Improving Downstream Process Related Manufacturability Based on Protein Engineering-A Feasibility Study

Abstract

While bioactivity and a favorable safety profile for biotherapeutics is of utmost importance, manufacturability is also worth of consideration to ease the manufacturing process. Manufacturability in the scientific literature is mostly related to stability of formulated drug substances, with limited focus on downstream process-related manufacturability, that is, how easily can a protein be purified. Process-related impurities or biological impurities like viruses and host cell proteins (HCP) are present in the harvest which have mostly acid isoelectric points and need to be removed to ensure patient safety. Therefore, during molecule design, the surface charge of the target molecule should preferably differ sufficiently from the surface charge of the impurities to enable an efficient purification strategy. In this feasibility study, we evaluated the possibility of improving manufacturability by adapting the surface charge of the target protein. We generated several variants of a GLP1-receptor-agonist-Fc-domain-FGF21-fusion protein and demonstrated proof of concept exemplarily for an anion exchange chromatography step which then can be operated at high pH values with maximal product recovery allowing removal of HCP and viruses. Altering the surface charge distribution of biotherapeutic proteins can thus be useful allowing for an efficient manufacturing process for removing HCP and viruses, thereby reducing manufacturing costs.

Keywords: Fc‐fusion protein; anion exchange chromatography; codon usage; manufacturability; protein surface charge.

FIGURE 1

Schematic depiction of the various protein variants. GLP1‐RA is an activating ligand for the GLP1‐receptor. Two GLP1‐RA variants, designated variant A and B, were used. L1: linker sequence between GLP1‐RA and the Fc‐domain. L2: linker sequence between the Fc‐domain and FGF21. The L2 sequence is unchanged throughout all constructs.

FIGURE 2

Sequences of the different L1‐linker and the C‐terminal charged tags. The protein linker sequence is highlighted in yellow. Stretches of more than six consecutive adenines are highlighted in red (with maximal eight adenines being present in the constructs). Changes for the adapted constructs are marked in cyan. In the adapted constructs the AAG codon was consequently used for all lysines.

FIGURE 3

Expression levels of the different protein variants as measured in the cleared supernatants. Remarkable is the loss of expression level for the variants with lysine‐residues inserted in the linker sequences, even after replacement of the AAA codons with AAG codons. Displayed are mean values +/− 3 standard deviations based on duplicate measurements.

FIGURE 4

SDS‐PAGE under non‐reducing and reducing conditions (A) and isoelectric focusing (B) of the initial variant A and altered versions of hereof. Expression was done in HEK 293 cells. The label of the protein does not consider which expression vector was used. Protein samples used were from elution of the protein A column.

FIGURE 5

Contour plots showing the dependence of (A) HCP clearance and (B) the yield in the AEX‐chromatography step as a function of pH and the load ratio (g target protein/L resin). For these experiments, initial variant B (isolectric point 6.62) was used. The HCP levels were measured in the flow‐through from the AEX‐chromatography step, using load with representative HCP levels. The step yield is given as the percentage of the recovered protein from the AEX‐chromatography step. The preferred operating conditions for the corresponding issues are marked with a green box. An elevated pH value is preferred for HCP removal whereas a lower pH is beneficial for the step yield.

FIGURE 6

Impact of the isoelectric point of the protein of interest on the yield for AEX‐chromatography. The different protein variants were arranged according to their calculated isoelectric points. The AEX‐chromatography step was either done at pH 6.0 (black bars) or pH 8.0 (grey bars). Note: Since the protein expression was very poor for the variants A‐L(K8), A‐L(K12), B‐L(K8), A‐L(K15), and B‐L(K12) the load ratio for these variants was only between 13 and 25 g/L (compared to 40 g/L for all other variants). Yet, achieved step yields were still high for some of these variants, despite underloading the resin.

FIGURE 7

Dose‐response curves from a luciferase gene expression assay with HEK293 cells overexpressing the human FGFR1 and β‐klotho proteins showing the biological activity of selected variants. Measured luminescence (in relative units) as a function of protein concentration as determined with the luciferase activity assay. The average of four biological replicas ± SEM is shown. Showing non‐modified FGF21 (green) vs. initial variants A and B as well as modified variants thereof. For readability purposes, dose‐response curves are separated into Figure 7 (A) and (B).