Validation of tissue microarrays for immunohistochemical profiling of cancer specimens using the example of human fibroblastic tumors

Tissue microarrays allow high-throughput molecular profiling of cancer specimens by immunohistochemistry. Phenotype information of sections from arrayed biopsies on a multitissue block needs to be representative of full sections, as protein expression varies throughout the entire tumor specimen. To validate the use of tissue microarrays for immunophenotyping, we studied a group of 59 fibroblastic tumors with variable protein expression patterns by immunohistochemistry for Ki-67, p53, and the retinoblastoma protein (pRB). Data on full tissue sections were compared to the results of one, two, and three 0.6-ram core biopsies per tumor on a tissue array. Ki-67 and p53 staining was read as two categories (positive or negative). Concordance for this staining between tissue arrays with triplicate cores per tumor and full sections were 96 and 98%, respectively. For pRB staining was read as three categories (high, moderate, or negative), where concordance was 91%. The use of three cores per tumor resulted in lower numbers of lost cases and lower nonconcordance with standard full sections as compared to one or two cores per tumor. Correlations between phenotypes and clinical outcome were not significantly different between full section and array-based analysis. Triplicate 0.6-ram core biopsies sampled on tissue arrays provide a reliable system for high-throughput expression profiling by immunohistochemistry when compared to standard full sections. Triplicate cores offer a higher rate of assessable cases and a lower rate of nonconcordant readings than one or two cores. Concordance of triplicate cores is high (96 to 98%) for two category distinction and decreases with the complexity of the phenotypes being analyzed (91%).