Prevotella bivia Influences Antiretroviral Pharmacokinetics and Viral Replication in an Ex Vivo Vaginal Tissue Model

ABSTRACT

The vaginal microbiome plays a complex role in tenofovir's mucosal pharmacology. While the relative contribution of systemic versus local concentrations to PrEP PK/PD is an ongoing debate, the development of effective HIV prevention requires an understanding of events at the site of transmission. Our objective was to further understand the relationship between tenofovir mucosal pharmacology and non‐optimal vaginal microbiota using the ex vivo vaginal tissue model. Vaginal and cervical explants were produced from cadaver tissue using a biopsy punch. Explants were incubated with Prevotella bivia (10³–10⁵ colony forming units/ml) with and without tenofovir for 24 h. TFVdp and endogenous adenosine triphosphate (dATP) were quantified using liquid chromatography tandem mass spectroscopy. Explants were then challenged with 10⁶ TCID₅₀ the viral concentration where 50% of tissue cultures are infected, of HIV_JR‐CSF for 3 h. Explants were cultured on gel‐foam rafts for 48 h then collected for HIV RNA quantification using RT‐qPCR. TFVdp formation in vaginal tissue was approximately 76% lower in anaerobic conditions (p = 0.2) compared to aerobic. While dATP concentrations did not significantly differ between any Prevotella bivia concentrations and the Prevotella‐free control, TFVdp and TFVdp:dATP ratio in vaginal tissue decreased as Prevotella bivia concentrations increased, although not statistically. Unexpectedly, tenofovir efficacy increased as Prevotella bivia concentrations increased. The ex vivo tissue model was successful in demonstrating the pharmacology of TFVdp is affected by Prevotella bivia . Viral replication was also affected by Prevotella bivia ; therefore, further work is needed to fully understand effects on tenofovir pharmacodynamics.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    The effect of the microbiome on the variability in drug response is understudied. In clinical trials, dysbiotic microbiomes decreased topical tenofovir's efficacy to protect against HIV transmission. The route of administration seems to play a role as oral tenofovir is not affected by the microbiome. However, the importance of local antiretroviral concentrations is an ongoing debate.
• What question did this study address?

  • ○
    What is the effect of non‐optimal microbiomes, using Prevotella bivia , on tenofovir mucosal pharmacology and how reliable is the ex vivo tissue model for studying this effect?
• What does this study add to our knowledge?

  • ○
    Consistent with clinical data, there was a decrease in tenofovir's active metabolite, tenofovir diphosphate (TFVdp), and in the ratio of TFVdp to endogenous nucleotide dATP as Prevotella bivia concentrations increased. However, there was an increase in tenofovir efficacy as P. bivia concentrations increased. Viral replication also increased in the presence of Prevotella bivia , suggesting an interplay between viral replication and the vaginal microbiome.
• How might this change clinical pharmacology or translational science?

  • ○
    This work provides more evidence on the relationship between the vaginal microbiome and tenofovir mucosal concentrations and efficacy. Future work developing models to study this interplay between the vaginal microbiome and antiretroviral PK/PD is critically needed. This study showed that the ex vivo tissue model is a credible model and laid the groundwork for development of such models.

1. Introduction

The contribution of the microbiome to variability in drug response is relatively new and underexplored. Growing evidence suggests that inter‐individual variability in a drug's efficacy and toxicity is impacted by the microbiome, including the vaginal microbiome [1, 2, 3, 4, 5, 6]. However, there are ethical and regulatory limitations to studying drug‐microbiota interactions in humans, and in silico predictive models are limited by the lack of data on drug‐microbiota interactions, especially within the vaginal microbiome. Standardized methodologies are lacking; microbiomes are complex and constantly evolving [2, 5]. More pre‐clinical models are needed to further understand the drug‐microbiota interactions.

Globally, almost half of the 1.3 million new human immunodeficiency virus (HIV) infections in 2023 were in young girls and women [7]. Therefore, HIV prevention measures, specifically the prevention of transmission via vaginal intercourse, are critical to continue reducing new HIV infections.

The vaginal microbiome plays a complex role in tenofovir efficacy [6, 8]. A “healthy” vaginal microbiome consists mainly of Lactobacillus species, which maintain a low pH through the production of lactic acid and prevent overgrowth of pathogenic bacteria [5, 6]. Vaginal dysbiosis, also clinically known as bacterial vaginosis, occurs when anaerobic and/or pathogenic bacteria overgrows [9, 10]. Vaginal dysbiosis in women of reproductive age is associated with different genera including, but not limited to, Gardnerella, Prevotella, Sneathia, Megashaera, and Dialister [11]. Women with dysbiotic vaginal microbiomes had 3‐fold lower protection from HIV acquisition compared to women with Lactobacillus‐dominant microbiomes with a 1% tenofovir gel [3]. However, vaginal dysbiosis does not seem to affect oral tenofovir efficacy. Oral tenofovir disoproxil fumarate + emtricitabine (Truvada) in serodiscordant couples showed no difference in efficacy between vaginal microbiome community types [8]. These data suggest that the effect of the microbiome on tenofovir efficacy depends on the route of administration. Although the present study does not examine route‐dependent differences, it focuses on clarifying how antiretroviral pharmacology interacts with the vaginal microbiome within the mucosal compartment—an understanding that is critical for interpreting PrEP efficacy at the local level.

Here, we investigated the effect of the vaginal microbiome, using the species Prevotella, on tenofovir PK/PD in cervicovaginal tissues using an ex vivo human tissue model, regardless of administration route. While Gardnerella vaginalis is strongly implicated with decreased topical tenofovir efficacy, Prevotella species, especially Prevotella bivia ( P. bivia ), were selected for this study because it is a key anaerobic bacterium in vaginal dysbiosis; P. bivia produces metabolites that raise the vaginal pH and promote the growth of other anaerobic bacteria, especially G. vaginalis . Clinical data have shown that P. bivia is associated with decreased topical tenofovir efficacy [3, 9, 12]. In addition, we previously showed that P. bivia was negatively correlated with TFVdp:dATP ratio in a dose‐challenge study [13].

2. Methods

2.1. Ethics

This study was approved by the University of Minnesota Institutional Review Board. De‐identified, HIV‐negative vaginal and cervical tissues from deceased women were procured from National Disease Research Interchange (NDRI). Informed consent was collected from the participant prior to passing or the participant's family post‐death.

2.2. Tissue Collection and Processing

In brief, tissues were collected from the deceased between 11‐ and 14 h post‐death between July 2023 and February 2024. Tissues were collected in Iscove's Modified Dulbecco's Medium (Gibco, Waltham, MA). Two donors had both cervical and vaginal tissue collected while six contained vaginal tissue only. The tissue type is denoted by V (vaginal) or C (cervical) and their sequential donor number.

2.3. P. bivia Preparation

Anaerobic conditions were maintained using an AnaeroPack‐Anaero gas chamber (Mitsubishi Gas Chemical Company Inc., Tokyo, Japan). Proved to be highly effective in establishing and supporting the growth of clinically significant anaerobic bacteria, including P. bivia [14]. Every time the chamber was opened, a new sachet was added and closed within 1 min. Modified Reinforced Clostridial Medium (MRCM) was reduced by incubating overnight at 37°C in AnaeroPack to remove oxygen from the media. A lyophilized pellet of P. bivia (ATCC 29303) was reconstituted in reduced MRCM per manufacturer's instructions. Reconstituted P. bivia culture was used to inoculate cryobead tubes (Key Scientific, Stamford, Texas, United States) to create bead stocks. Bead stocks were flash frozen and stored at −80°C. P. bivia from a single bead was passaged up to five times on reduced trypsin soy agar (TSA) with 5% sheep blood agar plate.

2.4. P. bivia Quantification

A single colony of P. bivia, incubated in 3 mL of reduced MRCM for 24 h at 37°C in AnaeroPack, was grown to optical density at 600 nm (OD₆₀₀ = 0.138). This was serially diluted 10‐fold to a concentration of 10⁻⁵, then dilutions of 10⁻³, 10⁻⁴, and 10⁻⁵ were plated on reduced TSA with 5% sheep blood agar plates. Plates were grown overnight at 37°C in AnaeroPack and colonies subsequently counted. Colony forming units (CFU)/mL were calculated by the equation: $\frac{\mathit{\text{Number of colonies}}\times \mathit{\text{Dilution factor}}}{\mathit{\text{Culture volume}}\left(\mathit{mL}\right).}$

2.5. Comparing tenofovir diphosphate (TFVdp) Formation Between Aerobic and Anaerobic Conditions

Tissue culture media consisting of Iscove's Modified Dulbecco Medium, 10% fetal bovine serum, 2.5% nystatin, 1% penicillin and streptomycin, and 1% Minimum Essential Media was freshly prepared. Tissues were collected and processed as stated above from eight tissue donors. Explants were incubated in two 48‐well tissue culture plates with 300 μg/mL tenofovir in 500 μL of tissue culture media in each well, one plate in AnaeroPack and the other aerobically. Both conditions were incubated at 37°C with 5% CO2 for 24 h. Explants were then collected, weighed, snap frozen in liquid nitrogen, and stored at −80°C until drug quantification was performed.

2.6. Antiretroviral Quantification

Tenofovir diphosphate (TFVdp) and endogenous dATP were measured in tissue homogenates using validated liquid chromatography tandem mass spectrometry methods [1, 15]. The dynamic range for TFVdp and dATP was 0.02–100 ng/mL and 0.02–20 ng/mL, respectively. Any samples below the sample specific lower limit of quantification were reported as 50% of the lower limit of quantification. TFVdp and dATP concentrations were normalized for tissue weight and presented as fmol/mg.

2.7. Viral Propagation and Tissue Infection

Figure 1 depicts the experimental scheme. HIV‐1 JRcsf was chosen for viral challenge as it is both R5 and T‐cell tropic, which is characteristic of viruses responsible for > 90% of the sexual transmission of HIV infection [16]. HIV‐1_JRcsf (NIH AIDS Repository Program) was propagated in human peripheral blood mononuclear cells (PBMC) following 24 h phytohemagglutinin (PHA) stimulation. Virus was harvested on day five post‐infection. Tissue culture infectious dose 50% (TCID₅₀) titers of the harvests were determined in TZM‐bl cells [17, 18] using a luciferase reporter assay Bright‐Glo (Promega) and calculated using the Reed and Muench method [19]. To evaluate infection in explant tissues, explants were exposed to a high titer (10⁶ TCID₅₀) of HIV‐1_JRcsf.

FIGURE 1.

Experimental Design Schematic. On the day prior to viral incubation, tissue was processed to remove submucosa, and epithelial tissue explants were created with a biopsy punch. Explants were incubated for 24 h with and without 300 μg/mL of tenofovir and 10³ to 10⁵ CFU/mL of Prevotella bivia . Explants were collected for drug quantification before a three‐hour viral incubation of 10⁶ TCID₅₀ 10⁶/ per well of HIV_JR‐CSF. Baseline explants were collected immediately following viral incubation for baseline viral replication. Explants were placed on a gel foam raft in a 12‐well plate and incubated for 2 days. All explants were collected on day 2 for viral replication quantification. Supernatants were collected for p24 quantification (data not shown).

Prior to viral infection, tissues were received and processed as described above. P. bivia was diluted to 10⁵, 10⁴ and 10³  CFU/mL in a 48‐well plate with and without 300 mg/μL tenofovir. P. bivia concentrations were informed by the total bacterial load of the vaginal microbiome, which is estimated between 10⁶ and 10⁸ CFU/mL [20, 21]. An aerobic positive control plate was prepared for each donor. Every experiment, in both aerobic and anaerobic conditions, included a Prevotella‐free control containing only virus, a Prevotella‐free control (tenofovir only), and a viability control (no drug, bacteria, or virus). Tissue viability was assessed as described below. After a 24 h incubation with tenofovir, wells were inoculated with 100uL HIV‐1_JRcsf to achieve 10⁶ TCID50 per well. Tissues were incubated with virus for 3 h at 37°C in respective conditions.

Following viral incubation, two 12‐well plates with wells containing a total volume of 750 μL comprising culture media, P. bivia between 10⁵ and 10³ CFU/mL if applicable, and tenofovir at 300 mg/μL if applicable, were incubated in the AnaeroPack at 37°C overnight to reduce oxygen levels in the media (reduced media). A high concentration of tenofovir (300 ug/mL) was used to avoid TFVdp measurements below the limit of quantification.

Gel foam was cut into 3mm³ blocks, pre‐soaked in culture media for at least 30 min prior to use, and placed in each well. After viral incubation, all explants were washed three times with culture medium for ~30 s each. One explant from each donor was stored in RNAlater (Invitrogen) for quantification of baseline spliced viral RNA and the rest were moved onto gel foam rafts. Explants were collected after 48 h for spliced viral RNA quantification (see below).

2.8. Tissue Viability

Tissue viability in explants in culture was assessed quantitatively at baseline, at time of viral infection, and at end‐of‐culture period using the methylthiazolydiphenyl‐tetrazolium bromide (MTT; Sigma) assay as previously described [16, 22]. Briefly, explants were collected and incubated in 200 μL of 250 mg/μL of MTT for 3 h at 37°C. Explants were then transferred to 1 mL of methanol overnight to extract formazan. Formazan was quantified by measuring absorbance at 570 nm. Viability score was calculated by the following (average absorbance‐blank absorbance)/(tissue weight).

2.9. Spliced Viral RNA Assay

To assess viral RNA production, RNA was extracted using the Qiagen RNeasy Mini Kit per manufacturer's instructions and transcribed to cDNA using SuperScript VILO IV cDNA Synthesis Kit (Life Technologies). RNA was quantified using the Nanodrop (Thermo Scientific). A real‐time PCR assay was designed to specifically detect mRNA copies from the D4‐A7 splice site in the HIV‐1_JR‐CSF genome [16]. Forward primer 5‐CCT CAA GAC AGT CAG ACT CAT CAA G‐ 3; Reverse primer 5‐CTG CGG GTC CCC TCG TT‐3; probe‐ 5‐/FAM/TAC CAA AGC AAC CCTC/MGB/−3. A standard curve was calculated using a 10‐fold serial dilution of a cloned plasmid of the splice variant. The spliced mRNA copy number per milligram tissue was determined for each explant.

2.10. Biostatistical Analysis

Average TFVdp formation between anaerobic and aerobic conditions was compared using the Wilcoxon rank‐sum test. Variability was quantified by calculating the percent coefficient of variation (CV%) using the equation: $\frac{\mathit{\text{standard deviation}}}{\mathit{\text{mean}}}x100\%$. Outliers were determined for antiretroviral concentrations and viral replication by calculating Z‐scores: $\frac{\mathit{\text{sample}}-\mathit{\text{mean}}}{\mathit{\text{standard deviation}}}$. Any samples more than 1.5 standard deviations away from the mean were considered outliers. TFVdp, dATP, and TFVdp:dATP ratio were compared among P. bivia concentrations and the tenofovir‐only control by Kruskal‐Wallis test. HIV RNA fold change was calculated by dividing RNA of each sample by the positive control at day zero. Viral replication was reported as mean fold change in viral RNA. Percent inhibition was calculated by $\frac{\left(\mathit{\text{control}}-\mathit{\text{sample}}\right)}{\mathit{\text{control}}}x100\%$. Spearman rank correlation test was used to assess correlations between percent HIV inhibition, TFVdp, dATP and TFVdp/dATP ratio.

3. Results

3.1. Donor Characteristics

Tissues were collected from eight donors between July 2023 and February 2024. Most tissue donors were post‐menopausal and white (Table S1). The average time between death and tissue processing was 43.5 h, with a range of 31–53 h.

3.2. Tissue Viability Between Anaerobic and Aerobic Conditions

The mean MTT viability scores in anaerobic and aerobic conditions were 0.11 and 0.10, respectively. Mean MTT viability scores in aerobic and anaerobic conditions increased 112% and 60% between day −1 and 0, respectively. Viability continued to increase slightly through day 2 for both conditions (anaerobic, 16%; aerobic, 5%; Figure S1). There were no distinct trends in viability over time between any of the tissue donors. The CV% was 63% for aerobic conditions and 46% for anaerobic conditions.

3.3. TFVdp and dATP Concentrations Between Aerobic and Anaerobic Conditions

TFVdp and dATP were quantifiable in 90% and 63% of samples, respectively. The mean TFVdp concentration in aerobic and anaerobic conditions was 985.1 and 440.7 fmol/mg, respectively (Figure S2). TFVdp was 76% lower in anaerobic conditions; however, this was not statistically different between conditions (p = 0.66). There was high inter‐donor variability with a CV% of 151% and 207% for aerobic and anaerobic conditions, respectively. The mean (CV%) concentrations of dATP in aerobic and anaerobic conditions were 28.3 (117%) and 14.0 fmol/mg (114%; p > 0.90).

3.4. Changes in TFVdp, dATP, and TFVdp:dATP in the Presence of P. bivia

For V15, our one premenopausal donor, TFVdp and TFVdp:dATP concentrations were 2.55 standard deviations away from the mean, so it was removed from these analyses. While P. bivia 10³ CFU/mL had slightly higher TFVdp concentrations compared to the Prevotella‐free control, TFVdp concentrations trended lower as P. bivia concentrations increased (Figure 2A). However, there was no statistical difference in TFVdp concentrations between any of the four conditions (p = 0.62). TFVdp concentrations and P. bivia were not correlated (R = −0.21, p = 0.25). Inter‐donor variability of TFVdp concentrations was high with CV% ranging from 164% to 304% for all Prevotella conditions and the Prevotella‐free control.

FIGURE 2.

Changes in TFVdp, dATP, and their ratio by Prevotella bivia condition (A) TFVdp fmol/mg (left) and (B) dATP fmol/mg (middle) were quantified in explants after a 24 h incubation with 300 μg/mL tenofovir and varying concentrations of Prevotella bivia (ATCC 29303; 10³–10⁵ CFU/mL). Explants were incubated in an anaerobic chamber at 37°C. TFVdp and dATP were quantified in tissue by liquid chromatography tandem mass spectroscopy (C). TFVdp: DATP ratio (right) is calculated by dividing TFVdp by dATP concentration for each sample. N = 8 de‐identified tissue donors. The black horizontal lines denote the mean concentration. TFVdp, tenofovir diphosphate; dATP, endogenous deoxyadenosine triphosphate; CFU/mL, colony forming units per one milliliter.

There was no correlation between dATP and P. bivia concentrations (R = 0.07, p = 0.68). Both the lowest and highest P. bivia concentrations were lower than the Prevotella‐free control (Table 1; Figure 2B). P. bivia 10⁴ CFU/mL resulted in dATP over two‐fold higher than the Prevotella‐free control (20558.6 and 14261.7 fmol/mg, respectively). The CV% ranged from 97 to 134% with and without Prevotella, respectively. There were no statistical differences in median dATP concentrations between the four conditions (p = 0.36; Table 1).

TABLE 1.

Effect of Prevotella bivia on tenofovir diphosphate, endogenous adenosine, and their ratio in anaerobic conditions in tissues from postmenopausal women.

	Prevotella‐ free	103 CFU/mL a   Prevotella bivia	104 CFU/mL Prevotella bivia	105 CFU/mL Prevotella bivia	Kruskall Wallis p
TFVdp b  fmol/mg mean (CV%)	210.1 (206)	232.1 (160)	64.5 (147)	15.4 (166)	0.62
dATP c  fmol/mg Mean (CV%)	15.5 (126)	9.1 (104)	23.1 (87)	13.4 (102)	0.36
TFVdp: dATP Mean (CV%)	59.2 (245)	77.5 (212)	8.8 (168)	3.4 (120)	0.42

^a Colony forming units per one milliliter.

^b Tenofovir diphosphate.

^c Endogenous deoxyadenosine triphosphate.

The ratio of TFVdp:dATP followed the same trend as TFVdp, where TFVdp:dATP trended to decrease as P. bivia increased (Table 1). However, there was no significant correlation between TFVdp:dATP and P. bivia concentrations (R = −0.2, p = 0.26). The TFVdp:dATP for 10⁴ and 10⁵ CFU/mL P. bivia conditions were over six‐fold and 17‐fold lower than the Prevotella‐free control, respectively.

3.5. HIV RNA Inhibition

As one vaginal donor (V08) had significantly higher viral replication than the rest of the donors, this outlier was removed for all analyses. Viral replication from baseline to day 2 for the Prevotella‐free control was 3.2. In the tenofovir‐free controls, both the 10³ and 10⁵ CFU/mL conditions of P. bivia saw lower viral replication compared to the Prevotella‐free control (Figure 3). Meanwhile, 10⁴ CFU/mL Prevotella without tenofovir had higher viral replication compared to the Prevotella‐free control (Figure 3).

FIGURE 3.

Changes in HIV RNA fold change by Prevotella bivia concentration HIV RNA was extracted from vaginal tissue and quantified using qPCR. The fold change of HIV RNA was calculated for each Prevotella bivia concentration (10³–10⁵ CFU/mL) by dividing RNA concentrations of the sample by the positive control RNA at day zero. The percent difference between each drug and no drug conditions is calculated using the equation for percent difference. The dashed line is at a fold change of 1, which is considered a productive infection of an explant (A). HIV RNA fold change in anaerobic conditions and (B) HIV RNA foldchange in aerobic control. TFV, tenofovir; HIV, human immunodeficiency virus; RNA, Ribonucleic acid; CFU/mL, colony forming units per milliliter.

All three P. bivia conditions without tenofovir had higher viral replication than the control, consistent with the initial hypothesis that the presence of Prevotella would increase viral replication (Table 2). As P. bivia concentrations increased, tenofovir efficacy tended to increase, though not significantly (Table 2). Tenofovir decreased viral replication by 33% in the lowest P. bivia concentrations compared to a 38% increase in the Prevotella‐free control; however, this difference was not significant (p = 0.054, Figure 3). Unexpectedly, the positive control had an increase, although not significantly, in viral replication when tenofovir was added (Table 2; R = 0.26, p = 0.66). Only two cases had productive infection in the Prevotella‐ and tenofovir‐free positive control (Figure 3). Notably, productive infection in the virus only control was limited to a single donor under aerobic conditions.

TABLE 2.

Viral Replication and Prevotella bivia relationship in anaerobic conditions. The mean fold change in HIV‐1 RNA is shown between the control and three P. bivia concentrations with and without tenofovir. The percent change in viral replication is shown in the rightmost column. A negative number indicates lower viral replication with tenofovir present.

	Without tenofovir	With tenofovir	Change in viral replication
Control	1.29 (1.25)	1.9 (0.9)	+38%
103 CFU/mL	1.69 (0.99)	1.2 (0.29)	−33%
104 CFU/mL	1.89 (1.27)	1.34 (0.49)	−34%
105 CFU/mL	1.79 (1.12)	1.09 (0.6)	−48%

Abbreviations: CFU/mL, colony forming units per milliliter; HIV, human immunodeficiency virus.

4. Discussion

Many components in the microenvironment of the female genital tract may alter mucosal antiretroviral pharmacokinetics (PK) and pharmacodynamics (PD), including the vaginal microbiome [3, 23, 24]. Here, we sought to understand the relationship between the vaginal microbiome and tenofovir PK/PD using the ex vivo model. We hypothesized that TFVdp:dATP would be correlated with P. bivia . Consistent with our hypothesis, there was over a six‐fold and 17‐fold decrease in TFVdp:dATP in the 10⁴ and 10⁵ CFU/mL conditions, respectively. This same trend was seen with TFVdp concentrations, where there was over a three‐fold and 13‐fold decrease of TFVdp in the 10⁴ and 10⁵ CFU/mL conditions, respectively. These data suggest that P. bivia decreased TFVdp concentrations, thus decreasing the TFVdp:dATP ratio. For optimal efficacy in vitro, the ratio of TFVdp:dATP should be greater than 1 [25]. For all conditions, the ratio was magnitudes higher than 1, suggesting that the ratio of 1 is limited to in vitro conditions.

Despite the lower TFVdp:dATP, explants in P. bivia concentrations of 10⁴ and 10⁵ CFU/mL had the highest tenofovir efficacy. Another study found that specific anaerobic bacteria from vaginal swabs, although not Prevotella, decreased endocytosis of tenofovir in human vaginal epithelial cells [26]. Endocytosis of tenofovir into vaginal cells was inhibited by increased concentrations of adenine, the purine base of dATP in culture. This correlation increased the half‐maximal inhibitory concentration needed for protection against HIV transmission. This could suggest that as more dATP is present, less tenofovir is being endocytosed and, consequently, being converted to TFVdp. However, this study did not measure TFVdp or assess pharmacodynamics. Endocytosed tenofovir was not directly quantified in our study, but TFVdp is an indirect measure of tenofovir. In another study, different bacterial community profiles showed increased HIV copies per cell in an immortalized vaginal epithelial culture model [27]. In addition, there was also a decrease in topical tenofovir efficacy with bacterial vaginosis that included Prevotella genera. This suggests that our model is consistent with in vitro models where anaerobic bacteria are increasing HIV RNA in the tissue model. However, our model indicated an increase in tenofovir efficacy. There are many factors that could affect TFVdp concentrations and efficacy, including immune cell populations, hormones, and drug metabolizing enzymes, and transporters [28]. The role of immune cells in these explant models requires further investigation.

Despite no statistical difference, both TFVdp and TFVdp:dATP ratio were negatively correlated with P. bivia concentrations in postmenopausal donors (Figure 2C). All of our analyses were done with post‐menopausal tissues. While V15 was excluded as an outlier for analyses, it is important to note that it was the only pre‐menopausal donor. This is consistent with our previous studies that showed that premenopausal donors had higher TFVdp formation [29]. However, no significant conclusions can be made with an N of 1, so future studies investigating microbiome‐drug interactions in pre‐menopausal women are needed.

A single high concentration of tenofovir was used in this experiment as lower concentrations of tenofovir could lead to more variability in TFVdp formation and efficacy and the limited amount of tissue for each donor. The goal of this work was to understand how TFVdp and dATP are affected by the microbiome, and specifically P. bivia in this study. This approach allowed for less inter‐donor variability and more data on how TFVdp and dATP are affected by this species. Future work should focus on developing a dose–response curve for TFVdp with P. bivia present to predict an EC₅₀ and understand how P. bivia affects tenofovir efficacy. Moreover, community‐level interactions may also impact these relationships, leading to the development of more sophisticated models to understand inter‐species interactions on drug availability.

One consideration is the site in the genital tract where HIV transmission typically occurs. Endocervical tissue has a single columnar epithelium and the transformation zone has high cell turnover, facilitating viral entry [30]. Vaginal tissue and ectocervical tissue have a stratified squamous epithelium, which impedes HIV transmission [31]. In addition, cervical tissue drives local cell‐mediated immunity in the female reproductive system and immune cells, including antigen presenting cells and CD4+ T cells, are most prevalent in the transformation zone [32]. These data suggest that cervical tissue is the main site of HIV transmission, although vaginal transmission is possible. The findings presented here should be evaluated in cervical tissue to further understand how tenofovir exposure and efficacy are affected by the vaginal microbiome.

The ex vivo cervicovaginal tissue model is an important preclinical tool to understand mucosal dose–response relationships. Colorectal, ectocervical, and vaginal tissues have been used to determine the efficacy of many antiretrovirals including tenofovir, emtricitabine, dapivirine, maraviroc, and dipivefrine [33, 34, 35, 36, 37]. Our lab has an established ex vivo vaginal tissue model that has been used to predict TFVdp dose–response relationships and identify an EC₅₀ of 716 fmol/mg [16]. The validity of this model was confirmed with clinical data [6]. Steady state TFVdp concentrations after oral tenofovir dosing are estimated to be ~555 fmol/mg and these concentrations provided ~75% efficacy as seen in the Partners PrEP study [38, 39]. Modifying the model to examine microbiome interactions can provide insight into the role of the vaginal microbiome in PrEP mucosal pharmacology in the FGT. The results from this study suggest that TFV mucosal pharmacokinetics are affected by P. bivia ; more work studying PK/PD‐microbiome interactions is needed.

This study was limited by the small sample size and high variability; therefore more data on TFVdp and dATP in this model are needed to fully understand the effect of P. bivia on TFVdp PK/PD. The explant model has inherent inter‐donor variability which limits the interpretation because such high variability with a low sample size makes it difficult to determine statistical trends in the data. Other studies report high inter‐donor variability in the explant model ranging from 0.63% to 191% [33, 40, 41], but there is no defined level of acceptable variability. The field should work together to come to a consensus on an acceptable rate of inter‐donor variability in the model. Also, some donor tissue did not have a productive infection. This is consistent with previous studies, where infection rates were between 25% and 80% [40, 41, 42, 43, 44]. Without productive infection in all control samples, it is difficult to understand how P. bivia is affecting viral replication. Finally, here only one bacterial species was used, but the vaginal microbiome is a complex, ever‐evolving community. It is known that Lactobacillus species can be protective. Further work investigating other anaerobic species individually and combined in this model is needed.

In conclusion, P. bivia negatively affected TFVdp mucosal concentrations in our ex vivo tissue model. These findings seem to be consistent with clinical data and therefore support the credibility of the model for testing current and future PrEP candidates. Further work increasing the sample size, diversity of human donors, and testing other dysbiosis‐associated species is needed. This will allow for better translatability and increased understanding of how hormonal differences between pre‐ and postmenopausal tissues affect TFVdp concentrations and efficacy, especially with vaginal dysbiosis. With the growing interest in microbiome‐drug interactions, models that can capitulate these interactions are sorely needed. We believe our manuscript describes an early application of such a model and in part identifies areas where gaps exist.

Author Contributions

A.M.L. wrote the manuscript; A.M.L. and M.R.N. designed the research; A.M.L. and L.B.C. performed the research; A.M.L. analyzed the data; E.B.H. and C.S. contributed new reagents/analytical tools. All authors participated in the editing of the manuscript.

Funding

This work was supported by the National Institute of Allergy and Infectious Diseases at the National Institutes of Health (K08 AI134262 to M. R. N), the National Center for Advancing Translational Sciences at the National Institutes of Health (TL1R002493 and UL1TR002494 to A. M. L.) and a University of Minnesota Academic Health Seed Grant. Authors report no conflicts of interest relevant to this work.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Tissue Viability over length of experiment.

An explant is taken out of culture from aerobic (left) or anaerobic (right) conditions and incubated with 250 ug/ml of MTT for 3 h at 37C. Explants are transferred to 1 mL of methanol to extract formazan. Formazan is quantified by measuring absorbance at 570 nm. The absorbance is normalized by tissue weight.

Figure S2: TFVdp and dATP concentrations by environment condition.

A. TFVdp and B. dATP were quantified in explants after a 24 h incubation with 300 μg/mL tenofovir. Explants were incubated either in 5% CO2 (aerobic) or in anaerobic chamber at 37C. TFVdp and dATP were quantified in tissue by liquid chromatography tandem mass spectroscopy. N = 9 de‐identified tissue donors.

Table S1: Donor Characteristics.