Experimental and in silico modelling analyses of the gene expression pathway for recombinant antibody and by-product production in NS0 cell lines

Abstract

Monoclonal antibodies are commercially important, high value biotherapeutic drugs used in the treatment of a variety of diseases. These complex molecules consist of two heavy chain and two light chain polypeptides covalently linked by disulphide bonds. They are usually expressed as recombinant proteins from cultured mammalian cells, which are capable of correctly modifying, folding and assembling the polypeptide chains into the native quaternary structure. Such recombinant cell lines often vary in the amounts of product produced and in the heterogeneity of the secreted products. The biological mechanisms of this variation are not fully defined. Here we have utilised experimental and modelling strategies to characterise and define the biology underpinning product heterogeneity in cell lines exhibiting varying antibody expression levels, and then experimentally validated these models. In undertaking these studies we applied and validated biochemical (rate-constant based) and engineering (nonlinear) models of antibody expression to experimental data from four NS0 cell lines with different IgG4 secretion rates. The models predict that export of the full antibody and its fragments are intrinsically linked, and cannot therefore be manipulated individually at the level of the secretory machinery. Instead, the models highlight strategies for the manipulation at the precursor species level to increase recombinant protein yields in both high and low producing cell lines. The models also highlight cell line specific limitations in the antibody expression pathway.

Figure 1. Antibody-related species detected by western analysis in the different cell lines.

A. Example growth curve of cell line 2N2. A sampling point was chosen below the maximum viable cell concentration, where cell-specific productivity was observed to be stable. (Full growth curves are available in File S1) B. The different species that could be observed in intracellular or supernatant samples were identified by western blotting with a polyclonal antibody able to detect both heavy chain and light chain species. The species are (from top) assembled antibody (LHHL), half antibody (LH), single heavy chain (H), light chain dimer (LL) and single light chain (L). For clarity of species observation intracellular and supernatant blots are representative and not of the same exposure time, quantitative data was obtained by western blotting alongside internal standards and is given in Table 1. Full data tables are provided in File S5.

Figure 2. Comparison of productivity in the current model system with other reported expression systems.

Secretion rates observed in the cell lines used in this study are compared to a number of published rates for various eukaryotic expression systems , , . L (light chain), LL (light chain dimer), mAb (assembled monoclonal antibody).

Figure 3. Origin of the HL species.

A. Treatment of 2N2 samples with oxidising agents does not alter the ratio of half-antibody to full mAb. BMDB (1,4-bismaleimidyl-2,3-dihydroxybutane), NEM (N-ethylmaleimide), LHHL (assembled antibody), LH (half antibody species), LL (light chain dimer), L (light Chain). B. In cell culture supernatants the ratio of half- to full antibody does not change upon prolonged incubation at room temperature. Both analyses indicate that the half antibody species does not result from decay of the full antibody.

Figure 4. The fit of our models to the observed steady-state values determined experimentally.

Different antibody assembly models were fitted to the different observed steady-state data given in Table 1. Panels on the left-hand side (A and C) show the assembly routes for the two different models, with productive reactions (those leading to the formation of full antibody) displayed in red. The panels on the right-hand side (B and D) indicate the difference between the best-fitting parameter set in the biochemical models and the observed parameter values. The two models, as well as a model combining the two assembly pathways, can all reproduce the experimentally observed data for all four cell-lines with high similarity. Similar results were obtained for the engineering models. H.mRNA (heavy chain mRNA), L.mRNA (light chain mRNA), LHHL (assembled antibody), LH (half antibody species), H (heavy chain), HH (heavy chain dimer), HHL (heavy chain dimer +1 light chain), LL (light chain dimer), L (light Chain), mAb (assembled secreted antibody), qmAb (mAb secretion rate), qHL (HL secretion rate), qLL (LL secretion rate), qL (L secretion rate).

Figure 5. Comparison of pathway fluxes as a function of productivity.

The molecular flux through individual reactions was obtained from the biochemical models running at steady state. The colour of the arrows connecting the model species indicates the molecular flux from one species to the next in molecules per time unit. The left hand panel show absolute pathway fluxes for three cell lines. The right hand panel shows relative fluxes, normalised to the highest individual reaction flux observed for each cell line. The diagrams shown are for the “HL” assembly pathway. H.mRNA (heavy chain mRNA), L.mRNA (light chain mRNA), LHHL (assembled antibody), LH (half antibody species), H (heavy chain), HH (heavy chain dimer), HHL (heavy chain dimer + 1 light chain), LL (light chain dimer), L (light Chain), mAb (assembled secreted antibody).

Figure 6. Clustering analysis of cell-line specific parameters.

Parameters were obtained from the biochemical models at steady state for each of the four cell lines analysed. Parameters that occur in similar patterns across the different cell lines are closely clustered in this analysis. Statistically significant clusters containing the qmAb parameter are indicated in red. H.mRNA (heavy chain mRNA), L.mRNA (light chain mRNA), LHHL (assembled antibody), LH (half antibody species), H (heavy chain), HH (heavy chain dimer), HHL (heavy chain dimer +1 light chain), LL (light chain dimer), L (light Chain), mAb (assembled secreted antibody).

Figure 7. Predicted intervention points for the reduction of by-product export.

The predicted effect of variations in increased heavy- (left panel) or light-chain (right panel) expression is illustrated. While variations in light chain expression predominantly affect secretion of the L and LL by-product species, variation in heavy chain expression selectively affects the expression of fully formed antibody. qmAb (assembled antibody secretion rate), qHL (half antibody secretion rate), qLL (light chain dimer secretion rate), qL (light chain secretion rate).