Biological Classification of Breast Cancer by Real-Time Quantitative RTPCR: Comparisons to Microarray and Histopathology. Homo sapiens

Microarrays have shown that gene expression patterns can be used to molecularly classify breast cancers into distinct and clinically significant groups. In order to translate these profiles into routine diagnostics, we have recapitulated a microarray breast cancer classification using real-time quantitative (q)RT-PCR. We performed statistical analyses on multiple independent microarray datasets to select an “intrinsic” gene set that can classify breast tumors into four different subtypes designated as Luminal, Normal-like, HER2+/ER-, and Basal-like. A minimal gene set from the microarray “intrinsic” list, and additional genes important for outcome (e.g., proliferation genes), were used to develop a real-time qRT-PCR assay comprised of 53 classifiers and 3 housekeepers. We prospectively compared the expression data and classifications from microarray and real-time qRT-PCR using 123 unique breast samples (117 invasive carcinomas, 1 fibroadenoma and 5 normal tissues) and 3 cells lines. The overall correlation for the 50 genes in common between microarray and qRT-PCR was 0.76. There was 91% (114/126) concordance in the hierarchical clustering classification of the real-time qRT-PCR minimal “intrinsic” gene set (37 genes) and the larger (402 genes) microarray gene set from which the PCR list was derived. As expected, the Luminal tumors (ER+) had a significantly better outcome than the HER2+/ER- (p=0.043) and Basal-like tumors (p=0.001). High expression of the proliferation genes GTBP4 (p=0.011), HSPA14 (p=0.023), and STK6 (p=0.027) were significant predictors of relapse free survival (RFS) independent of grade and stage. Our study shows that genomic microarray data can be translated into a qRT-PCR diagnostic assay that enhances the ability to predict outcome and may therefore improve the standard of care in breast cancer. Keywords: other

基因芯片（microarray）研究表明，基因表达模式可用于将乳腺癌进行分子分型，划分为不同且具有临床意义的亚型。为将这些表达谱应用于常规临床诊断，我们采用实时定量聚合酶链式反应（real-time quantitative PCR, qRT-PCR）复现了基于基因芯片的乳腺癌分型方法。我们对多组独立的基因芯片数据集进行统计分析，筛选出一套“固有”基因集，可将乳腺肿瘤划分为四种亚型：管腔型（Luminal）、正常样型（Normal-like）、HER2阳性/雌激素受体阴性型（HER2+/ER-）以及基底样型（Basal-like）。我们从基因芯片的“固有”基因集中筛选出最小基因子集，并结合与预后密切相关的额外基因（如增殖相关基因），开发出一套包含53个分类基因与3个管家基因的实时qRT-PCR检测体系。我们纳入123例独特的乳腺组织样本（含117例浸润性癌、1例纤维腺瘤与5例正常组织）以及3株细胞系，前瞻性对比了基因芯片与实时qRT-PCR的表达数据及分型结果。基因芯片与qRT-PCR共有的50个基因的整体表达相关性为0.76。针对实时qRT-PCR的最小“固有”基因集（37个基因）与开发该PCR检测所用的更大规模基因芯片基因集（402个基因）进行层次聚类分型，两者的一致性达91%（114/126）。正如预期，管腔型肿瘤（雌激素受体阳性，ER+）的预后显著优于HER2+/ER-型（p=0.043）与基底样型肿瘤（p=0.001）。增殖相关基因GTBP4（p=0.011）、HSPA14（p=0.023）与STK6（p=0.027）的高表达，可独立于肿瘤分级与分期，作为无复发生存期（RFS）的显著预测因子。本研究证实，基因组基因芯片数据可转化为qRT-PCR诊断检测体系，提升乳腺癌预后预测能力，进而改善乳腺癌的临床诊疗标准。关键词：其他