Functional profiles of coronal and dentin caries in children

ABSTRACT Background: Dental caries results from a dysbiosis of tooth-associated biofilms and frequently extends through enamel into dentin which has a different structure and composition. Objective: To evaluate the metatranscriptome of caries to determine the metabolic potential of caries communities compared with health. Design: Samples from children, caries-free (CF: n = 4) or with coronal (CC: n = 5) or dentin (DC: n = 5) caries were examined for gene expression potential. Functional profiling was performed using HUMAnN2 (HMP Unified Metabolic Analysis Network). Results: There was increased gene expression diversity in DC compared with CC and CF. Genes in CF included alcohol dehydrogenase from Neisseria sicca, methylenetetrahydrofolate reductase from Streptococcus sanguinis and choline kinase from streptococci. Genes in CC mapped mainly to Streptococcus mutans. Arginine deiminase in DC mapped to S. sanguinis and Actinomyces naeslundii. Glycerol kinase genes mapped to S. sanguinis in all groups whereas glycerol kinase in DC were from Rothia, Prevotella and streptococci. Uracil-DNA glycosylase in DC mapped to Prevotella denticola and Actinomyces. Repressor LexA in DC mapped to Scardovia wiggsiae, Dialister invisus and Veillonella parvula. Conclusions: Functional profiling revealed enzyme activities in both caries and caries-free communities and clarified marked differences between coronal and dentin caries in bacterial composition and potential gene expression.


Introduction
Dental caries in children persists as a significant dental and public health challenge in selected, mainly lower income communities [1,2]. Caries results from an imbalance in the oral microbiota with increased acid production from bacterial metabolism after intake of dietary carbohydrates [3]. This increased acid production alters the composition of the microbiome with the suppression of acid-sensitive bacteria [4,5] which changes the bacterial community's ability to counterbalance the lower local pH in active caries. This pH buffering of the healthy microbiome has been found from bacterial ammonia production from arginine deiminase (ADS) and urease activities [6][7][8][9].
The low pH at the tooth surface leads to enamel demineralization, then enamel cavities which can progress into the tooth dentin. The composition of enamel differs from that of dentin, and caries progression in the less mineralized dentin is not associated with as low a pH as observed for enamel caries [10]. Several highly acidogenic and acid-tolerant bacteria have been associated with caries, including Streptococcus mutans and Streptococcus sobrinus and certain Lactobacillus, Bifidobacterium, Scardovia and Actinomyces [5]. The microbiota of dentin caries differs from that of enamel caries and includes Lactobacillus and Prevotella species [11] and caries progression can result from acidic and proteolytic activity [10].
The metatranscriptome refers to the gene expression of the whole microbial community [12]. Previous studies have shown differences in microbial gene expression based on sample site within the oral cavity [13] between health and periodontitis [14] and dental caries [15,16]. Microbial gene expression activity in dentin can differ markedly from that in enamel and closer to the tooth surface [16]. These studies focused on dental disease in adults.
The goal of this pilot study was to determine the in situ activity of the microbial community and associated species for dental caries in children. Samples representing health and caries at the tooth surface and deep in dentin were examined to represent different disease stages. In caries, one might expect increased gene expression related to acid production [17], whereas in health-associated microbiomes gene expression would reflect that of a stable community able to counter acidic attack [6]. Long-term we sought which metatranscriptome activities characterized different clinical states and so that future studies could use the information in risk assessment and as targets for improved anti-caries therapies.

Study population
Medically healthy 3-to 8-year-old children were recruited from the Community Dental Center in Waterville, Maine. Children in this study had not used antibiotics within the last 3 months, and the parent or guardian consented to the child's clinical examination and microbial sampling. The children's primary caregivers completed a survey that incorporated caries risk factors including: demographics, socioeconomic status, dental history and dietary habits [18,19]. The study was performed in full accordance with ethical principles including the World Medical Association Declaration of Helsinki and the study design, protocol, and informed consent were approved by the Institutional Review Board at the Forsyth Institute.

Clinical characteristics
The caries, plaque and gingival status of children were charted (Table 1). Children were measured at six monthly intervals for up to 2 years and data examined for development of new carious lesions as evidence of caries progression.

Microbiological methods
Sterile wooden toothpicks were used to collect biofilm samples from supragingival sites of incisors and molars, including from lesions if present, from coronal caries (CC: n = 5) and caries-free (CF; n = 4) children [20]. Dentin caries (DC: 2 = 5) samples were taken with an excavator during cavity preparation prior to restoring the tooth. Bacterial samples were collected into individual tubes containing 100 µl RNAlater® (Life technologies, New York). The sample tubes were frozen at −20°C at the clinic and shipped on dry ice to the microbiology laboratory where they were stored at −80°C until processed.

Sample preparation
Sample-containing tubes were centrifuged to collect cells and the RNAlater® was removed by two wash steps with 500 µl DEPC PBS followed by centrifugation at 12,000 × g for 10 min. The following kits and procedures were used to purify RNA following the manufacturer's instructions: MolYsis® (

Sequencing
Enriched mRNA was prepared for RNA sequencing (RNASeq) using 5 µl of undiluted 'mRNA' added to 13 µl of Fragment Prime Finish Mix and the Illumina TruSeq Stranded mRNA Low Sample Protocol was run according to the manufacturer's instructions (Illumina, San Diego, CA). The prepared samples were run on the Bioanalyzer and their molarity estimated based on the reverse transcription results using the Kapa Library Quantification Kit. The libraries were normalized, pooled and diluted to a 2 ŋM starting concentration. The pooled diluted library was then denatured, diluted, and loaded onto an Illumina NextSeq 500 platform following the manufacturer's protocol. The samples were sequenced on the NextSeq 500 using Illumina's High output 150 cycle v2 cartridge.

Bioinformatics analysis
The raw RNASeq reads (in FASTQ format) sequenced from the 14 samples were analysed with the software package 'HUMAnN2', the HMP Unified Metabolic Analysis Network (http://huttenhower.sph. harvard.edu/humann2) [21]. HUMAnN is a pipeline for efficiently and accurately profiling the presence/ absence and abundance of microbial pathways in a community from metagenomic or metatranscriptomic sequencing data. This process, referred to as functional profiling, aims to describe the metabolic potential of a microbial community and its members. HUMAnN used 'MetaPhlAn' [22] to map RNAseq reads to a set of unique clade-specific marker genes identified from~17,000 reference genomes (~13,500 bacterial and archaeal,~3,500 viral, and~110 eukaryotic). The reads mapping was performed using 'Bowtie2' software [23]. HUMAnN then took the organism-specific gene hits generated by MetaPhlAn and computed the gene family and pathway abundance, and pathway coverage. For comparison, relative abundance of read counts was based on 1 million reads per sample. The results were stratified by organism at the species level, or to genus if not species specific. The gene family definitions were based on the UniRef database [24] and the pathway definitions were based on the MetaCyc database [25].
Stratified species-level transcriptomic profiles were subject to alpha-and beta-diversity analyses, including the rarefaction curves of alpha diversity, principal component analysis (PCoA) beta-diversity of samples among different disease and health groups, and the sequence abundance heatmaps, using the scripts provided by the Quantitative Insights into Microbial Ecology (QIIME) software package [26].

Statistical analysis
Clinical and demographic data The caries charting measurements were used to calculate total decayed and filled teeth scores for primary (dft) and secondary (DFT) dentitions. A child was considered to have caries progression if a new cavity was detected in the 2-year monitoring period. Clinical and survey (demographic and diet) data were compared between caries-free, caries and dentin groups by the non-parametric statics Kruskal-Wallis test. Mean taxon levels of abundant species and enzymes detected were compared by clinical group using pairwise Mann Whitney U tests (SPSS version 22). Species that are scientifically represented among comparison groups were identified with the Linear Discriminant Analysis Effect Size (LEfSe) software package [22].

Study population
Data was analysed from 14 children ranging in age from 3.4 to 7.1 years (Table 1). Six children were girls, and most were within the normal weight range although three of the children sampled were overweight. Children were non-Hispanic, Caucasian, and mothers were the primary caregivers. Samples were obtained from four caries-free children and five children with coronal caries at their first visit, and from six dentin cavities from children at a follow-up visit. One child was sampled at tooth surface and at a later visit from a dentin cavity (#6).
There was no difference in background demographic or dental history of diet data between children in the three groups, except that parents of children with caries all reported having cavities. The children sampled with dentin caries had lower plaque and gingivitis scores, but sampling was at a follow-up (visit 3) when they had received oral hygiene instructions, not the first visit after enrolled in the study. There was no association with overweight status with caries in the sampled children (Table 1).

Expressed gene sequences mapped to microbial taxa
The alpha diversity of gene expression of taxa (operational taxonomic units, OTUs) from samples varied between clinical conditions (Figure 1a). While there was similar diversity in caries-free and coronal caries sites there was increased diversity in the dentin caries. Principal component analysis of species grouped the samples from tooth surfaces, caries-free and coronal caries, together (Figure 1b, purple circle). The dentin samples were quite widely spread from the coronal samples and from each other.
Gene expression sequences mapped to 164 taxa that comprised 108 named species, 12 phylotypes, 33 bacterial genus-level identifications and 13 viruses. The cladogram of LEfSe (Linear discriminant analysis effect size) analysis showed that Actinobacteria, Neisseriales and viruses were detected in the dentin caries, Betaproteobacteria in caries-free and with one species in Carnobacteriaceae that includes lactobacilli in coronal caries (Appendix Figure a). The bar chart of LEfSe analysis illustrated the wide diversity of taxa detected in dentin caries, whereas only taxa in Carnobacteriaceae and Granulicatella differentiated coronal caries, and Neisseriaceae, Neisseriales, Betaproteobacteria and All Bacteria characterized the caries-free samples (Appendix Figure b).

Expressed gene sequences mapped to enzymes (functional profiling)
RNA sequences were mapped in a total of over 1200 genes coding for enzymes. The mean levels of major genes detected in descending order (by read counts normalized by samples) in coronal caries (CC) are given in Figure 3a and supplemental Table 2. At an exploratory threshold (p > 0.05 Mann Whitney U test) several genes were elevated in either caries-free or caries (CC and DC). S. mutans sequences were mapped mainly to CC (Figure 3b) Caries-free (CF) Alcohol dehydrogenase mapped mainly to the species Neisseria sicca in one (#16) (29.3 units) of the caries-free children ( Figure 3c). S. sanguinis genes in three of the four caries-free samples accounted to the higher levels of methylenetetrahydrofolate reductase (NAD(P)H). Choline kinase genes were detected in Streptococcus infantis, Streptococcus mitis (mitis/oralis/pneumoniae) and Streptococcus sp. GMD4S mainly from one cariesfree child 17H.
Dentin caries (DC) Threonine synthase in S. mitis was the dominant gene observed for coronal caries (aforementioned). While not significantly different in overall levels between clinical groups, genes for ADS mapped to several species in dentin caries including S. sanguinis: dentin caries (DC) 14.3, caries-free (CF) 9.8 and coronal caries (CC) 4.7 (Figure 3d

Discussion
The ability to examine functional profiling of the oral biofilm community focuses on addressing the question 'What are the microbes in my community-ofinterest doing (or capable of doing)?' in addition to providing a new approach to examine the oral microbiota in health and disease [27]. Compared with 16S profiling, metatranscriptomic analysis can provide a better picture of the species that are metabolically active in different caries types [11] and thus revealing the community function which is more relevant to disease states. This approach revealed species identifications consistent with bacterial identifications from samples using culture and molecular methods, providing a measure of validity to the functional profiling approach to describing the microbiota. Considering genes from S. mutans and S. wiggsiae, while S. mutans was elevated in coronal and dentin caries compared with caries-free, S. wiggsiae was associated with dentin caries suggesting that S. wiggsiae was primarily active in dentin lesions. This latest information adds to our knowledge of the ecology of S. wiggsiae in dental caries. While there were few differences in gene expression between the coronal caries and caries-free samples, there was increased and more diversity in genes detected in dentin caries.

Sample site
The genes detected in coronal caries and caries-free children clustered together suggesting an overlap in gene expression of the microbiotas on the tooth surface. This contrasted with the genes of dentin caries, which differed from each other and reflected a wider range of metabolic capabilities. Differences in the microbiome composition from enamel (coronal) to dentin caries and increased diversity in dentin caries was observed in a metagenomic analysis [13] as in the current report. The pH of dentin caries was found to vary considerably ranging from pH 4.5 to over pH 7, with the different microbiotas being associated with various levels of acidity [28], consistent with the diversity in the microbiota in dentin caries sites observed of the current report.

Health and disease
Detection of genes in S. mutans, R. dentocariosa, several Actinomyces species and S. salivarius in coronal caries compared with caries-free sites is consistent with other reports [5,[29][30][31][32]. Detection of genes to these species fit with the expanded ecological plaque hypotheses that proposes blooming of Actinomyces and non-mutans streptococci in early stages of dental caries [17]. Under this hypothesis, these species set up the environment for more acidtolerant and acidogenic species that would include S. mutans, S. wiggsiae [33], Actinomyces sp. HOT 448 [34] observed in dentin caries in the current report. S. wiggsiae is a relatively newly recognized species that has been detected in caries in several reports from the US [5], Europe [35] and Brazil [36]. Other highly acidogenic, lactate-producing species in dentin included A. rimae and A. parvulum [37] although the acid tolerance of these species is unknown. Other less acidogenic species with genes detected in dentin, included Prevotella and Capnocytophaga species and S. longum [38], likely reflecting the higher pH deeper in dentin compared to the tooth surface, and proteolytic activity related to dentin caries [10]. Detection of genes in dentin mapping to A. oris, P. denticola, A. parvulum and the genus Rothia in dentin was as previously reported [11]. Abiotrophia defectiva was detected in caries-free sites, although not significantly associated with health. This previously 'aberrant Streptococcus species' is clinically associated with bacterial endocarditis and has been observed in health compared with early childhood caries [39] as in the current study.
Viruses. Genes mapped to several viral taxa as previously described [27]. In the current report the principal viruses detected from gene expression sequences were the murine osteosarcoma virus and the dasheen mosaic virus which is a plant virus. It is not clear what their role, if any, may be in dental caries, or oral pathology.

Functional profilescommunity gene expression
The greater diversity of genes that was detected in dentin compared with caries-free or caries samples from coronal sites was linked with detection of more enzymes and taxa in dentin. This is consistent with observations of acidogenic and proteolytic activity in dentin caries [10,40].

Health and disease
Alcohol dehydrogenase, methylenetetrahydrofolate reductase, and choline kinase genes were detected at higher levels in caries-free compared to coronal or dentin caries. Alcohol dehydrogenase mapped mainly to N. sicca which produces this enzyme [41], Alcohol dehydrogenase can play a role in neutralizing bacterial acid production through reduction of nicotinamide adenine dinucleotide (NAD + to NADH). This suggests an alternative mechanism to NH 3 production to counteract community bacterial acid in health. Genes for methylenetetrahydrofolate reductase, which is involved in methionine production, mapped principally to S. sanguinis and while the gene has been detected in bacteria no specific role in the oral biofilm community was found. Choline kinase genes mapped mainly to S. mitis as previously reported [42] and are considered to play a role in cell wall lysis. It is not clear what their role in caries-free sites would be.
The highest gene expression was in dentin caries including for ADS, glycerol kinase, uracil-DNA glycosylase and urease. ADS and urease have been associated with reduced caries activity based on ammonia production and raising the pH to counterbalance acidic conditions typical of active caries [43]. It seems likely these enzymes also involved in the higher pH found deep in in dentin compared to the tooth surface [28]. Linking ADS with S. sanguinis and urease with A. naeslundii in the microbial community is consistent with the metabolic properties of these species.
Glycerol kinase is part of the glycolysis pathway rather than that for refined carbohydrates as reported for dentin caries [44]. The roles of the other enzymes in caries is unclear.
The functional profiling approach used in this study could identify potential gene activity and associated species and indicate likely community activity in health and disease both previously recognized or reveal new enzyme activities. ' Next generation RNA sequencing (RNA-seq)' provides an unbiased and culture-free method to study the gene expression of a microbial community. The high depth of Illumina sequencing makes it robust to identify genes expressed a low level but play key roles in regulating metabolic functions and pathways. The relatively high cost associated with the Illumina sequencing has been a limitation for this technology, but costs have been reducing with the popularity and the increasing high throughput of this approach. This renders Illumina sequencing possible for metatranscriptomic studies as in this report.
A major limitation of the current report was the small number of sites examined so it is unclear how the findings are applicable overall. Further, unlike study of the transcriptome from bacterial culture that allows for a sufficient amount of RNA, the small clinical samples size led to limited amounts of RNA that might also impact the results. The transcriptome profile might only reveal the relatively abundant functionality, but not that of the less abundant, but nevertheless could potentially have more biologically significant impact. Moreover, the resultant DNA yield from the small sample sizes precluded simultaneous metagenome analysis. Additionally, samples were collected without consideration of food intake, and if enzyme activity related acid production was limited or enhanced to periods just after fermentable carbohydrate intake that could have been missed. We suggest future study to consider timing of cariogenic food intakes particularly for coronal caries.

Conclusion
Metatranscriptomic analysis by functional profiling confirmed previous observations about key bacteria and cariogenic mechanisms that were previously known, and revealed some new enzyme activities by species in both caries and caries-free communities. Comparing coronal and dentin caries clarified the marked differences between these communities in both bacterial composition and functional profiles.