Genomic and transcriptional changes in response to pinene tolerance and overproduction in evolved Escherichia coli

Abstract

α-Pinene is an important monoterpene, which is widely used as a flavoring agent and in fragrances, pharmaceuticals and biofuels. Although an evolved strain Escherichia coli YZFP, which had higher tolerance to pinene and titer, has been successfully used to produce high levels of pinene, the pinene titer is much lower than that of hemiterpene (isoprene) and sesquiterpenes (farnesene) to date. Moreover, the overall cellular physiological and metabolic changes caused by higher tolerance to pinene and overproduction of pinene remains unclear. To reveal the mechanism of Escherichia coli YZFP with the higher tolerance to pinene and titer, a comparative genomics and transcriptional level analyses combining with CRISPR activation (CRISPRa) and interference (CRISPRi) were carried out. The results show that the tolerance to pinene and the overproduction of pinene in E. coli may be associated with: 1) the mutations of the DXP pathway genes, the rpoA and some membrane protein genes, and their upregulations of transcription levels; and 2) the mutations of some genes and their downregulation of transcriptional levels. These comparative omics analyses provided some genetic modification strategies to further improve pinene production. Overexpression of the mutated cbpA, tabA, pitA, rpoA, sufBCDS, mutS, ispH, oppF, dusB, dnaK, dxs, dxr and flgFGH genes further improved pinene production. This study also demonstrated that combining comparative omics analysis with CRISPRa and CRISPRi is an efficient technology to quickly find a new metabolic engineering strategy.

Highlights

• •A genomics and transcriptional level analyses combining with CRISPRa and CRISPRi was carried out.
• •The mechanism of the tolerance to pinene and overproduction of pinene was obtained.
• •Some target genes difficultly found by rational analysis were identified.
• •Combining comparative omics analysis with CRISPRa/i is an efficient technology for metabolic engineering.

1. Introduction

α-Pinene is a natural and active monoterpene, which is widely used as flavoring agents [1], chemicals [2] and pharmaceuticals [3], especially as a good candidate for jet fuel replacement. As a result, the biotechnological production of pinene has received increasing attention. Yang et al. first engineered an E. coli to produce 5.4 mg/L pinene by the coexpression of a heterologous mevalonate (MEV) pathway and α-pinene synthase (Pt 30) from Pinus taeda [4]. Then, the combinatorial expression of Abies grandis geranyl diphosphate synthase (GPPS)- pinene synthase (PS) fusion proteins further enhanced pinene production to 32 mg/L in E. coli [5]. The directed evolution of α-pinene synthase (Pt1) from P. taeda resulted in α-pinene production to 140 mg/L in E. coli [6]. Dunlop et al. reported that E. coli cannot grow in the presence of 0.5% [7]. To avoid high toxicity of pinene on the growth of E. coli and biosynthesis of pinene, we first improved pinene tolerance to 2.0% and pinene production to 9.9 mg/L from 5.6 mg/L through 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 [8]. Then, we applied a combinatorial strategy of enzymatic directed evolution and modular co-culture engineering to further increase pinene production to 166.5 mg/L [8]. In addition, Corynebacterium glutamicum has also been successfully used as host for pinene biosynthesis. Coexpression of native 1-deoxy-d-xylulose-5-phosphate synthase (Dxs), isopentenyl diphosphate isomerase (Idi), P. taeda PS and A. grandis GPPS in C. glutamicum yielded pinene of 27 μg/g cell dry weight [9]. Although some progress has been made in pinene biosynthesis, the production level of pinene is much lower than that of hemiterpene (isoprene) (gram level) [8,10]. Thus, further work is required to increase the production of pinene.

In this study, we combined comparative genomics and transcriptional level analysis with CRISPR activation (CRISPRa) and inference (CRISPRi) to investigate the mechanism of E. coli YZFP with higher tolerance to pinene and overproduction. Based on these results, overexpression of the mutants further improved pinene production.

2. Materials and methods

2.1. Strains, plasmids and primers

The bacterial strains and plasmids used in this study are listed in Table 1. The primers used in this study are listed in Supplementary Table 1.

Table 1.

Strains and plasmids used in this study.

Name	Description	Sources
Strain
E. coli DH5α	F- endA1 glnV44 thi-1 recA1 relA1 gyrA96 deoR nupG Φ80 dlacZ ΔM15 Δ(lacZYA-argF) U169	Invitrogen
E. coli BW25113(PT5-dxs)	lacIqrrnBT14 ΔlacZWJ16hsdR514 ΔaraB ADAH33 ΔrhaBADLD78, Pdxs::PT5	[11]
E. coli YZFP	The adaptive laboratory evolution strain from E. coli BW25113 (PT5-dxs), PacrAB::P37, SS9::P37-ttgB, tolerance to 2.0% pinene and 35 μM fosmidomycin	[8]

E. coli PINE	Pinene producer, chemically induced chromosomal evolution (CIChE) strain from E. coli E. coli YZFP after integration of the TIGR-mediated gene cluster of the Abies grandis GPPSMut- Pinus taeda Pt1MUT gene cluster	[8]
E. coli MEVI	CIChE strain from E. coli YZFP after integration of the mevalonate pathway	[8]
Plasmid
pMEVIGPS	pBbA5C-MTSAe-T1f-MBI(f)-T1002i-trGPPSA.grandis-PSA.grandi coding for MEV pathway enzymes to produce pinene from glucose in E. coli, p1FA ori, PlacUVF promoter, Cmr	[8]
pU57-RMS	pUC57 harboring Ptet promoter and scRNA sequences	This study
pCRISPathBrick*	Expressing dcas9(K848A/K1003A/R1060A), p15A ori, Cmr	[12,13]
pBbB2K-GFP	Addgene plasmid #35345, Ptet promoter, Kanr, pBBR1 ori	[14]
pBbB2K-GFP*	pBbB2K-GFP harboring XhoI endonuclease site after tet promoter	This study
pBbB2K-dCas9*	pBbB2K containing the dcas9 (K848A/K1003A/R10R0A)	This study
pUC57-MCPSoxS	pUC57 harboring the fusion gene of the codon-optimized MCP and SOX with a (GGGS) linker, Ptet Promoter	This study
pBbB2K-dCas9*-MCPSoxS	pBbB2K-dCas9* containing the MCPSoxS sequences	This study
pTargetB	E. coli sgRNA expression vector, BglBrick vector, pMB1, Sper, scRNA	[8]
pTargetA	E. coli scRNA expression vector, BglBrick vector, Ptet promoter, Sper, pMB1 ori	This study
pZEABP	Constitute expression vector, pBR322 ori, P37 promoter, Ampr	[15]

2.2. Whole genome resequencing and data analyses

Genomic DNA from E. coli YZFP strain was extracted using TIANamp Bacterial DNA Kit (Tiangen Biotech Co., Beijing, China) according to the manufacturer's instruction. Genomic library construction and whole-genome resequencing were performed on the Illumina HiSeq 2500 sequencing platform by GENWIZ (Shuzhou, China). The paired-end reads from E. coli YZFP were aligned to the reference genome of E. coli MG1655 using the BWA software (version 0.7.12). Potential mutations including point mutation and insertion/deletions were identified using the Samtools software (version 1.1) and GATK module (Unified Genotyper).

2.3. Transcriptional level analysis

The total RNA from E. coli cells grown for 30 h in shake flasks was isolated using an RNA extraction kit (Dongsheng Biotech, Guangzhou, China), following the manufacturer's instructions. RNase-free gDNaseI (Dongsheng Biotech, Guangzhou, China) was treated during the isolation procedure to eliminate possible DNA contamination. Absence of DNA was verified by control PCR reactions using the RNA as a template. RNA concentration was quantified using a NanoDrop ND-1000 spectrophotometer (Thermo Scientific). Only the RNA samples with 260/280 ratio (an indication of protein contamination) greater than 1.8 and 260/230 ratio (an indication of reagent contamination) greater than 2.0 were used for the analysis. The first-strand cDNA was synthesized using an All-in-OneTM First-Strand cDNA Synthesis kit (GeneCopoeia, Guangzhou, China). Quantitative real-time PCR (qRT-PCR) was performed with the All-in-OneTM qPCR Mix kit (GeneCopoeia) by 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. All qRT-PCR quantifications were performed simultaneously with PCR amplified standard and no-template (water) controls. The expression levels were analyzed by the 2^-△△Ct method described by Livak and Schmittgen [16] and normalized by cysG gene expression. Finally, the ratio of the transcriptional level of the evolved E. coli YZFP strain and original E. coli BW25113 (P_T5-dxs) strain of >1.2 or < 0.8 (P < 0.05, Student's t-test) was considered significant differential expression. Three biological replicates for each sample were used for qRT-PCR analysis and three technical replicates were analyzed for each biological replicate.

2.4. CRISPRa and CRISPRi of the mutant genes

Dong et al. developed a CRISPR-Cas system based on the actor SoxS with simultaneous activation and repression [17]. We applied this CRISPR-Cas-SoxS system [17] for activating and repressing target genes in E. coli BW25113(P_T5-dxs). To enhance specificity and reduce off-target effects, dcas9* (K848A/K1003A/R1060A) was amplified from pCRISPathBrick* with primers dCas9*-F/dCas9*-R and then cloned into the ApaLI/XhoI sites of pBbB2k-GFP* to generate pBbB2k-dCas9*. P_tet-MCPSoxS fragment was synthesized by GENWIZ (Shuzhou, China) and ligate into pUC57 to obtain pUC57-MCPSOX. This fragment was cut from pUC57-MCPSOX with BamHI/XhoI, and then cloned into BglII/XhoI-digested pBbB2K-dCas9* to obtain pBbB2K-dCas9*-MCPSoxS (Supplementary Fig. 1A).

Scaffold RNA (scRNA) expressing vector pTargetA was derived from pTargetB [8]. P_tet-scRNA fragment was synthesized by GENWIZ (Shuzhou, China) and ligate into pUC57 to obtain pUC57-RMS. This fragment was cut from pUC57-RMS with NdeI/EcoRI, and then cloned into NdeI/EcoRI-digested pTargetB to obtain pTargetA.

The pTargetA-X series used in target single-gene activation or repression with a targeting N20 sequence of gene loci of interest, was obtained by inverse PCR using primes the target N20F and N20-genes from pTargetA, then cut with KpnI and self-ligation. For activation, the N20 sequence was designed to target at the non-template strand at the upstream of the promoter. For repression, The N20 sequence was designed to target at the 5’ end (about 100 bp downstream of ATG) of the gene on the non-template DNA strand. Because the scRNA fragment was flanked with BamHI and BglII, the scRNA expressing vector pTargetA-XY for multiple genes was constructed from pTargetA-X and pTargetA-Y by using the standard BglBrick assemble approach [14].

To investigate the effect of the activation and repression on growth, pBbB2K-dCas9*-MCPSoxS and pTargetA-X were co-transferred into E. coli BW25113(P_T5-dxs). A single colony was grown in 5 mL LB medium in a falcon tube at 37 °C overnight. The overnight cultures were inoculated into 50 mL of LB medium supplement with 0.5% pinene and 200 nM anhydrous tetracycline with a starting OD₆₀₀ of 0.09. The cultures were incubated at 37 °C (200 rpm) for 12 h. The strain harboring the empty vector pTargetA, E. coli BW25113(P_T5-dxs) (pBbB2K-dCas9*-MCPSoxS, pTargetA), was set as the control.

To investigate the effect of the activation and repression on pinene production, pBbB2K-dCas9*-MCPSoxS, pTargetA-X and pMEVIGPS were co-transferred into E. coli BW25113(P_T5-dxs) and then analyzed the production of pinene in shake flasks. The strain harboring the empty vector pTargetA, E. coli BW25113(P_T5-dxs) (pBbB2K-dCas9*-MCPSoxS, pTargetA, pMEVIGPS) was set as the control.

2.5. Pinene production in shake flasks

A single colony was inoculated into 5 mL of LB medium in a falcon tube, which was cultured overnight at 37 °C and 200 rpm. The overnight seed culture was then inoculated into 50 mL of SBMSN medium with a starting OD₆₀₀ of 0.1. SBMSN medium (pH 7.0) containing the following (g/L): sucrose 20, peptone 12, yeast extract 24, KH₂PO₄ 1.7, K₂HPO₄ 211.42, MgCl₂·6H₂O 1, ammonium oxalate 1.42, and Tween-80 2 mL. The main cultures were then incubated at 37 °C and 200 rpm until an OD₆₀₀ of 0.8 was reached. Then, the cultures were induced with 1 mM IPTG and 200 nM anhydrous tetracycline, and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30 °C and 130 rpm for 72 h.

2.6. Whole-cell biocatalysis for pinene production

Genes were amplified from E. coli YZFP genomes using the responding primers (Supplementary Table 1) and cloned into pZEABP to obtain pZEA-gene series plasmids.

E. coli PINE harboring the target plasmid pZEA-gene and E. coli MEVI harboring the pZEABP were used for whole-cell biocatalysis. They were 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 OD600 of 0.1. The cultures were incubated at 20 °C for 16 h. The E. coli PINE culture was mixed with E. coli MEVI culture at the inoculation ration 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). Finally, the above cells were resuspended in 10 mL phosphate buffer (0.1 M, pH 7.0) containing 20 g/L of sucrose, 10 mM MgCl₂ and 5 mM MnCl₂ to form the cell suspension (OD₆₀₀ = 30). The reaction mixture was overlaid with 20% dodecane. The catalysis was performed for 28 h at 30 °C and 130 rpm.

For inverse engineering using CRISPRi, E. coli PINE (pBbB2K-dCas9*-MCPSoxS, pTargetA-X) and E. coli MEVI (pBbB2K-GFP, pTargetA) were used for whole-cell biocatalysis as the above-mentioned process.

2.7. GC analysis

Pinene concentration in the dodecane layer was assayed as previously reported [8]. Take the dodecane layer and dilute it in ethylacetate spiked with the internal standard limonene. 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 a flow at 1 mL/min, oven at 50 °C for 30 s, ramp at 4 °C/min to 70 °C, and ramp at 25 °C/min to 240 °C.

2.8. Statistical analysis

All experiments were conducted in triplicate, and 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 9.0) package. Statistical significance was defined as p < 0.05.

3. Results and discussion

3.1. Whole genome resequencing

In our previous study [8], we obtain an evolved strain E. coli YZFP, which showed a higher tolerance to pinene and titer. To investigate the relationship between the genotype and the phenotype of E. coli YZFP, the whole genome of E. coli YZFP was re-sequenced using a 2✕150 paired-end (PE) configuration. The genome sequencing generated 13548860 clean reads and achieved a mean depth of 399✕ and the mapped reads ratio of 99.79% (Table 2). After alignment to the genome sequence of E. coli MG1655, a genome coverage of 99.36% with covered bases of 4612119 bp was obtained. A total of 349 single nucleotide variants (SNV) and 25 insertion/deletion (InDel) were identified. Of SNVs, 184 nonsynonymous mutation distributed among 175 mutated genes (Table 2 and Supplementary Table 2). These nonsynonymous mutant genes were mainly involved in the membrane protein transport processing (49 genes), the genetic information processing (32 genes), the carbohydrate metabolism (26 genes), the stress response (11 genes), the amino acid metabolism (10 genes), the cellular processes (9 genes), the human diseases (7 genes), the nucleotide metabolism (6 genes), the biosynthesis of other secondary metabolism (5 genes), the lipid metabolism (4 genes), the metabolism of cofactors and vitamins (4 genes), the metabolism of terpenoids and polyketides (4 genes), the glycan biosynthesis and metabolism (3 genes), the energy metabolism (3 genes), and the xenobiotics biodegradation and metabolism (2 genes) (Fig. 1).

Table 2.

The SNV and InDel mutants in E. coli YZFP^a.

Type	Amounts of mutant	Amounts of gene
Synonymous SNV	93	84
Nonsynonymous SNV	184	175
Stop gaining SNV	11	11
Intergenic SNV	2
Upstream SNV	14
Downstream SNV	15
Upstream and Downstream SNV	29
Total	349	270
InDel	25

^aMapped reads ratio of 99.79%, covered bases of 4612119 bp, unique read ratio of 97.77%, genome coverage of 99.36%, mean depth of 399✕. The SNVs and InDels were identified with a frequency of 1.0.

Fig. 1.

Pathway gene ontology enrichment analysis of the nonsynonymous mutant genes.

3.2. Comparative transcriptional level analysis of the mutant genes

Although comparative transcriptome analysis can provide more information, it is difficult to confirm that transcriptome changes are caused by gene mutations. Thus, we only analyzed the changes in transcriptional levels of the mutant genes. To investigate whether gene mutations caused changes in transcriptional levels in E. coli YZFP, the transcriptional levels of the nonsynonymous mutant genes (174 genes) and genes observed SNV in the upstream (8 genes) were determined using qRT-PCR and were compared with those in E. coli BW25113(P_T5-dxs). A total of 116 genes showed significantly different expression patterns. Of them, 96 genes were up-regulated and 20 genes were down-regulated in E. coli YZFP compared to E. coli BW25113(P_T5-dxs) (Fig. 2).

Fig. 2.

The relative transcriptional levels of the nonsynonymously mutated genes in E. coli YZFP compared with those in E. coli BW25113(P_T5-dxs). (A) Ratio was less than 1.7; (B) ratio was greater than 1.7; (C) ratio was less than 0.8. The data represent the means of three replicates, and error bars represent standard deviations.

3.3. Functional confirmation of the mutant genes

Dong et al. developed a CRISPR-Cas system based on the actor SoxS with simultaneous activation and repression [17]. Using this CRISPR-Cas-SoxS system enable programmable control over multiple genes with simultaneous activation and repression. Thus, we first constructed a dcas9* and SoxS expressing plasmid pBbB2K-dCas9*-MCPSoxS and scRNA expressing vector pTargetA (Supplementary Fig. 1A). Then, we characterized the functions of the CRISPR-Cas-SoxS system. We constructed the scRNA plasmid for CRISPRa and CRISPRi to mCherry, respectively. When N20 sequence was designed to target the P37 promoter or the CDS of mCherry, mCherry was repressed (Supplementary Fig. 1B). When N20 sequence was designed to target the upstream of the P37 promoter, mCherry was activated (Supplementary Fig. 1C). These results demonstrate that the CRISPR-Cas-SoxS system can function with activation and repression.

Thus, we applied this CRISPR-Cas-SoxS system [17] for activating and repressing target genes in E. coli BW25113(P_T5-dxs) to investigate their effects on growth and pinene production. Of 96 up-regulated gens, a total of 57 up-regulated genes with higher transcription level compared to E. coli BW25113(P_T5-dxs) was selected to be activated with CRISPRa. pBbB2K-dCas9*-MCPSoxS and pTargetA-X were co-transferred into E. coli BW25113(P_T5-dxs). The growth and pinene production were analyzed and compared to the control strain harboring the empty vector pTargetA. The results are presented in Table 3 and Supplementary Table 3. The activations of 23 genes resulted in increases both in tolerance to pinene and pinene production (Table 3). The activations of 9 genes only improved pinene production. The activations of 20 genes only enhanced the tolerance to pinene.

Table 3.

Effect of CRISPR activation of the up-regulated genes on growth and production of pinene in E. coli BW25113(P_T5-dxs).

Gene	Protein	Ratio of growtha	Ratio of pinene concentrationb
Carbohydrate metabolism
ulaE	l-xylulose 5-phosphate 3-epimerase	1.24 ± 0.01	1.06 ± 0.02
Metabolism of terpenoids and polyketides
ispH	4-hydroxy-3-methylbut-2-enyl diphosphate reductase, 4Fe-4S protein	1.07 ± 0.01	1.19 ± 0.01
dxr	1-deoxy-d-xylulose 5-phosphate reductoisomerase	–	1.34 ± 0.02
dxs	1-deoxyxylulose-5-phosphate synthase, thiamine triphosphate-binding, FAD-requiring	–	1.39 ± 0.02
ispU	1-deoxy-d-xylulose 5-phosphate reductoisomerase/undecaprenyl pyrophosphate synthase	–	1.28 ± 0.01
Genetic Information Processing
rpoA	RNA polymerase, alpha subunit	1.50 ± 0.01	1.09 ± 0.01
dosP	oxygen sensor, c-di-GMP phosphodiesterase,	1.56 ± 0.02	1.07 ± 0.01
mutS	methyl-directed mismatch repair protein	1.10 ± 0.01	1.29 ± 0.01
cbpA	DnaK co-chaperone; curved DNA-binding protein/uncharacterized protein	1.30 ± 0.01	1.04 ± 0.01
dusB	methyltransferase for 50S ribosomal subunit protein L11/tRNA-dihydrouridine synthase B	1.46 ± 0.02	1.23 ± 0.01
Membrane protein transport Processing
gsiA	glutathione ABC transporter ATPase	1.09 ± 0.01	1.05 ± 0.01
nlpA	cytoplasmic membrane lipoprotein-28	1.64 ± 0.02	1.03 ± 0.02
sufBCDS	SufBCD Fe-S cluster assembly protein	1.48 ± 0.02	1.32 ± 0.01
opgB	OPG periplasmic biosynthetic phosphoglycerol transferases I	1.10 ± 0.01	1.06 ± 0.01
setC	putative arabinose efflux transporter	1.14 ± 0.02	1.10 ± 0.01
oppF	oligopeptide ABC transporter ATPase	1.14 ± 0.01	1.13 ± 0.01
ypjK	CP4-57 prophage; putative inner membrane protein	1.16 ± 0.01	1.10 ± 0.02
btuB	vitamin B12/cobalamin outer membrane transporter	1.05 ± 0.02	1.11 ± 0.01
pitA	phosphate transporter, low-affinity; tellurite importer	1.15 ± 0.01	1.17 ± 0.01
cusA	copper/silver efflux system, membrane component	1.15 ± 0.01	1.05 ± 0.01
Cellular Processes
yehA	putative fimbrial-like adhesin protein	1.48 ± 0.03	1.24 ± 0.02
motB	protein that enables flagellar motor rotation	1.17 ± 0.02	1.16 ± 0.01
flgFGH	flagellar protein	1.56 ± 0.03	1.41 ± 0.03
ypjA	adhesin-like autotransporter	1.12 ± 0.02	1.07 ± 0.01
tabA	transcriptional repressor for divergent bdcA/biofilm modulator regulated by toxins; DUF386 family protein, cupin superfamily protein	1.05 ± 0.01	1.05 ± 0.01
Stress response
hslU	molecular chaperone and ATPase component of HslUV protease	1.53 ± 0.03	1.29 ± 0.02
flgFGH-sufBCDS- dusB-rpoA-yehA-hslU		3.55 ± 0.03	1.71 ± 0.02

^aThe ratio of OD₆₀₀ with CRISPRa and without.

^bThe ratio of pinene production with CRISPRa and without.-: means no change or negative effect.

As shown in Table 3, the activations of dxr, dxs, ispH and ispU improved the production of pinene. The dxr, dxs, ispH and ispU is involved in 1-deoxy-d-xylulose-5-phosphate (DXP) pathway. These activations may increase the availability of pinene precursor isopentenyl diphosphate (IPP), enhancing the production of pinene. The activation of ispH improved the tolerance to pinene and the production of pinene (Table 3). The ispH gene encodes 4-hydroxy-3-methylbut-2-enyl diphosphate reductase which catalyzes the formation of IPP. Li et al. reported that the activation of ispH eliminated the accumulation of the intermediate of 1-hydroxy-2-methyl-2-(E)-butenyl-4-diphosphate of DXP pathway, improving the production of carotene [18]. It was reported that the mutations of Dxs^{E289G/N455I/E598V} and Dxr^K37N/K217N resulted in a 60% increase in the production of isoprene in E. coli [19]. The upregulations of the DXP pathway genes may be related to that E. coli YZFP is tolerant to fosfomycin [8,20]. Fosfomycin is the DXP pathway inhibitor that inhibits Dxr and methylerythritol phosphate cytidyltransferase (IspD). The Fosfomycin resistant mutants show higher levels of Dxr and IspD [21].

Activation of the rpoA gene improved the tolerance to pinene and the production of pinene (Table 3). rpoA encodes RNA polymerase subunit alpha. The plasmid expression of the rpoA mutant has been successfully used to improve a lot of cellular traits of E. coli, including butanol tolerance [22] and the production of tyrosine [23]. This group developed the global transcription machinery engineering (gTME) to allow for global perturbations of transcriptome. Plasmid expression of the mutant rpoD or rpoS enhanced ethanol tolerance, lycopene overproduction [24] and the production of hyaluronic acid biosynthesis [25].

Activation of some membrane protein genes, such as gsiA, nlpA, sufBCDS, opgB, setC, oppF, ypjK, btuB, pitA and cusA, also improved the tolerance to pinene and the production of pinene (Table 3). These activations may be beneficial to the transport of metabolite, thereby improving tolerance and pinene synthesis. Many papers demonstrated that the membrane modified transport engineering is a powerful tool for improving tolerance and product biosynthesis [26,27].

All of 20 down-regulated genes were selected to be repressed with CRISPRi. And their effects of the repressions on growth and pinene production were analyzed. The results are presented in Table 4 and Supplementary Table 4. Repressions of 6 genes led to increases both in tolerance to pinene and pinene production (Table 4). Repressions of 3 genes only improved the tolerance to pinene (Supplementary Table 4).

Table 4.

Effect of CRISPR interference of the down-regulated genes on growth and production of pinene in E. coli BW25113(P_T5-dxs).

Gene	Protein	Ratio of growtha	Ratio of pinene concentrationb
Lipid metabolism
ydiJ	putative FAD-linked oxidoreductase	1.07 ± 0.01	1.06 ± 0.02
Metabolism of cofactors and vitamins
yjbQ	thiamine phosphate synthase	1.08 ± 0.02	1.08 ± 0.02
Genetic Information Processing
prpR	propionate catabolism operon regulatory protein	1.06 ± 0.01	1.07 ± 0.01
marR	transcriptional repressor of multiple antibiotic resistance	1.22 ± 0.02	1.09 ± 0.01
fabR	transcriptional repressor of fabA and fabB	1.09 ± 0.01	1.07 ± 0.02
Cellular Processes
cedA	cell division modulator/catalase HPII, heme d-containing	1.07 ± 0.01	1.07 ± 0.02
ydiJ-yjbQ-prpR-marR-fabR-cedA		1.26 ± 0.02	1.13 ± 0.02

^aThe ratio of OD₆₀₀ with CRISPRi and without.

^bThe ratio of pinene production with CRISPRi and without.-: means no change or negative effect.

Because the phenotype in E. coli YZFP indeed resulted from the full mutations of these mutant genes, we tested the effect of multiple genes. However, constructing the scRNA expressing vector for simultaneous activation or repression of all genes is rather time-consuming. Because the repressions of six genes improved the tolerance and titer, we assembled the scRNA expressing vector of six genes and investigated the simultaneous effect of them. For activation, we randomly selected six genes. The simultaneous activation of the flgFGH, sufBCDS, dusB, rpoA, yehA and hslU further improved growth and pinene production than single activation (Table 3). After simultaneous activation of the six genes, the OD₆₀₀ at 12 h was increased to 0.451 ± 0.013 from 0.127 ± 0.009 without activation (Suppl. Fig. 2). From the grow curve (Sppl. Fig. 2), we can see that the wild-type strain without activation cannot grow in the present of 0.5% pinene. After activation with the six genes, the wild-type strain can grow slightly. And the tolerant stain E. coli YZFP can grow well in the present of 0.5% pinene.

The same result of simultaneous repression of the ydiJ, yjbQ, prpR, marR, fabR and cedA was also observed (Table 4). These results further demonstrated that the phenotype of E. coli YZFP is caused by the mutant genes.

3.4. Omics-guided genetic modifications

The ultimate goal of systems metabolic engineering is to further improve the production of metabolite. The above results show that the activations of 26 up-regulated mutant genes improved the tolerance to pinene and pinene production. Thus, these genes were amplified from E. coli YZFP and then individually cloned into plasmid pZEABP to obtain pZEABP-gene. These plasmids were individually introduced into E. coli PINE. These strains E. coli PINE (pZEABP-gene) were individually co-cultured with E. coli MEVI (pZEABP) with whole-cell biocatalysis.

As shown in Fig. 3A, overexpression of the mutated cbpA, tabA, pitA, rpoA, sufBCDS, mutS, ispH, oppF, dusB, dnaK, dxs, dxr and flgFGH increased the production of pinene by 7%, 9%, 9%, 10%, 10%, 10%, 11%, 12%, 15%, 17%, 26%, 28% and 31%, respectively. Except for ispH, dxs and dxr, other genes cannot be identified by general rational analysis. Some target genes difficulty found by rational analysis were identified.

Fig. 3.

Pinene production in the modular co-culture system of the whole-cell biocatalysis. (A) Using E. coli PINE (pZEABP-gene) and E. coli MEVI (pZEABP). (B) Using E. coli PINE (pBbB2K-dCas9*-MCPSoxS, pTargetA-X) and E. coli MEVI (pBbB2K-GFP, pTargetA). The data represent the means of three replicates, and error bars represent standard deviations.

Table 4 shows that the repressions of 6 down-regulated mutant genes improved the tolerance to pinene and pinene production. Thus, we also assayed whether these repressions can improve the production of pinene in the modular co-culture system of the whole-cell biocatalysis. It can be seen from Fig. 3B that these repressions in the strain tolerant to pinene did not further improve the production of pinene.

In this study, the plasmid system was used for gene expression. However, the plasmid expression system has the drawbacks of genetic instability and antibiotics pollution problem. How about chromosome expression? In addition, the simultaneous activation of the multiple genes improved the production of pinene in E. coli BW25113(P_T5-dxs) (Table 3). Are there synergies effects between these genes in the modular co-culture system of the whole-cell biocatalysis? All these should be needed further to investigate.

4. Conclusions

In this study, we combined comparative genomics and transcriptional level analysis with CRISPRa and CRISPRi to investigate the regulation mechanism of E. coli YZFP with higher tolerance to pinene and pinene titer. The reasons for the increased tolerance to pinene and pinene production may be associated with: 1) the mutations of the DXP pathway genes (dxr, dxs, ispH and ispU), the rpoA and some membrane protein genes (gsiA, nlpA, sufBCDS, opgB, setC, oppF, ypjK, btuB, pitA and cusA), and their upregulation of transcription levels; and 2) the mutations of some genes (ydiJ, yjbQ, prpR, marR, fabR and cedA) and their downregulation of transcriptional levels. Overexpression of the mutated cbpA, tabA, pitA, rpoA, sufBCDS, mutS, ispH, oppF, dusB, dnaK, dxs, dxr and flgFGH genes further improved pinene production in the E. coli-E. coli modular co-culture system of whole-cell biocatalysis. This study also demonstrated that combining comparative omics analysis with CRISPRa and CRISPRi is an efficient technology to quickly find a new metabolic engineering strategy.

Appendix A. Supplementary data

The following is the Supplementary data to this article: