Engineering the 5′ UTR-Mediated Regulation of Protein Abundance in Yeast Using Nucleotide Sequence Activity Relationships

Abstract

The 5′ untranslated region (5′UTR) plays a key role in post-transcriptional regulation, but interaction between nucleotides and directed evolution of 5′UTRs as synthetic regulatory elements remain unclear. By constructing a library of synthesized random 5′UTRs of 24 nucleotides in Saccharomyces cerevisiae, we observed strong epistatic interactions among bases from different positions in the 5′UTR. Taking into account these base interactions, we constructed a mathematical model to predict protein abundance with a precision of R 2 = 0.60. On the basis of this model, we developed an approach to engineer 5′UTRs according to nucleotide sequence activity relationships (NuSAR), in which 5′UTRs were engineered stepwise through repeated cycles of backbone design, directed screening, and model reconstruction. After three rounds of NuSAR, the predictive accuracy of our model was improved to R 2 = 0.71, and a strong 5′UTR was obtained with 5-fold higher protein abundance than the starting 5′UTR. Our findings provide new insights into the mechanism of 5′UTR regulation and contribute to a new translational elements engineering approach in synthetic biology.

The Ongoing Quest to Crack the Genetic Code for Protein Production

“…The initiation mechanisms in the 5 0 UTR in eukaryotes seem more diverse and complicated than do those for prokaryotes. However, recent studies have shown that the 5 0 UTR sequence has great potential for tuning the expression in eukaryotes, such as S. cerevisiae or Chinese hamster ovary-S (CHO-S) cells (Ding et al, 2018;Petersen et al, 2018;Weenink et al, 2018). These studies provided modular 5 0 UTRs designs that work relatively well with low-context dependence on the downstream ORF.…”

“…Likewise, successful 5 0 UTR prediction algorithms based on multiple regression or machine learning approaches have been developed for yeast (Cuperus et al, 2017;Decoene et al, 2018;Ding et al, 2018). Such big-data analyses, based on randomized sequence libraries, seem a promising road toward better predictive algorithms for robust regulation of synthetic genes.…”

Designing Eukaryotic Gene Expression Regulation Using Machine Learning

“…Features may also be inferred from the original sequence, for example using biophysical models yielding free energy of the 5′-UTR [9,45,54,55], or hybridization energy and target accessibility of miRNA in the 3′-UTR [56]. An important feature is nucleosome occupancy [39,[57][58][59], which has been used to facilitate terminator design [60].…”

5′ untranslated regions: the next regulatory sequence in yeast synthetic biology

“…It was also demonstrated that the obtained model could be used for the forward engineering of 5 UTRs as regulatory elements for synthetic pathways. Ding et al (2018) developed a regression model with a prediction precision of 60%, which was then used as basis for an improved model based on nucleotide sequence activity relationships (NuSAR). After multiple iterative cycles of designing, screening and model reconstruction, the final model was able to explain 71% of the variation in protein synthesis.…”