• Open Access

Reducing saturated fatty acids in Arabidopsis seeds by expression of a Caenorhabditis elegans 16:0–specific desaturase

Authors


Correspondence (Tel +(509) 335 2293; fax +509 335 7643; email jab@wsu.edu)

Summary

Plant oilseeds are a major source of nutritional oils. Their fatty acid composition, especially the proportion of saturated and unsaturated fatty acids, has important effects on human health. Because intake of saturated fats is correlated with the incidence of cardiovascular disease and diabetes, a goal of metabolic engineering is to develop oils low in saturated fatty acids. Palmitic acid (16:0) is the most abundant saturated fatty acid in the seeds of many oilseed crops and in Arabidopsis thaliana. We expressed FAT–5, a membrane-bound desaturase cloned from Caenorhabditis elegans, in Arabidopsis using a strong seed-specific promoter. The FAT-5 enzyme is highly specific to 16:0 as substrate, converting it to 16:1∆9; expression of fat-5 reduced the 16:0 content of the seed by two-thirds. Decreased 16:0 and elevated 16:1 levels were evident both in the storage and membrane lipids of seeds. Regiochemical analysis of phosphatidylcholine showed that 16:1 was distributed at both positions on the glycerolipid backbone, unlike 16:0, which is predominately found at the sn-1 position. Seeds from a plant line homozygous for FAT–5 expression were comparable to wild type with respect to seed set and germination, while oil content and weight were somewhat reduced. These experiments demonstrate that targeted heterologous expression of a desaturase in oilseeds can reduce the level of saturated fatty acids in the oil, significantly improving its nutritional value.

Introduction

Plant seed oils are important sources both of high-calorie food and of fatty acids that are essential for human nutrition. The diversity of fatty acids stored in oilseeds worldwide exceeds 200 different types, varying in acyl chain length, degree of unsaturation and functional groups (Badami and Patil, 1980). Despite this diversity, the most nutritionally important and abundant fatty acids found in the major commercial food oils are of just a few types: palmitate, stearate, oleate, linoleate and linolenate. In seeds, the majority of these fatty acids are esterified to glycerol backbones to form triacylglycerols (TAG); these storage lipids accumulate in specialized lipid bodies with limited metabolic activity (Millar et al., 2000). These same few fatty acids also predominate in the membrane lipids of all plant tissues, where their essential roles in metabolism are well established (Van Meer et al., 2008).

Worldwide production of edible fats and oils is dominated by vegetable oils, accounting for nearly 85% of the total (Broun et al., 1999). Consumption of fats and oils has important effects on human health (Ramsden et al., 2009) that depend largely on their fatty acid composition (Wahrburg, 2004). Of the five most abundant fatty acids in plants, only palmitate (16:0) and stearate (18:0) are saturated hydrocarbon chains with no double bond between carbon atoms. A diet high in saturated fats raises the risk of cardiovascular disease (Tanaka et al., 2008), and the incidence of both type 2 diabetes mellitus and insulin resistance is increased by consuming high levels of saturated fatty acids (Lopez et al., 2010; Vessby et al., 2001). Because 16:0 (and shorter chain fatty acids) has detrimental effects on plasma cholesterol concentrations, dietary guidelines recommend replacing palmitate with unsaturated fats to guard against cardiovascular disease and diabetes (Dietschy, 1998).

Palmitate (16:0) is the most abundant saturated fatty acid in the seeds of canola (4% of total fatty acids), soybean (11%), and sunflower (7%) (Liu et al., 2002) and also in the model oilseed plant, Arabidopsis, where 16:0, 18:0 and 20:0 occur in a ratio of 6 : 2 : 1. In Arabidopsis, both 16:0 and 18:0 fatty acids are synthesized as ACP thioesters in the plastid. While 18:0 is efficiently desaturated to 18:1 before export from the plastid and can undergo additional desaturation in the endoplasmic reticulum (ER), saturated fatty acids exported from the chloroplast as CoA esters are largely incorporated into lipids without undergoing desaturation (Baud and Lepiniec, 2010).

Diverse strategies have been employed to reduce saturates in oilseed plants. To desaturate 16:0-ACP before export from the plastid, the gene for a 16:0-ACP desaturase isolated from cat's claw (Doxantha unguis-cati) was expressed in Arabidopsis (Bondaruk et al., 2007). The proportion of 16:1 Δ9 increased but the total saturated fatty acid content was unchanged. Another approach, expressing an ACP desaturase from castor bean that had been engineered to be a 16:0-specific desaturase, produced a minor reduction in 16:0, with 88% of the wild type 16:0 remaining in the seed (Nguyen et al., 2010). T-DNA disruption of the fatty acyl acyl-carrier-protein (FATB) thioesterase, which has high activity on saturated acyl-ACP substrates, reduced the seed 16:0 level to 48% of wild-type plants (Bonaventure et al., 2003); however, these fatb-ko plants had severe defects in morphology and reproduction, suggesting that a certain level of saturates may be critical for plant growth and seed development.

Metabolic processes associated with oil production, including elongation and desaturation of fatty acids, TAG synthesis and oil body formation are localized to the ER (Frandsen et al., 2001). Introducing a desaturase enzyme active in the ER, at the site of TAG synthesis, should minimize potential negative effects of lowered 16:0 (Bonaventure et al., 2003), because TAG in oil bodies is largely sequestered from metabolic activity (Millar et al., 2000). Also, heterologous expression using a seed-specific promoter confines changes in fatty acid composition to the seed. Constitutive expression of acyl CoA Δ9 desaturases in Nicotiana spp. produced little reduction in seed 16:0 levels; plants transgenic for the yeast OLE1 Δ9 desaturase retained 90% of the saturates (16:0 + 18:0) found in wild-type seed (Polashock et al., 1992); similarly, plants expressing the rat stearoyl-CoA desaturase still contained 95% of seed saturated fatty acids (Moon et al., 2000). Overexpression of the Arabidopsis ADS1 desaturase in Brassica juncea using a seed-specific napin promoter was slightly more successful, but transgenic plants still retained 83% of the parental level of saturated fatty acids (Yao et al., 2003).

The fat-5 gene of Caenorhabditis elegans encodes a desaturase enzyme, FAT-5, previously characterized in our laboratory (Watts and Browse, 2000). Like other eukaryotic membrane-bound desaturases, FAT-5 has four stretches of hydrophobic amino acid residues, indicative of transmembrane domains, and three histidine-rich regions that are critical to coordination of the iron atoms involved in oxidative desaturation at the active site (Shanklin et al., 1994). Based on its protein sequence similarity to other desaturases, FAT-5 likely acts on fatty acids esterified either to Coenzyme A or to glycerol molecules of lipids. While its exact substrate is unknown, FAT-5 almost exclusively desaturates saturated 16:0 fatty acid chains, introducing a double bond at the ninth carbon from the carboxyl group. Unlike most other eukaryotic 16:0 fatty acid desaturases that have been studied, FAT-5 has no activity on 18-carbon fatty acid chains in C. elegans (Brock et al., 2007) and very low 18:0 desaturation when heterologously expressed in Saccharomyces cerevisiae (Watts and Browse, 2000).

We report here that expression in Arabidopsis seeds of the FAT-5 enzyme led to a substantial reduction in saturated fatty acids; seed oil from our best plant line retained only 2.8% 16:0 fatty acid, one-third of the wild-type level. Although both storage and membrane lipids were markedly reduced in 16:0, the seeds were normal with respect to number, morphology and germination, although slightly reduced in oil content. These results demonstrate an approach that may help provide plant-derived food oils that are very low in saturated fats, providing opportunities to improve human health.

Results

Napin-directed expression of the FAT–5 desaturase in Arabidopsis seeds

In a previous study of fatty acid desaturase genes in C. elegans, we identified a cDNA encoding FAT-5, a desaturase primarily active on unsaturated 16-carbon fatty acids (Brock et al., 2007; Watts and Browse, 2000). To test its effectiveness in oilseeds, the fat-5 open reading frame was cloned into a plant transformation vector under the control of the seed-specific napin promoter (Ellerstrom et al., 1996). After Agrobacterium–mediated transformation of Arabidopsis, seed samples from 200 independent T1 transformed lines were analysed by fatty acid methyl ester derivatization and gas chromatography. Lines whose segregation indicated a single insertion were propagated and homozygous lines selected for further analysis. Levels of 16:1 desaturation product varied from 0.5% of total fatty acids, comparable to the wild-type parent, to as high as 2.4%. Lines with high 16:1 were reduced in 16:0 from the 8.6% of the total, as measured in wild-type Arabidopsis, but even the best lines retained 83% of the parental 16:0 (Figure 1). Apart from the change in fatty acid composition, the seeds of these plants were indistinguishable from the wild type in size, oil content and germination success.

Figure 1.

Napin expression of FAT-5. Proportions of 16:0 and 16:1 fatty acids in seeds of a wild-type Arabidopsis (WT) compared with a plant line expressing FAT-5 under napin promoter control. Values are mean ± SD (n = 10).

Screening for improved FAT–5 activity in yeast

Although napin-driven FAT–5 expression reduced the proportion of 16:0 and increased 16:1 in the seed, we sought a more substantial reduction in saturates by engineered alterations in fat-5. To facilitate testing of these modified genes, we carried out a functional screen for activity in yeast, using the S. cerevisiae mutant strain ole1, which lacks fatty acid desaturase activity (Stukey et al., 1989). One approach was to optimize both the codon usage and G+C content of the fat-5 open reading frame without changing the peptide sequence (Figure S1). However, the modified version of FAT-5 exhibited reduced activity when expressed in yeast or in Arabidopsis seed; it had only 90% of the activity of the original fat-5 sequence under all conditions tested.

Fatty acid desaturation and TAG synthesis occur at the ER (Thelen and Ohlrogge, 2002), Although the FAT–5 predicted protein sequence contains hallmarks of ER integral membrane desaturases (Watts and Browse, 2000), computer analysis of the predicted protein sequence with SignalP (Dyrløv Bendtsen et al., 2004) and WolfPSort (Horton et al., 2007) failed to discover either an ER-targeting sequence or ER retrieval and retention sequence. We attempted to optimize ER targeting and retention of the protein by changing the fat-5 open reading frame, adding nucleotide sequences to generate a number of fat-5 variants with fused N-terminal signal sequences or C-terminal ER retention domains; we expressed these in yeast using vector pMK195 (Overvoorde et al., 1996).

Following transformation of yeast mutant ole1, we selected for transformants expressing desaturase activity on medium lacking unsaturated fatty acids. Expression of FAT-5 with the native sequence converted 65% of the 16:0 naturally present in yeast to 16:1; we scored activities for altered protein sequences as percentages of this unmodified FAT-5 sequence. In-frame fusion of the tobacco PR1a signal peptide (Cornelissen et al., 1987) to the N-terminus of FAT–5 reduced activity to 61% of that of the unmodified enzyme. For ER retention domains, we tested the C-terminal -KSKIN and –YNNKL motifs of the Arabidopsis FAD3 and FAD2 desaturases, respectively (McCartney et al., 2004), as well as a consensus –KKSS motif for yeast ER proteins (Pelham, 2000). Fusion of ER retention domains to the protein C-terminus reduced FAT–5 activity to 79% of the native protein activity for –KSKIN, to 47% for -YNNKL, and to only 40% for –KKSS. When we transformed Arabidopsis with these constructs under control of a seed-specific promoter, desaturase activity in seed of T1 plants was less in all cases than the activity of the unmodified fat-5 sequence. We also tested constructs that replaced the last six amino acid residues of FAT–5 with ER retention domains, rather than simply adding the domain to the C-terminus. Unexpectedly, replacing the last six C-terminal amino acids with any one of the ER retention signals completely abolished FAT–5 activity in yeast.

We confirmed that the unmodified FAT–5 protein was active in yeast microsomes by conducting a desaturase assay. We isolated a microsome fraction by differential centrifugation from ole1 yeast expressing FAT-5 and measured activity by supplying 14C-labelled palmitoyl-CoA. Following saponification, fractions enriched for saturated and unsaturated fatty acids were separated by AgNO3 thin layer chromatography (TLC). Desaturase activity converted the supplied 16:0 to a 16:1 product, while neither boiled microsome preparations nor reactions without NADH, the electron source for most eukaryotic ER-type fatty acid desaturases, had detectable activity (Figure 2). Eukaryotic desaturases rely on a NADH–cytochrome b5 reductase–cytochrome b5 electron transport chain for reducing equivalents (Shanklin et al., 1994).

Figure 2.

Functional expression of FAT-5 in yeast microsomes. Phosphorimager of a thin layer chromatography plate showing 14C labelled 16:0 and 16:1. The products from 10, 30 and 60 min incubations of microsomes with substrate are shown. Microsomes were boiled before incubation for the ‘Boil’ lane, and NADH was omitted from the ‘No NADH’ lane.

Expression of fat-5 with the phaseolin promoter

Although the napin promoter from Brassica napus is well-characterized and effective (Broun and Somerville, 1997; Robert et al., 2005), we achieved only a modest reduction in 16:0 with napin-expressed FAT-5. We tested FAT-5 expression with the Arabidopsis seed albumin 2S2 promoter (Conceicao and Krebbers, 1994), orthologous to the napin promoter. The 16:0 proportion of total seed fatty acids in wild type was reduced from 8.6% to 5.8% or 66% of the wild-type level. We tested the Lesquerella fendleri fatty acid hydroxylase promoter, because it is specific to seed fatty acid modification (Broun et al., 1998) but 61% of wild-type level of 16:0 remained. Finally, we tested the phaseolin promoter from Phaseolus vulgaris (Sengupta-Gopalan et al., 1985), reported to be an especially strong seed-specific promoter in Arabidopsis (De Jaeger et al., 2002). After transformation of Arabidopsis with a phaseolin-driven fat-5 construct, we screened seed of 72 T1 plants for increased 16:1 and reduced 16:0. Levels of 16:0 below 5% of the total seed fatty acids were observed for several lines (Figure 3); when these lines were propagated, homozygous plants evinced further reductions in 16:0. A plant line whose parent co-segregated herbicide resistance with low 16:0 fatty acids at ratios indicating a single insertion site was chosen for detailed analysis. This line, named FAT-5 #43-4, with 2.8% of total seed oil as 16:0, was stably maintained in excess of five generations.

Figure 3.

Bulk seed analysis of phaseolin-FAT-5 lines. Seed of 72 primary transformant (T1) plants was analysed; the proportion of total fatty acids that was 16:0 (●) or 16:1(○) varies from equivalent to wild type (on the left) to highly modified (on the right). The arrow indicates line 43, which was propagated for further analysis.

Characterization of FAT-5-expressing seeds

As Arabidopsis with reduced levels of 16:0 have exhibited defects in seed lipid content, plant morphology and germination success (Bonaventure et al., 2003), we examined line FAT-5 #43–4 for similar effects. Homozygous plant lines expressing fat-5 were indistinguishable from their wild-type parent with respect to time of emergence, rosette size and bolting time. While the number of seeds per silique was equivalent to wild type (Table 1), the weight of individual transgenic seeds was 92% of wild type, and the amount of oil in this transgenic line was reduced to about 85% of the wild type. The success of seed germination on soil or on defined media was >90%, comparable to wild-type seed.

Table 1. Seed properties of wild-type and FAT-5 transgenic plants
CharacteristicPlant line
Wild typeFAT–5 line 43-4
  1. All values are means ± SD (n = 10).

Number of seeds per silique49.7 (±6.0)49.8 (±4.1)
Dry weight (μg/seed)22.7 (±0.6)21.1 (±0.5)
Oil (μg/mg seed)363.6 (±11.6)311.5 (±17.2)

Lipid analysis of seeds expressing FAT-5

The bulk of fatty acids in Arabidopsis seeds are in neutral storage TAG, while the metabolic activity of the seeds is more dependent on the much smaller polar membrane lipid fraction (Browse and Somerville, 1994). To compare changes in 16:0 and 16:1 between storage and membrane lipids in FAT-5 transgenics, we extracted total lipids from seeds, separated neutral and polar fractions and then separated major polar lipid species by TLC. In this experiment, the 16:0 portion extracted from whole seeds was measured as 3.0% of the total fatty acids compared to 8.6% in untransformed wild type (Figure 4a). In the FAT-5 transgenic seed, the level of 18:1 increased to 18.5%, while the sum of all saturates was reduced from 12.9% in wild type to only 6.7% in transgenic seed. The increase in 18:1 levels might be due to an increase in 18:1Δ11, an elongation product of 16:1Δ9 or an increase in 18:1Δ9 from desaturation of 18:0. GC/MS analysis distinguished the 18:1 isomers by chromatography retention time and spectral comparison with authentic standards. The 18:1Δ11 isomer was 1.2% of total fatty acids in wild-type seed, in agreement with the published value of 1.3% (Kachroo et al., 2007); it increased to 2.5% in FAT-5 seed. Thus, as in wild-type seeds, most of the 18:1 in FAT-5 seeds is 18:1Δ9 rather than the elongation product of 16:1Δ9. Analysis of the neutral lipid fraction of FAT-5 seeds was very similar to the total seed analysis; 16:0 made up 3.2% of total fatty acids in the neutral lipid fraction, while 16:1 was 5.2% of that fraction (Compare Figure 4a,b).

Figure 4.

Modification of fatty acid composition of seeds and seed lipid fractions by expression of FAT-5. Lipids were extracted from seeds of phaseolin-driven FAT-5 and wild-type plants and separated by column and thin layer chromatography to generate neutral and polar lipid fractions. a, whole seeds; b, neutral lipids; c, polar lipids; d, phosphatidylcholine. Data are mean ± SD (n = 3), except a, analysed once in this experiment.

Interestingly, the changes in fatty acid composition were greater in the polar lipid fraction of FAT-5 seeds; 16:0 in the polar fraction was only 2.7% of the total, reduced from the 16.2% found in wild type, while 16:1 was elevated to 11%, a tenfold increase over the 1.1% of wild-type seed (Figure 4c). We further analysed phosphatidylcholine (PC), the most abundant polar lipid (Browse and Somerville, 1994), and found that 16:0 was only 0.8% of the total fatty acids in that lipid, compared with wild-type levels of 10.5%. The 16:1 increased to 14%, from the 2.4% found in wild-type PC (Figure 4d). There were also changes in the 18-carbon and 20-carbon fatty acid composition of whole seed and neutral lipids; 18:1 increased with a concomitant decrease in 18:2, more evident in the polar lipid fraction and especially in PC (Figure 4).

Positional analysis of 16:0 in FAT-5 lipids

As PC is an important intermediate in TAG synthesis (Bates and Browse, 2011), we examined the positional distribution of fatty acids on this membrane lipid in transgenic seeds. To determine the regiospecificity of 16:0 and 16:1 in FAT–5 transgenic plants, we isolated seed PC and digested it with phospholipase A2, an enzyme that preferentially cleaves fatty acids from the sn–2 position of PC, releasing a free fatty acid and a lyso-PC molecule retaining the sn-1 acyl chain. This positional analysis shows a tenfold decrease in 16:0 at sn-1 in PC from FAT-5 seed compared with wild type. The distribution of the 16:1 in PC from FAT-5 seed is also notable; the desaturation product occupies both the sn-1 and sn-2 positions, with a slightly greater proportion of the 16:1 at sn-2 (Figure 5). This is consistent with the general observation that saturated fatty acids are predominantly found at the sn-1 position in phospholipids, whereas unsaturated fatty acids can occupy either the sn-1 or sn-2 position.

Figure 5.

Regiospecific analysis of phosphatidylcholine (PC). Mol% of fatty acids at sn-1 and sn-2 positions of PC in seeds of wild-type (WT) and FAT-5 plants. Values are one of duplicate experiments with equivalent results.

Reduction of 16:0 during seed development

Whole-plant reduction of 16:0 to very low levels has been shown to be detrimental to Arabidopsis development (Bonaventure et al., 2003). To examine if the reduction in 16:0, which might compromise membrane physiology and cell biology, occurred during development, we dissected developing seeds from siliques of wild-type and FAT-5 plants at two developmental stages, 9–10 and 12–13 days after flowering (daf) and then analysed their fatty acid and lipid compositions. When we express the values for 16:0 and 16:1 as their proportion of total 16-carbon fatty acids (Figure 6), it is clear that some desaturation by FAT–5 had already commenced at the earliest developmental stage analysed, 9–10 daf. In wild-type seed at this stage, 97% of the 16-carbon fatty acids was 16:0, while in FAT-5 seeds desaturation had reduced 16:0 to just 63% of total 16-carbon fatty acids. In FAT-5 seed, the 16:0 proportion decreases as the seeds mature, declining to 33% in mature seed, while in wild-type seed, the ratio of 16:0 is constant. The polar lipids are similarly affected during the developmental stages. At both 9–10 and 12–13 daf, the most abundant polar lipid, PC, contains ratios of 16:0 and 16:1 very similar to the neutral lipid fraction (Compare Figure 6a,b); however, the proportions are very different when seeds reach maturity. In PC of mature seed, the 16:0 contribution to 16-carbon fatty acids is reduced to about 6%, while the total lipids still retain 33% of their 16-carbon fatty acids as 16:0 (Figure 6).

Figure 6.

Fatty acid composition of developing seeds. Relative amounts of 16:0 and 16:1Δ9 as a percentage of the total 16-carbon fatty acids in lipid fractions from developing seeds harvested from wild-type and FAT-5 plants 9–10 days after flowering (daf), 12–13 daf and from mature seeds. (a) total lipids; (b) PC. Values are the mean of three replicates.

Discussion

Saturated fatty acids, especially unsaturated 16:0, may have significant detrimental effects on human health (Wahrburg, 2004). Replacing saturated fatty acids with poly- or mono-unsaturated fatty acids reduces not only plasma cholesterol levels (Dietschy, 1998) but insulin sensitivity in type 2 diabetes patients (Garg, 1998). Reducing saturates and generating oil enriched in desirable unsaturates has long been a goal in oilseed engineering. In this study, our objective was to reduce saturated fatty acids in the seed of Arabidopsis, a model oilseed plant, through seed-specific expression of a desaturase. We focused on palmitate (16:0), the most abundant saturated fatty acid in the oil of Arabidopsis, and generated plants expressing FAT-5, a C. elegans desaturase whose principal activity is on unsaturated 16-carbon fatty acids. Using a strong seed-specific promoter, we reduced the 16:0 proportion of Arabidopsis seed oil to about one-third the level of wild-type seed. Reduction of 16:0 begins early in seed development and is maximal in the mature seed, with no loss of seed viability and slight reduction of the total oil content.

Our ability to measure activity of FAT-5 in the desaturase deficient yeast strain, ole1, made it feasible to alter the expressed protein in an attempt to optimize its activity and to demonstrate microsomal desaturase activity. Codon optimization of the fat-5 open reading frame (Figure S1) did not produce increased activity. Examination of the FAT–5 predicted protein sequence revealed no prototypic motifs for an N-terminal signal sequence or C-terminal ER retention and retrieval domain. However, introducing selected motifs into the coding sequence actually reduced FAT–5 activity, and replacing the C-terminal five amino acids of FAT-5 with established ER retention signals abolished activity, suggesting an essential function for this motif in the native protein. Through yeast expression, we established that FAT–5 is active in yeast microsome preparations and that its activity was NADH-dependent (Figure 2), consistent with the desaturase being reduced by the NADH-cytochrome b5 reductase—cytochrome b5 electron transport chain. This activity in microsomes, as well as the desaturation activity in seed (Figure 3), and the incorporation of the 16:1 product into lipids, demonstrate that FAT-5 interfaces well with other components of lipid metabolism in heterologous systems.

Experiments by others using constitutive expression of Δ9 desaturases produced little effect in seed tissue; neither the yeast OLE1 desaturase (Polashock et al., 1992) nor the Rattus desaturase (Moon et al., 2000) reduced saturated fatty acids to <90% of the level in untransformed seed. Our initial success expressing FAT-5 using the napin promoter demonstrated that the desaturase was functional in Arabidopsis and showed that expression of FAT-5 lowered 16:0 (Figure 1). The napin promoter from Brassica napus has been used successfully to express a range of genes including several involved in lipid metabolism (Broun and Somerville, 1997; Robert et al., 2005); overexpression of the Arabidopsis ADS1 desaturase in B. juncea with the napin promoter reduced saturates to about 83% of the parental level (Yao et al., 2003). While our results were encouraging, we had not reduced the saturated fatty acids by even 20%. We tested the seed-specific promoter of the napin ortholog in Arabidopsis, 2S2 (Conceicao and Krebbers, 1994), as well as the L. fendleri fatty acid hydroxylase promoter (Broun et al., 1998), but neither promoter reduced the proportion of 16:0 fatty acid to even 60% of the wild-type level.

We were much more successful when we expressed fat-5 using the phaseolin promoter (Sengupta-Gopalan et al., 1985). Like napin, phaseolin is a seed albumin, and its expression is stringently limited to embryogenesis and microsporogenesis and not present in vegetative tissue. This regulation is maintained in Arabidopsis (Chandrasekharan et al., 2003). Analysis of the cis-regulatory elements within 500 bp upstream of the transcriptional start site revealed sequences sufficient for high levels of seed expression as well as regions that limited expression in other tissues (Bustos et al., 1991). The phaseolin promoter was modestly successful when used to express a epoxygenase gene (Hatanaka et al., 2004) but more effective when used to express a 16:0-ACP desaturase (Nguyen et al., 2010). Direct comparison of the level of reporter gene expression in Arabidopsis by seed-specific promoters has demonstrated that phaseolin is superior at achieving high expression (De Jaeger et al., 2002). In our experiment, phaseolin was very successful: in the best transgenic line analysed, the proportion of 16:1 increased to 6.5% in homozygous plants, while 16:0 levels in mature seeds were reduced from 8.6% of total lipids characteristic of wild type to 3% (Figure 4a).

Previous attempts to reduce 16:0 in plant leaves by desaturase overexpression used a cyanobacterial acyl lipid desaturase, either specifically targeted to chloroplasts (Ishizaki-Nishizawa et al., 1996) or using plastid-based expression (Craig et al., 2008). Both experiments produced plants with 16:0 reduced by only 30% relative to wild-type leaf tissue, with lesser changes in nonleaf tissues, and plant phenotypes were sometimes adversely affected (Craig et al., 2008). In similar experiments expressing desaturases that use 16:0-ACP thioesters as substrate, the level of 16:0 was not greatly reduced, even when seed-specific promoters were employed. These experiments produced both 16:1∆9 and 18:1∆11, the elongation product of 16:1∆9, at high levels (Bondaruk et al., 2007; Nguyen et al., 2010).

While there is a small increase in 18:1∆11 in FAT-5 seed, from wild-type levels of 1.2% to 2.5% of total fatty acid, 18:1∆11 is not a major product of FAT-5 expression. A mammalian, ER-type desaturase, when constitutively expressed in tobacco, reduced leaf 16:0 to about 80% of the wild-type level, but there was only a minor effect in seed tissue. In those seeds, 95% of the 16:0 remained (Moon et al., 2000). In contrast, our strategy of using an ER-type desaturase expressed under seed-specific promoter control reduced 16:0 to 33% of the wild-type level (Figure 4a).

The effect of 16:0 reduction by an insertional mutation in fatb, the acyl-ACP thioesterase responsible for export of 16:0 from the chloroplast, were less specific, in respect to both substrates and tissues. Plants mutant in FATB (fatb-ko) produced seed with 16:0 reduced to 44% of the wild-type control, but many seeds were deformed and germination was reduced by half. Those fatb-ko plants also had reduced 18:0, and saturated fatty acids were reduced in every tissue examined (Bonaventure et al., 2003). Although 18:1 increases in our FAT-5 seed samples, the 18:0 level is nearly identical to wild type (Figure 4), and the fatty acid changes are confined to the seed by our promoter choice; FAT-5 plants grow normally and produce seeds that germinate normally, even though both the weight per seed and oil content are somewhat reduced in the line subjected to detailed analysis (Table 1).

The sum of all saturated fatty acids is reduced in the whole seed of FAT-5 transgenic plants to half the total measured in wild-type seed (Figure 4a); almost all the decrease is due to reduced 16:0. As expected, the greatest increase in unsaturated fatty acid is 16:1, although a marked increase in 18:1 also occurs. We determined that most of the increased 18:1 is oleate (18:1∆9), because 18:1∆11 increases to only 2.5% of the total (from 1.2% in wild-type seed); the 18:1∆9 may arise from unexpected 18:0 desaturation by FAT-5, even though there is very little change in the proportion of 18:0, or it may be a metabolic response of the plant to altered levels of saturates. Both decreases in saturated fatty acids and increases in monounsaturated fatty acids are beneficial to human health (Nicklas et al., 2004). Oleate has long been established as an important counterpoint to the negative effects of saturated fatty acids in food (Damude and Kinney, 2008), and more recent studies have demonstrated beneficial effects of 16:1 in animal models (Yang et al., 2011), in cultured human cells (Dimopoulos et al., 2006) and in human diet (Mozaffarian et al., 2010). The goal of our bioengineering strategy was to reduce 16:0 such that desaturated fatty acids would be incorporated into TAG with only minor alteration of the less abundant, but more metabolically significant, polar membrane lipids of seeds. Surprisingly, we found that the fatty acids in the polar lipids are more strongly altered by transgene expression than are the storage lipids; polar lipids contained more than twice the proportion of 16:1 than the neutral lipid fraction, and the PC fraction of the polar lipids had a lower proportion of 16:0 than either total polar lipids or neutral lipids (Figure 4).

The high levels of 16:1 present in the seed phospholipids raises the question of the actual substrate for FAT-5 desaturase activity: the desaturase could use either fatty acids esterified to coenzyme A or act directly on a lipid substrate. Rapid acyl editing reactions which interchange fatty acids esterified to coenzyme A and lipid molecules in yeast (Tanaka et al., 2008) and Arabidopsis (Bates and Browse, 2011) makes determination of the substrate problematic. Indeed, the ratio of fatty acids in each lipid type depends not only on desaturation but also on specificities of acyl transferases, lipases and on the rate of lipid turnover (Bates and Browse, 2011). The composition of membrane lipids is more conservative than that of storage lipids, likely constrained by membrane function in key physiological reactions including trafficking, adaptation to temperature change and signalling (Cahoon et al., 2007). Indeed, although some unusual fatty acids can accumulate to high levels in seed TAG, they are generally excluded from membrane lipids in plant tissues, including seeds (Millar et al., 2000). The highly altered fatty acid composition of the FAT-5 seed membrane fractions was therefore unexpected, indicating that cell membranes tolerate significant changes in the proportion of 16:0 and still retain normal function. The tolerance of the FAT–5 seeds for very low levels of 16:0 in membrane lipids suggests that more intensive efforts to lower 16:0 in neutral lipids may also be tolerated by plant metabolism, even if those efforts produce changes in seed membrane lipids.

It is noteworthy that in the fatty acids of developing, as opposed to mature seed, the effects of FAT–5 were nearly equivalent on the neutral and polar lipid fractions (Figure 6). While 16:1 was much higher in developing seed of transgenic, as compared to wild-type plants, reductions in 16:0 were not as evident as they are in mature seed. Normal development of FAT-5 seed may be due to the greater similarity of their fatty acid composition to the wild type early in seed filling, before the seed reaches maturity (Figure 6).

The high concentration of 16:1 fatty acid in polar lipids (Figure 4) led us to examine whether the 16:1 was predominately at the sn-1 position or was also found at sn-2 position on the glycerol backbone of PC. In wild-type plants, 16:0 was positioned predominantly at sn-1; 16:1 was found at both sn-1 and sn-2 (Figure 5). The fatty acid composition at each position is influenced both by the specificity of acyl transferases (Kim and Huang, 2004) and also by lipid desaturases acting on phospholipids (Miquel and Browse, 1992). Although the 16:1 fatty acid is more abundant in the transgenic plants, it is distributed in a similar fashion to wild type; there is more 16:1 at the sn-2 than at the sn-1 position in both seed types (Figure 5), suggesting that the enzymology at work in wild-type plants can accommodate the greater proportion of 16:1 present in the transgenic lines. A similar positional profile was seen in tobacco leaves expressing the rat stearoyl-CoA desaturase: PC from leaves of plants expressing that desaturase had 16:1 distributed approximately equally between the sn-1 and sn-2 positions (Moon et al., 2000). As acyl editing via PC is the major route of TAG synthesis in Arabidopsis seeds (Bates and Browse, 2011), the capacity of PC to accommodate 16:1 fatty acyl chains at both positions suggests that it may be possible to increase the level of 16:1 that can be incorporated into TAG molecules in future lipid engineering experiments.

The negative health implications of oils high in saturates, including the effects on cardiovascular disease and type 2 diabetes, make research on methods that reduce saturated fatty acids in plants of high interest. Such research requires a more thorough understanding of the biochemistry and metabolism of oil seeds where information is still emerging (Lu et al., 2009), as well as the development of tools capable of altering the fatty acid composition of seeds. We focused here on a transgenic approach for specific reduction of 16:0, the fatty acid that contributes most to saturated fatty acid levels in several commercial vegetable oils. While natural variation or induced mutations can be exploited to obtain desirable seed qualities (Scarth and Tang, 2006), these approaches can not only be lengthy but may also produce plants with other undesirable traits (Bonaventure et al., 2003). In the model plant Arabidopsis, we have demonstrated that substantial reductions in saturated fatty acids are compatible with normal seed physiology and reproduction, with the implication that changes of this magnitude will be tolerated by crop oilseed species.

Experimental procedures

Cloning of fat-5 and transformation of Arabidopsis

The cDNA for fat–5 (W06D12.3, GenBank Accession No. AF260242) from C. elegans was subcloned from pMK195::fat-5 (Watts and Browse, 2000) into the pBART (Gleave, 1992) binary plant vector under control of the napin promoter (Ellerstrom et al., 1996) and into pDS-Red-PHAS under control of the P. vulgaris phaseolin promoter (Sengupta-Gopalan et al., 1985). The cDNA was PCR amplified from the source plasmid with specific primers and cloned into the pENTR-D-TOPO vector (Invitrogen, Carlsbad, CA) prior to insertion into Gateway-modified versions of pTLH458 with the Arabidopsis thaliana albumin 2S2 promoter (Guerche et al., 1990) and pBP73 with the L. fendleri FAH12 promoter (Broun and Somerville, 1997). Agrobacterium-mediated transformation of A. thaliana ecotype Columbia (Col-0) was carried out by floral dip (Clough and Bent, 1998). Seeds transformed with pBART and pBP73 constructs were selected using glufosinate (Finale; Farnam Companies, Phoenix, AZ); those transformed with pTLH458 were selected with glyphosate (Surrender Eraser; Control Solutions Inc., Pasadena, TX); those transformed with pDS-Red-PHAS were selected by red fluorescence. Plants were propagated in growth chambers under continuous fluorescent light (100–150 μmol/m2/s) at 22 °C.

Functional assay in yeast

Constructs for yeast expression were generated by PCR amplification from pMK195::fat-5 using sets of primers designed to modify the N-terminal and C-terminal sequences of the resulting protein. The amplified fragments were cloned into the pENTR-D-TOPO vector (Invitrogen) prior to insertion into a Gateway-modified version of pMK195 (Overvoorde et al., 1996). The S. cerevisiae strain L8-14C (ole1, Stukey et al., 1989) cannot synthesize unsaturated fatty acids due to disruption of the OLE1 desaturase gene and will not grow unless the media is supplied with unsaturated fatty acid. Yeast cells were grown at 30 °C in Yeast extract, Peptone, Dextrose medium containing 50 mm of either oleic or linoleic acid (NuChek Prep, Elysian, MN) and 1% tergitol. The functional assay for desaturase activity is based on rescue of ole1 following transformation with constructs expressing fat-5 and derivatives, testing for growth on medium without uracil, the auxotrophic marker for pMK195. For fatty acid analysis, clones were grown in liquid minimal medium without uracil for 1–2 days; clones with little or no desaturation activity were supplemented with 50 mm linoleic acid and 1% tergitol.

Desaturase assay

Yeast cultures were harvested and resuspended in 1 mL cold lysis buffer (20 mm tris–HCl pH 7.9, 1 mm EDTA pH 8, 5% glycerol, 10 mm MgCl2, 1 mm dithiothreitol (DTT), 0.3 m ammonium sulphate). Cells were lysed by vortexing with glass beads for 10 × 30 s pulses with 30 s gaps on ice. After a 2000 g, 4 °C spin to remove intact cells, the suspension was centrifuged at 10 000 g, 4 °C. The supernatant was subjected to ultracentrifugation at 100 000 g to obtain the microsomal fraction. The pellet was resuspended in 0.1 m phosphate buffer, pH 7.2 and protein concentration determined (Bradford, 1976). To measure desaturation, microsomes equivalent to 10 μg protein were added to 0.1 phosphate buffer pH 7.2 with 1000 U catalase (Worthington, Biochemical Corporation, Lakewood, NJ), 0.2 m coenzyme A (yeast; Sigma, St. Louis, MO), 8 mm ATP, 1 mm DTT, 10 mm MgCl2 and 60nCi palmitoyl coenzyme A [palmitoyl-1-14C]-(55 mCi/mmol; PerkinElmer, Norwalk, CT) in 100 μL total volume. The reaction was initiated by addition of NADH to a concentration of 1 mm and incubated at 22 °C for 60 min unless indicated otherwise. Enzyme activity was terminated by addition of 200 μL 2.5 m KOH in 75% ethanol and lipids were saponified by incubating at 85 °C for 1 h. After acidification with 280 μL formic acid, the fatty acids were extracted with hexane. Saturated and monounsaturated fatty acids were separated by TLC on silica gel plates (Whatman P60, Piscataway, NJ) impregnated with 10% AgNO3 in acetonitrile, with CHCl3/MeOH/HOAc/H20, 90 : 8 : 1 : 0.8 (v/v) as developing solvent. The plate was exposed to a phosphorimager screen and scanned on a SSI445 phosphorimager (GE Healthcare, Piscataway, NJ). Band intensities were quantified by ImageQuant software (GE Healthcare) to calculate the conversion of 16:0 to 16:1 as a measure of desaturase activity.

Fatty acid analysis

For Arabidopsis, 20–50 seeds were incubated in 1 mL of 2.5% (v/v) sulphuric acid in methanol for 1.5 h at 80 °C (Miquel and Browse, 1992). The resulting fatty acid methyl esters were separated by GC (Alltech EC wax; 30 m × 0.53 mm i.d. × 1.20 μm column) and identified by flame ionization detection. Chromatography parameters were 210 °C for 2 min followed by a ramp to 245 °C at 10 °C per min and a 4 min final temperature hold. For GC/MS, fatty acid methyl esters prepared as described were analysed using an Agilent 6890 GC equipped with an Agilent 5975 MS detector. After chromatography through a 30 × 0.25-mm SP-2380 column (Supelco, Bellefonte, PA), using helium as the carrier gas at 1.4 mL/min, with initial temperature of 120 °C for 1 min followed by an increase of 10 °C/min to 190 °C followed by an increase in 2 °C/min to 200 °C. The isomers of 18:1 were identified by comparison of chromatography retention times and mass spectra with those of authentic standards.

Lipid extraction and analysis

Total lipids were extracted from 50 mg dry seed aliquots according to a modified protocol (Bligh and Dyer, 1959) from the Kansas Lipidomics Center (http//www.k-state.edu/lipid/lipidomics). Seeds were added to 1 mL 80 °C isopropanol, 0.01% butylated hydroxytoluene (BHT) for 15 min and then homogenized by Polytron. Two millilitres of chloroform, 3 mL methanol and 1.5 mL water were added to form a single phase, and after mixing, 2 mL chloroform and 2 mL 0.88% KCl were added to achieve phase separation. The chloroform phase containing the lipids was removed and the aqueous phase back extracted with 2 mL chloroform. Combined chloroform phases were dried under N2 and resuspended in a small volume of toluene with 0.005% BHT. For analysis of developing seeds, the seeds were removed from approximately ten siliques either 9–10 or 12–13 days after flowering, transferred to 80 °C isopropanol for 10 min, then stored at −20 °C and extracted as for dry seeds. Neutral and polar lipids were separated using a 3-cm silica column: after equilibration with chloroform and chloroform/methanol, 99 : 1 (v/v) and loading of total lipids, the neutral lipids were eluted with 5 mL of chloroform/methanol, 99 : 1 (v/v) followed by elution of the polar lipids with 5 mL of chloroform/methanol/water, 5 : 5 : 1 (v/v/v). Phase separation of the polar lipid fraction was achieved by addition of 2 mL chloroform and 2 mL 0.88% KCl. The chloroform phase of polar lipids and the neutral lipid fraction were dried under N2 and resuspended in a small volume of toluene, 0.005% BHT. Fatty acid analysis was carried out on aliquots of total lipids, neutral lipids and polar lipids. To analyse PC, the polar lipids were separated by TLC. Silica plates (Whatman P60) were dipped in 0.15 m ammonium sulphate and dried at 110 °C for at least 3 h prior to use. Polar lipids were separated in acetone/toluene/water, 91 : 30 : 8 (v/v/v). Plates were stained with 0.005% primulin in 80% acetone and lipids visualized by UV light. Bands representing the predominant seed phospholipid, PC, were collected for fatty acid analysis.

Regiochemical analysis of PC

Total lipids from 100 mg dry seed were extracted and isolated as above. Polar lipids were separated by TLC (Silica gel 60: 20X20; EMD Chemicals, Gibbstown, NJ) with chloroform/methanol/acetic acid, 75 : 25 : 8 (v/v/v). The PC bands were collected and extracted with chloroform/methanol/water, 5 : 5 : 1(v/v/v) and processed as for polar lipid fraction above. The fatty acid at the sn-2 position was preferentially hydrolysed as follows. A 500 μg aliquot of PC was dried under N2, resuspended in 0.5 mL diethyl ether. To this was added 150 U of porcine pancreas Phospholipase A2 (Sigma) in 150 μL 50 mm Tris–HCl pH 8.7, 5 mm CaCl2. Samples were vortexed for 3–5 min at 22 °C to achieve approximately 50% digestion. The ether phase was dried under N2, then 4 mL of chloroform/methanol, 2 : 1 (v/v) and 1 mL 0.15 m acetic acid were added and the chloroform phase dried under N2. The digestion products were separated by TLC (Silica gel 60; EMD Chemicals) with a development in chloroform/methanol/acetic acid/water, 50 : 30 : 8 : 4 (v/v/v/v) for 10 cm, followed by full development in hexane/ether/acetate, 80 : 20 : 1 (v/v/v). After staining with primulin, bands representing PC, lysoPC and free fatty acids were collected, transmethylated and FAMEs analysed by GC.

Acknowledgements

This work was supported by Bayer CropScience, USDA-AFRI grant no. 2010-65115-20393, and the Agricultural Research Center at Washington State University. We thank Dr. Shuangyi Bai of Washington State University for his analysis of preliminary constructs in yeast, and Dr. Peter Denolf of Bayer CropScience for extensive advice and consultation.

Ancillary