De novo biosynthesis of cannabinoids and their analogs in Yarrowia lipolytica
Department of Chemical Engineering, Guangdong Technion-Israel Institute of Technology, Shantou, Guangdong, 515063, China
The Ruth and Bruce Rappaport Faculty of Medicine, Technion-Israel Institute of Technology, Haifa, 320002, Israel
School of Food Science and Pharmaceutical Engineering, Nanjing Normal University, Nanjing, 210023, China
College of Biological and Pharmaceutical Engineering, West Anhui University, Lu'an, Anhui, 237012, China
Chemistry and Chemical Engineering Guangdong Laboratory, Shantou, 5l5031, China
Oncology and Hematology Division, Cancer Center, Emek Medical Center, Afula, 1834111, Israel
Center for Lipid Engineering, Muyuan Laboratory, Zhengzhou, 450016, Henan, China
⁎Corresponding author. The Ruth and Bruce Rappaport Faculty of Medicine, Technion-Israel Institute of Technology, Haifa, 320002, Israel. gil_ba@clalit.org.il⁎⁎Corresponding author. Oncology and Hematology Division, Cancer Center, Emek Medical Center, Afula, 1834111, Israel. idan5161@gmail.com⁎⁎⁎Corresponding author. Department of Chemical Engineering, Guangdong Technion-Israel Institute of Technology, Shantou, Guangdong, 515063, China. peng.xu@gtiit.edu.cnAbstract
Cannabis sativa has long been a cornerstone of both medicinal and cultural practices, with its therapeutic use spanning over 2700 years. Central to its therapeutic effects are cannabinoids, which interact with the endocannabinoid system to influence various physiological processes such as anxiety, pain, and inflammation. Despite its benefits, cannabinoid production faces challenges and scarcity from plant extraction. This work leverages Yarrowia lipolytica as a platform for cannabinoid biosynthesis. By optimizing the precursor supply, engineering biomolecular condensate-like dual prenyltransferase expression and expanding endogenous metabolism with a noncanonical polyketide synthase, we achieved the de novo biosynthesis of various cannabinoids and their analogs. Our engineered Y. lipolytica produced ∼3.5 mg/L cannabigerolic acid, 18.8 mg/L orsellinic acid, and 0.5 mg/L cannabigerorcinic acid. Additionally, the CBGA titer reached 15.7 mg/L with olivetolic acid supplementation. This work demonstrates the versatility of Y. lipolytica as a promising host for the production of cannabinoids and their analogs, which opens avenues for further research and medicinal applications.
1Introduction
Cannabis sativa has played a significant role in social and religious practices throughout human history and has been employed therapeutically for over 2700 years, beginning in ancient China [[1], [2], [3]]. The endocannabinoid system (ECS) is a crucial physiological network implicated in numerous fundamental functions, including neuroplasticity, learning, memory, neuronal development, cell fate determination, nociception, inflammation, appetite, digestion, and the regulation of stress and emotions [4]. Phytocannabinoids are a diverse range of chemical compounds known to interact with cannabinoid receptors (CBRs) within the ECS [5,6]. Cannabidiol (CBD), or CBD-C5, distinguished by its pentyl (five-carbon) side chain, is a major nonpsychoactive constituent of C. sativa. CBD modulates CBR activation, either by enhancing or inhibiting receptor activity, and is implicated in various physiological processes such as anxiety, insomnia, chronic pain, and inflammation [7,8]. The U.S. Food and Drug Administration (FDA) has acknowledged its therapeutic potential by approving Epidiolex, a purified form of CBD, for the treatment of seizures associated with Lennox-Gastaut syndrome (LGS), Dravet syndrome (DS), and tuberous sclerosis complex (TSC) in patients aged one year and older [9,10].
In addition to CBD, other CBD analogs have drawn significant attention in medicinal and clinical research. Cannabigerolic acid (CBGA), another key cannabinoid, acts as the primary precursor for all pentyl side chain cannabinoids. Like CBD, CBGA exhibits anticonvulsant properties in animal models of epilepsy and functions as a dual PPARα/γ agonist, suggesting its potential in treating metabolic disorders such as diabetes and dyslipidemia [11]. Notably, CBGA also demonstrated the second-highest inhibitory activity of all the phytocannabinoids against the COVID-19 3C-like protease [12]. Furthermore, a 2022 preclinical study indicated that CBGA, along with cannabidiolic acid (CBDA) and tetrahydrocannabinolic acid (THCA), could prevent SARS-CoV-2 infection [13]. These findings highlight the significant promise of CBGA for therapeutic development, both as a precursor to other cannabinoids and as a bioactive compound with potential applications in metabolic and antiviral treatments.
Microbial metabolic engineering provides an alternative solution to overcome the bottlenecks of chemical synthesis or plant extraction, including environmental concerns or resource scarcity [14]. In addition, microbes offer some other advantages: they require less arable land, are not affected by climate change, are genetically tractable, and can be scaled up for large-scale production [15,16]. Additionally, microbes exhibit robust growth and high conversion rates when utilizing a wide range of low-cost renewable raw materials. Nevertheless, only a few groups have completed the de novo biosynthesis of cannabinoids in industrially-relevant hosts [[17], [18], [19]]. Gülck et al. reported cannabinoid biosynthesis using the tobacco plants (Nicotiana benthamiana) and S. cerevisiae, by feeding OLA as substrate, they got CBGA production at 1 mg/L in the tobacco plant and 5.3 μg/L in yeast [20]. Luo et al. first reported the de novo biosynthesis of tetrahydrocannabinolic acid (THCA) and cannabidiolic acid (CBDA) in S. cerevisiae, attaining titers of 7.2 mg/L CBGA, 8.0 mg/L THCA, and 4.8 mg/L THCVA [21]. Subsequently, their research group advanced the production of CBGA, reaching a significantly higher titer of 510.32 mg/L in S. cerevisiae [18].
Yarrowia lipolytica is a nonconventional yeast and is classified as "generally regarded as safe" (GRAS) (Table 1) [22]. Unlike S. cerevisiae, Y. lipolytica affords high acetyl-CoA flux due to its cytosolic ATP citrate lyase. It also accommodates high flux for malonyl-CoA and HMG-CoA, and can grow on a wide range of inexpensive raw materials [16,23]. The oil-accumulating property and expanded inner membrane space facilitate the regio- and stereoselectivity of various cannabinoid-active enzymes (Table 1) [16,24]. These attributes make Y. lipolytica a superior platform for producing cannabinoids and their analogs, when compared to other commonly used chassis strains (Table 1).Expression Platform E. coli S. cerevisiae Y. lipolytica Genetic tools ★★★★★ ★★★★★ ★★★ Genome annotation ★★★★★ ★★★★ ★★★ Acetyl-CoA/Malonyl-CoA/HMG-CoA flux ★★ ★★★ ★★★★★ P450 expression ★ ★★★ ★★★★★ Substrate flexibility ★★★★ ★★ ★★★★ Acid tolerance ★★★ ★★ ★★★★★ FDA safety ★ ★★★★★ ★★★★★ Hydrophobic lipid body environment ★ ★★★ ★★★★★
In this work, we achieved de novo biosynthesis of cannabinoid family CBD-C5 and CBD-C1 precursors, including cannabigerolic acid (CBGA), orsellinic acid (OSA), and cannabigerorcinic acid (CBGOA), in Y. lipolytica. By combining modular metabolic design, chromosomal targeting knockout, 26s rDNA recombination, and the Cre-loxP system, the engineered yeast produced 3.5 mg/L CBGA, 0.5 mg/L CBGOA, and 18.8 mg/L OSA. Additionally, CBGA production reached 15.7 mg/L with olivetolic acid supplementation, the highest titer reported in oleaginous yeast. This work underscores the potential of Y. lipolytica as a versatile chassis for the synthesis of complex terpenes or polyketides. To date, there have been no reports of producing these cannabinoids and their analogs in Y. lipolytica. This study not only demonstrates the potential of Y. lipolytica as a robust host for complex biosynthetic pathways but also paves the way for future research and medicinal applications. Optimizing production yields and scaling-up the process will be crucial steps toward commercial viability.
2Materials and methods
2.1Strains, plasmids, chemicals, reagents, and primers
The Yarrowia lipolytica strains utilized in this study are outlined in Supplementary Table S1. The primers used were synthesized by GENEWIZ, Inc. in Guangzhou, China. The primers used in cloning, gene sequencing, and colony PCR are detailed in Supplementary Table S2. Chemicals, including cannabidiol (CBD-C5, CAS No. 13956-29-1), cannabidiorocol (CBD-C1, CAS No. 35482-50-9), cannabigerolic Acid (CBGA, CAS No. 25555-57-1), cannabigerorcinic acid (CBGOA, CAS No.69734-83-4), cannabidiolic acid (CBDA, CAS No. 1244-58-2), orsellinic acid (CAS No. 480-64-8), and olivetolic acid (CAS No. 491-72-5), and Ro 48–8071 (inhibitor of ERG7, CAS No. 189197-69-1), were purchased from Cayman Chemical (USA), Shanghai Macklin and Aladdin Biochemical Co., Ltd. Yeast extract powder and yeast peptone were procured from Shanghai Sangon Biotech Co., Ltd. Amino acids including leucine and uracil were sourced from Shanghai Macklin Biochemical Co., Ltd. For genetic screening, Yeast Nitrogen Base (YNB), CSM-Ura and CSM-Leu were purchased from Sunrise Science (United States). Plasmid extraction and DNA gel purification kits were purchased from TIANGEN Biotech, Beijing.
2.2Recombinant plasmid construction
The genes PpLvaE, CsOLS, CsOAC, CsPT4, NphB, ScERG20, EfmvaE, and EfmvaS were codon optimized in Y. lipolytica. These optimized genes were synthesized from Shanghai Sangon Biotech. ArmB was synthesized by GENEWIZ in Suzhou. For cloning, YaliBrick plasmid pYLXP’ was used as the backbone. This plasmid offers a versatile framework and streamlined protocol for gene cloning and multiple gene expression. The codon-optimized gene sequences can be found in previous studies [25]. Three medium-strong promoters and terminators combinations were selected to optimize gene expression: TEF-XPR2, P4-PEX20, and pTHD1-MIG1t (Supplementary Fig. S1).
We describe two examples of plasmid construction: To create the single-gene expression plasmid pYLXP'-CsOLS, pYLXP’ plasmid was first linearized using the restriction enzymes SnaBI and KpnI. Next, CsOLS gene fragment was Gibson assembled to linearized pYLXP’ [26]. Finally, pYLXP'-CsOLS was validated by sequencing at GENEWIZ, Inc. in Guangzhou, China. To create multi-gene expression plasmid pYLXP'-CsOAC-CsOLS, we employed the subcloning method following YaliBrick assembly protocol. Specifically, AvrII and SalI digested donor vector (pYLXP'-CsOLS) was gel-purified and ligated with NheI and SalI digested recipient vector (pYLXP'-CsOAC). The multiple gene expression construct was verified by restriction enzyme digestion.
2.3Gene knockout in Yarrowia lipolytica
Cre-loxP recombination was used to chromosomal gene deletion and integration in Yarrowia lipolytica [23] (see Supplementary Fig. S2) by using plasmid pUrlp-X-loci [23].
The following example outlines the knockout process for the DGA1 gene. 1,000 bp of upstream and downstream homologous arm of DGA1 gene was amplified by PCR, designated as DGA1-Up and DGA1-Dw, respectively. Then pUrlp plasmid was linearized using AvrII and SalI, the resulting loxP-Ura-loxP cassette and vector backbone were gel-purified. Next, the linearized pUrlp, DGA1-Up, loxP-Ura-loxP cassette, and DGA1-Dw were Gibson assembled, yielding the knockout plasmid pUrlp-ΔDGA1.
Then the knockout cassette containing DGA1 homologous arm and loxP-flanked Ura3 marker was obtaind by gel-purifying the AvrII and SalI digested pUrlp-ΔDGA1. The the deletion cassette was transformed into Y. lipolytica Po1g. Positive transformants were screened with colony PCR. Later the Ura3 marker can be cured by expressing Cre recombinase from pYLXP’-Cre plasmid. pYLXP’-Cre plasmid can be cured by incubating the cells in rich YPD media at 28–30 °C for 24–48 h.
2.4Genomic integration in targeted loci of the engineered pathway
To maintain genetic stability, a variety of promoters and terminators were selected to control the expression of multiple genes. Three strong promoters (pFBA1, pTDH1, and pTEF) paired with three terminators (PEX20, MIG1, and XPR2) were used. We have pre-screened a number of integration sites including YALI0A08734g, YALI0C08701g (Ku70), YALI0C01221g, YALI0C06446g(DLD2), YALI0D15246g, YALI0E27654g(POX4), YALI0E20977g(ARO8), YALI0D07986g(DGA2), YALI0E32769g (DGA1), YALI0D00789g(GGPPS), YALI0E03212g(LDH) and YALI0F26169g, as described in detail in a previous report (Supplementary Fig. S2) [23]. Gene integraiton was achieved by using homologous recombination similar to gene knock out.
2.5Yeast transformation, sample preparation and extraction
The Lithium-Acetate method was used for Y. lipolytica transformation as previously reported by Ma et al., [24]. For metabolite extraction, 400 μL of whole-cell culture was collected and homogenized using a 3D centrifugal cryogenic sample grinder (Jingxin JXCL-6K, Shanghai, China) with 400 μL of ethyl acetate (0.05 % formic acid) and glass beads (0.5 mm) at −20 °C. The grinding was performed at 21 m/s for 40 s with a 20-s interval, repeating for 40 cycles. Then 300 μL of the organic layer was collected. Two additional extractions were performed, each with 400 μL of ethyl acetate (0.05 % formic acid), and the upper organic layer was combined. The organic fraction was then evaporated using an Eppendorf Speedvac concentrator, and the dried extracts were subsequently resuspended in methanol/H2O (80/20) and vortexed for 1 min [24].
2.6HPLC and LC-MS quantification of the cannabinoid products
The samples were analyzed with an Agilent 1260 Infinity liquid chromatograph equipped with a reverse-phase C18 column (ZORBAX Eclipse Plus C18, 4.6 × 100 mm with a particle size of 3.5 μm, Agilent) and a diode array detector at 210 nm and 270 nm.
The mobile phase consisted of two solvents: solvent A (purified water with 0.05 % formic acid) and solvent B (100% chromatography-grade methanol). Solvent B was initially set at 60 %, and gradually ramped to 100 % over 20 min. Then solvent B was maintained for 4 min to wash the column, followed by a ramp of solvent B from 100 % to 60 % over 1 min, and a stabilization period of 2 min at 60 %.
The LC-MS analysis was conducted using a Thermo Scientific Orbitrap Exploris 120 mass spectrometer equipped with a heatable electrospray ionization (HESI) source. Chromatographic separation was performed on an Accucore C18 HPLC column (2.6 μm, 2.1 × 150 mm). The mobile phase consisted of two components: solvent A (0.1 % formic acid in water), and solvent B (methanol). The elution profile was programmed as follows: at the beginning, solvent B was set to 5 % and maintained for 0.5 min to establish a stable baseline. Then solvent B was gradiently increased to 95 % at 13 min, and held constant until 16.9 min. At 17 min, solvent B was decreased to 5%. The flow rate was controlled at 0.3 mL/min, and injection volume 1 μL. Mass spectrometric detection was set to capture a mass range of 100–1000 m/z, employing collision energies of 20, 50, and 80 electron volts (eV) to facilitate the fragmentation of ions and increase the specificity of detection. The identification of compounds was achieved by automatic search from comprehensive databases and spectral libraries. Notable database includes mzCloud, KEGG, ChemSpider, mzVault spectral library, and BioCyc.
2.7Confocal laser scanning microscopy
Imaging was conducted using a Zeiss LSM 980 confocal laser scanning microscope equipped with Airyscan 2 technology. Yeast cells were prepared and mounted on glass slides for optimal visualization. The Airyscan 2 detector allowed for superresolution imaging, enhancing the spatial resolution beyond the diffraction limit of light with an improved signal-to-noise ratio. The system was calibrated and optimized for various fluorophores, including Nile red, TurboGFP and mScarlet channels, to ensure accurate and reproducible results. Z-stack imaging was performed to capture three-dimensional structures, and image processing was carried out using Zeiss ZEN software, enabling analysis and visualization of subcellular components labeled with TurboGFP and mScarlet or stained with Nile red.
2.8Statistics
Triplicate experiments were performed and the reuslts were reported as mean +/- standard deviations (SDs). A two-sided t-test was used for statistical analysis, with p-values calculated using Origin 2024 or GraphPad Prism 10.1.
3Results and discussion
3.1Optimizing hexanoyl-CoA and malonyl-CoA supply to improve CBD-C5 precursor olivetolic acid
The biosynthetic pathway for CBD-C5 involves three modules: the polyketide olivetolic acid (OLA) pathway; the isoprenoid pathway, which produces geranyl pyrophosphate (GPP); and a prenyl transfer step, which synthesizes cannabigerolic acid (CBGA), the precursor to various cannabinoids (Fig. 1a). OLA is derived from one hexanoyl-CoA and three malonyl-CoA units and is catalyzed by olivetol synthase (OLS), which extends one unit of hexanoyl-CoA with three units of malonyl-CoA. The tetraketide intermediate then cyclizes to form OLA by olivetolic acid cyclase (OAC). The supply of hexanoyl-CoA is the rate-limiting step for CBD synthesis [24]. Our group previously identified the LvaE gene encoding the short-chain acyl-CoA synthetase from Pseudomonas putida KT2440 (PpLvaE) [24], which efficiently converts hexanoic acid to hexanoyl-CoA and improves OLA production. CsAAE1 from Cannabis sativa (C. sativa) has been successfully expressed in yeast to achieve a trace amount of CBDs [21]. However, our previous work revealed that CsAAE1 was not superior to PpLvaE [24]. Therefore, PpLvaE was ultimately used to facilitate hexanoyl-CoA synthesis.
Hexanoic acid concentration above 1 mM were detrimental to cell growth (Fig. 1b and c). The integration of the OLA cassette into different chromosomal loci significantly improved OLA production. Strain YX101, with the OLA cassette integrated at the YALI0C08701g (Ku70) locus, produced 0.33 mg/L OLA with 1 mM hexanoic acid (HA) supplementation (Fig. 1b). Subsequent integration of a second OLA cassette at YALI0E03212g (LDH) in strain YX102 doubled OLA production to 0.72 mg/L (Fig. 1d). Further integration at the YALI0F26169g locus (ATP-dependent Lon protease) in strain YX103, combined with delayed HA supplementation (48 h postinoculation) at 1 mM, increased production to 2.46 mg/L at 144 h (Fig. 1d and e). OLA production was confirmed by HPLC and high-resolution LC-MS (Fig. 1e, Supplementary Fig. S3).
To mitigate acetyl-CoA flow to the lipid pathway and increase malonyl-CoA availability, we further knocked out DGA1 (diglycerol acyltransferase) and overexpressed ylACC1 (Y. lipolytica native acetyl-CoA carboxylase), resulting in strain YX104 achieving an OLA titer of 6.73 mg/L, a threefold increase over YX103 (Fig. 1d and e). This result indicates that diverting acetyl-CoA flux from TAG synthesis is effective in improving the OLA level in Y. lipolytica.
3.2Optimizing the GPP supply and prenyltransferase module to improve CBGA production
The geranyl diphosphate (GPP) supply is a rate-limiting factor in terpene biosynthesis (Fig. 1). The cytosolic farnesyl pyrophosphate synthase (ERG20) acts as a bifunctional catalyst, sequentially condensing isopentenyl diphosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) to form farnesyl diphosphate (FPP). Previously, Codruta et al. utilized a mutant version of ScERG20∗F96W/N127W from S. cerevisiae that preferentially produces GPP over FPP and achieved a significant increase in monoterpene titers [27]. Herein, we used both the ylERG20∗F88W/N119W mutant from Y. lipolytica and the ScERG20∗F96W/N127W mutant from S. cerevisiae, which are presumed to perform preferential GPP production (Fig. 1a).
In addition, to increase the production of IPP and DMAPP precursors, we also leveraged the heterologous genes mvaE and mvaS from Enterococcus faecalis to replace the native ERG10, ERG13, and tHMGR genes (Fig. 1) [28]. These enzymes catalyze the first three steps of the MVA pathway and have proven high efficiency in S. cerevisiae [28]. We integrated EfmvaE, EfmvaS, and ylERG20∗F88W/N119W at the YALI0A08734g locus, and ScERG20∗F96W/N127W at the geranylgeranyl pyrophosphate synthase (GGPPS) locus, and these strategies increased GPP flux in the peroxisome. Combined with prenyltransferase (PTase) integration, we evaluated the performance of the strains in CBGA production (Fig. 1, Fig. 3d).
Prenyltransferase (PTase) is another rate-limiting step in the CBGA pathway (Fig. 1a). The PTase from Cannabis sativa (CsPT4) has been reported to play a crucial role in CBGA biosynthesis. The first 77 amino acids coding for chloroplast-targeting signal were removed for better activity [21]. With this approach, we engineered a truncated CsPT4 devoid of the first 77 amino acids, named tCsPT4. Plant enzymes with truncated chloroplast-targeting sequences may present in the cytoplasm or endoplasmic reticulum (ER) of yeast cells. To determine the exact cellular localization of tCsPT4 in Y. lipolytica, we applied confocal microscopy to identify its spatial distribution.
As an ER-positive control, we fused TurboGFP to the C-terminus of two ER retention signals: ER1 or ER2 (Supplementary Fig. S4). These tags are known to localize enzymes to the ER lumen or the cytoplasmic face of the ER membrane in S. cerevisiae [29]. Under confocal fluorescence microscopy, our positive control TurboGFP-ER1/ER2 was successfully localized to the intracellular compartment, including the endoplasmic reticulum (Fig. 2a). For tCsPT4, we fused its C-terminus with mScarlet and coexpressed TurboGFP-ER1/ER2 and tCsPT4-mScarlet in the same strain (Fig. 2b, Supplementary Fig. S4). Merged confocal images revealed that tCsPT4 was predominantly localized on the expanded ER surface (Fig. 2b).
The peroxisomal β-oxidation pathway can directly supply acetyl-CoA from fatty acids degradation. The peroxisomal localization of specialized plant pathways has previously enabled efficient squalene and alkaloid production [30,31]. We next attempted to localize tCsPT4 to peroxisomes. We have validated that the SKL signal successfully directed proteins to the peroxisomes of Y. lipolytica (Fig. 2C).
We fused SKL with tCsPT4 and integrated this engineered PT4 at YALI0C01221g (encoding 26S proteasome non-ATPase regulatory subunit 10), resulting in strains YX105 (without PTS) and YX106 (with PTS) (Fig. 3a). Strain YX105, without the PTS signal, produced 7.77 mg/L OLA and 143 μg/L CBGA (Fig. 3a). In contrast, YX106, with a PTS signal, produced 5.56 mg/L OLA without detectable CBGA (Fig. 3a). These findings suggest that PTase peroxisome targeting is not suitable for producing OLA or CBGA. Next, we increased the GPP flux in YX105 by integrating EfmvaE, EfmvaS, and ylERG20∗F88W/N119W. The engineered YX107 produced 3.39 mg/L OLA and 335 μg/L CBGA (Fig. 3a, Supplementary Fig. S5), representing a 2.34-fold increase compared with those produced by YX105.
As PTase remains a bottleneck, we hypothesized that increasing the copy number of tCsPT4 would increase CBGA production. Y. lipolytica contains at least 200 copies of ribosomal DNA (rDNA) clusters [32]. These rDNAs can be leveraged for random integration of heterologous pathways (Fig. 3b) [33]. In order to pull away the GPP flux toward CBGA synthesis, we increased the copy number of tCsPT4 with the 26S rDNA and Cre-loxP Systems (Fig. 3b). The engineered strain YX108, produced 3.35 mg/L OLA and 974 μg/L CBGA (Fig. 3a,b,3c), representing a 3-fold increase in CBGA compared with that produced by YX107.
3.3Condensed phase vesicle-like dual PTase expression improved CBGA production
In strain YX108, despite increasing the GPP flux and the copy numbers of tCsPT4, CBGA production remained below 1 mg/L. We hypothesized that low CBGA production might be due to suboptimal catalytic activity of tCsPT4 or pathway incompatibility with the host organism. To address this challenge, we chose to co-express a different PTase, NphB from Streptomyces sp., which has complementary prenyltransferase activity, to improve CBGA production [[34], [35], [36]]. NphB is a soluble aromatic PTase initially characterized by Kuzuyama et al. [37]. Studies on mutant versions of NphB, specifically Y288A and G286S, have shown improved catalytic specificity [35]. We analyzed the subcellular distribution of NphB∗Y288A/G286S in Y. lipolytica with confocal fluorescence microscopy and confirmed NphB∗Y288A/G286S localized to expanded ER-like cell vesicles, similar to tCsPT4 (Fig. 2b). We subsequently integrated NphB∗Y288A/G286S into the chromosomal locus YALI0E20977g (ARO8) of YX108, generating strain YX109. This led to the production of 1.73 mg/L CBGA, which is almost 2-fold increase compared with that of YX108 (Fig. 3d).
As the endoplasmic reticulum (ER) is the location where OLA synthesis takes place (Fig. 1a), Zhang et al. reported that localizing PTase onto the ER surface can improve CBGA production in S. cerevisiae [18]. To improve the catalytic efficiency of tCsPT4 and NphB, we fused an ER signal tag (validated previously) to the N-terminus of both enzymes. We integrated NphB∗Y288A/G286S-ER, or tCsPT4-ER into the chromosomal locus ARO8 of YX108, generating strains YX110, and YX111. YX110 produced 1.45 mg/L CBGA, while YX111 produced 0.93 mg/L CBGA (Fig. 3d), indicating that ER targeting does not promote CBGA synthesis.
Notably, YX109, which expresses NphB∗Y288A/G286S without the ER tag, achieved the highest CBGA production (1.73 mg/L), representing a 12-fold increase compared with that of the original YX105 strain (Fig. 3d). Furthermore, after increasing the GPP flux in YX109 by replacing endogenous GGPPS with ScERG20∗F96W/N127W, we obtained strain YX112, which produces 2.85 mg/L CBGA, the highest titer compared with that of the previous strains (Fig. 3d).
Due to the fact that PTases, tCsPT4 and NphB∗Y288A/G286S, are spatially assembled together forming a condensed phase vesicle-like structure (Fig. 2b), we assume that dual PTase coexpression may be similar to liquid-liquid phase separation induced biomolecular condensates [38]. This condensed vesicle-like PTase expression significantly enhances CBGA production (2.85 mg/L) in Y. lipolytica. While the OLA precursor remains a bottleneck in the chassis strain (OLA titer less than 10 mg/L), we next probed the potential of the condensate-like PTase, by feeding 1 mM OLA to the strains YX108, YX109, YX110, YX111 and YX112 (Fig. 3e). Furthermore, we applied an ERG7 inhibitor (Ro 48–8071 fumarate which inhibits oxidosqualene cyclase) to increase the level of GPP precursor [39]. Surprisingly, strains YX109plus and YX112plus with condensed phase vesicle-like dual PTase expression, achieved the highest CBGA production (11.48 mg/L and 15.77 mg/L, respectively) (Fig. 3e). Notably, YX112 showed the highest increase, producing over 15.77 mg/L CBGA with an approximately 39 % conversion ratio from OLA to CBGA (Fig. 3e). Machine-learning based protein sequence analysis [40] indicates that tCsPT4 and NphB will form a biomolecular condensate at a very high probability.
This result demonstrated that condensed-phase vesicle-like PTase expression- significantly increased CBGA production in Y. lipolytica, which also reinforced that the OLA precursor is currently the rate-limiting step for further improving CBGA yield in Y. lipolytica.
3.4Biosynthesis of cannabidiorcol precursor orsellinic acid
Cannabidiorcol (CBD-C1 or O-1821) is an underexplored minor cannabinoid derived from C. sativa that is distinguished by a shortened methyl sidechain (C1) instead of the pentyl group seen in CBD-C5. Like CBD-C5, it is nonpsychoactive. However, CBD-C1 remains less studied than other cannabinoids, such as THC and CBD-C5, although its unique structure offers potential for therapeutic applications [41]. Cannabigerorcinic acid (CBGOA) serves as the precursor for methyl-chain cannabinoids such as CBD-C1, much like CBGA does for pentyl cannabinoids [42,43]. The precursor of CBGOA, orsellinic acid (OSA), is a dihydroxybenzoic acid with antioxidant and neuroprotective properties, commonly found in lichens as a fungal metabolite [43,44]. Therefore, engineering heterologously produced CBD-C1 analogs and their precursor OSA holds promise for overcoming the supply limitations and expanding their medicinal applications.
Compared with the OLA pathway, OSA synthesis does not require hexanoyl-CoA, making the process faster and more efficient (Fig. 4a). We identified an iterative type I polyketide synthase (PKS), orsellinic acid synthase (ArmB), from Armillaria mellea [45]. To date, there have been no reports of ArmB being used for heterologous expressions in yeast or E. coli. ArmB contains multiple domains, including ACP transacylase (SAT), ketosynthase (KS), malonyl-CoA transacylase (MAT), product template domain (PT), acyl-carrier protein (ACP) and thioesterase domain (TE) (Fig. 4b), with a MW approximately 238 kDa [46,47].
We next integrated this codon-optimized ArmB into the Ku70 locus in Y. lipolytica. Strain YX201 produced 2.19 mg/L OSA, which was validated by HPLC and high-resolution LC-MS (Fig. 5a and b, Supplementary Fig. S6).
Since OSA synthesis follows a Claisen condensation with three malonyl-CoAs as the extending units, increasing malonyl-CoA levels is crucial. We integrated ylACC1 gene cassettes into either the DGA1 or DGA2 locus of YX201 to increase the malonyl-CoA supply, creating strains YX202 and YX203. In the test tubes, YX202 produced 7.19 mg/L OSA, and YX203 produced 3.90 mg/L OSA (Fig. 5a and b), a more than 3-fold increase compared with that of strain YX201.
In the shake flask, we observed that the culture pH decreased to 4, potentially impairing membrane permeability and strain performance. To address this, we applied PBS buffer or CaCO3 and adjusted the agitation rates when cultivating YX202. CaCO3 effectively stabilized the pH, and the OSA titer peaked at 5.278 mg/L at 72 h and 150 rpm. However, OSA began to degrade after 72 h in the presence of CaCO3 (Fig. 5d–g), possibly due to endogenous enzyme activity. PBS was ineffective at both 150 rpm and 250 rpm. The YPD control group showed steady OSA production, exceeding 4 mg/L, despite pH fluctuations. Agitation at 150 rpm was more conducive to strain growth. The strain in YPD medium produced 4.23 mg/L OSA at 150 rpm (Fig. 5d–g). This result indicates that the CBD-C1 precursor OSA gene cluster can be functionally reconstituted in Y. lipolytica.
3.5Production of CBD-C1 analog cannabigerorcinic acid (CBGOA)
The downstream module of the CBD-C1 biosynthesis pathway closely resembles that of CBD-C5, with PTase transferring a GPP moiety to the OSA side chain (Fig. 4a). To facilitate CBGOA production, we employed the same strategy used for CBGA production. Specifically, we integrated EfmvaE, EfmvaS, and ylERG20∗F88W/N119W into the YALI0A08734g locus of strain YX202, resulting in strain YX204 (Fig. 5c,h, Supplementary Fig. S7).
Single and dual PTase expression were also performed in YX204. Similar to the CBGA results, the highest CBGOA production (409 μg/L) was observed in YX206 which was engineered to harbor a condensate-like dual PTase expression system (Fig. 2, Fig. 5h). Dual PTase was less effective for CBGOA than for CBGA, despite using the same enzymes. This discrepancy is likely due to the limited supply of OSA and the subtle structural difference in the substrate, as variations in the side chain geometry and length alter the position of the substrate-GPP alignment within the catalytic pocket, thereby shifting the catalytic efficiency. Thus, further AI-assisted engineering of the PTase or mining of novel PTases that prefer OSA as a substrate is critical to enhancing its performance and delivering more potent CBD-C1 analogs.
3.6Fed-batch cultivation in shake flasks
To optimize the production of the YX112 and YX206 engineered strains, we conducted fed-batch cultivation in flat shake flasks (Fig. 6a and b). For strain YX112, the medium was supplemented with 1 mM hexanoic acid on Day 1. The OD600 exhibited steady growth, reaching approximately 52 by Day 6. The olivetolic acid concentration increased continuously, peaking on Day 4 at approximately 9.84 mg/L, after which it decreased due to its conversion to CBGA. By Day 6, the olivetolic acid titer stabilized at 7.64 mg/L. CBGA production began accumulating on Day 2, with a final titer of approximately 3.537 mg/L by Day 6 (Fig. 6a).
For strain YX206, glucose served as the sole substrate for orsellinic acid (OSA) and CBGOA production. The glucose concentration was maintained at no less than 30 g/L throughout the cultivation period. Although the OD600 increased steadily, it progressed at a slower rate than YX112 did, reaching approximately 28% by Day 5. Orsellinic acid production continuously increased, with a final titer of 18.87 mg/L. However, CBGOA production was less efficient than CBGA production was, resulting in a final titer of approximately 541 μg/L by Day 5 (Fig. 6b).
4Conclusions
In this study, we successfully engineered the nonconventional oleaginous yeast Y. lipolytica for the de novo biosynthesis of CBD and its analogs, including olivetolic acid (OLA), cannabigerolic acid (CBGA), orsellinic acid (OSA), and cannabigerorcinic acid (CBGOA). By optimizing the polyketide pathway for olivetolic acid (OLA) production and boosting geranyl pyrophosphate (GPP) precursor, we overcame the pathway bottlenecks and achieved successful microbial synthesis of cannabinoids. The strategic engineering of prenyltransferases, in particular, we observed a condensate-like expression of dual PTase in the engineered strain, which significantly improved the CBGA titer to 15.7 mg/L with OLA supplementation.
Our work represents the first reported production of these cannabinoids and their analogs in Y. lipolytica, highlighting the yeast's potential as a versatile chassis for the synthesis of complex terpenes and polyketides. The remarkable titer improvement of CBD analogs demonstrates the effectiveness of our metabolic engineering strategies. These findings not only demonstrate the feasibility of producing medically relevant cannabinoids in a microbial host but also lay the groundwork for future efforts to optimize production yields and scalability.
Three primary bottlenecks may be limiting cannabinoid titers in Y. lipolytica: (1) insufficient precursor supply due to competing fluxes of malonyl-CoA and acetyl-CoA from the endogenous lipid pathway, (2) suboptimal catalytic efficiency and localization of PTase, and (3) increased concentrations of hexanoic acid or olivetolic acid may disrupt cell membrane or cause the cell lose the proton motive force. To address these challenges, we propose the design of novel hexanoyl-CoA and malonyl-CoA pathways to increase precursor availability. Additionally, enzyme engineering strategies, such as the directed evolution of PTase to increase activity and substrate specificity, could further improve prenylation efficiency. Finally, adaptive laboratory evolution (ALE) may increase strain tolerance to hexanoic acid and redirect carbon flux toward cannabinoid biosynthesis. Resolving these limitations will be critical in advancing from laboratory-scale proof-of-concept to an industrially viable bioprocess.
Our study contributes to the development of sustainable and scalable methods for cannabinoid production in Y. lipolytica, with potential applications in pharmaceuticals and therapeutics. The approaches described here can also be extended to the biosynthesis of other complex natural products, paving the way for innovative applications in biotechnology and medicine.
Data availability
All data from the current study is available from the corresponding author upon request.
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 ASupplementary data
The following is/are the supplementary data to this article:
Acknowledgments
We thank the Core Facility team of the Department of Biotechnology and Food Engineers, Guangdong-Israel Institute of Technology for their service and technical support. The funding is supported by the 10.13039/501100001809National Natural Science Foundation of China (general grant 22378083) and the Li Ka-shing Foundation (EN2400022) and the Muyuan Laboratory (Program ID: 12106022401).