Oral microbiome diversity, community- and taxon-level differences by oral human papillomavirus (HPV) and race/ethnicity Authors: Morgan C Byrd, Shelly Han, Oyomoare Osazuwa-Peters, Deesha Bhaumik, Darien Weatherspoon, Nicolas F Schlecht, Nosayaba Osazuwa-Peters Published: 2026 ABSTRACT AbstractIntroductionOral microbiome differences by race/ethnicity may contribute to differences in oral human papillomavirus (HPV) infection, the precursor to HPV-associated oropharyngeal cancer. Oral HPV is most prevalent among non-Hispanic Black individuals, yet HPV-attributable cancers are more common in non-Hispanic White individuals.MethodsUsing the 2009–2012 NHANES data (n = 5,534), our study examined oral microbiome diversity, community-, and taxon-level differences by any oral HPV type and race/ethnicity using oral rinse samples.ResultsSignificantly greater alpha diversity was seen among HPV-positive Non-Hispanic Black individuals [Shannon/Inverse Simpson’s Indices (p < 0.001)]. Beta diversity also varied significantly by race/ethnicity and oral HPV status (PERMANOVA, p < 0.001). Microbiome topic modeling identified two microbial communities (topics 13 and 36) depleted in HPV-positive individuals, characterized by reduced Streptococcus, Actinobacillus, and Neisseria (Beta > 0.4), but this pattern was seen only among Non-Hispanic Black individuals. Taxon-level differential abundance analysis identified Sneathia and Odoribacter as enriched in HPV-positive participants (p < 0.01).ConclusionThese findings suggest a race/ethnicity-linked interplay between oral rinse microbiome samples and oral HPV. Microbial differences by race/ethnicity and oral HPV status may reflect underlying social, environmental, or behavioral factors and highlight the need for longitudinal studies to clarify causal pathways and inform future targeted interventions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13027-026-00757-4.Keywords: Human papillomavirus viruses, Microbiota, Head and neck neoplasms, Ethnicity, Racial groups, Molecular Epidemiology Introduction Human papillomavirus (HPV)-associated oropharyngeal cancer has continued to increase in incidence, and it is now the leading HPV-associated cancer in the U.S [1, 2]. Incidence of HPV-associated oropharyngeal cancer increased by 225% from 1988 to 2004, rising from 0.8 to 2.6 per 100,000 individuals [3]. Persistent oral HPV infection, the key precursor to HPV-associated oropharyngeal cancer, exhibits a notable variation across racial/ethnic groups in the United States; prevalence is highest among non-Hispanic Black adults (9.7%) followed by non-Hispanic Whites (7.3%), Hispanics (7.0%), and non-Hispanic Asians (2.9%) [4]. Despite this, the proportion of oropharyngeal cancers attributable to HPV is disproportionately greater among individuals who are Non-Hispanic White (70.2%), compared to individuals who are Hispanic (61.3%), non-Hispanic Asian (55.8%), and Non-Hispanic Black (46.3%) [5]. Additionally in 2017, Non-Hispanic White individuals had the highest incidence of HPV-associated oropharyngeal cancer (8.6 per 100,000 individuals), a rising trend, while Non-Hispanic Black individuals had a decrease in incidence (5.1 per 100,000 individuals) and Hispanic individuals remained steady (4.0 per 100,000 individuals) [6]. There is emerging evidence that the oral microbiome may be linked to oral HPV infection, but findings from existing cross-sectional studies remain inconsistent. Feng et al. found that participants with oral HPV had lower bacterial diversity [7]. Specifically, HPV-positive participants exhibited reduced microbial diversity, with fewer distinct bacterial types (observed amplicon sequence variants [ASVs]) and lower overall diversity [7]. In contrast, a smaller cohort found no association with alpha diversity, though differences in microbial composition were observed [8]. These inconsistencies may reflect methodological differences or unexamined racial heterogeneity. However, it is unclear whether oral microbiome diversity and composition vary by HPV status between and within racial/ethnic groups, a critical gap given the known racial differences in both HPV prevalence and HPV-associated oropharyngeal cancer. Moreover, few studies have gone beyond diversity metrics to assess which specific bacterial taxa differ in abundance by HPV status across racial groups, limiting our understanding of the microbiome’s role in mediating these differences. In this study, we use nationally representative cross-sectional data to describe the differences in oral rinse microbiome samples using alpha and beta diversity by combined race/ethnicity and HPV status. Race/ethnicity is treated as a proxy for differences in social, behavioral, and structural exposures, rather than as a biological determinant. Additionally, we identify differentially abundant taxa that may reflect microbiome patterning associated with these exposure differences and contribute to observed racial disparities in HPV-associated head and neck cancer epidemiology, despite higher oral HPV prevalence in minority populations. Methods Study population National Health and Nutrition Examination Study (NHANES) is a continuous, cross-sectional survey beginning in 1999 designed to monitor the health and risk factors for diseases among the civilian noninstitutionalized United States population [9]. NHANES is conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention. NHANES collects data every two years using a stratified, multistage probability sample to reflect a representative sample of the population. Participants complete home interviews using a Blaise format computer-assisted personal interview (CAPI) system. After completion of the home interview, participants are invited to participate in a health examination conducted in mobile examination centers (MEC). At the MEC, participants undergo additional interviews, have a physical examination, and then provide biological samples. Informed consent was obtained twice, once prior to the home interview and again prior to the health examination. Oral microbiome measures Oral microbiome measures were obtained from the oral rinse samples collected in the MEC [10]. DNA was extracted from the oral rinse samples and sequencing of 16S rRNA V4 region was performed on Illumina HiSeq 2500 (2 × 125 bp). Each plate of samples included oral rinse samples, blank samples, and two types of quality controls (oral artificial community and artificial gut samples). Reads were processed to amplicon sequence variants (ASVs) using DADA2 and classified to the genus level using the SILVA v123 database [11]. All taxa that were missing a phylum level identification were removed (2750 ASVs). Four alpha diversity measures [10], reflecting the richness and/or evenness of a single participant’s sample, were calculated: observed number of ASVs, Faith’s Phylogenetic Diversity [12], Shannon-Wiener index [13], and the Simpson Index [14]. These four alpha diversity metrics were calculated based on rarefaction values from 2,000 to 10,000. For this analysis, we used 2,000 rarefaction values, as the lowest rarefaction depth ensures that the sample with the least sequencing depth determines the number of reads, and averaged ten subsamples. Three beta diversity measures [10], reflecting the dissimilarity between pairwise samples, were calculated: unweighted UniFrac, weighted UniFrac [15], and Bray-Curtis dissimilarity [16]. Race/ethnicity & oral HPV status Race/ethnicity was self-reported by participants during the household interview to collect demographic information. Participants were asked to select the category that best describes their race and whether they are of Hispanic/Latino origin. NHANES then groups participants into a combined category that include individuals who are: Non-Hispanic White, Non-Hispanic Black, Mexican American, Other Hispanic, and Other/Multi-Racial [17]. Oral HPV status was determined using the oral rinse collected during MEC/laboratory analysis. The presence of 38 HPV types, including oncogenic types and non-oncogenic types, was detected by the Roche Linear Array HPV Genotyping Test using previously described methods [18]. Samples were classified as “positive”, “negative”, “not evaluated”, or “missing”. Those that were not tested or missing were excluded from the analysis. When sufficiently powered, we completed analyses for any oral HPV and high-risk oral HPV (genotypes 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68) [19]. For diversity analyses we created a combined race and oral HPV status variable: Non-Hispanic White/HPV Positive, Non-Hispanic White/HPV Negative, Non-Hispanic Black/HPV Positive, Non-Hispanic Black/HPV Negative, Other Hispanic/HPV Positive, Other Hispanic/HPV Negative, Mexican American/HPV Positive, Mexican American/HPV Negative, Other and Multi-Racial/HPV Positive, and Other and Multi-Racial/HPV Negative. Other covariates As oral microbiome data is only publicly available for NHANES cycles 2009–2010 and 2011–2012, we used these cycles for this analysis. Variables that are associated with oral microbiome and HPV were included in the analysis. From demographics data, this included age, educational attainment, and poverty income ratio. From MEC/laboratory components we included human immunodeficiency virus (HIV) status. From the questionnaire data, we included lifetime smoking. We used the provided 2-year MEC weights to compute a 4-year MEC weight for combining two NHANES cycles [20]. Diversity analysis Alpha diversity measures were visualized using boxplots of the combined race/ethnicity and oral HPV status variable accounting for survey weights. Kruskal-Wallis tests were performed to assess overall differences in alpha diversity across combinations of race/ethnicity and oral HPV status and stratified Kruskal-Wallis tests to evaluate differences by oral HPV status within each racial/ethnic group and by race/ethnicity within each HPV status group. Beta diversity measures the dissimilarity between participants. We used a weighted principal coordinate analysis (PCoA) to visualize how oral rinse microbiome samples differ by combined race/ethnicity and oral HPV status using the “wcmdscale” function from the “vegan” package (v 2.6.10). Survey weights were normalized to sum to 1 to account for complex sampling design while maintaining numerical stability. Weighted Permutation Multivariate Analysis of Variance (PERMANOVA) was performed using “fast.adonis” function and package (v 0.0.1.0) with 1,000 permutations to test for significant differences in the community microbiome composition by combined race/ethnicity and oral HPV status and to identify the proportion of variance explained by this variable (R2) [21]. Latent Dirichlet Allocation (LDA) topic modeling Latent Dirichlet Allocation (LDA) topic modeling was used to identify community-level patterns in the microbiome, allowing us to infer ecological structure from genus-level raw counts [22]. To determine the optimal number of latent topics, we used the “FindTopicNumber” function from the “ldatuning” package (v 1.0.2) applying CaoJuan 2009 and Arun2010 metrics with Variational Expectation Maximization (VEM) [23, 24]. The number of topics was selected based on the lowest values of both CaoJuan2009 and Arun2010. We fit the final LDA model using the LDA function from the “topicmodels” package (v 0.2.17) with k = 36. We extracted beta and gamma probability matrices using the “tidytext” package (v 0.4.2). To test for differences in topics by oral HPV status and race/ethnicity, we performed a differential abundance (DA) analysis using the “linda” (linear models for difference abundance analysis, LinDA) function from the “MicrobiomeStat” package (v 1.2). We enabled winsorization (is.winsor = TRUE) to reduce the influence of extreme values, stabilizing the variance and improving model robustness. LinDA adjusts for multiple hypothesis testing using the Benjamini Hochberg correction with adjusted p-values (q-values) less than 0.05 considered statistically significant [25]. In addition to characterizing the overall topic structure, we also evaluated the most influential genera for each topic and tested their associations with HPV status. Differential abundance (DA) analysis using ADAPT To complement the community-level analysis, we used ADAPT to identify individual taxon-level differences. ADAPT is an R package for DA analysis that accounts for compositional nature and the excessive number of zeros in metagenomic count tables, allowing it to achieve a good balance between false discovery rate and detection power [26]. It identifies invariant taxa for normalization and then uses Tobit models to compare normalized taxa abundances between different conditions [27]. We adjusted the p-values for multiple hypothesis testing using Benjamini-Hochberg correction [25]. We used ADAPT to test for genus-level differences by oral HPV status using relative abundance data adjusting for covariates. Together LDA and ADAPT allow us to evaluate both macroscopic (community-level) and microscopic (individual-taxon level) associations with oral HPV and race/ethnicity. P-Values < 0.05 were considered statistically significant and all analyses were conducted using RStudio Version 4.4.2. This data is publicly available and deidentified and was thus exempt from review by the Institutional Review Board. This report is compliant with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist. Results Study population characteristics A total of 20,293 participants were included in the NHANES 2009–2012 survey cycles. After excluding participants under 20 years of age and those with missing oral microbiome data, oral HPV status, and covariate information, the final analytical cohort consisted of 5,534 participants (Supplemental Fig. 1). When accounting for complex survey weights, 50.5% of the population were female and the majority had some college or college degree (64.0%) (Table 1). Most participants were oral HPV-negative (93.0%) and reported smoking fewer than 100 cigarettes in their lifetime (57.1%). The racial and ethnic distribution was as follows: Non-Hispanic White (65.0%), Non-Hispanic Black (11.6%), Mexican American (9.7%), Other/Multi-Racial (7.6%), and Other Hispanic (6.1%). Racial and ethnic distributions by any oral HPV, high-risk HPV, and HPV genotypes 16 and 18 are detailed in Supplemental Table 1. Table 1Characteristics of adults aged 20 years and older included in analysis from National Health and Nutrition Examination Survey (NHANES) cycles 2009–2012Unweighted NWeighted %Weighted SE (%) Race/ethnicity Mexican American8879.71.7 Other Hispanic5296.11.1 Non-Hispanic White226665.02.8 Non-Hispanic Black117811.61.3 Other & Multi-Racial6747.60.8 Sex Male271149.50.7 Female282350.50.7 Age category < 2000.00 20–39285949.41.5 40–59267550.61.5 60+00.00 Smoking <100 cigarettes in lifetime320757.11.5 >=100 cigarettes in lifetime232742.91.5 Educational attainment <9th grade4084.40.4 9-11th grade79611.30.9 HS diploma117520.31.1 Some college or AA degree173132.11.0 College graduate142431.91.9 Oral HPV status Negative511493.00.4 Positive4207.00.4 High-risk oral HPV status Negative532395.90.3 Positive2114.10.3 High-risk oral HPV genotypes 16/18 Negative546998.70.2 Positive651.30.2 HIV status Negative550499.60.1 Positive300.40.1 Characteristics of adults aged 20 years and older included in analysis from National Health and Nutrition Examination Survey (NHANES) cycles 2009–2012 Alpha diversity We visualized Observed Number of ASVs (Fig. 1A), Shannon Index (Fig. 1B), Inverse Simpson’s Index (Fig. 1C), and Faith’s Phylogenetic Diversity (Fig. 1D) alpha diversity measures by combined race/ethnicity and oral HPV status. Across all combinations of racial/ethnic groups and oral HPV status, each alpha diversity measure differed significantly, indicating that at least one group comparison showed a statistically significantly difference in microbial diversity. When comparing oral HPV status within each race/ethnicity category, a significant difference in alpha diversity was observed only in Non-Hispanic Black individuals, but not among other racial/ethnic groups, between people with and without any oral HPV infection [Shannon Index and Inverse Simpson’s Index (p < 0.001)]. When comparing races/ethnicities within each oral HPV status group, a significant difference in alpha diversity was observed for all four metrics (p < 0.01). Similar patterns were also observed for high-risk oral HPV (Supplemental Fig. 2), however, due to small sample sizes in some subgroups, these results should be interpreted with caution. Alpha diversity measures of the oral rinse microbiome samples: (A) Observed Number of ASVs, (B) Shannon Index, (C) Inverse Simpson’s Index, (D) Faith’s Phylogenetic Diversity, by combined race and HPV status among 5,534 adults 20 years and older from National Health and Nutrition Examination Survey (NHANES) cycles 2009–2012. Boxes are colored by race/ethnicity. Kruskal-Wallis tests were used to assess differences in alpha diversity. When comparing all races/ethnicities within each oral HPV status group, all four metrics were significant (p < 0.01). When comparing oral HPV status within each racial/ethnic group, a significant difference was observed between HPV positive and HPV negative in Non-Hispanic Black individuals (but not for other racial groups) for Shannon Index and Inverse Simpson’s Index (p < 0.001) noted with the bracket and *. Sample sizes are detailed in Supplemental Table 1 Beta diversity Weighted PCoAs of Bray-Curtis dissimilarity, unweighted UniFrac, and weighted UniFrac distances showed shifts in the microbiome community by combined race/ethnicity and any oral HPV status (Supplemental Figs. 3A-3C) [1] and high-risk oral HPV status (Supplemental Figs. 4A-4C). Using univariable PERMANOVA, oral rinse microbiome beta diversity significantly differed by combined race/ethnicity and oral HPV status for all three metrics (p < 0.001). The proportion of variance explained by this variable for any oral HPV was small (Bray-Curtis R2 0.97%; unweighted UniFrac R2 1.12%; weighted UniFrac R2 1.42%), and similarly small for high-risk oral HPV (Bray-Curtis R2 0.91%; unweighted UniFrac R2 1.13%; weighted UniFrac R2 1.40%), indicating subtle shifts in microbial communities across groups. Samples sizes were insufficient to adjust for covariates in the PERMANOVA. Topic modeling and LDA analysis Our LDA model revealed 36 topics based on CaoJuan2009 and Arun2010 metrics (Supplemental Fig. 3). In this context, topics represent latent microbial community structures defined by co-occurring taxa rather than discrete biological pathways. We identified significant/differentially abundant topics by oral HPV status and race/ethnicity, independently. For HPV status, two topics, Topics 13 and 36, were identified to be differentially abundant using LinDA, and both were significantly depleted in the oral rinse microbiome samples among those who were HPV-positive (Fig. 2A). These topics were prioritized for further investigation because they were the only significant topics associated with HPV-positivity, the primary exposure of interest. Latent Dirichlet Allocation (LDA) topic modeling identifying significant topics among (A) HPV-positive participants compared to the reference of HPV-negative participants and racial/ethnic groups compared to the reference of Non-Hispanic White: (B) Non-Hispanic Black, (C) Mexican American, (D) Other Hispanic, (E) Other/Multi-Racial. The x-axis represents the log2 fold change, with positive values indicating topics that are enriched (increasing in abundance) and negative values indicating topics that are depleted (decreasing in abundance). Black bars indicate topics that are significant at adjusted p < 0.05 (differentially abundant) using LinDA, linear models for differential abundance analysis, and gray are those that are non-significant In racial/ethnic comparisons, individuals who are Non-Hispanic Black (Fig. 2B), Mexican American (Fig. 2C), Other Hispanic (Fig. 2D), and Other/Multi-Racial (Fig. 2E), exhibited different combinations of enriched and depleted topics, compared to non-Hispanic White individuals. Topic 13 was significantly depleted in the oral rinse microbiome samples among Non-Hispanic Black individuals, similar to the oral HPV-positive topic analysis, but enriched in all other racial/ethnic groups. Conversely, Topic 36 was not significantly depleted among any racial/ethnic group, but was enriched among Mexican American individuals, Other Hispanic individuals, and Other/Multi-Racial individuals. Based on these differentially abundant topics, we further explored the bacterial taxa most strongly associated with each topic to provide biological context. We conducted an additional LinDA to identify differentially abundant genera by oral HPV status for each topic, adjusting for age, sex, race/ethnicity, educational attainment, poverty income ratio, and HIV status. For Topic 13 and Topic 36, we visualized the 15 most representative genera based on topic-genera probabilities (Fig. 3A/B), determined from the extracted beta values and highlighted which of these were differentially abundant by HPV status using the log2 fold-change estimates (Fig. 3C). [A-B]. Latent Dirichlet Allocation (LDA) topic modeling of the top 15 most contributing genera by topic: (A) Topic 13 and (B) Topic 36. The x-axis represents the extracted beta values, with larger values indicating genera that contributed more to the topic. (C) Panel C shows linear models for difference abundance analysis (LinDA) of the top 15 genera most contributing to Topic 13 and Topic 36. Linear models identified genera that are differentially abundant by HPV status after adjusting for age, sex, race/ethnicity, educational attainment, poverty income ratio, and HIV status. The x-axis represents the log2 fold change, with positive values indicating enriched (increasing in abundance) genera in the oral rinse microbiome samples of HPV-positive participants and negative values indicating depleted (decreasing in abundance) genera in the oral rinse microbiome samples of HPV-positive participants. Taxa names are formatted as combined Family_Genus on the y-axis and ordered by magnitude of effect. Some genera appear in both topics; in these cases, the LinDA estimates are identical as LinDA is performed across the samples grouped by HPV status, independent of their topic and may contribute to many topics. Therefore, these are collapsed into a single panel to avoid many repeated genera Streptococcus and Actinobacillus were the most prominent contributors to Topic 13 while genera Neisseria dominated Topic 36. All three of these genera were found to be depleted in differential abundance genera analysis for each topic. Both topics shared the same three enriched genera, Actinomyces, Prevotella_7, and Veillonella. Most of the depleted genera were consistent across both topics though there were a few distinct differences. Topic 13 uniquely included Alloprevotella, Prevotella_2, Simonsiella, and Actinobacillus (Fig. 3A) whereas Topic 36 uniquely included Lachnoanaerobaculum, Oribacterium, Abiotrophia, and Mannheimia (Fig. 3B). ADAPT DA analysis We used ADAPT to identify genera that were differentially abundant by HPV status. In the initial model adjusting for age, sex, and race/ethnicity (Fig. 4A), ten genera were identified as differentially abundant. After additionally adjusting for poverty income ratio and HIV status (Fig. 4B), five of these genera remained differentially abundant: two were enriched and three were depleted. Enriched genera included Odoribacter (Log10 fold change: 0.87, adjusted p = 0.048) and Sneathia (Log10 fold change: 0.97, adjusted p = 0.048). Depleted genera included Neisseria (Log10 fold change: -0.22, adjusted p = 0.039), Actinobacillus (Log10 fold change: -0.41, adjusted p = 0.039), and Gemella (Log10 fold change: -0.12, adjusted p = 0.039) (Supplemental Table 2). Further adjustment for educational attainment or lifetime smoking attenuated all associations, with no genera remaining differentially abundant. Differentially abundant genera identified using ADAPT, comparing the oral rinse microbiome samples of HPV-negative (reference group) to HPV-positive participants. Models adjust for (A) age, sex, and race/ethnicity; (B) age, sex, race/ethnicity, poverty income ratio (PIR), and HIV status. Statistical significance is based on adjusted p-values < 0.05, corrected for multiple hypothesis testing using the Benjamini-Hochberg method. The y-axis displays -log10(p-value) for visual interpretability, with non-significant species shown in gray. The x-axis shows log10 fold change, where positive values indicate genera enriched in the oral rinse microbiome samples of HPV-positive participants (colored in navy blue) and negative values indicate genera depleted in the oral rinse microbiome samples of HPV-positive participants (light blue). Taxa names are formatted as combined Family_Genus Discussion In this analysis of 5,534 participants from NHANES 2009–2012, we identified significant differences in oral rinse microbial diversity and composition by both race/ethnicity and oral HPV status at the taxon- and community-levels. Across all racial/ethnic groups, all four alpha diversity metrics were significantly different by HPV status. When stratified by race/ethnicity, only Non-Hispanic Black individuals showed significant differences in two alpha diversity metrics between HPV-positive and HPV-negative participants. Microbial community differences, assessed through PERMANOVA, showed significant variation in beta diversity by combined race/ethnicity and oral HPV status (any oral HPV and high-risk oral HPV); however, the proportion of variance explained was very small (R2 0.91–1.42%) Visualizations revealed no obvious separations in microbial community structure; however, Non-Hispanic White participants, particularly those who were HPV-positive, exhibited larger 95% confidence ellipses across all three beta diversity metrics compared to other racial/ethnic groups, suggesting greater within-group variability. Additionally, topic modeling revealed distinct bacterial community structures associated with both race/ethnicity and oral HPV status. These patterns were supported by both taxon- and community-level analyses, highlighting specific genera associated with oral HPV status. We identified two microbial topic communities, Topic 13 and Topic 36, that were significantly depleted among HPV-positive participants. Topic 13 was primarily characterized by higher probabilities of Streptococcus and Actinobacillus, both of which were found to be depleted in the genus-level topic modeling. Veillonella was enriched and Gemella was depleted, although both contributed less strongly to this topic. These results are broadly consistent with findings from Börnigen et al. who observed reduced abundance of Streptococcus anginosus among 121 patients with HPV-associated oral, oropharyngeal, and unknown primary squamous cell carcinomas, suggesting a potential genus-level pattern in HPV-related oral microbial shifts, although direct comparisons are limited by differences in study population and disease status [28]. While species-level resolution was not available in our data and we were unable to specifically assess high-risk HPV genotypes, our results suggest a similar directional trend at the genus-level. To our knowledge, no prior studies have directly examined Actinobacillus in relation to oral HPV status. However, recent evidence indicates that Actinobacillus species are integral components of early biofilm formation and participants in multispecies communities with Streptococcus species [29]. Disruption of these communities has been associated with inflammatory states and altered mucosal environments, suggesting that depletion of Actinobacillus may reflect broader ecological shifts that could influence susceptibility to viral colonization [29]. Together, these findings support a potential association for Actinobacillus and oral mucosal health. Consistent with our findings, Zhang et al. reported increased abundance of the family Veillonellaceae and decreased abundance of the Gemella in HPV-positive participants [8]. In contrast, studies of HPV-associated oropharyngeal cancers have reported increased abundance of Gemella compared to HPV-negative disease, suggesting that Gemella may thrive in dysbiotic environments associated with later-stage disease rather than early HPV infection [30]. Notably, Topic 13 was also depleted among Non-Hispanic Black individuals, suggesting a potentially shared microbial signature or shift associated with both oral HPV infection and race/ethnicity. While the mechanisms underlying these patterns remain unclear, they may reflect broader social, behavioral, or structural factors that influence oral microbial ecology. In addition to these sociobehavioral influences, host infectious disease history, oral immune response, and potential viral reservoirs may also contribute to inter-individual variability in oral HPV acquisition and clearance, as well as observed microbial patterns [29]. Although we could not directly assess these factors in NHANES, our findings may, in part, reflect underlying differences in immune response and prior infections that shape both oral microbiome composition and HPV persistence. To date, no prior studies have stratified oral rinse microbiome differences jointly by HPV and racial/ethnic group. Topic 36, predominantly composed of Neisseria, was similarly depleted within the genus-level topic modeling and ADAPT differential abundance analysis. While Neisseria gonorrhoeae has been associated with HPV-associated oropharyngeal cancer, the role of commensal Neisseria (non-gonorrhoeae) species in oral HPV infection remains poorly understood [31]. Topic 36 was enriched in all racial/ethnic minority groups except Non-Hispanic Black individuals, potentially indicating heterogeneity in Neisseria-associated microbial patterns across racial/ethnic groups and suggesting that these taxa may reflect context-specific microbial community structures rather than a uniform protective or pathogenic role. ADAPT differential abundance analysis initially identified enrichment of Sneathia and Odoribacter among HPV-positive participants; however, these associations were attenuated after covariate adjustment and should therefore be interpreted cautiously. Sneathia has been strongly associated with HPV-positive women and implicated in both infection and the development of precancerous lesions, particularly in the vaginal microbiome [32, 33]. In these studies, Sneathia species express cytopathogenic (Cpt) toxin and invasion factors that facilitate disruption of the mucosal barrier [34], thereby creating a more permissive environment for HPV acquisition and persistence [35]. These virulence traits, together with their association with bacterial vaginosis among Black women at risk for HPV-related cervical cancer, suggest a plausible mechanism linking Sneathia enrichment to HPV susceptibility in mucous membranes. While these associations derive primarily from genital tract studies [35], consistent with Escobar Marcillo et al., we report an enrichment of this genus in oral microbiome samples among HPV-positive participants [36]. In contrast, Odoribacter is generally considered a commensal genus and has been associated with anti-inflammatory properties and suppression of bacterial virulence pathways, particularly within the gut microbiome [37]. To date, Odoribacter has not been linked to HPV infection in human studies. However, enrichment of this genus has been observed in mouse models who exhibit higher HPV viral loads, suggesting a possible, albeit indirect, role in modulating susceptibility [38]. These animal-based findings should be viewed as hypothesis-generating rather than confirmatory in human populations. Our findings highlight key differences in oral rinse microbiome diversity and composition by combined race/ethnicity and oral HPV status. However, several limitations should be considered when interpreting these results. First, due to the cross-sectional design, we were unable to assess temporality; it remains unclear whether oral HPV infection drives microbial changes or whether oral microbiome composition influences susceptibility to HPV, raising the possibility of reverse causation. In addition, these 2009–2012 data reflect the oral microbiome at a single time point, which is dynamic and may have systemic implications, including oral-non-oral bidirectional effects on viral infections, and may not fully represent post-SARS-CoV-2 characterization [39, 40]. We were also unable to evaluate HPV persistence and could not complete all analyses in high-risk oncogenic genotypes due to insufficient power. Second, while 16S rRNA sequencing provides information on taxonomic composition and relative abundance, it does not capture functional potential without metagenomic data. While computational approaches to infer function exist, such analysis could not be conducted due to the lack of publicly available sequence data. Additionally, sequencing the V4 region of the 16S rRNA gene limits species-level resolution; because our data were restricted to genus-level resolution, we were unable to examine species-specific associations. Furthermore, oral rinse samples integrate microbes from multiple oral niches, treating the oral cavity as a single homogenous environment. Consequently, site-specific associations, such as those involving the tongue or tonsillar regions, cannot be resolved and may dilute biologically relevant signals related to oral HPV infection. While we adjusted for key covariates, including age, sex, race/ethnicity, educational attainment, poverty income ratio, and HIV status, residual confounding by measured and unmeasured factors remains possible. Several health behavior variables that are associated with HPV status and oral microbiome composition, including alcohol consumption, were not included due to substantial missingness. Finally, while our alpha and beta diversity analyses accounted for NHANES complex survey design using survey weights, topic modeling (LDA) and differential abundance analyses (ADAPT) did not. Accordingly, these findings should be interpreted as exploratory and hypothesis-generating rather than fully generalizable to the broader U.S. population. Despite these limitations, our study has several strengths. Using data from a large, nationally representative population, we evaluated multiple metrics of alpha and beta diversity and integrated both taxon- and community-level analytic approaches. To our knowledge, this is the first study to examine oral rinse microbiome differences by combined race/ethnicity and oral HPV status, addressing an important gap in understanding potential microbial contributors to racial/ethnic disparities in oral HPV prevalence. These findings provide a foundation for future longitudinal studies investigating the role of oral rinse samples in HPV acquisition, persistence, and related oral health outcomes. Conclusion In this study of over 5,000 oral rinse microbiome samples from a nationally representative U.S. population, we identified distinct differences in microbial diversity and composition by combined race/ethnicity and oral HPV status. These findings highlight the importance of considering both host and microbial factors in the context of HPV infection. The observed variation in microbial communities may reflect underlying social, environmental, or behavioral exposures that contribute to differences in oral HPV prevalence and related outcomes across racial/ethnic groups. Our results highlight the need for longitudinal and functional capability studies to clarify the directionality of these associations and to better characterize the role of oral rinse microbiome samples in HPV acquisition and persistence. Although most oral HPV infections are transient, persistent infection with oncogenic HPV types is a key risk factor for oropharyngeal cancer, and microbial patterns associated with oral HPV may represent an early context relevant to these pathways. Improved understanding of these microbial dynamics could inform microbiome-targeted prevention strategies, particularly among populations disproportionately affected by oral HPV. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements Not applicable. Author contributions MB: Conceptualization, Data Curation, Formal analysis, Methodology, Writing – original, Writing – review & editing; SH: Data Curation, Methodology, Writing – original, Writing – review & editing; OOP: Data Curation, Formal analysis, Methodology, Writing – original, Writing – review & editing; DB: Data Curation, Formal analysis, Methodology, Writing – original, Writing – review & editing; DW: Data Curation, Methodology, Writing – original, Writing – review & editing; NS: Data Curation, Methodology, Writing – original, Writing – review & editing; NOP : Conceptualization, Data Curation, Methodology, Writing – original, Writing – review & editing, Supervision; All authors gave their final approval and agree to be accountable for all aspects of the work. Funding The authors declare that no funds, grants, or other support were received for the preparation of this manuscript. Data availability All data is publicly available through NHANES. Declarations Ethics approval and consent to participate This data is publicly available and deidentified and was thus exempt from review by the Institutional Review Board. All NHANES participants complete informed consent prior to participation. Competing interests Dr. Nosa Osazuwa-Peters has served as a scientific advisor to Navigating Cancer, has received fees from Merck, and reports receiving grants from the National Institute of Health/National Institute of Dental and Craniofacial Research (K01 DE030916, R21 DE032531, and R01 DE032216) outside of the submitted work. Consent for publication Not applicable. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement All data is publicly available through NHANES.