Construction of plasmid-free Escherichia coli for the production of arabitol-free xylitol from corncob hemicellulosic hydrolysate

High costs and low production efficiency are a serious constraint to bio-based xylitol production. For industrial-scale production of xylitol, a plasmid-free Escherichia coli for arabitol-free xylitol production from corncob hemicellulosic hydrolysate has been constructed. Instead of being plasmid and inducer dependent, this strain relied on multiple-copy integration of xylose reductase (XR) genes into the chromosome, where their expression was controlled by the constitutive promoter P43. In addition, to minimize the flux from L-arabinose to arabitol, two strategies including low XR total activity and high selectivity of XR has been adopted. Arabitol was significantly decreased using plasmid-free strain which had lower XR total activity and an eight point-mutations of XR with a 27-fold lower enzyme activity toward L-arabinose was achieved. The plasmid-free strain in conjunction with this mutant XR can completely eliminate arabitol formation in xylitol production. In fed-batch fermentation, this plasmid-free strain produced 143.8 g L−1 xylitol at 1.84 g L−1 h−1 from corncob hemicellulosic hydrolysate. From these results, we conclude that this route by plasmid-free E. coli has potential to become a commercially viable process for xylitol production.

engineered E. coli (CRP mutant) in conjunction with VMCQI was able to eliminate L-arabitol production from a mixture of D-xylose, L-arabinose, and glucose 7 . However, we used plasmid-based VMCQI in E. coli, almost all the L-arabinose in the corncob hemicellulosic hydrolysate has been reduced to L-arabitol in 5L-scale fermentation 2 .
To engineer E. coli for more economical production of arabitol-free xylitol from corncob hemicellulosic hydrolysate, the aim of this study is to perform multi-copy chromosomal integration of N. crassa XR genes controlled by constitutive promoter in order to achieve stable xylitol production. A plasmid that contained an IS5 sequence (one family of insertion sequences), the R6K ori (narrow-host-range replicon) and the N. crassa XR genes driven by a P43 promoter was constructed for chromosomal integration at multiple locations. Moreover, a mutant XR with much lower enzyme activity toward L-arabinose was achieved and a plasmid-free E. coli in conjunction with this mutant was constructed. The synergy manifested as increased selectivity such that L-arabitol formation was completely eliminated in xylitol production from corncob hemicellulosic hydrolysate.

Results and Discussion
Generating plasmid-free strains for economical production of xylitol. The strong constitutive promoter P43 was isolated from Bacillus subtilis 12 and has been studied extensively. However, little result is known about incorporating P43 in E. coli for recombinant protein generation or metabolic engineering. In search of a suitable promoter for more economical production of xylitol, the plasmid pCDF43 with P43 promoter was constructed. After 6 h of cultivation in shake flask, strains HK432 (containing pCDF43) expressed 2758.3 U/L of XR. When cultured in shake flasks, the xylitol productivity of this strain was 0.70 g L −1 h −1 and this productivity was similar with that of HK402 using the Trc promoter 2 . The results indicated that strain containing the P43 promoter was efficient in producing xylitol. Therefore, P43 promoter was considered as a strong candidate for the production of chemicals especially low value added commodity chemicals in E. coli.
Multi-copy integration of the gene of interest into the chromosome is an excellent alternative strategy that overcomes the drawbacks of plasmid-based systems. One system, namely the chemically inducible chromosomal evolution (CIChE) was developed to achieve a plasmid-free, high gene copy expression in E. coli 13 . However, strains obtained from these genomic integration carry many antibiotic resistance markers, which might carry a potential risk of spreading these markers to other microbes in nature and add to a rapid emergence of drug-resistant organisms. Therefore, the BICES (biomass-inducible chromosome-based expression system) was established successfully using a temperature-sensitive and counter-selective plasmid in E. coli for the production of 2,3-butanediol and acetoin 14 . But, a single gene copy in BICES might be responsible for weak expression. To solve this problem, a similar CIChE method was developed using triclosan-induced chromosomal evolution to integrate more gene copies without incorporating plasmid or antibiotic marker 15 . Yet, it is possible that triclosan resistance gene could transfer horizontally and spread triclosan resistance to wild-type bacteria 16 . Most recently, the CRISPR/Cas9 system has been used in E. coli for genome editing 17 , but the gene copy number of chromosomal integration always be single one using CRISPR/Cas9 system in one experiment and large heterologous gene may be not effective using this method 18 .
Thus, based on RecA-assisted recombination, we described a method for multi-copy chromosomal integration in E. coli. Using this method, strains IS5-B and IS5-C were obtained at 400 μ g mL −1 and 800 μ g mL −1 chloramphenicol, respectively (Fig. 1a). After five rounds of integrations, five strains IS5-a, IS5-b, IS5-c, IS5-d, IS5-e without antibiotic marker were obtained (Fig. 1b). The mobile insertion element IS5 is a genetically compact DNA sequence of 1195 bp which was found in variable copy numbers in the genome of E. coli strains with copy numbers vary from 11 in the sequenced E. coli strain MG1655 to 23 in W3110 19 . As integrated loci, insertion sequences were selected on the basis of minimizing unwanted alteration of cellular functions. Our results indicated that the transcription levels of XR genes and the enzyme activity of XR produced by these plasmid-free strains were enhanced after multiple genes integrated into the chromosome, but these parameters were not linearly proportional to the copy numbers of the XR genes (Fig. 2a). In this study, we found that if N. crassa XR genes were integrated next to each other, one of the genes could be deleted by the FLP recombinase and resulted in integration failure. Therefore, primers IS5-check-P1 and IS5-check-P2 (Fig. 1a) were used to verify the decentralized assembly strain (N. crassa XR genes were distributed randomly in the genome of E. coli). In fact, this decentralized assembly strain favors the stability of this strain, since direct repeats of DNA loci are known to be genetically unstable 20 . Effect of gene copy number on xylitol production. To investigate the effect of gene copy number on xylitol production, the strains IS5-a, IS5-b, IS5-c, IS5-d, IS5-e, IS5-B, and IS5-C were cultured separately, and shake flask assays were carried out using each strain. The strong constitutive promoter P43 contributed to high xylitol productivity in E. coli, but xylitol productivity was not linearly proportional to gene copy number (Fig. 2b). Interestingly, it appeared that the multi-copy genes integration into the chromosome, even at low copy number of XR genes like that of 2 or 4, could effectively lead to sufficient accumulation of xylitol.
In addition to gene integration, the method used in this study allows for the insertion of an exact number of copies of target genes using one vector (may be useful for large genes of interest), which is a feature that is particularly crucial when low gene dosage is needed. In fact, with regard to metabolic engineering in bacteria, low-copy plasmids can perform just as well or better than high-copy plasmids 21 and two analogous results have been observed in previous studies, in which the highest level of secreted keratinase was achieved in Bacillus licheniformis containing between 3 and 5 integrated kerA gene, while strain of Lactococcus lactis containing more than 6 copies of the tra904 allele in an intron-GFP cassette exhibited lower levels of keratinase production and reduced GFP expression 22,23 . Therefore, methods that provide direct control over gene copy number are especially advantageous for gene expression in many respects. Stability of the plasmid-free strains. Potential recombination events at the integration loci during fermentation are of concern given the large number of generations (approximately 120 generations). Strains IS5-d3 (t = 60), IS5-d6 (t = 120), IS5-C3 (t = 60) and IS5-C6 (t = 120) were tested for the stability of the integrated genes. Our results showed that the copy number of the integrated genes and xylitol productivity were not significantly decreased in these strains after many generations (Fig. 2). We therefore concluded that the Campbell-type integrations in these strains were relatively stable. Furthermore, we speculate that this method with multi-copy integration described in this study can be applied to other prokaryotic expression systems to achieve stable strains in high-cell-density fermentation, since insertion sequences including various families are widely distributed in many bacteria 24 . High XR total activity has no beneficial effects on xylitol productivity. Strain HK402 was induced with 0.01 and 0.05 mM IPTG respectively, to obtain different levels of XR total activity (highest 17.36 and 92.6 U/mL). Although, the maximum XR total activity of IS5-d (4.63 U/mL) is much lower than that in the plasmid-based system (HK402) with plasmid copy number of about 30, the strain IS5-d produced 110.1 g L −1 xylitol at 3.06 g L −1 h −1 from glucose-xylose mixture in 5 L scale fermentation experiment (Fig. 3a), a productivity that was slightly higher than that containing plasmids (2.4 and 2.48 g L −1 h −1 ) (Fig. 3b,c). With regard to the plasmid-free strain (IS5-d) and the plasmid-based system (HK402) induced with different concentrations of IPTG, results from the fermentation experiment indicated that a high concentration of XR in the pathway had no beneficial effects on the production of xylitol. In agreement with our results, strains with almost 2-fold difference of XR total activity had very similar xylitol production, and increasing IPTG concentration during growth resulted in an increase in XR total activity without an increase in xylitol production 25 . Studies suggested that availability of xylose and enzyme do not significantly limit the xylitol production 26 . Therefore, cofactors (NADPH) supply might limit the reduction of xylose to xylitol and would be accounted for our results. Low XR total activity is beneficial to decreasing arabitol. Strain HK402 gained high XR total activity, but produced about 5% arabitol 2 . As seen from Table 1, strain HK462 with the eight point-mutations of XR on plasmid pTrc99a can eliminate the formation of arabitol in shake flask fermentation, but still can produce a significant amount of arabitol in the bioreactor fermentation from corncob hemicellulosic hydrolysate. It seemed that more arabinose was metabolized in shake flask fermentation compared with the bioreactor using the same strain, and this indicated that low XR total activity might be beneficial to decreasing arabitol. We speculated that ( Fig. 4) arabinose was encompassed by so much XR in the bioreactor, and arabinose was difficult to metabolize by the enzymes responsible for the metabolism of arabinose. Thus, arabinose was reduced to arabitol by XR even with much lower enzyme activity toward L-arabinose. Conversely, arabitol was decreased about 8.8-fold using the plasmid-free strain IS5-d (with much lower XR total activity), and only about 9.6% arabinose was reduced to arabitol (Fig. 5a).
The selectivity of XR is significantly increased by point mutation. Many efforts have been carried out to switch the cofactor specificity of XR, and varying degrees of success were achieved 27 . Meanwhile, previous work suggested that cofactor induced conformational changes of XR might influence the complimentarity between D-xylose and active site, and cofactor binding could provide an additional screening mechanism of XR for recognizing the substrate more specifically 28 . As the conserved Ile-Pro-Lys-Ser motif was responsible for the cofactor specificity of XR 29 , the point mutation of KSN271-273RTT can influence the complimentarity between XR and cofactor. Thus, we speculate that this mutant might influence the selectivity of XR, and as expected the selectivity of RTT for xylose increased about 29.4% (Table 1). Different mutations always have a synergistic effect on enzyme in protein engineering. Combined mutant RTT with the mutant VMQCI which has a 50-fold lower catalytic efficiency toward L-arabinose evolved by Nair 11 , the mutant VMQCIRTT (eight point-mutations) was obtained by point mutations. The selectivity for xylose of VMQCIRTT was increased by 13-fold and 1.22-fold compared with the WT and VMQCI, respectively. Strain (HK462) containing plasmid with VMQCIRTT produced significantly lower amounts of arabitol than strain HK402, but could not eliminate arabitol formation completely (Table 1).
Engineering E. coli for completely eliminating arabitol in xylitol production from corncob hemicellulosic hydrolysate. In this study, we found that high XR total activity had no beneficial effects on xylitol productivity and low XR total activity was beneficial to decreasing arabitol. So a new strain IS5-M with four copies of VMQCIRTT genes was constructed for xylitol production. In batch fermentation (Fig. 5b), the new strain produced  (Fig. 5c). Although many results about xylitol production from hemicellulosic hydrolysate have been reported, most of the xylitol productivity were too low to be industrial-scale production of xylitol 1 . This might be ascribed to the fact that there were still some toxic components in the detoxified hemicellulose hydrolysate which negatively affected the fermentation performance. Lignocellulose pretreatments always go together with the formation of byproducts that inhibit the fermentation process of E. coli and yeast-based systems. It was indicated that detoxification using vacuum evaporation and activated carbon was considered to be an efficient and low-cost procedure to remove the inhibitors for xylitol fermentation 2 . For that the volatile matter such as formic acid, acetic acid and furfural can be detoxified by vacuum evaporation and the non-volatile matter (phenol compounds) can be treated with activated carbon. For more economical production of xylitol, corn steep liquor, glucose from corn and corncob hemicellulosic hydrolysate were used as nitrogen source, co-substrate and substrate, respectively. Thus, this plasmid-free E. coli is a candidate for industrial-scale production of xylitol from corncob hemicellulosic hydrolysate.  Table 2.

Construction of plasmid-free E. coli.
Plasmid containing the constitutive promoter P43 was constructed based on the vector pCDFDuet-1 using primers P43-P1, P43-P2; P43-XR-P1, P43-XR-P2; pcdf-P1, pcdf-P2; and was designated as pCDF43. For construction of plasmid-free E. coli, the plasmid pRC43 containing an IS5 sequence, the R6K ori, a chloramphenicol resistance gene flanked by FRT (Flp recognition) sites and N. crassa XR gene under the P43 promoter, was constructed using primers CM + R6K-P1, CM + R6K-P2; IS5-P1, IS5-P2; pCDF43-P1, pCDF43-P2 for repeated N. crassa XR genes integrations (Fig. 1). Using this plasmid, we have used two methods for chromosomal integration. One of these (Fig. 1a) is ideal for quick generation and selection of multi-copy functional modules by increasing chloramphenicol concentration, while the other (Fig. 1b) is suited for generating precise copies of functional modules through inducing a corresponding number of rounds of chromosomal integration by deleting the selectable marker.
Briefly, the host strain HK401 was made competent for plasmid transformation using calcium, then the plasmid pRC43 was transformed into HK401 by heat-shock for integration through single-crossover Campbell recombination and the integrants were visually screened on plates containing 34 μ g mL −1 chloramphenicol. The  Table 1. The study of mutant XRs in shake flask fermentation except where otherwise noted. a XR activity (U/mg pro) was determined using xylose and arabinose as substrate, respectively. b Selectivity = XR activity for xylose/XR activity for arabinose. c The highest XR activity (U/mL) in the fermentation using xylose as substrate. d Arabitol/Arabinose = arabitol in the production/arabinose in the hemicellulosic hydrolysate. e The experiments were carried out in 5 L-scale bioreactor. To minimize the flux from L-arabinose to arabitol, two strategies including: 1) increasing the specificity of XR; 2) decreasing XR total activity can be adopted.
Scientific RepoRts | 6:26567 | DOI: 10.1038/srep26567 transformed strain was selected and grown to stationary phase in 34 μ g mL −1 chloramphenicol. After one round, the integrant was denominated as IS5-a, and was made competent as the start strain for subsequent integration. To efficiently generate multi-copy function modules (Fig. 1a), plasmid pRC43 was transformed into IS5-a by heat-shock, then the culture (about 100 μ l) was sub-cultured into a new tube in which the chloramphenicol concentration was increased. The culture was allowed to grow to stationary phase and the process was repeated until the desired concentration was reached. The chromosome was expected to develop higher gene copy numbers by recA-dependent homologous recombination. To obtain accurate number of functional modules (Fig. 1b), plasmid pCP20 was transformed into IS5-a by heat-shock to eliminate chloramphenicol resistance with the aid of FLP recombinase produced from pCP20 31 . Then, the strain (without chloramphenicol resistance) was made competent for next round of integration and the process was repeated until the appropriate gene copy number was achieved. Real-time PCR analysis. Total RNA was isolated from strains and purified from genomic DNA using TRIzol reagent (Invitrogen, Carlsbad, CA). The cDNA was prepared from total RNA using AMV reverse transcriptase (Takara, Dalian, China). Quantification of cDNA targets was performed with One Step SYBR ® PrimeScript ™ RT-PCR Kit II (Takara, Dalian, China). A Bio-Rad CFX96 Real-Time PCR detection system (Bio-Rad, Hercules, CA, USA) was used to measure the expression levels of the target gene. CFX Manager Software (Bio-Rad Laboratories) was used for quantification and 16s RNA gene was used as the internal standard. Gene copy numbers in genomic DNA isolated from the appropriate strains using the Genomic DNA purification kits (Axygen) were measured by the above method. Primers QPCR-F, QPCR-R; 16sRNA-F, 16sRNA-R for Real-time PCR are listed in Table 2.  Enzyme activity assay. For enzyme activity assay, cultivated cultures were harvested by centrifugation (10,000 × g, 5 min), washed with potassium phosphate buffer (50 mM, pH 7.4) and disrupted by sonication. Cell debris was removed by centrifugation (10,000 × g) for 10 min and the supernatant was used for enzyme activity assays. For the stability assay of plasmid-free strains, overnight E. coli cultures were diluted to 10 3 colony forming units (cfu)/ml in LB and were grown to stationary phase (approximately 20 generations). Six identical transfers in LB medium were carried out to reach 120 generations (t = 120). At t = 0, t = 60 and t = 120, diluted cultures were plated onto LB agar plates and three colonies were randomly selected from each assay time and then cultured in LB. The cultures were subsequently used as seed cultures for xylitol production.
Effect of XR total activity on xylitol productivity. To investigate the impact of XR total activity on xylitol production, strains HK402 (induced with different concentrations of IPTG to obtain different levels of XR total activity) and IS5-d were used for xylitol production in 5L-scale batch fermentation. The modified M9 minimal medium containing 10 g L −1 peptone and 7 g L −1 yeast extract was used. The seed culture was inoculated (8% v/v) into the fermenter and the fermentation control settings were: 37 °C, stirring speed at 500 rpm, airflow at 0.8 vvm, and pH at 7.0 (maintained with NH 4 OH). When OD 600 reached about 15 (about 6 h), a mixture of xylose and glucose with the final concentrations of 100 g L −1 and 50 g L −1 respectively, supplemented with 20 g peptone, 10 g yeast extract and IPTG (without inducer for IS5-d) were added, then the temperature was decreased to 30 °C. Samples were collected at regular time intervals.
Using plasmid-free strain for xylitol production from corncob hemicellulosic hydrolysate. The corncob hemicellulosic hydrolysate (prepared using dilute acid) which was provided by Huakang Pharmaceutical Co. (Quzhou, Zhejiang province, China) was prepared by the method described in our previous work 2 and was concentrated under vacuum to achieve appropriate concentration of xylose. Batch xylitol production from corncob hemicellulosic hydrolysate by IS5-d was carried out in a 5 L-scale laboratory fermenter containing 3 L modified M9 minimal medium supplemented with 24 g L −1 corn steep liquor and the fermentation control settings were described above. When OD 600 reached about 15, corncob hemicellulosic hydrolysate supplemented with appropriate amount of glucose (for cell growth and NADPH supply) and 30 g corn steep liquor were added.  Increasing the selectivity of XR by point mutation. According to the previous results 27 , the point mutation (KSN271-273RTT) which can increase the activity of XR was carried out based on XR (VMCQI) by the Megaprimer PCR of Whole Plasmid method 32 using primers poit-M1-F and poit-M1-R. Thus, the VMCQIRTT (eight point-mutations) mutant was obtained and verified by DNA sequencing. The point mutation of the integrative plasmid pRC43 was also implemented using the primers poit-M1-F and poit-M1-R and the mutant plasmid was designated as pRC43M. The wild type of XR (WT) was obtained from XR (VMCQI) by point mutation using primers poit-M2-F and poit-M2-R, and the three point-mutations (RTT) was achieved from XR (WT) using primers poit-M1-F and poit-M1-R.
Engineering plasmid-free E. coli for arabitol-free xylitol production. The plasmid-free E. coli IS5-M with four copies of VMCQIRTT was obtained by the same method described above. Batch fermentation for IS5-M was performed according to the method above. Fed-batch fermentation was accomplished in a 15 L-scale bioreactor initially containing 6 L modified M9 minimal medium containing 24 g L −1 corn steep liquor. The culture conditions such as agitation speed, inoculum size, pH, and aeration rate, were the same as those of batch fermentation except that the temperature was controlled at 30 °C all the way. The agitation speed was controlled to maintain the dissolved oxygen