Summary
Our preliminary results support a prominent role for dysregulation of RNA binding proteins (RBPs) in cancer
progression. In fact, our analysis of the cancer genome atlas (TCGA) transcriptome data shows that RBPs, as
a group, are significantly more dysregulated in cancer than transcription factors. We propose a multi-faceted
set of computational and experimental studies to systematically identify the set of RBPs that causally
contribute to cancer progression and to characterize their downstream effector mechanisms. In one strategy,
we propose to identify dysregulated RBPs by first discovering their cis-regulatory recognition elements in the 3’
and 5’ UTR of genes that show dynamic mRNA expression—both between tumor vs. normal samples, and
across each of the 25 cancer cohorts in TCGA. This will be accomplished using information-theoretic
algorithms that discover de novo linear and structural RNA motif elements with high sensitivity and low false
discovery rates. We have previously shown that such RNA motifs are the binding sites for RBPs that modulate
mRNA stability, a subset of which regulate tumorigenesis and metastasis. The genes harboring these motifs
constitute an orphan RBP regulon (or RBP module) with a suspected role in cancer progression. In order to
identify clinically significant RBP modules, we propose to develop a computational framework that quantifies
the degree to which the expression of each module stratifies patient survival across the TCGA primary tumor
samples. Our preliminary results have led to the discovery of many such modules with remarkable stratification
of patient survival across multiple cancer types. For the subset of the most clinically prognostic modules, we
will identify their cognate RBPs using both biochemical and CRISPR-based parallel genetic screens. In a
complementary strategy, we will computationally identify such clinically prognostic RBP modules from a
compendium of ENCODE transcriptome data obtained following shRNA knockdowns of each of ~250 RBPs. In
order to identify RBPs that causally contribute to cancer progression, we propose to develop a parallel
CRISPR loss-of-function screen for all RBPs in mouse xenograft models of tumor formation and metastasis.
We will then conduct a more focused CRISPR screen on the top ~20 RBPs that show both significant patient
survival stratification and mouse in vivo tumor effects in our primary comprehensive screen. The top validated
RBPs will then be individually characterized for their roles in a variety of in vitro and in vivo cancer cell
phenotypes. Finally, we propose to develop a parallel mouse in vivo CRISPR epistasis platform to efficiently
determine the specific downstream genes through which the RBP exerts its effects on tumor formation and
metastasis. Our integrated computational/experimental strategy will expand our molecular understanding of a
largely unexplored domain of cancer pathway dysregulation and potentially reveal new principles at work.
Furthermore, our focus on causal pathways of cancer progression will impact the diagnostic, prognostic, and
therapeutic precision with which we approach clinical oncology.