ELO2 and ELO3, Homologues of theSaccharomyces cerevisiae ELO1 Gene, Function in Fatty Acid Elongation and Are Required for Sphingolipid Formation*

ELO2 and ELO3 were identified from the Saccharomyces cerevisiae genome data base as homologues of ELO1, a gene involved in the elongation of the fatty acid 14:0 to 16:0. Mutations in these genes have previously been shown to produce pleiotropic effects involving a number of membrane functions. The simultaneous disruption ofELO2 and ELO3 has also been shown to produce synthetic lethality, indicating that they have related and/or overlapping functions. Gas chromatography and gas chromatography/mass spectroscopy analyses reveal that null mutations of ELO2and ELO3 produce defects in the formation of very long chain fatty acids. Analysis of the null mutants indicates that these genes encode components of the membrane-bound fatty acid elongation systems that produce the 26-carbon very long chain fatty acids that are precursors for ceramide and sphingolipids. Elo2p appears to be involved in the elongation of fatty acids up to 24 carbons. It appears to have the highest affinity for substrates with chain lengths less than 22 carbons. Elo3p apparently has a broader substrate specificity and is essential for the conversion of 24-carbon acids to 26-carbon species. Disruption of either gene reduces cellular sphingolipid levels and results in the accumulation of the long chain base, phytosphingosine. Null mutations in ELO3 result in accumulation of labeled precursors into inositol phosphoceramide, with little labeling in the more complex mannosylated sphingolipids, whereas disruption of ELO2 results in reduced levels of all sphingolipids.

ELO2 and ELO3 were identified from the Saccharomyces cerevisiae genome data base as homologues of ELO1, a gene involved in the elongation of the fatty acid 14:0 to 16:0. Mutations in these genes have previously been shown to produce pleiotropic effects involving a number of membrane functions. The simultaneous disruption of ELO2 and ELO3 has also been shown to produce synthetic lethality, indicating that they have related and/or overlapping functions. Gas chromatography and gas chromatography/mass spectroscopy analyses reveal that null mutations of ELO2 and ELO3 produce defects in the formation of very long chain fatty acids. Analysis of the null mutants indicates that these genes encode components of the membrane-bound fatty acid elongation systems that produce the 26-carbon very long chain fatty acids that are precursors for ceramide and sphingolipids. Elo2p appears to be involved in the elongation of fatty acids up to 24 carbons. It appears to have the highest affinity for substrates with chain lengths less than 22 carbons. Elo3p apparently has a broader substrate specificity and is essential for the conversion of 24-carbon acids to 26-carbon species. Disruption of either gene reduces cellular sphingolipid levels and results in the accumulation of the long chain base, phytosphingosine. Null mutations in ELO3 result in accumulation of labeled precursors into inositol phosphoceramide, with little labeling in the more complex mannosylated sphingolipids, whereas disruption of ELO2 results in reduced levels of all sphingolipids.
In the yeast Saccharomyces cerevisiae, sphingolipids comprise approximately 10% of the total membrane lipid species (1). The hydrophobic moiety of these lipids is ceramide, which consists of a long chain base coupled to a very long chain fatty acid that is almost exclusively 26:0 1 or hydroxy 26:0 (2). Although sphingolipids are relatively minor membrane lipid species, they are highly concentrated on the plasma membrane and appear to be essential for a number of critical membrane and cellular functions (3)(4)(5). Inhibition of sphingolipid biosynthesis results in growth inhibition and cell death (6,7). Ceramide has also been implicated as a component of an essential cell signaling pathways in Saccharomyces (8).
In wild type cells, most fatty acids are 12-18-carbon species that are found in glycerolipids. Those species appear to be formed de novo by the well characterized soluble cytoplasmic fatty acid synthase complex. The very long chain (20ϩ carbon) fatty acids found in sphingolipids, however, are formed by membrane-bound fatty acid elongation systems that are not well characterized. These enzyme systems extend 14 -18-carbon fatty acids by 2-carbon units by a sequence of reactions similar to those catalyzed by fatty acid synthases, with the exception of one reduction step, which in mammalian cells appears to be mediated by cytochrome b 5 (9).
We recently identified a gene (ELO1) that encodes a membrane protein involved in the elongation of 14:0 to 16:0 (10). Comparison of the amino acid sequence of that gene with the recently completed Saccharomyces genome data base revealed two additional genes with high identity to ELO1. These are referred to as ELO2 and ELO3 in this paper, based on their function in fatty acid elongation. Both genes have been previously identified as open reading frames of unknown function; mutations in these genes induce pleiotropic phenotypes that appear to play a key role in membrane and cytoskeletal functions (11).
ELO2 was initially cloned by complementation of mutants of GNS1 (12). Those mutants confer resistance to echinocandins and have defects in ␤-glucan synthase activities. It was also reported as FEN1 (11,13), a gene whose mutants exhibited bud localization defects and resistance to the sterol isomerase inhibitor SR 31747. ELO3 was previously cloned and identified, respectively, as APA1 with mutant alleles that cause a decrease in the level of the plasma membrane ATPase (14), SUR4 (15), whose mutants suppress the reduced viability on starvation mutant phenotype (rvs161), and SRE1 (11) whose mutants suppress the effects of the steroid isomerase inhibitor, SR 31747. At least two laboratories have reported that simultaneous disruption of ELO2 and ELO3 produces a lethal phenotype (11,13) which indicates that their encoded proteins have related and overlapping functions. Similar studies in this laboratory also support the synthetic lethality of the double disruptions.
The characterization of the genes described in this paper suggests that they encode proteins required for the production of very long chain fatty acids. Each gene apparently encodes a single enzyme component of one or more systems that elongate C 16 and C 18 acids to C 20 -C 26 very long chain fatty acids. Disruption of either gene causes either the reduction or loss of 26:0, the end product of the elongation pathway, with a con-comitant reduction in ceramide synthesis and striking changes in sphingolipid composition.

MATERIALS AND METHODS
Bacterial and Yeast Strains-The strains used in this study and their genotypes are presented in Table I. Plasmids constructed for this study are shown in Table II. Standard yeast genetics methods were used for construction of strains bearing the appropriate mutations (16). Cell growth conditions and growth medium have been previously described (17). Escherichia coli strain DH5␣ was obtained from Life Technologies, Inc. Saccharomyces cerevisiae cells were cultured as described previously (10).
Lipid Analysis-Fatty acid analysis of long chain and very long chain fatty acid methyl esters was performed by HCl-methanolysis of whole cell lipids as described previously (10). Methyl esters were prepared from washed cells grown to a density of 2-3 ϫ 10 7 ml in 50 ml of CM or CM (ϪURA) medium containing either 2% glucose or 2% galactose. Gas chromatography temperature programming was modified to optimize analysis of very long chain fatty acids on a 0.32 mm ϫ 30 m Supelco-Wax10 column.
Sphingolipid Synthesis-Two-ml cultures were grown to 2 ϫ 10 6 cells/ml in CM medium at 30°C and then labeled with either 20 Ci/ml [ 3 H]serine for 6 h or 1 Ci/ml [ 3 H]dihydrosphingosine (10 M) for 30 min. Cultures were chilled on ice with an additional 0.5 ml of unlabeled stationary phase cells and centrifuged at 2800 ϫ g for 10 min at 4°C. Cells were washed 2 times with 5 ml of cold H 2 O and treated with 5% trichloroacetic acid at 4°C for 20 min. Lipids were extracted twice in 1 ml of ethanol/water/diethyl ether/pyridine/NH 4 OH (15:15:5:1:0.018) at 60°C as described (18). The [ 3 H]serine-labeled extract was subjected to mild alkaline methanolysis by one treatment of 0.5 ml of monomethylamine reagent as prepared by Clarke and Dawson (19) for 30 min at 52°C. The alkali-stable [ 3 H]serine-labeled lipids were dried under N 2 , resuspended in 0.1 ml of chloroform/methanol/H 2 O (16:16:5), applied to Whatman Linear K6D silica gel TLC plates (25 l/spot), and resolved in CHCl 3 /methanol/ 4.2 N NH 4 OH (9:7:2). The [ 3 H]dihydrosphingosinelabeled total lipid extracts were dried under N 2 , resuspended in 0.2 l, and subjected to TLC as described (7). Radioactive bands were quantified on a Molecular Dynamics PhosphorImager using a tritium screen and visualized by x-ray film after treatment with EN 3  Construction of GAL1-ELO2 Over-expression Strain-The plasmid pCRELO2 was digested with BamHI and SalI. The released 1.2-kb DNA fragment was ligated into plasmid YCpGAL1URA downstream of the GAL1 promoter sequences. This construct (YCpGALELO2(U)) was analyzed by diagnostic restriction enzyme digest to confirm the correct orientation of the ELO2 fragment in respect to the GAL1 promoter. That plasmid was electroporated into strain CSY3H for over-expression studies of the ELO2 gene in the elo3 Ϫ background.
Construction of the ELO3 Disruption Fragments-A 2.9-kb fragment containing the coding sequence for ELO3 was derived by PCR using primers ELO3A and ELO3B (Table III) and strain DTY10A genomic DNA as a template. The PCR product was digested with KpnI and SalI and ligated into plasmid YEp352 in which the HindIII site had been destroyed to produce YEpElo3. This plasmid was digested with HindIII followed by Klenow fill-in to remove the ELO3 coding region. A 1.2-kb DNA blunt-ended SalI-XhoI fragment containing the S. cerevisiae HIS3 gene was inserted into the vector creating plasmid YEpelo3HIS. The 2 The abbreviations used are: kb, kilobase pair(s); PCR, polymerase chain reaction; Elo1p, polypeptide encoded by the ELO1 gene; Elo2p, polypeptide encoded by the ELO2 gene; Elo3p, polypeptide encoded by the ELO3 gene; elo1⌬, a null, gene disrupted allele of ELO1; elo2⌬, a null, gene disrupted allele of ELO2; elo3⌬, a null, gene disrupted, allele of ELO3; GC/MS, gas chromatography/mass spectroscopy.  Construction of GAL1-ELO3 Over-expression Strain-A DNA fragment containing the GAL1 promoter derived from vector YCpGAL1 by EcoRI digestion was inserted upstream of the ELO3 mRNA coding sequences in plasmid YEpElo3 at a SacI restriction enzyme site. The 5Ј-untranslated region of ELO3, including its presumptive basal promoter elements (TATA), was removed from the plasmid by partial HindIII digestion. A 2.0-kb DNA fragment, containing the GAL1 promoter and adjacent ELO3 protein coding sequences, was released from the resulting plasmid by EcoRI digestion. The recovered fragment was ligated into complementary EcoRI sites on plasmid YCpGAL1 which contains the URA3 gene as a selectable marker. These constructs YCpGALELO3(U) were analyzed by diagnostic restriction enzyme digest to confirm the correct orientation of the EL03 fragment in respect to the GAL1 promoter. These constructs were electroporated into strain DTY004 for over-expression studies of the ELO3 gene in an elo2 Ϫ background.
Fatty Acid Analysis-Fatty acid methyl esters were extracted from washed yeast cell pellets as described previously (17). These were generated by boiling the cell pellets derived from 50-ml cultures containing 1-4 ϫ 10 7 cells/ml in a mixture of 3 N methanolic-HCl. The samples were then extracted with hexane/anhydrous ethyl ether (1:1). Fatty acid methyl esters were analyzed using a Varian 3400 CX gas chromatograph and with a 30 m ϫ 0.32 mm SupelcoWax-10 column with a film thickness of 0.32 m using helium as the carrier gas. The injector temperature (splitting ratio 50:1) for gas chromatograph was 240°C and the oven temperature was increased from 70 to 240 at 20°C/min. Gas chromatography data was collected and quantified using the Shimadzu EZChrom data system.
Mass Spectroscopy-Mass spectra data were collected and quantified on a Finnigan Mat 8230 mass spectrograph using the Finnigan Mat SS300 data system. The same gas chromatography column and temperature programming parameters were used to fractionate fatty acids prior to ionization. The ion source temperature for mass spectra was 250°C, and the filament emission current was 1 mA. Mass spectra were recorded at an ionization voltage of 70 eV. Ions from 35 to 600 amu were scanned at 1 s/decade and interscanned at 0.85 s/decade.
Polymerase Chain Reaction-The polymerase chain reaction (PCR) was performed according to standard protocols (20). All PCR reactions were performed in a "OmniGene" thermal cycler (Hybaid, Ltd.) using a heat-stable recombinant "Vent" DNA polymerase with 5Ј-3Ј-and 3Ј-5Ј-exonuclease activity. This allowed for greater fidelity of PCR products with an error rate of approximately 1 in 10 5 base pairs of DNA. All PCR reactions used either a 50 or 55°C annealing temperature. Extension times at 72°C were typically between 2.0 and 4.0 min or were adjusted to the size of the expected product. PCR reactions were routinely run for a total of 30 cycles. A list of all PCR primers used in this work appears in Table III.

RESULTS
Two genes were identified from the S. cerevisiae genome data base that had high identity to ELO1, a gene involved in the fatty acid synthase-independent elongation of 14:0 to 16:0 (10). ELO2 is located on yeast chromosome III at the locus designated YCR34W. ELO3 is located on chromosome XII at the locus designated YLR372W. Fig. 1 shows regions of homology between ELO1, ELO2, and ELO3 protein coding sequences. Elo2p and Elo3p are, respectively, 76 and 72% similar and 56 and 52% identical to that of Elo1p. The three genes contain multiple regions of contiguous identical residues throughout the protein sequence. Hydropathy analyses of ELO1p, ELO2p, and Elo3p by the TMpredict algorithm (Fig. 2) suggest that the three proteins contain five membrane-spanning regions. The identified core regions of these sequences align identically with previously predicted transmembrane regions of Elo1p (10). The regions of highest identity in all three genes lie between presumptive transmembrane-spanning regions II and III (Fig. 2). That region has 16 identical amino acid residues located between residues 185 and 200 of Elo3p (from the amino terminus) and contains a cluster of four consecutive basic residues followed by an HXXHH motif, which has been previously identified with fatty acid desaturase, ribonucleotide reductase, hemerithrin, and other ironcontaining proteins (9). Elo2p also has a hydrophobic stretch in that region that contains several polar residues, suggesting  (Fig. 2). Residues enclosed by the box contain the histidine-rich region common to all three proteins. that it might serve as a hydrophobic cleft associated with an active site of the enzyme. The amino-and carboxyl-terminal regions of Elo2p and Elo3p are most dissimilar to Elo1p. In all three proteins the C-terminal sequences that follow transmembrane segment V contain unusually high numbers of basic residues, suggesting that those domains have some homologous function. Both ELO2 and ELO3 peptide sequences lack a GSA motif near the carboxyl terminus of ELO1 that is proposed to be an NADPH binding site for a number of lipid biosynthetic enzymes (10).
Disruption of ELO2 or ELO3 Produces Changes in Long Chain  Fatty Acids-Analysis of total fatty acids from elo2⌬ and elo3⌬ strains showed only minor changes in the composition of C 14 -C 18 fatty acids (Table IV). Those species comprise approximately 95% of the total cellular fatty acids. Compared with wild type, elo2⌬ C 14 fatty acids were reduced by 45%, 16:1 was reduced approximately 8%, and 18:1 was increased by about 10%. The elo3⌬ strain showed complementary changes with increased 16:1 (ϳ30%) and decreased 18:1 (ϳ30%).  Fig. 3. The most abundant species in wild type are 26:0 and ␣hydroxy 26:0 (2,21). This was confirmed by GC/MS of total fatty acid methyl esters and by separate GC/MS analyses of authentic standards. They comprise approximately 2.2 and 0.9% of the total fatty acid mass in mid-log cells grown in glucose medium. Minor peaks representing 22-and 24-carbonsaturated, monounsaturated, and hydroxy fatty acid species were also identified in the wild type strains. These combined represent less than 0.6% of the total fatty acids and are apparently derived from metabolic intermediates in the fatty acid elongation pathway. GC/MS analysis also identified several prominent peaks with retention times similar to the long chain species. These are the free fatty acids 16:0, 16:1, 18:0, and 18:1. They are apparently equilibrium products of the transmethylation reaction and do not represent a unique pool of unesterified species in vivo. The relative levels of the esterified and unesterified forms were combined for each species in data shown in Table IV.

GC/MS Analysis of Wild
Disruption of ELO2 or ELO3 Produces Changes in the Very Long Chain Fatty Acid Chain Length Distribution-Quantitative gas chromatography and GC/MS analysis of fatty acid methyl ester fractions indicate that the 26:0 and HO-26:0 species are absent in the elo3⌬ strain. Sharply reduced levels of 26:0 (approximately 20% wild type levels) and HO-26:0 (approximately 40% of wild type levels) were found in the elo2⌬ strain (Fig. 3, Table V). The reductions in C 26 species were accompanied by increases in minor (and in some cases, undetectable) wild type species. Lipids from the elo2⌬ strain displayed new peaks with retention times between 9 and 11 min. GC/MS analysis indicated that they were hydroxy 16-and 18-carbon fatty acids with fragmentation patterns consistent with the ␣-hydroxy fatty acid standards. (In wild type, those species were not detected.) No C 20 fatty acids were detected in the elo2⌬ strain and the total C 20 -C 26 species were 30% of wild type levels.
The elo3⌬ strain contained elevated levels of C 20 and C 22 fatty acids (Table V). The most abundant species was 22:0 which averaged 3.1% of the total fatty acyl mass, a 10-fold increase over wild type levels. Large increases in the levels of hydroxy C 16 -C 24 fatty acids were also observed in this strain. Unlike the elo2⌬ strain, ELO3 disruption resulted in an approximate 20% increase in the total levels of very long chain species.

Over-expression of ELO2 or ELO3 Alters Fatty Acyl Composition of, but Does Not Compensate for, the Complementary
Gene Disruption-The synthetic lethality of elo2⌬ and elo3⌬ suggests that their encoded proteins have overlapping functions. To test whether increased activity of Elo2p and Elo3p can compensate for loss of the other's function, their genes were each placed under the control of the strong, inducible GAL1 promoter and transformed into cells with the disrupted homologue. The very long chain fatty acid distributions of these strains and wild type are shown in Table V and Fig. 4.
Very long chain fatty acid distributions of wild type, elo2⌬, and elo3⌬ controls (which did not contain the plasmid) on the non-fermentable carbon source, galactose, were similar to those found for these same strains when grown on glucose medium (Table V) (Table V). The strain also had slight, but significant, increases in hydroxy 26:0 (which is elevated to wild type levels) and 26:0 (which is elevated from 15 to 36% of the wild type levels).
More dramatic changes were observed on over-expression of Elo2p in the elo3⌬/GAL1-ELO2 strain. This failed to restore the missing 26-carbon species and also did not reduce the characteristic high levels of 22:0. Large increases in C 20 and C 24 carbon species were observed, however. There was a 44-fold increase in 24:0 and a 5-fold increase in hydroxy 24:0 (Table V, Fig. 4). Similar increases were observed in 20:1 (2-fold) and 22:1 (5-fold). Induction of Elo2p in this strain also produced high levels of a 24:1 species that was not detected in the elo3⌬ or wild type strains. All of the shorter chain hydroxy Ϫ16:0, Ϫ16:1, Ϫ18:0, and Ϫ18:1 species present in the elo3⌬ strain were reduced to undetectable levels in cells containing overexpressed Elo2p.
Sphingolipid Synthesis in Elo2⌬ and Elo3⌬ Strains-The effects of disrupting ELO1, ELO2, and ELO3 on sphingolipid biosynthesis are shown in Fig. 5, A and B 6 -9), and the lipid extracts were chromatographed by TLC either before or after mild alkaline methanolysis, respectively. The methanolysis deacylates fatty acids on phospholipids that are linked to the glycerol-head group moiety by O-acyl bonds, but leaves the N-acyl ester bonds of sphingolipids intact. A thin layer chromatogram of those lipids is shown in Fig. 5A.
The lipids in wild type and elo1⌬ cells had a similar pattern; most of the label incorporated into sphingolipids was found in the major inositol-containing species including inositol phosphoceramide as well as the more complex mannosylated forms, mannosyl phosphorylceramide and mannosyl diinositoldiphos-  The total fatty acid mass in the elo2⌬::HIS3 and elo3⌬::HIS3 lipids were normalized to that of wild type to indicate the relative mass of the very long chain species in those strains. The region that contains hydroxy 24:0 is within the break in the x axis and that species is not shown. FFA, free fatty acid.
phorylceramide. Sphingolipid intermediates were in low abundance with almost no label incorporation into the long chain bases, phytosphingosine, or dihydrosphingosine, and only a faint band for ceramide, a precursor composed of a long chain base and a very long chain fatty acid linked by an N-acyl bond. The elo2⌬ and elo3⌬ strains, by comparison, displayed large accumulations of label in the long chain bases, primarily phytosphingosine, and the ceramide band was absent. Label incorporation into the mature sphingolipids was greatly reduced. In elo3⌬ almost all of the label accumulated into the inositol phosphoceramide species, with very little labeling of the mannosylated forms, while in elo2⌬ there were reduced amounts of all of the sphingolipids. The major serine-labeled phospholipids, which run as their more polar glycerol-head groups after the alkaline hydrolysis procedure, were normal in all of the strains. Pulse labeling with [ 3 H]dihydrosphingosine gave qualitatively similar results where elo2⌬ and elo3⌬ strains accumulated phytosphingosine and were very deficient in ceramide and the inositol-containing sphingolipids compared with wild type and elo1⌬ cells.
Quantitation of the [ 3 H]dihydrosphingosine labeling by PhosphorImager is shown in Fig. 5B. The majority of the converted label (81%) was found in the sphingolipid fraction in the wild type and elo1⌬ strains. In these lipids, approximately 90% of the label was in the form of inositol phosphoceramide and mannosyl phosphorylceramide and 10% was in the form of mannosyl diinositoldiphosphorylceramide. Approximately 2% of the converted dihydrosphingosine label was in phytosphingosine, 10% in ceramide, and 8% found in phospholipid fractions. The label in phospholipids is derived from the degradation of dihydrosphingosine to an aldehyde intermediate which is then converted to a fatty acid. In elo2⌬ and elo3⌬, the amount of converted dihydrosphingosine label was reduced to approximately 30% of wild type due to reductions in ceramide and sphingolipids. Phytosphingosine was increased approximately 2-fold in the elo2⌬ strain and by 1.5-fold in elo3⌬, whereas the label incorporated into the phospholipid fractions was similar to that observed in wild type. A similar pattern of dihydrosphingosine labeling was found in wild type cells that were treated with the ceramide synthase inhibitor, australifungin, as seen in Fig. 5A. These data suggest that disruption of ELO2 or ELO3 has the effect of reducing the level of ceramide synthesis which results in the concomitant reduction in cellular sphingolipid levels.

DISCUSSION
The roles of Elo2p and Elo3p as components of the very long chain fatty acid elongation system are supported by several observations. Hydrophobicity analysis indicates that Elo2p and Elo3p are intrinsic membrane proteins with multiple membrane-spanning regions, which is consistent with the reported tight membrane association of fatty acid elongation activities (9). ELO2 and ELO3 encode polypeptides that have a high degree of homology to the ELO1 gene product which is involved in the highly specific elongation of 14 carbon fatty acids (10). Disruption of either gene alters the composition of very long chain fatty acids and causes the accumulation of intermediate length fatty acid precursors.
The lethality caused by the simultaneous disruption of the ELO2 and ELO3 genes indicates that their products have a high degree of overlapping functions. The different pattern of accumulation of very long chain fatty acid intermediates in the elo2⌬ and elo3⌬ strains, however, suggests that Elo2p and Elo3p may play roles in independent, and to some degree parallel, metabolic pathways. These may be located in different parts of the cell. Previous studies of fatty acid elongation enzyme activities in Saccharomyces suggest, in fact, that there may be elongation systems both in the endoplasmic reticulum and on the mitochondrial surface (22). Clues about the enzymatic characteristics of Elo2p can be seen from the analysis of the ELO2 ϩ , elo3⌬ strain. The absence of a C 26 fatty acid indicates that Elo2p cannot catalyze the elongation of 24:0 to 26:0. Elo2p apparently has the highest catalytic specificity for C 20 acyl-CoA, which results in the observed accumulation of 22:0. Experiments in which ELO2 is over-expressed in elo3⌬ cells indicate that Elo2p can convert 22:0 to 24:0 with less efficiency. This reduced activity toward C 22 substrates can be compensated for by over-expression of Elo2p which shifts in the accumulated species from 22:0 to 24:0.
Elo3p appears to act on a broader range of substrates. Twenty-six-carbon fatty acids are formed when the masking Elo2p activity is removed in the ELO3 ϩ , elo2⌬ disruption strain. Taken together with the absence of C 26 species in the elo3⌬ strain, this indicates that conversion of 24:0 to 26:0 is exclusively performed by the Elo3p-dependent elongation system.
The lower levels of sphingolipids associated with the disruption of either ELO2 or ELO3 genes appear to be the result of reduced ceramide synthesis. Ceramide is the sphingolipid precursor that is formed from the direct condensation of very long chain fatty acids with a long chain base by ceramide synthase. This sensitivity of sphingolipid biosynthesis to changes in the very long chain fatty acid distribution in the disrupted strains could be caused by several factors. Reduced levels of very long chain acids could limit the pool of precursors available to ceramide synthase, producing a rate-limiting step at ceramide formation. Ceramide synthase could also have an optimal substrate specificity for 26:0 CoA and reduced activity in the presence of shorter chain substrates found in the disrupted strains. A third possibility is that the spatial distribution of fatty acid precursors in the cell is altered by ELO2 or ELO3 null mutations, possibly due to different cellular locations of the ELO2 and ELO3 elongation systems. This might result in the inability of ceramide synthase to make contact with available substrates.
The marked changes in cellular sphingolipid fatty acid composition in elo2⌬ and elo3⌬ strains provide a rational explanation for the pleiotropic effects reported for mutant alleles of these genes. Mutations or disruptions of ELO2 and ELO3 were previously reported as affecting ␤-glucan synthase activity (12), the plasma membrane, (H ϩ )-ATPase (14), bud localization defects (13), and resistance to sterol synthesis inhibitors (11,13). All of these functions are apparently produced by the activities of intrinsic membrane proteins, which could be affected by changes in the composition of their lipid environment or by the absence of an essential sphingolipid-membrane protein interaction.
Although Elo2p and Elo3p appear to be essential for the formation of very long chain fatty acids, there are apparently a number of other components that catalyze other steps in the elongation cycle that have yet to be identified. The lack of information about the structure, function, and cellular location of these enzymes gives an unclear picture of how they are organized and how intermediates are transferred from one reaction center to the next. Until now, the intrinsic hydrophobicity of these enzymes has hindered previous attempts to purify and characterize the components of these systems. The identification of ELO2 and ELO3 in this paper provides new tools that can be used to resolve questions concerning the mechanism, organization, and structure of these complex systems.