Regulatory heterogeneity in glioblastoma multiforme informs novel drug target discovery


Year:

Session type:

Theme:

Yunpeng Liu1,Ning Shi2,Shan He2,Michael Hemann1,Aviv Regev1
1Massachusetts Institute of Technology,2University of Birmingham

Abstract

Background

Glioblastoma multiforme (GBM) is one of the most malignant forms of cancer. Bulk and single-cell transcriptome profiling have revealed high levels of both inter- and intratumour heterogeneity in GBM. The disease has been stratified into four molecular subtypes according to gene expression - Classical, Neural, Proneural and Mesenchymal, each of which exhibiting distinct mutational signatures and therapeutic responses. However, the underlying regulatory circuitry that gave rise to such heterogeneity and its implications for rational design of therapy are unclear.

Method

We have developed a nonlinear regression model to understand key regulatory networks across the four GBM subtypes. We first constructed a backbone network of transcription factor (TF) - target gene pairs inferred from chromatin landscape data and TF binding motifs, and applied nonlinear regression using expression profiles of each subtype to derive subtype-specific regulatory parameters for each TF-gene pair. Next, we mined for co-regulatory TF pairs using correlation analysis. Mechanisms responsible for subtype-specific behaviour of TFs were then inferred from expression, regulatory and co-regulatory signatures. Finally, we simulated the effects of perturbing druggable signature TFs and their partners by propagating changes in expression to the corresponding target genes and then to the protein signaling layer using a random walk-based algorithm. 

Results

We show that subtype-specific repurposing of TFs explains a significant proportion of subtype-specific transcription landscapes. At least two mechanisms for TF repurposing are implied - differential expression of the TF itself and differential partnering with co-regulatory TFs. Using effectors of the apoptosis pathway as readout in our in silico perturbation analysis, we show that targeting a subset of druggable signature TFs and/or their partners may be specifically beneficial for treating the corresponding subtypes of GBM.

Conclusion

Data-driven modeling of transcription regulation in GBM is capable of gaining new biological insight into the molecular underpinnings of its heterogeneity and aids rational design of subtype-specific targeted therapy.