Evaluation of the microbiome in men taking pre-exposure prophylaxis for HIV prevention

Abstract

Tenofovir-based regimens as pre-exposure prophylaxis (PrEP) are highly effective at preventing HIV infection. The most common side-effect is gastrointestinal (GI) distress which may be associated with changes in the microbiome. Dysbiosis of the microbiome can have numerous health-related consequences. To understand the effect of PrEP on dysbiosis, we evaluated 27 individuals; 14 were taking PrEP for an average of 171 weeks. Sequencing of 16S rRNA was performed using self-collected rectal swabs. Mixed beta diversity testing demonstrated significant differences between PrEP and non-PrEP users with Bray-Curtis and unweighted UniFrac analyses (p = 0.05 and 0.049, respectively). At the genus level, there was a significant reduction in Finegoldia, along with a significant increase in Catenibacterium and Prevotella in PrEP users. Prevotella has been associated with inflammatory pathways, insulin resistance and cardiovascular disease, while Catenibacterium has been associated with morbid obesity and metabolic syndrome. Overall, these results suggest that PrEP may be associated with some degree of microbiome dysbiosis, which may contribute to GI symptoms. Long-term impact of these changes is unknown.

Introduction

Annually, two million new HIV infections occur worldwide [1]. Pre-exposure prophylaxis (PrEP), a combined reverse transcriptase inhibitor (RTI) consisting of tenofovir disoproxil fumarate and emtricitabine (TDF/FTC) or tenofovir alafenamide and emtricitabine (TAF/FTC) is highly effective at preventing HIV infection [2]. PrEP has the potential to significantly address the HIV epidemic, particularly when prescribed for those who report behaviors associated with HIV infection. It is important to understand the potential side effects of PrEP. Current formulations of PrEP are known to have significant impacts including declines in kidney function [3, 4] and bone mineral density [5]. Longer-term consequences are largely unknown.

For people living with HIV/AIDS, TDF/FTC and TAF/FTC are often part of an antiretroviral (ARV) treatment regimen. These medications can have a significant impact on the gastrointestinal microbiome in HIV/AIDS patients [6]. In untreated HIV/AIDS patients, there is a depletion of CD4+ T cells in the gastrointestinal (GI) tract, along with increased mucosal permeability, increased inflammation and dysbiosis [9–11]. Current classes of ARVs include RTIs, protease inhibitors (PI), integrase inhibitors (II) and entry inhibitors. Although these medications are highly effective at treating HIV and generally well tolerated, there are side-effects that are class-specific. A study in 2017 explored the impact of combined ARV on the gut microbiome [6]. This study included people living with HIV/AIDS on ARV for a minimum of one year and undetectable viral load for at least six months. Participants on an ARV regimen including nucleoside RTI and IIs demonstrated a normalization of systemic inflammation via reduction of soluble CD14, intercellular adhesion molecule and vascular adhesion molecule plasma levels along with an increase in alpha diversity similar to healthy controls. The study also demonstrated that HIV patients on nucleoside RTI and PIs have a large reduction in multiple genera of gut flora, leading to a loss of alpha-diversity [6]. These studies give insights into some of the changes that can occur when the gut microbiome is exposed to ARV medications. However, few studies to date have been performed to explore the impact of PrEP on the microbiome of HIV-negative people [7, 8], though many HIV-negative individuals at risk of infection take this medication long-term as a preventative measure. Dube et al. demonstrated that PrEP use significantly reduced the Streptococcus genus and significantly increased the Erysipelotrichaceae family in four out of eight patients [7]. Fulcher et al demonstrated a significant increase in Streptococcus, Mitsuokella and Fusobacterium along with reduction of Escherichia/Shigella in the PrEP group [8]. It is interesting to note the differences between these studies, which further emphasizes the need for the evaluation of the microbiome during long-term PrEP use.

In the healthy gut, the dominant bacterial phyla in the gut microbiome include Firmicutes and Bacteroidetes, with smaller abundances of Actinobacteria and Proteobacteria. Interestingly, in men who have sex with men (MSM), there is an increase in the genus Prevotella and a decrease in Bacteroidetes when compared to non-MSM [12, 13]. This is important to recognize because although the Prevotella can commonly be observed in healthy individuals [14], it can be associated with multiple disease states including cardiovascular disease [13], periodontal disease [15], new onset rheumatoid arthritis [16], and insulin resistance in morbidly obese patients [17]. By understanding specific microbiome changes that occur in the MSM population, it can help lay a foundation of what alterations might occur when this population is exposed to PrEP.

To date, major side effects associated with PrEP use include kidney disease [3, 4], reduction in bone mineral density [5], and PrEP start-up syndrome [18]. PrEP start-up syndrome involves mild gastrointestinal symptoms that spontaneously resolve in 1–2 months after initiation of PrEP including nausea, vomiting, abdominal pain and diarrhea [18]. These side effects could be explained by dysbiosis of the gut microbiome occurring after PrEP initiation. Dysbiosis is an imbalance of the resident microbiome associated with negative health consequences which is typically due to a loss of commensal or a gain of pathogenic or potentially pathogenic bacteria (pathobionts) [19]. Dysbiosis may have numerous consequences including increased risk of infection, bacterial translocation, or increased inflammatory states [6, 20]. The goal of this study was to evaluate longer-term changes in the gut microbiome in HIV negative patients on PrEP while carefully controlling for confounders such as behavior, lifestyle, and sexual practices.

Methods

Study and sample collection: Patients presenting to the major PrEP program in Rhode Island were offered enrollment in the study. All participants were MSM and HIV-negative. A total of 16 participants taking PrEP daily for a minimum of one year were enrolled in the study. All control individuals had either never received PrEP or had been off the medication for at least six months, except for one individual who had taken PrEP within 1–5 months prior to enrollment. The control cohort were identified at time of presentation to clinic if they were MSM and HIV-negative then pursued for study enrollment. A total of 16 control patients not currently on PrEP were recruited from subjects who came in for routine sexually transmitted infection screening. Recruitment took place from February 2019 through June 2019. All participants provided written consent in order to enroll in the study. After consent, each patient completed a survey that was administered by research staff. The survey was composed of 17 questions that including demographic, behavioral, and sexual practice data (Supplemental Material 1). Rectal swabs were all self-collected. Patients were educated to place the Copan Fecal Swab two inches past the anal orifice into the rectum then rotate five times in each direction [21]. This approach has previously been reported to obtain sufficient sampling and microbiome analysis has been reproducible and comparable to fecal sample collection [21, 22]. The patients then placed the swabs in the collection tube supplied in the kit. All samples were placed into a −80C freezer within 30 minutes of sample collection. This project was approved by the Institutional Review Board at The Miriam Hospital.

Sequencing and analysis: For microbiome analysis, DNA was extracted from each swab using the Quick-DNA Fecal/Soil Microbe (ZYMO Research). In accordance with the Earth Microbiome Project protocol the hypervariable V4 region of the 16S rRNA genes was targeted during PCR amplification with an 806R reverse and 515F forward primer along with samples-specific ID barcodes [23–26]. The amplification was performed in triplicate using the Phusion High Fidelity Polymerase (New England BioLabs) under the following conditions: 3 min of 98°C, 35 cycles of amplification (45 seconds at 98°C, 60 seconds of 50°C, and 90s of 72°C), and a final elongation for 10 minutes at 72°C. After amplification, equal concentrations were pooled and purified using the Machery-Nagel NucleoSpin Gel and PCR Clean-Up kit (Machery-Nagel). Purified samples were quality controlled and sequenced on the Illumina MiSeq platform by the Rhode Island Genomics and Sequencing Center at the University of Rhode Island (Kingston, RI, United States). Both amplification and kit controls were included for sequencing to assess possible contamination levels.

Samples were sent for blinded analysis in two groups, without indication of which was the PrEP or control group. Sequencing and analysis were performed before the group identities were revealed. Data analysis was carried out using the QIIME2 (v 2019.7) pipeline. The raw paired-end FASTQ files were imported and demultiplexed with the demux plugin. Then sequences were filtered, trimmed, denoised, and merged using the DADA2 plugin [27]. Another plugin, phylogeny, was used to generate a phylogenetic tree, and taxonomy was assigned with a naїve Bayes classifier from the Silva database (version 132) with 99% sequence similarity trained on the 515F/806R region of the 16S rRNA region [28]. Taxonomic assignments were exported for further analysis in R (version 3.5.1), wherein the diversity metrics were calculated with the vegan and phyloseq packages (versions 2.5–3 and 1.26.1, respectively). Significant of differences in beta diversity were calculated with a PERMANOVA using the Adonis function within vegan. To determine whether any taxa were associated with PrEP use, we used the web-based Galaxy module Linear discriminant analysis Effect Size (LEfSe, p-values <0.05) [29]. This approach allowed for sequencing error correction. All figures and linear regressions were generated using Prism (ver. 8.4).

Results

In total, 32 MSM HIV-negative participants were enrolled in the study. Sixteen participants were not currently taking PrEP and 16 participants were active PrEP users. Samples were blinded for sequencing and analysis. Two participants in the PrEP group were excluded because their samples were unable to be extracted and sequenced. Three total participants in the control group were excluded during analysis; two were excluded because active Neisseria infections were identified after sequencing and one was excluded because there was a very large relative abundance of Streptococcus that made additional interpretation difficult. A total of 14 PrEP users and 13 controls were included in the primary analysis. Five of the controls had previously been on PrEP (one between 1–5 months ago, one between >5–12 months ago, two between 12–24 months ago and one between 24–36 months ago). The average age of the PrEP group was 36 years and the average age of the control group was 29 years. There was no difference observed between the PrEP and control group (p=0.0779). The average duration of PrEP use was 171 weeks (3.3 years).

For sequencing analysis, 1,725,679 raw reads were obtained across all samples (for per-sample raw reads, refer to supplemental 2). First, we assessed the diversity within the microbial communities and found that PrEP was not associated with a significant shift in the overall diversity for either Shannon Index or Observed ASVs (Figure 1). Next, we compared the microbial makeup between the control and PrEP groups using three measures of beta diversity. The Bray-Curtis Dissimilarity, which compares taxonomic abundances, as well as the unweighted UniFrac, which reflects the differences in the phylogenetic relatedness of taxa present or absent, both revealed significant dissimilarity between the control and PrEP microbiomes (p-value = 0.05 and 0.049 respectively). Interestingly, when giving weight to the abundance of taxa along with phylogenetic relatedness using a weighted UniFrac, the dissimilarity did not prove significant (p-value 0.686), suggesting that the differences are found in low-abundance taxa.

Figure 1:

Comparison of Shannon alpha diversity between the Control and PrEP group. There is no statistical difference between groups using a Mann-Whitney test.

Next, we conducted LEfSe analysis to determine if PrEP is associated with changes in microbial composition. We did not observe any impacts at the phylum level; however, among lower taxonomic levels, several biomarkers were associated with the use of PrEP. We found significant differences at the family level including a reduction of Pasteurellaceae, Corynebacteriaceae, and Clostridiales Family XI in PrEP users (LDA score > 3.0; Figure 3). At the genus level, we found that PrEP was associated with a significantly lower level of Finegoldia and Corynebacterium 1 along with elevated levels of Catenibacterium and Prevotella 2 (LDA score > 3.0; Figure 4). Despite previous studies finding differences in Streptococcaceae and Erysipelotrichaceae [7, 8], this cohort showed no difference in these taxa with the use of PrEP. The most prevalent rectal genus was Prevotella 9 for both groups with a non-significant increase in relative abundance in the PrEP group.

Figure 3A-B:

Relative abundance of Families in the Control and PrEP groups. (A) Taxonomic bar plots depicting the microbiome of Control and PrEP users. PrEP use was associated with lower levels of Pasteurellaceae, Corynebacteriaceae and Clostridiales Family XI than the Control group. (B) LDA Scores were calculated using LEfSe and indicated families positively and negatively associated with PrEP use (corresponding p-values of more abundant families are indicated on plot A with * indicating p < 0.05).

Figure 4A-B:

Relative abundance of Genera in the Control and PrEP groups. (A) Taxonomic bar plots depicting the microbiome of Control and PrEP users. PrEP use was associated with lower abundance of Corynebacterium 1 and Finegoldia as well as higher levels of Prevotella 2 and Catenibacterium compared to non-PrEP users. (B) LDA Scores were calculated using LEfSe and indicated genera positively and negatively associated with PrEP use (corresponding p-values of more abundant genera are indicated on plot A with * indicating p < 0.05 and ** indicating p < 0.01).

Overall, this data indicates that the PrEP group was associated with significant changes in beta diversity and microbial composition. However, human behavior can also impact microbial composition. We found that there were no differences between the groups when comparing antibiotic exposures, stool softener use, probiotic supplementation, gastrointestinal symptoms, lifestyle practices and several sexual behaviors including rectal douching, anilingus or saliva as lubrication, which are factors that impact the gut microbiome [30–32]. Another factor, the number of male receptive anal sex partners, did have a possible effect on the outcome. Patients in the PrEP group had more partners on average than the control group over the past three months (8.43 vs 2.43 respectively; p-value = 0.015). This factor is positively correlated with the abundance of Prevotella 2 but did not correlate with either Finegoldia or Streptococcus (p-value = 0.0029, 0.119, 0.998; Figure 5A–C).

Figure 5:

The number of male receptive anal sex partners compared to the relative abundance of key microbiota. Linear regression analysis revealed positive correlation between Prevotella 2 abundance and the number of male receptive anal sex partners (p-value = 0.0029). For Finegoldia and Streptococcus, there was no correlation between the microbial abundance and the number of partners (p-value = 0.119 and 0.998 respectively).

Finally, we explored the impact of STIs on the microbiome by examining all patients initially recruited for the study. There was a significant difference in the presence of STIs between the groups (p-value = 0.049). Out of all subjects, a total of six participants tested positive for STIs (four with rectal gonorrhea, one with urine gonorrhea and chlamydia, and one with syphilis), all of which were within the control group. To determine if this is a significant driver of a microbial shift, we compared patients with STIs to those without STIs. Using the Bray-Curtis dissimilarity analysis we found that there was no significant dissimilarity when comparing the participants with STIs to those without STIs (p-value = 0.40; Figure 6). This indicates that while STIs were more prevalent in the control group, it is not a determinant of microbial shifts within this cohort.

Figure 6:

This Bray-Curtis plot exams the beta diversity of participants with STIs (red) to those who do not (black). Six total participants had testing that revealed STIs at the time of analysis: four with rectal gonorrhea, 1 with urine gonorrhea and chlamydia, and 1 with syphilis. There was no significant difference between groups (p-value = 0.40).

Discussion

Few studies have evaluated the impact of PrEP on the gut microbiome in the non-HIV MSM population. Dube et al followed eight patients from prior to initiation of PrEP use through 48–72 weeks of continuous PrEP use and Fulcher et al compared non-PrEP users to PrEP users with median use of 41 weeks. There are noticeable microbiome differences found between the previous two studies and ours. Dube et al study demonstrated significant increases in Erysipelotrichaceae and significant reductions of Streptococcaceae with continuous PrEP use while Fulcher et al demonstrated a statistically significant increase in Streptococcaceae in the PrEP group. Our study did not demonstrate differences in Streptococcaceae and Erysipelotrichaceae with long term PrEP use. Both our study and Fulcher et al were cross sectional studies while Dube et al was a longitudinal study. Additionally, Fulcher age matched the groups while Dube et al did not. The observed phylogenic differences in the previous two studies could represent a more temporary change in the gut microbiome that normalizes after longer term PrEP use.

Our analysis demonstrated no difference in alpha diversity, but it did demonstrate significant difference in beta diversity with Bray Curtis and unweighted UniFrac although there was no significant in the weighted UniFrac likely indicating that the observed differences were arising in the less abundant ASVs. Accordingly, we found no statistically significant differences at the phylum level. At the Family level, significant differences observed included decreases in PrEP users of Pasteurellaceae, Corynebacteriaceae, and Clostridiales Family XI. At the genus level, there was a significantly lower level of Finegoldia along with elevated levels of Catenibacterium and Prevotella 2 with PrEP use. Prevotella 9 was the most abundant genus in both groups with a higher, non-significant abundance in the PrEP group. This was likely due to the high abundance of Prevotella copri as observed previously in MSM [33].

Prevotella 2 has been associated with increased cardiovascular disease risk [13]. The Prevotella genus has also been associated with periodontal disease [15], new onset rheumatoid arthritis [16], and insulin resistance in morbidly obese patients [17]. Catenibacterium has also been associated with insulin resistance in morbidly obese patients [17] along with morbid obesity and metabolic syndrome [34]. These shifts observed with long term PrEP use are important findings because of the possible health implications. Not only could some of these microbiome changes be associated with known side effects of PrEP use including kidney disease and reduction in bone mineral density, but can also inform about possible disease states that might develop with long term PrEP in patients at risk for cardiovascular disease, metabolic syndrome, insulin resistance, and obesity.

Interestingly, Finegoldia is an abundant anaerobe found in the penile microbiome [35, 36] and has been associated with increased seroconversion to HIV when present in the gut microbiome [37]. Our study demonstrated that with long term PrEP use, there is a significant reduction in Finegoldia. It is unclear mechanistically why this occurs, but it could be from class specific mechanisms of PrEP on this pathogenic bacterium. Although the cohort is different than our study’s group of interest, in HIV positive patients, ARV therapy has been shown to both increase [38, 39] and decrease Finegoldia abundance [33]. The conflicting data in the HIV cohort could represent class specific ARV differences on the gut microbiome, although few studies have specified what ARV therapies the patients were on.

Subgroup analysis was performed to examine if there were specific behaviors that distributed differently between non-PrEP and PrEP users. There were no differences between group behaviors when comparing antibiotic exposures, stool softener use, probiotic supplementation, gastrointestinal symptoms, lifestyle practices and a number of sexual behaviors including rectal douching, anilingus or saliva as lubrication. The behavior that was different between PrEP and non-PrEP users was that PrEP users had significantly more receptive anal sex partners. Furthermore, there was a positive association between receptive anal sex partners and Prevotella 2 abundance. This is an interesting finding, but also a possible confounder as Prevotella 2 was significantly increased in the long-term PrEP group. Future studies should aim to further examine the relationship between receptive anal sex and PrEP use to Prevotella 2 abundance.

The main limitation of our study was the small sample size. Ideally, we would have been able to perform analyses between 3 groups: current PrEP users, previous PrEP users, and a PrEP naïve group, but the sample size would be insufficient to make these associations. Self-collected rectal swabs were also a limitation of this study which could be the reason why two of the PrEP participants’ swabs were unable to be extracted and sequenced, but it was also a strength of the study because it helped to optimize recruitment enabling us to get to the sample size we did. Also, self-collected rectal swabs are highly correlative to clinically collected rectal swabs which further supported this collection procedure [21]. Another limitation is that sexual behavior is a driving phenotype and given the study design, we were unable to determine if the host’s behavior influenced microbiome changes or if the microbiome changed the host.

In summary, this study found there are significant changes to the gut microbiome with long term PrEP use. Future research directions include exploring how our observed microbiota changes in the setting of long-term PrEP use might be associated with specific long-term side effects of PrEP such as kidney disease and decreased bone mineral density. Additionally, future investigations can be aimed at exploring how microbiome changes in the setting of long-term PrEP use might be related to development of symptomatic STIs.

Supplementary Material

1694271_Supp_Info1

Supplemental File 1: Survey. This 17-point questionnaire obtained demographic, lifestyle behavioral, and sexual practice information from each participant over the past 3 months leading up to enrollment.

1694271_Supp_Info2

Supplemental File 2: Raw Read Counts. The number of raw reads per sample after 16S rRNA sequencing.

Figure 2A-C:

Comparison of Beta diversity between Control group (blue) and PrEP group (red). Using three different metrics (A) Bray-Curtis dissimilarity, (B) unweighted UniFrac, and (C) weighted UniFrac. PERMANOVAs were used to evaluate the dissimilarity between PrEP and non-PrEP users (* indicates p < 0.05). For both the Bray-Curtis and unweighted UniFrac, the groups were significantly dissimilar, clustering separately from one another (p-values = 0.05, 0.49). However, when considering the abundance of taxa using the weighted UniFrac, the dissimilarities are non-significant, suggesting that the differences associated with PrEP use lie within low-abundance taxa (p-value = 0.686).