Engineering the oleaginous red yeast Rhodotorula glutinis for simultaneous β-carotene and cellulase production

Rhodotorula glutinis, an oleaginous red yeast, intrinsically produces several bio-products (i.e., lipids, carotenoids and enzymes) and is regarded as a potential host for biorefinery. In view of the limited available genetic engineering tools for this yeast, we have developed a useful genetic transformation method and transformed the β-carotene biosynthesis genes (crtI, crtE, crtYB and tHMG1) and cellulase genes (CBHI, CBHII, EgI, EgIII, EglA and BGS) into R. glutinis genome. The transformant P4-10-9-63Y-14B produced significantly higher β-carotene (27.13 ± 0.66 mg/g) than the wild type and also exhibited cellulase activity. Furthermore, the lipid production and salt tolerance ability of the transformants were unaffected. This is the first study to engineer the R. glutinis for simultaneous β-carotene and cellulase production. As R. glutinis can grow in sea water and can be engineered to utilize the cheaper substrates (i.e. biomass) for the production of biofuels or valuable compounds, it is a promising host for biorefinery.


Establishing a heterologous gene expression platform in R. glutinis. The result of minimal inhibitory concentration (MIC) (Supplementary
) showed that the R. glutinis wild type cannot grow in neither 100 nor 200 μg/ml concentration of any of the three antibiotics tested (Zeocin, G418 and Hygromycin), so either concentration can be used to select the transformants. For initial transformation, three gene expression cassettes (Table 1) including phytoene desaturase gene (crtI), cellobiohydrolase (cbhI) and G418 selection marker gene (KanMx) were integrated into the R. glutinis genome, using frozen protoplast and lithium acetate competent cells 18,31 . The transformants were screened using YP2D supplemented with G418 (200 μg/ml) and the wild type without expression cassette was used as a control. A total of 200 transformants were randomly selected and subcultured for 3 generations to select stable transformants ( Supplementary Fig. S1). Then transformants containing both crtI and cbhI were validated by PCR using the gene specific primer pairs (Supplementary Table S2). Due to the high GC content of R. glutinis, the PCR amplification of specific genes from their genomic DNA was found to be a difficult task 22,32 . Hence, the genes were amplified using long-range primer pairs and modified touchdown PCR conditions and the wild type was used as a negative control. The results showed that P4-10-9 was successfully transformed and contained the crtI, cbhI and KanMx genes ( Supplementary Fig. S2). This is the first study to integrate both β-carotene biosynthesis genes and cellulase genes into the R. glutinis genome by electroporation. Our gene cassettes were flanked with the specific homologous recombination (HR) region at both ends but we could not confirm the inserts in those regions (data not shown). Previous studies suggested that Rhodotorula had a low HR efficiency 33,34 . This suggests that the genes were integrated into the R. glutinis genome by non-homologous  Fig. S3). Previous studies applied ATMT in some recalcitrant transformation hosts including R. glutinis 22,35 . However, ATMT is not suitable for the integration of multiple genes at the same time.
Improvement of β-carotene production by installing genes into R. glutinis genome. Based on a previous study 9 , we tested different carbon sources to compare the β-carotene production between wild type and P4-10-9. Three different carbon sources were tested: 2% glucose, 2% galactose and 2% glycerol. The HPLC data showed that P4-10-9 produced more carotenoids than the wild type using galactose or glycerol as a carbon source, compared to glucose (Fig. 1). Similar results were obtained by Johns, Love and Aves 23 , who found that the ICL promoter was inhibited by glucose. It might be because the crtI gene was driven by the ICL promoter, so P4-10-9 could produce more carotenoids.
To further improve the β-carotene production, additional functional genes including crtYB and crtE were integrated into the P4-10-9 genome using electroporation, and transformants were screened using zeocin. 72 transformants were selected and subcultured for 3 generations to obtain stable transformants. The genomic integration of crtYB and crtE was confirmed by PCR using long-range gene-specific primer pairs (Supplementary  Table S2). The PCR amplification ( Supplementary Fig. S4) confirmed the integration of crtE in transformant 72 and it was named as P4-10-9-72. Similarly, transformant 63 was integrated only with crtYB and named as P4-10-9-63Y. To confirm the β-carotene production improvement, transformants P4-10-9-63Y and P4-10-9-72 and wild type R. glutinis were cultured in 10 ml YP2Gly for 2 weeks and total carotenoids were extracted. The HPLC data showed that both transformants produced higher β-carotene than the wild type (Fig. 2). Especially, β-carotene production of P4-10-9-63Y was improved by 20 folds (4.50 ± 0.46 mg/g), thus confirming the importance of crtYB in β-carotene biosynthesis. However, the HPLC data of P4-10-9-72 showed a small β-carotene peak at around the 18 min retention time and several new peaks were observed between 15-17 min (Fig. 2), which might be due to the production of torulene and torularhodin. Indeed, Frengova and Beshkova 36 demonstrated the production of torulene and torularhodin by R. glutinis. Furthermore, transformant P4-10-9-72 showed red color and previous studies found that torulene and torularhodin showed characteristic red color 37,38 . Transformant P4-10-9-72 contains both crtI and crtE, which might improve the precursor availability for the torulene and torularhodin pathways 39,40 . The higher β-carotene producing transformant P4-10-9-63Y showed yellow color. Similarly, Bhosale and Gadre observed yellow color in R. glutinis after UV treatment, which produced increased β-carotene compared to the parent strain 41 . Hence, the higher β-carotene producing P4-10-9-63Y was selected as a host for further study.
The β-carotene production of transformant P4-10-9-63Y was further improved by transforming the upstream genes (tHMG1 and crtE) of the β-carotene biosynthetic pathway. A total of 146 transformants were selected on YP2D plates with hygromycin as a selective marker and subcultured for 3 generations. The PCR results ( Supplementary Fig. S5) showed that 43 transformants containing all the four β-carotene biosynthesis pathway genes and selected for functional assays. These 43 transformants could be stably subcultured on selective plates supplemented with G418, zeocin and hygromycin. The β-carotene production of 43 transformants were analyzed and almost every transformant showed improved β-carotene production (Fig. 3). We obtained transformants containing different gene arrangements ( Table 2). P4-10-9, P4-10-9-63Y and P4-10-9-63Y-14B were selected for further analysis because they contained different arrangements of tHMG1, crtI, crtE and crtYB. We measured the lycopene and β-carotene amounts and observed the color of colonies of transformants compared to the wild type ( Fig. 4). We reduced the growing time to avoid the decrease of β-carotene amount and used two carbon source according to Fig. 1. Therefore, we used 10 ml YP2G2Gly and cultured for 7 days. The transformant P4-10-9-63Y produced 8.4 fold higher β-carotene with a significant color change than the wild type ( Table 2, Fig. 4b,c). Therefore, the bi-functional enzyme crtYB (phytoene synthase and lycopene cyclase) appears to play an important role in the carbon flux from lycopene to β-carotene. In contrast, P4-10-9-63Y-14B contains tHMG1, crtI, crtE and crtYB, which improved the conversion of HMG-CoA to β-carotene. The engineered R. glutinis strain P4-10-9-63Y-14B produced 15.7 fold β-carotene and 2.7 fold higher lycopene compared to the wild type. This is the first study that showed a significant improvement of carotenoid production in the oleaginous red yeast.

Fold of carotenoid increase
Lycopene

β-carotene
Wild type --- Installation of cellulase ability in R. glutinis. To increase the value of R. glutinis as a host for biorefinery applications, the strain P4-10-9-63Y was selected as a host to be transformed with the cellulase genes to utilize the cellulosic biomass. Three types of cellulases (CBHI, CBHII, EgIII, EgI, EglA and BGS) were transformed into the R. glutinis genome according to our previous studies with other hosts 18,42,43 . The genomic integration of each gene was confirmed by PCR using gene specific primer pairs (Supplementary Table S2, [9][10][11][12][13][14][15][16][17][18][19][20]. The PCR data ( Supplementary Fig. S5) showed that several transformants contained all the cellulase genes. We first utilized mix colonies PCR to validate CBHII, EgIII, EglA and BGS genes and the results showed correct gene sizes as the positive control. Second, we tested CBHI gene by the single colony PCR and selected transformants containing the correct gene size. Then we further confirmed EgI in 43 transformants. The cellulase activity was screened using fast-cellulase screening (4% Sigmacell cellulose type 20 as the carbon source) and the data showed that almost all transformants had higher cellulase activity compared to the wild type strain (Fig. 3).
Demonstrating the biorefinery potential of the engineered R. glutinis. The engineered R. glutinis, P4-10-9-63Y-14B, −23C and −27C were selected based on their genotype (Fig. 5) and phenotype (Fig. 3). The four major properties, i.e., maximum β-carotene amount, total cellulase activity, total lipid and salt-tolerance of transformants, were analyzed. To analyze the maximum β-carotene amounts in selective transformants, we reduced medium from 10 ml to 7 ml because we speculated that the dissolved oxygen might improve the β-carotene amount. Therefore, the selective transformants were cultured in 7 ml YP2G2Gly for 1 week to analyze the maximum β-carotene amounts. Figure 6a showed the β-carotene amounts of selective transformants were increased compared to the wild type. The β-carotene amount of P4-10-9-63Y-14B was improved up to 27.13 ± 0.66 mg/g. To the best of our knowledge, this is the highest β-carotene amount ever produced by R. glutinis.
To investigate the total cellulase activities of P4-10-9-63Y-14B, −23C and −27C, we tested different cellulosic substrates including Sigmacell cellulose type 20, Sigmacell cellulose type 50 and cellulose microcrystalline. The 50-fold condensed supernatants of R. glutinis wild type, 14B, 23C and 27C were harvested and the same total protein concentration was used for the total cellulase activity assay 18 . Figure 6b and Supplementary Fig. S6 shows that we have successfully improved the total cellulase activity of R. glutinis by transforming these cellulase genes. The results showed that the total cellulase activities of 14B (0.017 U/μg), 23 C (0.016 U/μg) and 27 C (0.016 U/μg)  were only around 1.53-to 1.63-fold lower than the commercial enzyme Celluclast 1.5 L (0.026 U/μg) in Sigmacell cellulose type 20 substrate. This is the first study to engineer the cellulase genes into R. glutinis and significantly improved its total cellulase activities. Thus, the engineered R. glutinis strain might have potential to use the cellulosic substrate as a carbon source to produce a higher amount of β-carotene compared to the wild type ( Supplementary Fig. S7).
In order to make sure that the insertion of additional genes did not affect the intrinsic characteristics of R. glutinis, the total lipid content and salt-tolerance of 14B, 23C and 27C was analyzed. The total lipid weight indeed showed no significant difference between the wild type and transformants (Fig. 6c); the total lipid content was around 70% lipid yield of dry weight biomass, the same as described in a previous study 44 . We also tested the growth of the wild type and transformants in YP2D with 5% NaCl to study their salt tolerance. The results showed that all of the strains were able to grow well in YP2D with 5% NaCl (Fig. 6d). Although the initial growth of the wild type was faster than the transformants, all the strains reached similar OD 600 after 3 days culturing. We could not observe any significant growth difference between the wild type and transformants in YP2D (data not shown). Moreover, the total lipid content of both the wild type and the transformants were unaffected. So we suggest that the growth rate was not affected significantly. These results demonstrated that engineering the R. glutinis using our method did not strongly affect the lipid production and salt tolerance of R. glutinis.

Discussion
We have successfully integrated multiple gene expression cassettes including 4 β-carotene biosynthesis pathway genes and 6 cellulase genes into the genome of oleaginous red yeast R. glutinis. The transformant P4-10-9-63Y-14B produced up to 27.13 ± 0.66 mg/g of β-carotene amount. This was achieved by transforming four functional genes (tHMG1, crtI, crtE and crtYB) into the R. glutinis genome. A previous study achieved a similar result in Saccharomyces cerevisiae 27 . Although Li et al. 30 discovered several novel β-carotene improving genes, the β-carotene amount (5.9 ± 0.1 mg/g) produced in their engineered S. cerevisiae was still as low as that in Verwaal et al. (2007). Larroude et al. applied synthetic biology tools to engineer oleaginous yeast Yarrowia lipolytica to obtain the highest β-carotene producing yeast (89.6 mg/g) 29 . These researchers pointed out that β-carotene may be stored and protected in lipid droplets in oleaginous yeast 28 . Gao et al. and Larroude et al. utilized multiple-copy insertions to improve the β-carotene production. Although there is some difficulty in analyzing the copy number in our R. glutinis transformants (i.e. RNA extraction), the strategy in this study can improve β-carotene production by 7 to 15 folds. The highest β-carotene amount (DCW) in previous studies were found in Saccharomyces cerevisiae (18 mg/g) 45 , Escherichia coli (72 mg/g) 46 and Yarrowia lipolytica (89.6 mg/g) 29 . The β-carotene production of engineering strain might further improve by fed-batch cultivations in a bioreactor 28,29 . Our best producer transformant 14B is inferior only to E. coli and Y. lipolytica. Moreover, we might improve its β-carotene amount by a bioreactor in the future.
We improved the β-carotene production and installed cellulase ability (CBHI, CBHII, EgIII, EgI, EglA and BGS) at the same time. The cellulase from concentrated supernatants of P4-10-9-63Y-14B were only 1.53-fold lower than the commercial enzyme (Celluclast 1.5 L) purified from Trichoderma reesei. The commercial cellulases are produced by cellulolytic fungi, such as Trichoderma reesei and Aspergillus niger. Each host has its own limitation. For example, Neocallimastix patriciarum can produce highly efficient β-glucosidase but it is an obligate anaerobic rumen fungus 47 . Therefore, many studies engineered organisms with multiple cellulolytic properties to design an ideal cellulolytic host. Chang et al. 43 engineered a cellulolytic K. marxianus with the similar cellulase genes as in this study 43 . The total cellulase activity of R. glutinis can be further improved by synthetic biology strategy. The functional analysis suggested that transformants also maintained its original lipid yield and salt-tolerance ability. To produce β-carotene, we need to culture for 7 days to get the highest β-carotene amounts. Our transformants can reach the same OD600 as wild type after 3 days in YP2D with 5% NaCl. So we suggested that we can produce higher β-carotene amount even in salty condition. We confirmed that the total lipid content of both the wild type and the transformants were unaffected (around 70% lipid yield), thus further supporting our statement. From these experiments, it seems that our engineering strategy is good. Several reports have described the synthesis of carotenoids, cellulases or lipid from fungi. However, in previous studies the hosts were usually engineered to produce a single product. Our goal is to turn cellulolytic wastes into β-carotene and renewable energy by a single host. In conclusion, we have upgraded the potential applications of R. glutinis for biorefinery using a synthetic biology approach and demonstrated that R. glutinis is a potential host for biorefinery.

Materials and Methods
Strains, media, and growth conditions. The oleaginous yeast R. glutinis BCRC 22360 (Bioresource Collection and Research Center, Taiwan) was kindly provided by Dr. Hong-Wei Yen (Tunghai University, Taiwan). R. glutinis was cultured in different growth media at 30 °C with 300 rpm for several days. The basic growth media contained 1% Yeast Extract (BactoDifco) and 1% Peptone (BactoDifco) as a nitrogen source and 2% dextrose (2D), 2% galactose (2 G) or 2% glycerol (2Gly) as a carbon source. For screening the R. glutinis transformants, a previously described medium was used 48 . The lipid inducing media (70 g/L glucose, 0.75 g/L yeast extract, 1.7 g/L yeast nitrogen base without amino acids and ammonium sulfate, and 0.1 g/L (NH 4 ) 2 SO 4 , pH 5.6) were applied for analyzing the total lipid content in R. glutinis 24 .
Establishing a heterologous gene expression platform in R. glutinis. In this study, we intended to develop an efficient transformation tool for R. glutinis. R. glutinis electro-competent cells were prepared using two different methods. First, lithium acetate was used to prepare competent cells modified from previous studies for Kluyveromyces marxianus and Rhodotorula gracilis 18,49 . Briefly, R. glutinis cells were cultivated in 5 ml YP2D from the single colony at 30 °C with 300 rpm and then 0.2 OD cells were subcultured into 50 ml YP2D until reaching 0.6~1.4 OD. Cells were harvested at 3000 rpm for 3 min (4 °C) and washed with 5 ml ice-cold distilled H 2 O. Then, cells were resuspended in TMLSD buffer (10 mM Tris-HCl buffer (pH 8.0) containing 2 mM MgCl 2 , 100 mM lithium acetate, 270 mM sucrose, 10 mM dithiothreitol) and incubated at room temperature for 1 h. After the incubation, cells were harvested as described above and washed twice with TMS buffer (10 mM Tris-HCl buffer (pH 8.0) containing 2 mM MgCl 2 , 270 mM sucrose). Finally, competent cells were resuspended in TMS buffer and stored in −80 °C.
For the second method, we used the frozen protoplast protocol to engineer R. glutinis. The R. glutinis protoplast was prepared according to previous studies 31,50,51 with some modifications. Briefly, single colonies of R. glutinis were inoculated in 50 mL of YP2D medium at 30 °C for 15 h. Cells were harvested at 3000 rpm for 10 min and suspended in 20 mL distilled H 2 O. Cells were harvested again as described above and gently resuspended in 10 mL 1 M sorbitol, followed by harvesting and suspending in 10 mL of sorbitol, sodium citrate, EDTA and β-mercaptoethanol (SCEM). Then, cells were mixed with 40 μl of lyticase solution (25,000U/ml) and incubated at 30 °C for 1 h. After the lyticase digestion cells were suspended in SCEM (10 9 cells/ml), 1 ml cell suspension was added 0.5 ml of lytic enzyme solution (1.5% (w/v) Zymolyase 60,000) for overnight culturing. Cells were centrifuged gently at 300 g for 5 min in round-bottom plastic tubes and suspended in 10 mL 1 M sorbitol by gently tapping the tube. Then cells were centrifuged at 300 g for 5 min, and the supernatant was discarded. This procedure was repeated to remove lyticase thoroughly. Finally, cells were suspended in 2 mL CaST solution (CaCl 2 , sorbitol and Tris-HCl) along with 2 mL cell-storage solution and protoplast cells were stored at −80 °C. Transforming genes into the R. glutinis genome. The carotenoid biosynthetic pathway in Rhodotorula species has been well studied 3,36 . Generally, R. glutinis synthesizes β-carotene from the precursor acetyl-CoA ( Supplementary Fig. S8). To increase the β-carotene amount in R. glutinis, the carotenogenic genes including geranylgeranyl pyrophosphate synthase (crtE), phytoene desaturase (crtI) and phytoene synthase/lycopene cyclase (crtYB) from Xanthophyllomyces dendrorhous (red yeast) were selected and integrated into the R. glutinis genome. Similarly, for improving the metabolic flux of carotenoid pathway, a tHMG1 (a truncated 3-hydroxy-3 -methylglutaryl-coenzyme A reductase) gene was selected from Kluyveromyces marxianus 27,52 . The carotenogenic genes were in the pUC18 vector and the construction of each gene cassette was as in Chang et al. 52 . Briefly, the carotenogenic genes were cloned in between the yeast promoter and terminator using specific restriction enzymes. The gene cassettes were amplified using TransStart FastPfu Fly (Ultra) High-Fidelity DNA Polymerases and the primers pairs were listed in Table S2. For the simultaneous improvement of β-carotene production and cellulose utilization ability of R. glutinis, three types of fungal cellulase genes were selected and integrated into the R. glutinis genome 18,43 . The cellulase genes included two cellobiohydrolases (CBHI and CBHII, from Trichoderma reesei), three endo-β-1,4-glucanases (EgIII, from T. reesei; EgI and EglA, from Aspergillus niger) and β-glucosidases (BGS, from Neocallimastix patriciarum). Each gene was fused with a secretion signal (α-factor) at the N-terminal for efficient secretion out from the cell. Each gene was flanked with independent inducible promoters and terminators ( Table 1). The amplification of six cellulase gene cassettes used the same method as in the carotenogenic gene cassettes. Each cassette was in the pUC18 vector as in previous studies 18,43 . The gene cassettes were integrated using three selection marker genes: G418 resistance gene (neomycin phosphotransferase gene, KanMx), Zeocin resistance gene (Sh ble) and Hygromycin phosphotransferase gene (hph). The MIC of three different selection markers (i.e., Zeocin, G418 and Hygromycin) were tested on R. glutinis wild type using YP2D plates supplemented with different concentrations of antibiotics.

Electroporation. Electroporation
Transformant screening. The transformants were sub-cultured for 3 generations to select stable transformants. Then, transformants were mixed with QuickExtract ((QE), DNA Extraction Solution 1.0, Epicentre, USA) for rapid extraction of genomic DNA and used as a template for PCR verification. Each transformed functional genes were confirmed by PCR using the gene specific primer pairs (Supplementary Table S2). Finally, the transformants with integrated gene cassettes were screened for the β-carotene production and cellulase activity. β-carotene extraction and Analytical methods. To analyze the β-carotene production in transformants, single colonies were inoculated into 5 ml medium 48 and incubated at 30 °C, with 300 rpm for 7 days. Cells were harvested by centrifugation (6000 rpm, 10 min), lyophilized and suspended in 1 ml acetone. Cells were then subjected to mechanical disruption using MagNA Lyser (MagNA Lyser Instrument, Roche) at 6,000 rpm, 20 s for 3 times. For screening the transformants, 1 ml acetone was used in crude extraction. To estimate the maximum β-carotene amount of a transformant, the extraction step was repeated until colorless extracts appeared. The carotenoid extracts were analyzed using High-Performance Liquid Chromatography (HPLC) with a Nomura Chemical Develosil C30-UG Column (3 mm, ID 4.6 mm × L 250 mm-UG17346250W, Interlink Scientific Services, UK), using two buffer systems including buffer A: methanol/MTBE/water (81:15:4) and buffer B: methanol/MTBE/water (7:90:3) as mobile phases. The HPLC condition was described in previous studies 52,53 . The commercial free-form β-carotene and lycopene were used as the standards (Sigma-Aldrich Co. LLC, USA).

Cellulase activity assay.
To screen the transformants with higher cellulase activity, a cellulase activity assay was conducted as described in previous studies 18,54 . Single colonies of transformants were incubated in 5 ml YP2D2G for 3 days and cells were harvested and washed twice with PBS. The washed cells were inoculated into 5 ml YP supplemented with 4% Sigmacell cellulose type 20 as the sole carbon source under 30 °C with 300 rpm for 7 days. The initial and final reducing sugar concentrations were analyzed using the Dinitrosalicylic (DNS) colorimetric method 55,56 . Similarly, total cellulase activity was demonstrated using different cellulosic substrates such as Sigmacell cellulose type 20, Sigmacell cellulose type 50 (SigmaAldrich, St.Louis, MO, USA) and cellulose microcrystalline (Merck, Darmstadt, Germany) 54 . For the cellulose degradation assay, the transformants and wild type R. glutinis were cultured in 50 ml YP2D2G for 3 days and supernatants were concentrated 50-fold using Vivaspin 20 (10,000 molecular weight cut off, PES membrane, GE Healthcare) at 4 °C. The commercial cellulases were used as a positive control, including cellulase powder from Aspergillus niger (≥0.3 units/mg solid, Sigma Aldrich C1184, USA) and cellulase liquid from Trichoderma reesei (≥700 units/g, Celluclast 1.5 L, Sigma Aldrich C2730, USA). The total protein concentration was determined by Bio-Rad Protein Assay Kit using the bovine serum albumin (BSA) as a standard. The assay reaction contained 32 μg protein, sodium acetate (50 mM, pH = 5) and cellulose substrates (2%, final concentration). The reaction mixture was incubated at 50 °C overnight. The initial and final reducing sugar amounts were analyzed by the DNS method.
Lipid content and salt tolerance analysis. The total lipid content of R. glutinis was determined using the method for Rhodosporidium toruloides with some modifications 24,57 . Briefly, R. glutinis wild type and transformants were cultured in 5 ml lipid-inducing medium for 3 days and cells were harvested, frozen and then freeze-dried. Then, the dry-cells were suspended in 1 ml chloroform/methanol (2:1 volumetric) and homogenized using a MagNA Lyser. The homogenized samples were then mixed with 0.2 ml ddH 2 O and vortexed for 15 s. The organic layer was taken using a needle and washed with 0.1 ml NaCl (0.1%, w/v) solution. The extract was repeated until the clear organic layer appeared and was dried in a hood at room temperature overnight followed by 1 h at 80 °C in a pre-weighed tube. The total lipid weight was determined. To determine the salt-tolerance of transformants, 5% NaCl was added to the YP2D media and the growth OD was measured every 24 h in 4 days.