Abstract
Background: The microscopic distinction between benign and malignant thyroid lesions in clinical practice is still largely based on conventional histology. This study was performed to evaluate the diagnostic value of retinoic acid (RAR) and retinoid X receptor (RXR) immunostaining in a large panel of thyroid neoplasms. Retinoid receptor expression has never been described as a diagnostic instrument to differentiate between thyroid tissues.
Aim: The aim of our study was to investigate whether retinoid receptor expression is a sensitive indicator for the diagnosis of thyroid neoplasms.
Methods: We used tissuearrays containing 156 thyroid tissues: 79 benign (normal thyroid, multinodular goiter, and follicular adenoma (FA)) and 77 thyroid carcinomas (papillary thyroid carcinoma (PTC), follicular thyroid carcinoma, and follicular variant of PTC (FvPTC)). Immunostaining was done for RARα, RARβ, RARγ, RXRα and RXRβ. Staining was scored semiquantitatively based on ROC analyses and with hierarchical cluster analysis.
Results: In general, we found increased expression of cytoplasm RARα, cytoplasm RARγ, cytoplasm RXRβ and decreased expression of nuclear RARβ, nuclear RARγ and nuclear RXRα in carcinomas compared to benign tissue. We also found three proteins differently expressed between FA and FTC and four proteins differentially expressed between FA and FvPTC. Using unsupervised cluster analysis, we found that the combination of negative staining of membranous RXRβ and positive staining for cytoplasm RXRβ had a very high discriminating power to cluster malignant thyroid tissues and that the combination of a positive staining for nuclear RXRα and a negative staining for cytoplasm RXR β had a very high discriminating power to cluster benign tissues.
Conclusion: We conclude that identifying optimal antibody panels with cluster analysis increases the diagnostic value in the differential diagnosis of thyroid neoplasms. The combination of negative staining for membranous RXRβ and positive staining for cytoplasm RXRβ had the best accuracy (sensitivity 56%, specificity 99%).