The cell subtype of origin and not the tumor-promoting event is critical for the formation of breast cancer subtypes

Objective:

Breast cancer (BC) is stratified into subtypes with different treatment responses and clinical outcomes, suggesting distinct origins. However, the precise cell of origin (COO) for each subtype remains debated. While previous conclusions were drawn from mice, we here use human cells, which is relevant given the human-mouse differences.

Material and methods:

We previously demonstrated that mutations correlate with epigenomes (Polak et al, Nature, 2015). Here, random forest regression modeled mutational distributions from 532 BCs across 4 subtypes based on ChIP-seq reads from 6 normal breast cell types. For refining our approach 2,641 whole genomes across 30 cancer types and 98 epigenetically analyzed normal tissue types were used.

Results:

First, we confirmed that our method is able to reveal tumor origins across various cancer types by identifying a COO in 28/30 tumor types. Next, we focused on BC discovering that basal-like tumors originated from luminal progenitors and all other subtypes (LumA/B/HER2-enriched) from mature luminal cells. This association held true even when accounting for different mechanisms of homologous recombination repair (HRR). In addition, we found that the origin of basal-like BCs is independent of ethnicity.

Conclusion:

We here show that it is possible to infer the COO from epigenetic and mutation data. Our BC results indicate that each subtype derives from a cell type along the luminal differentiation hierarchy. In addition, the final histotype appears to be more strongly associated with the COO than the HRR inactivation event. In summary, our findings could lead to improved prevention and treatment of BC.