A Journey Through the Dark: Exploration of Long Non-Coding RNAs and tRNAs in Chinese Hamster Ovary Cells for Recombinant Protein Production

The importance of biotherapeutic proteins for the treatment of various diseases has grown exponentially over the last few decades, and this growth is predicted to continue in the coming years. Among these, monoclonal antibodies (mAbs) represent the largest class for both revenues and new approvals. Chinese hamster ovary cells (CHO) are the leading platform in industry for the production of these complex molecules requiring human-like post-translational modifications, in particular mAbs. Although CHO cells are capable of producing and secreting mAbs at acceptable yields, numerous attempts to increase the maximum viable cell concentration and productivity have been described in literature exploiting modification of culture conditions, changing the genetic makeup of the vector(s) used to drive expression of the gene(s) of interest, and by engineering host cells. The application of non-coding RNAs, such as miRNAs and siRNAs, to reprogram CHO cells has been explored and allows specific targeting of detrimental genes without loading an additional translational burden on the cell. However, Long non-coding RNAs (LncRNAs), non-coding transcripts >200 nucleotides in length have only recently emerged as key regulators of epigenetics, splicing, microRNAs and translation. Despite the potential for applications in cell engineering, these molecules remain largely unexplored in mammalian expression systems. Further, whilst mRNA translation of coding transcripts is a central regulatory step for cell growth, and thus the yield and quality of recombinant proteins, non-coding tRNAs are an important regulatory molecule in the decoding process. Although recombinant gene sequences are often codon optimized, we do not currently have all the information required around tRNA abundance, modifications and tRNA charging to fully harness codon usage in recombinant sequences. The work reported here presents the first lncRNA and tRNA expression landscape in CHO cells under a variety of conditions and discusses the implications of these on recombinant protein production. To investigate lncRNA and tRNAs in CHO cells, a CHO-S cell linegrown under batch and fed-batch culture was sampled at day 4 and 7 of culture while six IgG1-producing CHO cell lines cultivated in an ambr®15 system with different fed strategies were sampled before inoculation and at day 4, 7 and 12 of culture. The whole transcriptomes were investigated using a mouse microarray providing the surveillance of 24,881 mRNAs and 35,923 lncRNAs for CHO-S samples and RNA-Seq for the IgG-producing cell lines. tRNA abundances were quantified using a previously optimized ARM-Seq protocol. Thousands of differentially expressed lncRNAs were filtered by counting the occurrences of each transcript, assessing sequence conservation, secondary structure and RT-qPCR validation. The behaviour of a group of lncRNAs is described for the first time in CHO cells and the applications for cell engineering discussed. In particular, the CHO cell long non-coding RNA (lncRNA) transcriptome from cells grown in controlled miniature bioreactors is defined under fed-batch conditions using RNA-Seq to identify lncRNAs and how the expression of these changes throughout growth and between IgG producers. lncRNAs associated with productivity and growth characteristics are identified, in particular finding that Adapt15, linked to ER stress,GAS5, linked to mTOR signalling/growth arrest, and PVT1,linked to Myc expression, are differentially regulated during fed-batch culture and whose expression relates to productivity (Adapt15) or growth (GAS5, PVT1). Changes in (non)-coding RNA expression between the seed train and the equivalent day of fed-batch culture are also reported, showing large differences in gene expression between these. The ARM-Seq protocol allowed the identification of 4-5 fold more tRNAs compared to standard sequencing, and was applied to yeast and HEK293 cells to allow comparisons with CHO. Ultimately, tRNA quantifications were used in a translation elongation model to calculate the decoding speed of model recombinant proteins and to generate codon optimized versions based on tRNA abundances.