Whole saliva expression data from human populations stratified by psychosocial stress exposure

The objective of this study was analysis of the whole saliva transcriptome to search for biomarkers of psychosocial stressor exposure and substance use in young adults drawn from a population-based longitudinal cohort, the Oregon Youth Substance Use Project. We conducted genome-wide gene expression analysis on whole saliva RNA from 48 individuals stratified by psychosocial stressor exposure using an Affymetrix Gene ST 1.0 array. We applied Weighted Gene Co-expression Network Analysis (WCGNA) to characterize the high-level structure in the data and to relate expression patterns among samples to participant clinical characteristics. This approach finds clusters of correlated genes (modules) which can reflect clinical, histological, or intracellular organization and function. Using WCGNA, we constructed a gene expression network from saliva genome-wide gene expression data, which we found to be similar to a network we constructed from publicly available cell-free saliva genome-wide gene expression data. Functional characterization of the WGCNA modules suggested that the samples varied in composition between an expression profile similar to circulating antigen presenting cell types and a glycoprotein-containing expression profile. We identified a significant relation between one module of whole saliva genome-wide gene expression and smoking behavior, and one with psychosocial stressor exposure. The smoking-related module was related to mitochondrial gene expression, and ever-smokers were found to have reduced overall mitochondrial gene expression. To determine whether reduced expression could be explained by reduction in mitochondrial DNA, we measured mtDNA copy number variation within genomic DNA from 400 OYSUP participants. Surprisingly, we found that more frequent smoking is associated with increased mtDNA copy number. The network structure of whole saliva and its association with clinically assessed exposures and behaviors are reported here for the first time. Two groups of 24 individuals each were selected from a sample of 132 based on results from the Life Events and Difficulties Schedule, an assesment of chronic psychosocial stress exposure. The two groups were selected from each end of the stress exposure distribution, composing a high stress and low stress group. One sample (4H) failed and was removed from the study.