Ask the Expert: Developing a Whole-Genome CRISPR Screen to Improve Chinese Hamster Ovary Cells for Biomanufacturing

View PDF

Great strides have been made in increasing the yield of biotherapeutic proteins mainly through process improvements with a focus on media and feed strategies. But aside from single-gene knockouts to allow for metabolic selection systems, the Chinese hamster ovary (CHO) host cell line remains largely unchanged from what was used 30 years ago. In a webinar on 21 March 2018, Jamie Freeman (product manager at Horizon Discovery) focused on this as a bioproduction application of clustered regularly interspaced short palindromic repeats (CRISPR) screening technology.

Freeman’s Presentation
A few years ago, Horizon realized the need for making cutting-edge CHO cells accessible to companies of all sizes. Starting with an existing industry-standard platform — the glutamine synthetase (GS) knockout approach to metabolic selection — the company built a GS knockout. Adventitious-agent testing and good manufacturing practice (GMP) banking of the cell line were complete in 2016. Since then, more than 20 companies have licensed it, and a successful investigational new drug (IND) application validates regulatory acceptance.

Horizon is using that GS knockout as a platform cell line supported with robust protocols and technical support. Continuing to improve on that, the company intends to develop a panel of lines with improved capacity for manufacturing proteins.

First, CRISPR is used to identify genes that have a significant impact on protein expression. Once those targets are identified, genetic engineers use recombinant adenoassociated virus (rAAV) to reengineer those genes with a clear intellectual property position. A screen to identify essential genes and those that improve cell fitness was used as a proof-of-concept to validate the CRISPR screening conditions. A compatible fluorescence-based assay for antibody titer also has been validated.

Horizon can screen for improved titers as well, starting with a CHO line that expresses a given monoclonal antibody (MAb). The line is exposed to a CRISPR library and sorted using the fluorescence-based assay for those cells with higher MAb expression. Horizon then uses validated bioinformatics approaches to identify genes that improve titer through guide enrichment.

Questions and Answers
To which scale do you grow cells in the antibody titer screen? It was all done at shake scale in normal Erlenmeyer flasks.

Can you screen for attributes other than titer? We focused on titer because it is important to our customers and the industry. But we can look at other attributes such as cell growth, phenotypes, and robustness under different stresses.

You mentioned sorting based on MAb expression by fluorescence-activated cell sorting (FACS). Did you expect cell-surface MAb expression? A lot of work went into optimizing that assay to reduce background fluorescence and improve sensitivity. With a cold capture approach, we take a snapshot of cells secreting antibody, not just building it up within them.

Have you tried removing genes that cause problems during downstream purification? We are interested in that. It tends to be relatively product dependent, but if specific HCPs are problematic, we can look at removing those.

Have you tested the stability of new clones over multiple generations? We haven’t yet but will consider it when validating the hits. Stability studies take a long time, so we want to do those only for our lead candidates.

How did you determine improved cell fitness during initial screening after
15 generations? We used a bioinformatics approach to the relative proportion of different guides. If a guide does not affect cell growth, then its proportion stays the same. If you knock out a gene that improves fitness, then it would be over-represented.

Have any of your cell lines been used in perfusion systems at high densities? A perfusion bioreactor company is working with our GS cells to generate that data.

Can you use your library to screen other CHO cells? It depends on how divergent other lines are from the K1 line because the library is based on our sequence to our line. The more divergent the cells, the higher the risk that some guides will miss their targets. We have a strong library to use against other CHO cells with a high degree of conservation, but we haven’t looked in detail at how divergent different lines are.

Might knockout screening not identify any strong targets? We are mitigating that risk through the iterative approach. We can examine our data sets for effects on different biological pathways and protein–protein interactions, then try to combine those based on scientific rationale and perform a more objective screen-on-screen approach. Alternatively, we can rescreen the library on a background of cells with a knockout in genes that only have a moderate impact in isolation.

Is it possible to remove retrovirus sequences? It is. The difficulty is that these are often high–copy-number insertions, making rAAV a difficult technology to use. CRISPR is best for removing those kinds of sequences but may involve an intellectual property issue.

More Online
Find the full presentation of this webcast — with slides and audio — on the BPI website:

Leave a Reply