Individuality, Stability, and Variability of the Plaque Microbiome

Abstract

Dental plaque is a bacterial biofilm composed of a characteristic set of organisms. Relatively little information from cultivation-independent, high-throughput analyses has been published on the temporal dynamics of the dental plaque microbiome. We used Minimum Entropy Decomposition, an information theory-based approach similar to oligotyping that provides single-nucleotide resolution, to analyze a previously published time series data set and investigate the dynamics of the plaque microbiome at various analytic and taxonomic levels. At both the genus and 97% Operational Taxonomic Unit (OTU) levels of resolution, the range of variation within each individual overlapped that of other individuals in the data set. When analyzed at the oligotype level, however, the overlap largely disappeared, showing that single-nucleotide resolution enables differentiation of individuals from one another without ambiguity. The overwhelming majority of the plaque community in all samples was made up of bacteria from a moderate number of plaque-typical genera, indicating that the overall community framework is shared among individuals. Each of these genera fluctuated in abundance around a stable mean that varied between individuals, with some genera having higher inter-individual variability than others. Thus, at the genus level, differences between individuals lay not in the identity of the major genera but in consistently differing proportions of these genera from mouth to mouth. However, at the oligotype level, we detected oligotype "fingerprints," a highly individual-specific set of persistently abundant oligotypes fluctuating around a stable mean over time. For example, within the genus Corynebacterium, more than a dozen oligotypes were detectable in each individual, of which a different subset reached high abundance in any given person. This pattern suggests that each mouth contains a subtly different community of organisms. We also compared the Chinese plaque community characterized here to previously characterized Western plaque communities, as represented by analyses of data emerging from the Human Microbiome Project, and found no major differences between Chinese and Western supragingival plaque. In conclusion, we found the plaque microbiome to be highly individualized at the oligotype level and characterized by stability of community membership, with variability in the relative abundance of community members between individuals and over time.

Keywords: 16S rRNA; community dynamics; community ecology; human microbiome; oral microbiota.

Figure 1

Abundant genera in Chinese and Western plaque. Mean relative abundances of the most abundant 15 taxa in HMP and Jiang et al. data. (A) ^*Eren et al. (2014) re-analysis of Human Microbiome Project V1-V3 supragingival plaque data; (B) ^*Eren et al. (2014) re-analysis of HMP V3-V5 data; (C) ^†data from supplemental table S2 of Jiang et al. (2015); (D) our re-analysis of Jiang et al. data. ^‡Sequences in table S2 of Jiang et al. (2015) that were labeled to the family level only.

Figure 2

Compositional similarity between Chinese and Western plaque. Multidimensional scaling plot showing the relatedness among the re-analyzed ^*(Eren et al., 2014) HMP V1-V3 (dark blue), and HMP V3-V5 data sets (light blue), and our re-analysis of the Jiang et al. (2015) data (red). Each dot represents an individual sample. Sample distances were calculated by the Bray-Curtis dissimilarity index using the relative abundances of the 40 genera shared between all three data sets. Ellipses bound the covariance of each data set and are centered on the mean of that set.

Figure 3

Apparent similarity of samples depends on taxonomic resolution. Multidimensional scaling (MDS) plots of relatedness of samples from different individuals. Bray-Curtis dissimilarity index was calculated based on relative abundances in each individual of all (A) phyla, (B) genera, (C) OTUs at 97% identity and (D) oligotypes. For (C), 97% OTUs were created by clustering the oligotype representative sequences using the centroid clustering method according to the Bray-Curtis distances. From this clustered matrix we binned together all oligotypes that were at least 97% similar to any oligotype in the same bin but were not already included in another bin. The ellipses mark the covariance of each individual data set; they are centered on the mean of each individual and colored by individual.

Figure 4

Stable differences between individuals at the genus level. (A) Relative abundances for each individual at each time point, for all 17 genera with greater than 1% mean relative abundance over all 64 samples. Together these genera compose 93.8% of the data set. (B) Anomaly from the mean relative abundance for each sample from each individual. The mean relative abundance for an individual is marked by the dark line, and one and two standard deviations by the dark and light gray fields, respectively. Columns represent individuals, and rows represent genera, with colors as in (A).

Figure 5

Stable differences between individuals are clear at the oligotype level. (A) Relative abundances of the 24 Corynebacterium oligotypes. The smaller, black stackbar shows the Corynebacterium abundance relative to all taxa in the sample, while the larger, colored stackbar shows the abundance of each Corynebacterium oligotype relative to the total abundance of Corynebacterium in each sample. (B) Anomaly from the mean relative abundance for each sample from each individual. The mean relative abundance for an individual is marked by the dark line, and one and two standard deviations by the dark and light gray fields, respectively. Columns represent individuals and rows represent oligotypes, colored as in (A). (C) Exactly the same data and organization as (A), but colored by species instead of by oligotype. C. matruchotii oligotypes are colored blue; C. durum, green.