2.1. Analysis of OTU level difference of gut microbiota
Based on PLS-DA analysis, we found significant differences in gut microbiota among the three groups (Fig. 1a). OTU level difference analysis results showed that PCOS-IR group rank-bank curve was steeper than the control group, revealing that the abundance and diversity of gut microbiota decreased (Fig. 1b). No significant difference was observed between the PCOS-NIR and control group (Fig. 1c). In addition, in comparison with the control group, the sobs index of PCOS-IR group was significantly lower (P < 0.05), and the ace index of PCOS-IR group was lower without statistical difference (P>0.05), thus proving the above results. No significant differences were observed in the sobs and ace index between the PCOS-NIR and control group (P>0.05, Figs. 1d, e).
To provide reference for the microbial diagnosis of PCOS, we used ROC curve to evaluate the accuracy of the model. The larger the area under the curve, the higher the diagnostic accuracy. Species with the top 3 OTU level abundance were selected for the analysis, and we found that the AUC value of the PCOS-IR and control group was 0.66 (Fig. 2a), and the AUC value of PCOS-NIR and control group was 0.66 (Fig. 2b). AUC was greater than 0.65, indicating that this model has a potential value for clinical diagnosis, but this result needs to be verified by expanding the sample capacity.
2.2. Genus level analysis of gut microbiota
We found 168 same genera in three groups, five characteristic genus in PCOS-IR group, six characteristic genus in PCOS-NIR group, and 26 characteristic genus in control group (Fig. 3a). Further, we analyzed the proportion of gut microbiota in three groups. The top five bacteria in the PCOS-IR group were Bacteroides (26.02%), Faecalibacterium (13.15%), Prevoteila_9 (9.55%), Phascolarctobacterium (4.08%), and Lachnoclostridium (2.94%, Fig. 3b). The top five bacteria in the PCOS-NIR group were Bacteroides (22.53%), Faecalibacterium (14.52%), Prevoteila_9 (7.72%), Blautia (4.26%), and Phascolarctobacterium (3.67%, Fig. 3c). The top five bacteria in the control group were Bacteroides (20.83%), Faecalibacterium (15.12%), Prevoteila_9 (11.98%), Agathobacter (3.29%), and [Eubacterium] _ellgens_group (2.68%, Fig. 3d). The proportion of bacteria in each group differ. Further, we obtained the bacteria with statistical difference among the three groups. The bacteria with statistical difference were Ruminococcaceae_UCG-014, Lachnospiraceae_ND3007_group, norank_f_Ruminococcaceae, norank_o__Mollicutes_RF39, and Coprococcus_1 (P<0.05, Fig. 3e). We also compared the genus level of gut microbiota in the PCOS-IR and PCOS-NIR group. The results showed that the abundance of Ruminococcaceae_UCG-004 and Acetanaerobacterium in the PCOS-IR group was significantly lower than that of the PCOS-NIR group (P<0.05, Fig. 3f).
Furthermore, Lefse analysis was performed on the gut microbiota of the three groups. The comparison between PCOS-IR and control group revealed that g__Fusobacterium and g__Faecalibaculum were the unique genus of the PCOS-IR group (Figs. 4a, b). Comparison between the PCOS-NIR and control group showed that g__Tyzzerella, g__norank_o__Gastranaerophilales, g__unclassified_p__Firmicutes, g__Morganella, g__Phocea, g__Faecalibaculum, and g__unclassified_f__Desulfovibrionaceae were the unique genus of the PCOS-NIR group (Figs. 4c, d). In conclusion, differences were observed in the genus level abundance among the three groups.
2.3. Analysis of gut microbiota function and enzyme abundance
We used PICRUSt to analyze the KEGG pathway abundance of gut microbiota. The abundance of each pathway can be calculated according to the gut microbiota. KEGG pathway included organismal systems, metabolism, genetic information processing, environmental information processing, and cellular processes. In these pathways, no significant difference was observed in the functional abundance among the three groups (Figs. 5a–e). Considering that IR is closely related to metabolism, we further analyzed the attached pathway of metabolism. The results showed the functional abundance of isoquinoline alkaloid biosynthesis, styrene degradation, and atrazine degradation with statistical difference (P<0.05, Figs. 5f, g).
Further, we analyzed the enzyme abundance of gut microbiota among the three groups. The results showed that the top 10 enzymes with significantly difference in abundance were futalosine hydrolase (P<0.01), arogenate dehydrogenase (NADP(+)), (P<0.01), L-lysine 6-transaminase (P<0.05), homoserine O-acetyltransferase (P<0.05), aldehyde dehydrogenase (FAD-independent, P<0.05), kdo(2)-lipid IV(A) lauroyltransferase (P<0.05), scyllo-inositol 2-dehydrogenase (NADP(+), P<0.05), serralysin (P<0.05), arabinose-5-phosphate isomerase (P<0.05), and 3-aminobutyryl-CoA ammonia-lyase(P<0.05, Fig. 6).
2.4. Correlation analysis between gut microbiota and serum sex hormones
After clarifying the level and function differences of gut microbiota in the three groups, we further explored the relationship between gut microbiota and sex hormones. We first compared the hormone levels of the three groups, luteinizing hormone (LH) was significantly higher in the PCOS-IR and PCOS-NIR groups than in the control group (P<0.001, P<0.01), and LH/follicle-stimulating hormone (FSH) was significantly higher (P<0.001, P<0.0001). In addition, FINS and HOMA-IR significantly increased in the PCOS-IR group compared with PCOS-NIR group (P<0.0001, P<0.0001). No significant differences in FSH, estradiol (E2), dehydroepiandrosterone (DHEA), prolactin (PRL), testosterone (T), and FBS were observed among the three groups (P>0.05, Figs. 7a–j). Although T and DHEA in PCOS group had no significant difference compared with the control group, they were higher than the clinical androgen threshold in normal women, and this result is consistent with the biochemical hyper androgen manifestations of PCOS .
Based on correlation analysis, the LH level was negatively correlated with Blautia and Agathobacter abundance (P<0.05), LH/FSH was positively correlated with norank_f__Prevotellaceae abundance (P<0.05), T level was negatively correlated with norank_f__Prevotellaceae and Roseburia abundance (P<0.01), DHEA level was positively correlated with Ruminococcus]_torques_group, norank_f__Lachnospiraceae and Parabacteroides abundance (P<0.05), and HOMA-IR was positively correlated with Anaerostipes and norank_f__Lachnospiraceae abundance (P<0.05) in PCOS-IR patients (Fig. 8a). FINS level was positively correlated with Faecalibacterium abundance (P<0.001), HOMA-IR was positively correlated with Faecalibacterium abundance (P<0.01) and negatively correlated with Klebsiella, Lachnospira, and Prevotella_2 abundance (P<0.01, P<0.05), and LH/FSH was positively correlated with Bacteroides and Lachnoclostridium abundance (P<0.05) in PCOS-NIR patients (Fig. 8b).