Identification of clinically predictive metagenes that encode components of a network coupling cell shape to transcription by image-omics

Abstract

The associations between clinical phenotypes (tumor grade, survival) and cell phenotypes, such as shape, signaling activity, and gene expression, are the basis for cancer pathology, but the mechanisms explaining these relationships are not always clear. The generation of large data sets containing information regarding cell phenotypes and clinical data provides an opportunity to describe these mechanisms. Here, we develop an image-omics approach to integrate quantitative cell imaging data, gene expression, and protein-protein interaction data to systematically describe a "shape-gene network" that couples specific aspects of breast cancer cell shape to signaling and transcriptional events. The actions of this network converge on NF-κB, and support the idea that NF-κB is responsive to mechanical stimuli. By integrating RNAi screening data, we identify components of the shape-gene network that regulate NF-κB in response to cell shape changes. This network was also used to generate metagene models that predict NF-κB activity and aspects of morphology such as cell area, elongation, and protrusiveness. Critically, these metagenes also have predictive value regarding tumor grade and patient outcomes. Taken together, these data strongly suggest that changes in cell shape, driven by gene expression and/or mechanical forces, can promote breast cancer progression by modulating NF-κB activation. Our findings highlight the importance of integrating phenotypic data at the molecular level (signaling and gene expression) with those at the cellular and tissue levels to better understand breast cancer oncogenesis.

© 2017 Sailem and Bakal; Published by Cold Spring Harbor Laboratory Press.

Figures

Figure 1.

Integrating imaging and expression data. (A) The three-way relationship between cell shape, signaling states, and cancer progression. (B) Workflow for linking cell shape to transcription and patient outcome. (C) Representative images of different breast cancer lines (BCLs) to illustrate the variation in nucleus/cell area ratio (N/C area) MDA-MB-453, CAMA1, hs578T, and HCC1143 cells, where cell lines to the left have the highest N/C area and cell lines to the right have the lowest N/C area. Red: DAPI, cyan: DHE. Scale bar = 30 µm.

Figure 2.

Shape-gene interaction network. A network of the interactions between the proteins encoded by shape-correlated genes, selected TFs, and shape features. Node size and font size represent the betweenness of a node, which reflects the centrality of the node.

Figure 3.

Analysis of the shape-gene network. A plot of the properties of gene nodes summarizing degree, closeness, and stress. Protein names for nodes that have high values for any of these features are shown.

Figure 4.

SMAD3-NF-κB subnetwork. Proteins that are in a direct path from a phenotypic feature to RELA, SMAD3, or YAP1 (Supplemental Table S6) and their interactions based on STRING. Edges in dashed lines are based on gene expression and indicate feedback from a TF to proteins encoded by shape-correlated genes.

Figure 5.

The expression profiles of shape-correlated genes that drive transcriptional activities of SMAD3 and RELA in luminal versus basal breast cell types. (A) Representative examples of luminal vs. basal shapes. Red: DAPI, blue: DHE. Scale bar = 50 µm. (B) Clustering of 18 BCLs based on the expression of shape-correlated genes in Figure 4 separates cell lines into luminal (green) and basal (blue) subtypes. The basal cluster includes only basal cell lines. The luminal cluster includes mostly luminal cell lines and the basal A cell lines HCC70 and HCC1954. (C) Networks of the expressed genes in the luminal/basal clusters in B, where differentially expressed genes between luminal and basal clusters are represented as circles. Genes that have a higher average expression difference between luminal and basal cluster have a larger node size. Node color indicates the average expression values in each cluster.

Figure 6.

The prognostic value of the morphological metagenes and their associations with the clinical parameters in the METABRIC data set. (A) Association between tumor grade and cell W/L, cell area, NF, and protrusion area metagenes. All these associations are significant, with P-value < 0.0005 using the Jonckheere–Terpstra test. Error bars indicate the standard error of the mean (SEM). (B) Association between tumor grade and NF SD, protrusion area SD, cell W/L SD, and cell area SD metagenes. These associations are significant, with P-value < 0.0005 using the Jonckheere–Terpstra test. Error bars indicate the SEM. (C–F) Kaplan–Meier curves to illustrate the disease-specific survival probabilities of patient groups in discovery and validation cohorts in the METABRIC data set stratified by (C) cell area, (D) cell W/L, (E) NF SD, and (F) cell area SD metagenes.

Figure 7.

Derivation of NF-κB response metagene and its association with the clinical parameters in the METABRIC data set. (A) CAMA1 and SUM159 cells stained with anti-RELA/NF-κB antibody (−/+TNF). Scale bar = 50 µm. (B) Representation of seven morphological BCL features and RELA response (fold change +TNF/−TNF) using PhenoPlot (Sailem et al. 2015), where BCL glyphs are positioned based on the value of cell W/L (x-axis) and protrusion area (y-axis). Cell line label color indicates molecular subtype. Red: luminal, green: basal A, and blue: basal B. (C) Association between Pam50 subtype and NF-κB response metagene (Jonckheere–Terpstra test P-value < 0.0005). Error bars indicate the SEM. (D) Association between tumor grade and NF-κB response metagene (Jonckheere–Terpstra test P-value < 0.0005). Error bars indicate the SEM. (E) Kaplan–Meier curves to illustrate the disease-specific survival probabilities of patient groups in the discovery cohort in the METABRIC data set, stratified by NF-κB response metagene.