Reshaping the 2‐Pyrone Synthase Active Site for Chemoselective Biosynthesis of Polyketides

Abstract

Engineering enzymes with novel reactivity and applying them in metabolic pathways to produce valuable products are quite challenging due to the intrinsic complexity of metabolic networks and the need for high in vivo catalytic efficiency. Triacetic acid lactone (TAL), naturally generated by 2‐pyrone synthase (2PS), is a platform molecule that can be produced via microbial fermentation and further converted into value‐added products. However, these conversions require extra synthetic steps under harsh conditions. We herein report a biocatalytic system for direct generation of TAL derivatives under mild conditions with controlled chemoselectivity by rationally engineering the 2PS active site and then rewiring the biocatalytic pathway in the metabolic network of E. coli to produce high‐value products, such as kavalactone precursors, with yields up to 17 mg/L culture. Computer modeling indicates sterics and hydrogen‐bond interactions play key roles in tuning the selectivity, efficiency and yield.

Implementation of novel synthetic pathways into metabolic network is challenging. Empowered by protein engineering and synthetic biology, a whole‐cell transformation process is reported to utilize engineered 2‐pyrone synthase with novel reactivity in metabolic networks to convert carboxylic acids and intracellular malonyl‐CoA directly into the pyrone products with high yields, catalytic efficiency and good selectivity.

Introduction

Metabolic engineering utilizes the endogenous metabolic network of living organisms to offer a cost‐effective generation of valuable products. Traditional metabolic engineering approaches rely on natural enzymes’ reactivities to optimize naturally evolved pathways or redirect flux towards non‐natural products.[ ¹ , ² ] In addition, the space of possible metabolic pathways can be further expanded through high‐level metabolic engineering by incorporating novel reactions from enzymes of modified substrate scopes and even novel chemistry from de novo designed enzymes. However, implementation of novel synthetic pathways into metabolic networks is challenging since they often interfere with endogenous metabolism, leading to side reactions and dead‐end metabolites. ^[3] On the other hand, while engineered enzymes sometimes may have high catalytic efficiency outside the cells, they may not be compatible with other enzymes in vivo and thus become a bottleneck of biosynthetic pathways. Because of these limitations, higher‐level metabolic engineering is difficult when the active sites of the proteins are redesigned to exhibit novel or enhanced reactivities and then such engineered proteins are incorporated into metabolic networks to produce valuable products.[ ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ ]

A primary example of such valuable products generated via metabolic engineering is the polyketide family. ^[10] Polyketides are a large class of biomolecules from bacteria, fungi and plants that have shown many clinically important biological properties, such as anticancer, antimicrobial, antioxidant and anti‐inflammatory activities.[ ¹¹ , ¹² , ¹³ ] Polyketides can be readily produced in vivo by incorporating key enzymes such as coenzyme A ligases and polyketide synthases (PKSs) in metabolic pathways without direct acyl coenzyme A supplement. Type I and type II PKSs are megaenzyme systems with multiple catalytic domains, which have been engineered to produce novel products by domain exchange and refinement in metabolic engineering,[ ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ ] and the type III PKSs are the simplest ones as homodimeric proteins of ∼40 kDa, with many type III PKSs sharing >50 % identity in both gene and protein sequences.[ ¹⁸ , ¹⁹ ] Although type III PKSs have been widely applied in whole cell transformation, triketide products, namely styrylpyrones (Figure 1A), are usually generated as derailed side products,[ ¹⁹ , ²⁰ ] which hampers the chemoselective production of target polyketides and complicates the separation processes. Moreover, PKSs suffer from inherently slow kinetics.[ ²¹ , ²² ] Therefore, to unlock the potential of PKSs, it is important to understand the structural characteristics of different type III PKSs and achieve increased chemoselectivity and catalytic efficiency for the target polyketides generation.

Figure 1.

TAL conversion pathway to value‐added products and 2PS‐catalyzed TAL generation. (A) TAL C6‐functionalized natural products. (B) Crystal structure of homodimer 2PS bound with intermediate acetoacetyl‐CoA. (C) TAL conversion pathway. Molecules which are used as food preservatives, additives, and fragrances are shown in blue, and molecules which are used as bifunctional intermediates and building blocks are shown in red. (D) 2PS‐catalyzed TAL generation pathway and mechanism.

One of the simplest triketide products is triacetic acid lactone (6‐methyl‐4‐hydroxy‐2‐pyrone, TAL) that is generated by 2‐pyrone synthase from Gerbera hybrida (Figure 1B). TAL can be produced via microbial fermentation of plant cell wall hydrolysates and further converted to high value‐added products, such as additives, fragrances, and pharmaceuticals, which fulfills the long‐term demand for carbon‐based products.[ ²³ , ²⁴ ] To obtain such products, however, TAL has to be derivatized through extra synthetic steps that include protection and deprotection under high pressure and long reaction time (Figure 1C).[ ²⁴ , ²⁵ ] In contrast, C6‐functionalized TAL‐derived pyrone products widely exist in nature such as the styrylpyrone moiety (Figure 1A). For instance, Katsumadain A, B and C, which were isolated from Alpinia katsumadain, are Chinese herbal drugs used as an anti‐emetic and stomachic agent, and kavalactones from Kava (Piper methysticum), which is an ethnomedicinal shrub with well‐established anxiolytic and analgesic properties.[ ¹¹ , ²⁶ ] Although styrylpyrones have shown prominent pharmaceutical properties, such products are mainly obtained by extraction from plants. ^[27] Recently, Weng and co‐workers revealed the first biosynthetic pathway of psychoactive kavalactones in kava and discovered styrylpyrone synthases (SPSs) involved in the process that exclusively produce triketide styrylpyrones as main products. Taking advantage of the discovery, about 2 mg/L pyrone product was produced in E. coli and yeast from p‐coumaric acid, ^[11] but the potentials of heterologous production for styrylpyrones in high yields have not been fully explored, especially with engineered PKSs. Therefore, to overcome the limitations mentioned above and to advance both protein and metabolic engineering for high‐value products generation from renewable sources,[ ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ ] we herein report engineering of 2PS as a highly selective and efficient enzymatic system for the production of C6‐functionalized TAL‐derived pyrone products from acyl‐CoAs, as well as the construction of an E. coli whole‐cell transformation system that allows the use of the engineered enzymes to directly generate pyrones in vivo by feeding the corresponding carboxylic acids. We also apply computational modeling to gain insights into how engineered 2PS works to achieve high yields and chemoselectivity for different pyrone products.

Results and Discussion

Computational Design through Structural Homology Modeling

2PS catalyzes the formation of TAL from an acetyl‐CoA and two molecules of malonyl‐CoAs through the mechanism shown in Figure 1D. The nucleophilic attack from the thiolate of Cys169 to acetyl‐CoA triggers the decarboxylation of malonyl‐CoA to generate an enolate species, which is followed by Claisen condensation to produce a diketide intermediate, acetoacetyl‐CoA. One more round of such a process using the acetoacetyl‐CoA intermediate as the substrate would produce a triketide intermediate, which cyclizes to produce TAL as the final product. We commenced our study by searching for key structural features of several PKSs that produce aromatic polyketides and engineering 2PS by homology modeling. Besides SPS, chalcone synthase (CHS) ^[34] and stilbene synthase (STS) ^[35] accept aromatic primer CoAs and generate tetraketides. For all four type III PKSs, two phenylalanine residues (Phe220 and Phe270 in 2PS) act as gatekeepers that distinguish substrate specificity, and they share an identical catalytic triad Cys‐His‐Asn (Figure 2A). ^[13] To generate C6‐functionalized TAL‐derivatized aromatic pyrones including styrylpyrones, a larger active site capable of accommodating aromatic primer CoAs rather than acetyl‐CoA is necessary. However, too large of a pocket would lead to potential side reactions, such as the generation of tetraketide intermediate, which would further lead to various side products. ^[35] Therefore, proper pocket size is important to achieve high reactivity while keeping good selectivity. To engineer 2PS to produce aromatic styrylpyrones, we aligned its structure with those of other type III PKSs that can generate similar products, aiming to identify the key residues in the active sites responsible for the functions. As shown in Table 1 and Supplementary Figure S1, SPS, CHS, and STS share more than 65 % sequence identity with 2PS. Moreover, all four enzymes share high‐level structural similarities with the pairwise C_α root‐mean‐square deviation (RMSD) relative to 2PS below 1.0 Å. Although several key residues that can differentiate substrate and influence the chemoselectivity have been identified among type III PKSs,[ ¹⁸ , ¹⁹ , ³⁴ ] residues exclusively responsible for styrylpyrone production in SPS have not been reported yet from the recently discovered SPS.

Figure 2.

2PS active site and homology modeling. (A) 2PS key active site residues (PDB: 1EE0). (B) Overlay of naringenin (CHS product, in magenta) and resveratrol (STS product, in green). (C) Structure of resveratrol and naringenin. (D) Key residues comparisons for homology modeling.

Table 1.

Similarity comparison among PKSs.

Protein	Original Product	Protein Sequence Identity	Cα RMSD [Å]
2PS (PDB: 1EE0)	TAL	100 %	0
SPS (PDB: 6OP5)	Styrylpyrone	68 %	0.932
CHS (PDB: 1CGK)	Naringenin	70 %	0.637
STS (PDB: 1U0W)	Resveratrol	66 %	0.658

Despite the high similarities in protein sequence and structure among the four type‐III PKSs, differences in the active site residues can be identified by carefully comparing those structures. For example, by overlaying the structure of 2PS with those of product‐bound STS and CHS (Figure 2B), we found that two bulky and hydrophobic residues, Leu261 and Ile343 in 2PS, may have strong steric clashes with resveratrol (product of STS) and naringenin (product of CHS) (Figure 2B and 2C). At these two locations, two smaller and conserved residues (Gly261, Ser343) are present instead in STS, CHS and SPS, which allow the aromatic primer CoAs to bind (Figure 2D). Therefore, we redesigned 2PS by introducing L261G/A/V and I343S mutations to make the two locations smaller, like those in STS, CHS and SPS. In addition, since Leu202 next to the above two residues is known to control the substrate specificity,[ ¹⁹ , ³⁶ ] we replaced it with a smaller conserved threonine in CHS and STS through L202T mutation. To further reduce the steric conflict, we replaced Ile201 that is next to Leu202, with a smaller valine conserved in STS, CHS and SPS. Based on these designs, we have made a variant called 3AP (3‐triketide, Aromatic Primers) that contains I201V, L202T and I343S mutations. In addition, we have made further variants to study the steric influences where L261 is mutated to either alanine, glycine or valine and they are called 3AP‐L261A‐2PS, 3AP‐L261G‐2PS and 3AP‐L261V‐2PS, respectively.

Condition Optimization and Kinetics

All the designed variants were constructed, expressed and purified in E. coli (Supplementary Figure S2). We first tested the 3AP‐L261A‐2PS reactivity and compared it with wild type 2PS (WT‐2PS) and WT‐SPS under the conditions shown in Table 2. Interestingly, 3AP‐L261A‐2PS generated styrylpyrone (1) in 60 % yield (Table 2, entry 1) while WT‐2PS produced only 7 % yield (Table 2, entry 2) and WT‐SPS displayed 23 % yield. (Table 2, entry 3). This product (1) is an important precursor that is naturally generated by SPS and can be further converted to other natural products such as 5,6‐dehydrokavain by 4‐OH methylation chemically or enzymatically.[ ¹¹ , ³⁷ ] In the absence of any enzyme, no product was detected (Table 2, entry 4), indicating that the enzymes are responsible for the observed product. We further screened the reaction condition to improve the yield of the transformation. When the malonyl CoA concentration was changed and the ratio between the starter CoA and extender CoA was screened (Table 2, entry 5–6), a ratio of 1 : 3 still achieved the highest yield, while increasing malonyl CoA concentration did not improve the yield further. The yield increased from 60 % at pH=7 (Table 2, entry 1) to 67 % at pH=8 (Table 2, entry 8), whereas the yield was only 25 % at pH=6 (Table 2, entry 7), indicating that the lower pH of the reaction buffer has a major impact on the reactivity of 3AP‐L261A‐2PS. The effect of reaction temperature was then tested (Table 2, entry 9–10). The yield decreased from 60 % (Table 2, entry 1) at 37 °C to 19 % at room temperature (Table 2, entry 9) but didn't increase when the temperature was increased to 55 °C (58 %, Table 2, entry 10).

Table 2.

Condition optimization of in vitro activity assay.

Entry	Catalyst	pH	Temp [°C]	Malonyl‐CoA [μM]	Yield [%][a]
1	3AP‐L261A‐2PS	7	37	150	60±2.8
2	WT‐2PS	7	37	150	7±0.4
3	WT‐SPS	7	37	150	23±0.6
4	No enzyme	7	37	150	n.d.[b]
5	3AP‐L261A‐2PS	7	37	100	43±2.2
6	3AP‐L261A‐2PS	7	37	200	59±0.4
7	3AP‐L261A‐2PS	6	37	150	25±1.3
8	3AP‐L261A‐2PS	8	37	150	67±0.9
9	3AP‐L261A‐2PS	8	25	150	19±0.3
10	3AP‐L261A‐2PS	8	55	150	58±1.7
11	3AP‐L261G‐2PS	8	37	150	90±1.9
12	3AP‐L261V‐2PS	8	37	150	64±1.0
13	3AP‐Leu261‐2PS	8	37	150	18±0.8
14	3AP‐L261G‐2PS	7	37	150	86±1.7
15	1 mg 3AP‐L261G‐2PS cell lysate	7	37	150	85±0.6

[a] Yield was determined by HPLC compared to a product standard curve (Supplementary Figure S3); three parallel experiments were conducted for each entry. [b] not detected.

After the initial reaction condition optimization, we investigated the reactivity of different 3AP mutants. When the steric hindrance at residue 261 was further decreased from 3AP‐L261A‐2PS to 3AP‐L261G‐2PS, a further increase in the yield to 90 % (Table 2, entry 11) was observed. We further tested the reactivity of 3AP‐L261V‐2PS, and the increased sterics decreased the yield to 64 % (Table 2, entry 12). When we kept the Leu261 in the 3AP mutant (I201V/L202T/I343S), the yield dropped drastically to only 18 % (Table 2, entry 13). Therefore, increasing steric hindrance from Gly, Ala, Val to Leu261 in 3AP‐2PS dropped the overall yield correspondingly. We further tested the performance of 3AP‐L261G‐2PS at neutral pH because a high performance at neutral environment is important for further whole‐cell transformation using microorganisms like E. coli due to its optimal growth environment. ^[38] The yield remained at a similar level (86 %) at pH 7 (Table 2, entry 14). As a result of such homology modeling, we have generated a 3AP‐L261G‐2PS mutant which displayed high reactivity toward product (1) generation with 13‐fold higher than WT‐2PS, 4‐fold higher than WT‐SPS under the optimal reaction conditions.

With the successful design of a highly active mutant, we wonder whether such transformation can be directly achieved in a whole‐cell bioconversion system. To find out, we expressed the 3AP‐L261G‐2PS in BL21(DE3) E. coli, lysed the grown cell pellet, carried out the enzymatic transformation, and quantified the yield using the protocol described in Supporting Information. Product (1) was produced in 85 % yield under optimal conditions (Table 2, entry 15). This whole‐cell transformation provides a convenient and straightforward method to generate styrylpyrone products from acyl‐CoA molecules.

To characterize such a promising system, we further monitored the kinetics of different mutants toward product (1) production and calculated the Michaelis–Menten parameters using different concentrations of cinnamoyl‐CoA. While it took more than 2 hours for both WT‐2PS, WT‐SPS and 3AP‐L261A‐2PS to reach the maximum conversion, 3AP‐L261G‐2PS could achieve the reaction in only 15 minutes with a higher yield of 79 % vs. 1.3 % for WT‐2PS, 5 % for WT‐SPS and 15 % for 3AP‐L261A‐2PS (Supplementary Figure S4A‐S4B). We further compared the k _cat and K _M between 3AP‐L261G‐2PS and WT enzymes toward cinnamoyl CoA (Supplementary Figure S4C‐S4F). The k _cat of 3AP‐L261G‐2PS was over 310‐fold greater than that of WT‐2PS (Table 3) and the overall catalytic efficiency (k _cat/K _M) was enhanced by around 70 folds. In comparison with WT‐SPS which is our mimicking target enzyme to produce styrylpyrone (1) in nature, The k _cat of 3AP‐L261G‐2PS was over 58‐fold greater than that of WT‐SPS (Table 3) and the overall catalytic efficiency (k _cat/K _M) was improved by 30 folds. More importantly, the catalytic efficiency of the engineered 3AP‐L261G‐2PS is even higher than those of many native type III polyketide synthases’ reactivity (3600 s⁻¹ M⁻¹ for WT CHS and 4200 s⁻¹ M⁻¹ for WT STS) that have been frequently applied in metabolic engineering researches.[ ³⁵ , ³⁹ , ⁴⁰ ] This provided a highly promising platform for higher level metabolic engineering where 2PS mutant could potentially produce styrylpyrone products naturally generated by SPS. Further kinetic experiments have been conducted on other 3AP mutants (Table 3) and the increased sterics at 261 position not only decreased the transformation yield, but also lowered the turnover number and catalytic efficiency (Supplementary Figure S5).

Table 3.

Michaelis–Menten parameters comparison using different concentrations of cinnamoyl‐CoA.

Protein	K M [μM]	k cat [min−1]	k cat/K M [s−1 M−1]
WT‐2PS	1.6±0.5	(3.2±0.3)×10−2	(3.4±0.6)×102
WT‐SPS	3.7±0.5	(17.2±0.6)×10−2	(7.7±0.3)×102
3AP‐L261G‐2PS	7.1±1.1	9.9±0.6	(2.3±0.4)×104
3AP‐L261A‐2PS	4.2±0.8	1.87±0.14	(7.4±1.5)×103
3AP‐L261V‐2PS	2.1±0.5	(1.35±0.1)×10−1	(1.07±0.27)×103
3AP‐L261‐2PS	1.4±0.3	(4.6±0.2)×10−2	(5.68±1.16)×102

Substrate Scope

After designing the highly active variant and identifying optimal conditions, we examined the reaction scope to see if other larger acyl‐CoAs could fit into the pocket to produce TAL‐derived pyrone products (Table 4). When the sterics increased from malonyl‐CoA to methylmalonyl‐CoA, WT‐2PS failed to generate the corresponding methylated styrylpyrone (2), while WT‐SPS generated only 4 % yield. In contrast, 3AP‐L261G‐2PS still produced the methylated product with ∼58 % yield, suggesting that the designed variant can accommodate a bulkier substrate. In addition, the reactivity towards p‐coumaroyl‐CoA as the substrate was compared among these enzymes. WT‐SPS generated the yangonin precursor product, bisnoryangonin 3 in 16 % yield, while the yield for 3AP‐L261G‐2PS and WT‐2PS was 7 % and 5 %, respectively. Notably, an HPLC peak close to triketide was observed in the 3AP‐L261G‐2PS catalyzed p‐coumaroyl‐CoA system, but not in WT‐2PS or WT‐SPS system (Supplementary Figure S6A). We found this peak to be a tetraketide side product using LC–MS/MS (HPLC integration 1 : 3=tetraketide:target triketide product) (Supplementary Figure S6B‐D). Furthermore, we tested the reactivity of 3AP‐L261G‐2PS using CoAs containing middle length carbon chains such as C₆ hexanoyl‐CoA, C₈ octanoyl‐CoA and C₁₂ lauroyl‐CoA. Such CoAs could be enzymatically synthesized from middle chain fatty acids, and the acids could be extracted from biomass directly. ^[41] The generated 6‐alkylpyrone products are important natural product precursors such as pseudopyronines (Figure 1A) ^[42] and can be used as a potent and selective G‐protein‐coupled receptor agonists. ^[43] Interestingly, all three enzymes can convert hexanoyl‐CoA to product 4 in >94 % yield. For octanoyl‐CoA, 3AP‐L261G‐2PS achieved the highest yield in 88 % with good chemoselectivity (<5% relative HPLC area for tetraketide side products, Supplementary Figure S7), while WT‐2PS and WT‐SPS showed yields of 80 % and 54 %, respectively. However, when the chain length was extended to lauroyl‐CoA, no product was detected for all three enzymes, suggesting that lauroyl‐CoA may be too long to fit into the binding pocket correctly. It is noteworthy that such 6‐alkylpyrone products would be even more challenging to be converted from TAL by post‐extraction synthetic derivatizations to elongate the C6 substituent chain. Finally, we compared the reactivity towards TAL production. WT‐2PS gave full conversion (100 % yield) in 30 minutes, while WT‐SPS achieved 16 % yield and 3AP‐L261G‐2PS generated 70 % yield with the fastest initial rate of 3.3 μM/min in the first 3 minutes (Supplementary Figure S8).

Table 4.

Substrate scope using various CoAs by in vitro enzymatic assays.

Protein	(2) Yield [%][a]	(3) Yield [%][b]	(4) Yield [%][b]	(5) Yield [%][a]	TAL Yield [%][b]
WT‐2PS	n.d.[c]	5±0.5	100±1.2	80±1.1	100±1.2
WT‐SPS	4±0.4	16±2.7	94±1.6	54±3.0	16±2.7
3AP‐L261G‐2PS	58±1.2	7±0.4	100±0.7	88±1.1	70±1.5

[a] Yield was estimated using styrylpyrone (1) and alkylpyone (4) as the calibration‐curve standard for (2) and (5) correspondingly. [b] Yield was determined by HPLC compared to product standard curve (Supplementary Figure S3, S9–S10). Three parallel experiments were conducted for each reaction. [c] not detected by LC–MS.

Heterologous Production of Styrylpyrones in E. coli by Metabolic Engineering

To establish more practical application platforms to produce these pyrone products from more available and cheaper carboxylic acids than CoAs, we explore supplementing the E. coli metabolic network with coenzyme A synthesis pathway to convert carboxylic acids into their corresponding CoAs in situ, which can then be further coupled with the engineered 2PS to produce novel polyketides. We successfully generated an E. coli Rosetta2(DE3) strain that contained the engineered 3AP‐L261G‐2PS gene in pET‐16b vector, as well as malonyl CoA synthetase gene (MCS) ^[44] and 4‐coumarate ligase (4CL) gene ^[45] in a pRSFDuet‐1 vector (Figure 3A). MCS can convert malonic acid, ATP and coenzyme A to malonyl‐CoA, while 4CL catalyzes the formation of aromatic‐CoAs from cinnamic acid and phenylpropanoic acid with different substitutions, which would expand the substrate scope of the engineered 2PS.[ ⁴⁶ , ⁴⁷ ] Malonic acid is listed as one of the top 30 chemicals that are produced from biomass and aromatic acids can be generated from biomass‐derived carbon sources.[ ²³ , ⁴⁸ ] Such whole‐cell transformation not only offers a much more simplified and straightforward synthetic pathway of pyrone production from carboxylic acids, but also can be particularly useful to upgrade biomass‐derived products to enhance bioeconomy.

Figure 3.

E. coli strain construction for whole‐cell transformation. (A) plasmid information. (B) whole‐cell transformation titer. Products are quantified in product standard curve (Supplementary Figure S3, S9 and S11–S12).

We first began with control experiments to test the engineered 2PS reactivities in vivo and the effect of intracellular thiolases in 2 mL small culture (Supporting Information Table S1), since it has been shown that some polyketoacyl‐CoA tholases in E. coli can produce polyketides including styrylpyrones without PKSs.[ ²¹ , ⁴⁹ ] In this system, however, E. coli containing only the 4CL and MCS genes failed to generate styrylpyrone products that can be detected by HPLC from cinnamic acid or p‐coumaric acid. When WT‐2PS was expressed together with MCS and 4CL, it produced trace amount (41.1 μg/L) of product (1) but no detectable product (3). Only when 3AP‐L261G‐2PS gene was expressed together with 4CL and MCS, it showed significant amount of product generation with 3.2 mg/L titer for product (1) and 17.3 mg/L titer for product (3). Although the yield for in vitro enzymatic experiments of 3AP‐L261G‐2PS showed higher reactivity for cinnamoyl‐CoA than p‐coumaroyl‐CoA, the whole‐cell transformation titer did not follow this trend. This might result from the 4CL innate substrate preference and higher reactivity toward its native substrate p‐coumarate over cinnamate. ^[50]

We further tested 1‐liter large‐scale culture whole‐cell biotransformation (Figure 3B) by feeding E. coli with 1 mM p‐coumaric acid and 3 mM malonic acid. Product (3) was observed from HPLC with the calculated titer of 8.9 mg/L. As E. coli is known to produce a small amount of cellular malonyl‐CoA, we hypothesized that 3AP‐L261G‐2PS could use such intracellular malonyl‐CoA directly in E. coli without feeding it with malonic acid. ^[51] The HPLC titer remained at 8.9 mg/L (4.1 mg/L isolated yield) with feeding only 1 mM p‐coumaric acid without malonic acid. Such phenomenon has been observed where MCS has been applied to increase intracellular malonyl‐CoA supply for TAL production. Although this strategy did increase the malonyl‐CoA supply, TAL production titer didn't improve, probably due to higher ATP requirements for MCS reactivity. ^[21] Previously native SPS with kava 4CL was applied to heterologously produce bisnoryangonin in E. coli using 1 mM p‐coumaric acid and in S. cerevisiae using 2 mM p‐coumaric acid, resulting in HPLC yields of 0.9–3.7 mg/L and 2–2.2 mg/L, respectively. ^[11] In comparison with these reported microorganism systems, our system showed at least 4‐fold increase in product yield in E. coli. When cinnamic acid was supplied, the product (1) was observed by HPLC, with calculated titer of 3.6 mg/L (2.0 mg/L isolated yield). In addition, phenylpropanoic acid and 3‐(4‐hydroxyphenyl)propanoic acid were tested to produce dihydro‐kavalactone precursors. The titer calculated by HPLC turned out to be 4.0 mg/L (2.0 mg/L isolated yield) and 17 mg/L (8.9 mg/L isolated yield), respectively. Interestingly, only <5 % (by HPLC area integration) tetraketide products could be detected from 3‐(4‐hydroxyphenyl)propanoic acid system by LC–MS/MS (Supplementary Figure S13), which indicates that both the rigidity and hydroxyl group are important for tetraketide production from p‐coumaroyl‐CoA and p‐coumaric acid.

Besides aromatic acids, the middle‐chain fatty acid such as hexanoic acid was also tested in our system. Recently, PKS systems have been studied using hexanoyl‐CoA and malonyl‐CoA because the hexanoyl moiety may be produced by fatty acid synthase (FAS), which might provide novel metabolic pathways to generate novel products by combining FAS and PKS reactivities.[ ⁵² , ⁵³ ] Our engineered E. coli system produced the triketide lactone product (4) from hexanoyl‐CoA and malonyl‐CoA with 2.0 mg/L titer.

We also measured the cell weight used for 1‐liter biotransformation and around 4‐gram pellet in wet weight (1.25 gram in dry weight) was harvested and utilized after overnight cell growth. For the whole‐cell biocatalysis, the amount for each of the products per gram wet cell weight is shown in Figure 3B. Different whole‐cell transformation conditions were surveyed to enhance both substrate utilization and product titer (Supporting Information Table S2). Although the titer in 2 mL small cultures of cinnamate can be further increased to around 10 mg/L by providing 5–10 mM cinnamate and medium pH adjustment to optimal neutral pH, the substrate conversion toward the target product (0.47 %–0.93 %) was lower than when 1 mM cinnamate was used (1.68 %).

Structural Explanations

After demonstrating the reactivity both in test tubes and in E. coli, we tried to understand the influences of different mutations we have designed into 2PS by control experiments. When one out of the four mutations in 3AP‐L261G‐2PS (I201V/L202T/L261G/I343S) was restored to the original residue in WT‐2PS, the yield all decreased to different extent (Table 5). For example, when malonyl‐CoA was used as an extender for cinnamoyl‐CoA conversion, restoring Gly261 back to Leu261 resulted in significant decrease in the reactivity, from 86 % to only 18 % yield. When a larger extender such as methylmalonyl‐CoA was used, not only Leu261 (from 58 % to 6 %) but also Ile343 (from 58 % to 15 %) showed a major decrease in the product (2) production. These reactivity differences confirmed our initial designs to remove steric hindrance, especially the steric clashes at the Leu261 and Ile343 positions.

Table 5.

Control experiments on different mutants.

Protein	Malonyl‐ CoA [%][a]	Methylmalonyl‐ CoA [%][b]
WT‐2PS	7±0.4	n.d.[c]
3AP‐L261G‐2PS	86±1.7	58±1.2
V201I/L202T/L261G/I343S‐2PS	66±1.0	48±0.9
I201V/T202L/L261G/I343S‐2PS	67±1.6	54±0.2
I201V/L202T/G261L/I343S‐2PS	18±0.8	6±0.4
I201V/L202T/L261G/S343I‐2PS	73±3.1	15±0.9

[a] Yield utilizing malonyl‐CoA as extender unit was determined by HPLC compared to product standard curve. Three parallel experiments were conducted for each reaction using biological replicates. [b] Yield utilizing methylmalonyl‐CoA as extender unit was estimated using styrylpyrone (1) as the calibration‐curve standard. Three parallel experiments were conducted for each reaction using biological replicates. [c] not detected by LC–MS.

Besides steric influences, Ser343 was known to be involved in key hydrogen bonding networks of both CHS and STS to control chemoselectivity of such enzymes. ^[35] For example, an inspection of the SPS crystal structure indicates that Ser339, the residue in SPS that corresponds to Ser343 in 2PS, has hydrogen bonding interactions with Glu192 and catalytic residue Cys164, which correspond to Glu197 and Cys169 respectively in 2PS (Supplementary Figure S14). To better understand the function of Ser343 of 3AP‐L261G‐2PS, a mutant of 2PS (I201V/L202T/L261G/I343G) without hydrogen bonding side chain at residue 343 was tested in both activity assay and kinetic studies. Although the sterics at 343 position was further decreased, this mutant could only produce product (1) in 63 % yield without the overproduction of tetraketide side products under standard activity assays. More importantly, it was shown that the reaction kinetics was slower than 3AP‐L261G‐2PS and the catalytic efficiency (5.1×10³ s⁻¹ M⁻¹) was only 1/4 of 3AP‐L261G‐2PS (Supplementary Figure S15). All these evidence supports the function of Ile343Ser mutation in 3AP‐L261G‐2PS beyond sterics.

Given observation of these results shown above, we used protein‐ligand docking and molecular dynamics simulations (Supplementary Figure S16–S17) to gain further insights. Styrylpyrone product (1) was docked into both the WT‐2PS and the predicted 3AP‐L261G‐2PS structural model (Figure 4A, 4B). Although the predicted model of 3AP‐L261G‐2PS is similar to the structure of WT‐2PS, with the C_α RMSD being only 0.44 Å, the styrylpyrone products such as (1) and (3) failed to be docked into the active site of WT‐2PS due to steric restrictions. In contrast, product (1) successfully entered the enlarged pocket of 3AP‐L261G‐2PS (Figure 4C). This difference in the size of the pocket may account for the higher reactivity of 3AP‐L261G‐2PS toward styrylpyrone production than WT‐2PS.

Figure 4.

Protein‐ligand docking and molecular dynamics simulations. (A) Overlay between overall structures of WT‐2PS (in blue) and predicted and optimized 3AP‐L261G‐2PS structural model (in brown yellow). (B) Styrylpyrone product (1, in red sphere) was docked into 3AP‐L261G‐2PS structure. (C) Interactions between (1) and active site residues of 3AP‐L261G‐2PS. (D) Interactions between (3) and active site residues of 3AP‐L261G‐2PS. (E) Interactions between (5) and active site residues of 3AP‐L261G‐2PS. (F) Interactions between product from C₁₂ aliphatic CoA and active site residues of 3AP‐L261G‐2PS.

We also performed docking and molecular dynamics simulations with different substrates. Interestingly, the yield and selectivity of p‐coumaroyl‐CoA were low, even if its structure is similar to that of cinnamoyl‐CoA. From the docking analysis (Figure 4D), we could clearly see a strong hydrogen bonding interaction between Glu197 and 4‐hydroxyl substituent on the aromatic ring of p‐coumaroyl‐CoA and such H‐bond persists along the simulation trajectory after it has been formed (Supplementary Figure S16B). Such a strong interaction may bury p‐coumaroyl‐CoA deeper into the pocket, move it further away from the catalytic triad, and thus result in lower yield and selectivity than when cinnamoyl‐CoA is used. To test this hypothesis, we generated a new mutant called 3AP‐L261G‐E197D‐2PS to weaken the hydrogen bonding interaction between Glu197 and p‐coumaroyl‐CoA by shortening the side chain of Glu197. Although the in vitro reactivity was not further enhanced, the selectivity was improved to generate bisnoryangonin 3 as the major product (HPLC integration 1 : 12=tetraketide:target triketide product) (Supplementary Figure S18), which showed the critical role of Glu197 for the chemoselectivity of p‐coumaroyl‐CoA.

To explain the reactivity among different lengths of the aliphatic carbon chain CoAs, the products of C₈ octanoyl‐CoA and C₁₂ lauroyl‐CoA were docked into 3AP‐L261G‐2PS (Figure 4E, F). Products from octanoyl‐CoA could be docked in the correct orientation as pyrone ring interacts with the catalytic Cys‐His‐Asn triad. In contrast, the C₁₂ carbon chain of lauroyl‐CoA was too long to be placed into the active site with the correct orientation, which accounts for the inactivity of 3AP‐L261G‐2PS toward lauroyl‐CoA, but high activity toward octanoyl‐CoA.

Conclusion

In summary, we have successfully engineered 2PS enzyme to produce various C6‐functionalized TAL‐derivatized pyrone products under mild aqueous conditions with high chemoselectivity and reaction yield. We accomplished the goal by removing steric hindrance of Ile201, Leu202, Leu261 and Ile343 to smaller residues through I201V, L202T, L261G and I343S mutations, and by introducing potential hydrogen bonding interactions with I343S mutation. Computational modeling of the engineered mutants has revealed that properly enlarged active sites resulted in high yield for various substrates compared to WT‐2PS and good chemoselectivity for triketide generation. It also showed the importance of Glu197 in controlling the chemoselectivity of p‐coumaroyl‐CoA. These gained insights make it possible to engineer one 2PS scaffold toward many other tetraketides production. Furthermore, we have developed a whole‐cell transformation process utilizing engineered enzyme with novel reactivity from modified substrate scopes to convert carboxylic acids directly into the final pyrone products in a complex metabolic network, resulting in comparatively high yields empowered by protein engineering and synthetic biology. Such system holds significant promise to generate more complex natural polyketide products from only one enzyme scaffold in metabolic engineering.

Conflict of interest

Prof. Y. Lu, Y. Zhou and E. N. Mirts have filed a patent application on E. coli metabolic engineering of pyrone production. The authors declare no other competing financial interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Zhou Y., Mirts E. N., Yook S., Waugh M., Martini R., Jin Y.-S., Lu Y., Angew. Chem. Int. Ed. 2023, 62, e202212440; Angew. Chem. 2023, 135, e202212440.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.