The strains used in this study were listed in Supplementary Table 1, E. coli DH5α served as the host for recombinant DNA manipulation and plasmid construction. For cis-3-HyPip production, E. coli BL21(DE3) and derived strains were used. The plasmid pEcgRNA carrying the ccdB gene was constructed and maintained in E. coli DB3.1. The DNA polymerases used for polymerase chain reaction (PCR) including 2 × Phanta Max Master Mix, 2 × Taq Master Mix and ClonExpress One Step Cloning Kit were purchased from Vazyme (Nanjing, China). Molecular biological reagents, such as T4 DNA ligase and DNA gel extraction kit, were obtained from TaKaRa (Dalian, China). Isopropyl β-D-1-thiogalactopyranoside (IPTG), ampicillin, spectinomycin, and kanamycin were provided from Sangon Biotech (Shanghai, China). All other chemicals were purchased from Sigma-Aldrich (Shanghai, China) or Sangon Biotech and were of analytical grade (Shanghai, China).

All strains were grown at 37°C and shaken at 220 rpm in lysogeny broth (LB). Based on plasmid maintenance requirements, ampicillin (100 μg/mL), kanamycin (50 μg/mL), or spectinomycin (50 μg/mL) were added to the medium. For recombinant strain expression, overnight cultures were inoculated into medium [12 g/L tryptone, 24 g/L yeast extract, 0.25% (w/v) glycerol, 2.3 g/L KH₂PO₄, and 12 g/L K₂HPO₄, 2.0 mM MgSO₄, and trace metal mix containing 0.05 mM FeCl₃, 0.02 mM CaCl₂, 0.01 mM MnCl₂, 0.01 mM ZnSO₄, and 2 mM each of CoCl₂, NiCl₂, Na₂MoO₄, Na₂SeO₃, and H₃BO_3]. Cultures were shaken at 200 rpm and incubated at 37°C until the OD₆₀₀ reached 0.4–0.6. Cells were then induced with 0.3 mM IPTG for 24 h at 20°C. After the end of the cultivation, cells were harvested by centrifugation and washed twice with 0.85% NaCl solution.

The strains and plasmids used in the present study are described in Supplementary Table 1. The primers used for gene cloning in this study were synthesized at General Biol (Chuzhou, China). Primer sequences are listed in Supplementary Tables 2–6. The sequences of all constructed plasmids were verified using DNA sequencing.

The recombinant plasmids were introduced to E. coli BL21 (DE3) and then the cells were induced with 0.3 mM IPTG for 24 h at 20°C and collected by centrifugation. Under the same reaction conditions, the ability of recombinant strains was evaluated in terms of product yield. The reaction system consisted of 50 mM L-lysine, 60 mM 2-OG, 10 mM Fe²⁺, 10 mM L-ascorbate (VC), 0.1% (wt/v) trition and cells with a final OD₆₀₀ of 30 in PBS buffer (50 mM). The reaction was carried out in a 50 mL Erlenmeyer flask for 20 h at 30°C with shaking at 250 rpm and then terminated by boiling in a water bath for 3 min. All reactions were carried out in triplicate.

Electroporation of competent E. coli BL21(DE3) carrying pEcCas was carried out as described previously (Li et al., 2021). For λ-Red induction, L-arabinose (30 mM final concentration) was added to the culture. For genome editing, electroporation was performed using MicroPulser (Bio-Rad) with the default E. coli program (2 mm, 1.8 kV, and 6.1 ms). Each electroporation reaction included donor DNA and the plasmid pEcgRNA-ΔsucA. The electroporation mixture was suspended in 1 mL of LB medium right away. Cells were recovered by incubating for 2 h at 37°C before spreading on LB plates containing kanamycin (50 μg/mL) and spectinomycin (40 μg/mL) and incubating overnight at 37°C. Colonies were chosen at random and successful transformation was confirmed using colony PCR and DNA sequencing.

A colony of the edited clone containing both pEcCas and pEcgRNA-ΔsucA was inoculated into 2 mL LB medium containing rhamnose (20 mM) and kanamycin (50 μg/ml) to eliminate the pEcgRNA-ΔsucA and pEcCas series plasmids (Figure 1). The culture was shaken at 220 rpm for 8 h before being transferred into a liquid LB medium without antibiotics and then grown for another 8 h. The cells were spread on LB plates containing 10 μg/mL sucrose and incubated overnight at 37°C on LB plates with or without kanamycin (50 μg/mL). LB plates, LB plates with kanamycin, and LB plates with spectinomycin were used to screen for positive colonies. The pEcgRNA-ΔsucA-cured colonies were used in the next round of aceA gene editing. The methods used for aceA gene editing were identical to those used for sucA deletion.

For the biotransformation, cells were suspended in the reaction mixture containing 0.5% (wt/v) glucose, 50 mM L-lysine, 10 mM Fe²⁺, 10 mM VC, 0.1%(w/v) Triton X-100 and cells with a final OD₆₀₀ of 30 in PBS buffer (50 mM). The bioconversion reactions were performed at 30°C and 250 rpm on a shaker. Aliquots were taken at various time points, centrifuged, and heated at 100°C for 5 min to terminate the reaction. At least three independent experiments were performed for each strain.

Cell growth was evaluated by measuring the absorbance at 600 nm using an UV spectrophotometer. For Fmoc derivatization, 50 μL of the culture supernatant was diluted with 250 μL of 100 mM borate buffer (pH 9.0). Diluted samples were derivatized with 10 mM Fmoc-Cl (TCI, Tokyo, Japan) in acetone (300 μL) and vigorously mixed and then reacted for 30 min. These mixtures were quenched with 600 μL of 25% acetonitrile/250 mM borate buffer (pH 5.5).

The amounts of L-lysine, L-Pip, and cis-3-HyPip were determined using a high-performance liquid chromatography system (Agilent 1260, Agilent Technologies, Santa Clara, CA, United States) with an HC-C18 column (Agilent Technologies, Santa Clara, CA, United States, 4.6 × 250 mm, 5 μm) and a flow rate of 1.0 mL/min at 40°C. The solvent gradient used was as follows: Buffer A [0.1% trifluoroacetic acid (TFA) water], B (0.1% TFA acetonitrile); 0–5 min, 30% B; 8–15 min, 40% B; 15–20 min, 40–50% B; 20–25 min, 50% B; 25–30 min, 50–90% B; 30–38 min, 90% B; 38–41 min, 90–30% B; and 41–45 min, 30% B.

In our previous study, a dual-enzyme system containing SpLCD and GetF was constructed and reaction parameters were systematically optimized to biosynthesize cis-3-HyPip from L-lysine (Hu et al., 2022). However, the development of this whole-cell cascade catalyst is challenged by difficulty in balancing the rates of the two main reaction steps, namely, eliminating intermediate pipecolic acid accumulation and shifting the equilibrium toward cis-3-HyPip production. Tuning translation rates by varying the strength of RBS sequences has been widely used as an effective approach to balance the expression of biosynthetic enzymes in metabolic pathways and optimize the synthesis of target metabolites due to convenience and high efficiency (Liu et al., 2019b; Li et al., 2020).

In the present study, the translation initiation rates of GetF and SpLCD were refined to enhance the titer of cis-3-HyPip. The expression levels of GetF and SpLCD were fine-tuned using RBS sequences with different translation initiation rates, and six RBS sequences were preliminarily selected in this study. T7 RBS originally existed in the Duet vectors, and five were from the iGEM Standard Parts and the RBS strength order was RBS 29 > 30 > RBS(T7) > 32 > 64 > 31 (Zhu et al., 2021). Using strain 1, which harbors pETduet-1-RBS(T7)-getf-RBS(T7)-splcd, as a starting strain, the GetF expression was combined with RBS sequences of various strengths to construct five strains (Figure 2A). As shown in Figure 2B, the efficiency of whole-cell catalysis was enhanced by increasing the strength of the RBS sequence. Finally, the highest activity was exhibited by the cells containing RBS 29 and the cis-3-HyPip productivity increased 30% through RBS optimization. Based on these results, the strain 2 harboring pETduet-1-29-getf-RBS(T7)-splcd as a starting strain, the SpLCD expression were combined with RBS sequences of various strengths to construct other five strains. As shown in Figure 2C, we found that the titers were much more impacted by the type of RBS used for GetF expression than that used for SpLCD expressions. Furthermore, it also showed that a moderate translational initiation rate of SpLCD is more suitable than the high or low rates, and that neither higher nor lower initiation rates not improved whole-cell activity. Similar results were also reported in previous studies, high-intensity RBS regulates the expression of non-rate-limiting enzymes, which may not result in a high yield of target products. Instead, medium-intensity RBS was observed to achieve a relatively highest yield (Alagesan et al., 2018; Sun et al., 2020). Finally, 32 mM cis-3-HyPip was obtained using strain 2 as the biocatalyst (Figure 2D). These results indicated that regulating gene expression by optimizing RBS strength could be utilized to solve the effect of rate-limiting steps and was useful for the target product yield improvement (Ge et al., 2022).

GetF catalyzes the conversion of L-Pip to cis-3-HyPip, which requires 2-OG as a co-substrate, resulting in a high cost. Several studies have linked glutamate to the 2-OG conversion reaction (Wu et al., 2018; Sun et al., 2019). This reaction also produces H₂O₂, which can impair enzymatic activity (Niu et al., 2014; Jing et al., 2020). In the TCA cycle, 2-OG is an important intermediate metabolite. To simplify the technological processes and lower the overall production costs, a whole-cell biocatalyst was reconstructed by using a metabolic engineering strategy in E. coli. This intervention could introduce a 2-OG accumulation system, eliminating the need for additional 2-OG supplementation (Figure 4A).

To engineer the metabolic flux of the TCA cycle, genes encoding sucA and aceA were separately knocked out in E. coli BL21(DE3). In this study, the sucA and aceA genes were knocked out in BL21(DE3) using the pEcCas/pEcgRNA system (Li et al., 2021). As shown in Figure 3A, the PCR amplification fragment was 4,563-bp (lane 2), indicated that sucA was not knocked out, while the PCR amplification fragment was 1,761-bp (lane 1), indicating that the sucA was knocked out. The PCR amplification fragment was 1,586-bp (lane 3), indicating that that it had been knocked out. By contrast, the 2,891-bp PCR amplification fragment (lane 4) indicated that aceA had not been knocked out (Figure 3B).

Three E. coli mutants were obtained through knockout, and the plasmid pETduet-029-getf-RBS-splcd was transformed into these engineered strains, respectively. In the resulting recombinant strains (strain 12, strain 13, and strain 14) showed higher production of cis-3-HyPip than the starting strain 2. As shown in Figure 4B, the production of cis-3-HyPip by strain 2, and strain 12(ΔsucA), strain 13(ΔaceA), and strain 14(ΔsucAΔaceA) was 3.1, 18, 4.4, and 20.1 mM. The strain 12(ΔsucA) showed the more obvious effect of increasing the yield of cis-3-HyPip. The deletion of aceA, the gene encoding the key enzyme for glyoxylate bypass, also caused a moderate increase in cis-3-HyPip yield, suggesting that aceA only played a minor role in 2-OG catabolism and only sucA mutation can completely block the conversion of 2-OG to succinate. These results indicate that using the redesigned TCA pathway to produce cis-3-HyPip in E. coli was effective. The TCA cycle can be engineered to adopt an alternative route without having significant impact on cell physiology (Lin et al., 2015).

As previously reported, increasing the expression of genes encoding citrate synthase (gltA) and isocitrate dehydrogenase (icd) can increase the carbon flux to 2-OG (Zhang et al., 2018; Sun et al., 2019). Here, we used pCDFDuet-1, which contains the genes for citrate synthase (gltA) and isocitrate dehydrogenase (icd), to strengthen the TCA cycle (Figure 5A). Strain 15 [E. coli BL21(DE3) △sucA △aceA pETduet-1-29-getf-RBS(T7)-splcd &pCDF-glta-icd] was constructed on the basis of strain 14, and its produced slightly more than strain 14 (Figure 5B), indicating that overexpression of gltA and icd strengthen TCA cycle activity, thereby increasing the yield of cis-3-HyPip.

When glucose is taken up by E. coli, it is first broken down into pyruvate via glycolysis. Pyruvate is further converted to acetyl-CoA, which then enters the TCA cycle (Figure 5A). To increase the flux of the TCA cycle, different concentrations of glucose were also optimized. Our results showed that different glucose concentrations had a significant impact on cis-3-HyPip production in strain 15. As the glucose concentration increased from 0.1 to 0.7%, the production of cis-3-HyPip increased in concert (Figure 5C). At glucose concentrations higher than 0.7% (such as 0.7–1% w/v), no significant increase of cis-3-HyPip production was observed (Figure 5C). Such effects of glucose on cis-3-HyPip production were not observed in strain 2, which the TCA cycle was not coupled with the GetF catalyzed reaction. These results further demonstrated that the flux of the TCA cycle can be coupled with the catalyzed reaction (Ruan et al., 2019; An et al., 2020), and that the efficiency of the whole-cell biocatalysis for production can be improved by pushing the TCA cycle (Lin et al., 2015; Sun et al., 2019).

Under the optimized conditions described above, a curve for whole-cell biocatalyst mediated conversion of L-lysine into cis-3-HyPip was obtained. As shown in Figure 6, at 0–20 h, the yield of cis-3-HyPip accumulated rapidly up to 21 mM. Subsequently, the rate of product generation increased slightly and slowly after 20 h. The product yield stabilized during the final phase (>40 h), which may be due to the decrease of thermal stability of GetF was the reason for the loss of catalytic activity. Finally, the reaction reached equilibrium when 33 mM cis-3-HyPip was produced after bioconversion for 50 h. Compared to that in previous studies (Table 1), the engineered strain used in this study displays the highest yield reported to date via whole-cell biotransformation, and without the addition of 2-OG.