Depletion of gut microbiota influents glucose metabolism and hyperandrogenism traits of mice with PCOS induced by letrozole

Forty female C57BL/6J mice aged 3 weeks were acquired from Shanghai Chengqin Laboratory Animal Co Ltd. In a vivarium designed to maintain specific pathogen-free (SPF) conditions, five mice were accommodated per cage. The vivarium was equipped with an automatic lighting system that provided a 12-hour light and 12-hour dark cycle, with the light period lasting from 06:00 to 18:00. The mice had unrestricted access to water and food. Weekly measurements were taken to determine the body weights of mice.

A total of forty mice were stochasticly allocated into four separate groups (each containing ten mice): control, letrozole (LET), antibiotics (ABX), and antibiotics plus letrozole (ABX_LET). To induce the PCOS phenotype (19), mice in LET and ABX_LET groups were orally administered with 1 mg/kg of letrozole (Jiangsu Hengrui Pharmaceuticals Co., Ltd., China) dissolved in 0.5% Carboxy Methyl Cellulose (CMC) (Sigma, USA) over a 21-day period, once daily. Control and ABX groups received vehicle only (0.5% CMC). ABX and ABX_LET groups were treated with an antibiotic cocktail composed of vancomycin hydrochloride(0.5g/L), ampicillin sodium(1g/L), neomycin sulfate(1g/L), and metronidazole(0.2g/L) (20) daily in their drinkable water, while there was free access to water in the control and LET groups. And one mouse from the ABX group died.

Mice were subjected to a 12-hour fasting period prior to undergoing the IPGTT procedure. Blood glucose levels were determined by extracting blood samples from tail vein with a blood glucose meter (Accu-Chek Performa; Roche, USA). After measuring the fasting glucose levels, glucose solution was injected intraperitoneally in mice at a dosage of 2 g/kg body weight for IPGTT, and tail blood samples were collected at intervals of 15, 30, 60, 90, and 120 minutes after glucose administration to measure glucose level. The fasting blood samples were gathered into tubes for the purpose of measuring insulin levels.

After completing the treatments, blood samples were gathered and subjected to centrifugation at a speed of 2500 rpm at 4°C for a duration of 30 min. Afterwards, the liquid above was meticulously transferred to cryogenic vials and promptly preserved at -80°C for subsequent examination. According to the manufacturer’s instructions, the levels of testosterone (CEA458Ge, Cloud-Clone Corp., China), luteinizing hormone (LH) (CEA441Mu, Cloud-Clone Corp., China), and insulin (80-INSMSU-E01, ALPCO, USA) underwent detection through utilizing enzyme-linked immunosorbent assay (ELISA) kits.

Dissected ovaries were fixed in 4% paraformaldehyde, hydrated in gradient ethanol, and embedded in paraffin at the end of the experiment. The samples then underwent slide processing in preparation for H&E staining, with the purpose of analyzing any pathological structural changes. The cystic follicles (CF) and corpus luteum (CL) numbers were counted.

Prior to sacrifice, the fecal samples were collected and preserved at -80°C immediately till following work. DNA was extracted from stool samples using a method described earlier (27). In accordance with the manufacturer’s instructions, the V3-V4 regions sequencing library of 16S rRNA gene was built(part no.15044223 Rev. B; Illumina Inc., CA, USA) based on the previous publication with some modifications (28).

The plasmid from the Ruminococcus strain, which contained the full-length 16S gene at a concentration of 10⁹ copies/ml was diluted at different gradients to obtain concentrations of 10⁸, 10⁷, 10⁶, 10⁵, 10⁴, 10³, and 10² copies/ml. The qPCR was conducted on a LightCycler 96 system using a 20μL reaction mixture comprising template (20 ng), primer Uni331F (5’-TCCTACGGGAGGCAGCAGT-3’), primer Uni797R (5’-GGACTACCAGGGTATCTAATCCTGTT-3’), and supermix (Bio-Rad), with Standard and sample DNA replicated twice (29, 30). A PCR reaction was performed using the following conditions: 95°C for 5 minutes, followed by 39 cycles of 95°C for 20 seconds, 60°C for 60 seconds, 95°C for 10 seconds, 65°C for 1 minutes, 97°C for 1 seconds, and 37°C for 30 seconds. A standard curve was generated by performing a linear fit of the copy number and CT value of the plasmid across various gradients. The DNA samples’ copy number was determined through a standard curve by utilizing Opticon Monitor 3.1. And the resulting conversion was copies per gram of wet feces.

Quantitative Insights Into Microbial Ecology2 (QIIME2, v2018.11) (31) was applied for analysis of raw sequences. The “Demux” and “cutadapt” plugins were used for sequence quality and primer removal, respectively. Meanwhile, DADA2 (32) was used to obtain mass amplicon sequence variants (ASVs) by filtering, denoising, and merging (33). The construction of a phylogenetic tree for each ASV was achieved by utilizing “FastTree”, while the taxonomic classification was assigned through the SILVA132 16S rRNA database (34). The absolute abundance matrix of ASVs was obtained by multiplying the original abundance matrix by the number of bacteria (copies/g). ASVs number, Shannon index, and Simpson index were employed to assess the Alpha diversity of samples. Based upon the Principle Coordinate Analysis (PCoA) of Bray Curtis distance, Beta diversity was evaluated. Permutational multivariate analysis of variance test (PERMANOVA; 9,999 permutations) was employed to determine the significance of differences in intestinal microbiota between groups. Differences were defined significant when P was <0.05.

Using the “random Forest” package in R (v 4.2.0, USA), the method of random Forest and cross validation is utilized to find the ASVs that can distinguish different groups. Use the “pheatmap” package in R to visualize the absolute abundance of different ASVs across various groups. Spearman analysis was conducted on the correlation between ASV abundance and total testosterone and glucose metabolism by Origin (MicroCal, USA). When P was <0.05, the correlation was considered significant.

PICRUSt2 (35) was used to build the prediction spectrum of gene function in the bacterial domain. And STAMP (36) was applied to analyze the PICRUSt2 prediction results, acquire metabolic pathway information with significant differences, and generate a visual representation of the findings. Clustering and correlation analysis and calculations were proceeded through the “stats” package in R.

The statistical significance of physiological and biochemical data among the different groups was assessed using one-way analysis of variance (ANOVA) with SPSS 22.0 software (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 8.0 (GraphPad Software, Inc., San Diego, CA). The data conforming to a normal distribution was analyzed using the One-way ANOVA method. The K-S test was applied for statistical analysis of the data disconfirming to normal distribution. In order to compare the differences between the four groups, the P values were corrected through the Tukey post-test. A two-tailed unpaired student t test was employed to assess the differences between the two groups. Statistical significance was considered when the P value was equal to or less than 0.05.

Letrozole was used by intragastric administration (at a concentration of 1 mg/kg diluted in a solution of 0.5% CMC.) to induce PCOS in C57BL/6 mice in letrozole (LET) and antibiotics plus letrozole (ABX_LET). Mice from antibiotics (ABX) and antibiotics plus letrozole (ABX_LET) groups were treated with an antibiotic cocktail, while the control and LET groups received distilled water. At the conclusion of the treatments, blood samples were acquired and the concentrations of testosterone, luteinizing hormone (LH) and insulin were tested. Hyperandrogenism is currently considered to be one of the most essential characteristics of PCOS, while elevated luteinizing hormone (LH) level is considered to be one of the features of neuroendocrine disorders of PCOS (37). Therefore, fasting serum sex hormone levels of four groups of mice were detected respectively. The LET group demonstrated a significant increase in both total testosterone (TT) and LH levels when compared to the Control group, and the ABX_LET group exhibited significantly lower TT levels compared to the LET group (Figures 1A, B). But antibiotics treatment did not affect LH levels in ABX_LET group (Figure 1B). It is noteworthy that the TT and LH levels in ABX_LET group were significantly elevated compared to the ABX group (Figures 1A, B), suggesting additional factors influencing the androgen levels in PCOS mice. The changes in ovarian structure were determined by observing the ovarian tissue sections of mice. The morphological structure of the ovarian tissues of Control and ABX group was generally normal, with no cystic follicles and multiple luteal bodies (Figure 1C). Compared with Control group, the ovarian tissue in the LET group displayed aberrant overall structure, characterized by polycystic changes, a significant increase in number of cystic follicles, a decrease in number of luteum present in the ovary (Figures 1C–E). Compared with LET group, the ovarian structure of mice in ABX_LET group was not significantly improved, and no significant reduction of cystic follicles was observed (Figures 1C–E). The findings validated the successful induction of hyperandrogenemia and neuroendocrine symptoms in mice with PCOS, and suggested that the gut microbiome may participate in the onset of hyperandrogenism instead of LH level and ovarian structure.

After modeling by letrozole for 21 consecutive days, the glucose metabolism of mice in the four groups was evaluated by IPGTT, body weight, and fasting insulin level. The four groups showed no significant differences in body weight. (Figure 2A). The fasting blood glucose levels, 90-minute postprandial blood glucose levels, and the area under the curve during IPGTT exhibited significantly higher values in the LET group compared to the Control group (Figures 2B, E). The ABX_LET group displayed lower levels of blood glucose at 15, 90, and 120 min and area under the IPGTT curve than the LET group (Figures 2C, E) (all P<0.05). The result above demonstrated that the mice belonging to the LET group exhibited glucose tolerance impairment and antibiotic treatment effectively alleviated glucose tolerance. However, the comparison of fasting insulin levels between four groups showed that no significant difference was found between the Control group and the LET group, nor between LET and ABX_LET group (Figure 2F), indicating that gut microbiota in PCOS mice may not affect their fasting insulin level. Interestingly, it was also found that compared to ABX group, the blood glucose level of mice in ABX_LET group increased significantly at 90 min and the area under the IPGTT curve showed a significant difference (Figures 2D, E), and a significant increase in fasting insulin level was observed in ABX_LET group compared to ABX group(Figure 2F), which suggested other factors that affected the glucose metabolism of PCOS mice.

With an aim of exploring the effects of antibiotics on the diversity of gut microbiota of letrozole-induced PCOS mice, sequencing was conducted on the 16S rRNA gene V3-V4 region of stool specimens acquired. The findings of the study suggested that the administration of antibiotics resulted in a markedly decrease in the quantity of amplicon sequence variants (ASVs), as well as the Shannon and Simpson indices of the intestinal flora in comparison to the two groups that did not receive antibiotic treatment (Figures 3A–C). Moreover, the LET group exhibited a remarkable elevation in the Shannon and Simpson indices, in comparison to the two groups without letrozole induction (Figures 3A–C). These results indicated that combined antibiotic intervention to gut microbiota could alter the diversity of intestinal flora in letrozole-induced mice.

Beta diversity is frequently used to indicate the resemblance of the constituent part of intestinal flora across groups. To this end, the absolute abundance matrix of ASV was utilized to compute the Bray Curtis distance, followed by the execution of Principal coordinate analysis (PCoA) and PerMANOVA analyses on the basis of this distance metric. The findings revealed that the intestinal flora of the Control and LET groups exhibited significant distinctions from the ABX_LET group along the first principal coordinate (PC1) axis, which accounted for 45.63% of the total microbiota variance. The clustering analysis revealed that the flora structure of the LET group and the ABX_LET group exhibited discernible differences, indicating that pseudo germ-free treatment caused a significant influence on the gut flora structure of letrozole-induced mice. Notably, along the PC2 axis, there was a clear distinction of gut microbiota between the ABX_LET group and the ABX group, which accounted for 14.95% of the total microbiota variance (Figures 4A, B). These findings suggested that both antibiotic intervention and letrozole modeling independently contributed to changes in the composition of gut flora structure of mice. A random forest model was established on account of the gut microbiota data obtained from four groups of mice through machine learning methods to further analyze the ability of ASVs for the purpose of recognizing different groups. According to the Gini index ranking in the random forest model, when the model error rate was the lowest, Mean Decrease Gini value exceeding 0.1 was selected. A total of 43 key ASVs contributing to the differentiation of Control, LET, ABX, and ABX_LET group were found (Figure 4C). Among them, 27 ASVs may contribute to the distinction between Control and LET group, while 29 ASVs may contribute to the distinction between LET and ABX_LET group. And 8 ASVs may contribute to the distinction between ABX and ABX_LET group (Figure 4C). The LET group exhibited a significant increase in the absolute abundance of five ASVs compared to the Control group, including bacteria belonging to the unclassified Clostridia_vadinBB60_group, Enterorhabdus, and Muribaculaceae (Figure 4D). The absolute abundance of these ASVs was in positive correlation with the TT levels and the area under the IPGTT curve. In addition, the absolute abundance of two ASVs was significantly decreased in LET group, and these microorganisms belonged to the unclassified Clostridia_vadinBB60_group and Enterococcus. The absolute abundance of these two ASVs was significantly negatively correlated with the total testosterone level. Compared with LET group, ABX_LET group significantly reduced the absolute abundance of 5 ASVs, which were ASV62 (unclassified Clostridia_vadinBB60_group), ASV69 (unclassified Enterorhabdus), ASV91 (unclassified Enterorhabdus), ASV51 (unclassified Muribaculaceae), and ASV 52 (unclassified Muribaculaceae). Furthermore, these ASVs with decreased abundance were positively correlated with mice serum total testosterone levels and area under the IPGTT curve (Figure 4D).

For the purpose of forecasting and comparing the gut flora function of mice in different groups, utilization of PICRUSt2 was implemented based on KEGG database. A total of 25 metabolic pathways displayed significant differences between the LET group and the ABX_LET group, among which the metabolic pathways in the LET group were all elevated compared to those in the ABX_LET group, including the polysaccharide biosynthetic metabolic pathway, lysine synthesis pathway, insulin signaling pathway, linoleic acid metabolism, ether ester metabolism, primary and secondary bile acid synthesis pathway, steroid synthesis pathway, etc. (Figure 5). The absolute abundance of these pathways was positively correlated with the area under the IPGTT curve and insulin level (Figure 5).