All strains and plasmids used in this study are listed in Table 1. E. coli DH5α was used as the host to propagate the plasmid DNA. E. coli BL21 was used as an expression host. The compatible plasmids pETDuet-1 and pACYCDuet-1 were used as expression plasmids. Luria-Bertani (LB) medium (10 g/l tryptone, 5 g/L yeast extract, and 10 g/l NaCl) was used for strain culture. M9Y medium (10/L glucose, 6.8 g/l Na₂HPO₄, 3.4 g/l KH₂PO₄, 0.5 g/l NaCl, 1.0 g/l NH₄Cl, 5 g/l yeast extract, 2 mM MgSO₄⋅7H₂O, and 0.1 mM CaCl₂; pH 7.2), LBG medium (LB medium with 10 g/l glucose; pH 7.2), and MSY media (10 g/l glucose, 3.8 g/l Na₂HPO₄, 1.5 g/l KH₂PO₄, 1.0 g/l (NH₄)₂SO₄, 0.2 g/l MgSO₄⋅7H₂O, 5 g/l yeast extract, and 2% (v/v) trace element solution, pH 7.2) were used (Qiu et al., 2004), and the composition of trace element solution was as follows: 100 mg/l ZnSO₄⋅7H₂O, 30 mg/l MnCl₂⋅4H₂O, 300 mg/l H₃BO₃, 200 mg/l CoCl₂⋅6H₂O, 10 mg/l CuSO₄⋅5H₂O, 20 mg/l NiCl₂⋅6H₂O, 30 mg/l NaMoO₄⋅2H₂O, and 0.5 M HCl. Ampicillin (100 μg/ml), kanamycin (50 μg/ml), spectinomycin (50 μg/ml), chloramphenicol (25 μg/ml), and isopropylthio-β-galactoside (IPTG) were added if necessary. Unless otherwise stated, incubation was carried out at 37°C and 220 rpm, and the IPTG concentration was 1 mM.

The FastPure plasmid mini kit, gel DNA extraction mini kit, bacterial DNA isolation mini kit, ClonExpress II one-step cloning kit, and FastPure cell/tissue total RNA isolation kit V2 were purchased from Vazyme Biotech (Nanjing, China). PrimeSTAR HS (premix), a competent cell preparation kit, and restriction enzymes (NdeI, XhoI, and DpnI) were purchased from Takara Biotech (Dalian, China). Primer synthesis and sequencing were performed by Shanghai Biotech (Shanghai, China).

All primers used in this study are listed in Table 2. The primers ribA-F/R, ribB-F/R, ribC-F/R, ribD-F/R, and ribE-F/R were used to obtain the gene fragments ribA, ribB, ribC, ribD, and ribE with the E. coli BL21 genome as the template. The artificial operon ribC–ribE–ribB–ribD–ribA was obtained by ligating the gel extraction products of the five genes through overlap-PCR with the primers ribC-F and ribA-R. After the plasmid pETDuet-1 was digested by NdeI and XhoI, the plasmid pET-AE was constructed by ligating the artificial operon ribC–ribE–ribB–ribD–ribA with the digested plasmid pETDuet-1 using a ClonExpress II one-step cloning kit (Figure 3A). The plasmid pAC-Z was obtained by amplifying the zwf gene from the E. coli BL21 genome with the primer zwf-F/R and conducting homologous combination with pACYCDuet-1, which was digested with NdeI and XhoI (Figure 3B).

The plasmid pET-AE was transformed into E. coli BL21 competent cells to generate the R1 strain by the CaCl₂-heat shock method (Fu et al., 2021). The R2 strain was constructed by introducing the plasmid pAC-Z into the R1 strain to increase the carbon metabolism level and facilitate carbon flow through the PPP.

A single colony was picked from agar plates and incubated in LB media containing appropriate antibiotics until the OD₆₀₀ values reached 0.6. A final concentration of 2 mM IPTG was added to the cultures to induce the expression of the target proteins. After 6 h of incubation, the cells were collected through centrifugation at 12,000 rpm for 10 min. Phosphate buffer (PBS, pH 7.4) was used to resuspend the obtained cells, and ultrasonic treatment was performed by a sonicator (SCIENTZ-II D, China) to release the proteins (5 s sonication at 150 W, 10 s pause). The resultant crushing solution was centrifuged (12000 rpm, 10 min, 4°C). The precipitate was resuspended in PBS buffer (pH 7.4).

Approximately 10 μl of the precipitate and supernatant were subjected to SDS-PAGE analysis performed on a 12% running gel. Coomassie brilliant blue G-250 was used to stain the resolved proteins.

The CRISPR/Cas9 system is composed of pCas and pTargetF plasmids. Owing to its efficiency and simplicity, it is widely used in genomic editing (Liu et al., 2017; An et al., 2020).

The designed N20 sequence was evaluated with a web-based tool.¹ sgRNA was obtained through reverse-PCR using primers with modified N20 at the 3′ end and pTargetF as the template. The PCR products were digested by DpnI at 37°C for 1 h to remove the template. The products were recovered and purified for the transformation of E. coli DH5α competent cells and incubated on an LB agar plate (50 μg/ml spectinomycin). Positive clones were screened out through sequencing.

pCas was transformed into E. coli BL21 competent cells. The positive clone was incubated in LB medium with kanamycin (50 μg/ml) until the OD₆₀₀ reached 0.2. L-arabinose (30 mM) was added to the culture to induce the expression of Gam, Beta, and Exo. The cultures were collected for the preparation of electrocompetent cells after the OD₆₀₀ reached 0.5. Donor DNA as the repair template which was used for homologous recombination was constructed as follows: approximately 500 bp of the upstream and downstream homologous arms were obtained through PCR, and E. coli DL21 genome DNA was used as template. The PCR products were ligated to generate donor DNA through overlap-PCR. The concentrations of sroG-sgRNA and donor DNA were measured with a microspectrophotometer (Thermo Scientific NanoDrop, United States). Approximately 400 ng of gRNA and 2 μg of donor DNA were added to E. coli BL21 electrocompetent cells, which harbored pCas, after electroporation at 2.5 kV in a 2 mm Gene Pulser cuvette (Bio-Rad, United States). The culture was separated on an LB agar plate (50 μg/ml spectinomycin and kanamycin) and incubated at 30°C overnight. Colony PCR and sequencing were carried out to screen the positive clones.

pCas contains a Cas9 protein from Streptococcus pyogenes MGAS5005 with its native promoter, the λ-Red gene under the control of the L-arabinose-inducible promoter (ParaB), temperature-sensitive replicon repA101ts for self-curing, and a sgRNA containing an N20 sequence targeting the pTarget pMB1 replicon (sgRNA-pMB1) under the control of the IPTG-inducible promoter (Ptrc; Jiang et al., 2015). Plasmid curing was performed as follows: the positive clone was incubated in LB medium containing 0.5 mM IPTG and 50 μg/ml spectinomycin and cultured overnight at 30°C. pCas was cured on LB medium overnight at 42°C. When a strain was subjected to further genomic editing, pCas curing was not performed.

The R3 strain was generated by deleting the FMN riboswitch from primitive E. coli BL21 with the CRISPR/Cas9 system according to the above method. The R4 strain was constructed by cotransforming pET-AE and pAC-Z into the R3 strain.

Given that the binding of FMN to the FMN riboswitch represses ribB expression, the effect of the FMN riboswitch deletion on ribB transcript levels was detected through quantitative real-time PCR (qPCR). The E. coli BL21, R2, R3, and R4 strains were incubated in LB medium at 37°C. When the OD₆₀₀ reached 0.6, 2 mM IPTG was added to the LB medium to induce the transcription of ribB. The cells were harvested when O₆₀₀ reached 1.0. Total RNA was extracted from the E. coli cells with a bacterial total RNA isolation kit. After cDNA was synthesized with an M5 Super plus qPCR RT kit (Mei5bio, Beijing, China) with the total RNA as the template, qPCR was performed on a LightCycler480 (Roche, Basel, Switzerland) with 2 × RealStar Green Fast Mixture (GenStar, China).

Differences in ribB gene transcript levels between the engineered strains and wild-type E. coli BL21 were measured in triplicate, normalized to the 16S internal control, and calculated according to the 2^−∆∆CT method (Livak and Schmittgen, 2001).

Batch shake-flask culture was used to compare the RF production capacities of different strains. The engineered R4 strain was used to optimize the fermentation conditions (medium, glucose concentration, incubation temperature, and IPTG concentration). The R4 strain was inoculated into 10 ml of LB medium and cultured overnight as a preseed culture. Preseed cultures (0.5 ml) were added to 50 ml of LB and cultured to mid-exponential growth phase at 220 rpm as the seed cultures. One percent of each seed culture (v/v) was transferred to LB (or M9Y or MSY) medium to produce RF. When the OD₆₀₀ reached 0.6, IPTG was added to induce the expression of target genes for producing RF. If necessary, appropriate antibiotics were added to the medium. Time-course analysis of RF production in a 1 L shake flask was also performed according to the above procedure.

OD₆₀₀ was measured with a PGENERAL New Century T6 spectrophotometer (Beijing, China). The concentration of glucose was measured by the 3,5-dinitrosalicylic acid method (Hu et al., 2008). The concentration of RF was determined by high-performance liquid chromatography (HPLC; Agilent 1,260 Infinity, CA, United States). Bacterial cultures were diluted before centrifugation because RF is poorly soluble under neutral or acidic conditions and form crystals. The supernatants were filtered (0.22 μm inorganic membrane) and analyzed under the following conditions (Pedrolli et al., 2015a): mobile phase, 18% (v/v) methanol–20 mM formic acid–20 mM ammonium formate (pH 3.7); flow rate, 0.8 ml/min; injection volume, 10 μl; column, ZORBAX SB-C18 (4.6 × 150 mm; Agilent, CA, United States); and RF detection at 445 nm.

All experiments were performed in triplicate. The results are presented as the mean ± SD of three independent experiments. SPSS 17.0 software was used for statistical analysis. Data were graphed using Origin 8.5 software. The plasmid construction map was generated with ChemDraw Professional 2017 software.

For the construction of the RF-producing strain R1, the plasmid pET-AE was transformed into E. coli BL21 competent cells. The R2 strain with enhanced glucose utilization ability was obtained by transforming the plasmid pAC-Z into the R1 strain.

The expression of the target proteins (RibA, RibB, RibC, RibD, RibE, and Zwf) in the recombinant strain R2 was verified through SDS-PAGE. The electrophoresis results are shown in Figure 4A. The apparent molecular weights of the six proteins were 22, 23.4, 23, 40, 16.2, and 52 kDa, respectively, which were consistent with the expected values, indicating that the target proteins were correctly expressed in the recombinant strain R2.

The RF production ability of the recombinant strains was measured through HPLC, and the OD₆₀₀ was measured with a spectrophotometer. The key fermentation parameters (OD₆₀₀ and RF titer) in the strains R1 and R2 are listed in Table 3. After 72 h of incubation, the engineered strain R1 produced 182.65 ± 9.04 mg/l RF, and the OD₆₀₀ reached 7.18 ± 0.36. The engineered strain R2 produced 319.01 ± 20.65 mg/l RF, and the OD₆₀₀ reached 8.76 ± 0.12, showing 74.66 and 22.01% increases compared to the R1 strain, respectively. These results suggested that the overexpression of the key genes in E. coli BL21 conferred on the strain the ability to produce RF and that the overexpression of zwf on this basis further enhanced RF production and promoted the growth of the recombinant E. coli BL21.

To relieve FMN inhibition, a total of 223 bp of nucleotides located upstream of the RBS sequence of ribB were deleted. sgRNA was obtained through reverse-PCR using the primer sroG-sgRNA-F/R and designated sroG-sgRNA. The positive clones were screened out through sequencing after plasmid extraction. The primers sroG-armL-F/R and sroG-armR-F/R were used to amplify the upstream and downstream homologous arms, sroG-Up and sroG-Down (with length of approximately 500 bp), with E. coli BL21 genome DNA as the template. The primers sroG-armL-F and sroG-armR-R were used to obtain approximately 1-kb-long donor DNA through overlap-PCR. After cotransformation with sroG-sgRNA and donor DNA, the positive clones with 1,026 bp bands were screened out through colony PCR, whereas the negative clones had 1,243 bp bands (Figure 4B). The sequencing results further confirmed the successful knockout of the FMN riboswitch. Before proceeding to the next step, strain R3 was successively subjected to IPTG induction and 42°C culture to cure the sroG-sgRNA and pCas. After incubation in 5 ml of LB medium for 24 h, the color of the supernatant turned yellow, indicating that the deletion strain R3 had derepressed the inhibition of RF biosynthesis by FMN compared to the wild-type E. coli BL21 (Figure 4C).

Northern blot analysis showed that the deletion of the FMN riboswitch in E. coli CmpX13 led to the constitutive synthesis of ribB mRNA (Pedrolli et al., 2015a). To compare the effects of FMN riboswitch deletion and ribB overexpression on ribB transcript levels, the ribB transcript levels of E. coli BL21, R2, R3, and R4 were quantified through qPCR. As shown in Figure 4D and Table 4, compared to those of E. coli BL21, the ribB transcript levels of R2, R3, and R4 improved 2.78, 2.57, and 3.05-fold, respectively. These results suggested that the deletion of the FMN riboswitch could derepress the effect of FMN on ribB transcription. However, knockout of the FMN riboswitch on the basis of the overexpression of ribB resulted in only a slight increase in the transcript level of ribB (an increase of 9.7%). This result is consistent with that reported by Pedrolli et al. (2015a).

After 72 h of incubation in M9Y medium, the FMN riboswitch deletion strain R3 accumulated 4.46 ± 0.23 mg/l RF, increasing by 3.8-fold compared to the control strain E. coli BL21 (0.93 ± 0.31 mg/l). In addition, no significant difference was observed in OD₆₀₀ values between the R3 strain and E. coli BL21 (Table 3), suggesting that the deletion of the FMN riboswitch did not affect the growth of the strain. The engineered strain R4 produced 437.58 ± 14.36 mg/l RF, and the OD₆₀₀ value of the cells reached 8.29 ± 0.29. Compared to the engineered strain R2, the RF titer of the engineered strain R4 increased by 0.37-fold (Table 3). This result showed that the deletion of the FMN riboswitch on the basis of the overexpression of key genes was effective in enhancing the RF production capacity of the engineered strain.

Riboflavin (RF) production can be further improved by optimizing fermentation conditions, such as culture medium, glucose concentration, culture temperature, and IPTG concentration (Figure 5). The engineered strain R4 was chosen for the optimization of the culture conditions. M9Y, MSY, and LB culture media were used to screen the optimum medium for RF production in this study. As shown in Figure 5A, MSY was the most suitable medium for the fermentation of the R4 strain for RF production, and the titer of RF in MSY medium after incubation for 72 h was 505.01 ± 45.88 mg/l. A variety of metal elements (e.g., Co²⁺, Mn²⁺, etc.) in the trace element solution in MSY may be cofactors for key enzymes in the de novo synthesis of RF. Therefore, MSY medium was effective in enhancing RF production in the engineered strain R4. Thus, MSY was selected as the fermentation medium, and a glucose concentration gradient (10, 20, 30, and 40 g/l) was set up to screen the optimal glucose concentration. As shown in Figure 5B, the optimum glucose concentration was 20 g/l, and the titer of RF reached 537.88 ± 10.26 mg/l. A temperature gradient (28, 30, 37, and 42°C) was used to determine the optimum fermentation temperature. MSY containing 20 g/l glucose was used as the fermentation medium. As shown in Figure 5C, the highest RF titer was obtained at 37°C, which was consistent with the optimum growth temperature for E. coli. The titer of RF reached 537.88 ± 10.26 mg/l in the engineered strain at 37°C. Since strain R4 contained compatible pET-AE and pAC-Z plasmids under the control of the T7-Lac promoter, an induction concentration gradient of IPTG (0.1, 0.25, 0.5, 1.0, 1.5, and 2.0 mM) was set up for the identification of the most suitable IPTG induction concentration. The medium was MSY (20 g/l glucose), and the temperature was 37°C. As shown in Figure 5D, the highest RF titer was obtained at 2 mM IPTG. The RF titer reached 611.22 ± 11.25 mg/l. Taken together, the optimum fermentation conditions of the engineered strain R4 for the production of RF were as follows: medium, MSY; glucose, 20 g/l; temperature, 37°C; and IPTG concentration, 2 mM.

For fed-batch cultivation, 1% (v/v) seed cultures were added to 250 ml of MSY with 40 g/l glucose in a 1 l shake flask (10 g/l glucose was added to the initial medium and feed medium separately), and the mixture was shaken at 37°C and 220 rpm. 3-(N-morpholino) propanesulfonate (MOPS, 50 mM) was added to the cultures to maintain a neutral pH. IPTG at a final concentration of 2 mM induced the expression of the target genes. Ammonia was used to adjust the pH to 7.2 during fermentation. As shown in Figure 6, with the addition of glucose, the RF titer and the OD₆₀₀ values of the cells increased quickly. The RF titer was the highest after induction with IPTG for 96 h (1574.60 ± 109.32 mg/l with a yield of 12.00 ± 0.83 mg/g glucose). Meanwhile, the OD₆₀₀ value of the cells reached a maximum of 13.9 ± 0.87. Afterward, these values began to decline due to the depletion of glucose and the accumulation of byproducts.