Multiple mechanisms contribute to increased neutral lipid accumulation in yeast producing recombinant variants of plant diacylglycerol acyltransferase 1

The apparent bottleneck in the accumulation of oil during seed development in some oleaginous plant species is the formation of triacylglycerol (TAG) by the acyl-CoA–dependent acylation of sn-1,2-diacylglycerol catalyzed by diacylglycerol acyltransferase (DGAT, EC 2.3.1.20). Improving DGAT activity using protein engineering could lead to improvements in seed oil yield (e.g. in canola-type Brassica napus). Directed evolution of B. napus DGAT1 (BnaDGAT1) previously revealed that one of the regions where amino acid residue substitutions lead to higher performance in BnaDGAT1 is in the ninth predicted transmembrane domain (PTMD9). In this study, several BnaDGAT1 variants with amino acid residue substitutions in PTMD9 were characterized. Among these enzyme variants, the extent of yeast TAG production was affected by different mechanisms, including increased enzyme activity, increased polypeptide accumulation, and possibly reduced substrate inhibition. The kinetic properties of the BnaDGAT1 variants were affected by the amino acid residue substitutions, and a new kinetic model based on substrate inhibition and sigmoidicity was generated. Based on sequence alignment and further biochemical analysis, the amino acid residue substitutions that conferred increased TAG accumulation were shown to be present in the DGAT1-PTMD9 region of other higher plant species. When amino acid residue substitutions that increased BnaDGAT1 enzyme activity were introduced into recombinant Camelina sativa DGAT1, they also improved enzyme performance. Thus, the knowledge generated from directed evolution of DGAT1 in one plant species can be transferred to other plant species and has potentially broad applications in genetic engineering of oleaginous crops and microorganisms.

increased TAG accumulation in Saccharomyces cerevisiae strain H1246 (MAT␣ are1-⌬::HIS3, are2-⌬::LEU2, dga1-⌬::KanMX4, lro1-⌬::TRP1 ADE2, designated 4 delta) (4). Further analysis revealed that the ninth predicted TMD (PTMD9) contains key amino acid residue substitutions leading to higher performance in BnaDGAT1. Two variants (L441P and I447F) with amino acid residue substitutions in that region have been found to increase oil content when the encoding cDNAs were expressed in yeast and Nicotiana benthamiana leaves (17). Despite this, little is known about how these amino acid residue substitutions affect DGAT1 performance. A number of questions can be posed in this regard. Is the observed increase in oil content attributable to the increased enzyme activity, increased polypeptide accumulation, or both? Are the specific amino acid residue substitutions associated with the changes in the kinetic properties of the enzyme? Because the BnaDGAT1-PTMD9 is moderately conserved among various DGAT1, does this domain contain key amino acid residues influencing DGAT1 activity in a range of oleaginous plant species? Finally, could the beneficial amino acid residue substitutions identified in BnaDGAT1-PTMD9 be used to guide the modification of another plant DGAT1?
To answer these questions, three BnaDGAT1 variants (L441P, I447F, and F449C) with amino acid residue substitutions in PTMD9 and one out-group variant, V125F, were chosen for characterization. These recombinant variants were characterized as having higher enzyme activity and/or exhibiting increased polypeptide accumulation in the yeast system. Possible reduced substrate inhibition also appeared to contribute to increased oil content in yeast. Sequence alignment of DGAT1 revealed that the PTMD9 is conserved in many plant species. When the amino acid residue substitutions that led to increased enzyme performance were introduced into a recombinant Camelina sativa DGAT1 (CsDGAT1B, GenBank TM accession number XM_010417066), the performance of CsDGAT1B was improved.

BnaDGAT1 variants increased yeast neutral lipid content through increased activity and/or polypeptide accumulation
Previously, directed evolution of BnaDGAT1 generated numerous variants resulting in increased TAG accumulation in yeast strain 4 delta (4,18). To investigate the underlying mechanisms responsible for the increased neutral lipid content, several BnaDGAT1 variants were assessed for in vitro enzyme activity and protein production level. The relative levels of DGAT activity and protein accumulation varied among the variants (supplemental Fig. S1A). When compared with the WT BnaDGAT1, the majority of the enzyme variants displayed higher activity, increased polypeptide accumulation, or higher activity combined with increased polypeptide accumulation (supplemental Fig. S1A). Interestingly, some BnaDGAT1 variants resulted in decreased in vitro enzyme activity and/or decreased polypeptide accumulation, although the introduction of these variants resulted in increased TAG accumulation in yeast compared with WT BnaDGAT1 (supplemental Fig.  S1A). This might be explained by the dynamic changes in enzyme activity of BnaDGAT1 variants in yeast over their growth phase (supplemental Fig. S1B). Based on these initial results (supplemental Fig. S1), it was hypothesized that the increased TAG production (as reflected by increased neutral lipid production) in yeast is due to increased enzyme activity, increased polypeptide accumulation, or a combination of enzyme activation and increased polypeptide accumulation. To test this hypothesis, six BnaDGAT1 variants with widely distributed amino acid residue substitutions throughout the entire polypeptide were chosen from our previously established directed evolution library for characterization. Neutral lipid content in yeast cultures producing these variants was confirmed to be equal to or higher than yeast producing the recombinant WT enzyme by the Nile Red assay (Fig. 1A). Because the enzyme activities and protein accumulation of the BnaDGAT1 polypeptide varied during yeast growth, their production profiles were examined. The production of the variant enzymes followed a similar pattern after induction. The activity of the recombinant enzymes increased markedly during the log phase and then decreased after reaching the stationary phase (Fig. 1B). The highest activity of each variant occurred at the late log or early stationary phase. Significant increases in DGAT activity were observed for all variants relative to that of the WT enzyme except for K289N, which displayed relative activities over time similar to the WT enzyme (Fig. 1C). The microsomes displaying the highest activity of each variant were analyzed for their corresponding recombinant protein accumulation (Fig. 1D). When compared with recombinant WT BnaDGAT1, all recombinant BnaDGAT1 variants exhibited equal or higher band density with the exception of K289N. The enzyme activities were then normalized to the Western blot results, and the normalized relative activity of each variant was similar to or higher than that of the WT enzyme (Fig. 1E).

Impact of amino acid substitutions in the ninth predicted transmembrane domain on enzyme activity and accumulation in yeast
In a previous study, directed evolution of BnaDGAT1 revealed that PTMD9 contains important amino acid residue substitutions affecting DGAT1 performance (17). To further investigate the effects of mutations in PTMD9 on enzyme activity and accumulation in yeast, three variants, including L441P, I447F, and F449C, with amino acid residue substitutions in BnaDGAT1-PTMD9 were characterized in detail. Similar to previous results, variant L441P or I447F displayed increased enzyme activity relative to that of the WT enzyme (supplemental Fig. S2). It is interesting to note that L441P, I447F, or F449C had very different effects on enzyme activity ( Fig. 1B and supplemental Fig. S2) and accumulation (Fig. 2, A and B), although all amino acid residue substitutions are located very closely in PTMD9 of BnaDGAT1. The normalized relative activity of L441P or F449C was significantly higher than that of WT BnaDGAT1. I447F, however, displayed normalized activity comparable with that of WT enzyme (Fig. 2C).
Altered polypeptide accumulation levels for the recombinant BnaDGAT1 variants in yeast 4 delta, however, might be attributable to variations in cDNA expression. DNA copy number and expression level of each BnaDGAT1 variant were assessed by qPCR. All plasmids containing BnaDGAT1 and its variants

Plant DGAT1 variants with enhanced performance
showed similar copy number (supplemental Fig. S3A). Consistent with previous results (17), BnaDGAT1 and its variants displayed high expression levels in the yeasts (supplemental Fig.  S3B), and no significant difference was observed for the selected BnaDGAT1 variants compared with that of WT DGAT1.

Kinetic characterization of microsomal recombinant BnaDGAT1 variants
To kinetically characterize the effects of the single amino acid residue substitutions in BnaDGAT1-PTMD9, the activities of microsomal recombinant BnaDGAT1 variants (L441P, I447F, F449C, and the out-group variant V125F) were examined over a range of oleoyl-CoA concentrations (Fig. 3), which were below the critical micelle concentration of oleoyl-CoA (about 30 M) (19). Variant V125F, I447F, or F449C exhibited an enzyme activity dependence on increasing acyl donor concentration similar to that of microsomes containing the recombinant WT BnaDGAT1. DGAT1 activity increased markedly at low oleoyl-CoA concentration, ranging from 0.1 to 5 M, and then the enzyme activity reduced gradually with further increases in oleoyl-CoA concentration. Variant L441P, however, displayed high activity at all oleoyl-CoA concentrations examined, suggesting that this enzyme variant may have a better tolerance for increased oleoyl-CoA concentration.
The initial reaction velocity data of BnaDGAT1 and its variants were fitted to the Michaelis-Menten or substrate inhibition equation, and the substrate inhibition kinetics was the preferred model for all of these BnaDGAT1 variants (Fig. 3, A-E). The apparent kinetic parameters were calculated and are summarized in Table 1. Kinetic parameters are indicated as being "apparent" because the enzyme, acyl acceptor (sn-1,2-diacylglycerol), and triacylglycerol product are all insoluble, contributing to a potentially complex kinetic situation. Variant L441P displayed weak substrate inhibition for oleoyl-CoA, with K i app value increasing as much as 6.4-fold, compared with the WT BnaDGAT1 value of about 13.9 M. There was also a slight increase (

Plant DGAT1 variants with enhanced performance
BnaDGAT1 variants were slightly higher than that of WT BnaDGAT1 with the exception of L441P. The catalytic efficiencies (V maxЈ app /K m app ) of variant L441P and V125F were improved to 162 and 184%, respectively, over WT BnaDGAT1.
The N-terminal fragments of BnaA.DGAT1.b (GenBank TM accession number AF164434.1), another DGAT1 isoform from B. napus, and mouse DGAT1 (20) were shown to bind acyl-CoA in a cooperative manner (21). Recently, it has been reported that DGAT1s from Corylus americana and maize displayed sigmoidal kinetics in response to increasing acyl-CoA at lower concentration (10). A closer look at the initial reaction velocities at lower oleoyl-CoA concentration (Fig. 3, F-J) showed that microsomal recombinant WT BnaDGAT1 and its variants in fact exhibited sigmoidal kinetics.

Sequence alignment of the ninth predicted transmembrane domain of BnaDGAT1 among various DGAT1s
Given the fact that amino acid residue substitutions in BnaDGAT1-PTMD9 improved enzyme performance, it was hypothesized that PTMD9 would also be present in other DGAT1s. Amino acid sequence alignment of 43 DGAT1s suggested that PTMD9 of BnaDGAT1 is moderately conserved, with 59.6% sequence identity among all species and 79.2% sequence identity among DGAT1s from the plant kingdom (Fig. 4). Different conservation levels were observed for the beneficial amino acid residue substitutions in BnaDGAT1-PTMD9, including L441P, I447F, and F449C. According to the phylogenetic tree, BnaDGAT1 Leu-441 and Ile-447 are highly conserved among the DGAT1 from the order Rosids and land plants, respectively, whereas BnaDGAT1 Phe-449 is more divergent. The high conservation of the beneficial amino acid residues in the predicted TMD9 suggested the possibility of modifying DGAT1 from other species by using the information obtained from BnaDGAT1.

Creation of high-performance C. sativa DGAT1 variants
It was hypothesized that the amino acid residue substitutions resulting in the improved performance of BnaDGAT1 could be installed in DGAT1 enzymes for other oil crops. C. sativa is an emerging oil crop, which possesses important agronomic traits and is considered as an ideal platform for the production of oil feedstock for biofuels and industrial compounds (22). Three C. sativa DGAT1 (CsDGAT1A, -B, and -C) cDNAs were identified from developing C. sativa (cv. CAME) seeds (23). The deduced amino acid sequences of the three CsDGAT1 genes shared 98.5% sequence identity and around 85% sequence identity when compared with BnaDGAT1 (supplemental Fig. S4). To determine whether the beneficial amino acid residue substitutions in BnaDGAT1-PTMD9 also could improve a DGAT1 from another species, the corresponding substitutions of two conserved BnaDGAT1 variants (L441P and I447F) in

Plant DGAT1 variants with enhanced performance
BnaDGAT1-PTMD9 and one out-group variant V125F (in the first predicted TMD) were introduced into CsDGAT1B (I144F, L460P, and I466F; supplemental Fig. S4). As shown in Fig. 5A, the three single CsDGAT1B variants (I144F, L460P, and I466F) and a double mutant L460P/I466F were produced in yeast 4 delta to determine the effects on neutral lipid content. The introduction of all four recombinant CsDGAT1B variants resulted in increased neutral lipid accumulation when compared with the effect of recombinant WT CsDGAT1B.
Microsomal fractions containing recombinant CsDGAT1 variants were prepared from corresponding yeast cultures harvested with similar optical density values at 600 nm (A 600 , about 7.5) at mid-log phase and were used to determine enzyme activ-

Plant DGAT1 variants with enhanced performance
ity. All of the variants exhibited higher DGAT activity than the WT enzyme, increasing as much as 6-fold for the most effective CsDGAT1B variants, L460P and L460P/I466F (Fig. 5B). Substrate inhibition kinetics were also observed for microsomal recombinant CsDGAT1B and the variants of CsDGAT1B (Fig.  5C). The apparent kinetic parameters were calculated based on the substrate inhibition equation (

A model that takes into consideration both sigmoidal and substrate inhibition kinetics
The observation of both sigmoidal and substrate inhibition kinetics for BnaDGAT1 and CsDGAT1B suggests that these DGAT1s are regulated by their substrates. Indeed, the classical sigmoidal (Equation 1) or substrate inhibition (Equation 2) model alone is not ideal for the prediction of the enzyme kinetics of the observed behaviors of BnaDGAT1 or CsDGAT1B, where v represents the reaction velocity, S represents the substrate concentration, S 0.5 represents the substrate concentration resulting at 50% of V max , and n represents the Hill coefficient, where v represents the reaction velocity, S represents the substrate concentration, K m represents the substrate concentration resulting at 50% of V max , and K i represents the inhibition constant. Thus, a new kinetic model (24) that takes into account both substrate inhibition and sigmoidicity is shown in Equation 3, where v represents the reaction velocity, S represents the substrate concentration, K m represents the substrate concentration resulting at 50% of V max , K i represents the inhibition constant, and n represents the Hill coefficient. This model predicts an initial lag in activity (sigmoidal) before reaching maximum activity, which is followed by a gradual reduction in activity due to the impact of the substrate inhibition effect (Fig. 6). This model resulted in a better fit for the kinetics of all DGAT1 variants (similar or slightly higher R 2 , supplemental Table S1) when compared with the results obtained from the substrate inhibition model (Figs. 3 and 5).

Discussion
Amino acid residue substitutions in DGAT leading to enhanced activity have been reported in DGAT1s from maize (11), Tropaeolum majus (25), C. americana, and soybean (10). However, due to the absence of a three-dimensional structure of this enzyme, the exact mechanisms for the effect of the beneficial amino acid residue substitutions on DGAT function remain elusive. It has been suggested that the amino acid residues potentially affecting enzyme function may be far away from each other in the amino acid sequence (10) but may in fact move close to each other to contribute to the active site after polypeptide folding.
Recently, PTMD9 of BnaDGAT1 was shown to contain key amino acid residues influencing the activity of the enzyme (17). Amino acid residue substitutions in recombinant BnaDGAT1-

Plant DGAT1 variants with enhanced performance
PTMD9 variants generated through directed evolution resulted in increased neutral lipid accumulation in S. cerevisiae strain 4 delta. In the current study, the effect of amino acid residue substitutions in PTMD9 on DGAT1 performance was investigated. None of the amino acid residue substitutions of the studied BnaDGAT1 variants reside in the putative functional motifs of DGAT1 (14), and the substitutions were between the amino acid residues with similar properties. Three BnaDGAT1 variants with amino acid residue substitutions in PTMD9 (L441P, I447F, or F449C) and one out-group variant V125F were characterized in detail by analyzing their enzyme activities, production levels, and enzyme kinetics. As shown in Figs. 1 and 2 and supplemental Fig. S2, BnaDGAT1 variants were found to boost neutral lipid production in the yeast transformants for different reasons. The amino acid residue substitutions affected BnaDGAT1 activity and/or polypeptide production. Variant L441P exhibited increased activity, whereas variant I447F accumulated to higher levels than WT BnaDGAT1. In contrast, variant F449C exhibited both increased enzyme activity and polypeptide accumulation. These differences occurred despite the fact that amino acid residues 441, 447, and 449 were in close proximity in PTMD9. It should be noted that the increased enzyme activity for these variants during the log phase was not tightly associated with changes in the level of polypeptide accumulation (Fig. 2). Because the DGAT assay was conducted using yeast 4 delta microsomal fractions, it is possible that the measured enzyme activity was affected by variations in the endogenous diacylglycerol content of the microsomes, although a relatively high amount of exogenous diacylglycerol was added to the reaction mixture.
The increased BnaDGAT1 polypeptide accumulation observed for variant I447F and F449C (Figs. 1 and 2) may be related to enhanced protein synthesis and/or the increased cellular stability of the variant polypeptide. The expression of WT BnaDGAT1 and its variants may also be controlled by transcriptional regulation. Although expression and plasmid copy number of WT BnaDGAT1 and that of its variants in yeast were at similar levels (supplemental Fig. S3), we still cannot completely rule out the possibility of variations in cDNA expression affecting polypeptide accumulation levels, especially considering that the BnaDGAT1 variants were carried by a high-copy plasmid and expressed under the control of the strong Gal1 promoter. For BnaDGAT1 variants with increased activity (e.g. L441P), in addition to contributing to a more favorable conformation in support of catalysis, it is also possible that the increased enzyme activity may be related to an altered interaction with a hypothetical modulator. Indeed, certain protein partners and modulatory molecules have been shown to be required for DGAT function (14, 26 -29). It is also known that DGATs can be regulated at the post-translational level via phosphorylation and/or ubiquitination (25,30). Putative phosphorylation and ubiquitination sites in BnaDGAT1 were predicted by NetPhosK (31) and UnPred (32), respectively, and among the selected variants, only residue Lys-110 is identified as being related to ubiquitination. No differences in protein accumulation were observed, however, between variants K110N/L441P and L441P (data not shown). It should also be noted that SDS-PAGE and subsequent Western blotting of microsomal pro-teins, prepared from yeast cells harvested at different culture times, did not reveal additional polypeptide fragments lower in molecular mass than full-length BnaDGAT1 (or the variants examined), suggesting that the enzyme was subject to little or no proteolysis. Nonetheless, it would be worthwhile to conduct a comprehensive study on the effects of amino acid residue substitutions on the protein stability of the BnaDGAT1 variants.
The enzyme kinetics of BnaDGAT1 were also affected by amino acid residue substitutions in PTMD9. When enzyme activity was determined at increasing acyl-CoA concentration, an eventual decrease in enzyme activity was observed for variant V125F, I447F, F449C, or WT BnaDGAT1 (Fig. 3, A, B,  D, and E). This might be explained by the effect of substrate inhibition. Previously, many acyl-CoA-dependent enzymes (33)(34)(35)(36)(37)(38), including DGATs (39 -41), have been found to be inhibited at relatively high concentrations of acyl donor. It is interesting to note that L441P displayed high activity at all oleoyl-CoA concentrations examined (Fig. 3C), suggesting that this enzyme variant may have a better tolerance of increasing oleoyl-CoA. An eventual decrease (to around 45% of the activity at 15 M oleoyl-CoA) in enzyme activity, however, was observed for L441P with a further increase in oleoyl-CoA concentration to 60 M (data not shown). The formation of acyl-CoA micelles (42) and the inhibition of enzyme activity by the detergent effect of acyl-CoA (39) might occur at such high concentration, together with substrate inhibition, resulting in the final decrease in enzyme activity. In other examples, the removal of substrate inhibition led to a concomitant decrease in enzyme activity and/or increase in K m value (43). Compared with WT BnaDGAT1, variant L441P displayed weak substrate inhibition accompanied by increased activity and similar K m value (Table 1). Variant L441P may be particularly useful for increasing seed oil content when combined with a strategy for increasing the concentration of the acyl-CoA pool.
The beneficial amino acid residue substitutions in BnaDGAT1-PTMD9 are conserved among various species (L441 and I447 in Fig. 4), suggesting that the knowledge generated from BnaDGAT1 variants might be transferred to other DGAT1s. Introducing the equivalent amino acid residue substitutions into CsDGAT1B resulted in enhanced neutral lipid accumulation in yeast 4 delta (Fig. 5A), which may be explained by the increased performance of the CsDGAT1B variants relative to that of WT CsDGAT1B (Fig. 5B). CsDGAT1B and its variants were found to exhibit substrate inhibition kinetics (Fig.  5C), and more importantly, CsDGAT1B variant L460P (Table  2) also displayed the improved catalytic properties (decreased substrate inhibition and increased enzyme activity). Together, the similarity in enzyme activity effects and kinetics between BnaDGAT1 variants and the corresponding CsDGAT1B variants demonstrated that the beneficial amino acid residue substitutions in BnaDGAT1 could be transferred to a DGAT1 from another plant species.
Sigmoidal kinetics for BnaDGAT1, CsDGAT1B, and their variants were observed at lower oleoyl-CoA concentrations (Figs. 3 (F-J) and 5D). The observed sigmoidal kinetics are consistent with a recent investigation of the allosteric properties of WT BnaDGAT1 (44). In addition, Roesler et al. (10) reported Plant DGAT1 variants with enhanced performance the sigmoidal kinetics for DGAT1s from C. americana and maize when varying the oleoyl-CoA concentration (0.1-5 M for C. americana DGAT1 and 1-10 M for maize DGAT1). The investigators modified C. americana DGAT1 by DNA shuffling and identified nine kinetically improved C. americana DGAT1 variants with increased substrate affinity and/or cooperativity (10). However, unlike Roesler et al. (10), the current study did not identify any trends in S 0.5 or Hill coefficient values for the BnaDGAT1 (or CsDGAT1B) variants compared with WT BnaDGAT1 (or CsDGAT1B). This discrepancy might be explained by the difference in the degree of modification of the DGAT1 sequence. In our study, error-prone PCR was used to introduce mutations, and by controlling the reaction conditions, the mean number of amino acid residue substitutions was estimated to be Ͻ3.8 in the resulting BnaDGAT1 variants (4). The BnaDGAT1 and CsDGAT1B variants examined only contained 1 or 2 amino acid residue substitutions. These very limited changes in amino acid sequence may not have resulted in large changes in enzyme conformation. In contrast, the C. americana DGAT1 variants generated by Roesler et al. (10) contained 6 -17 amino acid residue substitutions, probably resulting in more extensive effects on enzyme conformation and subsequent impacts on kinetic parameters. In addition, it would be useful to purify the most interesting BnaDGAT1 variants for additional kinetic studies under conditions where the variants are not potentially influenced by other proteins in the microsomes.
The combined substrate activation and inhibition kinetics observed for the WT plant DGAT1s and the variants might be explained by the presence of more than one acyl-CoA-binding site in the enzyme. Sigmoidal kinetics suggest that there is a cooperative effect when more than one substrate molecule binds to an allosteric enzyme (45). In addition to the proposed acyl-CoA-binding motif FYXDWWN in the hydrophobic segment (46), the hydrophilic N-terminal domain has been shown to associate with acyl-CoA via an allosteric interaction (14,20,21,44). Substrate inhibition could potentially arise when two substrate molecules bind to the enzyme to form a catalytically inactive ES2 complex by blocking second substrate binding (24) or product release (47) or when the substrate binding leads to the formation of less thermodynamically favored ES complex that turns over slowly (48 -50). In this study, it is hypothesized that the substrate inhibition of plant DGAT1 is caused by the formation of ES2 complex. This assumption is supported by the decrease in substrate inhibition observed when replacing the bulky aliphatic Leu-441 in BnaDGAT1 or Leu-460 in CsDGAT1B with Pro (Figs. 3 and 5), which might disturb the further binding of acyl-CoA to ES complex. It should be noted that the inhibition of DGAT activity observed at higher concentrations of acyl-CoA probably does not involve the allosteric acyl-CoA-binding site in the hydrophilic N-terminal domain of BnaDGAT1 because a truncation of BnaDGAT1, devoid of the hydrophilic N-terminal domain, also displayed substrate inhibition (44). Detailed structural information will be necessary to gain more insight into how the amino acid residue substitutions in BnaDGAT1 contribute to changes in the degree of substrate inhibition.
In conclusion, our results suggested that the beneficial amino acid residue substitutions in BnaDGAT1-PTMD9 led to enhanced enzyme performance in yeast via different mechanisms. The increased TAG accumulation in yeast producing these variants was due to variant enzymes with increased enzyme activity, increased protein accumulation, and/or reduced substrate inhibition. We also demonstrated that the beneficial amino acid residue substitutions in a DGAT1 from one species improved a DGAT1 from another species, which points out the possibility of systematically engineering DGAT1 by taking advantage of the amino acid residue substitution database generated based on the BnaDGAT1 variants. The fact that a single amino acid residue substitution can lead to an improved DGAT1 suggests that non-transgenic approaches, such as targeting-induced local lesions in genomes (TILLING) (51) or clustered regularly interspaced short palindromic repeats (CRISPR) (52) may represent useful methods for improving DGAT action in planta. Indeed, genome editing has recently been used to introduce loss-of-function mutations into DGAT1 and phosphoplipid:diacylglycerol acyltransferase genes in C. sativa, thereby leading to a reduction in seed TAG content (53).

Construct preparation, yeast transformation, and heterologous expression of DGAT1 variants
BnaDGAT1 variants together with WT BnaDGAT1 were recloned into the pYES2.1/V5-His TOPO yeast expression vector (Invitrogen, Burlington, Canada), under the control of the galactose-inducible GAL1 promoter. The stop codon of each gene was eliminated from the sequences for in-frame fusion with a C-terminal V5 tag encoded on the pYES2.1/V5-His TOPO vector. The CsDGAT1 variants were chemically synthesized by Invitrogen and inserted into the pYES2.1 vector.
After the integrity of all constructs was confirmed by sequencing, they were transformed into the quadruple mutant strain S. cerevisiae H1246 (4 delta, MAT␣ are1-⌬::HIS3, are2-⌬::LEU2, dga1-⌬::KanMX4, lro1-⌬::TRP1 ADE2) using an S.c. EasyComp transformation kit (Invitrogen). Yeast transformed with pYES-LacZ was used as a control. The recombinant yeast cells were first grown in liquid minimal medium (0.67% (w/v) yeast nitrogen base and 0.2% (w/v) SC-Ura) with 2% (w/v) raffinose for 24 h. An aliquot of the yeast cell culture (at a starting A 600 value of 0.4) was used to inoculate minimal medium containing 2% (w/v) galactose and 1% (w/v) raffinose (referred as induction medium). Cultures for all experiments were grown at 30°C with shaking at 220 rpm. To generate time course production profiles for the various BnaDGAT1 variants, cells were harvested by centrifugation (at 3000 ϫ g for 5 min at 4°C) every 2 h starting at an A 600 of 5 until the stationary growth phase.

Analysis of neutral lipid content in yeast
The Nile Red fluorescence assay was conducted as described previously (18). Yeast cultures were aliquoted (100 l/well) into 96-well solid black plates (Corning Inc.) and incubated with 5 l of Nile Red solution (0.1 mg/ml in methanol) for 30 s at room Plant DGAT1 variants with enhanced performance temperature. The fluorescence was measured before and after the addition of Nile Red solution with excitation at 485 nm and emission at 538 nm using a Synergy H4 Hybrid reader (Biotek, Winooski, VT). The Nile Red values were calculated based on the change in fluorescence over A 600 (⌬F/A 600 ).

Microsome extraction
Microsomal fractions were isolated from recombinant yeast cells as described previously (18). In brief, the recombinant yeast cells were resuspended in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.9, containing 10 mM MgCl 2 , 1 mM EDTA, 5% (v/v) glycerol, 300 mM ammonium sulfate, and 2 mM dithiothreitol) and homogenized in the presence of 0.5-mm glass beads by a bead beater (Biospec, Bartlesville, OK). The crude homogenate was centrifuged at 10,000 ϫ g for 30 min to sediment cell debris and glass beads. The supernatant was transferred into an ultracentrifuge tube and centrifuged at 105,000 ϫ g for 70 min to pellet the microsomes using an ultracentrifuge (Optima MAX-XP, Beckman-Coulter, Mississauga, Canada). The resulting microsomal fraction was resuspended in 3 mM imidazole buffer (pH 7.4) containing 125 mM sucrose. All procedures were conducted at 4°C. The concentration of crude protein was quantified by the Bradford assay (Bio-Rad, Mississauga, Canada) using BSA as a standard (54).
For kinetic studies of recombinant BnaDGAT1 variants, enzyme assays were allowed to proceed for 1 min using 0.2 g of microsomal protein. Recombinant CsDGAT1B variants were assayed for 4 min, and the quantity of microsomal protein used was as follows: for WT CsDGAT1B, 10 g of microsomal protein; for I144F and I466F, 2.5 g of microsomal protein; for L460P and L460P/I466F, 1 g of microsomal protein. The concentration of [1-14 C]oleoyl-CoA was varied from 0.1 to 25 M while sn-1,2-diolein was held constant at 333 M. Kinetic parameters were calculated by fitting the data to the Michaelis-Menten, substrate inhibition, or allosteric sigmoidal equation using the program GraphPad Prism version 6.0 (GraphPad Software, La Jolla, CA).

Western blotting
Equivalent amounts of microsomal proteins (10 g) from yeast strain 4 delta producing recombinant BnaDGAT1 variants and LacZ were resolved using 8 -16% gradient Mini-Protean TGX precast gels (Bio-Rad) and electrotransferred (overnight at 30 mA and 4°C) onto polyvinylidene difluoride membrane (GE Healthcare). The target C-terminally tagged recombinant BnaDGAT1 variants were detected using anti-V5-HRP-conjugated antibody (Invitrogen). To ensure equal protein loading, yeast constitutively producing Kar2p protein was used as an internal standard. Kar2p polypeptide was detected using a rabbit polyclonal anti-Kar2p (Santa Cruz Biotechnology, Inc.) as the primary antibody, followed by HRP goat anti-rabbit IgG (HϩL) secondary antibody (Invitrogen). Both HRP-conjugated antibodies were detected by chemiluminescence (FluorChem SP, Alpha Innotech Corp., San Leandro, CA) using an ECL Advance Western blotting detection kit (Amersham Biosciences). The band densities of BnaDGAT1 variants and internal standard were quantified with ImageJ software (55). The relative DGAT polypeptide accumulation level was calculated based on the density of the DGAT band after normalizing to internal standard.

Quantitative RT-PCR
Quantitative RT-PCR was performed on a StepOnePlus realtime PCR system (Applied Biosystems, Foster, CA) using the Platinum SYBR Green qPCR Master Mix (Invitrogen) as described previously (56).

Plasmid copy number assay
Plasmid copy numbers were quantified according to Karim et al. (57). Total DNA was extracted using glass bead beating, followed by phenol-chloroform extraction (58). Plasmid copy number was calculated using the standard curve method. To quantify the copy number, a section of the BnaDGAT1 gene (avoiding the mutated region) on the plasmid was targeted (with primers 5Ј-GGCCAATCCT-GAAGTCTCCTACT-3Ј and 5Ј-TGGGAGCAAGCATG-AAATACG-3Ј) and compared with a genomic target in the single copy ScALG9 gene (Saccharomyces Genome Database (SGD) accession number S000005163, with primers 5Ј-GCCGTTGCCATGTTGTTGTA-3Ј and 5Ј-GACCCAGTG-GACAGATAGCG-3Ј). Standard curves were created using plasmid containing both BnaDGAT1 and ScALG9 genes in the pYES2.1 vector. A serial dilution of this plasmid with concentration ranging from 5 ϫ 10 2 to 5 ϫ 10 7 copies/l was used to create standard curves for both BnaDGAT1 and ScALG9. Copy number here refers to the copy number per haploid genome rather than per cell, considering that there are two copies of the genome per cell during certain phases of cell division.

Gene expression analysis
Gene expression of BnaDGAT1 and its variants in yeast was analyzed as follows. Total RNA was isolated from yeast cells at the mid-log phase using the RNeasy kit according to the manufacturer's instructions (Qiagen, Toronto, Canada). First-Plant DGAT1 variants with enhanced performance strand cDNA was synthesized in a 10-l reaction mixture with 1 g of total RNA using the SuperScript III first-strand cDNA synthesis kit (Invitrogen). The relative expression levels of the BnaDGAT1 variants in yeast were calculated using the comparative Ct method (2 Ϫ⌬⌬Ct method) (59). The results are presented as -fold differences in gene expression after normalizing to the yeast stably expressing ScACT1 gene (SGD accession number S000001855). cDNA of variant V333I was used as a calibrator to normalize for plate-to-plate variation. The primers for BnaDGAT1 and its variants were the same as the primers we used to quantify plasmid copy number. The primers for ScACT1 cDNA were 5Ј-TCGTTCCAATTTACGCTGGTT-3Ј and 5Ј-CGGCCAAATCGATTCTCAA-3Ј.

Amino acid sequence analysis
Multiple amino acid sequence alignments of 43 DGAT1 proteins from different species (supplemental Table S2) were conducted using ClustalW in MEGA 7 with default settings (60). The alignment was used to construct a neighbor-joining tree using the same software with 1000 bootstrap repetitions. The topology organization of BnaDGAT1 was predicted using Phobius (61).

Statistical analysis
All experiments were repeated at least twice (n ϭ the number of independent experiments). Data are shown as means Ϯ S.D. when n was Ն3, or mean Ϯ range when n ϭ 2, unless otherwise stated. Statistical analysis was performed using a one-way analysis of variance (ANOVA), and mean separation was determined using the LSD test using the SPSS version 16.0 statistical package (SPSS, Chicago, IL). Means were considered significantly different at p Ͻ 0.05.
Author contributions-Y. X. performed all experiments, analyzed the data, and prepared the initial draft of the manuscript. R. J. W., Y. X., and G. C. designed the research approach. G. R. isolated the cDNA of CsDGAT1B and designed CsDGAT1B variants. G. C., M. S. G., K. M. P. C., S. S., L. W., M. J. L., and J. O. contributed valuable discussion during this study. All co-authors contributed to further editing of the manuscript.