Biomarkers of malignant transformation of oral submucous fibrosis

Abstract

A biomarker for malignant transformation of oral submucous fibrosis comprises a change in the salivary microbiome, wherein the characteristic bacterial species in the altered salivary microbiome include at least one or a combination selected from the group consisting of Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp.HMT-300, Mitsuokella sp.HMT-131 and Treponema sp.HMT-927.

Description

Technical Field

[0001] The present application relates to a biomarker for malignant transformation, and in particular to a biomarker for malignant transformation of oral submucous fibrosis. Background Art

[0002] Oral cancer is a general term for malignant tumors that occur in the oral cavity, and most of them belong to oral squamous cell carcinoma (OSCC). Oral cancer has become a global burden, especially in Southeast Asia. In Taiwan, China, mortality caused by oral cancer is also a socioeconomic problem because oral cancer patients are more likely to be young men compared with patients with other types of cancer. Oral submucous fibrosis (OSF) is an inapparent and chronic oral stromal premalignant disorder that can affect various parts of the oral cavity and sometimes the pharynx, making the texture of the mucosa, epithelium and stroma fibrotic, hypocellular, and hypovascular. According to global evidence, OSF lesions are accompanied by atypical high rates of malignant transformation. OSF patients usually have the habit of chewing betel quid, and have a high rate of malignant transformation compared with healthy people, with a risk ratio of 23.3-27.5, and oral cancer develops faster. According to studies from India and Taiwan, patients with oral squamous cell carcinoma in an OSF background (OSCC-OSF) are younger. Therefore, identifying patients at high risk of OSCC among OSF patients is a significant clinical challenge.

[0003] Invasive biopsy is currently the standard method for diagnosing oral cancer. However, OSF patients usually have limited ability to open their mouths, which often makes oral assessment difficult, and there is still a lack of non-invasive tools for effectively screening large groups (especially those with OSF). Although related studies in the past have reported biomarkers that can be used to predict OSF carcinogenesis, most biomarkers rely on the collection of invasive test samples followed by immunochemical staining or tissue microarray analysis. Studies have confirmed that saliva is one of the sources for detecting oral cancer biomarkers, but for high-risk OSF patients, no effective saliva biomarkers have been found. Since collecting saliva is a non-invasive test sample collection method, if biomarkers are detected in the saliva of OSF patients and effectively predict the occurrence of oral cancer, it can promote early intervention for this high-risk group.

[0004] There is increasing evidence that microbiota dysbiosis is closely related to human diseases and their development. The oral cavity provides a complex habitat for the growth of diverse bacterial communities, such as the mucosal surfaces of the buccal area, tongue and gums, and the surfaces of teeth. The bacterial community in the oral cavity is very complex, consisting of more than 700 bacterial species, showing a dynamic growth appearance in a specific oral microenvironment, and the oral microbiome provides various health benefits to the human host. Although the diversity, composition and function of the oral microbiome change with the development of oral cancer, there is currently no detection data on the characteristics of the salivary microbiome of OSCC in high-risk OSF groups. Therefore, this application intends to conduct a comprehensive comparison of the salivary microbiomes of male OSF patients with the habit of chewing betel nut and OSCC-OSF patients to explore the role of the microbiome in the development of OSF cancer. Summary of the invention

[0005] One purpose of the present application is to compare the salivary microbiota of OSF patients and OSCC-OSF patients and to observe the changes in the salivary microbiota during the malignant transformation of OSF to OSCC-OSF.

[0006] Another object of the present application is to provide biomarkers for malignant transformation of OSF, so as to identify patients at high risk of OSCC among OSF patients, and then detect and treat the high-risk patients.

[0007] In order to achieve the above objectives, the present application provides a biomarker for malignant transformation of oral submucous fibrosis, comprising changes in the salivary microbiome, wherein the characteristic bacterial species in the changed salivary microbiome include at least one selected from the group consisting of Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp.HMT-300, Mitsuokella sp.HMT-131 and Treponema sp.HMT-927 and combinations thereof.

[0008] In one embodiment, the featured strains further comprise Dialister micraerophilus and / or Mollicutes sp. HMT-504.

[0009] In one embodiment, the characteristic strains further include at least one selected from the group consisting of Mycoplasma faucium, Prevotella denticola, Peptostreptococcaceae sp. HMT-369, Prevotella sp. HMT-315, Clostridiales sp. HMT-093, Eubacterium saphenus, Catonella sp. HMT-451, Treponema sp. HMT-237, Selenomonas sputigena, Haemophilus pittmaniae, Prevotella baroniae, Actinomyces sp. HMT-169, Absconditabacteria (SR1) sp. HMT-874, Treponema sp. HMT-270, Mollicutes sp. HMT-906, Bacteroidetes sp. HMT-280, Treponema sp. HMT-238 and Treponema sp. HMT-258, and combinations thereof.

[0010] In one embodiment, the characteristic bacterial species further comprises at least one selected from the group consisting of Clostridiales sp. HMT-876, Corynebacterium durum, Gracilibacteria sp. HMT-871, Megasphaera sp. HMT-123, Mobiluncus mulieris, Negativicutes, Peptostreptococcaceae and Selenomonas sp. HMT-478 and combinations thereof.

[0011] In one embodiment, the characteristic bacterial species is associated with an oral health condition.

[0012] In one embodiment, the signature species is associated with smoking status.

[0013] In one embodiment, the alteration in the saliva microbiome comprises a change in relative abundance.

[0014] In one embodiment, the alteration of the salivary microbiome comprises an alteration in the biomass of the microorganisms.

[0015] In order to achieve the above objectives, the present application also provides a biomarker for malignant transformation of oral submucous fibrosis, comprising changes in metabolic pathways, wherein the metabolic pathways comprise at least one selected from the group consisting of synthesis of S-adenosyl-L-methionine, synthesis of norspermamine, synthesis of L-ornithine, synthesis of pyrimidine deoxyribonucleotides and formaldehyde metabolism, and combinations thereof.

[0016] In one embodiment, the metabolic pathway further comprises at least one selected from the group consisting of synthesis of L-histidine, synthesis of nicotinamide adenine dinucleotide, and synthesis of 1,4-dihydroxy-6-naphthoic acid, and combinations thereof.

[0017] Brief description of the attached figure

[0018] Figure 1 Null model randomization rate distributions of microbiome assemblies of OSF and OSCC-OSF patients are shown.

[0019] Figure 2A Shown are the results of principal coordinate analysis of salivary microbiome samples from OSF and OSCC-OSF patients.

[0020] Figure 2B Shown are the results of Adonis analysis of saliva microbiome samples from OSF and OSCC-OSF patients.

[0021] Figure 3 The ratios of core bacterial species in saliva samples of OSF and OSCC-OSF patients and their Venn diagrams are shown.

[0022] Figure 4 Characteristic bacterial species with significant differences in relative abundance in saliva samples of OSF and OSCC-OSF patients identified by LEfSe analysis.

[0023] Figure 5 The Kruskal-Wallis test showed that the characteristic bacterial species with significant differences in number per milliliter of saliva samples were identified between OSF and OSCC-OSF patients.

[0024] Figure 6 Venn diagram showing bacterial species associated with oral health and smoking status.

[0025] Figure 7 Displays the ROC curve for the Random Forest algorithm.

[0026] Fig. 8A Showing the interactive network structure among salivary microbiota of OSF patients.

[0027] Figure 8B Showing the interactive network structure between salivary microbiota of OSCC-OSF patients.

[0028] Fig. 9 Metabolic pathways showing significant differences in abundance between saliva samples of OSF and OSCC-OSF patients. DETAILED DESCRIPTION

[0029] Some embodiments that embody the features and advantages of the present application will be described in detail in the following description. It should be understood that the present application can have various changes in different aspects without departing from the scope of the present application, and the descriptions and drawings therein are essentially used for illustration and are not used to limit the present application in any way.

[0030] The main purpose of this application is to systematically analyze the correlation between the saliva microbiota of patients with oral submucous fibrosis and oral cancer (OSCC-OSF) and the malignant transformation of oral cancer, to understand the members and key species of the salivary microbiome in promoting or alleviating the malignant transformation of precancerous lesions, and to develop biomarkers for microbiome liquid biopsy as a strategy for determining the transformation of oral precancerous lesions into oral cancer precision medicine.

[0031] First, clinical samples were collected. The exploratory cohorts consisted of 52 clinical samples from National Cheng Kung University Hospital in Taiwan Province, China, including saliva samples from 18 OSF patients and 34 OSCC-OSF patients. The validation cohorts consisted of clinical samples from Chi Mei Hospital in Taiwan Province, China, including saliva samples from 10 OSF patients and 11 OSCC-OSF patients. These participating patients were all males who chewed betel nut and had not received antibiotic treatment for at least one month before the collection of saliva samples. They were also asked not to eat, drink, or use oral hygiene products for at least 1 hour before the collection of saliva. The microbiome in saliva was then analyzed by amplicon sequencing and quantitative polymerase chain reaction (qPCR, also known as real-time PCR), and statistical analysis of the composition and microbial biomass data, as well as machine learning analysis, were performed.

[0032] Ecological mechanisms, including deterministic processes (such as host selection, immune response, etc.) and stochastic processes (such as environmental factors, stress, chemical exposure, etc.), will simultaneously affect the structure of bacterial communities. To assess their relative advantages in controlling the salivary microbiome, this application quantified the null-model-based stochasticity using Bray-Curtis and Weighted UniFrac distances. Figure 1Figure 2 shows the null model randomization rate distribution of the microbiome assembly of OSF and OSCC-OSF patients calculated according to the Bray-Curtis and weighted UniFrac distance functions, respectively. As shown in the figure, all average randomization rates in both OSF and OSCC-OSF cohorts exceeded 50% in both distance measures, indicating that randomization processes contribute more to salivary microbiome assembly than host selection, that is, randomization processes dominate in the formation of the salivary microbiome. It is worth noting that the average randomization rate of the OSCC-OSF group is higher than that of the OSF group, which also indicates that the random advantage may be enhanced due to oral carcinogenesis.

[0033] Further observations were made to see whether oral health status and smoking status would cause microbiome composition variation. This application used four distance measurement methods to compare the microbiome structure of OSF and OSCC-OSF. The four distance measurement methods were Jaccard distance, Bray-Curtis distance, unweighted UniFrac distance, and weighted UniFrac distance. Unweighted and weighted UniFrac distances belong to phylogenetic distance measures, and the weighting here is weighted for the richness of microorganisms. Figure 2A The principal coordinates analysis results of the salivary microbiome samples of OSF and OSCC-OSF patients calculated according to the Jaccard, Bray-Curtis, unweighted UniFrac and weighted UniFrac distance functions are shown. Figure 2B The Adonis analysis results of OSF and OSCC-OSF patient saliva microbiome samples calculated according to Jaccard, Bray-Curtis, unweighted UniFrac and weighted UniFrac distance functions are shown. In the principal coordinate analysis, Figure 2AAs shown, unweighted and weighted UniFrac distances are better able to distinguish OSF and OSCC-OSF in the reduced coordinate space than Jaccard and Bray-Curtis distances. When using UniFrac distance, the difference in the first two principal coordinates (44.8%) is greater than the difference when using Jaccard and Bray-Curtis distances (<26.1%). A distance-based Adonis analysis was then performed to test whether the changes in microbiome composition were attributed to differences in the characteristics of the participating patients. Since all patients in this application chewed betel nut, the betel nut factor was excluded from the analysis. As Figure 2B As shown in the results of the test using Bray-Curtis and weighted UniFrac distance, the patient's age (Age), smoking status (Smoking), drinking status (Alcohol) or oral health status, such as periodontal condition (periodontal condition) and OSF carcinogenesis (OSFcarcinogenesis), were not significantly associated with changes in microbiome composition. However, according to the test results using Jaccard and unweighted UniFrac distance, OSF carcinogenesis (OSF vs. OSCC-OSF) and smoking status significantly affected the changes in microbiome composition (P<0.05). In other words, oral health status and smoking status do cause changes in microbiome composition.

[0034] The present application also uses tag-encoded high-throughput sequencing (HTS) of the V3-V4 region of the 16S rRNA gene amplicon to analyze the bacterial population in the saliva samples of 18 OSF patients and 34 OSCC-OSF patients, and obtains amplicon sequence variants (ASVs) to study microbial diversity. In HTS, a total of 6160 unique ASVs were detected from 52 samples, and from the analysis of ASV distribution, it can be seen that the ASVs in the saliva microbiome are significantly different between individuals. These ASVs, which accounted for approximately 99.9% of the total reads, could be classified into 12 bacterial phyla, including the following major taxa (>1%): Firmicutes (35.8±12.8%), Bacteroidetes (21.9±10.6%), Proteobacteria (21.0±14.0%), Saccharibacteria (TM7) (7.3±9.5%), Actinobacteria (6.0±5.5%), Fusobacteria (5.5±4.3%), and Spirochaetes (1.6±2.6%). Regarding species annotation, 408 and 470 taxa were detected in OSF and OSCC-OSF samples, respectively, of which 51 (12.5%) and 28 (about 6%) taxa had a prevalence greater than 75% in patients in OSF and OSCC-OSF samples, respectively, and were therefore considered members of the core saliva microbiome of each group. These core species accounted for the majority of the saliva microbiome and accounted for 64.5% and 52.7% of the average read richness in OSF and OSCC-OSF, respectively. Figure 3 The proportion of core bacterial species in saliva samples of OSF and OSCC-OSF patients and their Venn diagrams are shown. Among the core bacterial species, 25 species were present in both groups, and another 26 and 3 species were present in the OSF and OSCC-OSF groups, respectively. The names of each species are listed below the Venn diagram. In addition, the proportion of core bacterial species in the total taxa was 12.5% ​​and 6% in OSF and OSCC-OSF samples, respectively, which also showed that OSF cancerization would reduce the proportion of core bacterial species. Therefore, compared with OSF, the number and richness of core bacterial species in OSCC-OSF were reduced, indicating that OSF cancerization had a dispersive effect on the composition of the microbiome.

[0035] The present application then used different statistical analysis methods to identify unique microorganisms associated with oral health status in OSF and OSCC-OSF groups. First, linear discriminant analysis (LDA) effect size (LEfSe) was used for analysis. Figure 4 That is, the relative abundance of characteristic bacterial species in saliva samples of OSF and OSCC-OSF patients identified by LEfSe analysis was significantly different. A total of 42 species were identified with differential abundance between OSF and OSCC-OSF (LDA score>2). Among them, three species, namely Haemophiluspittmaniae, Prevotella sp.HMT-309 and Treponema sp.HMT-270, were significantly more abundant in OSCC-OSF patients (P<0.05). The prevalence of H.pittmaniae and Prevotella sp. HMT-309 in OSCC-OSF samples (genus Q2 prevalence (25-50%)) was significantly higher than that in OSF samples (genus Q1 prevalence (<25%)). The other 39 species, belonging to Prevotella (9 species), Treponema (6 species), Selenomonas (3 species) and other 21 genera (one species from each genus), were significantly more abundant in OSF patients (P<0.05), and the prevalence of these groups was significantly reduced in OSCC-OSF samples. Only 6 species, Mycoplasma faucium, Prevotella denticola, Clostridiales sp.HMT-093, Prevotella baroniae, Prevotella oulorum and Selenomonas sputigena, were core species (genus Q4 prevalence (>75%)), and their prevalence in OSCC-OSF samples decreased the most (41.0%±6.4%) compared with transitional groups (prevalence <75%).

[0036] The present application converts the relative abundance of components into microbial load, that is, the number of bacteria per milliliter of saliva, by multiplying the relative abundance corrected by 16S rRNA copy number by the qPCR quantitative cell number corrected by 16S rRNA copy number, and then uses the absolute microbial load to identify the characteristic bacterial species of OSF and OSCC-OSF. Figure 5The Kruskal-Wallis test showed that the number of characteristic bacterial species per milliliter of saliva samples in OSF and OSCC-OSF patients was significantly different. According to the data of absolute microbial counts, the number of 15 genera and 23 species was statistically different between the two groups (P < 0.05, effect size In at least one group, the prevalence of individual species was >33%. Moreover, these 23 species can be mapped to Figure 4 The LEfSe identification results showed that both LEfSe analysis and Kruskal-Wallis test consistently identified these 23 strains as characteristic strains of OSF and OSCC-OSF.

[0037] According to the aforementioned Figure 2B In addition to oral health status, smoking status also significantly affects changes in microbiome composition. Based on the Kruskal-Wallis test of absolute microbial biomass data, the bacterial species associated with oral health status and smoking status can also be further analyzed. Figure 6 A Venn diagram showing the bacterial species associated with OSF canceration and smoking status. The names of the related bacterial species are listed below the Venn diagram. Figure 6 It can be seen that 28 bacterial species are particularly related to the patient's smoking habits, of which 12 species in 8 genera are also related to OSF cancerization. In addition, 11 bacterial species are particularly related to OSF cancerization. Among them, those marked with a single asterisk are the core bacterial species of OSF, and those marked with a double asterisk are the characteristic bacterial species identified by the aforementioned statistical analysis method. There are 5 double asterisks, namely Mitsuokella sp.HMT-131, Porphyromonas catoniae, Prevotellamultisaccharivorax, Prevotella sp.HMT-300 and Treponema sp.HMT-927, all of which are characteristic bacterial species related to smoking habits and OSF cancerization.

[0038] The present application further uses machine learning to identify the characteristic strains of OSF and OSCC-OSF, and the machine-learning feature-selective algorithm used detects 15 strains that affect the discrimination efficiency, of which 5 strains, including Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp.HMT-300, Mitsuokella sp.HMT-131 and Treponema sp.HMT-927, can correspond to the characteristic strains identified by LEfSe and Kruskal-Wallis statistical analysis ( Figure 4 and Figure 5 ), and also correspond to the characteristic bacteria species related to smoking habits and OSF canceration ( Figure 6 ). Two other species, Dialister micraerophilus and Mollicutes sp.HMT-504, were also detected in the LEfSe analysis ( Figure 4 ). The remaining eight species were detected by the machine learning feature selection algorithm but not by the statistical analysis method, including Clostridiales sp. HMT-876, Corynebacterium durum, Gracilibacteria sp. HMT-871, Megasphaera sp. HMT-123, Mobiluncus mulieris, Negativicutes (unclassified), Peptostreptococcaceae (unclassified), and Selenomonas sp. HMT-478.

[0039] In other words, multiple statistical analyses and machine learning methods consistently identified five bacterial species, including Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp. HMT-300, Mitsuokella sp. HMT-131, and Treponema sp. HMT-927, and these five bacterial species are all related to the oral health status and smoking status of OSF patients who chew betel quid, and can be used as biomarkers for OSF malignant transformation. Therefore, the present application provides biomarkers for OSF malignant transformation, including changes in the saliva microbiome, wherein the characteristic bacterial species in the changed saliva microbiome include at least one selected from the group consisting of Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp. HMT-300, Mitsuokella sp. HMT-131, and Treponema sp. HMT-927, and combinations thereof.

[0040] In one embodiment, the characteristic bacterial species may also include Dialister micraerophilus and / or Mollicutes sp. HMT-504, which are bacterial species consistently identified by LEfSe statistical analysis and machine learning methods.

[0041] In one embodiment, the characteristic strains may also include a strain selected from Mycoplasma faucium, Prevotella denticola, Peptostreptococcaceae sp. HMT-369, Prevotella sp. HMT-315, Clostridiales sp. HMT-093, Eubacterium saphenus, Catonella sp. HMT-451, Treponema sp. HMT-237, Selenomonas sputigena, Haemophilus pittmaniae, Prevotella baroniae, Actinomyces sp. HMT-169, Absconditabacteria (SR1) sp. HMT-874, Treponema sp. HMT-270, Mollicutes sp. HMT-906, Bacteroidetes sp. HMT-280, Treponema sp. HMT-238 and Treponema sp.HMT-258 and combinations thereof, that is, species consistently identified by LEfSe and Kruskal-Wallis statistical analysis.

[0042] In one embodiment, the characteristic bacterial species may also include at least one selected from the group consisting of Clostridiales sp. HMT-876, Corynebacterium durum, Gracilibacteria sp. HMT-871, Megasphaera sp. HMT-123, Mobiluncus mulieris, Negativicutes, Peptostreptococcaceae and Selenomonas sp. HMT-478 and combinations thereof, that is, bacterial species additionally identified by the machine learning method.

[0043] For the 23 characteristic bacterial species that were consistently identified using LEfSe and Kruskal-Wallis statistical analysis, the relative AUC (AUC) was calculated by calculating the area under the receiver operating characteristic (ROC) curve (AUC) based on the relative abundance and absolute microbial count. relative ) and absolute AUC (AUC absolute) values ​​were used to evaluate the effectiveness of the 23 characteristic bacterial species in distinguishing OSF and OSCC-OSF. In addition to cross-validation of the exploration group dataset, the independent saliva microbiome dataset of the validation group was also used for evaluation, and the results are shown in Table 1 below. The average AUCs of the exploration group and validation group datasets were 0.68 and 0.56, respectively. Among them, the characteristic bacterial species Treponema sp.HMT-927, which is related to both oral health status and smoking status, showed the highest distinguishability in the exploration group dataset (relative AUC = 0.748, absolute AUC = 0.758), and had high predictive performance in the validation group dataset (relative AUC = 0.709, absolute AUC = 0.682). The best performing strain in the validation dataset was Absconditabacteria (SR1) sp. HMT-874 (relative AUC = 0.755, absolute AUC = 0.809), which is a taxon in the candidate phyla radiation group and is also identified as a characteristic strain associated with oral health status ( Figure 6 ). However, there was no statistical significance between the AUC derived from relative abundance and absolute microbial load for either the exploration dataset (H = 0.008, P = 0.923) or the validation dataset (H = 0.088, P = 0.767). This observation also shows that the distinguishability of characteristic species yielded similar results regardless of the quantitative method used in the study (relative abundance or absolute microbial load).

[0044] Table 1

[0045]

[0046] The machine learning analysis also achieved high discrimination efficiency by combining the characteristic bacterial species with the host clinical and lifestyle characteristics. The random-forest algorithm trained with the data of the bacterial species detected by feature selection, Faith's phylogenetic diversity (Faith's PD) and lifestyle (current smoking habits) was the most effective in discriminating between the OSF and OSCC-OSF groups. Figure 7 The ROC curve of the random forest algorithm (5-fold cross validation) is shown. The sensitivity is the vertical axis and the false positive rate (1-specificity) is the horizontal axis. The more the ROC curve deviates from the diagonal line (also known as the chance line), the larger the area under the ROC curve is, which also represents a higher accuracy. Figure 7As shown in FIG. 1 , the ROC curve obtained by cross-validation of the trained random forest algorithm has a mean 5-fold cross-validation accuracy of 85.1% and an AUC of 0.88, indicating that the trained random forest algorithm has a fairly high accuracy.

[0047] To evaluate the interactions between bacterial species in saliva, this application further performed SparCC analysis using the microbiome datasets of the OSF and OSCC-OSF groups. Fig. 8A Showing the interactive network structure between salivary flora of OSF patients, Figure 8B The interactive network structure of the salivary flora of OSCC-OSF patients is shown in the figure. As can be seen from the figure, the microbial interactions of OSF patients (7173 lines connecting 362 nodes) are much more complex than those of OSCC-OSF patients (2695 lines connecting 351 nodes), and the complexity is nearly 3 times. Regarding the network formed by the characteristic bacterial species as nodes and their highly correlated neighbors (SparCC correlation coefficient>0.6), Fig. 8A and Figure 8B The results showed that the network structures of the two salivary microbiota ecosystems were completely different. Fig. 8A As shown in the figure, the OSF group has a higher connectivity, and there are more positive relationship lines (solid lines, SparCC correlation coefficient>0.6) than negative relationship lines (dashed lines, SparCC correlation coefficient<-0.6), indicating the existence of a unique bacterial community symbiosis pattern. In addition to Eubacteriumsaphenus, Mycoplasma faucium, Catonella sp.HMT451, and Clostridiales sp. HMT-093, several characteristic species, including Treponema (HMT-927, HMT-258, HMT-237, and HMT-238) and Prevotella (P.denticola and HMT-315), are central hubs with high connectivity to positive relationship lines (solid lines) in the network. There are two other hub species, namely Neisseria sp. (unclassified) and Oribacterium sp. (unclassified), which are mostly connected to other species by negative relationship lines (dashed lines). These observations highlight the importance of regulating bacterial interactions in the salivary microbiome of OSF patients. Figure 8B As shown in the figure, there is no corresponding bacterial symbiosis pattern in OSCC-OSF patients, indicating that the network structure between bacterial communities has undergone substantial changes. The cancerization process has caused the original close interaction structure between OSF bacterial communities to disappear, confirming that the salivary microbiome does change with malignant transformation.

[0048] To compare the biochemical functions of OSF and OSCC-OSF saliva microbiomes, the present application further used LEfSe to discriminate the quantitative differences in metabolic pathways predicted by PICRUSt2 from the MetaCyc database. A total of 85.77% of the denoised sequences had high to medium quality (nearest sequenced taxon indicator (NSTI) score <0.15). To ensure appropriate prediction quality, denoised sequences with NSTI scores >2 (0.37% of total denoised sequences) were excluded from the PICRUSt2 analysis. Although many characteristic species were detected in the OSF and OSCC-OSF groups, only 9 metabolic pathways were quantitatively distinguishable. Fig. 9 Among the metabolic pathways with significant differences in abundance in saliva samples from OSF and OSCC-OSF patients, six were more closely associated with OSF, including the synthesis of L-ornithine, L-histidine, pyrimidine deoxyribonucleotide, and nicotinamide adenine dinucleotide (NAD), and formaldehyde oxidation and assimilation. In OSCC-OSF, three were more closely associated, namely the synthesis of 1,4-dihydroxy-6-naphthoate (an intermediate in the synthesis of vitamin K2), S-adenosyl-L-methionine (SAM), and norspermidine. Among them, OSCC-OSF had higher S-adenosyl-L-methionine and norspermine synthesis potential, but lower L-ornithine and pyrimidine deoxyribonucleotide synthesis and formaldehyde metabolism potential. These findings indicate that the salivary microbiome plays an important role in regulating microbial metabolism during oral carcinogenesis.

[0049] These nine metabolic pathways can also be classified into five types of metabolism. The synthesis of S-adenosyl-L-methionine, L-ornithine and L-histidine belong to amino acid biosynthesis. The synthesis of pyrimidine deoxyribonucleotides belongs to nucleoside and nucleotide biosynthesis. Formaldehyde oxidation and assimilation belong to C1 compound utilization and assimilation. The synthesis of 1,4-dihydroxy-6-naphthoic acid and NAD belongs to cofactor, prosthetic group electron carrier and vitamin biosynthesis. The synthesis of norspermine belongs to amine and polyamine biosynthesis. The regulatory mechanism of salivary microorganisms on metabolic pathways needs further study.

[0050] In summary, since changes in the salivary microbiome are related to oral squamous cell carcinoma (OSCC), and most OSCCs are caused by precancerous lesions, the malignant transformation rate of patients with oral submucous fibrosis (OSF) is high, but how the salivary microbiome changes with malignant transformation is unclear. Therefore, this application compares the salivary microbiome of patients with oral submucous fibrosis and oral cancer (OSCC-OSF) and OSF patients to observe the changes in the salivary microbiome during the malignant transformation of OSF to OSCC-OSF. According to the results of high-throughput sequencing of the V3-V4 region of the bacterial 16S rRNA gene, oral health status and smoking status significantly affect the changes in the systematic composition of the salivary microbiome, resulting in a significant decrease in species richness and phylogenetic diversity (P<0.05). OSF malignant transformation increases the impact of randomness on changes in the composition of the salivary microbiome and completely changes the species symbiotic network pattern. Multiple statistical analyses and machine learning methods consistently identified five bacterial species that were significantly altered between OSF and OSCC-OSF, including Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp. HMT-300, Mitsuokella sp. HMT-131, and Treponemasp. HMT-927, and all of these five species were associated with oral health status and smoking status. Therefore, these five species can be used as biomarkers for OSF malignant transformation, and these biomarkers have a high accuracy in predicting oral cancer in both the exploration and validation datasets. In addition, OSCC-OSF has a higher potential for S-adenosyl-L-methionine and norspermine synthesis, but a lower potential for L-ornithine and pyrimidine deoxyribonucleotide synthesis and formaldehyde metabolism, which also indicates that the salivary microbiome plays an important role in regulating microbial metabolism during oral carcinogenesis. Therefore, the present application provides biomarkers for OSF malignant transformation, which are helpful for the study of cancer formation and treatment, and can identify patients at high risk of OSCC among OSF patients, thereby detecting and treating high-risk patients.

[0051] Although the present invention has been described in detail by the above embodiments, various modifications may be made by those skilled in the art without departing from the scope of protection claimed in the appended claims.

Claims

1. Use of a reagent for detecting changes in a salivary microbiome in the preparation of a reagent for detecting malignant transformation of oral submucous fibrosis into oral squamous cell carcinoma, wherein the change in the salivary microbiome is a change in relative abundance or a change in microbial abundance, wherein the characteristic bacterial species in the changed salivary microbiome are selected from at least one of the group consisting of Porphyromonas catoniae, Prevotella multisaccharivorax, Prevotella sp. HMT-300, Mitsuokella sp. HMT-131 and Treponema sp. HMT-927 or a combination thereof.

2. The method of claim 1, wherein the characteristic strains further comprise a strain selected from the group consisting of Mycoplasma faucium, Prevotella denticola, Peptostreptococcaceae sp. HMT-369, Prevotella sp. HMT-315, Clostridiales sp. HMT-093, Eubacterium saphenus, Catonella sp. HMT-451, Treponema sp. HMT-237, Selenomonas sputigena, Haemophilus pittmaniae, Prevotella baroniae, Actinomyces sp. HMT-169, Absconditabacteria (SR1) sp. HMT-874, Treponema sp. HMT-270, Mollicutes sp. HMT-906, Bacteroidetes sp. HMT-280, Treponema sp. HMT-238 and Treponema At least one or a combination of the group consisting of sp. HMT-258.

3. The use according to claim 1, wherein the characteristic bacterial species is related to oral health status.

4. The use according to claim 1, wherein the characteristic bacterial species is related to smoking status.

5. The use according to claim 1, wherein the change in the salivary microbiome is a change in relative abundance.

6. The method according to claim 1, wherein the change in the salivary microbiome is a change in the amount of microorganisms.

Images