New Insights into Vaginal Environment During Pregnancy

Abstract

During pregnancy, the vaginal ecosystem undergoes marked changes, including a significant enrichment with Lactobacillus spp. and profound alterations in metabolic profiles. A deep comprehension of the vaginal environment may shed light on the physiology of pregnancy and may provide novel biomarkers to identify subjects at risk of complications (e.g., miscarriage, preterm birth). In this study, we characterized the vaginal ecosystem in Caucasian women with a normal pregnancy (n = 64) at three different gestational ages (i.e., first, second and third trimester) and in subjects (n = 10) suffering a spontaneous first trimester miscarriage. We assessed the vaginal bacterial composition (Nugent score), the vaginal metabolic profiles (¹H-NMR spectroscopy) and the vaginal levels of two cytokines (IL-6 and IL-8). Throughout pregnancy, the vaginal microbiota became less diverse, being mainly dominated by lactobacilli. This shift was clearly associated with marked changes in the vaginal metabolome: over the weeks, a progressive reduction in the levels of dysbiosis-associated metabolites (e.g., biogenic amines, alcohols, propionate, acetate) was observed. At the same time, several metabolites, typically found in healthy vaginal conditions, reached the highest concentrations at the end of pregnancy (e.g., lactate, glycine, phenylalanine, leucine, isoleucine). Lower levels of glucose were an additional fingerprint of a normal vaginal environment. The vaginal levels of IL-6 and IL-8 were significantly associated with the number of vaginal leukocytes, as well as with the presence of vaginal symptoms, but not with a condition of dysbiosis. Moreover, IL-8 concentration seemed to be a good predictor of the presence of vaginal Candida spp. Cytokine concentrations were negatively correlated to lactate, serine, and glycine concentrations, whereas the levels of 4-hydroxyphenyllactate, glucose, O-acetylcholine, and choline were positively correlated with Candida vaginal loads. Finally, we found that most cases of spontaneous abortion were associated with an abnormal vaginal microbiome, with higher levels of selected metabolites in the vaginal environment (e.g., inosine, fumarate, xanthine, benzoate, ascorbate). No association with higher pro-inflammatory cytokines was found. In conclusion, our analysis provides new insights into the pathophysiology of pregnancy and highlights potential biomarkers to enable the diagnosis of early pregnancy loss.

Keywords: miscarriage; pregnancy; vaginal metabolome; vaginal microbiome; women’s health.

FIGURE 1

Box and whiskers representing vaginal molecules, whose concentrations showed highly significant differences (p < 0.0001) between the three trimesters of pregnancy. Differences were searched by Friedman test, followed by Dunn’s Multiple Comparison test. 1 = first trimester; 2 = second trimester; 3 = third trimester.

FIGURE 2

rPCA model built on the centered and scaled concentrations of the metabolites showing significant differences between groups (H vs I vs BV). (A) In the scoreplots, women with a healthy vaginal status (H), an intermediate flora (I) and a BV-related microbiota (BV) are represented in black, red, and green respectively, with lines connecting each subject to the median of its group. (B) The respective boxplots summarize the position of the groups along PC1. In the barplot (C), describing the correlation between the concentration of each molecule and its importance over PC1, dark gray bars highlight statistically significant correlations (p < 0.05).

FIGURE 3

Most significant vaginal metabolites showing different concentrations between women with a normal pregnancy and subjects who had a miscarriage. N = normal pregnancy; M = miscarriage. The graphs display the mean ± SD of each metabolite. Differences were searched by Mann Whitney test.

FIGURE 4

Scatterplots showing some of the most significant correlations between cytokine levels and metabolite concentration. Cytokines levels were positively correlated to glucose concentrations and negatively correlated with lactate levels (p < 0.0001). Raw data were transformed in ranks. Significant correlations were searched with Spearman coefficient after Benjamini-Hochberg correction.