Combinatorial engineering of enzyme and pathway for efficient β-farnesene bioproduction in Yarrowia lipolytica

Abstract

β-farnesene, a natural sesquiterpene compound, has gained significant attention due to its versatile applications in agriculture, industry, biofuels, and related fields. Microbial biosynthesis offers an environmentally sustainable approach for its commercial-scale production. In order to enhance its production efficiency, further exploration of key rate-limiting steps is required. Here, through directed evolution of the essential β-farnesene synthase, we obtained an optimal variant (AaFS^{T196A/M356T/E380G}), demonstrating 2.29-fold enhancement in β-farnesene titer relative to wild-type. Structural elucidation revealed that the distal mutations mediate allosteric modulation of the catalytic core significantly improving the conversion efficiency of farnesyl diphosphate (FPP) to β-farnesene. Then comprehensive pathway engineering, including the mevalonate pathway amplification, acetyl-CoA precursor enhancement, competitive pathway elimination, and auxotrophic restoration, were carried out in Yarrowia lipolytica, resulting in a high-performance strain FS18 capable of producing 3.41 g/L β-farnesene in shake-flask cultures. Notably, scale up fermentation in 5 L bioreactors yielded a titer of 45.69 g/L, the highest concentration reported in Y. lipolytica to date. This study provided mechanistic insights into terpene synthase engineering and a practical framework for high-level terpenoid biosynthesis in Y. lipolytica.

1. Introduction

Sesquiterpenes represent the largest group within the terpenoid family, with approximately 7000 known members widely distributed throughout the plant kingdom [1]. They exhibit rich chemical diversity and a broad spectrum of biological activities, leading to extensive applications in pharmaceuticals, food, biofuels, and other fields [2]. β-farnesene is an important sesquiterpene compound and occurs naturally primarily in plant essential oils, using as an aphid control agent [3]. Besides, β-farnesene also possesses favorable properties as a biofuel [4]. Additionally, β-farnesene serves as a precursor for the efficient synthesis of isophytol, which is used in vitamin E production [5]. The broad utility of β-farnesene has significantly expanded market demand. With the rapid advancement of biosynthetic technologies, the efficient production of β-farnesene using engineered microbial cell factories has become a reality.

Previous studies have achieved β-farnesene synthesis in various microbial hosts, including Escherichia coli, Saccharomyces cerevisiae, and Yarrowia lipolytica. For instance, You et al. established a balanced mevalonate (MVA) pathway in E. coli, achieving a β-farnesene titer of 8.74 g/L [6]. In S. cerevisiae, rewriting the central carbon metabolism coupled with enzyme engineering enabled a β-farnesene titer of 130 g/L after two weeks fermentation, representing the highest production level reported to date [7]. Compared to these two model strains, Y. lipolytica is a non-conventional oleaginous yeast [8]. Its advantages for industrial production stem from abundant acetyl-CoA supply, a broad spectrum of utilizable carbon sources, and high cellular stress tolerance [9]. Currently, β-farnesene production has been achieved in Y. lipolytica. Shi et al. restored the leucine biosynthesis pathway and attenuated lipid synthesis, achieving a β-farnesene titer of 22.8 g/L [10]. Bi et al. implemented strategies including enhancing the NADPH regeneration pathway and increasing intracellular acetyl-CoA supply, reaching a β-farnesene titer of 28.9 g/L [11]. Liu et al. achieved a β-farnesene production of 35.2 g/L in Y. lipolytica using oleate as the carbon source through peroxisomal compartmentalization and β-oxidation engineering [12]. Although β-farnesene production in Y. lipolytica has been established, further exploration of the common and key rate-limiting steps are still required to further enhance production efficiency.

Prior research has indicated that the biosynthesis of terpenoids primarily involves two parts: the supply of terpene diphosphate intermediates, such as GPP, FPP and GGPP; the enzymatic conversion of these intermediates into terpenoids [13]. To enhance the synthesis of various terpenoids, researchers have developed strategies targeting these two parts. For example, Zhang et al. increased limonene yield by 23.89 % by redirecting acetyl-CoA metabolic flux into the mevalonate (MVA) pathway, thereby enhancing the supply of GPP [14]. Wang et al. achieved an 11.88-fold increase in α-farnesene production through reinforcement of the entire MVA pathway to amplify FPP availability [15]. Additionally, Zhou et al. employed rational design of patchoulol synthase to optimize substrate binding affinity, resulting in a 95.8 % improvement in patchoulol yield [16]. Su et al. utilized the catalytic promiscuity of terpene synthases by using a highly active 1,8-cineole synthase as a template. They successively engineered it into efficient myrcene, limonene, and α-terpineol synthases, enabling high-titer production of all four terpenoids in yeast [17]. Tashiro et al. successfully obtained a pinene synthase mutant through directed evolution, which resulted in a pinene yield twice that of the wild-type and 4.4 times higher than the previously reported maximum [18]. Additionally, Cao et al. fused class I and II diterpene synthases (TPS and LPPS) from the sclareol biosynthetic pathway using GGGS as a linker. This fusion effectively enhanced substrate channeling between LPPS and TPS, leading to a 6.7-fold increase in sclareol production compared to the expression of individual synthases [19]. These studies have demonstrated that terpenoid biosynthesis can be enhanced by optimizing metabolic pathways to increase the supply of upstream metabolic fluxes, coupled with engineering downstream key synthases to improve the conversion efficiency of intermediate metabolites.

To maximize β-farnesene bioproduction in Y. lipolytica, we systematically identified and addressed the critical bottlenecks through integrated enzyme and pathway engineering (Fig. 1). Initially, through protein engineering, a high-efficiency variant AaFS^{T196A/M356T/E380G} with 2.29-fold enhanced catalytic efficiency was obtained. By systematic pathway optimization, strain FS16 is capable of producing 3.08 g/L β-farnesene in shake-flask fermentation. To address the challenges of scale-up production, citrate transporter knockout and auxotrophic complementation were subsequently implemented, resulting in the engineered strain FS18. This strain achieved a β-farnesene production titer of 3.41 g/L in shake-flask fermentation. Furthermore, through fed-batch fermentation, a β-farnesene titer of 45.69 g/L was achieved, representing the highest production level ever reported in Y. lipolytica. These findings establish a framework for overcoming rate-limiting steps in microbial terpenoid biosynthesis and demonstrate the considerable potential of Y. lipolytica as a platform for high-value isoprenoid production.

Fig. 1.

Engineering Y. lipolytica for β-farnesene production. Enhanced β-farnesene synthesis efficiency was achieved through enzyme engineering to optimize β-farnesene synthase activity, combined with metabolic pathway optimization and augmentation of precursor supply via strengthening the downstream synthesis pathway and boosting intracellular acetyl-CoA availability. FPP (farnesyl pyrophosphate), β-FS (β-farnesene synthase), G-6-P (glucose-6-phosphate), X-5-P (xylulose-5-phosphate), acetyl-P (acetyl-phosphate), PDC (pyruvate decarboxylase), IPP (isopentenyl diphosphate), DMAPP (dimethylallyl diphosphate), PK (phosphoketolase, specific for xylulose-5-phosphate), PTA (phosphotransacetylase), ADA (acetaldehyde dehydrogenase acylating), Ald (aldehyde dehydrogenase), ACS (acyl-CoA synthetase), IDI (isopentenyl diphosphate isomerase), ERG20 (farnesyl diphosphate synthase); mβ-FS (modified of β-farnesene synthase). Green font denotes heterologous enzymes.

2. Materials and methods

2.1. Culture conditions and strain preservation

The screening of β-farnesene synthase was conducted using the laboratory-preserved S. cerevisiae strain ySLQ04 [17], and the construction of a high-yield β-farnesene-producing strain was based on the Y. lipolytica strain YLLQ52 [20]. Consistent culture conditions were maintained throughout the experiments. The YPD agar plate, composed of 10 g/L yeast extract, 20 g/L peptone, 20 g/L glucose, and 20 g/L agar, was used for the activation of strains by streaking at 30 °C. The preparation of yeast competent cells was completed using YPD liquid medium. For the screening of positive strains, S. cerevisiae was cultured on SD-URA (Helix) medium, while Y. lipolytica was subjected to positive screening on YPD medium supplemented with a final concentration of 250 mg/L nourseothricin. The plasmid construction was carried out with E. coli DH5α as the host for amplifying the target plasmid, and positive clones were selected on LB medium containing a final concentration of 100 mg/L ampicillin. The constructed strains were mixed with glycerol at a 40 % ratio and stored in a −80 °C freezer for preservation.

2.2. Phylogenetic retrieval and functional screening of β-farnesene synthases

Using Artemisia annua β-farnesene synthase (GenBank: AAX39387.1) as a template, we retrieved 100 phylogenetically related sequences from the NCBI database. A maximum-likelihood phylogenetic tree was constructed using MEGA software (v11.0) for systematic analysis (Fig. S1). From the 10 major clades identified, 13 plant-derived β-farnesene synthase genes were randomly selected. These genes were synthesized and expressed in S. cerevisiae. The enzyme yielding the highest β-farnesene titer was chosen as the starting template for directed evolution. Additionally, a phylogenetic tree of the 13 selected β-farnesene synthase genes was reconstructed in MEGA based on ClustalW multiple sequence alignment for evolutionary assessment.

2.3. Construction of recombinant strains and plasmids

Following the identification of genomic integration sites, single-guide RNA (sgRNA) sequences (20 bp) were rationally selected using the CHOPCHOP web tool (https://chopchop.cbu.uib.no/) to ensure CRISPR-Cas9 targeting specificity. For S. cerevisiae and Y. lipolytica, sgRNA expression plasmids were constructed via XbaI/EcoRI restriction digestion and T4 DNA ligase-mediated assembly, utilizing the S. cerevisiae-specific plasmid p416 and the Y. lipolytica-optimized vector pCfB3405 as backbones, respectively, with the constructed plasmids listed in Table S1. For S. cerevisiae, the homologous arms (1.5 kb flanking regions), inducible promoter (GAL1), terminators (CYC1). The exogenous β-farnesene synthase gene was codon-optimized by GenScript (Nanjing, China) to enable its heterologous expression in S. cerevisiae. The optimized gene fragments were subsequently amplified using primers listed in Table S3. The gene repair fragments were then constructed by fusion PCR. Finally, the corresponding S. cerevisiae strains were ultimately constructed using the lithium acetate-mediated transformation method, with the genotypes of the generated strains provided in Table S2. For Y. lipolytica, genomic integration sites were rationally designed using the EasyCloneYALI toolbox [21]. Promoters, terminators, and homologous arms were PCR-amplified from the genomic DNA of Y. lipolytica strain YLLQ52 with primers listed in Table S3. The codon-optimized exogenous AaFS gene and its mutant variants synthesized by GenScript (Nanjing, China) were tailored for Y. lipolytica expression. PCR amplification was performed using primers listed in Table S3 (Phusion® High-Fidelity DNA Polymerase, Thermo Fisher Scientific). Recombinant Y. lipolytica strains were generated via lithium acetate-mediated transformation, following the protocol described by Gietz and Schiestl [22].

2.4. Shake-flask fermentation cultivation

The β-farnesene production capabilities of the engineered S. cerevisiae and Y. lipolytica strains were assessed using a minimal salt medium. The synthetic medium consisted of 7.5 g/L (NH₄)₂SO₄, 14.4 g/L KH₂PO₄, 0.5 g/L MgSO₄·7H2O, 20 g/L glucose, 2 mL/L of trace elements, and 1 mL/L of vitamins [23]. Single colonies from YPD agar plates were inoculated into test tubes containing 2 mL of the minimal salt medium and cultured in a shaking incubator at 30 °C and 250 rpm for 24–36 h. Subsequently, the cell growth density was measured, and an inoculum with an optical density at 600 nm (OD₆₀₀) of 0.1 was transferred into 20 mL of the minimal salt medium. The cultures were maintained in a shaking incubator at 30 °C and 250 rpm for 24 h. Isopropyl myristate (IPM) was then added to the cultures at a final volume of 10 % for product extraction. After 96 h of flask fermentation, the yield of β-farnesene was determined.

2.5. Product extraction and analysis

After fermentation, the culture broth was collected and centrifuged at 13,000 rpm for 10 min, resulting in three distinct layers: the bottom layer containing the microbial cells, the middle aqueous phase, and the top organic phase. The top organic phase was collected and diluted with n-hexane before being subjected to GC-MS analysis. Quantitative analysis was performed on an Agilent 7890A-5975C instrument equipped with an HP-5 capillary column (30 m × 0.25 mm × 0.25 μm) and an ESI mass spectrometer. The oven temperature was initially held at 50 °C for 1 min, then increased to 280 °C at a rate of 15 °C/min and held for 1 min, followed by a further increase to 300 °C at a rate of 20 °C/min and held for 2 min. Under these conditions, the samples were analyzed alongside a β-farnesene standard for quantification. For the detection of mevalonic acid and citric acid in the culture broth, the aqueous phase was obtained by centrifugation at 13,000 rpm for 10 min and then filtered through a 0.22 μm, 13 mm water filter. High-performance liquid chromatography (HPLC) was used for analysis, with an Aminex HPX-87H column (7.8 × 300 mm, 1250140, Bio-Rad) maintained at 63 °C. The mobile phase was 5 mM sulfuric acid, with an injection volume of 5 μL and a flow rate of 0.6 mL/min. Detection was carried out at 35 °C using a refractive index detector, and the chromatographic peak areas were analyzed using Agilent ChemStation software.

2.6. Directed evolution of β-farnesene synthase

To amplify the error-prone fragments of the AaFS gene, primers listed in (Supplementary Table S3) were used in conjunction with EasyTaq DNA polymerase (from Beijing GenScript Biotechnology Corporation) and 0.5 mM magnesium ions. The error-prone PCR fragments of AaFS were then recombined with the linear plasmid p416 using Gibson Assembly and transformed into S. cerevisiae strain ySLQ04 to construct a mutant library. Positive clones were selected on SD-URA plates and inoculated into 96-well plates containing 400 μL of minimal salt medium, which were then cultured in a temperature and humidity controlled shaker at 30 °C and 800 rpm for 36 h. Subsequently, 40 μL of the inoculum was transferred into a new 96-well plate containing 360 μL of minimal salt medium and cultured for 24 h. Then, 200 μL of isopropyl myristate (IPM) was added to the mutant strains, and extraction fermentation was carried out in a constant temperature and humidity shaker (30 °C, 800 rpm) for 96 h. Following fermentation, the cultures were centrifuged at 3500 rpm for 10 min, after which 15 μL of the supernatant was mixed with 15 μL of 75 % sulfuric acid solution and 4 mg/mL vanillin solution, incubated and agitated at 65 °C for 30 min, and the absorbance was measured at 454 nm. Strains with higher absorbance values were selected for flask rescreening. After confirming significant improvement compared to the wild-type strain, plasmid extraction and sequencing were performed to identify the mutation sites.

2.7. Molecular modeling of β-farnesene synthase

β-farnesene synthase was predicted using AlphaFold3 (v. 3.0.0) to generate a 3D structural model. The catalytic active site and access tunnel locations of AaFS were predicted using CAVER software (https://loschmidt.chemi.muni.cz/hotspotwizard/). The molecular structure of farnesyl diphosphate (FPP; PubChem CID: 445713) was retrieved from the PubChem database in SDF format. Protein structures underwent preprocessing in AutoDock Vina, including removal of crystallographic waters and addition of polar hydrogens. The conformation with the optimal binding energy was selected, and molecular interactions were visualized and annotated using PyMOL v2.5.4. Molecular dynamics simulations of the FPP-docked complexes of both wild-type AaFS and its triple mutant mFS. was then performed following the methods described previously [24].

2.8. Expression and purification of β-farnesene synthase

Genes encoding wild-type Artemisia annua β-farnesene synthase (AaFS) and its triple mutant AaFS^{T196A/M356T/E380G} were cloned into the pET28a vector. The resulting plasmids were transformed into E. coli BL21(DE3) for heterologous expression. Transformants were cultured in 10 mL LB medium supplemented with 50 μg/mL kanamycin at 37 °C. Subsequently, 5 mL of the seed culture was inoculated into 500 mL fresh LB medium containing 50 μg/mL kanamycin. When the culture reached an OD₆₀₀ of 0.5, protein expression was induced with 0.2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). The culture was then incubated at 18 °C for 20 h. Cells were harvested by centrifugation (5000 rpm, 10 min, 4 °C). The pellet was resuspended in lysis buffer [20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 5 mM imidazole, 10 % (v/v) glycerol] and subjected to ultrasonication on ice for 30 min (5 s pulse/5 s interval cycles) until complete cell disruption. The lysate was clarified by centrifugation (10,000 rpm, 10 min, 4 °C). The supernatant was loaded onto a Ni-NTA resin column (GenScript, Nanjing, China) and purified according to the manufacturer's protocol. Purified recombinant proteins were verified by SDS-PAGE analysis. Purified enzymes 150 μg were incubated with 50 μM farnesyl diphosphate (FPP; Sigma-Aldrich) in 100 μL reaction buffer 50 mM Tris-HCl (pH 8.0), 10 μM MgCl₂ at 30 °C for 2 h. Reactions were terminated by adding 200 μL quenching solution (1 M EDTA, 4 M NaOH). Reaction products were extracted with 200 μL hexane and quantified by gas chromatography (GC) based on standard curves generated using authentic β-farnesene.

2.9. Fed-batch fermentation of β-farnesene producing strain

Initially, single colonies of the strain were obtained by streaking on YPD agar plates. Three to five single colonies were then inoculated into 20 mL of minimal salt medium and cultured in a shaking incubator at 30 °C and 250 rpm for 36 h. The culture was then transferred to 100 mL of minimal salt medium at an OD₆₀₀ of 0.2 for fermentation for 24 h. Subsequently, the culture was transferred to a 1 L bioreactor at an OD₆₀₀ of 0.2. The 1 L bioreactor contained 400 mL of minimal salt medium, which included 7.5 g/L (NH₄)₂SO₄, 14.4 g/L KH₂PO₄, 0.5 g/L MgSO₄·7H₂O, 20 g/L glucose, 2 mL/L trace elements, and 1 mL/L vitamins. When the glucose in the bioreactor was depleted, the feed composition was 25 g/L (NH₄)₂SO₄, 15 g/L KH₂PO₄, 2.5 g/L MgSO₄·7H₂O, 10 mL/L trace elements, 5 mL/L vitamins, and 600 g/L glucose. The pH inside the bioreactor was maintained at 5.0 with 10 M KOH, and the dissolved oxygen was controlled at 30 %, regulated by the agitation speed ranging from 300 to 1000 r/min. The aeration rate inside the bioreactor was 1.0 vol per volume per minute (vvm), and the glucose concentration inside the bioreactor was maintained at 5 g/L. When the culture reached an OD₆₀₀ of 10, 40 mL of the organic solvent IPM was added, and the IPM content was maintained at 10 %. Samples were taken during the fermentation process to measure the cell density (OD₆₀₀), glucose, β-farnesene, and other metabolites. The fed-batch fermentation in the 5 L bioreactor followed the same inoculation steps as the 1 L bioreactor, with an initial volume of 2 L of minimal salt medium and the same feed composition as the 1 L bioreactor. The pH inside the bioreactor was maintained at 5.0 with 10 M KOH, and the dissolved oxygen was controlled at 30 %, regulated by the agitation speed ranging from 300 to 1000 r/min. The aeration rate inside the bioreactor was 1.0 vol per volume per minute (vvm), and the glucose concentration inside the bioreactor was maintained at 5 g/L. When the culture reached an OD₆₀₀ of 10, 200 mL of the organic solvent IPM was added, and the IPM content was maintained at 10 %. Samples were taken during the fermentation process to measure the cell density (OD₆₀₀), glucose, β-farnesene, and other metabolites.

3. Results and discussion

3.1. Directed evolution of β-farnesene synthase

In the biosynthetic pathway of β-farnesene, the catalytic efficiency of β-farnesene synthase (β-FS) play a crucial role [3]. Directed evolution serves as a powerful technique that optimizes enzymatic functional properties by mimicking the natural evolutionary process [25].

To acquire a starting β-FS for enzyme engineering, a phylogeny containing 100 β-FSs was built on the sequence of β-FS from Artemisia annua. Phylogenetic analysis revealed that these 100 β-FS genes clustered into ten major clades (Fig. S1). Given that β-farnesene synthase from both A. annua and Matricaria chamomilla var. have been demonstrated for β-farnesene production in S. cerevisiae, leading to 184 mg/L and 203 mg/L of β-farnesene respectively [5], these two were prioritized in next primary screening. Following this, we randomly selected one to two genes from each clade. This yielded a final set of 13 genes for evaluating their β-farnesene biosynthesis capacity. Considering the CEN/ARS-containing plasmid is unstable in Y. lipolytica [26], these genes were expressed based on the centromeric plasmid pRS416 in the S. cerevisiae strain ySLQ04, which features an overexpressed mevalonate (MVA) pathway [19]. Unexpectedly, the 10 phylogenetically distant genes showed either low or undetectable enzymatic activity and the highest titer of β-farnesene (205.29 mg/L) was achieved by the reported β-FS derived from A. annua (GenBank: AAX39387.1). The rest two β-FSs clustered close with AaFS also showed equivalent capacity for β-farnesene biosynthesis (Fig. 2A) To ensure the activity of AaFS in Y. lipolytica, the germacrene A synthase gene in the reported sesquiterpene producing strain YLLQ52 was replaced by AaFS [20], resulting in strain YlAaFS. This strain produced 409.1 mg/L of β-farnesene in shake-flask fermentation (Fig. 2B), indicating that AaFS is a suitable starting β-FS for enzyme engineering.

Fig. 2.

Directed evolution of β-farnesene synthase. (A) Phylogenetic analysis and functional characterization of β-farnesene synthase genes. Numbers on branches represent bootstrap support values, and bar lengths correspond to β-farnesene production levels (mean ± SD of three biological replicates). Colored labels denote the five functionally active variants in S. cerevisiae, while black labels indicate inactive variants. (B) β-Farnesene production in S. cerevisiae and Y. lipolytica with expression of AaFS. (C) Twenty-nine isolates from the secondary shake-flask screening showed higher performance than the control strain, data represent mean ± SD of three. (D) Reverse validation of high-activity β-farnesene synthase mutants in S. cerevisiae. (E) Effect of combinatorial mutations of β-farnesene synthase in S. cerevisiae. (F) Iterative combinatorial mutations of efficient β-farnesene synthase.

To further directed evolution of AaFS, pRS416 plasmid-based mutant library was constructed via error-prone PCR and transformed into ySLQ04, followed by high-throughput screening using the vanillin-sulfuric acid method [5] (Figs. S2 and S3). A total of 5000 transformants were screened over three rounds of testing (Fig. S4) and 160 superior mutants compared with the control were selected and reevaluated in shake flask fermentation (Fig. S5). Among these strains, 29 exhibited higher titers of β-farnesene than the control (Fig. 2C) and the plasmids from the top 4 strains were extracted and sequenced. 5 nonsynonymous mutations and 1 synonymous mutation were identified (Fig. S6). Genome integration based reverse engineering revealed that five key amino acid substitutions (T196A, S208P, M356T, E380G, and I507T) boosted β-farnesene production in S. cerevisiae. These mutants achieved titers of 338.49 mg/L, 223.00 mg/L, 237.24 mg/L, 294.98 mg/L, and 238.06 mg/L, representing increases of 78.40 %, 14.24 %, 21.54 %, 51.11 %, and 21.96 % over the control respectively (Fig. 2D).

The activity of an enzyme is often influenced by the synergistic effects of multiple amino acid sites [27]. In order to further enhance the catalytic efficiency of β-farnesene synthase, we performed combinatorial mutagenesis based on the five identified sites. We constructed 10 double-mutant strains through pairwise combinations. Notably, significant activity improvements were only observed when these sites were combined with T196A. Among them, the T196A/M356T double-mutant strain showed the highest β-farnesene production, reaching at 423.52 mg/L, which was 2.17 times higher than that of the wild-type enzyme (Fig. 2E). Subsequently, based on the T196A/M356T double-mutant, we constructed an additional 10 strains with triple and quadruple mutations. After 96 h fermentation, the T196A/M356T/E380G triple mutant strain exhibited the highest β-farnesene yield of 447.52 mg/L, representing a 2.29-fold increase compared to the wild-type enzyme. It was also 32.2 % and 5.67 % higher than the T196A single mutant and T196A/M356T double mutant strains, respectively (Fig. 2E). Furthermore, GC-MS analysis revealed that alterations in the amino acid sites of β-farnesene synthase AaFS did not alter the product spectrum (Fig. S7). Thus, these results demonstrated the successful engineering of AaFS^{T196A/M356T/E380G}, a highly efficient β-farnesene synthase variant, via iterative combinatorial mutagenesis (Fig. 2F).

3.2. Purification and characterization of β-farnesene synthase mutant

To further validate the enhanced catalytic efficiency of the triple mutant, both AaFS and AaFS^{T196A/M356T/E380G} were expressed in E. coli BL21(DE3) and purified. Consistent with the in vivo results, in vitro assays with purified enzymes demonstrated that the triple mutant catalyzed the conversion of FPP to β-farnesene with a specific activity of 94.97 nmol h⁻¹ mg⁻¹. This represents a 53.75 % increase in activity compared to wild-type AaFS (61.77 nmol h⁻¹ mg⁻¹) (Fig. 3A, S8). The difference of increase amplitude between in vivo and in vitro assays suggested that these mutational amino acids may affect the protein expression level. To further investigate the mechanism underlying the enhanced activity of the AaFS^{T196A/M356T/E380G} variant, we employed AlphaFold3 to predict the structural models of both the wild-type AaFS and the triple mutant.

Fig. 3.

Structural and functional characterization of β-farnesene synthase. (A) In vitro activity of AaFS^{T196A/M356T/E380G} mutant. Data represent mean ± SD of three biological replicates (***p < 0.01). (B) Molecular docking of substrate FPP into β-farnesene synthase. (C) Structural mapping of the T196A, M356T, and E380G mutation sites (D) The binding process between β-farnesene synthase (and its mutants) and FPP generated fluctuations within a 100-ns timeframe. (E) In the β-farnesene synthase mutant AaFS^{T196A/M356T/E380G}, the fluctuating region encompasses residues 477–485, with blue highlighting indicating the spatial range of this segment. (F) Heterologous expression of AaFS^{T196A/M356T/E380G} in S. cerevisiae and Y. lipolytica, ScmFS: S. cerevisiae expressing AaFS^{T196A/M356T/E380G}YlmFS: Y. lipolytica expressing AaFS^{T196A/M356T/E380G}. Data represent mean ± SD of three biological replicates.

Studies by Wang et al. on the mechanism of farnesene synthase revealed that the core of β-farnesene catalysis lies in the loss of the pyrophosphate group from FPP, in which magnesium acts as a key cofactor [28], critically facilitating the dissociation of the pyrophosphate moiety, stabilizing the carbocation intermediate during cyclization, and inducing conformational changes within the enzyme active site pocket [29]. Using Salvia officinalis sesquiterpene synthase (PDB: 1N24) [30] as a homology modeling template, molecular docking analysis revealed the structural details of the magnesium ion (Mg²⁺) binding environment with the pyrophosphate (PPi) group in β-farnesene synthase (AaFS). Docking results revealed that the key amino acid residues within the FPP-binding region are R290, D327, D331, N332, E479, R482, D551, and W554. These eight residues interact with the pyrophosphate group of FPP, while the three magnesium ions chelate the pyrophosphate group within the catalytic center of β-FS. Specifically, Mg1 is positioned above the pyrophosphate group and coordinates with E479 and D551. Mg2 and Mg3 reside below the pyrophosphate group, forming coordination bonds with the conserved aspartate residues D327 and D331 of the DDXXD motif (Fig. 3B).

Interestingly, the three key mutations (T196A/M356T/E380G) in the high-efficiency mutant AaFS^{T196A/M356T/E380G} are positioned distally, outside the catalytic center (Fig. 3C–S9). To investigate the mechanism by which these distal mutations enhance the catalytic efficiency of β-FS, we performed molecular dynamics (MD) simulations on both the AaFS-FPP and AaFS^{T196A/M356T/E380G}-FPP complexes to analyze differences in their dynamic behavior. The results revealed significant fluctuations in the amino acids region spanning residues 477 to 485. This region is located at the entrance to the active site and encompasses key residues (E479 and R482) that interact with the pyrophosphate moiety (Fig. 3D and E). Following 10 ns of simulation, notable changes emerged in the MD trajectory of the mutant. Specifically, at 15 ns, 30 ns, and 50 ns, this region adopted an expanded conformation (Figs. S10 and S11). This phenomenon suggests that the three distal mutations likely induced conformational changes in the entrance region. These changes affected the key pyrophosphate-interacting residues E479 and R482, facilitating both the entry/exit of the substrate FPP and the loss of the pyrophosphate group, thereby enhancing the catalytic efficiency of β-FS.

Furthermore, we expressed AaFS^{T196A/M356T/E380G} in the S. cerevisiae ySLQ04 and replaced the AaFS in strain YlAaFS with AaFS^{T196A/M356T/E380G}, resulting in strain ScmFS and YlmFS. Consistent with previous findings, the Y. lipolytica strain YlmFS exhibited superior production capacity, yielding 630.89 mg/L β-farnesene with a glucose-specific yield of 0.031 g/g. In contrast, the S. cerevisiae strain ScmFS produced 447.51 mg/L β-farnesene with a glucose-specific yield of 0.022 g/g (Fig. 3F). However, as shown in (Fig. S12), the relative increase of AaFS^{T196A/M356T/E380G} compared to AaFS on β-farnesene production in S. cerevisiae was 229 % but only 154 % in Y. lipolytica, indicating the potential limitations of this cross-host screening strategy applied in this study. This discrepancy can be attributed to several underlying factors, including differences in precursor level, variation in enzyme solubility, and distinct intracellular environments that collectively influence protein folding, functional expression and post-translational modification [31]. Therefore, employing the targeted host for targeted enzyme engineering should be the first option. For Y. lipolytica, co-expressing an essential on the plasmid may be a solution to solve the plasmid instability and plasmid-based screening in Y. lipolytica. Based on its higher β-farnesene production titer and carbon source conversion efficiency, we selected the Y. lipolytica strain YlmFS for subsequent construction of high-yield β-farnesene production strains.

3.3. Tuning downstream pathways for enhanced β-farnesene production

During metabolic pathway engineering, multi-enzyme complex assembly was used to minimize intermediate metabolite dispersion and enhance carbon conversion efficiency [32]. The ERG20 gene plays a central role in terpenoid biosynthesis by encoding farnesyl diphosphate synthase [2], which catalyzes the sequential formation of terpenoid precursors geranyl diphosphate (GPP) and FPP (Fig. 4A). To minimize intermediate metabolite diffusion and establish substrate channeling, we constructed multi-enzyme complexes by fusing AaFS^{T196A/M356T/E380G} with ERG20 via flexible linkers (GGGS)₃ in different orientations. Compared to individual expression of AaFS^{T196A/M356T/E380G}and ERG20, the fusion constructs AaFS^{T196A/M356T/E380G}-ERG20 and ERG20-AaFS^{T196A/M356T/E380G} exhibited 21.80 % and 10.25 % higher β-farnesene titers, respectively (Fig. 4B).

Fig. 4.

Enhancing the β-farnesene production by regulating the downstream biosynthetic pathway. (A) The schematic diagram of strengthening the downstream β-farnesene synthesis pathway from mevalonate. (B) Effects of fusion expression of β-farnesene synthase AaFS^{T196A/M356T/E380G} with ERG20 in different configurations on β-farnesene synthesis. (C) β-Farnesene production profiles by copy number optimization of fused mFSERG20. (D) Impact of overexpressing MVA pathway genes IDI and ERG12 on β-farnesene synthesis. Data are from three biological replicates.

To further optimize metabolic flux toward the target product, we intensified downstream pathway engineering by implementing multi-copy expression of the AaFS^{T196A/M356T/E380G}-ERG20 fusion complex. Starting from base strain FS04 (single-copy), strains FS06-FS09 were generated with increasing copy numbers. β-farnesene titers reached at 1.59 g/L (FS06), 1.80 g/L (FS07), 1.95 g/L (FS08), and 2.02 g/L (FS09), representing increases of 51.43 %, 71.43 %, 85.71 %, and 92.38 % over FS04, respectively. Conversely, mevalonate (MVA) accumulation showed a significant inverse correlation: FS04 accumulated 1.21 g/L MVA, while FS06-FS09 exhibited progressive reductions to 0.98 g/L, 0.56 g/L, 0.41 g/L, and 0.32 g/L versus FS04, respectively (Fig. 4C).

Notably, five copies strain FS09 yielded only a 70 mg/L β-farnesene increase over four copies strain FS08, while still accumulating 317.53 mg/L MVA. Consequently, subsequent efforts focused on optimizing precursor supply by redirecting MVA byproduct flux (Fig. 4A). ERG12 catalyzes the phosphorylation of mevalonate to mevalonate-5-phosphate [33]. IDI mediates the isomerization between isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) [34]. As reported, overexpression of these two genes can effectively reduce the accumulation of the MVA pool [35]. Strains FS10 and FS11 were constructed by overexpressing IDI and ERG12 sequentially in FS09. However, the β-farnesene titers in strains FS10 (2.07 g/L) and FS11 (2.10 g/L) exhibited only minor and statistically insignificant increases compared to that of FS09 (2.05 g/L). Similarly, MVA accumulation in FS09, FS10, and FS11 was 317.53 mg/L, 310.62 mg/L, and 307.03 mg/L, respectively, although a decrease in MVA accumulation was observed, no significant difference was detected (Fig. 4D). This result indicates that the strategy of reducing MVA pool accumulation by overexpressing ERG12 and IDI to enhance β-farnesene synthesis yielded minimal effects. Since pulling downstream MVA metabolism demonstrated limited efficacy in promoting β-farnesene production, it would be worthwhile to explore whether increasing the supply of the precursor acetyl-CoA and directing metabolic flux toward the MVA pathway could enhance β-farnesene synthesis.

3.4. Augmented acetyl-CoA supply drives efficient β-farnesene biosynthesis

Following the insignificant impact of downstream MVA pathway enhancement on β-farnesene production, we redirected efforts toward upstream metabolic regulation to increase intracellular acetyl-CoA supply, thereby promoting β-farnesene biosynthesis. Acetyl-CoA serves as the central initiating precursor for terpenoid biosynthesis, providing the foundational carbon source for all terpenoid scaffolds [36]. Robust acetyl-CoA supply provides essential metabolic driving force for terpenoid biosynthesis [37].

In this study, we introduced the pyruvate bypass and pentose phosphate pathway to boost intracellular acetyl-CoA supply (Fig. 5A). On the basis of strain FS09, we constructed strain FS12 by introducing acetaldehyde dehydrogenase (acylating) (ADA), and strain FS14 by incorporating xylulose-5-phosphate-specific phosphoketolase (PK) and phosphotransacetylase (PTA). The β-farnesene titers of strains FS12 and FS14 reached 2.26 g/L and 2.66 g/L, corresponding to increases of 13.57 % and 33.67 %, respectively, compared to FS09. Notably, strain FS13, which involved the introduction of PK alone, also exhibited a higher β-farnesene production than FS09. We speculate that Y. lipolytica may have an endogenous acetyl-phosphate transferase that catalyzes this conversion. In order to further enhance the production of β-farnesene, we combined these two heterologous pathways and constructed strains FS15, which produced 2.83 g/L β-farnesene (Fig. 5B).

Fig. 5.

Enhancing intracellular acetyl-CoA supply to boost β-farnesene synthesis. (A) Schematic diagram of the pyruvate bypass pathway and pentose phosphate pathway for acetyl-CoA synthesis and β-farnesene production, the green arrows indicate the carbon flux directed by the introduction of heterologous genes. PK (xylulose-5-phosphate specific phosphoketolase), PTA (phosphotransacetylase), ADA (acetaldehyde dehydrogenase acylating), and Ald (aldehyde dehydrogenase). (B) Impact of enhanced acetyl-CoA supply via heterologous pathway engineering on β-farnesene biosynthesis. Statistical analysis was performed using one-way ANOVA.

In the pyruvate dehydrogenase bypass, pyruvate is converted to acetaldehyde by pyruvate decarboxylase, then to acetate by aldehyde dehydrogenase, and finally activated to acetyl-CoA by acetyl-CoA synthase at the expense of two ATPs [38]. To save more energy for cell growth and β-farnesene production, the endogenous aldehyde dehydrogenase (ALD6) was disrupted in strain FS15, resulting in strain FS16. This strain achieved β-farnesene titer of 3.08 g/L, representing 8.83 % increase compared with FS15 (Fig. 5B). Unexpectedly, the OD₆₀₀ values and MVA concentrations of these engineered strains showed no significant changes, indicating that the increased acetyl-CoA precursor availability did not affect the growth or MVA accumulation in the β-farnesene-producing strain. These results confirmed that increase the supply of acetyl-CoA was an effective way for increase β-farnesene production.

3.5. Fed-batch fermentation for β-farnesene overproduction

To evaluate the potential of the high-yielding strain FS16, a scaled-up fermentation experiment was conducted in 1 L parallel bioreactors using glucose as the sole carbon source. The initial glucose concentration in the medium was 20 g/L. After 36 h of fermentation, when the initial glucose was depleted, a feeding medium was supplemented to maintain the residual glucose concentration at approximately 5 g/L. The fermentation parameters were controlled as follows: pH was maintained at 5.0, the aeration rate was set at 1.0 vvm, and the agitation speed ranged from 300 to 1000 rpm to keep the dissolved oxygen (DO) level at 30 %. The results showed that the concentration of β-farnesene increased continuously over the cultivation period, reaching a titer of 24.30 g/L at 228 h. The biomass entered the logarithmic growth phase after 36 h, reached a plateau at 144 h with an OD₆₀₀ around 100, and began to decline after 180 h, indicating the onset of the death phase. Notably, significant accumulation of citric acid was observed during the later stages of fermentation, particularly after 96 h, with a final concentration of 97.57 g/L (Fig. 6A).

Fig. 6.

Fed-Batch fermentation optimization for β-farnesene production. (A) Fed-batch fermentation profiles of strain FS16. The solid points represent the average titer, while the hollow points of the same color indicate the individual values from two biological replicates. (B) Schematic of the URA complementation and citrate transporter knockout in β-farnesene producing strain. (C) Effects of citrate transporter (CEX1) knockout and URA complementation on the β-farnesene engineered strain. (D) Fed-batch fermentation profiles of strain FS18 in a 5 L bioreactor. The solid points represent the average titer, while the hollow points of the same color indicate the individual values from two biological replicates.

Recent study revealed a key gene involved in citric acid transport, named citrate exporters 1 (CEX1) [39], which impacts citric acid production in Y. lipolytica. Zhang et al. showed that knocking out CEX1 reduces late-stage citric acid accumulation and promotes mevalonate production [40]. To address citric acid accumulation during fed-batch fermentation, we knocked out CEX1 in strain FS16 to create strain FS17. Meanwhile, to resolve the auxotrophic marker of FS17, we performed in-situ complementation of the URA3 gene, yielding strain FS18 (Fig. 6B). In FS17 and FS18, citric acid accumulation decreased by 64.86 % and 70.27 % compared to FS16, to 0.13 g/L and 0.11 g/L in shake flasks fermentation, respectively. CEX1 knockout did not significantly affect the cell growth or β-farnesene production of FS16 while URA3 complementation led to enhanced cell growth and increased β-farnesene production by 10.38 % to 3.41 g/L compared with FS16 (Fig. 6C).

Subsequently, we conducted fed-batch fermentation of FS18 in a 5 L bioreactor. The fermentation parameters were controlled as follows: the pH was maintained at 5.0, the aeration rate was set at 1.0 vvm, and the agitation speed was varied between 300 and 1000 rpm to keep the dissolved oxygen level at 30 %. The glucose concentration was controlled using the same feeding strategy as applied in the 1 L bioreactor fermentation. Citric acid was first detected at 60 h and accumulated gradually to a final amount of 2.58 g/L. The biomass of the strain entered the logarithmic growth phase after 36 h of fermentation, and the cell density reached a peak OD₆₀₀ value of 349.4 at 192 h. The β-farnesene concentration increased over time, reaching 45.96 g/L with a glucose conversion rate of 0.12 g/g (Fig. 6D). This represents the highest production level reported to date in the yeast Y. lipolytica.

4. Conclusions

In this study, we conducted directed evolution of the key enzyme in the β-farnesene synthesis pathway, β-farnesene synthase and obtained an efficient variant. Based on this mutant enzyme, systematic metabolic engineering including protein fusion expression, acetyl-CoA supply, citrate transporter deletion and uracil auxotrophy complementation, was performed, resulting in strain FS18. This strain produced 3.41 g/L and 45.96 g/L β-farnesene in shake-flask cultures and in a 5-L bioreactor via fed-batch fermentation respectively, both representing the highest reported sesquiterpene titer in Y. lipolytica. The enzyme and pathway engineering strategies used here offer valuable insights for terpene production in Y. lipolytica.

CRediT authorship contribution statement

Hongyang Chen: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation. Liqiu Su: Writing – review & editing, Formal analysis. Zhen Yao: Methodology. Kaizhi Jia: Writing – review & editing, Supervision. Zongjie Dai: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Qinhong Wang: Supervision, Resources.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the supplementary data to this article: