Functional annotation of the Arabidopsis P450 ... - CiteSeerX

1192

Biochemical Society Transactions (2006) Volume 34, part 6

Functional annotation of the Arabidopsis P450 superfamily based on large-scale co-expression analysis J. Ehlting*1 , N.J. Provart† and D. Werck-Reichhart* *IBMP (Institut de Biologie Moleculaire ´ des Plantes), CNRS (Centre National de la Recherche Scientifique), 28 rue Goethe, 67000 Strasbourg, France, and †Department of Cell and Systems Biology, University of Toronto, 25 Willcocks Street, Toronto, ON, Canada M5S 3B2

Abstract Cytochrome P450 mono-oxygenases play prominent roles in a diverse set of metabolic pathways, but the function of most of these enzymes remains obscure. A bottleneck in the functional genomics of this superfamily constitutes hypothesis generation to identify potential substrates (or substrate classes) individual P450s may act on. We used publicly available large-scale expression data to perform co-expression analysis comparing the expression matrix of each P450 with those from more than 4000 selected genes across thousands of microarrays. Based on functional annotations of co-expressed genes from a diverse set of databases, co-expressed pathways were thus identified for each P450. Using this approach, most P450s with known functions were placed into their respective pathways, thereby proofing the concept. As examples, pathway mapping results identifying novel P450s potentially acting on flower-specific monoterpenes and root-specific triterpenes are described. Co-expression results for all Arabidopsis P450s will be presented as a web resource on the ‘CYPedia’ web pages (http://ibmp.u-strasbg.fr/∼CYPedia/).

Introduction Plants synthesize a vast array of diverse organic compounds that serve important adaptive functions in protection against pests, as attractants for pollinators, as allelochemicals, as structural components and as signalling molecules, just to mention some [1]. Recent genome sequencing initiatives revealed the presence of large gene families that probably encode enzymes involved in the biosynthesis of these compounds [2–4]. Among these is the class of cytochrome P450 mono-oxygenases, which constitutes the largest family of plant enzymes, with 272 genes annotated in Arabidopsis and 457 in rice, compared with only 57 in the human genome [5–7]. The size of this family is thought to reflect the complexity of plant (secondary) metabolism, but the functions of the vast majority of the cytochrome P450 enzymes are still obscure. They catalyse reactions difficult to perform via chemical synthesis, including hydroxylations, N- or O-dealkylations, epoxydations, deformylations, dehydrations, desaturations, isomerizations, C-C cleavages, ring extensions, dimerizations and decarboxylations [8], and frequently constitute rate-limiting steps in their respective pathways. In Arabidopsis, more than 40 P450s have been characterized in some detail, and for 38 genes the biochemical

Key words: allene oxide synthase (AOS), Arabidopsis, co-expression analysis, cytochrome P450, metabolic pathway, root-specific gene. Abbreviations used: AOC, allene oxide cyclase; AOS, allene oxide synthase; BR, brassinosteroid; CYP, cytochrome P450; FunCat, Functional Catalogue; KEGG, Kyoto Encyclopedia of Genes and Genomes; LOX, lipoxygenase; MRN, marneral synthase; OPR, 12-oxophytodienoate reductase; TAIR, The Arabidopsis Information Resource; TTPS, triterpene synthase. 1

To whom correspondence should be addressed (email [email protected]).

C 2006

Biochemical Society

function has been elucidated. This makes Arabidopsis the plant with the highest number of characterized P450s. Despite this impressive body of work on the functional characterization of this gene family, it should be noted that more than 200 genes encoded in the Arabidopsis genome still await functional characterization. Several approaches are being pursued addressing these ‘orphan’ P450s. On the one hand, collections of mutants are being generated for reverse genetics. These include T-DNA (transfer DNA) insertion mutants from public large-scale projects [9,10], but also overexpression and down-regulation lines using 35 S- and RNA interference constructs for selected subfamilies. These mutant lines can then be used for phenotypic analysis. However, it can be expected that in many cases phenotypes are very subtle or are highly localized and may not be apparent by visual inspection. In particular, molecular phenotypes, typically detected by targeted or large scale metabolic profiling, may be highly specific to individual cells or tissues and might be difficult to be identified when using whole plant extracts. As a second approach, screens of biochemical functions are appealing to identify novel enzymatic functions of P450s. To this end, libraries of P450s overexpressed in yeast in combination with an appropriate cytochrome P450 reductase are becoming available. Medium- to high-throughput methods for biochemical assays are now being developed for screening compound libraries as potential substrates. However, from the primary structure, it is usually difficult to predict the potential substrate class, which makes it unfeasible to perform targeted screenings. In summary, multiple tools for functional genomics of P450s are readily available, but a crucial bottleneck to exploit these tools is hypothesis

8th International Symposium on Cytochrome P450 Biodiversity and Biotechnology

generation, i.e. identifying leads to the pathways each P450 could be involved in. Thereby the overall search space would be reduced, allowing the selection of the appropriate substances (or substance classes) for targeted enzymatic assays and metabolic profiling of mutants. In a similar way, information on the expression pattern is a helpful support to select appropriate organs, tissues and treatment for phenotypic analyses of mutants, again a huge challenge if every single tissue and treatment needed to be examined for every P450 mutant. The abundance of publicly available microarray data and the fact that many P450s are regulated on the transcriptional level may be exploited to address both bottlenecks. Therefore we generated extensive expression datasets based on thousands of microarrays and used these expression data to identify genes expressed in similar manners with each P450 across hundreds of biological samples. Based on the hypothesis that genes encoding enzymes acting in the same biochemical pathways are co-regulated, the knowledge available for these co-expressed genes can be used to place individual P450s into metabolic pathways and thereby generate hypotheses about their biochemical function.

Expression profiling Analysis of transcript abundance is a powerful tool to understand gene function and regulation. Over the last few years, several large-scale approaches have been established that include SAGE (serial analysis of gene expression), MPSS (massive parallel signature sequencing) and DNA microarrays [11]. Among the latter, in particular the Affymetrix R GeneChip platform has been used most extensively (e.g. [12]) and thousands of array hybridization data are publicly available [13]. Although probe sets for this array were not specifically designed to discriminate P450 family members, the Arabidopsis ATH1 array contains 259 probe sets representing 227 P450 genes (83% of all P450s). Among these, 192 genes are represented by specific probe sets probably not cross-hybridizing with other genes. For the Affymetrix R ATH1 GeneChip , more than 2500 microarray datasets covering more than 800 biological experiments are available online, e.g. from the ‘Genevestigator’ database (https:// www.genevestigator.ethz.ch/at/) [13]. We retrieved these expression data from the ‘Genevestigator Digital Northern’ database (https://www.genevestigator.ethz.ch/at/) and generated extensive expression matrices for all available P450s separated into (i) organ and tissue data (277 biological samples; Figure 1), (ii) stress treatments (60 biotic stress treatments and 179 abiotic stress treatments, each compared with control samples), (iii) hormone and other treatments (43 hormone treatments, 32 treatments with hormone-related substances and 44 other treatments, e.g. nutrient deprivation), and (iv) mutant wild-type comparisons (218 mutant samples compared with the equivalent wild-type samples). We selected a total of 218 Affymetrix probe sets representing the 227 recognized P450s and included in our data processing a rather rigorous background correction to avoid artificially high expression ratios at low signal intensities, thereby accepting potential loss of expression information close to back-

ground noise. Details will be published elsewhere (J. Ehlting, V. Sauveplane, A. Olry, N. Provart and D. Werck-Reichhart, unpublished work) and all expression matrices will be available online from the ‘CYPedia’ web pages (http://ibmp.ustrasbg.fr/∼CYPedia/). From these expression matrices, it becomes clear that individual P450s display a wide range of expression profiles in agreement with their wide range of biochemical and physiological function. For example, in the organ dataset, only five P450s were not detectably expressed in any sample, CYP71B21 (where CYP is cytochrome P450) [TAIR (The Arabidopsis Information Resource) accession number AT3G26190, Affymetrix ATH1 probeset 257632 at], CYP71B30P (TAIR accession number AT3G53290, Affymetrix ATH1 probeset 251978 at), CYP77A9 (TAIR accession number AT5G04630, Affymetrix ATH1 probeset 250838 at), CYP79A4P (TAIR accession number AT5G35920, Affymetrix ATH1 probeset 249673 at) and CYP96A14P (TAIR accession number AT1G66030, Affymetrix ATH1 probeset 256520 at), and three of these (CYP71B30P, CYP77A9 and CYP96A14P) were also not detectably responsive to any treatment or mutation. In contrast, 200 P450 probe sets (88% of the represented genes) yielded signal intensities more than 4-fold over background in at least one organ or tissue sample. However, a relatively large group of 50 P450 probe sets yielded very low or undetectable expression levels in most tissues and organs, while a third group of 59 probe sets indicate fairly constitutive expression of the represented genes in most organs with the exception of cell cultures and pollen (the lower half of Figure 1). The latter group may thus contain genes with house-keeping functions, while the former may represent genes encoding specialized functions not needed during normal development and may also contain recent gene duplicates, which have lost or are on their way to losing their function; this is particularly true for the five pseudogenes represented on the Affymetrix array, which all fall into this group. The remaining 91 P450 probe sets display predominant or exclusive expression in selected organs or tissues (Figure 1, top half). These include genes specifically expressed during flower development, or in coordination with seed maturation (Figure 1). A rather large group of 54 probe sets (representing 57 genes) show high expression predominantly in roots; among these are several genes encoding enzymes involved in the metabolism of sterols and BRs (brassinosteroids), but also P450s involved in the biosynthesis of other hormones such as abscisic acid, cytokinin and gibberellin. These include CYP710A1 (TAIR accession number AT2G34500, Affymetrix ATH1 probeset 266995 at) characterized as a sterol C-22 desaturase (β-sitosterol to stigmasterol) [14], CYP72C1 (SHK1, SOB7, CHI2; TAIR accession number AT1G17060, Affymetrix ATH1 probeset 262525 at), which controls BR homoeostasis by modulating BR concentration and is probably involved in BR degradation [15–17], and CYP90D1 (TAIR accession number AT3G13730, Affymetrix ATH1 probeset 256788 at), which is involved downstream in BR synthesis [18]. Also involved in the biosynthesis of isoprenoid C 2006

Biochemical Society

1193

1194


Figure 1 Organ expression matrix of the cytochrome P450 superfamily Data from published Affymetrix microarrays (representing 167 organ and tissue samples) were retrieved from the ‘Genevestigator’ database (https://www.genevestigator.ethz.ch/at/) [13]. Background was defined as the mean intensities of all experiments where a given probe sets was called ‘absent’. Where this was the case, the expression value was set to this background value. The log2 of the background was subtracted from the log2 of the mean signal intensity from replicate arrays and these log2 ratios were used for K-means clustering [41] to place all P450 genes into 20 clusters. Clusters are separated by horizontal green lines; vertical green lines separate groups of samples belonging to the same type as indicated on top of the heatmap. Signal intensity ratios (log2 over background) are colour-coded as indicated on the bottom of the heatmap. The equivalent fold changes are also indicated.

derivatives are the root-specific CYP707A1 (TAIR accession number AT4G19230, Affymetrix ATH1 probeset 254562 at) encoding abscisic acid 8 -hydroxylase involved in abscisic acid degradation [19–21] and CYP88A3 (KAO1; TAIR accession number AT1G05160, Affymetrix ATH1 probeset 264586 at) encoding a multi-functional ent-kaurenoic acid oxidase involved in gibberellin metabolism [22]. CYP735A2 (TAIR accession number AT1G67110, Affymetrix ATH1 probeset 264470 at), encoding a trans-hydroxylase for isopentenyladenine tri/di/mono-phosphates, which is involved in cytokinin metabolism [23], and CYP86A1 (TAIR accession number AT5G58860, Affymetrix ATH1 probeset 247765 at), encoding ω-hydroxylase for saturated and unsaturated C12 to C18 fatty acids [24,25], can also be found in the group of rootspecific P450s. Among the additional 51 root-specific P450 genes, members of the CYP705A family represent the largest group: C 2006

Biochemical Society

15 out of the 23 genes of this family with Affymetrix probe sets are root-specific. The remaining root-specific P450 genes are spread across several families and include members from more than 20 different subfamilies. Details on the information for these genes and for those characterized by other organ-specific profiles will be available from the ‘CYPedia’ web pages (http://ibmp.u-strasbg.fr/∼CYPedia/). When compared with data collected from a P450 boutique array (http://arabidopsis-p450.biotec.uiuc.edu/), the majority (74%) of the identified root-specific genes are also exclusively or predominantly expressed in roots (compared to shoots, leaves, stems or flowers) using this platform. An additional eight apparently root-specific genes (15%) are not expressed to detectable levels in any sample using the boutique array, and for only six genes (11%) was an apparent contradiction observed, e.g. CYP714A1 (TAIR accession number AT5G24910, Affymetrix ATH1 probeset 246978 at)


Figure 2 Co-expression analysis using CYP74A encoding AOS as a bait Data from published Affymetrix microarrays (representing 243 stress treatments) were retrieved from the ‘Genevestigator’ database (https://www.genevestigator.ethz.ch/at/) [13]. Background correction was performed as described in Figure 1. Mean log2 intensities from replicate control samples (where applicable) were subtracted from mean log2 intensities from replicate treatments to generate log2 ratios comparing treatment with control. The stress expression vector of CYP74A was compared with those of 4100 genes annotated in diverse databases to be involved in any metabolic pathway using the ‘ExpressionAngler’ algorithm (http://bbc.botany.utoronto.ca/) [26]. Expression profiles of co-expressed genes with a correlation coefficient of more than 0.575 are shown as a heatmap. Groups of samples are indicated on top of the heatmap. Signal intensity ratios (log2 ratios comparing treatment with control) are colour-coded as indicated on the bottom of the heatmap and the equivalent fold changes are also indicated. Genes encoding enzymes acting in the LOX pathway are highlighted in orange. Abbreviations: MGD, putative 1,2-diacylglycerol 3-β-galactosyltransferase; CLH, chlorophyllase; APS, sulfate adenylyltransferase.

appears flower-specific and is not expressed to detectable levels in other organs based on the P450 boutique array, while Affymetrix data suggest that it is expressed in roots as well as in flowers and particularly in late stages of seed development (see ‘CYPedia’ entry for CYP714A1; http:// ibmp.u-strasbg.fr/∼CYPedia/). Similar to the organ and tissue profiling, individual P450s also display unique and distinct expression patterns in response to biotic and abiotic stresses, to hormone and other treatments and as a consequence of mutations in other genes. All these profiles will be presented as searchable and browsable files using common formats on the ‘CYPedia’ web pages (http://ibmp.u-strasbg.fr/∼CYPedia/). In summary, these expression matrices provide valuable information for targeted analysis of mutants in P450s of interest and for metabolic profiling. However, per se, they may point to a physiological role of a gene of interest, but do not necessary indicate a biochemical pathway the gene might act in.

Co-expression analysis Given the large abundance of microarray data available, it is now feasible to identify genes that share over a wide range of experiments a common expression profile and may thus be expected to act in the same biological process or biochemical pathway. Several co-expression analysis tools have been de-

veloped. This includes the calculation of correlations between single genes across thousands of arrays (or subsets thereof) as presented by Zimmermann et al. [13] using the ‘GeneCorrelator’ tool (https://www.genevestigator.ethz.ch/at/), the calculation and visualization of correlation networks among groups of genes using the ‘Co-expression Gene Search’ algorithm or to retrieve a cluster of co-expressed genes from a hierarchical cluster analysis using ‘Cluster Extract’ (both resources from the ‘Platform for RIKEN metabolomics’ (http://prime.psc.riken.jp/) performed with array data from the ‘AtGenExpress’ consortium [12]. Using different sets of whole genome array data, the ‘ExpressionAngler’ algorithm calculates correlation coefficients of a gene of interest against all other genes represented on the arrays and visualizes the resulting co-expression matrices as heatmaps (http:// bbc.botany.utoronto.ca/) [26]. In order to focus our co-expression analysis, we selected approx. 4000 Arabidopsis genes that were annotated in a range of pathway annotation databases that include the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway modules (Kyoto University Bioinformatics Center; http://www.genome.jp/dbget-bin/get htext?A.thaliana. kegg+-f+F+D) [27], the ‘AraCyc’ pathway database (http://arabidopsis.org/tools/aracyc) [28], the MIPS (Munich Information Center for Protein Sequences) FunCat C 2006

Biochemical Society

1195

1196


Figure 3 Co-expression analysis using CYP705A12 (unknown function) as a bait Data from published Affymetrix microarrays (representing 105 treatments with hormones, hormone-related substances and nutrients) were retrieved from the ‘Genevestigator’ database (https://www.genevestigator.ethz.ch/at/) [13]. Background correction, ratio generation and co-expression analysis was performed as described in Figure 2. Expression profiles of co-expressed genes with a correlation coefficient of more than 0.600 are shown as a heatmap. Groups of samples are indicated on top of the heatmap. Signal intensity ratios are colour-coded as indicated on the bottom of the heatmap. Abbreviations: SQP, putative squalene mono-oxygenases; GSTU, putative glutathione transferase; SERAT, serine O-acetyltransferase; BFRUCT, β-fructosidase; LTP, lipid transfer family protein; SDR, short-chain dehydrogenase/reductase family protein.

(Functional Catalogue) (http://mips.gsf.de/proj/funcatDB/ search main frame.html) [29], the ‘The Arabidopsis Lipid Gene Database’ (AcylLipid) (http://www.plantbiology.msu. edu/lipids/genesurvey/index.htm) [30], and the ‘BioPathAt’ database (http://www.wsu.edu/∼lange-m/biochemical.htm) [31]. We retrieved expression data for these genes from the ‘Genevestigator’ database (https://www.genevestigator. ethz.ch/at/) [13] and added the processed data to the four P450 expression matrices described above. Using each P450 as a bait, we calculated Pearson correlation coefficients of mean-centred expression vectors [26] against each other gene, resulting in lists of co-expressed genes in each of the four datasets. Using manual curation, each co-expressed gene was given an annotation score, and based on the score and number of co-expressed genes that were placed into the same biochemical pathway within any of the annotation databases used, co-expressed biochemical pathways were identified for each P450 (J. Ehlting, V. Sauveplane, A. Olry, N. Provart and D. Werck-Reichhart, unpublished work). C 2006

Biochemical Society

The underlying hypothesis for this approach assumes that genes encoding enzymes of the same biochemical pathway are transcriptionally co-regulated and share a common expression profile. CYP74A encodes AOS (allene oxide synthase) and is a well-characterized P450 involved in the biosynthesis of jasmonate and other oxylipins [32–34]. When used as a bait for co-expression analysis, this gene was found to be tightly co-expressed across 243 stress treatment samples with several other genes, which have been characterized to encode enzymes of the oxylipin pathway (Figure 2). These include AOC (allene oxide cyclase) (AOC1), OPR (12-oxophytodienoate reductase) (OPR3), two LOXs (lipoxygenases) (LOX2 and LOX3), and HPL (hydroperoxide lyase) (HPL1/CYP74B2) [35]. Based on the annotation of these and other co-expressed genes, the top scoring co-expressed pathways identified were: ‘Lipid signalling’ (AcylLipid), ‘jasmonic acid biosynthesis’ (TAIR-GO and AraCyc) and ‘LOX pathway’ (AraCyc) in accordance with the actual biochemical function. Similar results were found


for most of the characterized P450s, thus proving the concept of this approach. However, such clear pathway mapping results were not obtained for all P450s, either due to a lack of informative expression data (resulting in no co-expressed genes), or due to the fact that most co-expressed genes have not been characterized and are members of large gene families themselves. Many of the above-mentioned root-specific P450s fall into this category, but for others, more specific, but divergent pathways are co-expressed, highlighting the ability of this approach to distinguish co-expression sets even within larger groups with apparently similar expression patterns. For example, along with seven other P450s, CYP705A12 (TAIR accession number At5g42580, Affymetrix ATH1 probeset 249202 at) shares a similar organ expression profile with genes encoding the TTPSs (triterpene synthases) (i) MRN1 [marneral synthase 1; TTPS5; TAIR accession number At5g42600, Affymetrix ATH1 probeset 249205 at], which produces a bicyclic triterpenoid intermediate that is further metabolized to form marneral [36], and (ii) TTPS6 (TAIR accession number At5g48010, Affymetrix ATH1 probeset 248729 at), which produces from oxidosqualene a tricyclic triterpenoid called thalianol [37]. MRN1 is tightly coexpressed with CYP705A12 also in the hormone dataset, together with another putative TTPS gene (PEN1, TTPS2; TAIR accession number At4g15340, Affymetrix ATH1 probeset 245258 at) [38] and other genes related to the biosynthesis of triterpenoid and/or steroids (Figure 3). These include a gene similar to squalene mono-oxygenase (SQP2; TAIR accession number At5g24140, Affymetrix ATH1 probeset 249773 at) [39], one similar to steroid sulfotransferase (TAIR accession number At1g28170, Affymetrix ATH1 probeset 245663 at) and a putative geranylgeranyl pyrophosphate synthase (GGPPS8; TAIR accession number At3g20160, Affymetrix ATH1 probeset 257117 at) [40]. Finally, MNR1 and TTPS6 are also co-expressed with CYP705A12 in the mutant dataset and are suppressed in det2 and ga1 mutants but induced in zorro mutants. Based on the functional annotation of these co-expressed genes, the top scoring co-expressed pathways are ‘Biosynthesis of steroids’ (KEGG), ‘Isoprenoid biosynthesis’ (FunCat) and ‘Pentacyclic triterpenoid biosynthesis’ (TAIR-GO). Together, these results thus suggest that CYP705A12 (and other P450s coexpressed with its genes) is involved in modulating triterpenoids possibly using as a substrate the product of MRN1, maneral.

Conclusions Using these examples it is clear that co-expression analysis across thousands of public microarray datasets in combination with detailed functional annotation is a powerful tool for hypothesis generation. Co-expressed genes and pathways and thus potential functions for all P450s will be found on the ‘CYPedia’ web pages as a resource for the plant science community (J. Ehlting, V. Sauveplane, A. Olry, N. Provart and D. Werck-Reichhart, unpublished work) (http://ibmp.ustrasbg.fr/∼CYPedia/). In addition, it is also possible to

browse through the biochemical pathways to find P450s coexpressed with genes from these pathways. This information can now be utilized to perform targeted phenotypic screens of mutants or to perform biochemical characterizations of recombinant proteins.

References 1 Croteau, R., Kutchan, T. and Lewis, N. (2005) in Biochemistry and Molecular Biology of Plants (Buchanan, B., Gruissem, W. and Jones, R., eds.), John Wiley and Sons, Rockville 2 D’Auria, J.C. and Gershenzon, J. (2005) Curr. Opin. Plant Biol. 8, 308–316 3 International Rice Genome Sequencing Project (2005) Nature 436, 793–800 4 The Arabidopsis Genome Initiative (2000) Nature 408, 796–815 5 Didierjean, L., Gondet, L., Perkins, R., Lau, S.M., Schaller, H., O’Keefe, D.P. and Werck-Reichhart, D. (2002) Plant Physiol. 130, 179–189 6 Nelson, D.R., Schuler, M.A., Paquette, S.M., Werck-Reichhart, D. and Bak, S. (2004) Plant Physiol. 135, 756–772 7 Schuler, M.A. and Werck-Reichhart, D. (2003) Annu. Rev. Plant Biol. 54, 629–667 8 Mansuy, D. (1998) Comp. Biochem. Physiol. C Pharmacol. Toxicol. Endocrinol. 121, 5–14 9 Alonso, J.M., Stepanova, A.N., Leisse, T.J., Kim, C.J., Chen, H., Shinn, P., Stevenson, D.K., Zimmerman, J., Barajas, P., Cheuk, R. et al. (2003) Science 301, 653–657 10 Rosso, M.G., Li, Y., Strizhov, N., Reiss, B., Dekker, K. and Weisshaar, B. (2003) Plant Mol. Biol. 53, 247–259 11 Pollock, J.D. (2002) Chem. Phys. Lipids 121, 241–256 12 Schmid, M., Davison, T.S., Henz, S.R., Pape, U.J., Demar, M., Vingron, M., Scholkopf, B., Weigel, D. and Lohmann, J.U. (2005) Nat. Genet. 37, 501–506 13 Zimmermann, P., Hirsch-Hoffmann, M., Hennig, L. and Gruissem, W. (2004) Plant Physiol. 136, 2621–2632 14 Morikawa, T., Mizutani, M., Aoki, N., Watanabe, B., Saga, H., Saito, S., Oikawa, A., Suzuki, H., Sakurai, N., Shibata, D., Wadano, A., Sakata, K. and Ohta, D. (2006) Plant Cell 18, 1008–1022 15 Nakamura, M., Satoh, T., Tanaka, S., Mochizuki, N., Yokota, T. and Nagatani, A. (2005) J. Exp. Bot. 56, 833–840 16 Takahashi, N., Nakazawa, M., Shibata, K., Yokota, T., Ishikawa, A., Suzuki, K., Kawashima, M., Ichikawa, T., Shimada, H. and Matsui, M. (2005) Plant J. 42, 13–22 17 Turk, E.M., Fujioka, S., Seto, H., Shimada, Y., Takatsuto, S., Yoshida, S., Wang, H., Torres, Q.I., Ward, J.M., Murthy, G., Zhang, J., Walker, J.C. and Neff, M.M. (2005) Plant J. 42, 23–34 18 Kim, G.T., Fujioka, S., Kozuka, T., Tax, F.E., Takatsuto, S., Yoshida, S. and Tsukaya, H. (2005) Plant J. 41, 710–721 19 Kushiro, T., Okamoto, M., Nakabayashi, K., Yamagishi, K., Kitamura, S., Asami, T., Hirai, N., Koshiba, T., Kamiya, Y. and Nambara, E. (2004) EMBO J. 23, 1647–1656 20 Okamoto, M., Kuwahara, A., Seo, M., Kushiro, T., Asami, T., Hirai, N., Kamiya, Y., Koshiba, T. and Nambara, E. (2006) Plant Physiol. 141, 97–107 21 Saito, S., Hirai, N., Matsumoto, C., Ohigashi, H., Ohta, D., Sakata, K. and Mizutani, M. (2004) Plant Physiol. 134, 1439–1449 22 Helliwell, C.A., Chandler, P.M., Poole, A., Dennis, E.S. and Peacock, W.J. (2001) Proc. Natl. Acad. Sci. U.S.A. 98, 2065–2070 23 Takei, K., Yamaya, T. and Sakakibara, H. (2004) J. Biol. Chem. 279, 41866–41872 24 Benveniste, I., Tijet, N., Adas, F., Philipps, G., Salaun, J.P. and Durst, F. (1998) Biochem. Biophys. Res. Commun. 243, 688–693 25 Duan, H. and Schuler, M.A. (2005) Plant Physiol. 137, 1067–1081 26 Toufighi, K., Brady, S.M., Austin, R., Ly, E. and Provart, N.J. (2005) Plant J. 43, 153–163 27 Kanehisa, M. and Goto, S. (2000) Nucl. Acids Res. 28, 27–30 28 Zhang, P., Foerster, H., Tissier, C.P., Mueller, L., Paley, S., Karp, P.D. and Rhee, S.Y. (2005) Plant Physiol. 138, 27–37 29 Ruepp, A., Zollner, A., Maier, D., Albermann, K., Hani, J., Mokrejs, M., Tetko, I., Guldener, U., Mannhaupt, G., Munsterkotter, M. and Mewes, H.W. (2004) Nucleic Acids Res. 32, 5539–5545 30 Beisson, F., Koo, A.J., Ruuska, S., Schwender, J., Pollard, M., Thelen, J.J., Paddock, T., Salas, J.J., Savage, L., Milcamps, A., Mhaske, V.B., Cho, Y. and Ohlrogge, J.B. (2003) Plant Physiol. 132, 681–697 C 2006

Biochemical Society

1197

1198


31 32 33 34

Lange, B.M. and Ghassemian, M. (2005) Phytochemistry 66, 413–451 Laudert, D. and Weiler, E.W. (1998) Plant J. 15, 675–684 Laudert, D., Schaller, F. and Weiler, E.W. (2000) Planta 211, 163–165 Park, J.H., Halitschke, R., Kim, H.B., Baldwin, I.T., Feldmann, K.A. and Feyereisen, R. (2002) Plant J. 31, 1–12 35 Feussner, I. and Wasternack, C. (2002) Annu. Rev. Plant Biol. 53, 275–297 36 Xiong, Q., Wilson, W.K. and Matsuda, S.P. (2006) Angew. Chem. 45, 1285–1288 37 Fazio, G.C., Xu, R. and Matsuda, S.P. (2004) J. Am. Chem. Soc. 126, 5678–5679

C 2006

Biochemical Society

38 Husselstein-Muller, T., Schaller, H. and Benveniste, P. (2001) Plant Mol. Biol. 45, 75–92 39 Schafer, ¨ U., Reed, D., Hunter, D., Yao, K., Weninger, A., Tsang, E., Reaney, M., MacKenzie, S. and Covello, P. (1999) Plant Mol. Biol. 39, 721–728 40 Lange, B.M. and Ghassemian, M. (2003) Plant Mol. Biol. 51, 925–948 41 Sturn, A., Quackenbush, J. and Trajanoski, Z. (2002) Bioinformatics 18, 207–208

Received 20 July 2006