Development of a computational glycan engineering tool for biologics manufacturers

Lead Research Organisation: University of York
Department Name: Biology


Biologics are currently the most successful class of pharmaceuticals, with seven of the top 10 drugs marketed in 2018 hailing from this class and grossing a total of $60 bn. The majority of these protein drugs are glycosylated, meaning they are decorated with carbohydrate chains also called glycans. The presence of glycans causes a high degree of variability, which stems from the inherent heterogeneity fostered by the biosynthetic machinery that builds glycans. Variability is a problem for these drugs because proteins with different glycan structures attached to them display functional differences. Consequently, biologics are always sold as a mixture of drug molecules with varying efficiency, and each batch of a biologic can be significantly different. Such batch-to-batch variation, and in particular the inability to systematically control it, does curtail the ability of the pharmaceutical industry to develop new biologics and in particular to generate competing off-patent products.
Glycan heterogeneity is not random, but rather controlled in a non-intuitive way by the organisation of a large number of biosynthetic enzymes in the Golgi apparatus. Through a BBSRC IB catalyst project and a BBSRC Doctoral Training Partnership-funded PhD project, we have recently developed a computational model that can efficiently describe the organisation of glycosylation enzymes in the Golgi. In a proof-of-principle theoretical study (funded by a BBSRC impact accelerator award) we were also able to show how to use this computational tool for predicting how enzyme levels would need to be altered to shift the set of glycan structures produced by a cell line. This computational tool could be used to inform companies how to alter their production cell lines to generate biologics with more beneficial glycan repertoires. The proposed study will validate the use of this modelling tool for predicting which glycan biosynthetic enzymes to overproduce in a biologic-producing cell line, and to verify that this intervention indeed shifts the glycan repertoire to the desired range of glycan structures.
The work will be carried out with an industrial partner to ensure that our validation is performed on examples with industrial relevance. Three biologics with increasing glycan complexity, which have been produced by the partner, will be used. Following modelling of the Golgi composition required for generating the glycans that our partner reports on its products, the model will be challenged with a more desirable set of glycans. The predicted change in enzymes will then be used to create synthetic DNA constructs in York, which can then be transferred into the production cells at the company. The alterations in enzyme levels as well as the new glycan repertoire will be investigated using protein and glycan analytical tools, including western blotting and mass spectrometry. The experimentally obtained results will be compared to those computationally predicted, to assess the potency of the computational modelling tool for the engineering of biologic glycan states. Once fully developed and validated, we intend to customize this computational tool for other pharmaceutical companies as well, and will therefore set up meetings with some of these during the course of the project to establish their specific needs.


10 25 50
Description Collaboration shifted from GSK to FDBK
An important lesson in the context of this translationally focussed project was the need to plan more flexibly with the industrial partner. Our project was planned with GSK as the partner, with the reasoning that it would not be logistically feasible for the PDRA to duplicate the work with two different partners within the timeframe allowed by the funding. At the application stage we decided to partner with GSK, given their record in academic collaborations, our own experience with them, and the size of the company allowing more flexibility in timing and resourcing the experiments planned at the company. However, between the time of application and the start of the project (start was somewhat delayed due to the Covid-19 pandemic) there was a change in GSK's operational practice. The consequences of this for our project only became apparent after a very long negotiation of the collaboration agreement resulted in a complete impasse, and forced us to look for an alternative partner. Fortunately, FDBK, who were also keen on seeing results from this project, and wrote a supporting letter for the original application, could be quickly involved, and with a no-cost extension we were able to finish the bulk of the work before the end of the grant (see also below).

Used computational modelling to establish enzyme overexpression strategies that cannot be deduced without modelling, while providing important new glycoform distributions
Using real glycan profile data of biologics produced in a bioreactor and in a currently used industrial production cell line under industrial conditions, we were able to calculate the changes needed in the levels of two key enzymes, Mgat1 and GalT, to shift the major glycan from the ungalactosylated to the fully galactosylated form. Interestingly, to achieve the desired glycan shift, enzyme level increases had to be very carefully tuned. When GalT levels were increased 20% less then calculated optimum in the simulation, this caused a dramatic change in the resulting glycan distribution, while a 20% more of the GalT level increase gave no significant change.

Computation shows that elimination of heterogeneity in industrial bioreactor runs is not feasible for most glycoforms
We tested a number of target scenarios in the computational modelling, altering different aspects of the glycan profile, thereby striving to achieve single glycoforms. We found that in all those cases the results, while the target glycan did represent the majority of the products, there were always other glycan present as well. Although these side-products represent a small proportion of the glycans, they highlight the fact that the Golgi apparatus is not built to generate homogeneous single glycoforms of proteins.

Partial completion of aims due to bioreactor experiments to confirm computational predictions not yet completed
The final confirmatory experiment of the project uses the production cell lines, after altering Mgat1 and GalT levels by stable transfection of plasmids expressing these two enzymes, to generate biologics in a bioreactor with the same conditions as previously. We did modify the cells in our laboratory and verified successful overexpression of the enzymes in question. The logistics of organising transfer of the cells to FDBK and finding a suitable time to use their bioreactors, meant though that the experiment is scheduled for after completion of the grant. Experimental confirmation of the computation will therefore be completed at the company after the grant is officially finished, hence the partial completion of the aims during this reporting period.

Other important outcomes detailed in further sections on Researchfish
A collaboration was started with a third company, Lonza Biologics, and we are currently in discussion with the company Ludger Ltd, to initiate a potential collaboration.
Method chapters were published in both a glycobiology themed and a cell biology themed book, to increase the reach of our methodology and thereby bolster the potential for translation.
The PDRA trained on the project was recruited to a senior research position in a biotech company.
Exploitation Route Once the final experiment is completed (see explanation above), we will negotiate with FDBK, how they can use these results.
We are also in discussions with Ludger Ltd about a collaboration that builds on results of these findings, pending the results of the final experiment.
Sectors Manufacturing, including Industrial Biotechology

Description Optimisation of CHO for Biotherapeutic Manufacture
Amount £3,608,961 (GBP)
Funding ID EP/V038095/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Public
Country United Kingdom
Start 10/2021 
End 09/2026
Description Improving therapteutic antibody production under industrial conditions by manipulating the Golgi apparatus 
Organisation Fujifilm
Department Fujifilm Diosynth Biotechnologies, UK
Country United Kingdom 
Sector Private 
PI Contribution Providing computational modelling and plasmid construction/know-how for the engineering of the Golgi in production cell lines.
Collaborator Contribution Bioreactor runs at different scales to test computational predictions, and using plasmids designed by academic partner. Quality analysis of produced antibody therapeutics to validate model predictions.
Impact none yet, work ongoing.
Start Year 2021
Description Lonza testing of patented cell mutant 
Organisation Lonza Group
Department Lonza Biologics
Country United States 
Sector Private 
PI Contribution We supplied Lonza Biologics with know-how and reagents to test a mutant cell line we identified to increase the production of biologics.
Collaborator Contribution Lonza Biologics generated mutants of their production cell line harbouring the mutations we identified. They are testing the effect of the mutation on the production of model biologics.
Impact We are awaiting results from Lonza's tests.
Start Year 2022
Description The disclosure relates to isolated eukaryotic cells, for example isolated mammalian cells, comprising one or more mutations in the Conserved Oligomeric Golgi 4 gene (COG4) wherein the eukaryotic cells expressing the mutated COG4 gene are associated with increased secretion of glycosylated proteins. 
IP Reference WO2022243286 
Protection Patent / Patent application
Year Protection Granted 2022
Licensed No
Impact A collaboration was set up with Lonza Biologics, in which they are currently testing the use of the patented mutation in their production cell lines.
Description Poster presentation at virtual Gordon Research Conference 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact PDRA on the grant presented a poster at an international conference
Year(s) Of Engagement Activity 2021