Metatranscriptomics of the human oral microbiome during health and disease

Abstract

The human microbiome plays important roles in health, but when disrupted, these same indigenous microbes can cause disease. The composition of the microbiome changes during the transition from health to disease; however, these changes are often not conserved among patients. Since microbiome-associated diseases like periodontitis cause similar patient symptoms despite interpatient variability in microbial community composition, we hypothesized that human-associated microbial communities undergo conserved changes in metabolism during disease. Here, we used patient-matched healthy and diseased samples to compare gene expression of 160,000 genes in healthy and diseased periodontal communities. We show that health- and disease-associated communities exhibit defined differences in metabolism that are conserved between patients. In contrast, the metabolic gene expression of individual species was highly variable between patients. These results demonstrate that despite high interpatient variability in microbial composition, disease-associated communities display conserved metabolic profiles that are generally accomplished by a patient-specific cohort of microbes. IMPORTANCE The human microbiome project has shown that shifts in our microbiota are associated with many diseases, including obesity, Crohn's disease, diabetes, and periodontitis. While changes in microbial populations are apparent during these diseases, the species associated with each disease can vary from patient to patient. Taking into account this interpatient variability, we hypothesized that specific microbiota-associated diseases would be marked by conserved microbial community behaviors. Here, we use gene expression analyses of patient-matched healthy and diseased human periodontal plaque to show that microbial communities have highly conserved metabolic gene expression profiles, whereas individual species within the community do not. Furthermore, disease-associated communities exhibit conserved changes in metabolic and virulence gene expression.

FIG 1

Ribosome quantification reveals that disease-associated periodontal microbiota are less diverse and contain fewer low-abundance species than do health-associated populations. (A) Number of distinct 16S rRNA sequences (OTUs) observed in healthy (blue) and diseased (red) samples with increasing numbers of sequences sampled from each population. Error bars indicate standard errors of the means (n = 10). (B) Shannon indices show that health-associated populations are more species rich than diseased populations (*, P = 0.03, paired two-tailed Student t test). (C) Beta diversity was measured using the unweighted Unifrac method to calculate relatedness of paired health-associated (blue) and disease-associated (red) microbial populations by assessment of shared and unique species in each community. Principal coordinates 1 and 2 are plotted. Mean diseased and healthy centroids (mean ± standard deviation) are indicated by ellipses. Distances between samples and corresponding centroids are shown as blue and red lines, respectively. Black lines show distances between paired populations from the same patient. (D) Mean Euclidean distance (mean ± standard deviation) from each sample to corresponding centroids and corresponding paired sample from same patient (**, P = 0.0005, paired two-tailed Student t test).

FIG 2

Differential expression of enzyme gene families in health and disease. Log₂ fold change during disease is plotted against the log₂ mean read counts per million total reads for each EC enzyme-encoding gene family. Gene families upregulated in health are shown in blue, while gene families upregulated in disease are shown in red.

FIG 3

Differential metabolic gene expression in the diseased periodontal microbiome. Metabolic network reconstruction. Black lines indicate enzyme-encoding genes that were expressed and unchanged in health and disease, red lines indicate genes upregulated during disease, and blue lines indicate genes upregulated during health. Colored regions identify different sections of the metabolic pathway map. Those highlighted in yellow represent important pathways that were upregulated in disease. Complete data showing all differentially regulated genes are available in supplementary files 2 and 3 at http://web.biosci.utexas.edu/whiteley_lab/pages/resources.html. THF, tetrahydrofolate metabolism; TCA, tricarboxylic acid.

FIG 4

Metabolic niche dynamics in diseased populations. (A) Production of butyrate is primarily due to F. nucleatum lysine fermentation. (B) Multiple species that vary among patients fill histidine degradation and tetrahydrofolate (THF) metabolic niches. (C) Multiple species that vary among patients carry out pyruvate fermentation. For panels A to C, community fold changes of EC enzyme-encoding gene expression are indicated at each arrow. (D) Different organisms fill virulence niches in diseased periodontal communities. In patient 1, Tannerella forsythia is the major source of collagenase expression, whereas collagenase expression is augmented by Prevotella tannerae in patient 2 and by Porphyromonas gingivalis in patient 3. Protease production follows similar patterns, whereby combinations of different species express proteases in each patient. In panels A to D, heat maps indicate relative normalized expression (log₂ reads per million reads in each sample) of different enzyme-encoding genes or virulence genes by species in each patient. Abbreviations of species names and the color scale for heat maps are indicated. CoA, coenzyme A.

FIG 5

EC expression is less variable than individual gene expression. (A) Variance estimations for genes in the metagenome determined in edgeR analyses. (B) Variance estimations for EC expression determined in edgeR analyses. (C) Variance estimations for genes randomly binned into 1,137 gene groups determined in edgeR analyses. Blue lines in panels A to C indicate the tagwise dispersion, while red lines show the common dispersions calculated with edgeR. BCV, biological coefficient of variance.