Decoupling Growth and Production by Removing the Origin of Replication from a Bacterial Chromosome

Abstract

Efficient production of biochemicals and proteins in cell factories frequently benefits from a two-stage bioprocess in which growth and production phases are decoupled. Here, we describe a novel growth switch based on the permanent removal of the origin of replication (oriC) from the Escherichia coli chromosome. Without oriC, cells cannot initiate a new round of replication, and they stop growing while their metabolism remains active. Our system relies on a serine recombinase from bacteriophage phiC31 whose expression is controlled by the temperature-sensitive cI857 repressor from phage lambda. The reporter protein expression in switched cells continues after cessation of growth, leading to protein levels up to 5 times higher compared to nonswitching cells. Switching induces a unique physiological state that is different from both normal exponential and stationary phases. The switched cells remain in this state even when not growing, retain their protein synthesis capacity, and do not induce proteins associated with the stationary phase. Our switcher technology is potentially useful for a range of products and applicable in many bacterial species for decoupling growth and production.

Introduction

Production of value-added chemicals in living cells holds promise to reduce the need for oil and divert to more sustainable production methods. The list of chemicals that can be produced through metabolic engineering grows constantly, and there are many scientific papers describing the biosynthesis route for some new product.¹ However, only a small fraction of these are actually produced commercially, and many bioproduction processes suffer from low yield, titre, and productivity.²

Forcing a growing cell to produce any product (protein or small molecule) at significant quantities inevitably generates a resource allocation conflict. Biomass increase and product formation both use the same general resources (carbon and energy) and usually also compete for the same key metabolite(s). This leads to a growth–production trade-off: high production strains grow slowly and rapidly growing strains have low product yields.^3,4

A two-stage bioprocess has been used to bypass this trade-off.⁵ In the first stage, the cells are grown at a maximal rate without any significant product production. At the desired moment, the growth is turned off, and production is induced so that most of the available resources can be used for product formation. This two-stage bioprocess has been successfully implemented in strains where the switch from growth to production is part of their natural regulation. Combining an aerobic growth stage with an anaerobic production stage in Corynebacterium acetoacidophilum resulted in a high concentration of succinate,⁶ and a pH-shift-induced production stage in Klebsiella pneumoniae led to the accumulation of 2-ketogluconic acid.⁷ Other bacteria, such as Escherichia coli, do not have such natural regulation, and their growth must be controlled by other means. Nutrient limitation (other than the main carbon source) has been used to curb E. coli growth while maintaining product synthesis.⁸⁻¹⁰ In one study, limiting nitrogen also inhibited carbon metabolism, but constraining phosphorus, sulphate, or magnesium allowed cells to keep their metabolism active.⁸

In recent years, several artificial switches have been built to stop growth while leaving cells metabolically active. Brockman and Prather blocked fructose-6-phosphate utilization in glycolysis and channelled its precursor, glucose-6-phosphate, toward the synthesis of myo-inositol.¹¹ As a result, the growth rate of the culture decreased 6-fold and myo-inositol production increased more than 2-fold. Later, they added a quorum-sensing module to regulate the switch in an autonomous manner.¹² Klamt and colleagues utilized the temperature-controlled expression of an essential TCA cycle gene to stop the growth;¹³ the resulting two-stage bioprocess showed an increase in titre and volumetric productivity of itaconic acid production. These “metabolic valve” approaches suppress some cellular pathway necessary for cell growth and channel the carbon flow to the product production pathway. Lynch and colleagues combined phosphate limitation with a metabolic valve approach to limit cell growth and adjust its metabolism, respectively. This led to the increased production of both small molecules¹⁴ and proteins.¹⁵ Optogenetic regulation has also been used to simultaneously limit growth and induce production of isobutanol in yeast.¹⁶ Li and colleagues used CRISPR/dCas9 to suppress the expression of several endogenous genes, which resulted in growth reduction and increased expression of GFP.^17,18 Notably, the best targets were several genes in the nucleotide biosynthesis pathway.

Despite multiple options for two-stage fermentations, there is room for improvement. In several cases, growth is not eliminated by switch but only slowed down, substrate uptake rate in production phase is low, improvement in protein expression is marginal, or the system exhibits a very long reaction time. These shortcomings hamper the usability of these two-stage fermentation solutions and restrict their application in larger scale.

Chromosomal DNA replication in bacteria starts from the well-defined origin of replication, oriC. In E. coli, the DnaA protein initiates replication by occupying its multiple DNA binding sites inside the oriC region and drives DNA unwinding.^19,20 In subsequent steps, additional proteins are recruited, but the replication initiation is essentially controlled by the oriC–DnaA interaction.

Here, we describe a novel approach to decouple growth and production, which does not require direct modification of any metabolic pathway. During the switch from growth to production, we induce the removal of oriC from the chromosome with the help of a site-specific serine recombinase. This prevents further initiation of chromosomal DNA replication, and eventually, the cells stop growing. Cellular metabolism, however, remains active, and the cells keep expressing proteins long after cell growth has stopped.

Results and Discussion

Temperature-Induced Excision of oriC in a Switcher Strain

We reasoned that the removal of oriC from the genome will stop replication initiation and lead to cessation of growth. To achieve this, we redesigned the vicinity of oriC in the E. coli genome by adding serine recombinase recognition and cleavage sites (attB and attP) on either side of it (Figure 1a). We also included a GFP reporter gene downstream of the attB site in such a configuration that GFP would be expressed only after excision of oriC; this allowed us to monitor cells in which there was a successful recombination between attB and attP (i.e., switched cells) (Figure S1a).

Figure 1.

Expression of integrase leads to excision of oriC in the switcher cells. (a) Schematic setup of the switcher strain: serine recombinase recognition sites attB and attP are integrated on either side of oriC in the bacterial chromosome in an orientation that leads to the excision of oriC upon recombination. (b) Control (black) and switcher (orange) cultures were pregrown at 30 °C (white plot area). At timepoint 0 h, the temperature was changed to 37 °C (gray plot area). The number of CFUs was determined. The geometric mean of three independent experiments is plotted; error bars indicate standard deviation. (c) PCR was used to test switching on the chromosomal DNA using primers either unique to unrecombined (oriC retained) or recombined (oriC excised) DNA sequences and visualized using an agarose gel electrophoresis, M—DNA marker. Full uncropped gels are presented in Figure S2a.

The expression of a recombinase that is used to remove oriC from the genome must be tightly controlled to allow normal cell growth during uninduced conditions. We used the lambda phage cI857 transcriptional repressor²¹ and its target promoter to control the expression of serine recombinase from bacteriophage phiC31 (phiC31 integrase).²² cI857 is a temperature-sensitive mutant that represses its target promoter at 30 °C, but the repression is relieved at 37 °C. We generated an E. coli switcher strain in which initiation of the chromosomal replication can be eliminated by changing the culture temperature from 30 to 37 °C. In its chromosome, oriC sequence is flanked by phiC31-integrase-specific attP and attB sites, and it carries a medium-copy expression plasmid pAJ35 encoding an inducible phiC31 integrase (Figure S1b, File S1). As a control strain, we used the same bacterial strain but carrying plasmid pAJ27, which expresses only 20 N-terminal amino acids of phiC31 integrase and cannot catalyze the recombination (Figure S1b, File S2).

To test the removal of oriC, we first determined the number of colony-forming units (CFUs) of switcher and control cultures. Bacterial cells devoid of oriC cannot multiply and form visible colonies on a solid medium, and the lack of colonies is therefore an indication of the functional switch. At the same time, these cells remain intact and even increase their mass (see Figure 2c). Therefore, the usual correlation between optical density and CFU measurements does not hold in switched cells. The strains were precultured overnight at 30 °C, diluted in the fresh medium, and incubated at 30 °C for 2 h before shifting the temperature to 37 °C. After the initial lag phase, the number of control culture CFUs started to increase as expected, whereas the switcher CFU count began to decrease (Figure 1b). Four hours after the temperature shift, the switcher CFU count is more than 2 orders of magnitude lower than at the time of the temperature shift, indicating the effective excision of oriC.

Figure 2.

Switching enables the selection of final cell density. (a) Control strain (dashed lines) and switcher strain (solid lines) were inoculated from the overnight culture in various dilutions into fresh medium in a 96-well plate. At timepoint 0 h, the temperature was changed from 30 to 37 °C to induce the switching. At the end of the experiment, the cultures were diluted and spotted to LB agar plates. (b) Number of CFUs was determined after incubating the plates for 20 h at 30 °C (control and switcher, striped or filled bars, respectively) or 37 °C (switcher, empty bars). The dashed line indicates the detection limit. (c) Growth of control and switcher cells was monitored at 37 °C for 6 h using time-lapse microscopy. Hourly snapshots of cultures up to the 4 h timepoint are presented here. Full-length time-lapse videos are presented in Files S3 and S4. The mean of six biological replicates from two independent experiments is plotted on panel (a); shading indicates standard deviation. The mean of three biological replicates is plotted on panel (b); error bars indicate standard deviation.

We verified the DNA rearrangement in switched cells by PCR amplification specific to either a preswitch or postswitch DNA configuration (Figures 1c, S2a). In the switcher culture, the preswitch configuration is detected before the temperature shift, while a postswitch-specific PCR product appears after the culture has been shifted to 37 °C. In control cells, only the product specific to the preswitch configuration is present. We also verified the sequence of the attR site in the genome, which is formed during switching after recombination between attB and attP (Figure S2b).

Switching Enables the Selection of Final Cell Density

After removal of oriC from the genome, the cells cannot initiate a new round of replication, and the cell density should stabilize at a submaximal level. To test this hypothesis, we grew the switcher and control strains on a 96-well plate and followed the culture density before and after shifting the incubation temperature from 30 to 37 °C (Figure 2a). A few hours after the switch, the growth of the switcher strain attenuates, and its cell density eventually stabilizes at values different from a control strain. The switcher strain becomes stationary, but this is different from the stationary phase in control cells. The latter stops growing because they have used up available nutrients (nutrient-depleted stationary phase), but the switcher cells cannot grow because of lack of DNA replication (switching-induced stationary phase). For the sake of clarity, we refer to the switching-induced stationary phase as the switched culture whose density has plateaued at a certain OD600 value. The plateau OD level reached depends on the cell density at the time of switch: the greater the dilution from the overnight culture, the lower the plateau. Notably, the final density of a less diluted switcher culture (1600×) can even exceed the density of the stationary-phase control strain. The measured OD values of the control culture decrease through the nutrient-depleted stationary phase, likely because of cell shrinkage, whereas the decrease in switcher cultures is less prominent.

To quantify the extent of switching, we measured the number of CFUs in cultures after 20 h of incubation at 37 °C (sampled from the end of the experiment described in Figure 2a). Samples were plated on lysogeny broth (LB) agar medium and incubated overnight at 30 or 37 °C as indicated on the figure. The switcher cells that have been incubated for 20 h at 37 °C on a 96-well plate and still form colonies on an agar plate at 30 °C have either not switched yet or have permanently lost the ability to switch (escaper mutants). On agar plates incubated at 37 °C, only escaper mutants can form colonies, meaning the removal of oriC is defective in these cells. Despite a similar culture biomass observed at the end of the experiment in Figure 2a (less than 2× difference), most cells in the switcher culture have switched, which can be seen because the number of CFUs on the 30 °C plate has decreased by up to 4 orders of magnitude compared to CFUs of the control on the 30 °C plate (Figure 2b). A fraction of escaper mutants seems to increase in a more diluted culture; in a 3200× dilution, they constitute approximately 1% of all cells in switching-favored conditions.

Because the serine recombinase reaction is effectively irreversible, the switched cells cannot return to growth even if the temperature is lowered back to 30 °C degrees (as was tested in CFU measurements). This is different from growth regulation systems depending purely on gene activation and/or repression because these cells can grow again after regulatory signalling is removed. Irreversible elimination of cell multiplication might be a useful feature for the development of living medicines, where bacterial cell growth must be kept under tight control.²³

To analyze the morphology of switching cells, we used time-lapse microscopy. The cells were pregrown at 30 °C in LB medium and transferred onto an LB agarose pad under a microscope at 37 °C. Both control and switcher cells keep growing for the first few hours (Figure 2c). The switcher cells start to express GFP, indicating that switching has taken place. The switchers also become elongated and form filaments at later timepoints, as seen in time-lapse videos (Files S3 and S4). This indicates that cell division is inhibited in switched cells.

Similar cell elongations have been shown by Nielsen and co-workers who used CRISPR-interference (CRISPR/dCAS9) mediated blocking of DnaA binding sites in oriC as a growth switch to generate a two-stage production system.¹⁷ This switch slowed down the growth but did not block it completely and resulted in a modest increase in protein production. Their approach was more efficient by targeting the genes pyrF,¹⁷pyrG, and cmk,²⁴ leading to stabilization of culture density at a submaximal level. Notably, blocking of pyrF also led to elongated cells similar to those in our experiments (Figure 2c), indicating that this is a common reaction to the disruption of DNA metabolism.

Several examples of two-stage bioproduction techniques use metabolic valves, where the flow of metabolites is diverted from growth to production during the switch.^11,13,25 While effective, these solutions are rather product-specific, and for almost every new type of product, a new growth switch must be built. Our switching approach blocks only DNA replication initiation and does not directly target other metabolic pathways, making it applicable for many products.

Switching Enhances Protein Expression

Stopping cell growth before all nutrients are depleted from a growth medium should leave more resources for protein production. We tested this assumption by inducing the expression of a red fluorescent protein (mRFP1) at the time of the switch and by following the mRFP1 fluorescence intensity (Figure 3a). In these experiments, the cells carry a reporter plasmid pAJ144 (Figure S1); the expression of mRFP1 is controlled by LuxR transcriptional regulator and can be induced by adding homoserine lactone (HSL) into the growth medium. Initially, mRFP1 accumulates in both control and switched cultures, but the accumulation stops in the control culture when it enters the nutrient-depleted stationary phase (OD curve plateaus). In contrast, the switched cells continue to synthesize mRFP1 long after they have reached their maximum culture density. By the end of the experiment, the level of mRFP1 intensity is almost 3 times higher in the switched culture compared to the control, even when the cell density in the switcher remains lower.

Figure 3.

Assessment of protein production capability of the switcher and control strains in LB. Overnight cultures carrying a reporter plasmid pAJ144 were diluted 2400× or 1600× as indicated in the figure panels. Control (black) and switcher (orange) cultures were pregrown at 30 °C on a 96-well plate. At timepoint 0 h, the temperature was changed to 37 °C to induce switching. The production of fluorescent mRFP1 protein was induced by adding HSL (final concentration 12 μM) at timepoint zero. (a,b) Changes in OD at 600 nm (dashed lines, left axis) and increase in fluorescence (excitation: 584 nm, emission: 607 nm) of mRFP1 protein (solid lines, right axis) were monitored. (c,d) Fluorescence-over-optical-density ratio of control (black) and switcher (orange) culture was calculated based on values in panels (a,b), respectively, and plotted. The mean of three biological replicates is plotted; shading indicates standard deviation.

The difference in protein expression is even more evident when the switching is timed so that the switched culture stabilizes at the same density as the control culture (Figure 3b). At the end of the experiment, the mRFP1 level in the culture with an initial 1600× dilution is 5 times higher than in the control. mRFP1 intensities normalized to OD are clearly higher in the switcher than in the control and very similar in both dilutions (Figure 3c,d), indicating a more favorable protein-to-biomass ratio. To exclude a possible bias in one specific expression system, we tested the protein synthesis capacity using a different reporter protein, promoter system, growth medium, and plasmid backbone and confirmed switcher-enhanced production of the protein of interest (Figure S3).

Changes in the cell morphology, including filamentation, have been associated with increased accumulation of a desired product.²⁶ To test if the elevated mRFP1 expression is dependent on cell elongation, we induced filamentation in the mRFP1-expressing cells by inducing SulA overexpression. SulA is a well-known FtsZ inhibitor, and its expression inE. colileads to cell filamentation.²⁷ SulA-overexpressing cells become filamentous but do not produce more mRFP1 (Figure S4), suggesting that reasons other than cell elongation are behind the increased protein expression in our switcher cells.

Lynch and colleagues used phosphorus limitation to constrain growth and metabolic valves to reconfigure metabolism. This separates growth limitation from metabolic rearrangement and allows use of the same growth-curbing mechanism for different types of products.^15,28 Removal of oriC in our switcher is also a general way to stop growth and should be combined with appropriate changes in gene expression to achieve better production of the desired product. Both our switcher (Figure 3) and a phosphate-limitation-based system¹⁵ allow protein accumulation after cell growth has stopped or significantly slowed down. This is in contrast to conventional protein expression where growth and production are coupled, and the cells stop protein synthesis as soon as they reach the nutrient-depleted stationary phase.¹⁵

Protein synthesis can only occur if cells continue to use the growth substrate and keep up active metabolism. We measured the glucose uptake in switched cells and compared it to both growing and N-starved control cells (Figure S5). N-starvation stops growth but also shuts down the cellular metabolism so that very little glucose is consumed.^8,29 Growing cells consumed glucose at the rate of 6.52 ± 0.05 mmol/gDCW/h, while the rate for N-starved cells was only 0.44 ± 0.19 mmol/gDCW/h. This is similar to values reported previously.⁸ The specific glucose uptake rate in switched cells was 1.41 ± 0.14 mmol/gDCW/h, more than 3 times higher than in N-starved cells. This is in par with the glucose uptake rate during sulfur starvation and exceeds the same value in phosphate limitation.⁸

Coupled growth and production is problematic also for another reason. High-level production is a burden to cellular metabolism due to high resource usage and possible toxicity. This generates a selection pressure to accumulate cheater mutants—nonproducer cells that have a growth advantage. Accumulation of cheaters is a well-recognized problem in bioproduction,³⁰ and special effort in coupling production to the expression of an essential gene is needed to counteract the problem.^31,32 When production induction is concurrent to or initiated by the growth switch, high-level production only occurs in switched cells, which will not grow anyway. Although some cells escape the switch and can grow at 37 °C (Figure 2b), the frequency of mutants should be independent of burden because growing cells do not express the product yet.

Switched Cells Retain a Protein Synthesis Capability Long into the Nongrowing Phase

Continuous accumulation of an expressed reporter protein, despite unchanging cell density (Figure 3), prompted the question regarding how long switched cells can retain the ability to induce protein expression. To find this out, control and switcher cultures were grown on a 96-well plate, and the expression of mRFP1 reporter was induced at different timepoints after the switch (Figure 4). Already 3 h after the switch, induction in the control is very weak but almost unaffected in the switcher (Figure 4b). At 6 h, the control culture fails to induce any detectable mRFP1 expression, while in the switcher, the mRFP1 induction is still strong. The addition of HSL, even after 20 or 30 h, results in mRFP1 induction, although at lower levels (Figure 4b, inset). There is also a noticeable increase in the mRFP1 signal in the uninduced switcher culture, which is probably caused by the partial derepression of the mRFP1 promoter. The productive window, during which the protein synthesis is possible, lasts significantly longer in switched cells, although their capacity to produce proteins diminishes over time. These results suggest that despite similar cell densities, the switched cells are in a different physiological state that is favorable for protein expression.

Figure 4.

Protein production capability of the switcher remains active over a long period of time in LB. Control (dashed lines) and switcher (solid lines) overnight cultures carrying a reporter plasmid pAJ144 were diluted 1600× and pregrown at 30 °C on a 96-well plate. At timepoint 0 h, the temperature was changed to 37 °C to induce switching. The production of fluorescent mRFP1 reporter protein was induced by adding HSL (final concentration 12 μM) at indicated timepoints. Changes in optical density (a) and fluorescence (b) (excitation: 584 nm, emission: 607 nm) were monitored. The subplot is a zoom-in of panel (b) to better visualize the induction at 20 and 30 h. The mean of three biological replicates is plotted; shading indicates standard deviation.

Switched Cells Enter into a State That Is Distinct from Normal Growing or Nutrient-Depleted Stationary Phase Cells

Our results above indicate that nongrowing switched cells (switching-induced stationary phase cells) are in a different physiological state compared to nongrowing control cells (nutrient-depleted stationary phase cells). To characterize this state, we analyzed and compared the proteome of switched and control cells. We designed the experiment so that the cell density in the switcher culture would either reach a similar level to the nutrient-depleted stationary phase of the control after the switch or stay at a significantly lower level. We collected samples of switched and control cells from growing, early-plateau, and late-plateau stages according to the growth curves of each culture and additionally from a switched culture at low plateau (Figure 5a).

Figure 5.

Comparison of proteomes of control and switcher cells. (a) Growth curves of control (black) and switcher (orange and purple) cultures. Two different initial dilutions from the overnight culture were used to inoculate switcher cultures to obtain final cellular densities after the switch that were either similar to the control (orange) or approximately at OD ≈ 1 (purple). Sampling points for mass spectrometric analysis are indicated with empty symbols. (b) PCA of proteomics data. (c) Hierarchical clustering heatmap of differentially expressed proteins. Proteins shown have statistically significant expression levels (multiple-sample ANOVA test FDR < 1%). Log2 transformed LFQ intensities were Z-scored across rows, and hierarchical clustering using Euclidean distances was performed with the result presented as a heatmap. (d) Comparison of protein expression levels between control growing and control early-plateau samples. Red dots represent proteins upregulated in early-plateau cells, and blue dots represent the downregulated ones (fold change > 2, FDR < 1%). (e) Comparison of protein expression levels between control growing and switcher early-plateau samples. Protein color coding is the same as in panel (d). Presented proteomic data were analyzed using Perseus software, and source data are available in File S10.

To create a low-dimensional representation of the variability between switcher and control samples, we performed a principal component analysis (PCA) (Figure 5b) and a t-distributed stochastic neighbour embedding (t-SNE) (Figure S6a) analysis using the whole proteome of each sample. Proteome profiles of switched cells are different from both growing and nutrient-depleted stationary-phase (plateaued) control cells. All samples from switched cultures cluster together; growing and nongrowing switched cells are notably more similar to each other than the respective samples of control cultures. Stable proteomic profiles support the hypothesis that switching leads to a distinct cellular state and that the switched cells stay in this state for at least 28 h.

To analyze the differences between switcher and control cultures at the resolution of individual proteins, we visualized all statistically significant differences in protein abundance as a heatmap (Figure 5c). Statistical significance of changes in abundance of proteins between sample groups was determined by a multiple-sample ANOVA test, and hierarchical clustering of the significant proteins was performed on log2 transformed intensities after Z-score normalization. A hierarchical clustering dendrogram supports a similar sample group formation as in PCA and t-SNE and corroborates distinct differences between control and switcher cells. All samples from switched cultures cluster together, and their protein expression pattern is clearly different from control samples.

We also analyzed to what extent the nongrowing switched cells express proteins that are associated with the nutrient-depleted stationary-phase control cells. We identified and color-coded proteins in the control strain that are up- or downregulated in the early plateau (early stationary phase) and compared these to growing cells (exponential phase) (Figure 5d). Mapping these proteins on a plot comparing the switcher early plateau to growing control cells reveals that the nutrient-depleted stationary-phase-specific up- or downregulated proteins do not follow the same pattern in switched cells but instead are distributed quite randomly (Figure 5e). We got similar results when comparing late-plateau samples of the switcher to growing control cells (Figure S6b,c). To exclude that the distribution of up- and downregulated proteins in switcher samples is a result of random sample point variability, we compared early- and late-plateau samples of the control culture (early and late nutrient-depleted stationary phase). Most of the proteins that are upregulated in the late stationary phase are also upregulated in the early stationary phase (Figure S6d).

Switched cells remain metabolically active even if a culture reaches the cell density normally characteristic of the nutrient-depleted stationary phase (Figures 3, 4). Corroborated by proteomics, switched cells are clearly different from nutrient-depleted stationary-phase cells (Figure 5). This suggests that the removal of oriC induces a state in which cells do not respond to signals that initiate the nutrient-depleted stationary phase in normal cells. Ribosome hibernation factors Sra, Rmf, Hpf, and YqjD, which are responsible for inactivating ribosomes in the nutrient-depleted stationary phase,³³ are not elevated in switched cells and display expression levels similar to growing cells (File S6). Our results also indicate that the lack of protein synthesis in the nutrient-depleted stationary phase is due to the physiological response and not because of a simple lack of carbon source or energy—the high-density switched culture kept synthesizing proteins, while a nutrient-depleted stationary-phase culture did not. Our switcher strain could, among other applications, be a useful tool to study the onset of the stationary phase and the role of chromosomal replication in it.

We identified proteins that are detected only in switcher samples and in none of the control samples and those that are upregulated in all of the switched samples compared to each control sample (File S5). We mapped these proteins onto metabolic pathways using the toolset at EcoCyc.org and identified pathways that are upregulated in switched cells (p-value cutoff 0.05) (File S5). In most cases, only one or two proteins in a pathway were upregulated in switched cells. However, three distinct group of pathways had more hits.

First, the expression of ribonucleoside reductases, which are responsible for dNTP synthesis, is elevated after switching (NrdA, NrdB, and NrdD; File S5). DNA replication cannot be initiated in switched cells, and the cells may perceive this as a lack of dNTPs. Transcription from the main ribonucleotide reductase operon nrdAB is stimulated by DnaA-ADP and inhibited by DnaA-ATP.³⁴ Normally, the DnaA-ATP level is high before DNA replication initiation, and it is converted into DnaA-ADP soon after initiation. This cycle is absent in switched cells, and a permanently low DnaA-ATP-over-DnaA-ADP ratio might be behind the elevated expression of Nrd proteins.

Second, a full operon of his genes from the l-histidine biosynthesis pathway is upregulated. Histidine biosynthesis pathway is closely connected to the nucleotide biosynthesis pathway by AICAR (5-aminoimidazole-4-carboxamide ribonucleotide). AICAR is an important precursor for nucleotide biosynthesis, and depletion of AICAR derepresses the expression of his operon.³⁵

Third, enzymes participating in fatty acid synthesis pathways are upregulated in switched cells. These include FabG, FabI, FabF, FabA, FabH, and FabD, which are involved in several synthesis pathways. Their expression is controlled by FadR and FabR transcriptional regulators whose activity is regulated by unsaturated fatty acids.³⁶ Notably, fatty acid synthesis can determine the cell size in E. coli,³⁷ so upregulation of fatty acid synthesis enzymes could contribute to the elongated cell phenotype.

Conclusions

Here, we describe a novel way to decouple growth and production by removing oriC from the E. coli genome. In the current version, the trigger for the switch is elevation of temperature, but this could become difficult to control in large-scale fermentations. Other types of regulation (small molecules, etc.) could be used to control serine recombinase expression, as long as it is tightly repressed in noninducing conditions. Commonly used isopropyl ß-D-1-thiogalactopyranoside (IPTG)-inducible tac promoter, for example, is not, and we were unable to build a viable strain where serine recombinase expression was controlled by a tac promoter.

We show how cells can remain metabolically active and keep synthesising proteins long after growth has stopped. For the purposes of bioproduction, our growth decoupling system could be combined with other measures that increase product synthesis, such as the up- and downregulation of specific metabolic pathways, substrates, or cofactor availabilities. Stopping growth is not a substitute but rather an addition to these methods to increase product titre, yield, and productivity.

Materials and Methods

Bacterial Strains, Plasmids, and Growth Medium

Strains and plasmids are listed in Table 1. E. coli DH5α was used for plasmid cloning and propagation. Genomic alterations and plasmid-based switcher experiments were performed in E. coli MG1655. E. coli was grown in LB supplemented with the appropriate amount of antibiotics (100 μg/mL ampicillin, 25 μg/mL chloramphenicol, 25 μg/mL kanamycin, and 10 μg/mL tetracycline) when necessary for the selection of strains and maintenance of plasmids.

Table 1. Strains and Plasmids^a.

strain	description	source/references
DH5α	F–endA1 glnV44thi-1recA1 relA1 gyrA96 deoR nupG purB20 φ80dlacZΔM15 Δ(lacZYA-argF)U169, hsdR17(rK–mK+), λ–	laboratory stock
MG1655	K-12 F– λ–ilvG–rfb-50rph-1	laboratory stock
bAJ78	MG1655 ΔoriC::cat-attP-oriC-attB, CmR	this study
bAJ83	bAJ78/pAJ27, CmR KanR	this study
bAJ84	bAJ78/pAJ35, CmR KanR	this study
bAJ85	bAJ78/pAJ27/pAJ144, CmR KanR TetR	this study
bAJ86	bAJ78/pAJ35/pAJ144, CmR KanR TetR	this study
Plasmids
pAJ27	KanR, PlacI-cI857, lambaPR(T41C)-phiC31Int(1-29) in pBR322, negative control, truncated phiC31 integrase	this study
pAJ35	KanR, PlacI-cI857, lambdaPR(T41C)-phiC31Int in pBR322	this study
pAJ144	TetR, LuxR-6His-mRFP1 in pJB866	this study
pAJ157	AmpR, Ptac-Crimson in p15A	this study
pAJ181	KanR, PlacI-cI857, lambdaPR(T41C)-sulA in pBR322	this study
pInt	KanR, PlacZ-phiC31Int in pACYC177	(38)
pKD46	AmpR, lambda-Red helper plasmid (recombinase)	(39)
pKD3	AmpR CmR, template plasmid for FRT-flanked cat cassette	(39)
pTc-Wasabi	PlacI-cI857, PR/PL-mWasabi in pETDuet-1	(40)

^aCm^R, chloramphenicol resistance; Kan^R, kanamycin resistance; Amp^R, ampicillin resistance; Tet^R, tetracycline resistance.

DNA Manipulations

Short oligonucleotide sequences for cloning and sequencing were ordered from Metabion International AG. E. coli MG1655 genomic in situ engineering was performed using recombineering. The Lambda Red recombination system was expressed from the pKD46 plasmid.³⁹ Synthetic terminators (L3S3P25 and L3S2P24⁴¹), attB and attP sites, and tac promoter (without operators, Table 2) were ordered as synthetic DNA from Twist Bioscience. Flanking 400 bp long homologous regions were amplified from the genomic DNA of E. coli MG1655. The chloramphenicol resistance marker with adjacent FRT sites was amplified from pKD3. Recombineering fragment f_bAJ78 (Figure S1a, File S6) was assembled using overlap extension PCR, gel purified, and transformed into E. coli cells by electroporation.

Table 2. Nucleotide Sequence of Functional Elements and Switch-Specific Testing Primers.

phiC31Int attB(TT)	GTGCGGGTGCCAGGGCGTGCCCTTGGGCTCCCCGGGCGCGTACTCC
phiC31Int attP(TT)	AGTGCCCCAACTGGGGTAACCTTTGAGTTCTCTCAGTTGGGGGCGT
lambdaPR(T41C)a	ATCACCGCAAGGGATAAATATCTAACACCGCGCGTGTTGACTATTTTACCTCTGGCGGTGATAATGGTTGCA
PlacI	GACACCATCGAATGGCGCAAACCTTTCGCGGTATGGCATGATAGCGCCCGGAAGAGAGTCAATTCAGGGTGGTGAAT
oAJ32	CTCGATTCTATTAACAAGGGTATCACC
oAJ91	GTGCGGGTGCCAGGGCGTG
oAJ297	GGGGAGCCCAAAGGTTACC

^aT41C mutation is indicated in bold.

Plasmids were constructed using the CPEC method.⁴²

Plasmid pAJ35 (Figure S1b, File S1) was constructed to express phiC31 integrase and cI857 repressor in the switcher strain. In the plasmid, the DNA sequence of phiC31 integrase²² was amplified from pInt plasmid and placed under the control of a mutated lambda P_R(T41C) promoter (Table 2).²¹ The P_R(T41C) promoter is regulated by the cI857 repressor. A temperature-sensitive mutant of phage lambda repressor cI587²¹ together with the constitutive P_lacI promoter (Table 2) was amplified from the pTcI-Wasabi plasmid. pInt plasmid was a gift from Michele Calos (Addgene plasmid # 18941; http://n2t.net/addgene:18941; RRID:Addgene_18941), and pTcI-Wasabi was a gift from Mikhail Shapiro (Addgene plasmid # 86101; http://n2t.net/addgene:86101; RRID:Addgene_86101).

Control strains carry plasmid pAJ27 (Figure S1b, File S2), which contains all of the same elements as pAJ35, except that phiC31 integrase ORF was truncated to translate only the first 29 amino acids (MDTYAGAYDRQSRERENSSAASPATQRSA).

For the determination of protein synthesis activity, a gene encoding red fluorescent protein mRFP1⁴³ was placed under the control of homoserine-lactone-inducible (HSL-inducible) Lux promoter in plasmid pAJ144 (Figure S1b, File S7). In plasmid pAJ157 (Figure S1b, File S8) the gene encoding the fluorescent protein Crimson was placed under the control of IPTG-inducible tac promoter.

For the expression of SulA protein, the phiC31-integrase encoding sequence in pAJ35 was replaced with the sulA gene from the E. coli MG1655 genome, forming plasmid pAJ181 (Figure S1b, File S9).

Validation of oriC Excision

Cultures of switcher and control strains were grown overnight at 30 °C, diluted 2000 times into 20 mL of fresh LB, and grown at 30 °C or 2 h. Thereafter, the temperature was shifted to 37 °C. Samples were taken hourly to count CFUs per volume of culture (CFU/mL) on LB-agar plates and to analyze the excision of oriC by PCR. Specific primer pairs were used to detect either excised (oAJ297 and oAJ32) or retained (unexcised) (oAJ91 and oAJ32) versions of oriC. The DNA sequence of excision site of oriC was verified by sequencing.

Measurement of Growth and Fluorescence

Cultures of switcher and control strains were grown overnight at 30 °C, diluted into LB as indicated in the figure legends, and grown on a 96-well plate at 30 °C for a further 4 h until the temperature was shifted to 37 °C (timepoint 0 h). Where appropriate, mRFP1 or Crimson synthesis was induced by adding 12 μM N-(3-oxooctanoyl)-l-homoserine lactone (Sigma-Aldrich, O1764) or 100 μM IPTG, respectively (final concentrations), either immediately prior to the temperature shift or at indicated timepoints. The growth curves and mRFP1- or Crimson-related fluorescence (reflecting protein level) were determined using a BioTek Synergy MX plate reader. The optical density at 600 nm and fluorescence intensity (excitation at 584/13.5 nm and emission at 607/13.5 nm to detect mRFP1, or excitation at 611/20 nm and emission at 650/20 nm for detecting Crimson) were measured in a culture volume of 100 μL.

Glucose Uptake Rate Determination

Three independent cultures of strains bAJ85 (control) and bAJ86 (switcher) were grown in LB medium at 30 °C for 4 h and then shifted to 37 °C and grown overnight for 12 h to induce switching. Because of switching, the culture density of bAJ86 strain remains in the range of OD 2.8–3.4, while bAJ85 grows up to OD ≈ 7. Next morning, the cells were pelleted for 5 min at 5000g, resuspended in 5 mL of MOPS medium lacking nitrogen source, pelleted again, and resuspended in 25 mL of MOPS 0.3% glucose medium (to a final cell density of OD600 ≈ 0.2) either supplemented with 9.52 mM NH₄Cl as a nitrogen source (control and switcher) or not supplemented (N-starved control). Antibiotics kanamycin 25 μg/mL and tetracyclin 10 μg/mL were supplemented for selection purposes. Cultures were incubated in shaking flasks at 37 °C, and samples for OD600 and glucose measurements were taken at regular intervals.

Culture samples for glucose measurements were cleared through a 0.2 μm filter, and the filtrates were stored at −20 °C pending analysis. Glucose concentration was measured using high-performance liquid chromatography (Shimadzu Prominence-I LC-2030C 3D plus system) using a Rezex ROA-Organic Acids H+ (8%) LC column 300 × 7.8 mm (00H-0138-K0; Phenomenex) and a LC guard column (03B-0138-K0; Phenomenex). Twenty microlitres of the sample were injected using an autosampler and eluted isocratically with 5 mM H₂SO₄ at 0.6 mL/min for 30 min at 45 °C. Compounds were detected by a refractive index detector (RID-20A; Shimadzu) and identified and quantified using standards in the concentration range of 0.039–20 g/L using the software LabSolution (Shimadzu).

To calculate specific glucose uptake rates, we first determined the specific growth rates as slopes from the linear part of the ln-transformed OD600 values plotted against time. Next, the measured OD600 values were transformed into cell dry weights (gCDW) by following the previously determined coefficient: 1 OD600 unit = 0.40 gCDW.^8,44 The glucose uptake rate was determined as the slope of glucose concentration in the medium plotted against gCDW values; we used 4–7 h timepoints for growing control, 2–12 h timepoints for switched cells, and 3–12 h timepoints for the N-starved control. Specific glucose uptake rate was obtained by multiplying the glucose uptake rate by the specific growth rate.

Fluorescence Microscopy

Samples for microscopy were collected from a 96-well plate reader experiment 5 h after the switch. From each culture, 0.7 μL was pipetted onto a piece of 1% LB-agarose pad. Imaging was performed with a Zeiss Observer Z1 microscope with a 100 ×/1.4 oil immersion objective and a Axiocam 506 mono camera (Zeiss). Phase-contrast and mRFP1-fluorescence (excitation 540–580 nm; emission 615–675 nm) images were recorded.

Time-Lapse Microscopy

Cultures of switcher and control strains were grown overnight at 30 °C, diluted 100 times into 3 mL of fresh LB, and grown at 30 °C for 2 h. From each culture, 0.7 μL was pipetted into a glass bottom dish (Thermo Scientific, 150682) and covered with a piece of approximately 4 × 4 × 1 mm 1% LB-agarose pad. A cover glass was placed on top of each pad, and the dish was covered with a lid to minimize drying. The prepared cells were observed using a Zeiss Observer Z1 microscope with a 100 ×/1.4 oil immersion objective and a Axiocam 506 mono camera (Zeiss). The switch was induced under the microscope from the beginning of imaging by maintaining the temperature of the agarose pad at 37 °C using a Tempcontrol 37–2 digital (PeCon). Images were taken every 2 min for 6 hours. Multiple positions were imaged in one experiment using an automated stage and ZEN software (Zeiss). The signals from GFPmut2 (excitation 460–490 nm; emission 509–550 nm) and mRFP1 (excitation 540–580 nm; emission 615–675 nm), together with phase contrast, were recorded.

Label-Free Proteomic Analysis

Growth Conditions

Triplicate cultures of switcher and control strains were pregrown overnight in sterile MOPS medium⁴⁵ with 0.3% glucose (MOPSglucose) at 30 °C and diluted 800 times into fresh MOPSglucose at the start of the experiment. Cultures were grown for 8 h at 30 °C and thereafter for 28 h at 37 °C. Samples were collected 4–7 h after the temperature shift (growing phase, OD ≈ 1.0), 11–17 h after the temperature shift (early plateau, OD ≈ 2.3–3.0), and 28 h after the temperature shift (late plateau, OD ≈ 2.0–2.8). To collect samples from the switcher low plateau, triplicates of switcher strains were grown overnight in sterile MOPSglucose and diluted 2400 times into fresh MOPSglucose. Cultures were grown for 8 h at 30 °C and then for 11.5 h at 37 °C. Low-plateau samples were collected 11.5 h after the temperature shift. At indicated timepoints, 1 mg dry weight of cells calculated from OD at 600 nm⁴⁶ were harvested by centrifugation at 600 RCF for 7 min. The samples were then washed with phosphate-buffered saline and pelleted by centrifugation again, and the pellets were flash-frozen in liquid nitrogen and stored at −80 °C.

Proteomics Sample Preparation and Nano-LC–MS/MS Analysis

Sample preparation was conducted using the methods described.⁴⁷ Frozen bacterial pellets were thawed and resuspended in lysis buffer (4% sodium dodecyl sulfate, 100 mM Tris, pH 7.5, 10 mM dithiothreitol), heated at 95 °C for 5 min, and sonicated. The protein concentration was determined by tryptophan fluorescence, and 30 μg of total protein was loaded into 30 kDa-cutoff Vivacon 500 ultrafiltration spin columns (Sartorius). Samples were digested for 4 h on a filter with 1:50 Lys-C (Wako) and thereafter overnight with 1:50 proteomics-grade dimethylated trypsin (Sigma-Aldrich) as described for the filter-aided sample preparation protocol.⁴⁸ Peptides were desalted using C18 StageTips,⁴⁹ eluted, dried, and reconstituted in 0.5% trifluoroacetic acid. Nano-liquid chromatography with tandem mass spectrometry (LC–MS/MS) analysis was performed as described previously⁵⁰ using an Ultimate 3000 RSLCnano system (Dionex) and a Q Exactive mass spectrometer (Thermo Fisher Scientific) operating with top-10 data-dependent acquisition.

MS Raw Data Processing

Mass spectrometric raw files were analyzed using MaxQuant software⁵¹ package v. 1.5.6.5 (File S10). Label-free quantification with the MaxQuant LFQ algorithm was enabled with default settings. Methionine oxidation, glutamine/asparagine deamidation, and protein N-terminal acetylation were set as variable modifications, while cysteine carbamidomethylation was defined as a fixed modification. The search was performed against the UniProt (www.uniprot.org) E. coli K12 reference proteome database (September 2015 version) using the tryptic digestion rule (including cleavages after proline). Only identifications of at least seven-amino-acid-long peptides were accepted, and transfer of identifications between runs was enabled. Protein quantification criteria were set to one peptide with a minimum of two MS1 scans per peptide. Peptide-spectrum match and protein false discovery rate (FDR) were kept below 1% using a target-decoy approach. All other parameters were default. Summed peptide peak areas (protein intensities) were normalized using the MaxLFQ algorithm.⁵²

Proteomics Data Analysis

Proteomics data were analyzed using Perseus software⁵³ v. 1.6.15.0. Normalized LFQ intensity values were used as the quantitative measure of protein abundance. Protein identifications classified as “Only identified by site” and “Contaminants” were excluded from further analysis. LFQ intensity values of the whole proteome of each sample were used to conduct PCA and t-SNE analyses. Rtsne package and R plugin for Perseus were used to perform t-SNE analysis in Perseus software with default parameters, except Perplex was set to 4.

Protein LFQ intensity values were log2 transformed,⁵⁴ and normal data distribution was verified from the histogram distribution plots of log2 transformed data for each sample (data not shown). Samples were allocated into groups (two strains, four conditions: control growing, control early plateau, control late plateau, switcher growing, switcher early plateau, switcher late plateau, and switcher low plateau).

For the hierarchical clustering analysis, only proteins with complete data for all 21 samples were included. First, proteins with statistically significant changes in abundance between sample groups were identified using a multiple-sample ANOVA test, with p-values adjusted for multiple testing by the Benjamini–Hochberg permutation FDR at 1%. Statistically significant proteins were subjected to Z-score normalization. Hierarchical clustering analysis using Euclidean distances was performed and presented as a heatmap using Perseus software.

Proteins with three valid values in at least one sample group were used for further analysis. A two-way Student’s t-test was used to compare sample groups (Benjamini–Hochberg FDR < 0.01) (File S10). Of the statistically significant proteins, proteins with more than a 2-fold difference between LFQ intensities (|log2(LFQ a/LFQ b)| > 1) were interpreted as biologically significant.

The list of genes for differentially expressed proteins was used for enrichment analysis. Enrichment analysis for pathways was conducted using the SmartTables function of Pathway Tools v. 19.0 (available at biocyc.org)^55,56 with a threshold of Fisher’s exact test (p < 0.05).

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssynbio.1c00618.

• Schematic maps of switcher constructs and plasmids, sequence verification of the switched cells, protein synthesis capacity of the switcher, SulA-mediated cell elongation does not increase protein production efficiency, consumption of glucose in control and switched cells, and t-SNE analysis of switcher and control samples and distribution of up- and downregulated proteins in late-plateau cells (PDF)

• Sequence of plasmid pAJ35 (TXT)

• Sequence of plasmid pAJ27 (TXT)

• Time-lapse video of control cells (MPG)

• Time-lapse video of switcher cells (MPG)

• List of proteins up- or downregulated in switcher, pathways upregulated in switcher, and list of proteins up- or downregulated in nutrient-depleted stationary phase of control (XLSX)

• Sequence of oriC locus surroundings of bAJ78 (TXT)

• Sequence of plasmid pAJ144 (TXT)

• Sequence of plasmid pAJ157 (TXT)

• Sequence of plasmid pAJ181 (TXT)

• Summary of results of label-free proteomic analysis and source data for Figures 5 and S6 (XLSX)

Author Contributions

A.J. conceived the study. All authors designed and performed experiments and analyzed the data. A.J., M. Kasari, and V.K. wrote the manuscript. All authors read and approved the final version.

The authors declare the following competing financial interest(s): A. Joers, M. Kasari, V. Kasari, and M. Karmas are inventors of the priority patent application (2019175.5) filed by the University of Tartu. The application covers the use of oriC removal for producing products in bacterial cells. A. Joers, M. Kasari, and V. Kasari are co-founders and shareholders of Gearbox Biosciences, a company established to commercialize the technology described in this publication and in the related patent application.

Notes

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE⁵⁷ partner repository with the dataset identifier PXD029931 and 10.6019/PXD029931.