Association of general health and lifestyle factors with the salivary microbiota - Lessons learned from the ADDITION-PRO cohort

Abstract

Introduction: Previous research indicates that the salivary microbiota may be a biomarker of oral as well as systemic disease. However, clarifying the potential bias from general health status and lifestyle-associated factors is a prerequisite of using the salivary microbiota for screening.

Materials & methods: ADDDITION-PRO is a nationwide Danish cohort, nested within the Danish arm of the Anglo-Danish-Dutch Study of Intensive treatment in People with Screen-Detected Diabetes in Primary Care. Saliva samples from n=746 individuals from the ADDITION-PRO cohort were characterized using 16s rRNA sequencing. Alpha- and beta diversity as well as relative abundance of genera was examined in relation to general health and lifestyle-associated variables. Permutational multivariate analysis of variance (PERMANOVA) was performed on individual variables and all variables together. Classification models were created using sparse partial-least squares discriminant analysis (sPLSDA) for variables that showed statistically significant differences based on PERMANOVA analysis (p < 0.05).

Results: Glycemic status, hemoglobin-A_1c (HbA_1c) level, sex, smoking and weekly alcohol intake were found to be significantly associated with salivary microbial composition (individual variables PERMANOVA, p < 0.05). Collectively, these variables were associated with approximately 5.8% of the observed differences in the composition of the salivary microbiota. Smoking status was associated with 3.3% of observed difference, and smoking could be detected with good accuracy based on salivary microbial composition (AUC 0.95, correct classification rate 79.6%).

Conclusions: Glycemic status, HbA_1c level, sex, smoking and weekly alcohol intake were significantly associated with the composition of the salivary microbiota. Despite smoking only being associated with 3.3% of the difference in overall salivary microbial composition, it was possible to create a model for detection of smoking status with a high correct classification rate. However, the lack of information on the oral health status of participants serves as a limitation in the present study. Further studies in other cohorts are needed to validate the external validity of these findings.

Keywords: biomarker; microbiota; saliva; smoking; type 2 diabetes (T2D).

Figure 1

Association of sample variables. Heatmap of sample variables compared by chi-squared tests. The p-values are visualized on a log scale (-log10(pvalue)). A value of 1.30 represent a 0.05 significance cut-off, and values above 1.30 are considered statistically significant.

Figure 2

Predominant genera. Stacked bar charts of the 20 most abundant genera expressed as mean relative abundance. (A–F): glycemic status, smoking, sex, weekly alcohol intake, HbA_1c, and weekly activity level.

Figure 3

Beta diversity: Bray-Curtis dissimilarity calculated from Hellinger transformed total sum scaled data was used as beta-diversity measure and visualized with principal coordinate analysis (PCoA). (A–F): glycemic status, smoking, sex, weekly alcohol intake, HbA_1c, and weekly activity level.

Figure 4.1

Genera significantly associated with glycemic status, smoking and sex. Relative abundance of most abundant genera (panels A, C, E) and differential abundant genera (panels B, D, F). Comparison of groups were performed with non-parametric tests (Kruskal-Wallis). Unadjusted p-values as well as Benjamini-Hochberg corrected q-values are included in the plot. (A, B), glycemic status, (C, D), smoking status, (E, F), sex. All tests are included in Supplementary File 2 as well as a parametric approach (DESeq as implemented in DAtest) in Supplementary file 3 ).

Figure 4.2

Genera significantly associated with alcohol intake, Hb1Ac and physical activity. Relative abundance of most abundant genera (panels A, C, E) and differential abundant genera (panels B, D, F). Comparison of groups were performed with non-parametric tests (Kruskal-Wallis). Unadjusted p-values as well as Benjamini-Hochberg corrected q-values are included in the plot. (A, B) weekly alcohol intake, (C, D) Hb1Ac, (E, F) weekly activity level. All tests are included in Supplementary File 2 as well as a parametric approach (DESeq as implemented in DAtest) in Supplementary File 3 ).

Figure 5

Classification potential of the salivary microbiota. Receiver Operating Characteristics (ROC) curves from sparse partial least squares discriminant analysis where used to assess classification of the different groups. Area under the curve (AUC) were used to quantify the performance of the classifier. (A–F) glycemic status, smoking, sex, weekly alcohol intake, HbA_1c, and weekly activity level.