Members and sources of the toothbrush microbiota
Shotgun metagenomic sequencing of the DNA on toothbrushes yielded (5.07 ± 0.55) × 106 quality reads per sample, providing estimated sequencing coverage of 83.1 ± 3.7% (n = 34; mean ± st. error) (Additional file 1: Fig. S1). There were 258 microbial species and 113 genera identified using an assembly-free marker-gene analysis of the microbial communities (i.e., MetaPhlAn2; Additional file 2: Table S1). Prominent phyla included Actinobacteria, Firmicutes, and Proteobacteria. The most abundant taxa were consistent with those identified by metagenome assembly, which yielded 110 bins assigned to 35 genera (Additional file 3: Table S2). At the genus-level, there was a strong correlation between the average relative abundance based on the marker-gene approach and the frequency among bins generated from metagenome assembly (rho = 0.706, p < 0.001; Additional file 4: Fig. S2).
Toothbrush microbiomes contained a mix of human-associated taxa, primarily from oral body sites (Fig. 1a). The core microbiota (i.e., in > 75% of samples) included 8 predominant species frequently present in oral microbiomes (i.e., members of Streptococcus, Rothia, and Veillonella) and 2 others likely more associated with environmental origins (i.e., Klebsiella oxytoca and Stenotrophomonas maltophila) (Fig. 1b). More broadly, there were 37 relatively conserved bacterial species (i.e., in at least 50% of samples), of which 81.1% are common within the oral microbiome. These additional human-associated microbial species, many of which are found across multiple body sites despite occurring most frequently in oral microbiomes, included members of Actinomyces, Corynebacterium, and Prevotella, among others (Fig. 1b). The less-conserved toothbrush microbiota (i.e., in < 50% of samples) varied more in terms of possible origins, with only about half being frequently associated with the oral microbiome (Fig. 1a).
Despite notable overlap in membership with the human microbiota, toothbrush microbial communities had a widely distributed diversity (Fig. 1c). Variation in community composition and structure among toothbrush samples (Jaccard distance = 0.708 ± 0.029) exceeded that of person-to-person oral microbial communities (Jaccard distance = 0.563 ± 0.003) (p < 0.001; mean ± st. error). There was significant dissimilarity between the toothbrush microbiota and human oral (PERMANOVA R2 = 0.027), vaginal (PERMANOVA R2 = 0.062), skin (PERMANOVA R2 = 0.063), and gut (PERMANOVA R2 = 0.097) microbiota (p < 0.001 for all comparisons). Thus, oral microbiota were the least dissimilar taxonomic diversity with toothbrush microbial communities. Notably, out of the three primary oral sites sampled in the HMP-II (i.e., the buccal mucosa, supragingival plaque, and tongue), buccal mucosa was least dissimilar to toothbrushes (PERMANOVA R2 = 0.067, 0.131, 0.125, respectively; all p < 0.001), and they clustered closest together (Additional file 5: Fig. S3). Moreover, alpha-diversity (Shannon Index) of the toothbrush microbiota differed with that of each of the broadly characterized human body sites (Tukey’s p < 0.001 for all comparisons). While alpha-diversity on toothbrushes was greater than that of skin and vaginal microbiotas, it was lower than that of oral and gut microbiotas (Fig. 1d).
While about half of the toothbrush microbial communities clustered among oral microbiota in an ordination, another set appeared more closely related to skin microbiota and other microenvironments (Fig. 1c). The alpha-diversity of toothbrush-associated microbial communities was similar to that of indoor dust (Tukey’s p = 0.829) and tap water (Tukey’s p = 0.921), though not shower head biofilms (Tukey’s p < 0.001) (Fig. 1d). Nevertheless, the toothbrush microbiota had a structure and composition that was significantly different from dust (PERMANOVA R2 = 0.126, p < 0.001) and water (PERMANOVA R2 = 0.251, p < 0.001), as well as shower head biofilms (PERMANOVA R2 = 0.175, p < 0.001) (Fig. 1c, d).
Both toothbrushes and indoor dust are microbial “sinks” containing a mixture of human-associated and environmental microbiota. SourceTracker analysis [23] suggested that, on average, 47.2% of the toothbrush microbiota was derived from humans (almost all oral origin), which is greater than the ~ 30% previously estimated for indoor dust [3, 4]. SourceTracker corroborated the PCoA ordination clustering revealing a bimodal distribution of putative human-derived taxa on toothbrushes, indicating that they were generally either covered with or lacking (i.e., > 75% or < 25%) such members (Fig. 2a). Accordingly, there was a strong relationship between taxonomic diversity on toothbrushes and the proportion of taxa attributed to the human microbiome, primarily the mouth (Fig. 2b; PERMANOVA R2 = 0.215, p < 0.001). The set of toothbrushes with greater than 50% of microbiota putatively derived from the human microbiome contained high relative abundances of members of Streptococcus, Rothia, Veillonella, Actinomyces, and Neisseria (Fig. 2c). Alternatively, key taxa from the set of toothbrushes containing less niche-specific microbiota included members of Klebsiella, Acinetobacter, Stenetrophomonas, Pseudomonas, and Enterobacter (Fig. 2c). The putatively non-human-derived fraction of the microbiota could be attributed to sink tap water (10.6%), shower head biofilms (5.2%), and possibly local environmental sources (i.e., 37.1% were of unknown origin), which may collectively drive the aforementioned large variation in taxonomic diversity (i.e., Fig. 1c). We note that since we used publicly available “training data” from different subjects and locations than the toothbrush samples, the SourceTracker predictions are approximate.
Overall, while toothbrushes contained a distinct microbiota (PERMANOVA p < 0.001) with most frequently occurring members putatively derived from the oral cavity (i.e., human oral microbiota was the greatest predicted source fraction and had the lowest PERMANOVA R2 compared to all other sample types), their alpha-diversity was most similar to other microbial communities found in the built environment.
Resistome of the toothbrush microbiota
The toothbrush metagenomes and the subset of oral metagenomes (Fig. 3a) contained 176 antibiotic resistance gene (ARG) protein families (Additional file 6: Table S3). The latter included samples from human buccal mucosa (n = 11), keratinized gingiva (n = 1), saliva (n = 1), supragingival plaque (n = 9), and tongue (n = 12) oral sites (Additional file 7: Table S4). ARGs varied in drug class and mechanism (e.g., antibiotic inactivation, efflux, target alteration) for predicted resistance (Table S3). We note that this list includes some widely conserved genes that confer intrinsic resistance and those that confer resistance when certain point mutations are present (e.g., rpsJ, gyrB); ARGs presented in our analysis refer only to a particular gene detected from the Comprehensive Antimicrobial Resistance Database (i.e., amino acid sequence predicted to confer resistance) [24]. Toothbrushes contained 158 ARG families with 21.8 ± 3.0 different families per sample, which was significantly greater than the 53 ARG families and 13.6 ± 1.1 different families per sample in oral metagenomes (mean ± st. error; p = 0.042). Unlike the microbiota taxonomic profiles, the ARG profiles of toothbrushes tended to be more diverse than those of oral-associated counterparts (Fig. 3b). Such disparity and the lack of correlation between toothbrush taxonomic and ARG alpha-diversity (rho = − 0.062; p = 0.720) suggests possible microbial selection based on resistome.
Toothbrush resistomes were enriched with more ARG protein families than oral resistomes (Fig. 4a) despite having fewer that were conserved (i.e., in at least 50% of samples) (Fig. 3c) and exhibiting more person-to-person variation (toothbrush Jaccard distance = 0.794 ± 0.034, oral Jaccard distance = 0.607 ± 0.032, p < 0.001). While oral samples were enriched with an ARG encoding resistance to tetracycline (rpsJ), toothbrushes were enriched with a variety of multidrug resistance genes (oqxB, msbA, CRP, PhoP, marA, vgaC) and those that confer resistance to triclosan (fabI) and fosfomycin (PtsI), among other mechanisms (bacA, gyrB). Both sample types further contained enrichments in different proteins that encoded similar resistances, i.e., to drug classes including fluoroquinolones (patA and a parC in oral; emrR and a parC in toothbrush), macrolides (ermF in oral; mel and ErmX in toothbrush), and beta-lactams (CfxA6 in oral; ACT-35 in toothbrush). For all conserved ARGs, log-transformed reads per kilobase per million (RPKMs) did not significantly differ by sample type (q > 0.05 for all) (Fig. 4a). Thus, compared to oral microbiomes, those on toothbrushes contained enrichments in the presence, but not copy number, of a variety of ARGs, some of which encoded resistances to drug classes not commonly conferred by oral microbiota. Overall, while both oral and toothbrush metagenomes contained a variety of ARGs, the latter were more diverse and uniquely associated with multidrug resistance and resistance to non-clinical antimicrobials (e.g., triclosan).
Enrichments within the toothbrush resistome appeared to be most attributable to versatile environmental-derived taxa. The fractions of toothbrush microbiota that were putatively sourced from the human microbiome vs. alternative origin significantly correlated with ARG profile beta-diversity (Fig. 4b; PERMANOVA R2 = 0.141, p < 0.001). Most of the enriched ARGs, except mel, were more abundant on toothbrushes that contained a microbiota primarily derived from such alternative sources (Fig. 4c). In agreement, metagenome assembly uncovered 28 bins containing 29 ARGs, most of which were assigned to environmental taxa with broad niche ranges, such as members of Enterobacter, Klebsiella, and Pseudomonas (Fig. 5). While several bins that were assigned to oral-associated taxa carried ARGs, notably intrinsic resistance (e.g., Neisseria encoding rpsJ), only those with unpredictable origins were found to encode possible multidrug resistance. Taken together, compared to the more conserved oral-associated taxa on toothbrushes, emergence of non-niche-specific strains within the microbial assemblages likely accounted for the unique toothbrush resistome.
Factors shaping the toothbrush microbiome
We inferred a variety of factors that may influence the toothbrush microbiome structure and composition from the participant metadata (Fig. 6). Beta-diversity across microbiota was significantly, yet weakly associated with gender identity (PERMANOVA R2 = 0.055, p = 0.031), missing teeth (PERMANOVA R2 = 0.078, p = 0.004), and adenoid/tonsil removal (PERMANOVA R2 = 0.072, p = 0.010) (Additional file 8: Fig. S4). Similarly, beta-diversity across ARG profiles was weakly linked to having missing teeth (PERMANOVA R2 = 0.053, p = 0.030) and adenoid/tonsil removal (PERMANOVA R2 = 0.058, p = 0.013) (Additional file 8: Fig. S4). Perhaps related, the putative human-derived content of toothbrush microbiota was inversely proportional to having adenoids/tonsils removed (p = 0.011). No other metadata from the exploratory analysis appeared to be significantly linked to predicted source content of the toothbrush microbiota, perhaps a reflection of the relatively small sample size and wide variation across toothbrush microbiomes.
More broadly, alpha-diversity of microbial taxonomic and ARG profiles appeared to be weakly related to oral hygiene (Fig. 7). While there were trends for frequencies of flossing and mouthwash use being inversely associated with taxonomic alpha-diversity, possibly due to removal of potential microbial colonists, an increased frequency of toothbrushing and flossing positively associated with resistome alpha-diversity (Fig. 7). In addition to dental hygiene, microbiota alpha-diversity significantly correlated with gender identity (p = 0.045). Toothbrushes belonging to people that identified as women contained more diverse microbial communities than those belonging to men (Additional file 9: Fig. S5), even though gender identity was not significantly correlated with taxonomic (p = 0.144) or ARG profile (p = 0.418) alpha-diversity of oral metagenomes. Moreover, resistomes on toothbrushes belonging to women and persons taking non-antimicrobial oral medications, those that were stored on sink-top ledges, and those in bathrooms without windows, or with windows that were more frequently closed, exhibited trends for slightly more diverse ARG profiles than respective counterparts (Additional file 10: Fig. S6).
Regarding specific microbial responses to putative selective pressures, hierarchical all-against-all significance testing identified 12 variables significantly associated with relative abundances of 18 different species (Fig. 8a). For example, bathroom attributes (i.e., window presence, window open frequency, toothbrush storage location) correlated with relative abundances of commonly occurring bacterial species within Actinomyces and with Granulicatella adiacens. Whether or not a person had domestic pets in their home, had any missing teeth, or had adenoids or tonsils removed correlated with members of Acinetobacter, Kingella, Porphyromonas, and Pseudomonas, among other taxa. We further uncovered associations between relative abundances of 19 different ARG protein families and 7 variables, again including bathroom attributes, missing teeth, and domestic pets (Fig. 8b). While most of the ARGs that correlated with features associated with multidrug resistance, other drug classes included fluoroquinolones, fosfomycin, macrolides, tetracyclines, triclosan, and sulfonamides (Fig. 8b). Thus, a complex combination of personal health, dental hygiene, and environmental variables likely shapes taxonomic diversity and antimicrobial resistance in toothbrush microbiomes.