Probiotic Bifidobacterium lactis V9 may influence sex hormone levels in polycystic ovary syndrome through gut-brain communication

In the present study, clinical indices were measured in PCOS patients and healthy subjects (Table 1; see also Table S1 and Fig. S1 in the supplemental material). The average values determined for metabolic parameters, including levels of triglycerides (TGs), total cholesterol (TC), and fasting plasma glucose (FPG), in the PCOS group were 1.64 ± 0.26 mmol/liter, 4.73 ± 0.29 mmol/liter, and 5.98 ± 0.57 mmol/liter, respectively, and were significantly higher than those in the control group. Moreover, the metabolic levels of the typical gut-brain mediators peptide YY (PYY) and ghrelin reached 85.16 ± 10.11 pg/ml and 0.54 ± 0.03 ng/ml, respectively, in the control group and were significantly higher than those in the PCOS group. Additionally, we determined the levels of sex hormones, including luteinizing hormone (LH), follicle-stimulating hormone (FSH), prolactin (PRL), estradiol (E2), and testosterone (T), on day 3 of the menstrual cycle for each study subject. A much higher level of sex hormones was observed in the PCOS group than in the control group.

For each of the 64 subjects, the organismal structure of the intestinal microbiota was analyzed by sequencing the 16S rRNA gene amplicons. A principal-coordinate analysis (PCoA) was performed based on the weighted and unweighted UniFrac distances of the 16S rRNA sequence profiles at the genus level (Fig. 1A and B). The intestinal microbiotas from the control and PCOS groups were highly distinct with respect to organismal structure, with the subjects forming two clusters that corresponded to the two experimental groups. To correct for any possible population stratification caused by non-disease-related factors that could impact the structure of intestinal microbiota, an Adonis test was performed (Table S2 and S3). After correction, the effects associated with the non-disease-related factors disappeared. For example, the R² values which represented the correlation between the values representing certain host factors (i.e., age and HAIR-AN [hyperandrogenism {HA}, insulin resistance {IR}, and acanthosis nigricans {AN}] and acne values) and the host’s intestinal microbiota were 0.023, 0.023, and 0.011, respectively. Those values are far lower than the R² values determined for the disease factor, further confirming that PCOS was a significant factor explaining the observed variation in the intestinal microbiota. At the genus level, we observed that the abundances of Faecalibacterium, Lachnospira, Bifidobacterium, and Blautia were significantly higher in the control group than in the PCOS group, whereas those of Parabacteroides, Bacteroides, Lactobacillus, Oscillibacter, Escherichia/Shigella, and Clostridium were enriched in the PCOS group (Fig. 1C; Wilcoxon rank sum tests).

To further explore the differential intestinal microbes at the strain level, we binned the shotgun metagenomic sequencing data into 1,080 coabundance gene groups (CAGs) (Table S4), and the 151 CAGs containing more than 50 contigs were reassembled into MGS. Finally, 45 MGS were assigned specific taxonomic levels. Through a comparative analysis, we observed that 30 CAGs, including Subdoligranulum variabile MGS031, Prevotella copri MGS035, Eubacterium rectale MGS044, Eubacterium eligens MGS013, Dialister invisus MGS048, Collinsella aerofaciens MGS060, Bacteroides vulgatus MGS014, Bacteroides uniformis MGS022, Bacteroides sp. MGS027, Bacteroides caccae MGS018, and Alistipes putredinis MGS041, were enriched in the PCOS group (Fig. 2B), while 31 CAGs, including Bifidobacterium sp. MGS003, Faecalibacterium prausnitzii MGS004, Bacteroides sp. MGS008, Bifidobacterium animalis MGS015, Roseburia inulinivorans MGS148, Coprococcus eutactus MGS007, Bifidobacterium pseudocatenulatum MGS052, Roseburia intestinalis MGS026, Bacteroides stercoris MGS016, Coprococcus comes MGS023, Bacteroides plebeius MGS006, and Faecalibacterium prausnitzii MGS032, were enriched in the control group (Fig. 2A). Interestingly, we noted a lack of the beneficial microbes Faecalibacterium prausnitzii and Bifidobacterium sp., as well as the presence of the disease-related microbes Prevotella copri and Collinsella aerofaciens, in the PCOS patients.

To investigate the observed differences in the functional profiles of the intestinal microbiota between the PCOS patients and the healthy subjects, high-quality reads from all samples were obtained and annotated for protein-coding genes. Based on the results, a collective, nonredundant intestinal microbiota gene catalogue for PCOS was created. Next, for each sample, the reads were mapped to the collective gene catalog to reconstruct sample-specific gene profiles, and profiles were also generated using the Kyoto Encyclopedia of Genes and Genomes database orthologs (KO; Table S5). Metabolic pathways associated with fructose and mannose metabolism, the citrate cycle (tricarboxylic acid [TCA] cycle), flagellar assembly, bacterial chemotaxis, cationic antimicrobial peptide (CAMP) resistance, lipopolysaccharide biosynthesis, the phosphotransferase system (PTS), biotin metabolism, folate biosynthesis, thiamine metabolism, and one carbon pool by folate were enriched in the PCOS group, whereas the metabolic pathways associated with valine, leucine and isoleucine biosynthesis, propanoate metabolism, fatty acid biosynthesis, ABC transporters, and bacterial secretion systems were enriched in the control group (Fig. 3). Accordingly, we determined the concentrations of short-chain fatty acids, including acetic, propionic, butyric, and valeric acids (Table 1; see also Fig. S1), which are considered the main intestinal microbial beneficial metabolites, in the fecal samples of the participants. Unsurprisingly, the amounts of acetic, propionic, butyric, and valeric acids in PCOS patients were 24.59 ± 8.94, 13.93 ± 3.84, 5.05 ± 1.59, and 0.55 ± 0.29 μmol/g, respectively, and were significantly lower than those in the control group.

We further explored the correlations among the identified MGS, metabolic parameters, SCFAs, and sex hormones by constructing a network based on the determined Spearman’s rank correlation coefficients (Fig. 4; see also Table S6). As shown in Fig. 4, a generally positive correlation was observed between Bifidobacterium animalis and Faecalibacterium prausnitzii and the SCFAs as well as the gut-brain mediators, whereas a negative correlation was observed between these two species and the sex hormone levels. In contrast, the species Coprococcus eutactus, Collinsella aerofaciens, Prevotella copri, and Bacteroides caccae were positively correlated with the levels of sex hormones but negatively correlated with the levels of SCFAs and gut-brain mediators. By performing the Adonis test on the UniFrac distance matrix data shown in Fig. 1 using the same variables listed in Fig. 4, we linked the interpretations of the 16S amplicon sequencing data and the metagenomic sequencing data (Table S2 and S3). The variables (including TGs, TC, LH, T, PYY, ghrelin, and SCFAs) and the metagenomic species (including Bifidobacterium animalis MGS015, Bifidobacterium sp. MGS003, and Bifidobacterium sp. MGS009) were closely correlated with changes in the intestinal microbiota of the PCOS patients.

After demonstrating the alterations in the intestinal microbiome in the PCOS group at the taxonomic and functional levels, we further investigated the key roles of the beneficial microbes Bifidobacterium and Faecalibacterium prausnitzii, which were closely correlated with PCOS. Considering its safety and universality and the results of our previous studies (19, 20), we chose the probiotic Bifidobacterium lactis V9 strain for the subsequent experiments. Fourteen PCOS patients participated in a study to investigate the impact of Bifidobacterium lactis V9 on the intestinal microbiota of individuals with PCOS. By monitoring the sex hormone, signal peptide, and intestinal SCFA levels in the 14 volunteers during the Bifidobacterium lactis V9 consumption period, we observed dramatic changes in the indices mentioned above. The levels of LH and LH/FSH decreased significantly in 9 volunteers, whereas the levels of sex hormones increased markedly, as did the intestinal SCFA levels, including the levels of acetic acid, propionic acid, butyric acid, and valeric acid (Fig. 5). In contrast, the fluctuations in these indices were not notable in the remaining 5 volunteers. Accordingly, the 14 volunteers were divided into two groups, namely, the response group (group R) and the nonresponse group (group N).

Probiotics are defined as live microorganisms that confer health benefits to the gut microbiota of the host when present in adequate amounts. The host’s gastrointestinal tract is the main “battleground” where the probiotic is able to exert its health-promoting functions. Therefore, the effective colonization of a consumed probiotic in the host gut is crucial for its health-promoting characteristics. By performing quantitative PCR (q-PCR) with a strain-specific primer, we quantified the number of viable Bifidobacterium lactis V9 cells in the two groups. The results were highly consistent with the clinical and SCFA results. We observed that the average abundance of viable Bifidobacterium lactis V9 reached 6.93 ± 0.42 log CFU/g of feces in group R during the Bifidobacterium lactis V9 consumption phase (Fig. 6). More convincingly, no significant decline in the abundance of viable cells was observed during the washout phase in group R. In contrast, few live probiotic Bifidobacterium lactis V9 organisms survived in the guts of the participants in group R in both the consumption and washout phases. These results indicate a robust correlation between the abundance of the colonized probiotic Bifidobacterium lactis V9 and the PCOS-related clinical indices and intestinal SCFAs.

Based on the results described above, we further addressed the issue of how the intestinal microbiota responded to the intake of the probiotic Bifidobacterium lactis V9 in group R and group N. Unsurprisingly, marked changes in the structure of the intestinal microbiota were observed in group R through a principal-coordinate analysis (PCoA) based on UniFrac distance (Fig. 7A; see also Fig. S2). In this group, increased abundances of the genera Bifidobacterium, Faecalibacterium, Butyricimonas, and Akkermansia were observed in addition to decreases in the abundances of the genera Collinsella, Coprococcus, Klebsiella, Clostridium, Actinomyces, Streptococcus, Eubacterium, and Ochrobactrum (Fig. 7C; see also Table S7). In contrast, the influence of the probiotic Bifidobacterium lactis V9 on the gut microbiotas in group N was limited (Fig. 7B), and individual factors dominated the structure of the intestinal microbiotas throughout the experiment.

The experimental cohort consisted of 64 subjects who were divided into two groups.

The first group (case group) consisted of 38 PCOS patients (aged 22 to 38 years) who were recruited from the Hainan General Hospital, located in Haikou, Hainan Province of China. The diagnosis of PCOS was confirmed by a B-ultrasound scanner and diagnosis of serious oligomenorrhea and by assessing the gonadal hormone levels of the individuals. The second group (control group) consisted of 26 healthy individuals (aged 23 to 30 years). After being informed of the experimental guidelines and details, 14 volunteers agreed to participate in the subsequent probiotic Bifidobacterium lactis V9 intake experiment, which was designed to evaluate the impact of Bifidobacterium lactis V9 on the intestinal microbiome of individuals with PCOS. All 14 subjects were asked to consume a total of 10.6 log CFU of Bifidobacterium lactis V9 once daily for 10 weeks. During the experiment, the subjects were asked to avoid ingesting any other probiotic products or antibiotics. Fecal samples were collected at weeks 0, 2, 4, and 10. Additionally, to evaluate the extent of Bifidobacterium lactis V9 colonization, we set up a 3-day washout period after every sampling time point, and fecal samples were also collected during the washout periods. The study was approved by the Ethical Committee of Hainan University (Haikou, China) (HNU-EC-2017011), and informed consent was obtained from all 64 volunteers before they enrolled in the study. Sampling and all described subsequent steps were conducted in accordance with the approved guidelines.

Fecal samples were collected from each individual in the morning before their first meal. After the weight of the fecal sample was determined, a sample protector (CWBIO, China) was added at a ratio of one part fecal sample to five parts of the sample protector, after which the samples were stored at −20°C until further processing. All samples were used for high-throughput sequencing of the V3-V4 region of the bacterial 16S rRNA gene, and 40 samples (including 20 from the PCOS patients and 20 from the healthy control individuals) were selected for deep shotgun metagenomic sequencing.

The gut microbiota of each patient was isolated, and a QIAamp DNA stool minikit (Qiagen, Hilden, Germany) was used for metagenomic DNA extraction. The quality of the metagenomic DNA was assessed by spectrophotometry. We selected the bacterial V3-V4 region of the 16S rRNA gene for high-throughput sequencing (45). After PCR amplification, the amplicons were quantified, and all PCR products were subsequently pooled in equimolar ratios to reach a final concentration of 100 nmol/liter each. Next, samples were loaded onto an Illumina MiSeq high-throughput sequencing platform for sequencing (17). The QIIME (v1.7) (46) platform was used for the bioinformatics analysis. Representative operational taxonomic units (OTUs) were selected and annotated by the Ribosomal Database Project (RDP) for further microbial structure and taxonomic analyses (47).

We selected 40 samples, including 20 samples from PCOS patients and 20 samples from healthy subjects, for deep shotgun metagenomic sequencing using an Illumina HiSeq 2500 instrument. Libraries were prepared with a fragment length of approximately 300 bp. Paired-end reads were generated using 100 bp in the forward and reverse directions. The reads were trimmed using Sickle and were subsequently aligned to the human genome to remove the host DNA fragments. An average of 19.17 Gb of high-quality paired-end reads, were obtained from each sample, totaling 815.32 Gb of high-quality data that were free of human DNA and adaptor contaminants (seein the supplemental material). Table S3

The shotgun reads were assembled into contigs and scaffolds using IDBA-UD (48), and then the contigs were used to predict the functional genes by the use of MetaGeneMark (49). Finally, a nonredundant gene catalogue that contained 2,382,195 genes was constructed using CD-HIT (50).

The abundances of the genes were determined by aligning the reads to the gene catalogue using Bowtie2 (51). Subsequently, for any sample N, we calculated the abundance in two steps as follows: step 1 (calculation of the copy number of each gene), bi=xiLi; step 2 (calculation of the relative abundance of gene i), ai=bi∑ibi (where a_i represents the relative abundance of gene i, b_i represents the copy number of gene i from sample N, L_i represents the length of gene i, and x_i represents the number of mapped reads).

For the metagenomic species (MGS) analysis, the coabundance principle and canopy clustering algorithm were used to generate CAGs by binning the shotgun reads. Any CAG member that had more than 50 contigs was considered to represent an MGS. After reassembling, the MGS were assigned to a given genome when more than 80% of the subgenes matched the same genome using BLASTn at a threshold of 95% identity over 90% of the gene length. If more than 80% of the genes from an MGS had the same taxonomic level of assignment, then that MGS was identified as representing the same microbe.

The annotated amino acid sequences were aligned against the Kyoto Encyclopedia of Genes and Genomes (KEGG) databases using BLASTp (E value of ≤e−5 with a bit score higher than 60). The annotated sequences were assigned to a KEGG orthologue (KO) group according to the highest score (52). The reporter Z-scores were calculated to reveal the differences in enriched metabolic pathways between the control and PCOS groups as previously described (53). Accordingly, a reporter score of >2.3 (90% confidence according to the normal distribution) was used as a detection threshold to significantly differentiate between pathways (3).

The Bifidobacterium lactis V9 strain-specific primer was designed according to a specific fragment (317 bp) in the complete sequence of the strains. The primer sequences were as follows: V9F, 5′-ACCCATCTCATCCAAACCAAG-3′; V9R, 5′-CACCCAAATGTAAGGAAAGTCG-3′. The specificity of the Bifidobacterium lactis V9 strain-specific primer was confirmed by PCR using DNA from other Bifidobacterium strains and common intestinal microbes. Each PCR mixture (20 μl) contained 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl₂, a 200 μM concentration of each deoxynucleoside triphosphate (dNTP), 1.5 U Taq DNA polymerase (TaKaRa, Japan), 0.4 μM concentrations of the primers, and 10 ng template DNA. The amplification program consisted of 1 cycle of 94°C for 2 min; 28 cycles of 94°C for 45 s, 60°C for 30 s, and 72°C for 45 s; and, finally, 1 cycle of 72°C for 7 min. PCR products were electrophoresed at 50 V in a 1.0% agarose gel. The specific band was then extracted, purified, and sequenced to confirm the specificity further.

To quantify the number of Bifidobacterium lactis V9 cells in the present study, q-PCR was performed in an ABI Step-One detection system (Applied Biosystems). To ensure that the quantified Bifidobacterium lactis V9 cells were alive, we treated the samples with propidium monoazide and phorbol myristate acetate (PMA) (30 μM; 8-min exposure time), according to the manufacturer’s instructions. The reaction mixture (20 μl) contained 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl₂, a 200 μM concentration of each dNTP, 0.4 μl of SYBR green I (Invitrogen), 0.4 U Taq DNA polymerase (Hot Start version) (TaKaRa), 0.2 μM concentrations of the specific primers, and 2 μl of template DNA. The amplification program consisted of 1 cycle of 94°C for 20 s; 40 cycles of 94°C for 5 s, 60°C for 30 s, and 72°C for 45 s; and, finally, 1 cycle of 72°C for 180 s. Fluorescence intensities were detected during the last step of each cycle. To distinguish the targeted PCR product from nontargeted PCR products, melting curves were obtained after amplification by slow heating from 58 to 95°C in increments of 0.2°C with continuous fluorescence collection. At the moment when the digital signals were captured, they were transformed into values representing the amounts of Bifidobacterium lactis V9 in each sample according to the quantified standard curve and, finally, converted into values representing the amount of Bifidobacterium lactis V9 in each gram sample.

Blood samples were collected at the Hainan Provincial People’s Hospital for the measurements of clinical indices, including metabolic parameters (TGs, TC, and FPG), sex hormones (LH, FSH, LH/FSH, E2, PRL, and T) and gut-brain mediators (PYY and ghrelin). The sex hormones were measured on day 3 of the menstrual cycle. The blood biochemical indices were measured on an automatic biochemical analyzer (BS-220; Mindray, China), and the sex hormones were tested using an automated immunoassay system (AIA-1200; Tosoh, Japan). The ghrelin and PYY levels were determined using commercially available enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturer’s instructions (12). To determine SCFA levels, we used approximately 250 mg of frozen fecal samples. The concentrations of SCFAs were determined using a 1:25 dilution of 500 μl of supernatant. Gas chromatography-mass spectrometry (GC-MS) determinations were performed using a Varian Saturn 2000 GC-MS instrument with an 8200 CX solid-phase microextraction (SPME) autosampler (54).

All statistical analyses were performed using R software. The pipeline used in the present study has been uploaded to the GitHub, and the link is https://github.com/zhjch321123/PCOS_Probiotics.git↗. The PCoA analysis was performed in R using the ade4 (55) package. The differential abundances of genera, genes, and KOs were tested with the Wilcoxon rank sum test and were considered significantly different with P values of <0.01. For pathway bubble plot construction, the package ggplot2 was used. For boxplot construction, the package ggpubr was used. The heat map was constructed using the “pheatmap” package, and the Sankey diagram was built using the “riverplot” package. The networks were calculated using the Spearman rank correlation coefficient and were visualized in Cytoscape (version 3.4).

The sequence data reported in this paper have been deposited in the NCBI database (accession no. PRJNA513209↗ [metagenomic sequencing data]).

The sequence data reported in this paper have been deposited in the NCBI database (accession no. PRJNA513209↗ [metagenomic sequencing data]).