Alterations in oral bacterial communities are associated with risk factors for oral and oropharyngeal cancer

Abstract

Oral squamous cell carcinomas are a major cause of morbidity and mortality, and tobacco usage, alcohol consumption, and poor oral hygiene are established risk factors. To date, no large-scale case-control studies have considered the effects of these risk factors on the composition of the oral microbiome, nor microbial community associations with oral cancer. We compared the composition, diversity, and function of the oral microbiomes of 121 oral cancer patients to 242 age- and gender-matched controls using a metagenomic multivariate analysis pipeline. Significant shifts in composition and function of the oral microbiome were observed with poor oral hygiene, tobacco smoking, and oral cancer. Specifically, we observed dramatically altered community composition and function after tooth loss, with smaller alterations in current tobacco smokers, increased production of antioxidants in individuals with periodontitis, and significantly decreased glutamate metabolism metal transport in oral cancer patients. Although the alterations in the oral microbiome of oral cancer patients were significant, they were of substantially lower effect size relative to microbiome shifts after tooth loss. Alterations following tooth loss, itself a major risk factor for oral cancer, are likely a result of severe ecological disruption due to habitat loss but may also contribute to the development of the disease.

Figure 1

Oral microbial community taxa and functional pathways differentially abundant in cancer. (a) Taxa (genera and OTUs) and (b) pathways differentially abundant in oral cancer microbiomes as determined by a multivariate model incorporating case/control status, tumor HPV status, tooth loss, periodontal disease, and other demographic and clinical covariates (see Methods). Differences are significant at FDR q < 0.25, and n = 121 cases, 242 controls (see Table 1, Supplementary Datasets 4 and 5).

Figure 2

Microbial clades and pathways associated with oral cancer and with clinical, environmental, and demographic covariates. (a) Clades (tree includes all microbes in the dataset) and (b) pathways (shown on the KEGG BRITE hierarchy) differentially abundant with respect to oral cancer (highlighted points) and a subset of non-cancer covariates in this cohort (outer bands), comprising periodontal health, tumor HPV status, tobacco usage, and tooth status; multivariate model is as in Fig. 1 and Methods. Highlighted associations are significant at FDR q < 0.25, n as in Table 1.

Figure 3

Covariation of microbial community beta-diversity with non-cancer covariates including tooth loss, periodontal health, tumor HPV status, and tobacco usage. Ordination by non-parametric multidimensional scaling of samples’ Canberra dissimilarities, with oral/oropharyngeal cancer status indicated by shape and color stratified by (a) tooth loss status, (b) HPV positivity, (c) periodontal health, and (d) tobacco usage. Complete tooth loss represents the largest determinant of variability in the cohort’s oral microbial communities, with smaller effects of cancer case/control status and other covariates.

Figure 4

Oral cancer and tooth loss significantly affect microbial community alpha- and beta-diversity. (a) Within-sample inverse Simpson alpha-diversity across all samples, and (b) between-sample Bray-Curtis dissimilarity between all pairs of samples within each phenotype. Stars indicate significant differences by Wilcoxon rank sum test (binary oral cancer status) or for Cuzick’s trend test (ordinal tooth loss status), both at p < 0.05.

Figure 5

Gene-level analysis of the serine biosynthesis, cysteine biosynthesis, and threonine biosynthesis pathways. The results of our study showed significant changes in the serine biosynthesis pathway (M00020) in cancer patients as compared with healthy individuals, cysteine biosynthesis pathway (M00338) in periodontitis patients, and the threonine biosynthesis pathway (M00018) in patients with complete tooth loss. Here, these pathways are shown on a gene level by visualizing the absolute difference in relative abundances of the genes (red color scheme), revealing that most genes from these pathways have been altered in their abundance between both conditions on a gene level. Additionally, we demonstrate that some of the genes from the threonine biosynthesis are either specific for Gram-negative (green) or Gram-negative/-positive bacteria (blue), while the serine and cysteine biosynthesis pathways consists of universal genes present in eukaryotes and prokaryotes.