abundance of fermentation pathways in LC86 group (p = 0.020) implies that LC86 could enhance gut microecology by stimulating fermentation activity. Although differences in the aminoacyl-tRNA charging pathway were subtle (p = 0.031), its slightly reduced abundance in LC86 group suggests that LC86 may have optimized protein synthesis efficiency. Additionally, the alcohol degradation pathway was notably lower in LC86 group (p = 0.045), indicating a possible decrease in harmful metabolite degradation. Furthermore, the significant increase in fatty acid and lipid biosynthesis pathways in LC86 group (p = 0.045) implies that the probiotic could enhance bone health by boosting lipid metabolism.

Analysis of KEGG Level 3 pathways evaluated the impact of LC86 on osteoporotic zebrafish ( Figures 6C, D ). Relative to CTL group, MC group exhibited significant reductions in sulfur transfer, taurine and hypotaurine, glycine, serine and threonine metabolism, and citrate cycle (all p < 0.05). Increases were observed in betaine biosynthesis, butaphosphan and neomycin biosynthesis, apoptosis, proteasome activity, and pathways related to pathogenic E. coli infection (all p < 0.05). Compared to MC, LC86 group showed higher activity in pathways associated with systemic lupus erythematosus, streptomycin biosynthesis, and myo-inositol phosphate metabolism (all p < 0.05), and reductions in the proteasome, betaine biosynthesis, lipopolysaccharide biosynthesis, photosynthesis-antenna proteins, NOD-like receptor signaling, oxidative phosphorylation, and styrene degradation (all p < 0.05).