Caries Increment and Salivary Microbiome during University Life: A Prospective Cohort Study

The purpose of this 3-year prospective cohort study was to explore the relationship between an increase in dental caries and oral microbiome among Japanese university students. We analyzed 487 students who volunteered to receive oral examinations and answer baseline (2013) and follow-up (2016) questionnaires. Of these students, salivary samples were randomly collected from 55 students at follow-up and analyzed using next-generation sequencing. Students were divided into two groups: increased group (Δdecayed, missing, and filled teeth (ΔDMFT) score increased during the 3-year period) and non-increased group (ΔDMFT did not increase). Thirteen phyla, 21 classes, 32 orders, 48 families, 72 genera, and 156 species were identified. Microbial diversity in the increased group (n = 14) was similar to that in the non-increased group (n = 41). Relative abundances of the family Prevotellaceae (p = 0.007) and genera Alloprevotella (p = 0.007) and Dialister (p = 0.039) were enriched in the increased group compared with the non-increased group. Some bacterial taxonomic clades were differentially present between the two groups. These results may contribute to the development of new dental caries prevention strategies, including the development of detection kits and enlightenment activities for these bacteria.


Introduction
Dental caries caused by some pathogens of the oral microbiome is a widespread disease. In the Global Burden of Disease 2015 Study, untreated caries in permanent teeth was the most prevalent condition, affecting 2.5 billion people worldwide [1]. The cause of dental caries is multifactorial. In addition to major microbial species mutans streptococci (predominantly Streptococcus mutans and S. sobrinus), physical, biological, environmental, behavioral, and lifestyle-related factors are risk factors for dental caries [2].

DNA Extraction and NGS Analysis
At follow-up, unstimulated saliva (>1 mL) was randomly collected into sterile ice-chilled 15 mL tubes from students (from 09:00 to 16:00) before the dental examination and frozen at −80 • C until analysis. To prevent possible protein dilution, unstimulated saliva was collected instead of stimulated saliva [15][16][17]. A random number list was used for selecting participants for saliva collection. The saliva of selected participants was not collected at baseline. Saliva DNA was collected using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions using sterile equipment and DNA removal reagents for laboratory instruments. Collected DNA was stored at −20 • C for further analysis. For amplified bacterial DNA, V3 and V4 regions of the 16S rRNA gene were amplified using primers 357F (5 -TCGTCGGCAGCGTCAGATGT GTATAAGAGACAGCCTACGGGNGGCWGCAG-3 ) and 781R (5 -GTCTCGTGGGCTCGGAGATG TGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3 ) at Okayama University Hospital Biobank (Okayama University Hospital, Okayama, Japan) according to the standard protocol using the MiSeq platform (MiSeq Reagent V3 600 cycles, Illumina, San Diego, CA, USA). Quality of raw sequence reads was checked using FastQC (version 0.11.3, Babraham Bioinformatics, Cambridge, UK) and analyzed using USEARCH (version 8.0.1623, https://www.drive5.com/usearch/) at the Oral Microbiome Center (Taniguchi Dental Clinic, Kagawa, Japan). After removing chimeric reads, duplicated reads, and short reads <400 bp, preprocessed reads of each sample were clustered into operational taxonomic units (OTUs) at 97% level of nucleotide similarity using UCLUST algorithm to determine the number of OTUs. Furthermore, these reads were analyzed to identify human oral taxa using the Human Oral Microbiome Database (version 14.5; http://www.homd.org/).

Statistical Analyses
We estimated the sample size for the saliva examination based on a previous study [4] using G*Power (version 3.1.9.4, Düsseldorf, Germany). The difference in α diversity (Simpson index; caries group: 0.075 ± 0.019, healthy group: 0.099 ± 0.037) was selected as the primary outcome [4]. To calculate the effect size, mean and standard deviation of 0.024 ± 0.018 was considered to detect a difference in α diversity between the two communities. Based on the data, the minimum sample size required was 52 to provide a power of 91% with an alpha of 0.05 by t-test. Participants were divided into two groups based on the change in DMFT score during the 3-year follow-up period; participants with ∆DMFT > 0 were categorized into the "increased group," while participants with ∆DMFT = 0 were categorized into the "non-increased group" [18]. The normality of the data was confirmed by the histogram and quantile-quantile plot. The Mann-Whitney U test and Fisher's exact test were used to determine the presence of significant differences in variables of oral examination and questionnaire between increased and non-increased groups. Associations between variables and dental caries were examined in a series of logistic regression models, and the odds ratio (OR) and 95% confidence interval (CI) were calculated. Logistic regression models were reviewed for goodness-of-fit and validated using the Hosmer-Lemeshow statistic. A p-value < 0.05 was considered significant. Statistical analyses were performed using SPSS (version 25.0; IBM, Tokyo, Japan).
Alpha diversity was determined using R (version 3.4.3; The R Project for Statistical Computing, http://www.R-project.org). The species richness of saliva microbiota of individuals was measured by Chao1 and the abundance-based coverage estimator (ACE) indices. The diversity of the saliva microbiota was measured by Shannon and Simpson indices. Microbial beta diversity was visualized by principal coordinates analysis using the Calypso software tool (http://bioinfo.qimr.edu.au/calypso/). The Mann-Whitney U test was used to assess significant differences in the abundance of taxa between increased and non-increased groups. Rarefaction curves were calculated using Calypso to compare microbial richness among samples. To compare the microbial composition between increased and non-increased groups, relative abundances (%) were calculated from the taxonomic abundance count divided by preprocessed reads. Linear discriminant analysis (LDA) effect size (LEfSe) methods were used to identify taxa with differentiating relative abundance using the online interface Galaxy (http://huttenhower.sph.harvard.edu/lefse/). The threshold for the logarithmic LDA score for biomarker discovery was set at 2.0 [19,20].

Participant Characteristics
The study flowchart is shown in Figure 1; 293 students (follow-up rate: 13.4%; 149 males, 144 females; mean age and standard deviation at baseline: 18.2 ± 0.77 years) were analyzed (Table 1). At follow-up, the DI-S score of selected students (n = 55) was significantly lower than that of all participants (n = 293; p < 0.05, t-test). No significant differences were observed in the other parameters evaluated between selected students and all participants (p < 0.05, Mann-Whitney U test and Fisher's exact test).

Participant Characteristics
The study flowchart is shown in Figure 1; 293 students (follow-up rate: 13.4%; 149 males, 144 females; mean age and standard deviation at baseline: 18.2 ± 0.77 years) were analyzed (Table 1). At follow-up, the DI-S score of selected students (n = 55) was significantly lower than that of all participants (n = 293; p < 0.05, t-test). No significant differences were observed in the other parameters evaluated between selected students and all participants (p < 0.05, Mann-Whitney U test and Fisher's exact test).
In all participants, DMFT scores, DI-S score at baseline, and the percentage of students who had knowledge of fluoride at follow-up were significantly higher in the increased group than those in the non-increased group (p < 0.05; Table 2). In the logistic regression analysis, increase in dental caries was significantly related to DMFT score (OR: 1.191, 95% CI: 1.033-1.373, p = 0.016) and DI-S score (OR: 3.6, 95% CI: 1.355-6.913, p = 0.007) at baseline. participants. At baseline, 2187 students received the oral examination. Because follow-up oral examination was not mandatory, only 487 students received a second oral examination. Of these students, 55 students were randomized and consented to saliva collection. Figure 1. Flowchart of study participants. This flowchart shows the process for selecting analyzed participants. At baseline, 2187 students received the oral examination. Because follow-up oral examination was not mandatory, only 487 students received a second oral examination. Of these students, 55 students were randomized and consented to saliva collection. In all participants, DMFT scores, DI-S score at baseline, and the percentage of students who had knowledge of fluoride at follow-up were significantly higher in the increased group than those in the non-increased group (p < 0.05; Table 2). In the logistic regression analysis, increase in dental caries was significantly related to DMFT score (OR: 1.191, 95% CI: 1.033-1.373, p = 0.016) and DI-S score (OR: 3.6, 95% CI: 1.355-6.913, p = 0.007) at baseline.

Salivary Microbiome Analysis
In our study, 3,195,127 quality-filtering reads (58,093 ± 16,532) from 55 saliva samples were used for analysis. A total of 196 OTUs were obtained from saliva samples. Of these, 13 phyla, 21 classes, 32 orders, 48 families, 72 genera, and 156 species were identified. There were no significant differences in the number of each taxonomic level between increased and non-increased groups (p ≥ 0.05). Furthermore, there were no significant differences in the species richness of individuals between increased and non-increased groups (Figure 2a; p ≥ 0.05). Moreover, no significant differences were observed in species diversity and bacterial communities between the two groups (Figure 2b,c; p ≥ 0.05). The increased group was defined as ∆DMFT > 0 during the 3-year follow-up period; The non-increased group was defined as ∆DMFT = 0 during the 3-year follow-up period; DMFT: decayed, missing, and filled teeth; DI-S, debris index-simplified. 1 Data are expressed as mean ± standard deviation. 2 Data are expressed as n (%). 3 Fisher's exact test or Mann-Whitney U test.
differences in the number of each taxonomic level between increased and non-increased groups (p ≥ 0.05). Furthermore, there were no significant differences in the species richness of individuals between increased and non-increased groups (Figure 2a; p ≥ 0.05). Moreover, no significant differences were observed in species diversity and bacterial communities between the two groups (Figure 2b,c; p ≥ 0.05).

Figure 2.
Comparison of the oral microbiome diversity between increased and non-increased groups. (a) Saliva microbiome diversity between increased (red) and non-increased groups (blue) was compared. Alpha diversity metrics for Simpson, Shannon, Chao1, and ACE indices were calculated and illustrated by box plots. There were no significant differences in diversity between increased and non-increased groups (t-test, p ≥ 0.05). (b) Rarefaction curves of increased (red) and non-increased (blue) groups based on the observed operational taxonomic unit (OTU). There were no significant differences in species richness between increased and non-increased groups (t-test, p ≥ 0.05). (c) Principal coordinates analysis (PCoA) based on the Bray-Curtis index. Increased (red) and non-increased (blue) groups did not tend to separate.

Microbial Composition
There were no significant differences in microbial composition at class and order levels between increased and non-increased groups. At the phylum level, the increased group had a higher population of Proteobacteria than the non-increased group (Table 3; p = 0.029). At the family level, the (a) Saliva microbiome diversity between increased (red) and non-increased groups (blue) was compared. Alpha diversity metrics for Simpson, Shannon, Chao1, and ACE indices were calculated and illustrated by box plots. There were no significant differences in diversity between increased and non-increased groups (t-test, p ≥ 0.05). (b) Rarefaction curves of increased (red) and non-increased (blue) groups based on the observed operational taxonomic unit (OTU). There were no significant differences in species richness between increased and non-increased groups (t-test, p ≥ 0.05). (c) Principal coordinates analysis (PCoA) based on the Bray-Curtis index. Increased (red) and non-increased (blue) groups did not tend to separate.

Microbial Composition
There were no significant differences in microbial composition at class and order levels between increased and non-increased groups. At the phylum level, the increased group had a higher population of Proteobacteria than the non-increased group (Table 3; p = 0.029). At the family level, the increased group had a higher population of Prevotellaceae than the non-increased group (Table 3; p = 0.007). At the genus level, the increased group had a higher population of Actinobaculum, Dialister, and Alloprevotella than the non-increased group (Table 3; p < 0.05). At the species level, the increased group had higher populations of Neisseria sicca, Alloprevotella sp., Dialister invisus, Cardiobacterium hominis, Acinetobacter sp., Gracilibacteria (GN02) [G-1], Neisseria elongate, Actinomyces graevenitzii, Anaerolineae [G-1], Dialister pneumosintes, Haemophilus haemolyticus, Actinobaculum sp., Corynebacterium matruchotii, Prevotella pleuritidis, and Neisseria sp. than those in the non-increased group (Table 3; p < 0.05). Low percentages of S. mutans in saliva were detected in both groups. In LEfSe analysis, Prevotellaceae and Veillonellaceae at the family level and Alloprevotella and Dialister at the genus level were enriched in the increased group compared with the non-increased group (Figure 3).  (Table 3; p < 0.05). Low percentages of S. mutans in saliva were detected in both groups. In LEfSe analysis, Prevotellaceae and Veillonellaceae at the family level and Alloprevotella and Dialister at the genus level were enriched in the increased group compared with the non-increased group (Figure 3).

Discussion
Various groups have provided information about the relationship between dental caries and the oral microbiome [4,[7][8][9][10][21][22][23][24][25]. However, there is less information about this relationship in young adults or university students. In the present 3-year cohort study, the abundance of several taxa was significantly higher in the increased group than that in the non-increased group. In addition, poor oral hygiene and caries experience at baseline were significantly associated with an increase in dental caries during university life.
In this study, the observed taxa number was not significantly different between increased and non-increased groups. Moreover, the microbial structure was similar between the two groups. Jiang et al. reported that the microbial composition was similar between caries progression and non-progression groups among older people [23]. The findings from our study support these results.
The percentages of several taxa were significantly higher in the increased group than those in the non-increased group. In particular, further analysis using LEfSe methods revealed families Prevotellaceae and Veillonellaceae and genera Alloprevotella and Dialister were significantly enriched in the increased group compared with the non-increased group (LDA scores ≥ 3.0). Prevotellaceae was detected in carious dentin or saliva of participants with caries progression [24,25]. Furthermore, Eriksson et al. also reported that Veillonellaceae and Dialister were enriched in the saliva of patients with caries progression using the LEfSe method [20]. Alloprevotella was detected in the saliva of children with caries progression using LEfSe [4]. Using non-culture methods, non-mutans streptococci have been detected in carious dentin or saliva [26]. Furthermore, Prevotellaceae and Veillonellaceae species are known to produce acid [5,27]. These results indicate that several acid-producing bacteria or non-mutans streptococci may be associated with caries progression, which suggests new targets for preventing caries progression.
In this study, there were no significant differences in microbiome diversities between increased and non-increased groups. Previous studies exploring the oral microbiome to elucidate potential targets for caries prevention mainly targeted children with deciduous dental plaque or saliva. For example, a 2-year cohort study revealed that microbial diversity in saliva did not differ between caries progression and non-progression groups [10]. However, some cohort studies reported that salivary microbiome diversity was significantly different between participants with increased dental caries compared with those with non-increased dental caries [7,21,22]. According to these reports, whether salivary microbiome diversity is associated with caries progression remains controversial. Furthermore, few studies included young adults as subjects. A cross-sectional study exploring salivary microbiomes among families, including young adults, reported that dental caries did not impact microbial diversity [28]. While our findings support the results of this study, further cohort studies are required to investigate the relationship between microbiome diversity and an increase in dental caries.
Oral condition is also associated with caries progression. In a cohort study of adult participants in Sweden, DMFT scores of the caries progression group were higher than those in the non-progression group at baseline [29]. In addition, a previous study reported that the increased caries group had worse oral hygiene than that of the non-increased caries group [30]. Here, we demonstrated that DMFT and DI-S scores were associated with an increase in dental caries in the logistic regression analysis, supporting the results of these previous studies.
In this study, a low abundance of mutans streptococci was detected among the two groups. Mutans streptococci are traditionally recognized as the most common cause of dental caries [31]. Several studies also reported that there was no difference in the abundance of these species between caries-active and caries-free groups [7][8][9]32]. Furthermore, Takahashi and Nyvad proposed the "Ecological Plaque Hypothesis," wherein acid-producing bacteria, excluding mutans streptococci, are also considered to lead to an imbalance in mineralization [26]. However, we did not investigate the microbiome in dental plaques of participants with caries. Further studies are therefore required.
In this study, we used saliva samples for microbiome analysis. Saliva collection is non-invasive, simple, and effective for mass examination. Some studies reported that plaque bacteria are released into saliva [9,33], and saliva provides a niche for both anaerobic and aerobic bacteria [34]. Furthermore, saliva is a biomarker that reflects oral health and systemic condition [35,36]. Thus, saliva evaluation has several advantages, especially in large-scale epidemiological studies or caries screening.
Our study has some limitations. First, there may have been selection bias, given the low follow-up rate (13.4%). In addition, the number of participants who had regular dental checkups at baseline and DI-S scores at follow-up differed between selected participants and total participants. However, increased dental caries was not related to these variables in the logistic regression analysis. Therefore, the selection bias might be small in this study. Second, all participants were recruited from among students who attended Okayama University. Therefore, our results may not be generalizable. Third, we could not investigate potential confounders, such as social capital [37,38] or socioeconomic status [39]. Fourth, we did not use a specific primer and probe for detecting S. mutans. Using the Human Oral Microbiome Database, we could detect the taxa using obtained sequencing reads. However, a previous study mentioned that V1-V2 or V3-V4 hypervariable regions (approximately 400-460 bp) is the limit for identifying detailed level of taxa [19]. Therefore, a wider region or specific primer is needed [19]. Fifth, we did not set a specific time for saliva collection; that is, saliva was collected from 09:00 to 16:00. However, we randomly selected participants and collected saliva at the last step of the general health examination. Therefore, we do not consider that eating or drinking affected salivary analysis. Finally, we could not collect saliva at baseline; however, we checked caries increment using DMFT at baseline and follow-up. Although Belstrøm et al. reported the stable condition of the salivary microbiome [40], we cannot rule out the possibility of changes in the salivary microbiome during the 3-year study period.

Conclusions
Among Japanese university students, non-mutans streptococci bacteria in saliva were associated with increased caries in this 3-year prospective cohort study. However, there were no significant differences in salivary microbiome diversity between increased and non-increased groups. Funding: This work was financially supported by a Grant-in-Aid for Scientific Research (no. 15K11415) from the Ministry of Education, Culture, Sports, Science, and Technology, Tokyo, Japan.