Lipid biochemists salute the genome

Authors


For correspondence (fax +1 509 335 7643; e-mail jab@wsu.edu).

Summary

The biochemistry of plant metabolic pathways has been studied for many generations; nevertheless, numerous new enzymes and metabolic products have been discovered in the last 5–10 years. More importantly, many intriguing questions remain in all areas of metabolism. In this review, we consider these issues with respect to several pathways of lipid metabolism and the contributions made by the Arabidopsis genome sequence and the tools that it has spawned. These tools have allowed identification of enzymes and transporters required for the mobilization of seed storage lipids, as well as transporters that facilitate movement of lipids from the endoplasmic reticulum to the chloroplast in green leaf cells. Genomic tools were important in recognition of novel components of the cutin and suberin polymers that form water-impermeable barriers in plants. The waxes that also contribute to these barriers are exported from cells of the epidermis by transporters that are now being identified. Biochemical and genetic knowledge from yeast and animals has permitted successful homology-based searches of the Arabidopsis genome for genes encoding enzymes involved in the elongation of fatty acids and the synthesis of sphingolipids. Knowledge of the genome has identified novel enzymes for the biosynthesis of the seed storage lipid, triacylglycerol, and provided a refined understanding of how the pathways of fatty acid and triacylglycerol synthesis are integrated into overall carbon metabolism in developing seeds.

Introduction

Research in lipid metabolism provides the biochemical basis for understanding and appreciating the many roles of lipids in living organisms. For example, each membrane of a cell has a characteristic and distinct complement of glycerolipid molecules that vary in both headgroup structure and the length and degree of unsaturation of the fatty acid chains that constitute the hydrophobic portion of the membrane. What are the biochemical pathways that produce this diversity, and how are these pathways regulated at the organelle and whole-cell levels? Similar questions as to how biosynthesis relates to the resulting molecular structure and physiological function pertain in all areas of lipid biochemistry, and, of course, in other areas of metabolism as well.

In this review, we describe how completion of the Arabidopsis genome sequence (Arabidopsis Genome Initiative, 2000) has accelerated the discovery of important steps in the biosynthesis of acyl lipids – those lipids that are derived from fatty acids. The pathways include those leading to membrane glycerolipids and sphingolipids, to the barrier lipids, cutin, cuticular waxes and suberin, and the pathways for synthesis and utilization of triacylglycerols, the seed storage lipid. In each case, we give a necessarily brief synopsis of the metabolic pathway and then provide specific examples of biochemical discoveries that have been possible because of the availability of the genome sequence and tools derived from it. We do not discuss the many other classes of lipids found in plants, including sterol and isoprenyl lipids (Benveniste, 2004; DellaPenna and Pogson, 2006; Crowell and Huizinga, 2009), and, because of space constraints, we do not describe recent advances in the synthesis of the many lipid-derived signaling molecules that help control developmental and environmental responses of plants (Bargmann and Munnik, 2006; Boss et al., 2006; Browse, 2009; Li et al., 2009). However, the pathways that we do cover provide many wonderful examples of how the Arabidopsis genome sequence has changed the lives of lipid biochemists.

Global approaches to lipid metabolism

The availability of the Arabidopsis genome allowed an entirely new approach to analysis of lipid biochemistry and genetics. Where the sequence of a protein involved in lipid metabolism of any organism was known, the orthologous Arabidopsis gene could often be identified. A catalog of lipid metabolism genes was assembled based on a systematic search of the genome for genes involved in acyl lipid metabolism, based on examination of lipid biochemical literature; many genes previously annotated as of ‘unknown function’ were identified. In addition, by analysis of the genome as a whole, gene families could be detected. As many lipid biosynthetic genes are present as families of related genes in Arabidopsis, the ability to examine these families, rather than a single cDNA or genomic clone, is especially important. Expressed sequence tag (EST) data were analyzed to provide information on differential expression of the identified genes in Arabidopsis tissues (Mekhedov et al., 2000; Beisson et al., 2003). It became possible to compare the abundance of EST sequences present for genes active within single pathways, to compare the activities of individual genes among various tissues, and to detect tissue-specific regulation of family members. The information in the genome sequence became increasingly important in developing extended techniques for expression analysis, including microarray and co-expression analysis (described below). The accumulated information about lipid genes was collected in a curated database that is available to the public (http://www.plantbiology.msu.edu/lipids/genesurvey/index.htm), which, because of its focus on the genes of lipid metabolism, is more useful to lipid scientists than general sources such as the Arabidopsis Information Resource or GenBank. An updated version of this information will soon be available through the Arabidopsis Book (http://www.aspb.org/publications/arabidopsis/). A number of specific discoveries have been made as a result of access to this data, including analysis of plastid lysophosphatidate acyltransferase (Kim and Huang, 2004), examination of phosphatidylserine decarboxylase activity (Nerlich et al., 2007), and identification of the significance of some of the many annotated lipid transfer proteins (DeBono et al., 2009).

Knowledge of the genome spawned new systems approaches for biology. These include proteomics (Hajduch et al., 2006; Arai et al., 2008; Joyard et al., 2009), as well as the very through transcript profiling that is made possible by new high-throughput sequencing (Weber et al., 2007) and can be adapted to provide tissue- and cell-specific expression profiles. We confidently expect that these approaches will provide for many new discoveries in lipid metabolism and other areas of biology.

Seed oil mobilization

In Arabidopsis and other oilseed plants, the alpha and omega of life is oil. The very first emergence of plants from seed is powered by oil, and the purpose of the final metabolic rush of plants before they die is to store up enough oil to ensure survival and revival of their embryo. Coordinated activation of a number of metabolic systems is required to mobilize stored oil for use as energy in the developing embryo. Lipases hydrolyze stored triacylglycerol into its glycerol and fatty acid components. Fatty acids are transported to glyoxysomes, where they are activated to acyl CoA derivatives before conversion by the β-oxidation cycle into acetyl CoA, and thence via the glyoxylate cycle to four-carbon compounds. These are in turn transported to the mitochondria, and either used in respiration or returned to the cytosol for gluconeogenesis (Figure 1). A recent review provided an in-depth examination of the biochemistry of seed oil mobilization (Graham, 2008).

Figure 1.

 Outline of the pathway of oil utilization in Arabidopsis, indicating genes discussed in the text.
After fatty acids have been processed by the β-oxidation cycle, they are converted through the glyoxylate cycle to four-carbon compounds that may be respired or used for sucrose synthesis. TAG, triacylglycerol; TCA, tricarboxylic acid cycle; CoA, coenzyme A. See text for descriptions of SDP1, PXA1 and LACS6/7. LACS, Long-chain Acyl-CoA Synthetases; SDP, Sugar-Dependent; PXA, Peroxisomal ABC transporter.

Knowledge of the Arabidopsis genome sequence has magnified the power of traditional or forward genetics (Wallis and Browse, 2002), in which mutation effectively queries the organism about which genes are necessary for a particular process. The overpowering advantage of forward genetics is that it can reveal otherwise unpredictable requirements of the plant. For many years, map-based cloning was performed without full knowledge of the Arabidopsis genome sequence (Wallis and Browse, 2002), and the years of research effort expended to map even one mutant locus are testimony to the value obtained in understanding relationships between phenotypes and genotypes. Knowledge of the genome sequence, coupled with the availability of thousands of polymorphisms in the usual mapping partner of Columbia, Landsberg erecta (via the Cereon Arabidopsis polymorphism collection, http://www.arabidopsis.org/browse/Cereon/), reduced mapping times from years to months (Jander et al., 2002). A fine example of the use of this technique is discovery of a lipase important to the early development of Arabidopsis seedlings.

The cascade that provides the energy for plant germination begins when lipase enzymes hydrolyze triacylglycerol (TAG) molecules. Although the enzymology of the process is well known, it took a clever mutant screen to discover a lipase with physiological impact on Arabidopsis germination (Eastmond, 2006). A population of EMS-mutagenized Arabidopsis seeds was exposed briefly to light to induce germination, then incubated on plates with only basal salts in the dark. Those seeds that germinated but produced only short hypocotyls were rescued by transferring them to plates containing sucrose. These seedlings represented a population that required sucrose to support post-germinative growth. Biochemical analysis revealed that three mutants in one gene, SDP1, were deficient in lipase activity during seed germination. The mutant locus was identified using high-resolution mapping facilitated by the availability of the Arabidopsis genome sequence and Landsberg polymorphisms. While SDP1 is not a member of the recognized super-family of TAG lipases, it contains a conserved region with homology to recently characterized TAG lipases of the yeast Saccharomyces cerevisiae (TGL3, 4 and 5), Drosophila melanogaster (Bremmer lipase) and humans (adipose triglyceride lipase), but little homology outside that conserved region. Proteins with >60% amino acid identity to SDP1 are widely distributed in plants, including moss, legumes and rice (Eastmond, 2006), which is testimony to the significance of the gene for plant lipid biochemistry.

The acceleration of forward genetics methods and positional cloning of genes by use of the genome sequence is not simply due to the ready availability of position markers. In many cases, once a narrow interval is identified, the genomic sequence can be scanned for genes that are candidates for producing the mutant plant phenotype. This technique brings together the genome sequence itself, annotations of open reading frames accompanying the bare sequences, and information from many other organisms regarding the functions of genes sharing either general sequence similarity, or, as is often the case, conservation of a functional domain of the homologous protein sequences.

This candidate gene approach accelerated the identification and characterization of a locus responsible for transport of fatty acids to the glyoxysomes after they are released from triacylglycerols. In Arabidopsis, this process is mediated by PXA1 (also called CTS1 or PED3; Hayashi et al., 1998; Russell et al., 2000; Zolman et al., 2001). PXA1 mutants grow slowly and are reduced in size, indicating that the gene is important throughout the life of the plant. The locus was identified through forward genetics and positional cloning by three groups (Hayashi et al., 1998, 2002; Zolman et al., 2001; Footitt et al., 2002), each relying on a different aspect of the mutant phenotype, all of which pointed to a defect in transport of compounds to the peroxisomes or glyoxysomes. Plant lines homozygous for loss-of-function mutations require supplementation with either sucrose or short-chain fatty acids such as butyrate in order to germinate and establish seedlings. All three research groups identified the gene by mapping a narrow region of the chromosome responsible for the mutant phenotype, followed by examination of candidate genes within that region for the site of actual mutation. Candidate genes were identified by their sequence similarity to the yeast proteins Pxa1p and Pxa2p, which are known to participate in the import of long-chain fatty acids into peroxisomes. There was also evidence that the human homolog is responsible for the genetic disease X-linked adrenoleukodystrophy (X-ALD; Kemp and Wanders, 2007), showing the broad significance of the import mechanism in eukaryotes.

The availability of the entire genome sequence allows analysis of gene families that would otherwise not be possible. The number of members in a gene family, their relatedness to each other and to genes identified in other plants or more distantly related organisms, as well as information about which compartment of the cell they are probably targeted to, all contribute to understanding the specific functions of genes within families. For example, a crucial step in the utilization of fatty acids via β-oxidation is activation of the fatty acids freed by lipase to acyl CoA derivatives preparatory to entry into the β-oxidation cycle (Figure 1). In Arabidopsis, genomic analysis has played a very important role in the discovery of genes responsible for this process. A nine-member family of long-chain acyl CoA synthetases (LACS) had been identified in Arabidopsis; they are a sub-family of the large 64-gene AMP-binding protein family responsible for a variety of activities in Arabidopsis cells (Shockey et al., 2002). The LACS family was identified based on conserved sequence motifs, but the genes are not particularly similar, with only about 30% identity across the family. When the sequences of the nine LACS proteins were analyzed, it was clear that they could be apportioned into sub-groups according to sequence similarity, but it remains to be discovered whether that similarity corresponds to similarity of function. Sequence data indicated that two of the proteins (LACS6 and LACS7) were probably targeted to the peroxisomes; the sequences of these predicted proteins were also very closely related. Reverse genetics techniques were originally unrewarding: homozygous T-DNA insertional mutants in each of the corresponding genes had no detectable phenotype. Eventually this was shown to be an issue of functional redundancy, because a dramatic phenotype resulted when a line homozygous for insertions at both loci was isolated. The lacs6 lacs7 double mutant was defective in seed lipid mobilization and required exogenous sucrose in order to establish seedlings (Fulda et al., 2002), akin to sdp1 and pxa1 mutants.

Knowledge of the genome sequence and available polymorphisms accelerated the pace of forward genetic gene discovery in the β-oxidation pathways. By applying knowledge gained from other organisms, it was possible to avoid the last and most difficult mapping steps by analyzing candidate genes near where the mutation lay. Analysis of genes with overlapping functions within gene families was facilitated by knowledge not only of the number of family members but also their relatedness within the family.

Two pathways for biosynthesis of membrane lipids

Although fatty acids are core components of every membrane in the cell, and are also found outside cells in cuticular lipids, the fatty acid synthase responsible for de novo synthesis of 16- and 18-carbon fatty acids is localized in the plastid. Discovering the mechanisms of fatty acid and lipid transport between cell compartments is critical to developing a comprehensive understanding of lipid metabolism. In addition, we need to know how the requirements for biogenesis of various membranes of the cell are communicated to the plastid as signals that regulate the rate and products of the fatty acid synthase. As described below, knowledge of the Arabidopsis genome has allowed considerable progress with the first of these problems, while the second remains one of the major unanswered questions in lipid biochemistry.

Thirty years ago, results from radiotracer experiments on intact leaves and experiments using isolated spinach chloroplasts led to the recognition that plants use two distinct pathways for the synthesis of membrane glycerolipids (Roughan and Slack, 1982). The X:Y notation indicates a fatty acyl group containing X carbons and Y double bonds (cis unless specified), where 16:0 is palmitate, 18:0 is stearate, 18:1 is oleate, 18:2 is linoleate and 18:3 is α-linolenate. Both pathways are initiated by synthesis of 16:0-acyl carrier protein (16:0-ACP) from acetyl CoA by plastid acetyl CoA carboxylase and fatty acid synthase (Ohlrogge and Jaworski, 1997). This 16:0-ACP may be elongated to 18:0-ACP and then desaturated by a soluble desaturase (Lindqvist et al., 1996), such that 16:0-ACP and 18:1-ACP are the primary products of plastid fatty acid synthesis. The prokaryotic pathway (Roughan and Slack, 1982; Browse and Somerville, 1991) located in the plastid envelope uses 18:1-ACP and 16:0-ACP for the sequential acylation of glycerol-3-phosphate to PA (see Figure 2 for summary and definition of abbreviations). This PA is used for the synthesis of PG, and, via DAG, as a precursor for synthesis of the other major thylakoid lipids, MGD, DGD and SQD (Browse and Somerville, 1991; Ohlrogge and Browse, 1995). The eukaryotic pathway begins with the hydrolysis of 16:0- and 18:1-ACPs and export of these fatty acids to the cytoplasm as CoA thioesters. Recent evidence indicated that the majority of the acyl groups exported from the plastids are rapidly incorporated into PC of the endoplasmic reticulum (ER) by the action of acyl CoA:lysoPC acyltransferase (LPCAT) and possibly other enzymes (Bates et al., 2007). This makes 18:1 available for the synthesis of 18:2 and 18:3 by the action of the FAD2 and FAD3 fatty acid desaturases (Arondel et al., 1992; Okuley et al., 1994). The LPCAT reaction is reversible, allowing transfer of 18:2 and 18:3 to the acyl CoA pool. CoA esters from this diversified acyl CoA pool are used for the synthesis of PA in the ER. This PA gives rise to the phospholipids PC, PE, PI and PS that are characteristic of the various extrachloroplast membranes. In addition, however, the diacylglycerol moiety of PC is returned to the chloroplast envelope (Figure 2), where it enters the DAG pool and contributes to the synthesis of thylakoid lipids (Browse et al., 1986). Evidence from several Arabidopsis mutants indicates that this transfer of lipid from the endoplasmic reticulum to the chloroplast is to some extent reversible (Miquel and Browse, 1992; Browse et al., 1993).

Figure 2.

 An outline of the two pathways of glycerolipid synthesis in plant cells.
The prokaryotic pathway synthesizes thylakoid glycerolipids within the chloroplast (plastid). The eukaryotic pathway in the ER produces glycerolipids for both the thylakoids and other membranes of the cell. The diagram reflects the understanding before the availability of the Arabidopsis genome, and therefore does not include the new discoveries described in this review (see Figure 3).

Because of the substrate specificities of the plastid acyltransferases (Browse and Somerville, 1991; Ohlrogge and Browse, 1995), the PA made by the prokaryotic pathway has 16:0 at the sn-2 position and 18:1 at the sn-1 position of the glycerol backbone. In contrast, the acyltransferases of the ER produce PA that is highly enriched with 18:1 at the sn-2 position; 16:0, when present, is confined to the sn-1 position. This biochemical feature allowed quantification of the net fluxes through the two pathways by positional analysis of leaf lipids (Browse et al., 1986). In all plant tissues, the major glycerolipids when first synthesized contain high proportions of 16:0 and 18:1 acyl groups. Subsequent desaturation of the lipids to the highly unsaturated forms typical of the membranes of plant cells is carried out by membrane-bound desaturases of the chloroplast and ER (Shanklin and Cahoon, 1998). Cloning and characterization of the genes encoding these desaturases preceded availability of the genome sequence, with the exception of FAD4 and FAD5, whose identities were published more recently (Mekhedov et al., 2000; Heilmann et al., 2004; Gao et al., 2009).

In many species of higher plants, PG is the only product of the prokaryotic pathway, and the remaining chloroplast lipids are synthesized entirely by the eukaryotic pathway. In other species, including Arabidopsis and spinach, in which both pathways contribute approximately equally to the synthesis of MGD, DGD and SQD, the leaf lipids characteristically contain substantial amounts of 16:3 which is synthesized by desaturation of 16:0 at the sn-2 position of prokaryotic MGD (Ohlrogge and Browse, 1995). This synthesis of thylakoid lipids by the prokaryotic pathway in Arabidopsis (Figure 2) allowed isolation of mutants that are compromised in the synthesis of chloroplast lipids by the eukaryotic pathway. As described below, analysis of some of these mutants is now revealing the mechanism of lipid transfer between the ER and chloroplast.

The enzymes of galactolipid synthesis

The mechanism for transfer of lipids from the ER to the chloroplast and its relationship to the biochemistry of galactoglycerolipid synthesis are two areas in which information from the genome has been key to a recent series of important discoveries. MGD synthase, which transfers galactose from UDP-gal to DAG, was recognized as an enzyme activity of the chloroplast envelope in 1983 (Block et al., 1983), but it was not until 1997 that the enzyme was purified from cucumber and the corresponding cDNA cloned (Shimojima et al., 1997). The Arabidopsis genome contains three homologs of the cucumber gene, MGD1, MGD2 and MGD3. A forward-genetic screen for mutants deficient in chloroplast biogenesis identified a leaky allele of mgd1 (Jarvis et al., 2000), and a null mutant at this locus was embryo- or seedling-lethal (Kobayashi et al., 2007). The MGD1 protein is associated with the inner chloroplast envelope (Awai et al., 2001), and MGD2 and MGD3 have been localized to the outer envelope. Isolation and cloning of a dgd1 mutant and characterization of the DGD1 protein (Dörmann et al., 1995, 1999) established that the main reaction for DGD synthesis involves transfer of galactose from UDP-gal to MGD. A close homolog of DGD1, DGD2, encodes a second isozyme. Both the DGD1 and DGD2 proteins have been localized to the outer chloroplast envelope (Froehlich et al., 2001). A third enzyme activity of chloroplast envelope preparations leads to formation of oligogalactolipids (a DAG moiety esterified with three or more galactose groups) through repeated transfers of galactose from MGD (Benning, 2009). Although this galactolipid:galactolipid galactosyltransferase (GGGT) was once suggested to be the main source of thylakoid DGD, analysis of a dgd1 dgd2 double mutant, obtained through reverse genetics, has shown that it is a minor activity under most conditions (Kelly et al., 2003). However, this enzyme is responsible for the synthesis of trigalyctosyldiacylglycerol, which is characteristic of mutants in which transfer of lipid from the ER to the chloroplast is disrupted (see below).

The multiple galactosyl transferases and their locations in either the inner or outer chloroplast envelope could potentially provide many, possibly redundant, routes for the synthesis of thylakoid MGD and DGD. However, genetic and biochemical analyses have established that one series of reactions predominates for both the prokaryotic and eukaryotic pathways (Dörmann et al., 1995; Jarvis et al., 2000; Benning, 2008, 2009). DAG is first acted on by the MGD1 enzyme located in the inner envelope. Most of the MGD formed is transferred to the thylakoids, but approximately 30% is transferred to the outer envelope, where it is converted to DGD by the DGD1 enzyme. The DGD produced must then be transferred from the outer envelope, via the inner envelope, to the thylakoids (Figure 3). The MGD2, MGD3 and DGD2 isozymes act in a distinct pathway that provides DGD to extra-chloroplast membranes, particularly to provide the lipid molecules that replace phospholipids when plants are grown on low phosphate (Härtel et al., 2000). The role and identity of the GGGT enzyme is unknown at this time.

Figure 3.

 Schematic representation of newly discovered and proposed steps in the synthesis of chloroplast MGD and DGD by the prokaryotic and eukaryotic pathways.
ER, endoplasmic reticulum; iE, inner envelope; oE, outer envelope. For definition of other abbreviations, see Figure 2.

Transport of lipids from the endoplasmic reticulum to the chloroplast

The dgd1 mutant contains very low levels of DGD (approximately 15% of wild-type). A screen to identify suppressors with increased accumulation identified more than 20 mutants that had increased DGD but also significant levels of trigalactosyldiacylglycerol (Xu et al., 2003). These tgd mutants represent at least four distinct loci, and are compromised in the transfer of lipids from the ER to chloroplasts in the eukaryotic pathway. Cloning and characterization of the TGD1, TGD2 and TGD3 genes provided evidence that they encode the permease, substrate-binding and ATPase components, respectively, of an ATP binding cassette (ABC) transporter of the inner chloroplast envelope that binds and transports PA (Figure 3) (Xu et al., 2003, 2005; Awai et al., 2006; Lu et al., 2007). The proposed function of the TGD1–3 complex is analogous in some respects to the microbial transporters involved in export of lipopolysaccharide through the bacterial inner membrane, suggesting the possibility that the lipid-transfer mechanism in plants evolved from components made available by the ancestral endosymbiosis of photosynthetic bacteria that is believed to be the origin of plant chloroplasts (McFadden, 1999). GFP fusions with TGD4 appeared to be localized to the ER, and, on this basis, TGD4 has been suggested to be involved in establishment of contact sites between the ER and chloroplast and/or the transfer of PC, PA or other lipid molecules to the chloroplast envelope (Xu et al., 2008). To date, however, no components of the TGD pathway have been identified for the outer chloroplast envelope, or for the cytoplasm, which also contributes factors that are important for the transfer process (Benning, 2008, 2009).

We are still a long way from a complete understanding of the components and processes contributing to the transfer of glycerolipids from the ER to the chloroplast. However, the examples surveyed here show how genome tools (for mapping, gene identification and reverse genetics) and genomic tools (providing homology-based clues to protein function) are proving useful in solving long-standing questions in cellular metabolism. Other lipid-trafficking pathways, including those for vesicle transport between the chloroplast inner envelope and thylakoids, and for export of digalactosyldiacylglycerol (DGD) from chloroplasts/plastids to other membranes during phosphate stress, have been discussed in recent reviews (Benning, 2008, 2009).

The barrier lipids: suberin, cutin and wax

Knowledge of the Arabidopsis genome has opened doors to exploring the genetic basis of a myriad of biochemical pathways where progress had previously been slow. Many of these discoveries have been in understanding the synthesis of the complex lipids of suberin, cutin and wax, whose role in controlling water and ion transport and in protection of the plant from hazards is crucial for plant life.

Suberin

Suberin is important in the roots and other tissues of Arabidopsis, where it plays a critical role in controlling water and ion transport (Baxter et al., 2009). Research into suberin biosynthesis has advanced by combining biochemical analysis with genomic information. Tissue-specific expression analysis of large gene families has assisted in the choice of candidate genes, and promoter cloning based on the genomic sequence has been used to confirm both gene expression patterns as well as gene identity via complementation of mutant phenotypes. Suberin is a complex composite of aliphatic phenylpropanoid-derived aromatic molecules together with glycerol. Biosynthesis of its aliphatic component has recently been reviewed (Franke and Schreiber, 2007). Interest in the characterization of suberin synthesis has been revitalized by a fresh biochemical characterization of the suberin found in Arabidopsis root tissue (Franke et al., 2005). The aliphatic portion of suberin in this tissue contains a wealth of ω-hydroxy acids (43%) and α,ω-diacids (24%). Genomic tools were applied by candidate gene approaches, first to clone and characterize a cytochrome P450 ω-hydroxylase (Li et al., 2007; Hofer et al., 2008), and then to identify a fatty acid synthase/3-ketoacyl CoA synthase (FAE-KCS) enzyme responsible for elongation of fatty acids beyond a chain length of 22 carbons (Franke et al., 2009). Of the known ω-hydroxylases, CYP86A1 is strongly expressed in roots. When the suberin monomer composition of plants with T-DNA insertions in the gene was analyzed, 16- and 18-carbon ω-hydroxy acids were reduced to less than half of wild-type levels, and levels of α,ω-diacids were likewise reduced (Li et al., 2007; Hofer et al., 2008), indicating that CYP86A1 is responsible for a large proportion of the hydroxylation of suberin monomers.

The FAE-KCS mutant DAISY was isolated after examination of expression data to determine which KCS genes were strongly expressed in root tissue but were not known to participate in wax synthesis. A homozygous T-DNA insertion mutant in a gene chosen based on these criteria had roots that were 24% shorter when grown on agar plates (Franke et al., 2009). Characterization of suberin monomers in the root tissue of these plants demonstrated a reduction in the levels of fatty acids longer than 20 carbons; 22-carbon docosanoic acid was reduced to less than half the wild-type level, and monomers shorter than 22 carbons accumulated in the tissue. Detailed promoter–GUS fusion analysis of this DAISY gene indicates that it may also play a role in flower sepals and silique abscission zones. When Arabidopsis plants were treated with high levels of NaCl, DAISY expression specifically increased in those plant tissues where levels of suberin increase when plants are placed under osmotic stress, completing the link between suberin accumulation and abiotic environmental stress (Franke et al., 2009).

Although models have long proposed that the aliphatic and aromatic monomers of suberin are bound together by ester bonds, almost no direct evidence on the three-dimensional structure of the suberin polymer is available (Pollard et al., 2008). Co-expression analysis led to the recent identification of a mutant in a feruoyl CoA transferase that lacks almost all suberin aromatic esters, challenging the long-held assumption that the lamellar structure of suberin observed by transmission electron microscopy is due to alternating bands of aromatic and aliphatic domains (Molina et al., 2009). In addition, mutants in CYP86B1 are have greatly reduced levels of C22 and C24 ω-hydroxy and dicarboxylic acids (Compagnon et al., 2009; Molina et al., 2009). The seed suberin composition of cyp86b1 knockouts was surprisingly dominated by unsubstituted fatty acids that are incapable of forming polymeric linkages, and yet the total suberin aliphatic content remained similar, suggesting that suberin insolubility may not depend on an extended polymer structure.

Although the aliphatic monomers of suberin are bound together into polyesters, the presence of glycerol allows the formation of cross-linked, three-dimensional networks of suberin polymers. Biochemical evidence suggested that acyl CoA-dependent glycerol-3-phospate acyltransferases (GPATs) might be important in suberin synthesis. When probable GPAT genes (Beisson et al., 2003) were examined, an insertional mutation of one gene, GPAT5, was found to reduce the levels of suberin monomers without affecting other lipids. These mutants also showed decreased seed germination and increased seed coat permeability, revealing that suberin has a more significant role in the seed coat than had previously been established (Beisson et al., 2007).

Cuticle

Arabidopsis plants depend on their cuticle as a barrier to control the flux of water and to defend against pathogen attack and other hazards (Nawrath, 2006). The cuticle has an additional role in development, acting to prevent organ fusion. The composition of this cuticle layer is an elaboration of the cutin found in plant cell walls, supplemented with specialized waxes. Although the biochemistry of the composition, structure and biosynthesis of the cuticle has been studied for many years, wide gaps remain in our knowledge of this biochemistry. However, recent advances based on genomic analysis have added much to our knowledge, and suggest that much more will be forthcoming.

Synthesis of the Arabidopsis cuticle is a coordinated process, as shown through discovery of a transcription factor that affects both cutin and wax synthesis. A genome-wide approach was used to identify transcription factors based on homology with other characterized eukaryotic transcription factors (Riechmann et al., 2000). To functionally characterize individual transcription factors, plants that either over-expressed the identified genes or plants with insertion knockout mutations in these genes were analyzed (Broun et al., 2004). The approach was successful in identifying WIN1, an ethylene response transcription factor (ERF) whose over-expression produced a substantial increase in epicuticular wax accumulation (Broun et al., 2004). Genomic information was also important in analyzing plant lines that over-expressed the transcription factor. Microarray analysis of leaf samples from plants over-expressing WIN1 revealed a group of 12 genes that are strongly affected by its over-expression, including cytochrome P450 enzymes and the acyl CoA synthetase LACS2, as well as a lipase-like gene and genes involved in glycerol metabolism. Microarray data show that the genes involved in cutin synthesis are rapidly activated by the transcription factor WIN1, and that genes of wax synthesis are activated later, either through metabolic responses or through possible secondary transcription factors (Kannangara et al., 2007).

Cutin

Cutin is an ester-linked polymer whose principle components in many plants are ω-hydroxy fatty acids with chain lengths of 16 or 18 carbon atoms, often characterized by additional hydroxy or epoxy modifications within the hydrocarbon chains. Recently it has been shown that the Arabidopsis cuticle contains surprisingly high amounts of dicarboxylic acids, which may constitute 50% of cutin monomers in Arabidopsis (Bonaventure et al., 2004). This 18:2 α,ω-dicarboxylic acid monomer was once thought to be associated primarily with suberin.

The outline of cutin synthesis in Arabidopsis has been revealed by the success of forward genetic analysis, which accelerated as genome information became available. Genes identified by forward genetics include LACERATA, which encodes a P450 enzyme producing α-hydroxy fatty acid components of cutin (Wellesen et al., 2001). Similarly, the cytochrome P450 monooxygenase CYP86A2 gene is critical for cutin production (Xiao et al., 2004). The HOTHEAD oxidase has been proposed to synthesize dioic acids from α-hydroxy fatty acid precursors (Krolikowski et al., 2003; Kurdyukov et al., 2006b). The BODYGUARD protein is an epidermis-specific extracellular α/β-hydrolase-fold-containing protein that is required for proper cuticle formation and normal leaf morphogenesis (Kurdyukov et al., 2006a), but whose exact biochemical function is unclear. Identification of these and other genes was assisted by the availability of genome information.

While forward genetic analysis revealed much about the synthesis of the Arabidopsis cuticle, the genetic basis of many other biochemical reactions remained unknown. A tissue-specific microarray covering the whole Arabidopsis genome, extending the general approach of expression analysis for all lipid related genes described previously (Mekhedov et al., 2000; Beisson et al., 2003), has revealed much about the genes involved in production of the cuticle. The transcriptome of the stem epidermis was examined after careful removal of this cell layer (Figure 4) followed by RNA isolation and analysis using the Arabidopsis ATH1 genome array (Affymetrix, http://www.affymetrix.com) representing more than 23 000 Arabidopsis genes (Redman et al., 2004). Overall, approximately 1900 genes were up-regulated in the stem epidermis relative to the total stem (Suh et al., 2005). The experiment revealed that there were significant differences between various regions of the stem epidermis, and provided a substantial list of genes that are highly expressed in this tissue, about 40% of which are completely uncharacterized. In addition, 85 genes thought to be involved in lipid metabolism (Beisson et al., 2003) were identified as strongly expressed; this subset contained genes known to be involved in wax and cutin synthesis, as well as members of gene families whose expression suggested that they might also be important in synthesizing the stem cuticle.

Figure 4.

 Epidermis of Arabidopsis stems.
(a) Transverse section of stems viewed under a light microscope after staining with phloroglucinol. The epidermis is the transparent outermost single cell layer.
(b) Using manual dissection, epidermis can be isolated from the stem as a transparent film. This figure was originally published in Suh et al. (2005), and is reproduced with permission from the American Society of Plant Biologists.

One example of the discoveries resulting from these experiments is recognition that isozymes of glycerol-3-phosphate acyl transferase (GPAT) are required for normal cutin production. Two GPAT genes (GPAT4 and GPAT8) were found to be highly expressed in the epidermis. Insertional mutations in each of the genes separately produced no phenotype, but the double knockout gpat4 gpat8 had a large increase in cuticle permeability to dyes and water loss, as well as loss of cuticular ledges surrounding the stomatal pore, traits that are characteristic of defective cuticle formation (Li et al., 2007). The plants were also altered in their susceptibility to pathogen attack, another characteristic of defective cutin structure. Similar reverse-genetic approaches based on epidermis transcriptome analysis have been successful in analysis of suberin and wax synthetic pathways (Beisson et al., 2007; Bird et al., 2007).

Cuticular wax

Wax is another acyl component of the cuticle matrix in plants, a complex mixture of very long chain alkanes, fatty alcohols, aldehydes and ketones derived from very long chain fatty acids (VLCFAs) with lengths of 20–32 carbon atoms. VLCFAs are synthesized by an elongase complex in the ER acting on 16- and 18-carbon substrates. The complex performs four successive reactions, comprising 3-ketoacyl CoA synthase (KCS), 3-ketoacyl CoA reductase (KCR), 3-hydroxyacyl CoA dehydratase (HCD) and enoyl reductase (ECR). The substrate specificity of the elongation reactions resides in the KCS enzymes, and the other elongase reactions are non-specific (see below). In wax synthesis, the VLCFA products of elongation pass primarily through a decarboxylation pathway (about 80% of the VLCFAs), although a smaller flux through an acyl reduction pathway also occurs (reviewed by Samuels et al., 2008). The products of both pathways are transported through the plasma membrane by ABC transporters to the plant surface. In addition, lipid transport proteins are involved in extracellular transport of the waxes to the outside of the cell wall (DeBono et al., 2009; Lee et al., 2009).

In the two decades preceding completion of the Arabidopsis genome sequence, the biochemistry and genetics of fatty acid elongation had been characterized in yeast (Schneiter, 1999). With the availability of the Arabidopsis genome, it was possible to identify cognate genes, even whole families of genes in Arabidopsis, for which little or no previous genetic information was available (Sperling and Heinz, 2003; Dunn et al., 2004). Twenty-one KCS sequences are annotated in the Arabidopsis genome. The FAE type was first identified by a mutational exploration of VLCFA seed triacylglycerols (James et al., 1995), and it has been known for some time that the substrate specificity of the KCS enzyme determines the specificity of the elongase reaction (Millar and Kunst, 1997). Many of the KCS enzymes have been characterized by heterologous expression in yeast (Trenkamp et al., 2004; Blacklock and Jaworski, 2006; Paul et al., 2006). Recent analysis of the phylogeny, genomic organization and protein topology has divided the 21 genes into eight distinct sub-classes. Tissue-specific expression was measured by using the available genomic sequence to clone the promoters for each gene, followed by fusion to a reporter gene. Expression of some sub-families was limited to seeds or flowers, while complex expression patterns were observed for many of the genes. Twelve of the 21 genes were significantly expressed in the epidermis (Joubes et al., 2008). This complexity of expression suggests that considerable work remains in determining the functional roles of these genes.

Other components of the VLCFA elongase complex were readily identified in Arabidopsis by using information obtained from yeast and the Arabidopsis genome sequence (Table 1). These elongase enzymes are not specific to production of any particular product, but are engaged in elongation of all VLCFAs. The sequences of identified genes in yeast were used to detect homologous genes of Arabidopsis. The Arabidopsis genes were amplified and expressed in yeast, and their activity was confirmed through complementation of yeast mutants. Using a reverse-genetics approach, production of T-DNA insertions inactivating the genes (or multiple members of gene families) in Arabidopsis enabled investigators to examine the metabolic effects of interrupting the production of VLCFAs. The pleiotropic phenotypes of these mutants were similar to each other, with reduced cuticular wax, altered VLCFA composition in seed triacylglycerols, and altered sphingolipid composition; these changes are accompanied by abnormal morphology (ECR) or embryo lethality (HCD and KCR).

Table 1.   Yeast and Arabidopsis elongase genes
Yeast geneArabidopsis geneFunctionReferences
TSC13CER10Enoyl reductaseGable et al. (2004)
Zheng et al. (2005)
PHS1PAS23-hydroxyacyl CoA dehydrataseBach et al. (2008)
YBR159wAtKCR13-ketoacyl CoA reductaseBeaudoin et al. (2009)

Information from other plant species has been applied to gene discovery and analysis in Arabidopsis wax synthesis. Identification of an alcohol-forming fatty acyl CoA reductase (FAR) in jojoba (Simmondsia chinensis) (Metz et al., 2000) was critical in cloning the first wax synthesis gene for the steps beyond VLCFA elongation in Arabidopsis. When the deduced amino acid sequence of the jojoba FAR gene was used to search the Arabidopsis genome, eight genes with as much as 54% amino acid sequence identity were detected. The chromosomal location for one of these putative FAR genes matched the genetic map position of cer4, a locus that had been previously identified (McNevin et al., 1993). cer4 mutants are characterized by a loss of primary alcohols and wax esters, suggesting that the gene is important in the acyl reduction pathway. When the promoter sequence of the gene was cloned based on the Arabidopsis genome sequence and fused to a reporter gene, expression was detected in the epidermal tissues of leaves and stems, as well as in all cells of the elongation zone of root tips. When the cDNA was expressed in yeast, 24- and 26-carbon primary alcohols were produced, confirming the function of CER4 (Rowland et al., 2006).

A second gene of wax biosynthesis was cloned on the basis of its sequence similarity to a bacterial wax ester synthase/acyl CoA:diacylglycerol acyltransferase (WS/DGAT). Although there are more than ten Arabidopsis genes annotated as WS/DGAT candidates, only one gene was both akin to the bacterial wax synthase and highly expressed by epidermal tissues that were growing rapidly, according to a previous analysis (Suh et al., 2005). An Arabidopsis line containing a T-DNA insertion in this gene, WSD1 (At5g37300), lacked the glossy stem phenotype characteristic of many wax synthesis mutants, but biochemical analysis revealed that wax esters had been reduced to undetectable levels, and heterologous expression in yeast led to formation of wax esters (Li et al., 2008), confirming the function of the gene.

Although the biochemistry of wax synthesis pathways has been analyzed and it is clear that the products of synthesis must be exported from cells across the plasma membrane to the cuticle, the mechanism of that export was unknown until the first wax transporter gene, CER5, was cloned by a combination of positional cloning and examination of candidate genes. The candidates were chosen by homology to the human X-linked ALD locus, an ABC transporter that, if defective, causes lipid accumulation in the cytoplasm of cells; cer5 cells accumulate sheets of waxy inclusions in tissues that normally secrete wax. The CER5 gene is an important ABC transporter localized to epidermal cells (Pighin et al., 2004). Examination of expression data in stem epidermis (Suh et al., 2005) for other potential wax transport genes, including analysis of co-expression with CER5 (Pighin et al., 2004), led to the discovery of a second wax transporter gene, WBC11 (Bird et al., 2007). wbc11 mutants not only develop lipid inclusions within their cells, but a number of morphological defects are also evident, including stunted growth and mis-shapen leaves. The significance of the gene to developmental and biochemical processes in plants was revealed when mutants with reduced expression exhibited stem–stem and stem–leaf fusions, while the cuticular wax content of their stems was reduced to about half the wild-type levels.

Expression analysis led to the identification of yet another transport protein important to wax synthesis. The genomic sequence of Arabidopsis contains more than 70 genes annotated as lipid transfer proteins. These proteins have the in vitro property of transferring lipids between liposomes, and have long been considered candidates for important lipid transfer activities in plants. Five candidate genes were examined, each of which was highly expressed in the epidermis of rapidly growing stems (Suh et al., 2005). Subsequent examination of T-DNA insertions in the genes confirmed that one of them had only 50% of the wild-type level of alkanes and a 25% reduction in levels of total cuticular wax (DeBono et al., 2009). The protein had been previously characterized as a glycosylphosphatidylinositol-anchored protein in proteomics studies. Localization studies show that the protein, LTPG, is present in epidermal cell walls, not only at the plasma membrane surface but at punctuate inclusions within the cells, and is active in cuticular lipid transport.

This catalog of lipid genes guided many information-based reverse-genetic approaches, and additional information from expression analysis of specific tissues narrowed the search for the genes responsible for cuticle synthesis. Genes whose products are involved in lipid metabolism were discovered by identifying their co-expression with other genes. Discovery of the genetic basis of the biochemistry of VLCFA elongase reactions benefitted greatly from discoveries in the Saccharomyces cerevisiae genome; information from this organism and others continues to inform lipid genetics in Arabidopsis. Promoter analysis is a staple use of genome information for analyzing the genetics of cuticle formation. Fusion of the promoter derived from a gene of interest to a reporter gene sometimes confirms expectations but also can reveal unanticipated patterns of expression. Cloned promoters are also used to confirm mutant identity by expressing the open reading frame of candidate genes in complementation experiments.

Sphingolipids

Sphingolipids are by far the most structurally diverse class of lipid molecules (Pruett et al., 2008). The structural complexity of sphingolipids stems from variation of each of three components of these lipids. A long chain base (LCB) is amide-linked to a fatty acid, creating a ceramide; in addition, head groups are often substituted at the LCB terminal hydroxyl group. Serine palmitoyltransferase is the first enzyme in the pathway that synthesizes sphingolipid LCB, condensing palmitoyl CoA and serine to form 3-ketosphinganine; this product is reduced to form dihydrosphingosine, the simplest LCB. In plants, dihydrosphingosine is subject to hydroxylation at C4, and desaturation at C8 (commonly) or desaturation at C4 (uncommonly), producing additional forms of LCB. The acyl moiety bound to the LCB by an amide linkage is usually a VLCFA varying from 22 to 36 carbon atoms, although 16 carbon fatty acids are also present (Markham et al., 2006). Headgroup residues include phosphate (creating phytosphingosine-1-phosphate), phosphoinositol (creating inositolphosphorylceramide), glucose (creating glucosylceramide) or more complex molecules to create phytoglycolipids. Sphingolipids account for perhaps 40% of plant plasma membrane lipids (Sperling et al., 2005; Markham et al., 2006), and are therefore of considerable structural significance. In addition to their structural role, they are important molecules in plant reactions to drought (Coursol et al., 2005) and in control of programmed cell death (Petersen et al., 2008; Raffaele et al., 2008; Wang et al., 2008). While many distinct species of sphingolipid have been detected in Arabidopsis tissue (Markham et al., 2006), and progress in understanding the genetic basis of their synthesis is progressing rapidly, there are still considerable gaps in our knowledge.

Important findings regarding synthesis of LCBs have been based on genomic analysis of Arabidopsis using information first garnered from yeast and mammals. The serine palmitoyltransferase enzyme has two subunits. The LCB2 subunit was identified by its homology to a known mouse gene (Nishiura et al., 2000); LCB1 was identified from the genome sequence by homology to the yeast gene. A T-DNA disruption of the lcb1 locus was embryo-lethal (Chen et al., 2006). To analyze the biochemical effects of reduced LCB1, RNAi suppression lines were created. Suppression of lcb1 produced smaller plants with no reduction in the total LCB content of leaves, but there were alterations in the proportions of LCBs. The content of saturated LCB increased from about 7% of the total LCB in wild-type leaves to more than 30% in some RNAi lines, indicating that lowering the total level of LCBs available had the effect of reducing the flux through the pathways where desaturation of the LCB normally occurs (Chen et al., 2006). Initial examination of an insertion in the gene encoding LCB2 detected no phenotypic change, but examination of the Arabidopsis genome sequence indicated that there were in fact three copies of LCB2-like genes. One of the copies is incomplete, apparently the result of a defective gene duplication. T-DNA insertions in each of the remaining LCB2 genes showed no developmental phenotypes, but double homozygous mutant lines could not be obtained. Collectively, these experiments demonstrate that both subunits of serine palmitoyltransferase are essential for male gametogenesis (Teng et al., 2008).

Two homologs of the yeast LCB C4 hydroxylase were detected in Arabidopsis and named sbh1 and sbh2 (Chen et al., 2008). Mutations in each gene separately resulted in reduced 4-hydroxylation of LCBs, but only in the case of the double mutant were 4-hydroxy LCBs absent. Plants that were mutant in both genes were very small and failed to reach reproductive maturity. In plants without 4-hydroxy LCBs, all tissues were enriched in Δ-8 LCB, and the leaves contained 2–3-fold increases in total LCB content. Free LCB accounted for much of the increase in total LCB content, and the ceramide content of leaf tissue also increased. In wild-type plants, 13% of the total ceramides have 16:0 fatty acids; in the sbh1 sbh2 double mutant, 85% of the ceramides had 16:0 fatty acids. Most interestingly, examination of glycerolipids from double mutant lines showed that, while the levels of extra-plastidic glycerolipids were approximately the same as in the wild-type, chloroplast lipids were affected. The level of the major leaf lipid MGD was 33% lower than the wild-type, and the levels of both DGD and PG were reduced as well; the reduction came mostly at the expense of 16:0-containing species of chloroplast lipids (Chen et al., 2008), suggesting possible metabolic coordination between the ER and chloroplasts regarding sphingolipid and chloroplast lipid metabolism. These and other examples reveal how important the comparison of Arabidopsis and yeast sequences has been for elucidating the genetic basis of sphingolipid biochemistry.

Synthesis of seed oils

Plants store oil, starch and/or protein in the seed to provide carbon and energy for initial growth of the next generation. The amount of oil in seeds of different species varies from 1% to as much as 60% of seed dry weight. In almost all species, triacylglycerols (TAGs) are the major oil component stored in densely packed oil bodies of approximately 1 μm diameter (Kim et al., 2002). The oil bodies arise from the ER, the site of TAG synthesis, and are surrounded by a monolayer phospholipid membrane, containing specific structural proteins – the oleosins (Kim et al., 2002). The Arabidopsis genome contains a family of 16 oleosin genes, and a comprehensive analysis showed that the encoded oleosins are expressed in seeds alone (five genes), both seeds and pollen (three genes), or in the anther tapetum (eight genes), reflecting the importance of TAG reserves in all three of these tissues (Kim et al., 2002).

Vegetable oils constitute one of the world’s most important plant commodities, and are used in a wide range of foodstuffs. In addition, some plants produce oils containing derivatized fatty acids that are attractive feedstocks for the chemicals industry. The oils of major oilseed crops contain levels of 18:2 and 18:3 that affect the shelf life of food products made from them. These oils are typically stabilized by partial hydrogenation, but this results in the production of unhealthy trans fats (Steinhart et al., 2003). For these and other reasons (Singh et al., 2005; Napier, 2007), using our understanding of TAG synthesis pathways to engineer changes in the fatty acid composition of plant oils is an attractive goal. Much of our understanding of TAG synthesis comes from genetic and molecular investigations in Arabidopsis, so the availability of the genome, and genomics tools, have been just as valuable here as in other areas of lipid metabolism.

Many of the reactions of TAG synthesis are the same as those described for the eukaryotic pathway in leaf cells (Figure 2) (Browse and Somerville, 1991; Ohlrogge and Browse, 1995). As in the eukaryotic pathway, 18:1 produced in the plastids must be incorporated into PC of the ER for desaturation to 18:2 and 18:3 by action of the FAD2 and FAD3 enzymes. The 18:1 may be transferred directly into PC by the action of LPCAT and other acyl transferases (Kazachkov et al., 2008; Bates et al., 2009). Alternatively, 18:1 is used for DAG synthesis by reactions of the Kennedy pathway, and the 18:1-DAG is converted to PC by CDP-choline:diacylglycerol cholinephosphotransferase (CPT) (Browse and Somerville, 1991; Ohlrogge and Browse, 1995). It has been proposed that the CPT reaction is reversible and provides one mechanism for the production of polyunsaturated DAG for the synthesis of TAGs containing 18:2 and 18:3 (Browse and Somerville, 1991). Two enzymes specific to TAG synthesis are acyl CoA:diacylglycerol acyltransferase (DGAT) and phospholipid:diacylglycerol acyltransferase (PDAT) (Zou et al., 1999; Dahlqvist et al., 2000; Ståhl et al., 2004). The second of these transfers a fatty acid from the sn-2 position of PC, thus providing a source of lyso-PC for LPCAT (Kazachkov et al., 2008). In Arabidopsis, 20:1 is also a component of the seed oil, and this is synthesized by elongation of 18:1 by the enzymes of the ER described above.

The pathways of fatty acid and TAG synthesis must, of course, be integrated during seed development with sucrose import from the maternal plant, as well as the reactions of acetyl CoA synthesis by glycolysis and pyruvate dehydrogenase (Ruuska et al., 2002). These broader relationships have been difficult to explore in Arabidopsis by biochemical and genetic techniques. However, genomics methods, including microarrays and, more recently, high-throughput sequencing technologies and proteomics, provide indirect but workable approaches. The earliest application in oilseeds used cDNA microarrays to map changes in gene expression during seed filling in Arabidopsis (Ruuska et al., 2002). In this study, expression of approximately 35% of the genes changed more than twofold during seed development. Over the time course of the experiment, the expression patterns of plastid transporters and glycolytic enzymes demonstrated a shift from cytoplasmic to plastid pathways of carbon metabolism, indicating the importance of carbon flux into the plastid and of the plastid pyruvate kinase in the supply of acetyl CoA for fatty acid and TAG synthesis (Ruuska et al., 2002; Andre et al., 2007).

Many genes encoding enzymes of fatty acid synthesis exhibited a bell-shaped pattern of expression through seed development. A similar pattern was observed for genes encoding photosynthetic enzymes, providing a clue that CO2 fixation is important during storage TAG synthesis (Ruuska et al., 2002). This last result presaged the discovery of a biochemical pathway that allows green oilseeds to use ribulose bisphosphate carboxylase to re-fix a substantial proportion of the CO2 released during conversion of sucrose to fatty acids (Schwender et al., 2004).

The recent identification and characterization of the Arabidopsis ROD1 gene provides an example where direct use of the genome (for map-based cloning and candidate gene identification) and the extended use of genomics database tools (for clues to gene function) have contributed to a new understanding of TAG biosynthesis in seeds. The rod1 (reduced oleate desaturation 1) mutant was isolated 20 years ago as a plant whose seeds had a marked decrease in 18:2 + 18:3 fatty acids and a concomitant increase in 18:1 relative to wild-type (Lemieux et al., 1990). Importantly, PC in developing rod1 seeds was found to be more highly unsaturated than in wild-type, suggesting that the mutant is deficient in 18:1 transfer into PC and the reverse transfer of 18:2 + 18:3 into the TAG synthesis pathway. Labeling experiments with 14C-glycerol indicated that conversion of DAG to PC by CPT was the likely lesion in the mutant (Lu et al., 2009). However, sequencing of the two genes encoding CPT, At1g13560 and At3g25585, from the mutant revealed no changes from wild-type. Instead, mapping identified ROD1 as At3g15820, which was annotated as encoding a phosphatidic acid phosphatase-related protein. This annotation turned out to be misleading; ROD1 does not encode a PA phosphatase. Conventional searches of the non-redundant database (using BLAST and FASTA programs) identified homologs in many higher plants (all annotated as proteins of unknown function or phosphatidic acid phosphatase-related), but no matches were found in any other organism. However, when the position-specific iterated BLAST (PSI-BLAST) algorithm was used, the second iteration identified a mammalian phosphatidylcholine:ceramide cholinephosphotransferase. This enzyme, also known as sphingomyelin synthase (SMS), catalyzes transfer of the phosphocholine headgroup from PC to the alcohol of ceramide (Huitema et al., 2004; Sigal et al., 2005). Plants do not contain sphingomyelin, but the structure of ceramide is analogous in some respects to that of DAG, raising the possibility that ROD1 might catalyze the transfer of phosphocholine from PC to DAG in a reaction analogous to that of SMS in animals. This possibility was supported by further sequence analysis of the ROD1 protein. More importantly, assays of recombinant ROD1 expressed in yeast cells indicated that ROD1 does indeed encode phosphatidylcholine:diacylglycerol cholinephosphotransferase (PDCT), a previously unrecognized enzyme of lipid metabolism (Lu et al., 2009).

Conclusion

The genome cornucopia continues to deliver new tools for probing metabolism and other aspects of plant biology. We are only just starting to see the benefits from proteomics, metabolomics and systems biology approaches to data collection and analysis, while high-throughput sequencing technologies are increasing the reach of expression profiling. The modest collection of examples that we have been able to include in this review reveal the power of the approaches that have become available since completion of the Arabidopsis genome sequence. They also demonstrate that the biochemistry of metabolism, despite its ancient roots in the research enterprise, still holds some secrets and surprises that the genome can help us discover.

Acknowledgements

Like everyone else in the community, we owe a huge debt of thanks to the Arabidopsis Genome Initiative, and to the many other research groups who have contributed to the development of resources for the weed. We thank Deirdre Fahy for preparing the figures and Joyce Tamura for secretarial assistance. Research on lipid metabolism in our laboratory is funded by grants from the US National Science Foundation (grants MCB-0420199 and DBI-0701919), Agricultural and Food Research Initiative Competitive grant number 2010-65115-20393 from the USDA National Institute of Food and Agriculture, and by the Agricultural Research Center at Washington State University.

Ancillary