Developing a set of strong intronic promoters for robust metabolic engineering in oleaginous Rhodotorula (Rhodosporidium) yeast species

Background Red yeast species in the Rhodotorula/Rhodosporidium genus are outstanding producers of triacylglyceride and cell biomass. Metabolic engineering is expected to further enhance the productivity and versatility of these hosts for the production of biobased chemicals and fuels. Promoters with strong activity during oil-accumulation stage are critical tools for metabolic engineering of these oleaginous yeasts. Results The upstream DNA sequences of 6 genes involved in lipid biosynthesis or accumulation in Rhodotorula toruloides were studied by luciferase reporter assay. The promoter of perilipin/lipid droplet protein 1 gene (LDP1) displayed much stronger activity (4–11 folds) than that of glyceraldehyde-3-phosphate dehydrogenase gene (GPD1), one of the strongest promoters known in yeasts. Depending on the stage of cultivation, promoter of acetyl-CoA carboxylase gene (ACC1) and fatty acid synthase β subunit gene (FAS1) exhibited intermediate strength, displaying 50–160 and 20–90% levels of GPD1 promoter, respectively. Interestingly, introns significantly modulated promoter strength at high frequency. The incorporation of intron 1 and 2 of LDP1 (LDP1in promoter) enhanced its promoter activity by 1.6–3.0 folds. Similarly, the strength of ACC1 promoter was enhanced by 1.5–3.2 folds if containing intron 1. The intron 1 sequences of ACL1 and FAS1 also played significant regulatory roles. When driven by the intronic promoters of ACC1 and LDP1 (ACC1in and LDP1in promoter, respectively), the reporter gene expression were up-regulated by nitrogen starvation, independent of de novo oil biosynthesis and accumulation. As a proof of principle, overexpression of the endogenous acyl-CoA-dependent diacylglycerol acyltransferase 1 gene (DGA1) by LDP1in promoter was significantly more efficient than GPD1 promoter in enhancing lipid accumulation. Conclusion Intronic sequences play an important role in regulating gene expression in R. toruloides. Three intronic promoters, LDP1in, ACC1in and FAS1in, are excellent promoters for metabolic engineering in the oleaginous and carotenogenic yeast, R. toruloides. Electronic supplementary material The online version of this article (doi:10.1186/s12934-016-0600-x) contains supplementary material, which is available to authorized users.


Background
Red yeast species in the Rhodosporidium (teleomorph) genus, which was recently revised as genus Rhodotorula (anamorphic) regarding to the implementation of "One Fungus = One Name" nomenclatural principle [1], are outstanding producers of lipids and carotenoids [2,3]. More than 100 g/L of dry biomass with over 60% neutral lipids (triacylglycerol, TAG) content can be produced within a week when glucose was used as the carbon source [4][5][6]. To take advantage of their high metabolic flux and cell mass productivity, a number of laboratories have been engaged in establishing them as new platforms for synthetic biology and metabolic engineering. To date, several genetic manipulation tools have been reported, such as high efficiency transformation via Agrobacterium tumefaceins-mediated transformation (ATMT), high efficiency gene deletion, and constitutive and inducible promoter toolbox [7][8][9][10][11][12]. As oil producers, strong and robust promoters that function during lipid accumulation stage will be particularly useful.
The production of long chain acyl-CoA, triacylglycerol (TAG) and lipid bodies are the 3 major lipogenesis steps, and genes involved in these processes are the likely source of strong promoters in oleaginous yeasts. Acetyl-CoA carboxylase (Acc1) catalyzes the biotindependent carboxylation of acetyl-CoA to form malony-CoA, the first committed and rate-limiting step in fatty acid biosynthesis [13]. Proteomic studies showed that it is a highly abundant protein during lipid accumulation phase [5]. Likewise, the α and β subunits of fatty acid synthase (Fas2 and Fas1, respectively), fatty acid transporter (Fat1), ATP:citrate lyase (Acl1) and urea carboxylase/allophanate hydrolase (Dur1/2) were also found in high abundance during lipid accumulation phase [14,15]. Other known abundant targets are the perilipin, adipophilin and tail-interacting (PAT) family proteins, which serve as dynamic scaffolds regulating the formation, growth and degradation of lipid bodies [15][16][17][18][19].
Eukaryotic genes are often interrupted by spliceosomal introns, which vary greatly among different species in either density (number of introns per gene) or length [20]. Although introns are often perceived as junk DNA gained during genome evolution [21], some introns are known to regulate gene transcription and this effect is known as intron-mediated enhancement (IME) [22]. The concept of IME was initially reported in plant [22] and subsequently observed in other organisms such as flies [23] and fungi [24,25]. To our knowledge, two cases of IME have been reported in oleaginous yeasts where the 5′ introns significantly enhanced gene expression: one in fructose 1,6-bisphosphate aldolase gene (FBA1) of Yarrowia lipolytica and the other in d-amino acid oxidase gene (DAO1) of R. toruloides [12,26]. Despite Acc1 being the most abundant protein in R. toruloides [5], the 1.5-kb upstream DNA sequence of ACC1 (−1501 to −1 from the translational start site) showed little promoter activity (our unpublished data). In addition, the high intron density (an average of 6 introns per gene) [14] and strong enhancing effect of the DAO1 introns [12] suggested the global regulatory roles of introns in R. toruloides.
Here, we report the cloning and molecular characterization of 6 promoters from R. toruloides and demonstrate their applications in metabolic engineering.

Characterization of genes involved in lipid accumulation
Genomic sequences for acetyl-CoA carboxylase gene (ACC1), ATP:citrate lyase gene (ACL1), β subunit of fatty acid synthetase gene (FAS1), fatty acid transporter gene (FAT1) and urea amidolyase gene (DUR1) were identified by BLAST search against the public database as well as in-house EST and genome database of R. toruloides strains [27,28]. The amino acid sequences of known orthologous enzymes from Saccharomyces cerevisiae or Y. lipolytica were used as queries ( Table 1). The perilipin encoding gene of R. toruloides NP11 strain (lipid droplet protein 1 gene, LDP1) [15] was used to search for its counterpart in R. toruloides ATCC 10657. The putative homolog of ACC1, ACL1, FAS1, FAT1, DUR1 and LDP1 was found located in the genome sequencing scaffold No. 18, 9, 18, 9, (Table 1). Notably, the first intron was found to be located within the 5′UTR of both FAS1 and LDP1 (Fig. 1). The detailed structures and sequences of these genes are summarized in Table 1 and Additional file 1, respectively.

Analysis of promoter activity by luciferase reporter assay
Upstream DNA sequences of the above mentioned genes were amplified by PCR in two versions, with or without intronic sequence (Fig. 1), and fused to the codon-optimized luciferase reporter gene RtLUC2 (GenBank accession number KR258785) [12] in the binary vector pKCL2. pKCL2 allows site-specific integration of reporter cassettes at the CAR2 locus (phytoene synthase/carotene cyclase gene), which eliminates position effects caused by ectopic insertion of T-DNA into the chromosomes [8,12]. The names of intronic promoters were affixed with "in" to differentiate the promoters analyzed.
As reported, DUR1 and FAT1 were highly transcribed during lipogenic phase [14]. However, luciferase reporter assay revealed that none of DUR1, DUR1in, FAT1 and FAT1in promoters (Fig. 1a, b) displayed detectable activity throughout the cell culture (Fig. 2a, b). The presence of introns in FAS1 and ACL1 promoters weakened promoter activities at the initial stages of cell culture (day 1 and 2) (Fig. 2c, d). However, these repressive effects disappeared after day 3 when nitrogen levels became limited and lipid accumulation began to accelerate (Fig. 2c,  d). Thus, the FAS1in and ACL1in promoters (Fig. 1c, d) should be useful under circumstances where strong expression have undesirable effects at the early stages of cell culture. Otherwise, the intronless FAS1 and ACL1 promoters would be preferred for gene expression. Unlike the intronless ACC1 promoter (−1501 to −1, Fig. 1e), the ACC1in promoter (−1501 to +93) yielded significantly higher luciferase activity throughout the time course (Fig. 2e).

Transcription of ACC1, FAS1 and LDP1 mRNAs during lipid production
Gene expression of ACC1, FAS1 and LDP1 were investigated by qRT-PCR analysis, which was done by normalizing against the transcripts of actin gene (ACT1) using cycle threshold (Ct) method (2 −ΔCt calculation method). In good agreement with the luciferase reporter assays, mRNA levels of LDP1 were much higher than those of ACC1 and FAS1 throughout the 6-day culture period (Fig. 3a). To show the dynamic changes in gene expression, the relative mRNA levels of ACC1, FAS1 and LDP1 at different time points were compared using 2 −ΔΔCt calculation method, where each mRNA level at day 0 was set as 1. Overall, all 3 genes showed upward trends in transcription over the 6-day culture period, although the mRNA levels of ACC1 and LDP1 peaked earlier at day 2 and 4, respectively (Fig. 3b).

Performance of promoters in a lipid-less genetic background
To evaluate the performance of the above promoters in a non-oil accumulating genetic background, reporter constructs for LDP1 and LDP1in were transformed into a R. toruloides mutant named dlad, in which genes encoding 4 essential acyltransferase in lipid biosynthesis, acyl-CoA-dependent diacylglycerol acyltransferase (DGA1), phospholipid:diacylglycerol acyltransferase (LRO1), acyl-CoA:sterol acyltransferase (ARE1) and soluble diacylglycerol acyltransferase (DGA3), were sequentially disrupted (our unpublished data). This quadruple disruption mutant contains only 8.5% lipids of wild-type strain (WT) (our unpublished data). Similar to the situation in WT, both LDP1 and LDP1in promoters displayed significantly higher activity than the GPD1 promoter in dlad mutant. Moreover, the LDP1in promoter was more active than the LDP1 promoter ( Fig. 4) although the magnitude of enhancement appeared to be reduced (compare Figs. 2f, 4). These suggest that the lipogenic promoters such as LDP1 and LDP1in are largely regulated by nutrient levels independent of de novo TAG biosynthesis.

Metabolic engineering of lipid production using different promoters
Overexpression of Dga1, the key enzyme of TAG biosynthesis, has been demonstrated to significantly improve lipid yields in other oleaginous yeast like Y. lipolytica [13]. As a principle of concept, to demonstrate the application of the strong promoters characterized in this study, we compared the lipid production levels by overexpressing DGA1 using either GPD1 or LDP1in promoter (strain P GPD1 ::DGA1 and P LDP1in ::DGA1, respectively). Results showed that overexpression of DGA1 with either promoters significantly improved lipid content (Fig. 5a). More importantly, lipid content accumulated in the strain P LDP1in ::DGA1 was on average 21% higher than that in the strain P GPD1 ::DGA1 and 55% higher than that of WT strain (Fig. 5a). qRT-PCR analysis was used to identify the gene expression of DGA1 in the above three strains. When compared to WT strain on day 2, the use of GPD1 and LDP1in promoter resulted in 26-and 66-fold increase in the mRNA levels of DGA1, respectively (Fig. 5b). These data suggest that the promoters developed in this study, particularly the LDP1in promoter, are superior tools for metabolic engineering in oleaginous yeasts.

Discussion
Lipid biosynthesis is the dominant bioactivity in oleaginous yeasts. Therefore, promoters of genes involved in various steps of the pathway are particularly useful. Such examples include GPD1, PGK1 and PGI1 promoters as reported previously [7,11]. Here, we analyzed the promoters of 6 genes involved in either fatty acid biosynthesis, TAG biosynthesis or lipid body formation. Through luciferase reporter assay, the application of intronic promoters derived from ACC1, FAS1 and LDP1 genes were firmly established. Promoters are generally known to be located upstream of the transcriptional start site (TSS) and composed of a core promoter domain and several regulatory domains [29]. Although cis-acting elements may be located 50 or more kilobases proximal to the TSS, most eukaryotic promoters used to date are within 1 kb upstream from the TSS, including those reported promoters from R. toruloides [7,12]. However, increasing evidences have demonstrated that cis-acting elements may be located downstream of the TSS, in the 5′UTR, within an intron, or even downstream from the last exon of a gene [30]. The inclusiveness of introns in certain genes may improve their transcriptional and translational outputs [31]. Regulatory elements in the first intron within the 5′UTR and coding region have been found in higher eukaryotes elsewhere [32][33][34][35][36]. Although IME has been known for a long time, mechanistic studies are still lagging. It was recently proposed that introns may create a region of localized accessible chromatin to increase transcriptional initiation [37]. The introns characterized in this report with diverse roles in regulating gene expression should serve as good samples for the mechanistic studies of IME.
In this study, we found that the expression of 5 out of 6 genes were significantly regulated by introns: with strong enhancing effect on ACC1 and LDP1 genes and repressing effect on ACL1, FAS1 and DUR1 genes (Fig. 2). In fact, intron 1 of GPD1 gene (+3 to +131) also contains a repressing element (Additional file 2). These results, together with the previous report of DAO1 gene [12], showed that the involvement of introns in transcriptional control is a universal phenomenon in R. toruloides. Further studies in this area may facilitate the development of a more robust and efficient synthetic platform based on oleaginous red yeasts.

Conclusions
Intronic promoters of ACC1, FAS1 and LDP1 are superior tools for gene expression, metabolic engineering and synthetic biology in R. toruloides. Similar promoters may also be functional in other Pucciniomycotina species.

Strains, media and culture conditions
Rhodotorula toruloides strain ATCC 10657 was purchased from ATCC (Manassas, VA, USA). R. toruloides nonhomologous end joining-deficient mutant strain Δku70e [8] was considered as the wild-type strain. R. toruloides quadruple disruption mutant dlad (Δdga1Δlro1Δare1Δdga3) was generated by serial deletion of the four diacylglycerol acyltransferase genes, DGA1, LRO1, ARE1 and DGA3, in the host Δku70e through homologous recombination, in which the selectable marker cassette was recycled by activating Cre/loxP system [8]. A. tumefaciens strain AGL1 [38] and Escherichia coli XL1-BLUE were used for routine molecular cloning procedures.

Agrobacterium tumefaciens-mediated transformation
The binary vectors were electroporated into A. tumefaciens AGL1 (2.5 kV, 25 μF, 400 Ω) and subsequently selected with 2YT agar medium supplemented with streptomycin (100 µg/ml). Fungi transformation via ATMT was performed as described previously [7].  The dlad strain is mutant with targeted deletion of four diacylglycerol acyltransferase genes (our unpublished data). All reporter mutants were cultured in MinRL3 for 5 days and luciferase reporter assays were performed daily. Results were derived from three biological replicates and error bars indicate standard deviation

Isolation of genomic DNA and total RNA
Genomic DNA and RNA of R. toruloides were extracted as described previously [7]. The concentration and purity of the extracted DNA and RNA were analyzed using NanoDrop ® ND-1000 Spectrophotometer (Nanodrop Technologies, USA) and agarose gel electrophoresis.

Rapid amplification of cDNA ends (RACE)
The 5′ and 3′ end of cDNA sequences were determined by 5′ RACE and 3′ RACE using BD SMARTer ™ RACE cDNA Amplification Kit (Clontech, CA, USA) according to the manufacturer's instruction. Oligos ACC1U1 ( Table 2) was used as the specific primer for 5′ RACE of ACC1, LDP1U1 and LDP1L1 (Table 2) was used for 5′ RACE and 3′ RACE of LDP1, respectively.

Quantitative reverse transcription PCR (qRT-PCR)
Total RNA was treated with DNase I (Roche Diagnostics, Mannheim, Germany) followed by precipitation with ethanol to remove trace amount of DNA. cDNA was synthesized by reverse transcription using iScript ™ Reverse Transcription Supermix (Bio-Rad Laboratories, CA, USA). Real-time PCR was conducted in ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, CA, USA) using the ABI SYBR ® Select Master Mix (Applied Biosystems). PCR conditions were as follows: an initial 50 °C for 2 min and 95 °C denaturation step for 10 min followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 60 °C for 1 min. Triplicates were used for all analysis. The data was acquired using the software SDS 2.4 (Applied Biosystems) and relative gene expression levels were calculated against the reference encoding gene of actin (ACT1, GenBank acc. no. KR138696) [12] using the 2 − ΔΔCt method through the RQ Manager software v1.2.1 (Applied Biosystems).

Reporter assay
Strains bearing the integrated T-DNA at the CAR2 locus were identified by the albino phenotype followed by Southern blot verification. The strains were then cultured in YPD broth to mid-exponential phase and used for luciferase assay. Cells were washed twice with water and inoculated to the indicated medium at an optical density (OD 600 ) of 0.5 and cultured at 30 °C with agitation (250 rpm). Luciferase activity was determined by one-step measurement method as described previously [7] with some modifications. Briefly, cell culture (10 µl) was mixed with 85 µl of PBS buffer (pH7.4) and 5 µl of 10 mM D-luciferin (DMSO solution, catalogue No. L9504, Sigma-Aldrich), loaded in a well of FluoroNunc 96-well plate (Thermo Fisher Scientific, Langenselbold, Germany) for measurement of bioluminescence. Cell density was measured at OD of 600 nm with 10-20 fold dilution of cell culture in PBS buffer to a final volume of 100 μl, and loaded to a well of 96-well flat-bottom transparent plate (Nunc, Roskilde, Denmark). All data was measured and acquired with the Tecan Infinite M200 plate reader coupled with the iCycler version 3.0 software (Tecan, Salzburg, Austria). The relative promoter activity (RPA) was calculated by normalization against that of GPD1 promoter.