Plaque Microbiome in Caries-Active and Caries-Free Teeth by Dentition

Abstract

Objective:

Describe associations between dental caries and dental plaque microbiome, by dentition and family membership.

Methods:

This cross-sectional analysis included 584 participants in the Center for Oral Health Research in Appalachia Cohort 1 (COHRA1). We sequenced the 16S ribosomal RNA gene (V4 region) of frozen supragingival plaque, collected 10 y prior, from 185 caries-active (enamel and dentinal) and 565 caries-free (no lesions) teeth using the Illumina MiSeq platform. Sequences were filtered using the R DADA2 package and assigned taxonomy using the Human Oral Microbiome Database.

Results:

Microbiomes of caries-active and caries-free teeth were most similar in primary dentition and least similar in permanent dentition, but caries-active teeth were significantly less diverse than caries-free teeth in all dentition types. Streptococcus mutans had greater relative abundance in caries-active than caries-free teeth in all dentition types (P < 0.01), as did Veillonella dispar in primary and mixed dentition (P < 0.01). Fusobacterium sp. HMT 203 had significantly higher relative abundance in caries-free than caries-active teeth in all dentition types (P < 0.01). In a linear mixed model adjusted for confounders, the relative abundance of S. mutans was significantly greater in plaque from caries-active than caries-free teeth (P < 0.001), and the relative abundance of Fusobacterium sp. HMT 203 was significantly lower in plaque from caries-active than caries-free teeth (P < 0.001). Adding an effect for family improved model fit for Fusobacterium sp. HMT 203 but notS. mutans.

Conclusions:

The diversity of supragingival plaque composition from caries-active and caries-free teeth changed with dentition, but S. mutans was positively and Fusobacterium sp. HMT 203 was negatively associated with caries regardless of dentition. There was a strong effect of family on the associations of Fusobacterium sp. HMT 203 with the caries-free state, but this was not true for S. mutans and the caries-active state.

Knowledge Transfer Statement:

Patients’ and dentists’ concerns about transmission of bacteria within families causing caries should be tempered by the evidence that some shared bacteria may contribute to good oral health.

Introduction

Dental caries is the most prevalent oral disease in both children and adults and adversely affects quality of life (Frencken et al. 2017). The burdens of poor dental health are felt particularly among underserved cultural minority groups, such as those found in Appalachia, where rates of dental caries and other oral health problems are high and access to care is limited (McNeil et al. 2012). Cariogenesis is a multifactorial process that depends on behavior and the oral microbiome, factors that are shared with family members.

Oral microbiome diversity (the number of different species and evenness of their relative abundance) increases with age and dentition (Holgerson et al. 2020). Supragingival plaque from caries-active teeth is reportedly less diverse than that found in plaque from caries-free teeth. A study of 919 Chinese children reported reduced plaque diversity with caries in primary and mixed dentition (Shi et al. 2016), but a comparison of plaque from 64 caries-active and 64 caries-free Kuwaiti children found no difference (Qudeimat et al. 2021). Another Chinese study comparing dental plaque from 24 caries-active to 22 caries-free adults 60 y and older also found no difference in diversity (Jiang et al. 2019). A third Chinese study comparing plaque from 62 adults with decayed, missing, and filled teeth (DMFT) ≥8 to 29 adults with DMFT = 0, however, found no difference in plaque diversity but greater numbers of species (richness) in the caries group (Xiao et al. 2016).

Results from the few studies assessing the effects of family on oral microbiome diversity suggest that family is an important determinant of diversity. A Korean study of 7 sibling pairs, in which 1 child had experienced caries in 5 or more teeth and the other did not, simultaneously assessed the association of caries with diversity and the effect of family (Lee et al. 2016). While diversity was significantly lower among samples from caries-active children, the overall plaque composition was more similar among sibling pairs than nonsiblings. This finding is consistent with a study of 25 sibling pairs concordant and discordant for caries, in which there was a strong effect of sibship on the salivary metabolome after accounting for nested profiles of both dentition and dental decay (Foxman et al. 2014). Furthermore, this finding held true in a study assessing the supragingival plaque microbiome in twin pairs, where the similarity of the microbiome always increased with shared host genotype, regardless of caries state (Gomez et al. 2017).

Streptococcus mutans was the first species isolated from caries lesions and is strongly associated with caries (Clarke 1924; Loesche et al. 1975). However, a recent scoping review of human studies comparing plaque microbiome composition in caries-active and caries-free children and adults found S. mutans was significantly associated with caries in only half the studies reviewed (Bhaumik et al. 2021). This is consistent with the ecological hypothesis of caries, which posits that S. mutans is only part of a complex process that includes the dynamic microbial community and physiological and behavioral changes (Philip et al. 2018).

Our study objective is to provide further insight into the associations between dental caries and dental plaque diversity and composition, by dentition and family membership. We analyzed supragingival plaque samples from 584 children and adults participating in the Center for Oral Health Research in Appalachia Cohort 1 (COHRA1) between November 2003 and December 2009. Samples were collected from caries-free teeth (no lesions), caries active lesions, or both (n = 750). For this study, teeth with enamel (D2) or dentinal (D3) as classified using World Health Organization (WHO) criteria were considered caries active. We compare supragingival plaque diversity and composition from caries-active and caries-free teeth within the same individual, within families, and among unrelated individuals by dentition.

Methods

Study Population

The COHRA1 is a family-based study, conducted jointly by the University of Pittsburgh and West Virginia University. The protocol is described in detail elsewhere (Polk et al. 2008). Briefly, households with 1 parent with at least 1 biological child were recruited in Pennsylvania and West Virginia between November 2003 and December 2009. There were 3,652 participants in COHRA1. Participants completed comprehensive surveys of demographic information, medical and dental history, and hygiene habits and were examined by a dental health professional.

Dietary data were not collected, measures of fluoride in the water had high rates of missing values, and regular topical application of fluoride was not assessed. At the time of study collection, the International Caries Detection and Assessment System (ICDAS) was not yet published (Ismail et al. 2007); therefore, caries classification was based on WHO criteria: D1, precavitated (e.g., white spot) lesions; D2, enamel lesions; D3, dentinal lesions; and D4, pulpal lesions (World Health Organization 1979). Appendix Figure 1 lists WHO and ICDAS criteria and COHRA1 categorization.

The standardized periodontal and caries screen by calibrated dentists and hygienists included collection of supragingival plaque from any white spot, enamel, dentinal, or pulpal lesions present. Participants were instructed to not eat or brush their teeth 2 h before the appointment. Supragingival plaque samples for each person from each type of lesion were collected using a scaler, pooled by each type of lesion, and stored separately by lesion type. We did not have tooth location information. A pooled sample of plaque from caries-free supragingival surfaces was collected using a sterile Stimudent toothpick and placed in TSB-YE and 80% glycerol solution. Plaque from white spot, enamel, or dental lesions was collected using a sterile Gracey ½ curette tip. Samples were frozen in the glycerol solution at −80°C for at least 10 y before processing in freezers continuously monitored for temperature control. The study protocol was approved by the Institutional Review Boards at West Virginia University and the University of Pittsburgh. Written informed consent was obtained from all adult participants and parents of participating children at the time of specimen collection. This study conforms to STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cross-sectional studies.

Selection of Study Samples

We selected for inclusion supragingival plaque sampled from children with and without enamel or dentinal caries. We defined caries active as enamel or dentinal lesions (we observed no pulpal lesions). Caries free was defined as no lesions (no white spots). A sample from the enamel or dentinal carious lesion and the caries-free teeth (if available) was included from those with an active lesion (Fig. 1). Among the 568 children included, there were 136 sibling-pairs (n = 272), 62 children with multiple siblings (n = 218), and 78 singlets. In addition, there were 16 adults included (parents of children).

Figure 1.

Distribution of samples from caries-active lesions (enamel or dentinal) and caries-free teeth (no lesions). Between-individual comparisons are 428 caries-free samples to 156 caries-active lesions in all other individuals (boxes a to b, respectively). Within-individual comparisons are 56 caries-active lesions to 28 caries-free lesions (box c). Selected individuals participating in the Center for Oral Health Research in Appalachia Cohort 1 from Pennsylvania and West Virginia (n = 584).

DNA Extraction and Sequencing

Samples were thawed on ice, and DNA was extracted directly from thawed samples. If no liquid was present in the stored plaque sample following thawing (due to drying during storage), 300 µL of cell lysis solution (Promega) was added. Then, 100 µL of plaque-containing liquid was added to an enzyme mastermix containing lysozyme (15 mg/mL), mutanolysin (375 U/mL), RNase A (150 µg/mL), and lysostaphin (240 µg/mL) (all from Sigma-Aldrich). DNA was extracted using a Qiagen Qicube enzymatic lysis and DNA extraction protocol. DNA was quality control checked using a NanoDrop2000 (Thermo Scientific). The V4 region of the 16S ribosomal RNA (rRNA) gene was sequenced using the Illumina MiSeq platform, with a dual-indexing sequencing strategy (Kozich et al. 2013). The library preparation kit was MiSeq Reagent Kit V2 500 cycles (cat. MS102-2003; Illumina). Each plate of samples was submitted with a positive mock community control, a DNA extraction kit control, and a negative water control. Both positive and negative controls were sequenced simultaneously. Sequences have been submitted to the NCBI Sequence Read Archive (accession number: PRJNA811423).

Bioinformatics

We filtered sequences using the DADA2 package in R and assigned taxonomy using the Human Oral Microbiome Database (HOMD) reference database (Escapa et al. 2020). Briefly, DADA2 uses a naive Bayesian classifier method to assign taxonomy to the sequence variants (Callahan et al. 2016). We removed amplicon sequence variants (ASVs) that were not bacterial and collapsed ASVs assigned to the same species together. ASVs of unknown species were given their assigned ASV number from the DADA2 package. After quality filtering, there were 13,932,494 reads (1,125 to 60,402 per sample), with an average read count of 18,284, and our sequencing was sufficiently deep to detect most of the taxonomic diversity (Appendix Fig. 2). We identified 188 bacterial ASVs.

Statistical Analysis

To examine between-individual differences in plaque microbiomes, we included 428 plaque samples from caries-free teeth from individuals without an active caries lesion (cells in Fig. 1 marked with “a”) and the 156 samples from caries-active teeth from everyone else (if an individual had more than 1 sample collected, we randomly selected 1 caries-active lesion from their mouth) (cells in Fig. 1 marked with “b”). To compare within-individual differences among multiple active caries lesions in a mouth to a caries-free tooth, we compared plaque samples from the 28 individuals who had multiple active caries measured (cells in Fig. 1 marked with a “c”). Dentition was categorized as primary (primary teeth only within a mouth), mixed (primary and permanent teeth within a mouth), and permanent (permanent teeth only within a mouth).

We assessed α diversity using the Shannon Diversity Index and tested for differences in diversity using the Wilcoxon signed-rank test. We assessed β diversity using the Bray–Curtis dissimilarity metric, performed ordination analysis with Bray–Curtis distances, and drew principal coordinate analysis (PCoA) graphs. We tested statistically significant differences in β diversity by permutational multivariable analysis of variance (PERMANOVA: 9,999 permutations). For every individual with 2 carious lesions and 1 caries-free tooth, we calculated the Bray–Curtis dissimilarity distances separately between the carious lesion and caries-free tooth, then averaged them. To identify bacterial species driving differences in composition between plaque from caries-free teeth and caries-active teeth by dentition state, we used ALDEx2 software (Fernandes et al. 2013). ALDEx2 uses the centered log-ratio (clr) transformation that ensures the data are scale invariant and subcompositionally coherent, and it corrects for multiple hypothesis testing using Benjamini and Hochberg–corrected P value (Benjamini and Hochberg 1995). We set the false discovery rate to 0.1 and P < 0.01.

To account for the effect of dentition, DMFT, race, and family smoking on individual and family variation in plaque microbiomes, we fit a series of linear mixed models predicting the log-transformed relative abundance of selected species by caries status. Selected species were the top 4 significantly different species from the ALDEx2 analysis and 2 additional species of interest identified via literature review. Likelihood ratio tests were used to determine whether each model required a random intercept for the individual alone or a random intercept for individuals within family. We also fit similar models to assess whether the abundance of selected taxa varies between enamel and dentin caries compared to caries-free teeth (118 enamel caries, 67 dentin caries, and 565 caries-free teeth). Oral hygiene and dental insurance were not included in our final models as they were not statistically significantly associated with dental caries in our data set.

Results

We analyzed 750 supragingival plaque samples from 584 people (292 families) living in Appalachia between 2003 and 2009; 135 individuals had 2 or more samples where 1 or more samples were from a caries-active lesion (enamel or dentinal) and 1 or more from a caries-free (no lesions present) tooth (Fig. 1). There were 196 people with primary dentition (average number of decayed or filled teeth [d2ft, which does not count white spots]: 1.08; average age: 3.6 y), 282 with mixed dentition (average number of decayed, missing, or filled teeth [D2MFT, which does not count white spots]: 3.00; average age: 8.8 y), and 105 with permanent dentition (average D2MFT: 3.39; average age: 15.3 y). (Dentition was missing for 1 individual.)

There were no statistically significant differences in oral health and demographic measures between participants from West Virginia and Pennsylvania (Table 1). Most participants aged 11 y and over reported toothbrushing at least once a day (94.6% of participants with mixed dentition and 84.6% with permanent dentition). There was no measure of toothbrushing among children aged 1 to 10 y, but mothers reported that 96.1% of children with primary dentition and 93.0% with mixed dentition used toothpaste at least once a day (Appendix Table 1). Most participants flossed their teeth at least once a week (51.1% with primary dentition, 63.9% with mixed dentition, and 81.1% with permanent dentition).

Table 1.

Study Population Characteristics of Selected Participants from the Center for Oral Health Research in Appalachia Cohort 1 (COHRA1) Stratified by Dentition (n = 584).

		Dentition a
Characteristic	Overall	Primary Only	Mixed	Permanent Only
Number	584	196	282	105
Age category, n (%)
1–6	251 (43.0)	194 (99.0)	57 (20.2)	0 (0.0)
7–17	317 (54.3)	2 (1.0)	224 (79.4)	90 (85.7)
18+	16 (2.7)	0 (0.0)	1 (0.4)	15 (14.3)
D2MFT, mean (SD)	2.42 (3.42)	1.08 (2.39)	3.00 (3.05)	3.39 (4.94)
Sex, n (%)
Female	278 (47.6)	94 (48.0)	123 (43.6)	60 (57.1)
Male	306 (52.4)	102 (52.0)	159 (56.4)	45 (42.9)
Race, a n (%)
African American	90 (15.4)	20 (10.2)	45 (16.0)	25 (23.8)
Asian	3 (0.5)	1 (0.5)	2 (0.7)	0 (0.0)
Hispanic	7 (1.2)	0 (0.0)	6 (2.1)	1 (1.0)
White	483 (82.7)	175 (89.3)	228 (80.9)	79 (75.2)
Filled teeth, mean (SD)	1.26 (2.30)	0.31 (1.23)	1.74 (2.41)	1.76 (2.91)
Smoker in family, b n (%)
No	317 (56.1)	111 (56.9)	162 (57.9)	44 (49.4)
Yes	248 (43.9)	84 (43.1)	118 (42.1)	45 (50.6)
Site, n (%)
Pennsylvania	210 (36.0)	59 (30.1)	105 (37.2)	45 (42.9)
West Virginia	374 (64.0)	137 (69.9)	177 (62.8)	60 (57.1)

We selected supragingival plaque samples from children under the age of 14 y with 1 or more siblings (n = 568) and a sample of parents (n = 16) of included children from 3,627 participants in COHRA1.

D2MFT, decayed, missing, or filled teeth.

^aDentition was missing for 1 participant; race was not reported for 1 participant.

^bSmoking information was missing for 19 participants.

The most common cariogenic taxa found in all samples were Veillonella dispar (93.7%), Actinomyces oris (92.8%), and Veillonella parvula (90.1%). S. mutans was found in 37.7% (Appendix Table 2). A description of the abundances of ASVs is found in Appendix Table 3.

Plaque from caries-active lesions (enamel or dentinal) is less diverse and less similar than plaque from caries-free teeth and more similar than plaque from caries-free teeth within families (Fig. 2).

Figure 2.

Comparing α diversity (Shannon index) and β diversity (Bray–Curtis distance) between plaque from caries-active (enamel or dentinal) and caries-free teeth (no lesions), by dentition type. Individuals participating in the Center for Oral Health Research in Appalachia Cohort 1 from Pennsylvania and West Virginia (n = 584). (A) The α diversity among 156 caries-active and 428 caries-free teeth from different individuals by dentition. (B) The β diversity (pairwise comparisons) among 156 caries-active and 428 caries-free teeth from different individuals by dentition. (C) The β diversity (pairwise comparisons) of caries-active and caries-free teeth within the same individual by dentition, from 28 individuals with samples from 2 caries-active teeth and 1 caries-free tooth.* (See Fig. 1 for sample groups.)

We estimated the α and β diversity of plaque from caries-active and caries-free teeth using samples from 428 caries-free teeth and 156 caries-active teeth sampled from different individuals (Fig. 1). Across all samples, plaque from caries-free teeth were significantly more diverse than plaque from caries-active teeth (Shannon index: 2.46 vs. 2.09, respectively). This was true regardless of dentition (Fig. 2A). Among those with mixed and permanent dentition, the Bray–Curtis distance was smaller (showing more similarity) among caries-free teeth than caries-active teeth (Fig. 2B). The similarity of plaque composition significantly decreased from participants with only primary dentition to mixed to only permanent dentition for caries-active lesions. For caries-free teeth, the similarity of plaque composition significantly decreased from primary-only to mixed dentition but not from mixed to permanent-only dentition. The composition of caries-free and caries-active teeth varied significantly within each dentition state (PERMANOVA P = 0.001 for primary, mixed, and permanent; Appendix Fig. 3). Furthermore, the composition of caries-free teeth (PERMANOVA P = 0.001) and caries-active teeth (PERMANOVA P = 0.004) differed statistically by dentition (Appendix Fig. 4).

Overall, the plaque microbiome was slightly more similar among family members than among unrelated individuals (Bray–Curtis: 0.777 vs. 0.79; PERMANOVA P = 0.08). This was also true when restricting the analysis to carious lesions (PERMANOVA P = 0.01) but not when restricting the analysis to caries-free teeth (PERMANOVA P = 0.34) (Appendix Figs. 5 and 6).

Within individuals with primary dentition, caries-active teeth are more similar to each other than to their caries-free tooth (Fig. 2C).

We estimated β diversity using samples from the 28 people with samples from 2 caries-active teeth and 1 caries-free tooth (Fig. 2C). Among the 10 individuals with primary dentition, the 2 caries-active teeth were more similar to each other than when each was compared to the caries-free tooth (Bray–Curtis: 0.54, 0.64, respectively). However, the reverse was true among the 17 individuals with mixed dentition, as we found that the caries-free tooth and caries-active teeth were more similar to each other than the 2 caries-active teeth (Bray–Curtis: 0.70, 0.74 respectively). (There was only 1 individual with permanent dentition and multiple samples.)

S. mutans is more abundant in caries-active teeth, and Fusobacterium sp. HMT 203 is more abundant in caries-free teeth in all dentition types (Fig. 3).

Figure 3.

Bacterial species that occur significantly more often among caries-active and caries-free teeth by dentition. Bacterial species found more frequently in plaque from (A) caries-active teeth (enamel or dentinal) and (B) caries-free teeth (no lesions) by dentition. Only species that were statistically significantly different (Benjamini–Hochberg corrected P < 0.01) are shown. Species identified using ALDEx2 analysis of between-individual comparisons (156 caries-active teeth and 428 caries-free teeth). Selected individuals participating in the Center for Oral Health Research in Appalachia Cohort 1 from Pennsylvania and West Virginia (n = 584). (See Fig. 1 for sample groups.)

We used ALDEx2 to identify species that were significantly different in centered log-transformed relative abundance between the 2 tooth groups by dentition (Fig. 3). S. mutans was statistically significantly (P < 0.001 for all dentition types) more abundant in plaque from caries-active teeth regardless of dentition; V. dispar and V. parvulla were also more abundant in plaque from caries-active teeth, but the associations were statistically significant only in those with primary or mixed dentition (P < 0.001 for both species in primary and mixed dentition). Fusobacterium sp. HMT 203 was statistically significantly more abundant in plaque from caries-free teeth regardless of dentition (primary dentition: P = 0.01; mixed dentition and permanent dentition: P < 0.001); Cardiobacterium hominis was also more abundant in plaque from caries-free teeth but was statistically significant only in those with primary (P = 0.01) or mixed dentition (P < 0.001).

Linear mixed models predict changes in abundance of selected species by caries status while accounting for variation among individuals and families adjusting for dentition, DMFT, smoking in family, and race (Fig. 4).

Figure 4.

Comparisons of the relative abundance of selected species (from ALDEx2 results and literature) found in plaque from caries-active teeth and plaque from caries-free teeth (no lesions) by caries type (enamel or dentinal). Results from linear mixed-effect models. Models adjusted for dentition (primary only, mixed, permanent only); decayed, missing, and filled teeth (DMFT); household smoking; and race with either a random effect for each individual (denoted by *) or random effects for each individual within family (denoted by **). Selected individuals participating in the Center for Oral Health Research in Appalachia Cohort 1 from Pennsylvania and West Virginia (750 samples from 584 individuals).

We used results from the ALDEx2 analysis and the literature to select species for further analysis using linear mixed models. After adjustment for variation among individuals, dentition, DMFT, household smoking, and race, using a linear mixed model, plaque from caries-active teeth had statistically significantly greater abundance ofS. mutans than caries-free teeth (P < 0.001) but a statistically significantly lower abundance of A. oris, C. hominis, and Fusobacterium sp. HMT 203 (P < 0.001 for all 3) (Appendix Fig. 7). The mixed dentition state compared to primary dentition had a statistically significant increased impact on the relative abundance of S. mutans (P = 0.002), C. hominis (P = 0.003), and Fusobacterium sp. HMT 203 (P < 0.001). Those with household smoking had a statistically significant higher relative abundance of V. dispar (P = 0.049), and the race category of White/Other had a statistically significant lower relative abundance of Fusobacterium sp. HMT 203 compared to those who are Black/African American (P = 0.004) (Appendix Table 4).

When comparing dentinal lesions versus enamel lesions versus caries-free teeth (lesion free, no white spots) (Fig. 4), the relative abundance of S. mutans and V. dispar was greatest in dentinal caries > enamel caries > caries-free tooth after adjustment for variation among individuals, dentition, household smoking, and race. Dentition type had a significant impact on the relative abundance of S. mutans (P < 0.002) but not V. dispar.

Individual and Family Variation Differs by Species

In order to estimate the effect of family on the relationship between caries status and the abundance of selected species, we used likelihood ratio tests (LRTs) to compare linear mixed models with a random intercept for the individual to models with a random intercept for individuals within families. When predicting abundance of Fusobacterium sp. HMT 203, the random intercept for individuals within families improved model fit (LRT: P = 0.002). This was not the case for models predicting S. mutans, C. hominis, V. dispar, V. parvula, and A. oris.

Discussion

In this cross-sectional study of 750 supragingival plaque samples selected from 584 participants in the COHRA1 from November 2003 to December 2009, we uniquely compare the diversity of plaque microbiome from caries-active (enamel or dentinal caries) and caries-free teeth (no lesions, including no white spots) within the same individual, within families, and by dentition type. Regardless of the comparison, plaque from caries-active teeth was less diverse. The similarity in composition of caries-free and caries-active teeth decreased between primary and mixed and mixed and permanent dentition. In primary-only dentition, plaque from caries-active teeth was more similar than that from caries-free teeth, but the opposite was true in mixed and permanent-only dentition. Abundance of known cariogenic species S. mutans was greater in plaque from dentinal lesions and enamel lesions than in plaque from caries-free teeth in all dentition types, consistent with the literature (Forssten et al. 2010; Okada et al. 2012; Babaeekhou et al. 2020). Fusobacterium sp. HMT 203 had greater relative abundance in plaque from caries-free than caries-active teeth in all dentition types. Results from linear mixed models showed that there was an effect of family on the variation in Fusobacterium sp. HMT 203 abundance by caries status but not S. mutans abundance.

The decrease in similarity of plaque composition from caries-active teeth of mixed and permanent dentition compared to primary dentition suggests that there may be more pathways to caries in mixed and permanent than primary dentition. Consistent with this hypothesis, the abundance of Actinomyces sp. HMT 448, Parascardovia denticolens, and Rothia dentocariosa was significantly greater in plaque from mixed and permanent dentition caries-active teeth than primary dentition caries-active teeth. These species have been previously associated with dental caries (Munson et al. 2004; Oshima et al. 2015; Esberg et al. 2020). Furthermore, the abundance of Streptococcus salivarius and Prevotella denticola was significantly greater among caries-active teeth in the mixed-dentition state, but this was not the case for primary or permanent dentition. This difference might arise due to the eruption of primary teeth, during which supragingival and subgingival niches are introduced for bacterial colonization (Mason et al. 2018). Dentinal lesions were associated with higher abundances of cariogenic species when compared to enamel lesions, which could be due to the stage of cariogenesis (Richards et al. 2017). There are also known differences in the proteome of dental cementum (Giovani et al. 2021) and gingiva of primary teeth (Moriya et al. 2017), which may influence plaque composition, although further explorations of this association are required.

We found greater abundance of Fusobacterium sp. HMT 203 in plaque from caries-free than caries-active teeth in all dentition types. However, in an Indian study of 30 children, Fusobacterium sp. HMT 203 was a late colonizer and associated with severe caries and recurrent caries (Kalpana et al. 2020). These disparate findings may be related to strain variation. ASVs identified as Fusobacterium sp. HMT 203 in our study were closely related to Fusobacterium nucleatum subsp. vincentii. F. nucleatum plays integral and beneficial roles in biofilms that contribute to both periodontal health and disease (Brennan and Garrett 2019). We also found greater abundance ofC. hominis in caries-free teeth compared to caries-active teeth in primary and mixed dentition. C. hominis is a fastidious, Gram-negative bacterium primarily associated with endocarditis (Malani et al. 2006). However, a 2021 study (Yao and He 2021) reported a negative association between C. hominis and severe early child caries, consistent with our findings.

While the α diversity varied little by cigarette smoking exposure (Appendix Fig. 8), exposure to cigarette smoke modified β diversity among primary dentition: caries-free teeth were less similar than caries-active teeth in households with smokers (Appendix Fig. 9). Household smoking was associated with a statistically significant increase in the relative abundance of V. dispar but no other species. In adults, V. dispar has been previously reported as significantly higher in abundance in smokers when compared to nonsmokers (Al Bataineh et al. 2020; Jia et al. 2021). Our results suggest that household smoking may also affect the plaque microbiome composition in primary dentition, but more comprehensive studies are needed.

Data on the role of family members on the assembly of microbes in dental plaque are very sparse, and more discriminating methods than used here (e.g., metagenomics or whole-genome sequencing of specific species) are required to fully answer this question. However, results from our linear mixed models suggest that family members share Fusobacterium sp. HMT 203 in dental plaque from caries-active and caries-free teeth but not S. mutans. This is consistent with the observation that S. mutans amplitypes are transmitted among kindergarteners (Doméjean et al. 2010). Results of a prospective study of the salivary microbiome of 101 mother infant pairs further suggest that colonization with S. mutans results primarily from age-related changes in exposures such as introduction of foods and interactions with other children, as well as a potential for colonization (i.e., emergence of teeth) (Ramadugu et al. 2021).

Our study is one of the largest published to date of the supragingival plaque microbiome among sibling pairs. Furthermore, our sample design enabled us to compare the composition and diversity of the plaque microbiome from enamel and dentinal carious lesions to that from caries-free teeth from the same individual, to caries-free individuals, and to assess the effect of dentition and family. A strength in our analysis is our use of linear mixed-effect models that enabled control for potential confounders while assessing individual versus family effects on plaque microbiome composition. However, because the sampling was at a single point in time, all associations are correlational. Other limitations include lack of information on diet, fluoride exposure, and access to ongoing dental care. There was little variation in reported oral hygiene habits; we found no significant associations with oral hygiene habits reported. We were unable to evaluate the impact of family income due to high levels of missingness.

The plaque samples from teeth were pooled by type (caries-free, enamel lesions or dentinal lesions); the specific teeth from which samples were pooled were not recorded. Therefore, the microbiomes reflect composition of multiple sites. Furthermore, during the10 y of storage, there may have been loss of some species (Luo et al. 2016; Zhou et al. 2019), but there is no evidence that losses would have occurred differentially between carious lesions and caries-free teeth. It was not feasible to assess the microbiome at the time of sample collection, so it is impossible to quantify the impact of this storage. However, the observed microbiome compositions are consistent with similar plaque studies (Forssten et al. 2010; Okada et al. 2012; Kalpana et al. 2020). Dental lesions were classified using the WHO criteria (World Health Organization 1979), which is not as granular as ICDAS (Ismail et al. 2007) (see Appendix Fig. 1 for comparison of criteria used in this study and ICDAS). Only enamel or dentinal caries-active lesions were included, and samples from caries-free teeth excluded those with white spots. Therefore, our results do not provide information on the plaque microbiome during the earliest stages of caries development. We sequenced the V4 region of the 16S rRNA gene using a dual-index system; an advantage of this system is reduction in sequencing errors, but the resolution of streptococci to the species level is somewhat lower than using other regions. This might explain why we found no ASVs assigned to Streptococcus sobrinus. More in-depth sequencing would be required to assess whether family members shared the same strain S. mutans or other bacteria of interest. Nonetheless, this study provides new insights into the relative importance of family and individual effects on plaque microbiome composition.

Author Contributions

D. Bhaumik, contributed to data analysis, drafted and critically revised the manuscript; E. Salzman, E. Davis, F. Blostein, G. Li, contributed to data analysis, critically revised the manuscript; K. Neiswanger, R.J. Weyant, R. Crout, D.W. McNeil, M.L. Marazita, contributed to data conception and design, critically revised the manuscript; B. Foxman, contributed to conception and design, data analysis, drafted and critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.