Redesigning regulatory components of quorum-sensing system for diverse metabolic control

Investigating the regulation mechanism of luxR-luxI intergenic sequence

Autonomous and temporal control of multiple metabolic fluxes requires developing a series of QS promoters that can trigger the expression of each set of genes at a different time in a single cell. Such QS circuits could be created by in situ engineering of the natural QS system, which, however, are challenging due to the unknown regulation mechanism. In the natural QS system, the luxR-luxI intergenic sequence is responsible for binding with the LuxR-AHL complex and RNAP to initiate the transcription of promoter P_luxI (Fig. 1a)²¹. We hypothesized that mutation of this region is able to create a variety of QS variants that have a different binding affinity towards the LuxR-AHL complex and RNAP. When co-expressing these circuits in one cell, they will exhibit different QS behaviors. Random mutation of the luxR-luxI intergenic sequence has been proved to be ineffective in generating well-performed QS circuits²⁰. Rational engineering of this sequence requires a deep understanding of the functions of genetic components, which has yet been clearly elucidated so far.

It has been reported that there is a CRP-binding site located in the luxR-luxI intergenic sequence (Fig. 1a)²². The CRP-binding site is an important transcriptional regulation sequence (Consensus sequence: AAATGTGATCTAGATCACATTT) widely distributed in many bacteria^23,24. The CRP is an allosteric protein that usually binds to its signaling molecule cyclic AMP to form dimers or tetramers. This CRP-cAMP complex can bind to a recognition site in the target core promoters to assist RNA polymerase binding to the promoter²⁵. However, the exact function of the CRP-binding site in the lux-type QS system has not yet been proposed so far. Considering that the location of the CRP-binding site is very close to P_luxR, we hypothesized that P_luxR is a CRP-dependent promoter. The increased numbers of the CRP-binding site could recruit more CRP-cAMP complex to enhance the transcription of P_luxR by strengthening the interactions with RNAP. To test this hypothesis, we arranged in tandem multiple copies of the CRP-binding site (n = 0, 1, 2, 3, 4) in the luxR-luxI intergenic sequence (Supplementary Fig. 1a). The reporter gene mCherry was placed downstream of the promoter P_luxR. As we expected, with the increase of the CRP-binding site numbers, the transcriptional level of the gene under the control of the P_luxR was dramatically enhanced (Fig. 1b and Supplementary Fig. 2). The strain-carrying QS circuits with four copies of the CRP-binding site produced eightfold higher peak fluorescence than the strain harboring QS circuits without a CRP-binding site, indicating that the presence of more CRP-binding sites is beneficial for P_luxR transcription. In order to confirm how many CRP-binding sites are best, we increased the CRP-binding sites to 7. As shown in Supplementary Fig. 3, when the CRP-binding site varies from 0 to 7, the QS promoter harboring four CRP-binding sites exhibited the highest transcription level and dynamic range, indicating that four CRP-binding sites is the best. We further performed electrophoretic mobility shift assay (EMSA) experiments to confirm the result. As shown in Supplementary Fig. 4a, b, as the number of CRP-binding sites increases, the number of CRP proteins bound to the DNA increases too, indicating that more CRP-binding sites promoted the CRP protein binding to the DNA. We also used an atomic force microscope to observe the topographies of the CRP-DNA complex. As shown in Supplementary Fig. 5, when incorporating four CRP-binding sites in the locus, the DNA and CRP protein forms obvious multimers compared with that of one CRP-binding site in the locus. We believe that the formed multimers might promote the bounded CRP to recruit more RNAP for transcription. Since P_luxR is a type of CRP-regulated promoter, it is necessary to further explore which class it belongs to. The amino acid H159 in the CRP protein was substituted for Leu and the amino acid V287 in the alpha subunit of RNAP was substituted for Ala (Supplementary Fig. 6a, b). The results showed that both substitutions dramatically reduced the transcription of P_luxR (Supplementary Fig. 6c), indicating that P_luxR appears to be a class I CRP-dependent promoter.

Then, we began to investigate how the CRP-binding sites would control the transcription of P_luxI. The luxR and luxI genes were expressed under the control of a constitutive promoter to make sure the number of LuxR or LuxI is the same in all strains. The reporter gene eGFP was placed downstream of the promoter P_luxI. As shown in Fig. 1c, complete deletion of the CRP-binding site in the luxR-luxI intergenic sequence decreased the transcriptional intensity of P_luxI. The existence of one copy CRP-binding site (n = 1) promotes the transcription of P_luxI. However, further increase of the CRP-binding site numbers (n = 2, 3, 4) inhibited the gene expression under the control of the promoter P_luxI. We further performed EMSA experiments to confirm this result. As shown in Supplementary Fig. 7 and 8, the CRP protein indeed assists RNAP binding to the promoter P_luxR. When incorporating four CRP-binding sites in the locus, the binding band of RNAP with P_luxR showed no significant difference compared with that of one CRP-binding site in the locus. However, the binding band of LuxR and P_luxI was obviously shallowed. Based on these results, we believe that a high local concentration of CRP-cAMP might sterically hinder the binding of the LuxR-AHL complex to the promoter P_luxI.

We have demonstrated that an increase of CRP-binding site numbers activates the transcription of the P_luxR while inhibiting the transcription of P_luxI. It is worth noting that the activated P_luxR would further enhance the expression level of LuxR, leading to strong activation of the transcription of P_luxI⁹. When both P_luxR and P_luxI exist in the QS system, how the increased CRP-binding site numbers would control the transcription of promoter P_luxI should be further investigated. Thus, we placed LuxR under the control of the P_luxR promoter, and the expression of reporter gene eGFP was driven by the P_luxI promoter. We observed that the expression level of eGFP was increased by increasing the CRP-binding site numbers (Fig. 1d). The QS system with four copies of the CRP-binding site (n = 4) exhibited the highest P_luxI activity, which is 12.1-fold higher than that of the QS system with no CRP-binding site (n = 0). This result indicates that an increase of the CRP-binding site numbers could significantly expand the dynamic range of P_luxI when co-expressing the P_luxR controlled luxR in the system and luxI expressed from a constitutive promoter. We found that integration of more CRP-binding sites in the luxR-luxI intergenic sequence inhibits the transcription of P_luxI (in the absence of P_luxR) but significantly enhances the transcription of P_luxR (in the absence of P_luxI)_. However, when both P_luxI and P_luxR controlled luxR co-exist in the QS system, the activated P_luxR strongly activates P_luxI transcription. In this condition, an increase in the CRP-binding site numbers greatly improves the transcription level of P_luxI.

Next, we focused on identifying other target sites in the luxR-luxI intergenic sequence. It has been reported that the luminescence output varies over orders of magnitude between different Vibrio species²⁶. By sequencing and aligning the genomes of these diverse Vibrio strains, significant sequence differences were found in the lux box to the −10 site of the P_luxI promoter. Based on this, we hypothesized that the binding affinity could be altered by mutation of the lux box to −10 site sequence via directly using the homologous sequence from these different Vibrio strains. Thus, we used the natural QS system from V. fischeri ES114²⁷ as a template and created seven QS variants by replacing its lux box to −10 site sequence with the homologous one from other seven Vibrio species, including PP3²⁸, ES401²⁹, ET401³⁰, EM17²⁹, ET101³¹, H905²⁸, and MJ1¹⁹ (Fig. 1e). The lux box to −10 site sequence alignment showed that they shared high similarity (74.5%~95.7%) with V. fischeri ES114 (Supplementary Fig. 1b). To characterize circuit behaviors, we constructed eight pZE plasmids by individually employing these eight QS systems to drive green fluorescent protein (GFP) expression. The plasmid pCS-P_con-LuxR-LuxI was constructed for constitutively expressing genes luxR and luxI. It was co-transferred with each pZE plasmid into E. coli BW25113 (F’), yielding eight fluorescent strains. As we expected, these strains generated a variety of fluorescent profiles by continuous fluorescence measurements (Fig. 1f). Generally, the use of the homologous sequence to replace the natural one slightly decreased the transcription of P_luxI, the strain harboring the lux box to −10 site sequence from Vibrio ES401 exhibited the lowest peak fluorescence, which is 2-fold lower than that of the Vibrio ES114 (Fig. 1f). The EMSA experiments were carried out to test the relative binding strength of QS systems harboring different sources of lux box to −10 site sequence (Supplementary Fig. 9a). The results showed that the use of the homologous sequence to replace the natural one (ES114) decreased the relative binding strength of the QS system. The trends were well-matched with their transcriptional intensities (Supplementary Fig. 9b). These results indicate that mutation of the lux box to the −10 site region in the luxR-luxI intergenic sequence has only a slight inhibition effect on the transcriptional intensity of promoter P_luxI. However, we believe that a couple of this mutation with CRP-binding site number variation could further expand the diversity of QS variants. probably these mutations reduced the binding affinity between the promoter sequence and the transcriptional factor complex.

Establishing and characterizing a library of QS variants

We have created a series of QS variants that have different behaviors by mutation of the lux box to −10 site sequence and variation of the CRP-binding site numbers. Then, these two strategies were combined to construct a library of QS promoters with different triggering times (Fig. 2a). By altering the numbers of the CRP-binding site (n = 0, 1, 2, 3, 4) and the sequence of the lux box to the −10 site region (ES114(8), PP3(7), H905(6), EM17(5), ES401(4), MJ1(3), ET101(2), and ET401(1)), a total of 40 QS variants were generated. In order to obtain high dynamic range QS promoters, the P_luxR controlled luxR was also integrated into these systems. Continuous fluorescence measurements of these variants were performed to characterize their performance. As shown in Figs. 2b–d, these 40 QS variants exhibited a wide range of peak fluorescence intensities and triggering times. Generally, as the numbers of the CRP-binding site increased, peak GFP fluorescence increased too (Fig. 2b). The strain carrying the QS circuit with no CRP-binding site and the lux box to −10 site sequence from Vibrio ES401 produced the lowest peak fluorescence. The highest peak fluorescence was observed in the strain harboring the QS circuit with 4 CRP-binding sites and the lux box to −10 site sequence from Vibrio MJ1, which is 28.3-fold higher than that of the lowest strain (Fig. 2c). To compare the triggering times of these QS variants, we used the peak fluorescence value of the QS variant that has the lowest fluorescence signal in the library as a reference value. The “triggering OD” was defined as the cell density corresponding to the time of fluorescence reaching the reference value (Supplementary Fig. 10). As shown in Fig. 2d, triggering ODs of these variants ranges from 0.647 to 1.437, which covers the growth period from the early logarithmic phase to the stationary phase. We found that the triggering OD was decreased by increasing the number of CRP-binding sites. We tested the AHL accumulation rates of the QS systems harboring different CRP-binding sites (n = 0, 1, 2, 3, 4). As shown in Supplementary Fig. 11, the AHL accumulation rate was increased by increasing the number of CRP-binding sites, indicating that the decreased triggering OD resulted from the increased AHL accumulation rates. Altering the lux box to the −10 site sequence has only a slight effect on the triggering ODs under the same CRP-binding site number. These trends are consistent with our expectations based on the regulatory mechanism discovered in this study.

a Architecture of the QS variant in the library. b Gene expression dynamics of each QS variant in the library. Each QS variant was named P_QSMN, “M” represents the copy number of the CRP-binding site carried by the QS variant. “N” represents the source of the “lux box” to “−10” site sequence contained by the QS variant. c The peak fluorescence intensity of each QS variant in the library. d Triggering OD for each QS variant in the library. e Dynamic range for each QS variant. f Leaky expression for each QS variant. The leaky expression was exhibited as the ratio of comparing peak fluorescence intensity from strains carrying partial QS circuits in the absence of luxR and luxI to that of strains harboring entire QS circuits in the presence of luxR and luxI. Data shown are mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

To validate the function of these QS variants, we evaluated their dynamic range and leakiness. The dynamic range is defined as the ratio of the peak GFP fluorescence value of the strain carrying entire QS elements to that of the strain harboring partial QS circuits lacking the luxI gene. As seen in Fig. 2e and Supplementary Fig. 12a, we found that these QS variants exhibited a broad spectrum of dynamic ranges, ranging from 4.0-fold to 84.7-fold. It is worth noting that the highest dynamic range observed in the QS variant is 11.5-fold higher than the natural QS system without any modifications. For most promoters, improvement of their dynamic range usually leads to enhanced leakiness. Thus, we also evaluated the expression leakage of these QS variants, which was calculated as the ratio of comparing peak fluorescence intensity from the strains carrying partial QS circuits in the absence of luxR and luxI to the strains harboring entire QS circuits in the presence of luxR and luxI. Interestingly, except for QS variants lacking the CRP-binding site (n = 0), almost all of the QS variants in the library exhibited a declined leaky expression compared with the natural QS system from V. fischeri ES114 (Fig. 2f and Supplementary Fig. 12b). The leakiness of the QS variant was decreased by increasing the CRP-binding site numbers. Use of the homologous lux box to −10 site region sequences to replace the natural one also slightly reduced leaky expression. Especially, the QS variant harboring 4 CRP-binding sites and lux box to −10 site sequence from ES401 exhibited 6.7-fold lower leakiness than the natural QS system. These results indicate that variation of the lux box to −10 site sequence and increase of the CRP-binding site numbers reduced the leaky expression of QS systems, due to these mutations inhibited the transcription of P_luxI promoter. These finely tuned QS variants pave the way for implementing dynamic controls in complex metabolic networks.

Application of QS variants for autonomous and temporal regulation of multiple gene targets

Since our constructed QS variants have a high dynamic range, low leakiness, and different triggering times, those circuits can be used to simultaneously regulate three or more metabolic fluxes in a single cell, which has not yet been achieved before. We hypothesized that sequential transcription of multiple sets of genes can be achieved by regulating their expression under the control of the different QS variants. To test our hypothesis, we selected three QS variants with different triggering OD as 0.79, 0.89, and 0.92 to control the expression of three different reporter genes egfp, mCherry, and mOrange2, respectively. We constructed the plasmid pZE-luxR-P_QS37–luxI–eGFP-P_QS18–mCherry-P_QS13–mOrange2 and transferred it into E. coli, yielding strain CGemm (Fig. 3a). We measured the fluorescence of strains that carry no reporter proteins as negative controls (Supplementary Fig. 13). The normalized relative fluorescence intensities of these three reporter proteins in the strain CGemm were measured and compared (Fig. 3b). As we expected, the green fluorescent protein eGFP under the control of the QS variant P_QS37 was expressed first. Then, the expression of the red protein mCherry was initiated about one hour later. Finally, the yellow fluorescence produced by the protein mOrange2 under control of the QS variant P_QS13 was detected. The expression order of these three reporter genes observed through fluorescence characterization matches well with the predicted order. To further confirm the consistency, we used a laser scanning confocal microscope to directly observe the fluorescence of the strain CGemm. As shown in Fig. 3c, the green, red, and yellow fluorescence was sequentially observed in the fluorescent images, which is consistent with the predicted expression order. In order to confirm the sequential expression is not caused by the different maturation times of the reporter proteins, we placed the eGFP under the control of the QS promoters P_QS37, P_QS13, and P_QS18 by constructing plasmids P_QS37–eGFP-CO, P_QS13–eGFP-CO, and P_QS18–eGFP-CO, respectively. We tested the fluorescence profiles of the strains harboring these circuits. As shown in the Supplementary Fig. 14, when using P_QS37, P_QS13, and P_QS18 to drive the expression of the same reporter gene, the expression sequence is consistent with that in the Fig. 3b, c, indicating that the sequential expression is not caused by the different maturation time of the reporter proteins. These results indicate that our constructed QS variants are reliable tools to temporally and autonomously control the expression of multiple gene targets in a single cell.

a Architecture of the QS circuit that simultaneously regulates the expression of eGFP, mCherry, and mOrange2, under control of the P_QS37, P_QS18, and P_QS13, respectively. b Fluorescence profiles for eGFP, mCherry, and mOrange2 in the strain containing the designed QS circuit. For all three fluorescent proteins, the fluorescent intensity of each time point was normalized by their corresponding peak fluorescence value. c Laser scanning confocal images captured in a single cell containing the designed QS circuit. Images were taken at 7, 9, 12, and 22 h and under three different excitation wavelengths. Bar = 40 μm. The data were generated from three independent experiments. d Architecture of the QS circuit for simultaneous up- and downregulation of two different gene targets. e Peak fluorescent intensities of mCherry and eGFP when introducing CRISPRi into this system. Data shown are mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

The above-established QS circuits only have the function of activating transcription. To expand the functionality of the QS system, we tested the potential of using CRISPRi as a switch to achieve signal conversion. Thus, the AHL-induced upregulation was transduced into a downregulation function by integrating the CRISPRi system into QS circuits. To confirm whether simultaneously dynamic upregulation and downregulation can be achieved in an orthogonal manner, we constructed plasmid pZE-luxR-P_QS18–luxI–eGFP as the reporter of the upregulation function. To construct the downregulation function reporter, the plasmid plv-P_QS18–dCas9–sgRNA was constructed to repress the mCherry expression on the genome of E. coli BW25113 (F′) (Fig. 3d). As seen from Fig. 3e, in the presence of the AHL and CRISPRi system, the eGFP fluorescence intensity was up-regulated by 7.7-fold, and the mCherry fluorescence intensity was downregulated by 82.9% by the end of 24 h. This result demonstrates that coupling the CRISPRi system with the QS variants could generate a dynamic and bi-functional regulation tool for simultaneously up- and downregulate two sets of genes.

Dynamic regulation of salicylic acid biosynthesis

Next, we chose the biosynthesis of SA as a model to verify the effectiveness of designed genetic circuits in regulating intracellular metabolic fluxes. SA is a value-added chemical that has been used to produce medicines, perfumes, and dyes. In E. coli, biosynthesis of SA can be achieved by extending the shikimate pathway via overexpression of entC and pchB (Fig. 4a)³². However, SA is toxic to the E. coli strain and 1 g/L SA completely inhibits the cell growth³². We hypothesized that our designed QS variants could be used to improve SA titer by switching on the expression of pathway genes at appropriate time points. To test this hypothesis, QS promoters P_QS13, P_QS18, P_QS37, and P_QS33 from the library (Fig. 2b) were individually employed for driving the expression of entC and pchB, generating four SA upregulation modules. As a control, entC and pchB were constitutively expressed to provide a statically controlled SA upregulation module. As shown in Fig. 4b, the static expression strain only generated 295.1 mg/L SA. While, the strain harboring QS P_QS13 controlled pathway genes produced the highest SA titer of 523.2 mg/L, demonstrating a 72% increase compared with the static strain. It was also found that the strain carrying the QS variant with lower triggering OD produced less amount of SA. This demonstrates that delaying the expression of SA pathway genes is beneficial for improving the SA titer, probably due to the released metabolic burden to the host cell. Additionally, we observed no improvement in the SA titer in the absence of the luxI, indicating that the increase in SA production is induced by the QS dynamic regulation. We also tested the cell growth rates, glucose levels, and the SA titer of the SA-producing strains at different times (Supplementary Fig. 15a–c), the results showed that these factors make no contribution to the titer improvements.

a Schematic construction of layered dynamic control circuits for regulation of the SA biosynthetic pathway. b The effect of different triggering OD of QS variants driving entC-pchB expression on SA titer. Gray bars represent SA titers generated by the strains harboring partial QS circuits in the absence of the gene luxI. Red bars represent SA titers generated by the strains harboring entire QS circuits in the presence of the gene luxI. c The effect of different triggering OD of QS variants driving entC-pchB combined with different ppc downregulation levels on SA titer. d The effect of three-layered dynamic regulation strategy on SA titer in wildtype E. coli strain. The upregulation module containing SA pathway genes was controlled by the QS variant P_QS37, the downregulation module carrying the CRISPRi system was controlled by the QS variant P_QS18, and the APTA module was controlled by QS variant P_QS13, P_QS18, P_QS37, or P_QS33, respectively. e The effect of three-layered dynamic regulation strategy on SA titer in QH4 strain. The upregulation module containing SA pathway genes was controlled by the QS variant P_QS37, the downregulation module carrying the CRISPRi system was controlled by the QS variant P_QS18, the APTA module was controlled by QS variant P_QS13, P_QS18, P_QS37, or P_QS33, respectively. NA represents the SA titer generated by the control strain lacking the APTA module. Genes: tktA encodes transketolase, ppsA encodes phosphoenolpyruvate synthase, aroG encodes 3-deoxy-D-arabinoheptulosonate-7-phosphate (DHAP) synthase, aroL encodes shikimate kinase, ppc encodes phosphoenolpyruvate carboxylase, entC encodes isochorismate synthase, pchB encodes isochorismate pyruvate lyase. Metabolites: E4P D-erythrose-4-phosphate, PEP phosphoenolpyruvate. Data shown are mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Then, we aimed to precisely control SA production by using the QS variants to simultaneously regulate two metabolic fluxes. Having a sufficient supply of phosphoenolpyruvate (PEP) has been proved to be a key factor for the biosynthesis of the shikimate pathway-derived compounds³³. Except as a precursor for the shikimate pathway, PEP also serves as an essential metabolite in glycolysis and the TCA cycle³⁴. However, direct disruption of these competing pathways would impair cell growth^35,36,37. Especially, direct deletion of the phosphoenolpyruvate carboxylase (encoded by ppc) in E. coli significantly reduced cell growth³⁵. Alternatively, we attempted to dynamically downregulate ppc to balance the relationship between production and growth. The QS promoter P_QS18 was used to control the expression of the CRISPRi system. We designed four sgRNAs targeting different locations of ppc operon, including ribosome binding site (sgppc1), start codon region (sgppc2), coding sequence at 261-bp (sgppc3), and coding sequence at 1406-bp (sgppc4). This generated four different downregulation modules that can inhibit the expression of ppc at different levels (sgppc1 > sgppc2 > sgppc3 > sgppc4) (Supplementary Fig. 3a). Additionally, as a control, we constructed the fifth downregulation module by complete disruption of ppc. These five downregulation modules were combined with the former constructed five SA upregulation modules, yielding a library of 25 SA-producing strains. Generally, complete deletion or high-level downregulation (sgppc1 and sgppc2) of ppc resulted in only negligible SA titers, probably due to the cell growth retardation (Fig. 4c and Supplementary Fig. 16b). While higher SA titers were observed in dynamic downregulation of ppc at lower levels using sgppc3 and sgppc4. By quantitative real-time PCR analysis, the use of sgppc3 and sgppc4 inhibited the expression level of ppc by 42.4% and 33.6%, respectively (Supplementary Fig. 16a). Among the dual dynamically regulated strains that contain both upregulation module and downregulation module, downregulating the expression of ppc using sgppc4 and placing the SA pathway genes under the control of the P_QS37 is the optimal combination for SA production. This optimal SA producer generated 1010.6 mg/L SA after 48 h fermentation. As a control, SA biosynthesis was conducted by static overexpression of SA pathway genes and downregulation of ppc using sgppc4. As another control, SA pathway genes were placed downstream of the P_QS37, and this circuit was introduced into a host strain with ppc disruption. These two control strains only produced 470.8 and 35.5 mg/L SA, which were 2.1-fold and 28.5-fold lower than that of the optimal SA producer. These results demonstrate that the timing of target pathway expression and competing pathway repression must be carefully tuned to identify the optimal SA producer. We also tested the final titer relative to the total substrate consumed, the final titer relative to the biomass, and SA titer dynamics over cultivation time (Supplementary Fig. 15d–f), the results showed that all of the trends are consistent with the final titers observed in the Fig. 4c.

As the next step, we examined the effect of simultaneous dynamic control of three metabolic fluxes on SA biosynthesis. We previously constructed a chorismate-boosting plasmid pCS-APTA that carries aroL, ppsA, tktA, and aroG^fbr (a feedback inhibition resistant mutant of aroG)³⁸. The QS promoters P_QS13, P_QS18, P_QS37, and P_QS33 were individually used to drive the expression of the APTA module, generating plasmids pSA-P_QS13-APTA, pSA-P_QS18-APTA, pSA-P_QS37-APTA, and pSA-P_QS33-APTA. These four plasmids were individually transferred into the above-obtained optimal SA producer, yielding four SA producers that harbor QS circuits controlling three metabolic fluxes, as (1) the upregulation module containing SA pathway genes was controlled by the QS variant P_QS37, (2) the downregulation module carrying CRISPRi system was controlled by the QS variant P_QS18, and (3) the APTA module was controlled by QS variant P_QS13, P_QS18, P_QS37, or P_QS33, respectively. As seen from Fig. 4d, the addition of the third layer over-expresses APTA further enhancing the SA titers. Especially, the use of the P_QS33 for driving APTA module expression led to 1409.8 mg/L SA, showing a 43% increase compared with the strain that only has two layers of dynamic flux control. This indicates that the place of APTA downstream of the P_QS33 provided the optimal time for initiating the expression of the APTA module. Additionally, we tested the three-layered dynamic control circuits in a phenylalanine over-producer QH4, which is derived from the disruption of pheA and tyrA in E. coli ATCC31884. As shown in Fig. 4e, transform of the strain background from wildtype E. coli to QH4 led to a further increase in SA titers. The best-performed producer generated 2.1 g/L SA (546.5 mg/g CDW, 700 mg/L/day), which is the highest production achieved in shake flask experiments. Interestingly, we found that the optimal QS promoter for controlling the expression of APTA module in wildtype E. coli is P_QS33, with a triggering OD of 0.73. While, the highest SA titer in QH4 was observed when the APTA module was controlled by the QS promoter P_QS18, with a higher triggering OD of 0.89. This indicates that delayed overexpression of the APTA module in QH4 is beneficial for SA production. Owing to the carbon flux through the shikimate pathway in QH4 is higher than in wildtype E. coli strain, too earlier overexpression of APTA module in QH4 might bring metabolic burden to the host. Taken together, these results demonstrate that our designed QS variants are powerful tools for improving microbial production by simultaneously and dynamically regulating multiple metabolic fluxes.

Dynamic modular optimization of 4-hydroxycoumarin production

Modular optimization, including tuning gene expression level by altering plasmid copy number and engineering promoter or ribosomal binding site (RBS) strength, has been proved to be an effective strategy for pathway balancing^39,40,41. However, the currently developed modular optimization method only can modulate enzyme expression levels on a static mode, which is unable to respond to intracellular and environmental changes. We hypothesized that our designed QS circuits could be applied to dynamically optimize carbon flux distribution among different pathway modules to improve production. We utilize the 4-HC biosynthetic pathway as a demonstration to verify our hypothesis. 4-HC is a value-added compound used in the production of anticoagulant drugs, and also plays an important role in treating thromboembolic diseases^42,43. The 4-HC biosynthetic pathway consists of two modules, (1) EP module containing EntC and PchB that converts chorismate to SA, (2) PS module carrying salicyl-CoA ligase (SdgA) and biphenyl synthase (PqsD) that is responsible for generating 4-HC via condensation of SA with malonyl-CoA (Fig. 5a)³⁸. The incongruous expression of the EP and PS modules results in metabolic imbalance, which serves as a rate-limiting factor for 4-HC synthesis⁴⁴. Additionally, chorismate and malonyl-CoA are two essential metabolites in E. coli⁴⁵, indicating the trade-off between 4-HC production and cell growth. We hypothesized that the use of QS variants for dynamically modular optimization of EP and PS modules would improve 4-HC titer by balancing the carbon flux and initiating the expression of each module at the optimal time.

a Schematic of construction of dynamic control circuits for regulation of 4-HC biosynthetic pathway. b The effect of different triggering OD of QS variants individually driving entC-pchB expression and pqsD-sdgA expression on 4-HC titer. Con represents 4-HC titer generated by the control strain expressing EP and PS modules on constitutive promoters. Gene: entC encodes isochorismate synthase, pchB encodes isochorismate pyruvate lyase, pqsD encodes CoA ligase, sdgA encodes biphenyl synthases. Data shown are mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

To verify the effectiveness of dynamic modular optimization on 4-HC production, four QS promoters P_QS13, P_QS18, P_QS37, and P_QS33 with different triggering OD were selected. Each of these QS promoters was applied to drive the expression of the EP module and/or PS module, resulting in a library of sixteen 4-HC producers. These 4-HC producing strains are able to switch on the expression of EP and PS modules at different rates in a dynamic and autonomous mode. As shown in Fig. 5b, these strains produced varying 4-HC titers. The best producer harboring EP module controlled by the P_QS18 and PS module controlled by the P_QS37, generated the highest 4-HC titer of 288.8 mg/L, representing an 11.3-fold increase over the static producer. Generally, the EP module driving by QS promoter with higher triggering OD and PS module driving by QS promoter with lower triggering OD led to higher 4-HC titers, indicating that 4-HC production could be improved by fine-tuning the timing of each module expression. These results demonstrate the generalizability and effectiveness of our designed QS circuits for dynamic pathway engineering.