Clinical validation of a customized multiple signature microarray for breast cancer.

PURPOSE: Current histopathologic systems for classifying breast tumors require evaluation of multiple variables and are often associated with significant interobserver variability. Recent studies suggest that gene expression profiles may represent a promising alternative for clinical cancer classification. Here, we investigated the use of a customized microarray as a potential tool for clinical practice. EXPERIMENTAL DESIGN: We fabricated custom 188-gene microarrays containing expression signatures for three breast cancer molecular subtypes [luminal/estrogen receptor (ER) positive, human epidermal growth factor receptor 2 (HER2), and "basaloid"], the Nottingham prognostic index (NPI-ES), and low histologic grade (TuM1). The reliability of these multiple-signature arrays (MSA) was tested in a prospective cohort of 165 patients with primary breast cancer. RESULTS: The MSA-ER signature exhibited a high concordance of 90% with ER immunohistochemistry reported on diagnosis (P < 0.001). This remained unchanged at 89% (P < 0.001) when the immunohistochemistry was repeated using current laboratory standards. Expression of the HER2 signature showed a good correlation of 76% with HER2 fluorescence in situ hybridization (FISH; ratio > or =2.2; P < 0.001), which further improved to 89% when the ratio cutoff was raised to > or =5. A proportion of low-level FISH-amplified samples (ratio, 2.2-5) behaved comparably to FISH-negative samples by HER2 signature expression, HER2 quantitative reverse transcription-PCR, and HER2 immunohistochemistry. Luminal/ER+ tumors with high NPI-ES expression were associated with high NPI scores (P = 0.001), and luminal/ER+ TuM1-expressing tumors were significantly correlated with low histologic grade (P = 0.002) and improved survival outcome in an interim analysis (hazard ratio, 0.2; P = 0.019). CONCLUSION: The consistency of the MSA platform in an independent patient population suggests that custom microarrays could potentially function as an adjunct to standard immunohistochemistry and FISH in clinical practice.

Targets of genome copy number reduction in primary breast cancers identified by integrative genomics.

The identification of specific oncogenes and tumor suppressor genes in regions of recurrent aneuploidy is a major challenge of molecular cancer research. Using both oligonucleotide single-nucleotide polymorphism and mRNA expression arrays, we integrated genomic and transcriptional information to identify and prioritize candidate cancer genes in regions of increased and decreased chromosomal copy number in a cohort of primary breast cancers. Confirming the validity of this approach, several regions of previously-known copy number (CN) alterations in breast cancer could be successfully reidentified. Focusing on regions of decreased CN, we defined a prioritized list of eighteen candidate genes, which included ARPIN, FBN1, and LZTS1, previously shown to be associated with cancers in breast or other tissue types, and novel genes such as P29, MORF4L1, and TBC1D5. One such gene, the RUNX3 transcription factor, was selected for further study. We show that RUNX3 is present at reduced CNs in proportion to the rest of the tumor genome and that RUNX3 CN reductions can also be observed in a breast cancer series from a different center. Using tissue microarrays, we demonstrate in an independent cohort of over 120 breast tissues that RUNX3 protein is expressed in normal breast epithelium but not fat and stromal tissue, and widely down-regulated in the majority of breast cancers (>85%). In vitro, RUNX3 overexpression suppressed the invasive potential of MDA-MB-231 breast cancer cells in a matrigel assay. Our results demonstrate the utility of integrative genomic approaches to identify novel potential cancer-related genes in primary tumors. This article contains Supplementary Material available at http://www.interscience.wiley.com/jpages/1045-2257/suppmat.

c-erbB-2 (HER-2/neu) immunohistochemistry in invasive breast cancer: is there concordance between standard sections and tissue microarrays?

AIMS: Immunohistochemical detection of the 185-kDa transmembrane glycoprotein product of the proto-oncogene c-erbB-2 (also known as HER-2/neu), located on chromosome 17q21, is a well established method of evaluation in invasive breast cancer. This investigation is currently performed on standard sections cut from the tumour containing paraffin block. It is uncertain if concordant results can be obtained on tissue microarray (TMA) sections, a high throughput technique that is particularly advantageous in research and validation protocols. Our aim in this study was to compare the results of c-erbB-2 immunoexpression in standard sections of invasive breast cancers with those of TMAs. METHODS: Standard sections and TMAs constructed from archival paraffin-embedded breast cancers of 184 patients who had surgery in Singapore General Hospital during the period 1998-2002 were subjected to immunohistochemistry using the commercial antibody (A0485, Dako). c-erbB-2 over-expression was evaluated according to cytoplasmic membrane staining intensity, which was defined as 2+ and 3+ staining. RESULTS: Over-expression of c-erbB-2 protein was found in 21.2% (39/184) and 18.6% (34/183) of cases on standard sections and TMAs, respectively. There was substantial agreement between these two types of sections (k = 0.724) when positive and negative staining was considered. CONCLUSION: Immunohistochemistry on TMAs for c-erbB-2 expression in breast cancer is a reliable alternative to that performed on routine standard sections, as it is both cost effective and time efficient, especially in a research setting.

Cytokeratins in papillary lesions of the breast: is there a role in distinguishing intraductal papilloma from papillary ductal carcinoma in situ?

We studied 50 papillary lesions (25 papillomas and 25 papillary ductal carcinomas in situ, DCIS) diagnosed at Singapore General Hospital, for immunohistochemical expression of cytokeratin (CK) 5/6, CK14, and 34betaE12. The immunoscore (proportion of stained cells multiplied by staining intensity) was compared between the two groups. Cytokeratin expression was corroborated by confocal microscopy. Results were applied to a separate series of 43 papillary tumors from Hong Kong (HK). CK5/CK6, CK14, and 34betaE12 showed higher immunoscores in papillomas (mean values, 107.6, 186.6, and 113.1, respectively) than papillary DCIS (mean values, 12, 29.6, and 34.5, respectively; P<0.0001, P<0.001, and P<0.02, respectively). A cutoff immunoscore threshold of 50 appeared discriminatory between papilloma and papillary DCIS, and this value was applied to the HK cases, with CK5/CK6, CK14, and 34betaE12 correctly predicting 25 (89.3%), 26 (92.9%), and 27 (96.4%), respectively, of 28 HK lesions labeled as papillomas; while they corroborated 13 (86.7%), 13 (86.7%), and 5 (33.3%), respectively, of 15 HK cases diagnosed as papillary DCIS. Review of discordant cases showed that lesions were small, derived from core biopsies, or disclosed accompanying invasive carcinoma. When both SGH and HK cases were combined as a group, the sensitivity of an immunoscore of 50 or less in the diagnosis of papillary DCIS was 95%, 85%, and 62.5% for CK5/CK6, CK14, and 34betaE12, respectively, while the specificity was 86.8%, 94.3%, and 86.8%, respectively. CK immunohistochemistry can aid in evaluating papillary breast lesions. 34betaE12 does not appear as useful in identifying papillary DCIS.