Viroceuticals are emerging tools for vaccination, genetic therapies, and novel cancer treatments. While effective, many have large costs associated with their manufacturing and delivery. We have conducted novel studies on viroceutical products and production via mass spectrometry-based lipidomics. Lentivirus (LV) technology descends from human immunodeficiency virus (HIV), whose lipid envelope has been previously measured and shown to have a direct impact on its transduction efficiency. Madin-Darby Canine Kidney (MDCK) cells are a widely used host cell line for influenza A virus (IAV) production as an alternative to egg-based production and exhibit a pronounced heterogeneity that can significantly affect viral yields. We have developed a rapid, robust, and sensitive untargeted lipidomics pipeline to elucidate novel biomolecular signatures that affect both LV and IAV biotherapeutic production pipelines. The impact of 48 hours of LV production on the lipidome of HEK 293T cells was measured and 151 lipids were identified, 84 of which had fold changes with FDR-corrected P<0.05 compared to HEK 293T treated with media. LV contained 102 lipids, half of which were determined to be unique LV virion lipids after subtracting the media lipidome. We also investigated two MDCK clones (C59 and C113) that differ in biochemical properties and viral production attributes and examined their lipid dynamics upon influenza A virus (IAV) infection at 24, 48, and 72 hours. C113 had elevated levels of lipid species across all major lipid classes with the exception of ether lipids compared to C59. IAV infection in C59 led to lipid droplet (LD) accumulation, elevated levels of ceramides and diacylglycerols as well as lysophospholipid and phospholipid depletion. IAV infection in C113 led to a decrease in LDs and lysophospholipids. Lipidomic analysis of the purified progeny virions from C59 and C113 yielded only subtle differences with an overall strong positive correlation in lipid profile (R2 = 0.77), suggesting similar lipid raft domain composition between clones.

The production of monoclonal antibodies in CHO cells requires both high titer and favorable product quality. Understanding and controlling the complex, high-dimensional dependencies of quality attributes on genetic and bioprocess design is challenging—particularly when data and experimental capacity are limited, which is typical for bioprocess development. These multivariate relationships vary across cell lines, antibodies, pools, and clones, and we often lack prior knowledge of how a specific antibody’s quality profile will respond to different design decisions. Further, improving product quality often comes at the expense of reduced titer, making this a challenging multi-objective optimization problem. The ideal approach to model-based product quality optimization depends on the type and quantity of available data and the constraints on experimental scope. We employed distinct modeling strategies for three specific cases: when initial data was sparse and multiple sequential experiments were feasible, when initial data was plentiful but few validation experiments were feasible, and when initial data was plentiful and inference alone was required with experimental validation deferred. All strategies predicted quality attributes and titer of different mAbs from process conditions and plasmid design features as inputs. First, we used Bayesian optimization in the context of sparse data and iterative experiments to optimize charge variant profiles by changing media and feed types and quantities. For four combinations of antibody and cell line, we achieved a 22–25% reduction in acidic species while maintaining ≥3 g/L pool titer. Second, we applied a hybrid model of metabolism and post-translational modifications to optimize media choices, feeding strategies, and temperature and pH setpoints in the context of plentiful data and a single validation experiment. This yielded a 25% reduction in acidic species with no detrimental effect on titer. Third, when preparing to extend process optimization from charge variants to glycan distributions with plentiful initial data to inform future experimental plans, we used multivariate analysis via PLS regression to identify process variables and clonal phenotypes associated with favorable charge variant and glycan distributions. These results demonstrate that high quality, high titer processes may be best achieved through application of modeling strategies appropriately selected based on data availability and experimental scope constraints.

Elevated plasma and tissues histamine concentrations can cause severe symptoms in mast cell activation syndrome, mastocytosis or anaphylaxis. Endogenous and recombinant human diamine oxidase (rhDAO) can rapidly and completely degrade histamine, and administration of rhDAO represents a promising new treatment approach for diseases with excess histamine release from activated mast cells. We recently generated heparin-binding motif mutants of rhDAO with considerably increased in vivo half-lives in rodents compared to the rapidly cleared wildtype protein. Herein we characterize the role of an evolutionary recently added glycosylation site asparagine 168 in the in vivo clearance and the influence of an unusually solvent accessible free cysteine 123 on the oligomerization of DAO. Mutation of the unpaired cysteine 123 strongly reduced oligomerization without influence on enzymatic DAO activity and in vivo clearance. Recombinant hDAO produced in ExpiCHO-S cells showed a 15-fold reduction in the percentage of glycans with terminal sialic acid at Asn168 compared with CHO-K1 cells. Capping with sialic acid was also strongly reduced at the other glycosylation sites. The high abundance of terminal mannose and N-acetylglucosamine residues in the four glycans expressed in ExpiCHO-S cells compared to CHO-K1 cells resulted in rapid in vivo clearance. Mutation of Asn168 or sialidase treatment also significantly increased clearance. Intact N-glycans at Asn168 seem to protect DAO from rapid clearance in rodents. Full processing of all glycoforms is critical for preserving the improved in vivo half-life characteristics of the rhDAO heparin-binding motif mutants.

Enzymatic degradation of polysorbates (PS) in biologics formulations are often caused by hydrolytic host cell proteins (HCPs) and can lead to particle formation and reduced shelf-life. Because PS-degrading HCPs are critical even in miniscule amounts, their identification, characterization and removal pose great challenges to the biopharmaceutical industry. To this day, the occurrence of difficult-to-remove HCPs in the final drug product is mainly tackled by molecule-tailored optimization of downstream purification steps. Here, we present an alternative: the genomic knockout (KO) of 9 critical HCPs within a CHO host cell line. In the first part of this work potentially critical HCPs were identified in industrially relevant drug formulations, followed by artificial overexpression, purification and detailed characterization. Hydrolytic HCPs were ranked based on expression level, occurrence rate in monoclonal antibody (mAb) products, PS degradation activity and cellular localization providing a broad dataset for further risk management. In part two, we sequentially knocked out these PS-degrading HCPs giving rise to a novel CHO host cell line variant with substantially reduced hydrolytic activity. By means of precise genome editing tools, we knocked out 9 critical HCPs including large carboxylesterase gene clusters of >1 megabase in size. Surprisingly, all 9 HCP KOs could be combined in a clonal CHO cell line without significantly compromising cell culture performance. This paves the way for a potentially novel platform host cell line with improved properties for future biologics production. Additional KOs of the pro-apoptotic genes Bax and Bak1 further improved cell growth and viability of the multi-KO cell line. Finally, the suitability as host cell line was verified in two cell line development campaigns, where our multi-KO host cell line was able to deliver high mAb titers while maintaining high product quality. Most importantly, hydrolytic activity and thus PS degradation could be dramatically reduced. The presented multi-hydrolase KO strategy delivers a universal solution for the elimination of PS-degrading HCPs in a CHO-based bioprocess. Currently, we are further investigating the impact of repeated hydrolase gene editing on the geno- and phenotype of the multi-KO cell line, generating valuable insights for future multi-KO endeavors of other detrimental host genes and potential bioprocess optimizations.