Comparison of Lower Genital Tract Microbiota in HIV-Infected and Uninfected Women from Rwanda and the US

Introduction Previous studies have shown that alterations of the bacterial microbiota in the lower female genital tract influence susceptibility to HIV infection and shedding. We assessed geographic differences in types of genital microbiota between HIV-infected and uninfected women from Rwanda and the United States. Methods Genera of lower genital tract bacterial microbiota were identified by high-throughput pyrosequencing of the 16S rRNA gene from 46 US women (36 HIV-infected, 10 HIV-uninfected) and 40 Rwandan women (18 HIV-infected, 22 HIV-uninfected) with similar proportions of low (0–3) Nugent scores. Species of Lactobacillus were identified by assembling sequences along with reference sequences into phylogenetic trees. Prevalence of genera and Lactobacillus species were compared using Fisher's exact tests. Results Overall the seven most prevalent genera were Lactobacillus (74%), Prevotella (56%), Gardnerella (55%), Atopobium (42%), Sneathia (37%), Megasphaera (30%), and Parvimonas (26%), observed at similar prevalences comparing Rwandan to US women, except for Megasphaera (20% vs. 39%, p = 0.06). Additionally, Rwandan women had higher frequencies of Mycoplasma (23% vs. 7%, p = 0.06) and Eggerthella (13% vs. 0%, p = 0.02), and lower frequencies of Lachnobacterium (8% vs. 35%, p<0.01) and Allisonella (5% vs. 30%, p<0.01), compared with US women. The prevalence of Mycoplasma was highest (p<0.05) in HIV-infected Rwandan women (39%), compared to HIV-infected US women (6%), HIV-uninfected Rwandan (9%) and US (10%) women. The most prevalent lactobacillus species in both Rwandan and US women was L. iners (58% vs. 76%, p = 0.11), followed by L. crispatus (28% vs. 30%, p = 0.82), L. jensenii (20% vs. 24%, p = 0.80), L. gasseri (20% vs. 11%, p = 0.37) and L. vaginalis (20% vs. 7%, p = 0.10). Discussion We found similar prevalence of most major bacterial genera and Lactobacillus species in Rwandan and US women. Further work will be needed to establish whether observed differences differentially impact lower genital tract health or susceptibility to genital infections.


Introduction
Many reproductive-aged women have lower genital tract microbiota that can be classified as either "healthy" microbiota or bacterial vaginosis (BV) [1][2][3][4][5][6]. However, there is considerable heterogeneity within these classifications. For example, while healthy microbiota generally are comprised mostly of Lactobacillus spp., in different women L. crispatus, L. iners, L. jensenii or L. gasseri can be the dominant species [7][8][9]and there are significant differences in the beneficial effects of each [10][11][12][13][14][15][16]. BV can also vary dramatically between women. For example, Fredricks et al. showed, by sequencing 16S ribosomal RNA genes, that while Gardnerella vaginalis was a dominant bacterium in many cases of BV (comprising as much as 49% of the bacterial sequences), in some cases G. vaginalis was almost absent (1-2% of sequences) [8]. Similarly, while most women with BV had microbiota with substantial proportions of Prevotella sequences (25-30% of sequences), some had no detectable Prevotella sequences. A substantial heterogeneity in BV has been confirmed in other recent studies using molecular-based methods [7,17].
Cross-sectional and longitudinal epidemiologic studies show that BV is associated with increased susceptibility to infection with HIV [18,19] and other sexually transmitted diseases (STD) [20,21]. BV is also associated with increased genital shedding of HIV in women [22][23][24][25]. While BV is associated with these infections, an unresolved question is whether heterogeneity within BV is associated with HIV.
The 2010 UNAIDS Global Report estimated that in 2009 the prevalence of HIV in Rwandans aged 15-49 was 2.9% (2.5%-3.3%), five times higher than the estimated rate of 0.6% (0.4%-0.8%) in the United States (US) [26]. The rate of 1.9% (1.3%-2.3%) among young Rwandan women (15)(16)(17)(18)(19)(20)(21)(22)(23)(24) was 10 times higher than the rate of 0.2% (0.1%-0.3%) among young US women [26]. Differences in types of genital microbiota are potentially among a number of factors that contribute to these large disparities in HIV prevalence between Rwanda and the US. The purpose of the current pilot study was to investigate the heterogeneity of genital bacterial microbiota between HIVinfected and -uninfected Rwandan and US women. Organisms found to be more prevalent in one country, particularly among HIV-infected women compared to HIV-uninfected women, may contribute to HIV susceptibility or shedding. Associations between these organisms and HIV viral load in the female genital tract could be targeted in larger future studies. Although some genital bacteria are easily cultured (e.g. Gardnerella), it has become clear recently that many genital bacteria are either unculturable (e.g. Sneathia) or are difficult to culture since they are anaerobic (e.g. Mobiluncus and Atopobium). Therefore, we used high-throughput pyrosequencing of the 16S rRNA gene to identify the genital microbiota of women in this study. Previous 16S-based highthroughput sequencing studies of microbiota have been performed on US [7,27,28], Chinese [29], and Tanzanian women [17]. We directly compared the genital microbiota between US and Rwandan women, overall and stratified by HIV serostatus.

Study Population
The study population of 86 women came from two parent studies, the Women's Interagency HIV Study (WIHS) and the Rwandan Women's Interassociation Study and Assessment (RWISA). Both the WIHS and the RWISA are observational cohort studies of HIV-infected and uninfected women. Semiannual study visits included extensive interviews, specimen collection, and clinical examinations. The WIHS enrolled a total of 3766 women (2791 HIV-infected, 975 HIV-uninfected) in two waves, from 10/ 1994 to 11/1995 [30] and 10/2001 to 9/2002 [31], for the purpose of investigating the progression of HIV infection and associated diseases, such as diabetes and cardiovascular disease, among HIVinfected and demographically similar HIV-uninfected women in the US. The RWISA enrolled 936 women (710 HIV-infected, 226 HIV-uninfected) in 2005 for the purpose of investigating responses to antiretroviral therapy in Rwandan women, including potential impact of their experiences during the 1994 Rwandan genocide [32]. The 46 WIHS participants (36 HIV-infected, 10 HIVuninfected) included in the current study have been described in two previously conducted substudies [9,28]. The 40 RWISA participants were selected for the current study to have Nugent scores similar to the WIHS participants. All data on sexual behavior and STD history were missing for one Rwandan participant.

Ethics Statement
The study was conducted in accordance with protocols approved by the Rwandan National Ethics Committee and the Institutional Review Board (IRB) of Montefiore Medical Center, Bronx, NY, USA for RWISA participants and guidelines of the US Department of Health and Human Services and the IRBs of each of the WIHS institutions and the WIHS Executive Committee for WIHS participants. Written informed consent was obtained for all participants from both cohorts.

Laboratory methods
Genital tract samples for both studies were collected by cervicovaginal lavage (CVL) performed by irrigation of the cervix with 10 mL of nonbacteriostatic sterile saline, followed by aspiration from the posterior fornix. All CVL samples were gently vortexed to  Statistical methods Figure 1 shows the study selection and analytic design. Rwandan and US women were selected and frequency matched on the basis of Nugent score. Women were subsequently stratified by Lactobacillus-predominance versus non-Lactobacillus predominance.
There were two analytic aims, to investigate the heterogeneity of microbiota associated with BV (Aim 1) and to investigate the heterogeneity of healthy microbiota (Aim 2). For each aim we conducted two separate comparisons. First we compared all Rwandan and US women. Secondly, we compared by country and HIV status (four groups): HIV-infected Rwandan (RW+), HIV-uninfected Rwandan (RW-), HIV-infected US (US+), and HIV-uninfected US (US-) women. We compared prevalence, defined as the proportion of women in whom a given genus or species was present. A genus or species was considered present in a woman if at least 1% of all her sequences corresponded to it.
Fisher's exact tests were used to determine differences in the prevalence of individual genera and Lactobacillus spp. and to compare categorical demographic and behavioral characteristics. Mann-Whitney U tests were used to compare continuous demographic and behavioral characteristics. P values less than 0.05 were considered to be statistically significant and P values  Table 1 shows a comparison of demographic and behavioral characteristics between Rwandan and US women, stratified by Lactobacillus level. In general, US women had more sexual experience and higher proportions of STDs than Rwandan women. Differences between countries tended to be larger and more statistically significant among women with low Lactobacillus, especially for genital warts, syphilis, and PID, than among women with high Lactobacillus. The higher prevalence of HIV infection among US women versus Rwandan women overall (78% vs. 45%, p = 0.002) was reflected in each stratum and, again, was more pronounced among women with low Lactobacillus. None of the 18 HIV-infected Rwandan women had used antiretroviral medication (ART) compared to 19 (53%) of the 36 HIV-infected US women who were naïve to ART. Specimens from Rwandan women were collected before wide ART availability in Rwanda and between January 1995 and April 1997 for US women, when ART use was not yet optimal. CD4 + cell count was similar between RW+ and US+ women (218 vs. 254, p = 0.54). Log 10 HIV RNA viral load was also similar (4.77 vs. 4.61, p = 0.76).

Types of genital microbiota are similar between Rwandan and US women, overall
To compare the types of genital microbiota found in Rwandan and US women, genital tract samples obtained from each of the 86 women were used to identify bacterial genera by pyrosequencing of the 16S rRNA gene. Of the 86 prevalent genera identified across all 86 women (Table S1), 17 were unique to Rwandan women, 33 were unique to US women, and 36 were common to both groups. These 86 genera accounted for a median of 97% of all sequences (range 90-100%) among Rwandan women and 96% (90-100%) among US women (p = 0.12). Among Rwandan women, the median number of genera was 4.5 (range 1-15) compared to a median of 7.5 (range 1-23) for US women (p = 0.33). Figures 2 and 3 show the distributions of the most dominant genera in Rwandan and US women, respectively. Eleven genera comprised at least 10% of sequences in at least one Rwandan and at least one US woman: Lactobacillus, Prevotella, Gardnerella, Atopobium, Sneathia, Parvimonas, Megasphaera, Chryseobacterium, Gemella, Lachnobacterium, and Syntrophococcus (Figures 2 and 3). Proportions of these genera did not differ between Rwandan and US women, except for Lachnobacterium and Megasphaera (data not shown).
Prevalence of some microbiota associated with BV higher among US women Table 2 shows the 28 most prevalent genera (of 83) observed among the 48 women with low Lactobacillus. The table is grouped by high ($40%), moderate (16-39%), and low (10-15%) overall prevalence, and further stratifies by geography and HIV status. Rwandan women had lower prevalence of 10 genera (Megasphaera, Lachnobacterium, Allisonella, Mobiluncus, Propionibacterium, Citrobacter, Achromobacter, Escherichia/Shigella, Enterobacter, and Phocaeicola) and higher prevalence of Mycoplasma and Eggerthella, compared to US women. Prevalences for the other 16 genera were not significantly different between the two countries. Results were similar for the four group comparison (RW+, RW2, US+, US2). Of genera observed in both countries, Megasphaera, Lachnobacterium and Allisonella were higher in US+ and US2; Asaccharobacter was higher in RW+ and US+; Mycoplasma was highest in RW+; Mobiluncus and Dialister were highest in US2; and Propionibacterium was highest in US+.

Prevalence of L. crispatus is low among HIV-infected women
Among the 38 women with high Lactobacillus, only 17 genera in addition to Lactobacillus were observed. Of those, only three were observed at prevalences higher than 10% (Gardnerella 21%, Prevotella 16%, and Ureaplasma 11%). Therefore, we restricted analysis to Lactobacillus spp. among these women (Table 3, top), contrasting them to the 26 women with 1-49% of sequences corresponding to Lactobacillus (Table 3, bottom). Both halves of Table 3 show that overall L. iners was the dominant Lactobacillus spp. The diversity of Lactobacillus spp. was similar among the 18 Rwandan women and 20 US women with high Lactobacillus and the 10 Rwandan women with low Lactobacillus. However, there was almost no diversity in Lactobacillus spp. among the 16 US women with low Lactobacillus. Taking HIV into account revealed significantly lower prevalence of L. crispatus among all HIVinfected women with low Lactobacillus (p = 0.03). Among women with high Lactobacillus, the prevalence of L. crispatus was lower among Rwandan HIV-infected women but not US HIV-infected women.

Discussion
This exploratory pilot study was the first direct comparison of the types of lower genital tract microbiota of US and African women using pyrosequencing of the 16S rRNA gene. The major finding was an overall similarity in the types and prevalences of genital microbiota between US and Rwandan women, despite differences in sexual behaviors and STD history between the groups. This finding is congruous with previous studies. A large epidemiological study of West African women with BV also showed their genital microbiota was similar to that of women in developed countries [35]. Another study comparing phenotypic characteristics of microbiota, including production of acid and H 2 O 2 , from East African commercial sex workers and Canadian adolescents also noted similarity, reinforcing what previous work had suggested regarding the most frequently observed vaginal microorganisms across all groups of women studied around the world [36]. This suggests the possibility that the heterogeneity of the most common microbiota associated with BV is not largely impacted by sexual history or geography.
Regarding genera with dissimilar prevalences among US and Rwandan women, the higher frequency of Mycoplasma among HIV-infected Rwandan women could have important implications for HIV transmission. For example, a prospective study of women in the East African countries of Zimbabwe and Uganda found that the risk of HIV acquisition was significantly higher in those with M. genitalium [19]. Likewise in Kenya, women with high M. genitalium organism burdens were more likely to have measurable cervical HIV-1 DNA than were M. genitalium-negative women [37].
The coliform bacilli Escherichia and closely related Shigella, Enterobacter, and Citrobacter are members of the normal intestinal flora of humans [38,39]. A limited number of species, including E. coli, Enterobacter aerogenes, Enterobacter cloacae, S. dysenteriae, and S. flexneri are responsible for most of the infections which cause major health problems around the world. In the vagina, these bacteria are the primary cause of urinary tract infections, particularly E. coli [40,41]. Interestingly, these genera were prevalent only in US HIV-infected women in our study. This finding is intriguing, especially given the differences in sexual histories and STD prevalence between US and Rwandan women. Vaginal colonization with coliforms could be a marker of recent sexual activity. Future studies could investigate the association between these coliforms and markers of recent sexual activity such as PSA, RSID, and Y chromosome PCR.
No studies have shown any human disease to be associated with Lachnobacterium, Allisonella, Asaccharobacter, or Phocaeicola. Propionibacterium spp. can cause skin infection [42]. Achromobacter spp. have been associated with cystic fibrosis and immunosuppression [43]. However, the clinical significance, if any, in the human female genital tract for these bacteria is unknown. Studies of the gut microbiome show that individuals can have highly divergent types of microbiota, although the functional profiles are similar [44]. It is possible that in the genital tract, some of the differences we observed in types of microbiota also have similar functions.
Colonization with lactobacilli, particularly the H 2 O 2 -producing species L. crispatus and L. jensenii, has a protective effect against the acquisition of BV [10][11][12][13][14][15][16]. Macklaim et al. showed that L. iners has the ability to survive in the vagina under varying conditions [13]. L. iners was the dominant species among all women in our study, followed by L. crispatus and L. jensenii. The lower prevalence of L. crispatus that we observed among all HIV-infected women from Rwanda, and among some HIV-infected women from the US, could be clinically important if validated in a larger study.
It should be noted that this study was exploratory and limited by small sample size. However, although our results may not be generalizable to all Rwandan and US women, our samples were obtained from two established parent studies (WIHS and RWISA), including both HIV-infected and uninfected women, both with sound protocol and infrastructure [30][31][32]. Except for Lactobacillus spp., our analysis was conducted at the level of genera. It is therefore possible that while we observed similarities between countries in the most common genera, there could be important differences between countries at the species level. We have identified several genital bacterial microbiota (Mycoplasma, Escherichia/Shigella, Enterobacter, Citrobacter) which might be targeted in future studies using genus or species-specific PCR to investigate the association between bacterial load and HIV shedding and markers of recent sexual activity. Despite these limitations, this study revealed surprising similarities in the microbiota of US and Rwandan women, as well as some interesting differences that warrant further exploration. Additionally, it would be informative to investigate the clinical significance of these organisms.