Fig. 1

A gene expression signature of breast cancer metastases versus paired primary largely reflects non-breast versus breast tissue differences. a Heat map of differential expression in metastasis versus paired primary for a set of 3929 differentially expressed genes with p < 0.001 (paired t-test using log2-transformed data). Each metastasis expression profile was centered on its primary pair (not shown). Breast cancer expression data involving 152 metastases with paired primary are from the Breast International Group (BIG) [4]. Yellow, high expression in metastasis versus primary; blue, low expression in metastasis. Alongside the differential genes from the breast cancer dataset are the corresponding differential expression patterns of normal breast tissues (relative to non-breast tissues) from the GTEx dataset [19]. Heat maps of gene expression-based signatures of immune cell infiltrates [20] (taking the average log2 fold change from paired primary for the immune cell type marker genes) are also shown (NK cells natural killer cells). b Venn diagrams showing the overlapping genes between BIG breast cancer metastasis and GTEx normal breast (respectively considering high and low gene lists from each dataset). Breast cancer metastasis genes are from part a. GTEx normal breast genes by p < 0.000001, comparing log2 expression between normal breast and other normal tissues. Enrichment p values by chi-square test. From parts a-b, the observed global associations of the differential metastasis versus primary genes with both tumor biopsy site and GTEx breast expression patterns indicate that most of the observed expression differences in metastases would involve differences in non-cancer cells between the metastasis biopsy site and the breast, respectively, rather than representing changes intrinsic to cancer cells