A Sphingolipid Desaturase from Higher Plants

A recently cloned cDNA from sunflower codes for a fusion protein composed of an N-terminal cytochromeb 5 and a domain similar to membrane-bound acyl lipid desaturases. For a functional identification, homologous cDNAs from Brassica napus and Arabidopsis thaliana were expressed in Saccharomyces cerevisiae, and sphingolipid long chain bases were analyzed. The expression of the heterologous enzyme results in significant proportions of new Δ8,9-cis/trans-phytosphingenines that accompany the residual C18-phytosphinganine predominating in wild-type yeast cells. These results represent the first identification of a gene coding for a sphingolipid desaturase and for a stereounselective desaturase showingtrans-activity from any organism. Furthermore, this fusion protein is a new member of the cytochrome b 5superfamily. The formation of the two regioisomeric phytosphingenines in the transformed yeast sheds new light on the factors controlling regioselectivity.

A recently cloned cDNA from sunflower codes for a fusion protein composed of an N-terminal cytochrome b 5 and a domain similar to membrane-bound acyl lipid desaturases. For a functional identification, homologous cDNAs from Brassica napus and Arabidopsis thaliana were expressed in Saccharomyces cerevisiae, and sphingolipid long chain bases were analyzed. The expression of the heterologous enzyme results in significant proportions of new ⌬ 8,9 -cis/trans-phytosphingenines that accompany the residual C 18 -phytosphinganine predominating in wild-type yeast cells. These results represent the first identification of a gene coding for a sphingolipid desaturase and for a stereounselective desaturase showing trans-activity from any organism. Furthermore, this fusion protein is a new member of the cytochrome b 5 superfamily. The formation of the two regioisomeric phytosphingenines in the transformed yeast sheds new light on the factors controlling regioselectivity.
Sphingolipids are ubiquitous membrane components in eukaryotic cells and in a few bacteria (1). The hydrophobic moiety is a long chain base (LCB 1 ; 2-amino-1,3-dihydroxyalkane) carrying N-acylated fatty acids of 14 -26 carbon atoms to form a ceramide. Depending on the source, this basic structure can be modified by differences in chain length, degree of unsaturation, methyl branching, and insertion of additional hydroxy groups. Complex sphingolipids such as cerebrosides and phytoglycolipids are generated by the addition of various sugar residues and other polar head groups to the ceramide.
In animal cells, these membrane lipids and in particular their catabolites serve as intra-and intercellular second messengers regulating cell growth, differentiation, apoptosis, and pathogenic defense (2,3). On the other hand, far less is known about the role of sphingolipids in plants and fungi. However, the lethality of Saccharomyces cerevisiae mutants defective in LCB biosynthesis (4) and the toxicity of compounds interfering with their synthesis in plants suggest that they are essential also in these organisms (5).
Very recently, genes for the ceramide hydroxylation were identified (6,7), but the genes responsible for the desaturation of the LCB were still unknown. Enzymatic studies with animal systems showed that the ⌬ 4 -trans double bond is inserted into free ceramide (N-acylated sphinganine) and that this conversion is catalyzed by a desaturase and not by a dehydrogenase or an oxidase (8,9). In contrast to animal and yeast cells, plant sphingolipids contain additional C-8-unsaturated LCB of cisand trans-configuration, which might be involved in chilling resistance (10), but nothing is known about the molecular mechanism responsible for their formation.
Here we describe the first cloning of a sphingolipid desaturase from plants that catalyzes the formation of unsaturated LCBs.
DNA Amplification from Brassica napus-A ZAP cDNA library of developing siliques of B. napus cv. Ascari (12) was used for PCR amplification of a 571-bp DNA fragment with the degenerated primers 5Ј-G(G/C)(A/T/G/C)TGGTGGAA(AG)TGG-3Ј (forward) and 5Ј-GG(A/ G)AA(A/T/G/C)A(A/G)(A/G)TG(A/G)TG-3Ј (reverse). The 5Ј-end was amplified with the T3 primer (Stratagene) and the specific reverse primer 5Ј-TTATGCGTCCATTTCCACCA-3Ј. cDNA isolated from mRNA of developing B. napus (cv. Drakkar) embryos was used to amplify the 3Ј-end with the specific forward primer 5Ј-TTCTTTGGCGGGTTGCAGTT-3Ј and a poly(dT) primer. Specific primers (5Ј-AACCATCTCTGTTTCAA-C-3Ј and 5Ј-CAAGTGATGATGAGTTAC-3Ј) derived from the 5Ј-and 3Ј-untranslated regions were used in PCR experiments, and a fulllength clone of 1502 bp was isolated and sequenced. mRNA isolation, cDNA synthesis and tailing, PCR amplification, and DNA sequencing were carried out as described (13).
DNA Amplification from Arabidopsis thaliana-A BLAST search (14) in the non-redundant data base of the GenBank TM EST Division revealed some A. thaliana EST clones 2 aligning with the B. napus amino acid sequence amplified above. Specific primers derived from the A. thaliana EST clones EMBL accession no. T42569 (forward) 5Ј-GC-GATTCAAGGCAAGGTCTAC-3Ј and EMBL accession no. F13717 (reverse) 5Ј-TTAGCCATGAGTATTCAAAGC-3Ј were used for the PCR amplification of a 1269-bp DNA fragment from a ZAP cDNA library of A. thaliana (ecotype Columbia), which was received from the Genetic Stock Center of the Max-Planck Institute (Köln, Germany). PCR with the above mentioned primers in combination with T7 or T3 primer (Stratagene), respectively, amplified the corresponding 5Ј-and 3Ј-ends. All fragments were subcloned and sequenced as described before (13).
Plasmid Construction and Yeast Transformation-A 1.2-kilobase pair FAD2 sequence was amplified from cDNA of developing B. napus cv. Drakkar embryos by PCR with oligonucleotide primers (5Ј-CCGG-* This work was supported by Bundesministerium fü r Bildung, Wissenschaft, Forschung und Technologie Grant BEO/0311156; the Fonds der Chemischen Industrie; and the plant breeder companies Norddeutsche Pflanzenzucht, Deutsche Saatveredelung, and Kleinwanzlebener Saatzucht. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AJ224160, AJ224161, and U79010.
TACCATGGGTGCAGGTGGAAG-3Ј and 5Ј-CCGAATTCTCATAACT-TATTGTTGTACCA-3Ј) designed from the FAD2 cDNA sequence of A. thaliana (15). The 1.4-kilobase pair sphingolipid desaturase sequences were amplified by PCR from cDNA of developing B. napus (cv. Drakkar) embryos with the primers 5Ј-CCGGTACCATGTCGGAGCAGACAAA-G-3Ј and 5Ј-CCGAATTCCTAGCCATGAGTATTCAGA-3Ј and from the A. thaliana (ecotype Columbia) ZAP cDNA library mentioned above with the primers 5Ј-CCGGTACCATGGCGGAAGAGACGGAG-3Ј and 5Ј-CCGAATTCTTAGCCATGAGTATTCAAAGC-3Ј. For ligation into the yeast expression vector pYES2 (Invitrogen), these primers were extended by a KpnI or EcoRI restriction site (underlined). The amplified DNAs were ligated into the pGEM-T vector (Promega) and verified by sequencing (Applied Biosystems) using the ABI PRISM Dye Primer and Terminator Cycle Sequencing Ready Reaction Kits (Perkin-Elmer). The excised KpnI-EcoRI fragments were inserted into the KpnI-EcoRI site of the expression vector pYES2 (Invitrogen), and the resulting plasmids (pBnDES8 and pAtDES8) were transformed into S. cerevisiae strain INVSc1 (Invitrogen) using the polyethylene glycol method (16).
Growth and Galactose Induction-Cultures of transformed S. cerevisiae were grown at 30°C for 24 h in complete minimal-dropout-uracil medium containing 2% (w/v) raffinose (17). During early logarithmic growth, test cultures were induced with 1.8% galactose (final concentration) and grown aerobically to saturation for another 24 h.
Determination of the Double Bond Position in Phytosphingenines-One aliquot (360 g) of the peracetylated mixture of phytosphinganine and phytosphingenine isomers was de-O-acetylated (0.02 M sodium methoxide, 1 h at room temperature) prior to lead tetraacetate oxidation (22). The resulting aldehydes were extracted with chloroform, reduced with lithium aluminum hydride, and separated into two aliquots; one (200 g) was used to produce the nicotinates (22), and the other (140 g) was subjected to potassium permanganate oxidation (22). The resulting fatty acids were transformed to methyl esters by etheral diazomethane.
Determination of the Double Bond Position in Fatty Acids-Methyl esters of total fatty acids, obtained by acid methanolysis of cell pellets (about 300 mg) from wild type (pYES2) and transgenic yeasts (pBnDES8 and pAtDES8) were purified by preparative silica gel TLC in petroleum ether/diethylether (9:1, v/v; R F ϭ 0.4), yielding about 1 mg of purified fraction. Part of it (about 0.2 mg) was subjected to a modified von Rudloff oxidation (22). Briefly, the fatty acid methyl esters were dissolved in methanol (0.5 ml) and oxidized with 0.5 ml of a solution of 10 mg of KMnO 4 in 2 ml of 0.1 M H 2 SO 4 . After 30 min at 37°C, 8 ml of water were added, and organic products were extracted twice with 2 ml of chloroform. After concentration to dryness, 0.5 M NaOH in methanol/ water (1:1 (v/v), 1.0 ml) was added followed by 5 mg of NaBH 4 . After extraction, the resulting dihydroxy fatty acids were treated with diazomethane in diethylether for reconversion to methyl esters. Following removal of solvents, bis(trimethylsilyl)trifluoroacetamide (30 l; Sigma) was added, and the mixture was kept for 16 h at 60°C for formation of trimethylsilyl derivatives of bis-hydroxylated fatty acid methyl esters. Fatty acid pyrrolidides were prepared from another aliquot of the original fraction (0.2 mg) following an established procedure (24).

PCR Cloning and Sequence
Analysis-In previous experiments, we had cloned a cDNA from sunflower that codes for a fusion protein between an N-terminal cytochrome b 5 and a putative membrane-bound desaturase domain, but so far a biochemical function had not been assigned to this C-terminal domain (13). In continuation of these investigations, we carried out additional PCR experiments with cDNA from developing B. napus embryos and detected another cDNA closely related to the previously cloned sunflower sequence (13). 3 We isolated a full-length clone of 1594 bp corresponding to an open reading frame of 449 amino acids. Estimates of gene copy number indicated 6 -8 gene copies/haploid genome of B. napus. This number corresponds to the copy numbers found for other lipid desaturases in B. napus, which all are present as small gene families (25).
Based on substantial similarity of some EST clones of A. thaliana (EMBL accession numbers T42569 (36 -102), N37558 (42-204), F13728 (122-210), T42806 (375-405), F13717 (430 -458) (numbers in parentheses refer to the amino acid positions given in Fig. 1) to the B. napus sequence, we also carried out PCR experiments with an A. thaliana ZAP cDNA library. A full-length clone of 1678 bp was obtained, which corresponds to an open reading frame of the same length as the B. napus sequence. 4 At the N terminus, both open reading frames are nine amino acids shorter than the sunflower protein, but they also show seven highly conserved amino acid residues characteristic for cytochrome b 5 (13). The two predicted polypeptides of 51-52 kDa showed 76% identity to each other and 65% identity to the sunflower protein (Fig. 1).
The Cytochrome b 5 -Desaturase Fusion-Acyl lipid desaturases catalyze the oxygen-and electron donor-dependent insertion of double bonds into fatty acid residues. The immediate electron donor for many microsomal desaturases is cytochrome b 5 . Therefore, a fusion between cytochrome b 5 and a desaturase as found in these new sequences may have a functional advantage. A fused cytochrome b 5 was detected in the ⌬ 9 -acyl-CoA desaturase from yeast (13,26) and a red alga (27), in the ⌬ 6 -acyl lipid desaturase from plants (28,29) and the nematode Caenorhabditis elegans (30), as well as in the sphingolipid ␣-hydroxylase of yeast (7) and in the TU-36B gene of Drosophila melanogaster (31). Therefore, the desaturase-like proteins encoded by the two above mentioned sequences represent new members of the cytochrome b 5 superfamily (Fig. 2). The hemoprotein nature of the N-terminal domain from the homologous sunflower fusion protein has been confirmed by redox absorbance difference spectra of the recombinant cytochrome b 5 expressed in Escherichia coli (13), and the characteristic histidine boxes conserved in all lipid desaturases (32) are also present in this new class of desaturases. The highest similarity (58 -60% identity) was found for the ⌬ 6 -desaturase from borage 5 (28) (Fig. 1), but only limited resemblance (Յ22% identity) exists to the ⌬ 6 -desaturase from the moss Physcomitrella patens (29) and the nematode C. elegans (30).
Functional Expression in S. cerevisiae and Fatty Acid Analysis-In order to elucidate the catalytic function of these new fusion proteins, they were expressed in yeast cells, which provide a membrane-bound redox system suitable for functional cooperation with heterologous desaturases (33). As a control, we inserted a full-length cDNA clone for the microsomal ⌬ 12desaturase from B. napus into a yeast-E. coli shuttle vector under the transcriptional control of the inducible yeast GAL1 promotor. Galactose-induced cultures of transformed yeast cells were used to prepare fatty acids from membrane lipids for subsequent analysis (18). S. cerevisiae clones harboring the B. 3 The nucleotide sequence for the cytochrome b 5 fusion protein from sunflower can be accessed through the EMBL Data Base under EMBL accession number X87143 (13). 4 The nucleotide and amino acid sequences for the sphingolipid desaturases from B. napus and A. thaliana have been deposited in the EMBL Data Base under EMBL accession numbers AJ224160 and AJ224161. 5 The nucleotide sequence for the ⌬ 6 -acyl lipid desaturase from borage is available in the EMBL Data Base under accession number U79010 (28). napus FAD2 gene were capable of forming 11% of ⌬ 9,12 -C 16and ⌬ 9,12 -C 18 -dienoic fatty acids, proving and confirming the capability to express a microsomal ⌬ 12 -desaturase of plant origin (33). Therefore, the two newly isolated cDNAs coding for the fusion proteins of A. thaliana and B. napus were ligated into the same site of the vector described above, and the derived plasmids pBnDES8 and pAtDES8 were transformed into S. cerevisiae. Fatty acid analysis of transformed cells harboring pBnDES8, pAtDES8, or uncut pYES2 (negative control) showed identical wild-type patterns for all three cultures. To confirm that the monoenoic fatty acids (16:1, 18:1) observed in the two transgenic strains are in fact the same regioisomers as formed in wild type yeast, i.e. palmitoleate and oleate with ⌬ 9 double bonds, we carried out a detailed localization of the double bond in each of these two fatty acids from these strains. For this purpose, the fatty acid methyl ester fraction was subjected to a modified von Rudloff oxidation converting the original double bond into a pair of vicinal trimethylsilyloxy groups. Upon GC-MS, the derivatives of the two monounsaturated fatty acids in all three strains were identified as methyl 9,10di-(trimethylsilyloxy)-hexadecanoate (retention time 20.15 min) and methyl 9,10-di-(trimethylsilyloxy)-octadecanoate (retention time 23.73 min). Fragmentation between the vicinal trimethylsilyloxy groups results in characteristic fragments for the carboxyl ends (C-1-C-9 at m/z ϭ 259 from both ⌬ 9 -16:1 and ⌬ 9 -18:1) and for the methyl ends (C-10 -C-16 at m/z ϭ 187 for 16:1; C-10 -C-18 at m/z ϭ 215 for 18:1). The negative result of a single ion monitoring experiment to detect corresponding fragment ions for carboxyl and methyl ends resulting from ⌬ 6 -16:1 (m/z ϭ 217 and 229) and ⌬ 6 -18:1 (m/z ϭ 217 and 257) excluded the formation of these particular regioisomer (for discussion, see below). The analysis of pyrrolidides also showed only ⌬ 9 -16:1 regioisomers (fragments m/z ϭ 168, 182, 196, 208, 222, and 236 (24)) and no fragments diagnostic for the ⌬ 6 -16:1 isomer (m/z ϭ 166, 180, 194, 208, 222, and 236). The same holds true for the ⌬ 9 -18:1 derivative. Furthermore, transformed and wild-type cells, which incorporated exogenously added linoleic acid (⌬ 9,12 -18:2) into their membrane lipids (34), did not show any ⌬ 15 -or ⌬ 6 -desaturation activity with a polyunsaturated substrate. These results indicated that the new fusion proteins are not involved in the desaturation of fatty acids.
LCB Analysis-To investigate a possible modification of sphingolipid LCB, yeast cells were directly subjected to strong alkaline hydrolysis to liberate the free LCB from their complex sphingolipids (19). The extracted LCB were converted to dinitrophenyl derivatives (11) and analyzed by reversed-phase HPLC (Fig. 3). The main LCB of S. cerevisiae is C 18 -phytosphinganine (D-ribo-2-amino-1,3,4-trihydroxyoctadecane), whereas sphinganine and C 20 -phytosphinganine are only minor components under nonstressed growth conditions (35,36). Yeast cells transformed with the empty vector (control) showed the wild-type LCB pattern (Fig. 3B). Transformants expressing pAtDES8 and pBnDES8 accumulated in addition to phytosphinganine new LCB in yields of 59 and 31%, respectively (Fig. 3,  C and D). Their shorter retention times in reversed-phase HPLC suggested that they may be unsaturated derivatives of C 18 -phytosphinganine. . For the localization of the double bond, N-acetylated trihydroxybases were subjected to two different degradation protocols, i.e. to lead tetraacetate and to von Rudloff oxidation (22). Lead tetraacetate fragmentation/reduction resulted in alcohols, which were separated by GC-MS as their nicotinates into three peaks. The first (retention time 16.30 min) was a saturated alcohol, the second was identified as a derivative of 6-pentadecenol  /z ϭ 192, 204, 218, and 232). The von Rudloff oxidation, which yields aliphatic acids from the alkyl portion between the double bond and the terminal methyl group, re-sulted in methyl esters of nonanoic (7.5 min) and decanoic (9.0 min) acid in relative proportions of 1:3. These two independent sequences prove the formation of two regioisomeric phytosphingenines with a ⌬ 8 or a ⌬ 9 double bond in relative proportions of 3:1.

GC-MS Analysis of Phytosphingenine Regioisomers-Un-
NMR Analysis of Phytosphingenine Stereoisomers-For assignment of the double bond stereochemistry (cis/trans), the peracetylated LCB fraction was separated by preparative silver nitrate TLC into three components with R F values of 0.46 (125 g), 0.44 (300 g), and 0.38 (100 g), which were used for recording 1 H NMR spectra. The component with the highest R F value of 0.46 turned out to be the acetylated phytosphinganine (data not shown). The spectrum (Fig. 4) of the fraction with R F ϭ 0.44 displayed nonresolved multiplet signals for olefinic protons at 5.308 ppm, which showed cross-peaks in the COSY experiment to the ␣-methylene protons at 1.879 ppm characteristic for trans-configuration. In the fraction with R F ϭ 0.38 the olefinic signals were shifted to higher field (5.275 ppm), whereas the ␣-methylene protons appeared at lower field (1.927 ppm) as typical for a cis-double bond (37).

DISCUSSION
In view of the fact that only one additional heterologous gene is expressed in the transformed yeast cells, obvious questions arising concern the origin of both the cis/trans-stereoisomers and the ⌬ 8 /⌬ 9 -regioisomers. The presence of cis-and transdouble bonds in the newly formed products could be ascribed either to the activity of a genuine yeast cis/trans-isomerase similar to fatty acid isomerases found in other organisms (38) or to the lack of absolute stereospecificity of the desaturase itself as known from some monooxygenases (39). Reinvestigation of the trihydroxybases from A. thaliana sphingolipids by HPLC and GC-MS confirmed that cis-and trans-isomers are present in ⌬ 8 -phytosphingenines and that the double bond is restricted to C-8. In addition, the cis/trans ratio of A. thaliana phytosphingenines (Fig. 3E) is similar to that found in each of the ⌬ 8 -and ⌬ 9 -phytosphingenines extracted from pAtDES8 transformants (Fig. 3C). This suggests that formation of both stereoisomers may be catalyzed by a single desaturase of plant origin, which does not show absolute stereospecificity as typical for all the fatty acid desaturases cloned so far.
The presence of two regioisomeric ⌬ 8 -and ⌬ 9 -double bonds in phytosphingenine can be explained as follows. The plant LCB desaturase has a regioselectivity comparable with that of a fatty acid desaturase, which introduces a double bond in the ⌬6-position. The acyl desaturase measures this distance from the oxygen-functionalized carbon atom (⌬-desaturase) of the acyl group, which is usually the carboxyl carbon (C-1) of the fatty acid. On the other hand, this carbon atom does not necessarily have to carry an additional carbonyl oxygen. Desaturation of alkenyl glycerol ethers has shown that the oxygen of an ether bridge is sufficient to identify C-1 of the 9-cis-alkenyl chain for correct alignment and regioselective insertion of the second double bond by the ⌬ 12 -desaturase (40). Therefore, we assume that the LCB desaturase measures the ⌬ 6 -position from the oxygen-carrying carbon atom next to the alkyl chain in the LCB (Fig. 5). In plants "C-1" can only be C-3 of sphinganine and sphingosine (4-trans-sphingenine), resulting in the new double bond at C-8, whereas formation of ⌬ 8 -phytosphingenine by hydroxylation at C-4 can only occur subsequent to C-8 desaturation of sphinganine. The reason for this order of events is at present not understood. On the other hand, in the transformed yeast, the strict sequence of C-8 desaturation followed by C-4 hydroxylation does not exist, and the heterologous plant LCB desaturase also has access to phytosphinganine. In this case, C-4 and not C-3 of the LCB is used as "C-1" for alignment of the alkyl chain. Insertion of the double bond in the fixed distance gives ⌬ 9 -instead of ⌬ 8 -phytosphingenine. This picture (Fig. 5) explains the sequence similarity between the LCB ⌬ 8 -desaturase and the fatty acid ⌬ 6 -desaturase (28), but in contrast to the fatty acid desaturases, the LCB desaturase lacks stereospecificity. Therefore, this desaturase is the first trans-desaturase to be cloned. Furthermore, its activity does not depend on the presence of an additional cis-double bond in allylic position as required by most other acyl lipid desaturases including the ⌬ 6 -desaturase. Because of the sequence similarity of the LCB desaturase to ⌬ 6 -acyl group desaturases and the possibility that a ⌬ 6 -hexadecenoyl-CoA could be used by the yeast serine palmitoyl transferase for a direct formation of an unsaturated ⌬ 8 -LCB, we carried out a detailed analysis of the location of double bonds in monounsaturated C 16 and C 18 fatty acids of the two transgenic yeast strains. Both strains contained only the ⌬ 9 -regioisomers as found in wild type and thus rule out the possibility that the ⌬ 8 -double bond in LCB is carried over from the ⌬ 6 -unsaturated fatty acid precursor. Despite the fact that the sphingolipid desaturase accepts a satu- FIG. 5. Working hypothesis for the biosynthesis of phytosphingenine in plant as well as in yeast cells expressing a heterologous sphingolipid desaturase from plants. The desaturase measures the ⌬ 6 distance from the oxygen-functionalized carbon atom next to the double bond to be introduced. When expressed in yeast and depending on the hydroxylation status of C-4, the sphingolipid desaturase from higher plants introduces a C-8 (left branch) or a C-9 double bond (right branch) of cis-and trans-configuration into LCB leading to the formation of ⌬ 8 -and ⌬ 9 -phytosphingenine (t18:1 8 and t18:1 9 ). In higher plants, the desaturation of phytosphinganine (t18:0) seems to be blocked (right branch), because only C-8-unsaturated LCB have been shown to occur (10). R, different fatty acyl residues of the basic ceramide structure. rated substrate, it is not related to the other desaturase groups that use saturated substrates (acyl-ACP-, acyl-CoA-, desC-, and the senescence-induced desaturase (41)). Its similarity to the ⌬ 6 -acyl group desaturase suggests that structural features controlling regioselectivity rather than those allowing proton abstraction from saturated substrates were more suitable for evolutionary modification. On the other hand, in plants, LCBs with ⌬ 8 -double bonds are far more widespread than the rare introduction of ⌬ 6 -double bonds into acyl groups. Therefore, the evolutionary correlation between these two groups of desaturases (i.e. whether one evolved from the other or whether both represent independent lines from a common ancestor) is an interesting but open question.