Exploring the Anal Microbiome in HIV Positive and High-Risk HIV Negative Women

Abstract

This exploratory study sought to characterize the anal microbiome and explore associations among the anal microbiome, risk factors for anal cancer, and clinical factors. A pilot sample of 50 HIV infected and high-risk HIV negative women were recruited from the former Women's Interagency HIV Study. Microbiome characterization by 16S rRNA gene sequencing and datasets were analyzed using QIIME 2™. Composition of the anal microbiome and its associations with anal cancer risk factors and clinical factors were analyzed using linear decomposition model and permutational multivariate analysis of variance. Composition of the anal microbiome among HIV positive and high-risk negative women was dominated by Bacteroides, Prevotella, and Campylobacter. The overall taxonomic composition and microbial diversity of the anal microbiome did not significantly differ by HIV status. However, the abundance of Ruminococcus 1 belonging to the Rumincoccaceae family was associated with HIV status (q = .05). No anal cancer risk factors were associated with the anal microbiome composition. Clinical factors marginally associated with the anal microbiome composition included body mass index (BMI; p = .05) and hepatitis C virus (HCV; p = .05). Although HIV and risk factors for anal cancer were not associated with the composition of the anal microbiome in this pilot sample, other clinical factors such as BMI and HCV, may be worth further investigation in a larger study. Future research can build on these findings to explore the role of the microbiome and HIV comorbidities in women.

Introduction

Anal cancer in the general population is uncommon but the incidence among HIV positive individuals is alarmingly higher than the uninfected population.¹ Although HIV positive men who have sex with men (MSM) carry the heaviest burden for anal cancer, HIV positive women remain a high-risk group. The incidence of anal cancer among HIV positive women is 30 per 100,000 person-years compared to two per 100,000 person-years among HIV negative women.^2,3

Growing research has examined the role anogenital microbiome plays in mediating disease manifestations among HIV positive individuals.⁴ These manifestations include HIV transmission and progression, chronic immune activation, and metabolic and neoplastic mechanisms^5–11 Evidence has shown the microbiome may influence cancer susceptibility via either the specific metabolites or toxins produced or the inflammatory milieu stimulated by pathogen growth-hallmarks of cancer.^5,6,12 Of specific interest, although in a small sample of HIV positive MSMs, anal squamous intraepithelial neoplasia was found to be associated with a depletion of Bifidobacterium and an enrichment of Peptostreptococcus; particularly Ruminococcus and Pseudomonas were predictive of precancerous high-grade intraepithelial lesions.¹³

Although limited, prior studies examined the microbial composition in individuals living with HIV compared to uninfected controls. Several studies report no difference in the anal microbiome between HIV positive and HIV negative MSMs, where the anal microbiome is mainly composed of Firmicutes and Bacteroidetes, phyla typically dominating the gut microbiome.^14–16 Yet, other studies report a shift in the gut microbiome from Bacteroides predominance to an enrichment in Prevotella in HIV individuals.^7,17 Whereas, one study reports changes in the microbial composition among HIV positive individuals may be attributed to sexual practices and lifestyle versus HIV infection alone.¹⁸

Although research in the role of the microbiome in HIV infection and carcinogenesis is growing rapidly, one notable limitation in this field is the anal microbiome profile has been largely examined in HIV positive men; whereas, the composition of the anal microbiome in HIV positive women, another high-risk group, is still largely understudied.

The primary aim of this study was to characterize the taxonomic composition and diversity of the anal microbiome in a sample of HIV positive women and a control sample of HIV negative women. Additionally, we explored associations among the anal microbiome, clinical factors, and established risk factors for anal dysplasia and anal human papillomavirus (HPV) for HIV positive women and high-risk HIV negative women (e.g., CD4 count, HIV viral load, history of cervical dysplasia, age, smoking status, number of lifetime sexual partners, and history of sexually transmitted infections).^19–22

Methods

Study population

After approvals from the Institutional Review Board of Emory University and the research oversight committee at the recruitment site were obtained, we recruited 50 HIV positive (n = 28) and high-risk HIV negative women (n = 22) from the Atlanta, Georgia site of the formerly known Women's Interagency HIV Study (WIHS). Now the MACS/WIHS Combined Cohort Study (MWCCS), at the time of data collection, WIHS was an ongoing, multi-site, prospective study of the natural history of HIV infection and related disease conditions in HIV positive women and high-risk HIV negative women.²³ High-risk HIV negative women in the WIHS cohort met the following eligibility criteria: (1) injection drug use; (2) having a sexually transmitted disease; (3) unprotected sex with three or more men or protected sex with five or more men; or (4) history of exchange sex for money or goods.^23,24

Women enrolled in the parent WIHS study have an interview, physical examination, and blood work done every 6 months. The interview includes self-reported data on the participant's general medical history, antiretroviral therapy (ART), obstetric and gynecological history, use of drugs, alcohol, and cigarettes, sexual behaviors, health beliefs, and health care use. The physical examination conducted at each 6-month WIHS study visit includes sexually transmitted infections (STI) testing and a cervical Pap test with HPV testing and colposcopic examination and biopsy when indicated. All WIHS participants have blood and urine samples collected to gather complete blood count, CD4 counts, HIV viral loads, and further STI tests.

Study procedures and covariate data

Enrolled WIHS participants, who had an upcoming WIHS study visit, were contacted via telephone by the research team where details of the study were described, what participation entails, and asked whether they would be interested in providing an anal swab during their scheduled WIHS study visit (WIHS visit 44). Interested participants completed a verbal consent form over the telephone and were enrolled into our study. Participants were given a $45 gift card for participation at completion of the study visit. Recruitment, enrollment, and data collection spanned for 6 months in 2016.

All risk factors for anal cancer and clinical factors were obtained from the WIHS database of study assessments completed at study visit (WIHS study visit 44). Risk factors for anal cancer included CD4 count, HIV viral load, history of abnormal cervical cytology, age, smoking status, number of lifetime sexual partners, and history of sexually transmitted infections. At the time of collection, HPV was not available in the WIHS repository, thus abnormal cervical cytology was used as a proxy for HPV status. Additionally, we collected clinical factors that have been shown to be related to the microbiome such as body mass index (BMI), antibiotics use, diabetes, and history of hepatitis C virus (HCV).

Specimen collection

At the scheduled WIHS study visit, participants who consented for the substudy had two anal swabs collected by a research nurse practitioner before the WIHS related cervical Pap test. A Catch-All™ sample collection swab (Epicentre) was inserted 3–4 cm into the anal canal and rotated for about 10 s. After collection, each swab was placed in a MoBio Power bead Tube (MoBio Laboratories, Inc., Carlsbad, CA), repeatedly pressed and rotated for 20 s, placed upright on dry ice, transported to a laboratory, and stored at −80°C until DNA extraction.

Data processing

Bacterial DNA was extracted from anal swabs at the Emory Integrated Genomics Core laboratory (EIGC) using the MoBio PowerSoil^® DNA Isolation Kit (MoBio Laboratories, Inc.). PCR amplification of the 16S rRNA V3–V4 gene regions was performed with previously described PCR primers and methods as illustrated in the 16S Metagenomic Sequencing Library Preparation manual.^25,26 Samples were PCR-amplified in duplicated aliquots to control for variation due to random PCR amplification artifacts and for the detection of potential contamination or barcode binning errors.

All the samples were barcoded with oligonucleotide tags, pooled together in equimolar ratios, and the pooled DNA sample was sequenced on the Illumina MiSeq instrument.²⁵ These samples were completed in one batch. Each sample was exported with two sequencing reads (forward and reverse) on the Illumina MiSeq platform, resulting in 9.7M total sequencing reads for further analysis (138K reads per sample on average).

The QIIME 2™ standard pipeline^27,28 was used to process 16S rRNA V3–V4 sequences to get the feature operational taxonomic unit (OTU) table (a matrix of sequence read counts for each OTU observed in each sample) for further analysis. As all the 16S rRNA sequences from the EIGC were demultiplexed (divided into separate files for each barcoded sample), we directly conducted sequence quality control (QC) using DADA2 software, a method for modeling and correcting sequencing errors at each nucleotide (so that each OTU is an exact sequence variant).²⁹ The QC can trim sequences and filter PhiX reads (i.e., a subset of reads containing PhiX bacteriophage sequences) and chimeric sequences (sequences formed from two or more biological sequences).²⁷

One sample with very large sequencing reads (2,821,491) was excluded for being an outlier, resulting in 49 samples imported into QIIME 2 for QC. Among these 49 anal microbiome samples, the sequencing depth ranged from 62,088 to 301,037, with a mean count of 140,843 per sample. The primers were removed from the sequence reads by trimming the reads at the 17th and 21th base pair for forward and reverse reads, respectively. The sequence tails with pool quality scores <20 (corresponding to more than 1% sequencing error) were removed by truncating the reads at the 260th and 240th base pair for forward and reverse reads, respectively.

After the DADA2 process, 1,249 OTUs were identified. Total read counts per OTU, among all samples, ranged from 2 to 55,062; OTU frequencies/sample ranged from 9,894 to 51,219. To identify taxonomic composition of the samples, we assigned taxonomy to the representative sequences for each OTU using a trained Silva 132 99% classifier.³⁰ According to the Silva nomenclature, when each OTU is named to the lowest taxonomic level that it can be confidently assigned and a strain, clone, or uncultured genus-level group (UCG) identifier is included in the name as appropriate.³¹

Statistical analyses

Based on the rarefaction curve analysis via default median read counts ( = 25,000 in our sample), 16S rRNA samples with ≥3,500 read counts were adequate for microbial diversity and abundance analysis. Therefore, we kept all 49 samples for analysis. The alpha-diversity metrics included Chao-1 index, Shannon's index, and Pielou's evenness. Each alpha-diversity metric was compared between different groups (defined by HIV status or risk factor levels) using the Wilcoxon test. Beta-diversity metrics included the Bray–Curtis and weighted UniFrac distance measures. Principal coordinates analysis (PCoA) was conducted to visualize sampling clustering based on each beta-diversity metric.

Linear decomposition model (LDM)³² and permutational multivariate analysis of variance (PERMANOVA) implemented in adonis2³³ were both used to test associations of specific OTUs or the composite microbial taxonomic profiles respectively with HIV status (positive vs. negative), CD4 count, HIV viral load, cervical cytology, HCV, age, smoking status, number of lifetime sexual partners, and history of sexually transmitted infections, BMI, and diabetes.

While PERMANOVA provides only a global test of any association of the total microbial taxonomic composition (microbial community level) with the factor of interest based on a prespecified distance measure, the LDM provides both the global test and tests of individual OTUs (e.g., differentially abundant OTUs across sample groups), where the latter controls the false discovery rate at nominal level 20%.³² Both PERMANOVA and the LDM allow adjustment of confounding covariates. We also agglomerated the OTU table to the genus level and tested individual genera using the LDM. Our analyses were performed using QIIME 2 and R 3.3.3.

Results

Participant characteristics

The demographic and clinical factors are listed in Table 1 and compared between HIV positive and negative women. Of the 49 women included for analysis, the mean age was 47.45 years (standard deviation = 9.16). HIV positive women were older than their HIV negative counterparts (p = .07). The majority of the sample (91.8%) self-identified as African American/Black. The mean CD4+ count among HIV positive women was 669.35 where 85.2% had an undetectable HIV viral load (<20 copies/ml). HIV negative women reported more female sexual partners than their HIV positive counterparts (p = .00). No other factors were found to be significantly different among the sample.

Table 1.

Demographic and Clinical Characteristics

Variable	Overall, N (%)	HIV positive (n = 27)	HIV negative (n = 22)	p *
Mean age (SD)a	47.45 (9.16)	49.59 (8.88)	44.82 (8.98)	.07
BMI, median (IQR)b	33.1 (9.45)	33.1 (10.8)	32.95 (8.4)	.52
UD HIV viral load (%)		23 (85.2)	N/A
Mean CD4+ count (SD)		669.35 (328.3)	N/A
Mean nadir CD4+ count (SD)		232.0 (206.9)	N/A
Race/ethnicity
Caucasian/White	2 (4.1)	0 (0)	2 (4.1)	.27
African American/Black	45 (91.8)	26 (53.1)	19 (38.8)
Other	2 (4.1)	1 (2)	1 (2)
Smoking status
Current smoker	23 (46.9)	15 (30.6)	8 (14.3)	.41
Former smoker	13 (26.5)	6 (12.2)	7 (14.3)
Never smoker	13 (26.5)	6 (12.2)	7 (14.3)
Abnormal cervical Pap test
Yes	4 (8.2)	3 (6.1)	1 (2)	.62
No	45 (91.8)	24 (49)	21 (42.9)
Recent gonorrhea since last visit
Yes	0 (0)	0 (0)	0 (0)	—
No	49 (100)	27 (49)	22 (51)
Recent chlamydia since last visit
Yes	0 (0)	0 (0)	0 (0)	—
No	49 (100)	27 (49)	22 (51)
Recent genital warts since last visit
Yes	0 (0)	0 (0)	0 (0)	—
No	49 (100)	27 (49)	22 (51)
Hepatitis C status
Negative	40 (81.6)	22 (44.9)	18 (36.7)	.64
Positive	9 (18.4)	5 (10.2)	4 (8.2)
Antibiotic use
Yes (%)	5 (10.2)	2 (4.1)	3 (6.1)	.81
Diabetes
Yes	10 (20.4)	5 (10.2)	5 (10.2)	.49
No	39 (79.6)	22 (44.9)	17 (34.7)
Lifetime partners, median (IQR)b
Male	10 (25)	10 (20)	18 (62)	.38
Female	0 (.75)	0 (0)	0 (2)	<.001

^* Chi square test, unless otherwise noted.

^a Two-sample Kolmogorov–Smirnov test.

^b Mann–Whitney U test.

BMI, body mass index; IQR, interquartile range; N/A, not applicable; SD, standard deviation; UD, undetected.

Taxonomic analysis

In this study, 1,249 OTUs were identified, which were classified among a total of 13 phyla and 195 bacterial genera. The dominant phyla in our sample were Bacteroidetes, Firmicutes, Proteobacteria, and Fusobacteria and the dominant genera included Bacteroides, Prevotella, and Campylobacter (Fig. 1).

FIG. 1.

Abundance of taxa and taxonomic composition of microbial communities by Genus level and HIV status. The dominant genera included Bacteroides, Prevotella, and Campylobacter. NEG, HIV negative; POS, HIV positive.

Diversity analysis

Alpha diversity estimates were not associated with HIV status (Fig. 2). As observed in the two-dimensional PCoA plots based on the Bray–Curtis and weighted UniFrac distance measures, there were no detected separations of the anal microbiome profiles with respect to HIV status (positive vs. negative; Fig. 3). The global test with adonis2 confirmed that there was no significant difference detected between groups using either distance metric (i.e., Bray–Curtis [p = .37]; weighted UniFrac [p = .26]).

FIG. 2.

Alpha diversity by HIV status. Alpha diversity estimates, Chao-1 index (p = .062), Shannon's index (.385), and Pielou's evenness (p = .713) were not associated with HIV status.

FIG. 3.

Ordination plots and global test p value with adonis. Based on the Bray–Curtis and weighted and UniFrac distance measures (p = .371 and p = .263, respectively), there were no detected separation of the anal microbiome profiles by HIV status.

Abundance analysis

The global test with the LDM showed a nonsignificant difference (p = .11) in the presence/absence of OTUs between HIV positive and HIV negative women and in the relative abundance (p = .52) of OTUs. No individual OTUs in either the presence/absence scale or the relative abundance scale showed any significant difference (q > .5) between the two groups. After aggregating the read counts to the genus level, we identified a total of 167 genera (we excluded 28 genera that could not be confidently assigned a genus level). Using the LDM, Ruminococcus 1 (q = .05) was found to be significantly associated with HIV status where HIV positive women showed significantly decreased abundances in Ruminococcus 1 compared to HIV negative women (Fig. 4). Of some interest, Ruminococcaceae UCG-002 and Ruminococcaceae UCG-014 were also detected at q value <.25 using the LDM.

FIG. 4.

Differential abundance detected by HIV status. Using a linear decomposition model, Ruminococcus 1 (p = .0004, q = .052) was associated with HIV status where HIV positive women showed significantly decreased abundances. UCG, uncultured genus-level group.

Diversity and abundance analysis by risk and clinical factors

Age was not found to be associated with alpha diversity estimates; however, age may be associated with the overall microbial community composition of the anal microbiome (beta-diversity) (PERMANOVA [Bray–Curtis] p = .04 and the LDM p = .16). BMI and HCV infection, respectively, showed marginal associations with overall microbial community composition (PERMANOVA [Bray–Curtis] p = .05 and .05, respectively; and the LDM p = .14 and .15, respectively). Smoking status, number of sexual partners, viral load, and nadir CD4+ were not associated with the anal microbiome based on PERMANOVA and the LDM after adjusting for age (beta diversity analysis; Table 2).

Table 2.

Association Tests Among Beta-Diversity Metrics and Risk/Clinical Factors (Global p Values)

	adonis2: Bray–Curtis	LDM-omnibus
Age	0.038	0.163
BMI	0.054	0.144
Smoking status	0.611	0.9
HIV viral load	0.311	0.54
Diabetes	0.742	0.693
Lifetime sexual partners	0.722	0.699
HCV	0.052	0.146
CD4+	0.127	0.37
CD4+ NADIR	0.918	0.91

BMI, body mass index; HCV, hepatitis C virus; LDM, linear decomposition model; NADIR, patient's lowest reported CD4+ count.

Discussion

We examined the taxonomic profiles (beta-diversity and individual OTUs) and microbial diversity (alpha-diversity) of the anal microbiome in HIV positive women and a comparative group of high-risk HIV negative women. In previous literature, the anal microbiome has been characterized in HIV positive and negative men and MSMs but there is a paucity of research that characterizes the anal microbiome in HIV positive women. We found the anal microbiomes of HIV positive and negative women were dominated by genera bacteria Bacteroides, Prevotella, and Campylobacter similar to previous findings in HIV positive and negative men.^14,15,34 The anal microbiome did not differ in taxonomic composition, or alpha diversity by HIV status; again, similar to prior findings in HIV positive and negative MSMs.^14,15

When tested for differences in abundances at the genus level, Rumincoccus 1, a genera in the Ruminococcaceae family was significantly lower in HIV positive women. Species of the Ruminococcaceae family are commonly found in significant numbers in the human gut microbiota.¹⁶ Small but detectable differences have been previously reported for the gut microbiota between HIV positive, HIV negative, and ART-treated populations; yet, the exact gut microbiome configuration between these populations are inconsistent among studies.^{11,14,15,34,35}

Wang et al. found, among a sample of HIV positive and negative women on stable ART, Ruminococcus (anaerobic and Gram-positive gut microbes also in the Ruminococcaceae family) was enriched among their sample of HIV positive women.³⁵ Yet, among HIV positive male counterparts, McHardy et al. found the rectal microbiome in HIV positive men who were not receiving combination antiretroviral therapy (cART) to be significantly different in their HIV negative counterparts.³⁴

Specifically, Ruminococcous was found to be depleted in HIV positive subject not receiving cART. It is suggested that HIV infection or the HIV treatments may modulate the intestinal or gut microbiome via the underlying inflammatory state driven by HIV infection.^7,34 Notably, Wang et al.³⁵ and McHardy et al.³⁴ collection methods for the anal microbiome were different—self-collected fecal samples versus clinician applied enema, respectively—which is also different from our study. Thus, it cannot be ruled out conflicting findings may be due to the varying sampling methods of the anal and fecal microbiota, which have distinct microbiota signatures.^36,37

We did not find risk factors related to anal cancer in HIV positive women to be significantly associated with the anal microbiome. Although we had a limited sample size, BMI and HCV, were found to be marginally associated with the overall microbial community composition (beta-diversity metrics). Prior research supports differences in the microbiome by BMI³⁸; additionally, the use of ART has been associated with altered body fat distribution.³⁹ Similarly, Wang et al. found a positive association between Collinsella and BMI among their sample of HIV positive and negative women.³⁵ The relationship between the anal microbiome and BMI observed in our study is similar to emerging evidence of the relationship between the gut microbiome and obesity.^38,40 The common underpinnings to obesity is insulin resistance and inflammation.⁴¹

Specifically, a link between the gut microbiome and a proinflammatory state was first demonstrated in mice models in Cani et al.'s study.⁴² Endotoxemia was detected in mice models that were fed a high-fat diet, which was also associated with decreased abundance of Gram-negative Bacteroides and Gram-positive Clostridia. It is postulated a dysregulated microbiome increases gut permeability-begetting increased metabolic endotoxemia and inflammation. Since the anal microbiome has anatomical interactions with the gut microbiome³⁶ further research is needed to examine whether and how the anal microbiome mediates or moderates relationships among ART, HIV infection, and BMI.

HIV and HCV coinfections are common, but research that examines the role of the anal microbiome and liver diseases is limited. However, the “gut-liver axis” has been investigated and its role in the gut microbiota composition in liver diseases.⁴³ Common gut composition in patients with liver disease includes Enterobacteriaceae, Clostridiales XIV, Lachnospiraceae, and Ruminococcaceae.⁴⁴ Gut microbiota is significantly impacted by liver diseases; moreover, HIV impacts the gut microbiome.^4,43 This “double-hit” starts a cascade that has been investigated in its role in carcinogenesis. An altered gut microbiome has been linked among HCV infection, immune response, and obesity in liver carcinogenesis; additionally, it is suspected the main action of the gut microbiome is the promotion of liver cancer rather than carcinogenisis.^44–47

It is important to note, the microbiome is influenced by the anatomic site of collection (i.e., anal vs. rectal vs. gut) including collection methods (i.e., fecal vs. swab).³⁷ Thus, it is with caution to extrapolate relationships with the gut microbiome and disease to the anal microbiome and disease. Yet, given the proximity and interaction of the gut and anal microbiome, the impact of the anal microbiome and its link to carcinogenesis is an emerging and interesting topic for future research.

We were unable to evaluate our a priori objectives due to limited access to covariates to sufficiently evaluate (i.e., HPV status and history of anal intercourse) and only 8% of our sample reported history of abnormal cervical dysplasia. Additional limitations include the small pilot sample size where preliminary associations can only be captured. However, this pilot study provided information regarding feasibility and preliminary data for larger studies. Additionally, microbiome characterization via 16S rRNA sequences is limited in resolution for identifying bacterial species at the species level, even when sequence variants are used to define OTUs.

We did not collect other clinical factors that may have influenced the anal microbiome such as dietary intake, history of anal intercourse, and menopausal status particularly since HIV positive women in our sample were older than HIV negative women. Moreover, we could not explore gender differences in our sample; however, our study contributes to the literature since the anal microbiome has been predominantly explored in HIV positive men.

Our study did not collect history of nonalcoholic fatty liver disease (NAFLD). This liver disease is a cardiometabolic syndrome characterized by obesity, diabetes, and dyslipidemia, which can progress to nonalcoholic steatohepatitis (NASH). NASH is an inflammatory state that can lead to progressive liver damage to liver cancer. Recent studies have shown the gut microbiome influences the proinflammatory state, thus impacting NAFLD and its progression to NASH.⁴⁷ This is notable because the mean BMI of our sample was 33.5 kg/m², reaching the criteria for obesity, which may have impacted the results of the study.

Conclusion

Although exploratory, our findings are suggestive that the composition, abundance, and diversity of the anal microbiome composition in HIV positive women and a comparative group of high-risk negative women do not differ. A genera of the Ruminococcaceae family, commonly found in the gut, was depleted among HIV positive women, whereas another study reported an enrichment of another Ruminococcus genus in HIV positive women.³⁵ Although risk factors for anal cancer were not associated with the anal microbiome, we did reveal clinical factors such as BMI and HCV may be associated with the anal microbiome.

Our study warrants research on a larger scale to further examine associations among the anal microbiome and risk factor for anal cancer among high-risk groups, such as HIV infected women. This field of research is promising in developing more personalized therapeutic tools that identifies and targets specific bacterial strains to identify and modify the disease process in at risk populations.

Authors' Contributions

J.W., D.W.B., and T.G. designed the study. J.W. and A.F. were responsible for data collection and management. J.B., D.T., C.I.J., Y.L., and Y.H. conceived and performed data cleaning and data analysis with input from J.W. and J.B. drafted the article with critical review and editing from D.W.B., T.G., D.T., C.I.J., A.F., Y.L., and Y.H.

Author Disclosure Statement

No conflicts of interest declared for all authors.

Funding Information

Supported by Winship Cancer Institute No. IRG-14-188-01 from the American Cancer Society.