Metabolic engineering of Saccharomyces cerevisiae for neoxanthin production

Abstract

Background

Xanthophylls, a subclass of oxygenated carotenoids, are highly valued for their wide range of applications in the food and pharmaceutical industries, particularly due to their antioxidant properties and potential health benefits. Among these, neoxanthin, a less studied xanthophyll, has demonstrated significant therapeutic potential, including antioxidant and anticancer activities. Neoxanthin is also the primary precursor for the synthesis of other valuable compounds, such as fucoxanthin and β-damascenone, which are important in the cosmetic and pharmaceutical sectors.

Results

In this study, we report the first heterologous production of neoxanthin in Saccharomyces cerevisiae through a combination of metabolic and enzyme engineering. First, a S. cerevisiae strain was engineered to produce neoxanthin by expressing genes from the β-carotene and violaxanthin biosynthesis pathways. Following this, the VDL1 gene from Phaeodactylum tricornutum, responsible for converting violaxanthin into neoxanthin, was expressed, resulting in the production of 0.18 mg/g_DCW of neoxanthin. To further enhance production, a pulse-fed galactose strategy was employed during shake-flask growth, leading to a 2.5-fold increase in neoxanthin yield. Additionally, transmembrane peptides were incorporated into the yeast cells to improve the accumulation of carotenoids, generating an increase of 3.8-fold, achieving a final production of 0.7 mg/g_DCW of neoxanthin.

Conclusions

This is the highest reported yield of neoxanthin produced by engineered microorganisms, and the strategies employed here have considerable potential for scaling up production of this carotenoid.

Background

Xanthophylls, a subclass of oxygenated carotenoids, are recognized for their wide-ranging applications in the food and pharmaceutical sectors, primarily as natural colorants and antioxidants [52]. These compounds are essential in various biological mechanisms, including photosynthesis in plants and photoprotection in microorganisms [9]. Apart from their biological significance, carotenoids are valued for their antioxidant properties, contributing to human nutrition and health [23]. Within xanthophylls, neoxanthin, although less studied, is a promising carotenoid due to its unique characteristics and potential health benefits. Kotake-Nara et al. [36] indicated that neoxanthin can inhibit the proliferation of human prostate cancer cells. Furthermore, neoxanthin offers greater protection than β-carotene or lutein against oxidative stress in HepG2 cells [62] and kidney cells [22]. Additionally, neoxanthin serves as a precursor to fucoxanthin, another carotenoid with significant therapeutic potential [13, 17, 44, 46].

Despite its presence in various plant sources, the concentration of neoxanthin is relatively low and subject to seasonal fluctuations. In plants, carotenoid levels can vary by 20–30% depending on the season [2], and neoxanthin typically accounts for only 9 to 14% of total xanthophylls [50]. Additionally, the extraction of neoxanthin from plant sources presents several challenges: (i) low natural abundance requires large amounts of biomass, (ii) seasonal variability affects carotenoid content and yield consistency; (iii) plant cultivation and harvesting entail high resource and labor; (iv) extraction methods often rely on organic solvents, raising environmental and safety concerns. These limitations compromise scalability, cost-effectiveness, and sustainability. In contrast, microbial hosts, particularly Saccharomyces cerevisiae, represent a promising platform for the biosynthesis of neoxanthin, enabling controlled, reproducible, and season-independent production, and thus providing a more stable and efficient alternative to traditional extraction methods.

Saccharomyces cerevisiae is a well-established model organism known for its rapid growth, well-characterized genetics, and ease of manipulation, making it an ideal candidate for metabolic engineering aimed at enhancing carotenoid production [65]. Recent advancements in synthetic biology and genetic tools have facilitated the optimization of yeast strains for efficient carotenoid biosynthesis [64]. Various molecular approaches have been employed to increase carotenoid yield, including adaptive laboratory evolution (ALE) for β-carotene [56] and lycopene [59]. López et al. [38] demonstrated a remarkable increase in β-carotene production through a gene dosage strategy using CRISPR/Cas9, enhancing yield from 4 to 32 mg/g. The ability to modulate gene expression, engineer pathways, and optimize fermentation conditions has significantly improved both the yield and diversity of carotenoids produced by S. cerevisiae.

In the realm of carotenoid biosynthesis, various xanthophylls have been successfully synthesized in yeast, including astaxanthin [63, 67], zeaxanthin [57], and, more recently, violaxanthin [14, 15]. Neoxanthin biosynthesis follows the β-xanthophyll pathway, initiated from β-carotene (Fig. 1). Initially, β-carotene hydroxylase (CrtZ) catalyzes the hydroxylation of β-carotene, producing zeaxanthin as the pathway's first xanthophyll. Subsequently, zeaxanthin is converted to violaxanthin through the action of zeaxanthin epoxidase (ZEP). The enzymes responsible for converting violaxanthin to neoxanthin remained elusive for years.

Fig. 1

Expression of engineered neoxanthin biosynthetic pathway in S. cerevisiae. Carotenoid precursor FPP is generated by the endogenous mevalonate pathway, which is potentiated by the overexpression of tHMG1. The β-carotene pathway incorporates the heterologous genes CrtE, CrtYB and CrtI to synthesize β-carotene from FPP. The introduction of the β-xanthophyll pathway leads to the conversion of β-carotene into neoxanthin. The sequential hydroxylation and epoxidation in both rings of β-carotene by CrtZ and ZEP yield zeaxanthin and violaxanthin, respectively. The tautomerization of violaxanthin catalyzed by VDL1 yield neoxanthin. FPP farnesyl diphosphate, GGPP geranylgeranyl diphosphate, CrtE geranylgeranyl diphosphate synthase, CrtYB bifunctional lycopene cyclase/phytoene synthase, tHMG-CoA HMG-CoA reductase truncate version, VDL1 violaxanthin de-epoxidase like 1, ZEP zeaxanthin epoxidase, CrtZ β-carotene hydroxylase, CrtI phytoene desaturase

Functional studies have suggested the involvement of ABA4, NSY, and NXD1, which are genes that may encode proteins involved in neoxanthin biosynthesis in land plants [3, 10, 48, 49]. However, when these enzymes were expressed in a violaxanthin-producing Escherichia coli strain, they failed to produce neoxanthin [32], indicating that the mechanism in plants may require additional components or function differently outside of the native context. Notably, land plants lack VDL genes, which are present in certain algae. In this context, Dautermann et al. [18] identified violaxanthin de-epoxidase-like 1 (VDL1) as essential for neoxanthin synthesis in chromalveolate algae and proposed that neoxanthin biosynthesis evolved independently in plants and algae through convergent evolutionary processes. More recently, Higuchi et al. [32] assessed the expression of Phaeodactylum tricornutm VDL1 gene in a violaxanthin-producing strain of E. coli, achieving a neoxanthin concentration of up to 0.25 mg/L, thereby confirming the enzymatic activity associated with this gene.

In this study, we engineered the yeast Saccharomyces cerevisiae to accumulate neoxanthin by employing a combination of enzyme engineering and metabolic engineering techniques. Through the screening of various truncated enzyme variants, we established an initial strain capable of producing neoxanthin. The overproduction of neoxanthin was achieved by anchoring key enzymes to membranes and adjusting gene dosage. Lastly, culture optimization strategies were implemented to enhance carotenoid accumulation. To our knowledge, this study marks the first instance of heterologous neoxanthin production in yeast and reports the highest yield of microbially produced neoxanthin to date.

Methods

Genes and plasmids

VDL1 of Phaedofactylum tricornutum VDL1 (PtVDL1) and Thalassiosira pseudonana (TpVDL1) genes were codon-optimized and synthesized by Genscript (Nanjing, China). Genes coding for CrtE, CrtYB, CrtI, tHMG1 of Xanthophyllomyces dendrorhous and tHMG1 of S. cerevisiae were kindly provided by Dr. Lopez [38]. Genes coding for CrtZ of Pantoea ananatis, tZEP of Haematococcus lacustris, tFD3 (ferredoxin truncate version) and tRFNR1 (ferredoxin-NADPH oxidoreductase truncate version) of Arabidopsis thaliana were facilitated by Dr. Cataldo [14, 15]. The integration sites were amplified by PCR from the plasmid library established by Mikkelsen et al. [45]. The promoters GAL1, GAL10 and GAL7 were PCR-amplified from the genomic DNA of S. cerevisiae, and the GAL2 promoter was PCR-amplified from the genomic DNA of Saccharomyces eubayanus. VDL1 with transmembrane fusion were constructed by Gibson assembly. For the construction of VLD1 fusion proteins containing transmembrane domains, the nucleotide sequence encoding the transmembrane residues of the ZEP enzyme was amplified by PCR, while the peptide of rat fyn kinase aminoacidic sequence was obtained from Werner et al., [66].

Yeast integrative expression vectors were constructed using Uloop assembly as described by Pollak et al. [53], with the plasmid library pCA from Pollak et al. [54]. L0 plasmids were assembled through the Gibson assembly technique [27], which included the removal of BsaI and SapI recognition sites, along with the construction of genes and integration sites. Genes and the L0 backbone vector were amplified using Phusion High Fidelity Polymerase (Thermo Scientific, USA), following the manufacturer’s instructions. The resulting PCR products were purified with the Wizard SV Gel and PCR Clean-Up System kit (Promega, USA). Purified DNA fragments were combined with 1.33× Gibson master mix—containing isothermal buffer, T5 exonuclease (0.005 U/μL), Phusion DNA polymerase (0.03 U/μL), and Taq DNA ligase (5.3 U/μL)—in a final volume of 10 μL and incubated at 50 °C for 60 min. The reaction products were then transformed into E. coli Top10 cells (Thermo Fisher Scientific, USA). Finally, assembled plasmids were purified using the E.Z.N.A Plasmid Mini Kit (Omega Bio-Tek, USA) and verified by sequencing (Macrogen, South Korea). A list of all primers used in this study is provided in Table S1.

Strain construction

The CEN.PK113-5D strain (MATa; ura3-52; TRP1; LEU2; HIS3; MAL2-8^C; SUC2) was used as the parental strain for the construction of the β-xanthophyll-producing strains (Table 1). Integrative plasmids were linearized by NotI digestion (New England Biolabs, USA), and transformed using the CRISPR/Cas9 method, following the protocol provided by Tom Ellis lab, which is freely accessible on the Benchling webpage [58]. Briefly, yeast transformation was performed using the PEG/LiAc/SS carrier DNA method [28] combined with CRISPR/Cas9 to integrate markerless integration cassette strains. Plasmids containing ZEP (zeaxanthin epoxidase) displayed high toxicity in E. coli and proved unclonable through Uloop or Gibson assembly. To overcome these issues, ZEP expression cassettes were assembled and genome-integrated using a modified yeast homologous recombination method [14].

Table 1 Strains constructed in this study

Yeast transformants were plated on a synthetic complete medium without uracil (SC-URA; 1.8 g/L yeast nitrogen base, 5 g/L ammonium sulfate, 0.8 g/L CSM-Ura mixture, 20 g/L of glucose, and 20 g/L of agar) and incubated for three days at 30 °C. Genomic PCR was employed to verify the integrations of plasmids and expression cassettes. For subsequent strain transformations, selected transformants were cultured in YPD liquid medium (1% yeast extract, 2% peptone, 2% glucose) and were plated on both YPD and SC-URA plates. This process was repeated until no colonies appeared on SC-URA plates, indicating the loss of the CRISPR/Cas9 plasmid used for transformation. All the strains constructed in this study are listed in Table 1.

Shake flask cultures

For precultures, single colonies were selected from agar plates and inoculated in tubes containing 5 mL of YPD medium (1% yeast extract, 2% peptone, 2% glucose). These cultures were incubated overnight at 30 °C and 160 rpm in an orbital shaker. For shake flask cultures, the precultures were transferred to 100 mL baffled shake flasks containing 25 mL of YPDG (1% yeast extract, 2% peptone, 2% glucose and 2% galactose) at an initial optical density (OD₆₀₀) of 0.1. The cultures were then grown under the same conditions for 72 h. Biomass concentrations were determined by measuring OD₆₀₀ using a Genesys 20 spectrophotometer (Thermofisher, USA), with a linear relationship of 1 OD₆₀₀ = 0.4 g_DCW/L determined experimentally.

For kinetic studies, pre-cultures were inoculated into 125 mL baffled shake flasks containing 30 mL of YPD₁G₁ (2.6% yeast extract, 2% peptone, 1% glucose and 1% galactose) or YPD₂G₂ (2.6% yeast extract, 2% peptone, 2% glucose and 2% galactose) with an initial OD₆₀₀ of 0.1. The cultures were grown under the same conditions for 120 h. In the case of cultures with YPD₁G₁, after glucose consumption, pulses of 10 g/L of galactose were added to the flasks at 24 h intervals until 96 h of cultivation.

Carotenoid extraction

For each sample, 30 mg of biomass were centrifuged, and the resulting supernatant was discarded. Carotenoid extraction was carried out through sequential homogenization of the cells in hexane and ethanol. Initially, the cells were mixed with 400 μL of 0.5 mm zirconia/silica beads (BioSpec products, USA) and 1 mL of hexane, followed by lysing for 10 min using a Genie cell disruptor (Scientific Industries, USA). The disrupted mixture was then centrifuged, and the supernatant was collected in a 1.5 mL microcentrifuge tube. The resultant pellet underwent a second extraction with 1 mL of ethanol 100%, utilizing the same disruption program. Both hexane and ethanol extractions were repeated until a white pellet was obtained (usually, a total of 4 extractions were needed). The supernatant fractions were evaporated in a HyperVAC-MAX centrifugal vacuum evaporator (Gyrozen, South Korea) at 60 °C and 1000 rpm for the hexane fraction and at 65 °C for the ethanol fraction. Finally, all carotenoid dry fractions were resuspended and combined in a total volume of 1 mL of acetone, followed by centrifugation at 20,000×g to eliminate insoluble material. The acetone supernatant phase was analyzed by spectrophotometry and UPLC-MS.

Carotenoids analysis

Total carotenoids were quantified by measuring absorbance at 453 nm using a Genesys 20 spectrophotometer (Thermo Fisher, USA). The total carotenoid concentration was determined using a calibration curve from 0.2 to 6 mg/L of β-carotene standards (Sigma-Aldrich, USA). Individual carotenoid identification and quantification were conducted using UPLC-MS on a Dionex Ultimate 3000 system (Thermo Fisher Scientific, USA) coupled to an Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific, USA) equipped with electrospray ionization (ESI), and a C30 reverse-phase column (YMC, Japan). Two mobile phases were employed: mobile phase A (acetonitrile: formic acid: water, 100:0.1:0.1 v/v) and mobile phase B (acetone: formic acid: water, 100:0.1:0.1 v/v). The elution gradient was as follows (min-%A): 0–70; 21–70; 25–30; 27–5; 40–1; 56–1; 57–70; 65–70, with a flow rate of 1 mL/min. The column temperature was maintained at 35 °C, and detection was performed at 453 nm. Carotenoid concentration was determined using calibration curves from 1 to 20 mg/L of β-carotene, β-cryptoxanthin, zeaxanthin, antheraxanthin, and violaxanthin standards (Carotenature, Switzerland). Unidentified carotenoids concentrations were estimated using the mean response factor from identified carotenoids. Mass spectra parameters were as follows: positive polarity, scan range from 400 to 1000 m/z, resolution of 140,00 sheath gas flow of 60 AU, auxiliary gas flow of 40 AU, sweep gas flow of 0 AU, spray voltage of 7 kV, capillary temperature of 350 °C, auxiliary gas temperature of 300 °C and S-lens RF level of 100.

Bioinformatic and statistical analysis

For the design of truncated enzyme variants, putative cleavage sites of signal peptide and plastid transit peptide of diatom enzymes were determined analyzing the full protein sequences with SignalP5.0 server [4] and TargetP2.0 server [6] Consensus transmembrane regions of the ZEP enzyme were predicted by TOPCONS server [61].

Experimental data were obtained from three independent shake flask cultures of five different colonies and are presented as mean ± standard deviation for strain screening. For kinetic studies, three independent shake flask cultures of the best-performing strain were analyzed and results are also presented as mean ± standard deviation. Statistical significance was assessed using an unpaired Student’s t-test, with p-values < 0.05 considered significant. All analyses were performed using GraphPad Prism 8.

Results

VDL1 of Phaeodactylum tricornutum catalyzes neoxanthin production, with N-terminal truncation leading to increased enzyme activity

Neoxanthin is biosynthesized through the β-xanthophyll pathway (Fig. 1). Its formation involves the opening of one cyclohexenyl 5-6-epoxide ring of violaxanthin and generation of an allene bond, a reaction catalyzed by violaxanthin de-epoxidase-like enzyme (VDL1). Two VDL1 homologs were evaluated in this study: one from Phaeodactylum tricornutum and the other from Thalassiosira pseudonana. To assess their activity, a violaxanthin-producing strain was constructed through sequential transformation of a β-carotene-producing strain (β-CAR), incorporating the β-xanthophyll pathway intermediates zeaxanthin (ZEA) and violaxanthin (VIO) (Figs. 1 and 2B).

Fig. 2

Characterization of neoxanthin-producing strains. A Evaluation of different N-terminal truncated variants of VDL1 of Phaeodactylum tricornutum (NEO.P) and Thalassiosira pseudonana (NEO.T) in the violaxanthin producing strain (VIO). B Compared carotenoid accumulation in strains constructed to produce β-carotene (β-CAR), zeaxanthin (ZEA), violaxanthin (VIO) and neoxanthin (NEO.P80). Carotenoid levels were measured in five independent colonies per strain, error bars correspond to the standard deviation. C Growth and carotenoid accumulation kinetics of NEO.P80 strain. Error bars correspond to the standard deviation of triplicate shake flask cultures of the best-performing neoxanthin-producing clone were analyzed. (*) indicate statistically significant differences (p < 0.05)

Considering the plastidial localization of VDL1 enzymes, various N-terminal truncations targeting the predicted signal and transit peptides were tested (Fig. 2A). No statistically significant differences in total carotenoid content were observed among the strains expressing VDL1 variants (Table S2). Both full-length and N-terminally truncated forms of the enzymes were expressed in the VIO strain. Notably, the VDL1 homolog from T. pseudonana (NEO.T) failed to produce detectable levels of neoxanthin in either configuration. In contrast, the P. tricornutum VDL1 homolog (NEO.P) exhibited enzymatic activity, and truncation of the first 80 N-terminal residues (NEO.P80) led to a two-fold increase in neoxanthin production. The presence of neoxanthin was confirmed via HPLC–UV and UPLC–MS analysis (Fig. S1).

A comparative carotenoid profile analysis was performed between the neoxanthin-producing strain NEO.P80 and the parental β-CAR, ZEA, and VIO strains (Fig. 2B). The β-CAR strain accumulated 1.2 mg/g_DCW of β-carotene, with a titer of 9.7 mg/L and biomass production of 8.1 g/L. In contrast, the ZEA strain, expressing the β-xanthophyll pathway, accumulated 4.8 mg/g_DCW of β-xanthophylls, reaching a total carotenoid titer of 14 mg/L, while biomass decreased to 2.8 g/L. Subsequent integration of zeaxanthin epoxidase (ZEP) and its redox partners (FD3 and RFNR1) to obtain the VIO strain resulted in reduced carotenoids levels (3.6 mg/g_DCW), lower titers (6.1 mg/L) and biomass (1.6 g/L). Upon integration of the tr80-PtVDL1 gene to produce neoxanthin, the NEO.P80 strain accumulated 3.3 mg/g_DCW of total carotenoids, reaching a titer of 5.3 mg/L, with biomass production remaining at 1.6 g/L.

The expression of the β-xanthophyll pathway in β-carotene-producing strains led to a notable reduction in final biomass after 72 h of culture. This biomass decrease was consistently observed in the ZEA, VIO, and NEO strains. The VIO strain showed a shift in the carotenoid profile: β-carotene (0.89 mg/g_DCW) became the major compound, followed by violaxanthin (0.44 mg/g_DCW), while zeaxanthin levels dropped nearly 30-fold compared to the ZEA strain. Similarly, NEO.P80 strain showed a total carotenoid titer and overall carotenoid composition comparable to the VIO strain, with a 15% less in violaxanthin production (0.38 mg/g_DCW) and generation of 0.19 mg/g of neoxanthin due to VDL1 reaction. The introduction of the violaxanthin and neoxanthin biosynthetic modules resulted in the appearance of several unidentified carotenoids in both strains. Figure S2 shows a comparative chromatogram ZEA and NEO. P80 strains. To further characterize the NEO.P80 strain, growth kinetics and violaxanthin and neoxanthin accumulation were monitored over time (Fig. 2C). After 72 h of culture, neoxanthin levels reached 0.18 mg/g_DCW (Fig. 2C). Given that galactose consumption was incomplete at this time point (Fig. 2C), we hypothesized that extending the culture duration might enhance neoxanthin production.

Galactose pulses improve the production of neoxanthin in the NEO.P80 strain

Kinetic analysis of the NEO.P80 strain (Fig. 2C) suggested the potential for increased carotenoid accumulation, thereby improving neoxanthin yield. To explore this possibility, the culture duration was extended from 72 to 120 h (Fig. 3A). After 120 h, galactose was fully consumed, resulting in a total carotenoid yield of 2.5 mg/g_DCW. However, this prolonged cultivation did not significantly enhance neoxanthin production, which increased only slightly from 0.18 to 0.19 mg/g_DCW (Fig. 3B).

Fig. 3

β-carotenoids accumulation kinetics with different growth culture strategies. A Growth and total carotenoids production of the NEO.P80 strain with 2% of galactose in shake flasks. B β-carotenoids accumulation kinetics in the NEO.P80 strain with 2% of galactose. C Growth and total carotenoid accumulation kinetics with galactose pulses in shake flasks. D β-carotenoids accumulation kinetics with 10 g/L galactose pulses in shake flasks. Error bars correspond to the standard deviation of triplicate shake flasks of the best-performing neoxanthin-producing clone

Both kinetic profiles (Figs. 2C, 3A) revealed an extended lag phase in galactose consumption, which delayed the onset of carotenoid biosynthesis. Active galactose uptake (0.22 g g_DCW⁻¹ h⁻¹) and carotenoid production began only after 48 h. To overcome this limitation, a galactose pulsing strategy was implemented (Fig. 3C). Cultures were initiated with 1% glucose and 1% galactose, and additional galactose (10 g/L) was added in pulses to maintain a continuous supply for carotenoid synthesis. This approach successfully reduced the lag phase from 48 to 24 h, with an initial specific galactose uptake rate of 0.10 g g_DCW⁻¹ h⁻¹ for the first 30 h, and the uptake rate increased with the galactose pulses to 0.30 g g_DCW⁻¹ h⁻¹. Under this culture condition, the total carotenoid yield doubled to 5.2 mg/g_DCW (Fig. 3D). Consequently, neoxanthin production increased to 0.57 mg/g_DCW, representing a 2.5-fold improvement compared to the non-pulsed condition (Fig. 3D).

Although the implementation of the galactose pulsing strategy significantly improved overall carotenoid yields and neoxanthin accumulation, the relative proportion of neoxanthin remained low compared to other β-xanthophyll intermediates. This observation suggested metabolic bottlenecks may be limiting flux toward neoxanthin synthesis. To address this, we explored a gene dosage strategy aimed at enhancing the conversion efficiency of violaxanthin to neoxanthin by introducing extra copies of key pathway enzymes, specifically the truncated VDL1 variant (tr80-PtVDL1) and CrtZ.

Impact of constitutive tr80-VDL1 and an extra copy of CrtZ on neoxanthin and violaxanthin biosynthesis

The NEO.P80 strain primarily accumulated violaxanthin and β-carotene as the most abundant carotenoids. To enhance neoxanthin production, a second copy of the truncated 80-residue variant of PtVDL1 (tr80-PtVDL1) was integrated, generating the NEO.2 strain. In parallel, the NEO.Z strain was constructed by introducing an additional copy of CrtZ, aiming to reduce β-carotene accumulation by converting it to zeaxanthin and thereby redirecting the β-xanthophyll flux toward neoxanthin synthesis (Fig. 4A).

Fig. 4

Effect of tr80-VDL1 and CrtZ extra copy in the strain NEO.P80. A Neoxanthin and violaxanthin accumulation in NEO.P80, NEO.2 and NEO.Z. Carotenoid levels were measured in five independent colonies per strain. B Shake flask culture of the neoxanthin accumulation in the NEO.2 strain with the galactose pulse protocol. C Growth and total carotenoids kinetics in the NEO.2 strain with the galactose pulse protocol. Triplicate shake flask cultures of the best-performing NEO.2 clone were analyzed. D β-carotenoids accumulation kinetics of the NEO.2 clone. Error bars corresponde to the standard deviation

Although no significant differences in total carotenoid yield were observed among the strains (Table S2), changes in the relative distribution of xanthophylls were detected (Fig. 4A). The NEO.2 strain exhibited a decrease in violaxanthin levels (0.28 mg/g_DCW) compared to NEO.P80, along with a slight, but not statistically significant, increase in neoxanthin accumulation (from 0.19 to 0.22 mg/g_DCW). In contrast, the NEO.Z strain showed a higher accumulation of violaxanthin (0.49 mg/g_DCW), resembling to the VIO strain, and a marked decrease in neoxanthin content relative to NEO.P80 (from 0.19 to 0.097 mg/g_DCW).

In view of the moderate improvement in neoxanthin levels in the NEO.2 strain, this variant was selected for evaluation under the galactose pulse protocol (Fig. 4B, C). However, implementation of this strategy did not result in a significant increase in neoxanthin yield when compared to the parental NEO.P80 strain (Figs. 3D, 4D).

Considering that VDL1 is a soluble enzyme lacking transmembrane domains (Fig. S3), it may exhibit limited interaction with the membrane-bound carotenoid substrates. We hypothesized that anchoring the truncated tr80-PtVDL1 variant to cellular membranes could enhance substrate accessibility and thereby improve neoxanthin production.

Targeting tr-80-ptVDL1 to cell membrane increases neoxanthin production

The VDL1 was predicted to be a soluble protein, likely exhibiting limited interaction with membranes compared to other carotenoid biosynthetic enzymes containing transmembrane domains [66] (Fig. S3). To enhance neoxanthin production, we targeted the truncated variant tr80-PtVDL1 to cellular membranes, hypothesizing that increased substrate accessibility would improve enzymatic activity (Fig. 5A). For membrane anchoring, transmembrane peptides were fused to either the N- or C-terminus of tr80-PtVDL1. Two different transmembrane domains were tested: one derived from the ZEP enzyme (TM, Table 1) and the other from rat Fyn kinase, as described by Werner et al. [66]. Additionally, two distinct linkers, one flexible and one rigid, were employed to connect the TM peptides to tr80-PtVDL1 (Table 1). The Fyn peptide included its own specific linker, as reported by McCabe and Berthiaume [43].

Fig. 5

Effect of tr80-PtVDL1 membrane targeting on neoxanthin production. A Schematic representation of the constructed tr80-PtVDL1 membrane tag variants. B Neoxanthin production using the different tr80-PtVDL1 membrane tag variants. Carotenoid levels were measured in five independent colonies per strain. C Growth and total carotenoids kinetics in the NEO.VDL1FL_TM strain with the galactose pulse protocol. D β-carotenoids accumulation kinetics of the NEO.VDL1FL_TM strain. *p < 0.05, **p < 0.01, error bars corresponde to the standard deviation

No significant differences in total carotenoid yields were observed among the variant strains (Table S2). Among the various transmembrane peptide and linker combinations, the NEO.VDL1FL_TM strain exhibited the highest neoxanthin accumulation, showing a 2.5-fold increase relative to the parental NEO.P80 strain (0.5 mg/g_DCW, Fig. 5B). Other variants also demonstrated improved neoxanthin production compared to the untagged tr80-PtVDL1, with approximately 1.9-fold increases observed for TM_RLVDL1 and NEO.VDL1FYN strains. However, the N-terminal fusion with the Fyn peptide (NEO.FYNVDL1) resulted in loss of enzyme activity.

As a result of the enhanced neoxanthin production in the NEO.VDL1FL_TM strain, we applied the galactose pulsing protocol (Fig. 5C, D). This approach further increased neoxanthin accumulation, reaching a maximum of 0.7 mg/g_DCW, corresponding to a 1.4-fold improvement over the NEO.P80 strain under identical cultivation conditions.

Discussion

Most enzymes involved in the β-xanthophylls pathway are derived from plastids, including ZEP and VDL1. The ZEP enzyme was utilized to produce violaxanthin in S. cerevisiae by Cataldo et al. (2020), where truncation variants were studied. Similar to ZEP, VDL1 is also located in plastids. However, this enzyme originates from the diatom P. tricornutum, which is classified as a heterokont. The structure of plastids in this organism differs from those in plants. Heterokonts are characterized by possessing chloroplasts with four membranes, including a chloroplast endoplasmic reticulum membrane [5]. Consequently, proteins destined for these chloroplasts require a bipartite targeting sequence in their amino acid chain: a signal peptide that mediates entry into the intermembrane space of the endoplasmic reticulum, and a transit peptide that facilitates subsequent translocation across the outer chloroplast membrane. Following import, both the signal and transit peptides are cleaved, yielding the mature, functional form of the protein [41]. In this context, we evaluated truncated variants of VDL1 (Fig. 2A): the full-length version (NEO.P0 strain), a version lacking the signal peptide (NEO.P20 strain), and a version lacking both the signal and transit peptides (NEO.P80 strain). Both the full-length enzyme and the version lacking only the signal peptide displayed some activity; however, the most effective variant expressed in yeast was the construct lacking both targeting peptides. This truncated form, NEO.P80, mimics the mature enzyme after import into the chloroplast, suggesting that removal of the targeting sequences is necessary for full enzymatic activity in a heterologous system.

Notably, some strains such as β-CAR, ZEA and NEO.P80 showed high variation in carotenoid across colony replicates (Fig. 2), which may reflect biological variability among yeast transformants. Such variability is common in S. cerevisiae and can result from transformation-associated heterogeneity, including differences in plasmid copy number, genomic integration sites, and stochastic gene expression [11, 47]. The strains in this study were generated via double-site homologous recombination, a method intended to generate single-copy integrations without replication of the locus. This approach reduces the likelihood of looping-out events, thereby increasing genomic stability and population uniformity. However, even in single-copy strains, high clonal variability has still been documented in yeast [8], potentially leading to transformants with distinct metabolic outputs. The experiments designed to explore different culture conditions were carried out using replicates of the best-performing clone to eliminate clonal variability interference in the results (Figs. 2C, 3, 4C, D and 5C, D),

Carotenoid production in S. cerevisiae can be achieved through various metabolic engineering strategies. In this study, the production of β-xanthophylls was achieved by constructing a β-carotene-producing strain (β-car, Fig. 2B), which served as the precursor for zeaxanthin (strain ZEA, Fig. 2B), violaxanthin (VIO, Fig. 2B), and ultimately the target compound neoxanthin (NEO.P80, Fig. 2B). Overexpression of the β-carotene pathway genes, including CrtE and CrtI, along with reduction of the mevalonate pathway bottleneck via additional copies of HMG1 [38], and the introduction of the zeaxanthin biosynthetic enzyme CrtZ, led to enhanced production of this xanthophyll, which became the dominant product (Fig. 2B). The high efficiency of CrtZ in converting β-carotene to zeaxanthin has been previously reported [14, 15]. However, the production of polar β-ring carotenoids such as zeaxanthin resulted in reduced yeast growth, in contrast to strains producing only non-polar carotenoid as β-carotene. As xanthophyll biosynthesis progressed toward neoxanthin, a further decrease in biomass production was observed. This may be due to the distinct orientation of polar versus non-polar carotenoids within the membrane. Polar β-xanthophylls are known to interact with membrane phospholipids via their polar β-rings, adopting a diagonal transmembrane orientation within the lipid bilayer [29], whereas non-polar carotenes are embedded within the hydrophobic core of the membrane, adopting more flexible and undefined orientations. These xanthophylls also alter membrane properties [30], reducing membrane fluidity by rigidifying it in the liquid crystalline phase through interactions with phosphatidylcholine headgroups, while non-polar carotenes have considerably lower effects on membrane properties. Consequently, the incorporation of xanthophylls may compromise yeast growth by impacting membrane dynamics. This negative effect on biomass is particularly evident in strains VIO and NEO.P80, where unidentified carotenoids, presumably hydroxy-keto derivatives produced by the promiscuous activity of CrtZ [25], increase the proportion of polar to non-polar carotenoids. To further characterize the strain NEO.P80, growth and production kinetics were carried out (Fig. 2C).

After the depletion of glucose and partial consumption of galactose, the strain transitioned into a lag phase during which no carbon sources were utilized. Interestingly, the ethanol generated during glucose fermentation remained unconsumed. Active galactose uptake begins only after 48 h of cultivation, suggesting an adaptation during the lag phase. To the best of our knowledge, this phenotype of slow galactose consumption, characterized by delayed utilization beginning around 48 h, has not been previously reported. The metabolic response of S. cerevisiae to mixed carbon sources such as glucose and galactose is well described. It includes glucose-mediated repression of galactose catabolism [37], regulation of the GAL regulon, and defined galactose uptake following glucose depletion [51]. The metabolic shift of glucose to galactose consumption depends on initial relative concentrations of these sugars [21]. In S. cerevisiae, glucose is the preferred carbon source and actively represses alternative sugar metabolism via the transcriptional repressor Mig1. Upon glucose depletion, activation of Snf1 kinase leads to Mig1 inactivation and de-repression of the GAL regulon. In the presence of galactose, Gal3 binds and sequesters the Gal80 repressor, releasing the activator Gal4 to induce GAL1, GAL7, and GAL10 [37]. These genes encode enzymes required to convert galactose to glucose-1-phosphate for glycolysis [33]. This metabolic switch typically involves a lag phase of 2–6 h [24], depending on strain and culture conditions. In wild-type S. cerevisiae, the specific galactose uptake rate is approximately 0.5 g g_DCW⁻¹ h⁻¹ [12, 19], while strain NEO.P80 displays markedly lower rates (0.17 g g_DCW⁻¹ h⁻¹) The unusual phenotype observed in NEO.P80 kinetics (Fig. 2C) suggests an altered metabolic state, probably due to the presence of toxic or inhibitory compounds in response to galactose induction, leading to reduced growth. To investigate this phenotype further, cultures were extended from 72 to 120 h (Fig. 3A). The kinetics were recapitulated, with an extended lag phase in galactose consumption. Once the strain overcomes the lag phase, galactose consumption and carotenoid biosynthesis rates remain constant, suggesting a possible adaptation to the toxic compound or inhibitory mechanisms.

Although xanthophyll production was initially hypothesized to impair growth (Fig. 2B), the cultivation kinetics (Figs. 2C, 3A) do not support this assumption. Growth arrest occurs immediately after glucose depletion and before the onset of carotenoid biosynthesis. Carotenoid production coincides with the initiation of active galactose consumption at approximately 48 h. These observations led to the hypothesis that galactose-induced toxicity or inhibition precedes carotenoid synthesis. One plausible hypothesis to explain this phenomenon is that the expression of one or more heterologous proteins may lead to protein misfolding, triggering cytosolic unfolded protein response (CUPR) and inducing metabolic stress [20, 35]. Recombinant protein production in S. cerevisiae has been reported to activate the unfolded protein response (UPR) in the endoplasmic reticulum, which can impair cellular growth [60]. A comparable phenomenon has been described for the cytosolic unfolded protein response (CUPR), where the accumulation of misfolded heterologous proteins in the cytosol negatively impacts cell growth and physiological performance [26] Moreover, the xanthophyll biosynthetic pathway is driven by strong GAL promoters (GAL1, GAL2, GAL7, GAL10) [34], which could exacerbate protein misfolding and CUPR activation. An alternative hypothesis is that metabolic insufficiency may limit galactose incorporation due to sequestration of Gal3 and Gal4 proteins for transcription of heterologous genes rather than activation of GAL1, GAL7, and GAL10 [1]. The neoxanthin-producing strain carries thirteen heterologous genes integrated into the genome, all under GAL promoter control (GAL1, GAL2, GAL7, GAL10), potentially depleting the pool of available transcription factors. Future investigations employing comparative transcriptomics and proteomics will help elucidate the underlying inhibitory mechanisms in the strain. To test whether this stress could be mitigated, cultures were initiated with lower galactose concentrations and supplemented with pulses to sustain gene expression (Fig. 3C). During the first 30 h, galactose consumption occurred at 0.1 g g_DCW⁻¹ h⁻¹, the same as to the late phase of cultures with 2% initial galactose, and carotenoid production correlated with this rate. A subsequent 10 g/L galactose pulse further increased carotenoid yield. These results indicate that reduced initial galactose levels may limit the inhibitory mechanisms. Lower galactose concentrations have been shown to reduce expression of some GAL genes, such as GAL1, compared to higher concentrations such as 20 g/L [31]. In accordance with the proposed hypotheses, a lower galactose concentration may reduce excessive heterologous protein expression and prevent the activation of the CUPR, or reduce transcription factor titration, thereby eliminating the adaptation phase associated with metabolic inhibition.

To improve neoxanthin production, increasing gene dosage is a commonly employed strategy [16, 38,39,40, 67]. In the NEO.P80 strain, a second copy of the tr80-PtVDL1 gene was integrated to catalyze the conversion of violaxanthin to neoxanthin, alongside an additional copy of CrtZ to increase metabolic flux toward neoxanthin synthesis. However, these genetic modifications did not result in higher neoxanthin accumulation; instead, violaxanthin levels increased (Fig. 4A). Dautermann et al. [18] assessed the enzymatic activity of VDL1 in vitro and confirmed its role in the tautomerization of violaxanthin to neoxanthin. Their study also demonstrated that VDL1 can catalyze the tautomerization of neoxanthin back to violaxanthin, with a product ratio of approximately 25:75 (neoxanthin:violaxanthin) at chemical equilibrium. Neoxanthin and violaxanthin ratio in the NEO.P80 strain remained within this reported ratio (approximately 30:70, neoxanthin:violaxanthin, Fig. 4C), therefore, even with increased enzyme expression, violaxanthin formation remains unavoidable due to this intrinsic reversibility. Future efforts should thus prioritize protein engineering to enhance the enzyme’s specificity and catalytic efficiency toward neoxanthin formation.

Despite the inherent reversibility, we hypothesize that tr80-PtVDL1 remains a limiting factor in neoxanthin synthesis, possibly due to poor accessibility to its membrane-associated precursors. tr80-PtVDL1 encodes a cytosolic, soluble enzyme, as confirmed by TOPCONS analysis, which revealed the absence of transmembrane domains (Fig. S3). Since neoxanthin precursors are embedded in the membrane, spatially separated from the cytosolic enzyme, substrate accessibility may be restricted. To address this, membrane-targeting tags were fused to the enzyme to improve substrate proximity and xanthophyll production [42, 55]. The membrane-associated reaction could present a new condition characterized by a different equilibrium constant and a new neoxanthin-to-violaxanthin ratio limit.

Two types of tags were tested: a ZEP-derived transmembrane peptide and the fyn peptide. The ZEP tag was used to colocalize tr80-PtVDL1 with the ZEP enzyme, while the fyn peptide served as a lipid-anchoring domain representative of a helical structure [43]. Both membrane-targeting strategies led to a significant improvement in neoxanthin yield. Notably, the highest yield was observed when the ZEP transmembrane peptide was fused to the C-terminus of tr80-PtVDL1 via a flexible linker. In contrast, attaching the fyn peptide to the N-terminus reduced enzymatic activity, suggesting possible interference with substrate access to the active site. To investigate this, a structural model of VDL1 was generated based on the known structure of VDE [7] (Fig. S4), revealing the likely entrance path for carotenoid substrates. These findings indicate that both the position and flexibility of the linker used for membrane anchoring can critically influence enzyme performance, potentially by facilitating or obstructing substrate access to the active site.

Conclusions

To date, recombinant neoxanthin production has only been demonstrated in E. coli [32]. In this study, we employed various metabolic engineering strategies to enhance neoxanthin biosynthesis in S. cerevisiae. Strategies included increasing gene dosage of the truncated PtVDL1 enzyme, implementing a galactose pulsing feeding regime to sustain prolonged induction of carotenoid biosynthesis, and engineering membrane localization of the soluble VDL1 enzyme via fusion with heterologous transmembrane peptides to improve substrate accessibility. The combined application of membrane anchoring, and galactose pulse feeding resulted in a marked increase in neoxanthin titer, achieving 0.7 mg/g_DCW, representing a 3.7-fold improvement compared to the parental strain without these modifications. These findings highlight the importance of enzyme spatial organization relative to hydrophobic substrates in the carotenoid biosynthetic pathway. Further optimization may be achieved by integrating complementary strategies such as accelerated laboratory evolution to enhance strain robustness, targeting carotenoid biosynthesis to specific organelles (e.g., peroxisomes) to increase substrate channeling, or directed protein engineering of key enzymes to improve catalytic efficiency and pathway flux.