Methods to characterize the vaginal microbiome in a rhesus macaque model of simian human immunodeficiency virus (SHIV) transmission uncover epithelium-associated enrichment of Prevotella

Highlights

• •Prevotella spp were enriched at the vaginal epithelium, where they likely modulate HIV transmission.
• •Functional predictions suggest that Prevotella spp enrichment is associated with increased sialidase generation.
• •Traditional sampling methods to characterize the vaginal microbiome, as well as methods utilizing cervicovaginal fluid, are likely inadequate to fully characterize the epithelial niche.

Abstract

The vaginal microbiome plays a crucial role in maintaining mucosal integrity and mitigating pathogen transmission, yet its comprehensive characterization remains challenging due to limited sampling and analysis methods. In this study, we aimed to characterize bacterial and fungal taxa diversities in the vaginal microbiomes of Simian Human Immunodeficiency (SHIV)-infected rhesus macaques, as well as their metabolic activities, using three sampling methods. The cervicovaginal lavage (CVL), vaginal swab, and vaginal mucosal tissue methods offer novel insights into microbial diversity and their potential impacts on HIV transmission. Using 16S rRNA and Internal Transcribed Spacer (ITS) sequencing, we assessed bacterial and fungal community composition and abundances, respectively, across all sampling methods. PICRUSt2 was used for functional predictions, and a modified glycosidase assay to further characterize glycan-degrading enzymatic activity in CVL samples. Our findings reveal that tissue samples were uniquely enriched for microbial taxa such as Prevotella spp. and Helicobacter spp., showing notable abundance differences compared to CVL and swab samples. Tissue samples exhibited higher alpha diversity and distinct metabolic prediction profiles, particularly elevated sialidase activity. While fewer differences were found in fungal microbiome composition and diversity, marked correlations were observed between bacterial and fungal taxa, emphasizing complex interkingdom interactions. These results highlight the significance of sampling methods in microbial ecology studies, which should be carefully considered due to their potential influence on pathogen transmission risk.

Introduction

Currently, over 40 million people are living with Human Immunodeficiency Virus (HIV) (WHO, 2025). HIV can be transmitted through various routes, but most transmissions occur globally via the vaginal route (Korenromp et al., 2024). This is in part due to HIV primarily targeting CD4+ T cells, in which the vaginal sub-epithelium is enriched (Hladik et al., 2007). Additionally, the vaginal squamous epithelium is embedded with dendritic cells, which facilitate the trans-infection of CD4+ T cells in the sub-epithelium by viral transfer (Shey et al., 2015). HIV infection begins with the interaction of its highly glycosylated envelope glycoproteins with the host cell, initially binding to the CD4 receptor and coreceptor CCR5/CXCR4. This interaction promotes the fusion of the viral envelope with the CD4+ T-cell, allowing viral entry. Once inside, HIV integrates its genome into the host DNA, establishing infection and enabling replication (Wilen et al., 2012).

The vaginal epithelium plays a crucial role in innate immune function as a physical barrier to prevent pathogens from establishing a productive infection. This barrier is populated with various immune cells, including Langerhans cells, CD4+ T cells, and α4β7⁺ lymphocytes (Shey et al., 2015; Byrareddy et al., 2014). In addition to cellular defenses, a protective mucus layer overlays the epithelium. This mucus is composed primarily of glycoproteins known as mucins, which enhance host defense by inhibiting the passage of pathogens such as viruses. Mucins undergo extensive glycosylation, resulting in complex carbohydrate chains covalently attached to a protein core (Reily et al., 2019). These carbohydrate-rich structures retain water, forming a gel-like matrix that traps and immobilizes pathogens and provides a habitat for a diverse microbial community of microorganisms - bacteria, fungi, and viruses, that can be either beneficial or harmful to the host. Collectively, this microbial ecosystem is known as the vaginal microbiome (Vagios and Mitchell, 2021).

Significant heterogeneity exists in human microbiomes. A common microbiome type that includes the key beneficial and abundant bacterium in a healthy vaginal microbiome is Lactobacillus. One of its primary protective functions is the production of lactic acid, which maintains the vaginal pH between 3.5 and 4.5 (Liu et al., 2023). This acidic environment creates unfavorable conditions for pathogenic bacteria, thereby contributing to host defense. Another beneficial characteristic of Lactobacillus is its low immunogenicity. Unlike certain other bacteria, Lactobacillus species lack several pattern-recognition receptor agonists, such as lipopolysaccharide, which are commonly associated with pro-inflammatory immune responses (Decout et al., 2024). In contrast, bacteria like Sneathia elicit strong immune activation. Vaginal antigen-presenting cells can recognize Sneathia antigens, leading to the upregulation of pro-inflammatory cytokines such as IL-1α, IL-1β, and IL-8 (Theis et al., 2021). This inflammation can be detrimental to immune defense, as it increases HIV transmission by creating an environment conducive to infection (Theis et al., 2021). Chlamydia is a sexually transmitted infection (STI) that facilitates HIV transmission by disrupting mucosal barriers and promoting the recruitment of CD4+ T cells to the site of infection (Tan et al., 2009; Dzakah et al., 2022). Other STIs, such as gonorrhea and syphilis, also elevate HIV transmission risk through their ability to induce local inflammation (Jarvis and Chang, 2012; Wu et al., 2021). Inflammatory responses at mucosal surfaces increase the number of CD4+ T cells, thereby providing more target cells, thus making HIV transmission easier (Masson et al., 2015). Although some healthy vaginal microbiomes are dominated by Lactobacillus and have minimal Sneathia, disturbances in this microbial balance can lead to vaginal dysbiosis and the development of bacterial vaginosis (BV).

In addition to the two well-characterized vaginal microbiome phenotypes, recent studies have begun to recognize more heterogeneity in human vaginal microbiomes, including a phenotype with a notably higher diversity state (Ravel et al., 2011; Gosmann et al., 2017). Generally, these higher-diversity vaginal microbiomes are associated with a greater risk of HIV acquisition, although the consistency of this correlation requires further investigation (Gosmann et al., 2017). The rhesus macaque (RM), a common model for human infectious disease, including HIV transmission, has been characterized as also typically having a higher diversity of vaginal microbiomes, though also with heterogeneity (Rhoades et al., 2021; Spear et al., 2010). This microbiome type includes high abundances of genera associated with BV, like Sneathia and Prevotella, as well as Mobiluncus and Gardnerella (Gosmann et al., 2017). Additionally, RMs typically have lower levels of Lactobacillus spp (Rhoades et al., 2021; Spear et al., 2010). Similar microbiome compositions are present in other macaques, including cynomolgus and pigtailed macaques, with studies consistently finding high Sneathia, Prevotella, and Mobiluncus; and low levels of Lactobacillus (Nugeyre et al., 2019; Spear et al., 2012). For this reason, it has been proposed that RMs and other macaques are suitable models for studying the more diverse subtype of vaginal microbiome (Rhoades et al., 2021). Notably, macaques release little or no glycogen from their vaginal epithelial cells, an adaptation present in humans, thought to facilitate Lactobacillus colonization, thereby keeping the vaginal pH low (Mirmonsef et al., 2012; Mirmonsef et al., 2014). Whether the species that colonize the RM vaginal niche rely on alternative energy sources remains unknown.

Despite the differences between human and RM vaginal microbiomes, macaque models remain the gold standard for HIV transmission studies, due to similar genetics and mucosal immune responses. Since the vaginal microbial niche influences HIV transmission risk and remains poorly characterized in this model, we sought to build upon previous findings characterizing the role of the vaginal microbiome composition in HIV transmission (Shedlock et al., 2009). Currently, there is a paucity of information on the vaginal epithelium, as most studies focus on sampling the vagina with an assumption that the microbial communities would be homogenous across niches. We hypothesized that different sampling methods, representing different vaginal regions, would capture unique bacterial and fungal taxa based on spatial location due to potential differences in oxygen concentrations, pH, and innate immune functions. To test this hypothesis, we collected cervicovaginal lavage (CVL), vaginal swabs, and vaginal mucosal tissue samples from SHIV-infected RMs, representing potentially unique vaginal niches. We then performed 16S rRNA sequencing and internal transcribed spacer (ITS) sequencing to identify the bacterial and fungal microbiome compositions, respectively. Additionally, we employed PICRUSt2 to predict potential microbial metabolic functions and developed a modified glycosidase assay to assess the glycan-degrading capacity of bacteria present in CVL samples. Our PICRUSt2 analysis revealed that sialidase-producing taxa were more prevalent in tissue samples compared to CVL and swab samples. These findings suggest that the bacterial microbiome, particularly Prevotella species enriched at the epithelial surface, may play a role in modulating glycosylation patterns and, therefore, pose a possible HIV transmission risk. These findings highlight a critical limitation of traditional sampling methods, which may fail to fully capture microbiome compositions relevant to vaginal HIV transmission risk. Therefore, the development of more comprehensive sampling strategies, including tissue-based assessments, is essential for accurately characterizing the vaginal microbiome and its role in HIV transmission and pathogenesis.

Methods

Ethics statement and animal experimental details

All animal procedures occurred in compliance with guidelines established by the Animal Welfare Act and the National Institutes of Health (NIH) for Housing and Care of Laboratory Animals, 8th Edition. Five female, Indian-origin, outbred, pathogen-free rhesus macaques (Macaca mulatta) were included in this study (Table 1). The macaques were housed in compliance with the regulations under the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals in the nonhuman primate facilities at the Department of Comparative Medicine, University of Nebraska Medical Center (UNMC), Omaha, Nebraska, USA. Animals were maintained in a temperature-controlled (72°F) indoor climate with a 12-hour light/dark cycle. The animals were fed a daily monkey diet (Purina) supplemented with fresh fruit or vegetables and water ad libitum. This study was reviewed and approved by UNMC Institutional Animal Care and Use Committee (IACUC) protocol number 18–108–08-FC and the Institutional Biosafety Committee (IBC). UNMC has been accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International.

Table 1.

Animal Characteristics, including animal ID, date of birth, gender, and necropsy details.

SHIV study animal details
Serial number	Animal ID	Date of Birth	Gender	Date of Necropsy
1	JJ2	6/1/08	Female	1/19/23
2	067	8/2/11	Female	1/25/23
3	088	10/21/11	Female	1/25/23
4	091	11/8/11	Female	1/4/23
5	127	9/9/08	Female	1/19/23

In this study, we used three different strains of SHIV, which include SHIV-1H, SHIV-4TH, and SHIV-5H (Dutta et al., 2018). The viral stock was generated using CD8+ T cell-depleted rhesus macaques PBMCs and titrated by TZM-bl assay as described by us previously (Dutta et al., 2018; Siddappa et al., 2011; Siddappa et al., 2009). In brief, each of the SHIV clones mentioned above was transfected into HEK293 cells using jetPRIME transfection reagent (Polyplus Transfection, NY, USA), and the culture supernatant was harvested after 72 hrs. Next, CD8+ T-cell-depleted PBMCs from uninfected rhesus macaques were stimulated with concanavalin A (25ug/ml; Sigma-Aldrich) were cultured for 72 hrs with human recombinant IL-2 (20U/ml). Then, the cells were spinoculated with culture supernatant collected from HEK293 cells. Post-infection, cells were washed and cultured in RPMI supplemented with 20 % FBS and IL-2 (20U/ml) at a density of 2 million cells/ml. Culture supernatants were collected on alternate days, and viral titer was estimated using TZM-bl assays (Siddappa et al., 2009). The titer of all three viruses was normalized, and they were mixed in equal proportion based on infectious titer in TZM-bl assays (TCID50/ml). Using this stock, a weekly low-dose Intravaginal (IVAG) challenge was performed for ten consecutive weeks. Two macaques (JJ2 and 127) became productively infected with mixed challenge stock as described above after 3^rd challenge, and then another two monkeys (088 and 091) became positive after 5^th challenge, while the 5^th macaque (067) did not acquire the viral infection even after the 10^th challenge. For this animal, a high-dose intra-rectal (IR) challenge was carried out, after which the macaque became positive for the SHIV infection. The animals were considered positive for SHIV infection after detection of two consecutive weeks of plasma viremia by real-time quantitative reverse transcriptase polymerase chain reaction (RT-qPCR). The mix of three SHIV stocks was used in this study to understand the transmission efficacy of the viral strains based on the glycosylation pattern of the envelopes selected from the T/F HIV strains.

Two weeks post-infection, combined anti-retroviral therapy (ART) was initiated for all five macaques included in the study. A daily single dose of subcutaneous injection (1 ml/kg body weight) of ART was given. ART regimen consisted of two reverse transcriptase inhibitors (FTC: 40mg/ml and TDF: 20 mg/ml) and one integrase inhibitor (DTG: 2.5mg/ml) (Acharya et al., 2021). Peripheral blood from the femoral vein was collected to monitor SHIV viremia. After 12/16 weeks post ART, a complete suppression of plasma viremia was achieved, and ART was stopped, and the animals were monitored for rebound viremia for another 12 weeks and then euthanized. At the time of necropsy, CVL, vaginal swab, and vaginal tissues were collected for microbiome analysis (Fig. 1). The vaginal wall was collected, washed with RPMI, and the epithelial rugae were dissected off with a clean scalpel.

Fig. 1.

Study schema. At the time of necropsy, first, a vaginal swab was taken, followed by cervicovaginal lavage (CVL). Vaginal wall was collected at necropsy, washed, and vaginal epithelium rugae were dissected. DNA was isolated from all samples and was used for 16S rRNA and ITS sequencing. CVL fluid was utilized for a spectrophotometry-based glycosidase assay.

DNA isolation method

Immediately following collection, all three types of samples were snap frozen and stored at −80°C for later isolation. DNA was isolated from CVL and swab samples using spin column chromatography-based Stool DNA Isolation Kit (Norgen Biotek Corp; Product #27,600), a product specifically designed for the isolation of DNA for microbiome analysis, following manufacturer recommendations. Vaginal rugae were lysed using a TissueLyser LT (Qiagen, Germantown, MD, USA; Product #69,980), and then DNA was isolated using an AllPrep DNA/RNA Mini Kit (Qiagen, Germantown, MD, USA; Product #80,204) according to manufacturer instructions. Isolated DNA was initially quantified by the GE SimpliNano spectrophotometer and shipped on dry ice to LC Sciences, LLC (Houston, TX, USA) for 16S rRNA Sequencing and Internal Transcribed Spacer (ITS).

16S rRNA sequencing

Upon receipt, a combination of Agilent 2100 Bioanalyzer, spectrophotometer, and gel electrophoresis was used to determine the quality of the samples. A library was generated with Phusion Hot Start Flex 2X Master Mix by amplifying the highly conserved combined V3-V4 16S rRNA regions using PCR primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R(5′-GACTACHVGGGTATCTAATCC-3′). These primers have been in-silico validated and have been used to characterize Lactobacillus-dominated vaginal microbiomes (Van Der Pol et al., 2019; Väinämö et al., 2025). PCR conditions were: ° for 30 s; 35 cycles at 98°C for 10 s, 54°C for 30 s, and 72°C for 45 s; and then 72°C for 10 mins. A combination of 2 % gel electrophoresis and Qbit fluorometer (Invitrogen, USA) was used to monitor progress and ensure the library was of acceptable fragment size and concentration (>0.3ng/μL). PCR products were isolated by 2 % agarose gel and purified with AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, USA).

Sequencing was performed with NovaSeq 6000, 2 × 250 bp (NovaSeq 6000 SP Reagent Kit, 500 cycles), at a depth ranging from 57,324 to 69,687 pair-end reads per sample. The length of >99 % of amplicons was 400–500 bp, at their expected length. A sequencing lane was reserved for a quality control sample. Illumina base calling and data filtering programs were applied to raw data to remove low-quality reads. Table density (fraction of non-zero values) was 0.172. Paired-end reads were assigned to samples by barcodes, truncated, and pair ends merged. Dereplication was performed with DADA2. An in-house script was used to identify taxa using the Ribosomal Database Project (RDP, Greengenes), and NCBI 16S Microbial databases by LC Sciences. Diversity indices were determined by QIIME2. FASTQ files are available on the NCBI Sequence Read Archive with the BioProject number PRJNA1260605.

Internal Transcribed Spacer (ITS) sequencing

A similar procedure was utilized to amplify the ITS2 region of the small subunit rRNA gene for fungal microbiome compositions, using a separate aliquot of isolated DNA. DNA was amplified with primers fITS7 (5′-GTGARTCATCGAATCTTTG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′). PCR conditions were: 98°C for 30 s; 35 cycles at 98°C for 10 s, 52/54°C for 30 s, and 72°C for 45 s; and then 72°C for 10 mins. DNA was purified and analyzed as before. Sequencing depth ranged from 38,978 to 74,635 pair-end reads per sample. 26 % of amplicons were <200 bp; 29 %, 200–300, bp; 39 %, 300–400; and 5 %, 400–500 bp. Table density (fraction of non-zero values) was 0.179. Following preliminary analysis, it was determined that several of the most abundant amplicon sequence variants were derived from the host and were thus excluded based on sequence alignment with Taxon ID 9544 (Macaca mulatta) on NCBI. Following this exclusion, 70–88 % of tags were determined to be valid and were utilized for further analysis.

Functional prediction and correlation analysis

Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) with the castor R package (Douglas et al., 2020) allowed for the prediction of metabolic pathways corresponding to 16S rRNA gene sequencing data, which was used as the input to assign sequences to specific taxonomic groups. Based on these abundances, it uses a reference database to predict potential functions of the taxa. The abundances are then normalized from known phylogenetic relationships between taxa, where the taxa in the samples are finally assigned mechanisms based on this (Douglas et al., 2020). Further analysis was performed comparing bacteria genera abundance with either other bacteria or with fungal species abundances, and the results of this analysis were generated with RStudio and R version 3.5.0 (R Core Team 2018). Correlation matrices were produced using the R package ‘corrplot’ Version 0.92 (Wei and Simko, 2010).

Glycosidase assay

Glycosidase assays were performed by fluorescence spectroscopy in a black, clear-bottom 96-well microtiter plate (Corning, Corning, NY, USA; Product #3603) based on methods utilized by Moncla et al. and further described as follows (Moncla et al., 2015). A calibration curve was created by placing 200 μL of 1, 0.1, 0.01, and 0.001 mM solutions of 4-Methylumbilliferone (4-MU) (Sigma-Aldrich, St. Louis, MO, USA; Product M1381) into separate wells. 0.2 mM solutions of 4-Methylumbilliferyl-sugar derivatives (α-d-glucopyranoside M9766; β-d-glucopyranoside M3633; α-d-galactopyranoside M7633; β-d-galactopyranoside M1633; α-l-fucopyranoside M8527; α-d-N-acetylneuraminic acid M8639) were prepared. Negative controls were prepared by placing 50 μL of each sugar derivative and 50 μL ddH2O into the same well to create a concentration of 0.1 mM. Control CVL solutions were prepared by placing 50μL of each CVL and 50μL ddH2O into the same well. Control water and buffer solutions were prepared by placing 200μL of ddH2O and 200μL of buffer in separate wells. Experimental solutions were prepared by placing 50 μL of each CVL and 50μL of each sugar solution into the same well and incubating for 20 min at room temperature. All wells except 4-MU control solutions, water, and buffer only received 100 μL pH 9.2 borate buffer (Thermo Fischer Scientific, Waltham, MA, USA; Product # AC383875000) to quench reactions. All solutions were placed in duplicate. Solutions were read by a SpectraMax i3x (Molecular Devices, San Jose, CA); excitation 365 nm, bandwidth 9 nm, and emission 450 nm, bandwidth 15 nm. Six flashes were used, and the shake setting was on for 15 s, medium orbit. Relative enzyme activity was determined by comparing the mean relative fluorescent units (RFUs) of CVL with 4-MU-based reagent to CVL alone.

Statistical analysis

Paired t-test was used to compare sampling types, except for predicted gene values, for which ratio paired t-test were utilized, due to the several fold differences between sample types. Linear regression analysis was performed to determine Pearson correlation coefficients. Statistical analyses were performed using GraphPad Prism 7 for Mac OS X, unless specified above (ie, PICRUSt2, CorrPlot).

Results

Differences in bacterial composition among sampling methods

In clinical research, the most common way of testing for BV is by Gram staining, which involves using a vaginal swab sample to make a smear and determine the abundance of bacteria with peptidoglycan-containing cell walls, which retain the crystal violet stain. Additionally, functional changes in the microbiome have been observed via CVL sampling before, but not through tissue sampling (Moncla et al., 2015). To see whether CVL, swab, and tissue sampling would detect niche-specific vaginal microbiome compositions, we began by using 16S rRNA sequencing to determine the bacterial taxa within the vaginal microbiome. By performing this comprehensive approach, we found that there were differences between sampling method compositions, with certain taxa being more elevated in certain sampling techniques. Some of these prominent differences were that Streptococcus was much more abundant in CVL samples than in swab samples or tissue samples, and that Campylobacter was much more abundant in swab samples than tissue samples or CVL samples (Fig. 2A). Additionally, there were prominent differences of taxa within the tissue samples where far less Sneathia, Escherichia-shigella, Porphyromonas, Negativicoccus, and Streptococcus while presenting far more Helicobacter (Fig. 2A). By looking closer at the individual RMs, 067 has the highest percentage of Porphyromonas when compared to the other RMs for all forms of sampling (Fig. 2B). 088 has the highest percentage of Streptococcus when compared to the other RMs for CVL and tissue sampling, especially CVL (Fig. 2B). Bacteroides have the highest relative abundance in the tissue samples and more specifically in those of 088, 091, and 127 (Fig. 2B). Although these RMs had the highest percentage of Bacteroides relatively, in general, these percentages were not as high as previously mentioned taxa were within their respective sampling methods (Fig. 2B). Notably, there was considerable heterogeneity between samples, suggesting a general trend like what is seen in the “diverse” subtype of vaginal microbiomes.

Fig. 2.

Bacterial taxa are differentially dispersed across vaginal niches. 16S rRNA sequencing was performed from DNA obtained from swab (swb), CVL, and vaginal tissue (tis). Genus taxonomic ranking was assigned, revealing differences in (A) grouped genera across sampling types and (B) considerable differences between individual monkeys. Alpha diversity was also determined based on (C) unique reads, indicating operational taxonomic units, and (D) the Shannon diversity index was calculated to measure species richness and evenness. Beta diversity was characterized with (E) Bray-Curtis dissimilarity and (F) unweighted UniFrac metric, Principal Component Analyses (PCoA) Plots.

To determine overall differences between taxonomic communities, alpha diversities were calculated (Fig. 2C). Tissue samples had a significantly higher mean operational taxonomic units (OTUs) count at 719, while CVL and swab samples had 341 (p = 0.007) and 370.40 OTUs (p = 0.02) respectively (Fig. 2C). The Shannon index of the tissue samples was significantly higher at 6.15 than those of the CVL and swab samples which were at 3.52 (p = 0.01) and 3.88 (p = 0.03), respectively (Fig. 2D). Beta diversities were calculated including both Bray-Curtis dissimilarity which measures dissimilarity based on species present in a sample (Fig. 2E), and unweighted UniFrac metric, which differentiates samples with respect to phylogenetic relationship (Fig. 2F). With this latter measure, swab samples overlap with both tissue and CVL, but tissue and CVL do not overlap (Fig. 2F).

Role of Lactobacillus and significance of its abundance

Under homeostatic conditions, Lactobacillus is the most abundant genus present in most human vaginal microbiomes and plays a critical role in suppressing pathogenic bacteria's growth by lowering the microenvironment’s pH (Liu et al., 2023). However, studies have found that RM microbiomes have consistently had lower levels of Lactobacillus compared with this human microbiome phenotype. When we analyzed the abundance of different Lactobacillus spp., no species throughout all of the RMs and the sampling methods had a greater abundance than 1.28 % (Fig. S1). The highest relative abundance of Lactobacillus sp. l-YJ (1.28 %) and Lactobacillus sp. BL304 (0.03 %) was observed in the RM 088 tissue sample. Lactobacillus crispatus (0.73 %) and Lactobacillus kitasatonis (0.16 %) were most abundant in the RM JJ2 swab sample, while Lactobacillus intestinalis showed the highest abundance (0.24 %) in the RM 091 CVL sample. (Fig. S1).

Differences in fungal composition amongst sampling methods

To further understand the differences detected between the sampling methods, we also looked at the fungal taxa within the vaginal microbiome. No significant differences were found between sampling methods. However, we observed that within tissue samples, Ascomycota unclassified, Candida parapsilosis, Basidiomycota unclassified, and Aspergillus penicilloides were the most abundant taxa (Fig. 3A). Within CVL samples, Ascomycota unclassified, Aspergillus versicolor, Malassezia restricta, Alternaria unclassified, and Basidiomycota unclassified were the most abundant taxa (Fig. 3A). Within swab samples, Ascomycota unclassified, Kazachstania telluris, Basidiomycota unclassified, Malassezia restricta, and Pseudopithomyces unclassified were the most abundant taxa (Fig. 3A). When looking at the individual RMs, we observed that JJ2 has a high abundance of Kazachstania telluris, especially in tissue and swab samples (Fig. 3B). 127 has a high abundance of Preussia minima when compared to the other RMs, especially for CVL and swab sampling (Fig. 3B). Basidiomycota unclassified was much more prevalent in 091 tissue samples than any other RMs or sampling technique (Fig. 3B). Although these RMs had the highest percentage of Basidiomycota relatively, in general, these percentages were not as high as previously stated taxa (Fig. 3B).

Fig. 3.

Differential fungal diversity by sampling method. Internal transcribed spacer (ITS) sequencing was performed for the swab (swb), CVL, and vaginal tissue (tis) samples. Relative abundances were determined at the species level for (A) the group and (B) individual macaques.

Interpretations of bacterial taxa associations concerning bacterial and fungal taxa

CorrPlots were generated by performing linear regression with all samples to show associations between different bacterial taxa regarding other bacterial and fungal taxa. It was observed that Fusobacterium is positively correlated with Porphyromonas (r = 0.81, p = 0.0002) and Negativicoccus (r = 0.60, p = 0.02) (Fig. 4A). Porphyromonas is positively correlated with Prevotella (r = 0.76, p = 0.001) and Negativicoccus (r = 0.92, p < 0.0001). Negativicoccus is positively correlated with Prevotella (r = 0.71, p = 0.003). Eschericia_Shigella is positively correlated with Actinotignum (r = 0.75, p = 0.0009) (Fig. 4A). In regard to the correlations of bacterial taxa to fungal taxa, we noted that Helicobacter is positively correlated with Aspergillus penicilloides (r = 0.99, p < 0.0001) (Fig. 4B). Fusobacterium is positively correlated with Candida parapsilosis (r = 0.61, p = 0.01) (Fig. 4B). Actinotignum is positively correlated with Aspergillus versicolor (r = 0.68, p = 0.006) and Basidiomycota unclassified (r = 0.70, p = 0.003) (Fig. 4B).

Fig. 4.

Correlation analysis reveals distinct microbial communities. CorrPlot analysis comparing (A) the twenty most abundant bacterial genera to each other and (B) the twenty most abundant bacterial genera to the twenty most abundant fungal species in the microbiome.

Identification of predicted metabolic pathways

Additionally, we looked at the functionality of the taxa, using PICRUSt2, within the sampling methods and how they differed from each other (Fig. 5A). The top three most prominent pathways were neuraminidase (sialidase), Co/Zn/Cd cation transporter, cation efflux family, and ATP/ADP translocase (Fig. 5A). The most critical of these for our research is sialidase, given its pivotal role in the transmission of HIV. A ratio-paired t-test was employed to compare the predicted sialidase genes in each sample. The tissue samples (GeoMean: 1929.4) were significantly enriched with predicted sialidase genes compared to both the CVL (GeoMean: 199.8; p = 0.0393) and swab samples (GeoMean: 78.6; pCVL = 0.0102) (Fig. 5B).

Fig. 5.

The vaginal epithelium is enriched for bacteria with metabolic functions associated with increased HIV transmission risk. (A) PICRUSt2 analysis between sampling methods: swab (swb), CVL, and vaginal tissue (tis). (B) Ratio paired t tests to compare the sialidase genes predicted by PICRUSt2.

Identification of Prevotella species abundances

Because sialidase was notably elevated to such an extent, we looked further into the abundances of different Prevotella species, as these can produce sialidases (Briselden et al., 1992). More specifically, we looked at Prevotella_9_unclassified, Alloprevotella_unclassified, Prevotella_sp_DJF_B116, Prevotellaceae_UCG-001_unclassified, Prevotellaceae_UCG-004_unclassified, and Prevotellaceae_NK3B32_group_unclassified (Fig. 6A). These were consistently more elevated in tissue samples across different RMs, depending on the species, compared to CVL or swab samples (Fig. 6A).

Fig. 6.

Prevotella spp. are enriched at the vaginal epithelium. (A) Comparisons of multiple Prevotella species enriched in the vaginal tissue sample and (B) the results of a spectrophotometry-based glycosidase assay.

Investigation of predicted sialidase glycan interactions

Additionally, to better understand whether glycosidase activity remains localized near the epithelium, we performed a glycosidase assay to detect β-Glucosidase, β-Galactosidase, α-Glucosidase, Sialidase, α-Fucosidase, and α-Galactosidase in CVL fluids. We found that β-Glucosidase, β-Galactosidase, and Sialidase exhibited similar enzyme activities with relative fluorescent units (RFU) of 18.37, 13.66, and 13.96, respectively (Fig. 6B). In contrast, α-Glucosidase and α-Fucosidase displayed higher enzyme activities at 138.56 and 84.26 RFU, respectively. Finally, α-Galactosidase had the lowest enzyme activity among all the glycosidases, measuring just 3.33 RFU (Fig. 6B). These data suggest that sialidase is not released from the epithelial niche into the CVL. Therefore, measuring sialidase activity only in CVL may incompletely measure how BV impacts HIV transmission.

Examination of the theorized model of vaginal microbiome

Based on our data and the taxa identified across different sampling methods, we developed the following understanding of the vaginal microbiome. Bacteria such as Escherichia-Shigella, Porphyromonas, and Negativicoccus and fungi like Asomycota unclassified and Basidiomycota unclassified were abundant in all levels of the vaginal microbiome (Fig. 7). At the mucosal layer, bacteria such as Sneathia and Streptococcus and fungi like Aspergillus versicolor, Malassezia restricta, and Alternaria unclassified were most prevalent, evident by their abundance in the CVL samples (Fig. 7). In the stratum corneum layer, bacteria such as Sneathia and Fusobacterium and fungi like Kazachstania telluris, Malassezia restricta, and Pseudophithomyces unclassified were most common, as observed in the swab samples (Fig. 7). Finally, at the stratified squamous layer, bacteria such as Helicobacter and Fusobacterium and fungi like Candida parapsilosis, Aspergillus penicilloides, and Kazachstania telluris were most prevalent, as indicated by their abundance in the tissue samples (Fig. 7).

Fig. 7.

Microbial species are dispersed across diverse niches within the vagina. Relative enrichment of each species at each niche based on trends in our data.

Discussion

In our current study, we sought to determine whether there were spatial differences in vaginal microbiome compositions that may modulate HIV/SIV transmission by utilizing a rigorous sampling protocol gathering data from CVL and swab samples, which are more representative of sampling that can be obtained clinically (or experimentally with minimal distress to animals), and augmenting this with vaginal epithelial sampling. We found that there were significant differences in vaginal microbiome abundance and diversity of taxa between tissue samples versus the CVL and swab specimens. First, there were prominent differences between the relative taxa abundances of tissue samples compared to the taxa abundances in CVL and swab samples. For example, while Sneathia, Escherichia-Shigella, and Porphyromonas abundances were very similar between CVL and swab samples, they were all noticeably less abundant in tissue samples. The opposite was observed for Helicobacter, where it was noticeably higher in tissue samples than either CVL or swab samples. Another significant observation was that certain species, such as Prevotella_9, Alloprevotella, Prevotella_sp_DJF_B116, Prevotellaceae_UCG-001_unclassified, Prevotellaceae_UCG-004_unclassified, and Prevotellaceae_NK3B31_unclassified, were significantly more abundant in tissue samples than they were in CVL or swab samples. Few major differences were found in fungal taxa, although Ascomycota unclassified was slightly lower in tissue samples than it was in the other collection methods. When the bacteria/bacteria correlations and the bacteria/fungi correlations were determined, we observed that there were far more significant bacteria/bacteria than bacteria/fungi, suggesting a greater inter-relationship between bacteria than between bacteria and fungi. PICRUSt2 revealed many pathways that were enriched in tissue samples, with the greatest difference in sialidase expression. These analyses reveal that CVL and swab sampling may not be sufficient to characterize the vaginal microbiome. Our findings reveal distinct microbial profiles across different layers of the vaginal epithelium. Notably, Sneathia, a bacterium linked to bacterial vaginosis (BV) (Theis et al., 2021), exhibited lower abundance in tissue samples compared to CVL and swab samples. Given that Prevotella and Sneathia are both typically higher in BV, the data presented herein suggest a differential spatial niche of each taxon, with Prevotella abundance higher closer to the epithelium than Sneathia.

One concern raised by our study was the potential degradation of protective host vaginal mucins by glycosidase-producing bacteria. Although 16S rRNA sequencing has limited specificity regarding species and strain identification, it has previously been shown that Prevotella generates sialidases, as in a study, Prevotella and Bacteroides species accounted for increased sialic activity in 26 of 27 (96 %) women (Briselden et al., 1992). Similiar to this observation, another study suggested that Prevotella species produced varying amounts of sialidases that affected the expression of mucins (Ilhan et al., 2020). Gardnerella spp. are another producer of sialidases in the cervicovaginal environment, and Gardnerella spp. have been proposed to create scaffolding on vaginal mucosa for the attachment of other bacterial species, like Prevotella and Atopobium, leading to biofilm formation (Hardy et al., 2015; Swidsinski et al., 2008). Vaginal biofilms are significant because it has been considered a hallmark of BV and are particularly troublesome as they hinder antibiotic action, leading to persistence of BV-associated bacteria after treatment (Swidsinski et al., 2008). This effect of sialidases on mucins was also seen in another study where sialidases of mucin oligosaccharide-degrading strains cleaved mucin glycoproteins that had sialic acids with a ratio of at least 2:1 of moles of O-acetyl ester per mol of sialic acid, slower than sialic acids with a ratio of 1:1 or 0:1 (Corfield et al., 1992). Whether increased levels of these sialidase-producing enzymes may degrade vaginal barriers enough to increase pathogen transmission requires further investigation. It has previously been shown that during HIV infection, sialylated glycoproteins are reduced and that sialidase levels are elevated (Giron et al., 2024). Sialylated glycoproteins activate anti-inflammatory immune responses by inhibiting sialic-acid binding proteins known as Siglecs (Duan and Paulson, 2020). Since sialidases modulate these Siglecs, it was hypothesized that inhibiting sialidase with sialidase inhibitors such as oseltamivir and DANA would decrease HIV-mediated inflammation in vivo. This was tested using two cohorts of bone-marrow-liver-thymus humanized mice, which were either infected or not infected with HIV-1_SUMA (Singh et al., 2025). Sialidase inhibition significantly reduced markers of HIV-induced inflammation and immune activation, such as pro-inflammatory interferon responses (Singh et al., 2025). This further reduced HIV-induced inflammation and immune activation was correlated with a decrease in plasma viral loads (Singh et al., 2025). Altogether, the approach to normalize glycosylation patterns that was implemented in this study signifies a novel approach in preventing HIV-induced inflammation during HIV infection.

In addition, it is known that Pneumococcus commonly colonizes the nasopharynx, where a subset of the infections causes severe disease, such as CNS-invading disease like meningitis, via tissue invasion. Pneumococcus infection can also disrupt the blood-brain barrier through mechanisms similar to those observed during HIV infection (Rosenberg, 2004). This piggybacking of HIV and Pneumococcus into the CNS is currently accepted. Since sialidases are a family of exoglycosidases that catalyze the cleaving of sialic acid residues that are bonded to sialoglycoconjugates, like sialoglycoproteins, sialoglycolipids, and simple sialo-oligisaccharides, such as sialyl lactose (Rosenberg, 2004). Sialic acid is the predominant carbohydrate-derived molecule terminally attached to the cell surface of sialoglycolipids and sialoglycoproteins. These compounds generate a negatively charged sialic acid barrier on the cellular plasma membrane, as well as on the envelope of viruses like HIV, which acquire their envelope from the host cell during replication (Rosenberg, 2004).

Lastly, sialidase activity in the CVL is associated with microscopic findings of BV (Ferreira et al., 2021). Microscopic BV is often diagnosed using Nugent criteria and is characterized by the replacement of Lactobacillus sp. by an array of other bacteria. The presence of BV has been associated with poor pregnancy outcomes and increased risk for several sexually transmitted infections (STIs), including HIV. The vaginal microbiota of all reproductive-aged women fit into five bacterial community state types (CSTs), with these being: CST I, CST II, CST III, CST IV, and CST V. Of these four types, CST I, CST II, CST III, and CST V are all dominated with certain Lactobacillus spp. Most of these microscopic BV cases fit into vaginal community-state type IV (CST IV), which is known as “molecular-BV”. BV-associated bacterial species, such as Gardnerella spp., may act as sources of CVL sialidases (Ferreira et al., 2021). Fluorometric assays were performed using 2-(4-methylumbelliferyl)- α-d-N-acetylneuraminic acid (MUAN) for measuring sialidase activity in CVL samples on 140 participants (Ferreira et al., 2021). From these assays, it was found that cervical infections by C. trachomatis and N. gonorrhoeae were detected in 8 (5.7 %) and 2 (7.1 %) participants. Of these 140 participants, the 44 samples with molecular BV (CST IV) were tested using LEfSe (Ferreira et al., 2021). Sialidase activity in molecular-BV is also associated with changes in bacterial components of the local microbiome (Ferreira et al., 2021).

Our study has several limitations. Foremost, as a pilot investigation, it included only a small number of animals. A unique aspect of our work was the inclusion of vaginal rugae samples in addition to swabs and CVL. However, the RMs in our study were already infected with SHIV, since vaginal tissue sampling performed in this study can only be performed at necropsy. Larger cohorts and direct comparisons between infected and uninfected RMs would significantly improve interpretations and the broader relevance of these findings to HIV transmission research. Longitudinal sampling would also facilitate better understanding and interpretation of our findings, as vaginal microbiomes are living communities reacting to diverse stimuli, such as host hormone levels and infection status. Additionally, only a single vaginal epithelial sample time point was available for bulk characterization. Our analysis relied solely on 16S rRNA sequencing, which offers limited taxonomic resolution than other methods, and PICRUSt2 can only infer potential metabolic functions based on this data set. Emerging spatial sequencing approaches that integrate shotgun metagenomic sequencing rather than relying exclusively on 16S rRNA sequencing may be more appropriate to ameliorate several of these limitations in future studies.

In conclusion, our findings show that tissue sampling reveals important nuances in the vaginal microbiome that are often missed by CVL and swab-based methods. The observed differences in bacterial and fungal taxa, particularly the variation in sialidase activity associated with Prevotella, highlight the value of deeper, tissue-level profiling. Notably, Prevotella species can be enriched at other mucosal sites where they may also play a role in modulating barrier function related to both transmission risk and pathogenesis, possibly warranting similar studies (Johnson and Byrareddy, 2022; SD Johnson et al., 2022; SD Johnson et al., 2022). The approach in our study enhances the understanding of microbial dynamics across different epithelial layers and carries important implications for conditions such as bacterial vaginosis and the development of effective HIV vaccines. The elevated sialidase activity, which may facilitate pathogen transmission, underscores the pivotal role of microbiome composition in maintaining mucosal barrier function. These insights establish a foundation for future studies exploring how microbial interactions influence disease progression and may guide the development of strategies to prevent pathogen transmission at mucosal surfaces.

Author contribution

SB designed the study. AA led the in vivo study and specimen collection. SJ processed samples. SJ and RM performed data analysis and wrote the manuscript. SJ, RM, and PD developed and performed glycosidase assays. SB acquired funding, oversaw the study, and edited the manuscript. All authors contributed to manuscript development and reviewed and approved the final version.

Funding

This study was supported in part by NIH R01AI113883, R21AI11415 grants to SNB.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Siddappa Byrareddy reports financial support was provided by National Institutes of Health. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.