Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The association between ethnicity and vaginal microbiota composition in Amsterdam, the Netherlands

  • Hanneke Borgdorff,

    Roles Data curation, Formal analysis, Investigation, Project administration, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Amsterdam Institute for Global Health and Development, Amsterdam, The Netherlands, Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands

  • Charlotte van der Veer,

    Roles Investigation, Writing – review & editing

    Affiliations Public Health Laboratory, Public Health Service of Amsterdam (GGD), Amsterdam, The Netherlands, Department of Medical Microbiology and Infection Control, VU University Medical Center, Amsterdam, The Netherlands

  • Robin van Houdt,

    Roles Investigation, Resources, Supervision, Writing – review & editing

    Affiliation Department of Medical Microbiology and Infection Control, VU University Medical Center, Amsterdam, The Netherlands

  • Catharina J. Alberts,

    Roles Data curation, Project administration, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Infectious Diseases, Public Health Service of Amsterdam (GGD), Amsterdam, the Netherlands

  • Henry J. de Vries,

    Roles Conceptualization, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Infectious Diseases, Public Health Service of Amsterdam (GGD), Amsterdam, the Netherlands, Department of Dermatology, Academic Medical Center, Amsterdam, The Netherlands

  • Sylvia M. Bruisten,

    Roles Investigation, Resources, Supervision, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Public Health Laboratory, Public Health Service of Amsterdam (GGD), Amsterdam, The Netherlands

  • Marieke B. Snijder,

    Roles Data curation, Project administration, Resources, Writing – review & editing

    Affiliations Department of Public Health, Academic Medical Center, Amsterdam, The Netherlands, Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Academic Medical Center, Amsterdam, The Netherlands

  • Maria Prins,

    Roles Funding acquisition, Resources, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Infectious Diseases, Public Health Service of Amsterdam (GGD), Amsterdam, the Netherlands, Department of Internal Medicine, Division of Infectious Diseases, Academic Medical Center, Amsterdam, The Netherlands

  • Suzanne E. Geerlings,

    Roles Conceptualization, Supervision, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Internal Medicine, Division of Infectious Diseases, Academic Medical Center, Amsterdam, The Netherlands

  • Maarten F. Schim van der Loeff,

    Roles Conceptualization, Supervision, Writing – review & editing

    Affiliations Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Infectious Diseases, Public Health Service of Amsterdam (GGD), Amsterdam, the Netherlands, Department of Internal Medicine, Division of Infectious Diseases, Academic Medical Center, Amsterdam, The Netherlands

  • Janneke H. H. M. van de Wijgert

    Roles Conceptualization, Formal analysis, Funding acquisition, Methodology, Supervision, Writing – original draft, Writing – review & editing

    j.vandewijgert@liverpool.ac.uk

    Affiliations Amsterdam Institute for Global Health and Development, Amsterdam, The Netherlands, Center for Infection and Immunity Amsterdam, Academic Medical Center, Amsterdam, The Netherlands, Department of Clinical Infection, Microbiology and Immunology, Institute of Infection and Global Health, University of Liverpool, Liverpool, United Kingdom

The association between ethnicity and vaginal microbiota composition in Amsterdam, the Netherlands

  • Hanneke Borgdorff, 
  • Charlotte van der Veer, 
  • Robin van Houdt, 
  • Catharina J. Alberts, 
  • Henry J. de Vries, 
  • Sylvia M. Bruisten, 
  • Marieke B. Snijder, 
  • Maria Prins, 
  • Suzanne E. Geerlings, 
  • Maarten F. Schim van der Loeff
PLOS
x

Abstract

Objective

To evaluate whether ethnicity is independently associated with vaginal microbiota (VMB) composition in women living in Amsterdam, the Netherlands, as has been shown for American women.

Methods

Women (18–34 years, non-pregnant, N = 610) representing the six largest ethnic groups (Dutch, African Surinamese, South-Asian Surinamese, Turkish, Moroccan, and Ghanaian) were sampled from the population-based HELIUS study. Sampling was performed irrespective of health status or healthcare seeking behavior. DNA was extracted from self-sampled vaginal swabs and sequenced by Illumina MiSeq (16S rRNA gene V3-V4 region).

Results

The overall prevalence of VMBs not dominated by lactobacilli was 38.5%: 32.2% had a VMB resembling bacterial vaginosis and another 6.2% had a VMB dominated by Bifidobacteriaceae (not including Gardnerella vaginalis), Corynebacterium, or pathobionts (streptococci, staphylococci, Proteus or Enterobacteriaceae). The most prevalent VMB in ethnically Dutch women was a Lactobacillus crispatus-dominated VMB, in African Surinamese and Ghanaian women a polybacterial G. vaginalis-containing VMB, and in the other ethnic groups a L. iners-dominated VMB. After adjustment for sociodemographic, behavioral and clinical factors, African Surinamese ethnicity (adjusted odds ratio (aOR) 5.1, 95% confidence interval (CI) 2.1–12.0) and Ghanaian ethnicity (aOR 4.8, 95% CI 1.8–12.6) were associated with having a polybacterial G. vaginalis-containing VMB, and African Surinamese ethnicity with a L. iners-dominated VMB (aOR 2.8, 95% CI 1.2–6.2). Shorter steady relationship duration, inconsistent condom use with casual partners, and not using hormonal contraception were also associated with having a polybacterial G. vaginalis-containing VMB, but human papillomavirus infection was not. Other sexually transmitted infections were uncommon.

Conclusions

The overall prevalence of having a VMB not dominated by lactobacilli in this population-based cohort of women aged 18–34 years in Amsterdam was high (38.5%), and women of sub-Saharan African descent were significantly more likely to have a polybacterial G. vaginalis-containing VMB than Dutch women independent of modifiable behaviors.

Introduction

The majority of women of reproductive age have a vaginal microbiota (VMB) dominated by lactobacilli [1]. VMB not dominated by lactobacilli are increasingly being recognized as a cause of adverse reproductive health outcomes, such as increased acquisition and transmission of HIV (reviewed in [2]) and preterm birth (reviewed in [3]). The clinical condition known as bacterial vaginosis (BV) is thought to be the most common type of VMB not dominated by lactobacilli [1, 4]. Molecular studies in the last decade have consistently identified the following anaerobes with high relative abundance in BV cases: Gardnerella vaginalis, Atopobium vaginae, BVAB1-3, Mobiluncus species, Prevotella species, Sneathia/Leptotrichia species, Megasphaera species, among others [3]. Other types of VMBs not dominated by lactobacilli, such as VMBs containing a high relative abundance of pathobionts, have also been identified but less commonly [1]. Until recently, many research groups lumped all VMBs not dominated by lactobacilli together into one community state type (CST IV), which was originally described by Ravel et al in 2011 [5]. An expert consultation organized by the U.S. National Institutes of Health (NIH) in 2015 called for more detailed descriptions of the various types of VMBs not dominated by lactobacilli and, eventually, subcategorization into multiple CSTs [6]. This is deemed necessary to improve our understanding of the etiology and pathogenesis of different CSTs and their sequelae, and to develop targeted interventions.

Another important finding from molecular studies in the last decade is that not all lactobacilli are equal [1]: a VMB dominated by Lactobacillus crispatus is considered more beneficial than a VMB dominated by L. iners, because the latter is associated with a higher risk of developing a VMB not dominated by lactobacilli [7] and a higher prevalence of sexually transmitted infections (STIs) [8].

Known risk factors for BV are unprotected sexual activity (reviewed in [9]), smoking [10], some types of vaginal cleansing practices (reviewed in [2]), the presence of STIs [11, 12], low levels of sex hormones [1315], and African ethnicity (reviewed in [16]). Molecular studies in asymptomatic American women showed that African-American women are more likely to have a diverse VMB or a L. iners-dominated VMB than Caucasian or Asian-American women [5, 1719]. Molecular studies have also shown that the prevalence of L. crispatus-dominated VMB is much lower in sub-Saharan Africa than in other parts of the world [20, 21]. However, whether the association between ethnicity and VMB composition is based on sociodemographic, behavioral, environmental, or genetic factors is still a matter of debate [22, 23]. Molecular studies conducted to date were small and did not adequately control for confounding factors, and few studies have been done in multi-ethnic European populations [1]. Comparisons within and between multi-ethnic populations in different parts of the world may provide the new insights required to disentangle the effects of sociodemographic, behavioral, environmental, and genetic factors on VMB composition.

In this study, we characterized VMB compositions of women representing the six largest ethnic groups in Amsterdam, the Netherlands. In keeping with the recommendation by the recent U.S. NIH consultation, we provide detailed descriptions of all diverse VMBs that we identified. We also determined factors associated with VMB composition with a focus on ethnicity.

Materials and methods

Study population

We used baseline data and samples from the HELIUS study, a multi-ethnic cohort study in Amsterdam, The Netherlands [24]. The HELIUS study aims to unravel the causes of the unequal burden of disease across ethnic groups residing in Amsterdam, and was approved by the Medical Ethics Committee of the Academic Medical Center in Amsterdam (protocol number: 10/100; amendment 10/100# 10.17.1729; NL32251.018.10). All participants provided written informed consent.

People aged 18–70 years were randomly sampled by ethnic origin (Dutch, Surinamese, Moroccan, Turkish, and Ghanaian) through the municipality register of Amsterdam [25]. For the Dutch sample, we invited people who were born in the Netherlands and whose parents were born in the Netherlands. For the Turkish sample, Turkish ethnicity was defined as: (a) born in Turkey and having at least one parent who was born in Turkey (first generation) or (b) born in the Netherlands but both parents were born in Turkey (second generation). A similar definition was used for the other ethnic minority groups (Surinamese, Moroccan and Ghanaian). After data collection, the Surinamese group was further stratified according to self-reported ethnic origin into ‘African’ (of West-African descent), ‘South-Asian’ (of North-Indian descent), ‘Javanese’, or ‘other’. In the VMB sub-study, we included African Surinamese women and women of North-Indian Surinamese or Javanese Surinamese descent (collectively referred to in this paper as South-Asian Surinamese) as two separate groups.

Between January 2011 and December 2015, HELIUS baseline data were collected for 24,789 participants. For the VMB sub-study, a subsample of women enrolled between 1 January 2011 and 31 December 2013 was selected: all women aged 18–34 years who had provided a self-sampled vaginal swab and had completed the questionnaire on sexual behavior were eligible for selection, and seven female participants per ethnic group per life year (or all available participants if fewer than seven were available) were randomly selected [26].

Study procedures

Participants attended the study clinic for one study visit. They provided data on sociodemographics, behaviors, and medical history online or during the study visit. Participants who were unable to complete the questionnaire were offered assistance from a trained, ethnically matched interviewer. They were asked to bring a sample of their first morning urine to their study visit, which was tested by a urine reagent test strip (Combur 7, Roche). During the study visit, participants underwent a physical examination and biological sample collection. All women were asked to participate in the VMB sub-study by self-collecting a vaginal swab (Copan Diagnostics Inc., Murrieta, CA, USA) during the study visit, but refusal to do so did not render them ineligible for the HELIUS study. The vaginal swabs were kept at 2–8°C at the study site for up to six days, and then transported to, and stored at -20°C, at the Public Health Laboratory until testing. Women who were pregnant or less than 6 months postpartum were asked to postpone their HELIUS participation.

VMB characterization

Each vaginal swab was eluted in 800 μl phosphate-buffered saline and divided into four aliquots. One aliquot each was tested for Trichomonas vaginalis (Aptima Trichomonas vaginalis assay, HOLOGIC, Bedford, MA, USA), Chlamydia trachomatis and Neisseria gonorroeae (Aptima Combo 2, HOLOGIC, Bedford, MA, USA), and human papillomavirus (HPV) (SPF10-PCR-DEIA/LiPA25 system version 1, DDL, Voorburg, the Netherlands) [26], and the fourth aliquot was used for VMB characterization. Samples were lysed using chemical lysis (lysozyme, mutanolysin (Sigma Aldrich, USA, St. Louis), and lysostaphin (AMBI, USA, New York)) and bead beating (MagnaPure, Roche Diagnostics, Switzerland, Basel), as described by Ravel [5]. Lysed samples were further treated using proteinase K and RNase A (Thermo Fisher, USA, Waltham). DNA was extracted using the ChemaGen extraction robot (PerkinElmer, Germany, Baesweiler). Dual indexed universal primers (319F and 806R) were used for PCR amplification of the V3-V4 regions of the 16S rRNA genes as described by Fadrosh [27]. PCR products were pooled and normalized, and purified with Agencourt AMPure XP magnetic beads (BeckmanCoulter, Fullerton, USA). Paired-end sequencing was performed on the Illumina MiSeq instrument (San Diego, CA, USA).

Preprocessing of raw sequence data

Sequence data were preprocessed as described by Fadrosh [27] with minor modifications. Barcodes were trimmed off unaligned sequences. The unaligned sequences were then quality filtered by truncating reads with an average phred score <20 over a 30 bp sliding window with subsequent removal of sequences truncated to less than 75% of their original length as well as their paired sequences using Trimmomatic [28]. PANDAseq [29] was used for assembling paired sequences, error correction, and additional quality filtering: paired-end sequences that did not overlap or resulted in a sequence of <400 bp were removed. Sequences were matched to their corresponding barcode using the demultiplex tool in QIIME (version 1.8.0) [30] and removed when not matched to a barcode, being less than 75% of the original length after quality truncation, or having any ambiguous base calls. One mismatch was allowed during removal of PCR primer sequences (Tagcleaner [31]). The USEARCH tool in QIIME [32] was used for operational taxonomic unit (OTU) picking based on 97% sequence similarity and Greengenes version 13.8 was used as a reference for chimera detection [33]. Sequences were aligned to sequences in a vaginal reference package [34] using the PyNAST method in QIIME.

Taxonomic classification

Taxonomic classification of OTUs was performed using two methods: 1) pplacer suite [35] with the vaginal reference package [34] as reference taxonomy and 2) RDP Classifier in QIIME with Greengenes version 13.8 [33] as reference taxonomy. The 16S rDNA BLAST function of the NCBI database was used when the two methods produced discordant results or when identification at species level was not possible (for OTUs with >1% relative abundance in at least one sample only). OTUs that were assigned to the same species were merged. Extremely rare OTUs (a total read count of less than 10) were removed. This resulted in 455 OTUs, with 96% mapping to species level, 3% to genus level, and 1% to higher taxonomic levels.

Data analysis

Statistical analyses were performed using R version 3.1.3 (R Development Core Team, 2015) and STATA release 12 (StataCorp, College Station, TX, USA). All analyses were cross-sectional and contained one sample per participant. Samples with <100 high quality reads were excluded from subsequent analyses. Richness (number of OTUs) and diversity (Shannon Diversity Index) were calculated for each sample at a rarefied read count of 100. Hierarchical clustering was performed using Euclidean distance and complete linkage, and the resulting 20 clusters were visualized in a heatmap. Thirteen of these 20 clusters contained fewer than 10 samples. Clusters were therefore pooled based on microbiological characteristics to form meaningful VMB groups (see Table 1 for a detailed description of how this was done). The resulting eight VMB groups were described and their distribution across ethnic groups assessed. Sociodemographic, behavioral, and clinical characteristics were compared between HELIUS participants who did or did not provide a vaginal swab (to explore the magnitude and direction of participation bias), the six ethnic groups, and the three largest VMB groups (the other five VMB groups were too small to allow for multivariable modeling). The three largest VMB groups were: 1) L. crispatus-dominated VMB; 2) L. iners-dominated VMB; and 3) polybacterial G. vaginalis-containing VMB (Table 1). In unadjusted analyses comparing women who did or did not provide a vaginal swab, and the six ethnic groups, two-sided Chi-squared or Fishers’ exact tests were used for categorical data, and two-sided Kruskal Wallis tests for continuous data. Two separate multivariable logistic regression models were used to determine correlates of the L. iners-dominated VMB group and polybacterial G. vaginalis-containing VMB group, respectively, each compared to the L. crispatus-dominated VMB group. All covariates that were considered risk factors for a non-lactobacilli-dominated VMB a priori, and were associated with a VMB group in the unadjusted analyses at p<0.1, were included in the multivariable models. The proportion of women who used antibiotics in the past two weeks was small, and not statistically different among ethnic groups or VMB groups (see results). Furthermore, a sensitivity analysis excluding recent antibiotic users did not significantly alter our results and conclusions (data not shown).

thumbnail
Table 1. VMB cluster composition descriptions, Amsterdam, the Netherlands (2011–2015).

https://doi.org/10.1371/journal.pone.0181135.t001

Results

Study population

Of the 1,469 women aged 18–34 years who attended a HELIUS baseline visit before January 2014, 980 (67%) self-sampled a vaginal swab and 924 (63%) also completed the sexual behavior questionnaire. Up to seven women per ethnic group and per life year were randomly selected from these 924 women, which resulted in a sample size of 610 women. Dutch, African Surinamese, South-Asian Surinamese, and Ghanaian women were more likely to provide a vaginal swab (94%, 87%, 75%, and 71%, respectively) than Turkish and Moroccan women (37% and 49%, respectively; p<0.01; [26]).

The median age of the 610 selected women was 27 years (IQR 22–30), and did not differ between ethnic groups by design. However, other sociodemographic characteristics, substance use, reproductive health history, sexual behaviors, and clinical characteristics differed substantially (Table 2). Dutch women were more likely to have attained higher vocational or university education and to report use of tobacco, cannabis and alcohol, and less likely to report being religious and to have children. Turkish and Moroccan women were most likely to be married, Dutch and Surinamese women were most likely to currently use hormonal contraception, and Dutch and African Surinamese women were most likely to report sexual risk behaviors. Dutch and African Surinamese women also had the highest prevalence of any HPV and high risk HPV (both p<0.01) [24]. The prevalences of Neisseria gonorrhoeae, Chlamydia trachomatis, and Trichomonas vaginalis were low and not statistically significantly different across the ethnic groups.

thumbnail
Table 2. Characteristics of female participants by ethnic group, Amsterdam, the Netherlands (2011–2015).

https://doi.org/10.1371/journal.pone.0181135.t002

VMB compositions in the overall study population

The median number of high quality sequences per sample was 25,392 (IQR 4,654–43,798). Samples with fewer than 100 sequences (n = 64) were discarded from further analyses; these were equally distributed across ethnic groups (data not shown). Hierarchical clustering of the remaining 546 samples resulted in 20 clusters (Fig 1, Table 1), which were pooled into eight microbiologically meaningful VMB groups as outlined in Table 1.

thumbnail
Fig 1. Hierarchical clustering of participants by VMB composition.

AV = Atopobium vaginae; BVAB1 = BV-associated bacterium 1; GV = Gardnerella vaginalis; L = Lactobacillus; OTU = operational taxonomic unit; VMB = vaginal microbiota. A. The 20 VMB clusters obtained by hierarchical clustering of 546 vaginal samples. Clustering was based on the relative abundance of 455 OTUs. B. Heatmap showing the relative abundance of the most abundant OTUs using the color key on the left.

https://doi.org/10.1371/journal.pone.0181135.g001

Three VMB groups were dominated by lactobacilli: L. crispatus-dominated VMB (n = 131, 24.0% of women), L. iners-dominated VMB (n = 187, 34.2%), and ‘other Lactobacillus-dominated’ VMB (pooling clusters dominated by L. gasseri, L. jensenii, and L. delbrueckii; n = 18, 3.3%). Most women in these groups had a VMB with more than 50% relative abundance of the relevant Lactobacillus species, and women with less than 50% of that species typically had additional Lactobacillus species. Other bacteria were sometimes present but in much smaller relative abundance, most commonly G. vaginalis and Corynebacterium. Pathobionts (streptoccoci, staphylococci, Proteus, and Enterobacteriaceae) were present in 14 women in a relative abundance of 1–14%. Overall, for all ethnic groups combined, 61.5% of the women had a lactobacilli-dominated VMB and 38.5% had another type of VMB.

Three of the five VMB groups that were not dominated by lactobacilli resembled BV. These included ‘polybacterial G. vaginalis-containing VMB (pooling three polybacterial clusters with high relative abundance of G. vaginalis, A. vaginae and BVAB1 [4]; n = 134, 24.5% of women), ‘G. vaginalis-dominated VMB’ (n = 29, 5.3%), and ‘other anaerobic polybacterial VMB’ (pooling two polybacterial clusters with low relative abundance of G. vaginalis, A. vaginae and BVAB1 but high relative abundance of Mobiluncus, Prevotella species or other anaerobes; n = 13, 2.4%). All of these women had a polybacterial VMB characterized by the anaerobes that have typically been associated with BV (see Introduction and Table 1), often in combination with L. iners and several other minority taxa. Bifidobacteriaceae (not including G. vaginalis) were uncommon in these VMB groups. The final two of five VMB groups not dominated by lactobacilli included a VMB group dominated by Bifidobacteriaceae (not including G. vaginalis) or Corynebacterium (n = 14, 2.6%) and a VMB group dominated by pathobionts (n = 20, 3.7%). Overall, 32.2% of all women had a VMB resembling BV and another 6.2% had a VMB dominated by Bifidobacteriaceae (not including G. vaginalis), Corynebacterium, or pathobionts.

A total of 33 women (6.0%) distributed over all eight VMB groups had 1% or more relative abundance of a Streptococcus species, and an additional 18 women (3.3%) had 1% or more relative abundance of another pathobiont.

Unadjusted associations between ethnicity and VMB groups

The distribution of the eight VMB groups was significantly different among ethnic groups (Fig 2). The most prevalent VMB group in Dutch women was a L. crispatus-dominated VMB (38%), in South-Asian Surinamese, Turkish and Moroccan women a L. iners-dominated VMB (40%, 41%, and 35%, respectively), and in African Surinamese and Ghanaian women a polybacterial G. vaginalis-containing VMB (39% and 34%, respectively). Turkish and Moroccan women more often had a VMB dominated by Bifidobacteriaceae (not including G. vaginalis), Corynebacterium or pathobionts than women from the other ethnic groups (11% and 14%, respectively, versus 2–5%). When comparing the VMB group distribution of each ethnic group to the distribution in Dutch women (as shown in Fig 2), the Fisher’s exact p values were 0.41 for South Asian Surinamese women, <0.01 for African Surinamese women, 0.08 for Ghanaian women, <0.01 for Turkish women, and 0.01 for Moroccan women. The VMB of African Surinamese women had the highest richness (median 6.2, IQR 3.6–9.2) compared to a median richness of 4.2–5.2 in the other ethnic groups (overall Kruskal Wallis p = 0.02). The median Shannon Diversity Index [IQR] for Dutch women was 0.9 [0.3–1.5]. The median Shannon Diversity Indexes [IQR], and Kruskal Wallis p values for the other ethnic groups compared to Dutch women, were: South Asian Surinamese women 0.8 [0.2–1.4] (p = 0.31), African Surinamese women 1.5 [0.5–1.9] (p < 0.01), Ghanaian women 0.9 [0.2–1.7] (p = 0.83), Turkish women 1.0 [0.3–1.6] (p = 0.59), and Moroccan women 1.1 [0.4–1.7] (p = 0.16).

thumbnail
Fig 2. Prevalence of VMB groups by ethnic group.

L = Lactobacillus; GV = Gardnerella vaginalis; VMB = vaginal microbiota. *Excluding GV. A. Prevalence of six VMB groups by ethnic group. The error bars show the 95% confidence intervals for the prevalence of the three Lactobacillus-dominated VMB groups combined and the three VMB groups not dominated by lactobacilli combined for each ethnic group. B. The ‘Other VMB’ group was further subdivided into three subgroups: The prevalence of these subgroups is shown by ethnic group.

https://doi.org/10.1371/journal.pone.0181135.g002

VMB correlates in multivariable logistic regression models

VMB correlates were determined for the three largest of the eight VMB groups as described in the methods (Table 3). In adjusted analyses, being of African Surinamese and Ghanaian ethnicity, compared to Dutch ethnicity, was significantly associated with having a polybacterial G. vaginalis-containing VMB (adjusted odds ratio (aOR) 5.1, 95% confidence interval (CI) 2.1–12.0 and aOR 4.8, 95% CI 1.8–12.6, respectively), and being of African Surinamese ethnicity also with having a L. iners-dominated VMB (aOR 2.8, 95% CI 1.2–6.2). Other characteristics independently associated with having a polybacterial G. vaginalis-containing VMB at p<0.05 were being in a steady relationship for less than 2 years (aOR 3.5, 95% CI 1.4–8.5), inconsistent condom use with a casual sex partner in the past six months (aOR 3.2, 95% CI 1.0–9.9), and currently not using hormonal contraception (aOR 0.5, 95% CI 0.3–0.8, for current use).

thumbnail
Table 3. VMB correlates: Multivariable logistic regression models comparing each VMB cluster to the L. crispatus-dominated cluster.

https://doi.org/10.1371/journal.pone.0181135.t003

Discussion

More than a third (38.5%) of the women in our population-based study in Amsterdam had a non-lactobacilli-dominated VMB, of which 83.6% resembled BV and 16.4% had a VMB dominated by Bifidobacteriaceae (not including G. vaginalis), Corynebacterium, or pathobionts. Women of sub-Saharan African descent (either African Surinamese or Ghanaian) were significantly more likely to have a polybacterial G. vaginalis-containing VMB, and African Surinamese women were significantly more likely to have a L. iners-dominated VMB, than Dutch women after controlling for sociodemographic factors, sexual risk behaviors, vaginal cleansing practices, and hormonal contraceptive use.

While studies in multi-ethnic asymptomatic American populations have shown similar differences between African-American women and women of other ethnicities before [1, 5, 1719], our study is the first to show this in a multi-ethnic population living in Europe. We note some differences between our study and the American multi-ethnic studies: VMB dominated by lactobacilli other than L. crispatus and L. iners (most notably L. gasseri and L. jensenii) were less common, and the overall prevalence of a VMB resembling BV in White and Asian women was higher, in our study. Hypotheses to explain the association between ethnicity and VMB composition include differences in sociodemographic, behavioral, environmental, and genetic factors. Our results suggest that genetic differences might play a role. First, our sampling strategy balanced ethnicity and age, and we controlled our analyses for a large number of other sociodemographic factors and behaviors. Second, our study included two groups of women of sub-Saharan African descent: African Surinamese and Ghanaian women. The African Surinamese went from West-Africa to Suriname, a former Dutch colony in South America, in the nineteenth century, and many migrated from Suriname to the Netherlands in the 1970s and 1980s due to an unstable political situation in Suriname. Ghanaians migrated to the Netherlands (and also to other European countries such as the UK and Germany) between 1974 and 1983, and again in the early 1990s, for economic and political reasons. Despite different migration histories, both groups had higher adjusted odds of having a polybacterial G. vaginalis-containing VMB compared to Dutch women of a similar magnitude (aOR 5.1 and 4.8, respectively). Macro-environmental factors are therefore unlikely to play a role, but the role of cultural habits that women and their sexual partners take with them when they migrate, such as diet and sexual, vaginal, and penile hygiene practices, cannot be ruled out (vaginal hygiene practices were, however, not associated with VMB group in our study). A few studies have suggested that genetic differences in vaginal mucosal immune responses might favor colonization by L. iners over L. crispatus in women of African descent [3638] but this has not yet been sufficiently studied. We speculate that vaginal mucosal immune responses may have evolved differently in sub-Saharan Africa compared to Europe or the Americas as a result of differences in exposures at mucosal surfaces to types and quantities of pathogens and pathobionts.

The clustering we performed differentiated VMB groups not dominated by lactobacilli by bacterial diversity, the relative abundances of the BV-associated taxa G. vaginalis, A. vaginae, BVAB1, Mobiluncus species, and Prevotella species, and the relative abundance of Bifidobacteriaceae (not including G. vaginalis), Corynebacterium, and pathobionts. Others have also reported on VMB groups differentiated by bacterial diversity and relative abundance of typical BV-associated bacteria [1, 5, 1719], but different research groups are not yet using consistent VMB descriptions and categorizations making between-study comparisons difficult. VMBs dominated by Bifidobacteriaceae (not including G. vaginalis), Corynebacterium, and pathobionts have been infrequently reported thus far [1, 19, 39]. A Russian study isolated four Bifidobacterium species (B. bifidum, B. breve, B. adolescentis, and B. longum) in vaginal specimens of healthy women of reproductive age, and these were capable of inhibiting pathobionts in vitro [39]. While the clinical significance of the Corynebacterium VMB group is unclear, VMBs dominated by pathobionts are likely associated with adverse reproductive health outcomes; culture-based studies have consistently identified these pathobionts as causes of pregnancy complications and neonatal sepsis [3, 40]. Of note, only 6% of women in our study had at least 1% relative abundance of a Streptococcus species in their VMB, whereas a recent systematic review of culture-based studies concluded that the mean prevalence of rectovaginal carriage of S. agalactiae was 17.9% worldwide and 19.0% in Europe [41]. These figures cannot be compared directly due to differences in sampling (vaginal versus rectovaginal) and laboratory methods (16S sequencing versus culture). However, further research is needed to explain these differences and ensure that pathobionts are accurately identified in molecular studies. In addition, in our study, pathobionts occurred across VMB groups, including lactobacilli-dominated VMB groups. A better understanding of when pathobionts become clinically relevant (for example, at which relative or absolute abundance), and whether VMB groups or community state types adequately capture those situations, is urgently needed.

The five VMB groups not dominated by lactobacilli in our study might require different types of treatment or combinations of treatments. The three VMB groups resembling BV might respond to metronidazole treatment, but the efficacy might differ depending on the relative abundances of key taxa such as G. vaginalis, A. vaginae and BVAB1 and/or the presence of a biofilm (G. vaginalis is thought to initiate a vaginal biofilm by providing the scaffolding needed for other anaerobes to colonize) [42]. In contrast, VMBs with high relative abundance of pathobionts would require a different antibiotic treatment [40]. Clinicians should therefore be enabled to differentiate between these conditions.

Our study was cross-sectional. The vaginal microbiota often fluctuate over time, and the potential confounders that we included in our models (sexual risk behaviors, vaginal cleansing practices, and hormonal contraceptive use) are also time-dependent. Furthermore, vaginal microbiota are often influenced by the menstrual cycle [1, 18], and menstrual cycle data were not collected in the HELIUS study. Transient vaginal dysbiosis may not be clinically relevant, and data on the proportions of women who have persistent vaginal dysbiosis are needed. It should also be noted that the data on the potential confounders were self-reported, and may therefore suffer from social desirability bias.

The HELIUS study selected women at random from among those registered in Amsterdam in the age range and ethnic groups of interest, and we selected our subsample from the HELIUS sample. However, our subsample did suffer from participation bias and possibly reporting bias. The Turkish and Moroccan participants, in particular, were much less likely than the other groups to self-sample a vaginal swab and to complete a sexual behavior questionnaire [26]. They were also less likely to report sexual risk behaviors, which could be due to actual differences in behaviors or in reporting of those behaviors. Our results related to the Turkish and Moroccan groups should therefore be interpreted with caution. However, we believe that these biases did not influence our main finding that women of sub-Saharan African descent were much more likely to have a suboptimal VMB after controlling for potential confounding factors than Dutch women. In the three ethnic groups included in those comparisons (Dutch, African Surinamese and Ghanaian), vaginal sampling rates were 94%, 87% and 71%, respectively. None of the characteristics that differed between those who self-sampled and those who did not were associated with VMB composition in multivariable models, except that Ghanaian women who self-sampled were more likely to be in a steady relationship of longer duration and to have had at least one child than Ghanaian women who declined self-sampling [data not shown]. This suggests that Ghanaian women who participated may have a lower risk of having a VMB not dominated by lactobacilli than those who did not, which only strengthens our main finding.

We conclude that the overall prevalence of VMB not dominated by lactobacilli in women aged 18–34 years in Amsterdam was high, and that women of sub-Saharan African descent were more likely to have a polybacterial G. vaginalis-containing VMB than Dutch women independent of modifiable behaviors. Further research is required to determine if these women are also more likely to have persistent vaginal dysbiosis, and whether this contributes to the higher prevalence of adverse reproductive outcomes in these groups. Our findings support hypotheses implicating the contribution of genetic causes to these differences, such as those related to differences in vaginal mucosal immune response, and these should be further investigated.

Acknowledgments

We thank the study participants, the HELIUS team for study implementation, Jacques Ravel and members of his group for advice on vaginal microbiota DNA extraction and sequencing, Christina Gill and Bing Ma for advice on bioinformatics, Massimo Marzorati and Emma Hernandez-Sanabria for advice on ecological calculations, and the Public Health Laboratory of the Public Health Service of Amsterdam for storing, initial processing, and testing of vaginal swabs for sexually transmitted infections.

References

  1. 1. van de Wijgert JH, Borgdorff H, Verhelst R, Crucitti T, Francis S, Verstraelen H, et al. The vaginal microbiota: what have we learned after a decade of molecular characterization? PLoS One. 2014; 9: e105998. pmid:25148517
  2. 2. Low N, Chersich MF, Schmidlin K, Egger M, Francis SC, van de Wijgert JH, et al. Intravaginal practices, bacterial vaginosis, and HIV infection in women: individual participant data meta-analysis. PLoS Med. 2011; 8: e1000416. pmid:21358808
  3. 3. Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet. 2008; 371: 75–84. pmid:18177778
  4. 4. Fredricks DN, Fiedler TL, Marrazzo JM. Molecular identification of bacteria associated with bacterial vaginosis. N Engl J Med. 2005; 353: 1899–1911. pmid:16267321
  5. 5. Ravel J, Gajer P, Abdo Z, Schneider GM, Koenig SS, McCulle SL, et al. Vaginal microbiome of reproductive-age women. Proc Natl Acad Sci U S A. 2011;.108 Suppl 1: 4680–4687.
  6. 6. Martin DH, Marrazzo JM. The vaginal microbiome: Current understanding and future directions. J Infect Dis. 2016; 214(S1): S36–S41.
  7. 7. Verstraelen H, Verhelst R, Claeys G, De Backer E, Temmerman M, Vaneechoutte M. Longitudinal analysis of the vaginal microflora in pregnancy suggests that L. crispatus promotes the stability of the normal vaginal microflora and that L. gasseri and/or L. iners are more conducive to the occurrence of abnormal vaginal microflora. BMC Microbiol 2009; 9: 116. pmid:19490622
  8. 8. Borgdorff H, Tsivtsivadze E, Verhelst R, Marzorati M, Jurriaans S, Ndayisaba GF, et al. Lactobacillus-dominated cervicovaginal microbiota associated with reduced HIV/STI prevalence and genital HIV viral load in African women. ISME J. 2014; 8: 1781–1793. pmid:24599071
  9. 9. Fethers KA, Fairley CK, Hocking JS, Gurrin LC, Bradshaw CS. Sexual risk factors and bacterial vaginosis: a systematic review and meta-analysis. Clin Infect Dis. 2008; 47: 1426–1435. pmid:18947329
  10. 10. Hellberg D, Nilsson S, Mårdh PA. Bacterial vaginosis and smoking. Int J STD AIDS. 2000; 11: 603–606. pmid:10997505
  11. 11. Esber A, Vicetti Miguel RD, Cherpes TL, Klebanoff MA, Gallo MF, Turner AN. Risk of bacterial vaginosis among women with herpes simplex virus type 2 infection: a systematic review and meta-analysis. J Infect Dis 2015; 212: 8–17. pmid:25589333
  12. 12. Jamieson DJ, Duerr A, Klein RS, Paramsothy P, Brown W, Cu-Uvin S, et al. Longitudinal analysis of bacterial vaginosis: findings from the HIV epidemiology research study. Obstet Gynecol. 2001; 98: 656–663. pmid:11576584
  13. 13. van de Wijgert JH, Verwijs MC, Turner AN, Morrison CS. Hormonal contraception decreases bacterial vaginosis but oral contraception may increase candidiasis: implications for HIV transmission. AIDS 2013; 27: 2141–2153. pmid:23660575
  14. 14. Vodstrcil LA, Hocking JS, Law M, Walker S, Tabrizi SN, Fairley CK, et al. Hormonal contraception is associated with a reduced risk of bacterial vaginosis: a systematic review and meta-analysis. PLoS One. 2013; 8: e73055. pmid:24023807
  15. 15. Cauci S, Driussi S, De Santo D, Penacchioni P, Iannicelli T, Lanzafame P, et al. Prevalence of bacterial vaginosis and vaginal flora changes in peri- and postmenopausal women. J Clin Microbiol. 2002; 40: 2147–2152. pmid:12037079
  16. 16. Kenyon C, Colebunders R, Crucitti T. The global epidemiology of bacterial vaginosis: a systematic review. Am J Obstet Gynecol. 2013; 209: 505–523. pmid:23659989
  17. 17. Zhou X, Hansmann MA, Davis CC, Suzuki H, Brown CJ, Schütte U, et al. The vaginal bacterial communities of Japanese women resemble those of women in other racial groups. FEMS Immunol Med Microbiol. 2010; 58: 169–181. pmid:19912342
  18. 18. Gajer P, Brotman RM, Bai G, Sakamoto J, Schütte UM, Zhong X, et al. Temporal dynamics of the human vaginal microbiota. Sci Transl Med. 2012; 4: 132ra52. pmid:22553250
  19. 19. Fettweis JM, Brooks JP, Serrano MG, Sheth NU, Girerd PH, Edwards DJ, et al. Differences in vaginal microbiome in African American women versus women of European ancestry. Microbiology 2014; 160: 2272–2282. pmid:25073854
  20. 20. Gautam R, Borgdorff H, Jespers V, Francis SC, Verhelst R, Mwaura M, et al. Correlates of the molecular vaginal microbiota composition of African women. BMC Infect Dis. 2015; 15: 86. pmid:25887567
  21. 21. Jespers V, van de Wijgert J, Cools P, Verhelst R, Verstraelen H, Delany-Moretlwe S, et al. The significance of Lactobacillus crispatus and L. vaginalis for vaginal health and the negative effect of recent sex: a cross-sectional descriptive study across groups of African women. BMC Infect Dis. 2015; 15: 115. pmid:25879811
  22. 22. Ness RB, Hillier S, Richter HE, Soper DE, Stamm C, Bass DC, et al. Can known risk factors explain racial differences in the occurrence of bacterial vaginosis? J Natl Med Assoc. 2003; 95: 201–212. pmid:12749680
  23. 23. Kenyon C, Osbak K. Sexual networks, HIV, race and bacterial vaginosis. AIDS. 2015; 29: 641–642. pmid:25710290
  24. 24. Stronks K, Snijder MB, Peters RJ, Prins M, Schene AH, Zwinderman AH. Unravelling the impact of ethnicity on health in Europe: the HELIUS study. BMC Public Health. 2013; 13: 402. pmid:23621920
  25. 25. Stronks K, Kulu-Glasgow I, Agyemang C. The utility of 'country of birth' for the classification of ethnic groups in health research: the Dutch experience. Ethn Health. 2009; 14: 255–269. pmid:19052941
  26. 26. Alberts CJ, Vos RA, Borgdorff H, Vermeulen W, van Bergen J, Bruisten SM, et al. Vaginal high-risk human papillomavirus infection in a cross-sectional study among women of six different ethnicities in Amsterdam, the Netherlands: the HELIUS study. Sex Transm Infect. 2016; 92: 611–618. pmid:27188273
  27. 27. Fadrosh DW, Ma B, Gajer P, Sengamalay N, Ott S, Brotman RM, et al. An improved dual-indexing approach for multiplexed 16S rRNA gene sequencing on the Illumina MiSeq platform. Microbiome. 2014; 2: 6. pmid:24558975
  28. 28. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 2014; 30: 2114–2120. pmid:24695404
  29. 29. Masella AP, Bartram AK, Truszkowski JM, Brown DG, Neufeld JD. PANDAseq: paired-end assembler for illumina sequences. BMC Bioinformatics. 2012; 13: 31. pmid:22333067
  30. 30. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods. 2010; 7: 335–336. pmid:20383131
  31. 31. Schmieder R, Lim YW, Rohwer F, Edwards R. TagCleaner: Identification and removal of tag sequences from genomic and metagenomic datasets. BMC Bioinformatics. 2010; 11: 341. pmid:20573248
  32. 32. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics. 2010; 26: 2460–2461. pmid:20709691
  33. 33. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K, et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol. 2006; 72: 5069–5072. pmid:16820507
  34. 34. Srinivasan S, Hoffman NG, Morgan MT, Matsen FA, Fiedler TL, Hall RW, et al. Bacterial communities in women with bacterial vaginosis: high resolution phylogenetic analyses reveal relationships of microbiota to clinical criteria. PLoS One. 2012; 7: e37818. pmid:22719852
  35. 35. Matsen FA, Kodner RB, Armbrust EV. pplacer: linear time maximum-likelihood and Bayesian phylogenetic placement of sequences onto a fixed reference tree. BMC Bioinformatics. 2010; 11: 538. pmid:21034504
  36. 36. Nguyen DP, Genc M, Vardhana S, Babula O, Onderdonk A, Witkin SS. Ethnic differences of polymorphisms in cytokine and innate immune system genes in pregnant women. Obstet Gynecol. 2004; 104: 293–300. pmid:15292002
  37. 37. Doh K, Barton PT, Korneeva I, Perni SC, Bongiovanni AM, Tuttle SL, et al. Differential vaginal expression of interleukin-1 system cytokines in the presence of Mycoplasma hominis and Ureaplasma urealyticum in pregnant women. Infect Dis Obstet Gynecol. 2004; 12: 79–85. pmid:15739821
  38. 38. Ryckman KK, Williams SM, Krohn MA, Simhan HN. Racial differences in cervical cytokine concentrations between pregnant women with and without bacterial vaginosis. J Reprod Immunol. 2008; 78: 166–171. pmid:18336917
  39. 39. Korshunov VM, Gudieva ZA, Efimov BA, Pikina AP, Smeianov VV, Reid G, et al. The vaginal Bifidobacterium flora in women of reproductive age. Zh Mikrobiol Epidemiol Immunobiol. 1999; 4:74–78.
  40. 40. Garland SM, Chuileanna FN, Satzke C, Robins-Browne R. Mechanisms, organisms and markers of infection in pregnancy. J Reprod Immunol. 2002; 57: 169–183. pmid:12385841
  41. 41. Kwatra G, Cunnington MC, Merrall E, Adrian PV, Ip M, Klugman KP, et al. Prevalence of maternal colonisation with group B streptococcus: a systematic review and meta-analysis. Lancet Infect Dis. 2016; 16: 1076–1084. pmid:27236858
  42. 42. Machado A, Cerca N. Influence of biofilm formation by Gardnerella vaginalis and other anaerobes on bacterial vaginosis. J Infect Dis. 2015; 212: 1856–1861. pmid:26080369