A plant thiolase involved in benzoic acid biosynthesis and volatile benzenoid production

Authors

  • Alex Van Moerkercke,

    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands, and
    Search for more papers by this author
  • Ines Schauvinhold,

    1. Department of Molecular, Cellular and Developmental Biology, University of Michigan, Ann Arbor, MI 48109, USA
    Search for more papers by this author
  • Eran Pichersky,

    1. Department of Molecular, Cellular and Developmental Biology, University of Michigan, Ann Arbor, MI 48109, USA
    Search for more papers by this author
  • Michel A. Haring,

    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands, and
    Search for more papers by this author
  • Robert C. Schuurink

    Corresponding author
    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands, and
    Search for more papers by this author

*For correspondence (fax +31 205 257934; e-mail r.c.schuurink@uva.nl).

Summary

The exact biosynthetic pathways leading to benzoic acid (BA) formation in plants are not known, but labeling experiments indicate the contribution of both β-oxidative and non-β-oxidative pathways. In Petunia hybrida BA is a key precursor for the production of volatile benzenoids by its flowers. Using functional genomics, we identified a 3-ketoacyl-CoA thiolase, PhKAT1, which is involved in the benzenoid biosynthetic pathway and the production of BA. PhKAT1 is localised in the peroxisomes, where it is important for the formation of benzoyl-CoA-related compounds. Silencing of PhKAT1 resulted in a major reduction in BA and benzenoid formation, leaving the production of other phenylpropanoid-related volatiles unaffected. During the night, when volatile benzenoid production is highest, it is largely the β-oxidative pathway that contributes to the formation of BA and benzenoids. Our studies add the benzenoid biosynthetic pathway to the list of pathways in which 3-ketoacyl-CoA thiolases are involved in plants.

Introduction

In many plants, floral volatiles play a role in attracting pollinators (Stuurman et al., 2004). The production and emission of these volatiles are therefore tightly regulated, not only spatially and diurnally but also developmentally. Flowers of petunia (Petunia hybrida cv. Mitchell) emit benzenoid and phenylpropanoid-related volatiles (Verdonk et al., 2003), which are derived from the common precursor Phe (Boatright et al., 2004). Petunia hybrida cv. Mitchell has emerged as a model of choice to study the biosynthesis, emission and regulation of these classes of volatiles.

Based on their carbon skeleton, Phe-derived volatiles are categorised as C6–C1, C6–C2 and C6–C3 compounds. For the C6–C1 compounds, the benzenoids, the propyl side chain of Phe needs to be shortened by two carbon atoms. It has been reported that this shortening can occur either via a β-oxidative (Ribnicky et al., 1998; Jarvis et al., 2000; Hertweck et al., 2001) or a non-β-oxidative pathway with benzoyl-CoA and benzaldehyde as precursors for benzoic acid (BA), respectively (Figure 1). Whereas the β-oxidative pathway needs activation by coenzyme A (CoA), the non-β-oxidative pathway occurs both CoA dependently (Abd El-Mawla and Beerhues, 2002; Boatright et al., 2004) and CoA independently (Schnitzler et al., 1992; Boatright et al., 2004).

Figure 1.

 Schematic representation of the biosynthetic pathways leading to C6–C1, C6–C2 and C6–C3 compounds in petunia petals and the proposed biochemical reaction catalysed by PhKAT1.
The common precursor for all compounds is L-Phe. The C6–C1 compounds are derived from t-cinnamic acid and the propyl side chain can be shortened by two carbon atoms via either a non-β-oxidative or a β-oxidative pathway, with benzaldehyde and benzoyl-CoA as intermediates for benzoic acid (BA) production, respectively. The β-oxidative pathway is dependent on CoA activation of t-cinnamic acid and requires the action of a 3-ketoacyl-CoA thiolase (PhKAT1) for the production of benzoyl-CoA. Benzoyl-CoA is subsequently used to produce benzylbenzoate and phenylethylbenzoate using benzylalcohol and 2-phenylethanol as co-substrates, respectively. Benzoyl-CoA can also be converted to BA directly by the action of a thioesterase or indirectly via benzylbenzoate (not depicted; Boatright et al., 2004). Alternatively, BA can be produced via oxidation of benzaldehyde. Dashed arrows indicate unknown genes in petunia. See text for details. Abbreviations: BPBT, benzoyl-CoA:benzyl alcohol/phenylethanol benzoyltransferase; BSMT, benzoic acid/salicylic acid methyl transferase; CFAT, acetyl-CoA:coniferyl alcohol acetyltransferase; IGS, isoeugenol synthase; KAT1, 3-ketoacyl-CoA thiolase 1; PAAS, phenylacetaldehyde synthase; PAL, phenylalanine ammonia lyase.

Despite the importance of the C6–C1 compound BA in plant secondary metabolism (Wildermuth, 2006), only a few enzymatic steps leading to its synthesis have been characterised, whereas the corresponding genes remain elusive. For the non-β-oxidative CoA-dependent pathway the enzymatic activities of cinnamate:CoA ligase, cinnamoyl:CoA hydratase/lyase and benzaldehyde dehydrogenase, whose successive actions starting from t-cinnamic acid lead to the formation of BA, have been detected in Hypericum androsaemum L. cell cultures (Abd El-Mawla and Beerhues, 2002). Recently, the involvement of a snapdragon benzaldehyde dehydrogenase in the production of BA was shown by transient expression in petunia flowers (Long et al., 2009). Enzymatic activities for the other two pathways have not been described yet, although many labelling studies with precursors have suggested their presence in plants (Ribnicky et al., 1998; Boatright et al., 2004; Orlova et al., 2006).

In petunia, BA is a key intermediate in volatile benzenoid production. It serves as an immediate precursor for benzoic acid/salicylic acid methyl transferase (PhBSMT) leading to the production of methylbenzoate (MeBA) (Negre et al., 2003), the most abundant volatile emitted by petunia flowers (Kolosova et al., 2001a; Verdonk et al., 2005). BA is also a precursor of salicylic acid (SA), which can be methylated by PhBSMT to methylsalicylate (MeSA), one of the minor components in the floral headspace, but SA might also be formed from (iso)chorismate as in Arabidopsis (Wildermuth, 2001). In addition, BA can be converted to benzoyl-CoA, one of the substrates of benzoyl-CoA:benzyl alcohol/phenylethanol benzoyltransferase (PhBPBT), which in combination with benzylalcohol or 2-phenylethanol produces benzylbenzoate and phenylethylbenzoate in petunia, respectively. Although a benzoyl-CoA ligase has not yet been identified in petunia, its activity has been detected in Clarkia breweri flowers (Beuerle and Pichersky, 2002). Labelling studies in petunia showed that only the non-β-oxidative CoA-independent and β-oxidative CoA-dependent pathways are active (Boatright et al., 2004) and that the flux through these pathways differs in the light and dark period (Orlova et al., 2006). Furthermore, in vivo stable isotope labelling indicated that benzylbenzoate is also an intermediate between Phe and BA (Boatright et al., 2004).

In this study, we describe the identification and characterization of a petunia 3-ketoacyl-CoA thiolase (PhKAT1), which plays an important role in the β-oxidative pathway leading to the production of benzoyl-CoA from 3-oxo-3-phenylpropionyl-CoA (benzoylacetyl-CoA). To determine the role of PhKAT1 in floral volatile benzenoid production, we generated transgenic plants suppressing PhKAT1 expression (ir-PhKAT1). Analysis of these plants indicate that the β-oxidative pathway plays a major role in the production of BA and thus of precursors for C6–C1 volatiles. This 3-ketoacyl-CoA thiolase is present in the peroxisomes of petunia petals, adding the involvement of this organelle in volatile benzenoid biosynthesis.

Results

PhKAT1 expression is characteristic for a floral benzenoid-related gene

In order to elucidate steps in the β-oxidative pathway, we used our microarray data (Verdonk et al., 2005, 2006) to select genes whose expression correlated strongly with genes involved in floral scent production. One of these had high similarity (88%) to Arabidopsis thaliana 3-ketoacyl-CoA thiolase 5 (AT5G48880) and was tentatively named PhKAT1. Microarray analysis showed that expression of PhKAT1 was 3.6-fold (= 0.0003) higher in petals when scent emission started, 3 h before the dark period, than 9 h before the dark period, when scent emission was very low.

Figure 2(a) shows that PhKAT1 is predominantly expressed in petal limbs and petal tubes and that expression is hardly detectable in other tissues, including whole seedlings. PhKAT1 expression shows a diurnal rhythm with a peak before the dark period, which precedes the peak of ODORANT1 (PhODO1; Verdonk et al., 2005) expression by approximately 3–6 h (Figure 2b). This finding is consistent with the nocturnal emission of volatile compounds and rhythmic production of its precursors (Kolosova et al., 2001a; Verdonk et al., 2003). At the onset of volatile emission, PhKAT1 expression is highest when the flower is fully open after the anthers have dehisced (Figure 2c), correlating with the developmental pattern of volatile emission in petunia flowers (Verdonk et al., 2003).

Figure 2.

 Petal-specific, rhythmic and developmentally regulated expression of PhKAT1.
(a) Tissue-specific expression. Petunia hybrida cv. Mitchell plants were used to collect material from petal tubes (PT), 2-day-old flower limbs (petal), reproductive organs (f: pistils, m: stamens, m*: stamens with dehisced anthers), sepals, pedicels, stems, leaves, roots and whole seedlings. All tissues were collected 2 h before the onset of the dark period. For RNA gel-blot analysis, 8 μg total RNA was loaded per lane and the blot was hybridised with the PhKAT1 coding region. The blot was re-hybridised with an 18S rDNA probe to illustrate loading of the gels. Two independent experiments were performed and a representative blot is shown.
(b) Rhythmic expression in petal limbs. Petal limbs of 2-day-old flowers were collected with 3-h intervals for a 30-h period. The dark period in the growth chamber started at 12.00 h and ended at 20.00 h. A representative RNA gel-blot is shown as Supplementary Figure S1, but tissue collection and gel-blot analysis were performed twice (means and maximum values are shown). RNA gel-blot analysis was performed as in (a). Blots were hybridised with the PhKAT1 coding region and a PhODO1 3′-fragment. The blots were re-hybridised with PhFBP1 to determine relative mRNA levels. Black bars represent relative PhKAT1 and grey bars relative PhODO1 transcript levels in petunia petals.
(c) Developmentally regulated expression in petal limbs. Wild type flowers at different stages during flower development were used to collect petal tissue 2 h before the onset of the dark period. Two-day-old petal limbs, excluding petal tube tissue, of at least three flowers were pooled per sample. RNA gel-blot analysis was performed as in (a). The blot was re-hybridised with a PhFBP1 to illustrate loading of the gel. The same blot was re-hybridised with a PhKAT2 probe to show its distinct developmental expression pattern from PhKAT1. Letters indicate different developmental stages of the flower (a: 1 cm, b: 3 cm, c: 4 cm, d: flower almost open, e: flower opening (1 cm diameter), f: open flower (day 1), g: open flower with dehisced anthers and h: senescent flower. Two independent experiments were performed and a representative blot is shown. Abbreviations: FBP1: floral binding protein 1; KAT1 and 2: 3-ketoacyl-CoA thiolase 1 and 2; ODO1: ODORANT1.

By blasting the publicly available SGN (Solanaceae Genomics Network, http://sgn.cornell.edu/) EST collection with the full length PhKAT1 cDNA sequence, we identified an EST with high similarity to PhKAT1. We obtained its full-length cDNA sequence (75% identity at the nucleotide level and 78% identity at the amino acid level) and named it PhKAT2. We analysed the expression pattern using RNA-blot analysis and unlike for PhKAT1, transcript levels of PhKAT2 did not change during petal development (Figure 2c) and did not show a rhythmic pattern (Figure S1). Although this finding does not exclude a role for PhKAT2 in benzenoid biosynthesis, it strongly suggests no association with floral scent biosynthesis.

The petunia 3-ketoacyl-CoA thiolase protein is localised to peroxisomes in planta

PhKAT1 is predicted to encode a protein of 462 amino acids containing three conserved thiolase signatures (Figure S2). Typically for thiolases, PhKAT1 contains a peroxisomal targeting signal 2 (Gietl, 1990; Johnson and Olsen, 2003; Reumann, 2004; Carrie et al., 2007) at the N-terminus (Figure S2). To verify whether PhKAT1 is localised to the peroxisomes in planta, we constructed a PhKAT1–GFP fusion protein and expressed it in petunia petals by means of Agrobacterium-mediated transient transformation (Shang et al., 2007). Figure 3 shows that PhKAT1–GFP co-localises with the mCherry-peroxisomal marker px-rk (Nelson et al., 2007), indicating peroxisomal targeting of PhKAT1 in planta, in accordance with the localization of Arabidopsis 3-ketoacyl-CoA thiolases (Carrie et al., 2007).

Figure 3.

In vivo peroxisomal targeting of PhKAT1.
The mCherry peroxisomal marker px-rk (a) and 35S:PhKAT1–GFP (b) were transiently co-expressed in Mitchell flowers. The merged and bright field images are shown in (d) and (c), respectively.
(e, f, g) Magnification taken from the boxed area shown in (d). Confocal images were obtained with a Zeiss LSM 510 confocal laser scanning microscope. Co-localisation was seen several times in independent experiments. The scale bars represent 10 μm.

Silencing of PhKAT1 reduces C6–C1 benzenoid volatile emission

To investigate the role of PhKAT1 in floral scent production, we generated stable transgenic lines suppressing PhKAT1 expression by post-transcriptional gene silencing. For this experiment, an inverted repeat of 174 bp containing parts of the 3′-end of the cDNA (Figure S2) that has low similarity with PhKAT2, was constitutively expressed under control of the CaMV 35S-promoter in P. hybrida cv. Mitchell plants. Initial screening using RNA gel-blot and headspace analyses identified four independent silenced T0 lines with reduced PhKAT1 expression and volatile benzenoid emission (data not shown). Three lines were chosen for further analysis in the T1 generation. In this generation, one line seemed to have lost its phenotype and was discarded for further analysis, resulting in two independent silenced T1 lines (ir-PhKAT1 lines 4 and 7) and one transgenic non-silenced T1 line (ir-PhKAT1 line 6). For both silenced lines, expression levels were ca. 25% of wild type levels (Figure 4a,b). PhKAT2 levels remained unaltered in these lines, confirming gene-specific suppression of PhKAT1 (Figure 4a). Because we hypothesised that PhKAT1 produces benzoyl-CoA from 3-oxo-3-phenylpropionyl-CoA, we initially tried to specifically measure 3-oxo-3-phenylpropionyl-CoA thiolase activity, but attempts to do so were unsuccessful. Most thiolase activities, however, are measured with the general substrate, acetoacetyl-CoA, which we used. Overall thiolase activity in total protein extracts from petals harvested 1 h in the dark period, as measured by the decrease in acetoacetyl-CoA, was 50% lower in the ir-PhKAT1 line 7 (Figure 5), indicating a correlation with lower PhKAT1 transcript levels.

Figure 4.

 RNA gel-blot analysis of wild type and ir-PhKAT1 plants.
(a) PhKAT1 mRNA levels in petals of ir-PhKAT1 T1 and wild type plants. Total RNA was extracted from 2-day-old flowers (three per sample) 30 min before the onset of the dark period. The numbers on top represent independent transgenic T1 lines. A fragment containing the PhKAT1 full-length coding region was used as a probe. The blot was re-hybridised with PhKAT2 showing sequence-specific silencing of PhKAT1. Probing with PhFBP1 enabled quantification of PhKAT1 transcript levels. Representative blots are shown.
(b) Relative PhKAT1 mRNA levels in petals of ir-PhKAT1 and wild type plants (means and maximum levels are shown, = 2). Abbreviations: FBP1: floral binding protein 1; KAT1 and 2: 3-ketoacyl-CoA thiolase 1 and 2.

Figure 5.

 Thiolase activity is reduced in ir-PhKAT1 plants.
Crude protein extracts were prepared from 2-day-old wild-type flowers and flowers of ir-PhKAT1 line 7, harvested 1 h in the dark period. Thiolase activity was determined by measuring the decay of acetoacetyl-CoA. Products were separated and detected on HPLC as described in Burns et al. (2005). The mean values and standard deviations of four independent experiments are shown.

To examine the effect of reduced PhKAT1 expression on volatile emission, these T1 lines were used for headspace collections and subsequent GC-MS analyses. Emission levels of the C6–C2 compounds phenylacetaldehyde (Figure 6a), phenylethylacetate and 2-phenylethanol (Figure S3) were not reduced in the ir-PhKAT1 lines. Phenylethylbenzoate, which is produced from 2-phenylethanol and benzoyl-CoA, had on average a 3.2-fold reduced emission (Figure 6a), suggesting reduced benzoyl-CoA level in these lines. The emission of benzylbenzoate, which is also directly dependent on benzoyl-CoA as co-substrate, was on average 3.5-fold reduced (Figure 6a). Also, emission of MeBA (Figure 6a) whose production depends on BA, was lower in the ir-PhKAT1 lines (2.7-fold on average). Emission levels of the other C6–C1 compounds, benzaldehyde (Figure 6a), benzylalcohol and benzylacetate (Figure S3), were significantly less (2-fold, 5.4-fold and 4.5-fold on average, respectively) and the emission of MeSA was significantly lower only in line 7 (Figure S3). The C6–C3 compound isoeugenol had slightly higher emission in both silenced lines compared with the wild type, but not compared with the non-silenced transgenic line 6 (Figure 6a). Vanillin, which was shown to be synthesised from p-coumaric acid (Figure 1) in Vanilla planifolia (Podstolski et al., 2002), unlike the other C6–C1 compounds, remained unaltered in the silenced lines (Figure S3).

Figure 6.

 Effect of PhKAT1 silencing on the emission levels of selected compounds representing the C6–C3, C6–C2 and C6–C1 compounds and on internal benzoic acid (BA) and C6–C1 volatile pools.
(a) Quantified emission of volatile benzenoids/phenylpropanoid compounds in wild type and ir-PhKAT1 plants. Flowers of two independent silenced T1 lines (4 and 7), one Mitchell control line (wt) and one transgenic non-silenced T1 line (6) were used for headspace analyses. Volatiles were collected for 23 h and analysed using GC-MS. Emission levels are given per hour, given a constant emission rate, per gram fresh weight and standardised using p-cymene as internal standard. For each line, the average of seven experiments was taken, using three detached flowers per experiment. Standard errors are shown. Letters indicate significant differences among lines (anova, P < 0.05 according to least significant difference post-hoc analysis). The emission levels of the other volatile compounds are shown in Figure S3.
(b) Quantified internal levels of BA and C6–C1 volatile compounds. For each measurement, three flowers of ir-PhKAT1 line 7 and a wild type plant were collected 4 h after the onset of the dark period (mean ± SD, = 3). Internal BA levels were quantified using added 13C-BA and values are given in μg g FW−1 (mean ± SE, = 6).

To exclude that the decrease in emission of volatiles was due to reduced transcript levels of relevant benzenoid biosynthetic genes, we analysed their transcripts levels. Analysis of the transcript levels of PhBPBT, PhBSMT1 and petunia phenylacetaldehyde synthase (PhPAAS), which synthesise benzylbenzoate and phenylethylbenzoate, MeBA and phenylacetaldehyde respectively, showed that PhKAT1 silencing did not affect transcript levels of these genes (Figure S4).

Silencing of PhKAT1 influences benzoic acid levels and internal C6–C1 compound pools

Because MeBA emission levels were reduced (Figure 6a) and PhBSMT transcript levels unaffected (Figure S4) in ir-PhKAT1 lines, we hypothesised that the level of the precursor of MeBA, BA, would be lower in these lines (Figure 1). To test this hypothesis, we measured BA levels in wild type flowers and flowers of line 7 at a time point during the dark period, when volatile emission is high (4 h in the dark period). In wild type flowers, BA levels were approximately four times higher than in line 7 (86 ± 8 and 21 ± 6 μg g FW−1, respectively), illustrating the contribution of PhKAT1 and thus the β-oxidative pathway, to BA biosynthesis (Figure 6b).

As it was recently shown that for most compounds internal volatile pools are highest during the dark period (Orlova et al., 2006), we reasoned that the effect of PhKAT1 silencing on internal volatile pools would be most pronounced during the night, when emission peaks as well. Therefore, we decided to analyse internal C6–C1 volatile pools after 4 h in the dark period in petals of line 7 and wild type. Internal pools of benzylbenzoate, phenylethylbenzoate and methylbenzoate were indeed significantly reduced in line 7 (30-fold, 36-fold and 18-fold, respectively) and benzaldehyde internal pools were undetectable in the silenced line (Figure 6b).

Discussion

Petunia 3-ketoacyl-CoA thiolase 1, a peroxisomal protein, is involved in the central C6–C1 floral volatile biosynthetic pathway and contributes to benzoic acid formation in petunia flowers

Petunia has been used extensively as a model system to study the biosynthesis of flavonoids, particularly those involved in floral pigmentation (Koes et al., 2005). Recently, it has emerged as a model system to study the biosynthesis of benzenoid and phenylpropanoid-related floral volatiles (Schuurink et al., 2006; Pichersky and Dudareva, 2007). Previous work, however, has concentrated on identifying genes and enzymes involved in the final 1–2 steps resulting in the synthesis of the volatiles themselves (Negre et al., 2003; Boatright et al., 2004; Kaminaga et al., 2006; Koeduka et al., 2006; Dexter et al., 2007). In this study, we used functional genomics and targeted metabolomics to identify and probe the function of PhKAT1 that encodes an enzyme potentially centrally positioned in the pathway leading to the biosynthesis of BA (Figure 1). The homology of PhKAT1 to known thiolases led to the hypothesis that it could be involved in the β-oxidative shortening of the propyl side-chain of 3-oxo-3-phenylpropionyl-CoA, leading to the production of benzoyl-CoA, which is used as a co-substrate for the production of benzylbenzoate and phenylethylbenzoate (Boatright et al., 2004). This reaction would be analogous to the shortening by two carbons of the 3-ketoacyl-CoA substrates in fatty acid degradation, catalysed by bona fide thiolases (Germain et al., 2001). Furthermore, its expression levels and patterns suggested involvement in scent biosynthesis. To test this hypothesis, we stably silenced PhKAT1 in petunia plants.

In these transgenic plants, PhKAT1 transcript levels in the petals at the peak of volatile production were decreased by as much as 75% (Figure 4b) and overall thiolase activity was reduced by 50% (Figure 5). On the metabolite levels, we observed significant decreases in internal pools of BA, methylbenzoate, benzylbenzoate, phenylethylbenzoate and benzylaldehyde (Figure 6b) and significant decrease in emission of methylbenzoate, benzylbenzoate, phenylethylbenzoate and benzylaldehyde (Figure 6a), but not of the C6–C3 compound isoeugenol, the C6–C2 compounds phenylacetaldehyde (Figure 6a), 2-phenylethanol and phenylethylacetate (Figure S3) and the p-coumaric-derived C6–C1 compound vanillin (Figure S3). Interestingly, internal pool levels of the C6–C1 compounds were more reduced than the emission levels, which can reflect a constant emission rate of these compounds, depleting the internal pools even further. Although MeSA can be produced from BA via SA, the moderate reduction in MeSA emission levels suggests that SA is also produced via (iso)chorismate, as has been shown in Arabidopsis (Wildermuth, 2001). Similar results were seen in ir-PhODO1 plants, which show dramatically reduced BA levels but only slightly reduced MeSA emission levels (Verdonk et al., 2005). The existence of this route leading to SA and MeSA has yet to be shown in petunia.

The decreased levels of compounds that are derived from BA or benzoyl-CoA, unlike those that are not derived from these two precursors (the C6–C2 and C6–C3 compounds), strongly suggest that PhKAT1 is involved in BA biosynthesis, probably through the conversion of 3-oxo-3-phenylpropionyl-CoA to benzoyl-CoA. Efforts to develop an enzymatic assay to measure 3-oxo-3-phenyl-propionyl-CoA thiolase activity were not successful. Because most thiolases have activity towards acetoacyl-CoA, measuring acetoacetyl-CoA thiolase activity in a cell represents the activity of most if not all thiolases present. Therefore, because transcript levels were reduced with 75%, the residual thiolase activity of 50% indicates that either other thiolases are still active or the PhKAT1 protein levels were less reduced than its transcripts levels. Nevertheless, this result shows that reduction of PhKAT1 transcript levels reduces overall thiolase activity in petunia petals

Silencing of PhKAT1 did not alter 2-phenylethanol levels in these lines (Figure S3). Phenylethylbenzoate emission levels, however, were significantly reduced (Figure 6a). As phenylethylbenzoate is produced from 2-phenylethanol and benzoyl-CoA by PhBPBT (Boatright et al., 2004; Dexter et al., 2007), it is likely that reduced benzoyl-CoA levels in ir-PhKAT1 lines are the cause for lower phenylethylbenzoate emission, supporting the hypothesis that benzoyl-CoA is the product of PhKAT1. Similarly, in Streptomyces maritimus, the thiolase encJ knock-out strain KJ produces less benzoyl-CoA-derived enterocin (Xiang and Moore, 2003). Our results also showed a peroxisomal localization for PhKAT1 and thus synthesis of benzoyl-CoA in this organelle. This finding is consistent with the recent report that a mutation in a gene encoding the peroxisomal protein benzoyl-CoA ligase (AT1G65880) eliminates benzoyl-CoA (or BA) biosynthesis in Arabidopsis seeds (Kliebenstein et al., 2007).

In plants, β-oxidation is not only essential for fatty acid catabolism (Germain et al., 2001), but also for the production of indole acetic acid (Zolman et al., 2001), jasmonic acid (Afitlhile et al., 2005) and valine (Lange et al., 2004). In Arabidopsis, there are three 3-ketoacyl-CoA thiolase (KAT) genes, but only for AtKAT2 (AT2G33150) it is known that it is important for fatty acid β-oxidation, jasmonic acid biosynthesis and indole acetic acid production (Hayashi et al., 1998; Germain et al., 2001). In silico co-expression analysis groups AtKAT5 with genes of the flavonoid pathway, suggesting it has a role different from that in general fatty acid β-oxidation (Carrie et al., 2007). Because PhKAT1 is not expressed in seedlings (Figure 2a), it is not likely to be involved in fatty acid β-oxidation. In addition, it has been hypothesised that in some plant species biosynthesis of vanillin, which is a minor volatile in petunia, occurs by a process that mirrors fatty acid β-oxidation (Loscher and Heide, 1994; Podstolski et al., 2002). However, the silencing of PhKAT1 did not affect the emission of vanillin (Figure S3), excluding a role for PhKAT1 in vanillin biosynthesis in petunia flowers. It is likely that PhKAT2, which is already expressed early during flower development (Figure 2c), when no volatile benzenoids are produced (Verdonk et al., 2003) and that lacks rhythmic expression in petals (Figure S1), is involved in fatty acid degradation or the synthesis of other metabolites, but not in the synthesis of floral scent compounds.

What is the proportional contribution of β-oxidation to the synthesis of benzenoids in petunia flowers?

While elucidating the specific biosynthetic pathways leading to BA in plants has been a long and difficult process that is still incomplete, feeding experiments previously suggested the contribution of both the β-oxidative and the non-β-oxidative pathway with benzoyl-CoA and benzaldehyde as intermediates, respectively (Ribnicky et al., 1998; Abd El-Mawla and Beerhues, 2002; Boatright et al., 2004). Silencing of PhBPBT and subsequent metabolic flux analysis following feeding of labelled Phe to flowers of these transgenic plants indicated that both β-oxidative and the non-β-oxidative pathways contribute to the synthesis of benzenoid compounds in petunia (Boatright et al., 2004). Flux analysis of ir-PhBPBT transgenic flowers suggested that benzylbenzoate is a precursor to some of the free BA found in the petals (Orlova et al., 2006). Our results strongly support a major role for a peroxisomal thiolase in the synthesis of benzoyl-CoA and consequently for the synthesis of benzenoids. From our results it is clear that mainly the β-oxidative pathway contributes to BA formation in petunia flowers during the night (Figure 6b) although isotope labelling and modelling studies suggested a greater flux through the non-β-oxidative pathway (Orlova et al., 2006). The reduced internal BA pool levels (four-fold reduction) correlate perfectly with the reduced PhKAT1 transcript levels (four-fold reduction) in the dark period, when BA levels (Kolosova et al., 2001a), PhKAT1 transcript levels (Figure 2b) and volatile emission (Verdonk et al., 2005) are high. Our results however, do not exclude the contribution of additional routes leading to the biosynthesis of BA (Boatright et al., 2004; Orlova et al., 2006). Related to this outcome, an aldehyde oxidase that converts benzaldehyde to BA was identified in Arabidopsis recently (Ibdah et al., 2009).

Although benzylbenzoate, phenylethylbenzoate and BA levels are reduced in both the ir-PhBPBT and our ir-PhKAT1 plants, there are also several differences, not in the least because PhBPBT concerns an enzyme in a final step, i.e. making the volatiles benzylbenzoate and phenylethylbenzoate. The most striking difference is that silencing of PhBPBT increased benzaldehyde emission (Orlova et al., 2006; Dexter et al., 2008) and internal pools (Orlova et al., 2006) whereas benzaldehyde emission and internal pools decreased as a consequence of PhKAT1 silencing (Figure 6a,b). The immediate precursor of benzylbenzoate, benzylalcohol, was reduced in ir-PhKAT1 plants as well but increased in ir-PhBPBT plants Labelling studies indicated that benzylalcohol can be produced from both benzaldehyde and benzylbenzoate in petunia flowers (Boatright et al., 2004). Because the exact nature of the enzymes that are involved in benzaldehyde and benzylalcohol production in plants is not known, it is difficult to explain these different results. One interpretation is that accumulation of intermediates upstream of benzoyl-CoA in the ir-PhKAT1 plants could inhibit the activity of benzylalcohol and/or benzaldehyde-producing enzymes, as is the case for PhBPBT inhibition by coniferyl aldehyde (Dexter et al., 2007). Alternatively, accumulated intermediates could inhibit transcription of biosynthetic genes, as is the case for PhBSMT by Phe (Boatright et al., 2004). Finally, reduced benzoyl-CoA and/or BA levels in ir-PhKAT1 plants could enhance the formation of BA from benzaldehyde through the non-β-oxidative pathway in the dark period (Figure 1) as suggested for ir-PhBPBT plants in the light period (Orlova et al., 2006). This enhancement would deplete the benzaldehyde internal pool and consequently reduce benzylalcohol internal pool and emission levels. These ir-PhKAT1 lines can be important tools for future labelling studies to investigate the proportional contributions of the β-oxidative and non-β-oxidative pathway in more detail.

Regulation of floral scent production and the role of compartmentalization

In this study, we have shown that the expression pattern of PhKAT1 (Figure 2) is characteristic for a floral benzenoid-related gene (Kolosova et al., 2001a; Negre et al., 2003; Boatright et al., 2004; Underwood et al., 2005; Verdonk et al., 2005; Kaminaga et al., 2006; Koeduka et al., 2006; Dexter et al., 2007). We have further shown that expression of PhKAT1 peaks approximately 3–6 h earlier in petals than PhODO1 during the day/night cycle (Figure 2b), which precedes peak volatile emission by approximately 2–3 h (Verdonk et al., 2005). Reanalysis of our microarray experiments with ir-PhODO1 plants (Verdonk et al., 2005) showed that PhKAT1 expression is not influenced by PhODO1 silencing (ratio 1.04; = 0.44). Apparently, the expression of PhODO1 and PhKAT1 are under different transcriptional control mechanisms.

Volatile benzenoid/phenylpropanoid production and emission are spatially, developmentally and diurnally regulated. This fact makes sense as these volatiles have Phe as precursor in common with many other primary and secondary metabolites. Also methyl donors, supplied by the SAM cycle (Verdonk et al., 2003, 2005), are shared with other pathways (Negre et al., 2003; Schuurink et al., 2006). It is therefore expected that channelling and the existence of different substrate pools are part of the regulatory machinery of the benzenoid pathway. When fluxes are perturbed, for instance by the accumulation of intermediates as a consequence of petunia acetyl-CoA:coniferyl alcohol acetyltransferase (PhCFAT) silencing, this outcome has a direct effect on PhBSMT and PhBPBT enzymatic activity (Dexter et al., 2007). PhBSMT is likely to be localised in the cytosol in analogy to the snapdragon S-adenosyl-l-methionine:BA carboxyl methyltransferase (AmBAMT) (Kolosova et al., 2001b). However, it was modelled by Boatright et al. (2004) that a second, large pool of MeBA is stored in the vacuole. Here we show that PhKAT1 localises to the peroxisome, adding another layer of regulation. In analogy with fatty acid β-oxidation, CoA-activated compounds like benzoyl-CoA could be transported across the peroxisomal membrane by ABC-transporters (Footitt et al., 2002). Both in the cytosol and the peroxisomes, benzoyl-CoA can be converted to BA by the action of thioesterases (Figure 1), which would implicate the existence of different BA and benzoyl-CoA pools in the cell. Tilton et al. (2004) identified a peroxisomal acyl-CoA thioesterase that is likely not involved in fatty acid β-oxidation, but in another process in plants. The substrate for this thioesterase has not been identified yet and it remains to be seen whether multiple thioesterases that can act on benzoyl-CoA exist in petunia petals. Finally, the snapdragon benzaldehyde dehydrogenase, involved in the non-β-oxidative pathway, was recently shown to be located in the mitochondria (Long et al., 2009). Our finding that PhKAT1 and thus the β-oxidative pathway localises to the peroxisomes means that not only distinct routes with different enzymes are involved in the production of BA and volatile benzenoids, but that these enzymes are active in different cellular compartments.

Experimental procedures

Plant material, growth conditions and transformations

Petunia hybrida cv. Mitchell [P. axillaris × (P. axillaris × P. hybrida cv Rose of Heaven)] wild type and transgenic plants were grown in a greenhouse (16 h photoperiod, 500 μmol m−2 s−1 light intensity, 60–65% humidity and day/night temperatures of 22/17°C) during winter (T0 lines) and summer (T1 lines). For sample collection and volatile headspace analysis, plants were moved to controlled growth chambers (16 h photoperiod, 250–350 μmol m−2 s−1 light intensity, 70% relative humidity and constant temperature of 21°C) at least 3 days prior to the experiments.

Transgenic plants were obtained using a standard leaf disc transformation protocol (Horsch et al., 1985), with minor modifications. The shoot-inducing medium contained 500 mg L−1 carbenicillin, 300 mg L−1 kanamycin and 20 mg L−1 nystatin. The root-inducing medium contained 500 mg L−1 carbenicillin, 50 mg L−1 kanamycin and 20 mg L−1 nystatin. Media contained 3% sucrose (w/v) as the sole carbon source. Rooted plants were transferred to soil and screened for transgene integration by PCR using construct-specific primers. Next generation plants were obtained by manual self-pollination and subsequent selection on half strength MS (pH 5.8) agarose (0.7%) plates containing 80 mg L−1 kanamycin.

Constructs design

Constructs were produced using standard molecular biological methods. To create the PhKAT1 hairpin construct, a 174 bp fragment was amplified with forward primer 5′-aaaaagcaggctcgatctttgcaatgctcggga-3′ and reverse primer 5′-agaaagctgggtcaagttcactaaatcctgct-3′ that included AttB adapters (in bold; GATEWAY system, Invitrogen, http://www.invitrogen.com). This fragment was cloned in the pDONR207 entry vector (Invitrogen) and subsequently recombined with pK7GWIWG2(I) (Karimi et al., 2002), generating a PhKAT1 hairpin-RNAi construct under control of the CaMV 35S promoter. To create the CaMV 35S-driven C-terminal GFP fusion construct (35S:KAT1-GFP), the PhKAT1 CDS was amplified with the introduction of an NcoI site at the ATG start codon and an Ala-Gly-linker followed by an XbaI site excluding the stop codon (fw primer: 5′-atgccatggagaaagcaattcaaagg-3′ and rev primer: 5′-cctctagaacctgctttcgcatccttggataag-3′; restriction sites are underlined; the Ala–Gly-linker is in bold). The GFP CDS was amplified from plasmid SGFP2 (Kremers et al., 2007), introducing an XbaI site and Ala–Gly-linker adjacent to the ATG start codon and a SacI site beyond the stop codon (fw primer 5′-ggtctagagcaggtATGgtgagcaagggcgag-3′ and rev primer 5′-cgagtctTTActtgtacagctcgtccatgccgag-3′; restriction sites are underlined; the Ala–Gly-linker is in bold; start and stop codon are in upper case). Using a three-point ligation, both fragments were ligated NcoI–SacI in a shuttle vector between the CaMV 35S promoter and the nopalin synthase terminator (tnos). The 35S:PhKAT1–GFP:tnos cassette was excised using HindIII and SfoI and transferred to pBINplus (van Engelen et al., 1995) between HindIII and SmaI. The 35S:PhKAT1–GFP:tnos cassette was sequenced and the plasmid was transferred to Agrobacterium tumefaciens GV3101 (pMP90).

Volatile sampling, GC-MS analysis and benzoic acid measurements

For volatile sampling (headspace analysis), T1 transgenic lines (T0 selfings) and wild type plants were used. Headspace analyses were performed essentially as described by Verdonk et al. (2005), except that three detached flowers per desiccator were measured for a 23 h period, with seven replicas per line. GC-MS analysis was performed as described by Verdonk et al. (2005), but here 5 ng p-cymene was used as an internal standard. A synthetic mix with known amounts of volatile compounds was analysed using GC-MS, enabling quantification and identification of compounds.

For internal volatiles and BA measurements, one T1 transgenic line (line 7) and wild type plants were used. Flower limbs – three flowers per sample – were harvested 4 h into the dark period, pooled and frozen in liquid nitrogen and stored at −80°C. Internal volatiles were extracted using hexane as described by Boatright et al. (2004). Three independent replicates were taken for each plant. BA measurements were performed as described by Zhang and Zuo (2004) with a few modifications. Homogenised flower petals were extracted twice with ethyl acetate, after addition of 2N HCl and 13C-BA as the internal standard. Hexane-extracted and derivatised samples were analysed by means of GC-MS. For each plant, six measurements were performed.

RNA isolation and gel-blot analysis

T1 transgenic lines and wild type petunia plants were used for RNA gel-blot analyses. Independent sample collections for all gel-blot experiments were performed twice. With the exception of the developmental course, 2-day-old flowers were used in all experiments. All tissues except those for the time course were sampled two h before the onset of the dark period. For the time course, petal limbs of three flowers were taken at 3-h intervals for a 30-h period. During the experiment, the dark period in the growth chamber started at 12.00 h and ended at 20.00 h. For the developmental course, petal limbs (without petal tube) were harvested when the flower was 1, 3 or 4 cm long, opening (1 cm diameter), open (day 1), open with dehisced anthers or senescing.

Total RNA was extracted using Trizol reagent (Invitrogen). For RNA gel-blot analysis, 8 μg of total RNA per sample was loaded on gel. In order to normalise the RNA gel-bots, blots were re-hybridised with the petunia floral binding protein 1 (PhFBP1) (Angenent et al., 1992) probe for the time course and the developmental blot and with a 18S rDNA probe (fw primer: 5′-agcaggctaagtctcgt-3′; rev primer: 5′-agcggatgttgcttttagga-3′) for the tissue blot. For all cDNA probes, fragments containing the full coding sequence were used except for PhKAT2 (fw primer: 5′-tgcctctgtatctgacttg-3′; rev primer: 5′-ctactatgtgtggctttctc-3′) and PhFBP1 (fw primer: gttctttgtgatgctcg; rev primer: ctctcctgcaaatttgg). For the petunia ODORANT1 (PhODO1) probe, forward primer 5′-gtcacagcggcagcagctac-3′ and reverse primer 5′-ctaacttcctagtagttccagac-3′ were used, generating a 518 bp fragment, which does not include the R2R3-MYB domain.

Thiolase assay

For the thiolase assay, wild type flowers and flowers of line 7 (T2 generation) were used. Crude protein extracts were prepared from 2-day-old flowers, harvested 1 h in the dark period and used immediately. The extraction buffer consisted of 100 mm Tris/HCl pH 8.0, 14 mm 2-mercaptoethanol, 5 mm Na2S2O5, 10% glycerol, 5% PVPP and 0.2 mm PMSF. Thiolase activity was determined by measuring the decay of acetoacetyl-CoA. The final volume of the assay was 100 μl and conditions were as follows: 100 mm Tris pH 8.0, 50 mm KCl, 25 mm MgCl2, 50 μm CoA and 50 μm acetoacetyl-CoA. Products were separated and detected on HPLC as described in Burns et al. (2005). = 4; standard deviations are shown.

Agrobacterium-mediated transient transformation and confocal microscopy

Agroinfiltration of petunia petals was performed as described by Verweij et al. (2008). To enable co-localization, two separate A. tumefaciens GV3101 cultures harbouring the KAT1–GFP fusion construct or the peroxisomal marker px-rk, which contains mCherry with a C-terminal peroxisomal targeting signal 1 (Nelson et al., 2007), respectively, were mixed 1:1 (v/v) prior to infiltration. For confocal analysis, infiltrated petals were embedded in 80% glycerol to reduce light scattering of the conical epidermal cells. GFP and mCherry were imaged using a Zeiss LSM 510 confocal laser scanning microscope (Vermeer et al., 2008).

Isolation of the full-length cDNA of PhKAT2

Total RNA was extracted from petals using Trizol reagent as described above. One microgram of total RNA was used to synthesise cDNA using M-MuLV reverse transcriptase (Fermentas, http://www.fermentas.com). The full-length cDNA was obtained using a 5′-RACE kit (Invitrogen) and confirmed by nucleotide sequencing.

Acknowledgements

We would like to thank Ludek Tikovski, Harrold Lemereis and Thijs Hendrix for taking care of our plants, Gert-Jan Kremer for the SGFP construct, Erik Manders and Milou Schuurmans for helping us with the confocal microscope and Julian Verdonk for identifying the thiolase in the original microarray experiments.

Accession numbers: The PhKAT1 and PhKAT2 cDNA sequence data can be found in the GenBank database under the accession numbers FJ657663 and FJ657664, respectively.

Ancillary