Stephen Fedak

Abstract Repeat-associated non-AUG (RAN) translation allows for unconventional initiation at disease-causing repeat expansions. As RAN translation contributes to pathogenesis in multiple neurodegenerative disorders, determining its mechanistic underpinnings may inform therapeutic development. Here we analyze RAN translation at G 4 C 2 repeat expansions that cause C9orf72 -associated amyotrophic lateral sclerosis and frontotemporal dementia (C9RAN) and at CGG repeats that cause fragile X-associated tremor/ataxia syndrome. We find that C9RAN translation initiates through a cap- and eIF4A-dependent mechanism that utilizes a CUG start codon. C9RAN and CGG RAN are both selectively enhanced by integrated stress response (ISR) activation. ISR-enhanced RAN translation requires an eIF2α phosphorylation-dependent alteration in start codon fidelity. In parallel, both CGG and G 4 C 2 repeats trigger phosphorylated-eIF2α-dependent stress granule formation and global translational suppression. These findings support a model whereby repeat expansions elicit cellular stress conditions that favor RAN translation of toxic proteins, creating a potential feed-forward loop that contributes to neurodegeneration.

Abstract mRNA therapeutics are revolutionizing the pharmaceutical industry, but methods to optimize the primary sequence for increased expression are still lacking. Here, we design 5’UTRs for efficient mRNA translation using deep learning. We perform polysome profiling of fully or partially randomized 5’UTR libraries in three cell types and find that UTR performance is highly correlated across cell types. We train models on our datasets and use them to guide the design of high-performing 5’UTRs using gradient descent and generative neural networks. We experimentally test designed 5’UTRs with mRNA encoding megaTAL TM gene editing enzymes for two different gene targets and in two different cell lines. We find that the designed 5’UTRs support strong gene editing activity. Editing efficiency is correlated between cell types and gene targets, although the best performing UTR was specific to one cargo and cell type. Our results highlight the potential of model-based sequence design for mRNA therapeutics.

A CGG trinucleotide repeat expansion in the 5' UTR of FMR1 causes the neurodegenerative disorder Fragile X-associated tremor/ataxia syndrome (FXTAS). This repeat supports a non-canonical mode of protein synthesis known as repeat-associated, non-AUG (RAN) translation. The mechanism underlying RAN translation at CGG repeats remains unclear. To identify modifiers of RAN translation and potential therapeutic targets, we performed a candidate-based screen of eukaryotic initiation factors and RNA helicases in cell-based assays and a Drosophila melanogaster model of FXTAS. We identified multiple modifiers of toxicity and RAN translation from an expanded CGG repeat in the context of the FMR1 5'UTR. These include the DEAD-box RNA helicase belle/DDX3X, the helicase accessory factors EIF4B/4H, and the start codon selectivity factors EIF1 and EIF5. Disrupting belle/DDX3X selectively inhibited FMR1 RAN translation in Drosophila in vivo and cultured human cells, and mitigated repeat-induced toxicity in Drosophila and primary rodent neurons. These findings implicate RNA secondary structure and start codon fidelity as critical elements mediating FMR1 RAN translation and identify potential targets for treating repeat-associated neurodegeneration.

Abstract mRNA therapeutics are revolutionizing the pharmaceutical industry, but methods to optimize the primary sequence for increased expression are still lacking. Here, we design 5’UTRs for efficient mRNA translation using deep learning. We perform polysome profiling of fully or partially randomized 5’UTR libraries in three cell types and find that UTR performance is highly correlated across cell types. We train models on all our datasets and use them to guide the design of high-performing 5’UTRs using gradient descent and generative neural networks. We experimentally test designed 5’UTRs with mRNA encoding megaTALTM gene editing enzymes for two different gene targets and in two different cell lines. We find that the designed 5’UTRs support strong gene editing activity. Editing efficiency is correlated between cell types and gene targets, although the best performing UTR was specific to one cargo and cell type. Our results highlight the potential of model-based sequence design for mRNA therapeutics.