An integrated genomics approach identifies drivers of proliferation in luminal-subtype human breast cancer

Abstract

Elucidating the molecular drivers of human breast cancers requires a strategy that is capable of integrating multiple forms of data and an ability to interpret the functional consequences of a given genetic aberration. Here we present an integrated genomic strategy based on the use of gene expression signatures of oncogenic pathway activity (n = 52) as a framework to analyze DNA copy number alterations in combination with data from a genome-wide RNA-mediated interference screen. We identify specific DNA amplifications and essential genes within these amplicons representing key genetic drivers, including known and new regulators of oncogenesis. The genes identified include eight that are essential for cell proliferation (FGD5, METTL6, CPT1A, DTX3, MRPS23, EIF2S2, EIF6 and SLC2A10) and are uniquely amplified in patients with highly proliferative luminal breast tumors, a clinical subset of patients for which few therapeutic options are effective. This general strategy has the potential to identify therapeutic targets within amplicons through an integrated use of genomic data sets.

Figure 1. Patterns of genomic signature pathway activity in breast cancer

(A) Patterns of pathway activity (n=52) were determined for each sample in the published TCGA Breast Cancer cohort (n=476). Expression signature scores (y-axis) are median centered and clustered by complete linkage hierarchical clustering. (B) An ANOVA (P<0.0001) for all signatures according to PAM50 subtype followed by a Tukey test for pair-wise comparison demonstrates statistically significant differences in the levels of pathway expression between molecular subtypes. Box color indicates level of significance between subtypes as indicated by the legend.

Figure 2. Identification of genomic pathway-specific copy number alterations

(A) Schematic outlining strategy used to identify CNA associated with pathway activity. (B) For each signature, significant copy number gains and losses were calculated. The plot identifies those genes that had a positive Spearman rank correlation and have increased amplification frequency (q<0.01) (red), or that have a negative Spearman rank correlation and show an increased frequency of copy number losses in the top scoring (top quartile) samples with pathway activity (q<0.01) (blue). (C–E) A Spearman rank correlation was used to identify genes positively (black line) or negatively (dark blue) associated with pathway activity and a Fisher’s exact test was used to compare the frequency of copy number gains (red) or losses (light blue) for the (C) Her2 Amp (D) Her1-C2 signature, and (E) RB-LOH signature. Yellow arrows indicate known pathway drivers with q<0.01 for each analysis; black arrow indicates q<0.01 for a single analysis. In each figure, chromosomal boundaries are indicated by vertical black lines.

Figure 3. Identification of DNA copy number alterations in highly proliferative breast tumors

(A) Distribution of proliferation scores across all tumors and (B) by subtype. (B) Box and whisker plots indicate median score and the upper and lower quartile. Basal-like (n=88), HER2E (n=55), LumA (n=214) and LumB (n=119). (C) Highly proliferative tumors (top quartile) are comprised of Basal-like (49.6%), LumB (33.6%) and HER2E (16.8%). (D) Highly proliferative luminal tumors are restricted to LumB (68.0%) and HER2E (32.0%) samples. (E) Frequency of CNA in highly proliferative (black line) and all other samples (gray line). (F) Statistical analyses of CNA: positive correlation (black) and negative (dark blue) Spearman rank correlation and Fisher’s exact test of amplification (red) or deletion (light blue) frequency. (G) Frequency of CNA in highly proliferative luminal tumors; color key same as (E). (H) Statistical analyses of CNA in proliferative luminal tumors; color key same as (F). Chromosomal boundaries in (E–H) are defined by vertical black lines.

Figure 4. Identification of genomic pathway-associated essential genes in cell lines

(A) Schematic outlining strategy used to identify pathway-specific genetic dependencies. (B) A panel of 27 breast cancer cell lines with both expression data and data from a genome-wide RNAi screen was used to identify pathway-specific genes required for cell viability using a negative Spearman rank correlation (-log₁₀ P-values plotted); significant genes (P<0.05) are shown according to chromosome location. Vertical black lines indicate chromosomal boundaries. (C) ESR1 (D) ERBB2 and (E) STAT1 or JAK3 shRNA levels are inversely associated with the ER, Her2 or Stat1pathway scores.

Figure 5. Identification of essential genes amplified in highly proliferative luminal tumors

(A) Schematic outlining the integrated genomic strategy to identify essential genes amplified in highly proliferative luminal breast tumors. (B) Identification of 21 genes in amplified loci that are unique to highly proliferative luminal tumors and are specifically required for luminal cell line proliferation in vitro. mRNA expression of genes in red and blue were significantly associated with CNA status, with the subset highlighted in red being further validated in the METABRIC dataset; genes in black do not show a significant mRNA-DNA correlation. Candidate genes demarcated by (*) are located at cusp of a CNA segment and were originally excluded, but mentioned here. Genes identified by (#) were not included on mRNA expression microarrays, and the correlation between DNA and mRNA expression was not assessed.

Figure 6. Candidate gene amplification correlates with a poor prognosis

Amplification of (A) FGD5 (N_AMP=51, N_NoAMP=337), (B) METTL6 (N_AMP=51, N_NoAMP=337), (C) DTX3 (N_AMP=71, N_NoAMP=317) and (D) MRSP23 (N_AMP=127, N_NoAMP=261) correlated with poor disease-specific outcome in the luminal breast cancer patients in the TCGA dataset (n=388) while (E) CPT1A (N_AMP=111, N_NoAMP=277) amplification had no effect on prognosis. Consistent results were observed in the METABRIC dataset (n=1,333) for (F) FGD5 (N_AMP=42, N_NoAMP=1,218), (G) METTL6 (N_AMP=44, N_NoAMP=1,278), (H) DTX3 (N_AMP=67, N_NoAMP=1,266), (I) MRPS23 (N_AMP=266, N_NoAMP=1,062) and (J) CPT1A (N_AMP=241, N_NoAMP=1,029). Samples in the METABRIC dataset missing CNA calls were excluded. For each analysis, P-value determined by log-rank test and Hazard Ratio (HR) are reported.