Dr. Sheila Galvin

The primary aim of this study was to investigate if the microbiome of oral leukoplakia (OLK), the most common of the 11 oral potentially malignant disorders, was associated with changes in the microbiome and whether these changes differed with differing degrees of dysplasia. Secondary aims included examining if changes in the microbiome could be used to predict whether an OLK was low (LROLK) or high risk (HROLK), establishing how other factors such as smoking affected the microbiome and which clinical and patient factors were associated the degree of dysplasia of OLK on histological analysis. Following ethical approval swabs were taken from histologically proven OLKs and contralateral normal (CLN) sites in patients who met the inclusion criteria. Healthy controls (HC) were also sampled at various mucosal sites. After DNA extraction, the V1-V3 region of the 16S rRNA gene was sequenced. RStudio was used for statistical analysis. 189 patients with 225 OLKs were included in the analysis of clinical and patient features. The degree of dysplasia was found to be significantly associated with OLK site and clinical appearance with location on the tongue or floor of mouth and non-homogenous clinical appearance both significantly associated with severe dysplasia (both p=0.0005). OLKs on the dorsum or lateral tongue were 38.6 times more likely, and on the floor of mouth/ventral tongue 70.4 times more likely, to have severe dysplasia than OLKs on the buccal mucosa while non-homogeneity was associated with 25 times increased odds of having moderate or severe dysplasia. 202 CLN and 120 HC samples from 241 subjects were then examined to investigate the influence of smoking, alcohol consumption, oral hygiene practices and mucosal site on the microbiome of healthy mucosa. Analysis using the R package MaAsLin2 found that smoking and oral site had the most significant influence on the oral microbiome. Smoking was associated with a shift to a more periodontal pathogen-heavy microbiome dominated by Prevotella, Actinomyces, Porphorymonas, Fusobacterium, Tannerella, Parvimonas, Filifactor, Bacteroidales, Clostridiales and Peptostreptococcus species, and reduced in healthy mucosa-associated species such as those of Neisseria, Streptococcus and Rothia genera. Site specificity was also identified. 216 OLK sites and 202 CLN sites from 178 patients were then compared to establish if there were microbiome changes specific to OLK. After adjusting for multiple variables using the R package MaAslin2, OLK was consistently associated with increased abundance of five taxa: Bergeyella sp. HMT322, Streptococcus australis infantis clade 431, Gemella morbillorum, Leptotrichia sp. HMT215 and Leptotrichia sp. HMT392. OLKs were then divided into LROLK (no or mild dysplasia [n=91) and HROLK (moderate or severe dysplasia [n=125]) to investigate if differing degrees of dysplasia affected microbial colonisation. The microbial communities of LROLK and HROLK were found to exhibit significant changes, LROLKs were consistently associated with significantly increased abundances of Gemella morbillorum and Mogibacterium diversum, while HROLKs were associated with significantly increased abundances of Bergeyella sp. HMT322, Streptococcus australis infantis clade 431 and Leptotrichia sp. HMT215. Finally, predictive modelling using the top 50 microbial features from a preliminary modelling analysis could differentiate HROLK from LROLK with a sensitivity of 85.5% and specificity of 76.5%, while a combined model using both clinical/patient and microbial features could predict LROLK or HROLK with a sensitivity of 87.4% and specificity of 76.5%. Remarkably these models were considerably more accurate than models using clinical/patient factors alone. In conclusion, the microbiome of OLK was shown to differ significantly from that of normal mucosa. OLKs with different degrees of dysplasia were also found to exhibit significant differences in microbial composition.