• 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br Next we systematically correlated the quantitative proteo


    Next, we systematically correlated the quantitative proteo-types of the 96 individually measured breast cancer samples and ordered the resulting correlation coefficients according to the classical tumor subtypes (Figure 1B). Spearman correla-tion of proteomic profiles across the entire dataset was high (R > 0.97). The highest intra-group correlation of proteotypes was within luminal A subtype (R = 0.9900). Very high correlation was also observed within the luminal B subtype (both HER2 , R = 0.9866, and HER2+, R = 0.9878) and between luminal A and luminal B subtypes (R = 0.9865; Figure 1B). Furthermore, we found a high correlation of some samples of the HER2-enriched subtype with some (mostly lymph-node-positive) luminal B HER2+ samples, indicating that a higher degree of similarity in HER2+ tumors of luminal B and HER2-enriched subtypes were apparent from the proteotype. The group of tri-ple-negative tumors exhibited slightly lower inter- (R < 0.9852) and particularly intra-group (R = 0.9840) correlation, potentially indicating tumor heterogeneity not captured by the conven-tional tumor classification. In summary, we found that clus-tering by proteotypes closely recapitulates conventional tumor subtyping, but we also found that some of these subtypes are more heterogeneous (triple-negative tumors) than others (Figure 1B).
    Figure 1. Correlation of Breast Cancer Tissue Classification Based on Conventional Subtypes and Proteotypes
    (A) Unsupervised hierarchical clustering of 10 pooled samples from lymph-node-positive and negative samples of the five conventional breast cancer subtypes. Colors represent log2 protein intensities normalized to median. The sample pools are designated by the conventional subtype nomenclature and color coded as
    follows: luminal A (LA) (yellow); luminal B (LB) (light green); luminal B HER2 positive (LBH) (dark green); HER2 enriched (HER) (turquoise); and triple negative (TN)
    (pink). Lymph 104807-46-7 status is as follows: negative (N ) or positive (N+). The figure shows a high similarity of proteotype patterns of pairs of N+ and N tissues within each individual subtype. Furthermore, ER-positive and HER2-negative subtypes cluster together (see a close clustering of groups designated LA, LB); similarly, ER negative subtypes, HER and TN, cluster together.
    (B) Correlation matrix of 96 individual samples ordered according to their subtypes (LA [n = 48], LB [n = 16], LBH [n = 8], HER2 [n = 8], and TN [n = 16]; see Data S1 for more details on samples). Colors represent correlation of summarized log2 protein intensities normalized to median, scaled from blue (least correlated) through black to red (most correlated). Correlations of samples within each subtype are visible, most significantly for luminal A and luminal B and for correlation of both these subtypes. Triple-negative tumors show the highest intra-group heterogeneity.
    Pathways and Proteins Associated with Key Breast Cancer Characteristics
    Having a large, high-quality proteomic dataset at hand, we were interested in identifying pathways and proteins that are impor-tant for breast cancer biology and progression. We first identi-fied proteins that are differentially expressed in tumors of different ER status, tumor grade, HER2 status, or lymph node status (Data S3). We then used gene set enrichment analysis (GSEA) to find pathways that are enriched among the most differentially abundant proteins in these comparisons (Figure 2). Among these, there were several pathways known to be asso-ciated with the particular phenotype, for example, an enrich-ment of the nuclear factor kB (NF-kB) pathway in ER+ tumors, in agreement with its role in proliferation and metastasis of luminal tumors (Azim et al., 2015; Bouchal et al., 2015; Pratt et al., 2009). The list of pathways enriched in high-grade tumors was led by the MCM pathway, which includes pro-proliferation proteins of the MCM family regulating cyclin-dependent ki-nases and DNA replication (Shetty et al., 2005; Wojnar et al., 2010). In HER2+ tumors, we found an enrichment of proteins belonging to the VEGF pathway, namely seven upregulated subunits of eukaryotic translation initiation factors 2 and 2B, 
    which are known to be regulated by HER2 (Sequeira et al., 2009). In lymph-node-positive tumors, we found members of the CARM1 and regulation of the ER pathway (CARM_ER) to be enriched, potentially indicating an involvement of chro-matin-remodeling factors in breast cancer progression and metastasis (Wang et al., 2014). All these and further enriched pathways shown in Figure 2 could be highly relevant for breast cancer biology and warrant further investigation as potential targets of breast cancer therapy.
    Selection of Discriminant Proteins for Improved Classification of Breast Cancer Subtypes