Enhancing Production of Pinene in Escherichia coli by Using a Combination of Tolerance, Evolution, and Modular Co-culture Engineering

α-Pinene is a natural and active monoterpene, which is widely used as a flavoring agent and in fragrances, pharmaceuticals, and biofuels. Although it has been successfully produced by genetically engineered microorganisms, the production level of pinene is much lower than that of hemiterpene (isoprene) and sesquiterpenes (farnesene) to date. We first improved pinene tolerance to 2.0% and pinene production by adaptive laboratory evolution after atmospheric and room temperature plasma (ARTP) mutagenesis and overexpression of the efflux pump to obtain the pinene tolerant strain Escherichia coli YZFP, which is resistant to fosmidomycin. Through error-prone PCR and DNA shuffling, we isolated an Abies grandis geranyl pyrophosphate synthase variant that outperformed the wild-type enzyme. To balance the expression of multiple genes, a tunable intergenic region (TIGR) was inserted between A. grandis GPPSD90G/L175P and Pinus taeda Pt1Q457L. In an effort to improve the production, an E. coli-E. coli modular co-culture system was engineered to modularize the heterologous mevalonate (MEV) pathway and the TIGR-mediated gene cluster of A. grandis GPPSD90G/L175P and P. taeda Pt1Q457L. Specifically, the MEV pathway and the TIGR-mediated gene cluster were integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then evolved to a higher gene copy number by chemically induced chromosomal evolution, respectively. The best E. coli-E. coli co-culture system of fermentation was found to improve pinene production by 1.9-fold compared to the mono-culture approach. The E. coli-E. coli modular co-culture system of whole-cell biocatalysis further improved pinene production to 166.5 mg/L.

However, the production level of pinene is much lower than that of hemiterpene (isoprene) (Whited et al., 2010) and sesquiterpenes (farnesene) (Zhu et al., 2014) to date. Pinene is highly toxic to E. coli. E. coli growth is inhibited by 0.5% pinene (Dunlop et al., 2011). The inherent tolerance of E. coli may limit the production potential. It was demonstrated that increasing the tolerance of E. coli by overexpressing the efflux pump AcrBDFa (YP_692684) from Alcanivorax borkumensis significantly enhanced limonene production (Dunlop et al., 2011). Another reason for the lower yield may be that PS has a lower expression level and/or lower enzymatic activity in E. coli. Thus, we first combined tolerance engineering with directed evolution of the enzyme to improve pinene production in E. coli.
Recently, there has emerged a new modular co-culture engineering approach for engineering microorganisms. Modular co-culture engineering approaches divide a complete biosynthetic pathway into separate serial modules, which are introduced into different strains to accommodate individual modules for achieving designed biosynthesis (Zhang and Wang, 2016). The advantages of using modular co-culture engineering include the following: (1) reducing the metabolic burden on each host strain; (2) providing diversified cellular environments for functional expression of the different pathway genes; (3) reducing the undesired interference of different pathways; (4) easily balancing the biosynthetic pathway between individual pathway modules by simply changing the strain-to-strain ratio; (5) high-efficiency utilization of complex materials containing multiple active substrates; and (6) supporting the plug-and-play biosynthesis of various target products (Zhang and Wang, 2016). Thus, modular co-culture engineering was also used to enhance pinene production in E. coli.

Strains, Plasmids, and Primers
The bacterial strains and plasmids used in this study are listed in Table 1 Genetic Methods pMEVI was derived from pJBEI-6409 (Alonso- , which was obtained from Addgene. pJBEI-6409 contains six genes of the MEV pathway (atoB from E. coli, HMGS, and HMGR from Staphylococcus aureus and MK, PMK, and PMD from Saccharomyces cerevisiae, idi from E. coli, GPPS from A. grandis, and limonene synthase gene (LS) from Mentha spicata). The GPPS-LS gene cluster was removed from pJBEI-6409 to obtain pMEVI. The fusion gene cluster of the codon-optimized GPPS and PS from A. grandis with a (GSG) 2 linker was synthesized by Suzhou GENEWIZ, Inc. (Suzhou, China) and ligated into pQE30 to obtain pQE-GPPS-L-PS. The GPPS-PS gene cluster from pQE-GPPS-L-PS was inserted into the BamHI/XhoI sites of pMEVI to obtain pMEVIGPS. The evolved codon-optimized Pt1 (Pt1 Q457L ) from P. taeda was synthesized by Suzhou GENEWIZ, Inc. (Suzhou, China) and ligated into pQE30 to obtain pQE-Pt1 Q457L . The PS gene of pQE-GPPS-L-PS DNA shuffling was replaced with the Pt1 Q457L gene to obtain pQE-GPPS MUT -L-Pt1 Q457L .
The TIGR-mediated GPPS MUT -Pt1 Q457L gene cluster was cut from pQE-GPPS MUT -TIGR-Pt1 Q457L with EcoRI/HindIII and then cloned into EcoRI/HindIII-digested pHKKF3T5b to obtain pHKKF3T5b-GPPS MUT -TIGR-Pt1 Q457L . The MEVI operon was cut from pMEVI with EcoRI/XhoI and then cloned into EcoRI/SalI-digested pP21KF3T5b to obtain pP21KF3T5b-MEVI. Chromosomal integration was carried out by direct transformation as described by Chen et al. (2013). Chemically induced chromosomal evolution (CIChE) of the above construct was carried out by subculturing the resulting strains in 5 mL Super Optimal Broth (SOB) medium with increasing concentrations of triclosan in 15 mL culture tubes, as described by Chen et al. (2013). The strains were grown to the stationary phase in 1 µM triclosan for pP21KF3T5b-GPPS MUT -TIGR-Pt1 Q457L or 0.25 µM for pHKKF3T5b-MEVI. Fifty milliliters of the culture were subcultured into a new culture tube, in which the triclosan concentration was doubled to 132 or 32 µM and allowed to grow to the stationary phase. The process was repeated until the desired concentration was reached.
The recA gene of the CIChE strain was then deleted by the markerless deletion approach using the isopropyl β-D-1-thiogalactopyranoside (IPTG)-inducible ccdB as a counterselectable marker (Wei et al., 2016). Gene replacement of the native promoter of E. coli acrAB and the integration of ttgB from P. putida KT2440 were carried out by the CRISPR-Cas method as described by Jiang et al. (2015). To enhance specificity and reduce off-target effects, the cas9 on pCas (Jiang et al., 2015) was site-directed mutated into cas9(K848A/K1003A/R1060A) as described as Slaymaker et al. (2016) to obtain pCas * . To easily assemble the sgRNA sequence using the BglBrick standard method, the BglII site in the sgRNA plasmid pTargetF was first removed, and then a BglII site was added in the front of EcoRI site to obtain the sgRNA plasmid pTargetB.

Adaptive Laboratory Evolution for Improving Pinene Tolerance
A 1-mL culture of logarithmic phase E. coli was collected by centrifugation, washed twice with saline, and diluted to a cell concentration of 10 6 to 10 7 with physiological saline. Then, atmospheric and room temperature plasma (ARTP) mutagenesis was performed using an ARTP mutation system (ARTP-IIS, Tmaxtree Biotechnology Co, Ltd, Wuxi, China) with the following parameters: (1) the radio frequency power input was 100 W; (2) the flow of pure helium was 10 standard liters per min; (3) the distance between the plasma torch nozzle exit and the slide was 2 mm; and (4) the different treatment times were selected (10, 20, 40, 60, 80, 100, and 120 s). Ten microliters of the aforementioned cell dilution were evenly scattered on the slide and subjected to ARTP mutagenesis. After treatment, the slide was washed with LB medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl), transferred to 5 mL of LB medium with 0.5% pinene in a 15 mL falcon tube, and cultivated at 30 • C and 200 rpm for 24 h. The cultures were serially passed into fresh medium (initial OD 600 of 0.2) daily. Continuously repeating this transfer procedure at 0.5% pinene until OD 600 at 24 h did not increase further, the culture was then sequentially transferred to a pinene concentration of 1.0%, 1.5% and 2.0%. The cultures were frozen and stored at −80 • C at every pinene concentration.
The cultures of 2.0% pinene stored at −80 • C were transferred by the IPP/FPP sensor plasmid pP rstA -GFP. Single colonies were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 600 µL of LB medium and incubated at 30 • C and 200 rpm for 24 h on a Multitron shaker (Infors). The cells were harvested by centrifugation at 14000 × g for 2 min and then resuspended with 0.6 mL (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , pH 7.4). Then, 200 µL of the bacterial culture was transferred into a 96-well plate in which the OD 600 and fluorescence were read with the excitation at 485 nm and emission at 528 nm using a SynergyNeo2 multi-mode reader (SynergyNeo2, BioTek, USA).

Generating Random Mutagenesis Libraries Using Error-Prone PCR and Screening
The random mutagenesis libraries of the fusion gene cluster of AgGPPS-AgPS after optimization of the first 18 codons using the 6AA method (Boë et al., 2016) were constructed through errorprone PCR. The gene cluster of AgGPPS-AgPS was amplified from pQE-GPPS 6AA -L-PS using the primers EcoRI-GPPS/HindIII-PS. The error-prone PCR reaction mixture consisted of 5 mM MgCl 2 , 0.3 mM MnCl 2 , 0.2 mM each of dATP and dGTP, 1 mM each of dCTP and dTTP and Tag DNA polymerase. The PCR product was digested by EcoRI/HindIII, ligated into the EcoRI/HindIII sites of pQE30, and then transferred into the lycopene-producing strain E. coli LYCOP to generate the mutant library.
The mutant library was plated on LB agar with ampicillin and IPTG. The plates were incubated at 30 • C overnight. The mutant plasmid was isolated from the whiter colony and then transferred into component E. coli BW25113 (P T5 -dxs, pMEVI). The pinene productions of them were analyzed in a shake flask.

Generating Random Mutagenesis Libraries Using DNA Shuffling and Screening
DNA shuffling experiments were performed by the following steps: parental template preparation, DNase I digestion, primerless PCR and PCR with primers. The mutant plasmids from the 7 colonies resulted from error-prone PCR and were used as the template to amplify the gene cluster fragments with the primers EcoRI-GPPS/HindIII-PS. Following purification, 2 µg of the eight PCR products was mixed and treated with 0.02 U of DNaseI in 100 µL of the 10 × DNaseI buffer on ice for 2 min and terminated by the loading buffer containing SDS. The purified fragments of 50-300 bp were used in the primer-less PCR reactions to reassemble into full-length genes. The primerless PCR reaction mixture contained 0.5 mM each dNTP, 10 × Taq buffer and 0.5 µL Taq DNA polymerase (Takara). The PCR reaction conditions were as follows: 95 • C for 1 min, 35 cycles of 94 • C for 30 s, 45 • C for 30 s, 72 • C for 3 min, and final incubation at 72 • C for 8 min. The PCR products with the correct size were purified and subjected to PCR amplification using the same conditions with the primers EcoRI-GPPS/HindIII-PS. Finally, the mutated PCR products of the full-length gene were digested by EcoRI/HindIII, ligated into the EcoRI/HindIII sites of pQE30, and transferred into the lycopene-producing strain E. coli LYCOP to generate the mutant library.
The mutant library was plated on LB agar with ampicillin and IPTG. The plates were incubated at 30 • C overnight. Single colonies with a whiter color were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 1 mL of LB medium and incubated at 30 • C and 200 rpm on a Multitron shaker (Infors). After 8 h, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30 • C and 200 rpm for 48 h. The pinene concentration in individual wells was assayed using the concentrated sulfuric acid method as follows. One hundred microliters of the dodecane layer were mixed with 200 µL sulfuric acid, then inoculated for 5 min in boiling water, and the absorbance of the reaction solution at 450 nm was determined using a spectrophotometer (Shimadzu, Japan).

Creating TIGR Libraries and Screening
TIGRs were synthesized using PCR to assemble the oligonucleotides into chimeric DNA sequences as described by Pfleger et al. (2006) and Li et al. (2015). Briefly, 40 mmols of an equimolar oligonucleotide (A, B, C, and D in Supplemental Table 1) mixture was added to a mixture containing 2.5 units of Primer Star DNA Polymerase (Takara, Dalian, China). The assembly was conducted over 35 cycles of PCR for 10 s at 98 • C, 30 s at 72 • C, and 20 + 5 s/cycle at 72 • C. The assembly products were purified using a nucleotide removal column and amplified using the end-specific primers TIGRs-F(X)/TIGRs-R(A) and then cloned into the SacI/SalI sites of pQE-GPPS MUT -Pt1 Q457L to obtain the plasmid libraries pQE-GPPS MUT -TIGRs-Pt1 Q457L . The plasmid libraries were transferred into component E. coli BW25113 (P T5 -dxs, pMEVI) to generate the mutant library.
The TIGR library was plated on LB agar with ampicillin. The plates were incubated at 30 • C overnight. Single colonies were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 1 mL of LB medium and incubated at 30 • C and 200 rpm on a Multitron shaker (Infors). After 8 h, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30 • C and 200 rpm for 48 h. The pinene concentration in individual wells was assayed using the above concentrated sulfuric acid method.

Pinene Biosynthesis in Shake Flasks
For pinene fermentation production, a single colony was inoculated into 5 mL of LB medium in a falcon tube, which was cultured overnight at 37 • C. The overnight seed culture was then inoculated into 50 mL of SBMSN medium with a starting OD 600 of 0.1. SBMSN medium (pH 7.0) containing the following (g/L): sucrose 20, peptone 12, yeast extract 24, KH 2 PO 4 1.7, K 2 HPO 4 211.42, MgCl 2 ·6H 2 O 1, ammonium oxalate 1.42, and Tween-80 2. The main cultures were then incubated at 37 • C and 200 rpm until an OD 600 of 0.8 was reached. Then, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30 • C and 130 rpm for 72 h.
Co-culture of E. coli PINE and MEVI for Pinene Production E. coli PINE and MEVI cells were first separately grown in 5 mL SBMSN medium in a falcon tube at 37 • C overnight. The overnight culture was inoculated into 50 mL of SBMSN medium with a starting OD 600 of 0.1 and incubated at 37 • C and 200 rpm until an OD 600 of approximately 6.0 was reached. The cultures were then incubated at 20 • C and 200 rpm for 16 h. For pinene biosynthesis using co-cultures, the E. coli PINE culture and the desired amount of the E. coli MEVI culture were inoculated into the 30 mL SBMSN medium with a starting OD 600 of 0.1. The mixed culture was culture at 37 • C and 200 rpm until an OD 600 of 0.8 was reached. Then, the cultures were overlaid with 20% dodecane to trap pinene, and were incubated at 30 • C and 130 rpm for 72 h.

Whole-Cell Biocatalysis for Pinene Production
A single colony of E. coli PINE and MEVI was separately inoculated into 5 mL of SBMSN medium in a falcon tube, which was cultured overnight at 37 • C. The overnight cultures were then inoculated into 50 mL SBMSN medium with a starting OD 600 of 0.1. The cultures were then incubated at 37 • C and 200 rpm until an OD 600 of approximately 6.0 was reached. Then, the cultures were incubated at 20 • C and 200 rpm for 16 h. Finally, the E. coli PINE culture was mixed with the E. coli MEVI culture at the inoculation ratio of 2:1. The mixed cells were harvested by centrifugation (6000 × g at 4 • C) and washed twice with cooled phosphate buffer (0.1 M, pH 7.0).
For biocatalysis, the above cells were resuspended in 10 mL phosphate buffer (0.1 M, pH 7.0) containing 20 g/L of sucrose, 10 mM MgCl 2 and 5 mM MnCl 2 to form the cell suspension (OD 600 = 30). The reaction mixture was overlaid with 20% dodecane. The catalysis was performed for 28 h at 30 • C and 130 rpm.

GC Analysis
Five hundred microliters of the dodecane layer was placed in a 1.5-mL microcentrifuge tube and centrifuged at 25,000 × g for 1 min, and 50 µL of dodecane was diluted in 450 µL of ethyl acetate spiked with the internal standard limonene (10 µg/L). The samples were analyzed by GC-FID by using a standard curve of α-pinene (Sigma Aldrich). The GC-FID (Techcomp GC7900, Techcomp Ltd, China) was used with a TM-5 column (30 m × 0.32 mm × 0.50 µm). The inlet temperature was set to 300 • C, with the flow at 1 mL/min, the oven at 50 • C for 30 s, ramp at 4 • C/min to 70 • C, and ramp at 25 • C/min to 240 • C.

Quantitative Real-Time PCR (qRT-PCR)
The total RNA from E. coli cells grown for 24 h in shake flasks was isolated using an RNA extraction kit (Dongsheng Biotech, Guangzhou, China), following the manufacturer's instructions. The first-strand cDNA was synthesized using an All-in-One TM First-Strand cDNA Synthesis kit (GeneCopoeia, Guangzhou, China). The qRT-PCR was perfor1med with the All-in-One TM qPCR Mix kit (GeneCopoeia) on an iCycler iQ5 Real Time PCR system (Bio-Rad Laboratories, California, USA). The template was 100 ng of cDNA. The PCR conditions were as follows: 95 • C for 10 min, followed by 45 cycles of denaturation at 95 • C for 10 s, annealing at 60 • C for 20 s, and extension at 72 • C for 15 s. The primers for qRT-PCR are presented in Supplementary Table 1. The data were analyzed by the 2 − Ct method described by Livak and Schmittgen (2001) and normalized by cysG gene expression.
Gene copy numbers were measured by qPCR on genomic DNA isolated from the appropriate CIChE strains. qPCR was performed as described above. The primers QPt1F/QPt1R and QHF/QHR (Supplementary Table 1) were used to measure the copy number of Pt1 and HMGS, respectively.

Statistical Analysis
All experiments were conducted in triplicate, and the data were averaged and presented as the means ± standard deviation. Oneway analysis of variance followed by Tukey's test was used to determine significant differences using the OriginPro (version 7.5) package. Statistical significance was defined as p < 0.05.

Tolerance Engineering to Improve Pinene Production
To improve pinene tolerance, E. coli cells harboring pP rstA -GFP were treated with ARTP and then serially transferred into LB medium supplemented with increased concentrations of pinene of 0.5, 1.0, 1.5, and 2.0%. The culture was transferred daily. After the adaptive evolution at 2.0% pinene, the culture was streaked on LB plates for isolated colonies. It has been demonstrated that the IPP/FPP sensor plasmid pP rstA -GFP has been successfully used to test the intracellular IPP/FPP concentration and to screen the library with higher IPP/FPP concentrations (Dahl et al., 2013;Shen et al., 2016). Thus, we also used it to screen the library. Of the 670 clones, 14 strains with higher fluorescence strength (Supplementary Figure 1) were selected for further shake flask analysis. As shown in Figure 1, E. coli YZ-3 produced the highest level of pinene (7.3 ± 0.2 mg/L), which was 31% higher than the starting strain E. coli BW25113 (P T5 -dxs).
To improve pinene production, we investigated effects of efflux pumps on pinene production. Dunlop et al. (2011) reported that expressing some efflux pumps significantly improved pinene tolerance. Thus, we tested whether pumps that improved pinene tolerance also enhanced its production. As shown in Figure 2A, expressing native AcrB, AcrAB, or TtgB (NP_743544) from Pseudomonas putida KT2440 in E. coli YZ-3 using plasmid FIGURE 1 | Pinene production by the selected adaptive laboratory evolution strains harboring pMEVIGPS. E. coli BW25113 (P T5 -dxs, pMEVIGPS) was set as the control strain (CK). The data represent the means of three replicates and error bars represent standard deviations. resulted in increased pinene production. However, expressing A. borkumensis AcrBDFa (YP_692684) or P. putida KT2440 MexF (NP_745564) from did not improve pinene production. Therefore, we first replaced the native promoter of E. coli YZ-3 acrAB operon with the strong P37 promoter to obtain E. coli YZ-3-A, resulting in an increase in pinene production to 8.1 ± 0.2 mg/L from 7.3 ± 0.2 mg/L ( Figure 2B). Then, we integrated the ttgB from P. putida KT2440 under the control of the P37 promoter in E. coli YZ-3-A to obtain E. coli YZ-3-A-T. The modification further improved pinene production to 9.1 ± 0.2 mg/L ( Figure 2B). These results indicate that overexpressing some efflux pumps (E. coli acrAB and Pseudomonas putida KT2440 ttgB), whcih improved pinene tolerance, also enhanced its production.
To further improve pinene production, we isolated a mutant resistant to an inhibitor of biosynthetic pathway after ARTP mutagenesis. Isolating a mutant resistant to an inhibitor of biosysnthetic pathway is a common strategy used for strain improvement. In E. coli, the important precursors IPP and DMAPP are produced by the 1-deoxy-D-xylulose-5-phosphate (DXP) pathway. Fosmidomycin is the DXP pathway inhibitor that inhibits 1-deoxy-D-xylulose-5-phosphate reductoisomerase (Dxr) and methylerythritol phosphate cytidyltransferase (IspD) of the DXP pathway . Genes involved in the DXP pathway are essential for E. coli growth. The wildtype E. coli YZ-3-A-T can grow in the presence of 2% pinene (Supplementary Figure 2A), but does not grow in the presence of 35 µM fosmidomycin (Supplementary Figure 2B). After ARTP mutagenesis, cells grow well in the presence of 35 µM fosmidomycin (Supplementary Figure 2C). Overexpression of dxr or ispD in E. coli improved the fosmidomycin tolerance . This indicates that the fosmidomycin resistant mutants may show higher level of Dxr and IspD. Screening the fosmidomycin resistant mutants will increase the probability to obtain a mutant with higher IPP flux. Thus, to increase the probability to obtain a mutant with higher IPP flux, we screened the fosmidomycin resistant mutants using the IPP/FPP sensor. E. coli YZ-3-A-T cells harboring pP rstA -GFP were treated with ARTP. After ARTP mutagenesis, the cells were transferred into the LB medium supplemented with 35 µM fosmidomycin and 2.0% pinene. A total of 720 clones were screened for analyzing fluorescence strength in deepwell microplate cultures (Supplementary Figure 3). Twenty-one strains with higher fluorescence strength were selected for further shake flask analysis. As shown in Figure 3, Strain No. 19, which was denoted as E. coli YZFP, produced the highest level of pinene, which reached 9.9 ± 0.1 mg/L. In fact, our quantitative real-time PCR analysis also demonstrates that the dxs, dxr and ispD of the DXP pathway in E. coli YZFP showed higher transcription level than the wild-type strain (Data not shown, will be published in another paper).
To characterize the pinene tolerance, the growth of the above strains were compared in different concentrations of pinene. Figure 4A shows the growths of these strains in the presence of 2% pinene. The starting strain did not grow well in the presence of 2% pinene. The above engineered strains did grow well in the presence of 2% pinene. The maximum cell densities of the  three engineered strains were similar. The growth rate of E. coli YZFP was higher than that of the other engineered strains. These results indicate that the engineered strains have higher pinene tolerance than the starting strain. However, the maximum cell densities of the three engineered strains were lower than that of the starting strain in the absence of pinene ( Figure 4B). The reason may be that the three engineered strains produced higher level of IPP than the starting strain. IPP is toxicity to E. coli. We also investigated the genetic stability of E. coli YZFP. The strain can also grow well in the presence of 2% pinene and the level of pinene production remained constant after 20 rounds of subculturing in absence of selective pressure (data not shown).

Evolution Engineering to Improve Pinene Production
The lower expression level and/or lower enzymatic activity of GPPS and PS in E. coli may result in the lower yield of pinene production. Sarria et al. (2014) compared GPPSs and PSs from A. grandis and P. taeda and found that the combination of GPPS and PS from A. grandis was most suitable for pinene production. Thus, we first optimized the first 48 nucleotide sequences of A. grandis GPPS with the 6AA method to increase the expression level of the A. grandis GPPS-PS gene cluster in E. coli. The 6AA method substitutes all Arg, Asp, Gln, Glu, His, and Ile codons with the synonymous codon having the highest single-variable logistic regression slope (CGT, GAT, CAA, GAA, CAT, and ATT, respectively), while the other 14 amino acids were not changed from the wild-type gene sequence (Boë et al., 2016). The 6AA optimization increased pinene production from 5.6 ± 0.1 mg/L to 6.4 ± 0.3 mg/L ( Table 2).
Because pinene shares the same 5-carbon precursors IPP and DMAPP with carotenoids, a lycopene-producing strain E. coli LYCOP (Chen et al., 2013) was used to screen the error-prone  PCR mutant libraries of the GPPS-PS cluster from A. grandis after 6AA optimization. The higher the activity of the GPPS-PS cluster, the lower the intracellular precursor levels for lycopene biosynthesis, thereby reducing the pigmentation of the E. coli. Of approximately 1 500 colonies, 7 colonies with a whiter color were observed. The mutant plasmids were isolated from the 7 colonies and then were co-transferred with the MEV pathway plasmid pMEVI into E. coli BW25113 (P T5 -dxs). The pinene productions of them were analyzed in a shake flask, and the results are presented in Figure 5A. The strains harboring the mutant gene cluster produced higher pinene by 7.8-85.7% than that with the wild-type gene cluster. To increase the gene cluster activity, the 7 mutant gene clusters were used for DNA shuffling.
Because the colonies of E. coli LYCOP harboring the above mutant gene cluster became a faint color, it is difficult to discriminate these colonies by using the above carotenoid-based method. A more sensitive and quantitative screening method is needed. It is known that monoterpene can hydrate readily in the presence of acid catalysts, such as H 2 SO 4 (Robles-Dutenhefner et al., 2001). As a result, the initial reaction solutions turn yellow and then brown. After reaction with concentrated sulfuric acid in boiling water for 5 min, it was observed that the color of the reaction solution become darker as the pinene concentration increases and the absorbance at 450 nm is linearly related with pinene concentration (Supplementary Figure 4). Thus, the concentrated sulfuric acid method can quantitatively predict pinene concentrations.
After DNA shuffling, the mutant plasmids were transferred into E. coli LYCOP harboring pMEVI. Fifty colonies with a whiter color were used for assays of pinene production in a shake flask using the concentrated sulfuric acid method. The results are presented in Figure 5B. E. coli LYCOP harboring the mutant gene cluster produced higher pinene (6.5-10.1 mg/L) than those with the wild-type gene cluster. The mutant plasmid with the highest pinene production was isolated from strain No. 15 and then was co-transferred with pMEVI into E. coli BW25113 (P T5dxs). E. coli BW25113 (P T5 -dxs) harboring the mutant plasmid and pMEVI produced 12.4 ± 0.2 mg/L of pinene ( Table 2). The mutant plasmid with the highest pinene production was then sequenced. The two amino acid mutants (D90G and L175P) were observed in the CDS of GPPS from A. grandis. No mutant in the CDS of PS from A. grandis was observed. It has been reported that (-)-α-pinene synthase (Pt1) from P. taeda has the lowest K m for GPP among the known PSs (Phillips et al., 2003). Tashiro et al. engineered an E. coli with the highest yield of pinene so far using the pinene synthase mutant (Pt1 Q457L ) from P. taeda (Tashiro et al., 2016). Thus, replacing A. grandis PS in the mutant plasmid with P. taeda pinene synthase mutant gene (Pt1 Q457L ) yielded pQE30-GPPS mut -L-Pt1 Q457L . E. coli BW25113 (P T5 -dxs) harboring pQE30-GPPS mut -L-Pt1 Q457L produced a higher level of pinene (15.2 ± 0.2 mg/L) than those harboring pQE30-GPPS mut -L-AgPS, which achieved 12.4 ± 0.2 mg/L ( Table 2).
The unbalanced expression of multiple genes may overburden the cell and cause accumulation of toxic metabolic intermediates, resulting in reduced product titers. Pfleger et al. (2006) developed a combinatorial engineering approach for coordinating the expression of cascade enzymes. For this purpose, libraries of tunable intergenic regions (TIGRs) are generated that encode mRNAs with diverse secondary structures with RNase cleavage sites. The TIGR approach was applied to balance the gene expression of the MEV pathway using the TIGR approach, resulting in a 7-fold increase in mevalonate production.
Moreover, our previous paper demonstrated that the TIGR approach was more efficient compared to protein fusion for coordinating expression . Thus, we constructed a library of TIGRs to balance the expression of A. grandis GPPS D90G/L175P and P. taeda Pt1 Q457L . The library of TIGRS was inserted between GPPS D90G/L175P and P. taeda Pt1 Q457L to yield a series of operons. The functional operons from the libraries were screened by using the concentrated sulfuric acid method. A total of 768 colonies were used for the assay of pinene production in deep-well microplate cultures using the concentrated sulfuric acid method (Supplementary Figure 5). Forty-three strains with higher OD 450 were selected for further shake flask analysis. As shown in Figure 6, strain No. 6 produced the highest level of pinene (17.6 ± 0.2 mg/L). Thus, the TIGR-mediated plasmid was recovered from strain No. 6 and sequenced (Supplementary Table  2). We also retransformed the plasmid back to the host strain E. coli BW25113 (P T5 -dxs) and checked the pinene production. The resulting strain produced the same level of pinene (17.9 ± 0.1 mg/L), indicating that the pinene production improvement is the result of TIGR-mediated optimization.

Modular Co-culture Engineering to Improve Pinene Production
To take advantage of emerging co-culture engineering approaches to improve overall pinene biosynthesis in E. coli, the complete biosynthetic pathway was divided into the following two modules: the upstream module of the MEV pathway and the downstream module of the TIGR-mediated gene cluster of A. grandis GPPS Mut and P. taeda Pt1 MUT (Figure 7). The two FIGURE 6 | Pinene production by E. coli BW25113 (P T5 -dxs, pMEVI) harboring the selected TIGR-mediated gene cluster. The data represent the means of three replicates and error bars represent standard deviations. modules were integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then then evolved to a higher gene copy number by triclosan induction, respectively. Figure 8A shows the results of pinene production in CIChE strains of the TIGR-mediated gene cluster of A. grandis GPPS Mut and P. taeda Pt1 MUT without the MEV pathway. The maximum pinene production was obtained by the CIChE strains resistant to 32 µM triclosan. Thus, the recA gene of the CIChE strain resistant to 32 µM triclosan was deleted to obtain E. coli PINE. We determined the GPPS-Pt1 gene copy number in E. coli PINE. The copy number reached approximately 60 in the CIChE strain, which is the equivalent copy number of a high copy plasmid. Figure 8B shows the results of IPP/FPP concentration of the CIChE strains of the MEV pathway measured by the IPP/FPP sensor (pP rstA -GFP). As shown in Figure 8B, the maximum IPP/FPP production was obtained by the CIChE strains resistant to 0.5 µM triclosan. Thus, the recA gene of the CIChE strain resistant to 0.5 µM triclosan was deleted to obtain E. coli MEVI. We also determined the MEV pathway gene copy number in E. coli MEVI. The copy number reached approximately 4 in E. coli MEVI. Zhou et al. (2015) demonstrate that the modular co-culture engineering can be applicable to isoprenoids because their scaffold moleculars can generally permeate membranes. To demonstrate IPP can also cross cell membranes, we cultured E. coli (pP rstA -GFP) with the cell-free culture broth of E. coli MEVI and measured fluorescence strength. After addition of the cell-free culture broth of E. coli MEVI, E. coli (pP rstA -GFP) showed higher fluorescence strength (Supplementary Figure 6). Moreover, the E. coli MEVI: PINE co-culture produced higher level of pinene than E. coli PINE (Figure 9). These results indicate that IPP produced by E. coli MEVI diffused into E. coli PINE and was subsequently converted into pinene. We then optimized the E. coli MEVI: PINE co-culture system to further improve pinene production. To this end, different inoculation ratios between E.
coli MEVI and PINE were investigated. As shown in Figure 9, the highest pinene production of 64.9 ± 0.9 mg/L was achieved when E. coli MEVI and PINE were inoculated at a ratio of 1:2. Compared with the mono-culture strategy using E. coli PINE harboring pMEVI, the pinene production was increased by 1.9fold (from 22.3 ± 0.2 mg/L to 64.9 ± 0.9 mg/L). To test if all of IPP produced by E. coli MEVI were converted into pinene by E. coli PINE, we measured the IPP concentration in the broth of the co-culture and the E. coli MEVI after 28 h using the IPP sensor plasmid. The results showed that about 57.8% of IPP were converted by E. coli PINE (Supplementary Table 3). Thus, we introduced pQE-GPPS MUT -TIGR-Pt1 Q457L to overexpress the pinene biosynthetic pathway and checked the pinene production. The co-culture system after introducing the pinene biosynthetic pathway into E. coli MEVI produced higher level of pinene (60.2 ± 0.2 mg/L) than the control strain (52.1 ± 0.1 mg/L) harboring the empty plasmid (Supplementary Table 4). The result also demonstrates that not all of IPP can be converted in the co-culture system.
Biotechnological approaches for chemicals production can be broadly classified into fermentation and biocatalysis. In biocatalysis, cell growth and production phase are separated. In comparison to the fermentation bioprocess, whole-cell biocatalysis is an attractive method due to its great efficiency and relative simplification of downstream processing (Lin and Tao, 2017). The whole-cell biocatalysis processes comprise the following two stages: growth and conversion of the substrates. After the cells are cultured, they are harvested and washes with a buffer solution and suspended in the buffer for biocatalysis. Thus, the E. coli-E. coli modular co-culture system of whole-cell biocatalysis was used to further enhance pinene production. As shown in Figure 10, the highest pinene production of 166.5 ± 0.3 mg/L was achieved by the whole-cell biocatalyst after 28 h. The pinene titer obtained by the whole-cell biocatalysis was 2.6-fold higher than that produced by the fermentation process.

DISCUSSION
It has been reported that E. coli growth is inhibited by 0.5% pinene (Dunlop et al., 2011). We first improved pinene tolerance from 0.5 to 2.0% and pinene production by adaptive laboratory evolution after ARTP mutagenesis. In fact, improvements in tolerance are not sufficient to guarantee an increase production. Our results also demonstrate this point. Overexpreesion of A. borkumensis acrBDFa or P. putida KT2440 mexF that improved pinene tolerance did not improved pinene production (Figure 2A). To obtain a mutant with higher level of pinene production, we used the IPP/FPP sensor pP rstA -GFP to screen the mutants tolerant to 2% pinene. Tolerance engineering has also successfully been used to improve the production of limonene (Dunlop et al., 2011), amorphadiene , olefin (Mingardon et al., 2015), n-octane (Foo and Leong 2013). Although the level of pinene production reported in literatures did not inhibit growth, higher tolerance is beneficial to pinene production. Thus, the 2% pinene tolerant strain E. coli was used the parent strain in this study. To further improve pinene  production, we then expressed the efflux pumps in the pinene tolerant strain E. coli and subsequently selected a mutant resistant to fosmidomycin after ARTP mutagenesis. The pinene tolerant strain E. coli YZFP with higher level of pinene production was obtained through a two-step screening process. There is no directed evidence to prove the improved pinene production is the result of improved pinene tolerance.
Our study demonstrates that the overexpression of some efflux pumps improved pinene tolerance and production. Many groups also reported that overexpression of efflux pumps enhanced biofuel tolerance. Dunlop et al. (2011) reported that the overexpression of efflux pumps, such as A. borkumensis AcrBDFa, P. putida KT2440 MexF, P. putida KT2440 TtgB or E. coli AcrB, enhanced pinene tolerance. However, they did not investigate the effects of the pumps on pinene production. Our results demonstrate that overexpression of E. coli AcrAB and P. putida KT2440 TtgB enhanced pinene production (Figure 2). Overexpression of A. borkumensis AcrBDFa or P. putida KT2440 MexF did not improved pinene production (Figure 2A). However, Dunlop et al. (2011) reported that overexpression of A. borkumensis AcrBDFa enhanced limonene tolerance and yield. Overexpression of tolC together with ABC family transporters (macAB) or MFS family transporters (emrAB or emrKY) was found to improve amorphadiene titer by more than 3-fold . Overexpression of the native and evolved acrB improved olefin tolerance and production (Mingardon et al., 2015). Evolved AcrB variants with improved tolerance to pinene and n-octane have also been reported by Foo and Leong (2013). Taken together with these previous studies, our results show that a combination of the adaptive laboratory evolution with overexpression of some efflux pumps can improve pinene tolerance and production.
In this study, we reported a high-throughput screening method, which is known as the concentrated sulfuric acid method, for recombinant E. coli that overproduce pinene. We successfully applied the concentrated sulfuric acid method to FIGURE 9 | Effect of the inoculation ratio of E. coli PINE and MEVI on pinene production in the co-culture system. OD 600 (Gray bars), Pinene concentration (White bars). 0:1, only E. coli PINE; 0:1 (pMEVI), only E. coli PINE (pMEVI); others, the E. coli MEVI: PINE co-culture system with different inoculation ratio. The data represent the means of three replicates and error bars represent standard deviations.
FIGURE 10 | Time course of pinene production by the modular co-culture system of the whole-cell biocatalyst. The data represent the means of three replicates and error bars represent standard deviations.
screen the DNA shuffling library of the GPPS-PS gene cluster and the library of the TIGR-mediated GPPS-Pt1 gene cluster. Because limonene has the same properties as pinene, the concentrated sulfuric acid method can also be used to screen mutants for limonene production. Although the carotenoid-based method has been successfully used to screen isoprene synthase variants (Emmerstorfer-Augustin et al., 2016), the carotenoid-based method has a limitation when the colony has a faint color.
GPPS and PS have been identified as a major limiting factor in pinene production (Yang et al., 2013;Sarria et al., 2014;Tashiro et al., 2016). After directed evolution of the A. grandis GPPS-PS gene cluster using error-prone PCR and DNA shuffling, pinene production was increased by 1.2-fold ( Table 2). Two amino acid mutants were observed in the CDS of A. grandis GPPS. However, no mutant was observed in the CDS of A. grandis PS. Tashiro et al. evolved P. taeda Pt1 and constructed a recombinant E. coli with the highest pinene yield reported in literatures using the evolved variant (Tashiro et al., 2016). Using the A. grandis GPPS Mut -P. taeda Pt1 MUT gene cluster resulted in an increase in pinene production by 22.6% compared to using the A. grandis GPPS Mut -PS gene cluster ( Table 2).
GPPS and PS are inhibited by their substrate (GPP) or product (pinene) (Sarria et al., 2014). To overcome GPPS inhibition by GPP, GPPS was fused to PS, resulting in improved pinene production (Sarria et al., 2014;Tashiro et al., 2016). Our previous paper demonstrated that the TIGR approach was more efficient compared to protein fusion for coordinating expression . This study shows that using the TIGR-mediated gene cluster led to an increase in pinene production by 15.8% compared with the fused gene cluster ( Table 2).
In the present study, an E. coli-E. coli co-culture system was engineered to modularize the MEV and heterologous biosynthetic pathway. The MEV pathway and heterologous biosynthetic pathway (the A. grandis GPPS Mut -P. taeda Pt1 MUT gene cluster) was engineered in the pinene tolerance strain E. coli YZFP, respectively. The best co-culture system was found to improve pinene production by 1.9-fold compared to the mono-culture system. The modular co-culture can distribute the metabolic burden and allow for modular optimization by simply changing the strain-to-strain ratio. The E. coli-E. coli modular coculture system has been successfully used to improve 3-aminobenzoic acid , flavonoid (Jones et al., 2016), muconic acid (Zhang et al., 2015), and perillyl acetate (Willrodt et al., 2015), etc. In fact, the critical issue for modular co-culture engineering is the mass transfer of the pathway intermediate (IPP). It has been demonstrated that isoprenoids scaffold molecules can cross cell membranes (Zhou et al., 2015). Our results also demonstrate that IPP can cross cell membranes and secreted to the extracellular medium (Supplementary Figure 6 and Figure 9). Moreover, our results showed that the pinene tolerance strain E. coli YZFP (pP rstA -GFP) had higher fluorescence strength than the parent strain harboring pP rstA -GFP after addition the cell-free broth of E. coli MEVI (Supplementary Figure 6), indicating that E. coli YZFP shows greater membrane permeability than the parent strain. Our results demonstrate that there were still some IPP not to be converted into pinene by E. coli PINE (Supplementary  Tables 3, 4). Moreover, Overexpression of the pinene biosynthetic pathway in E. coli MEVI enhanced pinene production in the E. coli MEVI-E. coli PINE co-culture system (Supplementary  Table 4). However, overexpression of the pinene biosynthetic pathway in E. coli PINE did not enhance pinene production (Supplementary Table 4). Increasing the inoculation ratio of E. coli PINE and E. coli MEVI from 2:1 to 2.5:1 or 3:1 did not enhanced pinene production (Figure 9). These results indicate that the IPP transportation may be a key factor for further improving pinene production. Transporter engineering strategies have successfully been used to enhance the secretion of the pathway intermediates, improving production (Boyarskiy and Tullman-Ercek, 2015;Kell et al., 2015;Zhang et al., 2015). Thus, appropriate metabolite transporters engineering strategies may be used to further improve pinene production of the E. coli-E. coli co-culture system.
This study also demonstrated that whole-cell biocatalysis further improved pinene production by 1.6-fold compared to the fermentation process. The whole-cell biocatalysis has also successfully been used in many biotechnological production (Tao et al., 2011;Lin et al., 2015;Kogure et al., 2016;Chen et al., 2017;Lin and Tao, 2017). Kogure et al. (2016) also reported that the significantly higher shikimate productivity (141.3 g/L) was achieved by the whole-cell biocatalysis compared to that (78.8 g/L) achieved by the fed-batch fermentation accompanying cell growth. The pinene production improvement may be resulted from higher cell density (OD 600 of 30) and the growth-arrested cells used in the whole-cell biocatalysis.

CONCLUSIONS
Pinene tolerance and production were first improved via adaptive laboratory evolution and efflux pump overexpression. Through error-prone PCR and DNA shuffling, a GPPS variant was screened, which outperformed the wild-type enzyme. To balance the expression of multiple genes, a TIGR was inserted between A. grandis GPPS D90G/L175P and P. taeda Pt1 Q457L . To construct an E. coli-E. coli co-culture system to modularize the MEV and heterologous biosynthetic pathway, the MEV pathway and heterologous biosynthetic pathway (the A. grandis GPPS Mut -P. taeda Pt1 MUT gene cluster) was integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then evolved to a higher gene copy number by CIChE, respectively. The E. coli-E. coli modular co-culture system of whole-cell biocatalysis resulted in the highest pinene production of 166.5 mg/L. Our results demonstrate that the E. coli-E. coli modular co-culture system of the whole-cell biocatalysis is a promising approach for the production of pinene.

AUTHOR CONTRIBUTIONS
F-XN performed all of the experimental works. XH and Y-QW performed the pinene assay. J-ZL designed the study and wrote the manuscript. All the authors read and approved the final manuscript.