Gut Microbiota Modulation by Inulin Improves Metabolism and Ovarian Function in Polycystic Ovary Syndrome

To investigate the effects of inulin on metabolism and ovarian function in PCOS, we induced PCOS‐like phenotypes in 3‐week‐old C57BL/6 female mice using dehydroepiandrosterone (DHEA)^[16^] or in combination with a high‐fat diet (HFD)^[30^] (Figure S1A, Supporting Information). We observed that only PCOS‐like mouse model induced by the combination of DHEA and HFD exhibited significant metabolic abnormalities, including weight gain (Figure S1B, Supporting Information), impaired glucose tolerance (Figure S1C,D, Supporting Information), and insulin resistance (Figure S1E,F, Supporting Information). In the absence of HFD, DHEA treatment alone induced abnormal ovarian morphology (Figure S1G,H, Supporting Information), disrupted estrous cycle (Figure S1I, Supporting Information), and hyperandrogenemia (Figure S1J, Supporting Information) in mice. Given the high prevalence of metabolic dysfunction in clinical PCOS,^[4^] the PCOS‐like mouse model established by the combination of DHEA and HFD was selected for subsequent evaluations of the comprehensive effects of inulin on PCOS. Mice without inulin treatment were designated as PCOS_Ctr, while those treated with inulin in drinking water were designated as PCOS_Inu. A group of mice without induction of PCOS was designated as the normal control (NC) (Figure1A). To determine the optimal inulin dosage, we tested three concentrations: low (2% w/v), medium (4% w/v), and high (8% w/v) inulin in drinking water administered to PCOS mice (Figure S2A, Supporting Information). The low dosage did not significantly improve glucose levels in PCOS mice in the glucose tolerance test (GTT, Figure S2B,C, Supporting Information) and insulin tolerance test (ITT, Figure S2D,E, Supporting Information). In contrast, both the medium and high dosages significantly improved glucose metabolism in PCOS mice, with similar effects observed between the two dosages (Figure S2B–E, Supporting Information). Following the principle of using the minimum effective dosage, 4% (w/v) inulin was chosen for subsequent experiments.

Compared with the NC mice, PCOS_Ctr mice showed a significant increase in body weight (Figure 1B). Although inulin treatment reduced body weight in PCOS_Inu mice compared to PCOS_Ctr mice, the difference was not statistically significant (Figure 1B). Inulin treatment significantly improved glucose tolerance in PCOS_Inu mice compared to non‐treated PCOS_Ctr mice (Figure 1C,D). Similarly, inulin treatment improved insulin sensitivity assessed by ITT (Figure 1E,F). Moreover, inulin treatment lowered the fasting insulin levels (Figure 1G). Next, we assessed ovarian histopathology in the three groups. PCOS_Ctr mice showed an increased number of cystic follicles and a decreased number of corpora lutea, whereas the ovaries of NC and PCOS_Inu mice exhibited follicles at different developmental stages and a normal number of corpora lutea (Figure 1H,I). Importantly, inulin treatment restored serum testosterone levels to normal (Figure 1J). Moreover, inulin treatment reduced the levels of antimullerian hormone (AMH, Figure 1K) and luteinizing hormone (LH, Figure 1L). Although follicle‐stimulating hormone (FSH) levels were not significantly affected by inulin (Figure 1M), the LH‐to‐FSH ratios (Figure 1N) were significantly lower in PCOS_Inu mice compared to PCOS_Ctr mice. To evaluate ovarian function, we performed estrous cycle testing and found that PCOS_Ctr mice exhibited disrupted estrous cycles, whereas NC and PCOS_Inu mice displayed regular estrous cycles (Figure 1O,P).

It was reported that brown adipose tissue (BAT), a key thermogenic organ, can effectively rescue DHEA‐induced PCOS phenotypes in rats^[31^] and BAT activation is considered a potential therapeutic target for PCOS.^[32^] To investigate the role of BAT in our study, we analyzed thermogenic markers in BAT from the mice using qPCR. The results showed that relative expression levels of Pgc1α, Pparα, and Dio2 were significantly lower in PCOS_Ctr mice as compared to the NC mice. In contrast, inulin treatment significantly upregulated the expression of Ucp1, Pgc1α, Pparα, Dio2, and Cited1 (Figure 1Q). Furthermore, PCOS_Ctr mice exhibited lipid accumulation, as evidenced by enlarged lipid droplets in peri‐ovarian adipose tissue, while inulin treatment effectively reversed this phenotype (Figure 1R,S). Taken together, our findings indicate that inulin treatment significantly improves metabolic dysregulation and ovarian function in PCOS mice.

Since inulin, as a common dietary fiber, is known to influence gut microbiota composition,^[20^] we next performed high‐throughput 16S rRNA gene sequencing to profile gut microbial taxonomic signatures in each group of mice. At the phylum level, PCOS_Ctr mice exhibited a higher abundance of Firmicutes and a lower abundance of Bacterioidetes (Figure2A), along with a significantly increased Firmicutes‐to‐Bacterioidetes (F/B) ratio (Figure 2C). Inulin treatment reversed these alterations. At the genus level, the dominant genera differed significantly among the three groups (Figure 2B). Compared with NC mice, PCOS_Ctr mice exhibited gut dysbiosis, characterized by a decreased abundance of Muribaculaceae and Allobaculum, and an increased abundance of Lachnospiraceae (Figure 2B; Figure S3A,B, Supporting Information). It is noteworthy that inulin administration increased the abundance of Muribaculaceae (Figure 2B) and five genera associated with the bacteria known to produce SCFAs, including Allobaculum, Bacteroides, Akkermansia, Alloprevotella, and Bifidobacterium, as identified by Linear discriminant analysis effect size (LEfSe) (Figure S3A,B, Supporting Information).

When assessing microbial community diversity, we found no significant difference in alpha diversity between NC and PCOS_Ctr mice by the Shannon index. However, inulin treatment resulted in a significant decrease in the Shannon index in PCOS_Inu mice (Figure 2D). Interestingly, analysis using the Ace and Chao indices revealed that both PCOS_Ctr and PCOS_Inu mice had significantly lower microbial community richness compared with NC mice, with PCOS_Inu mice showing the lowest richness (Figure 2E,F). Based on the principal coordinate analysis (PCoA) plot of amplicon sequence variants (ASVs), there were significant differences in overall gut microbiota structure among the three groups (P = 0.001; R‐squared = 0.717) (Figure 2G). Notably, we observed shorter distances between samples from the NC and PCOS_Inu groups, as measured by weighted UniFrac distances (Figure 2H) or Jaccard distances (Figure S3C, Supporting Information), indicating that the gut microbiota structure in PCOS_Inu mice was more alike to that in NC mice.

The microbial functional profiling was further analyzed by metagenomic sequencing. Carbohydrate‐active enzyme (CAZy) family genes,^[33^] which encode carbohydrate‐active enzymes that degrade complex polysaccharides like dietary fiber into fermentable substrates, are essential for SCFAs production. Principal coordinate analysis (PCoA) based on the Bray‐Curtis distances of CAZy genes indicated significant differences among groups by Adonis (P = 0.001; R‐squared = 0.779) (Figure 2I). In addition, shorter distances were observed between samples from the NC and PCOS_Inu groups, based on the Bray‐Curtis distances or Jaccard distances of CAZy genes (Figure 2J; Figure S3D, Supporting Information), and Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs (KOs) (Figure S3E,F, Supporting Information). These results indicated that the gut microbial function of PCOS_Inu mice closely resembles that of NC mice. Moreover, the total abundance of CAZy genes was significantly lower in the PCOS_Ctr mice compared with NC mice, but inulin treatment significantly increased CAZy gene abundance in the gut microbiota (Figure 2K). Notably, CAZy genes involved in inulin metabolism were significantly enriched in PCOS_Inu mice (Figure 2L). Focusing on SCFAs metabolism, we annotated key enzymes involved in the production pathways of AA, PA, and butyric acid (BA), as previously described.^[34, 35^] The gut microbiota in PCOS_Ctr showed significantly decreased gene abundance related to BA production (Figure 2O), particularly the gene encoding butyryl–coenzyme A (butyryl‐CoA): 4‐hydroxybutyrate CoA transferase (4Hbt) (Figure S3G, Supporting Information). The production of AA, PA, and BA was promoted by inulin supplementation (Figure 2M–O). In addition, discriminatory genera enriched by inulin exhibited significantly positive correlations with AA and PA production (Figure S4, Supporting Information). Beyond genomic data, we measured SCFAs in cecal contents. Compared with NC mice, PCOS_Ctr mice had significantly lower total SCFAs (Figure 2P) and PA (Figure 2Q). In contrast, inulin enhanced total SCFAs in the cecum of PCOS_Inu mice (Figure 2P), with significant increases in PA and BA (Figure 2Q). Spearman's correlation analysis revealed that inulin‐enriched bacteria, such as Bifidobacterium, Alloprevotella, Allobaculum, and Muribaculaceae, were significantly positively correlated with AA and total SCFAs (Figure 2R).

In addition, we performed an alignment analysis based on a comprehensive antibiotic resistance database (CARD)^[36^] and a virulence factors database (VFDB).^[37^] From volcano plots, PCOS_Ctr mice showed more up‐regulated antibiotic resistance genes (ARGs) and virulence factor genes (VFGs) than NC mice, whereas inulin down‐regulated ARGs and VFGs in PCOS_Inu mice (Figure S5, Supporting Information). Using LEfSe for level‐three pathways in KEGG database, we found that numerous crucial metabolic pathways associated with carbohydrates, amino acids, and nucleic acids were mainly enriched in NC and PCOS_Inu mice, while pathways related to infection were enriched in PCOS_Ctr mice (Figure S6, Supporting Information). Overall, inulin altered the comprehensive function of the gut microbiota in PCOS mice, featured by significantly enhancing microbial capacity for carbohydrate metabolism and facilitating SCFAs production.

To elucidate the potential interactions in the symbiotic ecosystem of gut microbiota, we constructed a microbial co‐abundance network among species shared at least 80% of all samples. The SparCC correlation analysis and PERMANOVA were used to identify 42 co‐abundance groups (CAGs), visualized as a topological lotus diagram (Figure3A). We found that CAG12 was the group with the highest abundance and diversity of species, and it closely interacted with other CAGs, suggesting that CAG12 was an important core gut microbial community in the mice. Interestingly, several Bifidobacterium species were found within CAG12, and many species belonging to the Muribaculaceae family were also present in CAG12. Additionally, we observed a highly significant positive correlation (blue line) between CAG12 and CAG16. High abundances of Muribaculaceae and Bifidobacterium also existed in CAG16. Moreover, CAG16 contained high abundances of beneficial bacteria such as Parabacteroides distasonis^[38^] and Bacteroides thetaiotaomicron.^[39^] Meanwhile, CAG12 exhibited distinctly negative correlations (pink line) with CAG2 and CAG32, both enriched with species belonging to the Lachnospiraceae family. Importantly, compared to NC mice, the abundances of CAG12 and CAG16 were significantly decreased in PCOS_Ctr mice. Conversely, CAG2 and CAG32, which were negatively correlated with CAG12, were significantly enriched in PCOS_Ctr mice. However, inulin significantly promoted an increase in CAG12 and CAG16, accompanied by a decrease in CAG2 and CAG32 in PCOS_Inu mice. Additionally, CAG20 (represented by Muribaculaceae) and CAG17 (represented by Bacteroides acidifaciens) were significantly enriched in PCOS_Inu mice (Figure 3B). Collectively, inulin modulated the composition and interactions of microbiota within the gut ecosystem in PCOS mice. The complete matrix of the CAGs analysis is provided in Supplementary Data1.

The levels of SCFAs in the intestine have been strongly associated with the integrity of the intestinal mucosal barrier and susceptibility to inflammation.^[40^] Therefore, we utilized qPCR (Figure4A–C) and immunohistochemistry (Figure 4D–G) to examine the transcription and expression of the genes encoding for tight‐junction proteins, including Zo1, Occludin, and Claudin1, from the colonic tissues of mice in the study. There was a significant decrease in the expression levels of tight‐junction genes and proteins in PCOS_Ctr mice, which were upregulated by inulin treatment (Figure 4A–G). These data indicate that inulin treatment could markedly inhibit the integrity disruption of the intestinal barrier in PCOS mice.

Given the crucial role of intestinal barrier integrity in preventing the translocation of bacterial lipopolysaccharide (LPS) into systemic circulation,^[41^] we measured serum LPS‐binding protein (LBP), reflecting circulating LPS level. Consistent with the reduced expression of tight‐junction‐related genes and proteins observed in PCOS_Ctr mice, serum LBP levels were significantly elevated in PCOS_Ctr mice compared to both NC and PCOS_Inu mice (Figure 4H). Serum levels of pro‐inflammatory cytokines interleukin‐1 beta (IL‐1β) and interleukin‐18 (IL‐18), were also significantly elevated in PCOS_Ctr mice compared to both NC and PCOS_Inu mice (Figure 4I,J). To investigate the interplay between inulin and inflammation in PCOS, we administered LPS intraperitoneally to inulin‐treated PCOS mice, designated as PCOS_Inu+LPS. (Figure S7A, Supporting Information). Although LPS exposure did not result in significant growth in body weight (Figure S7B, Supporting Information), it induced dysglycemia (Figure S7C,D, Supporting Information), insulin resistance (Figure S7E,F, Supporting Information), increased cystic follicles and decreased corpora luteum (Figure S7G,H, Supporting Information), and disrupted estrous cycle (Figure S7I, Supporting Information) in PCOS_Inu+LPS mice. Notably, PCOS_Inu+LPS mice presented chronic low‐grade inflammation phenotype, similar to PCOS_Ctr mice, as evidenced by elevated serum levels of LBP, IL‐1β, and IL‐18 (Figure S7J–L, Supporting Information) and upregulated ovarian mRNA expression levels of Lbp, Il1b, and Il18 (Figure S7M, Supporting Information). These results demonstrate that LPS administration reverses inulin‐mediated amelioration of PCOS pathology, highlighting the significance of intestinal barrier integrity in suppressing inflammation.

Inflammation has been considered as one of the important factors in PCOS pathogenesis,^[42, 43^] however, it is not clear if innate immunity is involved in the inflammatory process in PCOS. The LEfSe analysis revealed significant enrichment of the NOD‐like receptor (NLR) signaling pathway^[44^] in PCOS_Ctr mice (Figure S6, Supporting Information). Thus, we assessed ovarian mRNA expression levels of Nlrp3, Nlrp1, and Nlrc4, among which the expression of Nlrp3 mRNA levels was upregulated significantly in both PCOS_Ctr and PCOS_Inu+LPS mice (Figure S7N, Supporting Information). Next, we assessed the upstream signaling molecules, including toll‐like receptor 4 (TLR4), myeloid differential protein 88 (Myd88), phosphorylated‐nuclear factor (NF)‐κB (p‐p65), as well as downstream pyroptosis‐related proteins, including NLR pyrin domain containing protein 3 (NLRP3), apoptosis‐associated speck‐like protein containing CARD (ASC), cysteinyl aspartate specific proteinase 1 (Caspase1), and gasdermin D (GSDMD) in ovarian tissue by western blot. We found that the expression levels of TLR4, Myd88, and p‐p65/p65 in the ovarian tissue of PCOS_Ctr mice were significantly increased compared to both NC and PCOS_Inu mice (Figure 4K,L). In addition, compared with NC mice, PCOS_Ctr mice showed increased expression levels of inflammasome NLRP3/ASC/cleaved‐Caspase‐1 and cleaved‐GSDMD, suggesting the activation of the pyroptosis pathway. However, inulin treatment significantly mitigated activation of the NLRP3 inflammasome and suppressed pyroptosis in the ovarian tissue of PCOS_Inu mice (Figure 4M,N). Our results suggest that the inulin treatment improves the integrity of the intestinal barrier and suppresses ovarian inflammation in PCOS mice.

To further investigate whether inulin has the equally health benefits for patients with PCOS, we performed a prospective self‐controlled clinical trial. Women diagnosed with PCOS, based on the 2003 Rotterdam criteria,^[45^] were enrolled in this study and received 10 g of inulin daily for three months. A total of 45 subjects were included in the study under strict inclusion and exclusion criteria, with a mean age of 29.53 ± 3.06 (Figure5A). As shown in Table1, the subjects had significantly lower sex hormones, including testosterone, dehydroepiandrosterone sulfate (DHEAs), and AMH at the end of the study. Moreover, the subjects had improved glucose metabolism at the end of the study, including lowered fasting blood glucose (FBG), fasting insulin (FIN), blood insulin level at the 2‐hour time point of the oral glucose tolerance test (INS‐2 h), and homeostatic model of assessment of insulin resistance (HOMA‐IR). In addition, there was a significant decline in body mass index (BMI) and total cholesterol (TC) levels as well as a decreased trend in low‐density lipoprotein (LDL). Overall, the metabolic status of study subjects improved significantly after inulin intervention.

It is known that gut microbiota play an important role in the host metabolism, to explore the effect of inulin, a soluble fiber, on the gut microbiota of the women with PCOS, we collected fecal samples from the study subjects at three time points, including pre‐inulin intervention (Pre), 1 month post‐inulin intervention (Post_1M) and 3 months post‐inulin intervention (Post_3M), and performed the 16S rRNA gene sequencing. Compared to the pre‐inulin intervention, we found the reduced F/B ratio in the Post_1M and Post_3M time points and the reduction was statistically significant at the 3‐month time point (Figure 5B,F). The microbial dysbiosis index, which was calculated based on the abundance of taxa, decreased significantly in Post_3M samples compared with Pre samples (Figure 5C). Furthermore, the Ace and Chao indices reflecting the alpha diversity of the microbiota community were significantly lower in Post_3M samples than in Pre samples (Figure 5D,E). However, the overall structure of gut microbiota did not differ significantly among Pre, Post_1M, and Post_3M groups, as displayed in PCoA based on the weighted UniFrac distance of ASVs with Adonis (P = 0.197; R‐squared = 0.020) (Figure S8A, Supporting Information). Aiming at identifying the relevant differential bacteria, we further analyzed the data at different taxonomy levels. At the phylum level, both Post_1M and Post_3M groups showed a higher abundance of Actinobacteriota than the Pre group (Figure 5F; Figure S8C–F, Supporting Information). At the genus level, both Post_1M and Post_3M groups had a higher abundance of Bifidobacterium but a lower abundance of Escherichia‐Shigella compared to Pre group (Figure 5G–I; Figure S8C–F, Supporting Information). In a random forest model, a group of seven genera was selected as the key genera with the most discriminability between the Pre‐ and Post group, with AUC values of 0.687 (Figure 5H; Figure S8B, Supporting Information). Bifidobacterium exhibited the highest mean decrease accuracy (MDA) (Figure 5H). As Bifidobacterium is an essential probiotic in the human intestine,^[46^] we investigated the association of the abundance of Bifidobacterium with the clinical parameters of patients with PCOS. We found a negative correlation between the abundance of Bifidobacterium and testosterone or DHEAs levels (Figure 5J). Our findings indicate that inulin benefits the management of PCOS and promotes the specific bifidogenesis of gut microbiota.

To prove the metabolic improvements in patients with PCOS by inulin are mediated by the gut microbiota, we designed cross‐species FMT experiments as illustrated in Figure6A, in which we transplanted human fecal microbiota to mice. To deplete the mouse endogenous gut bacteria, we treated the mice with an antibiotic cocktail for 2 weeks prior to FMT. The mice then received the pooled fecal microbiota from the donors either pre‐inulin intervention or post‐inulin intervention. The recipient mice were designated as FMT_Pre and FMT_Post, respectively. We sequenced fecal samples of the recipient mice three weeks after FMT along with the donor samples by shallow metagenomic sequencing method. The overall structure of the gut microbiota of the recipient mice was more similar to that of their donors in each group, however, the microbiota structure of the two FMT groups was very different (Figure 6B,C). Interestingly, FMT_Post mice had a significantly higher abundance of Bifidobacterium animalis compared with FMT_Pre mice (Figure 6D).

We found that there was a significant body weight loss in FMT_Post mice in comparison to FMT_Pre mice, despite all recipient mice being fed the same HFD (Figure 6E). Blood glucose levels were also significantly lower in FMT_Post mice than FMT_Pre mice both in the GTT and ITT tests (Figure 6F–I). Compared to FMT_Pre mice, the serum level of fasting insulin was lower in FMT_Post mice, although it was not statistically significant (Figure 6J). These data demonstrate that the human gut microbiota from post‐inulin intervention contributes to maintaining normal glucose regulation and insulin sensitivity in the recipient mice. Upon histology of ovaries in the recipient mice, FMT_Post mice showed normal structure of ovaries with follicles at all levels and corpora lutea. The hematogenous corpus luteum was found in FMT_Pre mice, but the number of corpora lutea was significantly reduced compared with FMT_Post mice (Figure 6K,L). In addition, the serum level of testosterone was significantly lower in FMT_Post mice than FMT_Pre mice (Figure 6M). We tested serum levels of AMH (Figure 6N), LH (Figure 6O), and FSH (Figure 6P) and found that AMH and LH/FSH ratio (Figure 6Q) were decreased in FMT_Post mice compared to FMT_Pre mice. In the estrous cycle test, FMT_Post mice exhibited regular estrous cycles, whereas FMT_Pre mice spent significantly more time in the estrus stage (Figure 6R,S). These results suggest the postinulin microbiota also improves ovarian dysfunction. We further measured BAT‐associated gene expression involved in thermogenesis to assess the BAT activity. The relative expression levels of Ucp1, Pgc1α, Pparα, and Dio2 were significantly higher in FMT_Post mice than in FMT_Pre mice (Figure 6T). Moreover, compared to FMT_Pre mice, smaller sizes of lipid droplets in peri‐ovarian adipose tissue were found in FMT_Post mice (Figure 6U,V), demonstrating that the post‐inulin human microbiota is conducive to enhancing BAT thermogenic activity and inhibiting lipid accumulation.

Furthermore, we analyzed the functional alterations of the gut microbiome by LEfSe for level‐three KEGG pathways between FMT_Pre and FMT_Post mice. Interestingly, we found that gut microbiota from the FMT_Post mice showed enrichment of butanoate metabolism and propanoate metabolism (Figure S9A, Supporting Information). We further measured the concentration of SCFAs in caecum samples. Compared with FMT_Pre mice, FMT_Post mice had significantly increased total SCFAs, including AA, PA, isobutyric acid (IBA), and isovaleric acid (IVA) (Figure7A,B; Figure S9B, Supporting Information). Furthermore, the mRNA expression levels of genes encoding tight‐junction proteins, including Zo1, Occludin, and Claudin1, and the immunohistochemical staining of these tight‐junction proteins in the colon revealed enhanced intestinal barrier integrity in FMT_Post mice compared to FMT_Pre mice (Figure 7C–I). With respect to inflammation, we observed significantly lower levels of pro‐inflammatory factors in FMT_Post mice than FMT_Pre mice, supported by the serum levels of LBP, IL‐1β, and IL‐18 (Figure 7J–L) and the ovarian mRNA levels of Lbp, Il1b, and Il18 (Figure S9C, Supporting Information). Moreover, the TLR4/Myd88/NF‐κB/NLRP3/GSDMD signaling pathways in the ovarian tissues were also evaluated between FMT_Pre and FMT_Post mice by western blot. FMT_Post mice exhibited significantly reduced expression of TLR4, Myd88, and p‐p65/p65 compared to those in FMT_Pre mice (Figure 7M,O), accompanied by downregulation of downstream NLRP3 inflammasome (NLRP3/ASC/cleaved‐Caspase‐1) and gasdermin‐D activation (cleaved‐GSDMD) (Figure 7N,P), which potentially underlie the ovarian functional improvements by the post‐inulin microbiota. These results support the notion that gut microbiota from the post‐inulin intervention patients with PCOS have the capacity to mitigate PCOS‐like phenotypes in recipient mice.