Introduction

Caffeoyl coenzyme A O-methyltransferases (CCoAOMTs) and caffeic acid O-methyltransferases (COMTs) are among the two most intensively investigated enzymes in plant secondary metabolism. This is due to their established role in lignin biosynthesis, both in gymno- and in angiosperms (Dixon and Reddy 2003; Davin et al. 2008). The cation-dependent CCoAOMT is considered specific for the methylation of caffeoyl CoA, a key metabolite in the biosynthesis of guaiacyl lignin, whereas the cation-independent COMT, due its preference for 5-hydroxyferulate-derived metabolites, in particular 5-hydroxyconiferylaldehyde, appears tightly associated with syringyl (S)-lignin (Li et al. 2000; Humphreys and Chapple 2002; Boerjan et al. 2003; Davin et al. 2008). S-Lignin is characterized by incorporation of aromatics with a higher phenolic substitution pattern. Double knockouts of Arabidopsis for both types of enzymes encoded by the genes At4g34050 (CCoAOMT1) and At5g54160 (COMT1, also termed AtOMT1 in Arabidopsis) result in a severe dwarf phenotype in a variety of plants without the development of reproductive organs (Do et al. 2007). Although a reduction in total lignin content and shifts in the ratio of guaiacyl to syringyl moieties are observed in single comt1 or ccoaomt1 plants of several species, both enzymes to some extent functionally complement each other, resulting in fertile plants with only minor differences with respect to the overall phenotype compared to wild type plants (Guo et al. 2001; Chen et al. 2006; Day et al. 2009).

Besides their established role in lignin formation, the corresponding enzymes COMT1 and CCoAOMT1 also methylate other soluble phenolic metabolites. COMT1 plays a role in the methylation of flavonols with ortho-dihydroxy groups, such as quercetin, resulting in accumulation of isorhamnetin glycosides in Arabidopsis flowers (Muzac et al. 2000; Goujon et al. 2003; Do et al. 2007; Matsuda et al. 2009). CCoAOMT1 contributes to the formation of soluble sinapoyl conjugates in Arabidopsis leaves (Do et al. 2007) and is essential for the biosynthesis of the coumarin scopoletin in Arabidopsis roots (Kai et al. 2008). Its role in the biosynthesis of acetosyringone, an early inducer of the Agrobacterium tumefaciens-derived vir-genes has also been proposed (Maury et al. 2010). Therefore, ccoaomt1 but also comt1 plants appear less susceptible to this pathogen. In all biosynthetic pathways investigated up to now, the role of CCoAOMT1 seems confined to the methylation of caffeoyl CoA only, consistent with its reported high substrate specificity for this compound in vitro (Pakusch et al. 1989).

Several paralogs encoding CCoAOMT (and COMT)-like enzymes are present in the genomes of dicots and monocots (Peng et al. 2010). CCoAOMT paralogs are characterized by lower sequence identities and by a broader substrate specificity than the CCoAOMT1-subtype, they catalyse the formation of 3′- and 5′-O-methylated flavonoids in epidermal tissues of the ice-plant, Mesembryanthemum crystallinum (Ibdah et al. 2003) or methylate anthocyanidin glycosides in several grape varieties (Hugueney et al. 2009; Lücker et al. 2010). One orthologue from Arabidopsis, AtTSM1, in a final methylation step modifies tris-5-hydroxyferuloyl spermidine to the N 5-mono-sinapoyl derivative (Fellenberg et al. 2008; Matsuno et al. 2009; Handrick et al. 2010). As in case of the attsm1 knockdowns, flowers of ccoaomt1 plants show a severe, but not a total reduction of hydroxycinnamic acid spermidine conjugates (HCAAs) (Fellenberg et al. 2009). These observations suggest a participation of CCoAOMT1 early in this pathway, catalysing the step from caffeoyl CoA to feruloyl CoA before conjugation to spermidine by a BAHD-like transferase, termed SHT (Grienenberger et al. 2009). A brief description of the corresponding pathway is illustrated in Fig. 1.

Fig. 1
figure 1

The late steps of the major biosynthetic pathway leading to the accumulation of HCAAs in anthers and pollen grains of wild type Arabidopsis flowers. Individual enzymes are marked

Full size image

Recent information from OMT-fingerprinting of Arabidopsis mutant and wild type flowers by affinity-based protein profiling (ABPP) demonstrated that besides CCoAOMT1 and AtTSM1, COMT1 is also constitutively present in high amounts during flower bud development (Wirsing et al. 2011). Therefore, COMT1 could theoretically compensate for the loss of CCoAOMT1 and/or AtTSM1 in Arabidopsis anthers, and participate in the methylation of residual HCAAs observed in the corresponding knockout and knockdowns, respectively.

By the use of specific polyclonal antibodies raised against recombinant Arabidopsis COMT1 and CCoAOMT1 and immunofluorescence microscopy, we characterize their distribution in stamens of Arabidopsis wild type and knockout lines and propose a role for both enzymes. These data are supported by qPCR and phenolic profiling data of isolated pollen grains.

Materials and methods

Plant material and culture conditions

Arabidopsis thaliana (Columbia 1092) and all knockout mutants (GK007F02.01 for the At4g34050 gene encoding CCoAOMT1 and SALK_135290 for the At5g54160 gene encoding COMT1) were obtained from the European Arabidopsis Stock Center (http://www.nasc.ac.uk) (Sessions et al. 2002; Alonso et al. 2003). Wild type and mutant plants were grown in fully climatized green houses at 22°C (day) and 18°C (night) under long day conditions with a 16-h-light/8-h-dark cycle. Homozygous ccoaomt1, comt1 and ccoaomt1/comt1 double mutant lines were isolated as published previously by Do et al. (2007). Pollen grains were harvested according to a method by Johnson-Brosseau and McCormick (2004) with slight modifications (Handrick et al. 2010) using polyester filters (NeoLab, Heidelberg, Germany) of 40 μm to remove the impurities and 11 μm to collect the purified pollen grains.

Real-time qPCR

Total RNA was isolated using a classical phenol/chloroform extraction, selective LiCl precipitation and 80% EtOH washing of the resultant RNA (Chomzynski and Sacchi 1987). Extracted RNA was treated with DNaseI. The quality and quantity was checked photometrically by the 260/280 nm ratio and agarose gel electrophoresis. cDNA was synthesized using Superscript III reverse transcriptase (Invitrogen) and was used as a template in qPCR reactions, whereas DNaseI-treated RNA was used as a control. Quantitative real-time RT PCR (qRT-PCR) was performed using gene-specific primers (Supplemental Table S1) with the fluorescent dye SYBR Green master mix (Applied Biosystems) according to the manufacturer’s instructions, and was run on a Multi Quant Mx3005 P (Stratagene). Non-specific product formation was excluded by the determination of melting curves. All quantifications were normalized to the 65-kDa regulatory subunit of protein phosphatase 2A (pp2A, At1g13320) using primers described previously (Fellenberg et al. 2009). Data were analysed with MxPro qPCR software (Stratagene) and EXCEL (Microsoft).

Production of recombinant proteins and polyclonal antibodies

For heterologous expression, the full length cDNAs of At4g34050, At5g54160 and At4g26220 were cloned in frame into the BamHI/HindIII site of the p QE 30 vector (Qiagen) and transformed into E. coli M15p[Rep4]. The resultant recombinant CCoAOMT1, COMT1, and CCoAOMT7 contained an N-terminal His-Tag. After purification by metal affinity chromatography, 1 mg of each protein was used to raise polyclonal antibodies in rabbits (Eurogentec, Genay, France). Antisera of the second bleed were used without further purification for western blots and immunolocalization.

Protein extraction and immunolabelling

For the isolation of total protein, 100 mg of plant material was homogenized in liquid nitrogen and transferred into 1 ml of buffer A [20 mM NaCl, 20 mM tris/HCl, pH 7.8, 5% (w/v) Polyclar AT (Serva, Heidelberg, Germany)], 1 mM EDTA, 10 μl ml−1 protease inhibitor (Sigma-Aldrich) and then incubated on a shaker for 20 min at 37°C. Samples were centrifuged for 5 min at 17,000g and the soluble protein of the supernatant was precipitated with 250 μl 30% (v/v) trichloroacetic acid. Total protein was centrifuged at 4°C and 17,000g for 5 min and the pellet was washed two times with ice-cold 80% (v/v) acetone and once with 100% acetone. The pellet was air-dried and resuspended in 200 μl buffer B [7 M urea, 2 M thiourea, 4% (w/v) CHAPS]. For SDS-PAGE the solubilized protein was mixed with SDS-PAGE sample buffer (Laemmli 1970) in a 1:1 ratio and was heated to 65°C for 10 min.

For western blots, denatured proteins were run on 14% SDS-PAGE gels (Laemmli 1970) and transferred onto a 0.22 μm nitrocellulose-membrane using NuSieve-Buffer and NuSieve-Antioxidant (Invitrogen) according to the manufacturer’s instructions. Membranes were blocked with 5% non-fat dry milk for 2 h, and incubated with the polyclonal anti-CCoAOMT1-antibody or anti-COMT1-antibody 1:1,000 in blocking solution over night. The first antibody was subsequently washed off and the membrane incubated for 1 h with rabbit alkaline phosphatase coupled anti-IgG antibody in blocking solution (1:5,000; Sigma-Aldrich), washed, and subsequently incubated with bromo-chloro-indolyl phosphate/nitroblue tetrazolium (BCIP/NBT; Sigma-Aldrich) for visualization. Controls were performed with rabbit pre-immune serum. Specificity and sensitivity of the polyclonal antisera were recorded with crude plant extracts as well as various recombinant OMTs from the lab collection at protein concentrations from 10 μg to 1 ng.

For immunolocalization, plant material was fixed in 4% (w/v) para-formaldehyde, 0.1% (v/v) Triton X-100 in PBS, dehydrated by a graded series of EtOH and embedded in PEG 1500 (Isayenkov et al. 2005). Sections of 2 μm thickness were blocked with 5% (w/v) BSA in PBS, incubated with the polyclonal antibody (1:1,000) overnight, washed and incubated with a goat anti-rabbit IgG antibody covalently linked to Alexa Fluor 488 (Invitrogen). Microscopy was performed on a fluorescence microscope AxioImager (Zeiss, Jena, Germany) using the proper filter combination. Micrographs were taken by AxioCam Mrc (Zeiss).

(U)HPLC(–MS), and HPTLC-based analyses of flower and pollen metabolites

Arabidopsis flowers and pollen grains were analysed from 80% (v/v) methanol extracts (15 mg tissue/100 μl 80% methanol). Plant material was homogenized in liquid nitrogen and suspended in 80% methanol in Eppendorf tubes under shaking for 30 min at RT. 5 mg of purified wild type and mutant pollen were extracted into 100 μl of 80% MeOH. Extracts were centrifuged for 2 min at 18,000g and the supernatant subsequently analysed on RP-HPLC on a 12.5 cm, 4 mm i.d., 5 μM Nucleosil C18-column (Macherey–Nagel, Düren, Germany) at a flow rate of 1 ml/min with a gradient from 5% (v/v) B (acetonitrile, 0.1% trifluoroacetic acid) in A (water, 0.1% trifluoroacetic acid) to 35% (v/v) B in A within 40 min. MaxPlot detection was performed between 220 and 600 nm. LC–MS analysis was performed on an Applied Biosystems 3200 Q-Trap® LC/MS/MS system hybrid QqQ-Lit mass spectrometer equipped with an ESI-Turbolon Spray™ interface operating in a positive ion mode. Parameters were as follows: ion spray voltage 3,500 V, nebulizing gas 40 psi, source temperature 450°C, drying gas 40 psi, curtain gas 25 psi. Hydrolysis of glycosides and metabolite extracts was performed by addition of 1 N HCl/50% MeOH at 80°C for 1 h, subsequent extraction of the resultant aglycone in ethylacetate and identification of quercetin and isorhamnetin with standard compounds obtained from Roth (Mannheim, Germany).

Results

Localization of CCoAOMT1 and COMT1 in flower organs

Ccoaomt1 as well as attsm1 knockdowns in all cases showed a severe, but not a complete reduction of methylated HCAAs (Fellenberg et al. 2009). Specifically, the observation of considerable amounts of tris-5-hydroxyferuloyl spermidine required the action of an OMT earlier in the pathway from caffeoyl CoA to feruloyl CoA, usually performed by CCoAOMT1. Based on similar substrate and position specificity in vitro and its reported role in lignin biosynthesis, COMT1 would be the ideal candidate.

Initial qPCR data of wild type plants suggested that the corresponding COMT1 transcript, similar to CCoAOMT1 was present in flower buds and open flowers (Fig. 2). We hypothesized that COMT1 could be present in Arabidopsis stamens at the site of HCA-conjugate biosynthesis. PCR performed from cDNA of isolated Arabidopsis stamens showed strong signals for the COMT1 transcript (data not shown). To more specifically localize the enzyme, polyclonal antibodies were raised against recombinant Arabidopsis COMT1 and also against CCoAOMT1, since a ccoaomt1 line investigated previously showed a strong reduction in HCAA formation (Fellenberg et al. 2009) and therefore, it is assumed that this enzyme is localized in the tapetum. The antibodies were tested with recombinant enzymes and with crude extracts of Arabidopsis wild type as well as comt1 and ccoaomt1 mutant flower buds.

Fig. 2
figure 2

Quantitative real-time PCR data of relative transcript levels of the At5g54160 gene encoding COMT1 (a), At4g34050 encoding CCoAOMT1 (b), and At1g67990 encoding AtTSM1 (c) in different Arabidopsis organs of fully developed 8-week-old flowering plants. The small subunit of pp2A was used as a reference gene

Full size image

Both antibodies on western blots showed the required specificity and sensitivity to be further used for immunolocalization studies. Crude protein preparations from Arabidopsis wild type flower buds showed the expected signals at 40 and 29 kDa when tested with a mixture of anti-COMT1 and anti-CCoAOMT1 antibodies (Fig. 3). Ccoaomt1 and comt1 knockout plants lacked the expected signals at 29 and 40 kDa, respectively (Fig. 3). No signals and therefore no cross-reaction with the abundant AtTSM1 (Fellenberg et al. 2008) were detected in these blots around 26 kDa. Down to one nanogram of recombinant enzyme could be detected specifically in case of the anti-COMT1 antibody (Supplemental Fig. S1). These data prove the sensitivity and selectivity of both antibodies. These protein data are consistent with the lack of any PCR signals of both full length transcripts in the corresponding knockout mutants (Fig. 4).

Fig. 3
figure 3

SDS-PAGE (a) and corresponding immunoblot (b) of COMT1 (40 kDa) and CCoAOMT1 (29 kDa) in flower buds of wild type (lane 1), ccoaomt1 (lane 2) and comt1 (lane 3) plants. Total proteins (15 μg each) were separated by SDS-PAGE and blotted onto nitrocellulose, immunolabelled by a combination of anti-CCoAOMT1 and anti-COMT1 polyclonal antibodies and visualized by an alkaline phosphatase-conjugated secondary antibody

Full size image
Fig. 4
figure 4

PCR of full length transcripts of flowers from wild type plants and ccoaomt1 (a) and comt1 (b) indicating the absence of any full length transcript in the individual knockout lines. M molecular mass marker (1,000 bp). The small subunit of pp2A was used as a control

Full size image

When flower buds were analysed by immunofluorescence microscopy with antibodies raised against CCoAOMT1 or COMT1 for organ and tissue-specific distribution, the results differ drastically (Figs. 5, 6). Whereas significant signals for CCoAOMT1 were observed in the tapetum of young flower buds (Fig. 5), there is no such signal for COMT1 (Fig. 6). Instead, strong signals are observed in the endothecium and the epidermal anther, both directly adjacent to the tapetum (Fig. 6). This shows that COMT1 is excluded from the tapetum and therefore is unlikely to be involved in the formation of HCAAs. Instead, the enzyme is prominent in not only all epidermal tissues of flower organs, specifically the petals (Fig. 6b), but also in sepals and the tip of the stigma, the latter is not shown. In this context, it is noteworthy that in contrast to COMT1, CCoAOMT1 cannot be detected in the endothecium (Fig. 6), although the secondary walls of this cell layer in Arabidopsis are lignified and have striated patterns similar to tracheary elements (Mitsuda et al. 2005). The presence of additional CCoAOMT1 signals in the vascular systems of all flower organs, including stigma, stamens, petals, and sepals is expected (Fig. 5) and consistent with CCoAOMT1-promoter-driven GUS-expression data (Do et al. 2007). As in the case of CCoAOMT1, signals for COMT1 were also observed in the vascular bundles of the stigma, but are less intense in stamens and sepals (Fig. 6). In summary, localization of CCoAOMT1 in the tapetum of young flower buds corroborates its role in HCA-spermidine conjugate formation. In contrast, COMT1 cannot be localized in the tapetum and its contribution to phenylpropanoid metabolism in this cell layer therefore seems unlikely. Significantly reduced CCoAOMT1 signals in the tapetum of old flower buds just before the petals appear suggest that this enzyme is degraded rapidly at later stages of flower development during anther dehiscence. In contrast, signal intensities of COMT1 persist even during later stages of flower development (data not shown).

Fig. 5
figure 5

Immunolocalization of CCoAOMT1 in flower buds of Arabidopsis. Sections of 2 μm thickness were immunolabelled with an anti-CCoAOMT1 polyclonal antibody and visualized by a secondary antibody conjugated to Alexa Fluor 488. Signals were observed in the tapetum (ta) and in vascular bundles (vb) of the stigma and stamen connective. Cross-sections through the whole flower bud (a) and pre-immune control (b)

Full size image
Fig. 6
figure 6

Immunolocalization of COMT1 in flower buds of Arabidopsis. Cross-sections of 2 μm thickness were immunolabelled with an anti-COMT1 polyclonal antibody and visualized by a secondary antibody conjugated to Alexa Fluor 488. No signals for COMT1 were observed in the tapetum (ta), but in the endothecium (en) and the epidermis (ep) of the anther. Cross-section of an anther (a), cross-section through a whole flower bud (b). COMT1 is detectable in epidermal tissues of stigma (st), anther (an), petal (pe), and sepal (se) as well as vascular tissue of the stigma. A cross-section of a comt1 anther indicates that the greenish fluorescence in pollen grains is not due to COMT1 signal (c), pre-immune control (d)

Full size image

Phenylpropanoid pattern of wild type, comt1, and ccoaomt1 anthers

The cell type-specific distribution of COMT1 has important consequences for the organ-specific metabolite pattern. If COMT1, in contrast to CCoAOMT1, does not participate in the methylation of HCAAs, pollen grains of comt1 lines should not show any effect on HCAAs. This is indeed the case. Phenylpropanoid pattern of wild type and comt1 are virtually identical, whereas a reduction in HCAAs is observed in pollen of the ccoaomt1 (Fig. 7). The HCAAs pattern of pollen from comt1/ccoaomt1 double mutants was identical to the pattern of the single ccoaomt1 line (data not shown). As in case of the HCAAs, the flavonoid profile of pollen grains is essentially determined by the tapetum (Hsieh and Huang 2007). If COMT1 is absent from the tapetum, the pollen grains should also not contain any isorhamnetin (quercetin-3′-O-methyl ether) glycosides, neither in the wild type nor in the comt1 plants. COMT1 was shown to convert quercetin efficiently into isorhamnetin in vitro (Muzac et al. 2000). UHPLC–MS data as well as HPLC–DAD and HPTLC analysis followed by hydrolysis of pollen extracts show that the flavonoid pattern of Arabidopsis pollen in contrast to flowers is fairly simple. It consists of only two major flavonol 3-O-diglucosides, one kaempferol and one quercetin glucoside, neither of which are methylated (Fig. 7; Supplemental Table S2). This flavonoid pattern is virtually identical in all lines investigated. No isorhamnetin glycosides were identified consistent with the absence of COMT1 from the tapetum. The same pollen flavonol diglucosides of Arabidopsis wild type pollen have recently also been characterized independently by Stracke et al. (2010). Among other flavonoids, both diglycosides can also be localized in HPLC chromatograms of whole flower bud extracts (Supplemental Fig. S2) where they occur in the same amounts and ratios than in pollen extracts of wild type, ccoaomt1, and comt1 flowers. This indicates that they are restricted to pollen grains and are not found in any other flower organ.

Fig. 7
figure 7

HPLC chromatograms of Arabidopsis pollen grains at 350 nm. Methanolic extracts (1 mg/20 μl 80% MeOH) of pollen grains were prepared and analysed by RP-HPLC. 1 quercetin-3-O-diglucoside, 2 kaempferol-3-O-diglucoside, 3 N 1,N 5,N 10-tris-(5-hydroxyferuloyl)-spermidine, 4 N 1,N 10-bis-(5-hydroxyferuloyl)-N 5-sinapoyl-spermidine. Wild type pollen grains (a), comt1 pollen grains (b), ccoaomt1 pollen grains (c)

Full size image

In summary, COMT1 is absent from the tapetum, and does not contribute to the phenylpropanoid pattern of the pollen tryphine (Ariizumi and Toriyama 2011). Strong signals in epidermal layers of flower buds support the perception that in flowers the enzyme is largely confined to isorhamnetin glycoside biosynthesis (Muzac et al. 2000; Matsuda et al. 2009).

Which enzyme compensates for CCoAOMT1 in the ccoaomt1 line?

Where do the remaining methylated HCAAs, like the feruloyl and sinapoyl derivatives in the tapetum of ccoaomt1 flower buds come from (Fellenberg et al. 2009), i.e. which enzyme is able to methylate caffeoyl CoA to feruloyl CoA? Barely measurable transcript levels of the genes At4g26220, At1g67980, At3g61990, At3g62000 and At4g24375, encoding five remaining CCoAOMT1-ortho- and paralogs in Arabidopsis flowers, virtually exclude these genes to participate in methylation of caffeoyl CoA in flowers (Fig. 8). Among these, the most prominent, At4g26220, encoding CCoAOMT7 is expressed poorly in flower buds and shows the highest transcript abundance in shoots (Supplemental Fig. S3; Raes et al. 2003). Nevertheless, in contrast to all other candidates, this gene was slightly, although not significantly up-regulated in ccoaomt1 flower buds (Fig. 8). Therefore, polyclonal, specific antibodies were also raised against recombinant CCoAOMT7. However, these did not show any signal in anthers or flowers of wild type and knockout mutants, although recombinant CCoAOMT7 could be detected down to about 10 ng by SDS-PAGE (data not shown). Due to its very low overall abundance in flower buds a significant contribution of CCoAOMT7 in HCAA formation can likely be excluded and its function in some aspect of lignin formation during Arabidopsis shoot development appears more likely (Minic et al. 2009).

Fig. 8
figure 8

Transcription levels of COMT1 and all CCoAOMT-like genes in Arabidopsis wild type and ccoaomt1 mutants. At4g34050 (a), At5g54160 (b), At1g67990 (c), At4g26220 (d), At3g61990 (e), At3g62000 (f), At1g67980 (g), At1g24735 (h). Notice the low expression levels of At1g67980, At1g61990, At1g62000, At4g26220, and At1g24735 compared to At4g34050, At5g54160, and At1g67990

Full size image

Discussion

A clear functional and spatial separation of COMT1 and CCoAOMT1 is evident in Arabidopsis anthers. Only CCoAOMT1 contributes to the HCAA-profile in the tapetum, methylating the precursor caffeoyl CoA. The same reaction in vascular tissues is required for the methylation of lignin precursors, where CCoAOMT1 is sufficient for the guaiacyl lignin monomer methylation. The contribution of CCoAOMT1 to the HCAA-methylation in anthers, based on its preference for caffeoyl CoA, therefore is not unusual and the drastic reduction of HCAAs in ccoaomt1 plants illustrates the requirement of this enzyme for their biosynthesis. COMT1 shows highest protein levels in epidermal tissues of all flower organs contributing to the flavonoid profile of these organs as suggested previously (Muzac et al. 2000). The enzyme cannot be detected in the tapetum, consistent with the characteristic unmethylated flavonol profile found in the tryphine of analysed pollen grains (Stracke et al. 2010; Fellenberg and Vogt, unpublished). In contrast to the formation of syringyl lignin, where COMT1 specifically methylates 5-hydroxyconiferylaldehyde, a tapetum-specific CCoAOMT1-like enzyme, AtTSM1 performs the required methylation of 5-hydroxyferulate moieties generated by a tapetum-specific cytochrome P450 monooxygenase, Cyp98A8 (Matsuno et al. 2009). The combination of these two tissue specific enzymes in an evolutionary young pathway adapted to pollen wall composition and function (Matsuno et al. 2009) replaces the ferulic acid 5-hydroxylase (F5H) and COMT1 enzyme combination generating sinapoyl signatures in lignin monomer formation. Interestingly, 5-hydroxylation, preceding methylation by COMT1 in lignin biosynthesis also is a relative advanced feature, initiated independently of guaiacyl lignin formation by a set of specific transcription factors NST1/SND1 regulating F5H (Zhao et al. 2010). Apparently, the formation of feruloyl units is highly conserved, whereas plants use different sets of enzymes to synthesize sinapoyl units.

The drastic reduction of the mono-sinapoyl HCAA (peak 4, Fig. 7) compared to virtually unchanged amounts of 5-hydroxyferuloylspermidine (peak 3, Fig. 7) in anthers and pollen of ccoaomt1 plants compared to wild type plants is unexpected, since, at a first glance, both peaks should be influenced equally by the knock out. No pile up of 5-hydroxyferuloylspermidine is observed as in case of the attsm1 knock-down plants, which are deficient in the final methylation step (Fellenberg et al. 2008). Currently, any explanation for this observation can only be speculative. Previous data from ccoaomt1 plants have shown that there is no transcriptional feedback inhibition on the late pathway genes, including AtTSM1, Cyp98A8 and SHT (Fellenberg et al. 2009). It is possible that all enzymes associated with this pathway are organized in a metabolon, as proposed for various pathways in plant natural product biosynthesis (Jørgensen et al. 2005). If CCoAOMT1 is critical for the functional integrity of this metabolon, a loss of this enzyme may result in an impaired channelling of earlier intermediates and therefore reduce the total flux through this pathway. The apparent lack of CCoAOMT1 could then partially be compensated for by AtTSM1, methylating caffeoyl CoA before or after conjugation to spermidine. The substrate profile of AtTSM1 in vitro (Fellenberg et al. 2008) and its high transcript levels in the tapetum of young flower buds are consistent with this assumption and may explain the similar, residual levels of both methylated HCAAs. Likewise, it was shown that the acyltransferase, SHT, also accepts caffeoyl CoA in vitro although it shows highest activity towards feruloyl CoA (Grienenberger et al. 2009). Therefore, the proposed order of methylation and hydroxylation steps could be affected and result in a reduced but not complete lack of tris-conjugated products. Alternatively, an insufficient compensation of the CCoAOMT1 loss by AtTSM1 could lead to a reduced pool of feruloyl CoA precursors resulting in lower amounts of HCAAs and, in parallel, also to an inefficient conversion of tris-5-hydroxyferuloyl spermidine by AtTSM1. Therefore, a direct effect on product channelling appears likely but may require a more thorough characterization of the metabolic grid of enzymatic steps leading to the extremely complex pattern of recently characterized soluble HCAA conjugates (Handrick et al. 2010).

Apart from AtTSM1, CCoAOMT1, and CCoAOMT7 the four remaining genes encoding CCoAOMT-like enzymes are unlikely to play any role in HCAA methylation. At3g61990 is expressed only in roots, the corresponding OMT has a reported membrane-spanning N-terminal sequence, and was recently shown to be involved in methylation of Glu6 in the N-terminal loop of AtPIP2 aquaporins specifically in roots (Sahr et al. 2010). Transcripts of At1g67980, At3g62000, and specifically At4g24375 are virtually absent in Arabidopsis flowers under standard greenhouse conditions (Fig. 8). Therefore, the proposed dual functional specificity of AtTSM1 in the mutant line appears most plausible and is also not unusual. Two steps of isoflavonoid O-methylation in Pisum sativum are also performed by a single enzyme (Liu et al. 2006). Unfortunately, a double knockout, attsm1 combined with ccoaomt1 which could provide further proof of this assumption, cannot be tested since only RNAi-plants and no knockout mutants are available for this gene (Fellenberg et al. 2008). A 50% reduction of RNAi-transcripts, as observed in some of these knock-down plants will not be sufficient to eliminate residual HCAA levels completely and provide unequivocal proof of the role of AtTSM1.