Environmental fungi modulate the vaginal mycobiome and cervical disease progression in Hispanic women Authors: Filipa Godoy-Vitorino, Daniela Vargas-Robles, Benjamin Bolaños-Rosero, Natalia Pagán-Zayas, Andrea Cortés-Nazario, Kara Wiggin, Sarah Allard, Josefina Romaguera, Jack A. Gilbert Published: 2026 ABSTRACT ABSTRACTThe vaginal mycobiome, though a minor component of the cervicovaginal ecosystem, plays a crucial role in reproductive health and disease. However, its composition and interactions with bacterial communities remain poorly understood, particularly among Hispanic women, who experience disproportionately high rates of human papillomavirus (HPV) infection and cervical cancer. We characterized the vaginal mycobiota across reproductive stages and examined its associations with cervical disease, HPV status, and bacterial community state types (CSTs) in 86 Hispanic participants from Puerto Rico using ITS1 amplicon sequencing. Amplicon sequence variants were inferred with QIIME2/DADA2 and taxonomically classified using the UNITE database, with diversity and discriminant taxa analyses applied to explore clinical and microbial associations. We detected 173 fungal species hypotheses, dominated by Candida albicans, Agaricomycetes sp., Scopuloides dimorpha, and Hortaea werneckii. While fungal composition did not differ significantly by reproductive stage, non-pregnant individuals exhibited greater inter-individual variability. Alpha diversity was reduced in high-grade squamous intraepithelial lesions compared with low-grade or normal cytology, and Candida parapsilosis prevalence was elevated in low-grade lesions. CST III, characterized by Lactobacillus iners dominance, showed greater dispersion variance than other CSTs. Collectively, these findings reveal a diverse vaginal mycobiome with stage- and disease-specific features and a notable contribution of environmental fungi that may influence cervical pathogenesis. This work provides foundational insight into cervicovaginal fungal ecology in a high-risk Hispanic population and highlights the importance of integrating bacteriome–mycobiome analyses in women’s health research.IMPORTANCEThis study demonstrates that environmental fungi may play an important role in shaping the vaginal microbiome and influencing cervical disease progression in Hispanic women. While most prior research has focused on bacterial communities, our findings highlight fungi as significant yet underexplored contributors to reproductive health. By linking environmental fungal exposures to shifts in the vaginal mycobiome, this work expands understanding of how external factors may affect disease risk. These results underscore the need to include fungal communities in human microbiome research and support more comprehensive approaches to women’s health. Ultimately, this study may inform improved risk assessment, prevention strategies, and the development of targeted interventions in diverse populations. INTRODUCTION The vaginal microbiome is a complex and dynamic ecosystem that plays a critical role in maintaining female reproductive health (1, 2). While bacterial communities have been extensively studied, the fungal component—known as the vaginal mycobiome—remains relatively unexplored (3, 4). Early research suggested that Candida albicans is the dominant fungal species in the vagina, yet recent advances in sequencing technologies have revealed a much broader fungal diversity, with potential implications for gynecological health and disease (5). Fungi, despite being a minor component of the human microbiome, have been found in several body sites, including the skin (6), gut (7), lung (8), oral cavity (9, 10), nose (11), and reproductive tract (3, 4), where they interact with bacteria, the host immune system, and environmental factors to influence health and disease. In the vaginal environment, fungi are thought to be present in low abundance under normal conditions, but they can become opportunistic pathogens under specific circumstances, such as immune dysregulation or microbial imbalance (5, 12). In the cervicovaginal niche, the mycobiome has long been overshadowed by the study of bacterial communities, mainly due to early reliance on culture-based methods that underestimated fungal diversity (5). However, next-generation sequencing and bioinformatics approaches have increasingly demonstrated that the vaginal mycobiota consists of more than just Candida species. Other fungi, including Saccharomyces cerevisiae, Malassezia restricta, and various filamentous fungi, have been detected, though their roles remain poorly understood (3, 4). These fungi may exist in a commensal state, contributing to vaginal homeostasis, or may play a more pathogenic role in conditions such as vulvovaginal candidiasis (13), bacterial vaginosis (BV) (14), and sexually transmitted infections (STIs) (15). The interplay between bacteria and fungi in the cervix is an emerging area of research, particularly in the context of human papillomavirus (HPV) infections. Candida albicans, the most well-characterized vaginal fungal species, has been shown to enhance inflammation and disrupt epithelial integrity, potentially facilitating HPV persistence and cervical lesion progression (12). Moreover, studies have demonstrated that bacterial dysbiosis, particularly a decrease in Lactobacillus species and an overgrowth of anaerobes such as Gardnerella vaginalis, is often accompanied by shifts in fungal composition. The presence of fungal species, such as Malassezia and Saccharomyces, in HPV-positive cervical samples raises questions about how fungi interact with bacteria and the host immune system to influence HPV persistence and disease progression (5, 16). Additionally, fungal colonization has been linked to increased levels of pro-inflammatory cytokines such as interleukin-17A, which may contribute to a more permissive environment for viral persistence and neoplastic transformation (12, 17). Given the high incidence of HPV and cervical cancer among Puerto Rican Hispanics (18), understanding the mycobiome in this population is crucial for advancing individuals’ health research. Hispanic women living in Puerto Rico have non-optimal bacterial communities distinct from other populations (19). Research indicates that the cervicovaginal microbiota of Hispanics in Puerto Rico is commonly dominated by Lactobacillus iners and anaerobic bacteria rather than by the protective Lactobacillus crispatus. These typical Puerto Rican profiles, often enriched with BV-associated bacteria such as Atopobium vaginae and Gardnerella vaginalis, have been linked to cervical intraepithelial neoplasia (CIN), particularly CIN3 precancerous lesions (19, 20). Interestingly, while no specific bacterial biomarkers were identified for HPV infections, fungal diversity was elevated in cases of high-risk HPV, suggesting a potential role for fungi in modifying disease susceptibility and immune responses (4). In this study, we analyze the vaginal mycobiota of 86 Hispanic individuals from Puerto Rico across different reproductive stages (pregnant, non-pregnant, and menopausal) and examine mycobiome diversity associations with cervical disease and HPV status among non-pregnant individuals. We identify key taxa associated with cervical lesions and explore their potential interactions with bacterial community state types (CSTs). By addressing this knowledge gap, we hope to contribute to a more comprehensive understanding of the cervicovaginal microbiome and its role in individuals’ reproductive health, particularly in these high-risk populations. MATERIALS AND METHODS Study design and participant sample collection This cross-sectional study included adult women attending gynecology and colposcopy appointments at the University of Puerto Rico (UPR) and San Juan City clinics, located in the metropolitan area of San Juan, Puerto Rico. Exclusion criteria were (i) a history of regular urinary incontinence, (ii) prior treatment for or suspicion of toxic shock syndrome, (iii) current candidiasis, (iv) active urinary tract infection, (v) being treated for any active STI, including gonorrhea or chlamydia, and (vi) cervicovaginal irritation at the time of screening. Eligible participants were adult women aged 21–60 years who self-identified as Hispanic and resided in Puerto Rico. Individuals were recruited from gynecology and colposcopy clinics at the University of Puerto Rico and San Juan City. All participants were required to be asymptomatic at the time of enrollment, able and willing to provide written informed consent in accordance with U.S. Health Insurance Portability and Accountability Act (HIPAA) regulations, and to comply with study procedures. Only those who met all study requirements and did not meet any of the predefined exclusion criteria were included in the study. Only asymptomatic participants were enrolled, and STI testing was not routinely performed in this population. Women who had received antibiotic treatment at least 2 months before enrollment were not excluded, as they represented a small proportion of the study cohort and were evenly distributed across groups. Data and biological samples were collected between 2018 and 2021 from participants aged 21 to 60 years. Participants completed a detailed metadata questionnaire capturing demographic data (age, birthplace, employment, and education), sexual history (age at sexual debut and number of partners), health and medical history, antibiotic and vitamin use, and body mass index (BMI). A sterile speculum was used to visualize the cervix. Cervical samples were obtained by gentle rotations in a circular motion in the posterior fornix—a sampling site selected to optimize HPV detection and provide a representative profile of the cervicovaginal microbiota as reported in previous studies (19)—using sterile Catch-All Specimen Collection Swabs (Epicenter Biotechnologies, Madison, WI) and immediately placed in MoBio bead tubes containing MoBio PowerSoil buffer (MoBio, Carlsbad, CA). Swabs were agitated for approximately 20 s in 750 μL of buffer before removal. Additionally, ~10 mL of cervical lavage fluid was collected by flushing the vaginal canal with Pure certified nuclease-free water (Growcells, Irvine, CA, USA) for future analyses. Lavage samples were transferred to 15 mL tubes, and pH was measured using Hydrion pH paper strips. All samples were labeled, placed on ice for up to 4 h, transported to the laboratory, and stored at −80°C until nucleic acid extraction and subsequent molecular analyses were performed. All processing and PCR procedures were performed in a single laboratory (F.G.-V.) to minimize technical variation. DNA extraction and sequencing Genomic DNA (gDNA) was extracted from posterior fornix cervical swabs using the Qiagen PowerSoil Kit (QIAGEN LLC, Germantown, MD, USA), following an optimized protocol described previously (16). No human DNA depletion or microbial/viral DNA enrichment steps were performed. Briefly, PowerBead tubes were homogenized for 10 min at 3,200 rpm using a Vortex-Genie 2 G560 (Scientific Industries, Inc., NY). Solutions C2 and C3 (100 μL each) were combined and vortexed for 5 s to facilitate cell lysis. DNA was eluted in 100 μL of sterile PCR-grade water pre-warmed to 55°C, which was allowed to remain on the filter for 5 min before the final centrifugation step to maximize yield. DNA concentrations were quantified using the Qubit dsDNA High Sensitivity Assay (Thermo Fisher Scientific, Waltham, MA), with yields ranging from 5 to 100 ng/μL. Genomic DNA was shipped to an external sequencing facility with average concentrations of 10–30 ng/μL. Negative controls were included throughout DNA extraction and library preparation, consisting of sterile swabs and molecular-grade water from the extraction kit processed in parallel with all samples. These controls did not yield detectable ITS reads and were therefore not included in downstream analyses. For 16S rRNA sequencing, extracted DNA was normalized to 4 nM during library preparation. The hypervariable V4 region of the 16S rRNA gene (~291 bp) was amplified using universal bacterial primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′), as established by the Earth Microbiome Project (https://earthmicrobiome.ucsd.edu/protocols-and-standards/16s/) (21). Sequencing was outsourced to Argonne National Laboratory (Illinois, USA) and performed on an Illumina MiSeq platform using a 2 × 250 bp paired-end kit, and bacterial diversity data were previously reported (19). To minimize potential bias in gDNA extraction, we used DNA extracted from the same patient swab samples that were previously utilized for 16S rRNA analysis. Clinical metadata collection Clinical data, including cervical cytology and pathology results, were obtained from patient medical records. Cytological findings were classified according to the Bethesda system as: negative for intraepithelial lesion or malignancy (NILM), low-grade squamous intraepithelial lesion (LGSIL), or high-grade squamous intraepithelial lesion (HGSIL) (22). Due to a low number of samples, patients with a cytology result of atypical squamous cells of undetermined significance (ASC-US) categorized as either HPV-negative or HPV-lr exclusively (n = 4) were reclassified as NILM. Patients identified as ASC-US with HPV-positive (n = 2) were classified as LGSIL. All sequencing reads and associated metadata were uploaded to QIITA (23) (study ID: 12871, https://qiita.ucsd.edu/study/description/12871). Raw internal transcribed spacer (ITS) sequencing data were deposited in the European Nucleotide Archive under accession number ERP136546. HPV genotyping and cytologic evaluation HPV genotyping was performed using SPF10-LiPA, a highly sensitive short polymerase chain reaction fragment assay (Labo Biomedical Products, Rijswijk, The Netherlands), based on Innogenetics-licensed technology. Amplified products from DNA extracted from swabs and used for both 16S and ITS sequencing were subjected to reverse hybridization for genotype identification by comparison to standardized kit controls as previously described (19). This assay enables detection of 25 mucosal HPV genotypes, including 11 low-risk (HPV-lr): 6, 11, 34, 40, 42, 43, 44, 53, 54, 70, and 74; and 14 high-risk (HPV-hr): 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68/73. Genotyping results were categorized into three groups: (i) HPV-hr, (ii) HPV-lr exclusively, and (iii) HPV-negative (Table 1). TABLE 1Demographic and clinical characteristics of the studied populationVariableLevelPregnantNon-pregnantMenopauseP valueOverall n 9631486Age (mean [SD])28.00 (4.92)36.17 (8.84)54.00 (5.19)<0.001a38.22 (10.89)Age range<50>50BMI (mean [SD])32.21 (10.55)28.85 (7.53)26.67 (4.42)0.23128.85 (7.53)Bacterial CST by 16S (%)I0 (0.0)11 (17.5)3 (21.4)0.016a14 (16.3)II0 (0.0)3 (4.8)1 (7.1)4 (4.7)III6 (66.7)19 (30.2)3 (21.4)28 (32.6)IV1 (11.1)29 (46.0)7 (50.0)37 (43.0)V2 (22.2)1 (1.6)0 (0.0)3 (3.5)Cervical disease status (%)Negative1 (11.1)43 (68.3)14 (100.0)<0.001a58 (67.4)Positive8 (88.9)20 (31.7)0 (0.0)28 (32.6)Cervical disease type (%)NILM1 (11.1)43 (68.3)14 (100.0)0.001a58 (67.4)LGSIL4 (44.4)10 (15.9)0 (0.0)14 (16.3)HGSIL4 (44.4)10 (15.9)0 (0.0)14 (16.3)HPV statusb (%)Negative2 (22.2)24 (38.1)7 (50.0)0.40833 (38.4)Positive7 (77.8)39 (61.9)7 (50.0)53 (61.6)HPV type (%)Negative2 (22.2)24 (38.1)7 (50.0)0.57733 (38.4)HPV-lr exclusively0 (0.0)3 (4.8)1 (7.1)4 (4.7)HPV-hr7 (77.8)36 (57.1)6 (42.9)49 (57.0)HPV-hr type richness total (mean [SD])2.00 (0.82)1.92 (1.06)2.57 (1.40)0.3522.02 (1.08)HPV-hr type richness mean (SD)1.71 (0.76)1.47 (0.61)2.33 (1.51)0.049a1.61 (0.81)HPV-lr type richness mean (SD)1.00 (0.00)1.29 (0.47)1.33 (0.58)0.6891.27 (0.46)Antibiotic use last 2 months (%)No9 (100.0)60 (95.2)14 (100.0)0.56783 (96.5)Yes0 (0.0)3 (4.8)0 (0.0)3 (3.5) a P value calculated using Fisher exact test and Kruskal-Wallis test. b A table with patient HPV risk type is given as supplemental data columns B–H (Text S1A). Demographic and clinical characteristics of the studied population P value calculated using Fisher exact test and Kruskal-Wallis test. A table with patient HPV risk type is given as supplemental data columns B–H (Text S1A). Sequencing the ITS1 genes Fungal community profiling was outsourced at the Argonne National Lab and performed by amplifying the internal transcribed spacer (ITS) region using a dual-indexed paired-end sequencing approach on the Illumina platform. The EMP.ITSkabir primer set, designed for broad coverage of fungal and other microbial eukaryotic lineages, was used to target the ITS1 region (24, 25). The forward primer (ITS1f) is located 38 bp upstream of the ITS1 region (25) and was synthesized with a 5′ Illumina adapter and linker sequence (AATGATACGGCGACCACCGAGATCTACAC GG CTTGGTCATTTAGAGGAAGTAA). The reverse primer (ITS2), identical to that described by White et al. (1990) (25), included the reverse complement of the 3′ Illumina adapter, a unique 12-bp Golay barcode for sample multiplexing, and a linker sequence (CAAGCAGAAGACGGCATACGAGATNNNNNNNNNNCGGCTGCGTTCTTCATCGATGC). PCRs were performed under standard cycling conditions optimized for fungal ITS amplification, and resulting amplicons were pooled, purified, and sequenced using paired-end chemistry. This protocol has been shown to enhance detection sensitivity and taxonomic resolution of fungal communities while minimizing amplification bias (26). Sequence processing Raw data were processed using QIIME2 (version 2023.7). Amplicon sequence variants (ASVs) were inferred using DADA2. Taxonomic classification was conducted using the UNITE dynamic database (version 9, 25 July 2023) and a pre-trained naïve Bayes classifier in QIIME2. The UNITE database assigns each fungal ASV to a species hypothesis (SH), which represents a cluster of sequences with ≥97% similarity and is used as a proxy for species-level identification in fungal ecology. Each SH is associated with a unique identifier (e.g., “SHXXXXXXX”) and serves as a standardized taxonomic unit when formal species names are unavailable or uncertain. These SH identifiers were used as the species-level taxonomic resolution for all downstream statistical analyses. The ASV table was then imported into R (version 4.1.2) for downstream filtering. In R, unclassified ASVs at phylum, class, and order levels were removed. Additionally, sequences corresponding to chloroplasts, mitochondria, archaea, and any non-Fungi sequence (k__Eukaryota_kgd_Incertae_sedis, k__Protista) were removed. A subset of the unclassified ASVs with the highest read abundance was manually curated using NCBI BLAST to confirm taxonomic assignments. We removed samples with fewer than 100 reads, taxa with fewer than two sequences or with no SH classification. No rarefaction was applied due to the low number of reads for an important number of samples, as this resulted in the discarding of a considerable number of samples. Among 102 sequenced samples, 16 were removed: 15 had fewer than 100 sequences, and one had a single ASV classified as Protista (SampleID: 2302). We analyzed the fungal communities compared to the bacterial profiles using CST groups as determined by VALENCIA (27). These are bacterial CSTs defined as CST I (Lactobacillus crispatus–dominated, protective), CST II (Lactobacillus gasseri–dominated), CST III (Lactobacillus iners–dominated, transitional), CST IV (diverse, Lactobacillus-depleted, anaerobe-rich/dysbiotic), and CST V (Lactobacillus jensenii–dominated). Statistical analyses The phyloseq package (28) in R was used to conduct vaginal mycobiota diversity analyses. The models assessed reproductive stage (pregnant, non-pregnant, and menopausal) across all patients, while cervical disease status (HGSIL, LGSIL, and NILM), HPV type status (HPV-hr and negative), and CST from bacterial microbiome (CST-I, CST-III, and CST-IV), only among non-pregnant individuals. Age and BMI were included as confounding variables in all models (alpha/beta diversity and discriminant taxa analysis), except for the reproductive stage variable, which did not include age due to high collinearity with the reproductive stage variable. Mycobial alpha and beta diversity were estimated at the SH level. To measure alpha diversity, the Shannon index was calculated and set as the dependent variable. Independent variables (group or cervical disease or HPV status, or CST) were evaluated using linear regression models. Normality in the distribution of the Shannon metric was verified and confirmed using “shapiro.test” R function. Shannon’s normality made it suitable to run linear regression models without any transformation. The “step” function in R was used to determine the best-fitted model by systematically eliminating non-relevant variables. We used “emmeans” and the “pairs” R function to conduct pairwise analysis of specific variables. Beta diversity was analyzed by computing Bray-Curtis and Jaccard distances, followed by permutational multivariate analysis of variance (PERMANOVA) (29) with the “adonis2” (30) R function for community composition differences and permutational analysis of multivariate dispersions (PERMDISP) (31]) to evaluate inter-individual dispersion with “betadisper” (30) R function. Non-metric multidimensional scaling was employed to ordinate beta diversity. Specifically, for the reproductive stage variable, two outliers (SampleID: 2031, 2113b) were excluded to enhance sample dispersion visualization. Additionally, to complement this analysis and specifically assess variance heterogeneity, Levene’s test was applied to the distances of each sample to its group centroid. This approach was motivated by visual inspection of beta diversity dispersion plots due to CST groupings. Levene’s test was conducted to formally evaluate differences in variance among CST groups. Discriminant taxa analyses were conducted at the species and genus levels. Discriminant taxa to each study variable were assessed using MaAsLin2’s (32), including only the taxa present in at least 10% of the samples and with a relative abundance of at least 10%. Prevalence analyses were performed at the Species level using Fisher exact test. P values were adjusted for multiple comparisons using the false discovery rate method in all analyses that involved generating multiple P values (e.g., pairwise comparisons, multiple prevalence calculations, or discriminant taxa analyses). All analysis files and scripts are available at https://github.com/UPR-MicrobiomeLab/ITS_86_HPV. RESULTS Study population The cohort included 86 Hispanic individuals: menopausal (n = 14), non-pregnant (n = 63), and pregnant (n = 9). Age significantly differed across groups (P < 0.001), while BMI did not (P = 0.231). Bacterial CSTs prevalence varied by group (P = 0.016), with pregnant individuals mainly exhibiting CST-III, and the other groups dominated by CST-IV. Pregnant individuals showed the highest prevalence of cervical disease (88.9%, P < 0.001), specifically of high-grade lesions (HGSIL; 44.4%, P = 0.001). HPV infection rates were similar among groups (P = 0.408), and most of them were high-risk HPV (HPV-hr, 92.5%). HPV-hr type richness (count of different HPV types in a sample) was significantly greater in menopausal individuals (P = 0.049). Antibiotic use in the last two months was infrequent and similar across groups (P = 0.567; Table 1). Fungal taxonomic diversity We analyzed 86 samples/individuals. Across these, we found 457 ASVs, corresponding to 173 SH, from which 120 (69%) occurred exclusively in single samples and were not shared between samples, indicating a highly heterogeneous and individualized fungal community with few recurrent taxa. The remaining 31% were shared by two or more samples (Text S1B). The 173 SHs represented 125 distinct genera, but seven samples (SampleID: 2218, 2031, 2113b, 2278, 2302, 2306, and 2308) showed an absolute dominance of a single SH, reflecting extremely low-complexity communities. Across the entire data set, only five SHs were detected in 10 or more samples: the human-associated yeast Candida albicans and the environmental yet human-associated yeast Hortaea werneckii (both Ascomycota), as well as three environmental filamentous fungi from the class Agaricomycetes—Agaricomycetes sp., Scopuloides dimorpha, and Trametes cubensis (Text S1B). Two individuals who presented only unclassified SHs and were removed then the analyses were performed at this specific taxonomic level (SampleID: 2128b, with the absolute dominance of one ASV with classification available up to Genus: Polyporales_sp and, SampleID: 2292, composed only of two ASVs classified as g__Aspergillus or f__Teratosphaeriaceae). Candida albicans was present in 45.3% of the samples (39 out of 86) and had the highest mean relative abundance of 43.1%, when present. The next most abundant SH was the non-human-associated Agaricomycetes_sp with an overall prevalence of 50.0% (43 out of 86) and a mean relative abundance of 29.8%, when present (Fig. 1). We identified a total of 11 human-associated taxa and 162 non-human-associated SHs (Fig. 1). Taxonomic composition and clinical metadata of vaginal samples across reproductive stages. Each column represents one sample, grouped into pregnant, non-pregnant, and menopausal stages (separated by dashed vertical lines). The top panel shows clinical metadata including cervical disease status (NILM, LGSIL, and HGSIL), HPV type status (negative, HPV-hr, and HPV-lr exclusively), and community state types (CSTs I–V). The taxonomic summary displays the relative abundance (%) of fungal taxa at three levels: all detected genera (GENUS), human-associated genera (GENUS human-associated), and human-associated species (SPECIES human-associated). Taxa with low relative abundance are grouped as “<1% abundance.” Color assignments are consistent across panels and correspond to the legends below. Sample IDs are displayed on the x-axis. Reproductive stage groups analysis (pregnant, non-pregnant, and menopausal comparison) We compared vaginal fungal diversity and composition among reproductive stages (pregnant, non-pregnant, and menopausal), including all individuals and then cervical disease (LGSIL, HGSIL, and NILM), HPV type status (HPV-hr, negative), and vaginal 16S CST (CSTs I, III, and IV), including only non-pregnant individuals. Reproductive stages were significantly associated with differences in beta dispersion (Jaccard, P = 0.004; Bray-Curtis, P = 0.046). Specifically, non-pregnant women showed a significantly greater inter-individual compositional heterogeneity than pregnant women (P adj = 0.025, Jaccard; Fig. 1B), while menopausal individuals showed a marginally significant trend toward greater heterogeneity compared to pregnant women (P adj = 0.074, Jaccard, PERMDISP test; Fig. 1B). The reproductive stages did not have a significant difference in fungal composition (PERMANOVA, P > 0.723; Fig. 1A; Text S2). No significant differences in alpha diversity were observed across reproductive stages (P > 0.050, Shannon index, linear model; Text S3). Discriminant taxa and taxa prevalence analyses suggested that no taxa were significantly associated (P < 0.050, P adj < 0.05) with the reproductive stage groups after P value adjustment for multiple testing. However, before P value adjustment, the relative abundance of the genus Recinicium (11/86) was greater in menopausal individuals (P = 0.027, P adj = 0.562; MaAsLin2; Fig. 2C; Text S4) and Candida tropicalis prevalence appeared marginally higher in pregnant (33.3%) than in menopausal (0%) or non-pregnant individuals (7.8%; P = 0.051, P adj = 0.568, Fisher’s exact test; Text S5), though these associations did not remain significant after adjustment. Vaginal mycobiome diversity by reproductive stage. (A) Ordination indicating centroids with an X symbol and 95% CI ellipse by using Jaccard dissimilarity, analyzed by PERMANOVA. (B) Distance to centroid using Jaccard dissimilarity, analyzed using PERMDISP test and its pairwise form. (C) Resinicium genus relative abundance with logarithmic transformation, identified by MaAsLin2 as significantly associated taxa prior to P value adjustment for multiple comparisons. Cervical disease and HPV status among non-pregnant individuals Cervical disease was evaluated among non-pregnant individuals due to the limited sample size and substantial physiological differences between the other groups (pregnant and menopausal; Table 1). While beta diversity and dispersion showed no significant difference between cervical disease groups (LGSIL, HGSIL, and NILM, P > 0.356, Bray-Curtis and Jaccard; Text S2), alpha diversity was significantly lower in HGSIL compared to LGSIL (P = 0.037) or NILM (P = 0.037), and LGSIL had greater alpha diversity than NILM (P = 0.049, estimated marginal means; Fig. 1C; Text S3). Candida parapsilosis (7/63) was at a greater proportion in LGSIL than NILM but not significantly after P value adjustment (P = 0.011, P adj = 0.497, MaAsLin2). Although no taxa remained significant after correction for multiple comparisons, prior to P adjustment, Candida parapsilosis was more prevalent in LGSIL samples (40.0%, 4 of 10) compared with NILM (11.1%, 7 of 43; P = 0.020) and HGSIL (0%, 0 of 10; P = 0.038; Fisher’s exact test; Text S6). HPV status did not show any significant difference for either beta using PERMANOVA or PERMDISP or alpha diversity (HPV-hr vs HPV negative, P > 0.327; Texts S2 and S3). The category “HPV-lr exclusively” was removed from this comparison due to the low sample number (n = 3). The only potential HPV-associated discriminant taxon was Aspergillus sulphureoviridis (7/60), being less abundant in HPV-hr than HPV-negative samples, but this was also not significant after P-value adjustment (P = 0.048, P adj = 0.765, MaAsLin2; Text S4). No taxa showed significant prevalence differences by HPV type status (Text S7). Fungal diversity analyses according to bacterial CSTs CSTs were determined previously for these samples from 16S sequencing (19) and were used as variables to associate with fungal diversity in this study. Neither beta using PERMANOVA nor PERMDISP nor alpha diversity was found to be significant (Text S2). However, visual inspection of the beta diversity dispersion plots suggested that CST III exhibited markedly greater variance in distance to its centroid compared to CST I and IV. Levene’s test applied to centroid distances revealed significant differences in variance across CSTs (P = 0.045, Levene; Fig. 3E), although pairwise comparisons between CST I and IV were not significant (P adj > 0.050, pairwise Levene; Fig. 3E; Text S2). The discrepancy between the results of these two variance testing methods (Levene’s test and PERMDISP) comes from the different aspects of dispersion each method evaluates. While PERMDISP compares the mean distances to the group centroid across groups, Levene’s test compares the actual variance of those distances within each group. As such, Levene’s test provides a complementary assessment of dispersion, capable of detecting group-level heterogeneity that may not be captured by the multivariate framework used in PERMDISP. Vaginal mycobiome variation by cervical disease, HPV type, and community state type (CST) among non-pregnant population. Shannon diversity index according to cervical disease status (NILM, LGSIL, and HGSIL) (A). Pairwise comparisons between adjusted group means were conducted based on estimated marginal means derived from the linear model. (B and C) Significant taxa identified as discriminant by MaAsLin2 previous P value adjustment, across cervical disease groups, Candida parapsilosis (B), genus Aspergillus (C). Or across HPV types Aspergillus sulphureoviridis (D). Distance to centroid using Bray-Curtis dissimilarity, analyzed using PERMDISP test and its pairwise form, Levene and pairwise Levene test between CSTs (E). CST groups did not show any discriminant taxa; however, the genus with the highest relative abundance, Aspergillus, was found to be at a lower relative abundance in CST-I compared to CST-II and IV (P = 0.133; Text S4). In terms of taxa prevalence, no differences among CST groups were detected after P value adjustment for multiple comparisons. The species named “Fungi_sp” was found to be most prevalent in CST-IV (20.7%), absent in II and III (P = 0.039, P adj = 0.914, Fisher’s exact test; Text S8). This taxon is composed of nine SHs. The most recent UNITE classification (https://unite.ut.ee/) classifies the two most abundant of these SH (>23% of abundance) as an uncultured Pyrenochaetopsis (https://unite.ut.ee/sh/SH0994492.09FU) and Resinicium (https://unite.ut.ee/sh/SH1252023.09FU), among others (Text S1). DISCUSSION To our knowledge, this is the first study in the Caribbean to comprehensively characterize the cervicovaginal mycobiome in a Hispanic population across reproductive stages and cervical disease states. Using high-resolution ITS1 sequencing, we delineate the diversity, composition, and clinical associations of vaginal fungal communities and explore their interactions with bacterial CSTs, HPV infection, and cervical lesions. The data presented here revealed a considerable fungal diversity in the cervicovaginal environment, with about 173 SH, which correspond to species-level OTUs. The majority of these taxa (69%) were unique to individual samples, indicating that vaginal mycobiota exhibit substantial inter-individual variability. This pattern aligns with previous reports describing the vaginal mycobiome as a heterogeneous and individualized ecosystem (24). Despite this diversity, only a handful of species, such as Candida albicans, Agaricomycetes sp., Scopuloides dimorpha, Trametes cubensis, and Hortaea werneckii were consistently detected across multiple individuals, underscoring their ecological dominance or resilience in the vaginal niche. H. werneckii. is, in fact, a halophilic black yeast and the agent of tinea nigra dark macules in the palms and soles, commonly confused with skin melanoma (33). Notably, C. albicans, a known opportunistic pathogen, was present in nearly half of the samples and exhibited the greatest relative abundance. Its prevalence reinforces its potential dual role as a commensal organism and a pathogenic driver under conditions of dysbiosis or immune imbalance. The detection of non-human-associate fungi, which are very common in the outdoor air in Puerto Rico (34), including members of the Agaricomycetes class, also suggests possible environmental contributions to the vaginal mycobiota—a phenomenon increasingly reported by others (10). Infants born in the aftermath of Hurricane Maria harbor more hurricane and asthma-related taxa, specifically Alternaria sp., in the nasal mycobiome (10). This reflects the continuous influx of environmental fungi into mucosal ecosystems. The adhesion of environmental fungi to epithelial surfaces (a critical step for colonization) potentially reduces fungal and bacterial biofilm formation and alters host-microbe interactions. Such mechanisms could also operate within the vaginal environment, where airborne or dermal fungi entering the mucosa may compete with established communities for adhesion sites and nutrients, thereby influencing colonization dynamics and community stability. Moreover, environmental fungi capable of forming hyphae may enhance surface hydrophobicity and secrete hydrolytic enzymes and toxins, thereby altering local immunity and promoting inflammatory responses. These processes could indirectly impact bacterial composition, potentially driving shifts toward dysbiosis. The presence of abundant environmental fungi in the vaginal tract, therefore, should not be dismissed as mere contamination but rather as a potential ecological force shaping the mucosal mycobiome. Our analysis revealed, however, no significant differences in alpha or beta diversity across CSTs, yet CST III displayed markedly greater fungal community variance among participants, suggesting that most detected fungi might be transient or opportunistic rather than members of a stable core mycobiome among the population. This increased variability may reflect the ecological characteristics of CST III, which is dominated by Lactobacillus iners, a species often associated with transitional or less stable microbiota states. Lactobacillus iners has been associated with Candida pathogenic hyphal transition and biofilm formation (35), which can provide antifungal resistance (36). Although no discriminant taxa were significantly associated with CSTs after multiple testing correction, we observed a trend toward lower relative abundance of the environmental genus Aspergillus in CST I, a Lactobacillus-dominated and protective environment, suggesting that beneficial bacterial communities may inhibit colonization by environmental fungi. Importantly, several of these observations were not statistically significant after P-value adjustment and should therefore be interpreted with caution. The loss of significance likely reflects both the multiple testing burden inherent to microbiome studies and the limited sample size, which may reduce statistical power to detect subtle but biologically relevant differences. This variability with the bacterial communities may indicate that the microbiota influences the stability and resilience of fungal colonization. Although no discriminant taxa were significantly associated with CSTs, the genus Aspergillus, a common fungus in the air, showed a trend toward lower relative abundance in CST I, a protective Lactobacillus-dominated environment, suggesting that beneficial bacterial communities may inhibit colonization by Aspergillus. Additionally, the enrichment of SH “Fungi_sp”—which included genera such as Pyrenochaetopsis and Resinicium—in CST IV, a CST often associated with dysbiosis, supports the notion that bacterial imbalance may facilitate the proliferation of otherwise rare fungal taxa. Pyrenochaetopsis is an environmental Ascomycete occasionally isolated from superficial human infections such as nail or skin lesions (37, 38). Resinicium, a wood-decaying Basidiomycete, has no record of human pathogenicity and appears at low abundance in infant gut or nasal samples as environmental spores (39). Both genera thus represent environmental fungi sporadically detected in human data sets, not established members of the human microbiome. Although the reproductive stage did not significantly influence fungal alpha diversity or taxonomic composition, we observed notable differences in beta dispersion, with non-pregnant individuals exhibiting greater inter-individual heterogeneity compared to pregnant women. This finding suggests that hormonal or immunological shifts during pregnancy may lead to more homogeneous fungal communities, which are often associated with Lactobacillus-rich epithelium in pregnancy, possibly through selective pressures favoring specific taxa. The marginally higher prevalence of Candida tropicalis among pregnant participants may reflect such host-driven ecological filtering. A major finding of this study is the reduction in fungal alpha diversity associated with HGSIL, suggesting that diminished mycobiome diversity may accompany or contribute to the progression of advanced cervical pathology. Lower diversity could reflect a less resilient microbial community more conducive to pathogen dominance and inflammatory processes that facilitate HPV persistence and neoplastic progression. Indeed, Candida parapsilosis was more prevalent in low-grade lesions (LGSIL), though not statistically significant after P adjustment, suggesting potential taxon-specific associations that warrant further investigation. HPV status (HPV-hr vs negative) did not significantly correlate with overall fungal diversity or community composition. However, detecting Aspergillus sulphureoviridis with higher abundance in HPV-hr samples hints at potential interactions between specific fungal taxa and viral infection dynamics. Such interactions may occur through modulation of mucosal immunity or direct effects on epithelial integrity, production of carcinogenic mycotoxins, or as previously suggested in bacterial-fungal-viral co-infection models. Humans, as holobionts (integrated organisms composed of host and microbial partners), are continuously exposed to environmental fungi through inhalation, ingestion, and skin contact, making such encounters inevitable. Airborne spores and hyphal fragments from genera such as Aspergillus, as shown here, are regularly inhaled or come into contact with the mucosal surfaces of the respiratory, gastrointestinal, and urogenital tracts (40, 41). These mucosal sites provide a moist and nutrient-rich microenvironment where fungi can transiently or persistently colonize. While most spores introduced into the body are efficiently cleared through mucociliary and immune mechanisms, a fraction can adhere to epithelial surfaces or germinate within the mucus layer, particularly when local immunity is compromised or the epithelial barrier is disrupted (42, 43). Disruption of the intestinal bacterial community can compromise epithelial barrier integrity and alter immune homeostasis, creating conditions that facilitate fungal translocation beyond the gut. Bacterial dysbiosis has been shown to increase mucosal permeability and reduce colonization resistance, thereby enabling opportunistic fungi such as Candida albicans to traverse the gastrointestinal barrier and disseminate systemically, including to the brain (43). The vaginal mucosa in this case may offer organic substrates and cellular debris that can sustain low-level environmental fungal growth in balance with bacterial communities, enabling the coexistence of certain environmental fungi as commensals even within asymptomatic hosts. The persistence of these fungi within mucosal niches reflects a finely tuned balance between immune recognition and tolerance. The host immune system detects fungal-associated molecular patterns while limiting unnecessary inflammation, allowing commensal species such as Candida, Malassezia, and Aspergillus to persist in low abundance without eliciting disease (44). However, perturbations such as antibiotic use, any behavioral variation such as douching habits, use of detergents, or chronic inflammation can disrupt bacterial communities and mucosal immune regulation, creating ecological space for environmental fungi to expand and potentially contribute to dysbiosis. Emerging research on the aerobiome–mycobiome continuum further suggests that inhaled bioaerosols can transiently influence mucosal immune responses and microbial composition (45). Recent hypotheses suggest that inhaled “aeronutrients,” including vitamins, fatty acids, and trace minerals, and “aeromicrobes,” such as environmental bacteria, may contribute positively to human health by supporting microbial diversity and metabolic balance within the respiratory and gastrointestinal tracts (46) with clear implications for urbanized environments and health. Building on this concept, we extend this framework to include airborne fungi as additional but potentially disruptive components of the human microbiome, including the vaginal niche. Airborne fungal spores and fragments, while ubiquitous, can interact directly with mucosal microbiomes, potentially altering their structure and function. Together, these results emphasize the intricate and dynamic relationship between environmental fungi and the host mucosa, where constant exposure, immune modulation, and microbial crosstalk govern the equilibrium between tolerance and inflammation. This study has several limitations that should be considered when interpreting the findings. Although we propose a possible environmental contribution to the vaginal mycobiota, we acknowledge that contamination cannot be entirely ruled out despite the inclusion of negative controls, which did not yield detectable ITS reads. Additionally, while key confounders such as age and BMI were included in all models, other factors known to influence the vaginal microbiome were not incorporated into the statistical analyses. These include sexual behavior (e.g., recent sexual activity, number of sexual partners, and condom use), hormonal influences (e.g., contraceptive use or hormonal therapy), and lifestyle factors such as smoking. Although some of this information was collected as part of a broader metadata framework, it was not included in the present analysis and may contribute to variability in fungal community composition. Therefore, the observed associations, including the potential role of environmental fungi, should be interpreted with caution, and future studies integrating these variables, as well as the study of environmental samples at the time of human sampling, as we did in another oral study (10), will be important to better disentangle their relative contributions. Collectively, these findings highlight the complex and potentially transient nature of fungal–bacterial interactions, with an important role of environmental fungi, and underscore the need for larger and longitudinal studies to better resolve these dynamics. Conclusion Our findings reveal the ecological complexity of the vaginal mycobiome and its potential influence on reproductive health and cervical disease progression. The reduction in fungal diversity associated with high-grade lesions, together with variability linked to bacterial community state types, highlights the need to consider interkingdom interactions as contributors to cervical pathogenesis. Environmental fungi within this niche may compete with resident mucosal species, shaping microbial resilience and community stability. Collectively, these insights emphasize that the vaginal mycobiome is not a passive inhabitant, but an active ecological partner shaped by host physiology and environmental exposure. Future longitudinal and mechanistic studies are essential to elucidate causal pathways and determine whether targeted modulation of fungal communities could inform new preventive or therapeutic strategies in HPV-related disease.