A whole-genome sequence and transcriptome perspective on HER2-positive breast cancers

Abstract

HER2-positive breast cancer has long proven to be a clinically distinct class of breast cancers for which several targeted therapies are now available. However, resistance to the treatment associated with specific gene expressions or mutations has been observed, revealing the underlying diversity of these cancers. Therefore, understanding the full extent of the HER2-positive disease heterogeneity still remains challenging. Here we carry out an in-depth genomic characterization of 64 HER2-positive breast tumour genomes that exhibit four subgroups, based on the expression data, with distinctive genomic features in terms of somatic mutations, copy-number changes or structural variations. The results suggest that, despite being clinically defined by a specific gene amplification, HER2-positive tumours melt into the whole luminal-basal breast cancer spectrum rather than standing apart. The results also lead to a refined ERBB2 amplicon of 106 kb and show that several cases of amplifications are compatible with a breakage-fusion-bridge mechanism.

Figure 1. Summary of biological and genomic features of the 64 sequenced HER2+ tumours.

(a) (From top to bottom) RNA expression groups; PAM50 subtypes; ER, PR and HER2 IHC statuses; HER2 CNV status; estimated ploidy; Fraction of genome altered (FGA) quartiles; number of large scale transitions (LST) quartiles; progenitor luminal gene signature (pLum) quartiles; mature luminal gene signature (mLum) quartiles; number of intrachromosomal SV in chromosome 17 (SV17) quartiles; (b) mutations, amplifications and homozygous deletions observed in a selected set of putative driver genes.

Figure 2. Examples of CNV patterns in the ERBB2 amplicon.

X axis: position on chromosome 17, the minimal region defining the ERBB2 amplicon is indicated by the blue shaded box. Y axis: tumoral integer copy number, computed in 1 kb binned read counts (Methods). Horizontal red segments indicate the closest sequence consistent with a breakage−fusion−bridge (BFB) fold and the purple curved line represents the corresponding folding pattern. Vertical lines indicate the location of detected breakpoints from discordant and clipped read pairs (Methods); plain lines correspond to intra-chromosomal events and dashed lines to inter-chromosomal events; line colour (and glyph on the bottom) correspond to the clipped reads orientation, red (<): reads clipped to the left, blue (>) reads clipped to the right. If the predicted BFB folding pattern is correct, clipped reads orientation should correspond to a left (resp. right) fold in the purple line. Black lines correspond to missed breakpoints, which (approximate) position could be inferred from a copy-number change. (a) A BFB consistent sequence composed of intra-chromosomal events (plain lines) only. (b) Same as previous with additional inter-chromosomal events (dashed lines) not involved in BFB. (c) A BFB consistent sequence composed of both intra- and inter-chromosomal events. The presence of inter-chromosomal events suggests that the amplification process may involve other chromosome(s). (d) Example of a focal amplification, unlikely to occur by a BFB mechanism, suggesting that other mechanisms (double minutes) may be involved.

Figure 3. Multiple correspondence analysis of selected biological and genomic variables.

Patients are represented by small grey dots and variables (categories) are represented by coloured squares, triangles and large dots. Categories indicated by triangles and squares were used for MCA analysis whereas categories indicated by larger dots were just projected on the resulting map. This map has a simple geometrical interpretation: a category is plotted at the centre of gravity of the patient points for those patients that choose that category (conversely, at a scaling factor, patient points are located at the centre of gravity of categories they choose). As an example, patients points have been linked to the RNA group they belong to (grey ellipses). Therefore the proximity of two categories suggests that they are chosen by a similar set of patients. Categories labels are as follows: RNA groups (RNA.A, RNA.B, RNA.C and RNA.D); PAM50 subtypes (LumB, Her2, Basal); ER status (ER.POS, ER.NEG), PR status (PR.POS, PR.NEG); TP53 mutations: (TP53.M, TP53.WT); fraction of genome altered (FGA.L (lower quartile), FGA.H (upper quartile)); number of interchromosomal SVs in chromosome 17 (SV17.L (lower quartile), SV17.H (upper quartile)); Number of Large Scale Transitions (BRCAness score) (LST.L (lower quartile), LST.H (upper quartile)); progenitor luminal signature score (pLum.L (lower quartile), pLum.H (upper quartile)); mature luminal signature score (mLum.L (lower quartile), mLum.H (upper quartile)); for clarity, interquartile points are not plotted.