The salivary microbiota is altered in cervical dysplasia patients and influenced by conization

This study supports the correlation between the salivary microbiota and cervical dysplasia and suggests that smoking influences the salivary microbiota. Cervical cancer is the fourth most commonly detected cancer in women worldwide [1]. The World Health Organization recently launched a global strategy to accelerate the elimination of cervical cancer [2, 3]. Despite the availability of vaccines for human papillomavirus (HPV) infection, which is the primary cause of cervical cancer, it is important to note that the current vaccines do not cover all oncogenic HPV types [4-6]. HPV testing and cytology testing (e.g., Pap smear) currently serve as the early screening methods for dysplasia and cervical cancer, after which histology can accurately identify the occurrence and stages of dysplasia. HPV testing and cytology testing have reported a significantly increased false rate compared with invasive histology diagnostic testing [7]. Hence, it is worth searching for a potential novel biomarker to improve the accuracy of early detection. Early detection, broader surveillance, and increased access to medical care and treatment would greatly assist cancer prevention, especially in low-income countries [8]. Emerging evidence supports that vaginal microbiota correlate to HPV infection and dysplasia [9-12]. However, data related to oral or salivary microbiota and cervical dysplasia are still limited. Whether oral microbiota play a role at different cervical dysplasia stages and how lifestyle influences salivary microbiota are important to evaluate. Poor oral hygiene and tooth loss are associated with an increased risk of oral squamous cell cancer [13]. Several specific oral bacterial pathogens, including Fusobacterium, Campylobacter, Prevotella, Pseudomonas, and Capnocytophaga, are correlated with lung, oral, esophageal, stomach, pancreatic, and colorectal cancers [14-16]. Notably, as saliva collection is noninvasive and the procedure is quicker, cheaper, and more convenient for the patient as compared to invasive processes such as blood collection, cytology, and histology testing. The salivary microbiota also serves as a biomarker for different cancers and systemic diseases diagnosis [14-19]. Other known risk factors for cervical cancer include smoking, increased parity, and infection with the human immunodeficiency virus (HIV) [20]. Numerous studies have shown that smoking is a significant risk factor for developing cervical abnormalities, including cervical dysplasia and cervical cancer [21, 22]. Furthermore, lifestyles, such as diet, smoking, and alcohol intake, have been suggested to affect the oral microbiota and different diseases [23-27]. Herein, saliva samples were collected from both individuals undergoing cervical dysplasia examination and volunteers visiting the dental clinic for dental examinations. The study investigated the salivary microbiota and its potential relation to the occurrence of cervical dysplasia. Additionally, the study compared the salivary microbes that differed between volunteer controls and participants undergoing vaginal examination with and without dysplasia or pre- and postconization (Figure 1A). Furthermore, the lifestyle of participants through questionnaire data was investigated, and factors that may potentially influence oral microbiota were identified. Significantly decreased microbial community richness and microbial diversity were observed in participants with and without dysplasia when compared to the control group (Figure 1B, Supporting Information: Figure S1A). Moreover, significantly decreased microbial diversity was observed in participants of the predysplasia (+) and postdysplasia (−) groups and in participants from different histological groups compared to the control group (Figure 1C, Supporting Information: Figures S1B–D). A significant dissimilarity was observed when we compared the control group with the groups with and without dysplasia (PANOSIM = 0.005) and with the predysplasia (+) and postdysplasia (−) (PANOSIM = 0.002) (Figure 1D,E). However, no significant differences were identified among the control and different histological groups (Supporting Information: Figure S1E). All oral samples from vaginal examination participants tested HPV negative. The PERMANOVA analysis considered the overall effect of all participants' lifestyles on the microbiota beta diversity and found that age and smoking could influence microbiota diversity (Supporting Information: Tables S1 and S2, Supporting Information: Figure S2). The top five main genera and species found belong to Streptococcus, Prevotella, Veillonella, Haemophilus, and Actinomyces (Supporting Information: Figures S3 and S4). When comparing the salivary microbiota of participants with and without dysplasia with healthy participants, up to 16 different genera were identified as having significantly different relative abundance (Figure 1F). A similar trend was observed when comparing groups of predysplasia (+) and postdysplasia (−). The relative abundance of Haemophlius and Alloprevotella was significantly increased among saliva samples from vaginal examination participants than those from the healthy participants (Figure 1G). Based on the reconstruction of unobserved states 2 (PICRUSt2) analysis, the pathways of primary immunodeficiency, viral carcinogenesis, proteoglycans in cancer, and retinoic acid-inducible gene I (RIG-I)-like receptor signaling pathways were significantly changed when comparing the salivary microbiota predicted function of participants with and without dysplasia with healthy participants (Figure 1H), as well as comparing the pre-ysplasia (+) and postdysplasia (−) participants with healthy participants. The high-relative-abundance bacterial genera, such as Streptococcus and Veillonella, were negatively correlated to the pathways of primary immunodeficiency, viral carcinogenesis, and proteoglycans in cancer. While other genera, such as Alloprevotella, were positively correlated with the viral carcinogenesis and proteoglycans in cancer, and Actinomyces was positively correlated with RIG-I-like receptor (RLR) signaling pathways (Figure 1I). Furthermore, when comparing the control group with the group with or without dysplasia, Prevotella, Haemophilus and Alloprevotella were found to be significantly increased in the latter groups compared to the control group (Supporting Information: Figure S5A–B). Moreover, Haemophilus and Alloprevotella were the genera significantly higher among participants from the predysplasia (+) and postdysplasia (−) groups than the control group (Supporting Information: Figure S5C–D). The area under the curve (AUC) of Haemophlius (p = 0.0039), Alloprevotella (p = 0.0076), and Prevotella (p = 0.0350) reached 0.748, 0.730, and 0.682, respectively, which indicates that these three genera could support the diagnosis of patients with dysplasia from healthy condition (Supporting Information: Figure S5E–G). Based on the similarities in the genera identified from different samples, three different salivary microbial types were generated (Figure 2A). Type 2 contained a higher abundance of Prevotella and Actinomyces, which closely interacted according to our network analysis. Type 3 contained a higher abundance of Haemophilus, and Type 1 microbiota type contained a similar abundance of all the major genera (Figure 2B–D). Participants in the groups with and without dysplasia had higher percentages of type 2 and type 3 than the participants in the control group (Figure 2E). A similar trend was observed when comparing predysplasia (+) and postdysplasia (−) groups with the control group (Figure 2F) and when comparing low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) groups with the control group (Figure 2G). Moreover, a tightly interactive network was formed by Prevotella, Actinomyces, and Haemophilus, indicating a potential relationship between salivary microbiota and their distinct contributions to various diseases and treatment outcomes (Figure 2D). All the comparisons were further evaluated without the dental examination control group (Figure 3A–D and Supporting Information: Figure S6). PERMANOVA analysis was also carried out to consider the overall effect of the participants' lifestyles for salivary microbial composition (Supporting Information: Tables S3 and S4). Only microbiota beta diversity between the predysplasia (+) and postdysplasia (−) participants was significantly altered (Figure 3D). In addition, Actinomyces was significantly increased in the dysplasia group, and bacteria, including Aggregatibacter, was significantly increased in the group without dysplasia (Figure 3E). Further, in the comparison of the predysplasia (+) and postdysplasia (−) participants, Stomatobaculum was significantly increased among the predysplasia (+) participants (Figure 3F). Actinomyces was found to be correlated with increased RLR signaling, which is elevated in patients with dysplasia (Figures 3G,H). The diagnostic ability of oral Actinomyces abundance alone reached 0.700 (p = 0.0370), which indicates that Actinomyces could help to diagnose dysplasia patients from the ones without dysplasia (Figure 3I). Notably, combining cytology and HPV infection testing with the oral Actinomyces abundance, the diagnostic ability improved to the classification of AUC = 0.945 (p = 0.0003) compared to conventional cytology combined with HPV infection testing (AUC = 0.928, p = 0.0005) (Figure 3J–K). Since smoking significantly influenced the salivary microbial beta diversity, as revealed by the PERMANOVA analysis (Supporting Information: Tables S1 and S2), the differences in salivary microbiota alpha and beta diversity that were affected by smoking (both previous and current smokers) were further investigated (Supporting Information: Figure S7). No significant alpha diversity was observed between smoker and nonsmoking groups (Supporting Information: Figure S7A–D). However, the smokers had significantly altered salivary microbiota beta diversity when compared with participants who never smoked (Supporting Information: Figure S7E,F). Furthermore, the salivary microbial types distribution among smokers and nonsmokers was also analyzed. The smokers had higher percentages of type 2 (containing a higher abundance of Prevotella and Actinomyces) than the nonsmokers (Supporting Information: Figure S7G). Interestingly, when we investigated the bacteria that were significantly altered between smokers and nonsmokers, Actinomyces was once again significantly increased among smokers compared to nonsmokers (Supporting Information: Figure S7H,I). Our study found that participants who visited the hospital for a vaginal dysplasia examination had significantly decreased oral microbiota richness and diversity compared to control participants who visited for a dental examination. This is in line with the earlier studies that link a decreased oral microbiota diversity to the occurrence of different kinds of cancers, such as throat cancer, lung cancer, and head and neck cancer [28-31]. We also found that the salivary microbiota diversity of patients who had recovered from cervical dysplasia after conization differed from that of patients with cervical dysplasia before treatment. Notably, significantly distinguished microbiota beta diversities were identified among the control group, the groups with and without dysplasia, and the groups predysplasia (+) and postdysplasia (−). By using the eHOMD database, Streptococcus, Prevotella, Veillonella, and Haemophilus were among the most abundant genera, and Haemophilus parainfluenzae, Veillonella atypica, and Rothia mucilaginosa were among the most abundant species in the present study, which is in line with previous publications [32-34]. Furthermore, three different salivary microbial types were identified, and both the percentage of Prevotella-enriched and Haemophilus-enriched types were increased in participants with and without dysplasia compared with the control group. Actinomyces, the key abundant genera in type 2, which had a positive correlation with Prevotella, showed a significantly higher abundance in patients with dysplasia than the ones without dysplasia. Notably, the increase of Prevotella, Haemophilus, and Alloprevotella in the saliva of current or past dysplasia patients could serve as a biomarker to distinguish them from the healthy participants. Furthermore, the increase of Actinomyces could serve as a biomarker for distinguishing the current dysplasia patients from past dysplasia patients. Similar to the diagnostic roles of salivary microbiota in other kinds of cancers, [14-19] our results expanded the use of the salivary microbiota as the noninvasive diagnostic biomarker for cervical dysplasia. Elevated levels of oral Prevotella, Haemophilus, and Alloprevotella have been linked to various oral diseases, including periodontal disease, tonsillar disease, and oral cavity squamous cell carcinoma [35, 36]. These same microbial types have also been associated with systemic diseases, such as gout, obesity, cardiovascular disease, and liver cancer [37-41]. Furthermore, increased levels of Actinomyces have been linked to suppurative, granulomatous inflammatory lesions and have been shown to induce cervicofacial, pulmonary, and abdominopelvic infections, as well as oral cancer [42-45]. Other studies imply that these four genera are all oral opportunistic pathogens and can be linked to inflammatory reactions and cancer development [16, 39, 46-57]. Notably, a significant increase of Prevotella and Actinomyces has been previously reported in vaginal or cervical exudates from women with cervical dysplasia or cervical cancer [10, 58, 59]. This finding suggests the potential for the transfer of these opportunistic pathogens between the oral and vaginal tracts. Oral sexual behavior is one of the main routes for the transfer of bacteria or viruses from the oral to the vaginal tract. Previous studies have shown that most oral HPV types match the co-occurring cervical HPV types, indicating the potential for transfer between these two sites [61]. Furthermore, oral intake of Lactobacillus probiotics has been shown to increase their relative abundance in the vagina [60]. In addition, metabolites or immune cells induced by oral microorganisms or virulent strains may enter the systemic circulation through the blood, leading to low-grade inflammation and promoting the development of chronic inflammatory diseases and cancer in organs beyond the oral tract [62, 63]. We identified potential microbial roles in differential microbial immune-related pathways, revealing that Toll and Imd signaling pathways, as well as RLR, showed increased activity in the dysplasia and pretreatment groups. This indicates a potential increase in oral inflammation [64]. Previous studies have indicated that increased oral inflammation can activate Th1 or Th17 cells, leading to bacteremia or systemic inflammation [17, 63, 65, 66]. Our study indicates that smoking can affect the composition of salivary microbiota, which is supported by previous studies [67-70]. Actinomyces, which is significantly increased among smokers in our study, had been linked to higher gingivitis risk among cigarette smokers in other studies [67-71]. Cigarette smoking is an independent risk factor for cervical dysplasia occurrence and development [21, 59, 72, 73]. We identified a significant positive relationship between the RLR pathway and Actinomyces among patients with dysplasia than the ones without, and among smokers than nonsmokers. This supports a potential role for oral microbiota in linking smoking, oral inflammation, and dysplasia development. Further mechanistic and longitudinal studies that include smokers are needed to track the complicated interactions and potential talk between oral microbiota and cervical cancer development. There are several limitations to the present study. First, the limited number of patients and the complex situation of the recruited participants may have affected the oral microbiota data and the identification of potential diagnostic bacterial biomarkers, which must be considered when interpreting the results. In addition, smoking, age distribution, and other living habits, such as the use of antibiotics, may affect the identified microbes [74]. However, the PERMANOVA analysis suggests a significant difference in microbiota between the participants with and without cervical dysplasia, as well as between predysplasia (+) and postdysplasia (−) groups considering age and other living habits as confounders. These results indicated that differential genera, such as Actinomyces and Stomatobaculum, may contribute to the occurrence and recovery of vaginal dysplasia. Moreover, the detailed oral conditions of the participants who underwent vaginal examination, such as the periodontal disease or the decayed, missing, and filled (DMF) index, were not well identified, which may also affect the oral microbiota [75, 76]. Further, the dental examination itself may also have potential selection biases for the result. Thus, the data without the dental examination group were presented. It is worth noting that the taxonomic classification at the species level using 16S rRNA gene sequencing may not be entirely reliable, and the use of PICRUSt2 to predict microbial function may not fully reflect the actual metagenome changes. However, using PICRUSt2 could provide novel insights into microbial functions that help to search for potential biomarkers [77]. Moreover, given that those four identified potential biomarker genera, including the Prevotella, Alloprevotella, Haemophilus, and Actinomyces, are all opportunistic pathogens in the oral cavity and can be linked to inflammatory reactions and cancer development, we believe that our results provide valuable information to identify additional potential salivary targets for smoking-related diseases and cervical dysplasia. Last, most of the participants without dysplasia previously experienced dysplasia, which might affect the representativeness of their data as a healthy state reference. However, our study provides one of the first insights into the association between salivary microbiota and cervical dysplasia and its treatment. Given that certain bacteria were identified in multiple comparisons, our study lays a valuable foundation for further research on the role of salivary microbiota in cervical cancer. Additional longitudinal studies focusing on the metagenome changes in the oral and vaginal tracts during disease development and recovery are urgently needed to gain a deeper understanding of the involvement of salivary microbiota. This study demonstrated that salivary microbiota profiles differed between participants with and without cervical dysplasia, as well as between patients before and after cervical conization. Moreover, the study suggests that Actinomyces in the salivary microbiota may link smoking, oral microbiota, and cervical dysplasia and could potentially serve as a diagnostic marker. This cross-sectional study involved 47 women who visited the Karolinska University Hospital in Stockholm, Sweden. In this study, 28 women visited due to potential dysplasia and 19 visited for follow-up examination after treatment with conization for HSIL (Figure 1A and Supporting Information: Table S5). Furthermore, 20 healthy women who volunteered for a routine dental examination at a dental clinic in Stockholm, Sweden, were also included in the study. All participants were asked to fill out a questionnaire regarding their lifestyles (Supporting Information: Table S6). The study was approved by the Regional Ethical Board in Stockholm, Sweden. All participants provided written informed consent to take part in the study. Detailed diagnostic and sample collection methods are available in the Supplementary material. DNA was extracted from all 67 saliva samples and used for HPV genotyping with the MAGPIX instrument, according to our published papers [6, 12, 78, 79]. Furthermore, the V3–V4 regions of the 16S rRNA genes were amplified using Illumina sequencing index-binding primer pairs 341F/805 R and sequenced on an Illumina MiSeq sequencing platform [37]. Thereafter, the sequencing data were analyzed using the QIIME2 platform and the Human Oral Microbiome Database (eHOMD) (V15.2). Detailed information about DNA extraction, oral HPV genotyping, oral microbiota sequencing, and bioinformatic analyses is available in the Supplementary material. The Mann–Whitney U test with multiple comparisons adjusted with the Benjamini–Hochberg FDR was performed to compare microbial alpha diversity and identify significantly altered genera and KEGG pathways between two groups. The Kruskal–Wallis test with the Tukey–Kramer post hoc test was used to test microbial differences among more than two groups. ANOSIM analysis based on Bray–Curtis distance matrices was used to identify beta diversity, and the adonis function from the R package “vegan” was used for PERMANOVA [80, 81]. The pairwise correlations (p < 0.05) were used to generate the co-occurrence network (Spearman's rank correlation coefficient indices > 0.5) and correlation heatmaps (Spearman's rank correlation coefficient indices > 0.4). Additionally, the AUC of the receiver operating characteristic was calculated to analyze the sensitivity and specificity of the diagnostic power. Conception and design: Shengru Wu, Lars Engstrand, Sonia Andersson, Juan Du. Sample collection: Liqin Cheng, Alexandra A. L. Pennhag, Miriam Mints, Sonia Andersson, Juan Du. Development of methodology: Shengru Wu, Liqin Cheng, Alexandra A. L. Pennhag, Maike Seifert, Juan Du. Acquisition of data: Shengru Wu, Liqin Cheng, Alexandra A. L. Pennhag, Maike Seifert, Unnur Guðnadóttir, Miriam Mints, Sonia Andersson, Juan Du. Analysis and interpretation of data: Shengru Wu, Liqin Cheng, Sonia Andersson, Juan Du. Writing of the manuscript: Shengru Wu, Juan Du. Review and/or revision of the manuscript: All authors. Juan Du and Shengru Wu are supported by the Swedish Research Council (2021-01683, 2021-06112), the Swedish Foundation for Strategic Research (SSF) [ICA16-0050], Svenska Läkaresällskapet [SLS-784981, SLS-960584], and the Karolinska Institute Foundation. A special acknowledgment to our loved Dr. Sonia Andersson, who recently passed away. We are very grateful for all her commitment and dedication to this project. Ferring Pharmaceuticals funded the center where this project was carried out. We thank all the support from the Centre for Translational Microbiome Research (CTMR), Karolinska Institute, Sweden. We thank Fadia Alkass and her colleagues for sample collection from the dental examination. We really appreciate all the support from the funding agencies and all the participants who provided their valuable samples. The authors declare no conflict of interest. The study was approved by the Regional Ethical Board at Karolinska Institute, Stockholm, Sweden (ethical permission number 2017/725-31 and 2019-04201). All the data generated from this study are included in this paper. The sequencing reads are available in the Sequence Read Archive (SRA) of NCBI under accession project number PRJNA863336 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA863336). Supplementary materials (figures, tables, scripts, graphical abstract, slides, videos, Chinese translated version, and updated materials) can be found in the online DOI or iMeta Science http://www.imeta.science/. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.