The Value of Vaginal Microbiome in Patients with Endometrial Hyperplasia

Objective To investigate the profiles of the vaginal microbiome in patients with endometrial hyperplasia and to explore the potential value of vaginal microbiome in the diagnosis of endometrial hyperplasia. Materials/Methods. 26 patients suffering from abnormal uterine bleeding (AUB) with thickened endometrium revealed by transvaginal ultrasonography were enrolled. Based on pathology, 12 patients with endometrial hyperplasia were classified as the Veh group and 14 patients with proliferative endometrium were classified as the Vne group. The vaginal samples were collected for the presence of microbial DNA by high-throughput next-generation sequencing of the 16S rRNA gene. The α-diversity and ß-diversity of vaginal microbiome were analyzed and compared between bacterial populations. The ROC curve was made to evaluate the feasibility of flora as a biomarker. Results The diversity of vaginal microbiome in the Veh group was significantly lower than that in the Vne group (P < 0.05). Lactobacillus was the most represented genus in the Veh group. The study's t-test between the two groups showed that Lactobacillus has the only significant difference in the abundance of the first 15 genera (P < 0.01). ROC analysis of the abundance of Lactobacillus showed that the area of AUC was 0.83, the sensitivity was 93.00%, and the specificity was 75.00%. Conclusion The study offers insight into the nature of the vaginal microbiome and suggests that surveying the vaginal microbiota might be useful for detection of endometrial hyperplasia.


Introduction
Endometrial hyperplasia (EH) is a common gynecological disease, which can progress or occur at the same time as endometrial carcinoma [1,2]. Sensitive and accurate diagnosis of true premalignant endometrial lesions and treated appropriately are required. e evaluation should include clinical documentation and transvaginal ultrasonography (TVS). e first two steps only evoke a clinical suspicion of endometrial hyperplasia and then obtain histopathological results through invasive procedures for the definite diagnosis. However, it is necessary to make clear that ultrasonographic measurements are less accurate in many conditions, such as obesity. For some patients, endometrial sampling is overtreatment.
Endometrial hyperplasia results from the continuous hyperestrogenic effect and lack of progesterone protection [3,4]. It has reported that there is a significant correlation between estrogen and vaginal microbiome by University of Arkansas for Medical Sciences' researcher [5]. Unfortunately, few studies on vaginal microbiome in patients with EH, especially the characteristics of EH vaginal flora, are still unclear. e purpose of this study is to explore the character and to access the value of vaginal microecology in patients with endometrial hyperplasia. thickness" revealed by TVS in the Department of Obstetrics and Gynecology of Shanghai Pudong Hospital from September 2017 to January 2020. e pathological results were obtained by diagnostic curettage or hysteroscopy, including 14 patients with proliferative endometrium as the Vne group and 12 patients with endometrial hyperplasia as the Veh group (simple hyperplasia, n � 4; complex hyperplasia without atypia, n � 3; and complex atypical hyperplasia, n � 5). e pathologic diagnosis for all patients was performed by two experienced gynecologic pathologists. e Ethical Committee of Shanghai Pudong Hospital approved the study protocol, and informed consent of all individual participants was obtained in the study. e inclusion criteria were as follows: (1)  e V3-V4 hypervariable regions of the bacteria 16S rRNA gene were amplified with primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) by the thermocycler PCR system (GeneAmp 9700, ABI, USA). e PCR reactions were conducted using the following program: 3 min of denaturation at 95°C, 27 cycles of 30s at 95°C, 30 s for annealing at 55°C, and 45 s for elongation at 72°C, and a final extension at 72°C for 10 min.

Processing of Sequencing Data.
Raw fastq files were quality-filtered by Trimmomatic and merged by FLASH with the following criteria. (i) e reads were truncated at any site receiving an average quality score <20 over a 50 bp sliding window. (ii) Sequences whose overlap being longer than 10 bp were merged according to their overlap with mismatch no more than 2 bp. (iii) Sequences of each sample were separated according to barcodes (exactly matching) and primers (allowing 2 nucleotide mismatching), and reads containing ambiguous bases were removed. Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff using UPARSE (version 7.1 http://drive5.com/uparse/) with a novel "greedy" algorithm that performs chimera filtering and OTU clustering simultaneously. e taxonomy of each 16S rRNA gene sequence was analyzed by the RDP classifier algorithm (http://rdp.cme.msu.edu/) against the SILVA (SSU132) 16S rRNA database using confidence threshold of 70%.

Data Processing.
e original image data (raw data) obtained from the second-generation sequencing is converted into sequence data by base recognition (base calling), the sequencing sequence is controlled by Trimmomatic software, spliced by FLASH software, OUTanalysis is carried out by UPARSE software, taxonomic analysis of OUT representative sequence is carried out by the RDP classifier Bayesian algorithm, and OUT is compared to SILVA database by Mothur analysis flow. Annotate the species (the threshold is set to 0.7). e difference of alpha diversity between groups was tested by the Shannon algorithm and Simpson algorithm, and the species composition was analyzed by R language based on the data table in tax summary folder. e diversity of beta was analyzed by NMDS (nonmetric multidimensional scaling analysis) statistical analysis of R language, the difference of bacteria between the two groups was tested by ANOSIM values based on Bray-Curtis dissimilarity at the genus level, and sequence data were mainly analyzed using the QIIME, Mothur1.30.2, SILVA132, UPARSE 7.0.1090, USEARCH 7.0, RDP classifier, and R packages (v3.2.0). In addition, sequencing data were also analyzed using the free online Majorbio I-Sanger Cloud Platform (https://www.i-sanger.com). en, we performed differential abundance analysis to distinguish which taxa contributed to the validated microbiome structural changes in the vaginal microbiome of the Veh group. e P value (P) < 0.05 was considered to reflect a statistically significant difference. e receiver operating characteristic (ROC) curve and area under the curve (AUC) were analyzed by R packages (v3.2.0) (R Development Core Team, Vienna, Austria).

Statistical Analysis.
e data were statistically analyzed by SPSS version 21.0 for Windows (SPSS Inc., Chicago, IL, USA) in which the data of age, body mass index (BMI), and endometrial thickness were in accordance with normal distribution, expressed by mean ± standard deviation (means ± SD), and compared between groups by the t-test.
e counting data of menopause, hypertension, and diabetes were expressed by percentage, the chi-square test was used for comparison between groups, and the rank-sum test was used for the number of births and abortions. e difference was statistically significant (P < 0.05).

Baseline Characteristics.
e age of the enrolled patients ranged from 31 to 62, with Vne patients between 38 and 62 years old (mean age:47.71 ± 6.78 years, mean ± standard error), Veh between 31 and 57 years old (mean age: 45.17 ± 6.21 years, mean ± standard error). e Veh group consisted of 4 simple hyperplasias, 3 complex hyperplasias without atypia, and 5 complex atypical hyperplasias; 14 proliferative endometriums were sorted as the Vne group. e patients`clinical data are listed in Table 1. e mean age, the case of menopause, hypertension and diabetes, BMI index, times of gravida and parity, Histotype and endometrial thickness between the Vne group and the Veh group were not statistically significant.

Sequencing Information.
In this experiment, a total of 1,337,846 high quality gene sequences (31,582-72,918) were obtained, with an average of 51,455 reads per sample. A total of 1,250 OTUs and 712 OTUs were detected in the Vne and Veh groups, respectively. Both groups shared 453 OTUs. A total of 516 genera of bacteria were detected in all vaginal samples, including 305 genera in the Vne group and 279 genera in the Veh group. Both groups shared 68 genera.

Difference of Bacterial Community between the Two
Groups. We first compared the overall microbiota structure between disease states by analyzing the α-diversity and ßdiversity.
e α-diversity (Shannon, Simpson, and Heip indices of the vaginal microbiota in OTU and genus level) in the Veh group was significantly lower (P < 0.05) than that of the Vne group (Tables 2 and 3). Significant differences were also found in ß-diversity based on the Bray-Curtis though NMDS was based on genus level (P � 0.008) between the Vne and Veh groups ( Figure 1). P values are reported combining the evidence across the Bray-Curtis, ANOSIM, and multiple displacement amplification: 999(stress: 0.138, R � 0.214, P � 0.008). NMDS is the evaluation of the rank information of the distance value. NMDS1 and NMDS2 axes do not have the weight of meaning.
e overall dimensionality reduction effect of NMDS is judged by the stress value.
Our data showed that the vaginal microbial community diversity and community evenness were significantly lower in the Veh group than in the Vne group in OTU and genus level.
We second conducted flora structure constituting ratio analysis.
e species composition of phylum in the Vne group from high to low is Firmicutes: 41.01%, Actinobacteria: 33.34%, Bacteroides: 15.59%, and other bacteria accounted for 10.05%. e phylum of bacteria in the Veh group is the same as that in the Vne group, but the constituent ratio is different. Among them, the proportion of Firmicutes increased to 64.92%, Actinobacteria decreased to 22.63%, Bacteroidetes decreased to 6.34%, and other bacteria accounted for 12.45% (Figure 2).
To get more insight into the characteristics of the patient's microbiome, we conducted a differential analysis of microbial abundance. In class, order, family, and genus levels, the Bacilli, Lactobacillales, Lactobacillaceae, and Lactobacillus were significantly higher in the Veh group than in the Vne group. After the overall microbiome assessment, only Lactobacillus has statistically significant different abundances of the top 15 bacterial genera. e abundance of Lactobacillus in the Veh group was significantly higher than those in the Vne group (P < 0.05) (Figure 3). rough the ROC prediction analysis of the patients with thickened endometrium with the abundance of Lactobacillus by R packages (v3.2.0), the ROC curve was obtained, which showed that the AUC area was 0.83, the sensitivity was 93.00%, and the specificity was 75.00%, as shown in Figure 4

Discussion
Here, we present a pilot high-throughput microbiome assessment of the female vaginal of patients diagnosed with a benign uterine condition (thickened endometrium). e dominant taxa in the vaginal microbiome were Lactobacillus, Gardnerella, and Prevotella, consistent with current vaginal microbiome literature [6]. Gardnerella was the dominant flora in the proliferative endometrium group, with uniform distribution of vaginal microorganisms and rich microbial diversity. e dominant vaginal flora of patients with endometrial hyperplasia was Lactobacillus, and the diversity of bacteria decreased significantly. e patients with endometrial hyperplasia were affected by high concentration of estrogen for a long time, which caused vaginal mucosal edema and increased vaginal mucosal permeability. Estrogen promotes the growth of epidermal cells and the increase of intracellular glycogen in the upper part of the vagina, which leads to the increase of Lactobacillus. ese bacteria ferment glycogen into glucose and finally transform into lactic acid, which maintains the low PH state of the vagina. At the same time, Lactobacillus inhibits the growth of other     bacteria by producing H 2 O 2 , which further strengthens the dominant position of Lactobacillus, resulting in a decrease in the diversity of bacteria [7]. In the past, we found only one study about the distribution of vaginal microorganisms in patients with endometrial hyperplasia. However, in contrast with our data, it reported that the lower genital tract microbiome structure of the hyperplasia cohort was not distinguishable from the benign cohort [8]. e vaginal microbiome of these hyperplasia patients resembled a benign microbiome signature. e discrepancy may be caused by the difference of sample sources, sample quantity, race, age, and menopausal years [9].
Our results indicate that endometrium hyperplasia can be distinguished by the vaginal microbial community diversity and community evenness. e vaginal α-diversity of the Veh group was significantly lower than the Vne group.
rough ROC curve analysis, we found that the abundance of Lactobacillus can be used as a landmark microorganism for differential diagnosis of endometrial thickening, with a sensitivity of 93.00% and a specificity of 75.00%. Our results suggest that the detection of Lactobacillus in the vagina is associated with the presence of endometrial hyperplasia. Since we do not have healthy asymptomatic patients in this research, we cannot assess whether this correlation exists or does it indicate illness status. e causal relationship needs further study. ere are few reports on the differential diagnosis of patients by microbiome identification [10]. It has an important value for avoiding medical overuse such as diagnostic curettage, hysteroscopy, and other invasive procedures and economic burden.
However, this study also has some defects. On the one hand, the small sample size of single center, and on the other hand, there are no confounding factors such as vaginal microbial PH value, menstrual interval, bleeding interval, estrogen level, menopausal time, uterine leiomyoma, adenomyosis, delivery mode, and so on. In addition, due to the heterogeneity and dynamics of vaginal microecology, longterm follow-up and microbial samples in different periods still need to be further studied.

Conclusions
We found a distinct microbiome signature in patients with endometrial hyperplasia. We have shown that the detection of Lactobacillus in the gynecologic tract was associated with the presence of endometrial hyperplasia in our study population. ese findings provide important insights into the etiology or manifestation of the disease with broad implications for biomarker development in the early detection of and screening for endometrial hyperplasia by noninvasive methods.
Data Availability e simulation experiment data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Hua Hu and Hu Zhang conceptualized and designed the study. Zhanpeng Zhu, Hu Zhang, and Qiao Feng involved in acquisition of sample. Qiao Feng and Hu Zhang analyzed the data and drafted the article. Haiyan Dai and Hua Hu revised it critically for important intellectual content. All authors approved the final version of the article to be published.