Efficient production of salicylic acid through CmeR-P_cmeO biosensor-assisted multiplexing pathway optimization in Escherichia coli

Abstract

To address the challenge of microbial tolerance in industrial biomanufacturing, we developed an adaptive evolution strategy for Escherichia coli W3110 to enhance its salicylic acid (SA) tolerance. Utilizing a CmeR-P_cmeO biosensor-enabled high-throughput screening system, we isolated an SA-tolerant variant (W3110K-4) that exhibited a 2.3-fold increase in tolerance (from 0.9 to 2.1 g/L) and a 2.1-fold improvement in SA production (from 283 to 588.1 mg/L). Subsequently, the designed sensors were combined with multi-pathway sgRNA arrays to dynamically modulate the other three branched-chain acid derivatives, achieving a balance between biomass growth and rapid SA production in the adaptively evolved strain, resulting in a maximum SA yield of 1477.8 mg/L, which represents a 30% improvement over the non-evolved control strain W3110K-W2 (1138.2 mg/L) using the same metabolic strategy. Whole-genome sequencing revealed that adaptive mutations in genes such as ducA* and anti-drug resistance C2 mutation genes (ymdA*, ymdB*, clsC*, csgB*, csgA*, and csgC*) play a key role in enhancing SA tolerance and productivity. Notably, the evolved strain W3110K-4 exhibits significant resistance to bacteriophages, making it a promising candidate for large-scale SA fermentation. This work develops and expands the CmeR-P_cmeO system, proposes new insights into improved strains through biosensor screening, guided multi-pathway metabolism, and adaptive evolution, and provides a paradigm for engineers to obtain engineered strains.

Introduction

Aromatic compounds are extensively employed across various industries, including food, pharmaceuticals, chemical engineering, and animal feed. In recent decades, significant advancements in genetic engineering, metabolic engineering, and synthetic biology have facilitated a gradual shift from traditional petrochemical and plant extraction methods to the development of efficient, non-toxic microbial cell factories for the synthesis of aromatic derivatives [1,2,3]. However, elevated concentrations of aromatic compound products can adversely affect the essential functions of microorganisms [4], particularly in the case of SA, a dibasic acid recognized for its antimicrobial properties [5,6,7]. These characteristics can substantially impede the progress of microbial platforms designed for the production of diverse aromatic compounds.

SA serves as a crucial precursor for many valuable metabolites and is a primary component in various pharmaceuticals [8, 9], including cosmetics, aspirin, and lamivudine (an antiretroviral drug) [6]. As one of the branched acid derivatives, the biosynthesis of SA primarily involves two distinct pathways: (i) the phenylalanine ammonia lyase (PAL)-initiated pathway and (ii) the chorismate pathway, which includes the phenylpropanoid cascade [10]. The central phenylpropanoid pathway has been shown to incorporate ¹³C-labeled phenylalanine (¹³C₆-Phe) into 4-hydroxybenzoic acid (4-HBA) rather than SA in Arabidopsis thaliana [11]. However, utilizing plants for the synthesis of aromatic compounds necessitates the use of substantial amounts of sulfuric acid, resulting in the generation of large volumes of high-concentration sodium sulfate waste, which imposes significant burdens on subsequent processing and environmental protection.

Consequently, microbial production systems, particularly those utilizing E. coli, have garnered attention as a more efficient and sustainable approach. The introduction of exogenous genes facilitates the production of various aromatic compounds from branched-chain acids [12], including para-hydroxybenzoic acid (PHBA), 3-hydroxybenzoic acid (3-HBA), 2-aminobenzoic acid (2-ABA), para-aminobenzoic acid (PABA), l-tyrosine, phenol, and 3-methylacetophenone (MA) [8, 13]. Previous studies have focused on constructing E. coli cell factories for the biosynthesis of SA, drawing upon biosynthetic strategies involving isochorismate synthase (ICS) and isochorismate pyruvate lyase (IPL) from Pseudomonas species. These strategies, along with modifications to the shikimate pathway and the incorporation of feedback-resistant genes such as aroG^fbr, have achieved SA yields ranging from 222.3 mg/L to 1480 mg/L [14]. However, these high-yield strains were derived from specific E. coli mutants (ATCC31882) that exhibit nutritional deficiencies, which may limit their growth and industrial applicability. This underscores the need for more universally applicable methods for engineering E. coli strains to enhance SA production without compromising the microorganism’s overall growth capacity [15]. Previous studies have concentrated on specific E. coli mutants and modifications to precursor metabolic pathways, suggesting that future research on branched-chain acid derivatives may lack universality. Therefore, we have redirected our efforts toward more universally applicable approaches that facilitate the development of advanced methods for engineering E. coli to enhance SA production.

To address these challenges, we developed an adaptive evolution (ALE) strategy in this study to enhance the tolerance of E. coli to high concentrations of SA [16]. Additionally, we integrated a biosensor for the screening of high producers and implemented dynamic regulation of branched-chain acid derivatives using a multi-pathway sgRNA array. The engineered strain exhibited significantly improved SA production and phage resistance, positioning it as a promising candidate for industrial-scale SA production. Whole-genome sequencing revealed key genes responsible for the enhanced tolerance and production, providing valuable insights for further metabolic engineering. This work contributes to the development of more efficient and sustainable microbial platforms for the industrial-scale production of SA, advancing biotechnological applications in the pharmaceutical and chemical industries.

Materials and methods

Strains, growth conditions, and transformations

Plasmids were constructed using E. coli strain JM109 as the host, while E. coli strain W3110 was employed for fluorescence characterization and SA production. Gene integration and knockout were performed using the CRISPR/Cas9 system, with gel images of the integrated knockout strains W3110 provided in the supplementary information. Unless otherwise specified, E. coli cells were cultured in liquid or solid Luria–Bertani (LB) medium at 37 °C for approximately 2.5 h, and competent E. coli W3110 was prepared using the CaCl₂ method. Following a 90-s heat shock transformation, cells were allowed to recover in an antibiotic-free LB medium for 1.5 h before being plated on a solid LB medium. When necessary, appropriate concentrations of kanamycin, spectinomycin, ampicillin, and chloramphenicol were added to the medium for plasmid selection [17].

Experimental materials

The Luria–Bertani (LB) medium contained 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl, utilized for inoculation and plasmid propagation. The M9 medium consisted of 20 g/L glucose, 6 g/L Na₂HPO₄, 0.5 g/L NaCl, 3 g/L KH₂PO₄, 2 g/L NH₄Cl, 246.5 mg/L MgSO₄·7H2O, and 14.7 mg/L CaCl₂ [14, 18], which was employed for fluorescence characterization of the CmeR-P_cmeO gene circuit. The modified medium, M9A, was formulated by adding 20 g/L glycerol, 2 g/L yeast extract, and 2 g/L MOPS to the M9 base, serving for SA production and the growth expression of knockout strains during whole genome analysis. When necessary, kanamycin, spectinomycin, ampicillin, and chloramphenicol were added to the medium. IPTG and 1 M arabinose stock solutions were prepared in sterile water. Unless otherwise specified, the final concentrations of kanamycin, spectinomycin, ampicillin, and chloramphenicol were 50, 100, 200, and 25 mg/L, respectively, with IPTG induced at a concentration of 0.5 mM. All chemical reagents for the media were obtained from the China National Pharmaceutical Group Chemical Reagents Co., Ltd. Antibiotics, IPTG, and arabinose were purchased from Shanghai Macklin Biochemical Co., Ltd. In contrast, analytical-grade methanol and formic acid were obtained from Shanghai Titan Technology Co., Ltd. and Shanghai Aladdin Biochemical Technology Co., Ltd.

Synthetic genes, oligonucleotides, and plasmids

The pchB (pf-5) and cmeR genes were purchased from Nanjing Jinshui Biological Technology Co., Ltd. [19], where they underwent codon optimization and were delivered in plasmid form as lyophilized preparations. The egfp gene [20] (GenBank accession number U55762). Primers were sourced from Genewiz Biotechnology Co., Ltd. The sequences of cmeR and pchB, along with all primers utilized in this study, are summarized in the supplementary data provided in Table S1. To investigate the regulatory effects of varying quantities of cmeO(TGTAATAAATATTACA) binding sites on cmeR protein activity, we synthesized seven P2 promoters containing the CmeR binding sites O1, O2, O3, O1O2, O1O3, O2O3, and O1O2O3. Using reverse PCR, these promoters were inserted alongside the egfp reporter gene into the medium-copy pACYCDuet-T7 plasmid, resulting in the generation of plasmids cmeO*pACYCDuet-P2-egfp, O2*pACYCDuet-P2-egfp, O3*pACYCDuet-P2-egfp, O1O2*pACYCDuet-P2-egfp, O1O3*pACYCDuet-P2-egfp, O2O3*pACYCDuet-P2-egfp, and O1O2O3*pACYCDuet-P2-egfp.

To examine the regulatory effects of CmeR protein expression levels on P2 activity, we synthesized three distinct combinations of strong and weak promoters and ribosome binding sites (RBS), including J23101-B0030, J23101-B0034, J23101-T7RBS, J23106-B0030, J23106-B0034, J23114-T7RBS, J23114-B0030, J23114-B0034, and J23114-T7RBS. Nine promoter combinations were created based on the hierarchy (J23101 > J23106 > J23114: B0030 > B0034 and T7RBS) [18], and reverse PCR was utilized to insert these combinations alongside the cmeR gene into the high-copy pUC19-PBBa-T7 plasmid, generating plasmids such as pUC19-PBBa-J23101-B0030-cmeR, pUC19-PBBa-J23101-B0034-cmeR, and others.

The entC gene (originating from E. coli K-12 MG1655) and pchB were inserted into the pTrc99a plasmid via reverse PCR, resulting in the production of the plasmid pTrc99a-B0030-entC-pchB for preliminary SA expression experiments.

By utilizing sgRNA targeting different genes and positions, along with the pTarget plasmid [21], we constructed a dynamic sgRNA array to regulate the expression of the plasmids trpE*pTarget-P_trc-B0030-entC-pchB-J23119-sgRNA1-P_O1O3-DCas9, trpE*pTarget-P_trc-B0030-entC-pchB-J23119-sgRNA2-P_O1O3-DCas9, and others (Tables 1 and 2).

Table 1 Engineering strains used in this study

Table 2 Plasmids constructed in this study

Fluorescence assays

To characterize the response of the biosensor, E. coli cells carrying genetic circuit plasmids were cultured overnight in an LB medium at 37 °C. The following day, 3 ml of M9 medium (pH = 7) containing varying concentrations of SA, and appropriate amounts of ampicillin and chloramphenicol were added to each well of a 24-well plate. Subsequently, 30 µL of overnight cultured seed solution was introduced into each well, and the mixture was incubated at 37 °C for 2 h, after which the working concentration of 0.5 mM IPTG was added. The plate was subsequently incubated on a shaking incubator at 37 °C for 18 to 24 h. The following day, the fluorescence of eGFP in the culture was measured using a Tecan SPARK multimode plate reader (excitation: 485 ± 10 nm; emission: 520 ± 10 nm), alongside OD600 measurements [17, 22, 23].

FACS

Fluorescence expression of the CmeR^F99A-P_O1O2 biosensor was measured using a FACSAria III flow cytometer (BD, USA). Cells were diluted in PBS at a 1:80 ratio to achieve an appropriate cell density (~ 1 × 10⁶ cells/mL). The event rate was maintained at approximately 2000 events per second, and a total of 1 × 10⁵ events were recorded per sample for statistical reliability. GFP fluorescence was detected in the FITC channel (485 nm excitation, 520/30 nm emission). To exclude debris, a forward scatter (FSC) threshold was applied, set to exclude particles smaller than 0.5 μm in diameter. For fluorescence normalization, data were calibrated using the background fluorescence of control strains (without the biosensor), and the fluorescence intensity of experimental strains was reported as fold-change relative to this baseline. The cells were gated based on GFP fluorescence, and side scatter (SSC) parameters were used to exclude aggregates, ensuring only single cells were analyzed. Following the first round of sorting, the top 500 colonies were selected based on fluorescence intensity, re-cultured, and subjected to a second round of FACS, followed by further colony screening. The final 20 colonies were selected for their fluorescence expression and SA yield, with the strain W3110K-4 exhibiting the highest fluorescence intensity chosen for subsequent fermentation.

HPLC

To produce SA, all transformants were inoculated into 10 mL of LB medium supplemented with an appropriate concentration of spectinomycin and incubated overnight on a shaking incubator at 37 °C and 200 rpm. Subsequently, 0.3 mL of the overnight culture was transferred into a 30 mL M9A medium containing appropriate antibiotics in a 500 mL baffled flask [14]. The culture was shaken at 200 rpm at 37 °C for 72 h, with a pH between 6 and 7. The timing of IPTG addition was conducted as required and will be detailed in the experimental section of the manuscript. The OD600 values were measured using a UV spectrophotometer (Aoyi, UV-1200). Samples were taken at 42 to 60 h, centrifuged at 12,000 rpm for 10 min, and 1 mL of the supernatant was filtered using a 0.22 μm polyethersulfone (PES) syringe filter for high-performance liquid chromatography (HPLC) analysis [15].

Samples and standards were analyzed using an Agilent high-performance liquid chromatography (HPLC) system (1260, Infinity II), equipped with a UV–visible detector and a reverse-phase Diamonsil C18 column (Diamonsil 5 μm, 250 mm × 4.6 mm) for SA. Solvent A consisted of 100% methanol, while solvent B composed 0.1% formic acid in water. The column temperature was maintained at 30 °C, and the injection volume was set at 10 μL. The flow rate of the mobile phase was set at 1 mL/min, with a gradient concentration as follows: 15% to 75% solvent B maintained for 25 min, followed by 75% to 90% solvent B for 1 min, and finally 90% solvent B for 4 min. SA was quantified based on the UV absorbance peak area measured at 300 nm [15].

DNA resequencing by illumina HiSeq/Novaseq/MGI2000 and data analysis

Following 14 h of cultivation in the fermentation medium, bacterial pellets were collected for whole-genome resequencing analysis. The samples were dispatched to GeneWiz Biotechnology Co., Ltd. in Hangzhou for DNA extraction and sequencing analysis, generating a mutation list for each evolved isolate. The MG1655 genomic sequence, with GenBank accession number NZ_AKBV00000001.1, served as a reference [24] (see Additional file 1 S8 and Additional file 2).

Next-generation sequencing library preparations were constructed following the manufacturer's protocol. For each sample, 200 μg genomic DNA was randomly fragmented by Covaris to an average size of 300–350 bp. The fragments were treated with End Prep Enzyme Mix for end repairing, 5′ Phosphorylation, and 3′ adenylated to add adaptors to both ends. DNA Cleanup beads were then used to select the size of the adaptor-ligated DNA. Each sample was then amplified by PCR for eight cycles using P5 and P7 primers, with both primers carrying sequences that can anneal with flowcell to perform bridge PCR and P7 primer carrying a six-base index allowing for multiplexing. The PCR products were cleaned up and validated using an Agilent 2100 Bioanalyzer. The qualified libraries were sequenced pair-end PE150 on the Illumina. HiseqXten/Novaseq/MGI2000 System.

Using fastp (V0.23.0) the sequences of adaptors, polymerase chain reaction (PCR) primers, N base more than 14, and Q20 lower than 40% were removed. The pipeline of Sentieon (V202112.02) was used to map clean data to reference genome, remove duplication and call SNV/InDel. Annotation for SNV/InDel was performed by Annovar (V21 Apr 2018). The-breakdancer and CNVnator were used to analyze genomic structure variation.

Compliance with biosafety regulations

While this study does not directly address ethical considerations, adherence to biosafety regulations was strictly observed throughout the experimental work. The adaptive evolution of E. coli strains and genetic modifications followed the guidelines set forth by local and international biosafety standards. All genetically modified organisms (GMOs) used in this study were handled in accordance with the relevant safety protocols, ensuring that all experiments were conducted in designated containment facilities suitable for such work. Additionally, the engineered strains, including those for SA production, were subject to thorough risk assessments to mitigate any potential hazards associated with their use. These measures were taken to ensure compliance with the biosafety regulations governing the manipulation and containment of GMOs.

Statistical analysis

Unless otherwise stated, the dose–response data points represent the average of three replicates, displayed either as three individual data points or as a single aggregated data point, with error bars indicating the standard deviation. A two-sided t-test in Origin was used to perform the statistical evaluations (P values). *P < 0.05 and **P < 0.025. Graphs of results were plotted using the Origin software.

Results and discussion

Adaptive evolution was employed to develop superior E. coli strains exhibiting enhanced tolerance to elevated SA concentrations

Previous studies have typically employed a two-plasmid system to enhance the expression of branched-chain acid precursors. This method, which involves overexpressing key enzymes and SA-related genes using PCS-APTA and PZE-EP plasmids, respectively, is critical for the synthesis of SA. Following this approach, researchers often knockout alternative branched-chain acid conversion pathways and sequentially express entC alongside the exogenous gene pchB to facilitate the conversion of branched-chain acids into SA [25, 26]. However, the overexpression of enzymes may induce metabolic stress in cells, potentially leading to premature cell lysis [27]. To further enhance SA production, we integrated aroL, tktA, aroG^S180F, and ppsA into the pseudogenes yeep, ygay, yneU, and yjiV within the W3110 genome. Additionally, we incorporated entC and pchB—utilizing entC from E. coli as the source of isochorismate synthase (ICS) activity (see Additional file 1 S1) and employing the optimized pchB gene from Pseudomonas fluorescens Pf-5 as the source of isobutyrate lyase (IPL) activity (see Additional file 1 Table S1) [14, 28])—into the pykA gene locus to generate the W3110K-2 strain (Fig. 1a). This modification resulted in a maximum SA yield of approximately 283 mg/L. In comparison to the traditional method employing the phenylalanine-producing strain ATCC31882, we adopted a gene integration approach. Although this method may reduce SA yield, it enhances the production potential of microbial cell factories. Furthermore, we explored the effects of IPTG supplementation at 0, 2, and 5 h post-inoculation on both SA yield and cellular growth rate. The W3110K-2 strain exhibited minimal variation in SA production relative to the timing of IPTG addition [8].

Fig. 1

CmeR-P_cmeO biosensor-assisted multiplexing pathway optimization of tolerant strains for efficient synthesis of SA. a Metabolic pathways associated with salicylate biosynthesis in E. coli. E4P: erythrose-4-phosphate; PEP: phosphoenolpyruvate; DAHP: 3-deoxy-d-heptulosonate-7-phosphate; shikimate: shikimic acid; chorismate: chorismic acid; Phe: l-phenylalanine; Trp: l-tryptophan; Tyr: l-tyrosine; isochorismate: isochorismate; SA: salicylic acid; aroL: shikimate kinase; ppsA: phosphoenolpyruvate synthase; tktA: transketolase I; aroG: 3-deoxy-d-arabino-heptulosonate-7-phosphate synthase; entC: branched-chain acid synthase; pchB: isochorismate pyruvate lyase; pykA: pyruvate kinase. b, f Overall workflow of the Adaptive Laboratory Evolution (ALE) process. c, d Design and construction of the CmeR system. e The CmeR^F99A-sgRNA device was designed and constructed to facilitate multiplexed module assembly. f Fluorescence-activated cell sorting (FACS) was performed, followed by characterization of the screened strains, which were subsequently applied for the production of SA

Considering the detrimental effects of SA on cell viability and DNA replication in E. coli [15], it adversely affects the bacteria, leading to a cessation of SA production. This concept has been proposed in previous studies; however, detailed investigations have yet to be conducted. We hypothesize that an E. coli strain capable of tolerating high concentrations of SA would be more advantageous for its synthesis. This study focuses on the adaptive evolution of E. coli W3110K-2 for SA production. While most studies utilize an initial concentration of 1 g/L SA [28] to inhibit the growth of E. coli, our preliminary toxicity assessment revealed that E. coli W3110K-2 exhibited significant growth inhibition at a concentration of 6.5 mM (0.9 g/L) SA (Fig. 6a). Given the adaptability concerns of E. coli, we selected a starting concentration of 3.5 mM SA for the tolerance assessment.

Previous studies have employed ALE to improve the tolerance of E. coli to octanoic acid (OCTA) [29] and butyric acid (BUT) [30] across 11 industrial chemicals, establishing a foundation for direct comparisons. Prior microbial engineering modifications have been applied to all compounds to enhance yield, highlighting their importance for bioproduction [31, 32]. Furthermore, there are two primary methods commonly employed in laboratory-based bacterial evolution: ultraviolet (UV) radiation-induced mutation and adaptive laboratory evolution (ALE). Although ALE requires more time and resources compared to UV-induced mutation, it is undoubtedly more effective in enhancing bacterial tolerance to substances such as SA and in providing more detailed insights into bacterial evolution. Given the critical role of ALE in advancing biological production, we have developed a novel method for adaptive evolution aimed at enhancing strains. Additionally, the pykF gene has been frequently observed to undergo mutations during the adaptive evolution process [33], and it will not be subjected to knockout or integration in future studies.

The study aimed to select E. coli strains that exhibit enhanced survival capabilities in environments with high concentrations of SA by progressively increasing the SA concentration. In a previous investigation, we found that the antimicrobial action of SA against E. coli is primarily mediated through two mechanisms: first, salicylate ions negatively affect cellular viability and DNA replication, which significantly contribute to its antibacterial effects [33]; second, the pH concentration decreases during fermentation, whereby lower pH levels inhibit bacterial growth, and elevated concentrations of H⁺ ions interact with salicylate ions to form higher concentrations of SA, thereby exacerbating the damage to bacterial cells. Consequently, to maximize the SA tolerance of E. coli, we conducted approximately 200 passages in LB medium containing SA at pH 6, a process that spanned roughly one and a half years (Fig. 1b). This effort resulted in the isolation of E. coli W3110K-3, which is capable of growing in LB medium with 2.1 g/L SA at pH 6.

Biosensor-assisted adaptive evolution strains were strategically redesigned to facilitate the screening and identification of highly efficient producer strains

Identifying a truly high-yielding strain for SA production from thousands of adaptive strains presents a significant challenge. To facilitate the rapid analysis of intracellular SA levels, we reconstructed the Campylobacter jejuni ter family repressor CmeR and developed a SA-responsive biosensor system, CmeR-P_cmeO [17, 34] (Fig. 1c), aimed at high-throughput screening and metabolic regulation of adaptive strains. To better address the variable relationship between transcription factor (TF) protein expression levels and binding proteins, we employed a dual-plasmid system to reconstruct the sensor system in the wild-type E. coli W3110 strain (Fig. 2a). This gene cassette comprises the SA-responsive repressor protein CmeR, the high-copy plasmid PUC19-PBBa-J23101-B0030-cmeR, a promoter P2 regulated by CmeR, and the green fluorescent protein (eGFP), resulting in the generation of the medium-copy plasmid pACYCDuet-P_cmeO-egfp [23]. The fluorescence intensity was normalized to OD600 to assess whether the dual-plasmid system affected the response expression of CmeR-P_cmeO. We measured the fluorescence intensity of the strain containing only pACYCDuet-P_cmeO-egfp, yielding a normalized intensity of 342,493 ± 177.15 au. Following the introduction of pUC19-PBBa-J23101-B0034-cmeR into the wild-type E. coli W3110 strain, the intensity decreased to 33,033 ± 564 au (Fig. 2c). Subsequently, we introduced a gradient concentration of SA (0 mM, 0.2 mM, 0.4 mM, 0.6 mM, 0.8 mM, 1.0 mM, 1.2 mM, 1.4 mM, 1.6 mM, 1.8 mM, 2 mM, 5 mM) into the culture medium to induce the biosensor system. The CmeR-P_cmeO biosensor system constructed in E. coli strains was capable of responding to 1.2 mM SA, with fluorescence intensity ranging from 33,033 ± 564.22 auto 118,636 ± 629.37 au (Fig. 2c). This indicates that the dual-plasmid system can effectively adapt to the CmeR-P_cmeO sensor system, which responds well to low concentrations of SA, consistent with findings from previous research. However, to achieve a better sensor system, further enhancements are necessary.

Fig. 2

Initial establishment and expansion of the CmeR-P_cmeO biosensor system. a Schematic representation of the re-established CmeR-P_cmeO biosensor system. The CmeR protein binds to P_cmeO to inhibit promoter expression, while salicylic acid (SA) interacts with the CmeR protein to alleviate this repression. b A systematic combinatorial analysis and optimization of cmeO for CmeR expression were conducted. c Dynamic performance of the wild-type CmeR-P_cmeO biosensor system in E. coli W3110 was assessed. d, e Optimization of the CmeR-P_cmeO biosensor system through cmeO binding sites was performed, and the data were obtained after 18 h of cultivation in 24-well plates

To further optimize the detection range of the CmeR-Pc_meO sensor system, we hypothesized that overexpressing the CmeR protein and enhancing its binding affinity to the binding sites may contribute to the limited detection range [35]. To validate this hypothesis, we modified the positions and quantities of the cmeO binding sites within the P_cmeO promoter to assess their effects on the response level to CmeR. The P2 promoter is a synthetic construct with the cmeO binding site located downstream of the -35 region (O3). We replaced the regions preceding -10 (O1) and between -10 and -35 (O2) with cmeO binding sites to evaluate the impact of binding site positioning on repression levels. Our findings indicated that both the O1 and O2 regions facilitated a broader detection range [23], demonstrating that variations in binding site positioning can significantly influence the binding affinity of CmeR. We then investigated whether the quantity of cmeO binding sites at the O1 position would affect the dynamic range. Based on the O1 configuration, we substituted the O2, O3, and O2O3 regions within the P_O1 promoter (Fig. 2b). This led to the identification of the optimal promoter, P_O1O3, which increased the fluorescence response range from 1.2 mM to 7 mM (Fig. 2c). While the O1O2 locus demonstrates elevated fluorescence expression, promoters containing the O1O3 locus exhibit greater fluorescence potential and encompass a broader dynamic response range.

To mitigate the excessive repression of the P_O1O3 promoter by CmeR protein levels, we utilized various ribosome binding sites (RBS) in conjunction with promoters of differing strengths (promoter strength ranking: J23101 > J23106 > J23114; RBS strength ranking: B0030 > B0034 > T7RBS) [36] (Fig. 3a). In the presence of SA, strains containing the plasmids pUC19-PBBa-J23114-B0030-cmeR and pACYCDuet-P_O1O3-egfp exhibited the highest unit fluorescence response range, with minimal fluorescence leakage. This response range peaked at 10 mM (Fig. 3b).

Fig. 3

Further optimization of the CmeR-P_cmeO biosensor system. a, b A systematic combinatorial analysis and optimization of the promoter and ribosome binding site (RBS) for CmeR expression were conducted using constitutive promoters with established relative strengths (http://parts.igem.org/Promoters/Catalog/Ecoli/Constitutive). c Discovery Studio was utilized to simulate the CmeR-SA complex, followed by validation and analysis using molecular docking software AutoDock Vina and ChimeraX 1.8, revealing the interactions between CmeR residues and SA. d Fluorescence signals of various mutants of CmeR were analyzed

To enhance the sensitivity and fluorescent response range of the CmeR-P_cmeO biosensor system, we designed the transcriptional regulator CmeR to modulate the system's dynamic performance. We hypothesize that the excessive sensitivity of CmeR to SA following induction leads to the premature release of CmeR [36]. To validate this hypothesis, we initially aimed to attenuate the interaction between the CmeR protein and SA. To achieve this, we employed molecular docking simulations. This study utilized Discovery Studio to simulate the CmeR-SA complex, followed by validation and analysis using molecular docking software AutoDock Vina and ChimeraX 1.8, which revealed the interactions between CmeR residues and SA (Fig. 3c, see Additional file 1 S3). Our simulations indicated that F99, S138, Y139, and C166 are the principal amino acid residues involved in the binding of CmeR to SA. Specifically, the interaction between F99 and SA predominantly manifests as a weak pi-pi T-shaped attractive force, while S138 and Y139 engage in conventional hydrogen bonding with SA, and C166 exhibits pi–alkyl interactions. This suggests that F99 and C166 play significant roles in the binding between CmeR and SA. To weaken the binding between CmeR and the promoter, we replaced F99 in the CmeR regulator with a nonpolar amino acid, such as alanine, or a negatively charged polar amino acid, such as aspartic acid or glutamic acid. Interestingly, substituting F99 with alanine resulted in a maximum eGFP expression induction that exceeded 12 times the original level in the presence of SA (Fig. 4a, b). This substitution also led to a 1.5-fold increase in the monomeric fluorescence response range, reaching 15 mM, while the leakage activity of P_cmeO was reduced to 50% of its original level (Fig. 3d). Furthermore, mutating F99 and C166 to polar amino acids revealed that the substitution of F99 with glutamic acid eliminated the detection capability of CmeR-P_cmeO for SA, whereas the sensitivity of C166 to SA remained unchanged when mutated to a polar amino acid. We speculate that F99 serves as the primary amino acid residue for the binding of CmeR to SA. In our experiments, we also tested shikimic acid, tyrosine, pyruvate, phenylalanine, branched-chain acids, and reported inducers such as tryptophan and glycerol, discovering that CmeR no longer responded to these molecules [17] (see Additional file 1 S4). In this experiment, we modified the relationship between the expression levels of the cmeR protein and its binding site. We employed a directed mutagenesis strategy to enhance the binding affinity of the cmeR protein for the target substance, thereby effectively optimizing the dynamic range of the biosensor. This led to the development of a new SA biosensor capable of linearly responding to the SA content in solution, which may serve as a valuable reference for future functional adjustments of biosensors. Subsequently, we transferred the engineered CmeR^F99A-P_O1O2 biosensor into an adaptive strain.

Fig. 4

Structural prediction and analysis of the regulator CmeR and the mutant CmeR^F99A. a The CmeR protein is depicted in cyan. b The CmeR^F99A protein is represented in blue. Homology modeling of CmeRF99A was performed using the CmeR protein template in Discovery Studio. Upon substitution of phenylalanine with alanine at position 99 (F99A), structural analysis revealed a distinct protrusion formation within the ligand-binding pocket. This conformational alteration significantly modified the interaction pattern between the mutant protein and SA, particularly in terms of binding orientation and potential contact residues

Screening promising SA producers from the adaptively evolved population assisted by the CmeR-P_cmeO biosensor

Utilizing the biosensor CmeR^F99A-P_O1O2, we isolated the strain with the highest yield of SA from the adaptively evolved population. The designed biosensor was transformed into the adaptively evolved strains. Following two rounds of fluorescence-activated cell sorting [37, 38] (FACS; see Methods; Fig. 5c) on the transformed population, 500 colonies were screened from a library of 10⁵ strains. These 500 colonies were subsequently co-cultured, followed by a second fluorescence selection of 20 colonies. Among the final 20 colonies, 9 exceeded a yield of 400 mg/L, representing an increase of more than 1.4 times compared to the average SA titer (245.6 mg/L) of the non-adaptively evolved population (Fig. 5c). These results indicate that the CmeR^F99A-P_O1O2 biosensor can effectively isolate high SA producers from a mixed strain population. From these 20 colonies, the strain W3110K-4, which exhibited the highest fluorescence expression of SA, was selected for fermentation. Notably, compared to the W3110K-2 strain, the W3110K-4 strain demonstrated a significantly enhanced growth rate under identical fermentation conditions, indicating that the ALE strategy effectively improved tolerance to SA (see Additional file 1 S5). This biosensor, combined with the adaptive evolution method, can facilitate the rapid identification of superior SA producers.

Fig. 5

Design and construction based on CmeR-P_cmeO system. a Schematic representation illustrating the application of CRISPRi-conjugated CmeR-P_cmeO. Three competitive modules involved in SA synthesis, namely trpE, pheA, and tyrA, were selected as regulatory targets. b, c The top 0.02% to 0.05% of cells exhibiting the highest fluorescence underwent two rounds of sorting following the installation of the CmeR^F99A-PcmeO device into the W3110K-4 strain. The final 20 colonies were selected for their fluorescence expression and SA yield, with the strain W3110K-4 exhibiting the highest fluorescence intensity chosen for subsequent fermentation. d Schematic representation of reporters employed to assess the performance of sgRNAs targeting trpE, pheA, and tyrA. The fusion protein consisting of eGFP and inactive variants of trpE*, pheA*, or tyrA* in the low-copy plasmid pTarget served as the reporter. Four sgRNAs were designed for each of the aforementioned genes and were placed under the control of the P_J23119 promoter. The specific sgRNA ligation base sequences are detailed in Table S1f (Annex 1). e Effects of various strategies on the SA titers of the W3110 strain. A two-sided t-test in Origin was used to perform the statistical evaluations (P values). *P < 0.05 and **P < 0.025

The adaptively evolved strain showed a better potential for SA production

To determine whether strains have evolved to tolerate high concentrations of the product and to assess their ability to produce SA, we initially replaced the entC and pchB genes within the pykA locus with CmeR^F99A. This modification also addresses the insufficient expression of SA biosynthetic enzymes while integrating sensor regulators into the genome. Subsequently, we constructed the pTarget-P_trc-B0030-entC-pchB-J23119-sgRNA4-P_O1O3-DCas9 plasmid [39, 40]. By employing a biosensor in conjunction with CRISPR interference (CRISPRi), we integrated the regulatory protein CmeR into the genome. The expression of Dcas9 is regulated by the promoter P_O1O3, and Dcas9 cannot be expressed in cells containing high levels of CmeR protein; thus, SA production commences only when the cell reaches a specific developmental stage. Once SA accumulates to a certain threshold, the CmeR protein responds to it and subsequently activates Dcas9, which then binds to various gene loci guided by distinct sgRNAs. Simultaneously, we modulated the metabolic flux of three branched-chain acid derivatives to precisely regulate SA production, facilitating further evaluation of the tolerant strains (Fig. 1e). Previous studies primarily achieved elevated SA production by disrupting the pheA and tyrA genes and by controlling the metabolic coupling between the phosphoenolpyruvate (PEP) and tricarboxylic acid (TCA) cycles. Tryptophan (Trp), a derivative of branched-chain acids, is a crucial metabolite for bacterial growth and development [21, 39, 41]. In preliminary trials, we successfully validated a dual plasmid-based expression system for SA in E. coli, designating the engineered strain as W3110K-1. This was accomplished using the plasmids pTrc-P_trc-entC-pchB and pACYCD-P_trc-aroG^s180F-ppsA-tktA-aroL, resulting in a maximum SA yield of 801.1 mg/L, which represents a significant improvement over the native W3110 strain (Fig. 5e). We observed that under enhanced branched-chain acid pathways, bacterial growth did not necessitate high levels of tryptophan, and excessive tryptophan could impose a burden on the cells. However, directly disrupting these competing pathways could hinder cell growth. Consequently, we attempted dynamic regulation to downregulate trpE, pheA, and tyrA to achieve a balance between production and growth [15, 28] (Fig. 5a). Concurrently, tryptophan, phenylalanine, and tyrosine serve as primary inhibitory products of aroG [27, 42]; their downregulation alleviates feedback inhibition on the shikimic acid pathway, potentially enabling the attainment or surpassing of plasmid expression levels. In this study, we designed four sgRNAs targeting different loci within the trpE, pheA, and tyrA gene operons. These targets include the promoter region (sg1), the ribosome binding site (sg2), approximately 200 bp of the coding sequence (sg3), and the terminal 200 bp of the target gene coding sequence (sg4) [20, 43] (Fig. 5d). The four distinct downregulation arrays can inhibit the expression of three target genes at varying levels (sg4 = sg2 > sg3 > sg1). Generally, while complete knockout or significant downregulation (sg4 and sg2) of trpE, pheA, and tyrA may lead to high yields of SA, the nutrient deficiencies resulting from these alterations can impede cell growth. Additionally, supplementing the medium increases production costs; therefore, these conditions are utilized solely as controls in the experiment. We linked the gene targeted for suppression to the eGFP protein using a helical linker (EAAAKEAAAK), allowing for clear visualization of protein expression flux via fluorescence. Subsequently, we combined this with the base sequence of the sgRNA that can be cleaved concurrently. The first gene is regulated initially, followed by the regulation of the second gene based on this framework, thereby enabling the simultaneous regulation of multiple metabolic fluxes. Fluorescence quantification analysis revealed that sg3 and sg1 inhibited trpE, pheA, and tyrA expression by 43.3%, 27.7%, 47.1%, 22.1%, and 42.5%, 26.4%, respectively (see attached file 1 S6b). Among the strains containing dynamic modulation modules, the combination of trpEsg3, pheAsg1, and tyrAsg3 was identified as the optimal configuration for SA production. We conducted an additional round of testing to evaluate the effects of IPTG addition at 0 h, 2 h, and 5 h post-inoculation on SA production and cell growth rates. Notably, the W3110K-5 strain exhibited the highest yield when IPTG was added at 0 h. The highest SA-producing strain reached its peak yield of 1477.8 mg/L after 42 h of fermentation (Fig. 5e). By reducing plasmid expression, we achieved the highest yield in shake flasks and shortened the fermentation cycle from 48 to 60 h to approximately 42 h, marking a significant breakthrough [15, 20]. As a control, the SA biosynthetic pathway was placed within a host in which the trpE, pheA, and tyrA genes were disrupted, resulting in an SA production rate of 945.8 mg/L, which constitutes only 60% of the yield from the W3110K-5 strain. These results indicate that reduced viability at high chemical concentrations is a critical limiting factor. By optimizing the timing of repression of complex metabolic and competitive pathways in tolerant cells, the production performance of the endogenous ALE strains can be enhanced through the use of tolerant ALE strains [32, 40].

Based on the existing central metabolic network competition pathway, the knockout of high-yield strains and the guidance of metabolic flux toward the desired synthetic products has been demonstrated as an effective strategy, aligning with the direction of previous research. In contrast to traditional methods, this study integrates the CmeR system of SA reaction with multi-array sgRNA regulatory components, enabling the engineered strain to produce a significant amount of SA. This approach redirects the central metabolic network from supporting cell growth to promoting SA production. Notably, this metabolic pathway transition does not occur at low production levels, thus circumventing the trade-off between cellular metabolic growth and product synthesis. Additionally, this engineered strain does not require extra media or fermentation time compared to the nutritional-defective strain, and the multi-array sgRNA can be replaced, allowing for metabolic flux to be directed toward the production of other aromatic derivatives. This provides potential regulatory methods for the production of various aromatic derivatives, thereby reducing industrial costs and enhancing suitability for industrial production. To further investigate the potential mechanisms of W3110K-5, whole-genome sequencing analyses of the W3110K-5 and W3110K-2 strains were conducted, employing a systematic workflow for further characterization and engineering.

Analysis of whole-genome resequencing in microorganisms

We conducted whole genome resequencing at the molecular level to elucidate the molecular mechanisms underlying SA synthesis promoted by the evolved strain W3110K-4. Each mutation was mapped to its corresponding gene; intragenic mutations were assigned to overlapping coding sequences, while intergenic mutations were attributed to the nearest gene coding region [24].

Comparative mutation analysis revealed key genetic regions associated with enhanced SA tolerance and yield, thereby elucidating the relationship between mutation sites and gene function. Preliminary analysis identified a total of 16 mutations in exonic genes between the W3110K-4 and W3110K-2 strains, which included three synonymous single nucleotide variants (SNVs), six nonsynonymous SNVs, and seven insertion/deletion (InDel) mutations, collectively impacting 23 genes. To identify the critical genes, we concentrated on the six nonsynonymous SNVs and the seven InDel mutations (adhP*, SNVs and InDel), while excluding synonymous mutation types (Fig. 6).

Fig. 6

Mutation analysis for tolerized isolates. a Growth states of W3110K-2 at different concentrations of SA. b The count of observed Indell and SNVs mutations for all mutated genetic features of this experiment. Numbers on the x-axis, y-axis coordinates indicate the number of times a single gene has been in different types of mutations, numbers on the z-axis coordinates different mutated genes (corresponding to (d), 18 = ducA). c Effect of W3110K-2 and W3110K-4 mutant genes on cell function. d A frequency heatmap of W3110K-4 mutated genes. The gene types are systematically organized based on their associated chemical groups (illustrated by the colors in the bottom row), with the color bar in the lower left corner providing a visual representation of the number of genes exhibiting mutations in the corresponding regions for each type. The bar chart on the right illustrates the growth conditions observed 12 h post-plasmid expression in the ALE experiments, specifically corresponding to the gene mutations for each type following the knockout of the respective genes. The color column located on the far left delineates the functional categories of the genes. The ybfC mutation cannot be categorized as a general tolerance, and there is currently no functional annotation available for it on ybfC uncharacterized protein YbfC [Escherichia coli str. K-12 substr. MG1655]—Gene—NCBI (nih.gov)

Analysis of SNVs and InDels

In previous fermentation experiments with the W3110K-5 strain, we observed that SA production reached its maximum yield between 48 and 60 h. Consequently, we focused our mutation analysis on growth metabolism and tolerance [44]. The primary nonsynonymous mutated genes include the unannotated protein gene ybfC, the global transcription regulator of carbohydrate metabolism mlC, the para-hydroxybenzoate efflux pump subunit aaeB, the two-component system sensor histidine kinase pmrB, the PTS trehalose transport protein subunit IIBC, the aspartate ammonia lyase aspA, the family protein fxsA, and the l-methionine/branched-chain amino acid transporter (collectively referred to as Group C1). Notably, the global transcription regulator mlC* coordinates the transcription of ptsG and serves as a regulatory factor for PEP biosynthesis and key developmental processes, which may be crucial for efficient SA synthesis. The para-hydroxybenzoate efflux pump subunit aaeB* functions as an aromatic amino acid efflux pump, while the two-component system sensor histidine kinase pmrB* is associated with ferritin transport and acid resistance genes, which may elucidate the ability of W3110K-3 to grow and develop under high concentrations of SA. Based on these hypotheses, we employed reverse engineering to introduce the mutations into the W3110K-2 and W3110C-1 (ΔentC-pchB::CmeR^F99A*W3110K-2) host strains, validating the relationships of several mutated genes with W3110K-4 (Fig. 7a). We conducted knockouts of these 12 genes in the W3110K-4 strain to observe their varying impacts on the host. Notably, the strain knocking out the Group C1 gene exhibited significantly slower growth and an unstable growth state in 1 g/L of SA compared to the W3110K-2 strain.

Fig. 7

Reverse engineering strategy for the 12 mutated genes. a Growth conditions of W3110 in the presence of 1 g/L salicylic acid (SA) over 24 h following the knockout of each of the 12 genes in W3110K-4. b Construction of plasmids for the 12 mutated genes using the medium-copy plasmid pACYCDuet, followed by their introduction into W3110K-4 to observe growth conditions and salicylic acid (SA) production yields. c Phage resistance profiles of W3110, W3110K-2, and W3110K-4. For the phage resistance assays, exponentially growing cultures (OD600 ≈ 0.6) were subjected to a tenfold dilution with PBS. A 2 μL aliquot of the diluted suspension was then pre-spotted onto LB agar plates. After allowing the surface to dry slightly, high-concentration phage suspensions were repeatedly spot-inoculated onto the E. coli area using sterile toothpicks to facilitate efficient phage–host interaction

To further investigate whether the insertion of genes affects SA biosynthesis and growth development, we conducted tests on the InDel mutated genes identified in the sequenced isolates. The InDel mutated genes encompass four primary categories: the ompT* family of efflux proteins, the biofilm-associated protein ymdA*, the O-acetyladenosine deacetylase ymdB*, the cardiolipin synthase clsC*, the curli small subunit csgB*, the curli large subunit csgA*, and the curli assembly chaperone csgC* (collectively referred to as Group C2), as well as the alcohol dehydrogenase adhP* and the C4-dicarboxylic acid transport protein dcuA*. Notably, the knockout of the dcuA* gene resulted in significantly slower growth during the early stages compared to the W3110K-2 strain. We also aimed to assess the impact of these mutations on the growth and development of strain W3110C-2. Subsequently, we integrated the C2 [45, 46] and dcuA* genes into the W3110K-2 strain through reverse engineering, yielding the W3110C-2 strain, which exhibited a growth rate in SA production comparable to that of the best-performing strains (Fig. 7a, b). During fermentation, the SA yield increased by approximately 1.3-fold compared to strain W3110K-1, with the W3110C-2 strain harboring the sgRNA array producing 1382.5 mg/L (Fig. 5e). Interestingly, these mutations do not have a direct association with SA production and cannot be explained by conventional rational analysis. Some of these genes enhance the strain's utilization of fermented substrates, thereby accelerating cell growth and ultimately increasing biomass concentrations, while others improve the strain's tolerance to acid and SA (see Additional file 1 S2). For instance, the knockdown of the ducA* gene results in a significantly low final biomass concentration of W3110K-4, and the knockout of the C2 group gene in W3110K-4 diminishes the strain's tolerance to acid and SA, which in turn reduces the penetration tolerance of E. coli and adversely affects SA yield. Therefore, it is intriguing to employ genome-wide analysis to investigate the generation mechanism of W3110K-5, which warrants further research. The W3110C-2 strain, which contains an sgRNA array, achieves a maximum yield of 1382.5 mg/L in approximately 42 h. In comparison, the strain W3110K-W2, which was developed by introducing the sgRNA array into W3110K-2, reaches a maximum yield of 1138.2 mg/L after 60 h. The W3110C-2 strain exhibits superior performance regarding biomass growth rate, growth cycle, and the ability to resume fermentation-coupled growth after encountering inhibitory levels of SA production. Conversely, the engineered adaptive evolved strain W3110K-5 surpasses the W3110C-2 strain, achieving SA production of 1477.8 mg/L in approximately 42 h (see Additional file 1 S9).

We conducted preliminary bacteriophage resistance profiling on the W3110K-4 strain. The strain was inoculated into 10 mL of LB medium and incubated at 37 °C with shaking at 200 rpm until it reached the mid-exponential phase (OD600 ≈ 0.6). Subsequently, a 1 μL high-titer phage was introduced. Bacterial growth was monitored every 6 h for the first 24 h, followed by monitoring at 12-h intervals from 24 to 48 h. For the phage resistance assays, exponentially growing cultures (OD600 ≈ 0.6) were subjected to a tenfold dilution with PBS. A 2 μL aliquot of the diluted suspension was then pre-spotted onto LB agar plates. After allowing the surface to dry slightly, high-concentration phage suspensions were repeatedly spot-inoculated onto the E. coli area using sterile toothpicks to facilitate efficient phage–host interaction. The plates were incubated at 37 °C for 48 h, with continuous monitoring of area integrity (Fig. 7c). Notably, the W3110-4 strain, which harbors mutations in ompT*, aspA*, fxsA*, and yjeH*, along with engineered strains W3110BC-3 and W3110C-3 that overexpress these genes, exhibited significantly enhanced phage resistance phenotypes. These genetic modifications (ompT* [45, 46], aspA* [47, 48], fxsA*, and yjeH* are proposed to mechanistically underlie the observed bacteriophage resistance.

ompT encodes the E. coli outer membrane protease, which cleaves T7 RNA polymerase. The fxsA gene is monocistronic and is not essential for growth. However, overexpression of fxsA is required for toxicity against T7 genes, potentially protecting against premature death caused by the expression of certain toxic genes from T7 phages [49]. The presence of these mutations indicates that our strains develop phage resistance through SA adaptive evolution, providing a valuable reference for large-scale industrial applications. Nevertheless, phage resistance remains an area that necessitates further research. While we speculate that the protein-modifying properties of SA are effective, the specific molecular mechanism underlying this phage resistance appears to be a consequence of SA's adaptive evolution. It remains unclear whether ALE from other aromatic derivatives will induce similar changes. Future detailed studies focusing on the phenotypic changes and understanding of these mutations will be beneficial. Furthermore, although our strains exhibit phage resistance, subsequent studies should investigate whether this resistance can be sustained during large-scale fermentation passages, which is crucial for both large-scale industrial production and advancements in biotechnology.

Conclusion

This study successfully constructed an efficient SA-producing strain by integrating adaptive evolution, multi-pathway regulation of the central metabolic network, and biosensor-assisted screening. A systematic reprogramming of the SA biosynthetic pathway was conducted in E. coli, involving fine-tuning of the central metabolic network, dynamic control of SA competitive pathways, and balancing precursor expression with product production. Subsequently, a biosensor-assisted high-throughput screening platform was designed and applied to explore further the potential for SA production and beneficial mutant genes. Ultimately, the optimally engineered strain W3110K-5 achieved a SA yield of 1477.8 mg/L within 42 h in shake-flask cultures, representing the most efficient, versatile, and suitable engineered strain for large-scale industrial production of SA.