A clinically relevant gene signature in triple negative and basal-like breast cancer

Abstract

Introduction: Current prognostic gene expression profiles for breast cancer mainly reflect proliferation status and are most useful in ER-positive cancers. Triple negative breast cancers (TNBC) are clinically heterogeneous and prognostic markers and biology-based therapies are needed to better treat this disease.

Methods: We assembled Affymetrix gene expression data for 579 TNBC and performed unsupervised analysis to define metagenes that distinguish molecular subsets within TNBC. We used n = 394 cases for discovery and n = 185 cases for validation. Sixteen metagenes emerged that identified basal-like, apocrine and claudin-low molecular subtypes, or reflected various non-neoplastic cell populations, including immune cells, blood, adipocytes, stroma, angiogenesis and inflammation within the cancer. The expressions of these metagenes were correlated with survival and multivariate analysis was performed, including routine clinical and pathological variables.

Results: Seventy-three percent of TNBC displayed basal-like molecular subtype that correlated with high histological grade and younger age. Survival of basal-like TNBC was not different from non basal-like TNBC. High expression of immune cell metagenes was associated with good and high expression of inflammation and angiogenesis-related metagenes were associated with poor prognosis. A ratio of high B-cell and low IL-8 metagenes identified 32% of TNBC with good prognosis (hazard ratio (HR) 0.37, 95% CI 0.22 to 0.61; P < 0.001) and was the only significant predictor in multivariate analysis including routine clinicopathological variables.

Conclusions: We describe a ratio of high B-cell presence and low IL-8 activity as a powerful new prognostic marker for TNBC. Inhibition of the IL-8 pathway also represents an attractive novel therapeutic target for this disease.

Figure 1

Principal biological phenotypes identified as metagenes among TNBC. Heatmaps of expression values of the 16 metagenes (upper panels) and the 355 individual Affymetrix probe sets (lower panels) are shown for the finding cohort (left panels, n = 394) and validation cohort (right panels, n = 185). The dendrogram at the left presents the results from hierarchical clustering of the metagenes. Three major clusters were observed representing (i) basal-like, apocrine, CLDN-CD24, proliferation, and adipocyte metagenes (ii) all five immune cell metagenes, and (iii) the IL-8 and VEGF metagenes, when the hemoglobin and stroma metagenes were left out which display some dataset-bias (see methods). In keeping with these three major phenotypes the samples were sorted according to (1.) Basal-like phenotype, (2.) low vs. high B-Cell metagene, and (3.) the expression value of the IL-8 metagene. (The 355 individual Affymetrix probesets and the respective metagenes are listed in the Additional file 4, Supplementary Methods).

Figure 2

Immunohistochemical analyses of the cellular source of expression of the B-Cell and IL-8 metagenes in TNBC. A) Detection of B-lymphocytes by a CD20 antibody (red staining) in a triple negative breast cancer from the Frankfurt cohort with high expression of B-Cell and IL-8 metagenes. B) An adjacent section of the same tumor as in (A) is stained with an IL-8 antibody demonstrating that carcinoma cells are the source of IL-8 expression (red staining). Note the strong IL-8 staining in rod-like structures in the carcinoma cells. Further analyses using antibodies specific for macrophages (CD68) also demonstrated that macrophages are not the cellular source of IL-8 expression in the tumor (Additional file 1, Supplementary Figure S15).

Figure 3

Distribution of the expression of the basal-like metagene among TNBC of cohort-A. The bimodal distribution of the expression of the basal-like metagene among the 394 TNBC samples in the finding cohort-A is shown. A mixture (black line) of two normal gaussian distributions (blue and red lines) was fitted to these data. The interception of the two gaussians was derived as a cutoff (0.0014) for the definition of basal-like tumors. Similar results were obtained for the validation cohorts-B, and -C, as well as from all samples combined.

Figure 4

Prognostic value of the combined B-Cell/IL-8 metagenes among TNBC. Kaplan Meier analysis of event free survival of 297 TNBC patients with follow up from the finding cohort A. Samples were stratified according to prognostic predictor of the combined B-Cell/IL-8 metagenes. "Good" refers to 95 samples with both high B-Cell and low IL-8 metagene expression whereas all other samples (n = 202) are referred as "Poor". A) Prognostic value of the B-Cell/IL8-metagene prognostic predictor in the 30 TNBC patients with follow up from the validation cohort-B. Samples were stratified as in (A). B) Prognostic value of the B-Cell/IL8-metagene prognostic predictor in the 75 TNBC patients with follow-up from the independent validation cohort-C. Samples were stratified as in (A). C) Prognostic value of the combined B-Cell/IL-8 metagenes among the subset of high grade (G3) TNBC tumors from all three cohorts -A, -B, and -C (n = 186). Samples were stratified as in (A). (Results from the individual cohorts are given in Additional file 1, Supplemental Figure S8). D) Prognostic value of the combined B-Cell/IL-8 metagenes among the subset of low to medium grade (G1 and G2) TNBC tumors from all three cohorts -A, -B, and -C (n = 77). Samples were stratified as in (A). (Results from the individual cohorts are given in Additional file 1, Supplemental Figure S8).