A BAHD acyltransferase is expressed in the tapetum of Arabidopsis anthers and is involved in the synthesis of hydroxycinnamoyl spermidines

Authors

  • Etienne Grienenberger,

    1. Institut de Biologie Moléculaire des plantes, Unité Propre de Recherche 2357 du Centre National de la Recherche Scientifique conventionnée à l’Université Louis Pasteur, 67084 Strasbourg Cedex, France,
    Search for more papers by this author
  • Sébastien Besseau,

    1. Institut de Biologie Moléculaire des plantes, Unité Propre de Recherche 2357 du Centre National de la Recherche Scientifique conventionnée à l’Université Louis Pasteur, 67084 Strasbourg Cedex, France,
    Search for more papers by this author
    • Present address: Viikki Biocenter, Department of Biological and Environmental Sciences, Viikinkaari 5D, PO Box 56, Helsinki, Finland.

  • Pierrette Geoffroy,

    1. Institut de Biologie Moléculaire des plantes, Unité Propre de Recherche 2357 du Centre National de la Recherche Scientifique conventionnée à l’Université Louis Pasteur, 67084 Strasbourg Cedex, France,
    Search for more papers by this author
  • Delphine Debayle,

    1. Plate-Forme d’Analyses Métaboliques de l’IBMP, Institut de Botanique, 67084 Strasbourg Cedex, France, and
    Search for more papers by this author
  • Dimitri Heintz,

    1. Plate-Forme d’Analyses Métaboliques de l’IBMP, Institut de Botanique, 67084 Strasbourg Cedex, France, and
    Search for more papers by this author
  • Catherine Lapierre,

    1. UMR206 Chimie Biologique, AgroParisTech, INRA, 78850 Thiverval-Grignon, France
    Search for more papers by this author
  • Brigitte Pollet,

    1. UMR206 Chimie Biologique, AgroParisTech, INRA, 78850 Thiverval-Grignon, France
    Search for more papers by this author
  • Thierry Heitz,

    1. Institut de Biologie Moléculaire des plantes, Unité Propre de Recherche 2357 du Centre National de la Recherche Scientifique conventionnée à l’Université Louis Pasteur, 67084 Strasbourg Cedex, France,
    Search for more papers by this author
  • Michel Legrand

    Corresponding author
    1. Institut de Biologie Moléculaire des plantes, Unité Propre de Recherche 2357 du Centre National de la Recherche Scientifique conventionnée à l’Université Louis Pasteur, 67084 Strasbourg Cedex, France,
      (fax +33 388 614442; e-mail michel.legrand@ibmp-ulp.u-strasbg.fr).
    Search for more papers by this author

(fax +33 388 614442; e-mail michel.legrand@ibmp-ulp.u-strasbg.fr).

Summary

BAHD acyltransferases catalyze the acylation of many plant secondary metabolites. We characterized the function of At2g19070, a member of the BAHD gene family of Arabidopsis thaliana. The acyltransferase gene was shown to be specifically expressed in anther tapetum cells in the early stages of flower development. The impact of gene repression was studied in RNAi plants and in a knockout (KO) mutant line. Immunoblotting with a specific antiserum raised against the recombinant protein was used to evaluate the accumulation of At2g19070 gene product in flowers of various Arabidopsis genotypes including the KO and RNAi lines, the male sterile mutant ms1 and transformants overexpressing the acyltransferase gene. Metabolic profiling of flower bud tissues from these genetic backgrounds demonstrated a positive correlation between the accumulation of acyltransferase protein and the quantities of metabolites that were putatively identified by tandem mass spectrometry as N1,N5,N10-trihydroxyferuloyl spermidine and N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine. These products, deposited in pollen coat, can be readily extracted by pollen wash and were shown to be responsible for pollen autofluorescence. The activity of the recombinant enzyme produced in bacteria was assayed with various hydroxycinnamoyl-CoA esters and polyamines as donor and acceptor substrates, respectively. Feruloyl-CoA and spermidine proved the best substrates, and the enzyme has therefore been named spermidine hydroxycinnamoyl transferase (SHT). A methyltransferase gene (At1g67990) which co-regulated with SHT during flower development, was shown to be involved in the O-methylation of spermidine conjugates by analyzing the consequences of its repression in RNAi plants and by characterizing the methylation activity of the recombinant enzyme.

Introduction

Acylation is a common and biochemically important modification of numerous plant metabolites. Plant BAHD acyltransferases were recently identified and demonstrated to share phylogenetic relationships. The BAHD family was named according to the first letter of the first four characterized members of the family (BEAT, AHCT, HCBT, DAT) (St-Pierre et al., 1998; St-Pierre and De Luca, 2000). BAHD acyltransferases catalyze the transfer of the acyl moiety to a wide range of acceptor molecules and are, therefore, involved in the biosynthesis of a large array of natural plant compounds such as lignin, phenolics, alkaloids, phytoalexins, anthocyanins and volatile esters (St-Pierre and De Luca, 2000; D’Auria, 2006). Over 50 BAHD acyltransferases have been assigned a function in numerous plant species, including dicotyledon and monocotyledon angiosperms and coniferous gymnosperms, on the basis of genetic and/or biochemical experiments (D’Auria, 2006).

We previously biochemically characterized the first identified BAHD acyltransferase of Arabidopsis thaliana as the hydroxycinnamoyl-CoA shikimate/quinate hydroxycinnamoyl transferase (HCT), which is involved in lignin synthesis. Hydroxycinnamoyl transferase was shown to catalyze the synthesis of shikimate and quinate esters of p-coumaric acid; these are the substrates of the 3-hydroxylase CYP98A3 (Schoch et al., 2001) and the precursors of guaiacyl and syringyl units of lignin. Repression of HCT resulted in marked changes in the amount and composition of lignin, thus demonstrating that HCT functions in phenylpropanoid metabolism in planta (Hoffmann et al., 2004; Besseau et al., 2007). Hydroxycinnamoyl transferase homologs have been identified in several plant species (Tsai et al., 2006; Wagner et al., 2007), consistent with the presence of lignin in all vascular plants.

Since the discovery of HCT, four other genes of A. thaliana have been biochemically demonstrated to encode BAHD acyltransferases. Three of these have been shown to acylate anthocyanin substrates (At3g29590, At1g03940 and At1g03495) (D’Auria et al., 2007b; Luo et al., 2007) and the fourth (At3g03480) to catalyze the synthesis of a volatile ester induced in leaves upon wounding (D’Auria et al., 2007a). All the BAHD acyltransferases use CoA esters as acyl donors and catalyzse either aromatic or aliphatic acylation of a variety of oxygen- and nitrogen-containing acceptors to produce esters and amides, respectively (D’Auria, 2006). This catalytic versatility makes functional predictions difficult to infer from the primary sequence alone. However, phylogenetic analysis of members of the BAHD family demonstrated that acyltransferase sequences are distributed into distinct clades, whose members display some similarity in their substrate preferences. This is the case, for instance, for acyltransferases that have anthocyanins as acceptor substrates and define a superclade of proteins (Luo et al., 2007). Hydroxycinnamoyl transferase (At5g48930) belongs to a distinct clade that also includes two other Arabidopsis members (At5g57840 and At2g19070). Several enzymes of this clade have been biochemically characterized from various plant species and shown to transfer an aromatic acyl group to various acceptors (Yang et al., 1997; Hoffmann et al., 2003; Niggeweg et al., 2004; D’Auria, 2006). Such similarity in the catalytic properties of closely related members of the BAHD family constitutes a valuable lead in the search for functions of unknown acyltransferase genes.

Here we report on the functional characterization of a BAHD acyltransferase of A. thaliana (At2g19070), which is phylogenetically closely related to HCT. Analysis of promoter–GUS gene fusion in transgenic Arabidopsis showed that the acyltransferase gene is specifically expressed in anther tissues at early stages of flower development. Histochemical sections of transgenic flower buds revealed that gene expression was restricted to tapetum cells of the anthers. The analysis of metabolic profiles obtained from flower buds of RNAi plants showed the specific decrease of peaks that were putatively identified by mass spectrometry as N1,N5,N10-trihydroxyferuloyl spermidine and N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine. These compounds were undetectable in a T-DNA insertion mutant line, thus confirming the involvement of the acyltransferase in their biosynthesis. Expression of At2g19070 in Escherichia coli enabled us to produce a recombinant protein that displayed acyltransferase activity. Spermidine was shown to be the only acceptor substrate and various hydroxycinnamoyl-CoA esters were efficient acyl donors. Therefore, the enzyme was named spermidine hydroxycinnamoyl transferase (SHT). Another gene, a putative methyltransferase (At1g67990) that is co-regulated with SHT during anther development, was shown to be implicated in the same biosynthetic pathway by analyzing the impact of its repression in RNAi plants and by characterizing the methyltransferase activity of the recombinant enzyme expressed in bacteria.

Results

At2g19070 is specifically expressed in anthers

As mentioned above, phylogenetic studies suggested that At2g19070 belongs to the BAHD acyltransferase gene family. Analysis of microarray data indicated a strong accumulation of At2g19070 transcripts at early stages of Arabidopsis flower development (https://www.genevestigator.ethz.ch/gv/index.jsp; http://www.bar.utoronto.ca/efp/cgi-bin/efpWeb.cgi) (Toufighi et al., 2005; Winter et al., 2007; Hruz et al., 2008).

The kinetics of At2g19070 gene product accumulation in Arabidopsis plants were investigated using a specific polyclonal antiserum raised against the purified recombinant protein expressed in Nicotiana benthamiana leaves (see Experimental Procedures). When protein extracts of different plant organs were immunoblotted with the antiserum, At2g19070 protein was solely detected in inflorescences, more precisely in flower buds at early stages of development (Figure 1a). Flower buds of varying age were analyzed and the protein was shown to accumulate during bud growth and to disappear later on, when the flower opened (Figure 1b). To study acyltransferase gene expression at the tissue level, promoter activity pattern was investigated by analyzing GUS staining in transgenic plants expressing promoter–GUS constructs. After incubation with GUS substrate, a strong coloration appeared in closed buds at the top of inflorescences (Figure 2a,b), thus confirming the protein blot data. A close-up of stained buds revealed that GUS staining was localized to the anthers (Figure 2b). Cross sections of flower buds showed that the coloration was restricted to anther tapetum cell layers surrounding the microspores (Figure 2c,d). These results suggested that the putative acyltransferase participates in the highly active tapetum metabolism that provides nutrients and materials for the formation of the pollen wall and pollen coat (Piffanelli et al., 1998; Scott et al., 2004).

Figure 1.

 Expression of At2g19070 in various organs of Arabidopsis.
Protein extracts from different plant tissues were separated by electrophoresis on SDS-PAGE and immunodetected with polyclonal antibodies raised against the purified recombinant protein.
(a) Expression levels in roots (R), leaves (L), stems (S) and flower buds (FB) were compared. The arrow indicates the migration of the protein at 51 kDa.
(b) Protein accumulation in flowers at the different stages of development shown in the corresponding upper photographs.

Figure 2.

At2g19070 expression analyzed by GUS staining of At2g19070 promoterGUS transgenic plants.
(a) and (b) Global view of GUS-stained Arabidopsis inflorescences.
(c) A GUS-stained section from a transgenic flower.
(d) Close-up view showing GUS staining of transgenic anther tissues (ta, tapetum cells).

Impact of At2g19070 repression and overexpression in transgenic Arabidopsis

The effects of At2g19070 repression were studied in silenced RNAi plants and in a T-DNA tagged mutant line (Alonso et al., 2003). To induce gene silencing, Arabidopsis plants were transformed with a hairpin construct containing a portion of the At2g19070 sequence driven by the CaMV 35S promoter. Among primary transformants, silencing efficiency was checked by immunoblotting of flower extracts using the specific antiserum raised against the recombinant protein. Out of about a hundred transgenic lines analyzed, a wide range of gene expression levels was measured as shown in Figure 3(a) for selected examples, but only 5–10% of plants were strongly silenced. This might be due to the poor expression of the 35S promoter in the tapetum as observed previously (Skirycz et al., 2007). Flower buds of the most efficiently silenced plants were extracted and their HPLC profiles (Figure 4b, c) were compared with that of wild-type flower tissues (Figure 4a). Among the peaks resolved by the HPLC gradient, a majority were identified by mass spectrometry as flavonol derivatives (data not shown) and were present in similar amounts in RNAi and control extracts. In contrast, two peaks eluting at 30 and 33 min, respectively, were strongly decreased in RNAi plant extracts compared with the control profile. Mass values of 721 and 735 Da were determined from the mass of the pseudo-molecular ions observed upon positive ([M+H]+ at m/z 722 and 736; Figure 4g) and negative electrospray ionization ([M−H] at m/z 720 and 734, not shown) of the two compounds. The mass values, retention times and UV spectra of these compounds (Table S1 in Supporting Information) strongly indicated that they are trisubstituted hydroxycinnamic acid spermidines. Such molecules have been isolated from various plant sources and their structures elucidated from 1H-NMR, 13C-NMR and mass spectrometry data (Meurer et al., 1988; Bokern et al., 1995; Zamble et al., 2006). Moreover, it has been shown that these compounds display characteristic fragmentation patterns of [M+H]+ ions when subjected to collision-induced dissociation. The fragmentation patterns of [M+H]+ ions from compounds 1 and 2 are shown in Figure 4(h) and are characteristic of triacylated spermidine conjugates (Meurer et al., 1988; Bokern et al., 1995; Zamble et al., 2006) (Table S1). Compound 1 yielded fragment ions at m/z 193, revealing the presence of hydroxyferuloyl residues whereas ions at m/z 193 and 207 were detected in the fragmentation pattern of compound 2, indicating the presence of both sinapoyl and hydroxyferuloyl moieties. Major ions at m/z 530 (compound 1) and 544 (compound 2) were generated by the loss of a hydroxyferuloyl moiety, and the additional loss of a second acyl moiety was likely to have yielded ions at m/z 352, 338 and 321 (Figure 4i). For the localization of hydroxycinnamoyl residues in compound 2, fragments at m/z 250 and 278 were diagnostic of the cleavage between C4 and N5 of the acylated spermidine molecule (see the putative fragmentation scheme presented in Figure 4i) and suggested that N10 is predominantly substituted by the sinapoyl residue. The minor signal at m/z 264 in the compound 2 fragmentation pattern (Figure 4h) may be explained by a fragment comprising either a sinapoyl group and the three-carbon chain or a hydroxyferuloyl group and the four-carbon chain, and thus reveals the presence of a minor isomer bearing the sinapoyl moiety in the N1 position. This heterogeneity is reminiscent of what has been described for spermidine conjugates of Quercus pollen (Bokern et al., 1995). In conclusion, mass spectrometry data enabled us to putatively identify compound 1 as N1,N5,N10-trihydroxyferuloyl spermidine and to show that compound 2 probably includes two isomers, a predominant one N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine and a minor species, N1-sinapoyl-N5,N10-dihydroxyferuloyl spermidine.

Figure 3.

 Expression of acyltransferase protein in different Arabidopsis lines.
Protein extracts from flower buds were immunoblotted using specific antibodies raised against the recombinant protein.
(a) Expression levels in wild-type (wt) and distinct RNAi lines (lanes 1–6) were compared.
(b) Expression in knockout (KO) and ms1 mutants compared with their respective controls.
(c) Protein accumulation in the overexpressing line (OE) compared with the wild type.
(d) Position of the T-DNA insertion in the KO mutant line.

Figure 4.

 Identification and quantification of spermidine conjugates in different Arabidopsis lines.
Flower bud HPLC profiles of wild-type (a), two RNAi lines differently repressed (b, c), a knockout (KO) mutant (d), ms1 mutant (e) and an overexpressing line (f) are presented.
(g) Peaks 1 and 2 from the wild-type profile were analyzed by positive electrospray ionization (ESI) mass spectrometry.
(h) Fragmentation patterns of [M+H]+ species obtained by collision-induced dissociation of compounds 1 and 2.
(i) Putative fragmentation scheme and m/z values of compounds 1 and 2. The scheme and m/z values of fragments are presented for the major compound, N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine (peak 2) and, when different, for N1,N5,N10-trihydroxyferuloyl spermidine (peak 1), m/z values of fragments are given in brackets.

To confirm the function of At2g19070 in vivo, a T-DNA insertion mutant was analyzed (Alonso et al., 2003). The mutant line (SALK_055511C) contains the T-DNA in the first exon sequence of the gene at position 91 (Figure 3d). The homozygous state of the insertion was verified by PCR (not shown) and, consistently, the protein was undetectable by immunoblotting in mutant flower buds (Figure 3b). As expected, the phenolic profiling of the KO mutant line demonstrated the complete absence of the two spermidine derivatives in bud extracts (Figure 4d).

At2g19070 was overexpressed in Arabidopsis transgenic plants under the control of the tapetum-specific TA29 promoter (Koltunow et al., 1990). Gene overexpression in flower tissues was checked by immunodetection of the protein (Figure 3c). Examination of the metabolite content of transgenic flower buds (Figure 4f) revealed changes in the spermidine conjugate ratio: peak 2 was barely affected compared with wild-type (Figure 4a) but peak 1 was significantly increased, indicating that methylation of the trihydroxyferuloyl precursor became a limiting step in acyltransferase overexpressing plants.

Taken together these data suggest involvement of the acyltransferase in the biosynthesis of the flower spermidine derivatives. The structure of the metabolites affected by At2g19070 deregulation indicates that the function of the acyltransferase is likely to transfer hydroxycinnamoyl moieties on the polyamine acceptor molecule. This assumption was investigated further by assaying the activity of the recombinant enzyme produced in bacteria.

Expression of At2g19070 in Escherichia coli and substrate specificity of the recombinant acyltransferase

In order to study the enzymatic properties of the putative acyltransferase, the coding region of At2g19070 was cloned into a vector that introduced a N-terminal GST tag and then expressed in bacteria. The recombinant protein was purified by affinity chromatography and acyltransferase activity was tested with different polyamines and hydroxycinnamoyl-CoA esters as substrates. The identity of reaction products was confirmed by retention times, UV spectra and liquid chromatography-tandem mass spectrometry (LC/MS/MS) analysis, and relative catalytic activities with the different substrates were estimated from the relative quantities of reaction products. Spermidine was the only polyamine that appeared to be efficiently acylated, since acyltransferase activities measured with the diamine putrescine and the tetraamine spermine as acceptor substrates were only 1–3% of the value measured with spermidine. These results are in accordance with flower extract analysis that did not detect putrescine or spermine conjugates. Therefore, the acyltransferase was named spermidine hydroxycinnamoyl transferase (SHT). Several hydroxycinnamoyl-CoA esters were substrates and yielded triacylated spermidines (Figure 5). Feruloyl-CoA ester appeared to be the most efficient donor substrate, sinapoyl-CoA was a poor substrate whereas hydroxyferuloyl-CoA did not produce trihydroxyferuloyl spermidine. As one example, Figure 6 presents reaction product characterization after SHT incubation in the presence of feruloyl-CoA and spermidine. N1,N5,N10-triferuloyl spermidine was the major reaction product but mono-acylated and di-acylated products were also detected in smaller amounts, indicating that these intermediates were rapidly acylated further. Similar patterns of reaction products were observed after incubation with other hydroxycinnamoyl-CoA esters but with lower yields (data not shown). Spermidine hydroxycinnamoyl transferase substrate preference data may indicate that, in the biosynthetic pathway, hydroxylation and methylation steps (to form the hydroxycinnamoyl-CoA substrate) precede the transfer reaction catalyzed by SHT and that, subsequently, further hydroxylations and methylation of the intermediate amide lead to the major flower end product, N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine.

Figure 5.

 Substrate preference of spermidine hydroxycinnamoyl transferase (SHT) with different CoA esters as acyl donors.
The different CoA esters shown on the left were incubated in the presence of spermidine. At the end of the incubation, reaction products were quantified by HPLC. Mean relative activity values and standard errors were calculated from duplicates.

Figure 6.

 Characterization by mass spectrometry of spermidine hydroxycinnamoyl transferase (SHT) reaction products after incubation in the presence of feruloyl-CoA and spermidine.
The incubation mixture was analyzed by ultra performance liquid chromatography tandem mass spectrometry (UPLC/MS/MS). Positive electrospray ionization was used and [M+H]+ ions with m/z of 322 (monoferuloyl spermidine), 498 (diferuloyl spermidine) and 674 (triferuloyl spermidine) were monitored after incubation in the presence (upper traces) or absence (lower traces) of recombinant SHT.

Mutants affected in MS1 and SHT genes display similar chemotypes

The MALE STERILITY 1 (MS1) gene encodes a PHD-finger transcription factor that regulates pollen and tapetum development (Ito et al., 2007; Yang et al., 2007). MALE STERILITY 1 has been shown to control the expression of a set of genes associated with pollen wall and coat formation (Alves-Ferreira et al., 2007; Ito et al., 2007; Yang et al., 2007). Among these numerous genes involved in pollen development, it is worth noting that MS1 transcriptionally regulates the At2g19070 gene. Accordingly, no SHT protein accumulation could be detected by immunoprobing of ms1 flower tissue extracts with our specific antiserum (Figure 3b). Moreover, as shown in Figure 4(e), the metabolic profile of ms1 flower buds strikingly resembled that of the sht KO mutant (Figure 4d), particularly with respect to the total absence of acylated spermidines. Mass spectrometric analysis of major peaks of the profile demonstrated that they correspond to the same flavonoids as those present in wild-type tissues (data not shown).

Impact of SHT gene repression on pollen development

Pollen grains from wild-type and sht KO mutant plants were examined by light and scanning electron microscopy. Wild-type Arabidopsis pollen exhibited a characteristic shape (Edlund et al., 2004; Morant et al., 2007) as shown in Figure 7 (left-hand photographs). Mutant pollen wall displayed irregularities and depressions that are barely visible at low magnification using light microscopy (Figure 7d) but are clearly apparent on scanning electron micrographs (Figures 7e and 7f).

Figure 7.

 Comparison of wild-type and knockout (KO) mutant pollen grains by light and scanning electron microscopy.
Arrows indicate depressions only visible on mutant pollen.
(a), (d) Control and mutant pollen examined using a stereomicroscope.
(b, c, e, f) Scanning electron micrographs of wild-type and mutant pollen.

Fluorescence microscopy revealed striking changes in the autofluorescence of pollen grains of different genotypes. A strong decrease in fluorescence of KO mutant pollen (Figure 8b) was observed compared with the wild type (Figure 8a) and, in contrast, the fluorescence intensity of pollen grains of the SHT overexpressing line (Figure 8c) was enhanced. When pollen grains were washed with aqueous methanol (see Experimental Procedures), analysis of the pollen wash by HPLC demonstrated the presence of acylated spermidines as the major fluorescent compounds released from all genotypes apart from the KO mutant. Spermidine hydroxycinnamoyl transferase overexpressing pollen grains displayed increased fluorescence, probably resulting from the accumulation of trihydroxyferuloyl spermidine (Figure 8c, peak 1). It is noteworthy that the fluorescence of pollen grains of a chalcone synthase RNAi line was increased compared with wild type due to the absence of UV-absorbing flavonoids, although the quantity of spermidine conjugates was essentially unchanged (Figure 8d). Thus, altogether these results indicate that spermidine conjugates are, at least in part, responsible for pollen autofluorescence due to their outermost location in the pollen coat.

Figure 8.

 Spermidine conjugates of the pollen coat contribute to pollen grain autofluorescence.
Pollen grains from various genotypes were examined by epifluorescence microscopy (left-hand photographs). Pollen was washed with aqueous methanol and the metabolite content in the solvent was analyzed by HPLC. Products were detected by their fluorescence and HPLC profiles are presented on the right. Peak 1, N1,N5,N10-trihydroxyferuloyl spermidine; peak 2, N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine.
(a) Wild-type, (b) knockout (KO) mutant, (c) overexpressing line, (d) CHS RNAi line.

The At1g67990 O-methyltransferase gene is involved in flower spermidine conjugate biosynthesis

Among the numerous genes that are co-expressed with SHT in stamen tissues (http://www.genevestigator.ethz.ch/gv/index.jsp; http://bbc.botany.utoronto.ca/ntools/cgi-bin/ntools_expression_angler.cgi), one gene, annotated as encoding a putative methyltransferase (At1g67990), appears to be a good candidate for catalyzing the last methylation step in the biosynthesis of acylated spermidine conjugates. Therefore, we cloned the corresponding cDNA and expressed it in bacteria. The recombinant protein was found to catalyze the methylation of caffeoyl-CoA, trihydroxyferuloyl spermidine and tricaffeoyl spermidine. Methyltransferase activity with the latter compound as substrate is illustrated in Figure 9: after incubation in the presence of the enzyme, tricaffeoyl spermidine was trimethylated into triferuloyl spermidine. Mono- and dimethylated intermediate products were also identified in the incubation medium. Moreover, the function of At1g67990 in spermidine conjugate biosynthesis in vivo was confirmed by analyzing the impact of gene repression in RNAi plants. The tapetum-specific promoter that proved efficient for SHT overexpression (Figure 3c) was used to drive the expression of a hairpin construct that contained an At1g67990 sequence (see Experimental Procedures). Profiling of flower extracts demonstrated a decrease of sinapoylated spermidine conjugate and a correlated accumulation of trihydroxyferuloyl spermidine (Figure 10). These data indicate that the O-methyltransferase is involved in the last step of the biosynthesis of N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine. During the completion of this work, results showing that At1g67990 belongs to the CCoAOMT gene family and methylates various caffeic acid esters and trihydroxyferuloyl spermidine in vitro were independently obtained by another group (Fellenberg et al., 2008). In that study, At1g67990 gene repression in RNAi plants under the control of the CaMV 35S promoter resulted in changes in the relative amounts of the two spermidine conjugates that were comparable to those we observed (Figure 10). These results unequivocally demonstrate that At2g19070 (SHT) and At1g67990 (CCoAOMT) are genes that are tightly co-regulated during anther development and are both implicated in the biosynthesis of spermidine conjugates of Arabidopsis pollen as illustrated in Figure 11.

Figure 9.

O-methyltransferase activity of At1g67990 protein expressed in bacteria.
Tricaffeoyl spermidine synthesized by spermidine hydroxycinnamoyl transferase (SHT) was used as the substrate. The incubation mixture was analyzed by ultra performance liquid chromatography tandem mass spectrometry (UPLC/MS/MS) and ions with m/z values of 632 (tricaffeoyl spermidine), 646 (monomethylated product), 660 (dimethylated product) and 674 (trimethylated product = triferuloyl spermidine) were monitored.
(a) Control without enzyme.
(b) Incubation in the presence of the recombinant enzyme.

Figure 10.

 Comparison of HPLC profiles obtained from flower bud extracts of wild-type (a) and At1g67990 RNAi line (b).
Peaks 1 and 2 were identified as trihydroxyferuloyl spermidine and dihydroxferuloyl-sinapoyl spermidine, respectively, as indicated in Figure 4.

Figure 11.

 A putative scheme of the biosynthetic steps leading to spermidine conjugates in Arabidopsis flowers.
The three acyltransferase reactions catalyzed by spermidine hydroxycinnamoyl transferase (SHT) in vitro are indicated. In vitro and in vivo data have demonstrated that At1g67990 is involved in the last O-methylation step. Other unknown methyltransferases may be involved in O-methylation of tricaffeoyl spermidine since this compound did not accumulate in plants silenced for At1g67990.

Discussion

The results presented here identify At2g19070 as encoding an acyltransferase specifically expressed in tapetum cells of the anthers (Figure 2). This specific pattern of expression has not to our knowledge been previously reported for any BAHD acyltransferase and constitutes a first clue for a function of At2g19070 in pollen development. Moreover, in contrast to other Arabidopsis acyltransferase genes implicated in the biosynthesis of flavonoids or green leaf volatiles that are strongly induced by stress (D’Auria et al., 2007a; Luo et al., 2007), At2g19070 appeared not to be responsive to stress as suggested by array data (Toufighi et al., 2005; Winter et al., 2007; Hruz et al., 2008). This was confirmed by using promoter–GUS transgenics where no change in GUS expression was recorded when leaves were challenged with pathogens or wounded (data not shown).

Anther tissues play a crucial role in pollen development (Owen and Makaroff, 1995). The tapetum cell layers supply the materials necessary for the formation of the pollen cell wall that is built up after the release of haploid microsporocytes in the anther locule. The outer wall of the pollen grain, the exine layer, displays a high degree of mechanical resistance and is, structurally, the most complex type of plant cell wall. Sporopollenin is the major biopolymer of the exine and is actively synthesized by tapetum cells during the free microspore stage. Sporopollenin polymer consists mainly of long-chain fatty acids and phenolic compounds, the exact structure of the molecular network remaining largely unknown (Piffanelli et al., 1998; Blackmore et al., 2007; Morant et al., 2007).

At the late stage of pollen grain development, tryphine is deposited on the surface and within the chambers of the exine and constitutes the pollen coat (Scott et al., 2004; Murphy, 2006; Blackmore et al., 2007). Tryphine contains a range of lipids, glycolipids, flavonoids and proteins which are involved in pollination and pollen–stigma interactions. First, the pollen coat protects pollen from excess desiccation after anther dehiscence. Its adhesive properties allow the pollen to stick to insect vectors for dispersal and to attach to the dry surface of stigmas, where it promotes the hydration necessary for pollen germination and tube growth (Edlund et al., 2004; Murphy, 2006). Unlike the exine, the pollen coat is readily extractable by organic solvents, without altering the intracellular content of the pollen grain (Piffanelli et al., 1998; Murphy, 2006). Here we showed that acylated spermidines were easily extracted from the pollen coat by aqueous methanol (Figure 8), and thus the outermost location of the compounds may offer the pollen grain protection against environmental stresses. For example, acylated spermidines have been reported to have antifungal activity (Walters et al., 2001) and may therefore constitute a shield against pathogen attack. The phenylpropanoid pathway is known to be particularly active in the tapetum cells of the anthers (Herdt et al., 1978; Piffanelli et al., 1998) leading, in particular, to the accumulation of flavonol derivatives. The importance of flavonols in the reproductive tissues appears to be greatly variable depending on plant species: for instance, flavonols are required for maize and petunia fertility (Mo et al., 1992) whereas they are dispensable for Arabidopsis pollen fertility (Burbulis et al., 1996). The accumulation of polyamine conjugates in the reproductive organs of plants has also been associated with fertility (Martin-Tanguy et al., 1982). A N1-spermidine feruloyl-transferase activity has been evidenced in tobacco callus extracts but not characterized at the molecular level (Negrel et al., 1991). Hydroxycinnamic acid spermidines have been found in the pollen of various plant species (Meurer et al., 1988; Bokern et al., 1995), thus suggesting a general role for these phenylpropanoid compounds in pollen function. However, the genes involved in their biosynthesis remained unknown and genetic evidence of their function was lacking until now.

Due to their phenylpropanoid moieties, acylated polyamines display a maximum of absorbance at a wavelength of about 320 nm and may protect plant cells from UV irradiation. A role in UV protection of leaf tissues has been previously proposed for phenolic compounds such as flavonoids and sinapoylmalate. Arabidopsis mutants that are compromised in their ability to produce flavonoids or sinapate esters have been shown to be more susceptible to UV irradiation than wild-type plants (Landry et al., 1995). We were unable to detect any increase of the sensitivity to UV for mutant pollen defective in SHT gene expression and consequently lacking triacylated spermidines. No effect of UV irradiation could be observed on pollen germination and tube growth either in vitro on synthetic medium or in vivo on the flower stigmas (data not shown). Moreover, mutant pollen fertility appeared similar to that of wild-type pollen. Thus, the biological importance of spermidine conjugates in pollen development remains elusive even though microscopic observations have shown some defects in the mutant pollen (Figure 7).

Many genes required for the formation of the pollen wall and pollen coat have been identified (Scott et al., 2004; Blackmore et al., 2007). Male Sterility 1 (MS1) is a transcriptional regulator that is expressed in tapetum and controls a whole set of genes involved in pollen development (Alves-Ferreira et al., 2007; Ito et al., 2007; Yang et al., 2007). Among these genes, it is interesting to note that At2g19070 is transcriptionally regulated by MS1 along with genes encoding enzymes for lipid and phenylpropanoid metabolism, cytochrome P450s and pollen coat proteins. The ms1 pollen is sterile and presents severe aberrations in exine structure that have been linked to defects in the biosynthesis of sporospollenin (Ito et al., 2007; Yang et al., 2007). The ms1 mutant was also shown to display impaired pollen coat development that was thought to be due to compromised lipid synthesis (Yang et al., 2007). Here we have shown that ms1 is unable to synthesize acylated spermidine (Figure 4e), consistent with the repression of the SHT gene and the lack of accumulation of SHT protein. It is noteworthy that, in contrast to ms1, pollen deficient in acylated spermidines as a result of SHT gene knock-out is less profoundly affected in its development and is fully fertile. This suggests that ms1 pollen sterility is due to the pleiotropic effects of this mutation.

The protein encoded by At2g19070 was produced in bacteria and was demonstrated to catalyze the transfer of various hydroxycinnamoyl groups onto the three amine functions of spermidine (Figure 5). Among the different CoA esters that were tested as acyl donors, feruloyl-CoA proved the best substrate, giving rise to N1,N5,N10-triferuloyl spermidine. Mono- and diacylated spermidines were detected in the incubation medium (Figure 6) but in lower amounts compared with triferuloyl spermidine, indicating a rapid acylation rate of the reaction product intermediates. Very low activity was measured with the diamine putrescine or the tetraamine spermine as substrates. The acyltransferase is, therefore, a spermidine hydroxycinnamoyl transferase (SHT). Metabolic profiling of Arabidopsis flowers revealed the occurrence of three spermidine conjugates, the predominant one being identified by LC/MS/MS as N1,N5-dihydroxyferuloyl-N10-sinapoyl spermidine. These findings indicate that, in the biosynthetic pathway of the flower products, the acyltransferase reaction is followed by three hydroxylations at the 5-position of each of the three aromatic rings and by the O-methylation of one hydroxyferuloyl moiety at the N10 or N1 position of the molecule. The O-methyltransferase gene involved in the last biosynthetic step has been identified by demonstrating the impact of its repression on accumulation of spermidine conjugate in flower tissues of RNAi plants (Figure 10), and by showing the capacity of the recombinant protein to methylate acylated spermidine conjugate in vitro (Figure 9). Our current understanding of the biosynthesis of the compounds is summarized in Figure 11. Further work is needed to elucidate the complete biosynthetic pathway.

Experimental procedures

Plant material and culture conditions. Arabidopsis thaliana Columbia 0 (Col-0) and Landsberg erecta ecotypes were grown. For germination, seeds were surface sterilized and placed on MS medium (Duchefa, http://www.duchefa.com/) supplemented with 10 g L−1 sucrose and 10 mg L−1 phosphinotricin or kanamycin if required. After cold treatment for homogenous germination (2 days at 4°C in the dark), the seeds were exposed to 20°C and 70 μmol m−2 sec−1 light intensity under a light/dark cycle of 12 h/12 h. Twelve days later, the plants were transferred to a growth chamber under a light/dark cycle of 16 h/8 h.

Production of transgenic Arabidopsis plants expressing At2g19070 promoter–GUS fusion.  A 2-kb DNA fragment situated upstream of the At2g19070 coding sequence was amplified by PCR from Col-0 genomic DNA using the primers 5′-AGTGGATCCTCTGCTCACCTAACCTAGTCGACA-3′ and 5′-GAGCCCGGGAACACAAACCCCCTTTCTTTTTCT-3′, containing BamHI and XmaI restriction sites, respectively. The PCR product was inserted in pBI101 vector (Clontech, http://www.clontech.com/) and the construct was used to transform Arabidopsis plants.

GUS staining.  For histochemical staining of GUS activity, flower buds were detached and fixed in a solution containing 50 mm sodium phosphate buffer (pH 7) and 1% glutaraldehyde for 10 min. Flower buds were then vacuum infiltrated in 50 mm sodium phosphate buffer (pH 7), containing 1 mm ferrocyanide, 1 mm ferricyanide, 10 mm EDTA, 0.01% Triton X-100 and 1.5 mm X-Gluc substrate for 10 min before 6 h incubation at 37°C. After successive washes with 50%, 70% and 96% ethanol solutions, tissues were directly observed under a stereomicroscope or gradually embedded in Paraplast before transverse sections (10 μm thick) were made with the Leica RM2155 microtome (http://www.leica.com/) and mounted on slides with Eukitt (Electron Microscopy Sciences, http://www.emsdiasum.com/). Observations were made with a Nikon E800 microscope (http://www.nikon.com/).

Epifluorescence, light and scanning electron microscopy.  For epifluorescence microscopy, pollen was deposited between a microscope slide and a cover slide in a droplet of PBS buffer, and observed with a Nikon E800 microscope using a Nikon UV-2A filter with excitation wavelength in the range 330–380 nm. A Leica Macro Fluo stereomicroscope was used for pollen observations at low magnification. For electron scanning microscopy, we used a Hitachi TM 1000 SEM (http://www.hitachi.com/) at an accelerating voltage of 1 or 2 kV, depending on the magnification.

Production of recombinant proteins and polyclonal antibodies.  The full coding sequences of At2g19070 and At1g67990 were cloned in pGEX-KG using NcoI/XhoI and NcoI/HindIII restriction sites, respectively. The inserts were sequenced and the recombinant plasmids were used to transform E. coli BL21-G612. Conditions of expression of the recombinant proteins and purification by affinity chromatography were as previously described (Hoffmann et al., 2003).

The expression in N. benthamiana leaves was carried out using the pBin61 binary vector. At2g19070 coding sequence with six additional histidine codons ahead of the stop codon, was PCR amplified using 5′-AGGGATCCCATGGCTCCCATAACTTTTAGAAA-3′ and 5′-GCCCCCGGGGGAATGATGATGATGATGATGATATCTTCATAAAAG-3′ oligonucleotides as forward and reverse primers and inserted in pBin61 BamHI and XmaI restriction sites. Conditions of expression of the recombinant protein and co-expression of a viral silencing suppressor were as previously described (Voinnet et al., 2003). Leaf tissues were ground in liquid nitrogen and extracted with 50 mm sodium phosphate buffer, pH 8.0, containing 500 mm NaCl and 5 mm imidazole. The recombinant protein was purified by affinity chromatography on Ni beads (HisTrap FF, Amersham Biosciences, http://www1.gelifesciences.com/) following the manufacturer’s instructions and used to raise polyclonal antibodies in rabbit by the procedure described previously (Besseau et al., 2007). Anti-SHT antibodies were used at 1/10 000 dilution in protein blot experiments after overnight pre-incubation with an acetonic powder of Arabidopsis leaves to eliminate any aspecific signal.

Protein gel blot analysis.  The basic procedures for the electrophoresis of proteins under denaturing conditions and immunoblotting were as described previously (Geoffroy et al., 1990), except that phosphatase activity was detected with a chemiluminescent substrate (CDP-Star; Bio-Rad, http://www.bio-rad.com/).

Synthesis of CoA ester and tricaffeoyl spermidine substrates.  The CoA esters were chemically prepared according to published methods (Stockigt and Zenk, 1975; Negrel and Smith, 1984). Tricaffeoyl spermidine was produced by incubating recombinant SHT with caffeoyl-CoA and spermidine, and used as substrate for the O-methyltransferase reaction without purification.

Enzyme activity determination.  The SHT activity was measured in 50 μl 100 mm sodium phosphate buffer (pH 7) containing 1 mm dithiothreitol, 2 mm polyamine substrate, 0.5 mm CoA ester and 1–5 μg of protein. After incubation for 1 h at 30°C, the reaction was stopped by the addition of 15 μl acetonitrile and 1 μl 12 N HCl. The samples were centrifuged at 18 000 g for 5 min and 45 μl were analyzed by HPLC.

The O-methyltransferase activity was assayed with 1 mmS-adenosylmethionine (Adomet) and 0.5 mm CoA ester as substrates in 100 μl 100 mm sodium phosphate buffer (pH 7) containing 1 mm dithiothreitol, 1 mm MgCl2 and 5 μg of protein. After incubation for 1 h at 30°C, samples were processed as described above for SHT and 30 μl was analyzed by HPLC. For testing O-methyltransferase activity with tricaffeoyl spermidine as substrate, tricaffeoyl spermidine was first produced by incubating SHT with spermidine and caffeoyl-CoA in 100 μl final volume under the conditions described above for SHT assay. At the end of SHT incubation, MgCl2 and Adomet were added at 1 mm final concentration and the solution was incubated for an additional 30 min at 30°C in the presence of 5 μg recombinant methyltransferase. At the end of the incubation, 2-μl aliquots were analyzed by ultra performance (UP) LC/MS/MS.

Production of RNAi and SHT overexpressing plants  The At2g19070 and At1g67990 fragments were PCR amplified from pGEX-KG plasmids containing the cognate cDNAs. The sense and antisense primer sequences were 5′-GAGTCTAGACTCGAGGTAACGCCGAGGGAGTGGAA-3′ and 5′-CATGGATCCGGCGCGCCTAGCTTTAAGAGCTTTCAGTTGGA-3′, respectively for SHT, and 5′-GAGTCTAGACTCGAGATTACCTGACAAAGGCATTCTC-3′ and 5′-CTCGGATCCGGCGCGCCTGTGATCAACACCTGCTTTCTT-3′, respectively for CCoAOMT. The resulting PCR products were inserted in pFGC5941 binary vector (http://plantsci.missouri.edu/muptcf/pFGC5941.html) that contains a chalcone synthase intron. Gene fragments were introduced in opposite directions upstream and downstream of the intron using XhoI–AscI and BamHI–XbaI restriction sites. For At1g67990 silencing, the 35S promoter of the pFGC5941 was first removed using EcoRI and XhoI restriction sites and replaced, using the same restriction sites, by the TA29 promoter that had been PCR-amplified from TA29-pGREEN (kindly provided by Dr A. Skirycz, Potsdam, Germany). Gene fragments were then introduced using XhoI–AscI and BamHI–XbaI restriction sites.

For the production of SHT overexpressers, the coding sequence of SHT was PCR amplified from flower cDNA using the following primers: 5′-GAGCCATGGCTCCCATAACTTTTAGAAAATCTTA-3′ (forward), 5′-GAGGGATCCCTAGATATCTTCATAAAAGTGTTTCTTG-3′ (reverse). The sequence was introduced in the pGREEN plasmid downstream of the TA29 promoter using the NcoI and BamHI restriction sites.

Transformation of Arabidopsis (Col-0) was performed with Agrobacterium tumefaciens GV3101 strain by the floral dip method (Clough and Bent, 1998) and transformants were selected from soil-grown plants by spraying 300 mg L−1 Basta (glufosinate) herbicide solution.

Extraction of flower metabolites.  Samples (100 mg) of flower tissues were frozen in liquid nitrogen and quickly ground in 500 μl methanol. After centrifugation at 500 g, the supernatant was collected and the residual pellet was re-extracted with 200 μl of 70% methanol. The supernatants were combined, clarified at 13 000 g for 20 min at 4°C, and analyzed by HPLC or UPLC/MS/MS.

Analysis of flower extracts by HPLC and LC/MS/MS characterization of the metabolites.  For HPLC analysis, phenolic compounds were resolved on a RP C18 column (Novapak, 4 μm, 4.6 × 250 mm; Waters, http://www.waters.com/) using an increasing gradient of acetonitrile in water containing 0.1% formic acid. Gradient conditions at a flow rate of 1 ml min−1 were as follows: 100% solvent A to 50% solvent B for 50 min; 50% solvent B to 100% solvent B for 5 min; 100% solvent B to 100% solvent A for 5 min and then 10 min re-equilibration in 100% solvent A. Solvent A contained acetonitrile/water/formic acid (10:89.9:0.1, v/v/v) and solvent B acetonitrile/water/formic acid (80:19.9:0.1, v/v/v). The UV absorption spectra of compounds were recorded with a photodiode array detector (996 detector, Waters) and fluorescence was measured with a 474 detector (Waters) set on 315 nm for excitation and 405 nm for emission.

For analysis by LC/MS/MS, an Acquity UPLC system (Waters) coupled to a Quattro Premier XE triple Quadrupole MS system (Waters Micromass) was used. Products were resolved on an Acquity UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm) and pre-column (2.1 × 5 mm, 1.7 μm) at 25°C and a flow rate of 0.45 ml min−1. Samples of 3 μl were injected and an increasing gradient of acetonitrile in water and containing 0.1% formic acid was used for elution. Gradient conditions were as follows: 3–20% acetonitrile for 17 min; 20–40% for 20 min; 100% for 3 min and 3% for 5 min.

The mass spectrometer was run using the Mass-Lynx software. The electrospray ionization source conditions in positive and negative modes were optimized by direct infusion of kaempferol-3-glucoside and sinapic acid solutions at 5 μl min−1, mixed with the mobile phase through a T-piece. Nitrogen was used as the nebulizer gas and for desolvation at flow rates of 50 and 900 L h−1, respectively. The source capillary voltage was set to 3 kV and temperature to 135°C. Desolvation was performed at 400°C. For all compounds, the cone tension was optimized at 25 V. For collision-induced dissociation, argon was used as the collision gas at a pressure of 3 × 10−3 mbar. Full-scan (100–900 m/z in negative mode, 230–900 m/z in positive mode), selected ion monitoring (SIR), daughter scan and multiple reaction monitoring modes were used for analysis.

Acknowledgements

Thanks are due to Drs D. Werck-Reichhart and M. Matsuno for valuable scientific discussions and for sharing results before publication. We thank Drs P. Constabel and K. Richards for careful reading of the manuscript and Drs A. Skirycz and B. Mueller-Roeber (Postdam, Germany) for providing the tapetum-specific promoter. The assistance of D. Meyer in histochemical analysis, M. Alioua in DNA sequencing, and Dr M. Erhardt for electronic microscopy is gratefully acknowledged. We are grateful to the Salk Genomic Analysis Laboratory (La Jolla, CA, USA) for providing the T-DNA mutant and to the Nottingham Arabidopsis Stock Centre (UK) for distributing the seeds. The UPLC/MS/MS system was co-financed by the Centre National de la Recherche Scientifique, the Université Louis Pasteur, the Région Alsace, INRA and Tepral Company. This work was supported by doctoral fellowships of the Ministère de l’Education Nationale, de l’Enseignement Supérieur et de la Recherche to EG and SB.

Ancillary