Conjugated Fatty Acid Synthesis

Background: Momordica conjugase produces α-eleostearic acid, whereas Punica or Trichosanthes conjugases produce punicic acid. Results: Momordica conjugase residues 111 and 115 affect total eleostearic acid accumulation levels in transgenic Arabidopsis. Conclusion: Interactions of Momordica side chains 111 and 115 influence α-eleostearic acid versus punicic acid formation. Significance: Momordica conjugase mutant accumulated punicic acid in addition to α-eleostearic acid. Conjugated linolenic acids (CLNs), 18:3 Δ9,11,13, lack the methylene groups found between the double bonds of linolenic acid (18:3 Δ9,12,15). CLNs are produced by conjugase enzymes that are homologs of the oleate desaturases FAD2. The goal of this study was to map the domain(s) within the Momordica charantia conjugase (FADX) responsible for CLN formation. To achieve this, a series of Momordica FADX-Arabidopsis FAD2 chimeras were expressed in the Arabidopsis fad3fae1 mutant, and the transformed seeds were analyzed for the accumulation of CLN. These experiments identified helix 2 and the first histidine box as a determinant of conjugase product partitioning into punicic acid (18:3 Δ9cis,11trans,13cis) or α-eleostearic acid (18:3 Δ9cis,11trans,13trans). This was confirmed by analysis of a FADX mutant containing six substitutions in which the sequence of helix 2 and first histidine box was converted to that of FAD2. Each of the six FAD2 substitutions was individually converted back to the FADX equivalent identifying residues 111 and 115, adjacent to the first histidine box, as key determinants of conjugase product partitioning. Additionally, expression of FADX G111V and FADX G111V/D115E resulted in an approximate doubling of eleostearic acid accumulation to 20.4% and 21.2%, respectively, compared with 9.9% upon expression of the native Momordica FADX. Like the Momordica conjugase, FADX G111V and FADX D115E produced predominantly α-eleostearic acid and little punicic acid, but the FADX G111V/D115E double mutant produced approximately equal amounts of α-eleostearic acid and its isomer, punicic acid, implicating an interactive effect of residues 111 and 115 in punicic acid formation.

The crystal structures of the soluble desaturases from castor (Ricinus communis) and the English Ivy (Hedera helix) have given valuable insights into novel catalytic activities and speci-ficities of the soluble desaturases (15,16 ), but the lack of threedimensional structural information has hampered our understanding of the structure-function relationship of membranebound desaturases and FAD2-like divergent enzymes. Structure-function relationships in this class of enzymes have been explored by site-directed mutagenesis, domain swapping, and recently by directed evolution in vitro (17)(18)(19)(20)(21)(22). Amino acid residues adjacent to the conserved histidine motifs have been shown to contribute to reaction partitioning, and only a handful of amino acid positions have been shown to define substrate specificity (14,21). For instance, studies on Arabidopsis FAD2 and castor fatty acid hydroxylase identified as few as four amino acid positions that exert control over reaction partitioning between hydroxylase and desaturase activity (21,22).
Evolution of FADXs from FAD2s coupled with the high degree of functional plasticity exhibited by FADXs suggests that identification of key amino acid residues that define their functionality should be feasible. In this context, we sought to identify domains and residues of FADX that are vital for its conjugase activity. Thus, domains of the Momordica FADX were swapped with equivalent domains from the Arabidopsis FAD2, and the resulting chimeras were heterologously expressed in transgenic Arabidopsis seeds. Our results identify two amino acid locations in Momordica FADX, 111 and 115, that play key roles in partitioning of catalytic outcome between the formation of ␣-ESA and its isomer, punicic acid.
Site-directed Mutagenesis-Modified FADX genes containing single or multiple mutations were constructed using appropriate templates by overlap-extension PCR as described above. The primers used for site-directed mutagenesis are shown in supplemental Table S1.
Arabidopsis Growth and Transformation-The plants were grown at 22°C under a 16-h day/8-h night photoperiod. Plants were transformed by employing the floral dip method (29) using the Agrobacterium tumefaciens strain GV3101. Individual T1 seeds carrying the transgenes were identified based on DsRed2 expression (30).
Fatty Acid Analysis-Fatty acid methyl esters (FAMES) were prepared from seeds (single or three seeds) by incubation with 30 l of 0.2 M trimethylsulfonium hydroxide in methanol (31). After drying the samples under a nitrogen stream, the samples were resuspended in hexane. FAMES were analyzed with the use of either a Hewlett-Packard 6890 gas chromatograph-flame ionization detector (Agilent Technologies, Santa Clara, CA) fitted with 60 m ϫ 250 m ϫ 0.25 m SP-2340 capillary column (Supelco, Bellefonte, PA), or an HP7890A gas chromatographmass spectrometer (Hewlett-Packard, Palo Alto, CA) equipped with a HP5975C selective detector (GC/MS) and a 30 m ϫ 250 m ϫ 0.25 m HP-Innowax capillary column (Agilent Technologies). The injector was maintained at 225°C, and the oven temperature was varied from 100 to 240°C at 15°C/min, then held at 240°C for 6 min. GC peaks were identified by comparison with authentic standards and by characterization via GC/MS. CLN methyl esters were routinely identified by the presence of an abundant molecular ion at m/z ϭ 292 and specific isomers identified by GC retention times compared with an authentic standard mixture of CLN FAMEs. The standard for CLN was obtained from seeds that accumulate different geometrical isomers of C18-conjugated trienoic fatty acids, namely P. granatum, M. charantia and C. bignonoides. P. granatum seeds accumulate punicic acid (18:3 ⌬ 9cis,11trans,13cis ); M. charantia seeds accumulate ␣-eleostearic acid (18:3 ⌬ 9cis,11trans,13trans ); and C. bignonoides seeds accumulate catalpic acid (18:3 ⌬ 9trans,11trans,13cis ). The percentage values indicating the composition in the fatty acid samples reported are mean values based on at least three individual measurements.

Construction of FADX-FAD2
Chimeras-To identify structural elements that are functionally important in FADX and FAD2, we constructed chimeras comprising domains of Momordica FADX fused to those of Arabidopsis FAD2. These two enzymes share ϳ62% sequence identity. To construct chimeras in a rational fashion we first generated a topology model for Momordica conjugase based on online computational algo- rithms and previously established topological models (14, 32) (Fig. 2). There are no experimental data available for the topology of either FAD2 or FADX, and the structural models available for other membrane-bound desaturases are derived from their primary structure and in vivo activities in yeast (32). Based on the hydropathy plots, the integral membrane fatty acid desaturases and desaturase-like enzymes are predicted to share the same membrane topology (four transmembrane helices and two hairpin domains) and overall three-dimensional structures. Therefore, replacing a particular domain of one of these enzymes with the corresponding domain from the another enzyme is not expected to disrupt the overall structure, and quantification of the fatty acid accumulation would help us to identify domains containing amino acids that control particular properties. To implement this approach seven sites were selected as domain borders that isolate individual helices or hairpin domains within Momordica FADX that could be replaced with the corresponding FAD2 sequence as shown in Fig. 2.
Functional Characterization of FADX-FAD2 Chimeras in Arabidopsis fad3fae1-Because the expression of FADX from different origins in yeast results in accumulation of CLNs to Ͻ1% (8, 11, 13), we were concerned that potential losses of enzymatic activity resulting from the mapping experiments employing chimeric enzymes would preclude the quantitative product analysis. We therefore chose to express the chimeras and the wild-type (WT) enzymes in Arabidopsis under the control of the seed-specific phaseolin promoter (28). To assess conjugase function, these enzymes were transformed into fad3fae1 double mutant line, which lacks the activity of both the endo-plasmic reticulum -3 desaturase (FAD3) and the fatty acid elongation 1 (FAE1)-condensing enzyme, and contains elevated level of the FADX substrate, linoleic acid (ϳ52% of the total fatty acids, compared with ϳ19% in WT) (33).
Previous studies reported Momordica FADX-transformed fad3fae1 seeds accumulate 13% total ESA (34). Consistent with this, FADX-transformed fad3fae1 T1 seeds in this study showed 9.9% of total ESA accumulation with an isomer distribution of 7.7% ␣-ESA, 0.2% punicic acid, and 2.1% ␤-ESA. Chimeras 1-4 and 8 produced lower total ESA (Ͻ7.5%) compared with WT-FADX. Chimeras 5 and 6 accumulated ϳ11% total ESA, whereas Chimera 7 accumulated ϳ17% total ESA (Fig. 3A  and supplemental Table S2). Intriguingly, Chimera 7 displayed a different product distribution of CLN isomers relative to that of WT-FADX in accumulating 8.9% of punicic acid in addition to 7.8% ␣-ESA (Figs. 4 and 5A). In the absence of protein quantitation data we are unable to determine whether changes in ESA accumulation result from changes in enzyme activity or changes in enzyme levels.
Six Amino Acids Located in Transmembrane Helix 2 Play Important Role in Determining Desaturase and Conjugase Functionalities-Chimeras 6 and 7 are fusions between the N terminus of FAD2 and the C terminus of FADX, the difference being that Chimera 7 contains the first two helices and first portion of the tripartite conserved histidine motif 1 (13) of FAD2, whereas Chimera 6 only contains the first helix of FAD2, the balance of the protein consists of FADX. That Chimera7 produced a mixture of ␣-ESA and punicic acid, whereas Chi-mera 6 produced primarily ␣-ESA (like Momordica) suggests that both the ability to produce punicic acid and the increase in total ESA accumulation must be specified by the six amino acid differences between FAD2 and FADX in helix 2 and first histidine box (Fig. 2). To test this hypothesis, the FADX 6 mut was engineered consisting of V97L, F100A, A104C, M109I, G111V, and D115E (using Momordica FADX numbering), in which the six WT-FADX residues were replaced by the corresponding residues from WT-FAD2.
Analysis of the transformed seeds confirmed that the FADX 6 mut was functionally similar to Chimera 7 in producing 20.1% total ESA in fad3fae1 lines (Fig. 5A and supplemental Table S4); and 7.8% total ESA in fad2fae1 lines ( Fig. 5B and supplemental Table S5). Further, the FADX 6 mut like Chimera 7 produces both punicic and ␣-ESA (9.5% punicic acid and 9.5% ␣-ESA in fad3fae1 lines, and 4.1% punicic acid and 3.3% ␣-ESA in fad2fae1 lines). Expression of the FADX 6 mut in fad2fae1 Arabidopsis background shows that it also produces 7.1% of 18:2 ⌬ 9cis,12cis , i.e. it has a desaturase activity that is responsible for the production of substrate for its own conjugase activity.
Substitutions of G111V and D115E in Momordica FADX Are Sufficient for Significant Punicic Acid Accumulation-To determine whether any single amino acid of the six substitutions in FADX 6 mut has a major impact on product partitioning between ␣-ESA and punicic acid, we constructed a series of six drop-out FADX 5 mut genes in which each of the six amino acid substituents was individually converted back to its FADX equivalent. The constructs were transformed into both fad3fae1 and fad2fae1 Arabidopis backgrounds. Analysis of the transformed seeds showed that four of the six dropout mutants, i.e. FADX 6 mut-V97L (contains all of the six FAD2 substitutions except V97L), FADX 6 mut-F100A, FADX 6 mut-A104C, and FADX 6 mut-M109I had conjugase activities similar to that of the FADX 6 mut (Figs. 5 and 6), indicating that substitutions at positions 97, 100, 104, and 109 do not contribute to the increase in ESA accumulation or the product distribution of FADX 6 mut enzyme.
Although FADX 6 mut-D115E showed a modest reduction of ESA accumulation relative to the FADX 6 mut (15.0% versus 20.1%, supplemental Table S4), punicic acid accumulation was decreased by ϳ75% compared with FADX 6 mut (2.1% versus   Tables S4 and S5). This result suggests that position 115 also plays a role in imparting Trichosanthes-like conjugase activity to WT-FADX. However, the single mutant D115E displayed activity like WT-FADX, with respect to total ESA and punicic acid accumulation. This implies that D115E alone cannot enhance ESA accumulation.
We therefore investigated the possibility that D115E and G111V substitutions in combination are responsible for additional ESA accumulation by analyzing the expression of a FADX G111V/D115E double mutant. Analysis of the transformed seeds show that the conjugase double mutant produced 21.2% of total ESA compared with 9.9% for WT-FADX in fad3fae1 and 10.8% versus 1.6% in fad2fae1, respectively (supplemental Tables S4 and S5). Besides increased ␣-ESA, the double mutant resulted in significant punicic acid accumulation (10.5% punicic acid and 9.7% ␣-ESA in fad3fae1; and 5.8% punicic acid and 4.2% ␣-ESA in fad2fae1, Fig. 5). This result suggests that just two amino acid changes, G111V/D115E, are sufficient to impart significant Trichosanthes-like conjugase  functionality to the Momordica conjugase. Further, the double mutant had increased FAD2-like functionality and produced 6.4% of 18:2 ⌬ 9cis,12cis compared with 1.7% for WT-FADX ( Fig.  6 and supplemental Table S5).

DISCUSSION
Through our approach of analyzing the products of Momordica FADX-Arabidopsis FAD2 chimeric enzymes by expression in various Arabidopsis mutant backgrounds we identified Chimera 7, which accumulated higher ESA relative to WT-FADX. In this analysis our interpretation is restricted to accumulation of total ESA in transformed seeds because no determination of differential fatty acid turnover was made. Further analysis revealed that the increase of ESA upon expression of Chimera 7 was associated with substantially elevated punicic acid production and equivalent levels of ␣-ESA production relative to WT-FADX. Subsequent targeted mutagenesis identified residues 111 and 115 as the two key residues that influence both ESA accumulation and reaction partitioning. Expression of FADX G111V in fad3fae1-transformed seeds resulted in a 14-fold increase in punicic acid and 2-fold increase in ␣-ESA accumulation relative to WT-FADX. Although FADX D115E showed similar ESA accumulation as WT-FADX, the double mutant FADX G111V/D115E led to a 50-fold increase in punicic acid accumulation in the same background. This result implies that the difference with respect to partitioning of conjugase product into its geometrical isomers is directly determined by the identity of the amino acid residues at positions 111 and 115. Further, the bulkier residues in the G111V/D115E double mutant appear to interact because we observed an increase in punicic acid formation but no change in total ESA accumulation (relative to G111V). We further tested the substitution of Gly at position 111 with a range of side chains of different bulk; i.e. the larger Leu and Phe, and the smaller Ala compared with Val. We made each of these mutations with Asp or Glu at position 115. Although these changes had little effect on product partitioning, each of these substitutions resulted in a decrease in ESA accumulation relative to G111V and G111V/ D115E (supplemental Tables S4 and S5), suggesting that Val has an optimal side chain bulk at position 111.
Conjugases from Trichosanthes and Punica accumulate punicic acid in WT-Arabidopsis seeds when heterologously expressed under the control of napin promoter (13). The average amounts of punicic acid in Trichosanthes and Punica conjugase transformants have been reported to be 3.5% and 2.3%, respectively, with maximal amounts reaching 10.2% and 4.4%, respectively. Interestingly, in transformed fad3fae1 Arabidopsis seeds, besides 9.7% of ␣-ESA, the expression of FADX G111V/D115E results in the accumulation of an additional 10.5% of punicic acid. According to the predicted structural model (Fig. 2), residues 111 and 115 are located in or adjacent to the first of the three conserved histidine boxes. Specifically, position 115 immediately follows the first histidine residue of the histidine cluster, and residue 111 would therefore be positioned one helical turn before residue 115, suggesting that these side chains are adjacent to one another and bracket a conserved histidine within the active site. It is interesting to note that the other four residues in helix 2 that differ between FAD2 and FADX which do not influence functionality are not predicted to occur near the active site histidine cluster. The absence of a crystal structure for any membrane-bound desaturase represents a major impediment to understanding of the mechanisms underlying these results.
Various chemical mechanisms have been proposed to describe the biological origin of conjugated trienoic acids. It has been shown that during conversion of linoleic acid to calendic acid in C. officinalis developing seeds, hydrogen is abstracted sequentially from the C8 and C11 methylene groups that flank the cis-⌬9 double bond of linoleic acid, a mechanism known as "1,4-desaturation" (36,37). The production of ␣-ESA is proposed to involve the removal of hydrogens from the C11 and C14 methylene groups that flank the cis-⌬12 double bond of linoleic acids (34). In contrast, little mechanistic information is presently known about the formation of geometrical isomers of ESA such as punicic acid and ␣-ESA.
In previous work on the evolutionarily distinct acyl-ACP desaturases, for which the geometry of the substrate binding cavity is defined from crystal structures, the production of allylic hydroxy fatty acids was rationalized by a relative shift of the fatty acid substrate with respect to the active site Fe-O oxidant. The boomerang shape of the substrate binding cavity imposes an eclipsed conformation on the acyl chain that causes the pro-R hydrogens to be presented to the active site oxidant bound to the diiron active site (15,16). Although the amino acid sequences of acyl-ACP desaturases and integral membrane FAD2 desaturases differ, the common chemical imperative imposed by the stereochemistry of desaturation implies that the shapes of the substrate binding cavities relative to the active site iron oxidant species will be similar.
Based on the data presented here, and by analogy to mechanisms proposed for the acyl-ACP desaturases (38) involving shifting the register of the bound fatty acid substrate relative to the active site, the formation of either punicic acid or ␣-ESA can perhaps be rationalized in terms of distinct proposed substrate binding modes relative to the position of the oxidant and the envisaged bend in the substrate binding cavity. It is likely that punicic acid results from the placement of C12 and C13 of the 18:2 ⌬ 9cis,12cis substrate at the bend in the binding cavity wherein the bulkier residues constrain the product in a cis configuration at C13, whereas ␣-ESA is produced when the substrate 18:2 ⌬ 9cis,12cis binds deeper into the binding cavity wherein the smaller residues allow the product to adopt a more stable anti-conformation that results in pro-S hydrogen removal at C14 and thence trans-double bond formation. In the case of the single mutant, G111V, steric bulk presented by a larger Val residue causes the substrate to bind two carbons less deeply in the active site leading to conversion of 18:2 ⌬ 9cis,12cis to punicic acid, and in the case of the double mutant, G111V/ D115E, conversion of 18:2 ⌬ 9cis,12cis to punicic acid is much enhanced due to larger steric bulk presented by the presence of larger residues Val and Glu near the active site.
Position 111 was also found to be crucial for imparting FAD2-like desaturase functionality to Momordica conjugase (Fig. 6). In fad2fae1 lines, the single mutant G111V produced 3-fold more 18:2 ⌬ 9cis,12cis and 8-fold more ESA compared with WT-FADX. Perhaps this higher FAD2-like desaturase function attributable to the G111V mutation allows the enzyme to perform the two consecutive reactions to produce ESA from 18:1 without release of the product of the desaturation reaction. Such a mechanism could help explain higher ESA accumulation in seeds transformed with G111V. As described above, perhaps we can rationalize the formation of 18:2 ⌬ 9cis,12cis and 18:2 ⌬ 9cis,12trans in terms of distinct binding modes of the 18:1⌬ 9cis substrate relative to the position of the active site oxidant and the bend in the active-site cavity. In the case of less steric hindrance in the active site the more stable product 18:2⌬ 9cis,12trans can form.
In other FAD2-related work, subtle changes near the active site have been shown to play a key role in influencing desaturation and hydroxylation product partitioning (21,22). We therefore performed alignments of the amino acid sequences of helix 2 and the first histidine box of various FAD2s and FAD2-like enzymes that modify the C12 position of fatty acids (Fig. 7). Our sequence analysis shows that position 111 is mostly conserved as a hydrophobic residue Val in FAD2s and FAD2-like enzymes. It is replaced by a smaller hydrophobic residue Gly in Momordica; and by Ala in coco plum (Chrysobalanus icaco) and gopher apple (Licania michauxii) conjugases. Coco plum and gopher apple conjugases are closely related and have shown to accumulate ␣-ESA in soybean somatic embryos (39). The epoxygenase contains an Ile at this position. Position 115 is mostly conserved as Glu in FAD2s and FAD2-like enzyme. It is replaced by a smaller residue Asp in Momordica, coco plum, and gopher apple conjugases. Notably, most of the FAD2 desaturase-like variant enzymes that are bifunctional and possess residual FAD2-like desaturase activity (Trichosanthes conjugase, Punica conjugase, Lesquerella hydroxylase, castor hydroxylase, Claviceps hydroxylase, Crepis acetylenase) contain Val and Glu at positions 111 and 115, respectively (Fig. 7). Although the identities of amino acids at positions 111 and 115 are conserved as Val and Glu in tung conjugase and Crepis alpina acetylenase, tung conjugase has been shown to produce 18:2 ⌬ 9cis,12trans , and not 18:2 ⌬ 9cis,12cis , the FAD2-like desaturase product; and C. alpina acetylenase preferentially produces 18:2 ⌬ 9cis,12trans (10,40,41). This suggests that in the case of tung conjugase and C. alpina acetylenase some additional changes near the active site may be necessary for introducing preferential FAD2-like desaturase activity.
In summary, the data presented here support the view that a small number of residues at key positions in the vicinity of the active site can have a profound impact on catalytic outcome of the membrane-bound desaturase-like enzymes. These studies identified Momordica conjugase mutants, G111V and G111V/ D115E, which provide an increased understanding of the specificity determinants of fatty acid conjugases; in addition, their expression results in the accumulation of substantially elevated levels of ESA in Arabidopsis seeds relative to the WT-FADX conjugase. Such variants may have utility for the accumulation of industrially relevant levels of ESA in crop plants.