Quantitative proteomics reveals myosin and actin as promising saliva biomarkers for distinguishing pre-malignant and malignant oral lesions

Abstract

Background: Oral cancer survival rates increase significantly when it is detected and treated early. Unfortunately, clinicians now lack tests which easily and reliably distinguish pre-malignant oral lesions from those already transitioned to malignancy. A test for proteins, ones found in non-invasively-collected whole saliva and whose abundances distinguish these lesion types, would meet this critical need.

Methodology/principal findings: To discover such proteins, in a first-of-its-kind study we used advanced mass spectrometry-based quantitative proteomics analysis of the pooled soluble fraction of whole saliva from four subjects with pre-malignant lesions and four with malignant lesions. We prioritized candidate biomarkers via bioinformatics and validated selected proteins by western blotting. Bioinformatic analysis of differentially abundant proteins and initial western blotting revealed increased abundance of myosin and actin in patients with malignant lesions. We validated those results by additional western blotting of individual whole saliva samples from twelve other subjects with pre-malignant oral lesions and twelve with malignant oral lesions. Sensitivity/specificity values for distinguishing between different lesion types were 100%/75% (p = 0.002) for actin, and 67%/83% (p<0.00001) for myosin in soluble saliva. Exfoliated epithelial cells from subjects' saliva also showed increased myosin and actin abundance in those with malignant lesions, linking our observations in soluble saliva to abundance differences between pre-malignant and malignant cells.

Conclusions/significance: Salivary actin and myosin abundances distinguish oral lesion types with sensitivity and specificity rivaling other non-invasive oral cancer tests. Our findings provide a promising starting point for the development of non-invasive and inexpensive salivary tests to reliably detect oral cancer early.

Figure 1. Biomarker discovery strategy.

The parts making up our integrated strategy are shown and described in the text.

Figure 2. Bioinformatic analysis of differentially abundant proteins.

A. Significantly represented pathways of differentially abundant proteins compared to control proteins (those showing no abundance differences) grouped by IPA. Pathways with bars rising above the significance threshold are those containing sufficiently high numbers of proteins from our proteomic data such that they are deemed highly represented in a statistically significant manner (See Materials and Methods for details on significance calculations). Numerical values in parentheses for each pathway are the total number of proteins in the Ingenuity Pathway database mapping to that pathway. B. A significantly represented protein network (P = 1×10⁻³⁵) determined by IPA on differentially abundant soluble saliva proteins. Myosin subunits and actins are highly represented in this network and shaded in black. Other differentially abundant proteins identified in our proteomics study are shaded in gray; proteins not identified in our proteomics study are not shaded.

Figure 3. Summary of western blot validation results.

Box plots of actin and myosin western blot data in soluble saliva (panels A and B, mean of twelve samples tested in each subject group; the ratio for each sample being the mean of three technical replicates) and exfoliated cells in the same saliva samples (panels C and D, mean of twelve samples tested in each subject group). Each plot displays the median (solid horizontal line within each box), quartiles (area if each box), and range of the data (bars extending from each box). Western blots were performed for actin (A, C) and myosin (B, D).