Vaginal microbiota are associated with in vitro fertilization during female infertility

Abstract

The vaginal microbiome plays an essential role in the reproductive health of human females. As infertility increases worldwide, understanding the roles that the vaginal microbiome may have in infertility and in vitro fertilization (IVF) treatment outcomes is critical. To determine the vaginal microbiome composition of 1411 individuals (1255 undergoing embryo transplantation) and their associations with reproductive outcomes, clinical and biochemical features are measured, and vaginal samples are 16S rRNA sequenced. Our results suggest that both too high and too low abundance of Lactobacillus is not beneficial for pregnancy; a moderate abundance is more beneficial. A moderate abundance of Lactobacillus crispatus and Lactobacillus iners (~80%) (with a pregnancy rate of I-B: 54.35% and III-B: 57.73%) is found beneficial for pregnancy outcomes compared with a higher abundance (>90%) of Lactobacillus (I-A: 44.81% and III-A: 51.06%, respectively). The community state type (CST) IV-B (contains a high to moderate relative abundance of Gardnerella vaginalis) shows a similar pregnant ratio (48.09%) with I-A and III-A, and the pregnant women in this CST have a higher abundance of Lactobacillus species. Metagenome analysis of 71 samples shows that nonpregnant women are detected with more antibiotic-resistance genes, and Proteobacteria and Firmicutes are the main hosts. The inherent differences within and between women in different infertility groups suggest that vaginal microbes might be used to detect infertility and potentially improve IVF outcomes.

Keywords: IVF; community; infertility; vaginal microbiome.

Figure 1

Correlations among different measurements and composition of the vaginal microbiota. (A) Overview of the workflow for integrated analysis of the sample information, vaginal microbiota, and phenotype in 1411 infertile women. (B) Spearman correlations among 11 measurements. The size of the circle represents |r|, and the statistical significance of the p values was calculated using two‐sided hypothesis testing. (C) Composition of the vaginal microbiota at different levels (namely phylum, genus, and species) according to the median relative abundance. Taxa that made up <1% of the microbiota were combined and labeled as “Others.” (D) Spearman correlations among different genera. *p < 0.05, **p < 0.01, and ***p < 0.001. CS, chronic salpingitis; ED, endometritis; EDO, endometriosis; FB, fallopian problem; FTO, fallopian tube obstruction; HD, hydrosalpinx; INF, inflammatory; IUA, intrauterine adhesion; IVF, in vitro fertilization; OF, ovulation failure; PCOS, polycystic ovarian syndrome; PID, pelvic inflammatory disease; POI, premature ovarian insufficiency; SLP, salpingectomy; SU, scarred uterus.

Figure 2

Diversity and composition of the vaginal microbiota in reproductive‐age women. (A) Principal coordinates analysis (PCoA) of the 1391 samples based on unweighted UniFrac distances at the genus level according to community types of the vaginal microbiota. (B) Species composition of all retained samples (n = 1391) categorized by community state types (CST) assignment according to VALENCIA. CST I‐A almost completely Lactobacillus crispatus; CST I‐B less L. crispatus but still majority; CST II communities are dominated by Lactobacillus gasseri; CST III‐A almost completely Lactobacillus iners; CST III‐B less L. iners but still majority; CST IV‐B contains a high to moderate relative abundance of Gardnerella vaginalis and Atopobium vaginae; CST IV‐C1 dominated by Streptococcus spp.; CST IV‐C3 dominated by Bifidobacterium spp.; CST V communities are dominated by Lactobacillus jensenii. (C) Violon plot of alpha diversity based on the Shannon index. (D) Co‐occurrence bacterial networks in each sub‐CST sample at the species level. Each network was created by computing the co‐occurring bacteria with significant Pearson correlation coefficients. Node properties: (i) circle size, proportional to the normalized and standardized bacterial relative abundances; (ii) color, communities as retrieved by the Louvain algorithm. Edge properties: (i) thickness, proportional to p‐value of the Pearson correlation coefficient, from the most significant (thicker) to the less significant (thinner); (ii) color, red for positive and green for negative Pearson correlation coefficients. (E) The number of unique and shared edges among five CSTs. (F) The centralities (rank of the closeness) and discrepancies of nodes in five CST co‐occurrence networks.

Figure 3

Heatmap of the abundances of microbial taxa significantly correlated with four biochemical and seven clinical observations from 1391 women of reproductive age. (A) Vaginal samples were divided into nine subtypes. (B) The samples based on the abundances of the top 10 genera of vaginal bacterial communities are shown. (C) Complete linkage clustering of taxa based on Spearman's correlation coefficient profiles, which were defined as the set of Spearman's correlation coefficients calculated between the genus composition and the measurement scores of a sample. Yellow tiles indicate positive associations between these measurements and genera; green tiles indicate negative associations. *, **, and *** represent significant differences at p < 0.05, p < 0.01, and p < 0.001, respectively; abs (cor) represent the absolute value of Spearman's correlation coefficient. (The color key is indicated in the upper left corner). (D) Heatmap of the abundances of 15 genera from 1391 women of reproductive age (color key is indicated in the upper right corner). (E) Four biochemical and seven clinical observations for each of the 1391 samples. (F) Shannon diversity indices were calculated for 1391 vaginal samples.

Figure 4

Association of vaginal microbiota composition with in vitro fertilization (IVF) outcomes. (A) Analysis of women who became pregnant in each infertility group compared with control, some women were diagnosed with more than one symptom would be used to compute multiple times. The p value of each group compared to the control was presented in the column. (B) Analysis of pregnancy rate in top and bottom 50% sample for each measurement. Chi‐square test was used to compare IVF outcomes between each group and the control. The p values were presented above the columns of each measurement. (C) The distribution of pregnancy results for each community state type (CST). (D) The composition of the vaginal microbiome for each type at the genus (left panel) and species level (right panel). (E) Associations between five CSTs and bacterial taxa at the species level were identified by Linear discriminant analysis Effect Size (LEfSe) analysis. (F) Associations between the pregnant and nonpregnant women in CST IV‐B and bacterial taxa were identified by LEfSe. BMI, body mass index; c_, class; E2, estradiol; f_, family; g_, genus; T, testosterone; o_, order; P, progesterone.

Figure 5

Composition and function difference of vaginal metagenome between pregnant and nonpregnant women. (A) Composition of the vaginal microbiota at different levels (namely order, family, and genus) according to the median relative abundance. The top four taxa were presented, and the left microbiota was combined and labeled as “others.” (B) Relative abundance of the top two genera and one species in pregnant and nonpregnant women. Number of total annotated genes (C) and antimicrobial resistance genes (ARG) genes (D) in two groups of women. (E) The relative abundances of functionally and phylogenetically annotated orthology classified based on the (Kyoto Encyclopedia of Genes and Genomes) KEGG database.