Tissue microarrays: one size does not fit all

Abstract

Background: Although tissue microarrays (TMAs) are commonly employed in clinical and basic-science research, there are no guidelines for evaluating the appropriateness of a TMA for a given biomarker and tumor type. Furthermore, TMA performance across multiple biomarkers has not been systematically explored.

Methods: A simulated TMA with between 1 and 10 cores was designed to study tumor expression of 6 biomarkers with varied expression patterns (B7-H1, B7-H3, survivin, Ki-67, CAIX, and IMP3) using 100 patients with clear cell renal cell carcinoma (RCC). We evaluated agreement between whole tissue section and TMA immunohistochemical biomarker quantification to assess how many TMA cores are necessary to adequately represent RCC whole tissue section expression. Additionally, we evaluated associations of whole tissue section and TMA expression with RCC-specific death.

Results: The number of simulated TMA cores necessary to adequately represent whole tissue section quantification is biomarker specific. Although 2-3 cores appeared adequate for B7-H3, Ki-67, CAIX, and IMP3, even as many as 10 cores resulted in poor agreement for B7-H1 and survivin compared to RCC whole tissue sections. While whole tissue section B7-H1 was significantly associated with RCC-specific death, no significant associations were detected using as many as 10 TMA cores, suggesting that TMAs can result in false-negative findings if the TMA is not optimally designed.

Conclusions: Prior to TMA analysis, the number of TMA cores necessary to accurately represent biomarker expression on whole tissue sections should be established as there is not a one-size-fits-all TMA. We illustrate the use of a simulated TMA as a cost-effective tool for this purpose.

Figure 1

Simulated TMA template mask. (A) 19 equally spaced 0.6 mm viewing ports and the TMA template placed below the objective of a microscope, and (B) the order of the 10 cores used for quantification of tumor biomarker expression.

Figure 2

Box plots illustrating Wald chi-square statistics from 500 bootstrap samples. Testing associations of (A) B7-H3, (B) survivin, (C) Ki-67, (D) CAIX, (E) IMP3, and (F) B7-H1 expression with RCC-specific death using 1 and up to 10 cores. The lower, middle, and upper horizontal lines that comprise the boxes denote the 25^th percentile, median, and 75^th percentile of the distribution of test statistics from the 500 bootstrap samples. Using a significance level of 0.05, the critical value for a 1 degree-of-freedom chi-square test statistic is approximately 3.8. B7-H1 was modeled as dichotomous (positive versus negative), whereas B7-H3, survivin, Ki-67, CAIX, and IMP3 were modeled as continuous.