For calculation of the process time for both process types, we added a down time of 12 hours for media preparation, reactor setup, CIP, and SIP to the cultivation time. In chemostat cultivations with population collapse, we only considered the cultivation time before the decrease in cell density, which was the case with B3 <T7-GFP> after 48 hours.

As can be concluded from the productivity comparisons (Table 3), the fed-batch process with the T7 expression system and the model protein GFP were inferior to the chemostat process in terms of volumetric yield and STY. The reason for this can be attributed to the duration of the induction. In fed-batch cultivation, the production phase takes 19 hours. Chemostat cultivation can be extended to 48 hours because the volume in the reactor is kept constant.

As already mentioned, a longer chemostat process would not be possible under these circumstances because the process ultimately becomes unstable. However, a short-term chemostat cultivation of 48 hours could be interesting for easy-to-produce proteins. With 874 mg/L/h, the STY of B3<T7-GFP>wt in the chemostat tripled compared to the fed-batch cultivation. In the case of the host RNAP-dependent A1 expression system, the STY of 62.8 mg/L/h in fed-batch cultivation was relatively low compared to the STY of 286 mg/L/h of the conventional T7 expression system. In addition to the short production time of 19 hours, the A1 expression system was inherently weaker. However, the low titer can be compensated by continuous production. Due to the very long stable production time of 312 hours, the STY of the chemostat cultivation more than quadrupled compared to the fed-batch cultivation. Nevertheless, for the production of GFP, the conventional T7 expression system is still superior overall, in both fed-batch and chemostat mode, especially when considering short-term chemostat cultivation of the T7 wild-type strain (Table 3).

The situation is different with Fab. The extraordinary strength of the T7 expression system is a disadvantage for Fab production. In neither fed-batch nor chemostat mode, the T7 expression system was able to outperform the host RNAP-dependent A1 expression systems. Due to the mutation in lacI , the volumetric yield of BQ<A1-Fab>wt was reduced from 0.087 g/L to 0.035 g/L, as seen with derivative #E2.1. However, long-term cultivation without population collapse was possible. Due to the long production time of 120 h, the STY of the BQ<A1-Fab>wt was increased from 3.00 g/L/h in fed-batch mode to 4.33 g/L/h in chemostat mode for derivative #E2.1 when comparing the two strains to one another.

These results indicate that, in continuous E. coli bioprocesses, the choice of expression system depends on the recombinant protein to be produced. In the case of an easy-to-produce protein, such as GFP, the conventional T7 expression system is still the expression system of choice. The duration of the chemostat cultivation is limited by the inevitable population collapse, but this can be compensated by the extraordinary strength of the T7 system.

However, the moderately strong host RNAP-dependent A1 system is more suitable for the production of a challenging protein. As a result of the large metabolic load that is triggered, a reduction in the expression rate is advantageous. Thus, the low titer can not only be compensated, but even increased, based on the STY. As already mentioned, the time in which the actual fermentation occurs and recombinant protein is produced in fed-batch cultivation is relatively low. In addition, the cost of CIP and SIP must be included. Although these process steps must also be carried out in a long-term chemostat cultivation, the phase in which the actual production occurs is inevitably prolonged, which means that the CIP and SIP frequency can be reduced.

Nevertheless, only mutants that have adapted to long-term stability under production conditions can be cultivated in long-term chemostat mode. In wild-type strains, the population collapse and adaptive rescue during long-term chemostat cultivation does not allow long-term production. This phenomenon has been observed frequently in otherE. coli chemostat cultivations and prevents the direct use of non-adapted strains for this purpose.^{[5, 52, 53]}