The Oral Bacterial Microbiome of Interdental Surfaces in Adolescents According to Carious Risk

Adolescence is closely associated with a high risk of caries. The identification of specific bacteria in an oral microniche, the interdental space of the molars, according to carious risk can facilitate the prediction of future caries and the anticipation of the progression or stabilization of caries in adolescents. A cross-sectional clinical study according to the bacteriological criteria of interdental healthy adolescents and carious risk factors—low and high—using a real-time polymerase chain reaction technique was conducted. The presence of 26 oral pathogens from the interdental microbiota of 50 adolescents aged 15 to 17 years were qualitatively and quantitatively analyzed. Bacteria known to be cariogenic (Bifidobacterium dentium, Lactobacillus spp., Rothia dentocariosa, Streptococcus cristatus, Streptococcus mutans, Streptococcus salivarius, Streptococcus sobrinus, and Streptococcus wiggsiae) did not present differences in abundance according to carious risk. Periodontal bacteria from the red complex are positively correlated with carious risk. However, only 3 bacteria—S. sobrinus, E corrodens and T. forsythia—presented a significant increase in the highest group. Estimating the risk of caries associated with bacterial factors in interdental sites of molars in adolescents contributes to the better definition of carious risk status, periodicity and intensity of diagnostic, prevention and restorative services.

Dental caries is a biofilm-mediated, sugar-driven, multifactorial, dynamic disease that results in the phasic demineralization and remineralization of dental tissues [3]. Dental caries is the result of dissolution of the tooth mineral by a reduction in pH due to the sustained fermentation of carbohydrate by bacteria in a local biofilm structure that limits the ability of saliva to wash away or buffer the acidic metabolic products [4][5][6]. The biofilm diversity of tooth surfaces is influenced by This study investigated the changes in biofilm composition of interdental surfaces in posterior permanent teeth according to carious risk. Data to identify bacteria from the microbiome associated with caries risk will be emphasized.
The hypotheses are: (i) The quantification of the oral microbiome at the interproximal surface of the tooth could contribute to the assessment of carious risk and personalized clinical decision-making, (ii) the qualitative differentiated analysis of the microbiome could help to predict interproximal adolescence caries development, (iii) the identification of specific bacterial species could help to develop novel approaches in their diagnosis and management of carious lesions.

Materials and Methods
The Microbiota of Interdental Space of Adolescent according to Risk of Caries (MIARC) trial is a cross-sectional observational clinical study. The workflow of the experiment is described in Figure 1.

Study Population
Fifty subjects (male and female) were recruited between November and December 2018 from a pool of first-time volunteers who were referred to the Department of Public Health of the Oral Medicine Hospital of Montpellier, France. Written informed consent was obtained from all enrolled subjects in accordance with the Declaration of Helsinki and authorization of the ethics council. The 50 adolescents were selected based on inclusion criteria and carious risk classification to have 25 subjects in the high caries risk (HCR) group and 25 subjects in the low caries risk (LCR) group.
The inclusion criteria were: (i) Age 15-17 years old, (ii) presence of teeth (15, 16, 25, 26, 35, 36, 45, and 46), (iii) accessibility of the interdental space for the four sites (15-16, 25-26, 35-36, and 45-46) by the interdental brush in each subject, (iv) the presence of at least 22 natural teeth, (v) good understanding of the French language, (vi) one of the parents accepts the terms of the study and signs the written informed consent, (vii) adolescent accepts the terms of the study and signs the written informed consent.
The clinical inclusion criteria for each premolar-molar interdental site were: (i) Accessibility of the interdental site for the four sites (15-16, 25-26, 35-36, and 45-46) by the interdental brush in each subject, (ii) no interdental caries or prosthetics restorations, (iii) no interdental diastema, (iv) no clinical sign of inflammation such as redness, swelling, or bleeding on probing (BOP) after 30 s, (v) no pocket depth (PD) > 3 mm or clinical attachment loss (CAL) > 3 mm, and (vi) the subjects were judged to be free of gingivitis or periodontitis.
The exclusion criteria were: (i) Smokers, (ii) subjects with any other concomitant systemic disease, (iii) subjects with daily medication, (iv) subjects with an orthodontic appliance, (v) subjects who have taken antibiotics in the past three months, (vi) subjects regularly using interdental brushes and/or dental floss and/or mouthwash, and (vii) subjects unable to answer questions and non-cooperative.

Ethical Approval and Informed Consent
This study was carried out in accordance with the ethical committee of Sud-Est VI Clermont-Ferrand (Approval number: AU1371) and the National Commission of Informatics and Liberties, France (2116544 v 0). The National Agency for the Safety of Medicines and Health Products (ANSM) approved it on February 13, 2017 (ID-RCB ref: 2017-A00425-48). This study was registered with ClinicalTrials.gov (identification number ID: NCT03700840).

Classification of Subjects According to Carious Risk
One trained and calibrated dentist, experienced in the clinical indices, realized the clinical examination. Subjects were classified into two carious risk groups: Low and high risk. In this study, the classification of subjects according to their carious risk was adapted from the caries-risk assessment of the American Academy of Pediatric Dentistry that is based on biological, protective and clinical factors [27].
First, to know the biological and protective factors, the dentist interviewed the patient face-to-face. Questions referred to socioeconomic status (mother s and father's occupation based on French socio-economic classifications), snacks, brushing, regular visits and dental care (Table 1).
Then, a clinical examination was performed to fulfil the questions concerning the clinical factors (Table 1). The participants were asked to refrain from oral hygiene measures, eating and drinking for two hours before clinical examination and interdental sampling.
The presence of active caries and interproximal lesions were measured by the International Caries Detection and Assessment System (ICDAS). This clinical scoring system allows the detection and the assessment of caries activity. ICDAS was developed for use in clinical research, clinical practice and for epidemiological purposes. This scoring system can be used on coronal surfaces and root surfaces and can be applied to enamel caries, dentine caries, non-cavitated lesions (contrary to many systems) and cavitated lesions. The ICDAS II system has two-digit coding for the detection criteria of primary coronal caries. The first one is related to the restoration of teeth and has a coding that ranges from 0 to 9. The second digit ranges from 0 to 6 and is used for coding the caries. A compressed air syringe was used to dry the teeth during the ICDAS assessment [28].
The salivary tests were performed for all subjects using Saliva-Check BUFFER (GC, Sucy-en-Brie, France). The tests aimed to investigate hydration, salivary consistency, resting saliva pH, stimulated saliva flow, stimulated saliva pH and saliva buffering capacity. All tests were performed according to the instructions of the manufacturer to detect low, moderate, or high salivary risk.
The subject was classified as "high carious risk" if at least one "yes" was selected in the column "high risk". The subject was classified as "low carious risk" only if no "yes" was selected in the column "high risk" for the biological and clinical findings, and "yes" was selected for the protective factors.

Clinical Examination and Interdental Sample Collection
For all subjects, the same four interdental sites (15-16, 25-26, 35-36, and 45-46) were assessed (total 200 sites). The interdental diameter was determined using a dedicated probe-the CURAPROX IAP calibration probe (Curaden, Kriens, Switzerland). This probe is a graduated conical instrument with a rounded end. The working portion includes colored bands from the tip to the base corresponding to interdental brushes (IDB) by increasing diameter. The largest section of each colored band corresponds to the cleaning efficiency diameter of the respective brush. This made it possible to select the calibrated IDB (Curaden) appropriate to the diameter of the interdental space. Each previously selected tooth was isolated using sterile cotton rolls, and the interdental biofilm as removed using this sterile IDB. For each sample, the IDBs were placed in 1.5-mL sterile microcentrifuge tubes and stored at 4 • C for further laboratory treatment.
The Bleeding on Interdental Brushing Index (BOIB) [29] was evaluated on the four interdental sites (15-16, 25-26, 35-36, and 45-46), as was the bleeding response to the horizontal pressure applied in the interdental area by a calibrated IDB. After 30 s, bleeding at each gingival unit was recorded according to the following scale: 0, absence of bleeding after 30 s; and 1, bleeding after 30 s. Then, interdental diameters and the BOIB were evaluated for all other interdental sites.
Clinical measurements including first the Gingival Index (GI) and, second, the Plaque Index (PI), were performed on the 4 sides (buccal, lingual/palatal, mesial, distal) of 6 teeth (12,16,24,36,32,44) (Silness-Löe Index) [30]. To evaluate the GI, the tissues surrounding each tooth was divided into 4 gingival scoring units: Distal facial papilla, facial margin, mesial facial papilla and lingua gingival margin. A periodontal probe was used to assess the bleeding potential of the tissues with a score from 0-3 (0: Absence of inflammation to 3: Severe inflammation). The scores of the four areas of the tooth were summed and divided by four to obtain the GI for the tooth. The index for the patient was calculated by summing the indices for all six teeth and dividing by six. Then, the determination of the PI recorded both soft debris and mineralized deposits on the 6 teeth. Each of the four surfaces of the teeth gave a score from 0-3. The scores from the four areas of the tooth were added and divided by four in order to give the plaque index (0: no plaque to 3: abundance of soft matter within the gingival pocket and/or on the tooth and gingival margin). The index for the patient was obtained by summing the indices for all six teeth and dividing by six.

Total Deoxyribonucleic Acid (DNA) Extraction
Total DNA was isolated from the interdental brushes using the QIAcube HT Plasticware and Cador Pathogen 96 QIAcube HT Kit (Qiagen, Hilden, Germany), according to the manufacturer's guidelines. The elution volume used in this study was 150 µL. DNA quality and quantities were measured using an ultraviolet spectrophotometer at 260 and 280 nm. The DNA sample was considered pure if the A260/A280 ratio was in the range of 1.8-2 and the A260/A280 ratio was in the range of 2-2.2. Simplex quantitative real-time PCR assays were performed in a volume of 10 µL composed of 1× SYBR Premix Ex Taq™ (Tli RNaseH Plus) (TaKaRa, Shiga, Japan), 2 µL of DNA extract and each primer at 1 µM. The bacterial species-specific PCR primers used in this study were provided by Institut Clinident SAS (Aix-en-Provence, France) and manufactured by Metabion international AG (Planegg, Germany). The bacterial primers used in this study are derived from sequences already published and have been adapted to the real-time PCR conditions (Supplementary Table S1).
The assays were performed on the Rotor-Gene Q thermal cycling system (Qiagen, Hilden, Germany) with the following program: 95 • C for 30 s, followed by 40 cycles of 10 s at 95 • C, 10 s at the appropriate annealing temperature (Supplementary Table S1), and 35 s at 72 • C. For the total bacterial load and that of all species, a final melting curve analysis (70-95 • C in 1 • C steps at 5 s increments) was performed. Fluorescence signals were measured every cycle at the end of the extension step and continuously during the melting curve analysis. The resulting data were analyzed using Rotor-Gene Q Series software (Qiagen, Hilden, Germany).
Serial dilutions of a bacterial standard DNA provided by Institut Clinident SAS (Aix-en-Provence, France) were used in each reaction as external standards for the absolute quantification of the targeted bacterial pathogens. The standard bacterial strains used for standard DNA production came from DSMZ (Germany), CIP Collection of Institut Pasteur or from BCMM/LMG Bacteria Collection: Aa The pathogenic strains were cultivated on the appropriate selective media. The total number of cells (number of colony-forming units) was enumerated three times using a Neubauer chamber. Serial dilutions ranging from 10xE+2 to 10xE+12 cells were utilized, and each of these dilutions was enumerated in duplicate. The DNA from each of these dilutions was extracted. A standard curve for each pathogen was generated as a plot between the crossing point (cycle number) and the initial cell count. The absolute counts of pathogen were determined using these calibration curves [31].
The limit of quantification (LOQ) of the method is summarized in Supplementary Table S1.

Sample Size
With an alpha error of 5% (2-sided test), a power of 80%, an intraclass correlation coefficient of 0.8, and a mean difference of bacteria counts between the two caries risk groups of 1,300,000, a total of 200 sites (which means 50 subjects i.e., 25 subjects per caries risk group) was necessary.

Statistical Tests
The statistical analysis consisted of three main steps: Producing descriptive summaries of the data, modeling the data using a mixed (linear) model and assessing the correlations between bacterial abundances. Prior to these steps, we transformed the original count data to handle missing data points, namely, the measurements that fell under the quantification threshold (LOQ) of the quantitative real-time PCR device. The missing values for a given species were replaced by half of the corresponding quantification thresholds given in Supplementary Table S1. We performed simulations to ensure that this simple strategy provided a reasonable estimation of the mean and standard deviation of the original count distribution. To test for potential effects of gender, interdental space, and the location of each site, we used a mixed linear model for the log-count abundance of each species at a measured site. This model includes three categorical variables as fixed effects (gender, mouth location, and interdental space) and one categorical variable as a random effect (subject). This random effect was introduced for a subject to model the correlation between the four sites of a given subject. Each coefficient in the regression was tested against the null hypothesis, which indicates that the coefficient is zero using a likelihood ratio test, and we reported that p-values less than 0.05, 0.01, and 0.001 were low, medium, and strong evidence against the null hypothesis, respectively. To perform the correlation analysis, we used the residuals of the model described above to avoid over-estimating the inter-site correlation (sites from the same patient are positively correlated, and we observed that fixed effects can also induce a correlation among sites). The trees associated with the correlation plot were obtained by hierarchical clustering with complete linkage. The difference between the two groups of caries risk relative to age, gender, mouth location, interdental space, BOIB, PI and GI were tested with chi-square tests. Kruskal-Wallis tests were performed to compare the mean counts for the different bacterial species relative to each clinical characteristic.
All statistical analyzes and associated plots were performed using the R environment (R Core Team, 2015), specifically the lme4 package [32], to estimate the mixed model. Table 2 summarizes the age, the sex, and clinical assessments of the study groups. The two groups were each composed of 25 subjects. The age, gender and clinical characteristics were similar in the two groups (Mann-Whitney and chi-square tests). The low caries risk (LCR) group was composed of 18 females and 7 males, whereas the high caries risk (HCR) group was composed of 15 females and 10 males. The mean age was 16.98 ± 0.77 years. The mean number of teeth present was 27.6 ± 1.1 (excluding wisdom teeth). Missing teeth (1.3%) were due to orthodontic treatment or agenesis (0.95%) and caries (0.35%). The mean BOIB score in the low risk group was 96.52% and 96.55% in the high-risk group. No problems due to oral hygiene were observed (fair plaque index 1.0-1.9, gingival index 1.1-2.0). The interdental spaces studied had a diameter between 0.8 and 1.1 mm.

Quantification of the Total Genome Count and Bacteria Count According to Carious Risk
The mean counts for the total bacterial load and that of the 26 evaluated species in the interdental biofilm according to carious risk are reported in Figure 2 and Table 3. An average of approximately 10 9.4 bacteria was collected in one interdental space for both groups. The quantity of total bacteria and of bacteria tested was not significantly modified except for T. forsythia, E. corrodens, and S. sobrinus, which were significantly increased in the HCR group relative to the LCR group. T. forsythia was 26.3 times, E. corrodens 4.7 times and S. sobrinus 3.3 times higher in the HCR group than in the LCR group.    Table 4 describes the distribution of pathogens according to sites and subjects. Lactobacillus spp., S. salivarius, S. mitis, F. nucleatum, and Clostridium XIV were detected in all subjects and in all interdental spaces, whatever the carious risk. Other bacteria were not expressed in each interdental site nor in each subject. For example, P. gingivalis was detected in 11% of HCR subjects and 12% of interdental sites from HCR subjects, but this bacterium was not detected in LCR subjects. E. corrodens was observed in all subjects, in 100% of interdental sites from HCR subjects and in 96% of interdental sites from LCR subjects. In the HCR group, all species tested were detected. The most abundant species were F. nucleatum (10 7.4 bacteria in one ID space), S. salivarius (10 7.05 bacteria in one ID space) and S. sanguinis (10 7.05 bacteria in one ID space). The least abundant species were Clostridium IV (10 0.52 bacteria in one ID space), A. actinomycetencomitans (10 0.55 bacteria in one ID space), and S. sobrinus (10 0.56 bacteria in one ID space).

Pathogen Correlations According to Carious Risk
The dendrogram (Figure 4) underscores the correlations between the 26 pathogenic species and the ID sites for the HCR group and the LCR group. Even after the removal of the fixed effects related to interdental space and age, and the subtraction of the inter-site correlations, the matrix still revealed a strong correlation structure, which appeared as two groups (or clusters) of correlated species for the HCR group and one for the LCR group.

Discussion
This study provides a comprehensive survey of the interdental microbiota in adolescents, a target group that remains poorly explored, according to carious risk.  The first cluster of the HCR group was composed of C. gracilis, P. micra, C. rectus, P. gingivalis, A. actinomycetemcomitans, V. parvula, and F. nucleatum, whereas the second cluster was composed of Lactobacillus spp., Clostridium XIV, S. mitis, S. sanguinis, A. odontolyticus, and S. salivarius.

Discussion
This study provides a comprehensive survey of the interdental microbiota in adolescents, a target group that remains poorly explored, according to carious risk. To our knowledge, there is no scientific reference in the literature that jointly targets a cross-sectional clinical study (MIARC), according to the bacteriological criteria of interdental healthy adolescents, caries risk factors and the use of a real-time PCR technique. A few studies focused on the carious microbiota of adolescents, but these studies were concerned with the subgingival microbiota without specific location, during and/or after orthodontic treatment [33,34], and quantified few bacteria (A. actinomycetemcomitans, P. gingivalis, P. intermedia, T. forsythia) [35].
Most research investigating the commensal oral microbiome are focused on disease or are restricted in methodology. To diagnose and treat caries at an early and reversible stage, an in-depth definition of health is indispensable [36]. Our research option consisted of carrying out real-time PCR analysis of the interdental microbiota of caries-free sites in healthy adolescents with or without a carious risk. All pathogens considered in our study have been before identified in oral samples of children, adolescents, or young adults [37][38][39][40]. Our study did not only focus on cariogenic bacteria because the classification of oral bacteria seems more complex. Effectively, some studies demonstrated a positive association between periodontitis and caries, whereas others demonstrated a negative association [41][42][43]. These studies observed clinical signs, and only one study considered the microbiota, but the subjects were clinically affected by these oral diseases [44]. Qualitatively and quantitatively, the presence of 26 oral pathogens was analyzed. Among them were bacteria commonly classified as cariogenic bacteria-B. dentium, Lactobacillus spp., R. dentocariosa, S. cristatus, S. mutans, S. salivarius, S. sobrinus, S. wiggsiae; bacteria considered periodontopathogenic-bacteria from the purple complex (A. odontolyticus, V. parvula), the green complex (A. actinomycetemcomitan, C. orchrocea, E. corrodens), the yellow complex (S. mitis, S. sanguinis), the orange complex (C. rectus, C. gracilis, F. nucleatum, P. intermedia, P. micra, P. nigrescens), the red complex (P. gingivalis, T. denticola, T. forsythia); and others such as Clostridium IV and Clostridium XIV.
A particular focus of our study is a Caries Risk Assessment (CRA) system. Indeed, the carious lesion is a multifactorial disease principally due to carious biofilm and sugar consumption. Caries-risk assessment models currently involve a combination of factors including diet, fluoride exposure, a susceptible host, and microflora that interplay with a variety of social, cultural, and behavioral factors [26]. The initiation of dental caries results from the balance between risk and protective factors. This interplay between factors underpins the classification of individuals and groups into caries risk categories, allowing an increasingly tailored approach to care. The classification criteria for CRA systems were determined by combining scientific evidence and expert opinion. From these CRAs, practitioners analyze the various clinical and social factors of a patient and can thus assign a carious risk status [45]. For low-risk patients, there is no necessity for further preventive professional treatment, and they should be offered an extended follow-up [46]. For high-risk patients, preventive actions must be taken to reduce the incidence and severity of future carious lesions [46]. This individual scheduling of preventive and follow-up activity better appropriates the use of dental resources and lowers dental costs for certain individuals [46]. At present, the criteria needed to perform quantitative caries risk assessment evaluations are still incomplete [26].
Few CRA classifications consider the quantification of bacteria, and generally such bacteria are singular and unique. In these few CRA classifications, the quantification of S. mutans in the saliva has been considered [47]. However, the results concerning the link between S. mutans and the development of dental caries is not clear. Some studies demonstrated a real association between S. mutans and the carious lesion, whereas others revealed no clear association [48]. Moreover, the amount of S. mutans needed to initiate the carious lesion varies according to the study [49]. Thus, to use this bacterium or other bacteria as a biomarker in a CRA classification, it is necessary to identify specific biomarkers associated with carious lesions and to normalize their quantification.
Concerning bacteria classified as cariogenic bacteria (B. dentium, Lactobacillus spp., R. dentocariosa, S. cristatus, S. mutans, S. salivarius, S. sobrinus, S. wiggsiae), no significant difference was observed in the HCR subjects relative to the LCR subjects. Therefore, these bacteria do not represent good predictive factors for CRA. Cariogenic bacteria are acidogenic and acid-tolerant species [50].
For many years, S. mutans was considered the major oral pathogen implicated in the initiation of caries. However, this role has been questioned [51]. Effectively, caries have been observed without the presence of S. mutans [50,[52][53][54]. In our study, S. mutans was expressed in small quantities (10 1.97 for the LCR and 10 1.95 for the HCR group). Our results confirmed that S. mutans alone could not be used to predict carious risk as previously demonstrated by Gross and colleagues [50]. A CRA classification that uses the quantification of S. mutans in the saliva must be used with care [21]. In some studies, caries was associated with the absence of S. mutans and the presence of Lactobacillus [55,56], B. dentium [55], and S. wiggsiae [57].
S. sobrinus, closely related to S. mutans, is described as a major bacterium in the apparition of carious lesions [58]. This pathogen was significantly increased by three times in the HCR group relative to the LCR group, and it could also be a predictive marker for caries, as described by Gross and colleagues for the supragingival biofilm [50]. However, contrary to this previous study, this pathogen was expressed only in 20% of the subjects from the HCR group, which limits its detection and its potential to be an interesting predictive marker of caries.
S. salivarius has been associated with caries due to its high cariogenic capacity [55]. Moreover, in a rat model, this pathogen was able to induce caries. Contrary to these results, in our study, this pathogen was expressed in all subjects at equal levels in both groups [59]. These divergent results could be explained by the fact that some S. salivarius strains such as S. salivarius M18 have probiotic activity and help to fight against cariogenic bacteria [60].
S. wiggsiae is classified as cariogenic because it is able to tolerate acid and to produce acid from several sugars at low pH [65]. This pathogen was significantly associated with severe-early childhood caries and was also observed in adolescents presenting initial carious lesions with fixed orthodontic appliances [61,62]. Contrary to these studies, our results indicated that S. wiggsiae was highly expressed whatever the carious risk and in more than 84% of subjects in both groups.
S. cristatus is considered an important cariogenic species due to its association with childhood caries [57,63]. Our results are in accordance with those of Dzidic and colleagues because, the quantity of S. cristatus is 13 times higher in HCR subjects but not significantly [63]. This pathogen is known to be inversely correlated to the presence of P. gingivalis [64]. Our results are in accordance with this because P. gingivalis represented 10 0.35 counts in one ID space, whereas S. cristatus represented 10 4.42 counts in one ID space.
The relationship between Lactobacillus and caries is well established [65,66]. Lactobacillus spp. produces water-insoluble polysaccharides that promote bacterial attachment to the tooth surface. Therefore, other bacteria are able to fix the organic acids that are confined that modify the microenvironment to enrich the aciduric microflora [67]. Although lactobacilli play an important role in the prognosis of caries, they are unlikely to play an important role in the development of caries [68,69]. Therefore, the number of salivary lactobacilli may be indirectly related to the progression of caries. Our study revealed that Lactobacillus spp. was present in all caries-free subjects, but the rate was higher in the HCR group, although not significant.
B. dentium was isolated from the oral biofilm and from saliva [70]. Bifidobacterium was isolated from 80% of plaque samples from early childhood caries patients [55,71] and from the saliva of children with a higher frequency in caries-affected children than in caries-free children [72]. B. dentium was defined as a novel caries-associated bacterium with an acidogenic potential and a high fluoride tolerance [73]. Our results revealed that this pathogen is only in 24% of the subjects in the HCR and LCR groups, and no significant difference was observed between the two groups. Therefore, this bacterium cannot be used for CRA.
R. dentocariosa was identified as a cariogenic bacteria [74,75]. Jiang and colleagues observed that the quantity of this pathogen was higher in the saliva from children with caries than caries-free children. They suggested that it could be a predictive marker for CRA. However, our results indicated that R. dentocariosa was expressed in 96% of subjects from the HCR group and 100% of subjects from the LCR group with no significant difference in quantity between the two groups. Therefore, this pathogen alone cannot be used for CRA.
Studies agree that oral diseases are not only attributable to specific bacteria but that they result from dysbiosis of the oral microbiota. The apparition of oral disease could be due to the decrease of bacterial taxon and to the presence of key pathogens [76,77].
Concerning bacteria classified as periodontal bacteria, as previously described, bacteria from the blue, yellow, green, and purple complexes of Socransky are associated with oral health, whereas bacteria from the orange and red complexes are associated with dysbiosis of the microbiota and the apparition of gingivitis and, later, periodontitis [78]. Our study reveals that bacteria associated with Socransky complexes and thus with gingivitis and periodontitis are present in adolescents that present no signs of gingivitis. The higher quantity of bacteria from the red complex in the HCR group than the LCR group indicated a positive correlation between periodonthopathogenic bacteria and carious risk.
Previously, Durand and colleagues demonstrated a positive correlation between periodontal disease severity and the decayed, missing, filled teeth surfaces index [44].
Concerning the red complex, our results indicated that P. gingivalis was not detected in the subjects of the LCR group and was only detected in 11% of the interdental sites in the HCR group. This could be explained because in both groups the quantity of S. mutans (10 1.97 in the LCR group and 10 1.95 in the HCR group) and S. sanguinis (10 6.94 in the LCR group and 10 7.05 in the HCR group) were high compared with quantity of P. gingivalis (not detected in the LCR group and 10 0.68 in the HCR group). A previous study demonstrated that the growth of P. gingivalis was significantly inhibited by supernatants from S. mutans and S. sanguinis [79]. Moreover, S. cristatus (10 4.97 in the HCR group) could also inhibit the growth of P. gingivalis because S. cristatus is able to inhibit the expression of virulence genes of P. gingivalis and consequently delay the growth of dental plaques [80].
T. denticola was detected in low quantities (10 2.17 in the HCR group) and was higher in the HCR group than the LCR group. This bacterium was detected in interdental sites expressing bacteria from the red or the orange complex. This is in accordance with previous studies that demonstrated that T. denticola is unable to adhere on oral surfaces as it is only able to colonize to form oral biofilms when other periodontal bacteria are present. In subgingival dental biofilms, T. denticola is typically associated with P. gingivalis [81]. A chymotrypsin-like proteinase found within a high-molecular-mass complex on the cell surface of T. denticola mediates its adherence to other potential periodontal pathogens such as P. gingivalis, F. nucleatum, P. intermedia and P. micra [82].
T. forsythia was expressed in 96% of adolescents (83% of interdental sites) from the HCR group and was the only bacteria from the red complex that significantly increased in the HCR group relative to the LCR group. This bacterium could represent an interesting biomarker for the CRA system. However, as it was expressed in 93% of subjects and 83% of interdental spaces from the HCR group, some false negatives could appear. T. forsythia acts with P. gingivalis to initiate chronic periodontitis and was not described in carious lesions. In our study, T. forsythia was not correlated with P. gingivalis and so could act with other bacteria such as S. cristatus and R. dentocariosa, which are known to be implicated in caries [74,83]. Moreover, T. forsythia was correlated with T. denticola, confirming previous results demonstrating that T. denticola and T. forsythia are involved in protein-protein interactions through a leucine-rich repeat in proteins [84].
Bacteria from the orange complex do not seem to have a key role in carious risk because no significant difference was observed by comparing the HCR and the LCR groups. In fact, these bacteria have a key role in the initiation of subgingival microbiota dysbiosis [85], but no role in carious lesions was described.
P. nigrescens was previously detected in children caries [86], which is in accordance with our results indicating that this pathogen was detected in 76% of subjects from the HCR group. For this pathogen, sex significantly impacted carious risk. A higher quantity was detected in females than males and in the female HCR group relative to the female LCR group. Nakagawa and colleagues previously demonstrated that P. nigrescens was significantly increased in puberty compared with prepuberty in females but not in males. This could be associated with sex hormone modifications [87].
Contrary to the study of Gross and colleagues that demonstrated that C. rectus decreased as caries progressed, our study revealed no modification according to caries risk that could be explained because those authors sampled supragingival biofilm from carious lesions, whereas we analyzed interdental biofilms associated with a healthy surface [56]. However, our study revealed that C. rectus increased significantly in the HCR group with the BOIB, suggesting that this pathogen could also be associated with interdental inflammation. Indeed, C. rectus is known to be present in chronic gingivitis [88].
From the yellow complex, S. mitis and S. sanguinis were not significantly modified when the HCR and the LCR group were compared. This result aligns with those from previous studies indicating that S. sanguinis and S. mitis are significantly associated with dental health [89][90][91]. S. sanguinis is known to be a pioneer colonizer that permits the adhesion of other microorganisms and biofilm formation [50,92,93]. Our results indicated that the quantity of S. sanguinis is 90 times higher than that of S. mutans. This result is in accordance with the study of Caufield and colleagues, who demonstrated a higher level of S. sanguinis relative to the level of S. mutans in the saliva [94].
Of the green complex bacteria studied, only E. corrodens was significantly increased with carious risk. Moreover, this pathogen was expressed in each subject and in each interdental space from the HCR group and in 96% of subjects from the LCR group. These results are in accordance with those from Choi and colleagues, who demonstrated that this pathogen was present in the saliva from periodontally healthy subjects [95]. Previous studies have demonstrated that E. corrodens is important for the progression of periodontal disease in young subjects [95,96]. This bacterium participates in the early stage of biofilm formation by mediating the specific co-aggregation of bacteria from the oral cavity. It is a middle colonizer, linking early colonizers such as Streptococci and late colonizers such as P. gingivalis [97]. As our results indicated that this pathogen could be a key factor for the carious process and that it was phylogenetically close to the cluster (Lactobacillus spp., Clostridium XIV, S. mitis, S. sanguinis, A. odontolyticus, and S. salivarius) from the HCR group, it could participate, as previously described in periodontitis, in biofilm formation and could permit cariogenic bacteria to adhere to biofilm. In our study, females from the HCR group had a higher quantity of E. corrodens than females from the LCR group. This could be explained because females have higher caries prevalence than males [98][99][100][101]. Moreover, E. corrodens is associated with systemic diseases such as spinal, head, and neck infection or endocarditis [102][103][104][105][106]. Therefore, it could be a predictive marker of carious disease but also of systemic diseases.
Concerning the purple complex, V. parvula and A. odontolyticus were not linked with carious risk but were expressed in more than 90% of the interdental spaces tested. Groos and colleagues demonstrated that V. parvula increased with the carious lesion step [50], and Kanasi and colleagues identified it in carious lesions from children [107]. V. parvula is important in biofilm formation. It can co-aggregate with other microorganisms such as S. mutans. Indeed, V. parvula cannot fix itself on the surface of teeth and attaches to S. mutans [108]. A. odontolyticus also acts in the early formation of biofilm [109]. A. odontolyticus was previously detected in carious lesions [86,110,111].
A recent study demonstrated that Clostridium was positively and significantly correlated to caries and pigment in primary dentition [112]. However, even in our study, all subjects expressed Clostridium XIV, while only 13% of the HCR group and 9% of the LCR group expressed Clostridium IV, and no significant differences were observed between the HCR and LCR groups. However, Clostridium XIV was 2 times higher in the HCR than in the LCR group, which could be in accordance with the study by Li and colleagues [112].
The human oral microbiota has recently become a new focus for its involvement in systemic diseases [113,114]. Our work underlines the presence in the interdental niche of healthy adolescents of a number of oral bacteria described in the literature for their association with systemic diseases. Effectively, B. dentium is known to cause bacteraemia [115]. A. odontolyticus has been identified in actinomycosis throughout the body [116,117]. R. dentocariosa and S. mutans are known to be associated with bacteraemia and infective endocarditis [118,119]. E. corrodens has been described in cardiovascular diseases (CVD) [120,121], and C. ochracea has also been described in lupus [122,123]. F. nucleatum has been related to CVD, Alzheimer s, and adverse pregnancy [124]. P. gingivalis was linked to CVD, respiratory tract infection, cancer, Alzheimer s, rheumatoid arthritis, and adverse pregnancy [124,125]. CVD and Alzheimer s were also associated with P. intermedia, T. denticola, and T. forsythia, which was also implicated in respiratory tract infection [124]. C. rectus and V. parvula were linked to CDV [124]. P. nigrescens was discovered in rheumatoid arthritis [125]. Therefore, the accumulation of these pathogenic bacteria in the interdental space must be considered. These bacteria represent a risk-predisposing factor to the development of systemic diseases in the future. As interdental hygiene is often neglected (the toothbrush is not used enough), it is necessary to introduce interdental brushing as early as adolescence.

Conclusions
Estimating the risk of caries associated with bacterial factors in interproximal sites in adolescents will permit a more evidence-based strategy for medical referrals for certain individuals and contribute to a better definition of carious risk status, periodicity and intensity of diagnostic, prevention and restorative services. In our study, the analysis of the interdental microbiota from a sample of adolescents differentiated by level of caries risk highlights the potential role that three bacteria (S. sobrinus, E. corrodens, and T. forsythia) could have on the predictive development of interproximal caries. E. corrodens appears to be more interesting because it was found in all interdental sites from the HCR group and, it had the lowest p-value when comparing HCR and LCR groups. Moreover, the detection of S. sobrinus and T. forsythia could give false negatives because they were not detected in all the interdental sites and all the adolescents for the HCR group. While existing tools for caries risk assessment and interproximal adolescence caries prediction have restricted manageable clinical value, introducing a biological, non-subjective and personalized marker could be recommended to increase the CRA classification. Sensitive assessment of metabolic process using microbial biomarkers at the biofilm-enamel interface could thus perform it possible to specify endpoints prior to the clinical expression. The knowledge obtained from our trial contributes to generating active hypotheses related to the composition and variability of the oral microbiome in the interdental space of adolescents. These hypotheses can be prospectively explored in longitudinal cohort studies, using next generation sequencing techniques, essential to identify other biomarkers of the disease in real time.

Acknowledgments:
We acknowledge the support of our work by Institut Clinident sas (Aix-en-Provence, France) and all the participants.

Conflicts of Interest:
The authors declare no conflict of interest.