The Correlation Between Bacterial Vaginosis, High-Risk Human Papillomavirus Infection, and Cervical Intraepithelial Neoplasia Progression

Persistent infection with high-risk human papillomavirus (HR-HPV) is an important reason for the progression of cervical intraepithelial neoplasia (CIN). Bacterial vaginosis (BV) is a genital infection that frequently presents in women infected with HPV, but the correlation between BV and HPV during CIN development is still elusive. In this study, we enrolled 624 participants and obtained 423 samples of vaginal secretions from them, including 193 HPV-negative samples and 230 HPV-positive samples. We used 16S rRNA sequencing to measure the vaginal microbiota diversity in women with or without BV and HPV co-infection and then calculated risk factors for CIN progression by logistic regression. We found that condom use (OR=3.480; 95% CI=1.069-11.325; P < .05) was a protective factor against CIN, whereas BV (OR=0.358; 95% CI=0.195-0.656; P < .05) and HR-HPV infection (OR= 0.016; 95% CI=0.004-0.072; P < .001) were risk factors for CIN. BV and HPV infection could trigger an increase in the diversity of vaginal microbiota and decrease Lactobacillus domination, which is conducive to CIN regression. The depletion of the carbohydrate metabolism pathway may induce Lactobacillus reduction. Treating BV in the clinical setting could block CIN development and L. iners may be a crucial species during this process. and


Introduction
Cervical cancer affects millions of women worldwide, with an estimated incidence of 604 200 new cases and 310 000 deaths in 2020, with China having 109 700 (18.2%) new cases and 59 700 (17.6%) deaths.
[1] The prevalence of cervical cancer in China remains higher than in the Western World due to lower vaccine coverage among young women and less awareness of regular cervical cancer screening in adult.

Persistent human papillomavirus (HPV) infection causes cervical intraepithelial neoplasia (CIN) that may
develop into cervical cancer if not detected in its early stages during screening or treated well once detected. According to a systematic review by Yin et al, [2] the rate of high-risk HPV (HR-HPV) infection caused by types 16, 52, 58, 53, and 18 is up to 19.0% in mainland Chinese women. Infection with HR-HPV is an independent risk factor for CIN and cervical cancer. HPV infection is the most common sexually transmitted infection (STI) especially in women with multiple sex partners. [3] Other factors like smoking, contraception and antibiotic use can disrupt the clearance of HPV leading to a high viral load that can trigger cervical lesions formation. [4] Bacterial vaginosis (BV) occurs in the reproductive age group with 15-50% of women presenting with vaginal discharge, peculiar smell, itching, or increased vaginal pH levels. [3] There is no standard de nition of BV, but it is widely recognized that its development is a process characterized by the dysregulation of vaginal microbiota with increased levels of Gardnerella vaginalis (G. vaginalis), Atopobium, Prevotella, Sneathia, Peptostreptococcus, Megasphera, BV-associated bacterium 1 (BVAB1) to BVAB3, and decreased levels of Lactobacilli spp. [5,6] The classi cation of community state types (CSTs) summarizes ve types of vaginal microbiota in healthy women based on different domination of bacterium. CST I, II, III, V are dominated by Lactobacillus crispatus (L. crispatus), L. gasseri, L. iners, and L. jensenii, respectively; whereas CST IV is de ned as the depletion of Lactobacilli spp. [7][8][9] Women with CST IV vaginal communities are more likely to experience BV, abortion, preterm labor, increased susceptibility to human immunode ciency virus (HIV) and HPV infection, or delayed HPV infection clearance. [5,10] The percentage of women with CST IV cluster is signi cantly higher in Black and Hispanic women who also have a higher incidence of cervical cancer than that in Asian and White women. [11] The association between BV and CIN was described 30 years ago, but the mechanism remains elusive. Some research revealed that BV is an independent risk factor for persistent HPV infection. Greater microbiota species diversity is observed in women infected with HPV. The abundances of bacteria associated with BV like Sneathia, Prevotella, and Megashaera are signi cantly higher in women who are positive for HPV.
[6] Previous research focused on clarifying the association between the cervical cytological lesion and BV due to inadequate technological development. With the popularity of colposcopy, recent studies have relied mainly on biopsy under colposcopy, which is more accurate than a thin prep cytology test (TCT), to determine the stage of cervical lesions.
This study is focused on documenting the association of BV, CIN, and HR-HPV infection. We used 16S rRNA gene sequencing to display the vaginal microbiota in Chinese women and compared the composition of microbiota between women with different BV, HPV, or CIN statuses. We aimed to provide some evidence for the importance of managing BV clinically to regress persistent HPV infection that can lead to CIN development.

Participant Enrollment
This study was approved by the Ethics Committee of the Aviation General Hospital, Beijing, China (Ethical approval No. 2021-KY-01-02). Written informed consent was obtained from all participants and all methods were performed in accordance with the guidelines and regulations. Women attending the Aviation General hospital, located in Beijing, China were recruited in our study. The inclusion criteria included: (a.) age range from 25 to 65 years; (b.) untreated BV; (c.) voluntary signing of informed consent form and questionnaire; (d.) no plans to move out of Beijing for two years; and (e.) willingness to be followed up. The exclusion criteria included: (a.) current pregnancy; (b.) undergoing menstrual period; (c.) vaccinated against HPV; (d.) sexual activity, vaginal irrigation or drug application performed 48 hours before sampling; (e.) diagnosed with vulvovaginal candidiasis (VVC), trichomonas vaginitis, or other STIs like Neisseria gonorrhoeae, Chlamydia trachomatis, human immunode ciency virus (HIV), hepatitis B virus (HBV) or Treponema pallidum; (f.) accompanied by hypertension, diabetes, immune system disease or other diseases; and (g.) antibiotics use in the past 30 days.

Sample Collection
Participants were put in the lithotomy position on the gynecological examination bed after signing the informed consent form and nishing the questionnaire. A well-trained clinical doctor swabbed the vagina using two disposable sterile swabs to collect cervical and vaginal secretions that were then stored in 2 ml saline and 2 ml phosphate buffer saline (PBS, HyClone, USA) tubes, respectively. Samples were kept on ice and transferred to the laboratory for subsequent testing.

BV, HPV, and CIN Diagnosis
An experienced doctor made a diagnosis of BV if a patient met three out of the following four Amsel's criteria: (a.) increased, thin, and homogeneous vaginal discharge; (b.) vaginal pH greater than 4.5; (c.) presence of clue cells; and (d.) amine odor when potassium hydroxide (KOH) was added to the vaginal secretions. The Hybrid Capture 2 assay (HC2) was used to detect seventeen HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82) and six low-risk types (6, 11,42,43,44,81). Women who were positive for HR-HPV also accepted TCT to de ne the cytological lesion. If the results were atypical squamous cells of undetermined signi cance (ASC-US) or more severe lesions such as low-grade squamous intraepithelial lesions (LSIL) or high-grade squamous intraepithelial lesions (HSIL), they were recommended for colposcopy to de ne the grade of CIN.

DNA Extraction
FastDNA Spin Kit (MP Biomedicals, USA) was used for DNA isolation from vaginal secretions. We added 200 µl of the samples suspended in saline and 1 ml cell lysis solution (CLS-TC) to Lysing Matrix A (FastPrep) and then followed the protocol. DNA concentration and purity was monitored on 1% agarose gel. According to the concentration, DNA was diluted to 1 ug/ µl using sterile water.

Sequence Analysis
Paired-end reads was assigned to samples based on their unique barcode and truncated by cutting off the barcode and primer sequence. Paired-end reads were merged using FLASH (V1.2.7, http://ccb.jhu.edu/sofeware/FLASH). [31] Quality ltering on the raw tags were performed under speci c ltering conditions to obtain the high-quality clean tag according to the QIIME (V1.9.1, http://qiime.org/scripts/split libraries fastq.html) quality controlled process. [32,33] The tags were compared with the reference database (Silva database, https://www.arb-silva.de/) using UCHIME algorithm (UCHIME Algorithm, http://www.drive5.com/usearch/manual/uchime_algo.html) to detect chimera sequences, and then the chimera sequences were removed. [ Figure 2C, the abundance of Lactobacillus spp. in the HPV-negative group was much higher than that in the HPV-positive group.
Meanwhile, microbiota composition in the HPV-positive group was more complicated. Bacteria such as Gardnerella vaginalis, Ralstonia pickettii, Streptococcus anginosus, Prevotella bivia, Prevotella timonensis, Bi dobacterium dentium, and Sneathia sanguinegens presented more frequently in the HPVpositive group. We analyzed the OTUs numbers in each group and displayed them in a Venn graph in Figure 2D. The unique OTUs in the HPV-positive group were signi cantly higher than in the HPV-negative group. These data suggest that HPV infection could lead to an increase in vaginal microbiota complexity.

The Relative Abundance of Vaginal Microbiota of Different BV, HPV and CIN Statuses
To further estimate the impact of BV and HPV infection on vaginal microbiota changes during CIN progression, we divided all the samples into 6 groups according to the following BV, HPV or CIN statuses:

Normal, BV (BV infection), HPV (HR-HPV infection), B.H (BV and HR-HPV co-infection), H.C (HR-HPV infection combined with CIN), and B.H.C (BV and HR-HPV co-infection combined with CIN). To better
distinguish the differences between groups, we conducted random screening of the grouped data to balance the sample size of each group. A total of 342 samples were collected and the numbers of nal samples of each group were 65, 52, 71, 60, 48, and 46, respectively (Supplementary material S1, Database 1, Table 2, 3, 4, 5, 6 and 7 for detailed information). In Figure 3A, we displayed the top 19 vaginal microbiota at the species level. Lactobacillus spp. had the highest percentage relative abundance of up to 82.6% in the normal group. Women with HPV infection only showed little change in the Lactobacillus spp. genera in their vaginal microbiota, but the percentage of L.crispatus, which presented most in a healthy vaginal environment was partly replaced by L.iners. Lactobacillus spp. and lost its dominance (with less than 60% in composition) in the remaining three groups of women, who were all diagnosed with BV infection with a more complex bacterial composition.
We de ned the community state types (CSTs) according to the dominant (> 60% of relative abundance in one sample) Lactobacillus species type. CST I, II, III, V were dominated by L. crispatus, L. gasseri, L. iners, and L. jensenii, respectively, whereas CST IV was de ned as the depletion of Lactobacillus spp. Surprisingly, the CST II cluster was only shown in the normal group and CST V in the BV group. The transformation of CST I to CST III was most apparent in the HPV-positive group, which indicated that L.iners may be a key species during HPV infection. Women diagnosed with BV infection were more likely to be de ned as being in the CST IV cluster. Interestingly, the percentages of CST IV cluster among BV and HPV co-infected women with CIN were signi cantly lower than those in the B.H group, and CST III cluster percentage was increased, which also indicates the importance of L.iners.

Higher Microbial Diversity in Women with BV and HPV Co-infection
The observed species, Shannon diversity, and Chao1indexes were used to measure the alpha diversity of microbiomes in 6 groups. High variability was observed among the samples; thus outliers were not The beta diversity was measured with the PCoA based on the Bray-Curtis dissimilarity. All the samples were mainly separated into three groups related to CST clusters [P < .001 ( Figure 4D)]. P values were calculated by the ANOSIM test. Biodiversity was not dependent on BV, HPV, or CIN status (supplementary Figure 1).

Important Phylotype in CIN Progression
To nd out which bacteria are responsible for the community differences between groups, we used LEfSe to analyze the bacterial diversity and nd biomarkers. It combines statistical difference analysis and the impact score of the species on the grouping results while emphasizing statistical signi cance and biological relevance. Figure 5A and 5B mainly show the species with signi cant differences in linear discriminant analysis (LDA) score greater than 3.5, which is the marker with a statistical difference. Like a previous study, we found a signi cant enrichment of Lactobacillus genera in the normal group compared with the group with BV or HPV-infected women. Anaerobic bacteria associated with BV infection such as the Prevotellaceae family, Streptococcaceae family, Atopobiaceae family, and Enterobacteriaceae family were signi cantly different among BV and HPV co-infected women compared with those in the normal group. These data suggest that CIN development could not be attributed to a single species but the result of multiple species interaction.
Gene Functional Pathways of Bacterial Taxa Associated with BV, HPV or CIN To better understand the bacterial function during disease progression, we also explored microbiota function using PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States). [12] Gene we detected were matched with the KEGG (Kyoto Encyclopedia of Genes and Genomes) database, predicted raw data was available in Supplementary S4, Database 3. [13,14] The KEGG pathway we tested was presented in supplementary Figure 2. There were many non-human gene pathways in different groups including pathways related to metabolism, genetic information processing, environmental information processing, cellular processes, and organismal systems. There were 33 pathways changes in the B.H.C group compared with the ones with healthy women (P < .05). Pathways related to amino acid, energy, cofactors, vitamin metabolism, biosynthesis of other secondary metabolites, folding, sorting and degradation, and endocrine systems were enriched in the B.H.C group women, whereas carbohydrate metabolism (glycolysis/gluconeogenesis, galactose, and glycerol-lipid metabolism) pathways were depleted in the B.H.C group (Fig 6). Depletion of membrane transporters and carbohydrate metabolism pathways was more frequent in the diseased versus normal groups, which may indicate that bacterial functions were mainly associated with these two pathways.

Discussion
Studies on the association of vaginal microbiota, HPV infection, and CIN progression have been a hotspot since gene sequencing technology was developed. Vaginal microbiota changes such as depletion of Lactobacillus spp. and increased microbiota diversity are a well-accepted nding in women infected with HPV. [15] We found that the vaginal microbiota in women with BV or HPV infection was more complex and had a reduction of Lactobacillus spp. (Figure 2). This is in line with the current research trends in China and abroad. Condom use (OR=3.480; 95% CI=1.069-11.325; P < .05) was a protective factor, whereas BV (OR=0.358; 95% CI=0.195-0.656; P < .05) and HR-HPV infection (OR=0.016; 95% CI=0.004-0.072; P < .001) were risk factors for CIN in our survey ( Figure 1) and other research studies too. [4,[16][17][18] The prevalence of smoking and combined oral contraceptive (COC) use in Chinese women is lower than in western countries, so we could not make any correlation between them and CIN progression in our cohort.
This study was a cross-sectional analysis based on a large sample size of Chinese women focused on the association of vaginal microbiota disturbance and HPV infection. A surprising nding in our survey was the CST cluster distribution. CST II and V rarely appeared in our cohort especially after HPV infection with or without BV ( Figure 3B), which was not consistent with foreign studies done in White, Hispanic, and Black women. [11,19,20] This may be because of the ethnic and geographic diversity. [21] The distribution of CST IV cluster in women with CIN was slightly lower than expected probably due to women with severe BV or VVC in whom a colposcopy was not done to prevent the spread of infection. [22,23] Drugs used to treat the vaginitis may have inhibited the colonization of the vagina by anaerobic bacteria and candida, thus the relative abundance of BV associated bacteria and microbiota diversity in the H.C or B.H.C group (Figures 3A and 4) was slightly lower than in the BV group. The actual relative abundance and diversity could be higher than the available data, but we could conclude that BV-free women (H.C group) have signi cantly lower vaginal microbiota diversity than the B.H.C group, which is probably conducive to regression of CIN.
Additionally, women infected with HPV have a higher proportion of L.iners than healthy women, which led us to focus on the characteristics of this special Lactobacillus spp. Several studies reveal that L. iners is not only present in the vaginal ecosystem of healthy women, but also of those in a transitional stage or BV state. [24] Genome analyses show the different strains and complicated characteristics of L. iners due to the different morphology and Gram-staining properties. [25 -28] In 2011, Macklaim et al. [29] used whole-genome sequencing to de ne the smallest species of L. iners, L. iners AB-1, which is present in both healthy women and those treated with antimicrobials. One of the reasons for its persistence in the vaginal epithelia despite the development of BV and treatment with antibiotics is the presence of bronectin (Fn)-binding adhesins. [30] The percentage of samples dominated with L. iners was higher in the B.H.C group than in the B.H group (Fig. 3), which makes us question whether L. iners AB-1 was present in our sample. Further research is necessary to detect L. iners AB-1 in abnormal vaginal microbiota ora.
Our study was a cross-sectional analysis conducted in one hospital, which may have limited the observation of the dynamic changes of vaginal microbiota after HPV infection. We chose two specialists to diagnose BV to minimize the misdiagnosis rate, but it is hard to avoid missing the diagnosis of asymptomatic or molecular BV and reporting bias. We only performed 16S rRNA gene sequencing in our study, which failed to distinguish L. iners AB-1 from L. iners in samples. Therefore, further study is needed to better understand vaginal microbiota changes in HPV persistence and CIN progression.
The strengths of this study include the large sample size, which was enough to organize the samples into 6 groups. We also did functional prediction of bacteria, which provides a reference for further mechanism research. We found that BV and HPV co-infected women did not have a higher diversity of vaginal microbiota unless CIN occurred. For future studies, short-, medium-and long-term follow-ups are needed to observe dynamic changes of the vaginal microbiota and the disease process in women with BV or HPV infection. Further investigation is also needed to understand the characteristics of L. iners in both healthy and unhealthy states.
In conclusion, BV and HPV infection could trigger an increase in the diversity of vaginal microbiota, especially BV, which could decrease Lactobacillus spp. domination, which is not conducive to CIN regression. The depletion of carbohydrate metabolism pathways may induce Lactobacillus spp. reduction. Regulating BV in the clinical setting may block CIN development and L. iners may be a crucial species during this process. Odds Ratio of Possible Risk Factors for CIN Using Logistic Regression. Bars represented 95% con dence intervals. * P < .05 circles and overlapping parts represents the number of OTUs shared between the groups and the number of non-overlapping parts represents the number of unique OTUs of the group. P: HPV-positive; N: HPVnegative.  Predicted Bacterial Function Pathways Differences between the B.H.C and Healthy Women Groups. PICRUSt algorithm was used to predict bacterial gene functions from the KEGG annotated databases.
The bar chart on the left represents the percentage of the abundance of a certain metabolic pathway from all metabolic pathways in the two groups of samples. Corrected P values are on the right side. T-test (two-tailed) was used to calculate the P-value.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. S1Database1.xlsx S2Database2amdTable.zip S4Database3.txt SupplementaryFigure1.pdf