Longitudinal changes in gut microbiota across reproductive states in wild baboons

Abstract

Background: In humans and other mammals, female reproduction is linked to extensive changes in physiology, immunity, hormones, and behavior. These changes likely shape, and may be shaped by, the composition of gut microbial communities. Characterizing the dynamics of gut microbial change across reproductive states, including its relationship to female physiology, is important for understanding how the gut microbiota influences female and offspring health.

Results: Here we characterize longitudinal changes in gut microbiota across reproduction by combining 16S rRNA gene sequencing data from 4,462 stool samples (spanning 14 years of sample collection) with life history data on multiple reproductive events in 169 female baboons. These baboons were members of a well-studied, natural baboon population in Kenya where reproductive state (ovarian cycling, pregnancy, and postpartum amenorrhea) is tracked daily and microbiota data could be paired with measurements of fecal-derived estrogen, progesterone, and glucocorticoid levels. We found extensive changes in baboon gut microbiota as females transitioned between reproductive states. Pregnancy was linked to distinct patterns of ASV richness, community composition, and taxonomic abundances compared to postpartum amenorrhea and ovarian cycling. The most dramatic shifts occurred as females transitioned from the first to second trimester of pregnancy, with altered abundances of taxa that have been linked, in humans and model systems, to host immunity, weight gain, or hormone levels. Host identity was consistently the strongest predictor of gut microbiota composition across states, and this individual signature was strongest during pregnancy. Estrogen and progesterone levels had robust associations with the gut microbiota overall, but the microbial taxa involved in these associations are reproductive state-dependent. Glucocorticoid concentrations were not a major predictor of gut microbiota composition in any state.

Conclusions: Together, our results support the idea that gut microbiota contribute to the complex physiological changes necessary during pregnancy, but that microbial changes during pregnancy are somewhat unique to each female. Variation in steroid hormones drives some, but not all, of these relationships, emphasizing the importance of considering steroid hormone levels in studies of gut microbiota variation. Our results motivate future work on how gut microbiota contribute to reproductive outcomes, including both maternal and offspring health.

Keywords: Gut microbiota; Papio cynocephalus; pregnancy; primates; reproductive state; steroid hormones.

Figure 1.. Baboon reproductive states and sampling design for this study.

(A) Schematic showing reproductive states in baboons (ovarian cycling, pregnancy, and postpartum amenorrhea), mean lengths of each state (see Methods for description), and relative changes in estrogen, progesterone, and glucocorticoid hormones as female baboons transition between reproductive states. A full reproductive event is on average 638 days (range 333 to 1,084 days). Drawings within each state illustrate the external signals of that state in female baboons. Number of microbiota samples and female hosts are shown within each state. (B) Gut bacterial communities were characterized using 16S rRNA gene sequencing in 4,462 fecal samples collected longitudinally from each female host. The y-axis shows each female host, and each horizontal row of points represents the fecal samples collected for that female as a function of her age in years. Points are colored by reproductive state following the color scheme in (A). Median ages at menarche (4.5 years) and first live birth (5.97 years) are shown with red lines. (C) Number of samples collected (range=1-100, median=21) for each of the 169 female baboons in our dataset, colored by reproductive state.

Figure 2.. Gut microbiota change as females transition between reproductive states and are especially distinct during pregnancy.

(A) ASV richness is higher during pregnancy than ovarian cycling (β=12.8, p<0.001) or PPA (β=10.0, p<0.001). Full model results are in Table S1. (B) The number of bacterial ASVs change in abundance (i.e., fold change ≥1.5 or ≤−1.5 and q<0.05) as females transition between reproductive states, especially for members of the phyla Firmicutes, Bacteroidetes, and Actinobacteria. All changes in abundance are shown in reference to the 30 days prior to conception. Analyses were performed on the 401 ASVs present in 20% or more of samples. Model results used to generate this figure are in Table S3. (C) Average relative abundance of bacterial families changes from one reproductive state to the next. “Unidentified” taxa are those not identified to the family level and “Other” taxa aggregates rare taxa identified to the family level but present in <40% of samples. (D and E) Volcano plots showing the effect of pregnancy compared to (D) ovarian cycling and (E) PPA on the abundances of the 401 ASVs. Each point represents an individual ASV and the color of each point represents the ASV’s assigned phylum. Points above the horizontal red line are statistically significant (q<0.05). Model results are in Table S4. (F) The 50 ASVs with the greatest differences in abundance in samples collected during pregnancy compared to ovarian cycling. Point color represents the assigned phylum of each ASV. The black vertical lines at −1 and 1 represent the minimum possible fold changes. Model results are in Table S4.

Figure 3.. Gut microbiota change across pregnancy, with the “pregnancy microbiota” emerging during trimesters 2 and 3.

(A) Trimesters 2 and 3 have higher ASV richness than trimester 1 (β=19.1, p<0.001 in trimester 2, β=12.1, p=0.007 in trimester 3; Table S5). (B and C) Scatterplots showing the correlation between effect sizes for individual ASV abundances in (B) trimester 2 and trimester 3 compared to reference state trimester 1 and (C) trimester 1 and trimester 3 compared to reference state trimester 2. Each point represents one of the 401 ASVs present in 20% or more of samples, colored by the trimester(s) where there is a significant (q<0.05) difference in CLR-transformed relative abundance from the reference state. R² and p-values are from linear models on the points, with linear model best fit lines shown on the plots. (D, E and F) Volcano plots showing the differences between (D) trimester 2 compared to trimester 1, (E) trimester 3 compared to trimester 1, and (F) trimester 3 compared to trimester 2 on the abundances of the 401 ASVs. Each point represents an individual ASV and the color of each point represents the assigned phylum of the ASV. Points above the horizontal red line are statistically significant (q<0.05). Model results used to generate figures B-F are in Table S7.

Figure 4.. The pregnancy microbiota is more variable and personalized than other reproductive states.

Plots show a temporal autocorrelation analysis of Aitchison similarities between pairs of gut microbiota samples from (A) the same reproductive event in the same host, (B) different reproductive events in the same host, and (C) different hosts in the same social group. Points and lines are colored by reproductive state (green=ovarian cycling; purple=PPA; orange=pregnancy); bands around each line represent the 95% confidence interval. The x-axis shows how many months apart the samples were collected during the reproductive state in question (i.e., the reproductive-months lag). For instance, pairs of samples collected within 30 days of the same time point within a reproductive event are collected within a 1-month lag (the left-most data points on the x-axes); samples collected between 30 and 60 days of the same time point are collected within a 2-month lag (the second left most data points) and so on. Points higher on the y-axis have greater pairwise similarity, while points lower on the y-axis have lower pairwise similarity. These analyses include 4,124 samples from 115 females that had at least 10 total samples, including at least 2 samples from each reproductive state and only reproductive events with at least 2 samples. Statistical comparisons between points are in Table S12.

Figure 5.. Steroid hormones predict ASV richness.

Plots show ASV richness in fecal samples as a function of the concentration of (A) fecal estrogen metabolites (ng/g), (B) fecal progesterone metabolites (ng/g), and (C) fecal glucocorticoid metabolites (ng/g) in that sample. Hormone values are corrected for the time to extraction and assay and mean-centered to 0 (see methods). Each point represents a fecal sample, colored by reproductive state, with linear model best fit line for cycling samples in green, pregnancy samples in orange, and PPA samples in purple. Model results are in Table S1.

Figure 6.. Estrogen and progesterone predict relative abundances of ASVs, with different effects in each reproductive state.

(A, B, and C) Volcano plots showing the effect of fE during pregnancy (A; Table S15), PPA (B; Table S16), and cycling (C; Table S17) on the abundances of the 401 ASVs present in 20% or more of samples. Each point represents an individual ASV and the color of each point represents the ASV’s assigned phyla. (D, E, and F) Volcano plots showing the effect of fP during pregnancy (D; Table S15), PPA (E; Table S16), and cycling (F; Table S17) on the abundances of the 401 microbial ASVs. Each point represents an individual ASV and the color of each point represents the ASV’s assigned phyla. (G, H, I, and J) Scatterplots showing the correlation between effect sizes for individual ASV abundances with (G) fE in cycling and pregnant samples, (H) fE in cycling and PPA samples, (I) fP in cycling and pregnant samples, and (J) fP in cycling and PPA samples. Each point represents one of the 401 ASVs present in 20% or more of samples, colored by the reproductive states(s) where there is a significant (q<0.05) association between CLR-transformed relative abundance of that ASV and fE or fP. R² and p-values are from linear models on the points, with linear model best fit lines shown on the plots.