Salivary exon-specific markers improves discriminatory power for oral cancer detection

Objective: Saliva meets the demand for an easily accessible and noninvasive diagnostic medium for the detection of oral squamous cell carcinoma (OSCC). However, gene expression profiling for biomarker development can be challenging in saliva because it contains partially fragmented RNAs that are difficult to detect. For example, use of 3'-end biased microarray platforms is not ideal for the discovery of biomarkers in saliva as they do not provide full-length transcript coverage. To overcome these challenges, we developed a Salivary Exon Expression Profiling (SEEP) strategy, utilizing the Affymetrix All-Exon-Array platform to enhance oral cancer discriminatory power of the salivary transcriptome. This technology allows for the detection of all exons along full-length RNA and is ideally suited for biomarker discovery in saliva. Methods: Affymetrix's All-Exon-Array was used to profile mRNA signatures from 6 OSCC patients and 6 control subjects. Then, quantitative PCR (qPCR) analyses were performed on 30 oral cancer salivas and 30 controls to validate candidate exon biomarkers. Results: Using microarray raw data, probe sets were identified representing 6 exons mapping to the IL1B gene and 3 exons mapping to IL8. Primers designed to the two most highly expressed exons for each gene and subsequent qPCR analysis demonstrated significantly elevated expression in 30 OSCC patients when compared to 30 healthy controls (t-test, IL8, p < 0.02 and IL1B p < 0.04). When 3'-end exons were used alone to determine the predictive power of discriminating OSCC from healthy controls, a predictive accuracy measurement (Area Under the Curve value, AUC) of 0.58 was observed. In contrast, an AUC value of 0.80 was obtained when an additional highly expressed exon from each gene was included in the analysis. Conclusions: Saliva Exon Expression Profiling enhances oral cancer biomarker discovery in saliva through full-length coverage of RNA transcripts.

Supported by PHS grant R01 DE015970