Expression of Umbelopsis ramanniana DGAT2A in Seed Increases Oil in Soybean

Oilseeds are the main source of lipids used in both food and bio-fuels. The growing demand for vegetable oil has focused research toward increasing the amount of this valuable component in oilseed crops. Globally, soybean is one of the most important oilseed crops grown contributing about 30% of the vegetable oil used for food, feed and industrial applications. Breeding efforts in soy have shown that multiple loci contribute to the final content of oil and protein stored in seeds. Genetically, the levels of these two storage products appear to be inversely correlated with an increase in oil coming at the expense of protein and vice versa. One way to overcome the linkage between oil and protein is to introduce a transgene that can specifically modulate one pathway without disrupting the other. We describe the first transgenic soy crop with increased oil that shows no major impact on protein content or yield. This was achieved by expressing a codon-optimized version of a diacylglycerol acyltransferase (DGAT2A) from the soil fungus Umbelopsis (formerly Mortierella ) ramanniana in soybean seed during development resulting in an absolute increase in oil of 1.5% (by weight) in the mature seed.

conclusions were reached by Kroon et al. (2006) in their study of developing castor seed.
A similar study of castor development by He et al. (2005) reported DGAT1 may play a dominant role in ricinoleic acid production. Determining the contributions of DGAT 1 and 2 in oil production within the seed is a challenging task since the elimination of one gene by mutagenesis or gene silencing tends to result in at least partial compensation by the other (Zou et al., 1999;Routaboul et al., 1999). The functional convergence of these two protein classes ensures TAG production is enabled within the seed in order to sustain the storage of this energetically important molecule.
Two studies conducted under greenhouse conditions, in Arabidopsis and corn, have demonstrated increasing DGAT activity during seed development can impact the final oil content in seed (Jako et al., 2001, Zheng et al., 2008. In these studies endogenous DGAT1 genes were overexpressed in their native hosts and achieved absolute increases in oil of 9-12% for Arabidopsis and approximately 1% for corn. In this report we describe the transgenic expression of a fungal DGAT2A gene from U.
ramanniana that was optimized for expression in soybean, analysis of the developing and mature seed, and description of the phenotype observed in multi-year, multi-location field trials for a commercial quality event. The data presented supports the commercial viability of increasing oil in soybean by increasing DGAT activity in the seed.

Expression of UrDGAT2A in seed increases oil in soybean
We previously reported the identification of the DGAT2 gene family by purification of two proteins (DGAT2A and 2B) from Umbelopsis ramanniana and 7 Homogyzous R2 seed from multiple events (produced in the greenhouse) were analyzed for oil and protein content by near infrared transmittance (NIT). Statistical analysis of the data showed a 1-2% oil increase in homozygous seed with no significant change in protein content. A difference in the level of oil between homozygous, hemizygous and null seed was also observed (Fig. 2), indicating a positive correlation between genedosage and phenotype. Further testing was required to assess the impact on crop yield.

Developmental Analysis
Homozygous seed was further characterized at nine different stages during seed development ( Fig. 3A-C). The stages were selected based on seed weight and color (1-6 early development, 7 mid development, 8-9 late development). DGAT activity, seed weight, UrDGAT2A mRNA level (by RT-PCR), and UrDGAT2A protein presence (by Western) were also determined ( Fig. 3A,B). Data showed that UrDGAT2A mRNA increased as the immature green seed developed (up to stage 4) but declined before seed expansion was complete (stage 7) and was very low after the seed turned yellow and began to desiccate (stages 8-9). In contrast to mRNA, DGAT activity and UrDGAT2A protein levels increased early and reached a peak mid-development that remained throughout seed desiccation and in mature seed (Fig. 3A,B). The presence of DGAT activity in transgenic seed (Fig. 3A, striped bars) during desiccation (stage 8) and in mature seed (stage 9), differ from native DGAT activity levels in soybean (Fig. 3A, solid bars). Endogenous DGAT activity levels in soybean were similar to those reported by Weselake et al. (1993) for oilseed rape (Brassica napus L.) and safflower (Carthamus tinctorius) seeds where they exhibited a peak in activity early in development that decreased as the seed matured.
The presence of active UrDGAT2A in mature soybean seed enabled purification of the transgenic protein for characterization. Using mature seed as starting material, UrDGAT2A was initially purified from a combination of the 20K x g floating fat pad and the 200K x g membrane fraction obtained from high speed centrifugation of the 20K x g supernatant. Using these two fractions, 1-3 mg UrDGAT2A (at 75% purity) was obtained from 50 g of seed. To obtain protein with even higher purity, only the fat pad was used as starting material. This protocol eliminated several protein bands in the final 8 preparation and improved protein purity to >95%. The specific activity of UrDGAT2A protein purified from transgenic soybean was approximately the same as the fungal purified protein (Lardizabal et al., 2001). We also studied the kinetics of purified UrDGAT2A with respect to a labeled 1-[ 14 C]-18:1-CoA donor and the unsaturated 18carbon diacylglycerol substrates di18:1, di18:2, and di18:3. At saturating levels of 18:1-CoA, no significant differences were observed in the kinetic parameters of the three diacylglycerol acceptors (Table IA). Similarly, at saturating levels of diacylglycerol species no significant differences were observed with respect to 1-[ 14 C]-18:1-CoA (Table   IB). UrDGAT2A does not appear to demonstrate any preference with regard to these substrates, which comprise the majority of fatty acids found in soybean oil.

Protein Localization
Protein localization was assessed using both confocal microscopy and transmission electron microscopy (TEM). Confocal results showed UrDGAT2A was present in the endoplasmic reticulum (ER), the membranes of oilbodies, and the plasmodesmata connecting cells (Fig. 4). At stage 6, UrDGAT2A was mainly localized in the ER, in some small oilbodies, and in the plasmodesmata (Fig. 4A). As seed development progressed through stages 7 and 8, the signal for UrDGAT2A was stronger and more frequently observed in the membranes of oilbodies (Fig. 4B). Also, oilbody size and subcellular localization were different in cotyledons expressing UrDGAT2A compared to wild-type. In transgenics, although the majority of oilbodies were still localized at the periphery of the cell, their size was 1.2 to 2.3-fold larger, with a variable number of significantly larger oilbodies randomly dispersed throughout the cell. Larger oilbodies were prominent in stages 7 and 8 for which oilbodies with 13-14 µm radius were commonly observed (Table II). Protein bodies appeared to be smaller than those observed in wild-type cells but were more numerous. They were also frequently observed to be 1.4 to 2.5-fold smaller than the oilbodies. In addition, the size of storage vacuoles was smaller in cotyledon cells expressing UrDGAT2A. The distribution of lipid and protein from the abaxial to the adaxial region of the cotyledon appeared identical for wild-type and UrDGAT2A seed, with protein maximal in the abaxial region and lipid maximal in the adaxial region of the cotyledon.
TEM results corroborated the confocal results, and in addition revealed that the dotted-like pattern observed in confocal microscopy resulted from alternating regions of ER with and without UrDGAT2A. This data is consistent with the findings of Shockey et al. (2006) and Cahoon et al. (2007) who showed that native DGAT1 and DGAT2 were localized to distinct subdomains of the endoplasmic reticulum. In addition, results showed UrDGAT2A was uniformly distributed in the oilbody membrane. Some micrographs suggested oilbody coalescence, which would explain larger oilbody formation. A significant improvement in membrane perseveration was achieved by preparing the samples with high pressure freezing. Using this technique, we were also able to observe the presence of UrDGAT2A in the plasmodesmata (Fig. 4A, arrow).

Seed Analysis
To examine the transgenic trait in even greater depth, a single-copy, commercialquality event with a 1.5% oil increase was selected for additional studies. To determine the degree of processing of the UrDGAT2A protein in soybean, traditional Edman sequencing was attempted on the purified protein. Sequencing of the N-terminus failed, as it had with the native, fungal-purified UrDGAT2A, indicating it was blocked. Instead, sequence of the purified transgenic protein was analyzed using enzyme digestion, isotopic labeling, and MALDI-TOF mass-spectrometry and the overall sequence was confirmed by peptide mass fingerprinting (Courchesne et al., 1999). The N-terminal methionine was found to be truncated and the new N-terminal amino acid, alanine, was blocked by acetylation. The C-terminus was confirmed to be intact.
No significant changes in the content of protein or carbohydrates were detected in mature seed of the transgenic event relative to controls. We did observe a 15% decrease (on a % dry weight basis) in neutral detergent fiber (NDF, significant at p<0.05) in the transgenic positive relative to the control but saw no difference in acid detergent fiber (ADF). A small increase was observed in the amount of 18:1 fatty acid (+1.13%) with a corresponding decrease in the amount of 18:2 and 18:3 fatty acids (-0.72%, -0.40% = -1.12%) in the final TAG pool that were statistically significant (p<0.05). However, this small compositional difference is well within the range of normal soybean fatty acid compositions observed in different varieties (http://www.codexalimentarius.net).

Field Trials
To determine whether oil increases observed in greenhouse-grown R2 seed translated to increases in the field, the homozygous commercial-quality event containing a single insertion of the UrDGAT2A gene (same event used for seed analysis) were grown at 63 locations over three years along with the corresponding control lines and nontransgenic checks. The transgenic DGAT event was evaluated for oil, protein, and yield in five seasons of field trials in the U.S. and Argentina from 2004 through 2006. All field trials were conducted as split-plot designs with transgenic event as the whole plot and the negative and positive isolines as the split-plot. We did not observe any agronomic differences between the transgenic soybean plants expressing UrDGAT2A and the control plants during growth. They were identical with respect to germination, seedling vigor, plant height, time to flowering and time to harvest.
At all locations and generations, oil was increased significantly over the negative isoline by an average of 1.5% (p<0.05) indicating generational stability of the oil phenotype ( Fig. 5 A,B). Though there appeared to be a slight decrease in the average yield in 4/5 field trials, grain yield observed over five seasons in both the United States and Argentina showed no statistical difference between the transgenic event and its negative isoline control (Fig. 5C). No major impact on protein level was observed in transgenic seed indicating the oil increase associated with this transgene does not carry a (1:1) corresponding decrease in protein in contrast to results obtained from traditional breeding ( Fig. 5D) (Burton, 1984;Wilcox, 1998). In addition, when crossed into high oil soy germplasm, transgenic events increase oil in that germplasm, indicating the endogenous and transgenic mechanisms for achieving a high oil phenotype are complementary. Progeny of events crossed into a high protein germplasm also maintain the high protein phenotype while exhibiting an increase in oil level corresponding to the transgene donor parent (data not shown).

DISCUSSION
We report the first transgenic expression of DGAT2 in any plant for the purpose of increasing oil content in seed and the first transgenic high oil soybean. This was achieved using a fungal DGAT2 from Umbelopsis ramanniana that was optimized for expression in soy. Introduction of the transgene led to a 1.5% increase (by weight) in seed oil with no significant impact on yield or protein content. To determine whether the additional carbon found in oil came at the expense of another seed component, we also measured carbohydrates (sucrose, raffinose, stachyose) and acid and neutral detergent fiber (ADF and NDF, respectively). The only difference observed was a decrease in NDF (15% on a dry weight basis, p<0.05) which could have contributed to the oil increase in transgenic seed. We detected a similar reduction (13%) in NDF in the non-transgenic high oil soybean line DKB31-51, which was grown as a positive control for oil levels, however, the difference in this case did not meet the significance test. A trend toward lower fiber was observed in 5/6 events therefore these results could indicate that a reduction in fiber may support increased carbon flow to oil (data not shown). Another possibility is the loss in carbon came from several sources whose differences were not detectable. In addition to changes in the NDF fraction, the only other measurable difference we observed in transgenic soybean was a significant increase in oil body size. Tzen et al. (1993), reported that oil bodies from different species had average diameters that were different but the differences were within a narrow range (0.6-2.0 μm). While we observed a similar range of oil body sizes in control soybean (0.72 -2.5 μm), the oil bodies in our transgenic soy were significantly larger (1-14 μm). Relationships between oil body size and oil content have been observed in several species (Lambert et al., 2000, Mantese et al., 2006 where there is a tendency for higher oil genotypes to contain slightly larger oil bodies than lower oil cultivars. Other transgenic strategies designed to increase oil in plant seeds have also been reported. In Arabidopsis, Zou et al. (1997)   In summary, increasing the amount of oil in soybean adds significant value to the crop if it can be accomplished without a yield penalty or a significant decrease in protein content. Transgenic expression of a codon-optimized UrDGAT2A has produced the first high-oil soybean crop that accomplishes these objectives. The mechanism for achieving this goal was to significantly increase the amount of diacylglycerol acyltransferase activity specifically in developing seed.

Enzyme assays
UrDgat2A activity is measured as described in Lardizabal et al. (2006) with the following changes. When assaying developing seed of transgenic plants, CHAPS (0.06%) replaced Triton X-100 as the detergent in order to maintain the endogenous DGAT activity present in the seed. This allows for a more realistic comparison between total DGAT activity levels in different events and was used to select the most active events for advancement at R1. When assaying fractions during the purification of DGAT from transgenic seed, phosphatidic acid is replaced with phosphatidyl choline in order to reconstitute DGAT activity.

Purification from transgenic plants to >95%
Starting with 50g ground mature seed suspended in buffer, centrifugation (20K x g, 30 min) yielded a floating fat pad, liquid fraction containing soluble and membrane associated proteins, and a pellet. The purest form of UrDGAT2A was obtained using only the fat pad for purification. DGAT activity was solubilized from the fat pad with 5% Triton X-100 (5.5:1 detergent: protein, w/w) and ultracentrifugation (200K x g, 1 hour).
Two different resins were used in the purification, Reactive Yellow 86 (Sigma R-8504) and ceramic hydroxyapatite (BioRad 157-0020). The high speed supernatant from detergent solubilization was loaded onto 500ml Reactive Yellow 86 in Buffer A (100 mM Tricine pH 7.5, 0.1% Triton X-100, 10% glycerol) containing 75 mM NaCl, washed with loading buffer and eluted with buffer A containing 500mM NaCl. The eluate was concentrated approximately 12-fold in an Amicon stirred cell (YM30 membrane) and applied to a 20 ml ceramic hydroxyapatite column. The majority of UrDGAT2A activity was collected in the flow through while a significant amount of protein bound the column and was removed. The sample was dialyzed to lower the salt concentration and applied to a 20 ml Reactive Yellow-86 column in buffer A with 75 mM NaCl and eluted in buffer A with a gradient of 75-500mM NaCl. The yield varied for each preparation but 1-3 mg UrDGAT2A protein with >95% purity was typically recovered.

Antibody development and Western analysis
Rabbit antibodies used to detect UrDGAT2A in the developing seed profile were generated to a small KLH-linked peptide (Zymed, Inc), KYGQTKDEIIRELHDS, near the C-terminal end of UrDGAT2A (Fig. 3c). Goat antibodies generated to the full-length UrDGAT2A (expressed and purified from E. coli as an N-HIS fusion) were used to detect protein in different soy tissues (Fig. 3d). Western analysis was performed as described (Burdette, 1981), using PVDF as the blotting membrane, and a 1:5000 dilution of the primary and secondary antibodies. Blots were developed with 1-Step NBT/BCIP reagent for alkaline phosphatase (Pierce #34042).

Developmental analysis
Single copy homozygous events expressing UrDGAT2A and a null segregant were used in the experiment. Plants were grown in the greenhouse and seed was collected throughout development and stored at -80C. Nine stages were chosen for analysis based  Controls were used to determine sample autofluorescence, absence of crossover or bleedthrough at the settings used for data collection, potential unspecific binding and other artifacts. Linear unmixing confirmed the signal origin from double and triple-labeled samples. Oilbodies were selectively stained by Nile Red (Greenspan et al., 1985;Siloto et al., 2006) (Chameleon laser at 790-nm, 10%, BP 565-615 IR). Endoplasmic reticulum (ER), and phospholipid bilayer membranes in general, were stained with ER Tracker Blue-White DPX (1:1000, 30 min, in the dark; Chameleon laser at 760-nm, 12%, BP 480-520 IR filter, selected to exclude crossover from autofluorescence signal above 560 nm). For immunogold labeling, cotyledons were ultrarapidly frozen by high pressure freezing (Liberton et al., 2006), ultra-sectioned, immunogold labeled with goat antirabbit-gold conjugate secondary antibody, silver enhanced, contrasted with uranyl acetate and viewed with a Leo 912AB transmission electron microscope. Wild-type sections and sections incubated in the absence of primary or secondary antibodies were used as negative controls.

Construction of plasmid
The plasmid pMON70900 contains two separate T-DNAs that are each flanked by left and right border elements. One cassette carries the gene-of-interest (UrDGAT2A) and the other, a marker (CP4) used to select events during the transformation process. The gene-of-interest cassette utilizes the seed-specific promoter 7Sa' from Glycine max and the E9 3'UTR from Pisium sativum to drive expression of the codon-optimized UrDGAT2A gene. The CP4 selectable marker is contained in a second T-DNA cassette and is driven by the Figwort Mosaic Virus 35S promoter and petunia hsp70 5'UTR with transit peptide EPSPS from Arabidopsis thaliana. The first generation of transformed seed was screened for absence of the T-DNA containing the selectable marker so that only marker-free events were carried forward.

Plant transformation, zygosity assays and linkage Southerns
Soybean was transformed as described in Martinell et al. (2002). Leaves of developing R1 plants were analyzed in a PCR-based assay (http://www.twt.com/invader_chemistry/invaderchem.htm) designed to quantitate the amount of a specific DNA target, in order to determine if the plants were homozygous, heterozygous or null for the transgene. Events where the transgene and the selectable marker were linked (closely co-localized on the same chromosome) were identified by Southern analysis and eliminated.

Field Trial Data
Field trials were conducted as two row plots seeded at nine seeds per foot in twelve foot rows with a three foot border on thirty inch row spacing. Seed was sampled from every plot and submitted for proximate analysis using Near-Infrared Transmittance.
Oil and protein data were reported on a dry matter basis. Outlier analysis based on deleted studentized residuals using SAS was performed on all data prior to statistical analysis. Statistical analyses were run using mixed model procedures in SAS. The analysis was performed using a split-plot model to compare positive to negative isolines within the transgenic event. The analysis was run across locations, with locations, replications within locations, and their interactions with the fixed effects considered random effects. Event, isoline, and their interactions were considered fixed effects.              . Field results from five growing seasons. A, Higher oil levels were observed for transgenic seed relative to the negative isoline over five seasons of oil data consisting of 63 multi-replication locations indicating environmental stability for the transgene trait. B, The across locations analysis for oil shows that in five multi-location field trials in both the U.S. and Argentina transgenic seed was significantly higher in oil than the negative isoline control. Data represents results for consecutive, sequential generations. This seed was not planted in Argentina during the 2005 field season. C, The across locations analysis for yield show no significant differences between the transgenic line and the negative isoline. D, Seed protein was significantly impacted in 3 of 5 field trails but the largest significant difference was 0.44%. *Values are significantly different (p<0.05) from control. DMB, dry matter basis.