PUB - Publikationen an der Universität Bielefeld

Chinese hamster ovary (CHO) cells represent the most prevalent platform for biopharmaceutical protein production. Random integration of a transgene into the genome during cell line development often leads to a loss of productivity during upstream processing. This is due to loss of gene copies or epigenetic silencing of the recombinant promoters. Both effects depend on the integration sites. Therefore, expensive and labor-intensive stability tests and screenings have to be performed. By site-directed integration into stable integration sites, this problem can be solved. However, only few experimentally validated stable integration sites for CHO have been published. In this work, chromatin immunoprecipitation (ChIP) sequencing was used to predict epigenetically stable integration sites in the CHO-K1 genome. Using genome-wide analysis of the histone modifications H3K4me3, H3K9me3, and H3K27ac and their association with gene expression, 709 potentially stable integration sites with specific histone modifications were predicted. The enhanced green fluorescent protein (eGFP) gene was site-specifically integrated into three of these integration sites mediated by CRIPSR/Cas9. The integration sites were validated by targeted Cas9 sequencing using nanopores. The specificity of site-specific integration was between 23% and 73%. Compared to a cell pool with random integration, the site-specifically integrated cell pools were characterized in a long-term cultivation for 70 doublings. This cultivation was used to investigate productivity and long-term stability. At the beginning of the cultivation, all site-specifically integrated cell pools showed a higher productivity than the random cell pool, whose productivity had already dropped by 75 % by day 21. Two of the three site-specifically integrated cell pools proved to be epigenetically stable. One cell pool (B) showed at least a slower decline in productivity than the random cell pool. Pool-A could be identified as a real safe harbour, as neither gene copy loss nor transcript level loss occurred. At the end of the long-term cultivation, which would correspond to the period of final production volume in the biopharmaceutical industry, the productivity of Pool-A was 15-fold higher than in the randomly integrated cell pool. Thus, it was shown that active and stable integration sites can be predicted by histone modifications. In addition, long-read transcriptome sequencing was performed for the first time for CHO cells. Using full-length cDNA and native RNA sequencing on nanopores, 5800 new transcript isoforms and 400 unannotated transcripts were identified after correction with the identified splice sites from Illumina sequencing. Native RNA sequencing was also used to analyze the poly(A) tails of transcripts from CHO cells for the first time. A mean length of 113 nt was measured. Mitochondrial and ribosomal transcripts showed shorter poly(A) tails and transcripts associated with the cell periphery and endoplasmic reticulum (ER) showed longer poly(A) tails than average.