Interactions between the tumor and the blood systemic response of breast cancer patients

Abstract

Although systemic immunity is critical to the process of tumor rejection, cancer research has largely focused on immune cells in the tumor microenvironment. To understand molecular changes in the patient systemic response (SR) to the presence of BC, we profiled RNA in blood and matched tumor from 173 patients. We designed a system (MIxT, Matched Interactions Across Tissues) to systematically explore and link molecular processes expressed in each tissue. MIxT confirmed that processes active in the patient SR are especially relevant to BC immunogenicity. The nature of interactions across tissues (i.e. which biological processes are associated and their patterns of expression) varies highly with tumor subtype. For example, aspects of the immune SR are underexpressed proportionally to the level of expression of defined molecular processes specific to basal tumors. The catalog of subtype-specific interactions across tissues from BC patients provides promising new ways to tackle or monitor the disease by exploiting the patient SR.

Fig 1. Individual characteristics and SR markers of BC subtypes.

(A) Collection of biospecimen from BC patients and controls. (B) Individual characteristics of BC patients and controls. (C) Parallel plot displaying the repartition of BC patients across RNA-based subtyping schemes. (D) Sparse hierarchical clustering of BC patients based on genes expressed in tumor (upper) and the patient SR (lower). Clinicopathological and subtypes attributes are presented below the dendrogram. (E) Significant gene markers of subtypes in SR (false discovery rate, fdr ≤ 0.2). Blue and red shade correspond to under- and over- expression of the marker in a given subtype vs the others, respectively. Shading is proportional to the level of significance of the gene marker.

Fig 2. Gene co-expression networks, modules and associations with clinicopathological attributes of BC patients.

(A) Network visualization using the edge-weighted spring embedded layout from Cytoscape (v3.2.1) including the top gene connections (topological overlap > 0.1) in tumor. Each node (gene) is color-coded by the module to which it belongs, Keywords representing top pathway enrichments (biological processes) are indicated for each module. (B) Network visualization including the top gene connections in the patient SR. The legend follows Fig 2A. (C) Associations between tumor modules and clinicopathological attributes of patients. Associations were estimated using Pearson correlation (Student’s p) or ANOVA. Shading is proportional to -log₁₀(fdr) of the associations (fdr ≤ 0.15). HER2S: HER2 score; LUMS: luminal score; MKS: Mitotic kinase gene expression score; hrt: hormone replacement therapy (D) Associations between SR modules and clinicopathological attributes of patients. The legend follows Fig 2C.

Fig 3. Modules size and overlap in their gene composition across tissues.

(A) Histograms depicting number of genes composing modules in each tissue. Edges between modules indicate significant overlaps in gene composition (Fisher exact test, fdr < 0.01). (B) Expression heatmaps of the 47 genes included in both darkturquoise modules in tumor (upper) and SR (lower). Patients in both heatmaps are linearly ordered based on their ranksum of gene expression in tumors. Yellow vertical lines delimit the region of Independence (ROI₉₅) in tumor that contains 95% of randomly generated samples. Twenty genes out of the 47 common genes are involved in the type 1 IFN signaling pathway (IFN alpha signaling pathway is depicted on the right).

Fig 4. Subtype-Specific Matched Interactions across Tissue (ssMIxT).

(A) Schematic of ssMIxT analysis (B) Significant associations between modules in SR and tumor from BC patients by subtype (MIxT statistic, p-value < 0.005). SR and tumor modules with top pathway enrichment keywords are presented in rows and columns, respectively. Subtype(s) in which the significant associations are found are indicated in the table. Blue and red borders correspond to negative and positive correlations between ranksums, respectively. Findings discussed in the text are highlighted in orange.

Fig 5. Association between the brown tumor and green SR module for two distinct subtypes.

(A) Scatter plot of ranksums of the brown tumor module and the green SR module in ER-/HER2- patients. The top corner depicts the background distributions of the correlations coefficients between ranksums of every modules pairs across tissues in ER-/HER2- patients. (B) Scatter plot of ranksums of the brown tumor module and the green SR module in ER+/lumB patients. Legend follows Fig 5A (C) Expression heatmap of genes in the brown tumor module in ER-/HER2- patients. Patients are linearly ordered based on the ranksum of gene expression in the brown tumor module. Yellow vertical lines delimit the ROI₉₅ in tumor that contains 95% of the randomly generated samples. Genes that are positively and negatively correlated with the ranksum are represented in the right sidebar colored in red and blue, respectively. Top pathway enrichment keywords and representative genes are indicated on the left and right of the heatmap, respectively). (D) Expression heatmap of genes in the brown tumor module in ER+/lumB patients. Legend follows Fig 5C. (E) Expression heatmap of genes in the green SR module in ER-/HER2- patients. Legend follows Fig 5C. Top pathway enrichment keywords and representative genes are indicated on the left and right of the heatmap, respectively. (F) Expression heatmap of genes in the green SR module in ER+/lumB patients. Legend follows Fig 5E. (G) Clinical characteristics of ER-/HER2- patients ordered by the ranksum of gene expression in the brown tumor module. Legend follows Fig 1D. (H) Clinical characteristics of ER+/lumB patients ordered by the ranksum of gene expression in the brown tumor module. Legend follows Fig 1D. Asterisks represent the level of significance of the associations between the gene ranksums for the brown tumor module and clinicopathological attributes of patients. Associations were estimated using ANOVA (fdr < ***0.01). (I) Distribution of ranksums for ER-/HER2- patients and controls induced by the expression of genes in the green SR module. Patients are grouped according to the ROI₉₅ brown tumor module category as defined in Fig 5C. aov: analysis of variance (J) Distribution of ranksums for ER+/lumB patients and controls induced by the expression of genes in the green SR module. Patients are grouped according to the ROI₉₅ brown tumor module category as defined in Fig 5D.

Fig 6. Significant Matched Interactions across Tissue (MIxT) in basalL patients.

(A) The figure summarizes the two sets of significant MIxT in basalL patients detailed in Fig 6A and 6C. Top pathway enrichment keywords are presented for each module. Red and blue arrows correspond to negative and positive correlations between ranksums, respectively. (B) MIxT in basalL patients between the brown tumor module and the darkgreen SR module. Heatmaps are ordered by ranksum of gene expression in the brown tumor module. Asterisks represent the level of significance of the associations between the gene ranksums for the brown tumor module and clinicopathological attributes of patients (fdr < **0.05). Associations were estimated using ANOVA and Pearson correlation for categorical and continuous variable, respectively. Boxplots show the distribution of ranksums for the SR module in patients classified according to their ROI₉₅ tumor module category and controls. (C) The second set of MIxT in basalL patients between the darkgreen tumor module and four SR modules (darkgreen, green, greenyellow, darkgrey). Legend follows Fig 6B.