Background

Polyketide synthases (PKSs) are classified into three types based on their enzyme structures [1]. Type III PKSs are ubiquitous in plants, bacteria, and fungi [2,3,4]. As an important class of enzymes, they synthesize a variety of commonly used and valuable natural aromatic products (secondary metabolites), including curcuminoid, chalcone, stilbene, coumarin, and chromone [5, 6]. A type III PKS, chalcone synthase (CHS), is also the key enzyme involved in flavonoids biosynthesis [7, 8]. Type III PKSs generally catalyze the iterative condensation of malonyl-coenzyme A (CoA) with a CoA-linked starter molecule. They use free CoA thioesters as substrates, in contrast to type I/II PKSs, which employ acyl carrier proteins for catalysis. Type III PKSs also have broad substrate specificities, making them ideal for engineering the synthesis of diverse polyketide compounds [9]. However, despite their utility and simpler structures, our understanding of type III PKSs lags far behind that of type I/II PKSs.

Type III PKSs catalyze the formation of many valuable secondary metabolites; nevertheless, their catalytic efficiencies are limited and these products usually exist in low amounts in nature. Previous studies have been focused on improving the diversity of secondary metabolites produced, based on the broad source and substrate specificity of type III PKSs [6, 10]. Engineering strategies have also been applied to broaden the function diversity of type III PKSs to obtain various polyketide compounds [9]. The catalytic efficiency of these enzymes is critical for the recombinant overproduction of valuable secondary metabolites in microbial cell factories; however, there are few reports on type III PKSs engineering for improved catalytic activities [11], likely due to the hardly simulated intracellular environments in whole-cell biosynthesis (e.g., the low and dynamically variable concentrations of CoA precursors). Beneficial enzyme mutants obtained from in vitro engineering frequently perform poorly during in vivo biosynthesis because of the different work environment. Thus, in vivo-directed evolution under real background metabolic conditions is preferred for type III PKS biosynthesis, yet a lack of rapid screening methods retards the application of this strategy.

CHS is the most well studied among type III PKSs [12,13,14]. It catalyzes the condensation of one molecule of p-coumaroyl-CoA and three molecules of malonyl-CoA to form naringenin chalcone, which is converted to naringenin, either spontaneously or by chalcone isomerase (CHI) [8] (Fig. 1). Variation in the volume of CHS active sites has been shown to alter the size of acceptable starter units and result in the formation of different final polyketide products [13, 15]. Apart from chalcone products, CHS has been shown to produce the derailment byproducts p-coumaroyltriacetic acid lactone (CTAL) [16, 17] and bis-noryangonin (BNY) [18] via the lactonization of reaction intermediates (Fig. 1). So far no way has been reported to reduce byproducts formation, resulting in low naringenin chalcone yields. Besides CHS, byproducts formation frequently occurs in most other type III PKSs catalysis [18,19,30]. High-throughput screening methods are critical for successful in vivo-directed evolution. Although biosensors for specific final products are important tools for in vivo high-throughput screening [30, 31], strategies that are readily adapted to the screening of enzyme groups with similar functions are more favorable. In this study, a growth selection system used for the in vivo-directed evolution of CHS was developed based on the inhibition of cell growth by p-coumaroyl-CoA starter molecule.

In type III polyketide biosynthesis, starter acyl-CoA substrates are usually synthesized by acyl-CoA ligases with corresponding carboxylic acids as substrates, using whole-cell catalysts (Fig. 2A). Strain CUR01 [25] (Table S1) was used for naringenin chalcone and naringenin biosynthesis in this study. Two p-coumarate:CoA ligases 4CL2M and 4AT were overexpressed in CUR01. Upon p-coumaric acid supplementation, cell growth was significantly inhibited, especially in the group of 4CL2M (Fig. 2B), which indicated that accumulation of excessive intracellular acyl-CoAs can be cytotoxic [32, 33] and disrupt CoA metabolism. Furthermore, a derepression in growth was observed with CHS co-expression. But the inhibition in cell growth did not completely recover, indicating that CHS activity was insufficient (Fig. 2C). Thus, the extent of cell growth recovery under substrate p-coumaric acid supplementation can be used to screen CHS activity. Through iterative rounds of cell growth screening, the cells harboring improved CHS which exhibited higher capabilities of p-coumaroyl-CoA conversion would be enriched. This growth selection mechanism can be potentially extended to the ultrahigh-throughput screening of most type III PKS activities (Fig. 2A). In addition, this selection system was simple, rapid and low-equipment dependent.

Fig. 2
figure 2

Development of a growth selection system for CHS engineering. A Scheme of in vivo biosynthetic pathway and the growth selection principle; B growth inhibition of cells overexpressing p-coumarate:CoA ligase 4AT and 4CL2M when p-coumaric acid (4.5 mM) was supplemented. Cells expressing 4CL2M in the absence of p-coumaric acid was used as control; C co-expression of CHS alleviated growth inhibition. Relative growth rate is calculated as the percentage of cell growth in the presence compared to in the absence of p-coumaric acid; D flowchart of in vivo-directed evolution of CHS

Full size image

CHS engineering

The growth selection system was then applied to engineer CHS in a naringenin biosynthetic pathway (Fig. 2D) that consisted of a 4CL2M enzyme and the CHS random mutagenesis library in strain CUR01, with 4.5 mM p-coumaric acid as substrate. After enrichment through five rounds of screening, strains from the library were mounted on agar plates, five largest clones were selected, and the products were analyzed using HPLC. As shown in Fig. 3A, in contrast to the wild-type CHS, mutants CHS-1-1, -2, -3, -5, and -7 exhibited improved production of naringenin or CTAL byproducts, whereas no naringenin chalcone was detected (Fig. S2). Sequencing results revealed the amino acid substitutions of these mutants (Table 1).

Fig. 3
figure 3

Engineering CHS for improved catalytic efficiency. A Naringenin and CTAL productions from strain CUR01 harboring the naringenin biosynthetic pathway expressing CHS mutants selected from three rounds of random mutagenesis; B specific activities of CHS mutants selected from the 2nd and 3rd rounds of random mutagenesis; C naringenin and CTAL productions from strain CUR01 harboring the indicated CHS mutants; D naringenin and CTAL productions from strain CUR01 harboring CHS with T131 saturated mutated. The mutants displaying no activity were not shown; E naringenin and CTAL productions by the indicated purified CHS enzymes, using p-coumaroyl-CoA and malonyl-CoA as substrates. CHS-WT, CHS wild-type. Results are shown as mean ± SD from at least n = 3 biological replicates in each experimental condition, n.s., not significant, *p < 0.05, **p < 0.005, Student’s two-tailed t-test

Full size image
Table 1 Mutations of CHS mutants
Full size table

Mutant CHS-1-1 was used as the parent enzyme to construct a library for the second round of random mutagenesis. Two mutants that produced higher levels of both naringenin and CTAL were selected in this round, and mutant CHS-2-7 was used as the parent enzyme for a third round of random mutagenesis, during which three mutants, CHS-3-3, -15, and -27, were obtained. Mutant CHS-3-15 displayed an approximately threefold increase in naringenin production relative to the wild-type enzyme, making it the best CHS mutant obtained in this study (Fig. 3A). The specific activities of the purified CHS-2-7, 2-12, 3-15, and 3-27 mutants were then determined. Since naringenin chalcone was rapidly converted to naringenin (Fig. S2), naringenin production in the presence of two acyl-CoA substrates was used to evaluate enzyme activity. Compared to the wild-type CHS, the activities of CHS-2-7, 2-12, 3-15, and 3-27 were increased by 1.31-, 1.12-, 2.30-, and 2.13-fold, respectively (Fig. 3B).

Evaluation of byproduct formation by CHS mutants

Since the principle of our growth selection system was based on the consumption of the acyl-CoA starter molecule, it did not exclude mutants forming derailment byproducts, whereas elimination of byproducts formation and improvement of product specificity is of great necessity to increase biosynthesis of the desired product naringenin. E values [22] were introduced to evaluate the product specificities of the CHS whole-cell catalysts. Since naringenin chalcone was rapidly converted to naringenin and BNY (Fig. 1) was almost negligible, naringenin concentration [NAR] and CTAL concentration [CTAL] were used for E value calculation as follows:

$$E=\frac{\left[\text{NAR}\right]-[\text{CTAL}]}{\left[\text{NAR}\right]+[\text{CTAL}]}\times 100\text{\%}.$$

A positive E value inferred that the rate of naringenin production exceeded that of CTAL formation, whereas a negative E value inferred that the rate of CTAL formation exceeded that of naringenin production. It was found that E value of the best mutant CHS-3-15 was 50.1%, almost similar with that of wild-type enzyme (50.0%). Mutants from the first round of screening were found to produce high levels of CTAL byproduct (Fig. 3A). Among them, mutant CHS-1-3 exhibited a negative E value, indicating that the rate of CTAL formation exceeded that of naringenin chalcone production (Fig. S3A). To explore the contributions of two amino acid substitutions in mutant CHS-1-3 (Table 1) to CTAL formation, two single mutants—T131S and C341S—were constructed. In particular, T131S was proved critical to the reduction of naringenin and improved CTAL production (Fig. 3C and S3B). The T131 position is located in the substrate-binding pocket. Site-saturation mutagenesis of T131 was performed, and only six mutants showed detectable activity. Among them, substitutions S and Q produced less naringenin but much higher levels of CTAL than the wild-type enzyme. Both T131A and T131G produced almost no detectable naringenin but significant levels of CTAL, with E values close to − 100% (Fig. 3D and S3C). We further explored the product profile of purified T131A mutant in vitro. Similar to that in in vivo reactions, naringenin production was dramatically reduced, whereas high levels of CTAL were generated. Thus, this mutation significantly improved byproduct formation in vivo and in vitro (Fig. 3E).

Mutants CHS-1-1 and -7 shared an M159L mutation, whereas CHS-1–5 and -7 shared a K62N mutation. Thus, two mutants with single mutations—K62N and M159L—were constructed. Both of them were found to contribute to improvements in naringenin and CTAL production, and an additive effect was observed when both mutations occurred simultaneously (Fig. 3C).

Host-cell engineering for improved naringenin synthesis

As shown in Fig. 1, apart from type III PKSs, the availability of intracellular malonyl-CoA, CoA, and adenosine triphosphate (ATP) affected the biosynthetic efficiency of the final polyketide products. Therefore, the metabolic network of host cells plays an important role in type III polyketide biosynthesis, and the growth selection system could also be applied in host cell genome engineering. Gene targets that promote the conversion of acyl-CoA would improve the cell growth. The E. coli ASKA (A Complete Set of E. coli K-12 ORF Archive) [34] library was used to transform strain BW25113, which expressed the naringenin biosynthetic pathway, and the mutants with improved production were screened via our growth selection system. Three colonies were selected for naringenin and CTAL determination, although none of them contributed greatly to naringenin biosynthesis, the CTAL formations were improved. Sequencing results revealed that the overexpressed genes were hyfA, yigI and tesB (Table S3). Interestingly, both yigI and tesB encode thioesterases, and the overexpression of tesB resulted in a 6.23-fold increase in CTAL production (Fig. 4A). Gene tesB encodes a thioesterase II, which has a relatively broad substrate specificity and cleaves medium- and long-chain acyl-CoAs [35].

Fig. 4
figure 4

Directed genome engineering for improved naringenin synthesis. A Naringenin or CTAL productions of strain CUR01 harboring wild-type naringenin biosynthetic pathway co-expressed with the indicated proteins; B naringenin and CTAL productions of purified CHS in the presence of purified CHIL or TesB. Biosynthesis of naringenin by CHS3-15 mutant in strain NAR01, in the absence (C) or presence (D) of CHIL expression. The results are shown as mean ± SD from at least n = 3 biological replicates in each experimental condition, n.s., not significant, *p < 0.05, **p < 0.005, Student’s two-tailed t-test

Full size image

To further explore the effects of thioesterases on CHS activity, another seven genes encoding thioesterases in E. coli were individually selected for co-expression with the naringenin biosynthetic pathway. Notably, none of the thioesterases facilitated naringenin production, whereas three led to improved CTAL production (negative E values). Among all the tested thioesterases, the overexpression of TesB yielded the highest CTAL production, and the E value reached − 58.6% (Figs. 4A and S3D).

Exploring the byproduct formation mechanism in CHS catalysis

Unexpectedly, some thioesterases significantly improved the biosynthesis of CTAL byproducts in CHS reactions. To remove the intracellular background, in vitro reactions of purified CHS in the presence or absence of purified TesB were performed, using p-coumaroyl-CoA and malonyl-CoA as substrates. We found that in the presence of TesB, CTAL formation was improved while naringenin synthesis declined (Fig. 4B). Thus the in vitro enzymatic reactions confirmed the effect of TesB on CTAL production. The catalytic mechanisms of CHS revealed the formation of thioester intermediates during these reactions (Fig. 1). In naringenin biosynthesis, the thioester intermediate of CHS, compound 1, was not stable and quickly cyclized to form lactone CTAL upon CoA hydrolyzation, whereas CTAL could no longer be converted to naringenin (Fig. 1). Our results suggest that TesB, which catalyzes the hydrolysis of acyl-CoA substrates, could improve CTAL yields. This is likely owing to the release of thioester intermediates 1 from CHS enzymes, which were cleaved by TesB outside of the catalytic pocket, leading to the formation of CTAL, which could no longer be converted.

Previously, the mechanism of byproduct formation in CHS was suggested to involve a lactonization-type ring closure, other than Claisen condensation in the catalytic pocket [17, 36, 37]. Our results indicate that it may also result from the release of catalytic intermediates from the catalytic site. Since compounds 1 was not commercially available, direct thioesterase catalysis was not performed. However, both in vivo and in vitro reactions had supported our prediction that the catalytic intermediate of CHS, compound 1, was at least partially released to the environment and converted to CTAL byproduct when CoA was hydrolyzed. Therefore, strategies preventing the release of such intermediates should reduce byproduct formation and promote naringenin chalcone production.

The T131A mutant obtained in this study displayed dramatic changes in product and byproduct formation (Fig. 3D and E). To further understand the interactions between compound 1 and CHS wild-type or T131A mutant, molecular docking and 20-ns molecular dynamics simulations were performed for the CHS-compound 1 complexes using YASARA. The average binding energy for compound 1 against CHS wild-type and T131A mutant were -999.2 and -876.0 kJ mol−1, respectively (Fig. S4). The calculations indicated that compound 1 showed a weaker interaction with the active site of T131A mutant, comparing with wild-type CHS, thus leading to an easier release of compound 1 from T131A enzyme.

Reaction intermediates are more likely to be released when a protein structure is unstable. Therefore, the FireProt web tool [38] was employed to guide the design of more stable CHS mutants. Six single-point mutations were predicted using an energy-based approach (Table S4) and the corresponding CHS mutants were constructed and applied to the whole-cell biosynthesis of naringenin. Only one mutant—T145M—produced notably higher levels of naringenin, and it simultaneously produced less CTAL than the wild-type enzyme (Fig. 3E). The DynaMut tool revealed that T145M substitution led to the formation of new hydrophobic bonds despite removing weak hydrogen bonds (Fig. S5). The ΔΔGWT-MT was predicted as 0.818 kcal mol−1, indicating an increased stability of the T145M protein. Combination of CHS-3–15 mutant and T145M substitution also resulted in further improvement in naringenin and reduction in CTAL synthesis (Fig. S6). These results further indicated that improving CHS stability would reduce the chance of intermediate release and could be a practical way of minimizing byproduct formation. This finding can be potentially expanded to most type III PKSs for reducing byproduct formation.

Effect of CHIL in naringenin biosynthesis

As CHIL has been reported to interact CHS and facilitate naringenin chalcone biosynthesis [21, 22], CHIL from A. thaliana was introduced to further improve naringenin production in this study. The results of in vitro reactions confirmed a significant improvement in naringenin production in the presence of CHIL proteins, whereas CTAL levels decreased slightly (Fig. 4B). CHS reactions were then performed in the presence of both CHIL and TesB and we found that, compared to the presence of only CHIL, TesB supplementation further reduced naringenin biosynthesis while improving CTAL formation. However, with the presence of both CHIL and TesB, naringenin biosynthesis remained notably higher than in presence of TesB alone, indicating that CHIL protected the reaction intermediates from degradation by thioesterase (Fig. 4B). In a previous study [22], a binding site of compound 1 was found in the CTAL structure via computational modeling, therefore, we inferred that CHIL bound the compound 1 released from CHS, protecting it from degradation by TesB. The CHIL–CHS complex was modeled and then docked with compound 1 (Fig. 5), and we found that the binding sites of compound 1 in CHS and CHIL were rather close, facilitating the intermolecular transport of this reaction intermediate. Since CHIL interacted with CHS [21, 22], the bound compound 1 could later return to the CHS catalytic pocket again and finally be converted to naringenin chalcone. Therefore, CHIL improved naringenin chalcone synthesis through binding and stabilizing the open-ring configuration of the intermediates during catalysis. Our study provides indirect evidence for the mechanisms of byproduct formation by CHS and CHIL protection for improved naringenin chalcone synthesis, shedding light on develo** more efficient whole-cell catalysts for naringenin and downstream flavonoids.

Fig. 5
figure 5

Structure model of CHS–CHIL complex docked with compound 1. The yellow and purple proteins represent CHS and CHIL, respectively. Compound 1 was docked in both CHS and CHIL

Full size image

Development of highly efficient and specific naringenin whole-cell catalyst

The biosynthesis of naringenin and downstream flavonoid compounds is mostly achieved with whole cells. Various thioesterases in a genome can lead to high levels of byproduct formation, lowering the productivity of the final products. Therefore, the inactivation of these thioesterases is necessary for the high-efficiency synthesis of these compounds. As a proof of concept, strain NAR01 was constructed based on strain CUR01, in which five thioesterase genes, tesB, yigI, yciA, fadM, and paaI, were deleted to reduce CTAL formation. Naringenin biosynthesis was performed with strain NAR01 that harbors plasmids pTrc99a-4CL2M-CHIL/pTrc99a-4CL2M and pYk-CHS, which express the naringenin pathway containing the CHS-3–15 mutant (Fig. 3A), in the presence or absence of CHIL (Fig. 4C and D). The time series of cell growth, product and byproduct formation, and substrate consumption were determined. In the absence of CHIL, naringenin production in strain NAR01 reached 3.24 mM in the flask, with a E value of 78.9% (Fig. 4C), while the E value was only 50.1% for the same biosynthetic pathway expressed in strain CUR01. Under CHIL expression, naringenin production was further improved to 3.98 ± 0.09 mM (1082 ± 24 mg L−1), whereas the accumulation of CTAL was minimized (E value = 96.7%) (Fig. 4D). This was the highest production in flasks reported to date [39,40,41] (Table S5). Therefore, a combination of endogenous thioesterases inactivation and CHIL co-expression finally resulted in an upgraded naringenin whole-cell catalyst with significantly improved product specificity. Combined with the previously reported strategies of engineering intracellular malonyl-CoA availability [42,43,44], the biosynthetic capability of the whole-cell catalyst was expected to be further improved.

Conclusions

In conclusion, owing to the inhibition of cell growth by the acyl-CoA starter molecule, a high-throughput screening strategy, coupled with growth-based whole-cell catalysts of type III polyketide naringenin, was developed. This growth selection system has greatly contributed to both enhanced activity and discovery of byproduct formation mechanism in CHS. Our study provided new insights in the catalytic mechanisms of CHS and shed light on engineering heterologous bio-factories to produce high-value type III polyketides at a large scale. Furthermore, the mechanism of this simple and rapid growth-based screening strategy is potentially applicable to the engineering of most type III PKSs, regardless of their final products.