A Higher Plant D 8 Sphingolipid Desaturase with a Preference for ( Z )-Isomer Formation Confers Aluminum Tolerance to Yeast and Plants [C][OA]

Three plant cDNA libraries were expressed in yeast ( Saccharomyces cerevisiae ) and screened on agar plates containing toxic concentrations of aluminum. Nine cDNAs were isolated that enhanced the aluminum tolerance of yeast. These cDNAs were constitutively expressed in Arabidopsis ( Arabidopsis thaliana ) and one cDNA from the roots of Stylosanthes hamata , designated S851 , conferred greater aluminum tolerance to the transgenic seedlings. The protein predicted to be encoded by S851 showed an equally high similarity to D 6 fatty acyl lipid desaturases and D 8 sphingolipid desaturases. We expressed other known D 6 desaturase and D 8 desaturase genes in yeast and showed that a D 6 fatty acyl desaturase from Echium plantagineum did not confer aluminum tolerance, whereas a D 8 sphingobase desaturase from Arabidopsis did confer aluminum tolerance. Analysis of the fatty acids and sphingobases of the transgenic yeast and plant cells demonstrated that S851 encodes a D 8 sphingobase desaturase, which leads to the accumulation of 8( Z / E )-C 18 -phytosphingenine and 8( Z / E )-C 20 -phytopshingenine in yeast and to the accumulation of 8( Z / E )-C 18 -phytosphingenine in the leaves and roots of Arabidopsis plants. The newly formed 8( Z / E )-C 18 - phytosphingenine in transgenic yeast accounted for 3 mol% of the total sphingobases with a 8( Z ):8( E )-isomer ratio of approximately 4:1. The accumulation of 8( Z )-C 18 -phytosphingenine in transgenic Arabidopsis shifted the ratio of the 8( Z ):8( E ) isomers from 1:4 in wild-type plants to 1:1 in transgenic plants. These results indicate that S851 encodes the ﬁrst D 8 sphingolipid desaturase to be identiﬁed in higher plants with a preference for the 8( Z )-isomer. They further demonstrate that changes in the sphingolipid composition of cell membranes can protect plants from aluminum stress.

Trivalent cations are toxic to most plant cells. The increased prevalence of soluble aluminum (Al 31 ) cations in acid soils is a major limitation to plant production around the world. Aluminum disrupts a range of cellular processes, including nutrient acquisition, cell wall loosening, nuclear division, cytoskeleton stability, cytoplasmic calcium homeostasis, hormone transport, and signal transduction (Taylor, 1988;Kochian, 1995;Matsumoto, 2000). Many of these symptoms occur rapidly and some workers have concluded that aluminum toxicity is initiated by interactions occurring in the extracellular compartment (Horst, 1995) and cell membranes. Aluminum accumulates rapidly in the highly charged cell wall and near the fixed charges and polar groups on the plasma membrane surface, which can displace calcium from critical sites in the apoplasm, alter physical properties of the plasma membrane, change membrane lipid composition, block ion channels, and disrupt signal transduction processes by interfering with phospholipase C metabolism (Haug and Caldwell, 1985;Rengel, 1992;Shi and Haug, 1992;Kinraide et al., 1994;Ryan et al., 1994;Jones and Kochian, 1995;Piñ a-Chable and Hernández-Sotomayor, 2001;Mantinez-Estévez et al., 2003;Stival da Silva et al., 2006). Whereas it remains unclear which, if any, of these reactions are primary causes for aluminum toxicity in plants, it is plausible that aluminum-dependent changes in cell membrane structure and function contribute to the overall stress encountered in acid soils. Consistent with this idea are reports demonstrating that aluminum can reduce membrane fluidity of the Archaebacterium Thermoplasma acidophilum by binding to the polar head groups of phospholipids (Deleers et al., 1986) and alter the lipid composition of plant roots (Lindberg and Griffiths, 1993;Zhang et al., 1997;Peixoto et al., 2001;Stival da Silva et al., 2006). Previous reports have also shown that genetically engineered changes to the lipid composition of plant and yeast (Saccharomyces cerevisiae) membranes can affect their susceptibility to chilling, photoinhibition, drought, fungal toxins, and ion toxicity (Avery et al., 1996;Nishida and Murata, 1996;Delhaize et al., 1999;Thevissen et al., 2000;Los and Murata, 2004;Zhang et al., 2005;Stival da Silva et al., 2006).
In contrast to the complexity of aluminum toxicity, the genetics of aluminum resistance can be relatively simple. For instance, the mechanism for aluminum resistance in some cereal species, such as wheat (Triticum aestivum; Raman et al., 2005) and barley (Hordeum vulgare; Minella and Sorrells, 1997) is controlled by single major genes. The aluminum resistance gene from wheat, TaALMT1, encodes a membrane protein that facilitates the aluminum-dependent release of malate anions from the root apices. These organic anions protect the root cells by chelating the aluminum cations in the apoplasm (Delhaize et al., 1993;Sasaki et al., 2004).
The aim of this study was to identify novel plant genes that confer aluminum tolerance using a strategy that does not make any assumptions about function. Using a yeast expression system, we isolated nine plant cDNAs that conferred increased tolerance to aluminum stress. One of these cDNAs from Styolsanthes hamata also enhanced the aluminum tolerance of Arabidopsis (Arabidopsis thaliana). We established that this cDNA encodes a D8 sphingolipid desaturase that preferentially produces the 8(Z)-isomer of phytosphingenine.

Plant cDNAs Conferring Aluminum Tolerance to Yeast
Bakers' yeast was transformed with cDNA libraries prepared from the nodules of soybean (Glycine max), roots of S. hamata, and cluster roots of white lupin (Lupinus albus). These libraries were chosen because they were either prepared from a species well adapted to acid soils (S. hamata) or from tissues predicted to contain proteins that facilitate organic anion efflux from root cells (soybean nodules; cluster roots of white lupin). Transformed yeast cells were screened on agar plates with aluminum concentrations sufficient to inhibit the growth of cells containing an empty vector. Plasmids isolated from aluminum-tolerant colonies were amplified in Escherichia coli and retransformed into wild-type yeast to confirm that the aluminumtolerance phenotype was caused by expression of the cDNA inserts and was not due to spontaneous mutations. Nine different cDNAs conferred increased aluminum tolerance to yeast cells. Figure 1, A to C, illustrates the increase in aluminum tolerance provided by one of nine cDNAs. The cDNAs were sequenced and their likely function determined by comparing their putative translation products with the nonredundant protein database using the BLASTx algorithm (http://www.ncbi.nlm.nih.gov/BLAST; Table I).
A cDNA Encoding a Lipid Desaturase Confers Aluminum Tolerance to Arabidopsis Nine cDNAs were expressed in Arabidopsis under the control of the constitutive cauliflower mosaic virus (CaMV) 35S promoter. Aluminum tolerance of several T 1 lines from each construct was compared with wildtype plants by estimating relative root growth on agar plates (data not shown). One cDNA from the S. hamata library, designated S851, enhanced the aluminum tolerance of all five independently transformed T 1 lines tested (data not shown) and this cDNA was investigated further. It is possible that some of the other cDNAs are able to increase the aluminum tolerance of T 2 material, but this was not tested in this study. We generated T 2 populations from the T 1 plants expressing S851 and selected two independent lines, At-S851-H4 and At-S851-H6, which were homozygous for antibiotic resistance. Expression of the S851 transgene in these two lines was confirmed by northern-blot analysis (data not shown). The aluminum tolerance of the At-S851-H4 and At-S851-H6 lines was estimated on agar in plates and pots. Relative root growth was consistently 20% to 100% greater in the homozygous transgenic lines compared to wild-type plants over a range of aluminum concentrations (Fig. 2). The stimulation of root growth observed in some treatments is likely to be caused by the often-reported amelioration of proton stress by low concentrations of aluminum (Kinraide, 1993).
In addition to aluminum, we tested whether yeast expressing S851 cDNA (designated here as Sc_ pYES3-S851) was more tolerant of other toxic cations than yeast cells transformed with the empty vector (designated here as Sc_ pYES3). Sc_ pYES3-S851 conferred enhanced tolerance to gadolinium (Gd 31 ), but no Figure 1. The S851 cDNA from S. hamata confers aluminum and gadolinium tolerance to transgenic yeast cells. Yeast strains transformed with an empty vector (Sc_ pYES3) or transformed with a vector containing the S851 cDNA (Sc_ pYES3-S851) were grown in SMM 2 ura medium. Cultures were diluted to an OD 600 5 1.0 in sterile water before a series of 10-fold dilutions was prepared for each of two independent cultures of each strain. Aliquots (10 mL) of each dilution were added to agar plates containing SMM 2 ura with a range of toxic cations. A, Control agar. B, 200 mM AlCl 3 . C, 400 mM AlCl 3 . D, 3.6 mM MnCl 2 . E, 700 mM GdCl 3 . F, 600 mM LaCl 3 . Results are typical of those obtained from at least two independent experiments. [See online article for color version of this figure.] consistent changes were observed for lanthanum (La 31 ) or manganese (Mn 21 ; Fig. 1, D-F). The homozygous Arabidopsis lines, At-S851-H4 and At-S851-H6, were also tested on agar plates containing a range of gadolinium concentrations, but neither showed greater tolerance than wild-type plants (data not shown).
A phylogenetic comparison between the predicted S851 protein and other D6 fatty acyl lipid desaturase and D8 sphingobase desaturase proteins of known function revealed that S851 grouped with most of the D8 sphingolipid desaturase proteins (Fig. 3). However, D6 fatty acyl lipid desaturases from Echium plantagineum and Borago officinalis also clustered with the D8 desaturase enzymes, making it difficult to predict S851 function from this cladogram. To establish the function of the S851 protein, we tested the ability of other D6 and D8 desaturase enzymes to confer aluminum tolerance to yeast. In addition, the fatty acid and longchain base compositions were measured in transgenic yeast and Arabidopsis expressing S851.
Initial analyses of the major fatty acid compositions in transgenic yeast and Arabidopsis showed no changes associated with the expression of the S851 cDNA (Table II). A cDNA encoding a D6 fatty acyl desaturase isolated from E. plantagineum (Zhou et al., 2006) was expressed in yeast (Sc_ pYES3-D6) and tested on agar plates containing toxic levels of aluminum. The Sc_ pYES3-D6 and Sc_ pYES3 cells were equally sensitive to aluminum and both grew significantly slower than the Sc_ pYES3-S851 cells (Fig. 4A). To confirm that the D6 fatty acyl desaturase was functional in yeast, we analyzed the lipid content of these cells grown in the presence and absence of linoleic acid, a substrate for D6 desaturase enzymes. When linoleic acid was excluded from the medium (control), the major fatty acid composition of all strains was similar (Fig. 5). When linoleic acid was included in the medium, it was activated to acyl-CoA and incorporated into membrane lipids where it accounted for over 40% of the total fatty acid content in all strains. The palmitoleic and oleic acid fractions showed a concomitant decrease of approximately 60%. Inclusion of linoleic acid was also associated with g-linolenic acid accumulation in the Sc_ pYES3-D6 strain only where it accounted for 1% of the total fatty acid content (Fig. 5). No g-linolenic acid was detected in the Sc_ pYES3 or Sc_ pYES3-S851 strains. This result confirms that the D6 desaturase enzyme from E. plantagineum was functional when substrate for the enzyme was available and indicates that S851 cDNA does not encode a D6 fatty acyl desaturase.
Yeast cells were then transformed with a known D8 sphingolipid desaturase from Arabidopsis (GenBank accession no. AF001394; strain Sc_ pYES3-D8) and its aluminum tolerance compared with control cells. The Sc_ pYES3-D8 strain was more tolerant to aluminum stress than the Sc_ pYES3 control (Fig. 4B). The finding that the aluminum tolerance of yeast is increased by expression of a known D8 desaturase is consistent with the hypothesis that S851 encodes a D8 sphingolipid desaturase, but not a D6 fatty acyl desaturase.

Sphingobase Analyses of Yeast and Plant Cells Expressing S851
Analysis of the sphingobases released from the Sc_ pYES3-S851 yeast strain identified D8-unsaturated long-chain bases C 18 -and C 20 -phytosphingenine, which were not present in the control strain (Fig. 6, A and B). These novel 8(Z)-and 8(E)-C 18 -phytosphingenines accounted for 3 mol% of the total sphingobases in Sc_ pYES3-S851 with an 8(Z):8(E) ratio of 4:1. Although the identities of most of the sphingobases in Figure 6 were confirmed by HPLC/mass spectrometry (MS) with electrospray ionization, the 8(Z):8(E) ratio of C 20phytopshingenines could not be estimated accurately because 8(E)-C 20 -phytosphingenine coeluted with the C 19 -phytosphinganine present in yeast. Changes in the sphingobase composition of Arabidopsis plants expressing S851 were also consistent with D8 sphingolipid desaturase activity. The main sphingobases in wild-type Arabidopsis plants, 8(Z)-C 18 -phytosphingenine, 8(E)-C 18 -phytosphingenine, and C 18 -phytosphinganine, comprised 15, 58, and 27 mol%, respectively, of the total with an 8(Z):8(E) ratio of approximately 0.3 ( Fig. 6C; Table III). This profile is generally consistent with previous analyses of the Arabidopsis ecotypes Columbia C24 (Sperling et al., 2005) and Wassilewskija (Bonaventure et al., 2003). In the At-S851-H4 and At-S851-H6 lines, these same sphingobases represented 43, 42, and 15 mol%, respectively ( Fig. 6, D and E; Table III). Separate analyses of the sphingobases in leaves and roots resemble the data obtained with whole plants (Table III). The increase in the 8(Z)-isomer of C 18 -phytosphingenine shifted the 8(Z):8(E) ratio closer to 1.0. These data confirm that S851 cDNA from S. hamata encodes a sphingolipid desaturase, which introduces a D8 double bond into phytosphinganine in both heterologous systems tested. The enzyme is stereo-unselective, but exhibits a strong preference for Z-isomer formation.

DISCUSSION
Nine plant cDNAs from three different plant libraries were isolated for their ability to confer aluminum tolerance to yeast cells. One cDNA (S851) that originated from the acid soil-tolerant forage species, S. hamata, also increased the aluminum tolerance of Arabidopsis Figure 3. Phylogenetic tree of D6 fatty acyl lipid desaturase and D8 sphingobase desaturase proteins. Phylogenetic and molecular evolutionary analyses of the proteins were conducted using MEGA, version 3.1 (Kumar et al., 2004). Default settings for the protein alignment and neighbor-joining phylogenetic calculations were used. All proteins have had their function verified experimentally as being a D6 fatty acyl desaturase enzyme (d6) or a D8 sphingobase desaturase enzyme (d8).  and At-S851-H6 (gray bars), and wild-type plants (black bars) were grown on sterile nutrient agar supplemented with 0, 300, 400, or 500 mM aluminum chloride (pH 4.8-4.9). Relative root growth (RRG), defined as (root length with aluminum)/(root length without aluminum), was estimated after 14 d by (1) growing seedlings on plates supported in an almost vertical orientation (plates) or (2) growing seedlings in sterile pots that forced the roots to grow directly into the agar (pots). Data show the mean and SEs of the RRG (n 5 20-30). To account for the accumulation of errors associated with deriving RRG, the SEs were calculated as follows: SE RRG 5 RRG [(SE x /x) 21 (SE y /y) 2 ] 1/2 , where x and y represent the mean root length in the control treatment and the mean root length in the aluminum treatment. The root growth in the zero aluminum (control) treatments for the wild-type and At-S851-H4 lines was (in mm): 16.0 6 0.1 and 19.8 6 0.1, respectively, for the plates and 16.6 6 0.2 and 14.7 6 0.2 for the pots. The root growth in the control treatments for the wild-type and At-S851-H6 lines was (in mm): 15.9 6 0.2 and 12.8 6 0.1, respectively, for the plates and 15.5 6 0.1 and 14.5 6 0.1 for the pots. seedlings. A previous attempt to isolate aluminum tolerance genes by screening a cDNA expression library in yeast identified several genes that were effective in yeast (Delhaize et al., 1999), but these did not prove to be effective when overexpressed in tobacco (Nicotiana tabacum) plants (E. Delhaize, unpublished data). Yeast is a powerful technique for isolating plant genes that confer aluminum tolerance regardless of their role in the plant from which they originate. However, it is clear that the transgenes will not always confer the same phenotypes in the single-celled system as they do in intact plants. For instance, many genes identified to be important for aluminum tolerance in yeast encode components of signal transduction pathways and cell wall metabolism (Kakimoto et al., 2005). Therefore, the phenotype conferred by a transgene may depend on a signal pathway or cell wall structure being conserved between the species. Despite this potential problem, there are examples of plant cDNAs conferring aluminum tolerance to both yeast and plants. For instance, Ezaki et al. (2000) showed that genes encoding a GDP dissociation inhibitor from Arabidopsis and a blue copper protein from tobacco enhanced the aluminum tolerance of transgenic Arabidopsis and yeast. Interestingly, one of the other cDNAs isolated here from white lupin (La97), which increased the aluminum tolerance of yeast, but not Arabidopsis, also encodes a GDP dissociation inhibitor. The reason La97 did not increase the aluminum tolerance of Arabidopsis might be due to subtle differences in functions of the two proteins or because the increase in tolerance was insufficient to be clearly identified in the segregating T 1 lines.
The putative protein encoded by S851 showed an equally strong similarity to a D6 fatty acyl lipid desaturase and a D8 sphingolipid desaturase. We demonstrated that the S851 protein did not have D6 fatty acyl lipid desaturase activity because the Sc_ pYES3-S851 yeast strain failed to accumulate g-linolenic acid under any conditions tested. Analyses of the sphingobases released from transgenic yeast and Arabidopsis indicated that S851 encodes a D8 (Z/E)-sphingolipid desaturase. We showed that another D8 sphingolipid desaturase enzyme from Arabidopsis also confers aluminum tolerance to yeast, which is consistent with the finding of Stival da Silva et al. (2006). By contrast, expression of a known D6 fatty acyl lipid desaturase in yeast provided no protection from aluminum stress.
Our results indicate that the sphingolipid composition can protect yeast and Arabidopsis from aluminum toxicity, but whether this occurs by altering membrane structure or by specific biochemical interactions is not clear. The Sc_ pYES3-S851 yeast strain also showed greater tolerance to toxic concentrations of another trivalent cation, gadolinium, which indicates that D8 unsaturated sphingolipids have the potential to provide tolerance to other ionic stresses.
Sphingolipids do not possess the ester-glycerol linkages common in most membrane lipids, but are composed of a long-chain amino alcohol base that forms an amide linkage to a fatty acid. This basic ceramide structure can be further modified by glycosylation, hydroxylation, and desaturation. D8 unsaturated sphingobases can exist as the E (trans)-or Z (cis)-isomer and all D8 sphingolipid desaturases isolated so far from higher plants preferentially form the E isomer (see Sperling and Heinz, 2003). The lipid desaturase characterized here is notable for being the first bifunctional D8 sphingolipid desaturase enzyme from higher plants to preferentially synthesize the 8(Z)-isomer of phytosphingenine.
A connection between D8 unsaturated sphingobases and aluminum tolerance in plants was previously investigated by Stival da Silva et al. (2006). They showed that the heterologous expression of a D8 (E/Z)-sphingolipid desaturase from Arabidopsis in hybrid maize (Zea mays) led to an 8-fold increase in the 8(E)-phytosphingenine content of a homozygous T 2 line, which changed the 8(Z):8(E) ratio from 5:1 (wild type) to approximately 1:3 (transgenic plants). However, those transgenic maize plants were scored as being more sensitive to aluminum stress. This contrasts with our findings, which show that expression of a similar desaturase from S. hamata increases the aluminum tolerance in Arabidopsis. There are several possible explanations for why one D8 sphingolipid Table II. Analysis of the major fatty acids in the total lipids extracted from yeast and Arabidopsis expressing S851 cDNA from S. hamata Fatty acids were analyzed as their methyl ester derivatives by gas chromatography. Results show the contribution of each major fatty acid as mol% of the total fatty acid extract. Data represent the mean and SE where n 5 3 for yeast, and n 5 7 to 9 for Arabidopsis. Dashes indicate that no peaks were detected.

Yeast and Plant Lines
Major desaturase enzyme confers aluminum tolerance, whereas another does not. For instance, maize has a naturally high content of 8(Z)-unsaturated bases and is generally more aluminum tolerant than Arabidopsis. These attributes could influence the magnitude of phenotype generated by expression of a D8 sphingolipid desaturase in maize. Furthermore, as noted above, the D8 sphingolipid desaturase from Arabidopsis preferentially forms the (E)-isomer of 8-phytosphingenine (Sperling et al., 1998;Sperling and Heinz, 2003), whereas the enzyme from S. hamata preferentially forms the (Z)-isomer. Introduction of a (Z)-double bond leads to a kink in the long-chain base, which is not generated by the formation of an (E)-double bond. This kink might induce some specific biochemical functions, generate changes to membrane structure, or possibly affect the functioning of the sphingolipid-rich lipid rafts (see below). The finding that both D8(E/Z) sphingolipid desaturase enzymes were able to increase the aluminum tolerance of yeast could be explained by the absence of any unsaturated sphingobases in wild-type yeast. Therefore, even a small accumulation of 8(Z)-phytosphingenine from either enzyme might be sufficient to improve its resistance to aluminum stress. Future studies will determine whether the stereochemistry of sphingobases influences their ability to confer aluminum tolerance by expressing the D8 sphingolipid desaturase genes from Arabidopsis and S. hamata in the same plant species. Participation of membrane lipids, not just sphingolipids, in the perception and response to environmental signals is well known. For instance, unsaturated acyl lipids can ameliorate the damage caused by chilling stress and photoinhibition at low temperatures (Cossins, 1994;Nishida and Murata, 1996;Murata and Los, 1997) as well as drought stress . Membranes become more rigid as temperature decreases and damage to cells can occur as membranes change from a liquid crystalline phase to a gel phase (see Los and Murata, 2004). Membranes with a higher percentage of unsaturated acyl lipids appear to incur less damage at low temperature, in part because fluidity is maintained.
In comparison to other membrane lipids, the functions of sphingolipids are poorly understood despite being a ubiquitous component of eukaryotic cells. Although in excess of 300 structurally different compounds have now been identified, sphingolipids typically constitute ,5% of total lipids in yeast and plants. Interest in sphingolipid metabolism has increased as their roles in cell growth, membrane stability, stress response, and apoptosis have been elucidated (Thevissen et al., 2000;Sperling and Heinz, 2003;Worrall et al., 2003;Lynch and Dunn, 2004). More recently, microdomains or rafts with a high sphingolipid-to-protein ratio have been detected in the plasma membranes of plant cells (Mongrand et al., 2004;Borner et al., 2005).  These microdomains are characterized by their insolubility to nonionic detergents and a high proportion of the proteins associated with them are involved with stress responses, cellular trafficking, and cell wall metabolism (Morel et al., 2006). Interestingly, about 16% of them also exhibit putative fatty acid modification sites. These sphingolipid-enriched lipid rafts appear to provide a platform for protein binding and organization and may constitute signaling centers for specialized physiological functions (Morel et al., 2006). Unsaturated sphingolipids may have specific roles in cellular responses to stresses (Ohnishi et al., 1988;Imai et al., 2000;Kawaguchi et al., 2000) and a few studies have examined these responses in detail. For instance, sphingosine-1-P can modulate stomatal closure by linking the perception of abscisic acid to reduction in guard cell turgor, whereas dihydrosphingosine-1-P, a structurally similar base without the D4 double bond, has no effect .
Desaturation of long-chain bases by sphingolipid desaturases probably occurs after ceramide formation, but the natural substrate of the desaturase encoded by S851 is unclear. García-Maroto et al. (2007) speculated that distinct D8 desaturases might have preferences for the cerebroside or glycosyl inositol phosphorylceramide groups of sphingolipids. The sphingobase phytosphinganine (t18:0) is a candidate substrate for S851 (Sperling et al., 2000). However, the relatively low activity of S851 in yeast (with relatively abundant t18:0) compared to the plant tissues suggests that it may prefer one of the cerebrosides, such as monoglycosylceramides, which are absent from yeast, but present in higher plants. This idea is also consistent with the stereoselectivity of the S851 enzyme because cerebrosides generally contain a greater proportion of 8(Z)unsaturated sphingobases.
the 8(Z)-isomer is paralleled by an increase in cerebrosides and how the two isomers 8(Z)-phytosphingenine and 8(E)-phytosphingenine affect membrane physiology and cellular metabolism.

Plant cDNA Libraries
The three cDNA libraries screened in this study were chosen because they were either derived from acid-tolerant species or because the libraries were prepared from tissues predicted to contain proteins that facilitate organic anion efflux from root cells as described below. Stylosanthes hamata is a widely used tropical forage plant and ecotype Verano is suited to the infertile and acidic soils of Central and South America as well as northern Australia. White lupin (Lupinus albus) forms specialized structures, called cluster roots, on its lateral roots during the onset of phosphorus deficiency (see Ryan et al., 2001). These specialized roots release large amounts of citrate into the rhizosphere, which mobilizes poorly soluble reserves of phosphorus from the soil. In the nodule cells of legumes like soybean (Glycine max), the peribacteroid membrane surrounds the compartment containing the nitrogen-fixing bacteroids. Proteins on this membrane control the transport of malate from the plant to the bacteria, which is energetically similar to the efflux of malate from the plant cells.
The cDNA library from the cluster roots of white lupin was prepared by first germinating seeds on river sand for 2 weeks. The seedlings were then transferred to nutrient solution adjusted to pH 6.0 that contained 625 mM KNO 3 , 250 mM CaCl 2 , 250 mM MgSO 4 , 12.5 mM FeCl 3 , 6.26 mM Na 2 EDTA, 11.5 mM H 3 BO 4 , 2.7 mM MnCl 2 , 0.35 mM ZnSO 4 , 0.3 mM CuCl 2 , and 0.03 mM (NH 4 ) 6 Mo7O 24 4H 2 O. The formation of cluster roots was induced by maintaining the phosphate concentration at 2.5 mM for the first 2 weeks and being omitted thereafter. The nutrient solution was aerated and changed every week as well as on the day before sample collection. After 5 to 6 weeks of growth, cluster roots at different stages of development were collected in liquid nitrogen. Total RNA was extracted from the frozen cluster root tissue with TRIzol LS reagent (Invitrogen) and mRNA was isolated with the Message-Maker mRNA isolation system (Gibco-BRL). cDNA was prepared using the SuperScript plasmid system for cDNA synthesis kit (Gibco-BRL) and size fractioned into long (.1.0 kb) and short (,1.0 kb) pools. cDNAs were directionally ligated into the SalI-NotI sites of a yeast (Saccharomyces cerevisiae) expression vector pYES3, which has the Gal-inducible promoter GAL1. pYES3 was modified from pYES2 as explained by Smith et al. (1995). The vectors were then cloned in Escherichia coli strain XL1-Blue (Stratagene). The plasmid library was recovered from E. coli and used to transform the yeast strain INVSc2 (MATa his3-D1 ura3-52; Invitrogen) as described by Gietz et al. (1995).
Primary transformants were selected on supplemented minimal medium without uracil (SMM 2 ura; Rose et al., 1990), washed off plates with sterile water, and stored in 15% glycerol at 280°C.
cDNA libraries from soybean root nodules and from the roots of S. hamata (ecotype Verano) were ligated into the SalI-NotI sites of pYES3 (Smith et al., 1995) and transformed into yeast (strain INVSc2) as described above.

Yeast Screen
Approximately 10 6 yeast transformants were screened as described previously (Delhaize et al., 1999) on SMM 2 ura with 2% Gal and buffered to pH 4.1 with 10 mM succinate. The selection medium also contained 500 or 600 mM AlCl 3 , which was sufficient to prevent growth of cells transformed with the empty vector. Plasmids were isolated from aluminum-tolerant colonies, amplified in E. coli, and then retransformed into wild-type yeast.

Preparation of pYES3-D6 and pYES3-D8 Yeast Strains
A clone of the coding region of the D6 fatty acyl lipid desaturase from Echium plantagineum (GenBank accession no. AY952780) was provided by Xue-Rong Zhou (Commonwealth Scientific and Industrial Research Organization Plant Industry). A full-length coding region of the D8 sphingolipid desaturase from Arabidopsis (Arabidopsis thaliana; At3g61580; GenBank accession no. AF001394) was amplified from RNA isolated from middle maturity developing embryos with the following primers: 5#-GTTCGTCGTCAATGGCGGAA-3# (forward) and 5#-CATTTAGCCATGAGTATTCAAAG-3# (reverse). Reverse transcription-PCR was performed using the SuperScript one-step reverse transcription-PCR with platinum Taq (Invitrogen) kit following the manufacturer's instructions. Briefly, a 50-mL reaction contained 25 mL 23 reaction mix and 200 ng RNA, 1 mL each of two oligo primers, and 2 mL platinum Taq enzyme mix. Thermal cycling was 50°C for 30 min (one cycle), 94°C for 2 min (one cycle), 94°C for 15 s, 58°C for 30 s, 68°C for 1.5 min (40 cycles), and 68°C for 7 min (one cycle). PCR fragments obtained were cloned into pGEM T Easy (Promega) and sequenced. The insert was subsequently excised with NotI from pGEM T Easy vector and ligated into the NotI site of pYES3 in a sense orientation relative to the GAL1 promoter and transformed into yeast strain INVSc2.

Arabidopsis Transformation and Measurements of Aluminum Tolerance
Plant cDNAs were cloned into the pART7 plasmid (Gleave, 1992) to generate an expression cassette with the cDNA under the control of the CaMV 35S promoter. The plasmid was digested with NotI and the fragment that contained the expression cassette was ligated into the NotI site of the binary vector pPLEX502 (Schü nmann et al., 2003). The binary vector was then transformed into Agrobacterium tumefaciens strain AGL1 by triparental mating. Arabidopsis (ecotype Columbia) was transformed by the floral-dip technique as described by Clough and Bent (1998). The seeds were germinated and screened on Murashige and Skoog medium containing 50 mg/mL kanamycin to identify transgenic plants. Two independent homozygous T 2 lines with single inserts were identified from two T 1 populations that displayed approximately 75% resistance to kanamycin (indicating a single insert). To screen the T 1 seed and homozygous T 2 lines for aluminum tolerance, seeds were sterilized in a container filled with Cl 2 gas for 3 h followed by germination on agar medium in either sterile plates or pots that contained 1.67 mM KNO 3 , 0.66 mM CaCl 2 , 0.66 mM MgSO 4 , 0.067 mM KH 2 PO 4 , 7.6 mM H 3 BO 4 , 1.8 mM MnCl 2 , 0.23 mM ZnSO 4 , 0.2 mM CuCl 2 , 16.67 mM FeCl 3 , and various concentrations of AlCl 3 with 5 mM succinic acid to buffer the medium at pH 4.8 to 4.9. Between 20 and 30 wild-type seeds and a similar number of seeds from one of the homozygous transgenic lines were spread along each half of a straight line across the middle of the agar plates. The plates were then held in a near-vertical position so that the line of seeds was horizontal. We noticed that the roots on these plates would sometimes lift off the agar and avoid the aluminum treatment. Therefore, aluminum tolerance was also measured in sterile pots, which forced the roots to penetrate the agar. In this arrangement, wild-type and transgenic seeds were spread around the periphery of the pots (6.5-cm diameter) to facilitate root length measurements at a later date. After positioning the seeds, the plates and pots were kept at 4°C for 2 d and then transferred to a temperature-controlled growth room (8-h darkness at 15°C and 16-h light at 20°C). Root lengths were measured after 14 d. These experiments were repeated at least three times for each homozygous line.

RNA Isolation and Northern-Blot Analysis
Total RNA isolated from Arabidopsis leaves was separated on a 1.5% denaturing formaldehyde gel, transferred onto a Hybond N 1 nylon membrane, and northern blots prepared according to Sambrook et al. (1989).

Fatty Acid Analyses
Total lipids of Arabidopsis leaves or yeast cells were extracted with methanol-chloroform according to the method described by Bligh and Dyer (1959). The fatty acid methyl ester (FAME) preparation and subsequent analysis of fatty acid composition by gas chromatography followed the method described by Liu et al. (2002). Briefly, after evaporating the solvent, lipid extracts were methylated in 2 mL of 0.02 M sodium methoxide for 1 h at 90°C. FAMEs were then extracted by adding 1.5 mL hexane and 2 mL water and vortexing. The upper phase, containing the FAMEs, was transferred to a microvial and separated in a SGE BPX70 column (0.25-mm diameter, 60-m length, and 2.5-mm film thickness) with gas chromatography (model 3400; Varian) using helium as carrier gas. Fatty acid composition was calculated as the percentage of each fatty acid represented in the total fatty acids.

Pretreatment of Yeast Cells in Linoleic Acid
Yeast cells were grown to OD 600 1.0 in 5 mL SMM 2 ura with 2% Glc at 30°C. Cells were collected by centrifugation, washed in sterile water, and resuspended in 5 mL SMM 2 ura containing 2% Gal, 0.5 mM linoleic acid (no. L1376; Sigma-Aldrich), and 1.0% NP-40, then incubated on a 20°C shaker for 3 d. Yeast cells were harvested by centrifuging and washed first with 1% NP-40, then with 0.5% NP-40, and finally with sterile water.

Preparation of Yeast and Plants for Sphingobase Analyses
Yeast cells transformed with the empty pYES3 vector (strain Sc_ pYES3) or with pYES3 containing S851 (strain Sc_ pYES3-S851) were grown aerobically at 30°C for 2 d in complete minimal medium (minus uracil) supplemented with 2% (w/v) raffinose for 24 h. Expression of the transgene was induced by a further 24-h growth after addition of Gal (final concentration 1.8% [w/v]). Cells were harvested by centrifugation for 10 min at 1,500g, resuspended in water, boiled for 15 min to inactivate lipases, and centrifuged again. Seeds (2.5 mg) of Arabidopsis (Columbia) wild-type and homozygous lines constitutively expressing S851, At-S851-H4, and At-S851-H6, were sterilized in 4% NaOCl and 0.02% (v/v) Triton X-100 for 10 min, washed three times in sterile water, and resuspended in 0.05% (w/v) agarose for plating. Plants were grown in a climate chamber for 3 weeks at 23°C, a 15-h light/9-h dark cycle, and 150 mE m 22 s 21 on sterile plates that contained 13 Murashige and Skoog salts, including B5 vitamins (M5519; Sigma), 1% (w/v) Suc, 2.3 mM MES-buffer adjusted to pH 5.8, and 0.8% plant agar. Homozygous lines were grown on plates supplemented with 50 mg kanamycin mL 21 . After 21 d, Arabidopsis plants were harvested, boiled in water for 15 min, and dabbed dry.
Sequence data from this article can be found in the GenBank/EMBL data libraries under the accession numbers provided in Table I.