Sequence optimization targeting mRNA stability enhances monoclonal antibody titers in CHO cells

Establishes mRNA stability as a tunable design dimension that can be optimized alongside translational metrics in codon optimization. Prior codon optimization has largely centered on translational metrics; here steady-state mRNA abundance is treated as an independent, optimizable objective within a multi-objective genetic algorithm. A combinatorial library of synonymous coding-sequence variants was integrated as single copies at a defined genomic locus in CHO cells with identical regulatory elements, and steady-state mRNA abundance was quantified by deep sequencing of gDNA and mRNA as a stability proxy.

Provides a machine-learning model that predicts mRNA abundance from coding sequence and can be used directly for sequence design. The predictor uses embeddings from a pre-trained nucleotide transformer and feeds the sequence-to-abundance mapping into a genetic algorithm, making abundance a computable optimization target. The model was trained on measured abundance data from the synonymous variant library and experimentally validated in high- and low-abundance optimized Trastuzumab designs.

High-abundance optimized sequences substantially improve expression performance at the cellular level. Relative to low-abundance designs, intracellular mRNA increased 41%, protein titer increased 59%, and cell-specific productivity increased 85%, with comparable viable cell densities. CHO cell lines were generated by targeted integration and protein titer and cell-specific productivity were measured, with low-abundance designs serving as the comparison.

High-abundance optimized sequences outperform benchmark sequences from commercial providers. Compared with benchmark sequences from two commercial providers, titer was 70% higher and cell-specific productivity 98% higher. Titer and cell-specific productivity were measured in targeted-integration cell lines within the same experimental system and compared directly against the commercial benchmarks.

The results apply to monoclonal antibody expression in CHO cells (IgG1 light chain, Trastuzumab-encoding sequences) under a controlled setting of single-copy integration at a defined genomic locus with identical regulatory elements; the authors propose potential applicability to other proteins and expression systems, but validation in this work is limited to that system.

This reading is at the summary level and does not include figures or supplementary material, so model architecture details, library size, statistical testing, and the robustness of abundance optimization across sequence contexts remain open questions to check in the full text; the scope of mRNA stability as a proxy and how well the approach transfers to other proteins and expression systems are also worth watching.