Based on the higher cell densities, BL21(DE3) fermentation broths had greater turbidity. All BL21(DE3) clones had high reduction of turbidity, at least 94%, except B<oFabx>, which was also the one that had the highest impact on growth. Notably, the highest reduction was achieved for B<dFabx>, which did not express any Fab and grew unaffected. In general, the HMS174(DE3) clones had lower centrifugation efficiency, reflecting the higher impact on cell growth.

LC-ESI-MS was used to evaluate the Fab products. We determined the molecular mass of both intracellular and extracellular Fab. All of FTN2 (100%) was the correctly folded full Fab both intracellularly and extracellularly, except for B<dFTN2>, in which a minimal amount (~1%) presented a mass shift of +148 Da.

In the case of BIWA4, all clones produced only the correct Fab, except for H<dBIWA4>, which had 70% with +2 Da in the extracellular fraction, and B<oBIWA4>, which had a mass shift of -178 Da for approximately 5% of the intracellularly retained Fabs. BIBH1 had a more pronounced mass shift, as the ratio of correctly processed Fab varied from 40% to 75% for both fractions. The observed mass shift was -17 Da for all clones. For B<dBIWA4>, the intracellular fraction had a mixture of -16 and +100 Da. Due to the low amounts that were gained for Fabx, there was only one MS result, which indicate a ratio of 30% -17 Da in the extracellular fraction of B<oFabx>. Wide variations were observed in terms of the ratio of free LC and LC fragments of total Fab. The ratio ranged from 0% for B<oBIWA4> up to 74% for H<oFTN2>; nevertheless, all combinations showed similar ratios in intracellular and extracellular fractions (Table 2). In general, the ranking for these impurities was FTN2>BIBH1>BIWA4, independent of the investigated host and leader combination. Due to the very low expression levels of Fabx, this point was only investigated for one exemplary combination.

In this study, we investigated how the combination of host strain, expressed Fab, and translocation pathway interact with and influence process efficiency. One of the main results was that the selected variations and their combined application led to scattered product yields and cell growth, as well as significant differences in DSP operability. The greatest influence was from the Fab molecule that was expressed, though all four candidates had the same backbone and only the CDA region varied. The highest product concentrations for all combinations were measured for FTN2, followed by BIWA4, BIBH1, and Fabx, with few exceptions. For all Fab-leader combinations, HMS174(DE3) was superior to its BL21(DE3) counterparts in terms of specific Fab titers. However, the BL21(DE3) host responded less strongly to the expression of the Fabs, and cells were generally more vital with better growth behavior. This was also confirmed by the DSP investigations, in which considerably less DNA was quantified in the supernatant, resulting in lower viscosity and better clearance (Fig. 3A and B). Considering the values from Table 1, linearity between clearance efficiency, viscosity, and DNA content is clearly apparent if the values for B<oFabx> are left aside. The least influence on process and yield was observed for the leader sequence. No general overriding behavior could be identified, only that OmpA^SS led to slightly higher expression levels in both strains in most cases. Taken together with DSP operability, the chosen procedure in the USP, with early induction and a long production phase independent of strain and Fab, resulted in very unfavorable distribution ratios of intracellular and extracellular Fabs, in the range of 60:40 to 40:60. One of the fractions would have to be omitted in DSP, which is not acceptable in view of the generally low product titers. Although high extracellular fractions were observed in both strains, there was a significant difference in extracellular DNA content, which was much lower in BL21(DE3), indicating increased release of Fab into the supernatant independent of cell lysis (Fig. 2A and Fig. 3C and D). This finding is clearly a positive characteristic attributable to the strain with high impact on process design and efficiency.

The traditional approach for process development focused on USP yield would favor H<oFab> clones as expression hosts due to the higher specific total Fab titers. However, the high cell lysis rates, high extracellular DNA concentrations, and high viscosities would negatively affect the DSP performance and pose great challenges for process development. DSP would be faced with cell broths containing high levels of extracellular product fraction and increased DNA and endotoxin levels, in combination with low centrifugation efficiencies due to the elevated viscosity. Therefore, a classical purification process for intracellular products would not be economic and would lead to high product loss. This could be circumvented by homogenizing the cell broth before harvesting. However, a drawback would be that DSP would be confronted with a broth containing media residue and extracellular metabolites that are normally removed by harvesting the cells. These impurities would need to be removed during subsequent unit operations.

These results highlight the importance of a holistic approach for process optimization, including integrating the upstream and downstream aspects. The parameters selected for the evaluation of DSP performance were found to be a good choice. These parameters allow for a comprehensive process evaluation and provide relevant information for integrated process optimization. In our specific case, due to the observed growth and product leakage kinetics, shifting the induction and harvest time points could be a starting point for optimization. Later induction would shorten the expression period and, therefore, could reduce cell lysis and product leakage.

The MS analysis showed that, in almost all cases, the correct regular Fab was expressed and the extracellular fraction was not subjected to changes. Only BIBH1 presented a significant mass shift caused by the cyclization of N-terminal glutamine to pyroglutamate, which is a well-described phenomenon in the literature for the production of recombinant antibodies [38, 39]. This represents another issue affecting DSP operability, as separation of product variants is challenging due to their small differences in charge. Furthermore, in all combinations except B<oBIWA4>, remarkable amounts of free LC and LC-fragments were found, which can impede DSP by limiting Protein L column capacities, as these impurities will compete with the product for binding [34].

Thus, we showed that strong impairment of growth by Fab expression leads to a high extracellular product fraction and contamination, resulting in high viscosity and low centrifugation efficiency. The chosen host and expressed Fab product exerted the main influences on growth, titer, and DSP operability. In particular, HMS174(DE3) clones had higher extracellular product fractions, DNA and endotoxin contents, and viscosities, and lower final CDM.

Our results indicate significant interdependencies between the two disciplines, and that decisions made at the USP level are especially important for the performance of the first steps of DSP. Therefore, the classical yield-oriented USP development approach would create major challenges for DSP development, as higher extracellular impurities, product localization, and poor manufacturability would lead to low overall product yields and high manufacturing costs. This underlines the need for a holistic and integrated process development approach to establish reliable and economical production processes, and to shorten the overall time to market. Our results provide the basis for further research into integrated and holistic process development. By monitoring the impact on growth and product localization caused by changes in the USP strategy, we are now able to better predict fluctuations in DSP.