Tear proteome and protein network analyses reveal a novel pentamarker panel for tear film characterization in dry eye and meibomian gland dysfunction

Dry eye and meibomian gland dysfunction are common ocular surface disorders. Discrimination of both conditions often may be difficult given the overlapping of signs and symptoms, and the lack of correlation with clinical parameters. A total of 144 individuals were included in this study. To search for proteome differences, tear proteins were collected by Merocel sponge and analyzed using 2D-PAGE. Comparative tear protein profile analysis indicated changes in the expression levels of fifteen proteins. Subsequent to MALDI-TOF/TOF protein identification, network analysis revealed expression/interaction connections with other proteins, thereby identifying additional putative markers. A screening validation assay demonstrated the discriminative power of six candidate biomarkers. A further validation study using multiplexed-like ELISA assays in tear samples collected with both sponge and capillary confirmed the high discriminatory power of five biomarkers: S100A6, annexin A1 (ANXA1), annexin A11 (ANXA11), cystatin-S (CST4), and phospholipase A2-activating protein (PLAA) with an area under ROC curve (AUC) ≥ 97.9% (sensitivity ≥ 94.3%; specificity ≥ 97.6%) when comparing dry eye and control individuals. This panel also discriminated between dry eye, meibomian gland dysfunction and control individuals, with a global correct assignment (CA) of 73.2% between all groups. Correct assignment was not found to be significantly dependent on the tear collection method.

Figure optionsDownload high-quality image (268 K)Download as PowerPoint slideHighlights
► We perform differential protein expression analysis of dry eye and meibomian gland dysfunction diseases.
► We examine and validate changes in tear protein profiles by 2DE-PAGE.
► Our model achieves an AUC of 97.9% for dry eye diagnosis with a pentamarker panel.
► The overall correct assignment for all the three studied groups is 73.2%.
► Tear collection method has no effect on the accuracy of our statistical model.