Fine-Tuning of the Fatty Acid Pathway by Synthetic Antisense RNA for Enhanced (2S)-Naringenin Production from l-Tyrosine in Escherichia coli

Abstract

Malonyl coenzyme A (malonyl-CoA) is an important precursor for the synthesis of natural products, such as polyketides and flavonoids. The majority of this cofactor often is consumed for producing fatty acids and phospholipids, leaving only a small amount of cellular malonyl-CoA available for producing the target compound. The tuning of malonyl-CoA into heterologous pathways yields significant phenotypic effects, such as growth retardation and even cell death. In this study, fine-tuning of the fatty acid pathway in Escherichia coli with antisense RNA (asRNA) to balance the demands on malonyl-CoA for target-product synthesis and cell health was proposed. To establish an efficient asRNA system, the relationship between sequence and function for asRNA was explored. It was demonstrated that the gene-silencing effect of asRNA could be tuned by directing asRNA to different positions in the 5′-UTR (untranslated region) of the target gene. Based on this principle, the activity of asRNA was quantitatively tailored to balance the need for malonyl-CoA in cell growth and the production of the main flavonoid precursor, (2S)-naringenin. Appropriate inhibitory efficiency of the anti-fabB/fabF asRNA improved the production titer by 431% (391 mg/liter). Therefore, the strategy presented in this study provided a useful tool for the fine-tuning of endogenous gene expression in bacteria.

INTRODUCTION

Biological synthesis has emerged as a highly promising alternative to the traditional extraction or organic synthesis of plant-specific secondary metabolites that have significant applications in medicine and agriculture, such as aromatic polyketides (1), flavanones (2, 3), and fatty acids (4). Among many heterologous hosts, Escherichia coli has become an attractive test subject due to its genetic tractability and favorable fermentation properties. However, developing efficient recombinant production platforms for synthesizing these value-added compounds often is limited by the availability of malonyl coenzyme A (malonyl-CoA) precursors derived from the host's native metabolic networks (3, 5). Therefore, it has become essential to rationally regulate the intracellular malonyl-CoA level in E. coli.

Previous studies have made significant advances in the control of intracellular malonyl-CoA through the amplification or deletion of particular pathways (3, 5, 6). For example, simultaneous deletion of the genes sdhA, adhE, brnQ, and citE and overexpression of acetyl-CoA synthase, acetyl-CoA carboxylase, biotin ligase, and pantothenate kinase (5), or the deletion of fumC and sucC and overexpression of acetyl-CoA carboxylase, phosphoglycerate kinase, glyceraldehyde-3-phosphate dehydrogenase, and pyruvate dehydrogenase complex (3), have significantly improved the availability of malonyl-CoA. As intracellular malonyl-CoA is directly associated with the synthesis of phospholipids and fatty acids, each of which play significant roles in cell growth, the tuning of malonyl-CoA into heterologous pathways via conventional gene knockout strategies yields significant phenotypic effects, such as growth retardation and even cell death (7).

Therefore, a fine-tuning system that balances the requirement of malonyl-CoA for cell growth and product synthesis is required. Such a tuning system would allow researchers to separate the consumption of malonyl-CoA in cell growth and flavonoid production. Cultures would accumulate biomass at wild-type rates, and, once a target cell density is achieved, the system would downregulate endogenous metabolism to redirect the flux away from growth into flavonoid formation. The antibiotic cerulenin has been shown to repress the fatty acid synthesis pathway to achieve high (2S)-naringenin titers (2, 8), as this inhibitor specifically inactivates the β-ketoacyl-acyl carrier protein synthases (KAS) I and II (encoded by fabB and fabF, respectively) (9). However, the cost and toxicity of cerulenin (more than $30 per mg) make it impractical in large-scale fermentation processes (6).

RNA-mediated regulatory mechanisms can be exploited for fine flux control (10). There are some successful applications of antisense RNA (asRNA) or other RNA-based gene expression control systems with equally diverse design rules for asRNA construction. Different aspects have been studied, such as binding interactions (11), the kinetics of hybridization (12), and the stability of constructs (13). Although there is general agreement that these RNA species bind to the 5′-untranslated region (5′-UTR), it is unclear how the start point, binding location, and RNA structure affect the inhibitory efficiency of specific RNA species. For example, successful designs with both short (∼20 nucleotides [nt]) (10, 14) and long (∼120 nt) regions complementary to the target mRNA (15) have been tried. Therefore, establishing an efficient asRNA system suited for the rational regulation of the intracellular malonyl-CoA level in E. coli remains a challenge.

Here, we explored the fine-tuning of the fatty acid pathway with asRNA and the impact of this method on heterologous pathway productivity using heterologous production of (2S)-naringenin in E. coli as a model system. (2S)-Naringenin may be responsible for a decreased risk of heart disease and diabetes and is the starting point for synthesis of a variety of other flavonoid molecules (2, 8). Synthesis begins with the enzymatic conversion of l-tyrosine by tyrosine ammonia lyase (TAL) to produce p-coumaric acid, which then is converted into its corresponding coenzyme A ester, coumaroyl-CoA, via 4-coumarate:CoA ligase (4CL). This compound subsequently is condensed with three malonyl-CoA units by chalcone synthase (CHS), and the resulting (2S)-naringenin chalcone is converted to (2S)-naringenin by the action of chalcone isomerase (CHI) (Fig. 1) (16).

FIG 1.

Heterologous biosynthesis of (2S)-naringenin from l-tyrosine in E. coli. The overall metabolic pathway performed heterologous biosynthesis of (2S)-naringenin from l-tyrosine in E. coli. TAL, tyrosine ammonia lyase; 4CL, 4-coumarate:CoA ligase; CHS, chalcone synthase; CHI, chalcone isomerase; MatB, R. trifolii malonate synthetase; MatC, R. trifolii malonate carrier protein; FabD, malonyl-CoA-acyl carrier protein transacylase; FabB, β-ketoacyl-acyl carrier protein synthase (KAS) I; and FabF, β-ketoacyl-acyl carrier protein synthase (KAS) II. The overall biosynthetic pathway was constructed using premade modules from a plasmid toolbox; FabB and FabF were repressed by asRNAs.

To establish an efficient asRNA system for the rational regulation of intracellular malonyl-CoA levels in E. coli, the relationship between sequence and function for asRNA was explored. It was demonstrated that the inhibitory efficiency of asRNA could be tuned by directing asRNA to different positions in the 5′-UTR of the target gene in E. coli cells. To confirm silencing, fatty acid and malonyl-CoA concentrations also were quantified. Based on these findings, the inhibitory efficiency of asRNA toward fabB and fabF was quantitatively tailored to balance the requirement of malonyl-CoA for both cell growth and product synthesis. Appropriate regulation of target genes by these asRNA devices improved the production titer of (2S)-naringenin by 431% (391 mg/liters). This is the highest production titer reported to date when starting from l-tyrosine.

MATERIALS AND METHODS

Strains, plasmids, and general techniques.

Luria broth (LB) (10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter NaCl) was used for strain construction. Morpholinepropanesulfonic acid (MOPS) minimal medium (17) supplemented with 5 g/liter glucose and an additional 4 g/liter NH₄Cl was used for asRNA expression and flavonoid production. Ampicillin (100 μg/ml), kanamycin (40 μg/ml), chloramphenicol (20 μg/ml), and streptomycin (40 μg/ml) were added when required. The primers and plasmids used in this study are listed in Tables 1 and 2, respectively.

TABLE 1.

Nucleotide sequences of primers

TABLE 2.

Plasmids used in this study

Plasmida	Description	Source or reference
pCDFDuet-1	Double T7 promoters, CDF ori, Smr	Novagen
pETDuet-1	Double T7 promoters, pBR322 ori, Ampr	Novagen
pACYCDuet-1	Double T7 promoters, P15A ori, Cmr	Novagen
pRSFDuet-1	Double T7 promoters, RSF ori, Knr	Novagen
pCDF-Trc	T7 promoter was replaced by trc promoter	This study
pCDF-Trc-4CL	pCDFDuet-1 carrying 4CL under trc promoter	This study
pCDF-Trc-TAL	pCDFDuet-1 carrying TAL under trc promoter	This study
pCDF-Trc-TAL-Trc-4CL	T7 promoter was replaced by trc promoter	This study
pACYC-matC-matB	pACYCDuet-1 carrying matB and matC	8
pET-CHS-CHI	pETDuet-1 carrying CHS and CHI	This study
pRSFDuet-1(Mut)	XhoI site of pRSFDuet-1 was removed	This study
pRSF-PTasRNA	pRSFDuet-1(Mut) carrying PTasRNA	This study
pRSF-PTasRNA(−158, M)	Expressing fabB asRNA sequences (−158 to M bp)	This study
pRSF-PTasRNA(−128, M)	Expressing fabB asRNA sequences (−128 to M bp)	This study
pRSF-PTasRNA(−98, M)	Expressing fabB asRNA sequences (−98 to M bp)	This study
pRSF-PTasRNA(−68, M)	Expressing fabB asRNA sequences (−68 to M bp)	This study
pRSF-PTasRNA(−38, M)	Expressing fabB asRNA sequences (−38 to M bp)	This study
pRSF-PTasRNA(0, M)	Expressing fabB asRNA sequences (0 to M bp)	This study
pRSF-PTasRNA(−87, Q)	Expressing fabF asRNA sequences (−87 to Q bp)	This study
pRSF-PTasRNA(−57, Q)	Expressing fabF asRNA sequences (−57 to Q bp)	This study
pRSF-PTasRNA(−27, Q)	Expressing fabF asRNA sequences (−27 to Q bp)	This study
pRSF-PTasRNA(0, Q)	Expressing fabF asRNA sequences (0 to Q bp)	This study
pRSF-PTasRNA(P, N)	Expressing fabF asRNA sequences (P to 113 bp) and fabB asRNA sequences (N to 147 bp)	This study

^aM indicates 847, 747, 647, 547, 447, 347, 247, or 147. Q indicates 813, 713, 613, 513, 413, 313, 213, or 113. P indicates −27, −57, or −87. N indicates −38, −68, −98, −128, or −158.

The compatible vector set (pETDuet-1, pCDFDuet-1, pRSFDuet-1, and pACYCDuet-1) was used to express multiple genes in one strain (Novagen, Darmstadt, Germany) (18). The gene copy numbers of pACYCDuet-1 (p15A origin), pCDFDuet-1 (CDF origin), pETDuet-1 (pBR322 origin), and pRSFDuet-1 (RSF origin) were assigned as 10, 20, 40, and 100, respectively (18). To remove the XhoI site from pCDFDuet-1 and pRSFDuet-1, the XhoI site (CTCGAG) was mutated to CTAGAG by PCR-based site-directed mutagenesis (QuikChange II XL site-directed mutagenesis kit; Agilent Technologies, Santa Clara, CA) with the primer pairs Pf_pCDF/Pr_pCDF and Pf_pRSF/Pr_pRSF, resulting in pCDFDuet-1(Mut) and pRSFDuet-1(Mut), respectively.

E. coli JM109 was used for plasmid propagation. E. coli BL21(DE3) was used to express asRNA and produce flavonoids. Cell growth was monitored by measuring absorbance at 600 nm (OD₆₀₀) with a UV-visible spectrophotometer (UVmini-1240; Shimadzu, Kyoto, Japan).

Construction of anti-fabB asRNA-expressing plasmids.

Plasmid pUC57-asRNA, which was synthesized by GenScript (Nanjing, China), contained asRNA sequences and the trc promoter, the lac operator, and the stem-loop structure of the rrnB terminator. DNA sequence for this synthesized asRNA is provided in the supplemental material (see Fig. S1). The asRNA sequences from pUC57-asRNA were digested with PfoI/BamHI and inserted into the corresponding sites of pRSFDuet-1(Mut) and pCDFDuet-1(Mut) to construct pRSF-asRNA and pCDF-asRNA, respectively.

The plasmid pUC57-PTasRNA (synthesized by GenScript [Nanjing, China]) included the PTasRNA sequence (Fig. 2D) (13), which contained the trc promoter, the lac operator, a 38-bp paired-terminal (PT) sequence, and the stem-loop structure of the rrnB terminator. DNA sequence for this synthesized PTasRNA is provided in the supplemental material (see Fig. S1). The PTasRNA sequence from pUC57-PTasRNA was digested with PfoI/BamHI and inserted into the corresponding site of pRSFDuet-1(Mut) to construct pRSF-PTasRNA (Fig. 2B).

FIG 2.

Construction of initial asRNA expression system. (A) The growth patterns of asRNA-regulated strains compared to the control were measured. (B) pRSF-PTasRNA is derived from pRSFDuet-1(Mut). (C) The transcriptional levels of fabB mRNA from asRNA-regulated strains also were calculated relative to the control. pRSFDuet-1(Mut), E. coli BL21(DE3) strains with pRSFDuet-1(Mut); pCDF-asRNA(147), E. coli BL21(DE3) strains with pCDF-asRNA(147); pRSF-asRNA(147), E. coli BL21(DE3) strains with pRSF-asRNA(147); pRSF-PTasRNA(147), E. coli BL21(DE3) strains with pRSF-PTasRNA(147). (D) Predicted secondary structure of the PTasRNA (13). Ptrc, trc promoter; lac^O, lac operator; PT, 38-bp paired-terminal sequence; rrnB, stem-loop structure of the rrnB terminator. Error bars are standard deviations from biological triplicates.

To express the anti-fabB asRNA sequence (with base pairs [bp] N and M, where N = −158, −128, −98, −68, −38, or 0 and M = 847, 747, 647, 547, 447, 347, 247, or 147) complementary to the 5′-UTR (N bp) and the fabB gene sequence (M bp), the primer pair Pf_fabB(N)/Pr_fabB(M) was used to amplify the asRNA sequence from genomic DNA of E. coli BL21(DE3). This amplicon then was cloned into the NcoI-XhoI sites of pRSF-PTasRNA to construct pRSF-PTasRNA(N, M). The primer pair Pf_fabB(0)/Pr_fabB(147) was used to amplify asRNA sequence from genomic DNA of E. coli BL21(DE3) into pRSF-asRNA, pCDF-asRNA, and pRSF-PTasRNA, resulting in pRSF-asRNA(147), pCDF-asRNA(147), and pRSF-PTasRNA(147).

Construction of anti-fabF asRNA-expressing plasmids.

To express the anti-fabF asRNA sequence (base pairs P and Q, where P = −87, −57, −27, or 0 and Q = 813, 713, 613, 513, 413, 313, 213, or 113) complementary to the 5′-UTR (P bp) and the fabF gene coding region (Q bp), primer pair Pf_fabF(P)/Pr_fabF(Q) was used to amplify the asRNA sequence from genomic DNA of E. coli BL21(DE3), which then was cloned into the NcoI-XhoI site of pRSF-PTasRNA to construct pRSF-PTasRNA(P, Q).

Construction of plasmids expressing anti-fabB and anti-fabF asRNA simultaneously.

The primer pair Pf_PTasRNA(BamHI)/Pr_PTasRNA(AvrII) was used to amplify the fabB asRNA sequence (N to 147 bp) from pRSF-PTasRNA(N, 147), which then was digested and inserted into the BamHI-AvrII site of pRSF-PTasRNA(P, 113) to construct plasmid pRSF-PTasRNAs(P, N) (P = −27, −57, or −87 and N = −38, −68, −98, −128, or −158).

Heterologous pathway construction and assembly.

The primer pair Pf_Trc(FseI)/Pr_Trc(EcoNI) was used to clone a sequence containing the trc promoter, multicloning sites, and the rrnB terminator from pTrcHis2B (2) into the EcoNI/FseI site in pCDFDuet-1 to construct pCDFD-Trc.

Codon usage-optimized TAL from Rhodotorula glutinis (GenBank accession no. KF765779) (2, 19) was amplified from pUC57-TAL (20) with the primer pair Pf_TAL(NcoI)/Pr_TAL(EcoRI) and then digested and inserted into the NcoI/EcoRI sites of pCDFD-Trc to construct pCDF-Trc-TAL. Codon usage-optimized 4CL from Petroselinum crispum (GenBank accession no. KF765780) (8) was amplified from pUC57-4CL (20) with the primer pair Pf_4CL(NcoI)/Pr_4CL(HindIII) and then digested and inserted into the NcoI/HindIII sites of pCDFD-Trc to construct pCDF-Trc-4CL. The pTrc-4CL region, including the trc promoter and 4CL, was amplified with the primer pair Pf_Ptrc4CL(EcoRI)/Pr_4CL(HindIII) and then digested and inserted into the EcoRI/HindIII sites of pCDF-Trc-TAL to construct pCDF-Trc-TAL-Trc-4CL.

Codon usage-optimized CHS from Petunia x hybrida (GenBank accession no. KF765781) (8) was digested from pUC57-CHS (20) and cloned into the NcoI/HindIII sites of pETDuet-1 to construct pET-CHS. Codon usage-optimized CHI from Medicago sativa (GenBank accession no. KF765782) (8) was excised from pUC57-CHI (20) and cloned into the NdeI/BlnI sites of pET-CHS to construct pET-CHS-CHI. Codon usage-optimized matB from Rhizobium trifolii (GenBank accession no. KF765783) (8) was excised from pUC57-matB (20) and inserted into the NdeI/KpnI sites of pACYCDuet-1 to construct pACYC-matB. Codon usage-optimized matC from Rhizobium trifolii (GenBank accession no. KF765784) (8) was excised from pUC57-matC (20) and inserted into the EcoRI/HindIII sites of pACYC-matB to construct pACYC-matC-matB. Sequences of these synthesized genes are shown in the supplemental material (see Fig. S1).

Culture conditions.

To investigate the effect of asRNA species on cell growth, target mRNA levels, and fatty acid content, cells were grown overnight in the absence of the inducer in 500-ml flasks containing 50 ml of MOPS medium at 37°C with shaking at 250 rpm. Cells were cultured at a starting OD₆₀₀ of 0.1 with medium containing 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at 37°C. Ten milliliters of cell culture was harvested in the mid-log phase of growth to quantify target mRNA levels and fatty acid contents. According to previous studies, the silencing efficacy of the various asRNAs was always calculated at mid-log phase (21, 22).

To investigate the effect of the various asRNA species on malonyl-CoA concentration, cells were cultured in 25 ml of MOPS medium at 37°C with 220 rpm orbital shaking. After an OD₆₀₀ of 1.65 was reached, an additional 25 ml of fresh MOPS medium was supplemented (18). asRNA expression was induced with 0.5 mM IPTG. Cultures subsequently were incubated at 30°C. Ten milliliters of cell culture also was harvested at the mid-log phase of growth for quantifying intracellular concentrations of malonyl-CoA according to previous studies, in which the silencing efficacies of asRNA typically were calculated at the mid-log phase (21, 22).

For flavonoid production, cells first were cultured in 25 ml of MOPS medium at 37°C with 220 rpm orbital shaking at a starting OD₆₀₀ of 0.1. After an OD₆₀₀ of 1.65 had been reached, an additional 25 ml of fresh MOPS medium was added (18). Cultures subsequently were incubated at 30°C for (2S)-naringenin production. IPTG, l-tyrosine, and sodium malonate dibasic were provided at a final concentration of 1 mM, 3 mM, and 2 g/liter (8), respectively. After a total fermentation time of 48 h, one milliliter of cell culture was collected to measure the concentrations of (2S)-naringenin and p-coumaric acid, and 10 ml of cell culture was collected to measure the malonyl-CoA concentration (2).

Measurement of flavonoid, p-coumaric acid, malonyl-CoA, and fatty acid concentrations.

Individual triplicate analyses were performed for each sample, and average values of the triplicates are reported. For quantification of the concentrations of (2S)-naringenin and p-coumaric acid, culture supernatant was extracted as indicated in a previous report (2). (2S)-Naringenin and p-coumaric acid were analyzed with an Agilent 1100 series high-performance liquid chromatography (HPLC) instrument equipped with a reverse-phase Gemini NX-C₁₈ column (5 by 110 mm) (8). (2S)-Naringenin was further identified by the area of major mass spectra signals ([M − H]⁻) using a liquid chromatography-mass spectrophotometer ion trap time-of-flight (LCMS-IT-TOF) (Shimadzu, Kyoto, Japan) instrument equipped with an Agilent Zorbax extend-C₁₈ column (4.6 mm by 150 mm; 5 μm) and an electrospray ionization (ESI) source (18) (see Fig. S2 in the supplemental material). Malonyl-CoA from cell culture was extracted and analyzed with an LCMS-IT-TOF equipped with a reverse-phase Gemini NX-C₁₈ column (5 by 110 mm) and an ESI source (6).

For quantification of free fatty acids, cells were harvested and prepared with a previously described protocol (23). The fatty acid content of each sample was quantified using a gas chromatography-mass spectrometry (GCMS)-QP2010 plus (Shimadzu) equipped with an Rtx-5 MS capillary column for analysis as described previously (24). One-tenth of a milligram of pentadecanoic acid was added as an internal standard (24).

RNA preparation and qPCR.

Total RNA was purified using the RNAprep Pure kit for cell/bacteria (Tiangen, Beijing, China). Purified RNA was treated with DNase I and further purified with the RNeasy minikit (TaKaRa, Dalian, China) to remove genomic DNA. Quantitative PCR (qPCR) assays were performed with the One Step SYBR PrimeScript RT-PCR kit II (TaKaRa, Dalian, China) on a LightCycler 480 II thermal cycler system (Roche, Mannheim, Germany). 16S rRNA was used for normalization of qPCR values with primers Pf_16S and Pr_16S (15). fabB and fabF were amplified with the primer pairs Pf_qfabB/Pr_qfabB and Pf_qfabF/Pr_qfabF, respectively.

RESULTS

Construction of initial asRNA expression system.

To evaluate the silencing effect of standard asRNA designs, fabB was used as an example. Growth patterns (Fig. 2A) and transcription levels (Fig. 2C) of fabB were measured in a culture of E. coli BL21(DE3) transformed with the recombinant plasmids pCDF-asRNA(147), pRSF-asRNA(147), and pRSF-PTasRNA(147) and in a control strain culture. Strains with pRSF-PTasRNA(147) (Fig. 2B) showed improved inhibition efficiency relative to strains carrying pRSF-asRNA(147) and pCDF-asRNA(147). This may have been caused by a higher plasmid copy number and the addition of paired-terminus antisense RNA (Fig. 2D). Therefore, the PTasRNA expression vector pRSF-PTasRNA(147) was used as the starter for the optimization of an asRNA regulation system.

Effect of the binding region on the inhibitory efficiency of asRNA.

The sequence that recognizes target mRNA is a crucial part of an asRNA molecule (13, 25, 26). To investigate the effect of the binding region on the inhibitory efficiency of asRNA, six anti-fabB asRNA variant sets binding to different regions of the target mRNA were designed (Fig. 3A). Strains containing each set of asRNA showed similar growth rates and final cell densities. However, they did display different cell growth patterns (Fig. 3C). Transcriptional analysis of fabB showed that the level of fabB mRNA was sensitive to the binding region toward the 5′-UTR of the target mRNA. Choosing an appropriate length (128 bp) for the 5′-UTR resulted in the lowest level of fabB mRNA (set R21-R28, 8% of the original level) (Fig. 3D). For the gene coding region, 147 bp was enough for successful downregulation with an asRNA, whereas other regions were less effective and resulted in slightly less efficient silencing.

FIG 3.

Effect of binding position of asRNAs on inhibition efficiency. (A) The binding sequences of six anti-fabB asRNA variants. The length of the 5′-UTR was set as 158 bp (R11-R18), 128 bp (R21-R28), 98 bp (R31-R38), 68 bp (R41-R48), 38 bp (R51-R58), or 0 bp (R61-R68). In each set, the length of the fabB gene coding region was set as 0 to 847 bp, 0 to 747 bp, 0 to 647 bp, 0 to 547 bp, 0 to 447 bp, 0 to 347 bp, 0 to 247 bp, or 0 to 147 bp (e.g., R11-R18). (B) The binding sequences of four anti-fabF asRNA variants. The length of the 5′-UTR was set as 87 bp (R11-R18), 57 bp (R21-R28), 27 bp (R31-R38), or 0 bp (R41-R48). In each set, the length of the fabF gene coding region was set as 0 to 813 bp, 0 to 713 bp, 0 to 613 bp, 0 to 513 bp, 0 to 413 bp, 0 to 313 bp, 0 to 213 bp, and 0 to 113 bp (e.g., R11-R18). (C) The growth patterns of E. coli strains with anti-fabB asRNA sets. The control was E. coli BL21 strains with pRSFDuet-1(Mut). (D) qPCR analysis for downregulation of fabB gene transcription by asRNA. (E) qPCR analysis for downregulation of fabF gene transcription by asRNA. Error bars are standard deviations from biological triplicates.

To confirm the results presented above, four anti-fabF asRNA variant sets binding to different regions of the target mRNA were designed (Fig. 3B). The transcription levels of fabF showed that strains in the same set had similar levels of fabF mRNA. Strains containing different sets showed different transcription levels (Fig. 3E). Choosing an appropriate length (57 bp) for the 5′-UTR in the second set (R21-R28) resulted in the lowest level of fabF mRNA (13% of the original level). For the gene coding region, 113 bp was enough for downregulation by the asRNA. Other regions were largely ineffective.

The effect of an asRNA system on fatty acid contents in E. coli.

Major fatty acid methyl esters, such as myristic acid methyl ester (C_14:0), palmitoleic acid methyl ester (C_16:1), cis-11-octadecenoic acid methyl ester (C_18:1), palmitic acid methyl ester (C_16:0), and stearic acid methyl ester (C_18:0), were quantified in cell cultures. For anti-fabB asRNA-regulated strains, the concentration of saturated fatty acid (myristic acid, palmitic acid, and stearic acid) in asRNA-regulated strains decreased slightly, while the concentration of unsaturated fatty acid (palmitoleic acid and cis-11-octadecenoic acid) decreased dramatically. Setting the binding region between the fabB 5′-UTR and the asRNA at 128 bp resulted in the lowest level of both saturated fatty acid (70% of the original level) and unsaturated fatty acid (14% of the original level) (Fig. 4A). This demonstrated that fabB plays an important role in the synthesis of unsaturated fatty acids.

FIG 4.

Effect of an asRNA system on fatty acid contents in E. coli. (A) The concentration of saturated fatty acids (myristic acid, palmitic acid, and stearic acid) and unsaturated fatty acids (palmitoleic acid and cis-11-octadecenoic acid) in E. coli strains with anti-fabB asRNA sets. (B) The concentration of saturated fatty acids (myristic acid, palmitic acid, and stearic acid) and unsaturated fatty acids (palmitoleic acid and cis-11-octadecenoic acid) in E. coli strains with anti-fabF sets. The control was E. coli BL21 strains with pRSFDuet-1(Mut). Error bars are standard deviations from biological triplicates.

For strains with anti-fabF asRNA, similar trends in the concentration of fatty acids were observed, with the exception of palmitoleic acid (Fig. 4B). The palmitoleic acid concentration was indistinguishable from that of the control. The concentration of saturated fatty acid (myristic acid, palmitic acid, and stearic acid) in engineered strains decreased slightly, while the concentration of cis-11-octadecenoic acid decreased dramatically. A binding region of 57 bp between the fabF 5′-UTR and the asRNA resulted in the lowest level of both saturated fatty acid (82% of the original level) and cis-11-octadecenoic acid (8% of the original level). These results showed that fabF played an important role in the synthesis of cis-11-octadecenoic acid.

Simultaneously repressing fabB and fabF by asRNA system.

To evaluate whether this asRNA system could simultaneously repress fabB and fabF, different antisense sequences of fabB and fabF were cloned into pRSFDuet-1(Mut) (see Materials and Methods for details). The transcriptional levels of target gene mRNA from asRNA-regulated strains were calculated relative to the control. As seen from Fig. 5, the transcriptional levels of target gene mRNA from asRNA-regulated strains decreased following the same trend as that described above. For strain fabF-R28+fabB-R28, the lowest transcriptional level of both fabF (18% of the original level) and fabB (12% of the original level) were observed (Fig. 5). This demonstrated that coexpression of two asRNAs could knock down both fabB and fabF.

FIG 5.

Simultaneously repressing fabB and fabF by asRNA system. The transcriptional levels of target gene mRNA from asRNA-regulated strains also were calculated relative to the control. Control indicates BL21(DE3) strains transformed with pRSFDuet-1(Mut). Designations using the format fabF-RX+fabB-RY were BL21(DE3) strains transformed with plasmids expressing fabF asRNA sequence (RX) and fabB asRNA sequence (RY) simultaneously, where X is 18, 28, or 38 and Y is 18, 28, 38, 48, or 58. Error bars are standard deviations from biological triplicates.

The effect of an asRNA system on the intracellular malonyl-CoA content.

To examine whether an asRNA system has the potential to enhance the synthesis of malonyl-CoA, the antisense sequences of fabB and fabF were cloned into pRSFDuet-1(Mut) to simultaneously silence fabB and fabF. The final OD₆₀₀ and the concentration of intracellular malonyl-CoA from asRNA-regulated strains were quantified. The concentration of malonyl-CoA varied depending on which binding regions within the 5′-UTRs of fabB and fabF were targeted (Fig. 6). While the silencing effect toward fabF was constant, malonyl-CoA accumulated with increased silencing of fabB, and maximum accumulation was observed when the binding region within the fabB 5′-UTR was set to −128 bp (0.88 nmol/mg dry cell weight [DCW]).

FIG 6.

Effect of asRNA system on the intracellular malonyl-CoA concentration. Control indicates BL21(DE3) strains transformed with pRSFDuet-1(Mut). Designations using the format fabF-RX+fabB-RY are BL21(DE3) strains transformed with plasmids expressing fabF asRNA sequence (RX) and fabB asRNA sequence (RY) simultaneously, where X is 18, 28, or 38 and Y is 18, 28, 38, 48, or 58. The effects of asRNA systems on malonyl-CoA concentration and the final OD₆₀₀ of engineered strains were investigated. Error bars are standard deviations from biological triplicates.

Similarly, while the silencing of fabB was constant, malonyl-CoA accumulated with increased silencing of fabF, and a maximum accumulation was observed when the binding region of the fabF 5′-UTR was set to −57 bp (0.88 nmol/mg DCW) (Fig. 6). This recombinant strain exhibited an 8.8-fold increase in the levels of intracellular malonyl-CoA compared to that of the wild-type strain (0.09 nmol/mg DCW). Furthermore, the asRNA system (fabF-R28+fabB-R28), which increased the malonyl-CoA concentration by 8.8-fold, decreased the final OD₆₀₀ by 28% (Fig. 6).

The effect of an asRNA system on (2S)-naringenin production.

Based on our previous study (20), a control strain overproducing (2S)-naringenin from l-tyrosine was constructed. The plasmids pCDF-Trc-TAL-Trc-4CL, pET-CHS-CHI, and pACYC-matC-matB were transformed into E. coli BL21(DE3), yielding a production titer of 91 mg/liter. To simultaneously repress fabB and fabF, the binding regions for the anti-fabB asRNAs were set as −38 to 147 bp, −68 to 147 bp, and −98 to 147 bp in the target mRNA, which yielded repression levels of 38%, 44%, and 62%, respectively, as indicated by the reduced mRNA level of the target gene. The binding regions of the anti-fabF asRNA were set as −27 to 113 bp, −57 to 113 bp, and −87 to 113 bp in the target mRNA, which yielded repression levels of 58%, 84%, and 72%, respectively.

When the inhibitory efficiency of asRNA toward fabF was set at 58% with increasing inhibitory efficiency of the asRNA toward fabB (see strains fabF-R38+fabB-R58 to fabF-R38+fabB-R38 in Fig. 7), the malonyl-CoA concentration remained at a low level (0.08 to 0.09 nmol/mg DCW) until an intermediate value of inhibition efficiency toward fabB was reached. After that intermediate value (see strain fabF-R38+fabB-R48), further increases in repression efficiency resulted in an increase in the malonyl-CoA concentration (0.09 versus 0.35 nmol/mg DCW). Furthermore, the (2S)-naringenin concentration decreased (391 versus 260 mg/liter), while the p-coumaric acid concentration increased. Similarly, when the inhibitory efficiency of asRNA toward fabF was set to 72% and 84% with an increasing inhibitory efficiency of asRNA toward fabB, an increase in the concentration of malonyl-CoA and p-coumaric acid and a reduced concentration of (2S)-naringenin also was observed. It was believed that the intermediate value of inhibition efficiency toward fabB and fabF approached the limit of malonyl-CoA saturation.

FIG 7.

Effect of various asRNA systems on (2S)-naringenin production. Control, strains containing the (2S)-naringenin heterologous pathway without an asRNA system. Designations using the format fabF-RX+fabB-RY indicate the control strain transformed with plasmids expressing fabF asRNA sequence (RX) and fabB asRNA sequence (RY) simultaneously, where X is 18, 28, and 38 and Y is 38, 48, and 58. The final OD₆₀₀ of cultures and concentrations of malonyl-CoA, p-coumaric acid, and (2S)-naringenin were measured from production strains after a total fermentation time of 48 h. Error bars are standard deviations from biological triplicates.

DISCUSSION

The small amount of intracellular malonyl-CoA in E. coli impedes the biological synthesis of fatty acids, polyketides, and flavonoids (8, 27). Previous approaches to the redirection of endogenous malonyl-CoA into heterologous pathways have relied on conventional gene knockout strategies (3, 5, 6). However, these strategies typically cause poor biomass accumulation and reduced cofactor regeneration (28). Here, an efficient asRNA-based strategy for cascading regulation of genes at the posttranscriptional level was developed. By simultaneously repressing fabB and fabF, the malonyl-CoA concentration increased by 8.8-fold while the final OD₆₀₀ decreased by 28% (Fig. 6). The presented strategy could be employed for the efficient production of many useful compounds that require malonyl-CoA as a precursor, such as the flavanones (2, 18), resveratrol (29), and polyketides (6).

In E. coli, malonyl-CoA is consumed only for synthesizing fatty acids (6, 7). Thus, inhibition of the fatty acid pathway would abolish a competing pathway consuming malonyl-CoA. It has been demonstrated that repressing the fatty acid pathway by antibiotic cerulenin, which specifically inactivates the β-ketoacyl-acyl carrier protein synthases (KAS) I and II, could trigger a large accumulation of malonyl-CoA (2, 8, 9). Hence, fabB and fabF were chosen as target genes for enhancing (2S)-naringenin production. However, in some particularly relevant studies, previous researchers have found that simultaneous deletion of genes sdhA, adhE, brnQ, and citE (5) or deletion of fumC and sucC (3) could increase the (2S)-naringenin level dramatically. It is proposed that a rational combination of these strategies could further increase (2S)-naringenin production.

RNA interference (RNAi) has emerged as a powerful tool for conditional gene silencing in eukaryotic systems (11, 30, 31). However, similar tools for bacteria that are as reliable and efficient as eukaryotic RNAi have been elusive (15). There has been increasing interest in the use of asRNA for various applications in bacteria (15, 22, 28). Although many studies find that the 5′-UTR, especially the ribosomal binding site, is the most sensitive for asRNA silencing (13, 15, 22), it is unclear how the start point, binding location, and RNA structure affect the inhibitory efficiency. Here, it was demonstrated that (i) the repression efficiency of asRNA could be tuned by directing asRNA to different positions in the 5′-UTR of the target gene, and the most efficient gene silencing was achieved by directing the asRNA to certain positions in the 5′-UTR of the target gene; and (ii) for the gene coding region, approximately 110 bp of target sequence is enough for downregulation of asRNA, whereas longer targeted sequences of the gene coding region result in a slight decrease in silencing efficiency (Fig. 3). It is believed that, by understanding these principles, researchers have the potential to rationally design highly active artificial asRNA.

The asRNA system described here increased the intracellular malonyl-CoA level from 1.1-fold to 8.8-fold compared to wild-type strains (Fig. 6). However, when placed in the context of flavonoid production, the accumulation of malonyl-CoA may not be accommodated by the capacity of the downstream pathway. This may result in suboptimal product titers (18, 32). It was demonstrated that the appropriate silencing of target genes (see strain fabF-R38+fabB-R48) would result in the best production titer (391 mg/liter) (Fig. 7). Further gene-silencing effects led to the accumulation of intermediate metabolites, such as malonyl-CoA and p-coumaric acid, and decreased the final product yield. The asRNA system described here could be helpful for balancing the expression of multiple enzymes and achieving optimized performance in other biological systems.

Researchers have made significant gains in the production of (2S)-naringenin from various sources, including d-glucose (84 mg/liter) (2), l-tyrosine (60 mg/liter) (16), and p-coumaric acid (474 mg/liter) (3). However, supplementation of p-coumaric acid as the precursor for biotransformation has hampered the broader adoption of microbial synthesis due to its high price and poor water solubility (2). Direct (2S)-naringenin production from d-glucose in E. coli suffers from a low turnover rate of l-tyrosine to (2S)-naringenin (2). The engineered strains described here possess robust cellular constitutions and exhibit efficient conversion of l-tyrosine to (2S)-naringenin. As numerous studies have improved the biosynthesis of l-tyrosine from d-glucose (33, 34), it is very likely that these strains can be further improved for direct flavonoid production from d-glucose. This would pave the way for the development of a simple and economical process for the microbial production of other flavonoids.

The asRNA design procedure described here provides a framework for optimizing and applying asRNA-mediated gene silencing for diverse applications. These artificial asRNAs, which enable fine-tuning of bacterial genes, have promising applications in synthetic biology and the characterization of genes that are essential for cell growth (13, 35, 36). It is believed that the tunable inhibitory efficiency of asRNA described here could facilitate the identification and characterization of new genes, especially essential genes, that have different phenotypes with different expression levels. In addition, the regulation of those essential genes at different levels could facilitate synthetic biology research as an efficient multilevel control unit.