Metabolic Engineering of Komagataella Phaffii for de Novo Synthesis of Retinol From Methanol

ABSTRACT

Retinol, a derivative of vitamin A with potent antioxidant and therapeutic properties, is in high market demand. In response to the low productivity of conventional methods, metabolic engineering has been explored for microbial retinol production. However, systematic engineering strategies for high‐level retinol synthesis in Komagataella phaffii remain limited. In this study, the methylotrophic yeast K. phaffii was developed as an engineered chassis for efficient de novo retinol biosynthesis. Based on a previously constructed β‐carotene‐producing strain, β‐carotene‐15,15′‐dioxygenase (Blh) and retinol dehydrogenase (RDH12) were screened and introduced to establish the synthetic pathway of retinol. To increase precursor supply, key genes in the β‐carotene biosynthetic pathway were overexpressed. The mevalonate (MVA) pathway was further optimized, and central carbon metabolism was reprogrammed to enhance metabolic flux toward retinol. Transport engineering was also performed to improve retinol secretion. Several candidate transporters were overexpressed, and the protein encoded by chr1‐4_0619 in K. phaffii was identified as an endogenous retinol transporter. The final engineered strain produced 3.38 g/L retinol with BHT supplementation in fed‐batch fermentation using a 1.5 L bioreactor. This work represents de novo microbial synthesis of retinol from a one‐carbon feedstock, demonstrating the formidable potential of K. phaffii as a sustainable chassis for retinol production.

Graphical Abstract and Lay Summary

In this study, Komagataella phaffii was successfully engineered to achieve de novo retinol synthesis from one‐carbon feedstock. By constructing and optimizing the retinol biosynthesis pathway, strengthening the β‐carotene synthesis module, upstream mevalonate (MVA) pathway, and central carbon metabolism, the precursor supply was significantly enhanced. Additionally, transporter protein engineering improved secretion efficiency, establishing K. phaffii as a powerful and efficient microbial chassis for sustainable retinol production.

1. Introduction

Vitamin A is an essential nutrient that supports normal physiological functions in humans. It plays key roles in vision, growth, and development, maintenance of epithelial tissues, and immune function [1, 2]. Deficiency in vitamin A may lead to growth retardation, night blindness, xerophthalmia, and keratomalacia [3]. Vitamin A comprises a family of fat‐soluble bioactive components, among which retinol exhibits the highest biological activity. Owing to its indispensable physiological functions, retinol is widely utilized in pharmaceuticals, dietary supplements, cosmetics, and animal feed [4, 5].

Vitamin A is found in animal tissues and is especially abundant in fish liver [6]. Although it can be extracted from animal sources, commercial production mainly depends on chemical synthesis. This is due to the limited availability of raw material sources, complicated purification processes, and high production costs associated with extraction [7]. Industrial vitamin A synthesis is dominated by two major routes: the Roche process and BASF process, both using β‐ionone as the starting material [8]. The Roche process is technically mature but requires over 40 raw and auxiliary materials, leading to substantial material consumption. The BASF process involves fewer reaction steps but requires highly toxic phosgene, which imposes strict safety and equipment requirements. Conventional chemical synthesis therefore faces significant limitations, including high environmental risks, process complexity, and substantial carbon footprints. With increasing demand for green manufacturing and rapid progress in synthetic biology, microbial production of retinol has gained attention as an alternative approach. This approach leverages metabolically engineered microbial cell factories and renewable carbon sources to enable efficient production, demonstrating notable advantages in environmental sustainability, process simplification, and long‐term viability [9].

To date, de novo biosynthesis of vitamin A, including retinol and its derivatives, has been achieved in Escherichia coli [10], Saccharomyces cerevisiae [11, 12], and Yarrowia lipolytica [13, 14]. When constructing microbial cell factories, a common strategy to achieve high expression of heterologous enzymes is to screen genes from different species and evaluate their compatibility with the host. In a β‐carotene‐producing Y. lipolytica strain, screening and overexpression of β‐carotene 15,15′‐dioxygenase (BCO) and retinol dehydrogenase (RDH) enabled the production of 348 mg/L retinol. Further strengthening of the upstream MVA pathway by overexpressing AtoB, HMGS, HMGR, and ERG12 increased the retinol titer to 583 mg/L [14]. The recombinant strain ultimately achieved a retinol yield of 5.89 g/L through fed‐batch fermentation. Sufficient precursor supply is also critical for high retinol production. Increasing the copy numbers of the key genes CrtYB and GGPPxd in Y. lipolytica increased β‐carotene accumulation to 822.7 mg/L, and subsequent introduction of heterologous retinol synthesis genes led to 389.6 mg/L retinol production in shake flask [13]. Retinol is lipophilic and can be transported by ABC transporters. In situ extraction with organic solvents [15] and identification of effective transport proteins [16] are effective strategies to enhance product secretion. Adding dodecane or olive oil as extraction agents in engineered S. cerevisiae increased the retinol titer, with 5% dodecane resulting in a 5.21‐fold increase compared to single‐phase culture [17]. Overexpression of snq2p in S. cerevisiae also enhanced β‐carotene secretion and production. The secretion level increased 4.04‐fold, and the intracellular level increased 1.32‐fold compared to the parent strain [18]. Furthermore, due to its isoprenoid side chain and unsaturated ring structure, retinol is prone to oxidation and degradation [19]. Therefore, the addition of antioxidants is crucial to reduce oxidative loss of retinol. Butylated hydroxytoluene (BHT) is a commonly used antioxidant for protecting easily oxidizable compounds. Supplementing 1% BHT in shake‐flask culture increased retinol production from 425.7 mg/L to 621.3 mg/L [20].

The methylotrophic yeast Komagataella phaffii (formerly Pichia pastoris) can utilize methanol as its sole carbon source for growth and biochemical production [21]. Its Crabtree‐negative phenotype allows high‐cell‐density fermentation under low osmotic conditions [22]. Combined with efficient editing tools [23, 24, 25] and strong methanol‐inducible promoters like AOX1, K. phaffii has become a promising chassis for producing high‐value compounds, such as lycopene [26], nootkatone [27], α‐santalene [28], and α‐farnesene [29]. Engineered K. phaffii cell factories hold significant potential to advance C1‐based biomanufacturing. Despite this potential, no studies have reported de novo retinol production in K. phaffii using methanol as the sole carbon source. Therefore, it is essential to systematically explore metabolic engineering strategies to evaluate the feasibility of achieving high retinol yields in K. phaffii.

In this study, we constructed a de novo retinol biosynthesis pathway in a β‐carotene‐producing K. phaffii strain by expressing β‐carotene‐15,15'‐dioxygenase and retinal dehydrogenase genes from various sources. We further optimized precursor supply and methanol utilization by overexpressing β‐carotene biosynthesis genes, systematically engineering of the MVA pathway, and remodeling central carbon metabolism. The transporter engineering identified the endogenous retinol transporter in K.phaffii and promoted the extracellular secretion of retinol. After coordinated optimization of carbon source utilization and antioxidant supplementation, scale‐up in a 1.5‐L bioreactor ultimately achieved a retinol titer of 3.38 g/L, representing the first reported de novo retinol synthesis in K.phaffii using methanol as the sole carbon source. This work provides a reference for the production and secretion of other high‐value compounds in K.phaffii.

2. Materials and Methods

2.1. Strains, Medium, and Culture Conditions

The parental K. phaffii strain used in this study is TTG, a β‐carotene‐producing strain derived from K. phaffii GS115. E. coli TOP10F’ (Invitrogen) was used for construction of plasmids. K. phaffii strains used in this study (Table S1) were grown at 30°C in YPD medium (1% yeast extract, 2% peptone, and 2% glucose). E. coli strains were grown at 37°C in LB medium or low‐salt LBL medium. Zeocin was added to YPD (100 mg/L) and LBL (25 mg/L) media for selection.

For shake flask fermentation, the K. phaffii strains were first grown at 30°C in BMGY medium (1% yeast extract, 2% peptone, 1.34% YNB (Yeast Nitrogen Base), and 1% glycerol) for 24 h. Subsequently, the cultures were transferred to BMMY medium (1% yeast extract, 2% peptone, 1.34% YNB, and 1% methanol) at an initial OD₆₀₀ of 0.5 and further incubated for 120 h. At the initiation of methanol induction, 10% (v/v) dodecane was added simultaneously to capture the products. The retinol titer was calculated based on the volume of the aqueous phase during the two‐phase fermentation.

2.2. Construction of Plasmids and Strains

Blh from Uncultured marine bacterium 66A03, RDH12 from H. sapiens, ybbO from E. coli, ENV9 from S. cerevisiae, XFPK from B. breve, and PTA from C. kluyveri were codon‐optimized for expression in K. phaffii and synthesized by Genewiz (Suzhou, China).

Primers used in this study are listed in Table S2. All plasmids used in this study were constructed using Gibson assembly. Gene deletions and integrations in the K. phaffii genome were performed via the CRISPR‐Cas9 or CRISPR‐Cpf1 system [30]. For endogenous gene overexpression and heterologous gene insertion, the target gene was amplified from the chromosome of K. phaffii or the corresponding species’ genome and flanked by the strong AOX1 or AOXm promoters and the AOX1 terminator. The DNA donors were assembled by overlap PCR.

2.3. Quantification of Retinoids and Carotenoids

The quantification of retinoids and carotenoids was performed on an Agilent HPLC (Agilent 1220 Infinity II, USA) system equipped with a Waters C18 column (4.6×250 mm²). Detection was carried out at 470 nm for carotenoids and 352 nm for retinoids. The separation was achieved using a mobile phase consisting of acetonitrile: water (9:1, v/v) and methanol: isopropanol (3:2, v/v) under gradient elution at a flow rate of 1 mL/min with the column temperature maintained at 40°C. The injection volume was set at 10 µL. Commercial retinol (Solarbio, Beijing, China), retinal (Aladdin, Shanghai, China), β‐carotene (Aladdin, Shanghai, China), and lycopene (Sigma‐Aldrich, Shanghai, China) were used as standards.

To quantify extracellular retinoids, the cell cultures were centrifuged for 5 min at 12,000 g, and the supernatant was filtered through a 0.2 µm membrane filter and subsequently analyzed by HPLC.

To extract intracellular retinoids and carotenoids, 400 µL of cell culture was centrifuged at 6000 g for 10 min to remove the supernatant. The cell pellet was collected and resuspended in 1 mL of dodecane along with 0.7 g of 0.5‐mm zirconium beads. The mixture was subjected to mechanical disruption using a bead‐beating homogenizer for four cycles. After homogenization, the sample was centrifuged at 10,000 g for 3 min. The supernatant was collected and analyzed using HPLC.

2.4. Fed‐batch Cultivations

A single colony isolated from a YPD plate was initially inoculated into 25 mL of YPD medium and cultivated at 30°C with shaking at 250 rpm for 16 h. Subsequently, 4% of seed culture was transferred into 50 mL of YPD medium and incubated under the same conditions for 20 h. Finally, 8% of secondary seed culture was inoculated into a 1.5‐L fermenter containing BSM medium.

Basal salt medium (BSM) (4% glycerol, 0.093% CaSO₄·2H₂O, 1.147% MgSO₄, 1.82% K₂SO₄, 0.413% KOH, 2.67% H₃PO₄, and 0.435% PTM1, PTM1: 0.05% CoCl₂·6H₂O, 0.6% CuSO₄·5H₂O, 6.5% FeSO₄·7H₂O, 0.002% H₃BO₃, 0.008% NaI, 0.5% H₂SO₄, 0.02% Na₂MoO₄·2H₂O, 2% ZnCl₂, 0.3% MnSO₄·H₂O, and 0.02% biotin) was used in fed‐batch cultivation.

The fermentation process was divided into three phases. In the batch phase (Phase I), initial cell cultivation was conducted in BSM medium at 30°C for 20 h, with dissolved oxygen (DO) maintained at 30%–40%. The pH was controlled at 5.5 using ammonia solution. During the fed‐batch phase (Phase II), limited glycerol feeding was performed to sustain cell growth until the OD₆₀₀ reached 220, after which glycerol feeding was terminated. In the induction phase (Phase III), methanol was added to promote retinol biosynthesis. The DO was maintained at 20%–30%, the temperature was adjusted to 25°C, and the pH was regulated to 6.0. At 24 h after methanol induction, 0.5% BHT was added and replenished every 48 h. When the OD₆₀₀ stabilized, 20% (v/v) dodecane was added.

2.5. Transcriptome Analysis

Cells were harvested after 48 h of methanol induction, and RNA sequencing was performed using the NextSeq 500 system (Illumina, CA) with three biological replicates per group. The reference genome sequence and annotation data of K. phaffii GS115 were obtained from the NCBI database.

Gene expression analysis was conducted using DESeq2. Differentially expressed genes (DEGs) were identified based on a fold change >1.5 and p< 0.05. Functional enrichment analysis of DEGs was performed using Gene Ontology (GO) and KEGG pathway analyses.

2.6. Molecular Docking

The transporter structures used for molecular docking were predicted from the AlphaFold protein structure database [31]. Molecular structure of retinol was retrieved from the ZINC database [32]. Molecular docking was performed using AutoDock Vina 1.2.7. The binding conformation with the lowest binding free energy between the receptor protein and ligand was selected as the optimal docking pose. The results of molecular docking were visualized using PyMOL 3.0.3.

3. Results and Discussion

3.1. Construction of the Retinol Synthesis Pathway

Retinol is synthesized from β‐carotene through two consecutive enzymatic reactions catalyzed by β‐carotene 15,15'‐dioxygenase (BCO) and retinol dehydrogenase (RDH) (Figure 1A). The β‐carotene‐producing strain TTG was selected as the parental strain to construct the retinol pathway. Heterologous expression of the blh gene encoding BCO from Uncultured marine bacterium 66A03 in TTG resulted in retinal production in the engineered BLH strain, reaching 48.53 mg/L. Notably, heterologous expression of blh in a β‐carotene‐producing S. cerevisiae strain resulted in detectable levels of both retinal and retinol [17]. In contrast, the K. phaffii system exclusively accumulated retinal with no detectable retinol formation. This species‐specific metabolic divergence may be attributed to the presence of highly expressed and active endogenous alcohol dehydrogenase (ENV9) in S. cerevisiae, whereas K. phaffii either lacks functionally active endogenous dehydrogenases capable of efficient retinal‐to‐retinol conversion or expresses them at insufficient levels.

FIGURE 1.

De novo retinoids biosynthesis in K. phaffii. (A) Synthesis pathway for retinol from β‐carotene. (B) HPLC spectra of retinoids. (C) Titers of retinoids in strains expressing RDH enzymes from different sources. (D) Titers of carotenoids in strains expressing RDH enzymes from different sources. Data were presented as the mean ± standard deviation (s.d.) (n = 3 biologically independent samples). Error bars represent s.d.

In order to establish a complete retinol biosynthetic pathway, three distinct RDHs were expressed in the BLH strain: ENV9 from S. cerevisiae, RDH12 from Homo sapiens, and ybbO from E. coli. As expected, all three RDHs successfully converted retinal to retinol when expressed in the BLH strain (Figure 1B). Among them, the BR strain, which expresses RDH12, produced the highest level of retinol (105.06 mg/L) and the lowest level of residual retinal (3.15 mg/L), demonstrating the highest efficiency of retinol production (Figure 1C). The titers of precursor β‐carotene and lycopene are shown in Figure 1D. The negligible accumulation of retinal in BR indicates that RDH12 catalyzes the conversion of retinal to retinol with high efficiency. Consequently, BR was selected for further engineering.

3.2. Optimizing Expression Forms of Blh and RDH12 for Retinol Synthesis Efficiency

The efficiency of retinol synthesis may be influenced by the loss of retinal due to diffusion, degradation, or conversion into byproducts. In order to enhance the utilization of intermediates during the catalytic process, adjacent proteins involved in sequential catalysis are usually expressed through linker region fusion or protein self‐assembly strategies [33]. To enhance the spatial proximity of Blh and RDH12 and minimize intermediate loss, two enzymes were fused in different orders using a flexible (GGGS) linker in the BLH strain (Figure 2A). However, fermentation results showed that the BLR strain produced less retinol and accumulated more β‐carotene than the BR strain (Figure 2B). This suggests that Blh activity was compromised, reducing the conversion of β‐carotene to retinal. Moreover, no retinol was detected in the RLB strain (Figure 2B), indicating a complete loss of Blh enzymatic activity.

FIGURE 2.

Optimization of the retinol synthesis module (A) RDH12 is positioned close to Blh via linker and targeting peptide. (B) Fusion expression decreased retinol titer. (C) Subcellular localization increased retinol titer. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Circles represent data points. Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Prediction using TMHMM‐2.0 indicated that Blh contains seven transmembrane domains, classifying it as a membrane protein (Figure S1). We hypothesize that fusion with RDH12 via a linker peptide likely altered the spatial arrangement of these transmembrane domains. Since the transmembrane regions of membrane proteins are often associated with their catalytic centers, this structural perturbation could explain the observed reduction or complete loss of Blh enzymatic activity.

Blh is primarily localized to the endoplasmic reticulum (ER) and the plasma membrane. To shorten the spatial distance between Blh and RDH12 without disrupting the native structure of Blh, RDH12 was targeted to the ER or plasma membrane by fusing with an ER‐targeting peptide (KDEL) or a plasma membrane‐targeting peptide (PMSeV‐C, abbreviated as PM), respectively (Figure 2A). The resulting PM strain (plasma membrane‐localized RDH12) and KDEL strain (ER‐localized RDH12) produced 111.32 mg/L and 106.68 mg/L of retinol (Figure 2C), respectively. The yield of retinol was slightly increased compared to the BR strain, but the effect was not significant. This indicates that subcellular localization is a potential effective strategy to improve the retinol biosynthesis efficiency, but the conversion efficiency of RDH12 without the localization peptide is already relatively high, sufficient to meet the conversion requirements of retinol.

3.3. Enhancing Precursor Supply via Key Gene Copy Number Modulation

β‐Carotene is the direct precursor for retinol synthesis, making its sufficient supply essential (Figure 3A). In the BR strain, lycopene accumulated to 50.64 mg/L, whereas β‐carotene was only 15.01 mg/L (Figure 1D), suggesting a potential bottleneck at the lycopene‐to‐β‐carotene step catalyzed by CrtYB. The low β‐carotene level, likely caused by insufficient CrtYB expression, appeared to limit the metabolic flux toward retinol. To address this, an additional copy of crtYB was introduced into the BR strain, generating the YB strain. As expected, the YB strain showed a decreased lycopene titer (16.31 mg/L) and an increased β‐carotene titer (18.31 mg/L), which consequently raised retinol production to 116.68 mg/L (Figure 3B).

FIGURE 3.

Optimization of the β‐carotene synthesis module. (A) Heterologous pathway in K. phaffii, the red genes are key heterologous gene for the synthesis of β‐carotene. (B) Titer of retinol and carotenoids in strains overexpressing β‐carotene synthesis genes. (C) Growth curve of strains overexpressing β‐carotene synthesis genes. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Circles represent data points. Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Unlike the BR strain, the YB strain accumulated moderate levels of β‐carotene but very little lycopene. To further redirect metabolic flux toward retinol, the copy numbers of crtE and crtI, which encode key enzymes for lycopene synthesis, were optimized in the YB strain. During this optimization, increasing the copy number of crtI consistently reduced retinol production, whereas amplification of crtE enhanced yields. These results indicate that CrtE, rather than CrtI, is the critical rate‐limiting enzyme in retinol precursor biosynthesis. By progressively increasing the copy number of crtE, the YE, YEE, and Y3E strains were constructed. The final retinol titer in the Y3E strain reached 125.61 mg/L (Figure 3B), representing a 19.56% increase over the parental BR strain. Furthermore, the multiple copies of the heterologous genes did not have a significant impact on the growth of the recombinant strain (Figure 3C).

3.4. Strengthening the MVA Pathway and Optimizing Central Carbon Metabolism for Retinol Production

Redirecting metabolic flux to efficiently produce target compounds is typically achieved through several strategies, including the enhanced expression of rate‐limiting enzymes [34, 35], attenuation of competing pathways [36, 37], and incorporation of heterologous biosynthesis pathways [38, 39].

The intracellular metabolic network of yeasts is tightly regulated. Overexpressing key genes in the upstream MVA pathway (Figure 4A) has become an effective strategy to increase terpenoid production [40]. Studies have indicated that excessive accumulation of MVA intermediates, such as dimethylallyl diphosphate (DMAPP) and isopentenyl diphosphate (IPP), can be cytotoxic and inhibit normal cell growth [41]. To balance these intermediates, IDI, a key enzyme catalyzing the reversible interconversion of IPP and DMAPP, was overexpressed in the Y3E strain. This adjustment effectively balanced the intermediate pools and increased the retinol titer to 135.68 mg/L in the resulting R02 strain (Figure 4B). Additional genes involved in converting acetyl‐CoA to farnesyl diphosphate (FPP) were overexpressed. Among these, overexpression of EGR20 and ERG13 increased retinol production, while other genes had negative effects. Subsequently, co‐overexpression of ERG13 in the R02 strain generated the R07 strain, which achieved a further increased retinol titer of 142.91 mg/L (Figure 4B). These results indicate that engineering the MVA pathway alone is insufficient for high‐yield retinol production, as the pathway is limited by the inherent catalytic efficiency at multiple nodes and its complex regulatory mechanisms.

FIGURE 4.

Optimization of the MVA pathway and Ac‐CoA synthesis module. (A) MVA pathway in K. phaffii, the red genes were overexpressed. (B) Overexpression of key genes in the MVA pathway increased retinol yield. (C) Heterologous Ac‐CoA synthesis pathways expressed in K.phaffii. Red for PK/PTA pathway and green for ACL pathway. (D) Introduction of heterologous Ac‐CoA synthesis pathways increased retinol yield. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Circles represent data points. Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

A sufficient supply of acetyl‐CoA, the starting substrate of the MVA pathway, is essential for increasing downstream carbon flux. Using a bottom‐up, modular, stepwise optimization strategy, we inferred that the critical bottleneck limiting retinol synthesis in the R07 strain likely resides upstream in the MVA pathway, specifically within the module responsible for converting methanol to acetyl‐CoA. Flux through this process is coordinately regulated at multiple key metabolic nodes.

Introducing a heterologous acetyl‐CoA synthesis pathway is a common approach to increase intracellular acetyl‐CoA levels [42]. The heterologous pathways directly synthesize acetyl‐CoA in the cytoplasm, which can avoid the subcellular compartmentalization and transmembrane transport losses of native synthesis pathway in K. phaffii, and improve the efficiency of carbon source flow towards retinol (Figure 4C). Incorporation of the heterologous phosphoketolase/ phosphotransacetylase (PK/PTA) pathway into the R09 strain to rewire central carbon metabolism increased the retinol titer to 155.07 mg/L (Figure 4D). Additionally, key enzymes in the pentose phosphate pathway (PPP) and the methanol assimilation pathway were overexpressed to promote regeneration of key intermediates and alleviate the accumulation of the toxic intermediate formaldehyde (Figure 5A). This strengthened the conversion of methanol to acetyl‐CoA and improved the cytosolic acetyl‐CoA supply. Among the targets tested, individual overexpression of DAS2 or DAS1 enhanced retinol production (Figure 5B). However, their co‐overexpression likely disrupted the balance of methanol metabolism, resulting in a significant decrease in production. Meanwhile, the biomass accumulation of the R12 strain was higher than that of the R09 strain (Figure S2A). This might be because the overexpression of the DAS2 promoted the conversion of formaldehyde, alleviating the toxicity caused by intracellular formaldehyde accumulation. These results underscore the importance of fine‐tuning metabolic flux for optimal retinol synthesis.

FIGURE 5.

(A) Methanol metabolism and PPP modules, genes with a blue background were overexpressed. (B) Retinol titer in strains overexpressed carbon metabolism genes. (C) Retinol titer in strains with the FPP branch pathway knocked out. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Circles represent data points. Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

FPP, a key precursor for the β‐carotene synthesis module, can be diverted by Dpp1 and Lpp1, which catalyze its conversion to the byproduct farnesol, resulting in carbon loss. To reduce this diversion of FPP, a previously established CRISPR‐Cpf1 system was used to knock out LPP1 and DPP1 in the R19 strain. The results showed that deletion of DPP1 increased retinol production by 3.77%, whereas deletion of LPP1 had no significant effect (Figure 5C), suggesting that Dpp1 is the primary enzyme responsible for FPP loss in this system. The deletion of DPP1 did not have a significant negative impact on the growth of the strain (Figure S2B).

3.5. Transporter Engineering for Promoting Retinol Secretion and Production

Transporter engineering can enhance both the secretion and synthesis of terpenoids by improving cellular tolerance to the target product and alleviating intracellular storage pressure. Previous studies have identified ABC transporter proteins in S. cerevisiae responsible for vitamin A transport [43], whose overexpression has been proven to effectively increase vitamin A production. To promote retinol export, we heterologously expressed three retinol transporters from S. cerevisiae (SNQ2, PDR5, and PDR10) and the retinol transport protein complex TTR‐RBP4 from H. sapiens in the R19 strain. To mitigate the potential growth burden of membrane protein overexpression, all transporter genes were placed under the control of the methanol‐inducible promoter AOX, thereby decoupling the production phase from the growth phase. Experimental results indicated that both SNQ2 and PDR5 exhibited strong transport activity in K. phaffii, with the respective strains achieving retinol titers of 213.48 mg/L and 212.97 mg/L (Figure 6A). Although the intracellular‐to‐extracellular retinol ratio did not change significantly after transporter expression (Figure S3A), the increase in total production suggests that timely efflux helps pull the intracellular synthesis flux toward retinol.

FIGURE 6.

Retinol transporters engineering. (A) Expression of heterologous retinol transport proteins improved retinol yield. (B) Expression of endogenous retinol transport proteins enhanced the proportion of retinol secretion. (C) Molecular docking simulation of endogenous candidate retinol transport proteins binding to retinol. From left to right, the first row shows docking simulations of proteins translated from genes chr4_0832, FragB_0065, and chr3_0935 with retinol; the second row shows docking simulations of proteins translated from chr1‐4_0619, chr2‐1_0050, and chr4_0859 with retinol. The numbers represent the binding free energy of the protein's interaction with retinol. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Even without introducing transporters, the R19 strain secreted 91.83% of retinol into the extracellular medium, indicating that K. phaffii has intrinsic retinol transport capability. To identify endogenous transporters, SNQ2 and PDR5 were used as reference sequences to screen the K. phaffii genome via NCBI BLAST. Six genes encoding potential ABC transporters were identified as candidates: chr4_0832, FragB_0065, chr3_0935, chr1‐4_0619, chr2‐1_0050, and chr4_0859. Each gene was overexpressed in the R19 strain under P _AOX1 , generating recombinant strains R26 to R31. Only the expression of chr1‐4_0619 enhanced retinol production (Figure 6B), while the other genes showed no observable effect. Although the total retinol increase induced by chr1‐4_0619 overexpression was smaller than that observed with SNQ2, the proportion of extracellular retinol secretion reached 93.55% (Figure S3B), higher than those of the R19 and R23 strain. Molecular docking of endogenous retinol transporters with retinol molecules revealed that the binding energy of the protein encoded by chr1‐4_0619 is −7.52 kcal/mol (Figure 6C). The low binding energy also suggests that it has a strong binding and transfer ability for retinol These results indicate that the protein encoded by chr1‐4_0619 functions as an efficient endogenous retinol transporter in K. phaffii. However, after knocking out chr1‐4_0619 in R19, the resulting R32 strain showed a reduction in retinol production but still retained certain capabilities for retinol synthesis and extracellular transport (Figure S4). This indicates that the protein encoded by chr1‐4_0619 only partially contributes to retinol transport, and there are likely undiscovered retinol transport proteins involved in supporting retinol transport in K.phaffii.

3.6. Optimizing Fermentation Conditions for Enhanced Productivity in Recombinant Strains

The key carbon source methanol has a dual effect on K. phaffii fermentation by influencing both cell growth and product synthesis. Adequate methanol supports cell growth and metabolic activity. In contrast, excessively high methanol concentrations can be cytotoxic, and the formaldehyde accumulated during its assimilation may lead to cell death. Therefore, determining the optimal methanol concentration is critical for fermentation process optimization. To evaluate its effect on the R23 strain, cultivations were performed at methanol concentrations of 0.75%, 1%, 1.5%, 1.75%, and 2%, while monitoring cell growth and retinol production. The highest retinol titer (267.66 mg/L) was observed at 1.75% methanol (Figure 7A). Concurrently, biomass increased progressively within the 0.75% to 1.75% methanol range, peaking at 1.75%. However, when the methanol concentration was raised to 2%, its toxic effects outweighed its supportive role as a carbon source, leading to inhibited cell growth (Figure 7B) and decreased retinol production. These findings demonstrate the necessity of balancing the promotion of cell growth against the avoidance of toxicity accumulation when optimizing methanol concentration.

FIGURE 7.

Optimization of shake flask fermentation conditions for R23 strain. (A) Effect of different methanol concentrations on retinol titer. (B) Growth curves of R23 strain under different methanol concentrations. (C) Effect of adding different metal ions on retinol titer. (D) Effect of adding different concentrations of BHT on retinol titer. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). Circles represent data points. Error bars represent s.d. Statistical significance determined by Student's t test is indicated with an asterisk (ns>0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Previous studies have reported that Fe^{2 +} and Zn^{2 +} may enhance retinol synthesis by modulating key enzyme activities [14], while antioxidants such as BHT can reduce oxidative degradation of retinol. To evaluate these effects, 1.5 mM Fe^{2 +}, 1.5 mM Zn^{2 +}, and 0.1% BHT were individually added to the shake‐flask fermentations. As expected, BHT supplementation increased the retinol titer, confirming that antioxidants effectively reduce oxidative losses. In contrast, the supplementation of Fe^{2 +} or Zn^{2 +} showed no positive effect (Figure 7C). Further optimization of BHT concentration revealed that retinol titer increased gradually between 0.1% and 0.5% BHT. However, the yield plateaued between 0.5% and 2%, indicating that 0.5% BHT is sufficient to suppress retinol oxidation, with no additional benefit at higher concentrations (Figure 7D). Under these optimized conditions, strain R23 achieved a retinol yield of 321.97 mg/L in shake‐flask when supplemented with 0.5% BHT and 1.75% methanol (Figure 7D).

3.7. High‐Density Fermentation to Expand the Production of Retinol

The retinol production capacity of the recombinant R23 strain was further evaluated using high‐density fed‐batch fermentation in a 1.5‐L bioreactor. During this fermentation process, cell growth and product accumulation were separated by switching the carbon source from glycerol to methanol. In the growth phase, glycerol served as the carbon source, repressing the expression of P _AOX1 ‐controlled genes while promoting rapid biomass accumulation. When the OD₆₀₀ reached 200, methanol induction was initiated to trigger the product accumulation phase, activating transcription from the AOX1 promoter while allowing continued biomass increase.

Previous studies have reported high retinol yields in high‐density fermentations of various microorganisms by adding dodecane during the exponential growth phase for in situ extraction. Following this approach, 20% (v/v) dodecane was added after 36 h of methanol induction, when the cells had adapted to the carbon source shift and reached an OD₆₀₀ of 219 with rapidly accumulating biomass. However, dodecane addition during the exponential phase significantly impaired cell growth and carbon source utilization. After dodecane was added, DO rapidly increased and remained around 80%, while biomass accumulation slowed, suggesting reduced methanol utilization. After 153 h of methanol induction, the OD₆₀₀ reached only 397, and the retinol titer was 1194.86 mg/L (Figure 8A). This may be explained by interference of dodecane with methanol utilization, as K. phaffii relies on methanol as its carbon source. Although dodecane had no adverse effects in shake‐flask cultures, its negative impact became evident during scale‐up, likely due to increased culture volume and non‐uniform mass transfer.

FIGURE 8.

Production of retinol in fed‐batch fermentation using the engineered strain R23 in a 1.5‐L fermenter. (A) Add dodecane during the logarithmic growth phase. Blue arrows represent the time of dodecane addition. (B) Add dodecane during the stationary growth phase. Blue arrows represent the time of dodecane addition. Data were presented as the mean ± s.d. (n = 3 biologically independent samples). bars represent s.d.

Although dodecane initially impaired cell growth, retinol continued to accumulate during the fed‐batch fermentation, indicating that the strain was still capable of utilizing methanol for product synthesis. To mitigate the inhibitory effect of dodecane on growth, its addition was postponed until the growth plateau. At 24 h after methanol induction, 0.5% (w/v) BHT was added and replenished every 48 h. At 108 h of induction, when the OD₆₀₀ had stabilized, 20% (v/v) dodecane was introduced. After 168 h of methanol induction, the biomass reached an OD₆₀₀ of 671, and the retinol titer reached 3379.38 mg/L (Figure 8B). This represents the highest retinol yield reported in K. phaffii to date and provides a new strategy for in situ extraction during high‐density fermentation of methanol‐utilizing strains.

4. Conclusions

This study represents the first achievement in the de novo synthesis of high‐value retinol from a one‐carbon feedstock, demonstrating the potential of K. phaffii as an efficient cell factory for retinol production. Building upon a previously constructed β‐carotene‐producing K. phaffii strain, an engineered yeast capable of de novo retinol biosynthesis was developed by integrating the retinol pathway. A modular metabolic engineering strategy was applied to enhance precursor supply and carbon source utilization, resulting in an 82.60% increase in retinol production. On this basis, overexpression of exogenous and endogenous transporters with retinol transport functions significantly promoted extracellular secretion. Notably, the protein encoded by chr1‐4_0619 was identified as an efficient native retinol transporter in K. phaffii. Through shake‐flask fermentation optimization and supplementation with the antioxidant BHT, the retinol titer reached 321.97 mg/L. During high‐cell‐density fed‐batch fermentation, further optimization of the timing of extractant addition increased the yield to 3379.38 mg/L. Overall, our constructed retinol cell factory demonstrates an efficient and promising platform for retinol biosynthesis. This work provides novel strategies and methods for the microbial production of high‐value compounds, particularly from one‐carbon feedstocks.

Although multiple strategies have been attempted in this study to improve retinol production, the production capacity of the recombinant strain constructed in this study still has room for improvement compared to previously reported highest retinol yields achieved using S. cerevisiae [44] or Y. lipolytica [13] as hosts. Future research may emphasize cofactor engineering to improve cofactor availability, directed enzyme engineering to enhance the catalytic efficiency of key enzymes, and pathway optimization to increase precursor flux, thereby further improving the retinol production capacity of recombinant K. phaffii strains. Additionally, in this study, an endogenous retinol transporter in K. phaffii was identified; however, more efficient retinol transporters remain to be explored. Future work could focus on further analyzing and comparing the structures of transporter proteins, paying special attention to conserved domains, and mining more efficient retinol transporters.

Author Contributions

Chengan Liu: Methodology and visualization. Yanxuan Wu: Writing – original draft; formal analysis, and data curation. Yuchen Jiang: Validation and visualization. Yuqi Zhou: Methodology. Shuli Liang: Methodology and conceptualization. Xinying Zhang: Methodology and investigation. Ying Lin: Conceptualization, supervision, and funding acquisition.

Funding

This work was supported by the Guangdong S&T Program (2024B1111150001).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: biot70288‐sup‐0001‐SuppMat.docx.

Data Availability Statement

Data will be made available on request.