Gut virome dysbiosis contributes to premature ovarian insufficiency by modulating gut bacteriome.

Premature ovarian insufficiency (POI) is characterized by a significant decline or cessation of ovarian function before the age of 40 in women, resulting in irregular menstruation or amenorrhea, diminished ovarian hormone levels, and reduced fertility. 1 , 2 As the leading cause of female infertility, POI can also trigger complications such as osteoporosis, anxiety, and increased cardiovascular risks. 2 , 3 Current standard clinical management for POI is limited to hormone replacement therapy (HRT) and assisted reproductive technology (ART), both of which remain suboptimal. While HRT alleviates symptoms, it neither restores fertility nor prevents ovarian senescence, and ART is costly and often ineffective. Emerging approaches such as stem-cell therapy remain experimental and lack robust clinical validation. Therefore, treatment options for POI remain limited and suboptimal. 4 The pathogenesis of POI is very complex, involving an interplay of genetic factors, autoimmunity, environmental exposure, and metabolic dysregulation. In this complex pathophysiology, the hypothalamic-pituitary-ovarian (HPO) axis hormones, particularly estrogen, are pivotal in regulating ovarian function. HPO axis dysfunction is implicated in POI pathogenesis, encompassing abnormal gonadotropin secretion, ovarian function decline, and premature follicular depletion. Nonetheless, the detailed pathogenesis of POI remains elusive and debated, 5-7 underscoring the need for more comprehensive research. In addition to endocrine regulation, mounting evidence implicates the gut microbiota as a critical extra-ovarian modulator of the HPO axis and ovarian function. The human gut harbors numerous bacteria, collectively known as gut bacteriome, which play essential roles in ovarian function, namely the gut-ovary axis. 8-10 Several studies have revealed a close association between gut dysbiosis and ovarian dysfunction, including ovarian insufficiency disorders. 8 , 11 , 12 For instance, in patients with polycystic ovary syndrome (PCOS), the abundance of gut B. vulgatus is significantly increased, which induces PCOS-like phenotypes in mice. 13 POI mouse models show a distinct shift in gut microbial composition, characterized by decreases in the commensal and potentially anti-inflammatory taxa (e.g., Helicobacter , Odoribacter , Alistipes ) and increases in the bacteria associated with metabolic or pro-inflammatory activity (e.g., Clostridium XIVa , Barnesiella , Bacteroides , Mucispirillum ). 14 Notably, gut microbiota changes in POI patients are closely associated with key sex hormones including follicle stimulating hormone (FSH), luteinizing hormone (LH), 17β-estradiol (E2), and anti-Müllerian hormone (AMH). 11 Sleep deprivation also disrupts the gut microbiome and accelerates POI progression. This effect appears to be mediated by circadian-driven gut barrier dysfunction and systemic inflammation, which compromise ovarian reserve. Transferring gut microbiota from sleep-deprived to normal mice can induce similar POI symptoms. 15 Furthermore, many enzymes produced by gut bacteria, collectively known as the gut estrobolome, are critical for estrogen metabolism and reabsorption. 16 Gut dysbiosis significantly correlates with abnormal levels of HPO axis hormones, including E2, FSH, LH, and progesterone. 11 Despite this progress, the causal relationships and underlying mechanisms linking gut dysbiosis and POI, particularly the characteristic changes and etiological roles of non-bacterial gut microbes, remain poorly understood. In addition to bacteria, the gut microbiome encompasses numerous non-bacterial microbes, particularly bacteriophages, which are pivotal in shaping bacterial communities and maintaining gut ecological stability. 17 Metagenomic analyzes have revealed distinct viral signatures in PCOS patients, suggesting a functional link between the gut virome and ovarian health. 18 Transferring fecal viral-like particles (VLPs) from mice enriched for A. muciniphila significantly enhances fertility in recipient animals. 19 Beyond reproductive disorders, gut virome abnormalities have also been reported across multiple metabolic and inflammatory diseases. 20-23 Recent studies indicate potential etiological roles of gut phages in obesity and colorectal cancer. 24 Collectively, these studies demonstrate that gut virome dysbiosis is associated with various diseases and plays a causal role in certain non-reproductive disorders. However, the role of the gut virome in the pathogenesis of POI remains poorly understood. Given the marked bacteriome dysbiosis in ovarian dysfunction and the close interdependence between phages and bacteria, investigating the gut virome in POI is warranted. Hence, we investigated the aforementioned issues using whole-genome shotgun sequencing of gut VLPs samples from 60 healthy reproductive-age women and POI patients, combined with fecal virome transplantation (FVT). Our study suggests that gut virome dysbiosis may contribute to POI and elucidates the underlying mechanisms. These findings support a possible role for the gut virome in regulating ovarian function and indicate its potential relevance for ameliorating ovarian dysfunction, including POI, and female infertility.

To characterize the gut virome in POI patients, we performed whole-genome deep sequencing of gut VLPs from 30 healthy reproductive-age women and 30 POI patients. The baseline clinical characteristics and dietary habits of the two groups were similar, except for POI-associated indicators (Table S1). Sequencing depth reached saturation for both groups without significant differences (Figure S1A, S1B). We identified 2,587 non-redundant contigs, of which 45% were viral sequences, with the remainder predominantly plasmids (1,288/1,423, 91%). Using stringent criteria, 43 551 viral operational taxonomic units (vOTUs) were identified. Most vOTUs (56.8%) were of high quality (genome completeness > 90%), while 14.9% were medium quality (50–90%) and 28.3% were low quality ( < 50%) (Figure S1C, S1D). The identified viral genomes predominantly ranged from 50 to 100 kilobase pairs (kbp) (65.2%), with smaller proportions of short viral contigs (200 kbp, 4.5%) representing the extremes of the genome-size distribution (Figure S1E). Over 90% of these genomes were double-stranded DNA (dsDNA) viruses, with minor proportions of single-stranded DNA (ssDNA) and rare retro-transcribing elements identified by automated annotation tools. The latter two categories were excluded from downstream analyzes. Among these viruses, 94.5% were bacteriophages, including 63 novel viral sequences absent in existing databases (MGV, GVD, and GPD) ( Figure 1A ; Figure S1F; Table S4). The top three functions enriched in gut virome genes were unknown functions, phage genome replication and assembly, and host defense inhibition, highlighting the current limited understanding of gut virome functions ( Figure 1B ). Characteristics of the Gut Virome in Reproductive-Age Women and Abnormalities in Composition and Abundance of Gut Virome in POI Patients. (A) Percentage distribution of virus types and their corresponding hosts in the gut virome. (B) Stacked bar chart of viral functional gene categories in the gut virome of healthy reproductive-age women (HC) and POI patients (POI). (C) Count, taxonomic classification, lifestyles, and genome sizes of gut phages (vOTUs), organized by their primary bacterial hosts. (D) Comparison of alpha diversity between HC and POI groups. Richness, Cliff's Delta: 0.458, 95% CI: 0.238 to 0.655; Shannon, Cliff's Delta: 0.716, 95% CI: 0.511 to 0.889; Pielou, Cliff's Delta: 0.542, 95% CI: 0.327 to 0.729; Simpson, Cliff's Delta: 0.622, 95% CI: 0.403 to 0.791. (E) Beta diversity analysis of gut vOTUs between HC and POI groups. (F) Distance-based redundancy analysis showing the association between vOTU profiles and primary HPO axis hormone levels. Arrows represent the direction and strength of associations. (G) Comparison of the percentages of temperate phages in the gut virome of HC and POI groups (Cliff's Delta: −0.416, 95% CI: −0.615 to −0.203). (H) Fold changes in the abundances of the top 50 vOTUs significantly associated with HPO axis hormone levels and a heatmap of partial Spearman correlations (adjusted for age and BMI) between these vOTUs and HPO axis hormones. (I) Relative abundance comparison of Faecalibacterium , Bacteroides , and Paraclostridium between HC and POI groups. n  = 30 per group for all panels. Two-tailed Wilcoxon rank-sum test for (D, E, G, and I); Partial Spearman’s correlation test with FDR adjustment for (H). * p  < 0.05; ** p  < 0.01; *** p  < 0.001. At the phylum level, most bacterial hosts for the vOTUs belonged to Firmicutes , Bacteroidetes , and Proteobacteria . At the genus level, over 80% of vOTUs were predicted to infect unknown, Faecalibacterium , Clostridium , Bacteroides , Lactobacillus , and Ruminococcus ( Figure 1C ). The major taxonomic categories of these phages included Straboviridae , Casjensviridae , Herelleviridae , and Peduoviridae . Additionally, crAss-like phages were also prominent, primarily targeting the Bacteroides genus . Phage types varied markedly across different hosts. For example, phages infecting Bacteroides , Akkermansia , and Streptococcus were predominantly virulent, whereas those infecting Roseburia , Mycobacterium , Lachnospiracea , and Agrobacterium were mostly temperate ( Figure 1C ). Furthermore, phage genome sizes varied significantly across different hosts. A significant positive correlation was observed between the genome sizes of phages and their bacterial hosts ( Figure 1C ; Figure S1G). Compared to healthy reproductive-age women, POI patients had a gut virome with significantly lower α -diversity and virus/bacteria ratio (VBR) ( Figure 1D ; Figure S2A). β -diversity analysis revealed a significant separation in gut virome composition between POI patients and healthy controls. Notably, healthy group samples clustered more tightly, indicating a less variable virome composition, whereas POI patient samples showed a more dispersed distribution, suggesting greater inter-individual variability ( Figure 1E ). Permutational multivariate analysis of variance (PERMANOVA) demonstrated that the POI disease state and associated sex hormone levels (E2, FSH, FSH/LH ratio, LH, and Progesterone) were significantly related to gut virome alterations, with the disease state (group) showing the largest contribution. In contrast, lipid levels (TC, LDL, and TG), age, and BMI did not show significant contributions (Figure S2B). Redundancy analysis (RDA) further revealed that E2 and FSH levels strongly correlated with gut virome composition variations in POI patients, while progesterone (PROG) levels and the FSH/LH ratio showed a stronger correlation with those in healthy reproductive-age women ( Figure 1F ). Additionally, lifestyle prediction indicated a marked increase in the proportion of temperate bacteriophages in POI patients ( Figure 1G ). Moreover, POI patients exhibited significant changes in the abundance of numerous gut phages at the family, genus, and vOTU levels (Figure S2C−S2E). Among them, 50 vOTUs significantly correlated with HPO axis hormone levels (progesterone, E2, LH, and FSH), primarily targeting Faecalibacterium , Bacteroides , and Paraclostridium . The abundances of these three bacterial genera were significantly decreased and strongly correlated with progesterone, E2, FSH, and LH levels, exhibiting trends contrary to those of their corresponding phages ( Figure 1H, 1I ; Figure S2F). SEM revealed a plausible association pattern in which reduced crAss phage abundance was linked to lower Bacteroides levels and, consequently, to decreased serum E2 concentrations (Figure S2G). These results indicated that gut virome dysbiosis is closely associated with ovarian dysfunction and the onset of POI. Pfam annotation of vOTU revealed that in healthy reproductive-age women, gut virome functions were predominantly associated with phage lysis and release, and DNA injection, while those in POI patients mainly involved host recognition and adsorption, host synthesis regulation, and phage adaptation and evolution (Figure S2H). A significant increase in the AcrIF1 gene level, which targets the type I-F CRISPR-Cas system, was observed in the gut virome of POI patients ( Figure 2A ). Additionally, POI patients exhibited significantly elevated gene levels of horizontal gene transfer (HGT) in their gut virome, which included increased levels of antibiotic resistance genes and multiple virulence factors (e.g., CagA, diphtheria toxin, and RTX toxin) ( Figure 2B, 2C ). In the gut virome of POI patients, we found a significant increase in the abundance, Shannon, and Pielou indices of AMGs. Notably, AMGs associated with energy and nucleic acid metabolism, folding, sorting and degradation, exhibited increased abundance, whereas those involved in amino acid metabolism decreased ( Figure 2D, 2E ; Figure S2I). Moreover, the proportion of virulent phages was significantly reduced in the gut virome of POI patients, accompanied by an elevated integrase to lysozyme gene abundance ratio, suggesting a preferential shift toward lysogenic cycles ( Figure 2F , Figure S2J). Characteristic Changes in Gut Virome Functions in POI Patients and Relationships Between Gut Phages and Bacteria. (A, B) Comparison of the relative abundance of the anti-CRISPR gene AcrIF1 (A) and HGT genes (B) in the gut virome between healthy reproductive-age women (HC) and POI patients (POI). (C) Comparison of the abundances of antibiotic resistance and virulence factor genes in the gut virome between HC and POI groups. (D, E) Comparison of the abundance (D) and functional classification (E) of AMGs in the gut virome between HC and POI groups. (F) Comparison of the percentages of virulent phages in the gut virome between HC and POI groups (Cliff's Delta: 0.360, 95% CI: 0.138 to 0.565). (G) Procrustes analysis of gut bacteria (species level) and virus (vOTUs level) in HC and POI groups. (H) Network showing associations between major vOTUs and their predicted bacterial hosts in HC and POI groups. Network edges represent conditional associations inferred using SPIEC-EASI, a compositionally aware method based on sparse inverse covariance estimation. (I) Distribution of correlations between major phages and their predicted bacterial hosts, as well as their respective host abundances. Correlations were calculated as the average correlation coefficients between the relative abundance of a bacterial genus and all host-assigned phages targeting it. The bar plots represent the relative abundance of host bacteria at the genus level. The dots indicate correlation coefficients. Lines ascending (descending) from left to right indicate strengthened (weakened) correlations in the POI group. n  = 30 per group for (A−G). Two-tailed Wilcoxon rank-sum test for (A, B, D, F, and G); Two-tailed Wilcoxon rank-sum test with FDR correction for (C and E); Spearman’s correlation test with Holm-Bonferroni adjustment for (I). P values are presented as adjusted P values, with the corresponding unadjusted p values shown in parentheses for (E). ** p  < 0.01; *** p  < 0.001; ns: no significance. We found that the bacterial and bacteriophage communities exhibited notable co-segregation in both healthy reproductive-age women and POI patients ( Figure 2G ). POI patients showed significantly strengthened positive correlations between diversity indices (Shannon, Simpson, Richness, and Pielou) of gut bacteria and those of gut bacteriophages, suggesting closer symbiotic relationships between them (Figure S2K). Furthermore, the number of significant correlations (r > 0.5, p  < 0.05) between gut viral vOTUs and bacteria in POI patients markedly increased from 322 to 646. The overall correlation patterns shifted from predominantly negative in the healthy group to primarily positive in POI patients. In the virus-bacteria correlation network, key bacterial nodes observed in healthy reproductive-age women (e.g., Klebsiella , Eggerthella , Ruminococcus , and Eubacterium ), which were negatively correlated with viruses, disappeared in POI patients. Instead, new key bacterial nodes with positive correlations to viruses emerged in POI patients, including Escherichia , Sporosarcina, Barnesiella , and Bifidobacterium ( Figure 2H ). Conversely, positive correlations between certain estrobolome bacteria (e.g., Akkermansia, Faecalibacterium, and Roseburia , all showing significantly decreased abundance) and their respective bacteriophages were weakened in POI patients ( Figure 2I ). These findings demonstrated significant alterations in the ecological relationships between gut bacteria and viruses/phages in POI patients. To elucidate the causal relationship between gut virome dysbiosis and POI, we performed fecal virome transplantation (FVT) (Figure S3). Endotoxin (LPS) concentrations in the administered VLP suspensions were maintained below 0.5 EU/mL across all groups (Figure S4A). FVT did not significantly alter liver and kidney morphology, function, or tissue inflammatory cytokine levels, nor did it affect the proportion of circulating lymphocytes, confirming the safety of this intervention (Figure S4B–S4G). Compared to untreated controls (Ctrl), antibiotic treatment (Ctrl + AB) reduced vaginal plug positive rates (without statistical significance), average lordosis quotient, and gut VLP numbers. Transplantation of gut bacteriome from healthy reproductive-age women (BH) significantly ameliorated these antibiotic-induced abnormalities. Co-transplanting gut VLPs from healthy reproductive-age women (BH + VH) further increased gut VLP numbers and enhanced sexual behaviors. Conversely, gut VLPs from POI patients (BH + VPOI) exerted opposite effects on sexual behaviors, which were abolished by heat inactivation (BH + hkVPOI) ( Figure 3A, 3B ; Figure S4H). Additionally, antibiotic treatment significantly prolonged estrous cycles and reduced pregnancy rates (without statistical significance) in plug-positive female rats. These effects were reversed by BH transplantation or BH + VH co-transplantation, with the latter showing more pronounced improvements. Conversely, gut VLPs from POI patients (BH + VPOI) counteracted these improvements, extending the estrous cycle and decreasing pregnancy rates, which were eliminated by heat-killing ( Figure 3C, 3D ). Influence of Gut VLPs on Reproductive Outcomes and Ovarian Function in Female Rats. (A) Comparison of the positive rate of vaginal plugs in female rats of each group. (B) Comparison of the average lordosis quotient across groups. Lordosis quotient: the number of lordosis responses of female rats divided by the number of mounting attempts by male rats, indicating the willingness of female rats to mate. (C, D) Comparison of the estrous cycle length (C) and pregnancy rate in female rats with vaginal plugs (D) across groups. (E−I) Comparison of serum levels of estradiol (E), follicle-stimulating hormone (FSH) (F), progesterone (PROG) (G), luteinizing hormone (LH) (H), and the FSH/LH ratio (I) in pregnant rats across groups. For serum estradiol levels in (E), Cohen’s d effect sizes for pairwise comparisons were as follows: Ctrl vs. Ctrl + AB, 1.15; Ctrl + AB vs. BH, −2.20; BH vs. BH + VH, −1.59; BH vs. BH + VPOI, 2.30; BH vs. BH + hkVPOI, −0.12; BH + VH vs. BH + VPOI, 3.88; BH + VH vs. BH + hkVPOI, 1.47; BH + VPOI vs. BH + hkVPOI, −2.41. For serum FSH in (F), Cohen's d effect sizes for pairwise comparisons were as follows: Ctrl vs. Ctrl + AB, −1.66; Ctrl + AB vs. BH, 3.46; BH vs. BH + VH, 1.22; BH vs. BH + VPOI, −11.62; BH vs. BH + hkVPOI, 0.03; BH + VH vs. BH + VPOI, −12.76; BH + VH vs. BH + hkVPOI, −1.05. (J) Representative images of ovarian tissue sections stained with Hematoxylin and eosin (H&E) (Left panel) and comparison of follicle numbers (Right panel) in pregnant rats among groups. Cohen’s d effect sizes for pairwise comparisons were as follows: Ctrl vs. Ctrl + AB, 1.13; Ctrl + AB vs. BH, −1.60; BH vs. BH + VH, −1.12; BH vs. BH + VPOI, 1.71; BH vs. BH + hkVPOI, −0.04; BH + VH vs. BH + VPOI, 2.85; BH + VH vs. BH + hkVPOI, 0.95; BH + VPOI vs. BH + hkVPOI, −1.77. Ctrl: untreated normal female rats; Ctrl + AB: female rats treated with antibiotics to deplete gut bacteriome; BH: female rats humanized with gut bacteriome from healthy reproductive-age women; BH + VH (BH + VPOI): BH group rats transplanted with gut VLPs from healthy reproductive-age women (POI patients); BH + hkVPOI: BH group rats transplanted with heat-killed gut VLPs from POI patients. For A−D, n  = 15 per group; For E−J: Ctrl = 7, Ctrl + AB = 6, BH = 9, BH + VH = 15, BH + VPOI = 5, BH + hkVPOI = 10. Chi-square test with Benjamini–Hochberg correction for (A, D); One-way ANOVA followed by Benjamini–Hochberg FDR–adjusted post hoc pairwise comparisons for (B, C, E−J); p values are presented as adjusted p values, with the corresponding unadjusted p values shown in parentheses. ns: no significance. HPO axis hormones and follicular development are essential for reproductive function. We observed that antibiotic treatment significantly decreased serum estradiol levels and follicle numbers while markedly increased serum FSH levels, without affecting progesterone (PROG), LH, or FSH/LH ratio. BH transplantation significantly mitigated the antibiotic-induced changes in serum estradiol, follicle numbers, as well as FSH, PROG, and LH levels, except for the FSH/LH ratio. BH + VH co-transplantation further enhanced these benefits, whereas BH + VPOI reversed them. Heat-inactivation of POI gut VLPs (BH + hkVPOI) eliminated this suppressive effect, making the levels of aforementioned parameters comparable to the BH group ( Figure 3E−3J ). Additionally, serum testosterone levels, placental aromatase (CYP19A1) expression, and ovarian morphology showed no significant differences across all groups (Figure S4I, S4J). These findings implied a mechanistic link between gut virome dysbiosis and POI development. The intricate interplay between gut phages and bacteria is widely recognized, but the extent to which phages shape the gut bacteriome remains incompletely understood. We found that the gut bacteriome of healthy reproductive-age women (BH) effectively mitigated antibiotic-induced decreases in amplicon sequence variants (ASVs) and α -diversity of the gut bacteriome (Ctrl + AB). Transplanting gut VLPs from either healthy women (BH + VH) or POI patients (BH + VPOI) further enhanced these effects, which were eliminated by heat inactivation (BH + hkVPOI) (Figure S5A; Figure 4A, 4B ). β -diversity analysis revealed that prior to FVT, the gut bacterial composition in rats receiving healthy gut bacteriome (BH-FMT) was highly congruent with that of healthy donors (HC) but differed from POI patients (POI) ( Figure 4C ). Compared to BH transplantation alone, BH + VH transplantation led to greater similarity with HC. However, BH + VPOI shifted the gut bacteriome toward a POI-like profile. This effect was not observed with heat-inactivated POI VLPs ( Figure 4D ). Before FVT (Day 21), the gut virome of BH + VH and BH + VPOI groups (pre-FVT) significantly differed from HC and POI. By Day 42 after FVT, however, the gut virome of BH + VH and BH + VPOI group closely matched their respective donors (HC or POI), with nearly complete overlap ( Figure 4E ; Figure S5B). Furthermore, this similarity persisted four weeks after the final FVT (Figure S5C, S5D). Profound Modulation of Gut Bacteriome and Estrobolome by Gut VLPs. (A, B) Comparison of Richness (A) and Shannon index (B) of gut bacteriome across groups. (C, D) Beta diversity analyzes of gut bacteriome in recipient female rats and donors 14 days after FMT but before FVT (C), and 14 days after FVT (D). (E) Beta diversity analyzes of gut VLPs in female rats of each group and donors before FVT and at the experimental endpoint. (F) Changes in the abundance of key bacterial genera in the gut estrobolome of female rats before and after FMT or FVT (Left Panel) and comparison of their average relative abundances across groups at experiment completion (Right Panel). (G) Multiple factor analysis (MFA) illustrating relationships between major bacterial genera in the gut estrobolome and serum sex hormone levels in pregnant rats. (H) Comparison of expression levels (Left Panel) and activity (Right Panel) of β -glucuronidase (GUS) in feces of pregnant rats across groups. (I) Scatter plots showing Spearman correlations between fecal GUS enzyme activity and estradiol levels in feces (Left Panel) and serum (Right Panel). HC: healthy reproductive-age women; POI: patients with POI; BH-FMT: female rats humanized with gut bacteriome from healthy reproductive-age women, subsequently subdivided into BH, BH + VH, BH + VPOI, and BH + hkVPOI groups for FVT experiments; Ctrl: untreated normal female rats; Ctrl + AB: female rats treated with antibiotics to deplete gut bacteriome; BH: female rats humanized with gut bacteriome from healthy reproductive-age women; BH + VH (BH + VPOI): BH group rats transplanted with gut VLPs from healthy reproductive-age women (POI patients); BH + hkVPOI: BH group rats transplanted with heat-killed gut VLPs from POI patients. Pre-FVT: female rats of the BH + VH and BH + VPOI groups before FVT. For A−D and F, n  = 6 per group; For E: HC, POI, BH + VH and BH + VPOI = 6, Pre-FVT = 12; For H, Ctrl = 7, Ctrl + AB = 6, BH = 9, BH + VH = 15, BH + VPOI = 5, BH + hkVPOI = 10. One-way ANOVA followed by Benjamini–Hochberg FDR–adjusted post hoc pairwise comparisons for (A−E, right panel in F, H); t -test for (left panel in F); Spearman’s correlation test for (I). p values are presented as adjusted p values, with the corresponding unadjusted p values shown in parentheses for (A-F, H). * p  < 0.05; ns: no significance. The gut estrobolome plays a critical role in estrogen metabolism. 8 BH transplantation significantly increased estrogen-metabolizing bacteria (e.g., Bacteroides and Akkermansia ) in antibiotic-treated female rats (before FMT group). BH + VH further increased their abundance, while BH + VPOI decreased it, an effect eliminated by heat inactivation (BH + hkVPOI) ( Figure 4F ). These bacteria, including Akkermansia , Bacteroides , and Alistipes , showed significant correlations with estrogen and progesterone levels ( Figure 4G ). β -Glucuronidase (GUS), a key estrobolome gene, 16 was significantly suppressed by antibiotics (Ctrl + AB) and restored by BH transplantation (BH). Healthy gut VLPs further enhanced the effects of BH, while gut VLPs from POI patients significantly diminished them. Heat-killed VLPs abolished these effects ( Figure 4H ). Additionally, estrogens levels in gut and serum were significantly positively correlated with gut GUS activity ( Figure 4I ). Consistent with these observations, mediation analysis indicated that the Shannon index of gut VLPs influenced E2 levels both directly and indirectly via gut GUS activity (Table S5). These results suggest that gut VLPs regulate estrogen levels by modulating estrobolome, potentially contributing to POI symptoms. Intestinal barrier injury caused by gut microbiome dysbiosis often results in enhanced oxidative stress and inflammation. We found that healthy gut bacteriome (BH) reversed antibiotic-induced decreases in colonic tight junction proteins (Occludin and ZO-1) without affecting estrogen receptor beta (Erβ). VLPs from healthy women (BH + VH) further enhanced these effects, while POI-derived VLPs (BH + VPOI) significantly attenuated them and increased inflammatory factor levels in serum and colon tissues of rats transplanted with healthy gut bacteriome. These effects of POI-derived VLPs were eliminated by heat-killing ( Figure 5A−5D ; Figure S5E, S5F). Gut VLPs-mediated Restoration of Intestinal Barrier and Attenuation of Ovarian Oxidative Stress and Inflammation in POI rats. (A, B) Representative immunohistochemical images of Occludin and ZO-1 in colon tissues across groups (A) and their quantitative comparison (B). (C, D) Comparison of the levels of inflammatory cytokines in serum (C) and colon tissues (D) across groups. (E) Principal component analysis (PCA) of gene expressions in ovarian tissues of rats across groups. (F) Gene Ontology (GO) enrichment analysis of differentially expressed genes between groups. (G) Heatmap showing the relative abundance of differential genes related to oxidative stress and inflammatory response. (H) Interaction network of genes involved in oxidative stress and inflammatory response constructed based on the STRING database. The red (blue) arrow indicates genes with significantly increased (decreased) levels in the group. (I) Comparison of enzyme activity (SOD, GSH, CAT) and MDA levels in ovarian tissues across groups. (J) Comparison of the levels of inflammatory cytokines in ovarian tissues across groups. (K) Heatmap of Spearman correlations between inflammatory cytokines, oxidative stress factors and HPO axis hormone levels, follicle numbers, reproductive-related indicators. ALQ: average lordosis quotient . (L) Heatmap of Spearman correlations between intestinal tight junction proteins (Occludin and ZO-1) and inflammatory cytokines, oxidative stress factors, HPO axis hormone levels, follicle numbers, and reproductive-related indices. Ctrl: untreated normal female rats; Ctrl + AB: female rats treated with antibiotics to deplete gut bacteriome; BH: female rats humanized with gut bacteriome from healthy reproductive-age women; BH + VH (BH + VPOI): BH group rats transplanted with gut VLPs from healthy reproductive-age women (POI patients); BH + hkVPOI: BH group rats transplanted with heat-killed gut VLPs from POI patients. For B−D, I and J, n  = 15 per group; For E−G, n  = 5 per group. One-way ANOVA followed by Benjamini–Hochberg FDR–adjusted post hoc pairwise comparisons for (B, IL-1β in C, IL-1β, IL6 in D, E, I, J); Fisher’s Exact Test followed by Benjamini-Hochberg FDR correction for (F); Kruskal–Wallis test followed by Benjamini–Hochberg FDR–adjusted pairwise Wilcoxon tests for (IL6, TNF- α in C, MCP-1, TNF- α in D); Spearman’s correlation test with Benjamini-Hochberg adjustment for (K and L). p values are presented as adjusted p values, with the corresponding unadjusted p values shown in parentheses for (B, I). * p  < 0.05; ** p  < 0.01; *** p  < 0.001; $ p  < 0.05 vs. other groups; ns: no significance. Transcriptomic analysis revealed that FVT significantly altered gene expression profiles in the ovarian tissues of recipient rats. Notably, the effects of healthy and POI gut virome showed opposing patterns. Compared to the gut bacteriome, the gut virome caused more substantial perturbations on ovarian gene expression profiles ( Figure 5E ; Figure S5G, S5H). Compared to the Ctrl + AB group, responses related to oxidative stress and inflammation were markedly suppressed in BH and BH + VH groups, with the BH + VH group showing a more pronounced suppression. Conversely, these responses were significantly enhanced in the BH + VPOI group ( Figure 5F, 5G ). Specifically, POI gut VLPs significantly upregulated pro-inflammatory and pro-oxidative stress genes in ovarian tissue (e.g., IL-6, IL-1β, Egfr, and Duox2) while downregulating antioxidant genes (e.g., Sod1, Gclc, Gclm, Prdx4, and Cat). In contrast, gut VLPs from healthy women (BH + VH) significantly suppressed the expression of pro-inflammatory genes (e.g., IL-18, Ccl2, and Cxcl6) and upregulated the aforementioned antioxidant genes. Gene interaction network analysis identified IL-6, IL-1β, Tp53, and Sod1 as key nodes connecting oxidative stress and inflammatory responses, which is consistent with previous reports ( Figure 5G, 5H ; Figure S5I, S5J). 44 , 45 ELIZA and biochemical assays further confirmed these findings regarding ovarian tissue oxidative stress and inflammation, which were abolished by heat inactivation ( Figure 5I, 5J ). Additionally, pro-inflammatory/pro-oxidative stress factors in different tissues were significantly negatively correlated with E2, progesterone, follicle number, and average lordosis quotient (ALQ), while significantly positively correlated with FSH, LH, and estrous cycle. Intestinal barrier integrity was significantly negatively correlated with MDA, inflammatory factors, FSH, LH, and estrous cycle, while significantly positively correlated with other indices ( Figure 5K, 5L ). Having established that gut virome dysbiosis promotes POI development, we used a CTX-induced POI rat model to explored the therapeutic potential of gut virome from healthy reproductive-age women (Figure S6A). LPS concentrations in the administered VLP suspensions were maintained below 0.5 EU/mL across all groups (Figure S6B). Compared to antibiotic-treated female rats (AB), CTX treatment significantly reduced mating willingness and success rates, as indicated by lordosis quotient and vaginal plug positive rate, in POI group rats ( Figure 6A, 6B ). CTX treatment also prolonged the estrous cycle and decreased pregnancy rates among female rats with vaginal plugs ( Figure 6C, 6D ). Moreover, CTX treatment significantly reduced serum estrogen and progesterone levels while increasing FSH, LH, and the FSH/LH ratio. We also observed a marked decrease in ovarian diameter, mature follicle number, and ERβ expression in colonic tissues ( Figure 6E−6H ; Figure S6C). Transplantation of healthy gut bacteriome (BH) significantly ameliorated the above pathological changes in POI rats. Co-transplantation of gut bacteriome and VLPs from healthy reproductive-age women (BH + VH) further improved the condition of POI rats, showing more pronounced amelioration of these pathological alterations (estrous-cycle recovery + 21.8% and E2 = 62.3 pg/mL) ( Figure 6A−6H ; Figure S6C). Additionally, healthy VLPs significantly increased Akkermansia muciniphila abundance and the levels and activity of GUS enzyme in the gut of BH group rats (BH + VH) (Figure S6D−S6F). Significant amelioration of POI Rats by Gut VLPs from healthy reproductive-age women. (A) Comparison of the average lordosis quotient across groups. Lordosis quotient: the number of lordosis responses of female rats divided by the number of mounting attempts by male rats, indicating the willingness of female rats to mate. (B) Comparison of the positive rate of vaginal plugs in female rats of each group. (C, D) Comparison of average estrous cycle length (C, BH + VH vs. POI, Cliff's Delta: −1.40, 95% CI: −1.98 to −0.82) and pregnancy rates in rats with vaginal plugs (D) across groups. (E, F) Comparison of serum levels of estradiol (BH + VH vs. POI, Cliff's Delta: 34.81, 95% CI: 22.24 to 47.39), progesterone (PROG, BH + VH vs. POI, Cliff's Delta: 48.89, 95% CI: 28.41 to 69.36), follicle-stimulating hormone (FSH, BH + VH vs. POI, Cliff's Delta: −20.10, 95% CI: −25.10 to −15.09), luteinizing hormone (LH, BH + VH vs. POI, Cliff's Delta: −5.04, 95% CI: −6.48 to −3.59) (E), and the FSH/LH ratio (F) across groups. (G) Gross morphological images of ovarian tissues (Left Panel) and quantitative comparison of ovarian diameter (Right Panel) among groups. (H) Representative images of ovarian tissue sections stained with Hematoxylin and eosin (H&E) (Upper Panel) and quantitative comparison of follicle numbers (Lower Panel) among groups. (BH + VH vs. POI, Cliff's Delta: 4.52, 95% CI: 3.15 to 5.89). (I) Principal component analysis (PCA) of gene expressions in ovarian tissues of female rats across groups. (J) Comparison of expression levels of important genes associated with inflammation and oxidative stress across groups based on transcriptomic sequencing data. (K) Comparison of enzyme activity (SOD, GSH, CAT) and MDA levels in ovarian tissues across groups. (L) Comparison of inflammatory cytokine levels in ovarian tissues across groups. AB: female rats treated with antibiotics to deplete gut bacteriome; POI: female rats treated with antibiotics to deplete gut bacteriome and injected intraperitoneally with cyclophosphamide (CTX) to induce POI; BH: POI rats transplanted with gut bacteriome from healthy reproductive-age women; BH + VH: POI rats transplanted with gut bacteriome and VLPs from healthy reproductive-age women. For A−D, G, K, and L: n  = 15 per group; For E, F, and H: AB = 11, POI = 5, BH = 9, BH + VH = 13; For I and J: n  = 5 per group. One-way ANOVA followed by Benjamini–Hochberg FDR–adjusted post hoc pairwise comparisons for (A, C, E−J, SOD, GSH, MDA in K, L); Chi-square test with Benjamini–Hochberg correction for (B, D); Kruskal–Wallis test followed by Benjamini–Hochberg FDR–adjusted pairwise Wilcoxon tests for (CAT in K). p values are presented as adjusted p values, with the corresponding unadjusted p values shown in parentheses. Principal component analysis (PCA) of transcriptome data showed that the ovarian gene expression profile in CTX-induced POI rats was distinctly different from that of other rats. The transplantation of healthy gut bacteriome and virome (BH + VH) significantly shifted this gene expression profile, making it resemble that of the normal rats with gut microbiota eliminated by antibiotics (AB) ( Figure 6I ; Figure S6G). In the ovarian tissues of CTX-treated POI rats, pro-inflammatory genes such as IL-6 and Ccl2 were significantly upregulated, while antioxidant genes such as Sod1 and Prdx4 were significantly downregulated, suggesting a marked increase in inflammation and oxidative stress. The healthy gut bacteriome and virome (BH + VH) reversed these abnormalities ( Figure 6J ; Figure S6H, S6I). Moreover, changes in these genes significantly correlated with the recovery of ovarian function in POI rats (Figure S6J). ELIZA and biochemical assays further showed that the healthy gut bacteriome and virome significantly reduced ovarian oxidative stress and inflammation in POI rats ( Figure 6K, 6L ). These findings demonstrate that modulation of ovarian oxidative stress and inflammation is a crucial mechanism by which the healthy gut virome ameliorates the pathological changes in POI rats.

This study was conducted in accordance with the principles of the Declaration of Helsinki and applicable local legislation with approval from the Ethics Committee of Dongying People’s Hospital (DYYX-2023-001; 2023.01.12) and Qilu Hospital of Shandong University (KYLL-2023ZM-461; 2023.02). Informed consent was obtained from all participants. Animal experiments complied with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) Guidelines and Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (NIH) and were approved by the Institutional Animal Care and Use Committee (IACUC) of Shandong University. Sample size was determined using G*Power software ( α  = 0.05, 1−β = 0.8). This study enrolled 30 newly diagnosed, treatment-naive idiopathic POI patients and 30 healthy reproductive-age women between February and October 2023, following the 2016 ESHRE guidelines. 25 POI patients were 24−40 years old with oligomenorrhea or amenorrhea for over four months and FSH > 40 IU/L. Control subjects were age-matched females with normal ovarian function. Exclusion criteria included recent gastrointestinal issues, chronic diseases, antibiotic treatment within 3 months, and autoimmune diseases requiring immunosuppressive therapy. Fecal samples were stored at −80 °C; serum was collected from fasting blood samples. Donor screening was conducted following the hospital’s standardized pathogen-testing protocol. Laboratory personnel were blinded to clinical status. Detailed protocols are provided in the Supplementary Materials & Methods. Demographic characteristics are presented in Table S1. Six-week-old female Sprague-Dawley rats were housed individually under standard SPF conditions with ad libitum access to food and water. All rats were stratified by body weight and randomized using computer-generated sequences. For the FVT experiment investigating the role of gut virome dysbiosis in POI, 90 rats were randomized into six groups (Ctrl, Ctrl + AB, BH, BH + VH, BH + VPOI, and BH + hkVPOI; n  = 15/group). For the therapeutic FVT experiment, 60 rats were randomized into four groups (AB, POI, BH, and BH + VH; n  = 15/group). Detailed housing conditions and group assignments are provided in the Supplementary Materials & Methods. Gut bacteriome and VLPs were isolated using an established protocol. Briefly, fecal samples were suspended in SM buffer, subjected to differential centrifugation to separate bacteria and VLPs, and sequentially filtered through 0.45 µm and 0.22 µm filters. VLPs were precipitated with PEG6000/NaCl (extraction efficiency > 80%). Bacteria and VLPs were enumerated using SYBR Gold staining and epifluorescence microscopy. Endotoxin was removed and monitored in VLP preparations for animal experiments. Detailed protocols are provided in the Supplementary Materials & Methods. Gut microbiota of female rats was depleted through daily oral gavage of an antibiotic cocktail for 14 days, 26 , 27 confirmed by 16S rRNA sequencing. For FMT, gut bacteriome from healthy women was gavaged at 4 × 10 8 /mL daily for one week. For FVT, gut VLPs from healthy women or POI patients were gavaged at 4 × 10 8 /mL every other day for 14 days. All VLP preparations were chloroform- and DNase I-treated and filtered through 0.22 µm membranes. Heat-killed VLPs (hkVPOI) served as controls. Donor screening followed standardized pathogen-testing protocols. Researchers were blinded to VLP sources. Detailed protocols are provided in the Supplementary Materials & Methods. The colonization of the gut virome after FVT was evaluated using the following methods. First, the abundance of gut VLPs was assessed using the SYBR Gold staining described above. Additionally, VLP sequencing was performed on fecal samples from both the recipient rats (BH + VH and BH + VPOI groups) at Day 21 and Day 42 during the experiment and donors (healthy reproductive-age women and POI patients). β -diversity analysis was conducted to assess the similarity of virome composition between groups. Finally, 16S rRNA gene sequencing was used to analyze the gut bacteriome composition of the rats in each group, investigating the impact of VLPs on the gut bacteriome and indirectly evaluating the colonization of the virome after FVT. The POI rat model was established using cyclophosphamide (CTX, Sigma, C0768-5G). On day -15, rats were given an intraperitoneal injection of CTX at 200 mg/kg, followed by daily injections of 8 mg/kg for 15 consecutive days. Subsequently, gut bacteriome depletion and FMT procedures were performed as described previously. In the experimental groups, POI model rats were gavaged with gut bacteriome (BH) or with both bacteriome and VLPs from healthy reproductive-age women (BH + VH). The control group (AB) received 0.9% saline instead of CTX and PBS for subsequent treatments. Genomic DNA was isolated from fecal samples using CTAB lysis, phenol extraction, and column-based purification. DNA integrity was assessed by agarose gel electrophoresis; concentration and purity were measured using a NanoDrop spectrophotometer. Detailed protocols are provided in the Supplementary Materials & Methods. Gut VLPs were treated with RNase A and DNase I, followed by Proteinase K/SDS treatment. DNA was extracted using phenol-chloroform-isoamyl alcohol and precipitated with sodium acetate/isopropanol. DNA quality was assessed by agarose gel electrophoresis and NanoDrop spectrophotometry. Detailed protocols are provided in the Supplementary Materials & Methods. The V3-V4 regions of the 16S rRNA gene were amplified and sequenced on an Illumina platform using the NEXTFLEX kit. Sequencing data underwent quality control using fastp (v0.23.2), and paired-end reads were merged using FLASH (v1.2.11). After demultiplexing and orientation adjustments, denoising was conducted using DADA2 in Qiime2 to yield Amplicon Sequence Variants (ASVs). After rarefying to 20,000 sequences per sample, 99.09% coverage was achieved. ASVs were taxonomically classified using the Silva database, and diversity analyzes were performed in R (v 4.2.0). VLP DNA libraries were constructed and sequenced on the Illumina NovaSeq platform (150 bp paired-end). After quality control and host sequence removal, data were assembled using MetaSPAdes. Viral sequences were identified using Genomad, 28 VirSorter2, 29 and VIBRANT, 30 with genome integrity assessed by CheckV. 31 High-quality viral genomes (≥90% completeness) were clustered into vOTUs at 95% similarity. Virus-host attribution, phage lifestyle, and auxiliary metabolic genes (AMGs) were determined using VPF-Class, 32 CHERRY, 33 PhaTYP, 34 and VIBRANT. 30 Horizontal gene transfer and virulence-associated genes were identified using HGTector2 35 and Virulence Factor Database (VFDB). 36 Analysis tools were selected based on community-accepted best practices for virome processing. Detailed protocols are provided in Supplementary Materials & Methods. A multi-layer decontamination workflow was applied to minimize potential host and bacterial contamination. Raw reads mapping to the human reference genome (GRCh38) and the SILVA database 37 were removed using Bowtie2. 38 Across all samples, host-derived reads accounted for a median of 0.8% (IQR 0.6–1.2%) and rRNA-derived reads were <0.05%. Following assembly, only high-confidence viral sequences (VirSorter2 categories 1-2) with <20% host contamination and ≥3 viral hallmark genes were retained. Plasmid-like sequences identified by VirSorter2 and CheckV were excluded ( n  = 1,288). After these stringent filtering steps, 551 high-confidence vOTUs remained for downstream analyzes. Total RNA was extracted from rat ovarian tissues using Trizol. cDNA libraries were constructed and sequenced on the Illumina platform. Clean reads were aligned using HISAT2, 39 gene expression was quantified using HTSeq 40 and normalized by FPKM. GO functional enrichment analysis was performed using ClusterProfiler 41 ; protein-protein interaction analysis used STRING 42 database and Cytoscape. Detailed protocols are provided in the Supplementary Materials & Methods. Causal structure models were constructed using the Peter–Clark algorithm ( α  = 0.05) in R. A structural equation model incorporating both direct and indirect paths was defined, and the standardized path coefficients were calculated using maximum likelihood estimation. Causal pathway diagrams were generated using the DiagrammeR package. Model fit and stability were evaluated using standard SEM diagnostics, including goodness-of-fit indices, bootstrap estimation of path coefficients, and inspection of residuals and modification indices. Sexually mature male and female SD rats, matched for age and weight, were paired overnight at a 1:1 ratio. Estrous cycles were monitored via vaginal lavage at 6 PM daily before pairing and at 8 AM the next morning after separation. Vaginal lavage fluid (100−200 µL) was stained with hematoxylin and eosin (H&E) and determined the estrous stage (Proestrus, Estrus, Metestrus, or Diestrus) based on cellular composition. Vaginal plugs were recorded before separating mated pairs; plugged female rats were designated as Day 1 of gestation, and females without vaginal plugs were remated. On pregnancy Day 20, rats were euthanized, and tissues collected. All behavioral assessments were double-blinded with good inter-rater reliability (ICC > 0.9). Detailed protocols are provided in the Supplementary Materials & Methods. H&E-stained liver and kidney sections were evaluated using a semi-quantitative scoring system (grades 1–4) by trained researchers. Liver scoring assessed hepatocyte architecture, sinusoidal changes, and necrosis. Kidney scoring assessed glomerular and tubular morphology, vascular congestion, and epithelial degeneration. Detailed scoring criteria are provided in the Supplementary Materials & Methods. Rat tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned. After deparaffinization and antigen retrieval in Tris-EDTA, sections were blocked with 5% BSA and treated with 3% H 2 O 2 . Primary antibodies were applied overnight at 4°C, followed by incubation with secondary antibodies at room temperature for an hour (Table S2). DAB (ZSGB-BIO, ZLI-9018) was used for color development, followed by hematoxylin counterstaining. Sections were then dehydrated, cleared, and mounted. Images were captured using a Carl Zeiss LSM710 confocal microscope (Carl Zeiss, Germany) or a Pannoramic SCAN II (3DHistech, Hungary). β -glucuronidase (GUS) activity was measured using spectrophotometry. Colonic contents were homogenized in PBS and centrifuged. Supernatant was incubated with 4-nitrophenyl β -D-glucuronide at 37°C, and absorbance was measured at 405 nm. Activity was calculated using a standard curve and normalized to protein concentration (BCA assay). Detailed experimental procedures are provided in the Supplementary Materials & Methods. Total RNA was isolated using TRIzol and reverse transcribed using PrimeScript RT Kit. qPCR was performed using TB Green Premix Ex Taq on a LightCycler 480 system with primers listed in Table S3. Gene expression was quantified using the 2 −ΔΔCt method. Detailed experimental procedures are provided in the Supplementary Materials & Methods. Inflammatory cytokines (IL-1β, IL-6, MCP-1, TNF- α ) were measured in rat serum, colon, and ovarian tissues. Oxidative stress markers (MDA, CAT, GSH, SOD) were quantified in ovarian tissues. Serum was obtained by cardiac puncture; tissue homogenates were prepared with protease inhibitor cocktail. Kit information is provided in the Supplementary Materials & Methods. Data are presented as mean ± SD or median with interquartile range (violin plots). Normality was assessed using the Shapiro-Wilk test. For two-group comparisons, Student's t -test or Wilcoxon rank-sum test was used as appropriate. For multiple-group comparisons, one-way ANOVA with false discovery rate (FDR; Benjamini–Hochberg)–adjusted post hoc pairwise comparisons or the Kruskal–Wallis test with FDR-adjusted pairwise Wilcoxon tests were applied. Multiple testing was corrected using Benjamini-Hochberg (FDR < 5%). Partial Spearman correlation analyzes were performed to assess virus–hormone associations while adjusting for age and BMI. Chi-square tests followed by Benjamini–Hochberg correction for multiple comparisons. PCA/PCoA significance was tested using Adonis or ANOSIM. Virus-bacteria networks were inferred using SPIEC-EASI. Cliff's Delta effect sizes with 95% CIs were computed for key indicators. Analyzes were performed using R v4.2.0 and GraphPad Prism (v9.4.1). Statistical significance was set at p  < 0.05 or adjusted p  < 0.05. Detailed procedures are provided in the Supplementary Materials & Methods.

This investigation revealed characteristic changes in the gut virome of POI patients. We demonstrated that dysbiotic gut virome induced HPO axis hormone dysregulation and POI-associated symptoms in recipient rats, primarily through remodeling the gut bacteriome (particularly the estrobolome), impairing intestinal barrier, and enhancing inflammation and oxidative stress. Moreover, gut VLPs from healthy reproductive-age women significantly ameliorated the condition of POI rats. These findings underscore the potential role of gut virome in the pathogenesis and treatment of ovarian dysfunction and reproductive diseases, particularly POI. The viral community in the human gut significantly outnumbers the bacterial population, with over 90% of these viruses being bacteriophages. 46 However, the hosts and functions of most gut phages remain largely unknown. 47 Previous studies employing the historical Caudovirales classification ( Siphoviridae , Podoviridae , and Myoviridae ) have reported these taxa as dominant in the human gut. Our findings are consistent with this ecological pattern when applying the updated Caudoviricetes taxonomy (e.g., Straboviridae , Casjensviridae , Herelleviridae , and Peduoviridae ). Our results revealed that over 90% of gut viruses were double-stranded DNA viruses, of which more than 94% were phages, corroborating previous studies. 48 The genome sizes of gut bacteriophages mainly ranged from 50 K to 100Kbp, possibly reflecting their adaptability to the gut environment. Additionally, the composition of gut bacteriophages exhibited substantial inter-individual variability, consistent with earlier reports. 49 This variability may be related to differences in individual diet, lifestyle, or gut microbiota composition, reflecting the co-evolutionary adaptation of phages and host bacteria to each person’s gut environment. 50 , 51 Our results indicated that phage genome sizes varied significantly and correlated with the genome sizes of their bacterial hosts, reflecting the deep evolutionary and functional connection between phages and their host bacteria. This correlation highlights the need for phages to match the complexity of their hosts to effectively infect, control, and utilize host resources. Furthermore, the high abundance of β -lactamase genes in phage genomes underscores the critical role of phages in the transmission of bacterial antibiotic resistance. However, this finding requires confirmation through transcript-level analyzes or functional assays in future studies. Extensive research has demonstrated the crucial role of the gut microbiota in regulating HPO axis hormones and ovarian function. The composition of the gut microbiota has been reported to directly affect ovarian function, including follicle development and ovarian weight. More than 60 genera in the human gut bacteriome express β -glucuronidase (GUS), collectively known as the “estrobolome”, 52 which increases estrogen bioavailability. 16 , 53 Additionally, recent research has confirmed that Eggerthella lenta in the gut can produce progesterone. 54 Consequently, gut dysbiosis is associated with ovarian dysfunction and HPO axis hormone abnormalities, which are implicated in various female reproductive disorders, such as PCOS, endometriosis, adverse pregnancy outcomes, and infertility. 55-57 Although gut viruses/phages vastly outnumber bacteria, their characteristic changes and relationships with ovarian dysfunction, HPO axis hormone imbalances, and related diseases remain elusive. We found that the dysbiotic gut virome is accompanied by reduced estrobolome abundance and lower GUS activity, suggesting that the gut virome may contribute to POI development through effects on the estrobolome and GUS activity. However, this mechanism requires future validation through GUS inhibition or related mediation experiments. Our study revealed characteristic changes in the gut virome of POI patients, including significantly reduced diversity and markedly altered community composition, particularly crAss-like phages and phages targeting estrobolome bacteria. We identified several gut bacteriophages that were significantly correlated with E2, progesterone, LH, and FSH levels, primarily from Straboviridae , Herelleviridae , Peduoviridae , and crAss-like phages. These findings align with previous observations in PCOS patients. 18 We found that female rats treated with antibiotics exhibited HPO axis hormone dysregulation, impaired mating behavior, and ovarian function, further underscoring the important role of the gut microbiota in maintaining HPO axis hormone homeostasis and ovarian function. Notably, our study advances beyond existing PCOS research in several key aspects. We provide the first systematic characterization of the gut virome in POI patients and employed VLP deep sequencing to establish associations between gut virome and HPO axis hormones. More importantly, through FVT, we moved from observational association to causal evidence, demonstrating that POI-derived gut virome can transfer disease phenotypes, whereas the healthy women gut virome ameliorates ovarian dysfunction. Both inflammatory enhancement and metabolic dysregulation play crucial roles in ovarian dysfunction, including POI. 58 Increased inflammatory factors (e.g., IL-1β and IL-6) lead to ovarian dysfunction and follicle loss. 59 Our results suggested that FVT modulated the immune response via the gut-ovary axis. The gut virome in POI patients exacerbated ovarian inflammation by activating the IL-6/IL-1β-Duox2 axis, which increases ROS production and weakens antioxidant capacity, thereby establishing a vicious cycle of inflammatory-oxidative stress. Conversely, the healthy virome exerts anti-inflammatory and antioxidant effects, while also activating xenobiotic response and cellular repair pathways, thus demonstrating therapeutic effects. However, activation of the IL-6/IL-1β-Duox2 axis is based on correlative inference, and future targeted or orthogonal assays will be required to validate this pathway. Additionally, many POI patients exhibit marked insulin resistance and lipid metabolism abnormalities. 60 Emerging evidence suggests that alterations in the gut virome, particularly the phageome, are closely associated with metabolic diseases such as obesity and type 2 diabetes. 61 Gut virome from healthy mice have been shown to reduce weight gain and normalize glucose tolerance in high-fat diet-induced obese mice. 62 These findings suggest a potential link between gut virome-mediated metabolic regulation and host ovarian function, yet the causal nature of this association remains to be elucidated. Therefore, further investigation is warranted to determine specific phage-bacterial interactions through which host metabolism and immunity are modulated, ultimately regulating ovarian function. Bacteria-phage interactions are crucial for maintaining gut microbiome homeostasis and host health. Phages can confer unique properties to their bacterial hosts through horizontal gene transfer (HGT), such as transferring AMGs, thereby influencing bacterial survival. 63 Research has shown that AMGs involved in sulfate metabolism are enriched in the gut phageome of centenarians, enhancing bacterial sulfur compound metabolism and maintaining gut barrier integrity. 64 In the gut phageome of POI patients, we observed a significantly increased diversity of AMGs, with a notable enrichment of AMGs involved in energy and nucleic acid metabolism but a reduction in AMGs associated with the metabolism of amino acid. These changes may reflect metabolic adaptations of the gut microbiome to the diet and intestinal environment of POI patients. Additionally, we found increased abundance of antibiotic resistance genes and virulence factors, suggesting a higher potential for antibiotic resistance and pathogenicity in the POI gut microbiota. 65 Such gut microbiota has a greater ability to disrupt the intestinal mucosal barrier, likely promoting systemic inflammation, which is a key mechanism in POI pathogenesis. However, these AMG- and HGT-related functional inferences remain largely descriptive. Their biological relevance requires validation using experimental approaches such as metatranscriptomics, reporter assays, or functional biochemical characterization. Our study also demonstrated that the relative abundances of gut bacteriophages were predominantly positively correlated with their bacterial hosts, consistent with previous studies. 66 However, certain estrobolome bacteria (such as Bacteroides , Akkermansia , and Faecalibacterium ) and their phages ( Straboviridae , crAss-like phages) showed a shift from positive correlations in healthy reproductive-age women to negative correlations in POI patients. This apparent paradox likely reflects indirect ecological interactions, such as niche release following phage-mediated lysis of competing taxa, rather than direct phage-host co-evolution. These patterns suggest altered ecological networks in POI, although they do not definitively establish causal predator-prey relationships. Despite ongoing discussions, experimental data elucidating the extent to which gut phages modulate bacterial communities remain limited. Our results showed that co-transplantation of gut bacteriome and VLPs from healthy reproductive-age women more significantly restored gut bacteriome diversity in antibiotic-treated rats compared to transplantation of gut bacteriome alone, achieving the highest similarity to donors. Conversely, gut VLPs from POI patients dramatically altered the gut bacteriome composition in recipient rats transplanted with gut bacteriome of healthy women, making it more akin to the dysbiotic gut bacteriome of POI patients. This observation suggests that the gut virome of POI patients effectively reshapes the gut bacteriome of healthy women into a POI-like dysbiotic state. A recent study reported that transplanting fecal VLPs from mice with enriched gut Akkermansia muciniphila significantly enhanced fertility in recipient mice. 19 We found that FVT from donors with low abundance of gut Akkermansia and Bacteroides (e.g., POI patients) impaired the colonization of these bacteria in recipients, whereas FVT from donors with high abundance of these bacteria (e.g., healthy reproductive-age women) facilitated their colonization. Previous studies have reported a significant reduction in Faecalibacterium in the gut in POI patients and Akkermansia in premature ovarian insufficiency mice. 11 , 14 Our findings provide mechanistic insight into these reported changes and emphasize the significant influence of gut phages on bacterial communities, highlighting their potential roles as key drivers for gut microbiota dysbiosis. Nonetheless, fully elucidating the microbial dependence of VLPs will require further experiments, such as antibiotic-treated rats receiving VH transplantation, or gnotobiotic models with defined bacterial colonization followed by VH transplantation to systematically dissect the causal relationships underlying phage–bacterial host interactions. Recently, accumulating evidence has highlighted the therapeutic potential of bacteriophages for non-infectious diseases. Phage cocktails targeting pathogenic bacteria, such as Klebsiella pneumoniae , Escherichia coli , and cytolytic Enterococcus faecalis , have demonstrated efficacy in mouse models of inflammatory bowel disease, Crohn's disease, and alcoholic liver cirrhosis. 67-69 Additionally, nanosilver-M13 phage complexes specifically eliminated Fusobacterium nucleatum in the gut, thereby enhancing the efficacy of immune checkpoint inhibitors in colorectal cancer treatment. 70 However, the therapeutic potential of gut bacteriophages for diseases related to the HPO axis and ovarian dysfunction has remained unclear until now. Our study showed that gut VLPs from healthy reproductive-age women significantly alleviate abnormalities in HPO axis hormone levels and ovarian function, thereby markedly improving the condition of POI rats. Notably, circulating E2 levels in healthy-VLP-treated rats increased to approximately 62.3 pg/mL, exceeding the commonly used clinical threshold of 50 pg/mL for defining hypoestrogenism, 71 thereby supporting the translational relevance of this hormonal improvement. These improvements were associated with increased abundance of gut estrobolome bacteria and reduced inflammation. These findings demonstrate the therapeutic potential of healthy gut phages for treating reproductive endocrine disorders related to HPO axis and ovarian dysfunction. Nevertheless, the limited observation window in this study precluded evaluation of long-term virome engraftment. Future studies employing longitudinal designs with extended follow-up are needed to assess engraftment durability. Similar to FMT, the potential risk of transmitting pathogens or unidentified eukaryotic viruses with FVT is a matter of concern. In this study, all donor samples were obtained from individuals without gastrointestinal symptoms who tested negative for common pathogens, including HIV, HBV, HCV, and HPV, in hospital clinical laboratories following standardized protocols. Routine quality controls ensured high analytical sensitivity (<10–100 copies/mL, well below clinical infection thresholds), utilizing spike-in controls to monitor extraction and amplification efficiency, as well as no-template and negative-serum controls to exclude cross-contamination. All procedures complied with national clinical testing standards and ISO 15189 quality management requirements. Moreover, we employed rigorous steps during VLPs extraction, which also removed a portion of eukaryotic viruses. Additionally, multiple studies have also shown that the proportion of eukaryotic viruses in the human gut is typically low (averaging less than 1.5%). Pathogenic eukaryotic viruses are rarely present in individuals without obvious gastrointestinal symptoms. 72-74 Furthermore, we detected only four eukaryotic viral genera in this study. Their average abundance was very low, and no viruses known to infect humans or mammals were detected (Table S6). The rats receiving FVT also did not exhibit obvious pathological changes in liver, kidney, and circulating lymphocytes compared to normal rats. These findings demonstrate the safety of FVT, which should be at least comparable to FMT currently employed in clinical practice. Nevertheless, our current understanding of viruses is still limited, and there may still be viruses in the gut that are either not identified or newly emerged (such as SARS-CoV-2). Therefore, the safety described here is only relative, and multicenter clinical trials are necessary to assess its safety and effectiveness. Additionally, the clinical application of FVT faces a series of technical and ethical challenges. In our cohort, fewer than 20% of candidate donors passed screening, underscoring the challenge of identifying eligible donors. Our batch-release criteria (endotoxin <0.5 EU/mL, pathogen-negative) and the necessity for viral mutation surveillance highlight the importance of standardized, GMP- compliant FVT protocols. Ethical concerns, including the protection of donor privacy, also warrant special attention. At the regulatory level, most countries have not yet developed dedicated policies for FVT. Nevertheless, the current knowledge from FMT clinical applications provides valuable insights for FVT. 75 , 76 Considering the absolute dominance of phages in the gut virome, we propose that FVT could be optimized as targeted gut phage transplantation (tGPT) based on specific phages. This strategy could fundamentally enhance the accessibility, standardization, and safety of FVT, while simultaneously reducing ethical concerns. To further strengthen its translational feasibility, tGPT would require well-defined phage selection criteria (host specificity, safety, genomic stability), optimized dosing and delivery strategies (e.g., enteric-coated capsules or rectal infusion), and resistance surveillance through periodic monitoring of bacterial CRISPR-Cas activity and receptor mutations. An early clinical roadmap could include phase I trials focused on safety with adverse event incidence as the primary endpoint. Phase II trials should assess efficacy with menstrual cyclicity restoration as the primary endpoint and hormonal parameters (E2, FSH, AMH) as secondary endpoints. Safety gates and stopping rules include SAE rates exceeding 10%, occurrence of ≥Grade 3 treatment-related SAEs, or primary endpoint response rates <10%. Notably, several FMT-based therapies have recently received FDA approval for clinical use, 77-79 lending cautious optimism to the future clinical application of FVT. The CTX-induced POI model recapitulates key pathological features observed in human POI patients, including decreased ovarian reserve, enhanced oxidative stress and inflammation, impaired follicular development, and altered sex hormone levels, and thus has been widely used. 80-82 However, this model rapidly induces ovarian dysfunction through pharmacological damage, simulating an acute ovarian damage process. This differs from the chronic progression of human POI. Moreover, the CTX-induced model cannot fully recapitulate the complex etiology of human POI, including genetic, autoimmune, metabolic, and environmental factors. 83-87 Therefore, caution must be exercised when extrapolating findings from animal models to human disease pathology. Nevertheless, given the absence of animal models that can completely simulate human POI, the CTX-induced POI model remains one of the most commonly used models for investigating ovarian dysfunction and therapeutic interventions, providing an effective tool to explore the relationship between gut virome and POI. Consequently, while careful interpretation is warranted, our animal experimental findings provide novel insights into the pathogenesis of POI and the development of microbiota-based therapeutic strategies. Future research should focus on developing animal models that more comprehensively replicate the human pathological process, thereby offering more robust tools for POI research. CTX is a commonly used chemotherapeutic agent that significantly alters the diversity and composition of the gut microbiota, such as reductions in Bifidobacterium and Lactobacillus , and increases in Firmicutes and Proteobacteria. 88 Human pathological states, such as hormonal levels, immune function, and metabolic status, also substantially influence the gut microbiome. 89 , 90 These effects involve multiple mechanisms, including direct cytotoxic effects of CTX on gut microbes, alterations in gut pH, oxygen, or nutrient levels, intestinal barrier damage, and host immune dysfunction. 89 , 91 Since bacteria serve as hosts for phages, changes in the gut bacteriome inevitably affect the composition and function of the gut phageome. 43 , 92 , 93 However, research on the gut virome remains limited compared with that on the gut bacteriome. The effects of drugs and host pathological states on gut microbiota are often confounded and distinguishing them is crucial for elucidating their respective mechanisms of action. In our experimental design, CTX was administered to rats in which the microbiota had been depleted by antibiotics but not reconstituted by microbial transplantation. This approach allowed us to isolate the acute host injury induced by CTX, rather than effects attributable to specific microbial grafts. Nevertheless, future studies should incorporate longitudinal sampling at multiple timepoints following FVT to distinguish acute CTX-induced injury from sustained therapeutic effects mediated by the virome. These investigations are important for understanding mechanisms of drug action, reducing adverse effects, and developing novel therapeutic strategies. In conclusion, our study revealed characteristic changes in the gut virome of POI patients and suggested that it may play an important role in the pathogenesis of POI, providing novel insights into the pathogenesis of POI. We also demonstrated the potential of phages for treating ovarian dysfunction-related diseases, including POI. Several limitations should be acknowledged. First, the relatively limited sample size may restrict the generalizability and translational relevance of our findings. Future studies should expand sample size through multicenter cohorts and pursue validation in non-human primate models or early-phase clinical trials. Importantly, this study was based on a cohort of reproductive-age women from Northern China with idiopathic POI, the generalizability of our findings to POI patients of other etiological types (e.g., autoimmune or infectious), ethnicities, and geographical regions require further investigations. Future work may also incorporate autoimmune or genetic POI models to better capture disease heterogeneity. In addition, detailed dietary information (e.g., dietary patterns, and food processing level) and genetic background of fecal donors were not included in the analysis. Therefore, future studies should incorporate standardized dietary records and host genotyping of relevant genetic variants (e.g., CYP19A1) to control for potential confounding influences on the gut virome. Furthermore, because oligomenorrhea or amenorrhea (≥4 months) is a defining feature of POI, cycle-based sampling would substantially limit enrollment and potentially confound disease-specific signals; therefore, menstrual cycle standardization was not applied. Future studies should account for hormonal fluctuations by including amenorrhea duration and serum hormone levels (FSH, E2, and AMH) as covariates. Despite these limitations, this research significantly advances our understanding of the regulation of HPO axis and ovarian function and opens new avenues for developing novel treatments for female infertility disorders.

No potential conflict of interest was reported by the author(s).

Supplementary materials-v1224_v1_clean.docx completed_ARRIVE_checklist.docx