Triggering outer membrane leakiness of a novel E. coli strain for recombinant protein production

Recombinant proteins in Escherichia coli are expressed inside the cell. With the growing interest in continuous cultivation, secretion of product to the medium is not only a benet, but a necessity in future bioprocessing. In this study, we present a novel E. coli production host for growth decoupled recombinant protein production that can leak up to 90% of recombinant protein to the extracellular space. We investigated the effects of the process parameters temperature and specic glucose uptake rate on physiology, productivity, lysis and leakiness. Two model proteins were used, Protein A and a VHH single-domain antibody, and performance was compared to the industrial standard strain BL21(DE3). We show that inducible growth repression in the novel E. coli strain enGenes-X-press, the effect of the metabolic burden on host can be greatly reduced compared to BL21(DE3). Furthermore, in both strains, increasing temperature and specic substrate enhanced productivity and leakiness. Using the enGenes-X-press strain, extracellular Protein A and VHH titer reached up to 349 mg/g and 19.6 mg/g, respectively, comprising between 80 and 90% of total soluble product, while keeping cell lysis to a minimum. BL21(DE3) leaked 198 mg/g and 3.9 mg/g of Protein A and VHH to the medium, accounting for only 56% and 34% of total soluble product, respectively.

allows for more exible change between high-and low-demand products, less down time of unit operations, less process variability and reduced degradation of the target molecule due to less residence time in the reactor [9][10][11]. Integrated continuous operation has been demonstrated in animal cell [12,13] and even E. coli processes [14]. However, in the latter, the product is typically located inside the cell, which requires cell disruption in downstream processing, leading to release of unwanted host cell proteins and other contaminants, like lipids and DNA [15]. If the target protein is produced as insoluble inclusion bodies (IBs), additional IB processing is needed. Thus, continuous manufacturing in E. coli is still hard to realize to date. One strategy to bring E. coli to continuous manufacturing is soluble, extracellular protein production.
Secretion of recombinant protein to the medium furthermore enhances solubility, stability and biological activity of the product [16]. This can be achieved either by one-step-secretion (directly from the cytoplasm to the extracellular space) via the T1SS or T3SS system, or by two-step-secretion: In the rst step, the protein is directed through the inner membrane via the Sec-or Tat-pathway. In the second step, the outer membrane (OM) is made permeable, or "leaky", to release the product to the medium [3]. Numerous studies on how to increase leakiness during cultivation exist and several reviews cover this research in detail [2,3,17].
One approach to increase leakiness is chemical permeabilization by addition of media supplements, like Triton-X, Tween or EDTA. However, those additives usually have detrimental effects on the viability of the cells and might harm the product [3]. Another approach is the generation of leaky E. coli mutants. Many expression systems that show permanently high leakiness have been engineered to date. Their outer membrane structure is usually altered by mutations in cell envelope genes and signal peptides are optimized for higher translocation e ciency [3,18,19]. Wacker Biotech GmbH developed a commercially available strain that can secrete several grams of different products per liter [20,21]. However, information on process design is not available for this proprietary strain. Different process parameters during cultivation, like temperature, speci c growth rate, pH and aeration, have been proposed to have an effect on the uidity, composition and therefore leakiness of the OM [22].
Especially temperature and growth rate were shown to have a signi cant effect on leakiness during cultivation [23][24][25][26][27][28]. Strategies based on these parameters are easy to implement, yet consistent understanding of the (possibly combined) effect of temperature and growth rate on product release, physiology and productivity is still missing.
In this study, we investigated the in uence of the process parameters cultivation temperature and speci c glucose uptake rate on OM leakiness of a novel E. coli expression host. The X-press strain is a proprietary expression technology developed by enGenes Biotech GmbH [29]. It carries a genomically integrated sequence coding for the bacteriophage-derived RNA polymerase inhibitor Gp2 under control of the araB promoter. This protein from the T7 phage inhibits the host RNA polymerase, while the T7 RNA polymerase stays unaffected. Thus, upon induction with L-arabinose, host mRNA levels and cell proliferation are reduced, while IPTG-induced target protein expression is enhanced. This approach to decouple growth from recombinant protein production has already been shown to increase speci c yield and product quality [29,30]. In previous, unpublished experiments, the X-press strain showed high tendency to leak periplasmic protein to the medium. Therefore, in the present research, we further investigated its response to the process parameters temperature and speci c substrate uptake rate in fed-batch cultivations under the controlled conditions of a benchtop bioreactor. We performed a screening Design of Experiments (DoE) to nd the adequate parameter space for enhancing leakiness while maintaining high productivity and viability. We compared the X-press strain to the industrial standard strain BL21(DE3), using two industrially relevant model proteins: Protein A (SpA) from Staphylococcus aureus and a VHH single domain antibody (VHH). The processes were analyzed with respect to physiology, productivity, lysis and leakiness. With this holistic approach, we aimed at 1) characterization of a novel E. coli expression host for growth decoupled protein secretion and 2) nding the parameter space that allows tight control of leakiness and productivity for possible application in continuous manufacturing.

Results And Discussion
In this study, we investigated the in uence of temperature and speci c glucose uptake rate on physiology, productivity, lysis and leakiness of two different E. coli chassis strains. These process parameters are important factors in bioprocess development and have been shown to have an impact on leakiness before [23][24][25][26][27][28]. Thus, understanding the impact of temperature and speci c glucose uptake rate on OM permeability is necessary for the successful control of product location. As a benchmark strain we chose E. coli BL21(DE3), since it is the most widely used E. coli strain for recombinant protein production due to its fast growth, low acetate production, diminished protease content and its powerful T7 expression system [31,32]. The responses of physiology, productivity, lysis and leakiness to three combinations of cultivation temperature and q S,0 (30/0.25, 25/0.13, 35/0.13) were rst studied using the model protein SpA. The most favorable production conditions were then tested again with the second model protein VHH.

Impact Of Process Parameters On SpA Production In BL21(DE3)
It has been long known that heterologous expression in plasmid-based E. coli systems has a grave impact on cell physiology, widely known as metabolic burden [33]. This burden is often associated with a decrease in growth rate or, ultimately, cell lysis [34,35]. We assessed the impact of the selected process parameters on cell physiology by measuring the biomass yield, Y X/S , and lysis. Yield reduction varied greatly between different cultivation conditions (Fig. 1). It increased with induction temperature, so that at 25 °C, biomass yield of BL21(DE3) was least affected, while at 35 °C, growth was fully arrested. We hypothesize, that this behavior stems from an increase in target gene transcript levels competing with host mRNA at elevated temperatures, which might also be re ected in SpA productivity. Indeed, at low temperature and q S,0 (25/0.13), biomass speci c, soluble SpA titer was lowest at 113 ± 7 mg/g ( Fig. 2A) after 12 h. Raising the temperature at low q S,0 from 25 to 35 °C drove SpA expression, so that the total titer after 12 h more than doubled to 240 ± 9 mg/g. It has been shown that the overall protein synthesis rate as well as plasmid replication are dependent on temperature [36,37]. In our experiment, 35 °C induction temperature might have resulted in higher plasmid copy number and concomitant high levels of target gene transcripts, competing for ribosomes with native mRNA, thus increasing recombinant protein expression and decreasing growth rate. At 25 °C, this reaction was possibly shifted in favor of host mRNA due to lower levels of plasmids, which could explain low productivity and little metabolic burden. The highest speci c SpA titer was 351 ± 17 mg/g 12 h after induction in cultivation 30/0.25, which was expected since more carbon was available for product formation. However, yield reduction was less than at 35 °C, indicating that a decrease in growth rate was not only mediated by foreign protein content, but as hypothesized, by the underlying temperaturedependent mechanisms at transcript level.
No lysis was detected at all conditions during SpA production with BL21(DE3). Leakiness in the reference strain reached 50-60% in both cultivations 30/0.25 and 35/0.13 after 12 h, accounting for 198 ± 9 mg/g and 133 ± 3 mg/g extracellular SpA titer, respectively. However, protein secretion commenced only between 4 and 8 hours (Fig. 2). In cultivation 25/0.13, no product at all was leaked to the medium. Those results are congruent with previous ndings, that state OM permeability increases with higher temperature and growth rates, respectively [23][24][25][26][27]. The similarity in leakiness between conditions 30/0.25 and 35/0.13 indicate a combined effect of temperature and substrate uptake rate: increasing temperature by 5 °C and feed rate by 0.12 g/g/h had the same effect on leakiness as an increase of 10 °C with no change in feed rate.
These results demonstrate, that controlling leakiness via temperature and speci c substrate uptake rate is possible for BL21(DE3) within the investigated parameter space. However, these process parameters have a grave impact on productivity as well, thus product location cannot be uncoupled from productivity. For BL21(DE3), this is a double-edged sword: increasing temperature and speci c glucose uptake rate greatly enhanced SpA titer, but the cells did not leak more than 60% of product to the medium. Hence, capturing the target protein from the cells or from the medium, respectively, would result in large product losses in both scenarios and therefore BL21(DE3) does not constitute an effective host for extracellular protein production. Impact of process parameters on SpA production in enGenes X-press By uncoupling growth from recombinant protein production via co-expression of Gp2, the novel enGenes X-press production host can achieve high speci c product yields and is suitable for expression of toxic proteins [29,30]. In preliminary studies, we observed that the X-press strain leaked up to 90% of periplasmic protein to the supernatant. In this study, we characterized the strain by investigating the response of physiology, productivity, lysis and leakiness to changes in temperature and speci c substrate uptake rate, in the same design space as the reference strain BL21(DE3).
Additionally, to investigate the growth repression induced by expression of Gp2 without metabolic burden from recombinant product, we cultivated the X-press strain without an exogenous plasmid and solely inducing Gp2 expression by addition of L-arabinose. After induction, the biomass yield was reduced up to half from 0.48 in the uninduced state to levels between 0.24 and 0.27, remaining almost constant throughout the cultivation (Fig. 3). We assumed that any additional reduction in biomass yield is caused by the metabolic burden of heterologous gene expression. During production of SpA in cultivation 25/0.13, the reduction of biomass yield was similar to the "basal" reduction by Gp2 expression, thus the metabolic load of SpA expression had little effect on growth. An additional reduction was observed at higher temperature and q S,0 . In both cultivations 30/0.25 and 35/0.13, biomass yield decreased throughout the cultivation to values between 0.03 and 0.1. Hence, the metabolic load of recombinant product expression still affected growth of the X-press strain, but it was largely mitigated by induced growth repression, so that variability in growth across different cultivation conditions was greatly reduced compared to the BL21(DE3) strain.
The effect of temperature and substrate uptake rate on total soluble SpA productivity of the X-press strain was similar to the BL21(DE3) reference strain. Generally, higher temperature and substrate uptake rate drove SpA production (Fig. 4). At low temperature and q S,0 , productivity was lowest, while increasing the temperature to 35 °C boosted nal productivity more than 2.5-fold, from 123 ± 4 mg/g to 314 ± 6 mg/g after 12 h of induction. As in BL21(DE3), this is likely due to higher protein translation and plasmid replication rates. The highest amount of total soluble SpA, 387 ± 12 mg/g after 12 h, was produced in cultivation 30/0.25, where more carbon was available for product formation.
The X-press strain did not lyse at low q S,0 , however, in cultivation 30/0.25, lysis increased towards the end of fermentation, so that 7% of cells were lysed after 12 h (Fig. 5). Thus, in the later stages of this cultivation, the amount of leaked protein is biased by product release by lysis.
Nonetheless, the X-press strain showed higher overall leakiness than BL21(DE3), though the triggering mechanisms remained similar. Increasing q S,0 and temperature individually both triggered leakiness, while simultaneous increase resulted in an ampli ed effect. Between 80 and 90% of SpA was found in the supernatant after 12 h at all conditions, except in cultivation 25/0.13, which only yielded 29% of extracellular product (Fig. 4). While lysis was low in cultivation 30/0.25, extracellular SpA reached 266 mg/g after 8 h, comprising 81% of total product.
From the results obtained in the SpA fermentations we deducted different approaches to extracellular production in the X-press strain: (1) Low q S,0 and high temperature are bene cial for maintaining a viable culture and boosting productivity and leakiness over extended fermentation times; (2) moderately increasing temperature and q S,0 rapidly enhances leakiness and productivity, but high viability might not be sustained for long fermentation times.

Production Of VHH In BL21(DE3) And X-press
The cultivation conditions that resulted in the highest productivity in each strain (T = 30 °C, q S,0 = 0.25 g/g/h) were repeated with a second model protein, a VHH single domain antibody, and fermentations were assessed after 14 h induction time. The results are summarized in Table 1. The biomass growth in both strains was less affected compared to the corresponding SpA cultivations. In the X-press strain, the biomass yield reduction was close to the "basal" growth repression by Gp2 induction. In BL21(DE3), biomass yield was reduced by less than 0.1 C-mol/C-mol. This was likely due to the much lower amount of produced recombinant product compared to SpA and, as a result, a lower metabolic load [34,35]. Total productivity of soluble VHH was greatly enhanced in the X-press strain compared to the reference strain. Although inclusion body formation was detected in both strains (Additional le 1), the induced growth repression and enhanced secretion ability of the X-press strain seemed to have a bene cial effect on solubility of VHH, which is more di cult to fold due to its disul de bridges [8,38]. Also the amount of secreted protein was greatly improved in the X-press strain and was comparable to the SpA cultivations, although lysis was negligible during VHH production. Overall, the cultivations with the second model protein con rmed that the selected settings of process parameters (T = 30 °C, q S,0 = 0.25 g/g/h) lead to e cient product secretion in the X-press strain, while product location in BL21(DE3) is une ciently partitioned both inside and outside the cell. The issue of insoluble product aggregation might be addressed in further development, for instance by inducer titration or similar approaches to ne tune expression levels and thus further enhance soluble productivity.

Cause Of Enhanced Leakiness In enGenes X-press
A plethora of leaky mutants have been described in literature before, and the increased secretion across the OM is most often due to mutations in genes related to membrane proteins, lipopolysaccharides or the peptidoglycan layer [3,18,19]. These genes were not manipulated during the construction of the enGenes X-press host. Thus, the question is raised, how Gp2 expression can have an impact on membrane properties. Clearly, inhibiting the host RNA-polymerase, a most central enzyme in cell proliferation, can disturb practically any metabolic pathway. So far, the chain of causality between Gp2 expression and increased membrane permeability remains obscure. At the time of preparing this manuscript, the effects of Gp2 at the transcriptome level were being investigated.

Conclusion
In this study, we present an approach to trigger periplasmic protein release in a novel E. coli strain solely via cultivation temperature and substrate uptake rate. We narrowed down the design space, in which extracellular protein production is favored without sacri cing viability: cultivation temperatures between 30 and 35 °C and q S,0 between 0.13 and 0.25 g/g/h enhance both leakiness and productivity while keeping lysis to a minimum. The process parameters both individually and interactively affected total product titer and leakiness in a positive manner in both investigated expression hosts. The process understanding gained in these fed-batch studies could ultimately be transferred to continuous cultivation, where steady-state would allow for tighter control of leakiness and productivity [39]. By inducible growth repression, the novel expression host enGenes X-press showed less susceptibility to the metabolic burden of recombinant protein production and thus allows for tighter process control due to reduced variability across different process conditions. Lastly, we showed that the X-press strain can achieve high titers of different classes of recombinant protein and leaks 80-90% of all soluble product. Therefore, this strain is a promising candidate for extracellular protein production in current fed-batch applications or for future continuous manufacturing.

Materials And Methods
Strains Two E. coli strains were used in this study: the X-press strain, a BL21(DE3) derivate patented by enGenes Biotech GmbH [29], and a state-of-the-art BL21(DE3) strain (New England Biolabs, Ipswich, MA). The Xpress strain carries a genomically integrated sequence coding for Gp2, a protein repressing cell growth by inhibition of RNA polymerase. Its expression is induced by L-arabinose, which cannot be degraded by Xpress due to a knockout of the araABCD operon. For determination of cell growth repression solely induced by Gp2 expression, the plasmid free X-press strain was used. For recombinant protein production, both strains were transformed with a pET30a plasmid containing a cer sequence for enhanced plasmid stability [40] and a kanamycin resistance marker. The plasmid carried the gene coding for 1) the IgGbinding domains of Protein A from Staphylococcus aureus (SpA) with the pelB signal sequence or 2) the anti-TNFRI VHH single domain antibody DOM101 with the ompA signal sequence [41]. Both proteins were His-tagged at the C-terminus. Protein sequences are listed in Additional le 2.

Media
The semi-de ned medium for the pre-culture contained 9.00 g/L glucose, 3.00 g/L KH 2 PO 4 , 4.58 g/L For bioreactor cultivations, de ned minimal media according to DeLisa et al. [42] was used, with glucose as carbon source. The initial glucose concentration was 20 g/L and the substrate feed had a glucose concentration of 400 g/L.

Bioreactor Cultivations
For the pre-culture, 500 mL of semi-de ned medium were inoculated with a frozen stock in a 2500 mL High Yield shake ask and incubated for 16 h at 37 °C and 230 rpm in an Infors HR Multitron incubator (Infors, Bottmingen, Switzerland).
The plasmid free X-press strain was cultivated in a stainless steel bioreactor with a working volume of 10 L (Biostat Cplus, Sartorius, Göttingen, Germany). The batch volume was 5 L. The culture broth was supplied with a mixture of air and pure oxygen at 10 L/min and stirred constantly at 1200 rpm. Dissolved oxygen (DO) was monitored using a uorescence electrode (Visiferm DO120, Hamilton, Reno, NV, USA) and kept above 35% by adjusting the amount of added pure oxygen. pH was monitored with an Easyferm electrode (Hamilton) and kept constant at 7.00 via addition of NH 4 OH (12.5%). The temperature was controlled with the built-in heat jacket and kept at 37 °C, except during induction (described below). CO 2 and O 2 in the off-gas were analyzed with an off-gas analyzer (M. Müller AG, Switzerland).
The recombinant protein production processes were carried out in a DASGIP parallel reactor system (Eppendorf, Hamburg, Germany) with four vessels containing 2 L working volume, aerated at 2 L/min. The batch volume was 1 L. Gas mixing and control of DO, pH and temperature (via heat blanket and cooling nger) were done analogously to the cultivations in the stainless steel bioreactor described above. Off-gas composition was analyzed with DASGIP GA gas analyzer (Eppendorf).
The batch was started by inoculating minimal media (90% of the batch volume) with the preculture (10% of the batch volume). Once glucose was depleted (detected by a DO spike), substrate was fed to reach a cell dry weight concentration of 50 g/L and 30 g/L in the growth repression and recombinant protein production processes, respectively. Subsequently, expression of Protein A or VHH was induced by addition of 0.5 mM or 0.25 mM IPTG, respectively. Additionally, Gp2 expression in the X-press strain was induced by adding 100 mM L-arabinose.

Design Of Experiments
To study the effect of temperature and substrate uptake on physiology, productivity, lysis and leakiness during SpA production, a full-factorial screening Design of Experiments (DoE) was performed. Since growth of the X-press strain is repressed by Gp2 expression during induction, we chose to apply a constant substrate feed rate in our experiments. In the DoE, this is re ected in the rst factor as the speci c glucose uptake rate with respect to biomass at start of induction (q S,0 ). It was set to 0. 13

Analysis Of Lysis
Under the assumption that released DNA is proportional to amount of lysed cells, quanti cation of lysis was adapted from Klein et al. [43]. For our calculations, we assumed a cellular DNA content of 31 mg/g, which was taken from literature [44]. VHH quanti cation was done analogously, with the exception that the cell pellet was sonicated in MES-Buffer (100 mM MES, 10 mM EDTA, pH 6.0) and HPLC analysis was performed with a cation exchange column (BioResolve™ SCX mAb; Waters). The loading buffer was 20 mM MES, pH 6.0 and VHH was eluted with a Na + gradient.
Inclusion body formation of VHH was analyzed qualitatively by SDS-PAGE. For this, the pellet obtained after homogenization was resuspended in 20 mL of Buffer A (50 mM TRIS, 0.5 M NaCl, 0.02% Tween, pH 8.0) and then centrifuged for 10 minutes (10,000 rcf, 4 °C). The resulting pellet was washed in 20 mL Buffer B (50 mM TRIS, 5 mM EDTA, pH 8.0) and 2 mL aliquots were centrifuged for 10 minutes (10,000 rcf, 4 °C). Subsequently, the pellet was resuspended in 1 mL ultrapure water, diluted with 1.5 × Laemmli buffer. A VHH standard (5 g/L) was diluted in 2 × Laemmli buffer. The samples and standard were then incubated at 95 °C for 15 minutes. 10 µL of sample and 5 µL of standard were loaded onto precast SDS gels (8-16%, Mini-PROTEAN TGX; Bio-Rad, Hercules, CA). Gels were run at 120 V for 30 minutes in a Mini-PROTEAN Tetra-Cell (Bio-Rad) and stained with Coomassie Blue. Images were captured and analyzed using the software Image Lab (Bio-Rad).

Calculation Of Leakiness
The quotient of soluble extracellular and total intracellular product (leakiness) in percent was calculated using Eq. 5: All data generated or analyzed during this study are included in this published article and its additional les.
Competing interests