Function and Localization of the Arabidopsis thaliana Diacylglycerol Acyltransferase DGAT2 Expressed in Yeast

Diacylglycerol acyltransferases (DGATs) catalyze the final and only committed step of triacylglycerol synthesis. DGAT activity is rate limiting for triacylglycerol accumulation in mammals, plants and microbes. DGATs belong to three different evolutionary classes. In Arabidopsis thaliana, DGAT1, encoded by At2g19450, is the major DGAT enzyme involved in triacylglycerol accumulation in seeds. Until recently, the function of DGAT2 (At3g51520) has remained elusive. Previous attempts to characterize its enzymatic function by heterologous expression in yeast were unsuccessful. In the present report we demonstrate that expression of a codon-optimized version of the DGAT2 gene is able to restore neutral lipid accumulation in the Saccharomyces cerevisiae mutant strain (H1246), which is defective in triacylglycerol biosynthesis. Heterologous expression of codon-optimized DGAT2 and DGAT1 induced the biogenesis of subcellular lipid droplets containing triacylglycerols and squalene. Both DGAT proteins were found to be associated with these lipid droplets. The fatty acid composition was affected by the nature of the acyltransferase expressed. DGAT2 preferentially incorporated C16:1 fatty acids whereas DGAT1 displayed preference for C16:0, strongly suggesting that these enzymes have contrasting substrate specificities.


Introduction
Triacylglycerols (TAGs) are the main form of fatty acid (FA) storage in most eukaryotes [1]. They accumulate in dedicated organelles called lipid droplets (LDs). LDs are composed of a neutral lipid core surrounded by a monolayer of phospholipids (PLs), in which various types of specialized proteins are embedded. LDs serve as an energy reservoir, used for example during seed germination in plants [2], and are the source of signaling molecules and substrates for membrane biogenesis [1].
Three enzymes catalyze TAG synthesis by the acylation of sn-1,2-diacylglycerol (DAG) [3]. Two acyltransferases catalyze acyl-CoA independent reactions. Phospholipid: diacylglycerol acyltransferase (PDAT, EC 2.3.1.158) uses PLs as acyl donors. Its activity has been reported in yeast and plants and corresponding genes have been identified [4,5]. DAG transacylase uses two DAGs as substrates to produce one TAG and one monoacylglycerol (MAG). This activity has been reported in various organisms [6,7] but to date, genes encoding this enzyme remain unknown. In contrast to PDAT and DAG transacylase, acyl-CoA: diacylglycerol acyltransferases (DGAT, EC 2.3.1.20) are acyl-CoA dependent enzymes, which catalyze the final and only committed step of the Kennedy pathway [8,9,10]. In seeds, the absence of DGAT activity leads to a significant reduction in TAG accumulation [11,12].
DGAT1 and 2 are found in mammals. These endoplasmic reticulum (ER) integral proteins [18,19] are both responsible for TAG production [20,21] and presumably have different physiological functions [22,23,24]. The most significant contribution to TAG synthesis in the yeast Saccharomyces cerevisiae is from Dga1p, a member of the DGAT2 family [25,26]. In the oleaginous yeast Yarrowia lipolytica Dga2p, a member of the DGAT1 family makes a major contribution to TAG synthesis [9].
In A. thaliana DGAT1 encoded by At2g19450 plays a major role in seed lipid accumulation. Its activity and function have been extensively characterized in planta [8,27]. DGAT1 inactivation results in a 20 to 40% decrease in seed oil content [12,28,29]. The observation that oil deposition was not completely abolished in this mutant background strongly suggested that other enzymes also contribute to TAG accumulation in seeds. Accordingly, a concomitant down regulation of PDAT1 and DGAT1 further decreased the seed oil content (by 70 to 80%), thus establishing that PDAT1 and DGAT1 play partially redundant functions in seeds [28].
DGAT2 coding sequences are found in the genome of several plants. This enzyme can incorporate unusual FAs into TAGs [30], such as eleostearic acid, a polyunsaturated FA found in Tung tree (Vernicia. fordii) [31], or ricinoleic acid, a hydroxy FA from Ricinus communis [32]. The A. thaliana genome encodes a putative DGAT2 (At3g51520), the functionality of which was recently demonstrated by transient expression in Nicotiana benthamiana leaves [33]. However, seeds from A. thaliana dgat2 mutants do not show altered TAG accumulation and introduction of the dgat2 mutation in a dgat1 mutant background did not aggravate the dgat1 seed phenotype, suggesting that DGAT2 is not involved in seed oil accumulation [28].
The current model of LD biogenesis proposes that LDs occur in specialized subdomains of the ER, gathering enzymes of the neutral lipid biosynthetic pathway [1]. Accumulation of lipids between the two membrane leaflets would lead to budding of LDs covered with a PL monolayer [1,34]. The presence of proteins normally associated with the ER membrane at the surface of LDs is in favor of the budding model. This has been verified for A. thaliana DGAT1 [27] and DGAT2 [35].
Here we report the functional characterization of A. thaliana DGAT2 in the yeast mutant strain H1246 which is unable to accumulate neutral lipids [26]. We demonstrate that yeast codon usage limits protein expression in this system and that expression of codon-optimized DGAT2 can restore TAG synthesis by functional complementation. A. thaliana DGAT1 was used as a positive control. We then show that the two DGAT proteins are associated with LDs. Both enzymes exhibit contrasted substrate specificity and induce squalene accumulation in LDs.

Cloning of DGAT sequences for expression in yeast
The A. thaliana DGAT1 (At2g19450) and DGAT2 (At3g51520) cDNAs (Fig. S1) were amplified with proofreading Pfu Ultra DNA polymerase (STRATAGENE, La Jolla, CA, USA) from a mixture of seed cDNAs (of the Ws ecotype) using the primers: 59-attB1-ATGGCGATTTTGGATTC-39 and 59-attB2-TCATGACATC-GATCCTTTTC-39 (DGAT1); 59-attB1-ATGGGTGGTTCCA-GAG-39 and 59-attB2-TCAAAGAATTTTCAGCTCAAG-39 (DGAT2). AttB1 and attB2 refer to the corresponding Gateway recombination sequences. BP recombination was used to introduce the PCR products into the pDONR207 entry vector (INVITROGEN, Carlsbad, CA, USA) for sequencing. Codon sequences were optimized for expression in S. cerevisiae (Fig. S1) by Eurofin MWG Operon (Ebersberg, Germany). These sequences were amplified using Phusion Hot Start polymerase (Thermo Scientific, Illkirch, France) and introduced into the BamHI/SmaI restriction sites of the pRT21 vector (also known as pNBT29) [36] to generate proteins fused with a C-terminal GFP. Corresponding constructs were made with a stop codon introduced for expression of DGAT alone without the GFP fusion. Details of the oligonucleotides used are shown in Table 1. All constructs were sequenced to confirm the absence of mutations in amplified sequences (Genoscreen, Lille, France).

Expression of DGATs in yeast
Details of the recombinant plasmids as well as an empty plasmid control used to transform the S. cerevisiae quadruple mutant (are1, are2, dga1, lro1) strain H1246 or the SCY62 wild-type (WT) strain [26] are given in Table 2. Cells were grown in Ura deficient medium as described in [37] and protein expression was induced with 2% galactose in the presence of 0.02% glucose.

Lipid droplet purification
Yeast LDs were separated by sucrose density gradients as previously described by Yu et al. [38]. Cells corresponding to 400 UA 600nm were harvested by centrifugation, washed with water and resuspended in Fat Body Buffer (FBB, 10 mM HEPES, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, pH 7.5) supplemented with protease inhibitors (cOmplete, Mini, EDTA-free, Roche, Indianapolis, USA). Cells were disrupted with a One Shot cell disruptor (Constant Systems Ltd, Daventry, UK) at a pressure of 2.97 kbar. Lysates were centrifuged for 10 min at 12 000 g. The supernatant volume was adjusted to 5.4 ml with FBB containing sucrose at a final concentration of 0.54 M. It was overlaid sequentially with FBB containing 0.27 M, 0.135 M and 0 M sucrose (361.8 mL). The gradients were subjected to ultracentrifugation for 90 min at 150 000 g and 4uC in a SW41 Ti swing-out rotor (Beckman Coulter, Villepinte, France). The floating lipid layer, or the corresponding volume for the strain transformed with the empty plasmid, was collected and stored at 280uC. The hydrodynamic diameter of purified LDs was assessed by dynamic light scattering on a Malvern HPPS (Malvern Instruments Ltd, Worcestershire, UK) according to Vindigni et al. [39].

Lipid extraction
Lipids from yeast cells or LDs were extracted as described by Folch et al. [40]. Cells were harvested by centrifugation and washed twice with a solution containing 0.5% BSA and 0.9% NaCl. After freeze-drying, 100 mg of dry cells or LDs purified from 1000 UA 600nm were resuspended in 5 ml of chloroform/ methanol 2/1 (v/v). After 1h incubation with shaking, the extract was centrifuged for 5 min at 900 g and the supernatant was recovered and mixed with 2.5 ml of 0.9% NaCl. The organic phase was collected after centrifugation at 900 g for 5 min and washed three times with chloroform/methanol/water 3/48/47 (v/ v/v) [41]. Solvents were evaporated under a stream of nitrogen.

Fatty acid determination
Total lipids. The FA content and composition of yeast strains were determined according to Browse et al. [45]. After freeze-drying, 20 mg of dry yeast were incubated in 2 ml of MeOH/H 2 SO 4 (100/2.5, v/v) at 80uC for 90 min after addition of heptadecanoic acid as an internal standard. After addition of 900 ml of hexane and 3 ml of 0.9% NaCl, FA Methyl Esters (FAMEs) were extracted into the organic phase and analyzed by gas chromatography with flame ionization detection (GC-FID) according to Froissard et al. [37].
Neutral lipids. Neutral lipids were obtained upon fractionation of total lipids using an Isolute SPE Aminopropyl column (ALLTECH France Sarl, Epernon, France) according to [46]. After transmethylation, FAMEs were identified and quantified by GC-FID.

Compositional analysis and quantification of the neutral lipid fraction
Neutral lipids were derivatized to TriMethylSylil-lipids (TMSlipids) prior to their analyses by GC-FID using 100 mL of reagent solution (n-methyl imidazole/N-methyl-N-trimethylsilyl-heptafluorobutyramide, 1/20, v/v). Analysis of TMS-lipid derivatives were performed on a Trace GC Ultra gas Chromatograph (Thermo Scientific, Illkirch, France) equipped with a fused-silica ZB5 HT capillary column (15 m60.25 mm I.D., 0.1mm film thickness; Phenomenex, Le Pecq, France). Samples were injected into an on-column detector and the FID system was set at 400uC. The oven temperature program was 100uC, increased to 370uC at 10uC/min and isothermal for 5 min at this final temperature (total run time 32 min). The carrier gas (H 2 ) flow was maintained constant at 1.5 mL/min. A mixture containing the following compounds: FAs (C16, C18 and C18:1), MAG (C16, C18 and C18:1), DAG (equivalent to 34 and 36 atoms of carbon), TAG (equivalent to 48, 50, 52 and 54 atoms of carbon), squalene, cholesterol and phytosterols (campesterol, stigmasterol and sitosterol) was used as the standard for further product identification [47].

Squalene identification
After extraction, 100 mg of lipids were resuspended in 300 ml hexane and dried over Na 2 SO 4 . Silylation was performed on 10 ml of the sample with 50 ml of BSTFA and 5 ml of pyridine (Sigma-Aldrich, L'Isle d'Abeau, France). Lipids were separated on a DB1 supelco capillary column of 15 m60.32 mm (carrier gas: helium at a constant flow rate of 1.5 ml.min 21 ) with a first increase from 45uC to 180uC (at 30uC.min 21 ) and a second rise from 180uC to 280uC (at 3uC.min 21 ). Detection was performed by mass spectrometry on a Varian 4000 ion trap operating in the electron impact mode (70 eV) with ions detected on a range of 50 to 800 m/ z. The presence of squalene in the sample was assessed by comparing the mass spectra of the component having an identical retention time to squalene with the mass spectra of squalene ($ 98% squalene, Sigma-Aldrich)

Confocal Image acquisition
Yeast were incubated for 20 min in a Nile Red solution (1 g/l from a 1 g/ml stock solution in acetone), and washed with PBS solution containing 10% v/v glycerol. Images were acquired using an inverted LEICA SP2-AOBS spectral confocal laser microscope (LEICA Microsystems, Mannheim, Germany) using an HCX PL APO 63 X water (long distance) 1.2 objective. GFP and Nile Red fluorescence were observed with a 488 nm light wavelength generated by an argon laser and an emission band of 5002520 nm and 6002650 nm respectively. Transmitted light was captured with a dedicated photomultiplier (PMT trans). Each image is the average of eight scans at a resolution of 5126512 pixels with 8x numerical zoom.

Results
Codon usage affects expression of DGAT2 and DGAT1 in S. cerevisiae The DGAT2 (At3g51520) sequence is highly similar to several active DGAT2 genes from plants [31,32]. However, inactivation of the DGAT2 gene did not significantly reduce seed oil content [28] and expression of the gene in S. cerevisiae did not lead to lipid accumulation [28,48,49,50]. Taken together, these findings firstly suggested that DGAT2 does not encode a functional DGAT protein [28]. Nevertheless, a very recent report has shown that DGAT2 can be actively expressed in N. benthamiana cells [33] and exhibits a higher specificity for C18:3 FA.
The function of A. thaliana DGAT2 was examined by heterologous expression in the H1246 S. cerevisiae strain. This strain lacks four acyltransferases and is completely defective in neutral lipid biosynthesis [26]. Codon usage can affect expression of heterologous proteins in S. cerevisiae [3,51], thus leading to the false conclusion that the expressed proteins are not active. Liu et al. [3] suggested that the lack of DGAT2 activity observed in S. cerevisiae could be due to differences in codon usage between this yeast and A. thaliana. To test this hypothesis and allow detection of the expressed proteins, we used WT or optimized versions of sequences encoding A. thaliana DGAT1 and 2 and harboring GFP tags for expression in S. cerevisiae. Accumulation of the two optimized DGATs in total yeast extracts was confirmed by Western blot analysis with anti-GFP antibodies (Fig. 1). When codon bias was not considered only slight expression of DGAT1 was observed, confirming previous results [27] and demonstrating the importance of codon optimization for plant DGAT expression in yeast. High molecular mass proteins were identified by immunoblotting which may reflect DGAT oligomerization as previously suggested for DGATs from other organisms [15,18,19,52].

DGAT2 restores neutral lipid accumulation in H1246
To study the effect of DGAT2 expression on the H1246 lipid profile, the quadruple mutant was transformed with an optimized version of the protein with no GFP tag (DGAT2 opt ). As a positive control, we used an H1246 strain transformed with an optimized version of DGAT1 (DGAT1 opt ). Total FAs from the different strains were quantified following transmethylation and GC-FID analysis. Expression of both DGAT1 opt and DGAT2 opt significantly increased total FA content in H1246 (Fig. 2.A). Cells expressing DGAT2 opt and DGAT1 opt contained 1.3 and 2.9 times more FAs, respectively, than H1246 transformed with an empty vector (negative control). Total lipids from the recombinant yeast were extracted and separated on TLC plates (Fig. 2.B). Extracts purified from both DGAT expressing strains contained TAGs. Densitometry showed that the DGAT2 opt expressing strain contained 3.3 times less TAGs than the DGAT1 opt expressing strain (Fig. 2.C). In addition, a lipid which accumulated in both these DGAT expressing lines had a retention factor identical to that of squalene (Fig. 2.B). Subsequent GC-MS analysis confirmed the identity of the product (Fig. S2). The highest squalene accumulation, determined by densitometric comparison of various amounts of squalene, reached approximately 2.7 g/kg in the DGAT1 opt expressing strain.
FA composition analyses of the neutral lipid fraction (TAGs) by GC-FID revealed differences depending on expression of DGA-T1 opt or DGAT2 opt . The DGAT2 opt expressing strain (Fig. 2.D) accumulated mostly C16:1 whereas the DGAT1 opt expressing strain preferentially accumulated C16:0. However, no significant differences were found for the TAG fraction (Fig. S3). TAG profiles of both DGAT1 opt and DGAT2 opt expressing cells showed a normal distribution centered on C50, close to 35% with also a high proportion of C48 (18%) and C52 (23%). A small proportion of C54 was also noted.

Expression of DGAT2 and DGAT1 in S. cerevisiae H1246 restores LD biogenesis
No LDs form in the H1246 mutant strain [26]. Neutral lipid production was re-established in yeast expressing optimized sequences encoding DGAT2 or DGAT1 as well as the nonoptimized sequence coding for DGAT1 (data not shown). We attempted to purify LDs from H1246 transformed with the empty pRT21 vector or pRT21-DGAT1 opt or pRT21-DGAT2 opt . No floating layer was recovered from the H1246 strain transformed with the empty vector. Using dynamic light scattering we determined that the floating layers recovered from DGAT1 opt expressing cells contained objects with diameters similar to LDs from a WT strain (SCY62, 330nm). In cells expressing DGAT2 opt we observed smaller objects (220 nm).
Lipids contained in the floating layers were extracted and separated by TLC (Fig. 2.E). Lipids from DGAT1 opt and DGAT2 opt expressing cells were mainly TAGs, squalene and sterols, but lacked sterol esters in contrast to lipids extracted from the WT strain. No lipids could be extracted from the top fraction of the gradients for the H1246 strain transformed with the empty vector.  Values are expressed in % of dry weight. Asterisks indicate statistically significant differences according to t-test (**P,0.01; ***P,0.001). (B) Separation of yeast lipids by TLC. Lipids were extracted from 100 mg of dry yeast. Lipids from 2 mg dry yeast were separated on a silica plate. This experiment is representative of three independent cultures. Vertical arrows indicate the direction of migration. (C) TAG and squalene contents. Relative amounts of TAGs and squalene to PLs were evaluated by densitometry after separation of yeast lipids by TLC (Fig. 2B). Values were calculated from three independent cultures. Asterisks indicate statistically significant differences according to a t-test (**P,0.01; ***P,0.001). DGAT2 and DGAT1 colocalize with intracellular neutral lipids Confocal fluorescence microscopy of DGAT1 opt -GFP and DGAT2 opt -GFP revealed the presence of the proteins in typical lipid particle structures after 8 h induction (Fig. 3.A). The same staining pattern was observed using the lipophilic fluorescent dye Nile Red which selectively accumulated in LDs (Fig. 3.A), thus confirming the localization of both DGAT proteins in this compartment. Similar results were obtained after 18 h induction, but neither GFP nor Nile Red were found in punctuate structures (Fig. 3.B).

Discussion
In A. thaliana, DGAT2 inactivation did not modify seed oil content. Furthermore, mutation of DGAT2 in the DGAT1 mutant background did not lead to any additional phenotypes, leading Zhang et al. [28] to conclude that DGAT2 does not play a substantial role in seed lipid accumulation. The possibility that DGAT2 could be expressed and active in vegetative tissues was very recently demonstrated by Zhou [33]. Despite the fact that A. thaliana DGAT2 shares 71% and 58% identity with active Tung tree [31] and R. communis [32] DGAT2, previous heterologous expression experiments did not lead to TAG accumulation in the yeast mutant strain H1246 [28,48]. Liu et al. [3] suggested that the A. thaliana DGAT2 gene may exhibit a strong codon bias affecting its expression in yeast. In this report, we demonstrated that using yeast codon usage is mandatory for DGAT2 expression (Fig. 1) and to restore TAG accumulation (Fig. 2.B) in H1246. Optimization of A. thaliana DGAT1 codon usage also significantly increased the rate of protein synthesis and enhanced TAG accumulation. Heterologous DGAT expression is thus controlled by subtle usage of tRNA via codon bias.
H1246 expressing DGAT1 accumulated FA levels of up to 7% of dry weight (Fig. 2.A). We also found a significant amount of squalene in this strain (Fig. 2.B). Squalene is a triterpene precursor in sterol biosynthesis with potential uses in nutrition, cosmetics, and medicine [53]. Up to 2.7 g of squalene per kg of dry biomass were produced in this strain. This value, obtained in nonoptimized conditions, is comparable to the reported yields of promising bioindustrial strains [53]. Spanova et al. [54] showed that the LD is a neutral lipid depository able to accommodate squalene. The WT yeast strain (SCY62) naturally accumulates squalene in LDs in addition to TAGs and sterols esters (Fig. 2.E). We found small amounts of squalene in the H1246 strain ( Fig.  2.B) even though it lacks LDs [26] (Fig. 2.E). These results are consistent with those of Spanova et al. [54] who demonstrated that in the absence of LDs, in a dga1Dlro1Dare1Dare2D mutant strain, squalene can be stored in microsomal and mitochondrial membranes. In contrast to DGAT1, however, LD biosynthesis was probably not high enough in DGAT2-expressing cells for significant amounts of squalene to be stored. The amount of sterols found in all cells and LDs (Fig. 2.B and 2.E) did not vary, strongly suggesting that sterol biosynthesis is uncoupled from the level of its precursor squalene.
C16 and C18 FAs were the major FA species acylated in TAGs from both DGAT expressing strains (Fig. 2D), DGAT2 preferred C16:1 whereas DGAT1 exhibited specificity toward C16:0. The method used for TAG profiling does not differentiate TAG containing saturated from unsaturated FAs [47], thus the results could not highlight differences between neutral lipids from strains expressing DGATs (Fig S3). Depending on the plant species, DGAT1 or DGAT2 is a major enzyme responsible for the accumulation of seed TAG. A. thaliana DGAT1 is involved in acylation of the sn-3 position of the glycerol. Stereochemical analysis showed that the sn-1 and sn-3 positions are usually saturated, while unsaturated species occur mainly at sn-2 [55]. In castor bean and Tung tree, DGAT2 is specific for unsaturated FAs [31,32]. The specialization of DGAT2 (C16:1) observed in the present study is consistent with these last findings. Similarly, transient heterologous expression of DGAT2 in tobacco leaves leads to the accumulation of TAG containing unsaturated FA [33] revealing possible expression under the subtle control of tissue specific factors. Thus, very long chain FA or polyunsaturated FA are potential acyl donors found in A. thaliana seeds and N. benthamiana leaves, but absent in S. cerevisiae. It is therefore difficult to compare enzyme specificity and velocity as reflected by lipid accumulation for DGATs expressed in heterologous systems. Overall, the results strongly suggest that these enzymes show specificity toward different acyl donors.
Our confocal microscopy investigations confirmed the colocalization of DGAT-GFP and neutral lipids. The punctuate structures observed after a short induction (8 h, Fig. 3.A) were typical of LDs, however Nile Red staining was more diffuse after longer induction times (18 h, Fig. 3.B). The lipids which accumulated in DGAT expressing yeast were different from the WT strain and lacked sterol esters. As previously suggested by Czabany [56], lipid composition affects LD protein content and as a consequence, the interactions between LDs and their environment could be modified. Nevertheless the objects purified from the yeast which had been induced for 18 h still had a typical LD hydrodynamic diameter. Thus it could be that close association of numerous LDs rendered individual droplets undistinguishable in vivo using light microscopy, explaining the shape of structures observed after 18 h induction (Fig. 3.B) The association of DGAT1 with LDs and the ER was previously demonstrated by Bouvier-Navé et al. [27]. DGAT2 localization to microsomes was deduced from in vitro assays by Zhou et al. [33], however its presence in other compartments cannot be excluded. Recent results by Kwiatkowska et al. tend to suggest multiple localizations of DGAT2 in germinated A. thaliana seeds [35]. These last experiments were based on the use of antibodies directed against human DGAT2, and the crossreactivity between plant proteins and human antibodies needs to be confirmed. Taken together, these reports and our present findings appear to support the general model proposing that TAG synthesis occurs at the ER with LDs budding from the ER [34].

Conclusion
We have described the enzymatic function and localization of A. thaliana DGAT2 and DGAT1 in the yeast quadruple mutant H1246. Functional complementation confirmed that DGAT2 has diacylglycerol acyltransferase activity. Differences observed at the level of FA composition suggest different substrate specificities for DGATs. Our study, together with previous reports from other teams, highlights the importance of tissue-specific factors and tRNA pools in regulating DGAT expression. Optimization of codon usage appears to be a valuable tool for increasing DGAT expression and subsequent lipid accumulation in yeast. Accumulation of squalene, a terpene with potential biotechnological applications, increased significantly by expression of Arabidopsis DGAT1. Overall, our results suggest that the nature of the DGAT expressed in yeast, is not only important for TAG accumulation, but also for the production of other lipids of interest.