Peroxisomal Δ32-enoyl CoA isomerases and evolution of cytosolic paralogues in embryophytes

Authors

  • Simon Goepfert,

    1. Département de Biologie Moléculaire Végétale, Biophore, Université de Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author
    • Present address: Research and Development, Philip Morris International, Quai Jeanrenaud 56, 2000 Neuchâtel, Switzerland.

  • Charles Vidoudez,

    1. Département de Biologie Moléculaire Végétale, Biophore, Université de Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author
  • Christian Tellgren-Roth,

    1. Bioinformatics Core, Center for Rural Health Research and Education, University of Wyoming, Laramie, WY 82071, USA
    Search for more papers by this author
  • Syndie Delessert,

    1. Département de Biologie Moléculaire Végétale, Biophore, Université de Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author
  • J. Kalervo Hiltunen,

    1. Biocenter Oulu and Department of Biochemistry, University of Oulu, FI-90014 Oulu, Finland
    Search for more papers by this author
  • Yves Poirier

    Corresponding author
    1. Département de Biologie Moléculaire Végétale, Biophore, Université de Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author

(fax +41 21 692 4195; e-mail yves.poirier@unil.ch).

Summary

Δ32-enoyl CoA isomerase (ECI) is an enzyme that participates in the degradation of unsaturated fatty acids through the β-oxidation cycle. Three genes encoding Δ32-enoyl CoA isomerases and named AtECI1, AtECI2 and AtECI3 have been identified in Arabidopsis thaliana. When expressed heterologously in Saccharomyces cerevisiae, all three ECI proteins were targeted to the peroxisomes and enabled the yeast Δeci1 mutant to degrade 10Z-heptadecenoic acid, demonstrating Δ32-enoyl CoA isomerase activity in vivo. Fusion proteins between yellow fluorescent protein and AtECI1 or AtECI2 were targeted to the peroxisomes in onion epidermal cells and Arabidopsis root cells, but a similar fusion protein with AtECI3 remained in the cytosol for both tissues. AtECI3 targeting to peroxisomes in S. cerevisiae was dependent on yeast PEX5, while expression of Arabidopsis PEX5 in yeast failed to target AtECI3 to peroxisomes. AtECI2 and AtECI3 are tandem duplicated genes and show a high level of amino acid conservation, except at the C-terminus; AtECI2 ends with the well conserved peroxisome targeting signal 1 (PTS1) terminal tripeptide PKL, while AtECI3 possesses a divergent HNL terminal tripeptide. Evolutionary analysis of ECI genes in plants revealed several independent duplication events, with duplications occurring in rice and Medicago truncatula, generating homologues with divergent C-termini and no recognizable PTS1. All plant ECI genes analyzed, including AtECI3, are under negative purifying selection, implying functionality of the cytosolic AtECI3. Analysis of the mammalian and fungal genomes failed to identify cytosolic variants of the Δ32-enoyl CoA isomerase, indicating that evolution of cytosolic Δ32-enoyl CoA isomerases is restricted to the plant kingdom.

Introduction

The β-oxidation pathway is the principal route of fatty acid degradation and occurs in the peroxisomes in plants. In many plant species, the fatty acids stored in triacylglycerol are the principal carbon reserve of the seed. Catabolism of the fatty acids through the β-oxidation cycle is thus of prime importance during seed germination for the provision of carbon and energy for growth before the establishment of photosynthesis (Graham, 2008). A basal level of β-oxidation is also required in all cells to allow recycling of the fatty acids from the turnover of membranes or from fatty acids that are not inserted into lipids (Mittendorf et al., 1999; Poirier et al., 1999). In addition, β-oxidation is involved in the generation of signaling molecules, such as members of the jasmonate family, and potentially other signals necessary for proper germination and seedling establishment (Baker et al., 2006; Nyathi and Baker, 2006).

The peroxisomal core β-oxidation cycle in plants comprises four enzymatic activities located on three proteins. The first enzyme is acyl CoA oxidase, converting acyl CoA into 2E-enoyl CoA. The second is a multi-functional enzyme that harbors enoyl CoA hydratase and 3-hydroxyacyl CoA dehydrogenase, catalyzing the successive conversion of 2E-enoyl CoA into 3-hydroxyacyl CoA and 3-ketoacyl CoA, respectively. The final enzyme is 3-ketoacyl thiolase, and is responsible for cleavage of 3-ketoacyl CoA to form acetyl CoA and an acyl CoA that is two carbons shorter and can re-enter the β-oxidation spiral. Several genes encoding enzymes of the core β-oxidation cycle have been identified in Arabidopsis thaliana and other plants (Goepfert and Poirier, 2007; Graham and Eastmond, 2002).

In contrast to the degradation of saturated fatty acids, which can be mediated completely by the four enzymatic activities of core β-oxidation cycle, degradation of cis(Z)-unsaturated fatty acids, which are very abundant in plant lipids, requires the presence of auxiliary enzymes. This is because core β-oxidation functions through a 2-trans(E)-enoyl CoA intermediate, and degradation of cis(Z)-unsaturated fatty acids leads to the generation of enoyl CoA intermediates that cannot be directly metabolized via the core β-oxidation enzymes.

At least three auxiliary enzymes have been identified in metabolism of unsaturated fatty acids (for review, see Hiltunen et al., 2003;Poirier et al., 2006). In eukaryotes, 2,4-dienoyl CoA reductase catalyzes the conversion of 2E,4Z- or 2E,4E-dienoyl CoA to 3E-enoyl CoA. Δ32-enoyl CoA isomerase converts 3Z- or 3E-enoyl CoA into 2E-enoyl CoA, the normal intermediate of the core β-oxidation cycle. In many organisms, Δ32-enoyl CoA isomerase activity has been found to be associated with the multi-functional enzyme of the core β-oxidation cycle as well as with mono-functional enzymes. A cucumber multi-functional protein has been reported to have isomerase activity in the sub-domain responsible for the hydration step of β-oxidation (Preisig-Muller et al., 1994). However, high Δ32-enoyl CoA isomerase activity in cucumber seedlings was also associated with mono-functional isomerases (Engeland and Kindl, 1991). Genes encoding mono-functional Δ32-enoyl CoA isomerase or 2,4-dienoyl CoA reductase activities have not yet been characterized in plants. The third auxiliary enzyme is Δ3,52,4-dienoyl CoA isomerase, catalyzing the conversion of 3,5-dienoyl CoA to 2,4-dienoyl CoA, and has been implicated in the degradation of conjugated fatty acids as well as cis-unsaturated fatty acids. A gene encoding a Δ3,52,4-dienoyl CoA isomerase has been identified in Arabidopsis and other plants (Goepfert et al., 2005). Furthermore, a mono-functional enoyl CoA hydratase II catalyzing the reversible conversion of 3R-hydroxyacyl CoA to 2E-enoyl CoA has recently been characterized in A. thaliana and shown to participate in the degradation of cis-unsaturated fatty acids (Goepfert et al., 2006).

Forward-genetic approaches have allowed identification of several enzymes involved in peroxisome biogenesis or the peroxisomal β-oxidation pathway, and have shed light on their contribution to the physiology of plants (Baker et al., 2006; Goepfert and Poirier, 2007; Graham, 2008), but no auxiliary enzymes have been identified through these screens. Genes encoding β-oxidation auxiliary enzymes may be difficult to identify through forward-genetic approaches because they may be part of a multi-gene family, or may have functions that are redundant with genes encoding other enzymes, including enzymes of the core β-oxidation pathway. Candidate genes encoding β-oxidation auxiliary enzymes can be identified to some extent by homology searches combined with the use of prediction tools for subcellular localization of proteins to the peroxisome. However, cloning and biochemical characterization of candidate peroxisomal β-oxidation auxiliary enzymes is essential as substrate specificity may be poorly predicted from the primary and secondary structure of proteins. For example, Δ32-enoyl CoA isomerases belong to the isomerase/hydratase superfamily (PFAM domain PF00378), a large family of enzymes sharing common topological features and catalyzing a wide diversity of reactions, ranging from 2-enoyl CoA hydratase to chlorobenzoyl CoA dehalogenase activities (Xiang et al., 1999). Finally, proteins that localize to the peroxisome cannot always be reliably identified through bioinformatics only, as non-peroxisomal proteins containing a predicted peroxisome targeting signal (PTS) as well as peroxisomal proteins with no evident PTS have been identified (Neuberger et al., 2004; Reumann et al., 2007). In this paper, we show that, while A. thaliana contains three proteins with Δ32-enoyl CoA isomerase activity, only two are targeted to the peroxisomes, while the third represents a tandemly duplicated paralogue that has lost its PTS and is located in the cytosol. Such cytosolic Δ32-enoyl CoA isomerases have arisen independently in several plants through gene duplication and C-termini divergence, and appear to be restricted to the plant kingdom.

Results

Identification of Arabidopsis genes encoding a putative Δ32-enoyl CoA isomerase

Identification of peroxisomal mono-functional Δ32-enoyl CoA isomerase was performed by searching the Arabidopsis genome for homologues of the human, yeast and rat mitochondrial and/or peroxisomal mono-functional Δ32-enoyl CoA isomerases using blastp and Clustal X. Candidate proteins were then examined for the presence of conserved catalytic residues based on data obtained from the crystal structure of human and fungal Δ32- enoyl CoA isomerases (Mursula et al., 2001; Partanen et al., 2004). Based on these criteria, putative genes encoding Δ32-enoyl CoA isomerases are At1g65520, At4g14430 and At4g14440, which are hereafter referred as AtECI1, AtECI2 and AtECI3, respectively.

The three AtECI genes consist of a single exon, and their expression is supported by several expressed sequence tags (data not shown). The AtECI1 protein comprises 240 amino acids, with a predicted molecular mass of 26 kDa and a pI of 6.7. The C-terminal tripeptide of AtECI1 is SKL, which is the best represented type 1 peroxisomal targeting signal (PTS1) among peroxisomal proteins (Reumann, 2004). The AtECI2 protein comprises 240 amino acids, with a predicted molecular mass of 26 kDa and a pI of 8.6. The C-terminal tripeptide of AtECI2 is PKL, which is a minor PTS1 (Reumann, 2004) that can also be found in the C-terminal part of the rat peroxisomal Δ32-enoyl CoA isomerase (Geisbrecht et al., 1999). The AtECI3 protein comprises 238 amino acids, with a predicted molecular mass of 26 kDa and a pI of 7.3. In contrast to the two other proteins, which contain a known PTS1, the C-terminal tripeptide of AtECI3 is HNL. While this tripeptide has not been identified as a PTS1 in plants, it resembles the C-terminal tripeptide HRL that is found on yeast peroxisomal Δ32-enoyl CoA isomerase, which is targeted into the peroxisome in a PEX8-dependent manner (Geisbrecht et al., 1998; Gurvitz et al., 1998).

Figure 1 shows the alignment of the three AtECI proteins. The AtECI2 and AtECI3 proteins share 77% identity over the whole protein. The largest divergence is seen in the C-terminal region encompassing the last 15 amino acids. AtECI2 and AtECI3 are located in tandem on chromosome 4, separated by only 1314 bp. Extensive sequence homology at the nucleotide level exist between these two genes in the open reading frame, with 77% identity over the first 809 bases. In contrast, little homology at the nucleotide level exists in the promoter region, the 5′ UTR, the last 33 coding bases, and the 3′ UTR (data not shown). AtECI2 and AtECI3 display 49 and 48% amino acid identity, respectively, with AtECI1 over the first 224 amino acids.

Figure 1.

 Sequence alignment of the three Arabidopsis ECI1 proteins.
The shading threshold for identity/similarity (black/grey boxes) is 50%.

Figure 2 shows a partial alignment between AtECI1, AtECI2 and AtECI3 and regions around the catalytic residues of the mammalian mitochondrial (mECI1) and peroxisomal (pECI1) Δ32-enoyl CoA isomerases, yeast peroxisomal Δ32-enoyl CoA isomerase (ScECI1p), rat mitochondrial enoyl CoA hydratase or crotonase (mECH1), the hydratase/isomerase domain of the multi-functional proteins from Arabidopsis (AtMFP2, AtAIM1) and cucumber (CsMFP-a), as well as the A. thalianaΔ3,52,4-dienoyl CoA isomerase encoded by AtDCI1. The rat mono-functional enoyl CoA hydratase and plant multi-functional proteins were included as they have been found to have residual isomerase activity together with the main hydratase activity (Kiema et al., 1999; Preisig-Muller et al., 1994). In the characterized mammalian and peroxisomal mono-functional isomerases, there is no corresponding base to the crotonase catalytic residue E144, which is, however, present in the hydratase/isomerase domain of the plant multi-functional proteins (Figure 2, arrow 1). In these mono-functional isomerases, the specific glutamate equivalent to the crotonase catalytic E144 is present in a different position in mitochondrial isomerases compared to peroxisomal isomerases (Figure 2, arrows 2 and 3). The Arabidopsis proteins AtECI1, AtECI2 and AtECI3 show high overall homology to peroxisomal mammalian isomerases, but have only one conserved glutamate residue in a position similar to that in the mitochondrial isomerases. The Arabidopsis AtDCI1 shares this conserved glutamate residue with the AtECI proteins.

Figure 2.

 Partial amino acid sequence alignment of AtECI1, AtECI2 and AtECI3 with characterized proteins that have Δ32-enoyl CoA isomerase activity.
The partial sequences of AtECI1, AtECI2 and AtECI3 are aligned with the hydratase domains of the Arabidopsis multi-functional proteins AtAIM1 (At4g29010) and AtMFP2 (At3g06860), the Cucumis sativus multi-functional protein CsMFP-a (Q39659), the Rattus norvegicus mitochondrial crotonase Rn_mECH (CAA34080), the mono-functional mitochondrial 2,3-enoyl CoA isomerases from Homo sapiens (Hs_mECI, P42126), R. norvegicus (Rn_mECI, P23965) and Mus musculus (Mm_mECI, P42125), the mono-functional peroxisomal 3,2-enoyl CoA isomerases from H. sapiens (Hs_pECI, NP_006108), R. norvegicus (Rn_pECI, NP_001006967), M. musculus (Mm_pECI, NP_035998) and S. cerevisiae (ScECI1p, NP_013386), and the Arabidopsis Δ3,52,4-dienoyl CoA isomerase AtDCI1. Only the part of the alignment covering the catalytic residues is shown. The hydratase catalytic glutamate of Rn_mECH1 and the aligned glutamate of the multi-functional proteins AtAIM1, AtMFP2, and CsMFP-a are indicated in turquoise and by arrow 1. The isomerase catalytic glutamate of the mitochondrial Rn_mECH1, Hs_mECI and ScECI1p is indicated in turquoise and by arrow 2 and aligned with the glutamate residue present in AtECI proteins. The isomerase catalytic glutamate of the peroxisomal Hs_pECI, Rn_pECI and Mm_pECI is indicated in turquoise and by arrow 3. The two aspartate residues important for the catalytic acitivity of AtDCI1 are shown in pink. The shading threshold for identity/similarity (black/grey boxes) is 50%.

Expression pattern of the AtECI genes

The expression pattern of the three AtECI genes during seedling development and in various tissues was analyzed by Northern blot analysis (Figure 3). The overall pattern of expression was similar for the three genes, except for the higher expression of AtECI2 in imbibed seeds compared to AtEC1 and AtECI3. In germinating seedlings, expression peaked 1–2 days after imbibition and then decreased and remained relatively constant until 7 days. The three genes are also expressed in roots and shoots of plants grown in either liquid medium or soil, as well as in leaves, stems and flowers from soil-grown plants. Analysis of the expression profile of the AtECI genes using the Genevestigator database (https://www.genevestigator.ethz.ch/) corroborated the rather ubiquitous expression profile observed by Northern blotting, and also revealed conditions under which the AtECI genes are expressed differentially, such as senescence or during silique development (data not shown).

Figure 3.

 Northern blot analysis of AtECI gene expression in Arabidopsis.
Total RNA of entire seedlings was extracted 0, 1, 2, 3, 4, 5 and 7 days after imbibition (first seven lanes), from tissues of plants grown in liquid medium containing 2% sucrose at 15 days after imbibition (whole plants, 15 days wp; leaves, 15 days lv; roots, 15 days rt), and from plants grown in soil for 35 and 40 days after imbibition (whole above-ground plant tissues, 35 days wp and 40 days wp; flower buds, 40 days fl; leaves, 40 days lv; stems, 40 days st). Membranes were hybridized with probes for AtECI1 (a), AtECI2 (b) or AtECI3 (c). Images of the gel stained with ethidium bromide (EtBr) are shown as a loading indicator.

Localization

To determine whether the three AtECI proteins are targeted into the peroxisomes, we addressed their subcellular localization through observation of fluorescent protein fusions, first in S. cerevisiae and subsequently in plants. In yeast, the chimeric AtECI proteins consisted of green fluorescent protein (GFP) at the N-terminus and the AtECI proteins at the C-terminus. The fusion constructs were expressed under the control of the constitutive S. cerevisae glyceraldehyde-3-phosphate dehydrogenase promoter in both wild-type cells and a pex5Δ mutant lacking the PEX5 receptor required for import of peroxisomal proteins containing a PTS1 (Van der Leij et al., 1993). In wild-type cells expressing either GFP–AtECI1 (Figure 4a) or GFP–AtECI3 (Figure 4m), fluorescence was present in a punctate pattern, while diffuse fluorescence was observed in cells expressing native GFP (empty vector) (Figure 4s). For cells expressing GFP–AtECI2, a punctate pattern was observed but a clear background of cytosolic expression was also present (Figure 4g). Only diffuse fluorescence without punctation was observed in pex5Δ0 strains expressing any of the three GFP–AtECI constructs or native GFP (Figure 4c,i,o). Thus, the punctate pattern observed in the wild-type strain is dependent on functional PEX5-mediated peroxisomal import machinery, indicating that AtECI1, AtECI2 and AtECI3 are peroxisomal proteins in yeast cells.

Figure 4.

 Subcellular localization of GFP–AtECI protein fusions within S. cerevisiae.
(a–v) Wild-type cells (WT), the pex5Δ mutant and the pex5Δ mutant expressing Arabidopsis PEX5 (AtPEX5) were transformed with the fusion constructs GFP–AtECI1, GFP–AtECI2 and GFP–AtECI3, or with unmodified GFP, and analyzed for fluorescence by confocal microscopy. For each image obtained by fluorescence microscopy, the corresponding image obtained by transmission microscopy is shown on the right. Scale bar = 5 μm.
(w–z) Wild-type cells were co-transformed with the construct GFP–AtECI3 and pALDsRedAKL, encoding a peroxisomal red fluorescent protein. Fluorescence was monitored with the red (w) and green channels (x), and cells were observed under white light (z). The overlaid red and green fluoresence images are also shown (y).

Localization of the AtECI proteins in plant cells was first tested by transient transformation of onion epidermal cells with constructs encoding fusions between enhanced yellow fluorescent protein (EYFP) at the N-terminus and the AtECI1, AtECI2 and AtECI3 proteins at the C-terminus. As a control, a plasmid carrying a fusion gene between an enhanced cyan fluorescent protein (ECFP) and peroxisomal malate dehydrogenase (MDH) from cucumber was used. All fluorescent proteins were expressed under the control of the CaMV 35S promoter. EYFP–AtECI constructs and ECFP–MDH were transformed either individually or together in onion epidermal cells, and fluorescence was examined by confocal microscopy after 12 h. Control experiments determined that fluorescence of EYFP and ECFP could be detected without cross-interference when expressed in the same cells (data not shown). Cells expressing the EYFP–AtECI1 and EYFP–AtECI2 constructs showed a punctate fluorescence pattern as expected for proteins located in the peroxisomes (Figure 5a,e). In cells co-transformed with EYFP–AtECI1 or EYFP–AtECI2 and ECFP–MDH constructs, the fluorescence pattern observed for EYFP–AtECI1 and EYFP–AtECI2 matched that observed with ECFP–MDH, indicating that AtECI1 and AtECI2 are peroxisomal proteins (Figure 5a–h). These results for AtECI1 and AtECI2 are in agreement with the study by Reumann et al. (2007). In contrast, transformation of onion cells with the EYFP–AtECI3 construct resulted in diffuse fluorescence throughout the cytosol and nucleus, indicating the absence of peroxisomal targeting (Figure 5i).

Figure 5.

 Subcellular localization of YFP–AtECI protein fusions within onion epidermal cells.
Cells were co-transformed with ECFP–MDH, encoding a cyan fluorescent protein fused to the C-terminus of the peroxisomal malate dehydrogenase, and either YFP–AtECI1 (a–d), YFP–AtECI2 (e–h) or YFP–AtECI3 (i, j), encoding a yellow fluorescent protein fused to the N-terminus of AtECI1, AtECI2 or AtECI3, respectively. Images were acquired using the yellow channel (a, e, i) and the cyan channel (b, f) and the two images were overlaid (c, g). Light transmission images of the cells are also shown (d, h, j). Scale bar = 10 μm.

The results obtained with onion epidermal cell transformation were further confirmed by transformation of the gene fusion constructs into A. thaliana root protoplasts. Expression of EYFP–AtECI1 (Figure 6a) and EYFP–AtECI2 (Figure 6c) lead to a punctate pattern of fluorescence as expected for peroxisomal localization and similar to the punctate pattern observed for the glyoxysomal(g)MDH–ECFP control (Figure 6i). However, expression of the EYFP–AtECI3 (Figure 6e) construct lead to a diffuse pattern of fluorescence that was similar to that observed for expression of unmodified EYFP (Figure 6g).

Figure 6.

 Subcellular localization of YFP–AtECI protein fusions within Arabidopsis root cells.
Cells were transformed with plasmids encoding the fusion proteins YFP–AtECI1 (a), YFP–AtECI2 (c), YFP–AtECI3 (e) or ECFP–MDH (i), or with a plasmid encoding unmodified YFP (g). For each image obtained by fluorescence microscopy, the corresponding image obtained by transmission microscopy is shown on the right. Scale bar = 10 μm.

In view of the different localization of AtECI3 in yeast and plants, a further control was performed to ensure that the punctate pattern observed for GFP–AtECI3 in yeast corresponded to the peroxisomes. Wild-type cells were co-transformed with the GFP–AtECI3 construct and plasmid pALDsRedAKL encoding the red fluorescent protein DsRed modified at the C-terminus with the peroxisome targeting signal AKL (Navarro et al., 2004) (Figure 4w–z). The punctate pattern obtained for the peroxisomal DsRed marker co-localized with GFP–AtECI3, demonstrating localization of AtECI3 in the peroxisome in yeast.

The differential targeting of AtECI3 in S. cerevisae and A. thaliana is dependent on PEX5

The targeting of AtECI3 to peroxisomes in yeast but not in onion cells or Arabidopsis indicated differences in the interaction between AtECI3 and the components of the peroxisome import machinery present in yeast and plant cells. PEX5 is an essential protein involved in import of proteins into peroxisomes via binding to PTS1 (Stanley and Wilmanns, 2006). Localization of the AtECI proteins in yeast was thus examined in strains expressing A. thaliana PEX5 instead of the endogenous S. cerevisiae PEX5. While expression of A. thaliana PEX5 in the yeast pex5Δ strain was sufficient to target both AtECI1 and AtECI2 to the peroxisomes (Figure 4e,k), it failed to target AtECI3 to the organelle (Figure 4q).

Δ32-enoyl CoA isomerase activity of the AtECI proteins in yeast

The in vivoΔ32-enoyl CoA isomerase activity of the proteins encoded by AtECI1, AtECI2 and AtECI3 was tested in vivo by complementation of the S. cerevisiae double mutant eci1Δdci1Δ, with deletions in the ECI1 and DCI1 genes, encoding Δ32-enoyl CoA isomerase and Δ3,52,4-dienoyl CoA isomerase, respectively. Use of the double mutant eci1Δdci1Δ was favored over the single mutant eci1Δ as yeast Δ3,52,4-dienoyl CoA isomerase has been shown to have low but significant Δ32-enoyl CoA isomerase activity (Gurvitz et al., 1999). Complementation was determined by synthesis of polyhydroxyalkanoate (PHA) in the peroxisome through polymerization of 3-hydroxyacyl CoA intermediates generated by β-oxidation of fatty acids via the activity of a PHA synthase from Pseudomonas aeruginosa modified for peroxisomal targeting (Mittendorf et al., 1998; Poirier et al., 2001).

The Δ32-enoyl CoA isomerase activity was monitored in S. cerevisiae synthesizing peroxisomal PHA by β-oxidation of the fatty acid 10Z-heptadecenoic acid. The pathway for degradation of 10Z-heptadecenoic acid is shown in Figure 7(b). The 3-hydroxyacyl CoA intermediates that can be polymerized by the PHA synthase are underlined and are identified by the prefix H, followed by the number of carbons in the chain and the number of unsaturated bonds. Degradation of 10Z-heptadecenoic acid generates the intermediate 2E,4Z-undecadienoyl CoA, which is converted to 3E-undecenoyl CoA by 2,4-dienoyl CoA reductase. Further metabolism of this intermediate requires Δ32-enoyl CoA isomerase to generate 2E-undecenoyl CoA. The complete degradation of 10Z-heptadecenoic acid thus yields a PHA containing 3-hydroxytridecenoic acid (H13:1), 3-hydroxyundecanoic acid (H11:0), 3-hydroxynonanoic acid (H9:0) and 3-hydroxyheptanoic acid (H7:0) (Figure 7b). In the absence of Δ32-enoyl CoA isomerase, complete β-oxidation of 10Z-heptadecenoic acid is blocked. Degradation of the saturated fatty acid tridecanoic acid does not require the participation of any auxiliary enzymes and generates a PHA containing the monomers H13:0, H11:0, H9:0, H7:0 and H5:0 (Figure 7a).

Figure 7.

 Degradation pathways of fatty acids in yeast and generation of PHA precursors.
Degradation pathways of tridecanoic acid (a), 10Z-heptadecenoic acid (b) and 10Z,13Z-nonadecadienoic acid (c). Acyl CoA intermediates generated by the degradation of fatty acids are indicated in italics, with the letter ‘C’ indicating the CoA ester that is the substrate for acyl CoA oxidase and the letter ‘H’ indicating the 3-hydroxyacyl CoA intermediate. The first and second numbers following ‘C’ or ‘H’ refer to the number of carbons and double bonds in the acyl CoAs, respectively. For unsaturated acyl CoA intermediates, the location and nature (Z or E) of the double bond is indicated by a prefix. For instance, 10Z-C17:1 stands for 10Z-heptadecenoyl CoA, and 6Z-H13:1 stands for 6Z-3-hydroxytridecenoyl CoA. 3-hydroxyacyl CoA precursors that can be incorporated into PHA are underlined. Core β-oxidation enzymes are indicated in bold: (1) acyl CoA oxidase (FOX1p), (2) multi-functional protein (FOX2p) with 2-enoyl CoA hydratase activity (2a) and 3-hydroxyacyl CoA dehydrogenase activity (2b); (3) ketoacyl CoA thiolase. The order of the numbers indicates the sequence of the reactions. 2× and 3× indicate two or three rounds through the β-oxidation cycle. The auxiliary enzymes involved are 2,4-dienoyl CoA reductase (SPS19p), Δ3,52,4-dienoyl CoA isomerase (DCI1p) and Δ32-enoyl CoA isomerase (ECI1p).

The AtECI1, AtECI2 and AtECI3 genes, as well as the S. cerevisiae ECI1 gene, were cloned into an expression vector and transformed into the eci1Δdci1Δ strain expressing a peroxisomal PHA synthase. The amounts of PHA produced from degradation of tridecanoic acid in eci1Δdci1Δ transformed with an empty control vector or the AtECI1, AtECI2, AtECI3 and ScECI1 genes were all comparable to that in wild-type cells (Figure 8b). In contrast, when 10Z-heptadecenoic acid was used as a carbon source, the amount of PHA synthesized in eci1Δdci1Δ was only 3% of the amount synthesized in wild-type cells or eci1Δdci1Δ transformed with yeast ScECI1 (Figure 8a) . Expression of the AtECI1, AtECI2 and AtECI3 genes in eci1Δdci1Δ cells led to synthesis of PHA to levels that were 13, 130 and 55%, respectively, of the wild-type levels (Figure 8a). Similar results were obtained when cells were grown in medium containing 10Z-pentadecenoic acid (data not shown). Together, these results indicate that the three AtECI proteins have Δ32-enoyl CoA isomerase activity towards both 3E-undecenoyl CoA and 3E-nonenoyl CoA.

Figure 8.

 Restoration of the Δ32-enoyl CoA isomerase-dependent degradation of unsaturated acids in yeast by the AtECI proteins.
PHA produced in wild-type or the double mutant eci1Δdci1Δ transformed either with an empty vector, or a vector containing either the ScECI1, AtECI1, AtECI2 or AtECI3 genes, were analyzed for cells grown in media containing either 0.05% 10Z-heptadecenoic acid (a) or 0.05% tridecanoic acid (b). The distribution of the odd-chain monomers present in the PHA is indicated by histograms, and the total quantity of odd-chain monomers recovered is indicated at the top of each panel (= 3). Statistical analysis revealed that the PHA content in cells containing either ScECI1, AtECI1, AtECI2 or AtECI3 genes was significantly higher than that in the eci1Δdci1Δ control cells transformed with an empty vector only for cells grown in 10Z-heptadecenoic acid (a) (Student’st test, P < 0.05).

Although the AtECI proteins show no significant homology to the A. thalianaΔ3,52,4-dienoyl CoA isomerase encoded by AtDCI1 when whole proteins are considered, the AtECI1 proteins contain a hydratase/isomerase domain and share a glutamate residue that has been found to be important for the catalytic activity of rat Δ3,52,4-dienoyl CoA isomerase (Figure 2). The potential presence of Δ3,52,4-dienoyl CoA isomerase activity in the AtECI proteins was thus tested in vivo by complementation of the S. cerevisiae double mutant eci1Δdci1Δ grown in the presence of the fatty acid 10Z,13Z-nonadecadienoic acid. The same assay has previously been used to reveal in vivoΔ3,52,4-dienoyl CoA isomerase activity for S. cerevisiae ScDCI1p and Arabidopsis AtDCI1 (Goepfert et al., 2005). The pathway of degradation of 10Z,13Z-nonadecadienoic acid by the peroxisomal β-oxidation pathway is shown in Figure 7(c). The intermediate 2E,5Z-undecadienoyl CoA can be degraded via two main pathways. In one pathway, involving the Δ3,52,4-dienoyl CoA isomerase, the intermediate 2E,5Z-undecadienoyl CoA is converted to the core β-oxidation intermediate 2E-undecenoyl CoA, thus leading to synthesis of the saturated PHA intermediate 3-hydroxyundecanoyl CoA (Figure 7c, left branch). In contrast, direct hydration of the 2E,5Z-undecadienoyl CoA intermediate by the multi-functional enzyme leads to the unsaturated PHA intermediate 5Z,3-hydroxyundecenoyl CoA (Figure 7c, right branch). Thus, the increased abundance of the PHA monomer H11:0 is a reflection of Δ3,52,4-dienoyl CoA isomerase activity (Goepfert et al., 2005). Figure 9 shows the GC/MS profile, indicating the H11:1 and H11:0 monomers present in the PHA synthesized in wild-type cells, or eci1Δdci1Δ transformed either with an empty vector, the Arabidopsis genes AtECI1, AtECI2 or AtECI3, or S. cerevisae ECI1, and grown in medium containing 10Z,13Z-nonadecadienoic acid. The increase in the PHA monomer H11:1 in cells expressing either AtECI1, AtECI2, AtECI3 or ScECI1p is consistent with in vivoΔ32-enoyl CoA isomerase activity of these proteins. In contrast, the lack of significant increase in the abundance of the H11:0 monomer in eci1Δdci1Δ cells expressing AtECI1, AtECI2, AtECI3 or ScECI1p indicated the absence of significant Δ3,52,4-dienoyl CoA isomerase activity of these proteins in vivo.

Figure 9.

 Expression of the AtECI1 proteins in yeast is not associated with Δ3,52,4-dienoyl CoA isomerase activity.
GC/MS (m/z 103) of the monomers 3-hydroxyundecanoic acid (H11:0) and 5Z,3-hydroxyundecenoic acid (H11:1) present in PHA synthesized in wild-type or the double mutant eci1Δdci1Δ transformed either with an empty vector (e.v.), or vectors for expression of ScECI1p, AtECI1, AtECI2 or AtECI3. All cells were grown in medium containing 0.05% 10Z,13Z-nonadecadienoic acid.

In vitro enzymatic activities

AtECI1, AtECI2 and AtECI3 were expressed in Escherichia coli under the control of the T7 promoter, and enzymatic assays were performed by monitoring in vitro the production of 3-keto-hexanoyl CoA from 3E-hexenoyl CoA in the presence of 2-enoyl CoA hydratase (crotonase) and 3S-hydroxyacyl CoA dehydrogenase (see Experimental procedures). Despite trying several distinct solubilization and renaturation protocols, no soluble ECI1 protein was detectable in cells expressing AtECI1. The bacterial extracts of cells expressing AtECI2 or AtECI3 lost activity upon purification or storage, indicating that the proteins produced were unstable. The same behavior has been reported for the S. cerevisiae ECI1p (Gurvitz et al., 1998). The highest activities that could be monitored were obtained with fresh extracts of transformed bacteria (Table 1). Conversion of 3E-hexenoyl CoA to 3-ketohexanoyl CoA was strictly dependent on protein extracts derived from cells expressing either yeast Δ32-enoyl CoA isomerase ScECI1p, AtECI2 or AtECI3. Conversion of 3E-hexenoyl CoA into 3-ketohexanoyl CoA was not recorded in the presence of only crotonase and 3-hydroxyacyl CoA dehydrogenase. No conversion was detectable using protein extracts from cells expressing similar amounts of AtDCI1, which encodes a peroxisomal Δ3,52,4-dienoyl CoA isomerase (Goepfert et al., 2005). Incubation of 2E-hexenoyl CoA or 2E-decenoyl CoA with the bacterial extracts in the presence of 3S-hydroxyacyl CoA dehydrogenase did not generate 3-ketohexanoyl CoA. Together, these data indicate that the AtECI2 and AtECI3 proteins possess Δ32-enoyl CoA isomerase activity but no detectable enoyl CoA hydratase activity.

Table 1. In vitro 3-enoyl CoA isomerase activity of AtECI2 and AtECI3
 Mean specific activity ± SD (μmol min−1 mg−1)
  1. Δ32-enoyl CoA activities of protein extracts from bacteria expressing AtECI2, AtECI3, ScECI1p or AtDCI1. Initial conversion velocities were monitored as described by Palosaari and Hiltunen (1990) using 50 μm 3E-hexenoyl CoA as substrate (= 3).

  2. SD, standard deviation; ND, not detectable.

AtECI20.61 ± 0.13
AtECI30.47 ± 0.01
ScECIp0.91 ± 0.28
AtDCI1ND

Phylogenetic analysis of embryophyte ECI genes

Analysis of plant genomes available at http://www.phytozome.net and in GenBank revealed that plants contain at least two proteins homologous to the A. thaliana ECI proteins, and that the ancestral gene common to all plants was an ECI2-like gene. The five dicotyledonous species analyzed, namely A. thaliana, oilseed rape (Brassica napus), poplar (Populus trichocarpa), soybean (Glycine max) and barrel medic (Medicago truncatula) all contain three ECI homologues. The genomes of the monocotyledonous species rice (Oryza sativa) and sorghum (Sorghum bicolor) contain two and three ECI genes, respectively, and two genes are present in the moss Physcomitrella patens. The phylogenetic tree (Figure 10) of these plant ECIs shows independent duplication of an ancestral ECI2-like gene in monocots, dicots and moss. A second duplication occurred subsequently in sorghum and all the dicotyledonous species analyzed. This second duplication involved the ECI2-like gene for all plants analyzed, except M. truncatula, in which the duplication occurred in an ECI1 homologue. Analysis of the rape and Arabidopsis ECI genes indicates that the duplication event leading to creation of AtECI2 and AtECI3 occurred before divergence of these two species. To identify possible functional changes after duplication, nucleotide substitution rates along all branches were calculated. For all genes analyzed, the Ka/Ks values ranged from 0.1 to 0.55, indicating negative purifying selection, i.e. selective pressure that reduces the number of non-synonymous substitutions retained in the genes compared with synonymous substitutions.

Figure 10.

 Phylogenetic tree of embryophyte ECI genes.
Sequences were aligned using muscle version 3.6 (Edgar, 2004) and the tree was constructed using MrBayes 3.1 (Ronquist and Huelsenbeck, 2003). Gene duplications are marked by a dot at the node. The numbers on the nodes are clade credibility values. The scale bar indicates 0.2 nucleotide substitutions per site. At, Arabidopis thaliana; Bn, Brassica napus; Gm, Glycine max; Mt, Medicago truncatula; Os, Oryza sativa; Pp, Physcomitrella patens; Pt, Populus trichocarpa; Sb, Sorghum bicolor.

A multiple sequence alignment of the ECI proteins from the eight selected plant species revealed greatest divergence at the C-termini (Figure S1). Analysis of the plant ECIs using the PTS1 Predictor program (http://mendel.imp.ac.at/mendeljsp/sat/pts1/PTS1predictor.jsp) revealed that rape BnECI3, barrel medic MtECI1B and rice OsECI2B are not predicted to contain a PTS1 and to be targeted to the peroxisome.

Discussion

The AtECI1, AtECI2 and AtECI3 genes were initially identified as potentially encoding mono-functional Δ32-enoyl CoA isomerases based on their amino acid sequence similarities to yeast and mammalian Δ32-enoyl CoA isomerases and the presence of a conserved glutamate in the hydratase/isomerase domain that has been found to be critical for catalysis in all characterized Δ32-enoyl CoA isomerases (Figure 2). In a study aimed at identifying potential plant peroxisomal proteins through a bioinformatic approach, Reumann (2004) identified the same genes as putative members of the Δ32-enoyl CoA isomerase clade. The three AtECI genes are expressed in a broad range of plant tissues but are particularly enhanced during the first two days of germination. Germination is a period of high β-oxidation requirement in oleaginous plants, such as A. thaliana (Graham and Eastmond, 2002). This expression pattern was also observed for genes encoding the auxiliary enzymes AtDCI1 and AtECH2 (Goepfert and Poirier, 2007; Goepfert et al., 2005).

Evidence that AtECI1, AtECI2 and AtECI3 exhibit Δ32-enoyl CoA isomerase activity was provided through complementation of the eci1Δdci1Δ mutant deficient in the degradation of cis-unsaturated fatty acids, as monitored by the synthesis of PHA. In addition to its role in biotechnology as a biodegradable elastomer, PHA synthesis has also previously been used to characterize the pathway of degradation of unsaturated fatty acids in Arabidopsis (Allenbach and Poirier, 2000) and yeast (Bogdawa et al., 2005; Robert et al., 2005). Expression of either AtECI1, AtECI2 or AtECI3 in the eci1Δdci1Δ mutant led to degradation of either 10Z-heptadecenoic acid or 10Z-pentadecenoic acid, fatty acids that require the action of a Δ32-enoyl CoA isomerase on the intermediates 3E-undecenoyl CoA or 3E-nonenoyl CoA, respectively, for complete fatty acid degradation and PHA synthesis. Similarly, complete degradation of the fatty acid 10Z,13Z-nonadecadienoic acid and generation of the intermediate 3-hydroxyundecanoyl CoA in the eci1Δdci1Δ mutant expressing either AtECI1, AtECI2 or AtECI3 further demonstrated the Δ32-enoyl CoA isomerase activity of these proteins on 3E,7Z-tridecadienoyl CoA.

In vitro enzyme assays confirmed the Δ32-enoyl CoA isomerase activity of AtECI2 and AtECI3 on 3E-hexenoyl CoA. Furthermore, no in vitro enoyl CoA hydratase activity was detectable for AtECI2 or AtECI3 on either 2E-hexenoyl CoA or 2E-decenoyl CoA. The absence of hydratase activity is in agreement with the absence of a second glutamate residue in the active sites of mono-functional enoyl CoA hydratases, such as crotonase, or in the enoyl CoA hydratase domain of multi-functional proteins (Kiema et al., 1999). In vitro activity measurements for either Δ32-enoyl CoA isomerase or enoyl CoA hydratase activity could not be performed for AtECI1 due to failure to solubilize the protein expressed in E. coli. However, based on the sequence similarity between AtECI1 and AtECI2/AtECI3, including the presence of the glutamate residue conserved in Δ32-enoyl CoA isomerases and the absence of the second conserved glutamate found in enoyl CoA hydratases, it is likely that AtECI1 is also an Δ32-enoyl CoA isomerase.

Together, these data show that AtECI1, AtECI2 and AtECI3 possess in vivoΔ32-enoyl CoA isomerase activity towards enoyl CoAs of 9–13 carbons. The in vitro assay of AtECI2 and AtECI3 further extends this range to include a six-carbon substrate. Furthermore, the failure to detect an enoyl CoA hydratase activity for AtECI2 and AtECI3 in vitro or Δ3,52,4-dienoyl CoA isomerase activity for all three Arabidopsis in vivo leads to the conclusion that AtECI1, AtECI2 and AtECI3 are Δ32-enoyl CoA isomerases.

For all fatty acids utilized in the in vivo complementation assay, expression of AtECI1 led to a substantially lower level of complementation compared to AtECI2, AtECI3 or ScECI1p. Although the reasons behind this lower apparent Δ32-enoyl CoA isomerase activity of AtECI1 are currently unknown, the fact that AtECI1 expressed in E. coli could not be found in a solubilized form suggest that perhaps the folding or stability of AtECI1 is sub-optimal not only in vitro but also in the yeast peroxisome.

In S. cerevisae and plants, β-oxidation occurs in the peroxisomes, and proteins involved in this pathway must be targeted to this organelle. AtECI1 and AtECI2 possess an SKL and a PKL C-terminal tripeptide, respectively, both of which are well-characterized PTS1s enabling targeting of proteins to the peroxisomes. AtECI1 and AtECI2 were targeted to the peroxisomes in both onion epidermal cells and Arabidopsis root protoplast. The presence of AtECI1 and AtECI2 in the peroxisomal proteome and fluorescent fusion protein localization (Reumann et al., 2007) have confirmed their peroxisomal localization. When expressed in S. cerevisae, both AtECI1 and AtECI2 were localized in the peroxisome, as shown by GFP fusion protein localization and the fact that these proteins complement the yeast Δeci1Δdci1 mutant, in agreement with the high degree of conservation in the mechanism of import of proteins into peroxisomes across plants, animals and fungi.

More surprising were the results obtained for the subcellular localization of AtECI3. This protein ends with the terminal tripeptide HNL. To our knowledge, this tripeptide has not been identified in peroxisomal proteins and has not been found to act as a PTS1. However, HNL resembles the terminal tripeptide HRL that was found to act as a PTS1 in the S. cerevisiae peroxisomal Δ32-enoyl CoA isomerase (Yang et al., 2001). Furthermore, the terminal tripeptide ANL in chloramphenicol acetyltransferase has been found to act as a PTS1 in plants (Mullen et al., 1997), and terminal tripeptides ANL or SNL are found in several peroxisomal proteins, such as mouse α-methylacyl CoA racemase (ANL; Genbank accession number O09174) and human d-aspartate oxidase (Q99489) and Dictyostelium discoideum catalase (SNL; Genbank accession number O77229). Both the localization of GFP–AtECI3 and the complementation of the Δeci1Δdci1 mutant showed that AtECI3 was localized to the peroxisome in S. cerevisiae. In contrast, the fusion protein EYFP–AtECI3 was localized to the cytoplasm in onion epidermal cells and Arabidopsis root cells, the latter being a tissue where all three AtECI genes are naturally expressed. Furthermore, although analysis of the Arabidopsis peroxisome proteome identified AtECI1 and AtECI2 as peroxisomal enzymes, AtECI3 was not identified as such in the same study (Reumann et al., 2007) or in a recent extension of the peroxisome proteome database (S. Reumann, University of Stavanger, Norway, personal communication). Altogether these data indicate that although AtECI3 is not a peroxisomal protein in plants, it can be targeted to this organelle in S. cerevisiae through a signal sequence that remains to be identified but could potentially be the terminal HNL.

Although the broad mechanisms of protein import into the peroxisome are well conserved between animals, plants and fungi, recognition and import into the peroxisomes of proteins carrying a variant of the canonical PTS1 is species-specific (Brocard and Hartig, 2006). An important factor in this species-specific recognition is the interaction between the PTS1 and the tetratricopeptide repeat domain found within the C-terminal half of the PTS1 receptor PEX5. Furthermore, successful recognition of the PTS1 by the PEX5 receptor and subsequent import of the protein into the peroxisome is strongly influenced by the nature of the amino acids adjacent to the C-terminal tripeptide. Thus, a complete PTS1 includes 12 amino acids divided into three regions: the most C-terminal tripeptide is responsible for interaction with the PEX5 cavity, a tetrapeptide immediately upstream is thought to interact with the surface of the PEX5 protein, and a flexible hinge allows accessibility and flexibility (Brocard and Hartig, 2006). The observation that AtECI3 was targeted to peroxisomes in yeast containing native S. cerevisae PEX5 but remained cytosolic in a pex5Δ mutant expressing Arabidopsis PEX5 indicates that, while there is effective interaction between AtECI3 and the yeast PEX5, no such interaction occurs between AtECI3 and A. thaliana PEX5. Conversely, the improved targeting of AtECI2 in yeast expressing Arabidopsis PEX5 compared to yeast PEX5 (compare Figure 4g and k) probably reflects a more effective interaction between AtECI2 and AtPEX5 compared to ScPEX5.

Phylogenetic analysis revealed the presence of numerous independent duplications of the ECI genes throughout plant evolution, and indicated that non-peroxisomal ECI homologues have probably arisen independently in rice, Medicago truncatula and the common ancestor of Arabidopsis and rape. In contrast, analysis of the genomes of human, mouse and rat on one hand, and of several fungi on the other, failed to identify proteins with homology to the respective mammalian or fungal Δ32-enoyl CoA isomerases that do not have a clear PTS1 (data not shown). The evolution of cytosolic Δ32-enoyl CoA isomerases thus appears to be restricted to the plant kingdom.

Gene duplication is considered to be the most important evolutionary process for generating novel functions. Accumulating evidence supports the notion that changing the subcellular location of a protein can also alter its function and that protein subcellular re-localization might be an important evolutionary mechanism for the retention of gene duplicates (Byun-McKay and Geeta, 2007). AtECI3 and the genes from rice, rape and Medicago truncatula encoding putative non-peroxisomal Δ32-enoyl CoA isomerases are under purifying selection, indicating that these genes have retained a function. In this context, it is interesting to speculate what the biological function of AtECI3-like proteins could be. It is possible that AtECI3 may still function in the cis/trans isomerization of double bonds, but on substrates present in the cytoplasm that are distinct from those present in the peroxisome. This is reminiscent of the demonstration that the regio-specificity of the Arabidopsis plastidial fatty acid desaturase FAD5 and its cytosolic homologues ADS1 and ADS2 for the Δ7 or Δ9 positions is dependent on the subcellular localization of the enzymes rather than their intrinsic enzymatic activity (Heilmann et al., 2004). However, given the high diversity in the reactions that are catalyzed by proteins within the hydratase/isomerase superfamily, ranging from isomerization to dehalogenation, hydration and carboxylation, it is possible that AtECI3 may also have other catalytic activities on other substrates.

Experimental procedures

Sequence analysis

Searches against the Arabidopsis sequence database were performed using the blastp algorithm available on the TAIR database (http://www.arabidopsis.org/Blast/). Alignment of the Arabidopsis ECI proteins and of the active sites among plant, yeast and mammalian Δ32-enoyl CoA isomerases was performed using the Clustal w2 program (http://www.ebi.ac.uk/Tools/clustalw2/index.html).

For construction of the ECI phylogenetic tree, embryophyte ECI genes were identified by a blast search at http://www.phytozome.net or in GenBank. Amino acid sequences were aligned using muscle (Edgar, 2004). Gaps from the protein alignment were introduced into the DNA sequences, and a phylogenetic tree was constructed using MrBayes (750 000 generations, sample frequency 100, burnin 2500, model GTR + γ + inv) (Ronquist and Huelsenbeck, 2003). Nucleotide substitution rates were calculated using PAML (Yang, 1997) and a method described by Liberles (2001).

Bacterial strains

Escherichia coli DH5α was used to maintain and propagate all plasmids according to standard procedures, with the exception of the Gateway® native plasmids pDEST™17 and pDONR™207, which were maintained in E. coli DB3.1™ (Invitrogen, http://www.invitrogen.com/). All Gateway® technology-related procedures were performed according to the manufacturer’s instructions. For protein expression, E. coli strain BL21 DE3 pLysS was used.

Yeast strains

Wild-type S. cerevisiae strain BY4742 (matα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) and the isogenic pex5Δ0 mutant (YDR244W::kanMX4) were obtained from EUROSCARF (http://www.uni-frankfurt.de/fb15/mikro/euroscarf/index.html). The isogenic strain eci1Δdci1Δ (eci1Δ::LEU2; dci1::kanMX4) was derived from eci1Δ as previously described (Robert et al., 2005). Plasmids were transformed into S. cerevisiae strains by the lithium acetate procedure (Gietz et al., 1992). Yeast strains were routinely propagated on selective medium comprising 0.67% yeast nitrogen base without amino acids (Difco, http://www.bd.com), 0.5% ammonium sulfate, 2% glucose and appropriate drop-out supplement (Clontech, http://www.clontech.com/).

Plant culture

Seeds of A. thaliana, accession Columbia (Col-0), were sterilized and plated on medium containing half-strength Murashige and Skoog (MS) salts, 1% sucrose and 0.8% agar. Plates were placed at 4°C for 48 h before placing them under constant illumination (70 μE m−2) at 21°C (defined as day 0 after imbibition; DAI). At 15 DAI, plants were transplanted into soil under the same light and temperature conditions. Whole plants were collected from plates between 0 and 15 DAI. Leaves, stems and flower buds were collected from plants grown in pots at 40 DAI. Roots for RNA extractions were harvested at 15 DAI from plants grown in liquid half-strength MS medium supplemented with 2% sucrose. Roots used for protoplast isolation were harvested at 15 DAI from plants grown vertically on solid medium containing half-strength Murashige and Skoog salts, 1% sucrose and 0.8% agar.

Yeast expression vectors

AtECI1, AtECI2 and AtECI3 coding sequences were PCR-amplified from A. thaliana Col-0 genomic DNA extracted with a Qiagen plant DNA mini kit (http://www.qiagen.com/) according to the manufacturer’s instructions. The PCR was performed using Expand HiFi polymerase (Roche, http://www.roche-applied-science.com) according to the manufacturer’s instructions using primers AtECI1-Fw (5′-GCAGGCTCGGCCGGATCCAATCAGTCACAAAGAGGAAGAGA-3′), AtECI1-Rev (5′-GCTGGGTCTAGAATTCAACTGACAAGTAGCCCAAGAA-3′), AtECI2-Fw (5′-CTTCTAGATCCTCCGATCACCAC-3′), AtECI2-Rev (5′-CATTGAATTCTCTTTCTCCC-3′), AtECI3-Fw (5′-CTTTCACTAGTTATTCTTGTACCATC-3′) and AtECI3-Rev (5′-AATGGAATTCGTTGACTTTTGTCG-3′). PCR products were digested with BamHI/EcoRI, XbaI/EcoRI or SpeI/EcoRI, respectively, and ligated into p413GPD adequately digested and dephosphorylated to generate plasmids p413-AtECI1, p413-AtECI2 and p413-AtECI3. p413-ScECI1 was created by insertion of the EcoRI–BamHI fragment from p423GPD-ScECI1 (Bogdawa et al., 2005) in p413GPD, placing ScECI1 under the control of the constitutive glyceraldehyde-3-phosphate dehydrogenase (GPD) promoter and the Cyc1 terminator (Mumberg et al., 1995). The cDNA from AtPEX5 (obtained from Alison Baker, University of Leeds, UK) was subcloned into the SpeI–BamHI sites of plasmid p415-GPD (Mumberg et al., 1995).

GFP fusion constructs

The PCR fragments described above for the yeast expression vectors were digested with BamHI and EcoRI and cloned into pUG36 adequately linearized to give pUG36-ECI1, pUG36-ECI2 and pUG36-ECI3. pUG36 (http://mips.gsf.de/proj/yeast/info/tools/hegemann/gfp.html) carries the enhanced green fluorescent protein (EGFP) (Cormack et al., 1997) under the control of the Met25 promoter and the Cyc1 terminator.

YFP fusion constructs

Plasmids pCAT-ECFP-MDH and pCAT-EYFP-Not have been described previously (Fulda et al., 2002). The DNA fragments amplified by Expand HiFi polymerase (Roche) from Arabidopsis genomic DNA using the primers ECI1-Not (5′-AGCGGCCGCTAATCAGTCACAAAGAGGAAGAGA-3′), Xba-Eco-ECI1 (5′-GCTGGGTCTAGAATTCAACTGACAAGTAGCCCAAGAA-3′), ECI2-Not (5′-AGCGGCCGCGATCACCACCATGTGTACG-3′), ECI2-Xba (5′-GCTTTCTCTAGAGCTTAGGTGTTGC-3′), ECI3-Not (5′-AGCGGCCGCAGTCCTTAACGTACGCACCA-3′) and ECI3-Xba (5′-CCCAACAAGACAATGAGAGTTTC-3′) were digested with NotI and XbaI and inserted into pCAT-EYFP-Not digested with NotI and XbaI, resulting in in-frame fusion of EYFP with the AtECI genes.

Production of the recombinant proteins

Bacteria carrying pDEST-ECIs plasmids were grown to saturation in selective LB medium at 37°C. This starter culture was used to initiate a selective NCYZM culture. Cells were grown at 29°C to an attenuance at 600 nm of approximately 0.6, induced with 1 mm isopropyl-1-thio-β-d-galactopyranoside (IPTG) and grown for an additional 4–8 h. Cells were collected by centrifugation for 5 min at 10 000 g and at 4°C and resuspended in 50 mm sodium phosphate buffer, pH 8.0, 0.3 m NaCl, 20% v/v glycerol, 1 mm phenylmethylsulfonyl fluoride (PMSF). Cells were lysed by four cycles of 10 sec sonication followed by 30 sec on ice. Protein concentration was measured using a Bio-Rad protein assay kit (http://www.bio-rad.com/) according to the manufacturer’s instruction.

Enzymatic assays

Enzymatic assays were performed as described previously (Palosaari and Hiltunen, 1990). Absorbance was measured using an Ultrospec 2000 (Amersham Biosciences, http://www5.amershambiosciences.com/). Isomerization of 3E-enoyl CoA to 2E-enoyl CoA can be followed by monitoring the production of 3-ketoacyl CoA by incubation of 3E-enoyl CoA with the protein extract in the presence of saturating quantities of 2E-enoyl CoA hydratase (crotonase) and 3S-hydroxyacyl CoA dehydrogenase. The 3-ketoacyl CoA produced, when coordinated with Mg2+ ions, gives a characteristic absorption peak at 303 nm. Likewise, 2E-enoyl CoA hydratase activity can be followed by the appearance of 3-ketoacyl CoA as a result of incubation of 2E-enoyl CoA with the protein extract in the presence of 3S-hydroxyacyl CoA dehydrogenase. The expression upon induction and the correct size of the chimeric proteins were verified by Western blot using anti-pentahistidine antibodies.

PHA production

PHA-synthesizing strains were generated using the yeast shuttle plasmid Yiplac128-PHA containing the PHAC1 synthase from P. aeruginosa modified for peroxisomal targeting by the addition, at the C-terminus of the protein, of the last 34 amino acids of Brassica napus isocitrate lyase (Mittendorf et al., 1998). For experiments analyzing PHA synthesis, a stationary phase culture of cells grown in selective medium containing 2% glucose was harvested by centrifugation at 4°C and 5000 g for 5 min, cells were washed once in water and resuspended at a 1:10 dilution in fresh selective medium containing 0.1% w/v glucose, 2% w/v Pluronic-127 (Sigma, http://www.sigmaaldrich.com/) and 0.1–0.05% v/v fatty acid. Cells were grown for an additional 3–4 days before harvesting for PHA analysis as previously described (Poirier et al., 2001). Fatty acids were purchased from Nu-Chek-Prep (http://www.nu-chekprep.com).

Expression of fluorescent fusion proteins in yeast and plant

Transient transformation of onion cells was performed essentially as previously described (Goepfert et al., 2005). Preparation of Arabidopsis root protoplasts was adapted from the method described by Birnbaum et al. (2003). Harvested roots were cut and incubated for 3 h at 29°C in a digestion solution containing 442 mm mannitol, 10 mm CaCl2, 10 mm KCl, 2 mm 2-morpholinoethane sulfonic acid pH 5.7, 1.5% cellulase (Onozuka, http://www.yakult.co.jp), 0.1% pectoylase (Sigma), 1% cellulase (Calbiochem, http://www.merkbiosciences.com) and 0.1% BSA. The preparation was filtered through a 50 μm cloth, and protoplasts were pelleted by centrifugation for 5 min at 500 g at room temperature and resuspended in 455 mm mannitol, 15 mm MgCl2. Protoplast were transformed as described by Yoo et al. (2007), and incubated overnight in the dark at 22°C before observation.

Saccharomyces cerevisiae cells transformed with fluorescent protein constructs were first grown overnight in selective medium containing 2% w/v glucose. Cells were harvested by centrifugation for 5 min at 5000 g at room temperature, washed once in water and resuspended at a 1:10 dilution in fresh selective medium containing 0.1% w/v glucose, 2% w/v Pluronic-127 (Sigma) and 0.1% v/v oleic acid. Cells were grown for 18 h in oleic acid-containing medium before observation by microscopy. For co-localization of GFP–AtECI3 to the peroxisomes, yeast were co-transformed with both GFP–AtECI3 and plasmid pALDsRedAKL encoding the red fluorescent protein DsRed fused to the peroxisome targeting signal AKL (Navarro et al., 2004).

Gene expression analysis

Total RNA was extracted using a hot borate protocol for seedlings from 0 to 5 DAI and the LiCl protocol for all other samples. Northern blots were prepared using 20 μg of total RNA. A [32P]dCTP-labeled probe from full-length cDNAs was produced by random priming with the Prime-a-gene® labeling system (Promega, http://www.promega.com/), and purified using a ProbeQant® kit (Amersham Biosciences). Pre-hybridization and hybridization were performed in 25 mm sodium phosphate buffer pH 7.5, 7% SDS, 1% BSA, 0.372 g l−1 EDTA at 65°C. Washes were performed at 65°C: twice for 10 min with 2 × SSC, 0.1% SDS, then twice for 10 min with 1× SSC, 0.1% SDS. Under these conditions, no cross-hybridization of the AtECI1, AtECI2 and AtECI3 genes was observed.

Acknowledgements

This work was funded by grants from the Ford National Swisse (grant number 3100A0-105874 to Y.P.), by grants from the Academy of Finland and the Sigrid Jusélius Foundation (to J.K.H.), and by a grant from the Idea Network for Biomedical Research Excellence (to C.T.R.). The authors are grateful to Pasqualina Magliano and Alison Baker (University of Leeds, UK) for help with the AtPEX5 constructs, Bulak Arpat and Arnaud Paradis for help with microscopy, and Marcello Russo (CNR Institute of Plant Virology, Bari, Italy) for providing the plasmid pALDsRedAKL. The authors are also grateful to Henrik Kaessmann for raising our awareness of the evolutionary implications of AtECI3.

Ancillary