Construction of a plasmid addiction system using grpE as selection marker in Escherichia coli and its application in phloroglucinol biosynthesis

Microbial synthesis of commodity chemicals often be conducted in recombinant plasmid-based expression systems, in which plasmids play pivotal roles on productivity. The recombinant plasmids always encounter instability, leading to losses in product recovery of entire process. To maintain the stability of plasmids, several mechanisms have been evolved. Plasmid addition system, selectively killing plasmid-free cells, is regard as a useful strategy to improve the proportion of plasmid-containing cells. In this study, a novel plasmid addition system using an essential gene grpE that encodes a molecular cochaperone as selection marker, avoiding use of antibiotics, was constructed in Escherichia coli . The solo copy of grpE gene on the chromosome was knocked out and relocated on multicopy plasmids. The generated strains can maintain high ratio of plasmid-harboring cells without antibiotics supplementation in mineral salts media and exhibit improved cell growth and increased tolerance to phloroglucinol. Using this system in phloroglucinol synthesis, it could signicantly increase the phloroglucinol titer from 0.75 g/L to 1.26 g/L, which was further increased to 1.78 g/L when biotin-[acetyl-CoA-carboxylase] ligase BirA was overexpressed. It can be expected that this system will be a powerful tool for microbial manufacture of important chemicals in E . coli . containing appropriate antibiotics or DAP with shaking at 30°C until an OD 600 ≈ 0.6-0.7 was reached. 1.5 mL of the two cultures were collected and mixed by brief vortex. The cells of the mixture were harvested by centrifugation, washed twice with 500 μL of double-distilled water, resuspended in 50 μL of 1% (w/v) DAP solution, spread on a LB plate without any antibiotics, and incubated at 30°C for 18 h. The transconjugants were harvested, washed twice with double-distilled water, spread on LB plates with Cm and Amp and incubated at 30 ℃ overnight to screen for clones that acquired Cm resistance via a single crossover event. Twenty of the growing clones were conrmed by PCR using gck-F/gck-RV as primers, and then replica-plated onto LB agar plates supplemented with Cm and Amp and NaCl-free LB-10% sucrose agar plates supplemented with Amp. Those colonies which were sensitive to Cm and tolerant of sucrose and Amp were further puried on NaCl-free LB-10% sucrose agar plates. Twenty of these colonies were randomly selected and their genotypes were examined by PCR using gck-F/gck-RV as primers. The resulting fragments were sequenced. One of the genotype-conrmed strains was assigned as PGH-0 and was selected for further study.


Introduction
The constantly growing demand for renewable green chemicals and fuels has stimulated great development of synthetic biology. Metabolic engineering of microorganisms by rewiring cells' metabolic ux towards desirable chemicals has been proved to be a promising strategy (Palazzotto et al. 2019;Wang et al. 2018;Westbrook et al. 2019). Microbial syntheses of important biobased chemicals often be conducted in well-established hosts such as Escherichia coli with plasmid-based expression systems , which highly rely on foreign genetic information encrypted in the plasmids (Kroll et al. 2010). Plasmids still supply some of the most versatile systems for the expression of foreign genes, which seriously depend on the plasmid stability and the copy number. Antibiotic resistance selection markers are always introduced to most of frequently used cloning or expression plasmids and antibiotics must be added to the media for plasmid maintenance as well as for easy transfer to the hosts (Lau et al. 2013;Summers 1991). Nevertheless, plasmid instabilities could not be totally avoided (Wang et al. 2020). Plasmid-free cells could decrease the product titer and the pro tability of the whole process (De Gelder et al. 2007;San Millan and MacLean 2017). Moreover, utilization of antibiotics is feasible at the laboratory scale, but are not applicable in the industrial procedures owing to the high costs and environmental issues. Chromosomal gene integration is an alternative used for decades to stabilize foreign genetic information.
However, it often results to low recombinant protein levels or poor product yields compared to the use of (multi-copy) plasmids .
Several mechanisms for plasmids stability have been evolved, one of which, plasmid addiction system (PAS), prevents the survival of plasmid-free cells (Kroll et al. 2011;Tsang 2017;Zielenkiewicz and Ceglowski 2001). PASs can be grouped into three major categories: (i) toxin/antitoxin-based systems, (ii) metabolism-based systems and (iii) operator repressor titration systems (Cranenburgh et al. 2004;Kawano et al. 2007;Kroll et al. 2010;Kroll et al. 2009;Ogura and Hiraga 1983). Various PASs have been successfully constructed. However encountering many problems such as invalidation during prolonged culture growth (Cooper and Heinemann 2000) , survival of plasmid-free cells by assimilating nutrition released from plasmid-containing cells (Gao et al. 2014) and so on. Another strategy of PAS is to knockout essential genes from the chromosomes and restore these genetic de ciencies on plasmids. Recently, two PASs using essential genes ispH and lpxA as selection marker respectively were constructed and were applied in the biosynthesis of different chemicals (Kroll et al. 2011;Zhao et al. 2019). Nevertheless, these genes are based on catabolism genes whose manipulation would cause metabolic imbalance of the host cells. To overcome these problems, an essential gene grpE was chosen for PAS construction. grpE encodes a cochaperone, functioning as the nucleotide exchange factor for the molecular chaperone DnaK and supporting bacterial growth at all temperatures (Ang and Georgopoulos 1989;Harrison et al. 1997). Generally, molecular chaperones protect cellular homeostasis of the host cells by promoting protein folding and disaggregation. It could be hypothesized that grpE expression would be helpful in response to environmental stresses and to biosynthesis of toxic products. Phloroglucinol (PG), a natural phenolic substance, can serve as antioxidant and low toxic smooth muscle antispasmodic (Cahen 1962;Kang et al. 2006;Singh et al. 2009). PG represents the scaffold of more than 700 natural products and has been demonstrated as building block for synthesis of many valuable chemicals traditionally derived from fossil resources including some important energy materials or explosives, such as 2,4,6-trinitro-1,3,5-benzenetriol and 1,3,5-triamino-2,4,6-trinitrobenzene (Hoffman and Fontes 2010;Singh et al. 2010;Wang et al. 2007;Yang and Cao 2012). Traditionally, PG was manufactured by chemical processes which always encountered problems related to safety, environmental pollution or low yield (Singh et al. 2009). Recently, natural PG was detected in the culture extract of Pseudomonas uorescens Pf-5 (Nowak-Thompson et al. 1994) and direct biosynthesis of PG from glucose in E. coli has been achieved by solo expression of phlD, a gene located in the phlACBDE cluster corresponding for the synthesis of 2,4-diacetylphloroglucinol in P. uorescens Pf-5 (Achkar et al. 2005;Nowak-Thompson et al. 1994;Siddiqui and Shahid Shaukat 2003). PhlD is a novel type III polyketide synthase, catalyzing the cyclization of three malonyl-CoA precursors to yield one PG as well as three CO 2 (Zha et al. 2006). Although the titers of PG in E. coli were remarkably enhanced by different genetic and metabolic engineering approaches (Cao et al. 2011;Liu et al. 2016;Rao et al. 2013;Zha et al. 2008;Zhang et al. 2017), the inherent toxicity of PG can enormously inhibit growth and viability of the hosts and limit the titers during bacterial fermentation. Hitherto microbially synthesizing PG are still far away from industrial pro tability.
The main objective of this study was to construct a novel PAS using an essential cochaperone encoding gene grpE as selection marker in E. coli BL21(DE3) and applied in synthesis of biobased chemicals in cheap mineral salts media without antibiotics addition. To verify the e ciency of the novel PAS in excluding competing plasmid-free cells and PG synthesis, PG producing pathway was reconstituted in the novel grpE based PAS and the bacterial growth and PG production was investigated in a cheap mineral salts medium without antibiotics stress.
The grpE knockout cassette (grpE k ) consisted of two DNA segments, upstream (5'-arm) and downstream (3'-arm) of grpE coding sequence respectively, which were ampli ed by common PCR using grpE k -F/grpE k -RV0 and grpE k -F0/greE k -RV as primers and the genomic DNA of E. coli BL21(DE3) as template and was generated by overlap PCR using grpE k -F/grpE k -RV as primers and those two PCR products as template. The grpE expression cassette (grpE e ) and the birA gene were PCR ampli ed from the genomic DNA of E. coli BL21(DE3) using grpE e -F/grpE e -RV and birA-F/birA-RV as primers, respectively. The phlD-marA fragment (pmf) was PCR ampli ed from pET-phlDmarA using pmf-F/pmf-RV as primers. These PCR products were puri ed and ligated to pEasy-Blunt cloning vector to form pEasyB-grpE k , pEasyB-grpE e, pEasyB-birA and pEasyB-pmf respectively, which were con rmed by DNA sequencing. grpE k was inserted into pRE112 to form pRE112-grpE k via SacI/KpnI. A cloning plasmid pUC57S-grpE m carrying mutated grpE expression cassette (grpE m ) in which the codons of grpE were synonymously mutated was purchased from Nanjing GenScript Biotech Co., Ltd. (Nanjing, China). grpE m was inserted into pKD46 via NcoI to generate pKD46-grpE m . pmf was inserted into pA-accADBC to generate pPGN-1 via NotI/A II. grpE e was inserted into pPGN-1 via XhoI/PacI to construct pPGNE-1. Biotin-[acetyl-CoA-carboxylase] ligase coding gene birA was inserted into pPGNE-1 by StuI/NotI to generate pPGNE-2.
All the strains and the plasmids used in this work were listed in Table 1. The oligonucleotides used in this study and the plasmid maps were attached in the Supplemental Table S1 and Supplemental Fig. S1, respectively.

DNA manipulations and transformation
DNA manipulations were carried out using routine methods. E. coli plasmids were puri ed using Monarch™ Plasmid Miniprep Kit (New England Biolabs, USA). PCR products were recovered using Monarch™ DNA Gel Extraction Kit (New England Biolabs, USA) or Monarch™ PCR & DNA Cleanup Kit (New England Biolabs, USA). Restriction endonuclease digestions and DNA ligations were performed according to the manufacturer's instructions. PCR ampli cations were performed in 20 or 50 μL of reaction systems containing 1× PrimeSTAR ® max premix (Takara, China), primers 0.5 μM each, and 1.0 μL of DNA template (about 100 ng). E. coli strains were transformed with plasmid DNA using the CaCl 2 method according to Sambrook et al (Sambrook and Russell 2001), or with minor changes which would be described as necessary.

Construction of ΔgrpE mutant of E. coli BL21(DE3)
Brie y, the grpE knockout strain of E. coli BL21(DE3) was constructed by allelic exchange method (Edwards et al. 1998). E. coli BL21(DE3) and E. coli χ7213 were respectively transformed by pKD46-grpE m and pRE112-grpE k to generate the recipient R-0 and the donor D-0 of conjugation. R-0 and D-0 were separately grown in LB broth containing appropriate antibiotics or DAP with shaking at 30°C until an OD 600 ≈ 0.6-0.7 was reached. 1.5 mL of the two cultures were collected and mixed by brief vortex. The cells of the mixture were harvested by centrifugation, washed twice with 500 μL of double-distilled water, resuspended in 50 μL of 1% (w/v) DAP solution, spread on a LB plate without any antibiotics, and incubated at 30°C for 18 h. The transconjugants were harvested, washed twice with double-distilled water, spread on LB plates with Cm and Amp and incubated at 30℃ overnight to screen for clones that acquired Cm resistance via a single crossover event. Twenty of the growing clones were con rmed by PCR using gck-F/gck-RV as primers, and then replica-plated onto LB agar plates supplemented with Cm and Amp and NaCl-free LB-10% sucrose agar plates supplemented with Amp. Those colonies which were sensitive to Cm and tolerant of sucrose and Amp were further puri ed on NaCl-free LB-10% sucrose agar plates. Twenty of these colonies were randomly selected and their genotypes were examined by PCR using gck-F/gck-RV as primers. The resulting fragments were sequenced. One of the genotype-con rmed strains was assigned as PGH-0 and was selected for further study.

Effects of grpE relocation on plasmid
To investigate the effects of grpE relocation on plasmid, the growth pro les of PGH-0 and BL21(DE3) with or without exogenous PG stress were examined. The selected strains were precultured in 5.0 mL of LB broth without any antibiotics. The precultures were inoculated into 250 mL ba ed Erlenmeyer asks containing 50 mL of mineral salts media supplemented with or without PG at an initial OD 600 = 0.1. Due to the limited solubility of PG in water (1.22 g PG/100 mL water), mother solution of PG was prepared by dissolving PG in dimethyl sulfoxide (DMSO) at a concentration of 250 g/L. 500 μL of PG mother solution was added the medium to make a nal concentration of 2.5 g/L. Fermentation broths supplemented with 500 μL of water or DMSO were used as controls and vehicles, respectively. The cultures were incubated at 30℃ for 36 h and sampled every 4 h, and the cell densities of the cultures were recorded.

Construction of recombinant PG synthesizing strains
The recombinant PG synthesizing strains were constructed by transforming the host strains via the routine CaCl 2 method with minor changes (Sambrook and Russell 2001): (i) the host strains were grown in LB broth at 30℃ before being rendered to competent cells; (ii) the transformants were selected on LB agar plates containing Cm at 42℃ overnight. pKD46 (containing a temperature sensitive R101 replication origin) derived plasmids could not replicate under higher temperature. To con rm the loss of pKD46-grpE m in E. coli PGNE-1 and E. coli PGNE-2 after being incubated at 42℃ overnight, PGNE-1 and PGNE-2 were respectively inoculated into 5.0 mL LB broth containing Amp or Cm and cultured with shaking for 14 h. The growth of the strains under different antibiotic stresses were investigated. PG synthesis in shaking ask cultivation PG synthesis was carried out in 250 mL ba ed Erlenmeyer asks containing 50 mL of mineral salts fermentation media supplemented with or without Cm using PGN-1, PGNE-1, and PGNE-2 as producers.
Single colony of each strain was inoculated in 5.0 mL LB broth overnight. The cultures were used as the seeds. The seed cultures were inoculated to the fermentation medium at an initial OD 600 = 0.1. The cultures were incubated in a gyratory shaker at 37°C and 200 rpm until an OD 600 ≈ 0.7 was reached. A nal concentration of 0.2 mM IPTG was added to the media, and the cells were induced at 30°C for 24 h. Cell density and PG production were monitored during the whole fermentation.

Analytical methods
Cell densities were determined by measuring the absorbance at 600 nm (OD 600 ) with Cary 50 UV-Vis spectrophotometer (Varian, USA). Samples were diluted with the appropriate medium as needed to ensure an absorbance at 0.20-0.80. The concentrations of PG in the media supernatant were quanti ed by cinnamaldehyde colorimetric assay. Brie y, cinnamaldehyde was dissolved in ethanol/HCl (v:v = 3:1) to make nal concentration of 10 mg/L. 5 μL of culture was added to 1 mL of cinnamaldehyde solution, fully mixed and incubated at room temperature for 15 minutes. Absorbance of the solution was measured at 446 nm (OD 446 ) using Cary 50 UV-Vis spectrophotometer. The concentration of phloroglucinol was determined using a standard curve. All the assays in this study were performed in triplet except for those speci cally mentioned.

Results
Construction of expression host with chromatinic grpE disrupted A chromatinic ΔgrpE deletion mutant of BL21(DE3) named PGH-0, harboring supplementary copies of allelic grpE m on the helper plasmid pKD46-grpE m , was constructed by allelic exchange. Considering the high e ciency of gene knockout in E. coli BL21(DE3), allelic exchange by conjugative transfer based on pRE112 suicide plasmid was used to disrupt the grpE gene. pRE112 replication initiation absolutely needs a phage protein Pir absent in most of E. coli strains including BL21(DE3) but present in χ7213 and C118. The principle of the construction is shown in Fig. 1b. A hybrid DNA fragment containing 5'-arm and 3'-arm adjacent to the coding sequence of grpE was obtained by PCR and inserted into pRE112 to generate pRE112-grpE k which would integrate into the genome of recipient strain by single crossover and endue the recipient strain with Cm resistance. Since deletion of the solo copy of grpE on the chromosome is lethal, extra copies of grpE allele must be transferred to the host cells before grpE knockout. The grpE allele grpE m was placed on the temperature sensitive helper plasmid pKD46-grpE m which conferred Amp resistance to the host cells, supported the vitality of the host cells at lower temperature (30℃ for example), and could be cured at 42℃. To avoid the allelic exchanges between linear pRE112-grpE k and pKD46-grpE m , the expression cassette of grpE m on pKD46-grpE m was chemically synthesized with synonymously mutated grpE whose expression was under the control of the promoter from chloramphenicol acetyltransferase gene Cat and the terminator from rnpB to make sure that grpE k fragments contained no similar sequences with pKD46-grpE m . After strain conjugation and the rst round of single crossover, the positive clones could grow on/in media amending with Amp and Cm, and was sensitive to sucrose due to the expression of sacB1 from pRE112. Twenty positive clones were con rmed by PCR (Fig. 1a and 1c) and then replica-plated onto LB agar plates. After second round of single exchange, two kinds of recombinants formed: (i) the positive mutants in which the coding sequences of grpE were knocked out; (ii) and the reverse mutants in which the grpE alleles were back to wild type (Fig.  1b). The positive mutants were screened by PCR ampli cation (Fig. 1d) and con rmed by DNA sequencing. One of the positive mutants was chosen and assigned as PGH-0 which was used for further study. It should be emphasized that PGH-0 harbors the helper plasmid pKD46-grpE m and should be grown preferably at lower temperatures (30℃ or lowers are proposed). grpE relocation on plasmid enhance the tolerance of host to toxic chemicals To examine the effects of grpE relocation on plasmid, the growth pro les of PGH-0 and BL21(DE3) were compared by adding PG solution (in DMSO) to the liquid media at nal concentration of 2.5 g/L of PG and 1% (v/v) of DMSO. It was found that higher stationary OD 600 values were obtained in PGH-0 compared with BL21(DE3) under PG and/or DMSO stresses (Fig. 2), which suggested that PGH-0 carrying multicopy of pKD46-grpEm exhibited increased tolerance to PG and DMSO.
Bacterial PG synthesis was enhanced by PAS In our previous work, PG biosynthesis was achieved in E. coli BL(DE3) harboring two plasmids, pET-phlDmarA and pA-accADBC, in which two kinds of antibiotics (Kan and Cm) must be added to the media for plasmid maintenance during fermentation (Cao et al. 2011). In this study, the functional phlD-marA fragments on pET-phlDmarA were PCR ampli ed and inserted into pA-accADBC to generate pPGN-1, which was transformed into E. coli BL21(DE3) to form PGN-1. Then, the grpE expression cassette (grpE e ) and the birA gene were sequentially inserted into pPGN-1 to generate pPGNE-1 and pPGNE-2, which were respectively transformed into E. coli PGH-0 to generate PGNE-1 and PGNE-2. PGNE-1 and PGNE-2 successfully lost the helper plasmid pKD46-grpE m after be growing at 42℃ for 14 h (Fig. S2). All the strains were grown in the mineral salts broth and were induced by 0.2 mM IPTG for 18 hours.
PGNE-1 and PGNE-2 grow faster and got higher stationary OD 600 and PG titer than PGN-1 whether with or without Cm stress, and even PGNE-2 displayed the best growth pro le and PG biosynthesis (Fig. 3). In terms of the same strain under different Cm stress, all the strains grew faster without Cm stress than those with Cm. Concerning with PG production, PG was produced during exponential phase. The time courses of PG production of these three strains were similar with those of cell growth. PGNE-2 got the highest PG titer, reaching 1.78 g/L PG without Cm stress. Whereas, the PGN-1 with Cm produced 0.75 g/L PG. An increase of 137.3% was achieved. PGNE-1 without Cm stress got a PG titer of 1.26 g/L, higher than that of PGN-1 with Cm stress.

Discussion
With exhaustion of fossil sources, microbial manufacture of important chemicals, biofuels, and biomaterials provides a promising sustainable avenue. Biotechnical processes often be conducted in plasmid-based expression systems which often encounters plasmid instability despite selection stresses were always utilized. To reduce the impacts of plasmid instability, several PASs have been developed. Nevertheless, a few drawbacks still need to overcome. In this study, a novel PAS using an essential gene grpE without downstream metabolic products as selection marker was developed. This PAS contains two components: that is an expression host with chromatinic essential gene disrupted and expression plasmids carrying the corresponding selection marker.
The genome of E. coli contains more than 4200 genes, about 300 of which are essential genes (Baba et al. 2006). To choose an essential gene for construction of the PAS, two major points were taken into consideration: (i) the nal product of this gene can't be easily transferred to other cells, even the host cells are lysed and (ii) overexpression of such a gene would not defect the growth of the host and had better protect the host cells from the toxicity of the metabolites. Based on these considerations, grpE encoding a cochaperone without catalytic synthesis activity to produce downstream chemicals was selected. GrpE plays the role of nucleotide exchange factor in the DnaK-DnaJ-GrpE chaperone system, a family of heatshock proteins, which assists in various cellular processes (Liberek et al. 1991). The innate grpE was a low-level expressed gene. When it was relocated on multicopy plasmid under the control of weak Cat promoter, it might be slightly overexpressed. It can be concluded that relatively low-level overexpression of grpE under weak promoters can increase tolerance to toxic PG and probably improved the productivity of metabolic engineered E. coli.
Generally, most of naturally occurring and arti cially constructed plasmids merely provide advantages to the hosts under speci c environmental conditions such as conferring resistance to antibiotics or utilizing certain substances but rarely concern viability or survival of the host cells. Any plasmids, especially for those arti cially constructed plasmids to complete heterologous metabolic pathway, may in ict an extra metabolic burden to the hosts in comparison to plasmid-free cells which always grow faster, reducing the proportion of product yielding cells and leading to losses of nal product recovery (Zielenkiewicz and Ceglowski 2001). To ensure the plasmids distribute equally in the two descendant cells, several mechanisms have been evolved: (i) site-speci c recombination systems involved in random distribution of high-copy number plasmids to the descendants, (ii) active partition process which precisely distributes low and median copy number plasmids to daughter cells and (iii) PASs that selectively kill the plasmidfree daughter cells (Zielenkiewicz and Ceglowski 2001). Antibiotic-free biosynthesizing processes using PASs are cost-friendly for industrial fermentation. Recently, several PASs have been developed (Kroll et al. 2011;Sánchez-Pascuala et al. 2017;Zhao et al. 2019). Most of them used genes encoding essential enzymes which catalyzed metabolism or catabolism of chemicals which were barely existed in the media. However, these systems are still not applicable in the industrial fermentation of toxic chemicals. The novel PAS developed here increases the tolerance of host to environmental stresses and is much faithfully bene cial for production of toxic chemicals such as PG.
PG is one of the structurally simplest polyketides and is an important platform chemical, which was synthesized by chemical methods traditionally. In 2005, Achkar et al. published the rst report of heterologously microbial synthesis of PG by solo expression of phlD in E. coli (Achkar et al. 2005). PhlD catalyzes three malonyl-CoA to form one PG. From then on, many efforts have been tried to increase the yield and productivity of PG in this bacterium. Overexpression of acetyl-CoA carboxylase complex encoded by accADBC to increase malonyl-CoA supply e ciently increased PG titer (Cao et al. 2011). Rao et al. increased the activity of PhlD by directed evolution of phloroglucinol synthase PhlD and improved the stability for practical PG production (Rao et al. 2013). Recently, PG biosynthesis was established in Arabidopsis thaliana (Abdel-Ghany et al. 2016), leading to accumulation of PG and PG-glucoside in the transgenic lines. However, hitherto the titers of PG still hardly meet the industrial application. This is mainly caused by the instability of the plasmids and the inherent toxicity of PG to the host.
Toxicity symptoms and growth reduction often be associated with the PG producing hosts. Enhancing the tolerance of the hosts to PG as well as other valuable metabolites is a promising strategy for higher titers. Overexpression of MarA, a DNA-binding transcriptional dual regulator involved in multiple antibiotic resistance, and GroESL, another molecular chaperone system, improved the E. coli tolerance to PG and the production as well (Cao et al. 2011;Zhang et al. 2017). Besides metabolic engineering, fermenting optimization are often be used to increase the production. Nutrient composition and fermentation parameters such as feeding strategies, temperature, pH, dissolved oxygen, and other variables signi cantly affect the cell metabolism and metabolite production (Kumar and Shimizu, 2011). Continuous or semi-continuous fermentation has been utilized for decades. In situ extraction has been proved to be a very effective method, which can abate the toxicity of the metabolites, prolong the production time, and improve the titer (Yang et al. 2017). Nevertheless, strain improvements are still the base of microbial synthesis. To increase the stability for PG biosynthesis, we disrupted an essential gene grpE on the chromatin and relocated it on plasmid to generate the rst PG producing strain PGNE-1, in which the PG titer was increased to 1.26 g/L from 0.75 g/L (produced by PGN-1 under antibiotic stress) in shaking ask fermentation. That was further increased to 1.78 g/L after overexpression of birA. BirA is a bifunctional protein that acts as regulator of the biotin operon and possesses biotin ligase activity to AccB, β subunit of acetyl-CoA carboxylase complex. Biotin serves as the prosthetic group covalently attached to a speci c lysyl residue of AccB. BirA-mediated biotinylation of AccB plays essential role for the acetyl-CoA carboxylase complex (Blanchard et al. 1999). The increase of PG titer in E. coli PGNE-2 may result from the improved activity of acetyl-CoA carboxylase complex by the increased biotinylating level of this enzymatic complex by overexpressed active BirA.
In summary, a novel PAS using an essential gene grpE as selection marker was constructed, which could eliminate those plasmid-free cell to increase the hoard-working portion. In this system, the selection marker was overexpressed at a low level, which could improve the cell growth and increase tolerance of the host to toxic metabolites. When this PAS was applied in PG microbial synthesis, it signi cantly increased the PG titers from 0.75 g/L to 1.78 g/L in shaking ask fermentation, an increasing of 137.3%. These results indicate that biosynthesis of toxic PG using grpE based PAS in E. coli is a versatile strategy, which may be expanded to more biobased chemicals.

Declarations
Availability of data and material All data generated or analyzed during this study are included in this published article.
Code availability Not applicable.
Authors' contributions HZ conceived of the study and designed components of the research. JW proposed the idea, performed the experiments, and prepared the manuscript draft. XM and YC analyzed and interpreted the data. All authors read and approved the nal manuscript. Consent for publication All authors declare that this article has not been published elsewhere and agree to publish it in this journal.
Con ict of interest The authors declare that they have no con icts of interests. deoR nupG purB20 φ80dlacZΔM15 Δ(lacZYA-argF)U169 hsdR17() λ -BL21(DE3) Initial strain; F -ompT gal dcm lon hsdS B () λ(DE3) Invitrogen  One hundred of clones were PCR examined. These two clones were tiny part of them. Lanes 3 and 4 were the controls using BL21(DE3) genomic DNA as template. (d) Agarose electrophoresis photos of the PCR products from the recombinant cells after the second round of single crossover using gck-F/gck-RV as primers. The PCR products were predicted to be 1300 bps long. The results were consistent with the prediction.

Figure 2
Time course of BL21(DE3) and PGH-0 growth under different toxic stresses. PGH-0 carrying multicopy grpE alleles relocated on plasmid obtained higher cell densities with or without toxic stress. Error bars indicate standard deviations (n = 3).

Figure 3
Microbial synthesis of PG was enhanced by the novel PAS. PGNE-2 and PGNE-1 grew faster and synthesized more PG than PGN-1 did. PGNE-2 and PGNE-1 produced 1.78 g/L and 1.26 g/L of PG without antibiotic stress respectively, while PGN-1 got an PG titer of 0.75 g/L under Cm stress. Error bars indicate standard deviations (n = 3).