Identification of Personalized Chemoresistance Genes in Subtypes of Basal-Like Breast Cancer Based on Functional Differences Using Pathway Analysis

Abstract

Breast cancer is a highly heterogeneous disease that is clinically classified into several subtypes. Among these subtypes, basal-like breast cancer largely overlaps with triple-negative breast cancer (TNBC), and these two groups are generally studied together as a single entity. Differences in the molecular makeup of breast cancers can result in different treatment strategies and prognoses for patients with different breast cancer subtypes. Compared with other subtypes, basal-like and other ER+ breast cancer subtypes exhibit marked differences in etiologic factors, clinical characteristics and therapeutic potential. Anthracycline drugs are typically used as the first-line clinical treatment for basal-like breast cancer subtypes. However, certain patients develop drug resistance following chemotherapy, which can lead to disease relapse and death. Even among patients with basal-like breast cancer, there can be significant molecular differences, and it is difficult to identify specific drug resistance proteins in any given patient using conventional variance testing methods. Therefore, we designed a new method for identifying drug resistance genes. Subgroups, personalized biomarkers, and therapy targets were identified using cluster analysis of differentially expressed genes. We found that basal-like breast cancer could be further divided into at least four distinct subgroups, including two groups at risk for drug resistance and two groups characterized by sensitivity to pharmacotherapy. Based on functional differences among these subgroups, we identified nine biomarkers related to drug resistance: SYK, LCK, GAB2, PAWR, PPARG, MDFI, ZAP70, CIITA and ACTA1. Finally, based on the deviation scores of the examined pathways, 16 pathways were shown to exhibit varying degrees of abnormality in the various subgroups, indicating that patients with different subtypes of basal-like breast cancer can be characterized by differences in the functional status of these pathways. Therefore, these nine differentially expressed genes and their associated functional pathways should provide the basis for novel personalized clinical treatments of basal-like breast cancer.

Fig 1. Comparison of differentially expressed genes (DEGs) in Basal and Luminal breast cancer (BC).

The left circle (blue) represents DEGs in basal type BC patients, and the right circle (orange) represents DEGs in luminal type BC patients. The overlapping and unique DEGs in two types of BC are shown using a Venn diagram.

Fig 2. Hierarchical clustering of luminal breast cancer samples.

A green-red heat map was used to visualize the clustering results. As illustrated, luminal type BC can be divided into multiple subgroups, indicated with different colors. Both similarities and differences were present between the subgroups. The red and green color key in the heat map represent up- and downregulated genes, respectively.

Fig 3. The clustering results for the basal-like subtype.

A heat map was used to visualize the clustering results for basal BC. Basal BC can be divided into 4 subgroups, indicated with different colors. CR and NOCR represent sensitive and drug resistant patients, respectively.

Fig 4. Pathway enrichment analysis.

This figure depicts the results of the KEGG functional pathway enrichment analysis with genes specific to subgroups 1 and 2, as well as the genes shared between these subgroups. The pathways in the blue box represent the pathways enriched for the subgroup 1-specific genes, the pathways in the green box represent those enriched for the subgroup 2-specific genes, and those in the purple box represent those enriched for the common genes. Only the top five pathways with the highest significance are listed in the figure; more detailed results are described in S2 Table.

Fig 5. Pathway deviation scores of the subgroups.

Subgroups 1 and 2 were marked with blue and red lines. To observe the deviation of subgroup 1 and subgroup 2 from the sensitive range, we also calculated the deviation scores of subgroup 3 and subgroup 4, marked in green (the deviation scores were the same for these two subgroups). The scores of the 16 pathways ranged from 0 to 1.6.

Fig 6. Degree distribution in the PPI network.

The network was constructed using candidate genes and common resistance genes. The degree of a gene represents the number of adjacent genes in the network that directly interact with that gene. The degree of all genes was converted using base the base 2 logarithm. Genes with a degree distribution of 4–6 were least common. These genes have a greater impact on the network and are considered to be important drug-resistance markers.