This Article Abstract Full Text (PDF) Supplemental Data All Versions of this Article: 18/9/2294    most recent tpc.106.043695v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Shockey, J. M. Articles by Dyer, J. M. Search for Related Content PubMed PubMed Citation Articles by Shockey, J. M. Articles by Dyer, J. M. Agricola Articles by Shockey, J. M. Articles by Dyer, J. M.

Tung Tree DGAT1 and DGAT2 Have Nonredundant Functions in Triacylglycerol Biosynthesis and Are Localized to Different Subdomains of the Endoplasmic Reticulum^[W]

^a U.S. Department of Agriculture, Agricultural Research Service, Southern Regional Research Center, New Orleans, Louisiana 70124
^b Department of Molecular and Cellular Biology, University of Guelph, Guelph, Ontario N1G 2W1, Canada

² To whom correspondence should be addressed. E-mail rtmullen{at}uoguelph.ca or jdyer{at}srrc.ars.usda.gov; fax 519-837-2075 or 504-286-4367.

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Seeds of the tung tree (Vernicia fordii) produce large quantities of triacylglycerols (TAGs) containing 80% eleostearic acid, an unusual conjugated fatty acid. We present a comparative analysis of the genetic, functional, and cellular properties of tung type 1 and type 2 diacylglycerol acyltransferases (DGAT1 and DGAT2), two unrelated enzymes that catalyze the committed step in TAG biosynthesis. We show that both enzymes are encoded by single genes and that DGAT1 is expressed at similar levels in various organs, whereas DGAT2 is strongly induced in developing seeds at the onset of oil biosynthesis. Expression of DGAT1 and DGAT2 in yeast produced different types and proportions of TAGs containing eleostearic acid, with DGAT2 possessing an enhanced propensity for the synthesis of trieleostearin, the main component of tung oil. Both DGAT1 and DGAT2 are located in distinct, dynamic regions of the endoplasmic reticulum (ER), and surprisingly, these regions do not overlap. Furthermore, although both DGAT1 and DGAT2 contain a similar C-terminal pentapeptide ER retrieval motif, this motif alone is not sufficient for their localization to specific regions of the ER. These data suggest that DGAT1 and DGAT2 have nonredundant functions in plants and that the production of storage oils, including those containing unusual fatty acids, occurs in distinct ER subdomains.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Engineering temperate oilseed crops to produce novel value-added oils has been a long-standing goal of academic researchers and the biotechnology industry. Many of these oils hold great promise for use in human and animal nutritional regimes, and several others may serve as renewable chemical feedstocks that could replace petroleum-based products in industrial applications (reviewed in Jaworski and Cahoon, 2003; Dyer and Mullen, 2005; Singh et al., 2005). For instance, the seed oils of many exotic plant species contain high amounts of unusual fatty acids (e.g., epoxy, hydroxy, conjugated, or acetylenic) that can serve as raw materials for the production of inks, dyes, coatings, and a variety of other bio-based products. Large-scale production of these oils through traditional farming is often impossible because of the poor agronomic traits of these plant species. Furthermore, efforts to transfer genes encoding the proteins responsible for unusual fatty acid biosynthesis to higher yielding plants have generally met with limited success, with much lower amounts of the desired fatty acid accumulating in the oils of transgenic plants (15 to 30%) compared with the native plant species (up to 90%) (Thelen and Ohlrogge, 2002; Jaworski and Cahoon, 2003; Singh et al., 2005). It is clear from these studies that additional genes and significantly more knowledge of seed oil biosynthesis are needed before plants can be engineered to produce industrially important oils.

There are three major biosynthetic events involved in the production of seed storage oils. The first involves the synthesis of fatty acids in plastids. The second involves the modification of these fatty acids by enzymes located primarily in the endoplasmic reticulum (ER). The third involves the packaging of the nascent fatty acids into triacylglycerols (TAGs), which subsequently accumulate in oil bodies that bud off from the ER. Although a significant amount of information is currently available regarding the synthesis and modification of fatty acid structures (including the synthesis of unusual fatty acids) (Ohlrogge and Browse, 1995; Shanklin and Cahoon, 1998), much less is understood about the enzymes and cellular mechanisms required for the selection and transfer of fatty acids into storage TAGs.

Biochemical analyses have shown that TAG is synthesized in the ER by at least two pathways. The first involves the acyl-CoA–independent transfer of fatty acids from phospholipids to the sn-3 position of diacylglycerol to form TAG. This reaction is catalyzed by phospholipid:diacylglycerol acyltransferase (PDAT) (Dahlqvist et al., 2000; Ståhl et al., 2004). TAG is also produced via three successive acylation reactions of the hydroxyl groups of glycerol, starting from glycerol-3-phosphate, with diacylglycerol acyltransferase (DGAT) catalyzing the committed step: the transfer of a fatty acyl moiety from acyl-CoA to the sn-3 position of diacylglycerol (Kennedy, 1961). As such, DGAT plays an essential role in controlling both the quantitative (Ichihara et al., 1988) and qualitative (Vogel and Browse, 1996; He et al., 2004a) flux of fatty acids into storage TAGs.

DGAT enzyme activity is encoded by at least two classes of genes in eukaryotic cells. The type 1 class of DGAT enzymes (DGAT1) was discovered first in mouse based on homology with mammalian acyl-CoA:cholesterol acyltransferase genes (Cases et al., 1998); subsequently, other DGAT1 genes were identified and characterized in several plant species (Hobbs et al., 1999; Routaboul et al., 1999; Zou et al., 1999; Bouvier-Navé et al., 2000; Nykiforuk et al., 2002; He et al., 2004b; Milcamps et al., 2005). For instance, the Arabidopsis thaliana DGAT1 gene has been shown to contribute significantly to seed TAG biosynthesis, both by overexpression (Jako et al., 2001) and through mutational downregulation studies (Katavic et al., 1995; Routaboul et al., 1999). The type 2 class of DGAT enzymes (DGAT2) also has been identified in a number of eukaryotes, including fungi, Caenorhabditis elegans, human, and Arabidopsis (Cases et al., 2001; Lardizabal et al., 2001). The physiological function(s) of these DGAT2 enzymes in plants, however, has not been determined. Characterizing the subcellular properties of these enzymes would provide new insight into the underlying mechanisms of oil biosynthesis. This knowledge may be especially important for the production of seed oils containing unusual fatty acids, because these structures are generally incompatible with normal membrane lipids and the spatial separation of lipid biosynthetic enzymes in the ER may provide an efficient mechanism for channeling these unusual fatty acids into storage oils.

We are studying the tung tree (Vernicia fordii) as a model system for the production of seed storage oils that contain high amounts of industrially important fatty acids, with the end goal of transferring appropriate sets of genes to high-yielding oilseed crops for industrial oil production. Tung oil contains 80% (w/w total fatty acids) -eleostearic acid, an unusual trienoic fatty acid (18:3^{9cis,11trans,13trans}) that imparts industrially useful drying qualities to the oil (Sonntag, 1979). We previously identified and characterized the enzymes and cofactor proteins required for the production of fatty acid components in tung oil, including fatty acid desaturase (FAD) and FAD-related enzymes (Dyer et al., 2002a, 2004), cytochrome b₅ (Cb5) (Hwang et al., 2004), and Cb5 reductase (Shockey et al., 2005).

Here, we provide a detailed functional and cellular analysis of tung DGAT1 and DGAT2. We show that tung DGAT1 and DGAT2 enzymes are each encoded by single genes, and that DGAT1 is expressed at similar levels in leaves, flowers, and developing seeds, whereas DGAT2 is strongly induced in seeds during tung oil synthesis. Furthermore, although both enzymes synthesized TAGs in a yeast functional complementation assay, DGAT2 showed a clear preference for the production of trieleostearin, the major TAG in tung oil. We also present data from subcellular localization and selective photobleaching experiments of DGAT enzymes transiently expressed in tobacco (Nicotiana tabacum) suspension cells indicating that they are localized to specific, dynamic regions of the ER, and furthermore, that these regions are not shared by the two enzymes. Investigation of the molecular targeting signals within DGAT1 and DGAT2 revealed that both proteins contain a similar cytosolically exposed C-terminal pentapeptide ER retrieval motif responsible for maintaining their steady state localization in the ER. Notably, this motif is not sufficient for targeting a reporter protein to the ER subdomains enriched with full-length DGAT1 or DGAT2 protein, indicating that additional portions of DGAT1 and DGAT2, including portions potentially involved in homotypic oligomerization, are required for targeting to ER subdomains. Together, these results suggest that DGAT1 and DGAT2 enzymes perform different roles in plant cells and that the production of TAGs by each enzyme occurs in specialized regions of the ER. The implications for plant lipid metabolism and the engineering of crops to produce industrially important oils are discussed.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
DGAT1 and DGAT2 Represent Distinct Gene Families in Higher Eukaryotes
To identify full-length coding sequences for type 1 and type 2 tung DGAT enzymes, sequences of high similarity in other DGAT genes were used to identify DGAT cDNAs by standard homology-based cloning techniques (details are provided in Methods). Alignment of the deduced amino acid sequences of DGAT1 or DGAT2 from tung, Arabidopsis, and rice (Oryza sativa) revealed that the proteins in each family share 50% identity (Figures 1A and 1B ), but comparisons of the two groups of proteins revealed striking differences. For instance, DGAT1 proteins are 500 amino acids in length with 10 predicted transmembrane domains (TMDs) (Figure 1A), whereas DGAT2 proteins are 320 amino acids long with two predicted TMDs (Figure 1B).

View larger version (69K): [in this window] [in a new window]   Figure 1. Sequence Alignments and Phylogenetic Comparisons of DGAT1 and DGAT2 Proteins. (A) Multiple sequence alignment of deduced amino acid sequences of DGAT1 proteins from tung (Vf), Arabidopsis (At), and rice (Os). Identical residues are shaded black, and similar residues are shaded gray. Transmembrane domains, as predicted by the TMHMM server (http://www.cbs.dtu.dk/services/TMHMM-2.0/) (Krogh et al., 2001), are underlined. (B) Alignment of DGAT2 protein sequences. Proteins are labeled as in (A). (C) Phylogenetic analysis of various cloned and annotated DGAT sequences from plants, humans, and/or S. cerevisiae. Each plant species is represented by either DGAT1 or DGAT2, or both, depending on the availability of database sequences. The branch lengths of the tree are proportional to divergence. The 0.1 scale represents 10% change. Bootstrap values are shown in percentages at nodes. Proteins used in the analysis were DGAT1 and/or DGAT2 sequences from tung (Vf), Arabidopsis (At), Brassica napus (Bn), cotton (Gossypium hirsutum; Gh), rice (Os), soybean (Glycine max; Gm), tobacco (Nt), wheat (Triticum aestivum; Ta), human (Hs), and yeast (Sc).

Tung DGAT1 and DGAT2 Are Encoded by Single Genes, and DGAT2 Is Highly Expressed during Tung Seed Oil Biosynthesis
To further investigate the differences in DGAT1 and DGAT2 gene families, the genes corresponding to the tung cDNAs were cloned. Comparison of the genomic architecture of each gene revealed that their intron/exon organizations were entirely different (Figure 2A ), and genomic DNA gel blot analyses indicated that single copies of each gene exist in the tung genome (Figure 2B).

View larger version (76K): [in this window] [in a new window]   Figure 2. Genomic Organization and Expression Patterns of Tung DGAT1 and DGAT2 Genes in Relation to Tung Seed Oil Biosynthesis. (A) Intron/exon structure of tung DGAT1 and DGAT2 genes. All exons (black boxes) and introns (lines) are drawn to scale, and the relative positions of restriction sites used for DNA gel blotting in (B) are shown. (B) DNA gel blot of tung genomic DNA digested with EcoRV, NcoI, or SacI and hybridized with digoxigenin (DIG)-labeled full-length ORF probes for either DGAT1 or DGAT2. DIG-labeled HindIII molecular mass markers, in kilobase pairs, are shown at left or right (M). (C) RNA gel blot analysis of DGAT1 and DGAT2 gene expression. RNA was extracted from leaves, flowers, and developing seeds, and DGAT1 and DGAT2 transcripts were characterized using gel blot analysis with DIG-labeled DGAT1- or DGAT2-specific probe. Ethidium bromide–stained rRNA from each sample is shown as a loading control. (D) Accumulation of tung oil in developing tung seeds. Tung fruit were harvested throughout the growing season (June 25 to September 3), and changes in total seed oil (percentage dry weight) and eleostearic acid (percentage [w/w] total fatty acid methyl esters) contents are shown.

DGAT1 and DGAT2 Show Minor Differences in Substrate Selectivity in Vitro
To investigate the functional properties of tung DGAT1 and DGAT2 enzymes, microsomal membranes were prepared from mutant yeast cells disrupted in endogenous DGAT and PDAT enzyme activities (dga1 lro1, strain SCY1998 [Oelkers et al., 2002]) but expressing myc epitope–tagged versions of either tung DGAT protein. Isolated microsomal membranes were then incubated with one of four different radiolabeled acyl-CoAs, palmitoyl-CoA (16:0), oleoyl-CoA (18:1), linoleoyl-CoA (18:2), or -linolenoyl-CoA (18:3), combined with one of three different unlabeled diacylglycerol substrates, sn-1,2-diolein (di18:1), sn-1,2-dilinolein (di18:2), or sn-1,2-dilinolenin (di18:3). Figure 3 shows that DGAT1 enzyme activity varied widely depending on the combination of substrates used, with 18:2-CoA plus di18:2 and 18:3-CoA plus di18:1 yielding the highest levels of activity. DGAT2 had lower levels of activity overall compared with DGAT1, possibly because of the relatively lower steady state levels of DGAT2 protein detected by anti-myc protein gel blotting (data not shown). However, DGAT2 did show a modest preference for 18:3-CoA as the acyl donor, possessing higher activity with this acyl-CoA (regardless of the acyl acceptor diacylglycerol species) than with any of the other acyl-CoAs tested (Figure 3). Eleostearoyl-CoA and eleostearate-containing diacylglycerol substrates could not be tested with DGAT1 or DGAT2 in this in vitro assay system, because these substrates are sensitive to heat, oxidation, and acidic pH (Sonntag, 1979), making their synthesis problematic.

View larger version (26K): [in this window] [in a new window]   Figure 3. Biochemical Activity of DGAT Enzymes in Vitro. Myc-tagged tung DGAT1 and DGAT2 were expressed individually in mutant yeast cells that lacked the ability to synthesize appreciable amounts of TAG, and then microsomal fractions were isolated and incubated with combinations of radiolabeled fatty acyl-CoAs and diacylglycerols. Fatty acyl-CoAs were palmitoyl-CoA (16:0-CoA), oleoyl-CoA (18:1-CoA), linoleoyl-CoA (18:2-CoA), and -linolenoyl-CoA (18:3-CoA). Diacylglycerols were diolein (di18:1), dilinolein (di18:2), and dilinolenin (di18:3). Amounts (in picomoles per hour) of radiolabeled TAG formed per 25 µg of yeast microsomal membrane protein minus the amount of TAG produced in control experiments lacking exogenous DAG substrates are shown for each condition (average ± SD of three independent experiments). See Methods for additional details.

Cultivation of wild-type yeast cells (harboring an empty pYes3 expression plasmid) in the presence of tung oil and lipase, followed by extraction of yeast lipids and analysis of fatty acid composition by gas chromatography (GC), revealed that eleostearic acid accounted for 22% of cellular fatty acid composition (Figure 4A , WT+pYes3). Separation and analysis of TAGs by HPLC with photodiode array detection (HPLC-PDA) demonstrated that eleostearic acid was incorporated into a variety of TAG species by the endogenous yeast DGA1 or LRO1 enzyme (Figure 4B, WT+pYes3). Cultivation of mutant yeast cells harboring disruptions in the DGA1 and LRO1 genes, however, showed a reduction in eleostearic acid content to 8% (Figure 4A, MUT+pYes3), and no TAGs were identified in these cells by either thin layer chromatography (TLC) (Figure 4A, MUT+pYes3) or HPLC-PDA (Figure 4B, MUT+pYes3). Transformation of mutant yeast cells with a plasmid-borne copy of the DGA1 gene restored TAG biosynthesis (Figures 4A and 4B, MUT+DGA1), thereby establishing a functional complementation assay for DGAT enzyme activity.

View larger version (23K): [in this window] [in a new window]   Figure 4. Functional Analysis of DGAT Enzymes in Vivo. Wild-type (WT) and mutant (MUT; lacking the ability to produce TAGs) yeast cells were cultivated in the presence of tung oil and a nonspecific lipase, cells were harvested, and lipids were extracted for the identification of eleostearoyl-containing compounds. Yeast cells were also transformed with an empty plasmid (pYes3) or a plasmid expressing yeast DGA1, tung DGAT1, or tung DGAT2. Strain and plasmid combinations are shown along the bottom of (A) and at right in (B). (A) The top panel shows the percentage of eleostearic acid in yeast cells complemented with various DGAT1 or DGAT2 enzymes, as determined by GC-FID analysis (average ± SD, n = 3). The bottom panel shows a TLC separation of various lipid classes. Positions of lipid standards are shown at left: diacylglycerol (DAG), free fatty acids (FFA), phospholipids (PL), sterol esters (ST), and triacylglycerols (TAG). Samples are representative of three independent experiments. Note the reduction of eleostearic acid (top panel) and the absence of TAGs (bottom panel) in the MUT+pYes3 yeast strain. (B) HPLC-PDA analysis of yeast lipids. Lipids were extracted from yeast cells and then separated and analyzed by HPLC with PDA detection. The PDA device monitored UV light absorbance from 210 to 345 nm, which allowed the detection of eleostearoyl-containing compounds by virtue of the characteristic UV light absorbance spectrum of eleostearic acid (see Supplemental Figure 1 online for a complete PDA data set of WT+pYes3 lipids and a comparison with a tung oil standard). Each chromatogram shown represents the 271-nm absorbance trace (max of eleostearic acid) extracted from the complete PDA data set. TAGs were identified by comparison of peak retention times and retention times of authentic standards, and assignments were confirmed by liquid chromatography-mass spectrometry (LC-MS) (see Supplemental Figure 2 online). Each TAG peak is labeled according to the three fatty acyl side chains, although stereospecific positions are not known. Fatty acid abbreviations are as follows: eleostearic (E), linoleic (L), oleic (O), palmitic (P), palmitoleic (Po), and stearic (S). Each chromatogram is representative of three independent experiments.

Together, the data presented in Figure 4 indicate that tung DGAT1 and DGAT2 produce different pools of TAGs under in vivo, steady state conditions and that DGAT2 shows an increased propensity for the formation of trieleostearin, the main component of tung oil.

DGAT1 and DGAT2 Proteins Are Enriched in Subdomains of the ER
Previous in vitro experiments demonstrated that DGAT enzyme activity was associated with microsomal membrane preparations (Figure 3) (Cao and Huang, 1986; Settlage et al., 1995; Lacey and Hills, 1996). To perform a more detailed investigation of the subcellular properties of these enzymes in plant cells, we expressed myc epitope–tagged tung DGAT1 or DGAT2 in tobacco BY-2 suspension cells and then examined their immunofluorescence staining patterns by confocal laser-scanning microscopy (CLSM). The addition of the myc tag provided a means to distinguish between tung DGAT1 and DGAT2 and endogenous tobacco DGAT proteins (Bouvier-Navé et al., 2000) but likely did not interfere with their folding or function, because both myc-tagged DGAT1 and DGAT2 possessed enzyme activities similar to their native protein counterparts when expressed in yeast (see Supplemental Figure 3 online).

As shown in Figure 5 , both myc-DGAT1 and myc-DGAT2 displayed reticular immunofluorescence patterns similar to the staining pattern of endogenous ER labeled with fluor-conjugated concanavalin A (ConA). Interestingly, close inspection of these cells at higher magnification revealed that myc-DGAT1 and myc-DGAT2 were not distributed uniformly throughout the ConA-stained ER but rather were enriched in distinct regions of the ER (Figure 5, white arrowheads). Other portions of the ER, however, contained relatively low amounts of myc-tagged DGAT1 or DGAT2 protein fluorescence compared with the ConA fluorescence in these same regions (Figure 5, black arrowheads). Similar results were observed when the ER in myc-DGAT1 or myc-DGAT2 cells was labeled with 3,3'-dihexyloxacarbocyanine iodide, another commonly used stain for the general ER (Sabnis et al., 1997) (see Supplemental Figure 4 online).

View larger version (123K): [in this window] [in a new window]   Figure 5. Subcellular Localization of DGAT1 and DGAT2 in Tobacco BY-2 Cells. BY-2 cells were transformed transiently with either myc-DGAT1 or myc-DGAT2, fixed in formaldehyde at 4 h after biolistic bombardment, and then processed for immunofluorescence CLSM. Hatched boxes represent the portion of the cell shown at higher magnification in the panels below. The yellow/orange color in the merged images indicates the colocalization of expressed myc-DGAT1 or myc-DGAT2 and endogenous ER stained with ConA conjugated to Alexa 594 in the same cells; white arrowheads also indicate colocalizations. Black arrowheads indicate regions of ConA-stained ER that contain relatively low immunofluorescence attributable to myc-DGAT1 or myc-DGAT2. Bars = 10 µm.

View larger version (33K): [in this window] [in a new window]   Figure 6. FRAP Analysis of GFP-DGAT2 in BY-2 Cells. (A) Selective photobleaching of GFP-DGAT2 fluorescence in a representative transiently transformed living BY-2 cell at 4 h after biolistic bombardment. The outlined area (ovals) represents a discrete region of the ER containing GFP-DGAT2 that was photobleached and then monitored over time for the recovery of fluorescence. The time lapse is indicated in seconds at top left of each frame in the top row of images, with the last prebleached image collected represented as 8 s. The 16-s time point represents the first image collected after the photobleaching event. The other image of the postbleach recovery series (124 s) illustrates the recovery of GFP-DGAT2 fluorescence in the ER subdomain. The portion of the GFP-DGAT2–transformed cell at each time point outlined by the hatched boxes in the top row of images is shown at higher magnification in the bottom row of images; arrowheads denote the specific region of the ER containing GFP-DGAT2 that was photobleached. Note that the CLSM settings (laser power and detection gain) for imaging GFP-DGAT2 fluorescence in the BY-2 cell shown here were higher than those used for most other cells examined during FRAP experiments; lower CLSM settings in these other experiments were necessary to reduce fluorescence oversaturation. Bar = 10 µm. (B) Fluorescence intensity recovery plots of GFP-DGAT2 in an ER subdomain in either the BY-2 cell shown in (A) (closed circles) or another representative GFP-DGAT2–transformed BY-2 cell that was fixed in formaldehyde at 4 h after biolistic bombardment and then subjected to photobleaching (open circles). Both fluorescence recovery curves are expressed as fluorescence percentage over the entire time series and represent the relative fluorescence intensity of a measured GFP-DGAT2–containing ER structure that was normalized to the nonbleached fluorescence in another area of the cell. One hundred percent fluorescence indicates the normalized prebleach fluorescence intensity.

View larger version (105K): [in this window] [in a new window]   Figure 7. Subcellular Localization of Coexpressed DGAT1 and DGAT2 in BY-2 Cells. BY-2 cells were cotransformed with either epitope- or GFP-tagged versions of tung DGAT1, DGAT2, or Cb5, fixed in formaldehyde at 4 h after biolistic bombardment, and then processed for immunofluorescence CLSM. Hatched boxes represent the portion of the cell shown at higher magnification in the panels below. Note that the fluorescence patterns attributable to coexpressed myc-DGAT1 and GFP-DGAT2 in the same cell are distinctly different; compare the nonoverlapping red and green fluorescence, respectively, in the merged images. Obvious colocalizations between coexpressed HA-DGAT2 and myc-DGAT2 in the same cell are indicated with white arrowheads. Arrowheads also highlight regions of the ER where HA-DGAT2 and myc-Cb5 colocalize in the same cotransformed cell. Endogenous ER in all cotransformed cells was stained with ConA conjugated to Alexa 594; fluorescence attributable to ConA staining is pseudocolored white. Bars = 10 µm.

The N and C Termini of DGAT1 and DGAT2 Are Oriented toward the Cytosolic Side of ER Membranes
To determine the topological orientation of DGAT1 and DGAT2 proteins in ER membranes, tobacco cells expressing N- and C-terminal myc-tagged DGAT1 (myc-DGAT1 and DGAT1-myc, respectively) or DGAT2 (myc-DGAT2 and DGAT2-myc, respectively) proteins were fixed and then differentially permeabilized with either Triton X-100 or digitonin before immunofluorescence microscopy. Triton X-100 permeabilizes all cellular membranes in tobacco cells (Lee et al., 1997), allowing the detection of epitopes located within the cytosol (e.g., endogenous tubulin) (Figure 8Aa ) and within subcellular compartments (e.g., endogenous calreticulin in the ER lumen) (Figure 8Ab) of the same cells. By contrast, digitonin permeabilizes only the plasma membrane (Lee et al., 1997), permitting the detection of cytosolic tubulin (Figure 8Ac) but not ER lumenal calreticulin (Figure 8Ad). Incubation of DGAT-transformed tobacco cells with digitonin resulted in the immunodetection of myc-DGAT1 (Figure 8Ae), myc-DGAT2 (Figure 8Ag), DGAT1-myc (Figure 8Ai), and DGAT2-myc (Figure 8Ak) but not endogenous calreticulin in the corresponding same cells (Figures 8Af, 8Ah, 8Aj, and 8Al). Thus, both the N and C termini of DGAT1 and DGAT2 are located on the cytosolic side of ER membranes, consistent with a predicted even (but different) number of TMDs for these two proteins (Figure 8B; see also Figures 1A and 1B).

View larger version (86K): [in this window] [in a new window]   Figure 8. Topological Orientation of DGAT1 and DGAT2 in ER Membranes. (A) Nontransformed ([a] to [d]) and transiently transformed ([e] to [l]) BY-2 cells were fixed in formaldehyde and permeabilized using either Triton X-100 ([a] and [b]), which perforates all cellular membranes, or digitonin ([c] to [l]), which selectively permeabilizes the plasma membrane, then cells were processed for immunofluorescence microscopy. Bar in (a) = 10 µm. (a) and (b) Immunostaining attributable to endogenous tubulin in cytosolic microtubules (a) and calreticulin in the ER lumen (b) in the same nontransformed, Triton X-100–permeabilized cells. (c) and (d) Presence of cytosolic tubulin (c) but lack of immunostaining attributable to endogenous calreticulin (d) in the same digitonin-permeabilized cells. (e) to (l) Immunostaining attributable to transiently expressed myc-DGAT1 (e), myc-DGAT2 (g), DGAT1-myc (i), or DGAT2-myc (k) but absence of fluorescence attributable to endogenous calreticulin in the corresponding ([f], [h], [j], and [l]) same digitonin-permeabilized cells. (B) Predicted topological maps of DGAT1 and DGAT2. Regions in DGAT1 and DGAT2 proposed to be hydrophobic membrane-spanning domains or hydrophilic domains facing the cytosol or ER lumen were identified using the TMHMM program (version 2.0). Based on the results presented in (A), the N and C termini of DGAT1 and DGAT2 are predicted to face the cytosol.

View larger version (31K): [in this window] [in a new window]   Figure 9. Localization of Various GFP-Cf9 Fusion Proteins in Onion Epidermal Cells. (A) Deduced C-terminal amino acid sequence alignment of various DGAT1 and DGAT2 proteins from tung (Vf), Arabidopsis (At), and rice (Os). The putative aromatic amino acid–enriched ER retrieval motif in both groups of proteins is boxed. (B) Schematic of the GFP-Cf9 chimeric proteins and their corresponding intracellular localization in transformed onion epidermal cells as ER, Golgi, or plasma membrane (PM). The GFP-Cf9 chimeric protein is drawn to scale and consists of an N-terminal Arabidopsis chitinase signal peptide (black box), the GFP (gray box), and C-terminal amino acid sequences from Cf9 (white box), including the protein's single TMD (lined box) and dilysine ER retrieval motif (underlined). The C-terminal Cf9 sequences were modified by replacing the protein's dilysine motif with either the tung DGAT1 or DGAT2 C-terminal sequence, including each protein's putative pentapeptide ER retrieval motif (underlined). Modified amino acid residues in the various GFP-Cf9 or GFP-Cf9-DGAT chimeric proteins are shown in boldface. (C) Localization of various GFP-Cf9 and RFP fusion proteins in onion epidermal cells. Onion epidermal peels were either bombarded with DNA encoding different GFP-Cf9 chimeric proteins, as illustrated in (B), or cobombarded with DNA encoding a GFP-Cf9 chimeric protein and either RFP-HDEL or RFP-RhoGTPase. Cells were then incubated for 8 h in the dark and visualized using either fluorescence or bright-field microscopy. Each representative micrograph is labeled at top left with either the name of the transiently expressed GFP-Cf9 fusion protein or the name of the coexpressed RFP fusion protein. Also shown are the corresponding bright-field micrographs of cells cotransformed with either GFP-Cf9 and RFP-HDEL or GFP-Cf9KKNN and RFP-RhoGTPase (two panels at top right). Arrowheads indicate individual Golgi complexes in cells transformed with GFP-Cf9-DGAT2 or GFP-Cf9-DGAT2+13. Bar = 100 µm.

Replacement of the -KKRY motif of GFP-Cf9 with the C-terminal 15 amino acid residues of tung DGAT1 resulted in localization of the protein in the ER (Figures 9B and 9C), demonstrating that this sequence could function as an ER retrieval signal (cf. fluorescence attributable to GFP-Cf9-DGAT1 with GFP-Cf9 and RFP-HDEL). Replacement of DGAT1 residues –1 to –5 or –6 to –10 with Ala had no effect on ER localization (GFP-Cf9-DGAT1A1 and GFP-Cf9-DGAT1A2), whereas replacement of the -YYHDL- sequence disrupted ER targeting (cf. fluorescence attributable to GFP-Cf9-DGAT1A3 with GFP-Cf9KKNN and RFP-RhoGTPase) (Figures 9B and 9C). These data clearly demonstrate that the upstream -YYHDL- pentapeptide motif functions as an ER localization signal. Furthermore, the construct GFP-Cf9-YYHDL also was localized to the ER (Figures 9B and 9C), demonstrating that the DGAT1 pentapeptide sequence could also function at the extreme C terminus, similar to previously reported ER resident membrane proteins (McCartney et al., 2004).

Replacement of the -KKRY motif of GFP-Cf9 with the C-terminal 9 or 13 amino acid residues of tung DGAT2 (which included the -LKLEI- pentapeptide motif), however, resulted in localization of the reporter protein to both the ER and small punctate structures (GFP-Cf9-DGAT2 and GFP-Cf9-DGAT2+13 [Figure 9C, white arrowheads]), suggesting that the reporter protein was inefficiently retrieved from post-ER compartments. Based on their morphology, distribution, and dynamic movements, these punctate structures were assumed to be Golgi bodies, a premise that was substantiated by the colocalization of GFP-Cf9-DGAT2 and GFP-Cf9-DGAT2+13 with the coexpressed Golgi marker protein RFP-BS14a (see Supplemental Figure 7 online). GFP-Cf9 was localized entirely to the ER, however, when the fusion protein contained the C-terminal 18 amino acid residues of DGAT2 (GFP-Cf9-DGAT2+18), indicating that the context conveyed by residues upstream of the putative ER retrieval motif was essential for its proper function. As expected, replacement of the residues within the -LKLEI- sequence with Ala resulted in the protein (GFP-Cf9-DGAT2+18A) being mislocalized to the plasma membrane rather than to the ER. Together, the results presented in Figure 9 demonstrate that although DGAT1 and DGAT2 each contains a C-terminal pentapeptide ER retrieval motif, the functioning of the DGAT2 motif may be more sensitive to proper protein context for correct presentation and interaction with the protein-sorting machinery.

The DGAT1 and DGAT2 C-Terminal ER Retrieval Motifs Are Not Sufficient for the Localization of Proteins to Subdomains of the ER
To determine whether the DGAT1 and/or DGAT2 C-terminal ER retrieval motifs also were involved in the localization of the DGAT proteins to subdomains of the ER, we coexpressed each GFP-Cf9 sufficiency construct (i.e., GFP-Cf9-DGAT1 or GFP-Cf9-DGAT2+18 [Figure 9B]) with its full-length protein counterpart (i.e., myc-DGAT1 or myc-DGAT2). As shown in Figure 10 , both full-length DGAT proteins were localized to distinct portions of the ER, as expected (cf. with Figure 5), whereas the coexpressed GFP-Cf9 sufficiency constructs were localized throughout the entire ER network, as indicated by their complete colocalization with the fluorescence attributable to ConA-stained ER in the same cells (Figure 10). Collectively, these data indicate that although the C-terminal ER retrieval motifs of DGAT proteins function as general ER localization signals, additional portions of the DGAT protein sequence are required for targeting to specific subdomains of the ER.

View larger version (130K): [in this window] [in a new window]   Figure 10. Localization of Coexpressed DGAT1 or DGAT2 and GFP-Cf9 Fusion Proteins in BY-2 Cells. BY-2 cells were cotransformed with either myc-DGAT1 or myc-DGAT2 and GFP-Cf9-DGAT1 or GFP-Cf9-DGAT2+18 (see Figure 9B), fixed in formaldehyde at 4 h after biolistic bombardment, and then processed for immunofluorescence CLSM. Hatched boxes represent the portion of the cell shown at higher magnification in the panels below. Black arrowheads indicate regions of ConA-stained ER that contain GFP-Cf9-DGAT1 or GFP-Cf9-DGAT2+18 but relatively low immunofluorescence attributable to myc-DGAT1 or myc-DGAT2, respectively; compare the colocalization of GFP-Cf9-DGAT1 or GFP-Cf9-DGAT2+18 and ConA-stained ER in the same cells. The yellow/orange color in the merged images indicates colocalizations of coexpressed myc-DGAT1 or myc-DGAT2 and GFP-Cf9-DGAT1 or GFP-Cf9-DGAT2+18, respectively, in the same cells; white arrowheads indicate colocalizations. Fluorescence attributable to ConA conjugated to Alexa 594 is pseudocolored white. Bars = 10 µm.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Conversely, initial reports describing the cloning of type 1 DGAT genes indicated that they constituted small, if not single-copy, gene families in plants (Bouvier-Navé et al., 2000; Nykiforuk et al., 2002; He et al., 2004a). These findings suggested that the determination of the biochemical characteristics and physiological roles of DGAT1 would be relatively easier than that of other acyltransferases and that significant control of TAG fatty acid content would be possible through the overexpression of single DGAT1 transgenes in new host species. However, the discovery of a second type of DGAT enzyme (Lardizabal et al., 2001) complicated this scenario. Indeed, because DGAT2 enzymes are widely conserved in nature (Figure 1), they are also likely to perform one or more important metabolic functions.

Tung represents an excellent model for characterizing the functional properties of DGAT enzymes, because there are single copies of DGAT1 and DGAT2 in the tung genome (Figure 2) and tung seeds produce TAGs containing high amounts of the unusual fatty acid -eleostearic acid (Sonntag, 1979). Expression of DGATs in yeast cells followed by in vitro analysis of enzyme activity in purified microsomal membranes revealed that DGAT1 and DGAT2 showed only minor differences in substrate selectivity (Figure 3), although eleostearoyl-containing substrates were not tested in this assay system. An in vivo functional assay in yeast cells, however, demonstrated that these enzymes show clear differences in their affinity for substrates containing eleostearic acid. That is, tung DGAT1 synthesized a pool of TAGs containing eleostearic acid similar to the pool produced by yeast DGA1 (Figure 4B), an enzyme that does not normally encounter eleostearic acid during TAG formation. Tung DGAT2, however, showed a clear preference for the synthesis of trieleostearin compared with yeast DGA1 and tung DGAT1. These data, coupled with the upregulation of tung DGAT2 during the synthesis of tung seed oil (Figure 2), suggest that tung DGAT1 and DGAT2 have nonredundant functions in plant cells and that DGAT2 plays a more important role in the production of trieleostearin in developing tung seeds.

Localization of DGAT1 and DGAT2 in Distinct Subdomains of the ER and Potential Role(s) of DGATs in TAG Metabolism
Inspection of the subcellular localization of tung DGAT1 and DGAT2 in tobacco BY-2 cells revealed that these proteins are enriched in distinct subdomains of the ER (Figure 5). In addition, FRAP experiments demonstrated that GFP-DGAT2 proteins diffuse in and out of the subdomains that they occupy, and at a rate of mobility comparable to that reported for several other ER subdomain–specific membrane proteins (Nehls et al., 2000; Klopfenstein et al., 2001; Snapp et al., 2003). These latter data indicate that the DGAT-containing subdomains are dynamic regions of the ER and not stable, immobilized aggregates of (transiently overexpressed) protein. Further evidence in support of this conclusion is that Cb5, a membrane protein known to be localized throughout the ER network in plant cells (Hwang et al., 2004), is not excluded from the DGAT1- and DGAT2-containing subdomains (Figure 7), as would be expected if they consisted solely of immobilized aggregates of DGAT proteins.

It is well established that the ER is a functionally complex and dynamic organelle with at least 16 morphologically different membrane domains that carry out a variety of biochemical functions (Staehelin, 1997; Levine and Rabouille, 2005). The enrichment of DGAT proteins in distinct subdomains of the ER suggests that these regions are dedicated to TAG biosynthesis. The existence of focal points for TAG biosynthetic activity would facilitate the separation of metabolic activities that function in membrane lipid and storage oil synthesis, as has been suggested previously (Fernandez and Staehelin, 1987; Vogel and Browse, 1996). However, perhaps the most surprising observation from our microscopic analysis of DGAT1 and DGAT2 is that these proteins do not localize to the same ER subdomains but rather are enriched in different locations within the ER (Figure 7). These data, coupled with data regarding the differential expression of DGAT1 and DGAT2 transcripts (Figure 2C) and biochemical differences in their enzymatic activity (Figures 3 and 4), suggest that these two enzymes produce different pools of TAGs within plant cells. Our results do not clarify, however, whether DGAT1 and DGAT2 are actually present in the same cells in developing tung seeds, leaves, or flowers or whether DGAT expression is restricted to different tissue types within these organs. These questions will need to be addressed using monospecific DGAT1 and DGAT2 antibodies and immunomicroscopic analysis of native tung tissues. Regardless, it is tempting to speculate that DGAT1 plays a more general role in TAG metabolism in plant cells, whereas DGAT2 has primarily evolved to accommodate the processing and accumulation of unusual fatty acids. In support of this premise, DGAT1 has been shown to play a role in the production of TAGs during cellular senescence in plants (Kaup et al., 2002), and additional studies in mammals have begun to highlight the existence of a dynamic pool of TAGs in cells associated with the temporary storage of fatty acids rather than the long-term storage of TAGs as a carbon and energy reserve (Murphy, 2001; Liu et al., 2004; Smirnova et al., 2006).

The roles of DGAT1 and DGAT2 in oil production, however, do not have to be mutually exclusive. In some plants, DGAT1 may play a more dominant role depending on gene expression patterns and protein accumulation, whereas in plants containing unusual fatty acids, DGAT2 may play a more important role. Nonetheless, the enrichment of enzymes for unusual fatty acid metabolism into distinct regions of the ER would provide an effective mechanism for excluding the unusual fatty acids from membrane lipids, which otherwise could be disruptive to the structure and/or functioning of the membrane lipid bilayer, and for channeling unusual fatty acids into storage oils.

Biogenesis of DGAT ER Subdomains and Implications for Oilseed Engineering
A major goal is to now determine how the ER subdomains containing DGAT enzymes are formed and what their roles are in storage oil biosynthesis. Toward this end, we investigated the intracellular sorting pathways by which the DGATs are targeted and retained within the ER. Each of the DGAT enzymes is predicted to contain multiple TMDs (Figures 1A, 1B, and 8B); therefore, these proteins are likely inserted cotranslationally into ER membranes by an N-terminal signal anchor (TMD) sequence and the signal recognition particle/Sec61p-dependent pathway (reviewed in Shan and Walter, 2005). Once in the ER, the DGATs must be retained there by either static- or retrieval-mediated mechanisms, and they may contain additional signals for sorting to their respective subdomains. Characterization of the topological orientation of DGAT1 and DGAT2 revealed that the N and C termini of both proteins are exposed on the cytosolic side of ER membranes (Figure 8), and targeting assays with a GFP-Cf9 reporter protein in onion epidermal cells revealed that both DGATs possess at their C terminus a pentapeptide ER retrieval motif (Figure 9) similar to that recently identified in FAD2 enzymes (McCartney et al., 2004). Notably, although the C-terminal ER retrieval motifs of DGAT1 and DGAT2 were sufficient for localizing GFP-Cf9 to the ER (Figure 9), neither signal alone was sufficient for localizing GFP-Cf9 to the subdomains of the ER that contain the full-length DGAT proteins (Figure 10). These data indicate that although the DGAT C-terminal ER retrieval signals function in steady state localization in the ER, additional portions of the DGAT proteins are required for targeting to their respective ER subdomains.

There are at least two mechanisms by which additional portions of DGAT protein sequence could contribute to the localization of the enzymes in specific ER subdomains. One possibility is that DGAT enzyme activity itself generates a local pool of TAGs and that DGAT enzymes have higher affinity for TAG-enriched regions of the ER membrane compared with the surrounding phospholipid bilayer. These local pools of lipids and DGAT proteins could serve as nucleating centers to attract additional enzymes and proteins (e.g., oleosins) involved in TAG metabolism. Although we did not test directly the localization of inactivated forms of the tung DGAT proteins, we did investigate the functional and cellular properties of the Arabidopsis DGAT1 and DGAT2 proteins as well as a modified version of tung DGAT1, specifically a chimera of tung DGAT1 and GFP (GFP-DGAT1). Arabidopsis DGAT1 was entirely functional in our yeast complementation assay and, similar to tung DGAT1, was localized to distinct regions of the ER in suspension-cultured plant cells (see Supplemental Figure 8 online). Arabidopsis DGAT2, on the other hand, showed no detectable activity in yeast complementation assays (similar to the findings of Lardizabal et al. [2001]) and was not localized to any ER subdomains but rather was found throughout the entire ER network in plant cells (see Supplemental Figure 8 online). These data suggest that the activity of DGAT enzymes is essential for their nucleation and/or targeting to specific subdomains of the ER. However, DGAT enzyme activity alone is not sufficient for targeting to ER subdomains, because GFP-tagged tung DGAT1 (GFP-DGAT1) possessed similar enzymatic activity to that of myc-DGAT1 (see Supplemental Figure 3 online) but was not enriched in any distinct regions of the ER (data not shown). Based on these results, we hypothesize that the addition of GFP to the N terminus of DGAT1 prevents its ability to form homooligomers, which, as discussed below, may be important for its proper localization within the ER.

Another mechanism by which additional portions of DGAT protein sequence contribute to their localization in certain regions of the ER, in addition to a prerequisite for enzyme activity, is through homotypic oligomerization. For instance, there is evidence that the retention of certain proteins within the Golgi occurs by the formation of homooligomers, which prevents further transport to other compartments within the endomembrane system (Nilsson et al., 1994). Similar mechanisms of homotypic oligomerization have been proposed to promote the localization of certain membrane proteins in the ER (Ivessa et al., 1992). Although the oligomerization status of plant DGATs is currently unknown, human DGAT1 forms homotetramers (Cheng et al., 2001). Thus, it is possible that the oligomerization of plant DGAT enzymes could nucleate the formation of dynamic, functional centers within the ER dedicated to TAG biosynthesis. The recruitment of additional interacting proteins, such as other acyltransferases and/or oleosins, could provide additional machinery for the specialized production and packaging of lipids destined for storage oils. Moreover, the inability of DGAT1 and DGAT2 to interact and form heterooligomers would provide a simple mechanism for the spatial segregation of these enzymes (and their putative associated protein complexes) in ER membranes. We are currently testing this hypothesis by investigating the ability of DGAT1 and DGAT2 to interact through yeast two-hybrid split-ubiquitin screening (Thaminy et al., 2004). Notably, recent characterization of an Arabidopsis LPAT enzyme revealed that this protein is also enriched in certain regions of the ER (Kim et al., 2005), and we are also determining whether these regions are equivalent to those containing DGAT1 and/or DGAT2.

Overall, the functional and cellular aspects of DGATs presented here have at least two important implications for the genetic engineering of crop plants that are tailored for the production of industrially important oils. First, it is clear from these studies that coexpression of DGAT2 and an enzyme responsible for the synthesis of an unusual fatty acid from the same species is likely to lead to an enhanced accumulation of this fatty acid in a transgenic host. This hypothesis is supported by the observation that co-overexpression of castor bean DGAT2, but not Arabidopsis DGAT2, with castor oleate hydroxylase (van de Loo et al., 1995) leads to substantial increases in hydroxy fatty acid content in Arabidopsis seed lipids (i.e., a 50 to 70% increased overexpression of oleate hydroxylase alone) (J.M. Shockey and J. Browse, unpublished data). Second, the optimal production of storage oils containing unusual fatty acids may be dependent on the ability of the expressed proteins to establish functional ER subdomains that are dedicated to the production of specific types of TAGs. The molecular mechanisms underlying the formation of these subdomains and their protein composition are the subject of future investigation.

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Recombinant DNA Procedures and Reagents
Standard recombinant DNA procedures were performed as described by Sambrook et al. (1989). Molecular biology reagents were purchased from New England Biolabs, Promega, Perkin-Elmer, Stratagene, or Invitrogen. Oligonucleotides were synthesized by either University of Guelph Laboratory Services or Invitrogen. DNA was isolated and purified using reagents from Qiagen and Promega. All DNA constructs were sequenced using dye-terminated cycle sequencing to verify that undesirable mutations were not introduced during amplification or cloning procedures. Mutagenesis was performed using appropriate complementary forward and reverse mutagenic primers and the QuikChange site-directed mutagenesis kit according to the manufacturer's instructions (Stratagene). Complete details of the oligonucleotide primers used in the gene cloning and plasmid constructions described below are available upon request.

Several conserved regions of the DGAT families identified previously served as starting points for the identification of tung (Vernicia fordii) DGAT1 and DGAT2 genes (Hobbs et al., 1999; Lardizabal et al., 2001). Sense-strand degenerate oligonucleotide primers were designed to the amino acid sequences -QSHAGLFNL- and -GYEPHSVLP-, which span residues 130 to 138 and 106 to 114 of Arabidopsis thaliana DGAT1 and DGAT2, respectively. Paired with a capped oligo(dT) primer, PCRs with tung seed cDNA library (Dyer et al., 2002a) DNA preparations yielded amplicons corresponding to the 3' ends of both DGAT cDNAs. Attempts at isolation of the remaining 5' cDNA ends from the seed cDNA library were unsuccessful. The 5' sequence of tung DGAT1 was determined through multiple rounds of adaptor-anchored PCR from tung genomic DNA (Universal GenomeWalker kit; Clontech). Genomic sequences were compared with the cDNA sequence of Arabidopsis DGAT1 until a putative start Met ATG codon was found. Genomic PCR and RT-PCR (Superscript III first-strand cDNA synthesis kit; Invitrogen) results were compared to determine the 5' untranslated region for DGAT1. The complete sequence for tung DGAT2 was determined by 5' rapid amplification of cDNA ends (5' RACE system; Invitrogen) using capped oligo(dT)-primed total RNA isolated from developing tung seeds. Finally, PCR products encompassing the full-length open reading frames (ORFs) were amplified, digested, and cloned into the pYes3-CT vector (Invitrogen) for expression in Saccharomyces cerevisiae.

Yeast and plant expression plasmids containing the tung DGAT1 or DGAT2 ORFs were constructed as follows. First, the tung DGAT1 and DGAT2 ORFs were amplified from pYes3-CT/DGAT1 and pYes3-CT/DGAT2 using PCR with appropriate forward and reverse primers that introduced in-frame XbaI or BamHI sites 5' and 3' of the start and stop codons of DGAT1 and DGAT2, respectively. Next, PCR products were gel-purified and subcloned into either XbaI- or BamHI-digested pRTL2/mycX, a modified version of the plant expression vector pRTL2N/S (Lee et al., 1997) that includes the 35S cauliflower mosaic virus promoter and sequences encoding an initiator Met, Gly linkers, and the myc epitope tag (underlined: MGEQKLISEEDLG-) (Fritze and Anderson, 2000) followed by in-frame XbaI and BamHI sites. The resulting plasmids were referred to as pRTL2/myc-DGAT1 and pRTL2/myc-DGAT2. pRTL2/GFP-DGAT2 was generated by ligating the BamHI fragment from pRTL2/myc-DGAT2 into BamHI-digested pRTL2/monoGFP-MCS, a plasmid containing a modified version of the GFP, in which the Leu at position 221 in the GFP ORF was mutated (via site-directed mutagenesis) to Lys. This modification to GFP disrupts the protein's dimerization domain (Zacharias et al., 2002) and helps to prevent alterations in the target membrane (e.g., aggregation) attributable to transient (over)expression of the membrane-bound GFP fusion protein (Lisenbee et al., 2003). pRTL2-HA-DGAT2 was constructed by replacing in pRTL2/myc-DGAT2 (via site-directed mutagenesis) sequences encoding the myc epitope tag with a HA epitope tag (underlined: MGYPYDVPDYAG-) (Fritze and Anderson, 2000). To construct pRTL2/DGAT1-myc and pRTL2/DGAT2-myc, the DGAT1 and DGAT2 ORFs were amplified using PCR and pRTL2/myc-DGAT1 and pRTL2/myc-DGAT2, respectively, as template DNA. Forward and reverse primers used in these PCRs introduced an in-frame XbaI site 5' of the DGAT1 or DGAT2 start codon and a NcoI site 3' of each ORF. The resulting PCR products were digested with XbaI and NcoI and ligated into XbaI-NcoI–digested pRTL2/X-myc, a modified version of pRTL2N/S with sequences encoding XbaI, NheI, and NcoI sites followed by a myc epitope tag and a stop codon. Construction of pRTL2/myc-Cb5, encoding an N-terminal myc-tagged version of the A isoform of tung Cb5, has been described (Hwang et al., 2004).

The yeast expression plasmid containing Arabidopsis DGAT1 (AtDGAT1) was generated as follows. An amplicon representing the At DGAT1 ORF was generated using gene-specific primers and Pfu Ultra (Stratagene). The template for this PCR was first-strand cDNA generated from developing Arabidopsis silique total RNA using Superscript III reverse transcriptase (Invitrogen). PCR products were T-tailed and cloned into the yeast expression plasmid pYes2.1 TOPO as described by the manufacturer (Invitrogen). pYes2.1/AtDGAT2 was generated by amplification of the Arabidopsis DGAT2 (AtDGAT2) ORF from ABRC clone U16101 (Newman et al., 1994), followed by T-tailing and cloning into pYes2.1 TOPO, as described above. Plant expression plasmids containing myc epitope–tagged versions of Arabidopsis DGAT1 or DGAT2 were constructed as follows. First, the Arabidopsis DGAT1 and DGAT2 ORFs were amplified from pYes2.1/AtDGAT1 and pYes2.1/AtDGAT2 using PCR with appropriate forward and reverse primers that introduced in-frame XbaI or BamHI sites 5' and 3' of the start and stop codons of At DGAT1 and At DGAT2, respectively. Next, PCR products were gel-purified, subcloned into pCR2.1 TOPO, and then cloned into either XbaI- or BamHI-digested pRTL2/mycX.

Plasmids coding for GFP-Cf9 mutants with altered amino acid residues at the C terminus of Cf9 (Figure 9) were constructed using site-directed mutagenesis and pRTL2/GFP-Cf9 or pRTL2/GFP-Cf9-KKNN (McCartney et al., 2004) as template DNA. pRTL2/RFP-HDEL encoding the RFP fused to the N-terminal Arabidopsis chitinase signal sequence and the C-terminal HDEL retrieval signal was constructed by amplifying (via PCR) sequences coding the chitinase signal sequence from pBIN/mGFP5-HDEL (provided by Jim Haseloff, University of Cambridge) (Haseloff et al., 1997