Introduction

Condyloma acuminatum (CA), commonly known as anogenital wart, is a sexually transmitted infection caused by human papillomavirus (HPV)1,2. CA is characterized by skin-colored fleshy papules in the anogenital region3,4 and has a prevalence of approximately 1% in Asia, with a higher incidence among men than women3. The pathogenesis of CA remains complicated, and the main cause is believed to be persistent HPV infection.

Previous studies have revealed that the female vaginal microenvironment is essential for the appropriate functioning of the vaginal barrier, which defends against sexually transmitted infections (STIs) and maintains vaginal health5,6. Vaginal microbiota (VMB) refers to the the microbial community in the vagina of reproductive-aged females and is typically dominated by Lactobacillus species7. VMB plays a significant role in the health and disease state of the female reproductive tract, is intricately associated with the vaginal epithelium and depends on host tissues for essential nutrients8,9 As a unique part of the human mucosal immune system, the vaginal mucosal barrier has a lower pH and is often exposed to inflammatory stimuli and pathogens10. Numerous studies suggest a link between changes in the vaginal microbiota and STIs, as well as various reproductive health outcomes11,12,13,14,15.

Despite a growing body of literature on the vaginal microbiome, most current research focuses on its general characteristics. In addition, alterations in microbial diversity and composition have been documented in various gynecological conditions, including HPV infection16. However, studies on the vaginal microbiota associated with HPV infection predominantly focus on cervical neoplasia caused by high-risk HPV17,18with limited attention given to the association between the vaginal microbiota and HPV-induced CA. Until now, only one study from Zhou19 revealed the profile of the vaginal microbiota in CA patients; however, it was limited to CA patients who had a single infection of HPV6 or 11 and could not mimic real clinical situations. Furthermore, a correlation analysis between differentially expressed microorganisms and clinical characteristics was lacking. This research gap highlights the need for targeted studies to investigate the relationship between the vaginal microbiome and HPV infection, particularly among populations presenting with CA.

Although most cases of CA are caused by low-risk HPV, many patients with CA are co-infected with both low-risk and high-risk HPVs, which can lead to persistent HPV infection. Knowledge is limited concerning the changes in the vaginal microbiota among female patients with CA. Additionally, the differences between the vaginal microbiota of patients who are persistently HPV positive (HPV_PP) and those who have turned negative (HPV_TN) are not well understood. Furthermore, systematic studies evaluating the various species of vaginal microbiota are lacking.

To address these issues, this study focuses on the differences in the vaginal microbiota between the HPV_PP group and the HPV_TN group, as well as the differences in the vaginal flora composition between the CA group and the healthy control. Moreover, we investigated species of the vaginal microbiota that may aid in the diagnosis and prediction of the outcomes of HPV infection. This study aims to contribute valuable knowledge that may enhance clinical management strategies and improve patient outcomes.

Results

Clinical characteristics of the recruited subjects

A total of 63 female patients with CA (CA) and 20 healthy females (CON) were included in this study. No significant differences were observed in age, BMI or marital status between the two groups. More detailed demographic and clinical characteristics of the recruited individuals are shown in Table 1. Additionally, 24 patients in the CA group had low-risk HPV genotyping, accounting for 38.10% of the total. Moreover, 26.98% of CA patients had mixed infections of low-risk HPV and HPV16/18, while the remaining patients were co-infected with low-risk HPV and other high-risk types. Most patients with CA (74.60%) had a disease course of less than 6 months at the time of the first visit. Furthermore, 77.78% of patients with CA did not have low-grade squamous intraepithelial lesions (LSIL). Approximately 22.22% of patients in the CA group underwent treatment for more than 12 months due to recurrence. We re-examined these patients at the end of 2 years after the first visit and found that 25.40% of them had persistent HPV positivity. For more detailed information, please refer to Table 2. The details of the clinical characteristics of each CA patient are listed in Supplementary Table 1.

Table 1 Characteristics of the study subjects. Comparison of demographic and clinical characteristics between the female CA group (n = 63) and CON group (n = 20). CA, condyloma acuminatum; CON, healthy female; BMI, body mass index; HPV, human papillomavirus. P < 0.05 was considered to indicate statistical significance.
Full size table
Table 2 Characteristics of patients with condyloma acuminatum (CA). Clinical details for patients in the CA group (n = 63). Percentages were calculated based on the total number of CA patients. LSIL, low-grade squamous intraepithelial lesion; CIN, cervical intraepithelial neoplasia; hr_hpv, high-risk HPV.
Full size table

Changes in vaginal microbial diversity in CA patients compared with those of the CON group

The alpha diversity of the vaginal microbiota indicated by Chao1 and ACE indices revealed that the species richness of the vaginal microbiota was greater in the CA group than in the CON group (P = 0.036 and 0.046, respectively) (Fig. 1a–b). Principal component analysis (PCA) and principal coordinate analysis (PCoA) were performed on the vaginal microbiota in the CON and CA groups based on the bacterial operational taxonomic unit (OTU) relative abundances (Fig. 1c–d), indicating that the samples from the CA group and CON group did not separate from each other. Furthermore, analysis of similarity statistics (ANOSIM) was performed to test the significance of differences among prior grouping strategies based on the PCoA results (Supplementary Fig. 1a–h). With the exception of the Jaccard index (OTU-based, P = 0.034; Supplementary Fig. 1 h), no significant differences were detected in Bray-Curtis dissimilarity at the order, phylum, genus, or species levels, or in unweighted/weighted UniFrac distances or OTU-based Bray-Curtis.

Fig. 1
Fig. 1
Full size image

Alpha and beta diversity of the vaginal microbiota between healthy control samples (CON) and the condyloma acuminatum group (CA). (a–b) Chao1 diversity and ACE diversity, which represent the community richness of the vaginal microbiota, were calculated. (c) Principal component analysis (PCA) of the vaginal microbiota based on the bacterial operational taxonomic units (OTUs) between the CON and CA groups. (d) Principal coordinate analysis (PCoA) of the vaginal microbiota based on the bacterial OTU relative abundances between the CON and CA groups.

Changes in the vaginal microbiota taxonomic composition in CA patients

According to the species annotation results, a corresponding histogram depicting relative abundance profiling was generated for each group at the phylum, class, order, family, genus and species levels. In Fig. 2, the histogram illustrates the average relative abundance of genera and species, thereby visualizing the composition and proportion of genera and species with higher relative abundance in each group. The five most abundant genera in the vaginal microbiota of both the CA and CON groups were Lactobacillus, Gardnerella, Atopobum, Prevotella and Sneathia (Fig. 2a). At the species level, uncultured_Firmicutes bacterium, Gardnerella vaginalis, Atopobium vaginae, Lactobacillus jensenii and Lactobacillus gasseri were the most abundant entities in the vaginal microbiota of both groups (Fig. 2b). In the HPV_PP and HPV_TN groups, the abundance of the genus Prevotella and the species Prevotella bivia was greater in the HPV_PP group than in the HPV_TN group, whereas the abundance of the genus Gardnerella and the species of G. vaginalis was lower in the HPV_PP group (Fig. 2c–d). However, due to the high conservation of the V3-V4 region among Gardnerella species and the lack of updated reference databases, our detection of G. vaginalis may represent a mixture of multiple Gardnerella species, including recently identified taxa.

Given the importance of Lactobacillus in the female vaginal ecosystem20we also focused on the relative abundance of Lactobacillus. The abundance of L. jensenii appeared to be greater in the CON group, whereas the abundance of L.gasseri was greater in the CA group (Fig. 2b). We further compared the relative abundance of various Lactobacillus species, including L. jensenii, L. gasseri, L. reuteri, L. delbrueckii, L. iners, L. murinus, L. mucosae, L. coleohominis, L. salivarius, L. plantarum, L. crustorum, L. ruminis and L. fermentum, in the CA and CON groups and in each subgroup using the Wilcoxon rank-sum test. The relative abundance of these Lactobacillus species did not significantly differ among subgroups categorized by HPV type, LSIL status or HPV clearance in addition to between the CA and CON groups (Supplementary Fig. 2).

Fig. 2
Fig. 2
Full size image

Comparison of the vaginal microbiota taxonomic composition between healthy control samples (CON) and the condyloma acuminatum group (CA) and between the HPV persistent-positivity (HPV_PP) group and HPV turn-negative (HPV_TN) group. (a) Histogram showing the abundance of vaginal microbial communities in both the CON and CA groups at the genus level. (b) Histogram showing the abundance of vaginal microbial communities in both the CON and CA groups at the species level. (c) Histogram showing the abundance of vaginal microbial communities in both the HPV_PP and the HPV_TN groups at the genus level. (d) Histogram showing the abundance of vaginal microbial communities in both the HPV_TN and HPV_PP groups at the species level. *Note: The detection of Gardnerella vaginalis in the present study may represent a mixture of multiple Gardnerella species.

In addition, linear discriminant analysis (LDA) effect size LEfSe was used to evaluate the differences in taxonomic composition between the CA and CON groups. LEfSe_data between the CA and CON groups, including the biomarker_name, log_value, groups, LDA_values and P value are presented in Supplementary Table 2. As shown in Fig. 3a–b, the abundance of the P. bivia, S. agalactiae, M. hominis, G.asaccharolytica, Bifidobacterium detium, Curvibacter gracilis, Delftia tsuruhatensis, Rhodococcus erythropolis and Mitsuarua sp 7 species increased, whereas the abundance of Sneathia sanguinegens, S. anginosus, Haemophilus parahaemolyticus, L. ruminis, Pseudomonas aeruginosa, L. salivarius and Pauljensenia hongkonggensis increased in the CON group (P < 0.05, |LDA score| >2). Furthermore, the 20 most abundant species in the vaginal microbiota of the CA and CON groups were selected for further comparison by the Wilcoxon rank-sum test. As shown in Fig. 3c, the abundance of P. bivia, G. asaccharolytica, S. agalactiae and M.hominis species were significantly greater in the CA group than in the CON group, among which P. bivia was the most abundant species. In contrast, S. sanguinegens was more abundant in the CON group than in the CA group.

Fig. 3
Fig. 3
Full size image

Relative abundance profile of vaginal microbiota between healthy control samples (CON) and the condyloma acuminatum group (CA). (a) Differentially expressed vaginal microbiota between the CON and CA groups chosen by |LDA score| > 2 and P < 0.05. (b) Cladogram showing the most differentially abundant taxa identified by LEfSe. (c) Boxplots showing the relative abundance of the top 20 species in the CON and CA groups. The detection of Gardnerella vaginalis in the present study may represent a mixture of multiple Gardnerella species.

Changes in the vaginal microbiota associated with clinical parameters

To explore the potential for specific bacterial species to affect the development and progression of CA, we further compared the changes in the top 20 species in different groups based on various clinical characteristics of the CA patients, as shown in Tables 1 and 2. We found that the relative abundance of P. bivia, Prevotella timonensis and Anaerococcus prevotii was significantly greater in the persistent-positive (HPV_PP) group than in the turn-negative (HPV_TN) group (P < 0.05; Fig. 4a). Additionally, the abundance of A. prevotii was greater in the samples from the HPV16/18 co-infected group (HPV16_18) than in those from the low-risk HPV-infected group (P < 0.05; Fig. 4b). Moreover, the relative abundance of the top 20 species across different groups was compared based on LSIL status, disease duration at the first visit and treatment duration and is presented in Supplementary Fig. 3. No significant differences were observed among the groups, except for an increased abundance of M. hominis in the Not_LSIL group (Supplementary Fig. 3a–c).

Fig. 4
Fig. 4
Full size image

Relative abundance of the top 20 species in different groups based on the status of HPV infection. (a) Relative abundance of the top 20 species in the group with HPV persistent-positive (HPV_PP) and the group with HPV turn-negative (HPV_TN). (b) Relative abundance of the top 20 species in the groups with different HPV genotypes of low-risk HPVs (Lr_HPV), co-infected with HPV16/18 (HPV16_18) and with other high-risk HPVs (Other_Hr_HPV). The detection of Gardnerella vaginalis in the present study may represent a mixture of multiple Gardnerella species.

Furthermore, an advanced correlation Link of the relative abundances among the top 15 species and clinical characteristics was performed using Omicstudio21. As shown in Fig. 5, S. agalactiae was positively related to treatment course (P < 0.05). Moreover, for noncontinuous variables, such as the HPV genotype, whether or not LSIL or HPV_PP. P. bivia, G. asaccharolytica, Allpscatdovia omnicolens, Fusobacterium gonidiaform, A. prevotii and S. sanguinegens were positively related to HPV_PP, and A. omnicolens was positively related to the status of LSIL (P < 0.05).

Fig. 5
Fig. 5
Full size image

An advanced correlation Link of the relative abundance among the top 20 species and clinical characteristics. Correlations of the relative abundances among the top 20 species and the clincal characteristics of age, BMI, disease course, treatment duration, HPV genotype, presence or absence of LSILs and the status of HPV_PP were analyzed using Mantel’s test.

Differences in the vaginal microbiota between the HPV_PP and HPV_TN group

The combined results shown in Figs. 4 and 5 suggest that the differential relative abundance of the vaginal microbiota may play an important role in the clearance of HPV and P. bivia and A. prevotii might be potential indicators for predicting HPV_PP. Therefore, we focused on these two groups and studied the relative abundance profile of the vaginal microbiota.

First, PCA was performed on the genital tract microbiota based on the bacterial OTUs between the HPV_PP and HPV_TN groups (Fig. 6a). ANOSIM of the microbial beta diversity in the vaginal microbiota revealed a significant difference between the HPV_PP and HPV_TN groups (R = 0.1741, P = 0.012; Fig. 6b). Furthermore, LEfSe was conducted to compare the taxonomic composition between the HPV_TN and HPV_PP groups, including the biomarker_name, log value, groups, LDA_values and P value, and the data are presented in Supplementary Table 3. As shown in Fig. 6c–d, at the species level, M. hominis was more prevalent in the HPV_TN group, whereas P. bivia, Schaalia radingae, Enterococcus faecalis, Negativicoccus succinicivorans, Porphyromonas asaccharolytica, Prevotella buccalis, Peptoniphilus coxii and Winkia neuii were more abundant in the HPV_PP group (|LDA score| > 2 and P < 0.05).

In addition, the predictive performance of P. bivia and A. prevotii abundance for persistent HPV infection was evaluated using logistic regression, and the a receiver operating characteristic (ROC) curve was generated with the pROC package. The area under the curve (AUC) was calculated to assess the discriminatory ability of the model. The AUC for P. bivia was 0.744, whereas the AUC for A. prevotii was 0.7247 (Fig. 6e), indicating that the increased relative abundance of P. bivia may serve as a predictive index for persistent positive HPV infection, given that P. bivia showed higher abundance in the CA group as previously illustrated.

Fig. 6
Fig. 6
Full size image

Relative abundance profile of the vaginal microbiota between the group with HPV persistent-positive (HPV_PP) and the group with HPV turn-negative (HPV_TN). (a) Principal component analysis (PCA) of the genital tract microbiota based on the bacterial operational taxonomic units (OTUs) between the HPV_PP and HPV_TN groups. (b) Anosim analysis of the microbial beta diversity vaginal microbiota based on the OTUs between the HPV_PP and HPV_TN. (c) Differentially expressed vaginal microbiota between the HPV_PP and HPV_TN groups chosen by |LDA score| > 2 and P < 0.05. (d) Cladogram showing the most differentially abundant taxa identified by LEfSe between the HPV_PP and HPV_TN groups. (e) Receiver operating characteristic (ROC) curves of the likelihood of persistent HPV infection predicted by the relative abundance of Prevotella bivia and Anaerococcus prevotii.

Changes in function between the HPV_PP and HPV_TN groups

Functional prediction analysis was conducted using LEfSe to identify the significant functional differences between the HPV_PP and HPV_TN groups. LEfSe_data illustrating functional changes between HPV_TN and HPV_PP groups, including the lineage, log_value, groups, LDA_values and P value, are presented in Supplementary Table 4. A total of 44 categories from the Kyoto Encyclopedia of Genes and Genomes22,23 (KEGG) differed significantly between the two groups (P < 0.05, |LDA score|>2; Fig. 7). Our analysis revealed that several functions, including signal transduction, drug resistance antimicrobial, xenobiotics biodegradation and metabolism, and the MAPK signaling pathway, were significantly enriched in the HPV_PP group. Conversely, several functions, such as necroptosis, the phosphatidylinositol signaling system, cell growth and death, were lower in the HPV_PP group than in the HPV_TN group.

Fig. 7
Fig. 7
Full size image

Differences in gene function between the HPV_PP and HPV_TN groups were determined using PICRUSt2. (a) The bar plots of KEGG categories showing significant differences between the HPV_PP and HPV_TN groups. (b) A cladogram illustrating the variations in gene function between the HPV_PP group (red) and the HPV_TN group (green).

Discussion

The vaginal microbiome is crucial for maintaining female reproductive health, with its composition significantly influencing susceptibility to various gynecological conditions, including HPV infections and associated diseases, such as CA. The interplay between the vaginal microbiota and HPV is particularly noteworthy, as disruptions in microbial balance can lead to an increased risk of HPV persistence and subsequent disease progression.

In our study, we identified specific microbial patterns linked to HPV infection and assessed how these patterns are related to clinical outcomes. The abundance of P. bivia, G. asaccharolytica, S. agalactiae and M. hominis was greater in the CA group compared to the CON group, with P. bivia being the most significant. Notably, the abundance of P. bivia also increased in the HPV persistent-positive (HPV_PP) group and the HPV-16/18 group. ROC analysis revealed that the relative abundance of P. bivia could predict persistent HPV infection, with an AUC of 0.744, indicating moderate discriminatory performance. Our findings indicated significant alterations in alpha diversity, composition and functional relative abundance of the vaginal flora in females with CA compared with those in healthy controls. These findings shed light on the potential effects of the vaginal flora, especially P. bivia, on HPV infection and persistence.

Prevotella is a common vaginal commensal bacterium and its increased presence in the vaginal mucosa is linked to bacterial vaginosis. Bacterial vaginosis is associated with an increase in vaginal pH, increasing a person’s susceptibility to infection. Lipopolysaccharide and ammonia produced by P. bivia are part of the vaginal mucus24. It is also associated with the production of epithelial cytokines and promotes the growth of other vaginosis-associated bacteria, such as G. vaginalis, which in turn stimulates the growth of P. bivia25,26. According to previous studies, the overgrowth of Prevotella can produce polyamines, increase the vaginal pH, inhibit Lactobacillus growth and promote G. vaginalis growth27,28. Several studies have shown that women with Prevotella dominance secrete sialidase, collagenase and fibrinolysins, leading to increased activation of cervical TLR4, NF-κB, C-myc and hTERT expression and a stronger inflammatory response7,29,30,31,32. This microbial dominance may lead to a greater possibility of persistent HPV infection and the development of cervical lesions32,33,34. However, clear evidence is still lacking regarding the function of P. bivia in the development of CA, and thus we could not clarify the causality between them. However, as suggested by several previous studies, a vicious cycle in which a vaginal a microbiota imbalance may lead to diseases such as CA, while CA can change mucosal metabolism or the immune response of the host system in turn. To our knowledge, this study systematically analyzed the species-level changes in the vaginal microbiota of female patients with CA infected with lr-risk HPV or co-infected with hr-HPV, providing insights that more accurately reflect real-world clinical scenarios. Previous studies have focused on either the changes in the vaginal microbiota associated with different grades of intracervical neoplasia caused by high-risk HPV17,18 or the changes in the vaginal microbiota caused by a single low-risk HPV infection19. Moreover, these studies have primarily focused on revealing differential profiles of the vaginal flora and have not thoroughly investigated the clinical significance of the differential genera or species19. Notably, our study also incorporated various disease outcomes, revealing that the relative abundance of P. bivia in the CA group or HPV_PP group was much greater than that in the CON group or HPV_TN group. In addition, the AUC for the ability of P. bivia to predict whether the patient’s HPV status could revert to negative was determined to be 0.744. That is P. bivia could serve as a candidate noninvasive biomarker in future clinical practice; however, experiments on larger samples and functional experiments in animal models are still needed to confirm this hypothesis.

A. prevotii, under the phylum Firmicutes, is a common oral anaerobic bacteria35. It can produce virulence factors that enhance its ability to adhere to epithelial cells and evade the host immune response, thereby exacerbating the infection36. In our study, although A. Prevotii was positively correlated with HPV_PP and LSIL, no differential was observed in relative abundance between the CA and CON groups. Therefore, many more samples are needed to explore its effect on CA development and persistent HPV infection.

Gardnerella plays a multifaceted role in vaginal health, serving not only as one of the hallmark bacteria associated with BV but also potentially influencing the diversity and overall health status of the vaginal microbiota37. It has historically been classified as a single species (G. vaginalis); however, recent studies have identified new species: G. leopoldii, G. picketii, G. greenwoodii, G. piotii and G. swidsinskii38,39. In addition, it was revealed that the coexistence of multiple clades of Gardnerella may contribute to BV and preterm birth by augmenting microbial load or enhancing synergistic pathogenicity. However, in our study, we failed to detect the relative abundances of these newly identified species of Gardnerella using 16S V3-V4 rRNA sequencing. Although in the HPV_TN group, Gardnerella and G. vaginalis was greater than that in the HPV_PP group, the differences was not significant. It is necessary to underscore the critical importance of undertaking a comprehensive investigation into the diversity of Gardnerella in our future studies by using full-length 16S rRNA or shotgun metagenomics.

In addition, this study reveals the functional profiles of the vaginal microbiota, providing valuable insights into the molecular mechanisms that may be related to HPV infection and persistence. The functions of necroptosis, cell growth and death, and the MAPK signaling pathway were altered in the HPV_PP group, suggesting a hypothesis that the vaginal microbiota, especially P. bivia and A. prevotii, may influence the proliferation and necroptosis of HPV-infected cells through these pathways. However, further functional gain- and loss-of-function experiments are needed to establish a causal relationship between P. bivia or A. prevotii and persistent HPV infection. The significant enrichment of bacteria such as P. bivia and A. prevotii in the HPV-positive group suggests a complex interplay between these microbes and the host’s immune response, potentially facilitating the ability of HPV to establish and maintain infection. The link between these bacterial species and immune modulation emphasizes their potential as new biomarkers for predicting HPV infection outcomes and guides therapeutic strategies focused on restoring a healthy microbiome. Moreover, the predicted signaling pathways implicated in persistent HPV infection, especially those related to drug resistance and cellular survival, open up exciting avenues for future research.

The limitations of this study warrant careful consideration. First, the relatively small sample size may limit the generalizability of the findings, as larger cohorts are often necessary to establish robust associations between vaginal microbiota composition and HPV infection status. Second, as mentioned above, 16S V3-V4 rRNA sequencing is highly conserved across multiple species of Gardnerella, and we should characterize these vaginal flora more precisely by using full-length 16S rRNA or shotgun metagenomics in our future studies. Furthermore, the absence of longitudinal data restricts our ability to assess temporal changes in the microbiota and their potential implications for HPV persistence and disease progression. Additionally, the lack of experimental validation of microbiota functions leaves a gap in understanding the specific mechanisms through which these microbial communities may influence host-pathogen interactions. Potential batch effects during sampling could also introduce variability that may confound the results. Finally, although the factors of age, BMI and marital status were matched between the CA group and the CON group, a history of contraceptives, sexual practices, and intimate hygiene also can affect the vaginal microbiome, introducing bias into the results. Addressing these limitations in future studies will increase the reliability and applicability of findings concerning the intricate interplay between the vaginal microbiota and HPV infection. However, despite these limitations and flaws, we believe that this study has substantial importance, as it is the first to analyze the role of specific microbial species, especially P. Bivia, in HPV infection and CA.

Materials and methods

Participants and specimens

This study was approved by the Institutional Review Board of the First Affiliated Hospital, School of Medicine of Zhejiang University under protocol numbers 2020-057. All methods were performed in accordance with the relevant guidelines and regulations. Written informed consent, was obtained from all participants. CA patients (between January 2021 and June 2022) were diagnosed clinically on the basis of their morphological characteristics, intravaginal location, and positive HPV genotyping. Other sexually transmitted infections (STIs), including HIV, syphilis, Ureaplasma urealyticum, Mycoplasma hominis, Chlamydia trachomatis, and Neisseria gonorrheae, were also screened. All patients reported no previous medical history of STIs. Patients with tumors or other immunocompromised diseases and those who recevied antibiotics within 3 months were excluded. Age-, BMI- and marital status-matched healthy controls were chosen from females who underwent routine health checkups in May and June 2022. These healthy females were confirmed to have no STIs by screening, as mentioned above, as well as by HPV genotyping. Vaginal secretion samples from the CA group (CA) and control group (CON) were collected, immediately preserved in DNA stabilization solution (TinyGene Bio-Tech Co., Ltd., Shanghai, China) and subsequently stored at −80 °C within 15 min. Among the 63 CA patients, 22 patients chose three rounds of photodynamic therapy (PDT, once a week), and the others opted for CO2 laser or cryotherapy as their therapy. Cervical exfoliated cells before therapy and at the end of 2 years after the first visit were delivered for HPV genotyping using an HPV GenoArray Test Kit (Hybribo, China), including 18 high-risk HPV types—HPV 16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82, and 7 low-risk HPV types—HPV 6, 11, 42, 43, 44, 81 and 83. Based on the results of the HPV status, whether negative or persistent positive at the end of 2 years, the CA group was divided into the HPV_TN or HPV_PP group.

Vaginal secretion DNA collection and processing

All samples were collected and stored at −80°C until they were delivered to TinyGene Bio-Tech Co. in January 2023 for further processing. DNA extraction, PCR amplification, sequence processing, and bioinformatics analysis were performed following the protocols outlined in a previous study40. In brief, the proteinase-K cleavage method was used for DNA extraction from the vaginal samples. The specific primers 341 F 5’-CCTAYGGGRBGCASCAG-3’ and 806R 5’-GGACTACNNGGGTATCTAAT-3’ with barcodes were used for PCR amplification of the V3-V4 region of 16S rRNA. All PCRs were conducted in a total volume of 30 µL. The PCR products from different samples were indexed and mixed in equal ratios. They were subsequently sequenced on the Illumina NovaSeq 6000 PE250 (Illumina, San Diego, CA) platform using a dual-index sequencing strategy, following the recommended Illumina protocol at TinyGene Bio-Tech Co. Raw Illumina read data were deposited in the EMBL database of Sequence Read Achieve (accession number PRJNA1033799). The OTU data and taxonomic classifications for each sample are provided in Supplementary Table 5.

Bioinformatics analysis and statistic analysis

A combination of software [mothur (version 1.39.5)41UPARSE (v8.1.1756)42and R (version 3.6.3), refer to https://intro2r.com/citing-r.html] was used to analyze the 16S rRNA amplicon sequences. After quality control, the demultiplexed reads were clustered at 97% sequence identity into operational taxonomic units (OTUs), and the chimera were removed, and the singleton OTUs were deleted using the UPARSE pipeline (https://drive5.com/usearch/manual8.1/uparse_pipeline.html, default parameters). The representative OTU sequences were assigned for taxonomy against the Silva 138.1 database43 with a confidence score ≥ 0.6 by the classify.seqs command in mothur.

The Student’s t-test was applied to investigate principal coordinates and alpha diversity (Shannon, Chao1, ACE, Simpson index and PD-whole tree indices) between the CA group and CON group or among subgroups. PCA was conducted utilizing the prcomp function from the stats package (version 3.6.3) in the R package (version 3.6.3), followed by visualization using the ggplot2 package (version 3.4.2). To assess the beta diversity, the Bray-Curtis and Jaccard distance matrices were calculated using the vegdist function from the vegan package (version 2.6-4, https://vegandevs.github.io/vegan/authors.html#citation), and then the PCoA was performed using the pcoa function from the ape package (version 5.7-1), which was visualized using ggplot2. Then, we conducted ANOSIM analysis using the anosim function in the vegan package of R (version 2.6-4) to test the significance of differences among a prior grouping strategy based on the PCoA result and visualized the results with ggplot2 (version 3.4.2). For the specific species, the Wilcoxon rank-sum test was used to compare the relative abundance between the CA and CON groups and among the subgroups44. OTUs were aligned utilizing the built-in reference database from the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2)45employing default parameters. Additionally, LEfSe version 1.046 was utilized to determine potential significant associations of the taxonomic composition and KEGG22,23 (https://www.kegg.jp/kegg/docs/relnote.html) modules. The Kruskal-Wallis test was used to control the false discovery rate (FDR) during multiple hypothesis testing in LEfSe, and a P < 0.05 and |LDA score| > 2 were used to screen the differential species or functions between groups. A P value less than 0.05 was considered to indicate statistical significance.

Bioinformatic analysis for the advanced correlation link and the ROC curve was performed at https://www.omicstudio.cn/tool21. The advanced correlation link was analyzed by Mantel’s correlation test using the R package (version 4.1.3) followed by visualization using the ggplot2 package (version 3.4.0). A pROC package in R which employs a generalized linear model (logistic regression) by default, was used to conduct the ROC analysis47. The relative abundance of P. bivia and A. prevotii as continuous predictor variables, and HPV status (HPV_TN or HPV_PP) as the binary response variable were uploaded to https://www.omicstudio.cn/tool/58. The ROC curve was generated by plotting the true positive rate (sensitivity) against the false-positive rate (1-sensitivity), and the AUC was subsequently calculated48.