De novo biosynthesis of taxifolin in yeast peroxisomes

Abstract

Background

Yeast peroxisomes have been engineered as ideal synthetic compartments to enhance the heterologous biosynthesis of natural products, particularly terpenoids and fatty acid derivatives. This advantage is primarily attributed to the rich acetyl-CoA pool generated from the spatially specific fatty acid β-oxidation within peroxisomes. However, their potential for flavonoid biosynthesis has been largely underexplored, primarily due to limited knowledge regarding precursor transport, cofactor availability, and the redox environment in peroxisomes.

Results

In this study, we successfully compartmentalized the biosynthesis of taxifolin, a dihydroflavonol, in Saccharomyces cerevisiae peroxisomes. The result indicated that flavonoid biosynthesis in peroxisome offers a more efficient approach compared to its synthesis in the cytosol. This study managed to expand the application scope of peroxisome compartmentalization to flavonoid biosynthesis. By reinforcing the rate-limiting steps, optimizing cofactor supply and activation of fatty acids, we accomplished the de novo synthesis of taxifolin in peroxisomes for the first time, attaining a titer of 120.3 ± 2.4 mg/L in shake-flask fermentation using a minimal medium.

Conclusion

These findings highlight the feasibility of peroxisomal compartmentalization for flavonoid biosynthesis, providing new insights and a framework for the biosynthesis of other high-value flavonoids using yeast peroxisomes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12934-025-02773-2.

Introduction

Taxifolin (dihydroquercetin) is a plant-derived secondary metabolite classified as a dihydroflavonol with potent antioxidant properties, making it a valuable bioactive compound in the food, pharmaceutical, and agricultural industries [1, 2]. Additionally, it has been identified as a promising therapeutic agent for managing metabolic syndrome and age-related disorders [3, 4]. Commercial productions of taxifolin predominantly rely on extraction from the roots of deciduous conifers [5], which typically yield 0.8–1.2% w/w. However, this approach faces significant challenges, including complex extraction processes, raw material scarcity, and seasonal variations [6, 7]. Conventional extraction methods are further hindered by lignocellulosic matrix interference and solvent-intensive processing, leading to suboptimal productivity [8]. To address these limitations, microbial biosynthesis has emerged as a sustainable alternative, successfully facilitating microbial production of taxifolin [9, 10]. Particularly, Saccharomyces cerevisiae (brewer’s yeast), which is classified as a Generally Recognized as Safe (GRAS) microorganism, provides substantial advantages for the scalable and safe production of taxifolin in both the food and pharmaceutical industries. Currently, the heterologous synthesis of flavonoids is predominantly designed to be localized in yeast cytosol. As a eukaryotic organism containing multiple organelles, strategies to enhance flavonoid production by fully utilizing organellar metabolic resources in S. cerevisiae remain to be investigated.

Taxifolin pathway consists of three sections: biosynthesis of the precursor molecules, p-coumaroyl CoA and malonyl-CoA; conversion of p-coumaroyl CoA and malonyl-CoA into (2S)-naringenin; and the hydroxylation of (2S)-naringenin to produce taxifolin. Last section is often regarded as the rate-limiting step in taxifolin production, subsequently, previous studies have developed several metabolic engineering strategies to improve the efficiency of this step. Notably, three enzymes are involved in the last section: α-ketoglutarate-dependent flavanone 3-hydroxylase (F3h), NADPH-dependent flavanone 3’-hydroxylase (F3’h), and cytochrome P450 reductase (Cpr). The catalytic activity of a combination of Citrus sinensis CsF3h, Silybum marianum SmF3’h, and S. cerevisiae Cpr has demonstrated superior efficiency, achieving a 38% mass conversion efficiency in the transformation of naringenin to taxifolin in S. cerevisiae [11]. Additionally, the production of taxifolin has been significantly improved through the combined use of semi-rationally designed mutant MdF3h^P221A — which possesses higher thermostability — along with other pathway genes in yeast. This approach achieved a production of 637.3 ± 20.4 mg/L taxifolin by feeding 800 mg/L (2S)-naringenin [8]. In another study, the cofactors NADPH and α-ketoglutarate were enhanced by overexpressing NADP⁺-dependent isocitrate dehydrogenase gene (IDH1), achieving a 14.1% increase in taxifolin production [9]. In addition, both p-coumaroyl CoA and malonyl-CoA are common precursors in flavonoid synthesis, and their availability may play a critical role in taxifolin production efficiency. Indeed, supplementing p-coumaric acid significantly enhances the synthesis of flavonoid [12]. A study has extensively explored the synthesis of p-coumaric acid, with the highest reported yield reaching 12.5 g/L in a 1-liter fermenter [13]. In our previous study, we also successfully constructed an efficient yeast cell factory for p-coumaric acid biosynthesis, which can support the precursor supply for flavonoid biosynthesis [14].

The expression of the citrate shuttle and ATP-citrate lyase (ACL) facilitates the transportation of mitochondrial citrate to the cytosol, where it is cleaved to generate acetyl-CoA, which serves as a precursor for malonyl-CoA synthesis. This metabolic enhancement aiming to increase the supply of malonyl-CoA in the cytosol significantly improves the synthesis of flavonoids, including taxifolin [9]. In addition to mitochondria, peroxisomes serve as important sources of acetyl-CoA in S. cerevisiae due to their unique fatty acid β-oxidation pathway. Notably, peroxisomes facilitate the conversion of fatty acyl-CoA through β-oxidation to maintain a high acetyl-CoA pool [15]. Therefore, as a precursor for terpenoids and fatty acid derivatives, an adequate supply of acetyl-CoA renders yeast peroxisomes as ideal subcellular compartments for the biosynthesis of these compounds [15, 16]. However, whether the peroxisomal acetyl-CoA pool is beneficial for flavonoid biosynthesis remains unknown. We think the potential use of peroxisomal acetyl-CoA in the synthesis of flavonoid offers several benefits. In contrast to mitochondria, which play significant roles in the tricarboxylic acid (TCA) cycle and oxidative phosphorylation, peroxisomes are non-essential for cell growth under glucose culture conditions. However, their abundance and size can be dynamically regulated to enhance metabolites production [17]. Peroxisomes facilitate high-density compartmentalization of enzymes, mitigating the interference with native metabolic pathways while protecting the cell from toxic enzymes or potentially harmful metabolic intermediates [18]. Additionally, peroxisomes establish a unique microenvironment characterized by a confined volume (usually 0.1–0.2 mm in diameter) and specialized redox homeostasis [19, 20]. Therefore, this study aims to reconstruct the de novo biosynthesis pathway of taxifolin within peroxisomes, thereby demonstrating the feasibility and potential of peroxisomal compartmentalization as a strategy for flavonoid production.

To achieve this, we developed a four-module peroxisome metabolic engineering strategy for the de novo biosynthesis of taxifolin in yeast (Fig. 1 and Figure S1). First, we demonstrated that peroxisomes offer superior benefits compared to the cytosol in flavonoid synthesis. Subsequently, we introduced the taxifolin biosynthetic pathway into peroxisome for its de novo production. Moreover, we attempt to regulate peroxisome biogenesis and fatty acid β-oxidation to promote taxifolin accumulation. Additionally, we enhanced the supply of cofactors involved in the metabolic pathway, leading to improved taxifolin production. Ultimately, utilizing the minimal medium in shake flask fermentation, we achieved a de novo taxifolin titer of 120.3 ± 2.4 mg/L. The metabolic engineering strategies described in this study provide a novel approach for leveraging peroxisomal compartmentalization in yeast to enhance de novo synthesis of taxifolin, offering new insights and a framework for the biosynthesis of other high-value flavonoids using yeast peroxisomes.

Fig. 1.

Modular engineering design for the compartmentalized synthesis of taxifolin in the peroxisomes of S. cerevisiae. E4P, erythrose-4-phosphate; PEP, phosphoenolpyruvate; DAHP, 3-deoxy-d-arabino-heptulosonate-7-phosphate; PPA, prephenate; PPY, phenylpyruvate; HPP, para-hydroxy-phenylpyruvate; α-KG, α-ketoglutarate; SUC, succinate; CIT, citrate; ISO, isocitrate; GAC, glyoxylate cycle. gpp1Δ, indicates deletion of GPP1 (encoding glyceraldehyde 3-phosphate phosphatase); CkPTA, encoding Clostridium kluyveri phosphotransacetylase; TAL, encoding transaldolase; TKL, encoding transketolase; LmXFPK, encoding Leuconostoc mesenteroides phosphoketolase; ARO1, encoding shikimate dehydrogenase; ARO2, encoding chorismate synthase; ARO3, encoding 3-deoxy-d-arabino-heptulosonate-7-phosphate synthase; ARO4*, encoding the mutant of (DAHP) synthase (Aro4^K229L); EcAORL, encoding Escherichia coli shikimate kinase II; ARO7*, encoding the mutant of chorismate mutase (Aro7^G141S); PHA2, encoding prephenate dehydratase 2; MtPDH1, encoding Medicago truncatula prephenate dehydrogenase; SbPAL*, encoding the mutant of Sorghum bicolor phenylalanine ammonia lyase (SbPal^H123F); AtCPR1, encoding Arabidopsis thaliana cytochrome P450 reductase; CYPc, encoding Cyp complex consisting of Populus trichocarpa cinnamic acid hydroxylase 1 (PtrC4h1), cinnamic acid hydroxylase 2 (PtrC4h2), and coumarate-3-hydroxylase (PtrC3h); FjTAL, encoding Flavobacterium johnsoniae tyrosine ammonia lyase; pdc5Δ, indicates deletion of PDC5 (encoding pyruvate decarboxylase); aro10Δ, indicates deletion of ARO10, (encoding phenylpyruvate decarboxylase); PaHPAB, encoding Pseudomonas aeruginosa reduced flavin adenine dinucleotide (FADH₂)-dependent 4-hydroxyphenylacetate 3-monooxygenase; SeHPAC, encoding Salmonella enterica NADH-dependent flavin oxidoreductase; ACC1*, encoding the mutant of acetyl-CoA carboxylase 1 (Acc1^{S569A, S1157A}); Ptr4CL5, encoding Populus trichocarpa 4-coumarate coenzyme A ligase; SjCHS1, encoding Sophora japonica chalcone synthase; MsCHI, encoding Medicago sativa chalcone isomerase; AtF3H, encoding Arabidopsis thaliana flavanone 3-hydroxylase; SmF3’H, encoding Silybum marianum flavanone 3’-hydroxylase; SmCPR*, encoding the mutant of Silybum marianum cytochrome P450 reductase (SmCpr^I453V); ACO2, encoding aconitase 2; IDP3, encoding peroxisomal NADP-dependent isocitrate dehydrogenase; TRY1, encoding prephenate dehydrogenase; ARO8, encoding aminotransferase I; PPP, pentose phosphate pathway

Materials and methods

Chemicals and reagents

The standards naringenin, eriodictyol, dihydrokaempferol, and taxifolin were obtained from MedChemExpress. Additionally, methanol was obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

Construction of strains and plasmids

All strains used in this study are listed in Table S1. All primers and the codon-optimized genes used in this study were synthesized by Sangon Biotech (Shanghai, China); the primers are listed in Table S3. The gene sequences used in this study are listed in Table S3.

Genetic engineering

All gene overexpression cassettes, comprising a promoter, a coding gene, a terminator, and upstream and downstream genomic flanking sequences to facilitate homologous recombination, were constructed using overlapping extension PCR [21]. The native promoters, coding genes, and terminators were cloned from the genomic DNA of E. coli strain DH5α and S. cerevisiae strain CEN.PK113-11C. The modified strain CEN.PK113-11C* was obtained by introducing the expressions of the Cas9 protein and performing a complete genomic knocking out of GAL80 in the wild-type CEN.PK113-11C strain [14]. The CRISPR-Cas9 genome editing software CHOPCHOP (v3) was used to identify two distinct guide RNAs (gRNAs) for each selected genomic locus [22] (Table S4). All the genes expressed in this study were all driven by GAL series promoters, unless specified, to achieve high expression levels under low glucose concentrations, thereby avoiding competition for carbon sources with cell growth processes [14]. All overexpression cassettes (Table S5) were integrated into their designated chromosomal loci to ensure stable and high-level expression of heterologous genes [14]. The purified overexpression cassettes and the gRNA plasmid were co-transformed into the modified strain CEN.PK113-11C* to facilitate targeted cassette integration via the CRISPR-Cas9 genome editing method [23]. For gene deletion, 500 ng of a repair fragment consisting of approximately 500 bp sequences homologous to the upstream and downstream regions of the chromosomal target site was used for the homologous repair of the genome break induced by the Cas9 nuclease cleavage.

Strain cultivation

Yeast cells were normally cultivated in YPD media supplemented with 20 g/L glucose, 20 g/L peptone, and 10 g/L yeast extract. Strains harboring URA3-, HIS3-based plasmids/cassettes, or both were selectively cultured in SD media (6.7 g/L yeast nitrogen base without amino acids and 20 g/L glucose) lacking uracil or histidine, respectively. The URA3 maker plasmid was eliminated by culturing strains in SD agar plates supplemented with uracil (0.02 g/L) and 5-FOA (1 g/L, 5-fluoroorotic acid).

Shake flask fermentations were performed in 20 mL of Delft-D minimal medium containing 2.5 g/L (NH₄)₂SO₄, 14.4 g/L KH₂PO₄, 0.5 g/L MgSO₄·7H₂O, 20 g/L glucose, 2 mL of trace metal, and 1 mL of vitamin solutions. Where necessary, the medium was supplemented with 40 mg/L histidine, 60 mg/L uracil, or both [24]. Cultures were inoculated at an initial OD₆₀₀ of 0.1 in 20 mL of minimal medium and incubated under continuous agitation at 220 rpm at 30 ℃ for 120 h.

Metabolite extraction and quantification

For the extraction of fermentation products, 0.2 ml of small glass beads were transferred into a 2 mL centrifuge tube, followed by the addition of 400 µL of fermentation culture and 400 µL of methanol. The mixture was agitated for 30 min and subsequently centrifuged at 13,000 × g for 5 min. The supernatant was collected using a 1 ml syringe, and the impurities were eliminated via filtering through a 0.22 μm filter membrane. The resulting mixture was then transferred into a sample vial for analysis via high-performance liquid chromatography (HPLC).

All samples were quantified using HPLC. The procedure employed was as follows: each sample was analyzed on an LC-2030 Plus HPLC (Shimadzu) equipped with a Poroshell 120 EC-C18 column (2.7 μm, 3 × 100 mm) (Agilent) connected to a photodiode array detector. The samples were analyzed with a gradient method using two solvents: deionized water with 0.05% formic acid (A) and acetonitrile with 0.05% formic acid (B). The key parameters of analysis are listed in Table S6. Liquid chromatography data analysis was conducted using the LabSolutions LC Workstation v.5 software. UHPLC/MS was used to confirm taxifolin production (Figure S3). The sample was analyzed on a 6530-UHPLC/MS (Agilent) equipped with a Poroshell 120 EC-C18 column (2.7 μm, 3.0 × 150 mm) (Agilent). The samples were analyzed with a gradient method using two solvents: deionized water with 0.1% formic acid (A) and acetonitrile with 0.1% formic acid (B). The key parameters of analysis are the same as above.

Fluorescence microscopy

Strains were cultured in YPD medium overnight and the inoculated at an OD₆₀₀ of 0.1 in Delft-D minimal medium of this study for fermentation. After 36 h, cultures were resuspended in ddH₂O before spotting onto plain glass slides for imaging on a Nikon AX series confocal microscopy with Nikon Spatial Array Confocal (NSPARC) detector. Images were analyzed using NIS-Elements Viewer (5.22 64-bit) software. The detection wavelength of the fluorescence microscope is 499–551 nm for eGfp. The detection wavelength of the fluorescence microscope is 592–721 nm for mCherry.

Results

Feasibility validation of compartmentalized flavonoid synthesis in peroxisomes

Firstly, a fluorescence colocalization assay was performed in strain CEN.PK113-11C* to validate the peroxisome-targeting function of ePts1. The N terminus of the Pex22, a peroxisomal protein, was fused to the C terminus of red fluorescent protein (mCherry, which could be regarded as a peroxisome marker protein; while ePts1 was fused with green fluorescent protein (eGfp) at the C-terminal [25]. Under the confocal microscope, two fluorescence overlapped with each other (Figure S2), demonstrating the peroxisome-targeting function of the signal peptides ePTS1. Malonyl-CoA and p-coumaric acid were the two precursors for taxifolin biosynthesis. Thus, the genes PaHPAB from Pseudomonas aeruginosa and SeHPAC from Salmonella enterica, which are involved in caffeic acid biosynthesis, were totally knocked out in the high caffeic-acid-producing strain RB197 (previously engineered in our studies) to redirect the metabolic flux toward p-coumaric acid synthesis [14] (Fig. 2A). The resulting strain ZZP4103 demonstrated efficient p-coumaric acid biosynthesis, achieving a cytosolic titer of 232.4 mg/L (Figure S4). Notably, the peroxisomal membrane exhibited non-specific permeability for small molecules (MW < 400 Da), thereby possibly facilitating the passive diffusion of cytosolic precursors, including p-coumaric acid, into the peroxisomal lumen [26]. The oxidation of fatty acids in peroxisomes results in the production of acetyl-CoA, which is vital for the glyoxylate cycle. This process enabled the regeneration of HS-CoA, which is essential for p-coumaroyl-CoA biosynthesis and plays a crucial role in the compartmentalized synthesis of flavonoids. Contrastingly, the endogenous acetyl-CoA carboxylase (Acc1) was localized in the cytosol, facilitating the biosynthesis of malonyl-CoA, which are large molecules (> 700 Da) unable to penetrate the peroxisomes. Therefore, sufficient production of malonyl-CoA within the peroxisomes is another crucial requirement for the efficient biosynthesis of taxifolin. Consequently, we targeted Acc1 for peroxisomal expression by fusing the ePts signal peptide to its C-terminus (Acc1-ePts1, abbreviated as Acc1p) to produce malonyl-CoA in peroxisomes of strain WQ3 [26]. Given that the direct detection of malonyl-CoA is challenging, we attempted to directly express genes 4CL and CHS to synthesize flavonoids. Co-expression of genes Ptr4CL5 from Populus trichocarpa and SjCHS1 from Sophora japonica in peroxisomes enabled the production of naringenin at a titer of 38.2 ± 4.5 mg/L in the WQ6 strain through shake-flask cultivation (Fig. 2B). Additionally, we introduced the Acc1 mutant, Acc1^{S569A, S1157A}p, to enhance flavonoid synthesis levels in peroxisomes, since this mutant can avoid being phosphorylated and exhibits high activity under conditions of low glucose concentration [27]. The naringenin titer produced by strain WQ7 was 1.7 times higher than that of strain WQ6, achieving a value of 60.6 ± 1.1 mg/L. These results indicated that the p-coumaric acid synthesized in the cytosol was capable of penetrating the membrane into peroxisomes, interacting with malonyl-CoA synthesized from acetyl-CoA to produce flavonoids. Consequently, our results validate the feasibility of utilizing acetyl-CoA in peroxisomes to synthesize flavonoids.

Fig. 2.

Biosynthesis of flavonoid naringenin in yeast peroxisomes. (A) Schematic illustration of naringenin biosynthesis in peroxisomes. (B) Comparison of distinct ACC1 on naringenin synthesis. (C) Comparison of different subcellular localizations of biosynthetic pathway on naringenin synthesis. Appending “p” to a gene name signifies that the encoded protein is localized to the peroxisome. Two ACC1 genes encoding S. cerevisiae, ScAcc1 and ScAcc1* (ScAcc1^{S569A, S1157A}), one 4CL gene encoding P. trichocarpa Ptr4Cl5, one CHS gene encoding S. japonica SjChs1, and one CHI encoding M. sativa MsChi were used as potential target genes. The other gene abbreviations used are listed in Fig. 1 legends. All shake-flask fermentations were conducted in Delft-D minimal medium supplemented with 40 mg/L histidine and 60 mg/L uracil unless stated otherwise. The fermentation broths were extracted after 120 h of growth for metabolite quantification. All data are presented as the means of three biologically independent samples, with the error bars showing the s.d

The flavonoid biosynthesis capacity in peroxisomes was evaluated by constructing strains WQ64 and WQ65 from strain WQ3 via the expression of Ptr4CL5, SjCHS1, and MsCHI in the cytosol and peroxisomes, respectively. The results revealed that the naringenin titer of strain WQ65 (60.5 ± 4.5 mg/L) was 26% higher compared to that of strain WQ64 (48.0 ± 3.9 mg/L) (Fig. 2C). Thus, flavonoid biosynthesis in peroxisome offers a more effective approach compared to its expression in the cytosol. Subsequently, we speculated that expressing flavonoid biosynthesis pathway in both peroxisome and cytosol is promising to enhance the synthesis efficiency of flavonoids.

Biosynthesis of taxifolin in yeast peroxisomes

Strain WQ7 was further selected for taxifolin synthesis in peroxisomes. Dihydrokaempferol was synthesized by expressing chalcone isomerase (Chi) and F3h from different plant species in peroxisomes of strain WQ7. The results demonstrated that strain WQ39 expressing genes MsCHI from Medicago sativa and AtF3H from Arabidopsis thaliana in peroxisomes achieved a dihydrokaempferol titer of 43.2 ± 1.1 mg/L, which was 1.7 times higher than that of strain WQ58 expressing SmF3H from Silybum marianum and MsCHI from Medicago sativa (Fig. 3A). In plants, enzyme F3’h is a cytochrome P450 monooxygenase (Cyp) that utilizes NADPH as the cofactor and requires co-anchoring to the endoplasmic reticulum (ER) with Cpr to complete electron transport and catalyze taxifolin production. Therefore, the enzyme S. marianum SmF3’h, along with two Arabidopsis thaliana Cpr (AtCpr1, and AtAtr2), were expressed in peroxisomes of strain WQ39 to generate strain WQ43, achieving a taxifolin titer of 19.8 ± 0.9 mg/L (Fig. 3C). Previous research indicates that the highly active S. marianum Cpr mutant SmCpr^I453V is highly suitable for SmF3’h of the same species [28]. Thus, SmF3’h and SmCpr^I453V were expressed in peroxisomes, resulting in a taxifolin titer of 57.6 ± 2.7 mg/L in strain WQ57 (Fig. 3C). Peroxisomes exhibit a high redox potential [29], which may be suitable for the Cyp function and taxifolin production. In addition to substrates and enzyme requirements, the synthesis of taxifolin involves cofactors such as NADPH and α-ketoglutarate, which should be in abundant supply for efficient synthesis of taxifolin within the peroxisomal lumen.

Fig. 3.

De novo biosynthesis of taxifolin in yeast peroxisomes. (A) Schematic illustration of taxifolin biosynthesis in peroxisomes. (B) Comparison of different F3H genes on dihydrokaempferol synthesis. Synthesis (C) and optimization (D) of taxifolin in peroxisomes. Two F3H genes encoding A. thaliana AtF3h and S. marianum SmF3h, one CHI encoding M. sativa MsChi, one F3’H gene encoding S. marianum SmF3’h, three CPR genes encoding A. thaliana AtCpr1 and AtAtr2, S. marianum SmCpr* (SmCpr^I453V), and one CHS gene encoding S. japonica SjChs1 were used as potential gene targets. The remaining gene abbreviations used are listed in Fig. 1 legends. Unless otherwise specified, all subsequent shake-flask fermentations were conducted in Delft-D minimal medium supplemented with 40 mg/L histidine and 60 mg/L uracil. The fermentation broths were extracted after 120 h of growth for metabolite quantification. All data are presented as means of three biologically independent samples, with the error bars showing the s.d

Given that Chs is the rate-limiting enzyme in flavonoid biosynthesis [30], two additional copies of SjCHS1 were introduced into strain WQ57 and expressed in peroxisomes. The titer of taxifolin in the modified strain WQ62 increased by 15%, achieving a value of 66.1 ± 2.4 mg/L (Fig. 3D). The condensation of p-coumaryl-CoA and malonyl-CoA is catalyzed by Chs, while 4-coumarate-CoA ligase (4Cl) is responsible for the formation of p-coumaryl-CoA. Consequently, the peroxisomal overexpression of extra double-copy and single-copy Ptr4CL5 resulted in the production of strain WQ84 and strain WQ85, respectively. However, this did not enhance their production (Figure S5), suggesting that 4Cl is not the rate-limiting step in flavonoid biosynthesis under the current conditions.

Impact of peroxisome biogenesis on taxifolin synthesis

Regulating the biogenesis of peroxisomes potentially enhances the compartmentalized synthesis efficiency of flavonoids. The underlying mechanisms involved in peroxisome biogenesis are highly conserved across the eukaryotic kingdom. The proteins involved in this process, designated as peroxins and encoded by a series of PEX genes, have been extensively studied [31]. The signal peptide recognition receptor of ePts1, PEX5 [15], was overexpressed in background strain WQ62, exhibiting the ability to improve the peroxisome localization efficiency of the pathway enzyme. However, this did not alter the yield of taxifolin (Figure S6). Upon binding PEX5 to the ePts1 of cargo proteins (enzymes), the resulting complex was recruited into the peroxisome via a docking complex composed of the conserved membrane proteins PEX13 and PEX14 [32]. Given that the single overexpression of PEX5 did not increase taxifolin production, we co-expressed PEX5, PEX13, and PEX14, aiming to synergistically enhance the peroxisomal import of heterologous proteins. However, this strategy failed to enhance the synthesis efficiency of taxifolin in strain WQ93, implying that the dominant factor affecting the peroxisome localization efficiency is the signal peptide ePts1 rather than the signal peptide receptor. The PEX11 family, composed of PEX11, PEX25, and PEX27, is typically the primary factor responsible for modulating the proliferation and morphology of peroxisomes [15]. Knockout of PEX11 has been reported to lead to the formation of larger but fewer peroxisomes, which tend to aggregate [15]. This renders cells unable to grow on fatty acids. Contrastingly, overexpression of PEX11 reduces peroxisome volume and increases their number, as well as promotes the extension of the peroxisome membrane [33]. Additionally, the PEX30 family, comprising PEX30, PEX31, and PEX32, regulates the size and proliferation of peroxisomes [34]. The PEX28 and PEX29 act on the upstream of the PEX30 family. Overexpression of PEX28 has been reported to increase the number of peroxisomes and increase the synthetic titer of fatty alcohols by 2.6 times [35]. Therefore, we attempted to enhance production by regulating the size and quantity of peroxisomes by regulating these genes. However, all these designs did not increase the production levels of taxifolin (Figure S6). The reason may be that in the medium with limited glucose as the only carbon source, S. cerevisiae peroxisome is difficult to proliferate a lot, leading to few impacts on the heterologous pathway by regulating the size and quantity of the limited number of peroxisomes.

Cofactor engineering improves taxifolin synthesis in peroxisomes

The biosynthesis of taxifolin involves multiple cofactors, which may be scarce in peroxisomes and insufficient to facilitate efficient synthesis. The F3h, a member of 2-oxoglutarate-dependent dioxygenase (2-Odd) superfamily, requires α-ketoglutarate as the cofactor. In yeast cytosol, prephenate dehydrogenase (Tyr1) converts prephenate to 3-(4-hydroxyphenyl)-pyruvate, while aminotransferase I (Aro8) is responsible for converting an amino group from L-glutamate to 3-(4-hydroxyphenyl)-pyruvate, resulting in tyrosine and α-ketoglutarate (Fig. 4A). The endogenous production of tyrosine provides a sufficient precursor pool for p-coumaric acid biosynthesis in the cytosol. Simultaneously, the molecular weight of the synthesized ketoglutarate permits its translocation into peroxisomes, where it serves as a cofactor in the catalytic activity of F3h. Subsequently, we overexpressed TYR1 and ARO8 in strain WQ62 to increase the carbon flux from the shikimate pathway to tyrosine while enhancing the supply of the cofactor α-ketoglutarate, thereby improving the synthesis level of taxifolin. Consistently, the resulting strain WQ73 produced 98.5 ± 2.6 mg/L taxifolin, achieving a 50.1% production efficiency (Fig. 4B). These results indicate that enhancing the supply of cofactors can increase the synthesis of flavonoids within peroxisomes.

Fig. 4.

Engineering cofactor supply for taxifolin production. (A) Schematic illustration of increasing the supply of α-ketoglutarate and tyrosine by optimizing the shikimate pathway. (B) Optimization of shikimate pathway to enhance taxifolin synthesis. (C) Schematic illustration of increasing the supply of peroxisomal NADPH and α-ketoglutarate. (D) Optimizing the supply of peroxisomal NADPH and α-ketoglutarate to enhance taxifolin synthesis. TRY1, encoding prephenate dehydrogenase; ARO8, encoding aromatic aminotransferase I; IDP2, encoding cytosolic NADP-specific isocitrate dehydrogenase; IDP3, encoding peroxisomal NADP-dependent isocitrate dehydrogenase; ACO2, encoding aconitase 2; CIT2, encoding citrate synthase 2; ACO1, encoding aconitase; ICL1, encoding isocitrate lyase 1; MLS1, encoding malate synthase 1; MDH2, encoding malate dehydrogenase 2. GAC, glyoxylate cycle; TCA, tricarboxylic acid cycle. All data are presented as the means of three biologically independent samples, with the error bars showing the s.d

The F3’h is an NADPH-dependent Cyp enzyme, requiring a substantial supply of NADPH to persistently sustain its high activity. In S. cerevisiae, glucose-6-phosphate dehydrogenase (Zwf1) and NAD-dependent butanediol dehydrogenase (Bdh1) are involved in the regeneration of NADPH [9]. Thus, we separately overexpressed both BDH1 and ZWF1 in strain WQ73 to produce strain WQ88 and WQ89, respectively. However, this approach yielded a decrease in taxifolin production of 87.7 ± 3.3 mg/L in strain WQ88 and 94.6 ± 4.7 mg/L in strain WQ89 (Figure S7). Given that the taxifolin biosynthetic pathway was reconstructed within peroxisomes, enhancing NADPH regeneration in the cytosol may pose challenges in its effective utilization by the compartmentalized pathway. Additionally, this approach could disrupt redox homeostasis by creating an imbalance in the distribution of reducing equivalents between cellular compartments. Therefore, subsequent studies should emphasize improving the regeneration of NADPH in peroxisomes.

Regeneration of NADPH in peroxisomes is dependent on isocitrate/α-ketoglutarate shuttle, which involves isocitrate dehydrogenase (Idp) distributed in the peroxisome and cytosol [36] (Fig. 4C). The IDP2 and IDP3 were co-expressed to promote the recycling of NADPH in peroxisomes, with the resulting strain WQ75 achieving a production of 61.2 ± 3.7 mg/L taxifolin, signifying a 40.0% decrease in yield production (Fig. 4D). We speculated that the strengthening of the isocitrate/α-ketoglutarate shuttle is associated with undesired redox imbalance in peroxisomes, thereby exacerbating the overall efficiency of the cell factory. To alleviate this imbalance, we examined the glyoxylate cycle of S. cerevisiae. A portion of the acetyl-CoA produced by β-oxidation enters the glyoxylate cycle and combines with oxaloacetate to generate citrate in peroxisomes. The citrate shuttle transports citrate to the cytosol, where it is subsequently converted into isocitrate by the enzymatic action of aconitase (Aco) [37]. Therefore, we utilized the citrate from the glyoxylate cycle to compensate for the isocitrate consumed by overexpressing IDP3 (Fig. 4C). The Aco2 was compartmentalized in peroxisomes in strain WQ75 to convert citrate into isocitrate without shuttle and further dually strengthen the supply of α-ketoglutarate and NADPH (Fig. 4D). Consistent with our expectations, while taxifolin production of strain WQ86 unmatched the production efficiency level of strain WQ73 (which lacked cofactor regulation), it showed an improvement compared to strain WQ75, achieving a production of 68.0 ± 5.2 mg/L. To optimize cofactor supply, we retained only the expression of ACO2 in peroxisomes, facilitating an increase in taxifolin production to 114.5 ± 3.6 mg/L (Fig. 4D). Our results demonstrated that the metabolic flux of the glyoxylate cycle was partially separated and replenished into the isocitrate/α-ketoglutarate shuttle through the compartmentalization of Aco2, which improved the regeneration of NADPH and α-ketoglutarate, thereby increasing the titers of taxifolin.

Malonyl-CoA supply improves taxifolin synthesis in peroxisomes

Previous studies systematically optimized modules III and IV by enhancing the expression pathways and ensuring a sufficient supply of cofactor, which has been shown to increase the efficiency of flavonoid biosynthesis in peroxisomes. To further enhance taxifolin production, we re-evaluated the role of module II in malonyl-CoA supply. In the engineered biosynthetic pathway for taxifolin, peroxisomal malonyl-CoA is generated through the carboxylation of acetyl-CoA (Fig. 5A). Peroxisomes conduct β-oxidative metabolism of fatty acids in S. cerevisiae, resulting in a pool of acetyl-CoA. Theoretically, enhancing the supply of acetyl-CoA derived from β-oxidation has the potential to enhance malonyl-CoA synthesis for subsequent flavonoid production. Thus, we intended to enhance fatty acid metabolism in peroxisomes to improve the production of taxifolin. Fatty-acyl CoA oxidase (Fox1) is a key rate-limiting enzyme in β-oxidation [38]. The activated acyl-CoA esters of long chain fatty acids can be transferred into peroxisomes through the activity of a specific transporter complex, the heterodimer Pxa1-Pxa2 on the membrane of the peroxisome [39]. However, applying this approach did not enhance the production of taxifolin, regardless of whether Fox1 or both Pxa1 and Pxa2 were expressed (Figure S8A). The titer of taxifolin was similar to that with constant acetyl-CoA. Meanwhile, alcohol dehydrogenase regulator 1 (Adr1) regulates the expression of fatty acid β-oxidation related genes (FOX1, FOX2) in peroxisomes to promote accumulation of acetyl-CoA [40], which may accelerate the consumption of ATP. The exchange of ATP between the peroxisome and the cytosol is mediated by peroxisomal adenine nucleotide transporter (Ant1) [41]. However, overexpression of ADR1 and ANT1 or only ADR1 did not improve the yield as well (Figure S8B). We speculated that the β-oxidation of fatty acids in the peroxisome might be inherently sufficiently active with no bottleneck in supply of ATP because the titer of taxifolin remained almost constant with acetyl-CoA increase.

Fig. 5.

Optimization of fatty acid β-oxidation to enhance taxifolin synthesis. (A) Schematic illustration of optimizing β-oxidation. (B) Enhancing fatty acid esterification to increase taxifolin synthesis in peroxisomes. PXA1, encoding subunit of heterodimeric peroxisomal ABC transport complex; PXA2, encoding subunit of heterodimeric peroxisomal ABC transport complex; ANT1, encoding peroxisomal adenine nucleotide transporter; FAA2, encoding medium chain fatty acyl-CoA synthetase; FAT1, encoding very long chain fatty acyl-CoA synthetase; FOX1, encoding fatty-acyl-CoA coenzyme A oxidase; FOX2, encoding 3-hydroxyacyl-CoA dehydrogenase and enoyl-CoA hydratase; FOX3, encoding 3-ketoacyl-CoA thiolase with broad chain length specificity. ACC1*, encoding acetyl-CoA carboxylase 1 mutant ScAcc1^{S569A, S1157A}; ATP, adenosine triphosphate; AMP, adenosine monophosphate

Additionally, research indicates that the dimer transporter Pxa1-Pxa2 catalyzes hydrolysis during the transport of very long chain fatty-acyl-coenzyme A (VLCFA-CoA) and long chain fatty-acyl-coenzyme A (LCFA-CoA), releasing hydrolyzed VLCFA and LCFA into the peroxisome [42]. The re-esterification of LCFAs and VLCFAs is catalyzed by very long-chain fatty acyl-CoA synthetase (Fat1), medium-chain fatty acyl-CoA synthetase (Faa2), or both [42]. Subsequently, we hypothesized that FAT1 and FAA2 were also involved in fatty acid β-oxidation. The strain WQ100 was produced by expressing FAT1 and FAA2 in strain WQ87, facilitating an increase of 5.1% in taxifolin production and achieving a production of 120.3 ± 2.4 mg/L (Fig. 5B). The results indicated that overexpression of FAT1 and FAA2 enhanced activation of fatty acids in peroxisomes, facilitating an increased level of fatty-acyl-CoA for β-oxidation.

Discussion

Recently, there has been a growing interest in compartmentalizing biosynthesis pathways in organelles, with significant potential for improving the titer, yield, and productivity [43]. Peroxisomes can support considerable capacity to compartmentalize heterologous enzymes, enabling physical separation between native cytosolic metabolism and engineered peroxisomal metabolism [18]. Leveraging the acetyl-CoA pool generated through the β-oxidation of fatty acids, peroxisomes serve as optimal compartments for the efficient biosynthesis of terpenoids and fatty acid derivatives, including cembratriene-ol, squalene, and fatty alcohols [17, 43]. We hope to expand the application scope of peroxisome compartmentalization to include the synthesis of other products rather than limiting it to the production of terpenes and fatty acid derivatives. We assume that acetyl-CoA is likely to be utilized for flavonoid biosynthesis in peroxisomes since it can be easily converted into malonyl-CoA, which is a precursor of the flavonoid skeleton. In the previous study, the prenylated flavonoid desmethyl xanthohumol downstream pathway from p-coumaric acid and the optimized MVA pathway were reconstructed in S. cerevisiae peroxisomes. However, the titer of desmethyl xanthohumol was much lower than those of the cytosolic pathway, which may be attributed to the unbalanced metabolic flux distribution of acetyl-CoA in peroxisomes where acetyl-CoA demanded be converted into both malonyl-CoA by Acc1 (for flavonoid skeleton synthesis) and DMAPP by MVA pathway (for isoprenoid precursor supply) [44]. Although the synthesis of prenylated flavonoids in peroxisomes did not meet expectations, we proposed that synthesis of flavonols compartmentalized in peroxisomes remained promising. In this study, we demonstrated the feasibility of using peroxisomes in the production of flavonoids. By integrating engineering strategies for peroxisomes, such as pathway expression enhancement, cofactor supply optimization, and regulation of peroxisome biogenesis, we achieved a production of taxifolin in peroxisomes, reaching a yield of 120.3 ± 2.4 mg/L. We achieved this titer value through de novo synthesis using a minimal medium, thereby demonstrating a synthesis efficiency compared to conventional methods [9, 45]. With the increasing demand for taxifolin, this study presents innovative engineering strategies for developing efficient microbial cell factories, facilitating the sustainable and environmentally friendly production of taxifolin.

By individually constructing naringenin biosynthesis pathways within the peroxisome and cytosol, we demonstrated that the naringenin titers in the peroxisomal engineering strain increased by 26% compared to those in the cytosol engineering strain. Besides, the downstream biosynthesis pathway from naringenin to taxifolin was further reconstructed in peroxisomes and cytosol, respectively, revealing that the yield of taxifolin in peroxisomal compartmentalization strain was 1.27-fold higher than that of cytosolic expression strain (Figure S9). Accordingly, compartmentalizing the flavonoid biosynthesis pathway in peroxisomes results in enhanced efficiency. In previous studies, the biosynthesis pathway was simultaneously expressed in the peroxisomes and cytosol to overproduce α-humulene [46], indicating that the acetyl-CoA pools in the peroxisome and cytosol are relatively independent. Additionally, the combined application of both pools simultaneously for synthesis exhibited higher efficiency than a single use of either of them [47]. Therefore, taxifolin production may be further enhanced by employing a dual cytosol-peroxisomal engineering in the future.

The PEX5, PEX13, and PEX14 are involved in the localization of heterologous proteins fused with ePts1 signal peptide [32]. In this study, we enhanced peroxisomal import of heterologous proteins by overexpressing them, with the results revealing an invariable yield of taxifolin. Furthermore, our results demonstrated that enhancing the import efficiency of enzymes by expressing transport-related protein receptors also did not contribute to increased production. Consequently, future research should perhaps focus on reconstructing the biosynthetic pathway by incorporating a more efficient signal peptide compared to ePts1. The PEX11 family and PEX30 family are important in the regulation and proliferation of peroxisomes. PEX11 deletion strains tend to exhibit peroxisome deficiency along with increased accumulation of reactive oxygen species (ROS) [18]. Conversely, overexpressing PEX11 can reduce peroxisome size while increasing their abundance. Additionally, PEX28 and PEX29, operating upstream within the PEX30 family, can be overexpressed to increase peroxisome proliferation. Surprisingly, there was no significant impact in the production level of taxifolin despite the regulation of peroxisome biogenesis through these genes. These observations were contrary to the findings in a previous study, where the titer of fatty alcohols was increased by 2.6 times via overexpression of PEX28 [35]. Our results highlighted that the production efficiency of heterologous pathway compartmentalized in peroxisomes may tend to be limited by the capacity of peroxisomes rather than their abundance. Recently, an engineered strain exhibiting a 137% increase in peroxisome functional capacity was constructed through the overexpression of eight PEX genes, resulting in enhanced pathway compartmentalization and an 80% increase in geraniol biosynthesis [18]. Subsequently, we suggest that further peroxisome engineering optimization of metabolic pathways should perhaps focus on improving the capacity of peroxisomes.

The heterologous expression of enzymes from other species in yeast is likely to be constrained by an insufficient supply of essential cofactors [14]. In this study, ARO8 and TYR1 were overexpressed to enhance α-ketoglutarate availability, resulting in a 50.1% increase in taxifolin production. Additionally, Aco2 was compartmentalized in peroxisomes to convert citrate into isocitrate to facilitate the dual function of NADPH and α-ketoglutarate regeneration, thereby enhancing the activity of F3h and F3’h to enhance taxifolin production. During this process, part of the metabolic flux from the glyoxylate cycle is redirected into the isocitrate/α-ketoglutarate shuttle, facilitating the supply of cofactor in peroxisomes.

Conclusively, this study demonstrates the successful compartmentalization of the taxifolin biosynthesis pathway in peroxisomes for the first time. It demonstrates that leveraging acetyl-CoA in peroxisomes to produce flavonoids offers superior benefits compared to that in the cytosolic pathway. The application scope of peroxisome compartmentalization has been successfully extended from the production of terpenoids and fatty acid derivatives to flavonoids. We observed that cofactor supply is equally critical for taxifolin biosynthesis in peroxisomes. Therefore, we developed a novel strategy to redirect a portion of the citrate metabolic flux from the glyoxylate cycle into the isocitrate/α-ketoglutarate shuttle, improving NADPH and α- ketoglutarate regeneration in peroxisomes. Both NADPH and α-ketoglutarate are essential cofactors in the action of F3h and F3’h and are closely associated with the synthesis efficiency of taxifolin. Through the comprehensive optimization of metabolic engineering strategies, including compartmentalization of the synthetic pathway, cofactor regeneration, and enhancement of rate-limiting enzyme, the taxifolin production was significantly improved. In the best-performing recombinant strain, the taxifolin yield reached 120.3 ± 2.4 mg/L, representing the highest yield achieved in shake-flask fermentation using a minimal medium (Fig. 5). This study provided a valuable foundation for the synthesis of more structurally complex flavonoids through peroxisome compartmentalization in the future.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Q.W. collected literature and prepared the initial draft of the manuscript. R.C. and L.Z. supervised and edited the manuscript. R.C. and Q.W. conceived the idea and developed the study design. R.C. and Q.W. performed the experiments and data analysis. L.Z. and R.C. provided technical support during the experimental phase. R.C. and Q.W. drafted and revised the final manuscript. The final manuscript was approved by all the authors.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.