FATTY ACID DESATURASE4 of Arabidopsis encodes a protein distinct from characterized fatty acid desaturases

Authors


(fax 517 353 9334; e-mail lastr@msu.edu).

Summary

Polar membrane glycerolipids occur in a mixture of molecular species defined by a polar head group and characteristic acyl groups esterified to a glycerol backbone. A molecular species of phosphatidylglycerol specific to chloroplasts of plants carries a Δ3-trans hexadecenoic acid in the sn-2 position of its core glyceryl moiety. The fad4-1 mutant of Arabidopsis thaliana missing this particular phosphatidylglycerol molecular species lacks the necessary fatty acid desaturase, or a component thereof. The overwhelming majority of acyl groups associated with membrane lipids in plants contains double bonds with a cis configuration. However, FAD4 is unusual because it is involved in the formation of a trans double bond introduced close to the carboxyl group of palmitic acid, which is specifically esterified to the sn-2 glyceryl carbon of phosphatidylglycerol. As a first step towards the analysis of this unusual desaturase reaction, the FAD4 gene was identified by mapping of the FAD4 locus and coexpression analysis with known lipid genes. FAD4 encodes a predicted integral membrane protein that appears to be unrelated to classic membrane bound fatty acid desaturases based on overall sequence conservation. However, the FAD4 protein contains two histidine motifs resembling those of metalloproteins such as fatty acid desaturases. FAD4 is targeted to the plastid. Overexpression of the cDNA in transgenic Arabidopsis led to increased accumulation of the Δ3-trans hexadecanoyl group in phosphatidylglycerol relative to wild type. Taken together these results are consistent with the hypothesis that FAD4 is the founding member of a novel class of fatty acid desaturases.

Introduction

Polar membrane glycerolipids are the predominant lipid building blocks of biological membranes. They are classified based on a polar head group in the sn-3 position of a glyceryl backbone. The two hydroxyl groups at the sn-1 and sn-2 glyceryl carbons are esterified to fatty acyl groups of different carbon chain length (typically 16–18 carbons) with variable numbers of double bonds (typically one-to-three cis double bonds per chain). The resulting combinatorial biochemistry leads to a plethora of possible molecular species in each glycerolipid class. However, not all possible molecular species are found in nature as a consequence of the specificity of the involved enzymes or as dictated by the biological or physiological functions of particular lipid molecular species. Photosynthetic membranes of plants and algae contain an interesting example of a specific type of phosphatidylglycerol (PtdGro) (Dubacq and Tremolieres, 1983). It is characterized by the presence of Δ3-trans hexadecenoic acid (16:1Δ3t; 16:1, number of carbons : number of double bonds; Δ3, trans double bond between carbon 3 and carbon 4 counting from the carboxyl end) in the sn-2 position of the glycerol backbone. The structure of this PtdGro is shown in Figure S1. The exclusive presence of 16:1Δ3t in chloroplast PtdGro initially gave rise to speculation that this molecular species is critical to photosynthesis as reviewed in McCourt et al. (1985).

The Arabidopsis thaliana (Arabidopsis) fad4-1 mutant (originally designated fadA) was discovered over 20 years ago because it lacks 16:1Δ3t and has elevated levels of palmitic acid (16:0) in leaf PtdGro (Browse et al., 1985). Surprisingly, this mutant had no major growth defects or photosynthetic impairments (McCourt et al., 1985). Despite the fact that fad4-1 was the first fatty acid mutant described for Arabidopsis, the desaturase responsible for the introduction of the Δ3-trans double bond into 16:0 at the sn-2 position of chloroplastic PtdGro remains elusive. This fact is especially noteworthy because of the large number of plant fatty acid desaturases and fatty acid modifying enzymes that have been identified and extensively studied (Shanklin and Cahoon, 1998; Sperling et al., 2003; Napier, 2007).

Fatty acid desaturases can be integral membrane proteins or soluble and are characterized by specifically spaced conserved histidine motifs that vary slightly in soluble versus membrane-bound fatty acid desaturases (Shanklin and Somerville, 1991; Fox et al., 1993; Shanklin et al., 1994). As directly shown for soluble desaturases and inferred for membrane bound desaturases, these conserved histidines coordinate two irons constituting the diiron centre crucial for fatty acid desaturase activity. Catalysis by this centre occurs through the activation and coordination of molecular oxygen that serves as oxidant in the formation of the double bond in fatty acid desaturases (Shanklin and Cahoon, 1998; Shanklin et al., 2009). Given that the Arabidopsis genome sequence is available (Arabidopsis Genome Initiative, 2001) and that fatty acid desaturases are conserved at the primary sequence level (Sperling et al., 2003), it was surprising that the gene encoding the missing Arabidopsis 16:1Δ3t PtdGro desaturase was not identified based on sequence similarity to other desaturases. To begin studying 16:1Δ3t formation in plants and characterizing the enzyme(s) responsible, we identified the FAD4 gene by map-based cloning.

Results

Identification of a FAD4 candidate gene

Prior to release of the Arabidopsis genome sequence, the fad4-1 mutation was mapped to Arabidopsis chromosome 4 close to the APETALA2 (AP2) locus (Hugly et al., 1991). However no gene with a sequence similar to known fatty acid desaturases was identified in this interval and we used map-based cloning to uncover FAD4. This was feasible because the homozygous fad4-1 mutant is easily recognizable due to lack of a 16:1Δ3t methylester-specific gas chromatography signal when total leaf fatty acid methylesters are analyzed (Browse et al., 1985). An F2 population of 100 plants of mutant phenotype (fad4-1/fad4-1) was generated from a cross between a fad4-1 mutant plant in the Col-2 background and wild-type Landsberg erecta (Ler) and the fad4-1 mutation was mapped to a 3.55 Mbp interval between markers F9F13 and T16L4 as shown in Figure 1(a). This mapping interval was confirmed and narrowed to a 1.75 Mb region using array genotyping of bulk segregants (Hazen et al., 2005) (nucleotides 12,340,173 to 14,090,173) as shown in Figure S2. Using an additional 300 homozygous F2 plants, the fad4-1 mutation was placed into a 0.52 Mbp interval between DNA markers F14M19 and M4122 (Figure 1a, Table S1). The approximate location of fad4-1 identified by these two methods was inconsistent with the previously reported location (Hugly et al., 1991).

Figure 1.

 Mapping of the fad4-1 mutation.
(a) First pass mapping of fad4-1 onto chromosome 4 (Chr. 4) in the vicinity of APETALA 2 (AP2). DNA marker names are indicated above the horizontal line. Recombinants between the respective DNA marker and fad4-1 per total chromosomes analyzed are indicated. The black bar on top indicates the interval for fad4-1 identified by array genotyping.
(b) Fine mapping between markers F14M19 and M4122. Additional DNA markers and the exact position of FAD4 are indicated. Recombinants between the respective marker and FAD4 are indicated below the line (of 2000 F2 plants assayed).
(c) Structure of the FAD4 gene locus (At4g27030.1). The solid black bar indicates the single exon annotated for this gene. The positions for the fad4-1 point mutation and the two T-DNA insertion alleles fad4-2 (SAIL_1250_C12) and fad4-3 (SAIL_515_B9) are indicated. (d) Genotyping of fad4-1 using a dCAPS marker. The sequence surrounding the fad4-1 mutation is indicated on the right (wild type on top, fad4-1 on the bottom). This is a G to A transition at nucleotide 560 in the At4g27030.1 full length cDNA listed at TAIR (http://www.Arabidopsis.org), predicted to cause a stop codon at amino acid residue 177.

Using the F14M19 and M4122 markers, a population of 2000 F2 plants was used for fine-mapping by scoring for recombination events within the interval flanked by the markers (Jander et al., 2002). Eighteen recombinants were phenotyped for the fad4-1 characteristic fatty acid methylester pattern and plants of the subsequent F3 generation tested to confirm for heterozygosity at the FAD4 locus in the original F2 recombinants. Of these, 11 could be placed on the left side of fad4-1 and 7 on the right side (Figure 1b). Using additional markers in this interval, the FAD4 locus was ultimately placed between newly generated markers F10M23b and T24A18 (Figure 1b, Table S1).

The interval between F10M23b and T24A18 on chromosome 4 contained 31 annotated genes (from At4g26880 to At4g27170) according to The Arabidopsis Information Resource (TAIR; http://www.arabidopsis.org) version 8. We reasoned that FAD4 would be co-expressed with genes encoding plastid fatty acid desaturases FAD5 (At3g15850) and FAD7 (At3g11170) and plastid glycerol-3-phosphate acyltransferase ATS1 (At1g32200). Twenty five of the initial 31 gene models in this interval (from At4g26940 to At4g27170) were included in the cluster analysis as shown in Figure S3. One gene, At4g27030, consistently clustered with the test set of fatty acid desaturases and the acyltransferase and contained a predicted chloroplast targeting signal. The inferred amino acid sequence also contained histidine motifs resembling those found in other fatty acid desaturases (Shanklin and Cahoon, 1998) as discussed below.

Sequencing the At4g27030-corresponding cDNA from the fad4-1 mutant revealed a G to A transition at nucleotide 560 in the At4g27030.1 full length wild-type cDNA listed at TAIR, which is predicted to cause a stop codon at amino acid residue 177 of the inferred amino acid sequence (Figure 1c,d). A dCAPS PCR marker (Neff et al., 1998) was designed and the mutation confirmed in the genomic sequence of fad4-1 (Figure 1d). To independently verify that loss of At4g27030 function is linked to the altered fatty acid profile observed for fad4-1, two independent T-DNA insertion alleles (Sessions et al., 2002) (Figure 1c; SAIL_1250_C12, fad4-2; SAIL_515_B9; fad4-3) were included in the analysis. The insertion sites are shown in Figure 1(c). All three available fad4 mutant alleles caused a similar change in the gas chromatography profile of PtdGro derived fatty acid methylesters as shown in Figure 2(b) for fad4-1.

Figure 2.

 Expression of At4g27030 in fad4-1 converts palmitate into Δ3-trans hexadecenoic acid in phosphatidylglycerol.
Gas chromatograms of fatty acid methylesters in phosphatidylglycerol (PtdGro) isolated from leaf extracts of wild-type Col-2 (a), fad4-1 (b) and a representative fad4-1 plant with high expression of the FAD4 wild-type coding sequence under the control of the 35S-cauliflower mosaic virus promoter (c). The signals for palmitoyl methylester (16:0) and Δ3-(E) hexadecanoyl methylester (16:1Δ3t) are labeled accordingly.
(d) Relative mol% palmitic acid (16:0, open bar) and Δ3-trans hexadecenoic acid (16:1Δ3t, solid bar) in phosphatidylglycerol isolated from leaves of the wild type (Col-2), fad4-1 and three representative fad4-1 lines expressing the FAD4 wild-type coding sequence. As shown in reverse transcriptase PCR experiments, two of the lines are higher expressers, one lower. Three replicates were averaged and the standard deviation is shown. In some instances, error bars are too small to be visible.

Overexpression of At4g27030 leads to synthesis of 16:1Δ3t in Arabidopsis

The influence of ectopic expression of At4g27030 was tested in transgenic Arabidopsis. Expression of the wild-type cDNA sequence under control of the 35S-cauliflower mosaic virus promoter (cFAD4) not only reduced levels of 16:0 in PtdGro, but also led to an increase of 16:1Δ3t in PtdGro (Figure 2c,d). Mass spectrometry analysis of the 16:1Δ3t methyl ester confirmed its identity and the position of the double bond (Figure S4). The extent of conversion of 16:0 to 16:1Δ3t in PtdGro was correlated with expression of FAD4 mRNA in independent transgenic lines as shown in Figure 2(d) and overexpression lines accumulated 16:1Δ3t in PtdGro in excess of that seen in wild-type Col-2. Taken together these data indicate that At4g27030 is the FAD4 gene. The previously published genetics results demonstrating intermediate levels of 16:13Δt in the FAD4/fad4-1 heterozygote (Browse et al., 1985) combined with our transgenic plant data (Figure 2d) indicate that the FAD4 protein is limiting for the conversion of 16:0 to 16:1Δ3t in PtdGro of Arabidopsis and is a candidate for the Δ3-trans desaturase itself or a component essential for enzymatic activity.

The FAD4 protein is chloroplast-localized

If FAD4 is directly involved in desaturation of 16:0 in plastid PtdGro, it should be in the chloroplast. While such a localization is predicted from the amino acid sequence (Emanuelsson et al., 2000), experimental evidence was sought. The 3′-end of the full-length FAD4 coding sequence was fused to the 5′-end of the yellow fluorescent protein (YFP) coding sequence. The resulting fusion coding sequence was transiently expressed in tobacco leaf cells under the control of the 35S-cauliflower mosaic virus promoter. As shown in Figure 3(a), the FAD4–YFP fluorescence co-localized with chloroplast chlorophyll fluorescence. In an independent approach, the FAD4 coding sequence was transcribed and translated in vitro in the presence of labeled methionine. The labeled FAD4 precursor protein (prFAD4) was imported into isolated pea plastids and was processed to a smaller mature form (mFAD4), which associated with membranes as shown in Figure 3(b). However, unlike the control protein ARC6, which is an inner envelope membrane protein with a large domain protruding into the intermembrane space (Vitha et al., 2003), FAD4 was not sensitive to trypsin digestion. It was previously shown that trypsin can penetrate the outer envelope membrane but not enter the stroma (Jackson et al., 1998). To validate this point, a stromal control protein, the small subunit of Rubisco, was insensitive to trypsin treatment (Figure 3b). Interestingly, while the FAD4 protein sequence contains four predicted transmembrane spanning domains (Hofmann and Stoffel, 1993), no domain of the protein seems to protrude sufficiently into the intermembrane space to be accessible to trypsin. Taken together these results indicate that FAD4 is imported into chloroplasts where it is inaccessible to trypsin and localizes to the membranes.

Figure 3.

 Localization of the FAD4 protein in plastids.
(a) Localization of a FAD4–YFP fusion protein during transient expression in tobacco leaf cells.
(b) Chloroplast import experiments with labeled FAD4 and control proteins ARC6 and Rubisco small subunit (SS). Chloroplasts were treated with (+) or without (−) trypsin (Tr). Total chloroplast membrane (P) or soluble (S) fractions were analyzed by SDS-PAGE and fluorography. TP, translation product; protein prefix pr, precursor; m, mature form; M, molecular weight markers.

FAD4 contains histidine motifs reminiscent of desaturases

The FAD4 gene is predicted to encode an integral membrane protein, which is 323 amino acids long and has a predicted molecular mass of 36.4 kDa. The protein sequence contains two histidine motifs, 229HAWAH233 and 258HAEHH262 (Figure 4a), whose sequence and spacing are reminiscent of, but not identical to conserved motifs in membrane-bound desaturases (Shanklin et al., 1994). Moreover, the FAD4 protein has other interesting differences from characterized membrane-bound desaturases. First, the position of the two motifs in FAD4 differs relative to the membrane-spanning domains in previously characterized desaturases. Second, while FAD4 contains a third histidine motif (170QGHH173; Figure 4a) its sequence diverges from the third histidine motif present in membrane-bound desaturases. Third, the inferred FAD4 primary sequence has little resemblance to that of known fatty acid desaturases beyond the presence of histidine motifs and membrane-spanning domains.

Figure 4.

 The FAD4 protein sequence clusters with proteins unrelated to taxonomically diverse homologues.
(a) Alignment of FAD4 predicted coding sequence with related proteins from flowering plants, moss and algae. Black bars indicate predicted membrane spanning domains. Histidine residues in three clusters are marked with asterisks. Open boxes indicate conserved residues and black boxes identical residues. Sequences from top to bottom are (GenBank Protein ID): Arabidopsis thaliana FAD4 (NP_194433), FAD4L2 (NP_179874) and FAD4L-1 (NP_176410); Vitis vinifera (CAO38982); Oryza sativus (BAF23067); Chlamydomonas reinhardtii (EDP01256); Physcomitrella patens (EDQ82472); Ostreococcus tauri (CAL55480).
(b) An unrooted tree indicating the relatedness of predicted FAD4 protein homologues in representative organisms. Boot strapping values >950 are marked by +, those between 500 and 950 are marked with an open circle and those under 500 by an open square. Protein sequences in addition to those already described above are (GenBank Protein ID): Trypanosoma cruzi (Q4DYH2); Leishmania major (Q4QCH1); Acanthamoeba polyphaga (minivirus) (Q5UR78); Bradyrhizobium japonicum (Q89SH1); Maricaulis maris (Q0AK39); Myxococcus xanthus (Q1D0D6); Stigmatella aurantiaca (Q098L6); Bt, Bos taurus (NP_001093785); Mm, Mus musculus (NP_663513); Hs, Homo sapiens (Q5TGE1); Dr, Danio rerio (NP_001073633); Drosophila pseudoobscura (Q29KL3); Caenorhabditis elegans (Q9XW52).

Discussion

The results described above identify At4g27030 as the FAD4 locus. Genetic mapping (Jander et al., 2002) and array genotyping (Hazen et al., 2005) narrowed the search to 31 genes and one gene in this region was found to have mRNA levels that are highly correlated (Toufighi et al., 2005) with three genes known to be involved in plastid lipid metabolism. At4g27030 was confirmed to be FAD4 by sequencing of the mutant allele, complementation by expression of the wild-type cDNA and phenotypic analysis of two independent T-DNA insertion lines. While map-based cloning of the gene was straightforward, definitively proving that FAD4 is the enzyme directly involved in the conversion of 16:0 to 16:1Δ3t in PtdGro turned out to be as challenging as expected. This is because membrane associated fatty acid desaturases are notoriously difficult to assay in vitro.

As active membrane-bound desaturases have been successfully expressed in S. cerevisiae, e.g. FAD5 (Heilmann et al., 2004), we attempted to assay FAD4 activity in baker’s yeast. Unfortunately, no evidence of enzymatic activity was observed. Approaches included expression of genes for recombinant proteins with or without amino terminal protein truncation using both a galactose inducible promoter and a constitutive glyceraldehyde-3-P dehydrogenase promoter and yeast cytochrome oxidase subunit IV mitochondrial target sequence. These constructs were transformed into a wild-type and a cardiolipin synthase deficient crd1 yeast strain, which produces higher levels of PtdGro that accumulates mostly in the mitochondria (Chang et al., 1998). Lack of activity in yeast could be due to several factors. First, it is possible that PtdGro with 16:0 in the sn-2 position was not available to the heterologously expressed FAD4 enzyme. Second the requisite electron-donating redox cofactor might not have been available to the enzyme. Finally, the protein might not have been inserted properly into a membrane to allow enzymatic activity.

Several lines of evidence support the hypothesis that FAD4 is the suspected 16:1Δ3t PtdGro desaturase. First, our results suggest that the FAD4 protein is limiting for the biosynthesis of 16:1Δ3t PtdGro as transgenic plants overexpressing FAD4 cDNA have increased levels of this product (Figure 2). This result is consistent with the observation that heterozygous F1 plants (fad4-1/FAD4) have a phenotype intermediate to that of wild type and homozygote mutants (fad4-/fad4-1) (Browse et al., 1985). Second, indirect evidence is obtained by analysis of the predicted FAD4 amino acid sequence. This contains features similar to known membrane-bound desaturases (Shanklin et al., 1994, 2009): four predicted membrane spanning domains and two histidine motifs that could potentially coordinate a metal centre involved in catalysis (Figure 4a). A third histidine motif potentially involved in metal binding consists of 170QGHH173 (Figure 4a). Although glutamine is found substituting for histidine in these motifs (Sayanova et al., 2001), known membrane-bound desaturases typically have two-to-three amino acids between the glutamine and the histidines (Shanklin and Cahoon, 1998). Another difference is that in characterized membrane-bound desaturases two histidine motifs are located between the second and the third membrane spanning domain, while the third is located at the C-terminus of the protein. In FAD4, only one proposed histidine motif is located between membrane spanning domains two and three, while the other two are located at the C-terminus (Figure 4a). Although these are clear differences, the predicted topology of FAD4 would still permit the interaction of the proposed three histidine motifs on one side of the membrane into which FAD4 is integrated and, therefore, the protein could potentially fold into an active site similar to that predicted for membrane-bound desaturases. Based on these sequence features, it seems likely that FAD4 is a metalloenzyme that evolved independently from characterized desaturases.

A third line of evidence that FAD4 encodes a desaturase is that no other mutants were recovered that caused a loss of 16:1Δ3t PtdGro during the extensive fatty acid mutant screen conducted by John Browse, Chris Somerville and coworkers (Somerville and Browse, 1991; Wallis and Browse, 2002). This is consistent with the hypothesis that FAD4 encodes the one protein uniquely required for the biosynthesis of this lipid species, though we cannot rule out the contribution of genetic redundancy. Other proteins, for example the electron carrier ferredoxin, may be required as well (Shanklin and Cahoon, 1998), though complete loss of function mutations in these would be expected to affect additional physiological processes. Fourth, labeling kinetics done in algae suggested that 16:0 esterified in PtdGro is the substrate for the biosynthesis 16:1Δ3t PtdGro (Ohnishi and Thompson, 1991) making it unlikely that a soluble desaturase is involved. In addition, desaturation at the PtdGro level is indirectly evident from studies in which a bacterial diacylglycerol kinase was targeted to chloroplast envelopes in transgenic tobacco (Fritz et al., 2007). As a result unusual Δ3-trans double bond containing 18-carbon fatty acids were found in the sn-2 position of PtdGro because ER-derived diacylglycerol moieties containing 18-carbon fatty acids at the sn-2 position were channeled into plastid PtdGro biosynthesis. Integral membrane proteins such as membrane-bound desaturases or FAD4 more likely work on fatty acids buried in the membrane when esterified in PtdGro than a member of the soluble desaturase class, which work on acyl–acyl carrier proteins in the plastid. As a side note, the observation of the Δ3-trans double bond in 18-carbon fatty acids of PtdGro in these transgenic lines also suggested that the respective desaturase can accept 16- or 18-carbon fatty acids in PtdGro as substrates. Finally, the FAD4 protein is located within the plastid (Figure 3) and associates with chloroplastic membranes as expected for an enzyme acting on a plastid-specific membrane lipid.

The FAD4 gene has two closely related paralogs in Arabidopsis, At1g62190 and At2g22890, for which no function is reported. Moreover this protein class, originally designated Kua proteins (Thomson et al., 2000), is highly conserved in organisms ranging from bacteria (but not cyanobacteria) to mammals, as illustrated in an unrooted tree with representative proteins shown in Figure 4(b). Unfortunately, definitive functional data are not available for these proteins. CarF from Myxococcus xanthus is a notable exception, having been genetically implicated in light-dependent regulation of carotenoid biosynthesis (Fontes et al., 2003). Based on genetic and protein–protein interaction experiments, CarF functions as a bacterial anti-antisigma factor possibly by physically interacting with the sigma factor CarR along with the antisigma factor CarQ (Galbis-Martinez et al., 2008). Of significance to our understanding of FAD4 is that site-directed mutation of histidine residues in each of the three motifs conserved between CarF and FAD4 led to loss of CarF function demonstrating the importance of these motifs (Thomson et al., 2000; Galbis-Martinez et al., 2008). A similar mutational analysis of histidines in membrane-bound desaturases contributed to the recognition that these histidine motifs bind transition metals crucial for catalysis (Shanklin et al., 1994). The identification of FAD4 from Arabidopsis will provide future insights into its catalytic role in the biosynthesis of 16:1Δ3t PtdGro in Arabidopsis and provide clues for the biochemical function of CarF and related Kua proteins in bacteria and animals.

Experimental procedures

Plant materials, mapping, DNA markers and sequencing

Arabidopsis thaliana plants were of the ecotypes Columbia-2 (Col-2) or Landsberg erecta (Ler). The fad4-1 mutant was obtained from John Browse (Washington State University, Pullman, WA) and insertional mutants were from the Arabidopsis Resource Center (Ohio State University, Columbus, OH). Plants were grown as previously described (Xu et al., 2002). For map-based cloning, fad4-1 plants in the Col-2 ecotype were crossed to Ler wild-type plants. The F2 plants were tested for their leaf fatty acid methylester profiles as described below. The location of fad4-1 on chromosome 4 was established with DNA polymorphism markers G4539 and T16L4 (Figure 1). For fine mapping, additional markers were generated as shown in Figure 1 with details listed in Table S1 by taking advantage of the Monsanto Polymorphism and Ler Sequence Collection [(Jander et al., 2002); http://www.arabidopsis.org/Cereon/index.jsp]. Genotyping at the fad4-1 locus was performed using the dCAPS marker (Neff et al., 1998) described in Table S1.

Both cDNA and genomic DNA of the fad4-1 mutant were sequenced. Primers 5′-TTTGACAACTTTCACCTGCA-3′ and 5′-TACGAAGAAATCTTTTCAGT-3′ were used to PCR amplify the single exon of At4g27030. The resulting PCR fragment was cloned into the sequencing plasmid pCR®2.1-TOPO (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions. Sequencing was done at the MSU Research Technology Support Facility (RTFS; http://www.genomics.msu.edu/).

Constructs for complementation and plant transformation

Total RNA was purified from plants using an RNeasy® plant mini kit (Qiagen, Valencia, CA, USA). First strand cDNAs were synthesized from 1–2 μg total RNA by using SuperScript III reverse transcriptase (Invitrogen) following the manufacturer’s instructions. The FAD4 coding-sequence was isolated by RT-PCR using the primers 5′-GCCGAAGATCTATGGCTGTATCACTTCCAACCAAG-3′ and 5′-AAAGGGTACCTTATGCTTGGTTGTTGGAGATTTCGG-3′. The PCR fragments were ligated into the pCR®-Blunt vector using the Zero® Blunt PCR Cloning Kit (Invitrogen) and sequenced. The plasmids were cut with BglII and KpnI and then inserted into the binary plant transformation vector pCAMBIAmcs1300 (http://www.cambia.org). The vector was introduced into A. tumefaciens strain GV3101 by electroporation and plant transformation was done by floral-dip (Clough and Bent, 1998). The T1 seeds were screened for resistance to Hygromycin B (30 μg ml−1) on agar-solidified MS medium. An estimation of FAD4 transcript levels in transgenic plants was obtained by semiquantitative RT-PCR using the primers mentioned above. Actin2 (AT3G18780) mRNA was tested for control purposes using the following primers: 5′-ATTCAGATGCCCAGAAGTCTTGTTC-3′ and 5′-GCAAGTGCTGTGATTTCTTTGCTCA-3′. PCR amplification was performed for 28 cycles with denaturation at 98°C for 30 sec, annealing at 55°C for 30 sec and elongation at 72°C for 60 sec, followed by 72°C for 10 min.

Transient expression in tobacco

For the generation of the FAD4–YFP fusion, the entire coding-sequence of At4g27030 was inserted into the XbaI and SalI sites of the plasmid pVKH18–EN6–Sec12–vYPF (unpublished; provided by Federica Brandizzi, MSU-DOE-Plant Research Laboratory, E. Lansing, MI) using the following primers: 5′-GCGCTCTAGAATGGCTGTATCACTTCCAACCAAG-3′ and 5′-GCCAGGGTCGACTTCCTTGCTTGGTTGTTGGAGATTTCGG-3′. The vector was introduced into A. tumefaciens strain GV3101 by electroporation and Nicotiana tabacum (cv. Petit Havana) was transformed according to (Batoko et al., 2000). The FAD4–YFP fusion protein was directly examined using a Zeiss LSM5 confocal microscope. Excitation light was provided by an argon laser at 514 nm. YFP fluorescence was observed with a band-pass filter of 530–600 nm and chlorophyll fluorescence with a 650-nm long-pass filter.

Chloroplast import assays

The cDNA encoding FAD4 was inserted into pCR8/GW/TOPO (Invitrogen) using a PCR-based strategy according to the manufacturer’s instructions. The Rubisco small subunit RCBS (Olsen and Keegstra, 1992) and ARC6 (Vitha et al., 2003) genes were used as controls. Isolation of pea chloroplasts, import assays and post-import treatment were done as previously described (Xu et al., 2005).

Lipid and fatty acid analysis

To isolate PtdGro, lipid extracts were prepared and chromatographed as previously described (Dörmann et al., 1995) on activated ammonium sulfate-impregnated silica gel TLC plates (Si250PA; Mallinckrodt, Baker, NJ, USA) by using a solvent system of acetone/toluene/water (91/30/7.5, v:v:v). Lipids were stained by exposure to iodine vapor for 30 sec and then silica material containing PtdGro was scraped with a razor blade into a glass reaction tube. Fatty acid methylesters were prepared according to (Dörmann et al., 1995) and analyzed as described in (Rossak et al., 1997). Positional analysis of the double bond in the 16:1Δ3t methylester was carried out as described by Francis, 1981.

Bioinformatics

The e-Northern tool provided by the Bio-Array Resource for Arabidopsis Functional Genomics (http://bar.utoronto.ca/) (Toufighi et al., 2005) was used to analyze coexpression of genes in the interval of interest with three known chloroplastic lipid metabolism genes, FAD5 (At3g15850), FAD7 (At3g11170) and ATS1 (At1g32200).

FAD4 homologues and paralogues were identified through a pre-computed BLAST search (BLINK) available for the public domain at the National Center for Biotechnology Information website (http://www.ncbi.nlm.nih.gov/). The multiple sequence alignment shown in Figure 4(a) was computed by ClustalW (2.0) (Larkin et al., 2007) and shading of the conserved amino acid residues was done using the BOXSHADE tool (3.21) both available at the Swiss EMBnet node server (http://www.ch.embnet.org/index.html). The multiple sequence alignment used in the phylogenetic analysis was performed by ClustalX (2.0) (Larkin et al., 2007). Phylogenetic analysis was computed by the PHYLIP package (3.68) using the neighbour-joining method based on comparison of 1000 bootstrap replications as previously described (Mamedov et al., 2005).

Acknowledgements

We are grateful to John Ohlrogge for his advice and critique of the manuscript, to Dan Jones for help with the mass spectrometry analysis and Barry Williams and Kristine Cox for help with yeast functional assays. We thank John Browse for providing the Arabidopsis fad4-1 allele, Federica Brandizzi for providing the plasmid VKH18–EN6–Sec12–vYPF and Peter Griac and Maria Simockova for the crd1 mutant. This work was funded by NSF Arabidopsis 2010 grant MCB-0519740 to RL, CB and others, by DOE grant DE-FD02–98ER20305 to CB and NSF grant MCB-0741395 to CB. JEF was funded by DOE grant DE-FG02–91ER20021 to Ken Keegstra.

Ancillary