Introduction

Prostaglandins are a class of important lipid mediators that exhibit diverse biological activities in humans and animals. Based on the different carbon structures, prostaglandins, including prostacyclopentanes and thromboxanes, can be categorized into types A, B, C, D, E, F, G, H and I1. Among these, prostaglandin F2α (PGF2α), a type of prostacyclopentane, demonstrates a wide range of biological activities2,3. Moreover, PGF2α derivatives, such as carboprost tromethamine and latanoprost, possess various physiological functions, which are used clinically to treat conditions such as postpartum hemorrhage in pregnant and glaucoma1,4.

The initial chemical synthesis of PGF2α, achieved in 1969 via a 20-step reaction, laid the foundation for its production5. Subsequently, the more researches has focused on improving the efficiency of this chemical synthesis process6,7. To date, the chemical synthesis remains the primary method for the industrial production of PGF2α. However, this approach has several limitations, such as complex reaction steps, high costs, and adverse environmental impacts. Focusing on solving these problems, Chen et al. successfully developed a method that integrates enzyme-catalyzed and chemical-catalyzed for the synthesis of PGF2α8. Despite this advancement, the modified chemical synthesis method still presents the certain drawbacks.

PGF2α is naturally produced by mammals and certain other organisms, such as Gracilaria vermiculophylla and Candida albicans. In this biosynthetic pathway, free arachidonic acid (ARA) serves as the direct precursor, which is first synthesized from membrane phospholipids by the action of phospholipase A2. As illustrated in Fig.1, free ARA is subsequently converted to PGF2α by a cascade enzyme system harboring prostaglandin H synthase (PGHS) and prostaglandin F synthase (PGFS)9. However, the natural yield of PGF2α from these organisms is low, making it insufficient to meet market demand.

Fig. 1: PGF2α synthesis via cell-free metabolic engineering platform.
Fig. 1: PGF2α synthesis via cell-free metabolic engineering platform.The alternative text for this image may have been generated using AI.
Full size image

A Diagram of cell-free PGF2α biosynthesis. The gray schematic (created in Biorender) illustrates the biosynthetic pathway for the conversion of ARA to PGF2α. Molecular structures of ARA, PGG₂, PGH₂, and PGF2α are displayed. The PGHS catalyzes the conversion of ARA to PGH₂ utilizing heme, O₂ and NADPH. The PGFS converts PGH₂ to PGF2α with NADH. B PGF2α production via cell-free platform from ARA. n = 3. C LC-MS/MS MRM chromatogram (EIC) of PGF2α showing the quantitative transition (m/z 353 > 309.2). The standard and the synthesized PGF2α are shown in the top and bottom panels of Fig. 1C, respectively.

Advances in metabolic engineering have enabled the production of PGF2α in genetically modified plants and microorganisms. In 2009, Mohamed et al. engineered Saccharomyces cerevisiae to produce PGF2α at 1927 ng/g dry cell weight (DCW) via introducing the PGHS-encoding gene from Mus musculus (MmPGHS) and the PGFS-encoding gene from Trypanosoma brucei (TbPGFS), with production achieved by feeding 10 μM ARA substrate10. Subsequently, the same group achieved de novo production of PGF2α in Arabidopsis thaliana at 0.76 ng/g wet cell weight (WCW) by co-expressing ∆-9 elongase, ∆-8 desaturase, ∆-5 desaturase, along with the MmPGHS and TbPGFS11. Beyond these systems, other chassis organisms including Marchantia polymorpha12, Escherichia coli13, and Fistulifera solaris14 have also been engineered for PGF2α synthesis. Nevertheless, these approaches face the challenges for scalable production, including slow plant growth, low production titer, and the added cost of ARA supplementation.

Yarrowia lipolytica has emerged as a promising industrial chassis for producing diverse value-added bioproducts, owing to its unique metabolic characteristics15,16. The key advantage of Y. lipolytica as chassis is its inherent abundance of acetyl-CoA, a central precursor for biosynthesizing a variety of derived chemicals17. Taking this advantage into account, our group previously engineered a strain of Y. lipolytica capable of de novo ARA production, achieving a titer of 118.1 mg/L ARA18. Nevertheless, the de novo biosynthesis of PGF2α from glucose in engineered Y. lipolytica remains unreported.

This study demonstrates the successful engineering of Y. lipolytica for PGF2α biosynthesis. Initial investigation involved the development of both in vitro cell-free and whole-cell catalytic systems to verify the biosynthetic pathway. Subsequently, systematic metabolic engineering enabled de novo production from glucose, achieving a PGF2α titer of 32.63 μg/L (approximately 3.58 μg/g DCW). To our knowledge, this work reports the highest titer of de novo PGF2α in engineered Y. lipolytica using glucose, thereby establishing a pioneering platform for the green and sustainable manufacturing of PGF2α and other prostaglandins.

Results

Biosynthesis of PGF2α by cell-free metabolic engineering

The characteristics of the PGF2α biosynthetic pathway have been summarized in detail19. To investigate the in vitro biotransformation of ARA to PGF2α, a cell-free platform utilizing Y. lipolytica lysates harboring PGHS and PGFS was established for PGF2α biosynthesis over a 30-min period (Fig. 1A). The 30-min reaction duration was selected based on observed plateauing of product formation and consideration of lysate stability limitations. In this study, the PGHS-encoding gene from Mus musculus (MmPGHS) and the PGFS-encoding gene from Trypanosoma brucei (TbPGFS) were codon optimized and coexpressed in Y. lipolytica Po1f (∆ku70) to construct the engineered strain PG01.

To optimize the cell-free system, in addition to adding ARA and Y. lipolytica lysates, hematin was added to increase the oxygen supply, and tryptophan was added to improve the PGFS activity. Cell-free reactions employed lysates from engineered Y. lipolytica strains expressing PGHS or PGFS. While individual enzyme concentrations were not quantified due to technical limitations, standardized preparation protocols ensured reproducibility. The cell-free transformation platform enabled 0.53 μg/L PGF2α production by strain PG01 via the addition of 100 μM ARA, 2 μM hematin, and 5 μM tryptophan (Fig. 1B). The cell-free system, utilizing lysates from the base strain PG01, successfully established a baseline for the catalytic competence of the enzyme pair, confirming that the pathway was functionally assembled. Herein, it is noteworthy that the primary objective was to characterize the enzyme activity directly. Therefore, lysates from later, optimized strains were not employed, as any performance improvement would be more appropriately attributed to and evaluated in the context of whole-cell metabolism. However, no synthesized PGF2α was detected in strain PG01 in the absence of exogenous ARA. To confirm the biosynthetic capability, the cell-free platform employing Y. lipolytica lysates was analyzed, and the production of PGF2α was successfully verified by LC-MS (Fig. 1C). Subsequently, the same cell-free platform with E. coli lysates was evaluated, yielding a PGF2α titer of 101 ng/mL (Supplementary Fig. 1).

Production of PGF2α by whole-cell catalysis

To verify whether the engineered Y. lipolytica is capable of producing PGF2α, the PGF2α biosynthetic pathway harboring MmPGHS and TbPGFS was constructed in Y. lipolytica (Fig. 2A).

Fig. 2: Whole-cell catalysis for PGF2α production in Y. lipolytica.
Fig. 2: Whole-cell catalysis for PGF2α production in Y. lipolytica.The alternative text for this image may have been generated using AI.
Full size image

A Diagram of PGF2α synthesis from ARA via whole-cell catalysis system. B PGF2α production in the original and engineered strains. C PGF2α production in the engineered strain PG02 under different ARA concentrations (nmol/L). D PGF2α production in the engineered strain PG02 grown on 100 μM ARA under time-course fermentation. Pairwise comparisons were analyzed by Student’s t-test. ***P < 0.001. n = 3.

The engineered strain PG01 was able to produce 0.28 μg/L PGF2α under the supplementation with 100 μM ARA (Fig. 2B). To enhance enzyme-host compatibility, the engineered strain PG02, was constructed by expressing GvPGHS from Gracilaria vermiculophylla alongside TbPGFS. This modification drastically increased PGF2α production to 23.80 μg/L under the same conditions, a significant improvement over PG01 (P < 0.001), indicating the superior efficiency of GvPGHS over MmPGHS in converting ARA. Notably, the parent strain Po1f (Δku70) produced only a basal level of 0.15 μg/L. These results collectively demonstrate the successful production of PGF2α in engineered Y. lipolytica strains equipped with the heterologous biosynthetic pathway and ARA supplementation. While extracellular PGHS/PGFS activity was not quantified in this work, future studies may explore extracellular secretion dynamics under engineered conditions.

The fermentation characteristics of engineered strain PG02 were further evaluated by investigating the effects of ARA concentration and fermentation time on PGF2α production. The titer of PGF2α produced by the strain PG02 was improved with the concentration of supplemented ARA, reaching a maximum of 41.84 μg/L at 200 μM. The highest conversion rate from ARA to PGF2α, however, was 0.067%, achieved at 100 μM ARA. Meanwhile, the PGF2α titer was also found to increase with fermentation time (Fig. 2C). Under the optimal condition of 100 μM ARA and 48 h of fermentation, strain PG02 produced 22.11 μg/L PGF2α (Fig. 2D).

Engineered Y. lipolytica for de novo PGF2α synthesis

Based on the above results of PGF2α production using Y. lipolytica via a cell-free system and whole-cell catalysis, we further designed and constructed the PGF2α biosynthetic pathway in Y. lipolytica using glucose as the sole carbon source.

Considering free ARA as a key intermediate metabolite, the PGF2α biosynthetic pathway was divided into the ARA module and PGF2α synthesis module (Fig. 3A). To construct an engineered Y. lipolytica strain that can produce high-level free ARA, the free ARA biosynthetic pathway, harboring ∆-9 elongase and ∆-8 desaturase from Isochrysis galbana20, ∆-5 desaturase from Mortierella alpina21 and Acot7 from Mus musculus, was introduced into the Po1f (∆ku70), resulting in the engineered strain PG03. Herein, it was shown that ARA is successfully synthesized in strain PG03 by GC analysis (Supplementary Table 1). To enhance the metabolic flux for ARA biosynthesis via improving the supply of linoleic acid (C18:2), ∆-12 desaturase from Y. lipolytica22 and Fusarium moniliforme23, converting oleic acid (C18:1) to linoleic acid, was overexpressed in PG03 to construct the strain PG03-1 and PG03-2, respectively. The intermediate metabolites of linoleic acid and eicosadienoic acid (C20: 2) were increased in the strain PG03-1 and PG03-2, respectively (Supplementary Table 1). However, ARA was not observed in the strains PG03-1 and PG03-2. It was supposed that ∆-8 desaturase was the rate-limiting step in the ARA biosynthetic pathway. Moreover, the PGF2α synthesis module, harboring GvPGHS and TbPGFS, was introduced into strain PG03 to construct strain PG04. Notably, PGF2α at 290.49 ng/L was successfully de novo produced by strain PG04 using glucose, which was significantly higher than the Po1f (∆ku70) (P < 0.001) (Fig. 3B).

Fig. 3: De novo production of PGF2α in Y. lipolytica.
Fig. 3: De novo production of PGF2α in Y. lipolytica.The alternative text for this image may have been generated using AI.
Full size image

A Diagram of the PGF2α biosynthetic pathway in Y. lipolytica using glucose. B PGF2α production by the engineered Y. lipolytica strains harboring ∆-8 desaturase from different organisms. Pairwise comparisons are analyzed by Student’s t-test. ***P < 0.001. n = 3.

To further optimize the catalytic efficiency of ∆-8 desaturase for increasing the production level of PGF2α, the other three ∆-8 desaturases, including ∆-8 desaturase from Pavlova salina21, EFD1 and EFD2 from Euglena gracilis24, were selected to reconstruct the PGF2α biosynthetic pathway, resulting in the strain PG05, PG06 and PG07, respectively. However, the titer of PGF2α produced by the above engineered strains PG05, PG06, and PG07 was notably lower than in the strain PG04, respectively (Fig. 3B). To explain this result, ARA content in engineered PG04, PG05, PG06, and PG07 was determined. ARA with a content of 0.55% was detected in strain PG04, however, it was not found in strains PG05, PG06, and PG07 (Supplementary Fig. 2). Therefore, we supposed that ∆-8 desaturase from I. galbana performs the higher catalytic efficiency.

Improved PGF2α production by the combinatorial fusion enzyme and organelle engineering

To further improve the production titer of PGF2α, the combinatorial metabolic engineering strategies were applied to regulate the metabolic flux in Y. lipolytica grown on glucose. ARA synthesis module was designed via enzyme fusion strategy by protein linker 2 A or GGGGS to improve the catalytic efficiency of ∆-8 desaturase and ∆-9 elongase. Herein, two engineered strains PG08 (harboring ∆9E-2A-∆8D) and PG09 (harboring ∆9E-GGGGS-∆8D) were constructed. Compared with strain PG04, strain PG08 achieved the decreased production titer of PGF2α at 101.28 ng/L. However, the PGF2α titer in the strain PG09 was significantly increased to 563.94 ng/L with an increase of 94.1%, indicating that the fusion of ∆-9 elongase and ∆-8 desaturase using linker GGGGS was capable of optimizing the substrate channeling for linoleic acid transformation (Fig. 4A).

Fig. 4: Compartmentalization engineering enhances PGF2α production in Y. lipolytica.
Fig. 4: Compartmentalization engineering enhances PGF2α production in Y. lipolytica.The alternative text for this image may have been generated using AI.
Full size image

A PGF2α production in different engineered Y. lipolytica strains. n = 3. B Subcellular localization of RFP in Y. lipolytica Po1f (∆ku70). C Subcellular localization of GvPGHS in Y. lipolytica Po1f (∆ku70). D Subcellular localization of TbPGFS in Y. lipolytica Po1f (∆ku70). E Subcellular localization of RFP-KDEL in Y. lipolytica Po1f (∆ku70). F Subcellular localization of RFP-OLE in Y. lipolytica Po1f (∆ku70). Pairwise comparisons were analyzed by Student’s t-test. **P < 0.01, ***P < 0.001.

In addition, we speculate that the heterogeneously expressed GvPGHS was localized to mitochondria and TbPGFS was localized to cytosol in Y. lipolytica by the PSORT II, respectively (Supplementary Fig. 3). To test the hypothesis that GvPGHS and TbPGFS are located in the putative organelles in Y. lipolytica, the red fluorescent protein (RFP) was fused to the C-terminus of GvPGHS or TbPGFS and the fusion proteins were expressed in the strain Po1f (∆ku70). Results from laser scanning confocal microscopy (LSCM) suggest that RFP in strain Po1f (∆ku70) is distributed in the cytosol (Fig. 4B) and provide preliminary evidence that GvPGHS-RFP is localized in the mitochondria (Fig. 4C) while TbPGFS-RFP is distributed in the cytosol (Fig. 4D). However, comparison with marker proteins that localize to these organelles would be needed to confirm subcellular localization of the proteins. Based on the above results of GvPGHS and TbPGFS location in Y. lipolytica, the organelle engineering strategy was applied for rewiring the metabolic flux and optimizing the conversion efficiency of free ARA to PGF2α. Herein, RFP-KDEL (RFP fused with the endoplasmic reticulum-targeting signal KDEL) and RFP-OLE (RFP fused with the lipid droplet-targeting signal OLE) were constructed and introduced into the strain Po1f (∆ku70), respectively. It was observed that RFP-KDEL localized to the endoplasmic reticulum (Fig. 4E), while RFP-OLE was targeted to lipid droplets (Fig. 4F). Given the metabolic characteristics of the PGF2α biosynthetic pathway, the PGF2α synthesis module (PGHS-PGFS) was relocated into the endoplasmic reticulum by fusing the KEDL signal to generate the strain PG10 (PGHS-KDEL-PGFS-KDEL). Meanwhile, the PGF2α synthesis module was relocated into the lipid droplet by fusing the OLE signal to generate the strain PG11 (PGHS-OLE-PGFS-OLE). The strains PG10 and PG11 produced 371.07 ng/L and 420.90 ng/L of PGF2α, respectively (Fig. 4A). The PGF2α titer of strains PG10 and PG11 was improved in comparison to the strain PG04, respectively. Especially, the strain PG11, harboring PGF2α synthesis module localized into lipid droplet, performed the significant production titer of PGF2α.

To maximize the metabolic flux towards PGF2α synthesis, we redesigned the PGF2α biosynthetic pathway in Y. lipolytica to construct the strain PG12, harboring linker GGGGS-mediated ARA synthesis module and PGF2α synthesis module localized in lipid droplet. The PGF2α titer of strain PG12 was improved to 792.38 ng/L after fermentation for 72 h (Fig. 4F). Moreover, we constructed the strain PG13 via the dual regulation of PGF2α synthesis module by the endoplasmic reticulum and lipid droplet engineering. The PGF2α titer of strain PG13 was further enhanced to 958.78 ng/L (P < 0.01), which was markedly higher than strain PG12 (Fig. 4F). The results suggested that the dual regulation of endoplasmic reticulum and lipid droplet engineering is an effective approach for boosting PGF2α production in Y. lipolytica.

Cofactor engineering enhances PGF2α production

Cofactors, such as NADPH, ATP, and acetyl-CoA, are involved in regulating the metabolic flux during microbial production of chemicals. The PGF2α biosynthetic pathway requires NADPH and acetyl-CoA. To test the potential of PGF2α synthesis in the engineered Y. lipolytica, we focused on engineering the pentose phosphate pathway (PPP) for increasing NADPH supply in the strain PG13, which was the primary source of NADPH in Y. lipolytica (Fig. 5A). This approach is supported by the work of Qiao et al.25, who demonstrated that coordinated overexpression of PPP genes (glucose-6-phosphate dehydrogenase (G6PD) and 6-phosphate gluconate dehydrogenase (PGD)) in Y. lipolytica significantly enhanced NADPH supply and lipid production. Herein, G6PD and PGD involved in the PPP of Y. lipolytica were co-overexpressed in the strain PG13 to generate the strain PG14. However, compared with the strain PG13 capable of producing 958.78 ng/L PGF2α, the strain PG14 produced a decreased titer of PGF2α at 941.53 ng/L (Fig. 5B).

Fig. 5: Cofactor engineering improved PGF2α production in Y. lipolytica.
Fig. 5: Cofactor engineering improved PGF2α production in Y. lipolytica.The alternative text for this image may have been generated using AI.
Full size image

A Diagram of cofactors for improving PGF2α synthesis in the engineered Y. lipolytica. The colored green lines represent the overexpressed genes for enhancing acetyl-CoA supply, and the colored blue lines represent the overexpressed genes for enhancing NADPH supply. B PGF2α production in engineered different Y. lipolytica strains. C NADPH levels in engineered different Y. lipolytica strains. D Acetyl-CoA levels in engineered different Y. lipolytica strains. Pairwise comparisons were analyzed by Student’s t-test. *P < 0.05, **P < 0.01. n = 3.

To explain the negative effect of NADPH improvement on PGF2α production in the strain PG14, NADPH content in the strains PG13, PG14, and Po1f (∆ku70) was analyzed, respectively. Compared with the strain Po1f (∆ku70), the strain PG14 synthesized a higher content of NADPH (1978.12 nmol/g vs 1426.28 nmol/g), meanwhile, the strain PG13 synthesized the decreased content of NADPH with 1185.09 nmol/g (Fig. 5C). Given the production capacity of PGF2α in both the engineered strains PG13 and PG14, we supposed that the synthesis of PGF2α requires NADPH, however, the excess supply of NADPH may limit PGF2α production in the engineered Y. lipolytica.

Acetyl-CoA, as an indispensable metabolic cofactor, is viewed as a precursor for PGF2α biosynthesis. Therefore, we next assessed whether the enhanced acetyl-CoA supply can boost PGF2α production (Fig. 5A). Herein, the ATP-citrate lyase (ACL2) from Y. lipolytica and carnitine O-acetyltransferase (Cat2) from S. cerevisiae were overexpressed in strain PG13, respectively. When overexpression of ACL2 was introduced in strain PG13, the resulting engineered strain PG15 performed a similar titer of PGF2α, compared with the strain PG13 (955.03 ng/L vs 958.78 ng/L). Cat2 overexpression is viewed as an effective strategy for facilitating the conversion of lipid-based feedstocks into cytosolic acetyl-CoA pools26. Surprisingly, compared with strain P13, the engineered strain PG16 harboring Cat2 overexpression performed a significant increase of PGF2α titer at 1955.05 ng/L (Fig. 5B). However, co-overexpression Cat2 and ACL2 in the strain PG13 led to the production of PGF2α at 1482.09 ng/L in the engineered strain PG17 (Fig. 5B).

To perform the effect of acetyl-CoA supply on enhancing PGF2α production, the content of acetyl-CoA in the strains Po1f (∆ku70), PG13, PG15, PG16, and PG17 was detected, respectively. The contents of acetyl-CoA in the above five engineered strains were 1705.04 pg/g DCW, 694.64 pg/g DCW, 667.96 pg/g DCW, 870.61 pg/g DCW, and 797.37 pg/g DCW, respectively (Fig. 5D). The results indicated that the PGF2α biosynthetic pathway requires the amount of acetyl-CoA and the production titer of PGF2α can be improved by enhancing the acetyl-CoA supply. Notably, the significant increase of PGF2α production in the strain PG16 was related to the enhanced acetyl-CoA supply by overexpressing Cat2. However, overexpression of ACL2 failed to enable the increase of acetyl-CoA supply and PGF2α production in the strain PG15.

Improving PGF2α production by medium engineering

Naturally, the fermentation performance of Y. lipolytica is influenced by the culture conditions. To evaluate the production potential of PGF2α by the engineered strain PG16, the different fermented conditions, including C/N ratio, 5-aminolevulinic acid hydrochloride (ALA), and oleic acid (C18:1, OA) addition, were considered for increasing PGF2α production.

We evaluated the effect of different C/N ratios on PGF2α synthesis in the strain PG16 (Fig. 6A). When the strain PG16 was fermented with C/N ratios of 10, 20, 40, 60, 80, 100, and 120, the production titer of PGF2α was 3.85 μg/L, 7.90 μg/L, 9.57 μg/L, 10.04 μg/L, 10.30 μg/L, and 10.52 μg/L, respectively. The results demonstrate that changing C/N ratios regulate PGF2α synthesis in the strain PG16. Especially, with the increase of the C/N ratio at 120, the highest titer of PGF2α was produced. However, the negative influence on the growth of strain PG16 with the increasing C/N ratio was showed. In addition, the cell morphology of strain PG16 under different fermented conditions was mainly filamentous in the shake flask (Supplementary Fig. 4).

Fig. 6: Fermentation performance of PGF2α-producing engineered Y. lipolytica strain.
Fig. 6: Fermentation performance of PGF2α-producing engineered Y. lipolytica strain.The alternative text for this image may have been generated using AI.
Full size image

A Effects of different C/N ratios on fermentation characteristics of strain PG16. B Effects of different ALA additions on fermentation characteristics of strain PG16. C Effects of different OA additions on fermentation characteristics of strain PG16. D Effect of different mediums on fermentation characteristics of strain PG16. Pairwise comparisons were analyzed by Student’s t-test. ***P < 0.001. n = 3.

Considering the demand of oxygen involved in the PGF2α biosynthetic pathway, adding ALA as a heme precursor was applied to enhance the supply of oxygen in the engineered strain PG16. The production titer of PGF2α in the strain PG16 was notably affected by adding the varied ALA (Fig. 6B). Especially, compared with no addition of ALA in the strain PG16 grown on YNB medium containing C/N ratio at 10, the addition of ALA with 0.5 mM led to a 2.11-fold improvement of PGF2α titer in the strain PG16 (4.13 μg/L vs 1.96 μg/L). When adding 1.5 mM ALA, the highest titer of PGF2α at 4.32 μg/L was produced by the strain PG16. However, the increased addition of ALA (2–2.5 mM) led to the decreased production titers of PGF2α. Moreover, we observed that the growth of the strain PG16 was not significantly impacted by the addition of ALA. In addition, the cell morphology of the strain PG16 under the varied ALA addition was mainly filamentous in the shake flask.

The potential effect of adding OA on PGF2α production by the strain PG16 was considered (Fig. 6C). Herein, the different additions of OA (0.2–1.0%, w/v) were added in the strain PG16 grown on YNB medium (Fig. 6C). As expected, the addition of OA resulted in the increased PGF2α titer by the strain PG16 after fermentation for 72 h. Especially, the addition of 0.2% (w/v) OA led to the significantly improved titer of PGF2α with 4.59 μg/L in the strain PG16. However, the increasing additions of OA (0.4–1.0, w/v) led to no significant improvement of PGF2α titer in the strain PG16. In addition, when fermented with varied OA additions, strain PG16 exhibited primarily filamentous cell morphology. Although the present study did not explicitly investigate the correlation between filamentous morphology and prostaglandin production, we speculate that morphological changes may influence metabolic efficiency and fermentation phenotype in Y. lipolytica. Morphological engineering represents a promising strategy for future optimization of prostaglandin biosynthesis. As a future strategy, morphological engineering could be leveraged to optimize prostaglandin biosynthesis.

Given the effects of the aforementioned fermentation conditions on the PGF2α titer of the engineered strain PG16, we further evaluated the production potential of PGF2α by the strain PG16 grown on YAO medium, containing C/N ratio at 120, ALA at 0.5 mM and OA at 0.2% (w/v). The remarkable PGF2α titer at 32.63 μg/L was produced by the strain PG16 fermented using YAO medium, which was 16.65-fold higher than that fermented using YNB medium (32.63 μg/L vs 1.96 μg/L) (Fig. 6D). Moreover, the cell morphology of the strain PG16 grown on YNB or YAO medium was mainly filamentous. Overall, the results demonstrated that the defined YAO medium can be used as an optimized fermentation condition for improving PGF2α production by the strain PG16.

Discussion

PGF2α plays a crucial role in regulating a variety of physiological processes1. The yeast Y. lipolytica provides a rich enzymatic environment and suitable metabolic conditions for producing non-native biochemicals27. However, there have been no reports on the biosynthesis of PGF2α in Y. lipolytica. In this study, PGF2α was first synthesized de novo in Y. lipolytica by integrating cell-free and whole-cell catalysis, optimizing substrate delivery and metabolic engineering pathways. The systematic metabolic engineering strategies have been employed in this research, including selecting and fusing key enzymes, employing subcellular engineering, increasing NADPH and acetyl-CoA supplies. Under the optimized fermentation conditions, the PGF2α titer in Y. lipolytica reached 32.63 μg/L, establishing the novel direction for high-efficiency PGF2α production in Y. lipolytica.

In vitro cell-free systems are laboratory-based platforms that harness extracted cellular components to perform biological processes in controlled environments28,29. For example, Yin et al.30 presented an efficient chemoenzymatic synthesis of PGF2α in just five steps on a 10 g-scale. Our study harnessed a cell-free enzymatic system to increase precursor availability for PGF2α production. By adding ARA into the transformation platform of strain PG01, the PGF2α titer of 0.28 µg/L was achieved. Takemura et al. first synthesized prostaglandins in Marchantia polymorpha L. by introducing a cyclooxygenase gene from red algae to accumulate ARA12. Our research introduced GvPGHS and TbPGFS in Y. lipolytica, which led to an 85-fold increase in PGF2α titer. Optimizing the process with 100 μM ARA and 48 h fermentation resulted in a maximum PGF2α titer of 22.11 µg/L. The research of Farida Asras et al. on the ARA-rich fungus Mortierella alpina 1S-4 and the research of Mohamed et al. on Arabidopsis thaliana support our approach11,31. The lower PGF2α titer achieved through whole-cell catalysis compared to in vitro catalysis with 100 μM ARA as the substrate, may be attributed to intracellular competition among different metabolic pathways.

Whole-cell biocatalysis offers distinct advantages through its ability to maintain the natural environment that not only prevents protein denaturation but also enables efficient cofactor recycling32,33. Kanamoto et al. engineered E. coli to express a red algae cyclooxygenase gene, enabling the engineered E. coli capable of producing PGF2α up to 2.7 mg/L using ARA as a substrate13. In this study, the PGF2α biosynthetic pathway was introduced into Y. lipolytica. Although we observed that Po1f(Δku70) also produced trace amounts of PGF2α, no previous has been reported on PGF2α production in Y. lipolytica. Previous studies have demonstrated the production of trace amounts of PGF2α in yeasts, such as S. cerevisiae and Candida parapsilosis34,35. It was speculated that Y. lipolytica itself may produce trace amounts of PGF2α, potentially accounting for the background levels of PGF2α detected via ELISA kit.

Optimizing the adaptation between functional enzymes and chassis is crucial for increasing the yield of target products36. However, the compatibility of dehydrogenases from different sources in host cells can vary significantly, with some enzymes performing poorly in specific chassis37. For example, despite screening multiple sources of PGHS and ∆-8 desaturase in our study, only the metabolic pathway optimization led to improved strain performance, whereas the tested ∆-8 desaturases from three different sources failed to enhance PGF2α production. This may be attributed to their poor adaptability in the Y. lipolytica Po1f (∆ku70) chassis. Similarly, Wang et al.38 optimized metabolic pathways in Y. lipolytica to efficiently synthesize ω6-PUFAs such as GLA and ARA by introducing and tuning ∆-6 and ∆-8 desaturase pathways. However, enzyme compatibility issues in certain chassis strains still limited the yield of target products. These findings highlight the necessity for careful selection and optimization of enzyme-chassis combinations to maximize metabolic efficiency and product yield.

Multi-enzyme co-localization strategies, including scaffold, cell compartment, or synthetic organelle engineering, offer substantial potential for optimizing metabolic pathways and improving the production of value-added chemicals39,40. Our research showed that the GGGGS linker outperforms the 2 A peptide in forming enzyme complexes and improving pathway efficiency, which may be the potential of the 2 A peptide for incomplete cleavage leading to nonfunctional aggregates41,42. Leveraging the subcellular compartments of yeast, due to the unique environments (such as redox potential, pH, etc.) and metabolites (such as cofactors, enzymes, precursors, etc.), enabled the enhanced production of target chemicals43,44. Inspired by the system of subcellular compartmentalization in yeast developed by Liu et al.45, the PGHS and PGFS were targeted to lipid droplets and the endoplasmic reticulum in this study. Combining the GGGGS linker with subcellular localization strategies significantly increased the PGF2α titer, highlighting the benefits of fusion enzyme and organelle engineering for bioproduction efficiency.

Cofactor engineering is vital for enhancing natural product synthesis in cell factories46. The overexpression of NADPH-generating genes in this study was designed to augment the reducing power specifically required by cyclooxygenase enzymes within the prostaglandin biosynthesis pathway, thereby enhancing the production of our target metabolite, PGF2α. In Y. lipolytica, overexpressing G6PD and PGD involved in PPP, resulted in increased NADPH content27. However, the overexpression of G6PD and PGD reduced PGF2α production in this study, possibly due to that Y. lipolytica itself has a sufficient supply of NADPH. Additionally, the strategies for channeling acetyl-CoA toward the desired products such as enhancing metabolic flux in different organelles, optimizing substrate utilization, and regulating protein acetylation levels, promote the production of its derivatives47. Following the successful expression of Cat2 by Yocum et al.48, our group’s previous research also demonstrated that overexpressing ACL significantly increased acetyl-CoA levels in the cytoplasm, which markedly increased the production of target chemicals49. In this study, acetyl-CoA levels were significantly increased by the overexpression of Cat2. Additionally, the results showed that Cat2 overexpression is more effective than ACL2 in elevating acetyl-CoA levels. While this study focused on the role of NADPH in a specific biosynthesis pathway, future research will explore the broader impact of NADPH modulation on overall lipid profiles under varying nutritional regimes.

Fermentation condition optimization is pivotal for enhancing the production of target products in engineered strains50. Nitrogen limitation was identified as a key factor that significantly increasing the PGF2α titer at a carbon-nitrogen ratio of 120:1. The addition of ALA (0.5 mM) and OA (0.2%), as precursors for heme and linoleic acid, respectively, further improved the PGF2α production. Through the combination of these factors, we achieved a PGF2α titer of 32.63 μg/L, demonstrating the significant impact of medium optimization on production titer. Furthermore, our research considered deletion of the PEX10 gene to enhance the flow of acetate-CoA precursors towards PGF2α. However, the results demonstrated that PEX10 knockout significantly compromises the growth of Y. lipolytica in both YNB and YOA media (Supplementary Fig. 5), potentially attributable to disrupted peroxisomal function and subsequent lipid accumulation resulting from PEX10 deletion51.

While trace PGF2α was detected in PO1f, the 175-fold increase in engineered strains confirms the establishment of a complete heterologous pathway enabling high-yield de novo synthesis. In summary, we have successfully engineered Y. lipolytica for the biosynthesis of PGF2α, developing a sustainable and eco-friendly approach to synthesize PGF2α from glucose for the first time at the high titer of 32.63 μg/L (Fig. 7).

Fig. 7: Overview of the improvement of PGF2α production by systematic metabolic engineering of Y. lipolytica.
Fig. 7: Overview of the improvement of PGF2α production by systematic metabolic engineering of Y. lipolytica.The alternative text for this image may have been generated using AI.
Full size image

The schematic illustrated the systematic metabolic engineering strategies and the resultant fold-improvement in PGF2α titer. Key metabolic engineering strategies included: introduction of PGF2α biosynthetic pathway, enzyme fusion of ARA synthesis module, reconstruction of PGF2α biosynthetic pathway, enhancement of acetyl-CoA supply, and optimization of fermentation conditions. These strategies collectively enhanced PGF2α titer from 42.86 ng/L in the initial chassis strain Po1f (Δku70) to 32.63 μg/L in the final engineered strain PG16, representing an over 760-fold increase. n = 3.

Materials and methods

Media and culture conditions

The plasmid-carrying E. coli DH5α cells were incubated for 16 h in LB media with ampicillin (100 μg/mL) at 37°C for plasmid propagation. The Y. lipolytica strains were cultivated in a shake flask at 30°C and 200 rpm utilizing YPD medium which contained 20 g/L glucose, 10 g/L yeast extract, and 20 g/L tryptone. The C/N ratio of the YNB medium used for PGF2α production was 10/1 and it comprised 8 g/L yeast nitrogen base without amino acids (YNB), 9 g/L yeast extract, and 50 g/L glucose. The YYG medium with different carbon-to-nitrogen (C/N) ratios are listed (Supplementary Table 2). A 100 mM solution of ALA was prepared with sterile water. Oleic acid (OA) was emulsified into 0.2 g/mL solution with 0.1% (v/v) Tween 80 solution, the emulsification was carried out using an ultrasonic crusher with 20% power ultrasound for 20 min. Both solutions were filtered, sterilized, and added to the YNB medium at the final concentration, resulting in the YNBA medium and YNBO medium. The YAO medium contained 0.66 g/L of YNB, 0.8 g/L of yeast extract, 50 g/L of glucose, 0.5 mM of ALA, and 0.2% (w/v) of OA. The LEU/URA medium used for yeast screening consisted of 6.7 g/L YNB, 20 g/L glucose, and 20 g/L agar, supplemented with 0.4 g/L leucine or uracil as necessary for selecting transformed Y. lipolytica strains. On YPD + FOA plates, which supplemented 1 g/L 5-fluoroorotic acid (5-FOA) and 20 g/L agar based on the YPD, the URA3 marker was removed and selected against.

A single yeast colony of the recombinant Y. lipolytica strain was aseptically transferred from the agar plate and pre-cultured in YPD liquid medium overnight, followed by inoculation into 5 mLof fresh YPD mediium and cultivation at 30°Cwith shaking at 200 rpm for 24 h. The resulting seed culture was inoculated into 15 mL YPD medium and cultivated at 30°C with shaking at 200 rpm for 48 h. Subsequently, the secondary seed culture were transferred to 50 mL of either YNB or YAO medium (initial OD600 = 0.2) for 72 h fermentation at 30°C with shaking at 200 rpm.

Construction of plasmids and strains

E. coli DH5α was selected for cloning and plasmid propagation. The Y. lipolytica Po1f (Δku70) strain served as the original strain. All yeast strains, plasmids and primers used in this study are listed in Supplementary Data 1-3.

The ∆-9 elongase encoding gene (∆9E, Genbank: GU812433) from I. galbana, ∆-8 desaturase encoding gene including Ig∆8D (Genbank: JN854293), EFD1 (Genbank: AF139720) and EFD2 (Genbank: GU812432) from E. gracilis, and Ps∆8 (Genbank: DQ995518) from P. salina, ∆-5 desaturase encoding gene (∆5D, Genbank: AY464949) from M. alpina, acyl-CoA thioesterases encoding gene (Acot7, Genbank: AB049821) from M. musculus, ∆-12 desaturase encoding gene (Fm∆12D, Genbank: DQ272515) from F. moniliforme, prostaglandin H synthase encoding genes including MmPGHS (GenBank: BC023322) from M. musculus and GvPGHS (GenBank: KJ415281) from G. vermiculophylla, PGF2α synthase encoding gene (PGFS, Genbank: AB034727) from T. brucei, and Cat2 encoding gene (Cat2, Genbank: YML042W) from S. cerevisiae were codon-optimized and synthesized by Genewiz, (Suzhou, China). ∆-12 desaturase encoding gene (Yl∆12D, Genbank: CP028449), G6PD encoding gene (G6PD, Genbank: YALI0E22649g), PGD encoding gene (PGD, Genbank: YALI0B15598g), and ATP citrate lyase encoding gene (ACL2, Genbank: YALI1D32268g) were amplified from the cDNA of Y. lipolytica Po1f (Δku70) strain. Plasmid pUC19 was utilized for plasmid construction, and the recombinant plasmids were assembled in DH5α using the ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China). All functional expression cassettes were amplified with primers and transformed into Y. lipolytica via Frozen-EZ Yeast Transformation II Kit (Zymo Research, Irvine, USA).

In vitro catalytic and whole-cell catalytic synthesis of PGF2α

The in vitro catalytic system of PGF2α synthesis was established based on previously reported protocols52, with appropriate modifications made to the reaction system in this study. The specific steps were as follows: engineered strains PG01 and PG02 were fermented according to the Media and culture conditions described earlier, and the organisms were collected by centrifuging. The organisms were then suspended in 100 mM Tris-HCl (pH=7.5) and the supernatant was obtained by ultrasonic homogenization and further centrifugation. For the cell-free biocatalytic reactions, lysates were prepared from the strain PG01. This approach was taken to specifically assess the intrinsic activity of the expressed enzymes (MmPGHS and TbPGFS). The in vitro catalytic system was prepared to a final volume of 500 μL, containing 100 mM phosphate buffer (pH=7.0), 2 μM hematin, 5 μM tryptophan, 100 μM ARA, and 300 μL of the cell lysis products. The reaction mixture was incubated at 37 °C for 30 min and the reaction was stopped by adding 100 μM hydrochloric acid. The sample was extracted using twice the volume of pre-cooled ethyl acetate, and the content of PGF2α was determined after nitrogen drying.

The subsequent use of the engineered strain PG02 was for whole-cell biotransformation, where its enhanced catalytic performance could be evaluated within the context of a metabolically active system. In the whole-cell catalytic system for PGF2α synthesis, the biosynthetic process was initiated by supplementing ARA substrate to a final concentration of 100 μM into the Y. lipolytica culture at the 48-h fermentation time point. The culture was subsequently maintained at 30°C with continuous shaking at 200 rpm for an additional 24 h to facilitate PGF2α biosynthesis, after which the product was quantitatively analyzed. Subsequently, varying concentrations of ARA (50, 100, 150, and 200 μM) were supplemented into 48-h Y. lipolytica cultures, followed by 24 h of additional fermentation at 30°C with shaking at 200 rpm for PGF2α production and detection. To investigate the influence of catalytic duration on whole-cell PGF2α synthesis, ARA was supplemented to a final concentration of 100 μM into 48-h Y. lipolytica cultures. PGF2α production was then monitored at different time intervals (1, 6, 12, 24, 36, and 48 h) post-substrate addition under standard fermentation conditions (30 °C, 200 rpm).

Lipid extraction and fatty acid quantification

The fermentation broth was centrifuged at 8000 rpm for 5 min, and the resulting precipitate was subsequently washed twice with distilled water. The harvested cells were pre-frozen at −80 °C for 2 h and subsequently lyophilized at −40°C for 24 h. Then, 0.05 g of the lyophilized cell pellets were suspended in 1.5 mL of methanol: chloroform solution (1: 2, v/v) and homogenized using a tissue grinder. The resulting mixture was vortexed at 200 rpm for 30 min using a Vortex Genie 2 (Model 3044) and then centrifuged at 12000 rpm for 10 min. The resulting supernatant was collected and evaporated under a gentle stream of nitrogen gas. The nitrogen-evaporated product was subjected to lipid extraction using 500 μL of n-hexane, followed by methylation with 100 μL of a methanol-potassium hydroxide solution. After vortexing the mixture for 1 min and subsequent centrifugation at 6000 rpm for 5 min, the hexane phase was collected for analysis using gas chromatography coupled with a flame ionization detector (GC-FID). The ARA content is expressed as a percentage of total fatty acids.

The GC-FID system was equipped with a DB-23 column (25 m × 320 μm × 0.25 μm) and operated with nitrogen as the carrier gas at a constant flow rate of 2.0 mL/min. The oven temperature program was initiated at 100°C, then increased to 200 °C at a rate of 25 °C/min with no holding time, followed by a second temperature ramp to 230 °C at 5 °C/min with a 3-min holding period. The FID port’s temperature was maintained at 250 °C, and the samples were injected at a volume of 1 μL with a shunt ratio of 10: 1.

Analysis and quantification of PGF2α

LC-MS served the critical role of structural verification, while ELISA provided robust quantification for physiologically relevant concentrations. The fermentation supernatant was diluted using the buffer and the PGF2α titer was quantified using the PGF2α ELISA kit (Cayman Chemical, Ann Arbor, MI, USA) according to the manufacturer’s instructions. PGF2α was analyzed qualitatively using LC-MS (Shimadzu, Kyoto, Japan) equipped using a Waters ACQUITY UPLCRBEHC18 column (2.1×100 mm, 1.7 μm). The mobile phases of gradient elution comprised methanol and 0.1% formic acid, with a starting concentration of 50% methanol that increased to 100% after 10 min. The flow rate was maintained at 0.25 mL/min, column temperature at 40 °C, and injection volume at 10 μL. The mass spectrometry conditions were set at a DL temperature of 270 °C, heat block temperature of 450 °C, nebulizing gas at 2.0 L/min, and gyring gas at 10 L/min. Negative ion mode was utilized with MRM monitoring. Quantitative ions: 353/309.2, Q1: 17 V, CE: 20 eV, Q3: 30 V. Qualitative ions: 353/193.15, Q1: 17 V, CE: 25 eV, Q3: 12 V. Qualitative ions: 353/335, Q1: 17 V, CE: 20 eV, Q3: 15 V.

Analysis of bioinformatics

The subcellular localization of GvPGHS and TbPGFS proteins in Y. lipolytica was predicted using the Protein Subcellular Localization Prediction Tool (PSORT) II (GenScript, USA). Multiple sequence alignment and construction of Neighbor-Joining evolutionary trees were performed using MEGA software (Mega Limited, Auckland, New Zealand) for the sequences of genes encoding ∆-8 desaturases from various organisms.

Analysis of fermentation parameters

The glucose content in the supernatant was measured using an SBA-40E biosensor (Biology Institute of Shandong Academy of Sciences, Jinan, China). Simultaneously, the pH value was determined using a calibrated pH meter (Mettler Toledo, Shanghai, China). After lyophilization for 24 h, the cell biomass was quantified through dry weight measurement. Specifically, the lyophilized cells were weighed using a pre-weighed plastic dish. After transferring the cells, the empty dish was re-weighed using an analytical balance (PL203, ±0.0001 g; Mettler Toledo, Switzerland) at room temperature. The dry weight, representing the biomass, was calculated by subtracting the empty dish weight from the initial total weight.

Cells morphology observation

Fluorescence microscopy of Y. lipolytica was performed using a Zeiss LSM710 inverted laser scanning confocal microscope (Carl Zeiss AG, Germany) at 67x magnification, and the images were processed using ZEN software (version 3.8; Carl Zeiss AG, Germany). The fermentation morphology of Y. lipolytica were observed using a light microscope equipped with a 40× objective lens and a 10× eyepiece, resulting in a total magnification of 400×. The images were subsequently processed using the TCapture V4.3.0.605 (Shanhan Optoelectronics, China).

Statistics & reproducibility

No formal sample size calculation was performed, and no data points were excluded. The experimental design did not include randomization or blinding of investigators. Data collection was carried out in triplicate. Raw data were processed in Microsoft Excel, and all statistical analyses as well as graphing were performed using OriginPro 2023 and GraphPad Prism 8.0.1. Results are expressed as mean ± standard deviation (s.d.) or median, based on data distribution. Pairwise comparisons were evaluated with Student’s t-test, and differences were considered significant at P < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.