In this study, we investigated the protein expression potential of two modified phage-derived promoters, T5 and A1, which were recognized by the σ70 E. coli RNAP. The promoter sequences were modified to contain one, two, or three lacO sites. We created seven promoter/operator constructs combined with the open reading frame of the model protein, GFPmut3.1 (Figure 1).
Figure 1 Schematic of GFPmut3.1 expression cartridges controlled by seven different promoter/operator combinations. The cartridges were integrated into the attTN7 site (indicated with <pointed brackets>) of the E. coli BL21 chromosome, or they were cloned into the pET30a-cer vector (indicated with round brackets (), but not shown in this figure). In two promoter/operator combinations, the wild-type lacI promoter (B, black) was exchanged with the lacIQ promoter (BQ, red). LacO1* is a 2-bp truncated version of wild-type lacO1. Sym-lacO is the perfectly symmetric lacO. The native, initially transcribed sequence of the A1 promoter, is labeled +1 T7A1 +20. Transcription is terminated by tZENIT (tZ) . The BL21(DE3) T7 expression system (B3<T7>) is used as a reference. The BQ-wt carried the wild-type sequence, with the lacIQ promoter.
Productivity of σ70 dependent promoter/operator combinations
The T7 expression system is known to provide high expression rates, even from a single target gene copy, when integrated into the E. coli chromosome. First, we wanted to check whether the same productivity could be reached with σ70 E. coli RNAP-dependent promoters in the same experimental set-up. Therefore, we compared the genome-integrated (indicated with pointed brackets: <>) and plasmid-based (indicated with round brackets) T5 and A1 promoter/operator combination expression systems to the T7 expression system. The cells were grown in fed-batch-like conditions, in micro-titer fermentations, over a period of 22 h. Expression of GFPmut3.1 was induced with 0.5 mM IPTG after 10 h.
Figure 2 Promoter activities of different promoter/operator combinations, under non-induced (0 mM IPTG) and induced (0.5 mM IPTG) conditions. The specific fluorescence of the reporter protein, GFPmut3.1 (YP/X), is given in relative fluorescence units per mg of cell dry matter [rfu/mg CDM]. This value was used to characterize (A) genome-integrated expression systems and . Error bars indicate standard error of the mean (n = 3). Expression system names are defined in Figure 1.
In all promoter/operator combinations, the cells maintained growth in the micro-titer fermentations. The average growth rate was µ = 0.05 /h, during the 12-h production period. We directly compared average growth rates between the T7 and the σ70 promoters (Figure S2 and S3).
On-line fluorescence measurements of the plasmid-based expression systems (Figure 2B) showed that all promoter/operator combinations, except B(3lacO-T5), expressed comparable amounts of GFPmut3.1. In contrast, with the genome-integrated expression systems (Figure 2A), we observed quite distinct differences between the different promoter/operator combinations. The A1 expression systems produced 1.5-fold GFPmut3.1 yields compared to the T5 expression systems. These results were consistent with previously published data [20, 21, 28]. In the genome-integrated T7 expression system, induction of GFPmut3.1 expression led to 145 rfu and a specific soluble GFPmut3.1 concentration of ~ 135 mg/g cell dry matter (CDM). The same experiment with the A1 expression systems yielded almost 50 rfu and a GFPmut3.1 concentration of 37 mg/g CDM. A comparison of protein solubility in the plasmid-based and genome-integrated systems indicated that a large proportion of insoluble GFPmut3.1 was produced in the plasmid-based expression systems. Conversely, over 90 % of the recombinant protein was soluble in the genome-integrated expression systems(Figure S4).
The reduced productivity observed with the plasmid-based B(3lacO-T5) and the genome-integrated B<3lacO-T5> might have been due to the presence of the perfectly symmetric lac operator (sym-lacO) , which replaced the initially transcribed sequence (ITS). This symmetric lacO could influence promoter escape, and therefore, productivity . This effect was less evident with the plasmid-based 3(lacO-T5) expression system, where the high plasmid copy number compensated for the reduced promoter activity. However, in the genome-integrated expression system, the promoter activity was quite low; therefore, we discarded the 3lacO version with the A1 promoter. For the one and two lacO promoter/operator combinations, we replaced sym-lacO with the native ITS of the A1 promoter (+1 T7A1 +20). This resulted in a 1.4-fold increase in productivity, in the case of the T5 promoter.
Basal expression in σ70 dependent expression systems
For challenging proteins, even low basal expression can have adverse effects on the host metabolism, or it may even be toxic to the host cell. Hence, in those cases, equipping the host with an expression construct, either plasmid-based or genome-integrated, can be rather difficult. This difficulty is typically represented by the low frequency of transformants or integrants, respectively. Thus, the tightness of gene regulation is an important quality criterion for expression systems.
In the plasmid-based systems, promoters that were controlled by one lac operator (1lacO) showed the highest basal expression, at a level of ~ 4 rfu/mg CDM, particularly under carbon-limited conditions (Figure 2B). The addition of a second lacO (2lacO) or an increase in LacI production, by introducing the lacIQ promoter, reduced the basal expression of the A1 promoter to 1 rfu/mg CDM. In constructs with the T5 promoter, only the inclusion of three lac operators (3lacO) reduced the basal expression to almost 0 rfu/mg CDM. In contrast to the plasmid-based expression systems, all genome-integrated systems showed that the promoter/operator combination significantly impacted the system leakiness (Figure 2A). Both an increase in the number of LacI molecules and the addition of a second lacO site reduced the basal expression of A1 expression systems from 4 rfu/mg CDM to nearly no significant background expression. Importantly, productivity was not affected. Although both promoters contained lacO sites at the identical position, only an increased level of LacI molecules or three lacO sites could sufficiently reduce basal expression in the T5 expression systems. Similar findings were obtained by Lanzer and Bujard . They concluded that the promoter strength was not correlated with effective repression. The host RNAP recognized the A1 promoter only half as efficiently as the T5 promoter . When one lacO site was located within the promoter sequence, between the -10 and -35 promoter elements, the host RNAP and LacI competed with each other for their respective binding sites, and this competition determined how efficiently promoter activity was controlled by the repressor. The RNAP and T5 promoter form a complex at one of the highest complex-formation rates known in nature . Thus, controlling this promoter requires either a high repressor binding affinity in the operators or a high concentration of repressor molecules.
Control of recombinant gene expression rate
The control of the transcription rate, also referred to as “tunability”, is used to fine-tune protein production. This fine-tuning is highly relevant in bioprocessing. Optimal bioprocesses are designed to maximally exploit cell synthesizing capacities for long periods to yield correctly folded, processed proteins. Depending on the physical properties and metabolic requirements of the desired product, transcription rates must be adapted to RNA stability, translation efficiency, protein folding, protein transport, and all other interactions in the system.
To evaluate the tunability of the promoter/operator combinations described herein, we tested a series of fed-batch-like microtiter cultivations at varying IPTG levels and benchmarked protein production to the genome-integrated T7 expression system. The range of IPTG concentrations for fully and partially induction with IPTG was determined in a preliminary experiment. The strains B<3lacO-T5> and B3<T7> were induced with following IPTG concentrations: 1.0, 0.5, 0.1, 0.05, 0.01, 0.005 mM IPTG (Figure S1). Based on these results, we decided on the concentrations 0.005, 0.01 and 0.5 mM IPTG. On-line fluorescence measurements and end-point flow cytometry analyses were used to characterize the different promoter/operator combinations.
Figure 3 Influence of lac operators on expression rate control, shown by the change in on-line GFPmut3.1 fluorescence in fed-batch-like microtiter cultivations. The dashed vertical lines indicate the time of induction. Induction was performed with 0 (gray, not induced), 0.005 (red), 0.01 (blue), or 0.5 mM (green) IPTG. A–D: The T5 promoter is controlled by: (A) three lacO, (B) two lacO, (C) one lacO, and (D) one lacO / lacIQ sequences. E–G: The A1 promoter is controlled by (E) two lacO, (F) one lacO, and (G) one lacO / lacIQ sequences. (H) The T7 expression system is used as a reference. The Y-axis scale is adjusted to the respective expression rates. The mean relative GFP fluorescence intensity (rfu) represents triplicate samples.
Expression systems controlled by one lacO site for gene regulation exhibited the highest basal expression and the least pronounced gradation of GFPmut3.1 expression at increasing inducer concentrations (Figure 3C, F). Although promoters controlled by two lacO sites showed sufficiently low basal expression, they also produced less protein at the lower inducer concentrations (Figure 3B, E). The promoter/operator combinations controlled by 3lacO-T5 and 2lacO-A1 led to a complete stop (plateau) of recombinant GFPmut3.1 production after a certain time, independent of the inducer concentration (Figure 3A, E). We did not observe this behavior in promoter/operator combinations with only one lacO site. The combination of promoters controlled by one lacO site and lacIQ repressor (Figure 3D, G) and the T7 expression system (Figure 3H) resulted in the desired system properties, including tunability and low system leakiness.
Figure 4 Flow cytometry analysis of single-cell expression of GFPmut3.1. Induction was performed with 0 (gray, not induced), 0.005 (red), 0.01 (blue), or 0.5 mM (green) IPTG. A–D: T5 promoter controlled by: (A) three lacO, (B) two lacO, (C) one lacO, or (D) one lacO / lacIQ sequences. E–G: A1 promoter controlled by: (E) two lacO, (F) one lacO, or (G) one lacO / lacIQ sequences. (H) The T7 expression system is used as a reference.
T7 expression systems exhibit an all-or-none induction phenomenon, where reduced expression in partially induced cultures results from the formation of subpopulations of fully induced and non-induced cells . Therefore, we investigated transcription rate tuning at the cellular level with flow cytometry analyses of all genome-integrated promoter/operator combinations (Figure 4). We confirmed that the all-or-none phenomenon occurred in genome-integrated T7 expression systems. In fact, we observed a mixture of fully, partially, and non-induced cells, particularly at very low inducer concentrations (Figure 4H, red line). In the B<2lacO-A1> expression system, flow cytometry analyses revealed that these expression systems stopped GFPmut3.1 production, although the cells continued to grow (Figure S2 and S3). This result indicated that there were two distinct subpopulations of producing and non-producing cells. We also observed this behavior in B<3lacO-T5> (Figure 4A, E). But the BQ<1lacO-A1> system showed different behavior. There, the induction of the gfpmut3.1 gene resulted in a homogenous population at any given IPTG concentration (Figure 4G). Consequently, this expression system provided proof that the expression rate was controlled on a cellular level.
Influence of LacI autoregulation on expression rate control
We assumed that the complete stop in productivity, observed when the B<3laco-T5> and B<2lacO-A1> systems were partially induced, was associated with the autoregulation of the lac repressor. The native lac operon is regulated by three lacO sites (Figure 5A). The LacI molecule simultaneously binds to two sites, either lacO1 and lacO3 or lacO1 and lacO2 . The lacO3 sequence overlaps with the 3‘ end of the lacI gene. When LacI binds to lacO1 and lacO3, it causes the DNA to form a loop. This results in truncated lacI mRNA molecules, which are degraded by the cell. This autoregulation of LacI production resulted in a constant level of ~10 LacI molecules per cell in the absence of an inducer [11, 31, 32].
Figure 5 Schematic of lac operators in the native lac operon (top panels) and its regulation of the gene of interest (bottom panels). Lacl (pink tetramer) production effects are shown, when the promoter for the gene of interest was regulated by (A) one lac operator or (B) two lac operators, respectively. Ka = association constant; red cross = stopped production
We hypothesized that, when the binding constant (Ka) of LacI to the lacO sites of the GOI was greater than the binding constant to the lacO sites of the lac operon, the first LacI molecules, which are not inactivated by IPTG, will preferentially bind to the lacO site of the GOI, instead of the lacO3/lacO1 within the lac operon. Hence, autoregulation of LacI would not intervene, and LacI molecules would continue to be produced. This would cause the whole system to become overregulated, which would result in a complete stop in production (Figure 5B).
To test this hypothesis, we compared the effect of autoregulation on LacI in B<2lacO-A1> and BL21 wild-type cells (BL21-wt). We estimated the LacI content of non-induced, partially-induced, and fully-induced cells with western blot analyses. The band intensities were quantified and normalized by the cell number (Figure 6).
Figure 6 Influence of additional lacO sites on cellular LacI concentrations. Proteins of ~ 1.2 x 107 cells were separated with SDS-PAGE and analyzed on western blots, probed with an anti-LacI antibody. (A) Western blot of BL21 wild-type cells and B<2lacO-A1> cells, which were grown without IPTG, 0.01 mM IPTG, or 0.5 mM IPTG. (M) PageRuler™ Plus Prestained Protein Ladder. (B) Fold changes of band intensities determined in panel A are shown relative to the levels observed in 0 mM IPTG BL21-wt cells. Error bars indicate the standard error of the mean (n = 3).
In fully induced (0.5 mM IPTG) BL21-wt cells, the number of LacI molecules was 3.3-fold greater than the number observed in non-induced BL21-wt cells. Partial induction with 0.01 mM IPTG only led to a 0.7-fold increase. The 3.3‑fold change in fully induced BL21-wt cells was consistent with previous results from Semsey et al. In that study, they measured an average of 15 LacI molecules per cell in the absence of inducer and ~40 molecules per cell in fully induced cells .
In B<2lacO-A1> cells, LacI numbers in non-induced and partially induced conditions were clearly higher than the numbers observed in uninduced BL21-wt cells. LacI yields were 2.4-fold greater in the absence of inducer and 3.2-fold greater in partially induced cells, relative to uninduced BL21-wt cells. In fully induced cells, LacI yields were 4.3-fold greater than those observed in uninduced BL21-wt cells, which was similar to the yield in fully induced BL21-wt cells.
Although the addition of 0.01 mM IPTG resulted in almost half-maximal GFPmut3.1 expression in B<2lacO-A1> cells (Figure 3E), it had little or no influence on the LacI levels. This suggested that LacI continued bind to lacO1/lacO3 in the lac operon; hence, it could maintain autoregulation under these conditions. In the fully induced state, the LacI concentrations are almost the same with a 4-fold increase regardless of whether it is the BL21-wt or the B<2lacO-A1> expression system. LacI therefore no longer binds to its operators and thus the expression of LacI itself is no longer inhibited. The small fold change of 4 results from the weak constitutive LacI promoter, which provides about one new mRNA per cell generation . Thus, the high LacI levels in non-induced and partially induced B<2lacO-A1> cells clearly supported our hypothesis that LacI autoregulation impacted the expression rate control in genome-integrated E. coli production strains (Figure 5).
The effect of LacI autoregulation was only observed in genome-integrated, host RNAP-dependent expression systems, which were controlled by two or three lacO sites. In contrast, this effect was not observed in plasmid-based, host RNAP-dependent expression systems or in the conventional T7 expression system. This discrepancy might be explained by differences in the balance between lacO sites and LacI concentrations. The T7 expression system harbors a second lacI gene sequence within its DE3 lysogen, which would, theoretically, double the LacI concentration per cell. The plasmid-based expression systems used in this study were based on the pET plasmid system, which encodes a second lacI gene sequence. In turn, depending on the plasmid copy number, that resulted in an extra 15 - 20 lacI gene sequences . However, the effect of LacI autoregulation on partially induced cells was also observed in plasmid-based expression systems, like the E. coli pAVEwayTM expression system, from Fujifilm Diosynth Biotechnologies (Hillerød, Denmark). In the pAVEwayTM expression system, transcription control was enabled by two perfectly symmetric lac operators, one positioned upstream and one downstream of the T7A3 promoter. The high affinity of LacI to the symmetric lacO sites, combined with the ability to form a DNA loop, resulted in very low basal expression, but also, a complete stop in productivity in partially induced cultures.
Considering the autoregulation of lac repressor synthesis, we identified BQ<1lacO-A1> as the σ70 promoter/operator combination that fulfilled the desired properties. It showed a high expression rate, negligible basal expression, and true tunability of the expression rate on a cellular level, even at low inducer concentrations, without a complete stop in productivity.