A cross-sectional pilot study of birth mode and vaginal microbiota in reproductive-age women.

Recent studies suggest that birth mode (Cesarean section [C-section] or vaginal delivery) is an important event in the initial colonization of the human microbiome and may be associated with long-term health outcomes. We sought to determine the association between a woman's birth mode and her vaginal microbiota in adulthood. We re-contacted 144 adult women from two U.S. studies and administered a brief survey. Vaginal microbiota was characterized on a single sample by amplicon sequencing of the V3-V4 hypervariable regions of the 16S rRNA gene and clustered into community state types (CSTs). We evaluated the association between birth mode and a CST with low relative abundance of Lactobacillus spp. ("molecular bacterial vaginosis" [Molecular-BV]) compared to Lactobacillus-dominated CSTs in logistic regression modeling which adjusted for body mass index, a confounder in this analysis. Twenty-seven women (19%) reported C-section. Overall, C-section showed a non-significant trend towards increased odds of Molecular-BV (aOR = 1.22, 95% CI: 0.45, 3.32), and Prevotella bivia was the strongest single taxa associated with C-section. However, because the two archived studies had different inclusion criteria (interaction p = 0.048), we stratified the analysis by study site. In the study with a larger sample size (n = 88), women born by C-section had 3-fold higher odds of Molecular-BV compared to vaginally-delivered women (aOR = 3.55, p = 0.06, 95% CI: 0.97-13.02). No association was found in the smaller study (n = 56, aOR = 0.19, p = 0.14, 95% CI: 0.02-1.71). This pilot cross-sectional study suggests a possible association between C-section and Molecular-BV in adulthood. However, the analysis is limited by small sample size and lack of comparability in participant age and other characteristics between the study sites. Future longitudinal studies could recruit larger samples of women, address the temporal dynamics of vaginal microbiota, and explore other confounders, including maternal factors, breastfeeding history, and socioeconomic status, which may affect the relationship between birth mode and vaginal microbiota.

Introduction

There is emerging evidence that babies born by Cesarean section (C-section) are more likely to develop metabolic and chronic disorders, including celiac disease, diabetes mellitus, obesity, food allergy, and asthma, in early childhood, compared to those born by vaginal delivery.[1-7] In addition, the association between C-section delivery and obesity has been shown to persist into adolescence and adulthood.[8] Several of these chronic conditions have been associated with deviations in the colonization of the gut microbiota, with observed decreases in overall bacterial diversity and lack (or delayed colonization) of protective bacterial taxa among individuals born by C-section.[9-11] While evidence to suggest the influence of birth mode on the gut microbiota is building, there is little existing research on the impact of birth mode on the microbiota of the female reproductive tract.

Reproductive-age women with Lactobacillus-dominated vaginal microbiota are at lower risk for bacterial vaginosis (BV), which reduces the likelihood of sexually transmitted infection (STI) acquisition and development of abnormal pregnancy outcomes.[12] While there are a number of studies on demographic and behavioral risk factors for BV in adult women [13-16], less is known about the early risk factors for BV. We hypothesized that a woman’s birth mode, that is, the method through which she was delivered by her mother, is an important early life factor in determining how a woman’s vaginal microbiota is initially seeded and transitions into adulthood. Thus, any differences in the composition of vaginal microbiota attributable to birth mode must persist through known hormonally-driven transitions in the microbiota during early childhood and puberty, including the longitudinal dynamics observed among reproductive-age women in menstruation and pregnancy.

While some prior studies have found short-term differences in microbiota composition and structure at specific body sites between C-section- and vaginally-delivered neonates and infants, [17-19] others found that differences in infant microbiota weeks after delivery could not be explained by birth mode.[20,21] However, these studies did not assess the infant girls’ vaginal microenvironments, and instead focused on the skin, oral, nasal, and/or gut microbiota. In addition, the infants were only observed for up to three years. Therefore, the purpose of this pilot study was to assess whether birth mode was associated with the composition of the vaginal microbiota of adult women in two study populations.

Methods

We used cross-sectional, secondary data and repository samples from the Hormonal Contraception Longitudinal (HCL) Study [22] and the Vaginal Microbiota 400 Woman Study (VM400) [23]. A study coordinator re-contacted participants to query on additional questions related to birth mode and other confounding variables described below. In this study, women were asked to recall whether they were born vaginally or by C-section (i.e., how their mother gave birth to them) and were informed that the question was unrelated to how the adult participants gave birth to their own children.

Secondary data sources

The HCL study aimed to evaluate how the initiation or cessation of hormonal contraceptives (HCs), including oral contraceptive pills, vaginal rings, and other methods, affect the vaginal microbiota and immune responses in the lower genital tract. Eligibility criteria included age (16 to 35 years old), not currently being pregnant, having a uterus (not post-hysterectomy), and not having any implanted uterine devices (such as Mirena or Paraguard). In addition, women diagnosed with illnesses that alter their immune system or hormone levels or who take medications that change their immune system or hormone levels were not eligible. This analysis focused on a mid-vaginal swab that was collected by a clinician at the baseline visit. Between 2011 and 2016, the HCL study enrolled 125 women in the Baltimore, Maryland area.

Women in the VM400 study were recruited for a cross-sectional study at two sites in Baltimore, Maryland and one site in Atlanta, Georgia. Participants ranged from 12 to 45 years old, menstruated regularly, had a history of sexual activity (but none reported in the past 48 hours), reported no vaginal discharge, had not taken antibiotics or antimycotics in the past 30 days, and were not pregnant at the time of enrollment. Participants self-collected two mid-vaginal swabs using the ESwab system (Copan). The 396 participants recruited between 2008 and 2009 represented four racial/ethnic groups (white, black, Asian, and Hispanic) in roughly equal proportions.

Processing of vaginal swabs

For the VM400 study samples, protocols for genomic DNA extraction from vaginal ESwabs (Copan) were carried out in 2009 as previously described.[23] The V3-V4 regions of the 16S rRNA gene were amplified by a one-step polymerase chain reaction (PCR) method.[24] Amplicon pooling, sequencing by Illumina MiSeq, and sequence data processing were conducted as described by Holm et al.[25]

For HCL vaginal samples, protocols were modified in year 2012, and genomic DNA was extracted from ESwabs with either the QS DSP Virus/Pathogen Midi Kit (Qiagen) on the QiaSymphony platform or with the MagAttract Microbial DNA Kit (Qiagen) using a custom automated protocol on the Hamilton Microlab Star if the samples performed poorly (<15,000 reads) in the first round of sequencing. For the QS DSP kit, samples were thawed on ice and a 500μl aliquot from the vaginal swab was used as input.[25] For the MagAttract kit, samples were thawed on ice and a 200μl aliquot from the vaginal swab was used as input for the kit following the manufacturer protocol. Cells were lysed by bead beating on the TissueLyser (Qiagen) at 20Hz for 20 minutes and the final elution volume was 110μl. The V3-V4 regions of the 16S rRNA gene were amplified by two-step PCR, with amplicon pooling, sequencing on an Illumina HiSeq 2500 instrument, and sequence data processing as previously described.[25]

For this analysis, amplicon sequence variants (ASVs) generated by DADA2 were taxonomically classified using the RDP Naïve Bayesian Classifier [26] trained with the SILVA v128 16S rRNA gene database [27] as implemented in the dada2 R package.[28] ASVs of major vaginal taxa were assigned species-level taxonomy using speciateIT.[29] Taxa for which the total read count across all samples was less than 10−5 were regarded as likely contaminants and removed. Individual samples with fewer than 1,000 reads were also removed from the analysis. Vaginal microbiota were broadly clustered into five groups or “community state types” (CSTs) based on their diversity and relative abundance of bacteria.[23] For the purposes of this analysis, a binary outcome variable was created to contrast a low-Lactobacillus CST versus Lactobacillus-dominated states. CSTs dominated by Lactobacillus—L. crispatus (CST I), L. gasseri (CST II), L. iners (CST III), or L. jensenii (CST V)—were grouped together. CST IV, termed “Molecular-BV” [30], is a low-Lactobacillus state characterized by the abundance of BV-associated anaerobic organisms, including Gardnerella, Megasphera, Sneathia, and Prevotella. CSTs were assigned for each sample in this study based on similarity to the centroid of each CST as determined from a pool of over 13,000 archived vaginal samples.[31]

Primary data collection

For the current study, the research team attempted to recruit all former HCL and VM400 study participants who had consented to re-contact for future research (over 450 women). Participants were contacted by phone or email and were given the option to complete a survey by phone or in an online survey collected and managed with REDCap.[32] Informed consent was received for the current research prior to administering the survey. While a few participants were younger than 18 years when enrolled in the parent studies, all were older than 18 at re-contact. Participants had the opportunity to ask questions and decline further involvement at any time.

In the short survey, participants reported their birth mode (C-section or vaginal delivery), age at first menstrual period, and weight status at menarche. As the VM400 study did not collect height and weight information for body mass index (BMI) calculation, former VM400 participants provided their current height, current weight, and estimated weight at the time of parent study sampling. In addition, former VM400 participants were queried on breastfeeding history questions, which had been included in baseline surveys administered to HCL participants. There were many missing values for breastfeeding history questions with only 37.5% of VM400 participants providing responses. Behavioral and demographic data collected in the parent studies were also included in the analysis.

The institutional review boards at the University of Maryland School of Medicine and Johns Hopkins University provided approval for the protocol, including the re-contact of parent study participants.

Statistical analysis

To compare distributions of selected health, behavioral, and demographic characteristics between women born by C-section and vaginal delivery, medians, frequencies, and percentages were calculated, and p-values were obtained using chi-square, Fisher’s exact, or Wilcoxon rank sum tests. Logistic regression was used to examine the association between birth mode and Molecular-BV. As the two archived studies differed in participant characteristics and inclusion criteria, we planned a priori to assess potential effect modification due to study site. Several confounders identified in the literature as being associated with birth mode and/or Molecular-BV were considered for inclusion in the final adjusted model, including race/ethnicity, age, body mass index (BMI) at the time of sampling, and age and weight status at the first menstrual period. Because L. iners-dominated CST III is often associated with transitions to a Molecular-BV state [33], we conducted sensitivity analyses in which CSTs III and IV were each individually compared to the other Lactobacillus-dominated states (CSTs I, II, and V) in a multinomial model. These analyses were conducted using STATA/SE 14.2 (Stata Corporation, College Station, Texas). In addition, classification and regression tree (CART) analysis (performed in R [R Foundation for Statistical Computing, Vienna, Austria]) and linear discriminant analysis effect size (LEfSe) [34] were used to determine whether specific taxa differed significantly in relative abundance between women born by C-section or vaginal delivery. A heatmap, also created in R, included the 25 most abundant bacterial taxa found in the samples and were color-coded based on the relative abundance of each taxa in the samples.

Ethics approval and consent to participate

The described research was performed in accordance with the Declaration of Helsinki and was approved by the institutional review boards of the University of Maryland College Park (reference #999286), University of Maryland Baltimore (HP-00073692 & HM-HP-00040935-14), and Johns Hopkins University (NA_00043112/CIR00024424). All participants provided either oral or written informed consent prior to completing phone or online surveys.

Results

A total of 144 women were enrolled in this analysis, 88 from the HCL study and 56 from the VM400 study. At the time of sampling (parent study enrollment), the average age was 27.5 years (standard deviation [SD] = 5.4, range = 17 to 44), 57.6% (n = 83) were nulliparous, and 9.2% (n = 13) reported no past sexual partners. No women reported current menstruation at the time of sampling, although one HCL woman reported that her enrollment visit coincided with the last day of her menstrual period. The median number of days since the end of the last menstrual period prior to sampling was 15 days in both studies.

Twenty-seven (18.8%) women reported birth by C-section. Women born by C-section were more likely to report higher BMI (≥ 30 kg/m2) and lifetime douching (both p < 0.05, Table 1). Women born by C-section and vaginal delivery were similar in terms of other demographic, behavioral, or health characteristics, including days since last menstrual period at time of sampling. However, there were significant differences between the former HCL and VM400 participants (S1 Table). Former HCL women were more likely to report younger age (≤ 23 years) at enrollment, BMI of < 25 kg/m2, past douching, using tampons only at last menstrual period, and younger age (≤ 12 years) at menarche (all p<0.05). Though not significantly associated when categorized (p = 0.22), there was a moderate difference in the number of years since participants were last pregnant between HCL (median = 2.8 years) and VM400 (median = 4.7 years) women when assessed continuously (p = 0.10). No VM400 women reported vaginal discharge at enrollment (an exclusion criterion); however, they were more likely to report vaginal itching or discharge in the prior two months compared to HCL women (both p < 0.05). The distributions of menstrual cycle phases were not statistically different between the studies (p = 0.40). In addition, ten (11.4%) HCL participants reported antibiotic use within the past two months of the baseline parent study visit (not statistically different between C-section- or vaginally-delivered groups), whereas antibiotic use within one month was an exclusion criterion for the VM400 study.

Table 1

Characteristics of study participants (N = 144).

	C-Section (n = 27)		Vaginal delivery (n = 117)
	n	%	n	%	P-valuee
Race/ethnicity					0.99
Black or Latina	9	33.3	40	34.2
Non-black and non-Latina	18	66.7	77	65.8
Age at study entry					0.66
17 to 23	10	37.0	33	28.2
24 to 30	10	37.0	51	43.6
31 & over	7	25.9	33	28.2
Body mass index a					0.03
≤24.9	16	59.3	58	50.0
25.0–29.9	2	7.4	31	26.7
≥30.0	9	33.3	27	23.3
Community state type					0.51
CST-I, L. crispatus-dominated	11	40.7	45	38.5
CST-II, L. gasseri-dominated	1	3.7	2	1.7
CST-III, L. iners-dominated	8	29.6	38	32.5
CST-IV, Low Lactobacillus	7	25.9	23	19.7
CST-V, L. jensenii-dominated	0	0.0	9	7.7
Bacterial vaginosis diagnosis within 2 months					0.61
Yes	1	3.7	6	5.1
No	26	96.3	111	94.9
Vaginal pH b					0.34
4.0–4.5	18	66.7	83	72.8
4.6–5.0	3	11.1	18	15.8
>5.0	6	22.2	13	11.4
Vaginal symptoms in prior 2 months
Discharge	4	14.8	15	12.8	0.76
Itching	2	7.4	4	3.4	0.31
Ever been pregnant					0.53
No	17	63.0	66	56.4
Yes	10	37.0	51	43.6
Parity					0.33
0	18	66.7	79	67.5
1	2	7.4	19	16.2
2+	7	25.9	19	16.2
Ever given birth vaginally					0.58
No	20	74.1	87	74.4
Yes	7	25.9	30	25.6
Years since last pregnancy b					0.75
Never pregnant	17	65.4	66	57.4
3 or less	4	15.4	23	20.0
more than 3	5	19.2	26	22.6
Age at menarche d					0.18
≤12	18	66.7	59	50.4
>12	8	29.6	54	46.2
Weight status at menarche c					0.49
Average or below	23	85.2	93	79.5
Overweight or above	4	14.8	20	17.1
Hormonal contraceptive use (current)					0.54
No	14	51.9	62	53.0
Yes	13	48.1	55	47.0
Number of sexual partners in the prior 2 months c					0.55
None	7	25.9	19	16.8
1	19	70.4	89	78.8
2+	1	3.7	5	4.4
Douched (ever)					0.03
No	17	63.0	98	83.8
Yes	10	31.0	19	16.2
Hygiene product use (2 months)
Feminine towellette	1	3.7	11	9.4	0.47
Hygiene spray	0	0.0	3	7.0	0.99
Hygiene powder	1	8.3	2	4.8	0.54
Other product	2	7.4	10	8.6	0.60
Sanitary product use at last menstrual period a					0.15
Tampon only	8	30.8	37	31.6
Sanitary napkin only	3	11.5	33	28.2
Tampon and sanitary napkin	15	57.7	47	40.2

a. Missing for 1 participant

b. Missing for 3 participants

c. Missing for 4 participants

d. Missing for 5 participants

e. P-values obtained from chi-square or Fisher's exact tests

As previously reported [23], five CSTs were identified in this study population (Fig 1). Samples identified as Molecular-BV (CST IV), representing 20.8% (n = 30) of the study population, were compared to samples with high-Lactobacillus states (i.e., CST I, II, III, and V). Overall, birth mode was not significantly associated with Molecular-BV (CST IV); however, the OR estimate suggested increased odds of CST IV with C-section delivery (aOR = 1.22, p = 0.70, 95% CI: 0.45–3.32) (Table 2). BMI, the only variable that was associated with both C-section delivery and Molecular-BV, was the only confounder included in the adjusted models. Due to significant effect modification based on parent study site (interaction p = 0.048), we also stratified the analysis. In the HCL study stratum, women born by C-section had 3-fold higher odds of having low-Lactobacillus vaginal communities compared to vaginally-delivered women (n = 88, aOR = 3.55, p = 0.056, 95% CI: 0.97–13.02). No association was found within the VM400 study stratum (n = 56, aOR = 0.19, p = 0.14, 95% CI: 0.02–1.71). Sensitivity analyses excluding L. iners-dominated CST III from the high-Lactobacillus reference showed similar trends for the relationship between Molecular-BV and C-section. In a multinomial logistic regression analysis of HCL study participants, the odds of having vaginal microbiota characterized as CST III was not associated with C-section (aOR = 1.29, p = 0.51, 95% CI: 0.31–5.22) and the odds of having CST IV retained its magnitude of association with C-section (aOR = 3.91, p = 0.06, 95% CI: 0.95–16.23) compared to other Lactobacillus-dominated CSTs.

Fig 1

Heatmap of bacterial relative abundance from cross-sectional sample previously collected from 144 women enrolled in the HCL and VM400 studies.

Table 2

Association between C-section delivery and molecular-BV, combined (N = 144) and stratified by parent study.

		Molecular-BV (n = 30) n (%)	Lactobacillus-dominated (n = 114) n (%)	Unadjusted		Adjusted
Study	Birth Mode			OR (95% CI)	P-value	aOR (95% CI)	P-value
HCL & VM400	C-section (n = 27)	7 (25.9)	20 (74.1)	1.43 (0.54, 3.79)	0.47	1.22 (0.45, 3.32)	0.70
	Vaginal (n = 117)	23 (19.7)	94 (80.3)	1 (ref)		1 (ref)
HCL	C-section (n = 16)	6 (37.5)	10 (62.5)	3.33 (1.00, 11.03)	0.049	3.55 (0.97, 13.01)	0.056
	Vaginal (n = 72)	11 (15.3)	61 (84.7)	1 (ref)		1 (ref)
VM400	C-section (n = 11)	1 (9.1)	10 (90.9)	0.28 (0.03, 2.38)	0.24	0.19 (0.02, 1.71)	0.14
	Vaginal (n = 45)	12 (26.7)	33 (73.3)	1 (ref)		1 (ref)

Crude and adjusted odds ratios (ORs and aORs), 95% confidence intervals (CIs), and p-values were calculated using logistic regression. Percentages shown are row percents. Adjusted model controlled for BMI (categorized ≤24.9 [ref], 25.0–29.9, & ≥30). Molecular-BV (CST IV) was compared to Lactobacillus-dominated states (CST I, II, III, & V).

Phylotype-level analyses confirmed that C-section delivery was not associated with high abundance of Lactobacillus species. In CART modelling, Prevotella bivia was the taxon with the strongest association with birth mode, and C-section-delivered women were more likely to have greater relative abundance (≥ 2.8%) of this bacteria (Fig 2A). High relative abundance of P. bivia was also associated with C-section in the LEfSe analysis; however, this result was non-significant (p = 0.23; Fig 2B). LEfSe also suggested that greater abundances of L. jensenii and L. iners were non-significantly associated with vaginal delivery.

Fig 2

Phylotype-level analyses results.

(A) In the pruned classification tree, each node shows the most likely birth mode (C-section or vaginal) given the bacterial relative abundance, the probability of C-section delivery in that node, and the percentage of observations in the node. P. bivia relative abundance was the taxon with the strongest association with birth mode history. If the relative abundance of P. bivia was less than 2.8%, the probability of birth mode being C-section was 16%. If the relative abundance of P. bivia was greater than 2.8%, the probability of birth mode being C-section-delivered was 71%. (B) The bacterial taxa with the highest effect sizes (linear discriminant analysis [LDA] scores > 4) reflect marked abundance in one birth mode group and not in the other. Three taxa were differentially abundant at this level, with P. bivia being more abundant in C-section-delivered group and L. jensenii and L. iners being more abundant in the vaginally-delivered group. However, these differences were not statistically significant (0.05 < p < 0.30).

Discussion

In this cross-sectional pilot study of 144 reproductive-age women, we found modest evidence to suggest that a woman’s vaginal microbiota in adulthood is associated with her mode of birth history. Among women recruited from the HCL study, being born by C-section was associated with a 3-fold increase in the odds of having the low-Lactobacillus CST IV (Molecular-BV) at a single time point in adulthood. P. bivia, a species often found in CST IV, was the bacterial taxon most strongly associated with birth mode, with higher relative abundances more often observed among C-section-delivered women.

We found qualitative interaction between study site and birth mode in the pooled analysis, which led us to stratify results by parent study. As with all subgroup analyses, the stratified results should be interpreted with caution as there is high likelihood of spurious findings. In addition, the sample sizes within strata were small, leading to concern for inadequate power. However, the low p-value of the test for interaction indicated low likelihood that the subgroup differences were due to chance.[35] The subgroup differences that were observed could be linked to marked differences between the study sites in age at sampling, weight, personal hygiene behaviors, and recent antibiotic use. As HCL participants were younger in age and were less likely to have given birth vaginally, we could reasonably hypothesize that the effect of birth mode could be stronger in this population. While the swabs used between the two studies were the same, the sample processing and sequencing methods differed. Our group’s recent paper comparing vaginal samples processed using the dual-indexing one-step and two-step library preparation methods on the Illumina MiSeq (one-step and two-step) and HiSeq (two-step only) platforms found complete within-subject agreement among the CST assignments from all three methods.[25] Therefore, we have evidence to suggest limited variability in the CST outcome measurements between studies. CST in this study was determined from one vaginal sample collected at baseline. Our group and others have confirmed that the vaginal microbiota is dynamic for many women and can change at various points in the menstrual cycle and in pregnancy.[33,36-40] Basing the overall CST assignment on one sampling day may have resulted in misclassification of the typical bacterial composition of the participants. However, longitudinal studies have also demonstrated that women tend to stay in the same plane of fluctuation over many months and possibly years of observation.[33,38] Women who are found in CST IV at a single time point tend to most often fluctuate between IV and the L. iners-dominated CST III, which may represent an enduring pattern that favors Molecular-BV.[33] In addition, from an epidemiologic perspective, if there was non-differential misclassification of CSTs, that would serve to dampen the point estimates toward the null.[41] A finding of higher likelihood of Molecular-BV with CST IV, even within a cross-sectional analysis, still provides evidence for concern and could be indicative of long-term risk for BV and STIs.[42] The characterization of the vaginal microbiota was based on relative abundances of bacterial taxa. Although absolute abundances estimating the total bacterial loads [43] would complement the relative abundance results, they were not calculated in this analysis. Absolute abundance data provided by species-specific qPCR would also allow for observation of rare taxa and their relationship to birth mode. Our phylotype-level results indicating a higher relative abundance of P. bivia among women born by C-section reiterates the Molecular-BV findings in the CST-level analysis. Prevotella spp. have been associated with both BV and increased inflammation.[44-48] Si et al. also reported that Prevotella is a heritable bacteria that is associated with genetic variants of pro-inflammatory cytokines (interleukin-5) and obesity risk.[49]

BMI was the only variable associated with both birth mode and CST in the pooled analyses, and thus it was the only confounder included in the adjusted model. There is evidence to suggest that overweight mothers are more likely to have overweight daughters.[50,51] This mother-daughter correlation of BMI led us to interpret this variable as a proxy for the mother’s BMI, and a mother’s BMI is often related to her delivery mode.[52,53] In addition, Mueller et al. demonstrated that overweight or obese offspring were more likely to be born to obese mothers by C-section, compared to children born vaginally to normal weight mothers.[54]

Our study is limited by missing variables on important maternal characteristics that may act as confounders, such as maternal age, the composition of the mother’s vaginal microbiota preceding delivery, and intrapartum antibiotic exposure among mothers who gave birth by vaginal delivery. We did not collect the indication for C-section delivery or whether the surgery was elective or emergent. While these, and likely other variables, are undoubtedly important potential confounders, it may be difficult for adult daughters to recall these details accurately. In addition, there were many missing values in the breastfeeding variable, prohibiting use of this variable in assessing confounding. Like birth mode, there are conflicting findings in the literature on the effect of breastfeeding on the gut microbiota,[19] so it is not certain whether this factor would influence our analysis. Though the inability to adjust for breastfeeding history and other birthing/infancy characteristics may indicate unresolved confounding, the finding that these factors are not reliably obtained from adult women through direct questioning will inform future studies.

Another limitation of this study was the lack of accounting for the sociodemographic characteristics of both the mother and adult daughter within the analysis. It is known that a mother’s socioeconomic status (SES) can influence whether she gives birth by C-section or vaginal delivery.[55] Indicators of SES, such as education and income, are also linked to BV prevalence.[56] While the racial/ethnic breakdown for re-contacted participants was similar between studies, it is likely that there were differences in SES between former HCL and VM400 participants. Unfortunately, SES information was not collected from VM400 participants at enrollment in the parent study or during the follow-up. Future studies should consider incorporating SES factors when assessing the intergenerational determinants for increased BV risk.

Our study has several strengths, including the efficient combination of archived samples and primary data collection to answer a novel research topic. To our knowledge, there are no available data on vaginal microbiota in adulthood as it relates to birth mode. Overall, the efforts to re-contact past participants were successful. We re-engaged over 72% of the former HCL participants who consented to be contacted for future research. However, former VM400 participants were less responsive, with less than 15% participating in the follow-up. The time interval between the end of the parent study and re-contact for this study on birth mode was much greater for VM400 than for HCL. Many participant phone numbers and email addresses had been disconnected. Self-selection bias was a minor concern, as women who participated in multiple follow-ups for the parent studies have proven to be very motivated and may be different from women who refused to participate. Only 37.5% of VM400 participants were able to report whether or not they were breastfed, which hindered further analysis of this variable. However, participants showed greater confidence when answering all other items, including their birth mode, in phone and online questionnaires, and in the few instances when participants were unsure of an answer, they were prompted to confirm their answers with a family member. The adult women were uniformly very confident in reporting their birth mode; any misclassification is likely to be independent of the Molecular-BV outcome.

There is little data published on how the vaginal microbiome is initially seeded. The establishment of stable adult vaginal microbiota may rely on several early environmental factors. There is conflicting evidence on the extent to which the acquisition of infant microbiota at several body sites is determined by birth mode.[57,58] While babies born by C-section delivery do not transit through the environment of the mothers’ vaginal bacteria, they still acquire human microbes shortly after birth, possibly through contact with the operating room environment [59] as well as through breastfeeding, diaper changes, and other close contact. After infancy, the vaginal microbiome undergoes significant transitions. In early childhood, girls tend to be colonized by stable aerobic, anaerobic, and enteric bacteria.[60] Important findings by Hickey et al. suggest that the vaginal microbiota of girls begins to resemble those of adults (typically dominated by Lactobacillus spp.) before menarche, while girls are still in the early and middle stages of puberty.[61] It is thought that the composition and function of the vaginal microbiota change in puberty due to increased estrogen production.[62] However, it remains unclear how girls transition and colonize adult-like vaginal microbiota, and further studies are needed to confirm the hypothesis that any effect of birth mode on the vaginal microbiota persists through infancy and puberty to adulthood.

Despite these controversies, there is evidence to support the hypothesis that birth mode influences health later in life. Babies born by C-section have an increased risk of childhood-onset type 1 diabetes and obesity at age 3 years, outcomes that may be related to the gut microbiota.[2,6] Most longitudinal studies investigating the effect of birth mode and neonatal exposures have followed infants for 3 years or less. For example, Chu et al. described a cohort study of 81 mother and infant dyads in which multiple body sites, including stool, oral gingiva, nares, skin and vagina, were sampled for 6 weeks.[20] The authors observed minor variations in the neonates’ microbiota community structure associated with C-section delivery in most body sites immediately after birth. However, while the infants’ microbiota matured substantially at each body site, the authors reported no discernable differences in bacterial community structure within gut, nares, and oral cavity between infants delivered vaginally or C-section at 6 weeks of age. In a recent pilot study, Dominguez-Bello et al. reported that C-section infants who were seeded with vaginal secretions from their mother had gut, oral, and skin bacterial communities enriched in vaginal bacteria after 30 days, similar to infants born vaginally.[63] These studies did not collect infant vaginal samples to determine how the vaginal microbiota undergoes reorganization in early life, which underscored the need for longitudinal studies assessing the seeding of bacteria in an infant girl’s vaginal microenvironment.

BV is highly recurrent, associated with a number of adverse reproductive health outcomes, and is not effectively treated or prevented by current therapies.[16,64] Birth mode may represent a potentially modifiable risk factor. Broadly, interventions to prevent BV may include reducing C-section rates, reseeding the infant girl microbiome, or probiotic therapies administered to C-section-born women at various phases of the lifespan. Conclusive evidence from interventional studies would provide the causal link between birth mode and the composition and structure of the vaginal microbiota.

In summary, this novel pilot study found a moderately significant association between the low-Lactobacillus Molecular-BV state and being born by C-section, which indicates that C-section delivery may be related to having a less protective microbiota in adulthood. Most covariates we analyzed, including race, factors of puberty, and personal hygiene behaviors, did not alter the strength of the relationship between adulthood CST and birth mode. However, BMI was shown to be associated with C-section delivery and CST in this study sample. Although this cross-sectional study of a relatively small convenience sample may be limited by unmeasured confounding, and we are unable to establish causality or characterize the temporal dynamics of the vaginal microbiota, results suggest that birth mode is associated with vaginal seeding in the infant vaginal microbiome that may persist to adulthood.

(DOCX)

Click here for additional data file.

11 Oct 2019

PONE-D-19-22430

A cross-sectional study of birth mode and vaginal microbiota in reproductive-age women

PLOS ONE

Dear Dr Brotman,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The reviewer's comments  are addressed below.

The comments and opinions of the reviewers were quite diverse and even opposed. However the subject of your study merits attention. Indeed, your report may initiate the discussion and encourage other researcher to conduct well designed, longitudinal, studies to elucidate the questions and hypothesis you raised in the submitted manuscript.

The manuscript will need a major revision, we agree with reviewer 1 that the reported study lacks power and that various confounders were not taken into account.   We also want to remind you of the journal’s criterion “experiments must have been conducted rigorously, with appropriate controls and replication, sample sizes must be large enough to produce robust results and methods and reagents must be described in sufficient detail for another researcher to reproduce the experiments described”.

We ask you to carefully address the comments of the first reviewer. Take into consideration to increase the sample size, to perform additional statistical analyses and to assess if the different procedural methods may have introduced a study to study variance.

We concur with reviewer 2 that a supplementary table presenting the study populations side by side may be helpful for the readers.

We would appreciate receiving your revised manuscript by 1st of December 2019. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Tania Crucitti

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide

3. Our internal editors have looked over your manuscript and determined that it is within the scope of our The Microbiome Across Biological Systems Call for Papers. This collection of papers is headed by a team of Guest Editors for PLOS ONE: Zaid Abdo, Colorado State University, USA; Sanjay Chotrimall, Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore; Noelle Noyes, University of Minnesotta, USA;  Pankaj Trivedi, Colorado State University, USA; and Thomas Dawson, A*STAR, Singapore. The Collection will encompass a diverse range of research articles about microbiomes and human health, the natural and built environment, and new technologies used to study microbiomes. Additional information can be found on our announcement page: https://collections.plos.org/s/microbiome. If you would like your manuscript to be considered for this collection, please let us know in your cover letter and we will ensure that your paper is treated as if you were responding to this call. If you would prefer to remove your manuscript from collection consideration, please specify this in the cover letter.

Additional Editor Comments (if provided):

You stated that all data are made available, however, you should clarify on how this will be in addition to the reads that you will deposit in the SRA achive.

You cite ref 22 when referring the HCL study, this reference is not easily accesible, please consider another reference or a website where the report can be accessed.

Under limitations of the study we also would  like you to discuss the use of relative versus absolute abundance data and possible impact on your results.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Reviewer #1: In this paper by Stennett and colleagues, attempts are made to associate birth mode with vaginal microbiota in adulthood. As currently presented, there are a number of issues with the article, particularly in the study design and subsequent interpretation of data, that prevent the paper being acceptable for publication. While the hypothesis of the study is interesting- that birth mode may influence likelihood of development of BV-like vaginal microbiota later in life- the study design is insufficient to enable the primary research question to be addressed in any meaningful way. The authors themselves acknowledge a lack of power which prevents associations (positive or negative) to be determined. However, there are other major concerns, some of which I have outlined below.

- The introduction fails to provide sufficient background and consideration of major known influences of vaginal microbiota composition. Chiefly, hormonally driven mechanisms of vaginal microbiota shaping throughout a woman’s life span, the lasting impact of pregnancy on the vaginal microbiota, menses etc. These are particularly important given the cross-sectional design and low power of the study, which makes the results highly prone to type I and II errors.

- While the multinomial logistic regression analyses provided a useful way adjust for 1 confounder (BMI), from what can be pertained from the methods, other likely confounders were not assessed and therefore could not be controlled for. For example, when in the menstrual cycle were base line samples for both cohorts collected? Were women asked if they had been recently pregnant? The latter is particularly pertinent given that women born by C-section were more likely to report giving birth within one year prior to enrolment in the parent study. Work from David Relman’s group and others have shown that pregnancy has a lasting impact on vaginal composition and is associated with a strong shift towards a “molecular BV” like composition for more than a year in some women. Thus, it is possible that the trend increase in molecular-BV type vaginal communities associated with delivery by C-section is due to a higher proportion of these women giving birth recently.

- Large amounts of missing data on breastfeeding at the time of sampling (72.5% missing) is also a major limitation. It is too superficial to discard its importance on vaginal microbiota composition by comparing it to the “conflicting findings in the literature on the effect of breastfeeding on the gut microbiota” as done so in the discussion. Breast feeding status would influence both hormonal levels as well as recommencement of menstruation and thus likely be an important confounder.

- Parts of the Methods are unclear and poorly described. For example, what is meant by “Taxa present at less than 10-5 across all samples were removed….”. Are you describing percentage of total read counts here?

- Given the difference in extractions, swabs used, PCR (one-step v two-step) sequencing platforms etc between the VM400 and HCL cohorts, how was study-to-study variance in resulting microbiota composition assessed? Where long-term reference samples used? If not, some samples should be split and analysed using both extraction and sequencing protocols side by side to ensure no bias was inadvertently introduced.

- Why were both chi-square and Fisher’s exact tests used? What were the assumptions for both?

- Was clustering and subsequent assignment of CSTs performed using the 25 most abundant bacterial taxa or all taxa?

- What are the “factors of puberty” that were considered in the adjusted models?

- Race and ethnicity seem to be used interchangeably throughout and it appears as though these terms were not used consistently across the VM400 and HCL cohorts.

- It is erroneous to suggest, as done so in the abstract and elsewhere in the paper, that Prevotella bivia or “molecular BV” are predictors of C-section. Care should be taken to not confuse interpretations of statistical testing with prediction of an event that occurred in the past (c-section)!

- The premise that birth mode may have a lasting impact on vaginal microbiota composition in later life demands more careful consideration and analysis of underlying clinical phenotypes at birth. While birth mode likely impacts early colonisation events of the vagina, other key exposures and events around the time of birth will also have a major influence (e.g. antibiotic exposure at time of delivery, being born prematurely, breastfeeding, etc). Without such detailed consideration, the findings presented herein remain spurious and unfortunately, not supported by the data presented.

Reviewer #2: Stennett et al have produced an interesting study that addresses a question that surprisingly has not been previously been looked at: whether differences in the vaginal microbiome stem from a woman's early life exposure to her mother's microbiota at the time of birth. The authors recontacted, and with full informed consent, surveyed women who had previously participated in two cross-sectional cohort studies (HCL and VM400) of the vaginal microbiome to determine the mode by which they were birthed. The study's primary analysis combined the cohorts to achieve greater statistical power, but each cohort was also examined separately. Multivariate logistic regression analysis found a non-statistically significant trend toward Molecular-BV with Cesarean section delivery, a trend that appeared to be driven largely by the HCL cohort. The last finding of interest in the analysis of individual bacterial taxa was that P. bivia was the best predictor of Cesarean birth.

This manuscript is well crafted and the analyses are appropriately conducted. The authors were cautious in their interpretation of the results and included a thorough discussion of the strengths and limitations of their study.

I recommend acceptance of the manuscript and offer a few minor recommendations to strengthen the manuscript, though I do not feel the manuscript's acceptance needs to be conditioned upon these modifications:

1) A reference that seemed to be missing and relevant is the Cell H&M paper by Si et al (2016) that described vaginal Prevotella as being a highly heritable. It is not clear at this point whether the P. bivia in the vaginal microbiotas of women born by Cesarean came from their mothers, but it is worth considering potential mechanisms that could support such a link.

2) As I was reading I frequently went looking for a table that described the two cohorts separately. It would have been nice to examine how relevant factors such as race and parity segregated across the cohorts in tabular format. Though a good amount of information was in the text, one's own curiosity and expediency of reference would be facilitated by a supplemental table.

3) Given a growing appreciation that having given birth has a long-lasting effect on the vaginal microbiome (MacIntyre et al. 2015 Scientific Report), I feel strongly that parity should be included in Table 1. It does not matter so much whether a woman has herself delivered vaginally or by Cesarean, but more whether she has had ever delivered a baby at term, or not.

4) Line 151 - was >> were

Reviewer #3: The current study represents an important step further in understanding the occurrence of BV. The field is slowly growing and the effect of C-section has been linked to number of serious health complications, but indeed as the authors point in their study not to much effort has been made to understand the effect of mode of delivery on the initial vaginal colonization and how this can affect women's life later. This makes the current study an interesting starting point to better understand the role of C-section and will clearly give the opportunities for longitudinal studies in the future.

The manuscript is clear, well organized and the the perform analysis is well selected.

I have only a few minor suggestions to the authors:

1. Discussion is rather long section, so I would advice the authors to shorter it a bit

2. Figures are not very clear, so authors should consider uploading figures with  higher resolution

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

10 Jan 2020

We uploaded this response to reviewers with formatting as well.

We thank the reviewers for their careful review and detailed feedback. The reviewers’ suggestions have led to significant improvements in this revision. Comments and response to reviewers are listed below.

Reviewer #1: In this paper by Stennett and colleagues, attempts are made to associate birth mode with vaginal microbiota in adulthood. As currently presented, there are a number of issues with the article, particularly in the study design and subsequent interpretation of data, that prevent the paper being acceptable for publication. While the hypothesis of the study is interesting- that birth mode may influence likelihood of development of BV-like vaginal microbiota later in life- the study design is insufficient to enable the primary research question to be addressed in any meaningful way. The authors themselves acknowledge a lack of power which prevents associations (positive or negative) to be determined. However, there are other major concerns, some of which I have outlined below.

Response: A limitation of our research is the relatively low participation in the follow-up survey among former VM400 participants. However, we believe the study still provides important information when interpreted in context. While this convenience sample is not the definitive study on this topic, this novel work will serve as the impetus for further research and will inform future study designs and sample size calculations. In the revision, we refer to the analysis as a “pilot” study.

Revisions: “A cross-sectional pilot study of birth mode and vaginal microbiota in reproductive-age women” Title

“Therefore, the purpose of this pilot study was to assess whether birth mode was associated with the composition of the vaginal microbiota of adult women in two study populations.” Introduction, line 74

“In this cross-sectional pilot study of 144 reproductive-age women, we found modest evidence to suggest that a woman’s vaginal microbiota in adulthood is associated with her mode of birth history.” Discussion, line 257

“In summary, this novel pilot study found a moderately significant association between the low-Lactobacillus Molecular-BV state and being born by C-section, which indicates that C-section delivery may be related to having a less protective microbiota in adulthood.” Discussion, line 389

1) The introduction fails to provide sufficient background and consideration of major known influences of vaginal microbiota composition. Chiefly, hormonally driven mechanisms of vaginal microbiota shaping throughout a woman’s life span, the lasting impact of pregnancy on the vaginal microbiota, menses etc. These are particularly important given the cross-sectional design and low power of the study, which makes the results highly prone to type I and II errors.

Response: In the introduction, we led with the public health significance of our study, describing research on long-term health effects of C-section delivery, and prior investigations on direct seeding of mothers’ microbiota to infants. In addition, some of the life course context is provided in the discussion—research characterizing the vaginal microbiota in early childhood (line 366) and puberty (line 361), as well as the longitudinal dynamics of the microbiota with respect to menstruation and pregnancy in reproductive-age (line 284). In the revised introduction section, we now introduce the concept that the effect of birth mode must persist through life course transitions.

Revisions: “Thus, any differences in the composition of vaginal microbiota attributable to birth mode must persist through known hormonally-driven transitions in the microbiota during early childhood and puberty, including the longitudinal dynamics observed among reproductive-age women in menstruation and pregnancy.” Introduction, line 65

2) While the multinomial logistic regression analyses provided a useful way adjust for 1 confounder (BMI), from what can be pertained from the methods, other likely confounders were not assessed and therefore could not be controlled for. For example, when in the menstrual cycle were base line samples for both cohorts collected? Were women asked if they had been recently pregnant? The latter is particularly pertinent given that women born by C-section were more likely to report giving birth within one year prior to enrolment in the parent study. Work from David Relman’s group and others have shown that pregnancy has a lasting impact on vaginal composition and is associated with a strong shift towards a “molecular BV” like composition for more than a year in some women. Thus, it is possible that the trend increase in Molecular-BV type vaginal communities associated with delivery by C-section is due to a higher proportion of these women giving birth recently.

Response: Menstruation at the time of sampling was an exclusion criterion for the VM400 study, and the former VM400 participants included in this analysis were sampled between 4 and 30 days (median=15, IQR=9 - 22) after their last menstrual period ended. In contrast, six out of 88 former HCL participants reported fewer than 4 days since last their last menstrual period (with 1 participant reporting that her period ended on the same day as her enrollment visit), and 6 participants reported greater than 40 days since last period at the enrollment visit (median=15.5, IQR=9 – 25). Several women in the HCL study were taking long-acting reversible contraceptives (LARCs), which affect regular menstrual cycles and contributed to the greater variability in this study population.

We assessed menstrual cycle continuously and categorized at several different cut points and found no significant differences by birth mode, Molecular-BV status, or parent study. In the revision, we noted in the results section that our analyses did not appear to be affected by menstrual cycle phase.

The years since last pregnancy variable was included in Table 1 in the previous submission. This variable was not associated with Molecular-BV in this sample; however, a greater proportion of women in the vaginal delivery group reported less than 3 years since last pregnancy compared to C-section-born women. In the revision, we discuss moderate differences in this variable between parent studies.

Revisions: “No women reported current menstruation at the time of sampling, although one HCL woman reported that her enrollment visit coincided with the last day of her menstrual period. The median number of days since the end of the last menstrual period prior to sampling was 15 days in both studies.” Results, line 188

“Women born by C-section and vaginal delivery were similar in terms of other demographic, behavioral, or health characteristics, including days since last menstrual period at time of sampling.” Results, line 194

“Though not significantly associated when categorized (p = 0.22), there was a moderate difference in the number of years since participants were last pregnant between HCL (median = 2.8 years) and VM400 (median = 4.7 years) women when assessed continuously (p = 0.10). No VM400 women reported vaginal discharge at enrollment (an exclusion criterion); however, they were more likely to report vaginal itching or discharge in the prior two months compared to HCL women (both p < 0.05). The distributions of menstrual cycle phases were not statistically different between the studies (p = 0.40).” Results, line 199

3) Large amounts of missing data on breastfeeding at the time of sampling (72.5% missing) is also a major limitation. It is too superficial to discard its importance on vaginal microbiota composition by comparing it to the “conflicting findings in the literature on the effect of breastfeeding on the gut microbiota” as done so in the discussion. Breast feeding status would influence both hormonal levels as well as recommencement of menstruation and thus likely be an important confounder.

Response: Low participation in the breastfeeding survey question is a limitation. However, we believe the data remain useful to inform future study designs. Adult participants were less confident reporting their breastfeeding history compared to their birth mode history. Researchers who are considering study designs that ask adult women to self-report these factors can expect lower participation on some birthing/infancy survey questions than others. Though incomplete variables may lead to analytic limitations, prospective studies that follow participants from birth to adulthood are infeasible.

Revision: “Though the inability to adjust for breastfeeding history and other birthing/infancy characteristics may indicate unresolved confounding, the finding that these factors are not reliably obtained from adult women through direct questioning will inform future studies Discussion, line 316

4) Parts of the Methods are unclear and poorly described. For example, what is meant by “Taxa present at less than 10-5 across all samples were removed….”. Are you describing percentage of total read counts here?

Response: We revised this sentence to be clearer about the rationale behind and the criteria for removing taxa and samples with few reads in the sequencing dataset.

Revision: “Taxa for which the total read count across all samples was less than 10-5 were regarded as likely contaminants and removed. Individual samples with fewer than 1,000 total bacterial reads were also removed from the analysis.” Methods, line 127

5) Given the difference in extractions, swabs used, PCR (one-step v two-step) sequencing platforms etc. between the VM400 and HCL cohorts, how was study-to-study variance in resulting microbiota composition assessed? Where long-term reference samples used? If not, some samples should be split and analysed using both extraction and sequencing protocols side by side to ensure no bias was inadvertently introduced.

Response: Our group recently published a comparison of sample processing and sequencing methods and reported that CST assignments were consistent across the cited methods. We have added this information to the discussion of inter-study differences in the revised manuscript.

Revision: “While the swabs used between the two studies were the same, the sample processing and sequencing methods differed due to advancements made in sequencing technology over time. Our group’s recent paper comparing vaginal samples processed using the dual-indexing one-step and two-step library preparation methods on the Illumina MiSeq (one-step and two-step) and HiSeq (two-step only) platforms found complete within-subject agreement among the CST assignments from all three methods.[25] Therefore, we have evidence to suggest limited variability in the CST outcome measurements between the studies.” Discussion, line 273

6) Why were both chi-square and Fisher’s exact tests used? What were the assumptions for both?

Response: To examine the association between two categorical variables, the chi-square test applies an approximation assuming the sample is large. The Fisher's exact test runs an exact procedure that is more appropriate for small sample sizes, especially when some of the expected cell counts in a 2x2 table are less than five. For example, there was only one C-section-delivered woman who was diagnosed with BV within two months prior to sampling. The Fisher’s exact test was more appropriate to assess the association between birth mode and prior BV diagnosis due to the low cell count.

7) Was clustering and subsequent assignment of CSTs performed using the 25 most abundant bacterial taxa or all taxa?

Response: The CST assignments were based on all taxa (excluding contaminants) identified in 13,000+ archived samples. The heatmap only shows the 25 most abundant taxa identified in the 144 samples included in this analysis.

8) What are the “factors of puberty” that were considered in the adjusted models?

Response: The puberty-related variables we considered were age and weight status at the first menstrual period. This has been clarified in the revision.

Revision: “Several confounders identified in the literature as being associated with birth mode and/or Molecular-BV were considered for inclusion in the final adjusted model, including race/ethnicity, age, body mass index (BMI) at the time of sampling, and age and weight status at the first menstrual period.” Methods, line 169.

9) Race and ethnicity seem to be used interchangeably throughout and it appears as though these terms were not used consistently across the VM400 and HCL cohorts.

Response: Thank you for pointing out this oversight. The instance in the methods where race was used without ethnicity has been corrected. Both VM400 and HCL collected this variable similarly; ethnicity (i.e., Hispanic/Latina heritage) was not collected separately from race in either study, which is more common in newer demographic questionnaires.

Revision: “Several confounders identified in the literature as being associated with birth mode and/or Molecular-BV were considered for inclusion in the final adjusted model, including race/ethnicity, age, body mass index (BMI) at the time of sampling, and age and weight status at the first menstrual period.” Methods, line 169.

10) It is erroneous to suggest, as done so in the abstract and elsewhere in the paper, that Prevotella bivia or “molecular BV” are predictors of C-section. Care should be taken to not confuse interpretations of statistical testing with prediction of an event that occurred in the past (csection)!

Response: We agree that this wording obscures the implied direction of association and have revised the abstract, results, and discussion.

Revisions: “In CART modelling, Prevotella bivia was the taxon with the strongest association with birth mode, and C-section-delivered women were more likely to have greater relative abundance (≥ 2.8%) of this bacteria (Fig 2A). High relative abundance of P. bivia was also associated with C-section in the LEfSe analysis; however, this result was non-significant (p = 0.23; Fig 2B). LEfSe also suggested that greater abundances of L. jensenii and L. iners were non-significantly associated with vaginal delivery.” Results, line 236

“(A) In the pruned classification tree, each node shows the most likely birth mode (C-section or vaginal) given the bacterial relative abundance, the probability of C-section delivery in that node, and the percentage of observations in the node. P. bivia relative abundance was the taxon with the strongest association with birth mode history.” Fig 2 caption, line 243

“P. bivia, a species often found in CST IV, was the bacterial taxon most strongly associated with birth mode, with higher relative abundances more often observed among C-section-delivered women.” Discussion, line 261.

11) The premise that birth mode may have a lasting impact on vaginal microbiota composition in later life demands more careful consideration and analysis of underlying clinical phenotypes at birth. While birth mode likely impacts early colonisation events of the vagina, other key exposures and events around the time of birth will also have a major influence (e.g. antibiotic exposure at time of delivery, being born prematurely, breastfeeding, etc). Without such detailed consideration, the findings presented herein remain spurious and unfortunately, not supported by the data presented.

Response: Certainly, there are considerable limitations to this secondary data analysis. However, we still consider the results of this pilot study to be informative as they provide a useful first look at a novel research topic.

Revision: “Although this cross-sectional study of a relatively small convenience sample may be limited by unmeasured confounding, and we are unable to establish causality or characterize the temporal dynamics of the vaginal microbiota, results suggest that birth mode is associated with vaginal seeding in the infant vaginal microbiome that may persist to adulthood.” Discussion, line 394

Reviewer #2: Stennett et al have produced an interesting study that addresses a question that surprisingly has not been previously been looked at: whether differences in the vaginal microbiome stem from a woman's early life exposure to her mother's microbiota at the time of birth. The authors recontacted, and with full informed consent, surveyed women who had previously participated in two cross-sectional cohort studies (HCL and VM400) of the vaginal microbiome to determine the mode by which they were birthed. The study's primary analysis combined the cohorts to achieve greater statistical power, but each cohort was also examined separately. Multivariate logistic regression analysis found a non-statistically significant trend toward Molecular-BV with Cesarean section delivery, a trend that appeared to be driven largely by the HCL cohort. The last finding of interest in the analysis of individual bacterial taxa was that P. bivia was the best predictor of Cesarean birth.

This manuscript is well crafted and the analyses are appropriately conducted. The authors were cautious in their interpretation of the results and included a thorough discussion of the strengths and limitations of their study.

I recommend acceptance of the manuscript and offer a few minor recommendations to strengthen the manuscript, though I do not feel the manuscript's acceptance needs to be conditioned upon these modifications:

1) A reference that seemed to be missing and relevant is the Cell H&M paper by Si et al (2016) that described vaginal Prevotella as being a highly heritable. It is not clear at this point whether the P. bivia in the vaginal microbiotas of women born by Cesarean came from their mothers, but it is worth considering potential mechanisms that could support such a link.

Response: We thank the reviewer for this valuable feedback. The results of this relevant paper have been added to the paper.

Revision: “Prevotella spp. have been associated with both BV and increased inflammation.[41-45] Si et al. also reported that Prevotella is a heritable bacteria that is associated with genetic variants of pro-inflammatory cytokines (interleukin 5) and obesity risk.[46]” Discussion, line 296

2) As I was reading, I frequently went looking for a table that described the two cohorts separately. It would have been nice to examine how relevant factors such as race and parity segregated across the cohorts in tabular format. Though a good amount of information was in the text, one's own curiosity and expediency of reference would be facilitated by a supplemental table.

Response: We agree that this table would be beneficial to the reader. S1 Table with the breakdown of the covariates by Parent study has been included with this revised submission.

Revision: “However, there were a few significant differences between the former HCL and VM400 participants (S1 Table).” Results, line 196

3) Given a growing appreciation that having given birth has a long-lasting effect on the vaginal microbiome (MacIntyre et al. 2015 Scientific Report), I feel strongly that parity should be included in Table 1. It does not matter so much whether a woman has herself delivered vaginally or by Cesarean, but more whether she has had ever delivered a baby at term, or not.

Response: We have added parity to our revised tables.

Revisions: See Tables 1 and S1.

4) Line 151 - was >> were

Response: We have made the correction.

Revision: “There were many missing values for breastfeeding history questions with only 37.5% of VM400 participants providing responses.” Methods, line 155

Reviewer #3: The current study represents an important step further in understanding the occurrence of BV. The field is slowly growing and the effect of C-section has been linked to number of serious health complications, but indeed as the authors point in their study not too much effort has been made to understand the effect of mode of delivery on the initial vaginal colonization and how this can affect women's life later. This makes the current study an interesting starting point to better understand the role of C-section and will clearly give the opportunities for longitudinal studies in the future.

The manuscript is clear, well organized and the perform analysis is well selected.

I have only a few minor suggestions to the authors:

1. Discussion is rather long section, so I would advise the authors to shorter it a bit.

Response: We looked for areas in the discussion with unnecessary details. For instance, we removed some detail in the strengths paragraph (starting on line 334).

Deletion: “In phone interviews, three women expressed being unsure about their answers initially, particularly their ages at menarche, and were given the opportunity to contact family members to confirm. All three re-contacted research staff to confirm their responses, with one woman changing her age at menarche response. Limited missing data on the questionnaires administered online (including few “I don’t know” responses and skips for items other than breastfeeding history) indicated that women who participated online were similarly confident in their answers.”

Revision: “However, participants showed greater confidence when answering all other items, including their birth mode, in phone and online questionnaires, and in the few instances when participants were unsure of an answer, they were prompted to confirm their answers with a family member.” Discussion, line 343

2. Figures are not very clear, so authors should consider uploading figures with higher resolution.

Revision: The revised versions of Figures 1 & 2 now have higher resolution.

Additional Editor Comments:

1. You stated that all data are made available, however, you should clarify on how this will be in addition to the reads that you will deposit in the SRA archive.

Response: All VM400 survey and sequence data used in this manuscript are submitted to dbGaP. We are currently authorized by the Johns Hopkins School of Medicine IRB to release three baseline HCL variables: race, age, and hormonal contraceptive status to dbGaP. However, additional survey data are available directly from the PI (Ghanem). We are in the process of receiving dbGaP ascension numbers for HCL.

Revision: “The VM400 data can be found at the National Center for Biotechnology Information (NCBI) Database of Genotypes and Phenotypes (dbGaP) under accession number phs001909.v1.p1. For the HCL study, dbGaP accession numbers are in process. Additional survey data and accession numbers (when available) can be accessed directly from the Principal Investigator (Ghanem).” Declarations, line 408

2. You cite ref 22 when referring the HCL study, this reference is not easily accesible, please consider another reference or a website where the report can be accessed.

Response: The reference has been updated so that it is easier to access.

Revision: “22. Tuddenham S, Ghanem K, Gajer P, Robinson C, Ravel J, Brotman R. P591 The effect of hormonal contraception on the vaginal microbiota over 2 years. Sex Transm Infect. 2019;95: A263–A264. doi:10.1136/sextrans-2019-sti.662”

3. Under limitations of the study we also would like you to discuss the use of relative versus absolute abundance data and possible impact on your results.

Response: We have revised the discussion to address this point.

Revision: “The characterization of the vaginal microbiota was based on relative abundances of bacterial taxa. Although absolute abundances estimating the total bacterial loads [43] would complement the relative abundance results, they were not calculated in this analysis. Absolute abundance data provided by species-specific qPCR would also allow for observation of rare taxa and their relationship to birth mode.” Discussion, line 292.

Submitted filename: Response to Reviewers 011020_RB_submit.docx

Click here for additional data file.

21 Jan 2020

A cross-sectional pilot study of birth mode and vaginal microbiota in reproductive-age women

PONE-D-19-22430R1

Dear Dr. Brotman,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Tania Crucitti

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

16 Mar 2020

PONE-D-19-22430R1

A cross-sectional pilot study of birth mode and vaginal microbiota in reproductive-age women

Dear Dr. Brotman:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Tania Crucitti

Academic Editor

PLOS ONE