Engineering ergothioneine production in Yarrowia lipolytica

Ergothioneine is a naturally occurring antioxidant that has shown potential in ameliorating neurodegenerative and cardiovascular diseases. In this study, we investigated the potential of the Crabtree‐negative, oleaginous yeast Yarrowia lipolytica as an alternative host for ergothioneine production. We expressed the biosynthetic enzymes EGT1 from Neurospora crassa and EGT2 from Claviceps purpurea to obtain 158 mg·L−1 of ergothioneine in small‐scale cultivation, with an additional copy of each gene improving the titer to 205 mg·L−1. The effect of phosphate limitation on ergothioneine production was studied, and finally, a phosphate‐limited fed‐batch fermentation in 1 L bioreactors yielded 1.63 ± 0.04 g·L−1 ergothioneine in 220 h, corresponding to an overall volumetric productivity of 7.41 mg·L−1·h−1, showing that Y. lipolytica is a promising host for ergothioneine production.

The amino acid-derived nutraceutical ergothioneine (ERG) has recently gained much scientific interest for its potential application in preventing or treating neurodegenerative and cardiovascular diseases [1,2]. Current scientific understanding attributes its capacity in ameliorating disease to its potent antioxidant properties [3], as oxidative damage is commonly present in chronic, inflammatory diseases [4]. While ERG is ubiquitous in nature [5], it is only produced by fungi and bacteria. To date, no higher eukaryotes have been reported to biosynthesize ERG.
Ergothioneine is biosynthesized from the precursors histidine, cysteine, and S-adenosylmethionine (SAM; Fig. 1). In the fungal pathway, histidine is methylated three times by Egt1 to form hercynine before the same enzyme attaches cysteine to generate hercynylcysteine sulfoxide. The b-lyase enzyme Egt2 then dissociates ammonium pyruvate from the intermediate, and a subsequent reduction of the sulfur produces ERG. The bacterial pathway (Fig. S1) has the disadvantage of using five enzymes compared to two, and ATP is used to form other pathway intermediates that include glutamate, which has to be cleaved off in later steps.
Both pathways have previously been used to produce ERG in cell factories (Table 1). Aspergillus oryzae produced 231 mg of ERGÁkg À1 solid media after integrating multiple copies of the EGT1 and EGT2 genes of Neurospora crassa [6]. Escherichia coli produced 24 mgÁL À1 of ERG after expression of the Mycobacterium smegmatis egtBCDE genes and the addition of amino acid precursors to the medium [7]. Further metabolic engineering for improved cysteine and SAM biosynthesis enhanced the production in E. coli to 1.3 gÁL À1 in 216 h when precursors were supplemented to the cells [8]. Previously, we screened various combinations of fungal and bacterial ERG biosynthesis enzymes to find the optimal combination for ERG production in Saccharomyces cerevisiae. Two copies of the N. crassa EGT1 and the Claviceps purpurea EGT2 gene produced 598 mgÁL À1 of ERG in an 84-h fedbatch cultivation where additional amino acid precursors were fed to the cells [9].
In contrast to S. cerevisiae, Y. lipolytica is a Crabtree-negative yeast, so it does not have an extensive overflow metabolism in the presence of sugar excess [17] and is therefore much easier to ferment at large scale. The aim of this study was to explore Y. lipolytica as the host for the production of ERG.

Strains and chemicals
The Y. lipolytica strain ST6512 [18] was used as the background strain for metabolic engineering. Escherichia coli DH5a was used for all cloning procedures, propagation, and storing of plasmids. ERG (catalog # E7521-25MG, ≥ 98% purity) was bought from Sigma-Aldrich (St. Louis, MO, USA). Synthetic genes were ordered through the Gen-eArt Gene Synthesis service of Thermo Fisher Scientific (Waltham, MA, USA) or the custom gene synthesis service of IDT (Newark, NJ, USA). Sequencing results were obtained through Eurofins Genomics (Ebersberg, Germany) using their Mix2Seq kit.

Cloning
Strain construction for the integrations in Y. lipolytica was performed using the EasyCloneYALI method [18]. After transformation with plasmids, E. coli was grown on LB plates with 100 mgÁL À1 of ampicillin. For the selection of Y. lipolytica strains after modification with Cas9 plus gRNA, YPD plates supplemented with 250 mgÁL À1 of nourseothricin were used. Strains were checked for correct genetic modification by colony PCR. A list of the used genes, primers, biobricks, plasmids, and strains can be found in Tables S1-S6.

Small-scale cultivation conditions
For checking initial ERG production, a single colony of the respective strains was inoculated in 5 mL of mineral medium in a 13-mL preculture tube and cultured two times overnight at 30°C and 250 r.p.m. ERG production of the strains was tested in mineral medium by cultivating for 48  at 30°C and 250 r.p.m. Precultures for phosphate-limited experiments were made by inoculating a single colony of ST10264 in 5 mL of mineral medium without phosphate, supplemented with 0.04 gÁL À1 of KH 2 PO 4 , and culturing the strain three times overnight at 30°C and 250 r.p.m. The correlation of biomass to phosphate concentration was performed by inoculating strain ST10264 in mineral medium without phosphate with either 20 gÁL À1 of glucose or 50 gÁL À1 of glucose, supplemented with various concentrations of KH 2 PO 4 , at OD 600 = 0.1 in 2 mL of medium in 24 deepwell plates. The strain was then cultivated for 72 h at 30°C and 250 r.p.m. The OD 600 values were determined using the NanoPhotometer Pearl (Implen, Munich, Germany). Phosphate-limited ERG production was determined by

HPLC analyses
Ergothioneine was quantified by HPLC, similar to our previous work [9]. The ERG extraction was performed similar to Ref. [19]. ERG concentrations were determined by taking a 1 mL sample of the cultivation broth and immediately boiling the sample at 94°C for 10 min, with subsequent vortexing at 1600 r.p.m. for 30 min using a DVX-2500 Multi-Tube Vortexer from VWR. The vortexed samples were centrifuged at 10 000 g for 5 min, and the supernatant was taken and analyzed using HPLC. If storage was necessary, samples were stored at À20°C. For HPLC analysis, the Dionex Ultimate 3000 HPLC system with the analysis software CHROMELEON (Thermo Fisher Scientific) was used. Samples were run on a Cortecs UPLC T3 reversed-phase column (particle size 1.6 µm, pore size 120 A, 2.1 9 150 mm). The flow rate was 0.3 mLÁmin À1 , starting with 2.5 min of 0.1% formic acid, going up to 70% acetonitrile, 30% 0.1% formic acid at 3 min for 0.5 min, after which 100% 0.1% formic acid was run for 4-9 min. ERG was detected at a wavelength of 254 nm. For analysis of bioreactor samples, we additionally quantified glucose, ethanol, pyruvate, and acetate concentrations by HPLC as described [20].
A single colony from a YPD plate with ST10264 colonies was used to inoculate 5 mL of mineral medium with 20 gÁL À1 of glucose but without phosphate, supplemented with 0.04 gÁL À1 of KH 2 PO 4 in a 13-mL preculture tube. The tube was incubated at 30°C and 250 r.p.m. two times overnight. Two milliliters of this preculture were used to inoculate 98 mL of mineral medium with 20 gÁL À1 of glucose but without phosphate, supplemented with 0.04 gÁL À1 of KH 2 PO 4 , in a 500-mL baffled shake flask. The shake flask was then incubated two times overnight at 30°C and 250 r.p.m. The culture was centrifuged at 3000 g for 5 min, the supernatant was decanted, and the pellet was resuspended in 50 mL of sterile MilliQ water. The precultured cells were then used to inoculate 0.7 L batch phase medium in a single 1-L Sartorius bioreactor. The starting OD 600 was 0.2. The stirring rate was set at 800 r.p.m., the airflow was set at 0.5 SLPM, the temperature was kept at 30°C, and pH was maintained at pH 5.0 using 2 M of KOH and 2 M of H 2 SO 4 . The stirring was controlled by the level of dissolved oxygen in the solution. If it dropped below 40%, the stirring was increased up to 1200 r.p.m. The feeding was started after 40 h when all the glucose was consumed. The airflow was set to 1.0 SLPM during the feeding. The starting feed rate at 40 h was 2.6 gÁh À1 (2.0 mLÁh À1 ). At 64 h, the feed rate was increased to 3.9 gÁh À1 (3.0 mLÁh À1 ). At 76 h, the feed rate was decreased to 3.3 gÁh À1 (2.5 mLÁh À1 ). At 124 h, the feed rate was decreased to 2.6 gÁh À1 (2.0 mLÁh À1 ) until the feeding was stopped at 172 h. Antifoam was added as necessary.
A single colony from a YPD plate with ST10264 colonies was used to inoculate 5 mL of mineral medium with 20 gÁL À1 of glucose but without phosphate, supplemented with 0.04 gÁL À1 of KH 2 PO 4 in a 13-mL preculture tube. The tube was incubated at 30°C and 250 r.p.m. two times overnight. Two milliliters of this preculture were used to inoculated 98 mL of mineral medium with 20 gÁL À1 of glucose but without phosphate, supplemented with 0.04 gÁL À1 of KH 2 PO 4 , in a 500-mL baffled shake flask. The shake flask was then incubated two times overnight at 30°C and 250 r.p.m. The culture was centrifuged at 3000 g for 5 min, the supernatant was decanted, and the pellet was resuspended in 50 mL of sterile MilliQ water. The precultured cells were then used to inoculate 0.6 L of batch phase medium in 1-L Sartorius bioreactors in triplicate. The starting OD 600 was 0.2. The temperature was kept at 30°C, and pH was maintained at pH 5.0 using 2 M of KOH. The dissolved oxygen level controlled a cascade for the stirring and airflow. If the dissolved oxygen level dropped below 40%, the stirring was first increased from 800 to a maximum of 1200 r.p.m., followed by the airflow from 1.0 SLPM to a maximum of 1.5 SLPM. The feeding was started after 40 h when all the glucose was consumed. The starting feed rate at 40 h was 2.6 gÁh À1 (2.0 mLÁh À1 ). At 64 h, the feed rate was increased to 3.3 gÁh À1 (2.5 mLÁh À1 ). At 76 h, the feed rate was increased to 4.0 gÁh À1 (3.0 mLÁh À1 ). At 88 h, the feed rate was increased to 4.7 gÁh À1 (3.5 mLÁh À1 ). Finally, at 100 h, the feed rate was increased to 5.3 gÁh À1 (4.0 mLÁh À1 ) until the feeding was stopped at 160 h. Antifoam was added as necessary.

Integration of the ergothioneine biosynthesis pathway
To implement ERG biosynthesis in Y. lipolytica, we chose the combination of enzymes that performed optimally in S. cerevisiae, viz. the EGT1 gene from N. crassa (Genbank accession XP_956324.3) and the EGT2 gene from C. purpurea (Genbank accession CCE33140.1) [9]. The genes were codon-optimized for Y. lipolytica expression and placed under the control of either TEFintron promoter or GPD promoter in the ST6512 strain. ST6512 is derived from Y. lipolytica W29 by integrating the cas9 gene and knocking out of the ku70 gene to reduce the nonhomologous end joining [18]. While both TEFintron and GPD promoters are strong constitutive promoters, TEFintron promoter is several-fold stronger [18]. The expression of a single copy of the pathway allowed Y. lipolytica to produce 141 or 158 mgÁL À1 ERG, depending on the promoter choice (Fig. 2). The higher titer was obtained for the strain expressing NcEgt1 under the control of TEFintron promoter, which is in line with the results in S. cerevisiae, where an additional copy of N. crassa EGT1 improved the production of ERG, but not an extra copy of C. purpurea EGT2 [9]. Interestingly, these titers, obtained in a simple small-scale batch cultivation, were 3-fold higher than the titers obtained for an analogous S. cerevisiae strain cultivated in simulated fed-batch medium and 9-fold higher than S. cerevisiae titers in batch medium [9]. As the next step, we integrated another copy of both EGT1 and EGT2 genes to obtain strain ST10264, which produced 205 mgÁL À1 of ERG (Fig. 2). Comparatively, the ERG titer in E. coli expressing the bacterial pathway on high-copy number plasmids was approximately 22 mgÁL À1 [8]. These results show that Y. lipolytica is a suitable host for the production of ERG.

Phosphate limits biomass accumulation but not ergothioneine production in small-scale cultivations
Biomass concentration should be carefully considered for a successful fed-batch cultivation of the ERGproducing ST10264 strain. Yarrowia lipolytica grows to cell densities as high as 80-120 gÁL À1 during fed-batch cultivation depending on the substrate, as evidenced in [13,21,22]. As the biosynthesis pathway for ERG requires oxygen, high cell density in fed-batch conditions can be disadvantageous for the production of ERG. The solution was to limit the final biomass by means other than the typical carbon limitation in fedbatch cultivations. Nitrogen limitation is often used in Y. lipolytica cultivations, but nitrogen limitation triggers downregulation of amino acid synthesis. The carbon flux is then redirected to lipid metabolism [23]. Since ERG is derived from the amino acids histidine and cysteine, and the amino acid-derived SAM, nitrogen limitation should not be used to limit the biomass concentration in fed-batch cultivation.
Therefore, it was investigated if the biomass of strain ST10264 could be limited by the phosphate concentration instead (Fig. S2). Phosphate is typically used to buffer the medium, so we used MES hydrate for pH buffering instead. We tested two glucose concentrations (20 or 50 gÁL À1 ) and different KH 2 PO 4 concentrations (0-0.3 gÁL À1 ) for cellular growth in 24-deep-well plates (Fig. S2). In parallel, we cultivated the same strain in shake flasks with 0.04 gÁL À1 of KH 2 PO 4 and 50 gÁL À1 of glucose to determine if ERG would be produced after the strain stopped growing. Indeed, while the OD 600 increased by only ca. 30% in Fig. 2. ERG production titers of engineered Yarrowia lipolytica strains (n = 3, error bars signify SD). The NcEGT1 and CpEGT2 genes were integrated in single copies under different promoters, or in two copies. A two-tailed Student's t-test was used to determine if the difference was significant between the control strain ST6512 and the strain with the integrated ERG biosynthesis pathway (***P-value < 0.0005). the second half of cultivation, ERG concentration doubled during the same period (Fig. 3B). These experiments indicated that phosphate limitation is a viable strategy for ERG fermentation.
Producing ergothioneine using phosphate-limited fed-batch fermentation When limiting the biomass using the phosphate concentration in 24-deep-well plates, the OD 600 values increased by 4-5 units per 0.01 gÁL À1 of KH 2 PO 4 in the KH 2 PO 4 concentration range of 0.01 gÁL À1 to 0.08 gÁL À1 (Fig. S2), corresponding to an increase of ca 0.48-0.60 gÁL À1 cell dry weight (CDW) per 0.01 gÁL À1 of KH 2 PO 4 . We then designed an initial fermentation experiment to confirm that the KH 2 PO 4 concentration limited the biomass to the same level in a bioreactor. Therefore, we added 0.24 g of KH 2 PO 4 per 1 L of final volume of fed-batch cultivation in a bioreactor, meaning that 0.34 gÁL À1 of KH 2 PO 4 was added to the starting volume of 0.7 L, to reach 15-20 gÁL À1 CDW during the fermentation. The results of this experimental phosphate-limited fed-batch fermentation in a single bioreactor are shown in Fig. S3. After the initial batch phase of 40 h, the glucose feeding was started at a rate of 1.4 gÁh À1 (2.0 mLÁh À1 feed solution) and was adjusted throughout the fermentation to ensure that the added glucose was not entirely consumed by the cells.
The strain produced ERG throughout the fed-batch fermentation, with the highest productivity of 10.07 mgÁL À1 Áh À1 ERG when phosphate was not limiting the growth. However, when phosphate became limiting, the productivity dropped to 2.77 mgÁL À1 Áh À1 ERG. Curiously, strain ST10264 reached 32-33 gÁL À1 CDW already at 76 h, and the biomass concentration did not increase until 100 h. Even though the biomass concentration was higher than the aimed 15-20 gÁL À1 of CDW, the dissolved oxygen level did not drop below 40%, meaning no oxygen limitation was observed. Therefore, 60 mg of KH 2 PO 4 was added to the bioreactor at 100 h to study its effect on biomass concentration, and the additional KH 2 PO 4 increased the CDW of ST10264 to 40 gÁL À1 . Furthermore, the productivity increased to 5.01-6.59 mgÁL À1 Áh À1 ERG upon addition of potassium phosphate, as calculated using either the lower ERG titers or higher ERG titers between 136 and 208 h. Surprisingly, the strain generated approximately 1.33 gÁL À1 of CDW per 0.01 gÁL À1 of KH 2 PO 4 , more than double the level of biomass compared to the experiment in 24-deep-well plates.
From the results of this trial fermentation, we estimated that a biomass concentration of approximately 60 gÁL À1 of CDW would not cause oxygen limitation issues. Hence, we added 0.5 g of KH 2 PO 4 per 1 L of the final volume of fed-batch cultivation in bioreactors, equivalent to 0.83 gÁL À1 of KH 2 PO 4 in the starting volume of 0.6 L. After inoculation at OD 600 = 0.2, the cells were grown for 40 h in the batch phase to accumulate enough biomass to start feeding. The dissolved oxygen level was controlled through a cascade of stirring (800-1200 r.p.m.) and airflow (1.0-1.5 SLPM). The feeding was started after the batch phase concluded at 2.6 gÁh À1 (2.0 mLÁh À1 ) and kept at that level for 24 h. After that, the feed rate was increased by 0.65 gÁh À1 (0.5 mLÁh À1 ) every 12 h until glucose started accumulating in the medium. The feed rate reached a maximum of 5.3 gÁh À1 (4.0 mLÁh À1 ) between 100 and 160 h, after which the feeding was stopped. Figure 4 details the results of the phosphate-limited fed-batch fermentation in 1 L bioreactors performed in triplicate.
Overall, strain ST10264 produced 1637 AE 41 mgÁL À1 of ERG after 220 h, out of which 26% was extracellular and 74% was intracellular. The overall productivity was 7.41 AE 0.19 mgÁL À1 Áh À1 ERG. Both the overall titer and productivity were higher than obtained with E. coli, which produced 1.3 gÁL À1 ERG with a productivity of 6.02 mgÁL À1 Áh À1 [8], and S. cerevisiae producing 598 mgÁL À1 ERG at a rate of 7.12 mgÁL À1 Áh À1 [9]. However, ERG yield was highest at 16.36 mgÁL À1 Áh À1 until 100 h, when phosphate became limiting, and glucose started accumulating in the medium. After phosphate became limiting, the ERG titers remained approximately the same between 100 and 160 h, while the volume inside the bioreactor increased at a rate of .0 mLÁh À1 . The overall productivity at 100 h was 12.73 mgÁL À1 Áh À1 , which dropped to 8.03 mgÁL À1 Áh À1 at 160 h. Thus, we estimated the productivity between 100 and 160 h to be 4.70 mgÁL À1 Áh À1 to account for the volume increase in this timeframe. After the feeding stopped and excess glucose was consumed by the cells, the ERG titer increased from 1285 to 1637 mgÁL À1 between 160 and 220 h, showing a productivity of 5.87 mgÁL À1 Áh À1 in this period.
The final biomass concentration of the fermentation was 60.6 gÁL À1 of CDW, close to the target of 60 gÁL À1 of CDW of the experimental design. However, the biomass concentration was 45-46 gÁL À1 CDW until 196 h. The lower than expected biomass concentration until 196 h could have been caused by increased use of phosphate for general metabolite production in the early fed-batch phase, as the ERG productivity until 100 h was 16.36 mgÁL À1 Áh À1 , compared to 10.07 mgÁL À1 Áh À1 in the early fed-batch phase of the fermentation with a lower amount of phosphate. The sudden increase in biomass concentration occurred when excess glucose in the medium reached below 10 gÁL À1 glucose and became limiting. We hypothesize that the lower glucose concentration initiated a change in metabolism that allowed the generation of additional biomass.
The production of ERG by Y. lipolytica had another notable advantage over E. coli and S. cerevisiae. In both the E. coli and the S. cerevisiae studies, the fermentations were supplemented with additional amino acid precursors to improve titers (Table 1) [8,9]. Furthermore, E. coli was additionally supplemented with ammonium ferric citrate. The fermentation media used here only contained an additional 1 gÁL À1 of pyridoxine, a cofactor for the CpEgt2 enzyme. Pyridoxine was supplemented to the media in the E. coli and S. cerevisiae studies at 10 and 0.5 gÁL À1 , respectively [8,9]. Thus, using Y. lipolytica as a host also improved the cost-effectiveness of ERG production.
In conclusion, we integrated the ERG biosynthesis pathway into oleaginous yeast Y. lipolytica. Phosphate limitation was a viable strategy to limit biomass accumulation in the bioreactors. ERG titer reached 1.63 AE 0.04 gÁL À1 after 220 h of fermentation in mineral medium with glucose as the only carbon source.

Conflict of interest
SH, DBK, and IB are named inventors on a European Patent application covering parts of the work described above.

Author contributions
IB and DBK conceived the study; IB and JLM supervised the study; SAH, IHJ, and JLM designed the experiments; SAH, IHJ, and MR performed the experiments and data analysis; SAH and IB wrote the manuscript; SAH, IB, DBK, IHJ, and JLM reviewed and revised the manuscript; IB and JLM provided resources for the study; IB acquired funding for the study.

Data accessibility
The data that support the findings of this study are available in Figs 1-4 and the supplementary material of this article.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Bacterial ergothioneine biosynthesis pathway. Fig. S2. The effect of phosphate limitation on the growth of the ergothioneine-producing Yarrowia lipolytica strain ST10264. Fig. S3. Ergothioneine production and biomass accumulation of ST10264 under phosphate-limited fed-batch conditions in a single 1 L bioreactor with an initial amount of 240 mg KH 2 PO 4 . Table S1. List with genes and their DNA sequences used in this paper. Table S2. List of primers used in this study for cloning purposes. Table S3. List of primers used in this study for sequencing. Table S4. List of biobricks used in this study. Table S5. List of plasmids used in this study, made by USER cloning.