The pathway of triacylglycerol synthesis through phosphatidylcholine in Arabidopsis produces a bottleneck for the accumulation of unusual fatty acids in transgenic seeds


(fax 509-335-7643; e-mail


Engineering of oilseed plants to accumulate unusual fatty acids (FAs) in seed triacylglycerol (TAG) requires not only the biosynthetic enzymes for unusual FAs but also efficient utilization of the unusual FAs by the host-plant TAG biosynthetic pathways. Competing pathways of diacylglycerol (DAG) and subsequent TAG synthesis ultimately affect TAG FA composition. The membrane lipid phosphatidylcholine (PC) is the substrate for many FA-modifying enzymes (desaturases, hydroxylases, etc.) and DAG can be derived from PC for TAG synthesis. The relative proportion of PC-derived DAG versus de novo synthesized DAG utilized for TAG synthesis, and the ability of each pathway to utilize unusual FA substrates, are unknown for most oilseed plants, including Arabidopsis thaliana. Through metabolic labeling experiments we demonstrate that the relative flux of de novo DAG into the PC-derived DAG pathway versus direct conversion to TAG is ∼14/1 in wild-type Arabidopsis. Expression of the Ricinus communis FA hydroxylase reduced the flux of de novo DAG into PC by ∼70%. Synthesis of TAG directly from de novo DAG did not increase, resulting in lower total synthesis of labeled lipids. Hydroxy-FA containing de novo DAG was rapidly synthesized, but it was not efficiently accumulated or converted to PC and TAG, and appeared to be in a futile cycle of synthesis and degradation. However, FA hydroxylation on PC and conversion to DAG allowed some hydroxy-FA to accumulate in sn-2 TAG. Therefore, the flux of DAG through PC represents a major bottleneck for the accumulation of unusual FAs in TAG of transgenic Arabidopsis seeds.


Triacylglycerols (TAGs) produced by oilseed plants are a major renewable source of reduced carbon for use as food, fuel and industrial feedstocks. Common oilseed crops such as canola (Brassica napus) and soybean (Glycine max) produce large yields of TAG for mostly food and cooking applications and have fatty acid (FA) compositions limited to 16- or 18-carbon acyl chains with zero to three double bonds (Dyer et al., 2008). However, within the plant kingdom there is a very large repertoire of unusual FAs that differ in chain length, number and position of double bonds, or that contain different functional groups such as hydroxy, epoxy, etc. These unusual FAs have many uses as industrial feedstocks for polymers, lubricants and resins. However, most plants that produce unusual FAs are either limited in oil yield or have undesirable agronomic features (Dyer et al., 2008). Most attempts to produce unusual FAs in transgenic oilseeds have resulted in low yields of the unusual FA in seed TAG. These low yields in transgenic oilseeds may be due to biosynthetic bottlenecks within the host plant that inhibit synthesis of unusual FAs or induce their turnover (Cahoon et al., 2007; Dyer et al., 2008; Moire et al., 2004). To circumvent these bottlenecks in the production of seed oils containing unusual FAs we need to understand not only the enzymes/genes involved in TAG synthesis and turnover, but also the relative fluxes through competing pathways of TAG biosynthesis.

The FA building blocks for TAG biosynthesis are produced within the plastid. Newly-synthesized FAs [18:1 >> 16:0 > 18:0, (FA, number of carbons:number of double bonds)] are exported from the plastid into the cytosolic acyl-CoA pool, where they can be directly incorporated into membrane lipid phosphatidylcholine (PC) by the PC:acyl-CoA acyl editing cycle (Li-Beisson et al., 2010). Acyl editing involves a very rapid cyclic deacylation and reacylation of PC (Bates et al., 2009, 2007). Extra-plastidic desaturation of 18:1 to 18:2 and 18:3 takes place while the FAs are bound to PC (Sperling et al., 1993). Thus, the PC acyl editing cycle incorporates nascent 18:1 into PC and releases polyunsaturated FAs (PUFAs) from PC, generating an acyl-CoA pool of newly synthesized FAs and further-modified FAs to be utilized for glycerolipid synthesis in the endoplasmic reticulum (ER) (Li-Beisson et al., 2010).

Triacylglycerol is assembled in the ER by acylation of diacylglycerol (DAG) by either the acyl-CoA-dependent diacylglycerol acyltransferase (DGAT) (Cases et al., 1998; Li-Beisson et al., 2010) or by the acyl-CoA-independent phospholipid:diacylglycerol acyltransferase (PDAT) (Dahlqvist et al., 2000; Li-Beisson et al., 2010). However, DAG can be produced through at least two pathways (Figure 1) (Bates et al., 2009). Each pathway can provide different molecular species of DAG and ultimately affect two-thirds of the TAG FA composition. First, de novo synthesis of DAG involves esterification of FA to the sn-1 and sn-2 positions of glycerol-3-phosphate (G3P) producing phosphatidic acid (PA), by acyl-CoA:glycerol-3-phosphate acyltransferase (GPAT) and acyl-CoA:lyso-phosphatidic acid acyltransferase (LPAAT), respectively. Subsequent removal of phosphate by phosphatidic acid phosphatase (PAP) produces de novo DAG (Li-Beisson et al., 2010). The second pathway of DAG synthesis utilizes de novo DAG to first synthesize PC. Diacylglycerol is regenerated from PC by the removal of the phosphocholine head-group to generate the substrate for TAG synthesis. Since PC is the site of ER FA desaturation and acyl editing the molecular species of PC-derived DAG can be different from the molecular species of de novo DAG utilized to synthesize PC. PC-derived DAG can be produced by the reverse action of the enzymes that produce PC, CDP-DAG:cholinephosphotransferase (CPT) (Slack et al., 1983, 1985) and phosphatidylcholine:diacylglycerol cholinephosphotransferase (PDCT) (Lu et al., 2009). A lipase-mediated pathway utilizing phospholipase C, or phospholipase D (Lee et al., 2011) and PAP may also be possible. Thus, the production of TAG through PC requires synthesis of DAG twice, first de novo DAG synthesis and conversion to PC, then conversion of PC back to DAG (Figure 1). The relative FA composition of de novo DAG versus PC-derived DAG depends on the acyl selectivities of GPAT and LPAAT, and the relative activities of PC-FA-modifying enzymes, PC acyl editing and the interconversion of DAG and PC for both PC synthesis and derivation of DAG from PC.

Figure 1.

 The flow of glycerol into triacylglycerol (TAG).
The model is based on Bates et al. (2009) where there is a small pool of de novo diacylglycerol (DAG) and a separate small pool of phosphatidylcholine (PC)-derived DAG involved in oil synthesis that rapidly exchanges with the bulk DAG pool in the oil body. The PC is also split into a small biosynthetically active pool and a larger bulk pool. The width of the arrows indicates relative flux. Dashed arrows indicate uncertainty in the amount of flux. Flux A: TAG synthesis through PC. Flux B: de novo DAG bypasses PC for incorporation into the other DAG pools. Flux C: TAG synthesis directly from de novo DAG.

Quantitatively, the ratio of de novo DAG or PC-derived DAG utilized for TAG synthesis is not known for most oilseeds, nor are the enzymes/genes that differentiate between the pathways. The use of in vivo metabolic labeling experiments has suggested that PC can be utilized for TAG synthesis on a species-dependent basis. The interconversion of PC and DAG for TAG synthesis has been proposed as a major route to provide PUFA-containing DAG through glycerol backbone labeling of excised soybean, linseed (Linum usitatissimum) and safflower (Carthamus tinctorius) cotyledons (Griffiths et al., 1988; Slack et al., 1978). Phosphatidylcholine head group labeling of linseed cotyledons suggested that the PC and DAG pools are in equilibrium (Slack et al., 1983). However, glycerol backbone labeling studies of avocado (Persea americana) mesocarp suggested that PC is not involved in TAG synthesis in this species (Griffiths et al., 1988). Additionally, acyl and backbone labeling studies of developing B. napus embryos (Perry and Harwood, 1993a,b) led the authors to suggest that TAG synthesis is controlled through the action of DGAT on the de novo DAG pool. Recently, characterization of the Arabidopsis thaliana rod1 mutant has provided genetic evidence that at least 40% of PUFAs (∼20% of total FAs) in TAG are derived from DAG–PC interconversions by PDCT (Lu et al., 2009).

Recently we developed a more quantitative model for the flux of DAG through competing metabolic pathways in developing soybean embryos (Bates et al., 2009). Through in vivo metabolic labeling and the analysis of glycerolipid synthesis rates, initial molecular species and FA stereochemical compositions we proposed a model containing kinetically and functionally distinct pools of DAG (Figure 1). In the model, even though TAG is accumulating to 20× that of membrane lipids, the structure of the pathway indicates that a de novo DAG molecule will flux through the large bulk PC pool prior to TAG synthesis. Therefore, de novo DAG is used primarily for PC synthesis and TAG is synthesized primarily from PC-derived DAG such that >95% of TAG synthesis is through PC-derived DAG in soybeans (Figure 1, flux A) (Bates et al., 2009).

Deciphering the relative flux through competing metabolic pathways is crucial for our understanding of TAG synthesis in plants that accumulate different FAs and our ability to engineer the FA composition of oilseed crops. Many unusual FA-synthesizing enzymes are homologs of the 18:1 desaturase FAD2 (Okuley et al., 1994), and also utilize PC as a substrate. For example, the castor bean (Ricinus communis) hydroxylase (RcFAH12) produces ricinoleic acid (18:1-OH; 12-hydroxy-9-cis-octadecenoic acid) from 18:1 esterified to the sn-2 position of PC (Bafor et al., 1991; Moreau and Stumpf, 1981; Vandeloo et al., 1995). During castor endosperm TAG synthesis ricinoleate only accumulates to ∼5% of FAs in PC (Stahl et al., 1995), but accumulates up to ∼90% of FAs in TAG and over 70% of seed TAG contain three ricinoleates (Lin et al., 2003). This requires that at least 90% of seed oleate fluxes through the sn-2 position of PC where it is hydroxylated by RcFAH12 and then is redistributed into the sn-1, -2 and -3 positions of TAG. Assays of castor endosperm microsomes suggest that ricinoleate is efficiently released from PC and the ricinoleoyl-CoA produced is utilized to synthesize de novo DAG. The di-ricinoleoyl de novo DAG is more efficiently utilized directly for tri-ricinoleoyl-TAG synthesis than PC synthesis (Bafor et al., 1991), suggesting that castor endosperm does not utilize PC-derived DAG for TAG synthesis. However, when RcFAH12 is expressed in transgenic Arabidopsis, hydroxy-FAs (HFAs; 18:1-OH and 18:2-OH) accumulate up to ∼12% in developing seed PC, while only accumulating to ∼17% in mature seed TAG. Approximately, half of the TAG species that accumulate contain only a single HFA which is located at over 70% in the sn-2 position, and total TAG accumulation is reduced by at least 20% (van Erp et al., 2011). These results indicate that there are important differences in the metabolic pathways that affect the accumulation of HFA (and TAG) between castor endosperm and transgenic Arabidopsis seeds. However, the basis of these differences remains unknown.

To fully understand the pathway of Arabidopsis TAG synthesis and to utilize Arabidopsis as a model for oilseed crop engineering, we need to understand the relative fluxes of TAG synthesis from de novo DAG or through PC-derived DAG, and how expression of enzymes for unusual FA synthesis (such as RcFAH12) affects each of these fluxes. Here we utilize developing Col-0, fae1 and CL37 seeds to determine the fluxes through competing TAG synthesis pathways in wild-type and HFA-producing Arabidopsis. The CL37 line expresses RcFAH12 under control of the phaseolin seed-storage-protein promoter in the fae1 background (Lu et al., 2006). The fae1 mutation reduces elongation of 18-carbon FAs and thus simplifies the HFA profile. Through [14C]glycerol labeling we demonstrate that TAG synthesis primarily utilizes PC-derived DAG in developing Col-0, fae1 and CL37 seeds. However, in CL37, HFA-containing de novo DAG is restricted in the flux through PC, reducing the total PC-derived DAG flux and total glycerolipid labeling. However, the PC-derived DAG pathway can utilize the (sn-2)-HFA-PC product of RcFAH12 to generate mostly (sn-2)-HFA-DAG for TAG synthesis. Therefore, the flux of TAG synthesis through PC is a major bottleneck for the accumulation of TAG containing three unusual FAs in Arabidopsis seeds.


The major flux of de novo DAG is for PC synthesis in developing Col-0 and fae1 seeds

Mature Arabidopsis seeds contain about 20 times more TAG than membrane lipids (Li-Beisson et al., 2010). Thus, around 95% of all de novo DAG backbones synthesized in developing seeds end up in TAG. The initial synthesis rate of TAG or PC from de novo DAG indicates the relative fluxes of TAG synthesis through de novo DAG or PC-derived DAG, respectively. To determine the rates of in vivo glycerolipid synthesis we followed [14C]glycerol into the backbones of newly synthesized glycerolipids in developing Col-0, fae1 and CL37 seeds. As demonstrated previously, [14C]glycerol can be rapidly incorporated into the backbones and acyl groups of newly synthesized membrane and storage lipids without labeling of the head-groups of polar lipids such as PC (Bates et al., 2009, 2007). Developing Arabidopsis seeds excised from siliques were found to take up [14C]glycerol and synthesize 14C-labeled lipids for at least an hour of continuous labeling (Figure S1 in Supporting Information). The patterns of backbone labeling in DAG, PC, TAG and phosphatidylethanolamine (PE) for Col-0 and fae1 were very similar (Figure 2) and are described together below; CL37 labeling is described in a later section.

Figure 2.

 Incorporation of [14C]glycerol into backbones of major developing seed glycerolipids.
(a–c) Zero to 60-min time course. (d–f) The first 15 min of each time course. (a), (d) Col-0. (b), (e) fae1. (c), (f) CL37. Symbols: phosphatidylcholine (PC), diamonds; diacylglycerol (DAG), squares; triacylglycerol (TAG), triangles; phosphatidylethanolamine (PE), X. Data represent the average ± SE for two labeling replicates.

Initially [14C]glycerol backbone labeling of DAG was highest, with all other products lagging. However, PC rapidly accumulated more backbone label than DAG and was still accumulating label linearly through 60 min (Figure 2), representing a classical precursor–product relationship of de novo DAG with PC. The TAG backbone labeling lagged significantly behind that of DAG and PC and slowly accelerated over 60 min (Figure 2), indicating that the de novo DAG pool is not the major source for TAG labeling. The relative slopes of PC and TAG backbone labeling over the first 10 min of [14C]glycerol labeling are a measure of the relative initial rates of PC/TAG synthesis from initially 14C-labeled de novo DAG. Col-0 and fae1 displayed initial PC/TAG synthesis ratios of about 14/1 and 15/1, respectively. These results indicate that PC is preferentially synthesized from de novo DAG over direct synthesis of TAG. Developing seeds of Col-0 and fae1 accumulate PC and PE in a molar ratio of ∼1.7/1 (Table S1). However, the relative rates of [14C]glycerol accumulation in PC/PE between 10 and 60 min in Col-0 and fae1 are ∼17/1 and ∼18/1, respectively (Figure 2). These results suggest that PC is synthesized from de novo DAG at over 10 times what is required for membrane biogenesis. Therefore, in developing Arabidopsis seeds which are actively accumulating TAG, the much higher rate of PC synthesis from de novo DAG over both TAG and PE suggest that the primary pathway of TAG synthesis is not directly from de novo DAG but is through PC-derived DAG.

In addition to labeling the backbone of glycerolipids, a small proportion of [14C]glycerol was metabolized to acetyl-CoA and incorporated into FAs, so that about 11% of the total lipid radiolabel at 60 min accumulates in the acyl chains (Figure S1). In contrast to glycerol backbone labeling which enters lipids solely through de novo DAG synthesis, newly labeled FAs are incorporated into lipids through three separate mechanisms: (i) de novo DAG synthesis, (ii) acyl editing of PC, (iii) sn-3 esterification of DAG to produce TAG (Bates et al., 2009, 2007). At early time points the ratio of acyl/glycerol labeling in each lipid class reflects the lag that glycerol takes through precursor pools to be incorporated alongside newly synthesized 14C-FA directly incorporated into lipids by each acyl transfer mechanism. Figures 3(a) and 3(b) show the percentage of each lipid class label that was in the acyl chains for Col-0 and fae1 lipids, respectively. [14C]FA are a small proportion of total label in DAG and PC because the backbone of these lipids is rapidly labeled with [14C]glycerol through de novo DAG and PC synthesis. However, [14C]FA were up to 40% of the total label in TAG, consistent with most backbone labeled TAG being derived from a more slowly labeled DAG pool. These results are consistent with the major flux of newly synthesized de novo DAG into PC rather than into TAG.

Figure 3.

 Percentage of total lipid class radiolabel in the acyl groups of phosphatidylcholine (PC), diacylglycerol (DAG) and triacylglycerol (TAG).
(a) Col-0.
(b) fae1.
(c) CL37.
Symbols: PC, diamonds; DAG, squares; TAG, triangles. Data represent the average ± SE for two labeling replicates.

Bulk DAG is distinct from de novo DAG and derived from PC

Since the major flux of glycerolipid synthesis in developing Arabidopsis seeds is for the production of TAG, the bulk cellular DAG must be utilized primarily for TAG synthesis, as demonstrated previously in soybean (Bates et al., 2009). The FA composition of de novo DAG can be determined by separating and analyzing the molecular species of DAG rapidly backbone labeled with [14C]glycerol. Molecular species of 14C-DAG from fae1 seeds labeled for 6 and 10 min were separated by argentation TLC based on the number of double bonds (Figure 4a). The proportion of backbone label in each band was used to calculate the FA composition of the de novo DAG pool. The results are compared to the composition of bulk PC and bulk DAG in Figure 4(b). [14C]glycerol backbone labeled DAG was dominated by molecular species containing 18:1 and 18:2 FAs (Figure 4a), corresponding to a FA composition of 48% 18:1 and 40% PUFA (Figure 4b). This is distinctly different from the bulk DAG mass composition of 18% 18:1 and 62% PUFA. The FA composition of the bulk DAG in fae1 is much more similar to PC than to de novo DAG (Figure 4b). The distinct difference in FA composition between de novo DAG and bulk DAG suggest that PC acyl editing and de novo DAG synthesis cannot provide all of the PUFA-containing DAG that accumulates in TAG and again supports the proposal that PC-derived DAG provides the major substrate for TAG synthesis in Arabidopsis.

Figure 4.

 Comparison of de novo diacylglycerol (DAG), bulk DAG and phosphatidylcholine (PC) from fae1.
(a) Molecular species of 6- and 10-min backbone labeled DAG. Molecular species are represented as the sum of two fatty acids (FAs).
(b) The FA composition of de novo DAG (calculated from a) as compared with the composition of bulk DAG and PC (calculated from Table S1).
The FA nomenclature is based on the number of double bonds: S, saturates; M, monoenes; D, dienes; T, trienes. Data represent the average ± SE for two labeling replicates and three mass replicates.

Flux of TAG synthesis through PC is reduced in CL37

Castor bean endosperm accumulates about 90% of seed FAs as 18:1-OH and over 70% of seed TAGs contain three HFAs (tri-HFA-TAG) (Lin et al., 2003). This requires that at least 90% of seed oleate fluxes through the sn-2 position of PC where it is hydroxylated by RcFAH12 and then is redistributed into the sn-1, -2 and -3 positions of TAG (Bafor et al., 1991). However, in our CL37 plants expressing RcFAH12, HFAs only make up about 17% of mature seed FAs and the percentage of oil per dry weight is reduced by about 30% from the parental line fae1 (Figure S2). Of the TAG species that do accumulate in CL37, about 48% contain no HFAs (0-HFA-TAG), 44% contain a single HFA (mono-HFA-TAG) and about 8% contain two HFAs (di-HFA-TAG). Tri-HFA-TAG was not detected. Regiochemical distribution of the HFAs within TAG indicates that the sn-2 position contains about 72% of the HFAs in mono-HFA-TAG and about 40% of the HFAs in di-HFA-TAG (van Erp et al., 2011). Therefore, the HFA-TAGs in mature CL37 seeds were synthesized primarily from mono(sn-2)-HFA-DAG, and the utilization of di-HFA-DAG for TAG synthesis is limited. To determine if the pathways of DAG metabolism affect the accumulation of HFA-containing TAG in transgenic Arabidopsis, we compared the flux of [14C]glycerol into normal and HFA-containing species of DAG, PC and TAG in developing CL37 seeds. At this stage of development HFAs have begun accumulating in membrane and storage lipids, but represent <4% of the total seed FAs (Table S1).

Figure 5 demonstrates the accumulation of [14C]glycerol in the backbones of combined HFA-containing and non HFA-containing lipid classes from CL37 as compared with that of fae1. Total incorporation of [14C]glycerol into the backbones of CL37 lipids was reduced by about 50% compared with the fae1 control by 60 min of labeling (Figure 5a). The reduction in total lipid labeling is due primarily to a reduction of approximately 75% in accumulation of label in PC (Figure 5b). Diacyglycerol labeling was also reduced by about 40% (Figure 5c), and TAG backbone labeling was reduced by about 20% (Figure 5d). These, results indicate that the approximately 50% reduction in [14C]glycerol incorporation into total lipids is primarily through a reduction in the utilization of de novo DAG for PC synthesis. However, even with the large reduction in PC labeling the relative DAG-PC labeling still retains a stronger precursor–product relationship than DAG-TAG labeling (Figure 2c,f). The ratio of initial synthesis of PC/TAG from de novo DAG over the first 10 min was reduced from ∼15/1 in fae1 (Figure 2b,e) to ∼4/1 in CL37 (Figure 2c,f). The reduction in PC glycerol backbone labeling appears to be specific for flux of DAG through PC for TAG synthesis and not total ER membrane lipid synthesis, because the relative rate of PC/PE labeling from 10 to 60 min was also reduced from ∼18/1 in fae1 to ∼6/1 in CL37, yet the mass ratios of PC, DAG and PE were not significantly changed from Col-0 and fae1 (Table S1). These results suggest that the flux of DAG through PC into TAG has been reduced by over 70%, but it still represents the major pathway for utilization of de novo DAG. Additionally, the flux of de novo DAG directly to TAG in CL37 was not increased to compensate for the decrease in the PC-derived DAG pathway, which may result in the reduced total oil phenotype in CL37 seeds.

Figure 5.

 Comparison of the accumulation of backbone label in fae1 and CL37 lipids.
(a) Total lipids.
(b) Phosphatidylcholine (PC).
(c) Diacylglycerol (DAG).
(d) Triacylglycerol (TAG).
Symbols: fae1, diamonds; CL37, squares. Data represent the average ± SE for two labeling replicates.

As in Col-0 and fae1, some [14C]glycerol is utilized for FA synthesis and the percentage acyl labeling represents the relative rates of backbone and newly synthesized FAs incorporated into lipids. Similar to Col-0 and fae1, the percentage of total label in the acyl groups of TAG in CL37 at early time points is much higher than DAG, indicating that de novo DAG is not the major substrate for TAG synthesis (Figure 3c). However, at later time points the percentage acyl labeling of CL37 TAG is much closer to the equilibrated PC/DAG acyl labeling than in Col-0 and fae1. These data indicate that in CL37 the relative acyl/glycerol labeling in TAG changes more rapidly than in Col-0 and fae1, due to either an increased flux of glycerol or reduced acyl labeling into TAG over the time course. The ratio of acyl/glycerol labeling in DAG and PC of CL37 is not significantly different from that in fae1. Therefore, the reduction in glycerol labeling has been matched by a reduction in acyl labeling through de novo DAG/PC synthesis and acyl editing.

HFA-containing de novo DAG has limited accumulation and conversion to PC

Figure 6 shows the relative accumulation of backbone-labeled 0-HFA-DAG and mono-HFA-DAG in CL37. Labeled di-HFA-DAG was not detected. At very early time points mono-HFA-DAG makes up a substantial proportion of labeled de novo DAG, and with extrapolation to time zero, mono-HFA-DAG contributes over 50% of all synthesized de novo DAG (Figure 6b). After reaching a quasi-steady state the [14C]glycerol labeled 0-HFA-DAG accumulates at about eight times the rate for mono-HFA-DAG (Figure 6a). These results suggest that the Arabidopsis GPAT and LPAAT enzymes can utilize HFA for at least a quarter of the total FAs esterified to G3P. However, the backbone labeled mono-HFA-DAG may be rapidly utilized or turned over such that it does not accumulate over time. Figure 6(c) demonstrates the positional specificity of HFA in backbone labeled mono-HFA-DAG. Backbone labeled mono-HFA-DAG had an approximately constant HFA stereochemistry of about 60% sn-1 and 40% sn-2 over the 60-min time course. Hydroxylation of PC and conversion to PC-derived DAG would produce sn-2 mono-HFA-DAG; therefore the constant stereochemistry indicates that the backbone labeled mono-HFA-DAG pool represents primarily de novo DAG over the 60-min time course. As both GPAT and LPAAT can utilize HFA as substrates it is possible that de novo di-HFA-DAG can also be produced, but it may be more rapidly turned over than mono-HFA-DAG such that it does not accumulate enough to measure it. Alternatively, LPAAT may have selectivity against HFA-LPA such that di-HFA-DAG is rarely formed. Additionally, the unequal stereochemistry suggests that LPAAT may have more of a selectivity against HFA-CoA than GPAT.

Figure 6.

 Relative [14C]glycerol backbone labeled zero hydroxy-fatty acid-diacylglycerol (0-HFA-DAG) and mono-HFA-DAG in developing CL37 seeds.
(a) Accumulation of [14C]glycerol in the backbone of 0-HFA-DAG (diamonds) and mono-HFA-DAG (squares).
(b) Percentage of total DAG labeling from (a).
(c) Distribution of HFA in the sn-1 (diamonds) or sn-2 (squares) position of mono-HFA-DAG.
Data represent the average ± SE for two labeling replicates.

To determine if the rapidly synthesized mono-HFA-DAG is quickly utilized by the flux of de novo DAG through PC, we determined the relative amount of backbone labeled PC with and without HFA that accumulate in CL37 (Figure 7). Only 0-HFA-PC and mono-HFA-PC were labeled. No 14C-di-HFA-PC was detected. As PC was synthesized from labeled de novo DAG (Figures 2c,f and 5b) mono-HFA-PC accumulated at an approximately constant 10% of total labeled PC (Figure 7). Hydroxylation of backbone labeled PC by RcFAH12 would be expected to lag similar to that of 18:1 desaturation on PC (Bates et al., 2009) and thus may not contribute to the radiolabeled HFA-PC at these time points. Therefore, even though HFA-DAG can be synthesized up to 50% of total DAG, its limited accumulation in the DAG pool or PC suggests that HFA-DAG is restrained in the flux through PC as compared with 0-HFA-DAG.

Figure 7.

 Relative [14C]glycerol backbone labeled zero hydroxy-fatty acid-phosphatidylcholine (0-HFA-PC) and mono-HFA-PC in developing CL37 seeds. Symbols: 0-HFA-PC, diamonds; mono-HFA-PC, squares. Data represent the average ± SE for two labeling replicates.

HFA-containing molecular species of TAG synthesized from de novo DAG

Triacylglycerol synthesized rapidly from initially labeled de novo DAG in CL37 consisted of 0-HFA-, mono-HFA- and di-HFA-TAG (Figure 8). Labeled tri-HFA-TAG was not detected. Over the time course each TAG species accumulated label at accelerating rates with 0-HFA-TAG > mono-HFA-TAG > di-HFA-TAG (Figure 8a). However, the relative amount of backbone labeled TAG species that accumulated at the earliest time points was very different from the later time points. At 3 min di-HFA-TAG contained the most label (∼60%), but the least label at 60 min (∼5%) when the relative accumulation of each TAG species has reached an approximate steady state (Figure 8b). These results indicate that di-HFA-TAG can be rapidly synthesized from labeled de novo DAG, yet is hindered in accumulation, similar to mono-HFA-DAG (Figure 6a). Considering that the PC/TAG ratio of fluxes from de novo DAG in CL37 is 4/1, then the ∼24% HFA-containing TAG that accumulates directly from de novo DAG (Figure 8b) may contribute <5% of the total TAG that accumulates in mature seeds, further suggesting that the PC-derived DAG and not the de novo DAG pathway provides most of the substrate for the highly sn-2 HFA-containing TAG accumulation in CL37 seeds.

Figure 8.

 Relative [14C]glycerol backbone labeling of different hydroxy-fatty acid (HFA)-containing species of triacylglycerol (TAG) from developing CL37 seeds.
(a) Accumulation of [14C]glycerol in 0-HFA-TAG (triangles), mono-HFA-TAG (diamonds), and di-HFA-TAG (squares).
(b) Percentage of total backbone labeled TAG species in (a).
(c) Regiochemistry of HFA-containing TAGs. Labeled TAGs from a secondary [14C]glycerol labeling of developing CL37 seeds 9–12 days after flowering (DAF) that had a similar ratio of labeled TAG species (Figure S3) were utilized. Only seeds labeled for 30 min or more contained enough radioactivity in the backbones of HFA-containing TAGs for the analysis.
Data represent the average ± SE for two labeling replicates.

Production of backbone labeled TAGs from the de novo DAG in Figure 6 suggests that di-HFA-TAG must be produced by sn-3 acylation of labeled mono-HFA-DAG with a HFA. Mono-HFA-TAG can be produced by either sn-3 acylation of mono-HFA-DAG with a normal FA or sn-3 acylation of 0-HFA-DAG with a HFA. Regiochemical analysis of TAGs from a second [14C]glycerol labeling of developing CL37 seeds 9–12 days after flowering (DAF) is shown in Figure 8(c). In this experiment the relative labeling of HFA-containing TAG species was similar to that shown in Figure 8(b) (Figure S3). Only labeling time points of 30 min or longer accumulated enough labeled TAG for analysis. For these time points the mono-HFA-DAG regiochemistry (Figure 6c) indicates that the PC-derived DAG flux has not substantially added to the labeled HFA-containing DAG pool by 60 min, so that essentially all the labeled HFA-containing TAG was produced from de novo DAG. The HFA regiochemistry of mono-HFA-TAG of about 60% sn-1/3 and 40% sn-2 is very similar to the mono-HFA-DAG stereochemistry of about 60% sn-1 and 40% sn-2 (Figure 6c), suggesting that labeled mono-HFA-TAG was not produced by sn-3 HFA acylation of 0-HFA-DAG. Additionally, the average 30 and 60 min HFA regiochemistry of di-HFA-TAG supports the hypothesis of sn-3 HFA acylation of the labeled mono-HFA-DAG. Together the high initial synthesis of di-HFA-TAG and the lack of sn-3 HFA mono-HFA-TAG suggests that sn-3 acylation of de novo DAG with a HFA was selective for mono-HFA-DAG over 0-HFA-DAG.


Utilization of PC-derived DAG for TAG synthesis in wild-type Arabidopsis

Through [14C]glycerol labeling of developing Col-0 and fae1 seeds we demonstrate: (i) kinetically, PC is synthesized from de novo DAG at over 14 times the rate of TAG synthesis from de novo DAG, and over 10 times what is required for membrane biogenesis (Figure 2); (ii) relative acyl/glycerol labeling indicates that PC and TAG are synthesized from different pools of DAG (Figure 3); and (iii) the FA composition of bulk DAG is more similar to PC than to de novo DAG (Figure 4b). Since developing seeds accumulate 20 times more TAG than membrane lipids, these results suggest that the major flux of TAG synthesis in Arabidopsis utilizes a de novo DAG, to PC, to PC-derived DAG, to TAG pathway (Figure 1, flux A).

We can more quantitatively discuss the fluxes through each TAG synthesis pathway by considering the accumulation of labeled PC, DAG and TAG through Figure 1 routes A, B and C, over the 60-min time course. At early time points we can assume that most labeled DAG is de novo DAG and labeled PC and TAG are immediate products produced through Figure 1 routes A and C, respectively. Therefore, the relative initial rate of PC/TAG synthesis (∼14/1) indicates a preference for over 93% of de novo DAG to flux through route A over route C. This may represent an upper limit to the flux of DAG through PC into TAG. However, as membranes do not have a large capacity for storage of neutral lipids and DAG can partition into the oil body (Kuerschner et al., 2008), the labeled DAG that accumulates at later time points is probably mixing with the PC-derived bulk DAG pool in the oil body. The labeled DAG that accumulates may be produced through route A as PC-derived DAG or route B by direct incorporation into the oil body. At 60 min of labeling DAG represents <33% of total labeled DAG/PC/TAG in Col-0 (Figure 2). If all the DAG that accumulates at 60 min is through route B and this DAG is not incorporated into PC prior to TAG synthesis, then as this accumulated DAG is incorporated into TAG it puts the lower limit of total DAG flux through PC at about 60%. However, a very large interconversion of DAG and PC is suggested by both the equilibration of the PC and DAG acyl/glycerol labeling by 60 min of labeling (Figure 3a,b), and by the similarity of bulk DAG and PC (Figure 4b). Therefore, the flux of DAG through PC is most likely closer to the upper limit of 93% through route A than the lower limit of 60%.

Flux through PC is a bottleneck in tri-HFA-TAG accumulation

The relative fluxes of TAG synthesis from de novo DAG or PC-derived DAG are especially important for the field of oilseed engineering. Many industrially useful FAs contain oxygenated functional groups and are efficiently removed from microsomal membranes in many plants (Bafor et al., 1991; Banas et al., 1992; Stahl et al., 1995). This may present a problem for accumulating DAG containing two unusual FAs if the transgenic host plant uses PC-derived DAG for TAG synthesis. We analyzed the flux of TAG synthesis in CL37 plants which express RcFAH12 to determine how HFAs are incorporated into TAG in transgenic Arabidopsis seeds. Proposed pathways of HFA-containing TAG synthesis and corresponding bottlenecks in CL37 are summarized in Figure 9. [14C]glycerol labeling demonstrated a reduction of about 50% in total labeled glycerolipid accumulation, which was mostly due to a reduction of about 75% in accumulation of labeled PC, and was not compensated for by an increase in TAG synthesis from de novo DAG (Figure 5). The reduction in PC synthesis may be due to turnover of HFA-containing DAG (Figure 9, flux 6) or PC. The HFAs were rapidly incorporated into the sn-1/sn-2 positions of de novo synthesized mono-HFA-DAG representing around 50% of total DAG (Figure 6b). However, mono-HFA-DAG did not efficiently accumulate in PC (Figure 7) or TAG (Figure 8) and accumulated eight times more slowly than 0-HFA-DAG (Figure 6a). These results suggest that the flux of de novo DAG through PC is a bottleneck for the accumulation of di-HFA-DAG for tri-HFA-TAG synthesis (Figure 9, flux 5).

Figure 9.

 Proposed pathways of hydroxy-fatty acid (HFA)-containing triacylglycerol (TAG) synthesis and bottlenecks in CL37.
The thickness of the arrows indicates relative flux. The red Xs indicates bottlenecks in the pathway in Arabidopsis.
Flux 1: HFAs are produced on sn-2 phosphatidylcholine (PC) by castor bean (Ricinus communis) hydroxylase (RcFAH12).
Flux 2: Most HFAs are incorporated into sn-2 TAG through the PC-diacylglycerol (DAG)-TAG flux.
Flux 3: HFAs are released from PC by the acyl editing cycle and are then available for de novo glycerolipid synthesis.
Flux 4: De novo glycerolipid synthesis produces similar amounts of DAG and mono-HFA-DAG with the HFA at either the sn-1 or sn-2 position. Di-HFA-DAG is not produced.
Flux 5: There is a bottleneck in production of HFA-PC from de novo HFA-DAG.
Flux 6: De novo HFA-DAG is rapidly turned over.
Flux 7: De novo di-HFA-TAG is rapidly produced and then turned over.
Flux 8: Re-utilization of HFAs released from glycerolipid turnover may cause a futile cycle of HFA-DAG synthesis and turnover.
Flux 9: Fatty acids (FAs)/HFAs released from glycerolipid turnover and/or acyl editing may be broken down by beta-oxidation or cause feedback inhibition of FA synthesis such that total oil is reduced in CL37.
Abbreviations: FAS, fatty acid synthesis; M-HFA-, mono-HFA-; D-HFA-, di-HFA-; 0-HFA-, zero-HFA-; G3P, glycerol-3-phosphate; LPA, lyso-phosphatidic acid.

The mechanisms that limit HFA-containing DAG or PC accumulation are unknown. However, the relative 80% drop in mono-HFA-DAG labeling as compared to 0-HFA-DAG, while mono-HFA-PC/0-HFA-PC labeling was approximately constant, suggests turnover of HFA-DAG prior to incorporation into PC (Figure 9, flux 6). Although, if interconversion of de novo DAG to PC and then to PC-derived DAG is very fast, then turnover of HFA-containing PC could reduce the production of PC-derived HFA-DAG. Any mechanism must completely turn over DAG/PC rather than just edit out the HFA, since this would not reduce the total incorporation of [14C]glycerol into DAG or PC. Turnover of de novo DAG containing HFAs may generate a futile cycle of DAG synthesis and breakdown (Figure 9, fluxes 6 and 8).

PC-derived DAG allows the accumulation of sn-2 HFA in CL37

We demonstrate that mono-HFA-DAG accumulates at ∼8% of de novo DAG and contains ∼40% sn-2 HFA (Figure 6). These results suggest that ∼3% of de novo DAG accumulates with sn-2 HFA. However, mature CL37 seeds accumulate ∼67% of the total HFA in the sn-2 position of TAG, which corresponds to ∼35% of total DAG containing a sn-2 HFA (van Erp et al., 2011). These results imply that de novo DAG is not the source of HFA-DAG utilized for (sn-2)-HFA-TAG synthesis. Together, the ratio of de novo DAG fluxes into PC/TAG of 4/1 in CL37, and the fact that (sn-2)-18:1-PC is the substrate of RcFAH12, suggest that most (sn-2)-mono-HFA-DAG is produced by the PC-derived DAG pathway after the action of RcFAH12 (Figure 9, fluxes 2 and 1, respectively).

HFA specific TAG synthesis and subsequent remodeling/turnover may further limit the accumulation of HFA in CL37

Interestingly, at very early labeling time points backbone labeled di-HFA-TAG accounted for more labeled TAG than 0-HFA-TAG plus mono-HFA-TAG (Figure 8b). However, di-HFA-TAG accumulated at a much slower rate than other TAG species (Figure 8a). Additionally, the regiochemical analysis of de novo synthesized HFA-containing TAGs suggests that sn-3 acylation with a HFA was specific to mono-HFA-DAG and not 0-HFA-DAG (Figure 8c). These results suggest that di-HFA-TAG may be specifically produced and subsequently turned over or remodeled (Figure 9, flux 7). Arabidopsis PDAT has been shown to utilize HFA substrates more specifically than non-HFA substrates (Stahl et al., 2004), and may be part of an endogenous mechanism to remove oxygenated FA from the ER membrane. Similar changes in labeled HFA-containing TAG species from castor endosperm demonstrate that tri-HFA-TAG is remodeled in microsomes but is stable in oil bodies (Mancha and Stymne, 1997). Rapid conversion of mono-HFA-DAG to di-HFA-TAG by HFA selective AtPDAT and then TAG remodeling in CL37 may explain both the reduction in mono-HFA-DAG accumulation in the de novo DAG pool and the rapid synthesis/turnover of di-HFA-TAG. Therefore, efficient incorporation of HFA-TAG into the oil body may also be required to accumulate high levels of HFA in transgenic oilseeds. Identification of the enzymes involved in glycerolipid turnover or remodeling may represent new targets for oilseed engineering.

Engineering of unusual FAs into transgenic oilseeds

The differences in the TAG biosynthetic pathways between Arabidopsis (utilization of PC-derived DAG) and castor (utilization of de novo DAG; Bafor et al., 1991), may be a major factor limiting the accumulation of tri-HFA-TAG (and total TAG) in CL37 seeds. CL37 contains around 30% less oil/dry weight than the fae1 parental line (Figure S2). Therefore, the FAs released from turnover of HFA-containing lipids (both HFA and non-HFA) may either be broken down through beta-oxidation (Moire et al., 2004) or reduce the total rate of fatty acid synthesis through feedback inhibition (Shintani and Ohlrogge, 1995) (Figure 9, flux 9). However, if castor homologs of DAG sn-3 acyltransferases (DGAT or PDAT) are co-expressed with RcFAH12 the amount of HFA in seed TAG is increased from about 17% to 25–28%, and the reduced oil phenotype is mostly recovered (Burgal et al., 2008; van Erp et al., 2011). Castor DGAT/PDAT have been shown to be selective for both HFA-containing DAG and acyl substrates (Burgal et al., 2008; van Erp et al., 2011), which may specifically capture de novo mono-HFA-DAG prior to turnover, increasing the flux of de novo DAG directly to TAG. If the castor DGAT/PDAT can increase the flux of HFA-containing TAG into the oil body it may be less available to remodeling and/or turnover lipases (Mancha and Stymne, 1997). Additionally, reduced turnover of HFA-containing glycerolipids may also alleviate the reduced oil phenotype by limiting either FA beta-oxidation or feedback inhibition of FA synthesis. The co-expression of RcFAH12 with HFA-specific DAG sn-3 acyltransferases increased the amount of di-HFA-TAG to ∼20%; however, tri-HFA-TAG was still <3% of total TAG (van Erp et al., 2011). These results imply that production of di-HFA-DAG through the predominantly PC-derived DAG pathway of TAG synthesis in Arabidopsis is a major limiting factor in CL37 based lines.

Accumulation of other PC-produced unusual FAs mostly at the sn-2 position of TAG has also been demonstrated in soybeans and Arabidopsis expressing the Calendula officinalis and Momordica charantia FA conjuases, respectively (Cahoon et al., 2006). We recently demonstrated that >95% of de novo DAG in soybeans also fluxes through PC prior to TAG synthesis (Bates et al., 2009). Therefore, the flux of DAG through PC for TAG synthesis may represent a bottleneck for engineering many oilseed plants with different kinds of unusual FAs as well as HFAs. Future engineering strategies that focus on replacing endogenous de novo DAG synthesis enzymes with castor HFA-specific homologs may be able to produce greater quantities of di-HFA- de novo DAG in transgenic oilseeds. However, the flux of de novo DAG through PC may need to be circumvented in transgenic plants utilizing a PC-derived DAG pathway to reduce the HFA-containing de novo DAG turnover and make more de novo DAG available for direct conversion to TAG. The complexity of different mechanisms of TAG synthesis (Li-Beisson et al., 2010) indicates that a single unusual FA engineering strategy may not apply to all host plants or unusual FA products. Further determination of the pathways of unusual FAs containing DAG production in additional plant species that accumulate unusual FAs may provide clues for future engineering of oilseed crop plants.

Oilseed engineering has been limited by uncertainties not only about the genes/enzymes involved in biosynthesis of TAG but also about the relative fluxes through competing metabolic pathways to produce TAG. The differences in fluxes between TAG synthesis pathways in different plant species (such as Arabidopsis and castor) add additional complexity to our attempts to engineer the FA composition of one plant to be similar to another. Here we demonstrate that the common oilseed model, Arabidopsis, can utilize PC-derived DAG for as much as 93% of total TAG synthesis. However, engineering Arabidopsis to produce HFAs in seeds causes a decrease in total glycerolipid accumulation. This decrease is mostly due to a reduction in PC synthesis from de novo DAG, and leads to reduction in total seed oil. The reduction in PC synthesis may be due to a bias against incorporation of HFA-containing de novo DAG into PC and cause a futile cycle of HFA-DAG synthesis and breakdown. Utilization of PC-derived DAG after synthesis of HFAs on PC allows accumulation of HFAs mostly at the sn-2 position of TAG. Therefore, the flux of DAG through PC may represent a key bottleneck in the production of tri-HFA-TAG in transgenic oilseeds utilizing a PC-derived DAG pathway of TAG synthesis.

Experimental Procedures


Plants were grown at 22–24°C under constant light of ∼120–160 μmol m−2 sec−1. Siliques of 9–10 DAF were chosen for labeling experiments and lipid mass analysis (siliques of 9–12 DAF were utilized for Figures 8c and S4). These experiments used: HPLC grade organic solvents, Fisher Scientific (; thin layer chromatography (TLC) plates, EMD silica gel 60 20 × 20 cm glass plates (EMD, and Whatman Partisil® K6 silica gel 60 Å 20 × 20 cm glass plates (Whatman,; [14C(U)]glycerol (specific activity 150 mCi mmol−1), American Radiolabeled Chemicals, Inc. (; Ecoscint liquid scintillation cocktail, National Diagnostics (; butylated hydroxytoluene (BHT), 2′,7′-dichlorofluorescein, Primuline, Phospholipase C (Bacillus cerus), and TAG lipase (Rhizomucor miehei), Sigma-Aldrich (

[14C]glycerol labeling of developing seeds

Individual time-course labelings were done in a single container removing aliquots of seed/medium at each time point. Replicates were radiolabeled separately. Seeds of 50 siliques were combined per single labeling time course. Developing seeds were harvested directly into 5 mm MES pH 5.8, 0.5% sucrose, 0.5× MS salts on ice. After harvest the seeds were pre-incubated in a lighted water bath at 23°C with constant shaking for 20 min. The labeling was started by removing all the medium and replacing it with the same medium as above but also containing ∼0.33 mm [14C]glycerol. At 3, 6, 10, 30 and 60 min an aliquot of seeds was removed and quenched in 85°C isopropanol + 0.01% BHT for 10 min. Lipids were extracted with a modified protocol from the Kansas Lipidomics Research Center ( All solvent volumes were doubled and 0.88% KCl was used for the phase separation instead of water.

Analysis of radiolabeled lipids

All radiolabel measurements were normalized to micrograms of chlorophyll (Arnon, 1949) per time point aliquot to correct for different number of seeds in each sample. Lipids separated by TLC were collected and transmethylated on the sillica in 0.5 ml 2.5% sulfuric acid in methanol. One milliliter of hexane and 1 ml of H2O were added to separate the fatty acid methyl ester (FAME) and glycerol fractions. The entire hexane phase and aqueous/silica phase were liquid scintillation counted separately as the acyl and backbone fractions, respectively, on a Tri-CARB® liquid scintillation analyzer (Packard Instrument Company).

Lipid class analysis

For Col and fae1, neutral lipids were separated by TLC on EMD plates by a single development in hexane/diethyl ether/acetic acid, (70:30:1, v/v/v). Total polar lipids were eluted from the origin of the neutral lipid TLC and further separated by TLC on Whatman plates as in Bates et al. (2007). The HFA-containing and non HFA-containing neutral lipids from CL37 seeds were separated on EMD plates with a double development system in different mixtures of chloroform/methanol/acetic acid (v/v/v). The first development was to 12 cm from the bottom of the plate in 97:3:0.5. Plates were dried in a vacuum for 15 min then the second development was to 19 cm in 99:0.5:0.5. Polar lipids from CL37 seeds were separated as above. All lipids were identified by co-migration with lipid standards after staining with 0.005% primuline in 80% acetone and visualization under UV light. A TLC example and further details about identification are given in Figure S4. All TLC solvent mixtures contained ∼0.005% BHT.

Molecular species analysis of radiolabeled DAG

Molecular species of radiolabeled DAG were separated as 1,2-diacyl-3-acetyl-glycerol (acetylated DAG) derivatives as in Bates et al. (2009). Lipid bands were identified by staining with 0.1% 2′,7′-dichlorofluorescein in 95% methanol and visualization under UV light. All lipid bands were collected and eluted from the AgNO3 silica by vortexing in 4 ml of hexane/isopropanol/H2O (6:4:0.5, v/v/v). Two milliliters of 1 m KCl was added to force a phase separation and precipitate AgCl2. The organic phase was collected and the aqueous phase was back extracted with hexane/isopropanol (7:2, v/v). The collected acetylated-DAG molecular species were dried under N2 and transmethylated as above. The entire aqueous phase was scintillation counted for backbone label as above and the acyl composition of the FAME was determined by gas chromatography as in Table S1 (supplement for quantification of lipid mass).

Regiochemistry of HFA-containing lipids

The HFA-containing lipids were separated from other lipids by the HFA neutral lipid TLC system and eluted from the silica as for the polar lipids above. The HFA-containing DAGs and TAGs were digested to monoacylglycerols (MAG) with Rhizomucor miehei lipase as in Cahoon et al. (2006). The HFA-MAG and MAG were separated by TLC with EMD plates developed fully in hexane/diethyl ether/acetic acid (35:70:1.5, v/v/v), and radioactivity in the backbones determined as above. Stereochemistry of radiolabeled mono-HFA-DAG was calculated as: %HFA at sn-2 of mono-HFA-DAG = [14C]HFA-MAG/([14C]HFA-MAG + [14C]MAG) × 100; the %HFA at sn-1 of mono-HFA-DAG = 100 − % at sn-2. The same calculation was used for mono-HFA-TAG with sn-1/3 analogous to mono-HFA-DAG sn-1. For di-HFA-TAG, the % HFA-MAG = the % of TAG species containing HFA at sn-2 and sn-1/3. The % MAG = the % of TAG species containing HFA at sn-1 and sn-3.

Separation of radiolabeled PC and HFA-PC

Phosphatidylcholine was collected from an aliquot of the CL37 extracts by TLC on EMD plates developed fully in chloroform/methanol/acetic acid/H2O (75:25:4:4, v/v/v/v) and eluted from the silica as above. Phosphatidylcholine was converted to diacylglycerols by digestion with phospholipase C (B. cerus) (Christie, 2003). The DAG and HFA-DAG products were separated by TLC by developing fully in hexane/diethyl ether/acetic acid (35:70:1.5, v/v/v).


We would like to thank Dr Mike Pollard for critically reading the manuscript and helpful discussions of the results. This work was supported by the US National Science Foundation (grant no. DBI–0701919) and by the Agricultural Research Center at Washington State University.