Classification and risk stratification of invasive breast carcinomas using a real-time quantitative RT-PCR assay

Abstract

Introduction: Predicting the clinical course of breast cancer is often difficult because it is a diverse disease comprised of many biological subtypes. Gene expression profiling by microarray analysis has identified breast cancer signatures that are important for prognosis and treatment. In the current article, we use microarray analysis and a real-time quantitative reverse-transcription (qRT)-PCR assay to risk-stratify breast cancers based on biological 'intrinsic' subtypes and proliferation.

Methods: Gene sets were selected from microarray data to assess proliferation and to classify breast cancers into four different molecular subtypes, designated Luminal, Normal-like, HER2+/ER-, and Basal-like. One-hundred and twenty-three breast samples (117 invasive carcinomas, one fibroadenoma and five normal tissues) and three breast cancer cell lines were prospectively analyzed using a microarray (Agilent) and a qRT-PCR assay comprised of 53 genes. Biological subtypes were assigned from the microarray and qRT-PCR data by hierarchical clustering. A proliferation signature was used as a single meta-gene (log2 average of 14 genes) to predict outcome within the context of estrogen receptor status and biological 'intrinsic' subtype.

Results: We found that the qRT-PCR assay could determine the intrinsic subtype (93% concordance with microarray-based assignments) and that the intrinsic subtypes were predictive of outcome. The proliferation meta-gene provided additional prognostic information for patients with the Luminal subtype (P = 0.0012), and for patients with estrogen receptor-positive tumors (P = 3.4 x 10-6). High proliferation in the Luminal subtype conferred a 19-fold relative risk of relapse (confidence interval = 95%) compared with Luminal tumors with low proliferation.

Conclusion: A real-time qRT-PCR assay can recapitulate microarray classifications of breast cancer and can risk-stratify patients using the intrinsic subtype and proliferation. The proliferation meta-gene offers an objective and quantitative measurement for grade and adds significant prognostic information to the biological subtypes.

Figure 1

Two-way hierarchical clustering of real-time quantitative reverse-transcription (qRT)-PCR data. (a) The sample-associated dendrogram groups the 126 breast samples profiled by qRT-PCR into the same classes seen by microarray analysis. Samples are grouped into Luminal (blue), HER2+/ER- (pink), Normal-like (green), and Basal-like (red) subtypes. The expression level for each gene is shown relative to the median expression of that gene across all the samples, with high expression represented by red and low expression represented by green. Genes with median expression are black and missing values are gray. (b) A minimal set of 37 'intrinsic' genes was used to classify tumors into their primary 'intrinsic' subtypes. The 'intrinsic' gene set was supplemented using (c) PgR and EGFR, and (d) proliferation genes. The genes in (c) and (d) were clustered separately in order to determine agreement between the minimal 37 qRT-PCR 'intrinsic' set and the larger 402 microarray 'intrinsic' set (see Additional file 1, Supplemental Figure 1). Overall, 114/123 (93%) primary breast samples were classified the same between microarray and qRT-PCR.

Figure 2

Grade and proliferation as predictors of relapse-free survival. A Cox regression model was used to determine probability of relapse over time. Kaplan-Meier curves show the time to event given different grades and levels of proliferation. The grade was scored as low (green), medium (red) or high (blue). The proliferation score was based on continuous expression data, and is shown as tertiles that correspond to low (green), medium (red), and high (blue) levels of expression. The proliferation meta-gene (log₂ average of the 14 proliferation genes) showed significant value in predicting relapse, even after correcting for other clinical parameters important for survival (Table 1). Furthermore, when we include both the grade and proliferation in a model for relapse-free survival, we find that the proliferation meta-gene is the better predictor (grade, P = 0.51; proliferation index, P = 0.047).

Figure 3

Intrinsic subtype stratified by the proliferation index. Tumors were given an 'intrinsic' subtype assignment based on the minimal 37-gene quantitative reverse-transcription-PCR classifier (Figure 1b). Patients were classified as having Luminal (estrogen receptor (ER)-positive) or HER2/Basal (ER-negative) subtypes. In order to have groups of similar size and because the subtypes largely follow ER status, tumors in the HER2 and Basal-like groups (both ER-negative) were combined. Continuous expression data for the proliferation meta-gene (log₂ average of the 14 selected markers) were used in a Cox regression model to determine the probability of relapse over time. Differences in relapse for low (green), medium (red), and high (blue) expression are shown as tertiles in the Kaplan-Meier plots. Stratification by proliferation added information for relapse in the Luminal subtype (P = 0.00039) but not the ER-negative subtypes (P = 0.74).

Figure 4

Stratification of five 'intrinsic' subtypes by the proliferation meta-gene. A large microarray breast cancer data set (337 samples × 16,000 genes) from women with early-stage disease was used to confirm the significance of the proliferation meta-gene to further risk-stratify the Luminal tumors. Tumors were classified as Basal, HER2+/ER-, Luminal A, Luminal B, and Normal-like. The microarray data for the proliferation meta-gene was then used in a Cox regression model to determine probability of relapse in women with the different tumor subtypes. Differences in relapse for low (green), medium (red), and high (blue) expression are shown as tertiles in the Kaplan-Meier plots. The Kaplan-Meier curves show that proliferation adds significant survival information, beyond that gleaned from the intrinsic subtype, only for patients with Luminal A tumors (P = 0.012).

Figure 5

Co-clustering of real-time quantitative reverse-transcription (qRT)-PCR and microarray data using 50 genes and 252 samples. The relative copy number (qRT-PCR) and R/G ratio (microarray) for each gene was log₂-transformed and combined into a single dataset using distance-weighted discrimination. Two-way hierarchical clustering was performed on the combined dataset using Spearman correlation and average linkage. (a) The sample-associated dendrogram shows the same classes as seen in Figure 1. Samples are classified as Basal-like (red), HER2+/ER- (pink), Luminal (blue), and Normal-like (green). The expression level for each gene is shown relative to the median expression of that gene across all the samples, with overexpressed genes in red and underexpressed genes in green. Genes with average expression are black. (b) The gene-associated dendrogram shows that the Luminal tumors and Basal-like tumors differentially express estrogen-associated genes (cluster 1); as well as basal keratins (KRT 5 and KRT 17), inflammatory response genes (CX3CL1 and SLPI), and genes in the Wnt pathway (FZD7) (cluster 3). The main distinguishers of the HER2+/ER- group are low expression of genes in cluster 1 and high expression of genes on the 17q12 amplicon (ERBB2 and GRB7) (cluster 4). The proliferation genes (cluster 2) have high expression in the estrogen receptor (ER)-negative tumors (Basal-like and HER2+/ER-) and low expression in ER-positive (Luminal) and Normal-like samples.