1 Introduction
Fed-batch bioprocessing is the most common cultivation method in
industrial microbial production of biopharmaceuticals. This batch-wise
process essentially includes the repetitive steps of media preparation
and reactor setup, fermentation, and subsequent cleaning in place (CIP)
and sterilization in place (SIP). In terms of time, the actual
fermentation, and particularly the production phase of the recombinant
protein, is relatively short. As a result, continuous production becomes
more and more interesting due to the greatest possible space time yields
and optimal use of the installed
assets.[1]In such chemostat cultivations, cells are maintained in a steady-state
growth environment by adding fresh medium to the reactor at constant
flow. Simultaneously, the cell suspension, and thus the recombinant
protein, is removed at the same rate.[2] The
growth rate (μ) can be specified depending on the dilution rate (D).
Through this process, stable volumetric productivity and high space-time
yield can be achieved [3]. In contrast to
fed-batch fermentation, the average residence time of a producing cell
is always the same, which can be advantageous in terms of product
quality. Examples of microbial continuous processes for the production
of recombinant proteins have already been described in the
literature.[4-6]
For industrial microbial production of recombinant proteins, theEscherichia coli strain BL21(DE3) is often used due to low
acetate formation and high production rates resulting from the
integrated T7 RNA polymerase (RNAP).[7, 8] In
combination with a pET series plasmid, which harbors the gene of
interest (GOI) under control of the T7 promoter, extraordinarily high
expression rates can be achieved after induction with the
non-metabolizable lactose analogue isopropyl-D-1-thiogalactopyranoside
(IPTG).[9]
Genetic heterogeneity caused by metabolic burden and toxicity can be
problematic at all industrial scales, especially for challenging
proteins.[10-12]In bacterial production processes,
challenging proteins impose adverse effects on host metabolism, even at
low concentrations. Escape variants, which have a growth advantage due
to mutations or plasmid loss, can overgrow high-performing producer
cells, reducing the overall product yield.[13-15]Therefore, longer production phases, or even continuous production mode,
are hardly feasible in such E. coli expression systems.
To obtain stable, high-yield, and predictable E. coli production
hosts, engineered producer strains must focus on reducing metabolic load
and genetic escape. Attempts to reduce process instability caused by
metabolic burden have been made on both the genetic and bioprocessing
levels. The metabolic burden and increased selection pressure can be
reduced by decoupling growth and production in cascading chemostat
cultivations with two bioreactors.[6] Genetic
escape can be reduced by removing insertion elements (IS), and by
deleting recA or error-prone DNA polymerase
genes.[16] Another promising strategy uses the
directed evolution approach combined with fluorescence-activated cell
sorting to select cells with a lower plasmid mutation
rate.[17]
In previous studies, we showed that genomic integration of the GOI under
the control of the strong T7 promoter reduces the metabolic burden
because plasmid-mediated metabolic load is eliminated, and strong
expression has been shown from even a single copy of the GOI. However,
after approximately seven doublings under production conditions,
mutations in the T7RNAP gene lead to a faster growing non-producing cell
population. This phenomenon can be excluded in systems using the host
RNAP-specific A1 promoter because full functionality of the host RNAP is
required for cell growth.[18]
In combination with a directed evolution approach, this would allow
characterization of mutated production hosts, which have adapted
themselves to the metabolic load triggered by recombinant protein
production and could potentially enable continuous protein production.
Moreover, a directed evolution approach can circumvent the complexity of
the process of biogenesis and its adverse effects on the host
cell.[19-21]
For example, in the study by Walker et al. [22],
derivatives of BL21(DE3) were adapted by directed evolution to produce
cell membrane proteins that are toxic to host cells. These strains are
currently widely used for production of a variety of membrane proteins
and toxic proteins.
In the present study, we performed iterated carbon-limited
fed-batch-like microbioreactor cultivations under production conditions.
The goal was to investigate whether and how long-term metabolic load
triggered by the production of recombinant proteins influences the
characteristics of mutations occurring in different genome-integratedE. coli production systems.[23, 24] We
compared the host RNAP-dependent BL21Q A1 expression
system (BQ<A1>) [18] with
the T7-based BL21(DE3) expression system
(B3<T7>). To study mutation characteristics based
on the protein of interest (POI), we used the easy-to-produce protein
GFPmut3.1 and the challenging protein Fab fragment dFTN2 as model
proteins. To investigate the underlying mutations in the different
clones in more detail, we performed comparative whole genome sequencing
analyses. We also performed long-term chemostat cultivations in
lab-scale bioreactors with the above-mentioned clones and additional
robust production strains obtained by the directed evolution approach.