This is the first study to use archival lesion swabs for the characterization of bacteria within OED. It is also the first to compare the diversity between progressing and non-progressing OED in a longitudinal design and to report on the functional potential of the bacteriome associated with risk of progression in OED. Our study has revealed that the bacterial make-up of the OED niche is similar to what has previously been described in the normal oral cavity, OPML and OSCC at the phylum and genus levels.29,55, 60–62 We also found that although participant factors such as age or smoking status have an overall larger impact on the oral microbiome than progression status, once these are controlled for there are also small differences in the microbiome that can be attributed to progression status (Fig. 1). In cross-sectional studies, differential relative abundance of certain genera, such as Streptococcus, Prevotella, Campylobacter, Pseudomonas, and Fusobacterium, is often seen between normal and diseased states.26,27,33,59 While there were some slight differences in relative abundances of these genera, none of them were significant between Ps and NPs. Even those taxonomic features that were determined to be significantly associated with either Ps or NPs (e.g., Lautropia, Unclassified Actinobacteria, and Cardiobacterium) had quite low relative abundances, and therefore are not likely to be large contributors to the niche. To determine whether these slight changes in abundances are significant at either genera or species level resolutions, it will likely take substantially larger sample sizes.
More recently, the literature has pointed to the role of the microbial metabolome and how the sum of the community as a whole may play a larger role in influencing the tissue microenvironment than any species alone.63,64 The concept of functional redundancy may explain how compositional variations of the microbiome associated with OPML and OSCC may collectively be contributing to a dysbiotic community. Our functional prediction analysis identified a number of EC numbers that were significantly differentially abundant with the time to progression in Ps only and are found to be expressed by multiple bacterial species that we identified to be a part of the oral microbiome niche. Of the top 10 significant EC numbers that we examined (Supp. Figure 4), three have previously characterized links to various cancers: EC:4.2.1.22 Cystathionine β-synthase, which is involved in the transsulfuration pathway that leads to production of the amino acid cysteine; EC:1.2.1.3 Aldehyde dehydrogenase (NAD(+); ALDH), which is involved in a variety of pathways that lead to the consumption of an aldehyde and the production of acetate and acids; and EC:6.3.5.4 Asparagine synthase (glutamine-hydrolyzing), which produces the amino acids asparagine and glutamate. These three EC numbers are also involved in amino acid metabolism, along with most of the other top 10 EC numbers – EC:2.7.7.83 UDP-N-acetylgalactosamine diphosphorylase, EC:1.3.1.31 2-enoate reductase, and EC.6.2.1.30 Phenylacetate–CoA ligase. Additionally, EC:3.2.1.21 β-glucosidase is involved in cyanoamino acid metabolism, as well as hydrolyzes cellulose to glucose molecules, which are often more highly taken up by tumour cells.65,66 Most of the top 10 EC numbers have increased abundance in NPs vs Ps, with a trend toward increased levels of the identified EC numbers with increasing time to progression. This could indicate that either a depletion of the chemicals produced by these enzymes or a build-up of the chemicals that are consumed by these enzymes may promote tumourigenesis. For example, ALDH deficiency, which leads to a buildup of acetaldehyde and is caused by a single nucleotide polymorphism (SNP) in the human genome has been widely linked to an increased risk of developing multiple cancers.67Acetaldehyde is a group 1 carcinogen,68 and the microbial origin of salivary acetaldehyde is well documented. Alongside human ALDH deficiency, several dominant members of the microbiome, such as Streptococcus and Neisseria, may be lacking an aldehyde dehydrogenase enzyme,69 which could also contribute to oncogenesis and allow for progression of pre-malignant lesions to carcinoma. However, as these enzymes participate in normal bacterial functions and the coefficients were quite low, it is difficult to conclude whether these are indeed a contributor to the microenvironment or just widely expressed among bacterial species and involved in some of the more abundant pathways, which highlights the need for additional studies.
Given that our study design compared samples taken at an early stage of disease (mild /moderate OED), the lack of significant differences between P and NPs may indicate that changes in overall diversity as well as taxonomic shifts occur at later stages in progression, or perhaps detectable only after OSCC has been established. Further to this, a study looking at OSCC, normal, and OPML found that while OSCC samples clustered based on beta-diversity, pre-cancer and normal samples were mixed, indicating that there was not a great difference in the diversity between these groups.60 This may support the notion that microbial changes in diversity change at a later stage. Larger studies that investigate the full spectrum of expertly graded OED are necessary.
A potential limitation of this work is that the samples were stored at ultra-low temperatures for a considerable period of time from date of collection to DNA extraction (mean 15.7 years, range 8.9–24.3 years). It is unknown how long-term storage may have affected such small samples with a potentially small biomass. However, this should also be viewed as strength, as older samples have longer follow-up and as a result, more robust outcome data. Alpha-diversity differed between Ps and NPs; however, beta diversity did not differ significantly among these groups alone (although it did in conjunction with other participant metadata). One of the reasons for this lack of significance in beta diversity may be due to a relatively small sample size. It is possible that the study suffered from a type II error due to lack of statistical power. However, prior to moving forward with a full-scale study, a pilot study of these small and invaluable samples was necessary. A significant strength is that this study examined patients with known outcomes, who had samples taken prior to developing disease. However, a limitation is that no longitudinal sampling points have been evaluated. A comparison between longitudinal samples can provide insightful results on the temporal changes. Future studies that employ repeated sampling are warranted. A caveat to this work is that 16S rRNA sequencing is not always capable of providing a high enough resolution to differentiate between closely related genera. In addition, functional profiles were established via prediction analysis (PICRUSt2).46. A metagenome sequencing approach may yield more comprehensive data for taxonomic assignment to the species level and provide more direct information on metabolic pathways for functional profiling based on pathway component genes.
In conclusion, for the first time, we have characterized the microbiome of low-grade OED with known outcome using 16S rRNA gene sequencing from annotated archival swabs. At the genus level, known oral cavity colonizers did not correlate with progression. The collective metabolic impact of the bacteriome trends toward a depletion of several enzymes that have been previously linked to cancer in progressing oral lesions but requires a larger sample size to show this more clearly. Having shown that quality NGS data can be obtained from archival oral swabs, larger prospective cohort studies to further explore the taxa and the function of the microbiome as a potential biomarker of risk in OED are warranted.