The Cancer Genome Atlas Network recently reported findings of analyses of primary breast cancers in a total of 825 patients using genomic DNA copy number arrays, DNA methylation, exome sequencing, mRNA arrays, microRNA sequencing, and reverse-phase protein arrays.1 Integration of information across these platforms allowed the investigators to gain deeper knowledge of previously defined gene-expression subtypes in breast cancer and confirmed the existence of four main breast cancer classes—luminal A, luminal B, HER2-enriched, and basal-like—when data from five of the platforms were combined, with each of these classes showing significant molecular heterogeneity. As noted by the authors, the finding of four main subtypes caused by different subsets of genetic and epigenetic abnormalities suggests that much of the clinically observed heterogeneity of tumor behavior occurs within rather than across these major biologic subtypes.
Other Key Findings
Major findings of the analyses included identification of somatic mutations in only three genes (TP53, PIK3CA, and GATA3) with an incidence of > 10% across all breast cancers. Numerous subtype-associated and novel gene mutations were identified, including enrichment of specific mutations in GATA3, PIK3CA, and MAP3K1 in the luminal A subtype. For HER2-classified tumors, two novel protein expression–defined subgroups were identified, possibly produced by stromal/microenvironment elements; specific signaling pathways characterized the two molecular subtypes, including an EGFR/phosphorylated EGFR/HER2/phosphorylated HER2 signature within the HER2-enriched subtype. Basal-like tumors showed many molecular similarities with high-grade serous ovarian tumors, suggesting a related etiology and the potential utility of investigating similar therapeutic approaches to the two cancers.
In addition to identifying nearly all genes previously implicated in breast cancer (PIK3CA, PTEN, AKT1, TP53, GATA3, CDH1, RB1, MLL3, MAP3K1, and CDKN1B), the investigators identified numerous novel significant mutations, including those in TBX3, RUNX1, CBFB, AFF2, PIK3R1, PTPN22, PTPRD, NF1, SF3B1, and CCND3.
Of note, RUNX1 and CBFB, both of which are rearranged in acute myeloid leukemia and mutations in which interfere with hematopoietic differentiation, contained 19 and 9 mutations, respectively. PIK3R1 contained 14 mutations, with clustering in the PIK3CA interaction domain similar to previously identified mutations in glioma and endometrial cancer. A significant exclusion pattern was observed among PIK3R1, PIK3CA, PTEN, and AKT1 mutations (P = .025). SF3B1 had 15 mutations, including recurrent K700E substitutions; SF3B1 mutations have also been identified in myelodysplastic syndromes and chronic leukocytic leukemia.
Luminal Tumors
Notable findings in luminal/estrogen receptor (ER)-positive subtypes included the greatest heterogeneity of gene expression and the largest number of significantly mutated genes. The analyses confirmed the high rate of PIK3CA mutation in luminal/ER-positive cancers, identified a high frequency of MAP3K1 and MAP2K4 mutations (which were largely mutually exclusive), and a higher rate of TP53 mutation in luminal B vs luminal A cancers. Additional investigation suggested that the TP53 pathway remains largely intact in luminal A cancers but often is inactivated in the more aggressive luminal B cancers. The tumor suppressor RB1 had higher levels in luminal A tumors, as well. Pathway differences associated with luminal B vs luminal A cancers included hyperactivation of transcriptional activity associated with MYC and FOXM1 proliferation.
Based on the genomic characterizations, potential therapeutic approaches include use of inhibitors of the activated PIK3CA kinase or its signaling pathway, AKT1 inhibitors, since 11 of 12 identified AKT1 variants were luminal, and PARP inhibitors for BRCA1/BRCA2 mutations. Based on amplification of genomic DNA copy number, other targets might include FGFR, IGFR1, cyclin D1, CDK4, and CDK6.
HER2-based Classification
Of all HER2-enriched tumors, 20% were ER-positive/HER2-negative, 68% were HER2-positive, and 9% were triple-negative (ie, negative for ER, HER2, and progesterone receptor). Analysis of these tumors strongly indicated that there are at least two types of clinically defined HER2-positive tumors. Not all clinically HER2 protein–positive tumors were of the HER2-enriched mRNA subtype and vice versa. When HER2-positive protein and HER2-enriched mRNA subtypes overlapped, there was a strong signal of EGFR/phosphorylated EGFR/HER2/phosphorylated HER2. Approximately 50% of clinically HER2-positive tumors fell into this category, with the remainder predominantly being of the luminal mRNA subtype.
Gene-expression analysis for these subtypes showed that genes largely tracked with ER status but also indicated that the HER2-enriched-mRNA-subtype/HER2-positive tumors had significantly higher expression of a number of receptor tyrosine kinases, including those for FGFR4, EGFR, and HER2, as well as genes within the HER2 amplicon (including GRB7); the luminal-mRNA-subtype/HER2-positive tumors showed higher expression of the luminal cluster of genes, including GATA3, BCL2, and ESR1. Other notable findings included significant enrichment of TP53 mutations in HER2-enriched or ER-negative tumors and exclusive presence of GATA3 mutations in luminal subtype or ER-positive tumors.
Findings in this study that suggest therapeutic targets in HER2-positive disease include a high frequency of PIK3CA mutations (39%), lower frequencies of PTEN and PIK3RI mutations, and genomic loss of PTEN and INPP4B. Other potentially druggable mutations included variants in HER family members, including two somatic mutations in HER2, two in EGFR, and five in HER3, with these findings suggesting that joint targeting of EGFR and HER2 could be beneficial. The HER2-enriched mRNA subtype showed high aneuploidy, the highest somatic mutation rate, and DNA amplification of other potential targets, including FGFRs, EGFR, CDK4, and cyclin D1.
Basal-like Tumors
Most of the basal-like tumors (80%) were triple-negative cancers. Basal-like tumors had a high frequency of TP53 mutation (80%), with inferred effect on TP53 pathway activation suggesting that most, if not all, basal-like cancers have loss of TP53 function. Although PIK3CA was the next most frequently mutated gene (9%), inferred PI3K pathway activity was highest in basal-like cancers; alternative means of activating this pathway might include loss of PTEN and INPP4B or amplification of PIK3CA.
Basal-like tumors had numerous features in common with serous ovarian tumors, including widespread genomic instability and common gains of 1q, 3q, 8q, and 12p, and loss of 4q, 5q, and 8p. When both types of tumor were compared with luminal tumors, they shared BRCA1 inactivation, RB1 loss and cyclin E1 amplification, high expression of AKT3, MYC amplification and high expression, and high frequency of TP53 mutation. Pathway analysis showed comparably high activity of HIF1-α/ARNT, MYC, and FOXM1 regulatory hubs in the two cancers.
As noted by the authors, the common findings of TP53, RB1, and BRCA1 loss with MYC amplification strongly indicate that these are shared driving events in carcinogenesis for the two types of cancer. Moreover, as suggested by the activity of platinum compounds and taxanes in both types, common therapeutic approaches should be considered in both.
Therapeutic Targets
With regard to potential therapeutic approaches, it was found that approximately 20% of basal-like tumors had a germline or somatic BRCA1 or BRCA2 variants, suggesting that 20% of patients might benefit from a PARP inhibitor or platinum compound. Observed amplifications and deletions in copy number analysis suggested a number of targets including PTEN and INPP4B loss, both of which have been shown to promote sensitivity to PI3K pathway inhibitors.
Many components of the PI3K and RAS-RAF-MEK pathway were amplified, though not typically mutated, including PIK3CA (49%), KRAS (32%), BRAF (30%), and EGFR (23%). Receptor tyrosine kinases that were amplified in some tumors and could constitute drug targets included those for FGFR1, FGFR2, IGFR1, KIT, MET, and PDGFRA. The high HIF1-α/ARNT pathway activity suggests that the basal-like cancers might be susceptible to angiogenesis inhibitors or bioreductive drugs that are activated under hypoxic conditions. ■
Reference
1. Cancer Genome Atlas Network: Comprehensive molecular portraits of human breast tumours. Nature 490(7418):61-70, 2012.
