Global Gene Expression Profiles Identify Metastasis Regulatory Networks

Circos diagram showing integrated global transcript and miRNA profiling analyses. From periphery inward: mRNAs with statistically significant eQTL peaks (red peaks); mRNAs with expression significantly associated with metastatic propensity (blue peaks); miRNAs with metastasis-associated expression (pink peaks); miRNAs predicted to target highly connected network hubs (blue line links miRNA and its predicted target mRNA, red point indicates mRNA position). Genes labeled in purple indicate module central node

Circos diagram showing integrated global transcript and miRNA profiling analyses. From periphery inward: mRNAs with statistically significant eQTL peaks (red peaks); mRNAs with expression significantly associated with metastatic propensity (blue peaks); miRNAs with metastasis-associated expression (pink peaks); miRNAs predicted to target highly connected network hubs (blue line links miRNA and its predicted target mRNA, red point indicates mRNA position). Genes labeled in purple indicate module central nodes (i.e., Cnot2). Genes labeled in green indicate hub transcripts with predicted MIR3470b MREs that are downregulated upon Mir3470b overexpression.

Metastasis is a systemic disease in which cancer cells break away from a tumor and migrate to other parts of the body, usually via the blood or lymphatic systems, to form new tumors. Metastatic tumors are difficult to treat and account for the majority of cancer-related deaths. Susceptibility to metastasis is known to have a genetic component, with some individuals more predisposed than others. However, because of the complex interchange between random genomic and epigenetic events that contribute to the disease, characterization of individual genes or small numbers of genes is not sufficient to understand the processes leading up to metastasis.

Systems biology approaches, such as global gene expression profiling and the mapping of transcriptional networks, have shown promise in identifying molecular nodes that may be critical to metastatic progression, as well as individual genes important to metastatic cascade. These systems-level approaches can guide improvements in diagnostic tools and interventions for metastasis.

Kent Hunter, Ph.D., Head of the Metastasis Susceptibility Section in CCR’s Laboratory of Cancer Biology and Genetics, and his colleagues used a recombinant inbred (RI) mouse line panel to explore global gene profiles predictive of metastasis susceptibility. RI panels are powerful tools for mapping complex traits. Because each recombinant genome is isogenic, genetic variations arising from random meiotic and environmental factors can be minimized through the phenotyping of many individuals within each subline. This gives researchers the opportunity to accurately map quantitative trait loci, thereby increasing the power to detect genotype-phenotype associations, such as those involved in metastasis. The RI panel used by Hunter’s group, AKXD, crossed with MMTV-PyMT transgenic mice yielded progeny with a more than 20-fold difference in metastatic propensity.

Hunter’s team previously demonstrated that heritable differences in miRNA expression underlie metastatic susceptibility, leading to the hypothesis that miRNAs regulate metastatic progression. The researchers, therefore, profiled mRNA and miRNA of their recombinant mice to identify co-expressed metastasis-driving transcriptional networks. The mRNA and miRNA analyses revealed seven mouse genes with human homologs that segregated tumor genomes within a human breast cancer gene repository (GOBO) into high- and low-risk groups with respect to an index of metastatic susceptibility called distant metastasis-free survival (DMFS).
Next the team identified 13 gene expression networks linked to metastatic susceptibility. Several of the networks, named after the most highly connected gene within each network, were found to contain genes that Hunter’s group and others have previously demonstrated to be associated with metastasis efficiency. Also, five of the 13 networks significantly discriminated DMFS high- and low-risk groups in two human breast cancer gene expression datasets (GSE2034 and GSE11121).

The team then chose the network with the highest discrimination index—Cnot2, a core component of the transcriptional regulatory CCR4-NOT complex—for further investigation. They found that CNOT2 expression was 16-fold lower in 53 invasive breast cancer samples than in six normal breast tissue samples, suggesting that its expression is negatively correlated with tumor progression. In vivo tests, performed to validate the causative role of Cnot2, showed similar results: mice implanted with tumor cells programmed to overexpress Cnot2 showed reduced tumor mass and pronounced metastatic suppression, whereas Cnot2 knockdown increased metastasis.

High throughput sequencing of small RNAs derived from the recombinant mouse tumors followed by eQTL analysis—a powerful approach for linking regulatory variants to gene expression—uncovered three miRNAs, Mir216a, Mir216b, and Mir217, as potential inhibitors of metastatic progression. Mice implanted with two different tumor cell lines programmed to reliably express Mir216a, Mir216b, and Mir217 showed pronounced metastatic suppression.

Hunter’s team employed a final computational approach that integrated their mRNA and miRNA profile data to search for miRNA metastasis master regulators. The researchers identified 153 network hub transcripts that significantly discriminated the metastasis index DMFS in the two human breast cancer gene expression datasets (GSE2034 and GSE11121) mentioned above. Eliminating any transcript hubs that did not have commonly shared, computationally predicted miRNA recognition elements, left 38 miRNAs that targeted 49 of the 153 hub genes. The team then looked for miRNAs that targeted mRNAs of the hubs more frequently than chance and found that 32 of the 49 target genes were targets of Mir3470a or Mir3470b. In vitro studies showed that these miRNAs were more highly expressed in metastatic cells than non-metastatic cells, suggesting they may be pro-metastatic. Furthermore, Mir3470a and Mir3470b overexpression downregulated several genes within the Cnot2 network. In vivo studies indicated that overexpression of Mir3470a led to a dramatic increase metastasis and overexpression of Mir3470b led to increased metastatic colonization.

In summary, Hunter’s team established a novel approach to identifying heritable disease susceptibility gene networks. Analysis of these networks, and follow up in vivo studies, revealed a novel regulator of metastasis, Cnot2, suggesting a role for modulation of RNA expression in tumor progression and malignancy. Correlation analyses between metastatic susceptibility and miRNA expression, eQTL analyses, and in vivo validation identified Mir216a, Mir216b, and Mir217 as inherited metastasis-suppressing loci. Computational screening also identified Mir3470a and Mir3470b as metastasis enhancers.

Summary Posted: 12/2013

Reference

Faraji F, Hu Y, Wu G, Goldberger NE, Walker RC, Zhang J, Hunter KW. An integrated systems genetics screen reveals the transcriptional structure of inherited predisposition to metastatic disease. Genome Research, December 9, 2013 PubMed Link