Convergent evolution in the BAHD family of acyl transferases: identification and characterization of anthocyanin acyl transferases from Arabidopsis thaliana


(fax +44 1603 450045; email


Members of the BAHD family of plant acyl transferases are very versatile catalytically, and are thought to be able to evolve new substrate specificities rapidly. Acylation of anthocyanins occurs in many plant species and affects anthocyanin stability and light absorption in solution. The versatility of BAHD acyl transferases makes it difficult to identify genes encoding enzymes with defined substrate specificities on the basis of structural homology to genes of known catalytic function alone. Consequently, we have used a modification to standard functional genomics strategies, incorporating co-expression profiling with anthocyanin accumulation, to identify genes encoding three anthocyanin acyl transferases from Arabidopsis thaliana. We show that the activities of these enzymes influence the stability of anthocyanins at neutral pH, and some acylations also affect the anthocyanin absorption maxima. These properties make the BAHD acyl transferases suitable tools for engineering anthocyanins for an improved range of biotechnological applications.


Anthocyanins are water-soluble pigments that that are produced by almost all vascular plants and are the major contributors to the orange, red, purple and blue colours of many flowers, leaves and fruits (Grotewold, 2006). In addition to their important roles in attracting pollinators and protecting plants against various stresses (Harborne and Williams, 2000), interest in these compounds has intensified because of their health-promoting properties as antioxidants, and for their anti-mutagenic and other protective activities (Galvano et al., 2004; Lamy et al., 2006). In many plant species, anthocyanins are modified by glycosylation, acylation or methylation, which may alter their light-absorption properties, their stability in solution, their bioavailability in the diet and their abilities to interact with other molecules (a process known as co-pigmentation;(Tanaka and Brugliera, 2006). There are two major types of acylation – aromatic acylation and aliphatic acylation. Aromatic acylation usually involves the addition of coumaroyl or sinapoyl groups and is proposed to stabilize and intensify blue tones of anthocyanins, whereas aliphatic acylation (normally malonylation) is important for water solubility, stability and resistance to enzymic degradation (Strack and Wray, 1994).

Synthesis of acylated anthocyanins requires the activity of anthocyanin acyltransferases, a group of flavonoid-specific, acyl CoA-dependent acyltransferases. The anthocyanin acyltransferases belong to the BAHD acyltransferase family, which was named based on the first letter of each of the first four enzymes characterized in this family (BEAT, AHCTs, HCBT and DAT) (St Pierre and De Luca, 2000). So far, cDNAs encoding two aromatic acyltransferases [Gt5AT from Gentiana triflora (Fujiwara et al., 1998) and Pf3AT from Perilla frutescens (Yonekura-Sakakibara et al., 2000)] and six aliphatic acyltransferases [Ss5MaT1 and Ss5MaT2 from Salvia splendens (Suzuki et al., 2001, 2004b), Dv3MaT from Dahlia variabilis (Suzuki et al., 2002), Sc3MaT from Senecio cruentus (Suzuki et al., 2003), and Dm3MaT1 and Dm3MaT2 from chrysanthemum (Dendranthema × morifolium) (Suzuki et al., 2004a)] have been cloned and characterized functionally (see Table 1). However, considering the range of acylated anthocyanins found in various plant species, a wide variety of anthocyanin acyltransferases with various specificities are yet to be isolated (Nakayama et al., 2003).

Table 1.   Cloned anthocyanin acyltransferases in plants
EnzymeShort nameSpecies of originReference
Aromatic anthocyanin acyltransferases
 Hydroxycinnamoyl CoA:anthocyanin 5-O-glucoside-6′′′-O-acyltransferaseGt5ATGentiana trifloraFujiwara et al. (1998)
 Hydroxycinnamoyl CoA:anthocyanin 3-O-glucoside-6′′-O-acyltransferasePf3ATPerilla frutescensYonekura-Sakakibara et al. (2000)
Aliphatic anthocyanin acyltransferases
 Malonyl CoA:anthocyanin 5-O-glucoside-6′′′-O-malonyltransferaseSs5MaT1Salvia splendensSuzuki et al. (2001)
 Malonyl CoA:anthocyanin 5-O-glucoside-4′′′-O-malonyltransferaseSs5MaT2S. splendensSuzuki et al. (2004b)
 Malonyl CoA:anthocyanin 3-O-glucoside-6′′-O-malonyltransferaseDv3MaTDahlia variabilisSuzuki et al. (2002)
 Malonyl CoA:anthocyanin 3-O-glucoside-6′′-O-malonyltransferaseSc3MaTSenecio cruentusSuzuki et al. (2003)
 Malonyl CoA:anthocyanin 3-O-glucoside-6′′-O-malonyltransferaseDm3MaT1Dendranthema × morifoliumSuzuki et al. (2004a)
 Malonyl CoA:anthocyanin 3-O-glucoside-3′′,6′′-O-dimalonyltransferaseDm3MaT2Dendranthema × morifoliumSuzuki et al. (2004a)

The conventional way to identify genes encoding defined activities in new species and the basis of gene sequence annotation is through sequence homology, especially in species with sequenced genomes (Hartmann et al., 2006; Hilson, 2006; Vij et al., 2006). However, in the case of the BAHD enzymes, there are problems associated with this approach. BAHD enzymes are very versatile in their substrate specificities, and enzymes with remarkably similar substrate specificities may have evolved independently. This has been demonstrated most effectively for two malonyl CoA anthocyanin acyl transferases in S. splendens. One transfers the malonyl group to the 6′′′ position of the glucose in cyanidin 5-O-glucoside, whereas the second transfers the malonyl group to the 4′′′ position on the sugar (Suzuki et al., 2001, 2004b). These two anthocyanin acyl transferases are separated phylogenetically, and belong to quite separate clades of the BAHD tree (see Figure S1), indicating that, despite their similar substrate specificities, they must have evolved from different ancestral proteins. This versatility complicates the prediction of function based on structural information alone, and necessitates the development of additional criteria for genomics-based predictions of function.

There are estimated to be about 88 genes encoding members of the BAHD family in Arabidopsis (Gang, 2005). So far, only three of them have been characterized functionally (D’Auria et al., 2002; Hoffmann et al., 2005; Negruk et al., 1996; Xia et al., 1996), and none of them encode anthocyanin acyltransferases. Our previous transcriptome study of the genes induced by the PAP1 transcription factor, which activates anthocyanin accumulation in Arabidopsis, suggested the involvement of several novel genes in the biosynthesis of anthocyanins (Tohge et al., 2005). These induced genes provide good clues for narrowing down the search for candidate genes that encode anthocyanin acyl transferases from within the BAHD gene family.

As the initial part of a project to identify the function of proteins of the BAHD family in Arabidopsis, we have identified and characterized genes encoding three anthocyanin acyltransferases in Arabidopsis by adopting a modified functional genomics strategy. By focussing on BAHD genes co-induced at times of high anthocyanin production, we have been able to identify acyltransferases with specificity for anthocyanins. This modification of traditional functional genomics strategies provides a useful refinement for identifying function in families of enzymes with significant catalytic versatility, such as those involved in plant secondary metabolism.


In Arabidopsis, the major anthocyanin that accumulates in leaves is cyanidin 3-sambubioside 5-glucoside (Bloor and Abrahams, 2002). This pigment is commonly modified by acylation at three positions: a coumaroyl group attached to the 6′′ position of the 3-O-glucoside, a malonyl group attached to the 6′′′′ position of the 5-O-glucoside and a sinapoyl group attached to the 2′′′ position of the 2′′-O-xylosyl group. An additional glucose residue is commonly linked to the coumaroyl residue (Figure 1a). We extracted anthocyanins from A. thaliana (Columbia) leaves grown under stressed conditions, and confirmed the identity of the major anthocyanins by LC/MS (Figure 1b). This acylation pattern means that Arabidopsis is likely to express a malonyl CoA anthocyanin acyl transferase, a coumaroyl CoA anthocyanin acyl transferase and a sinapoyl CoA anthocyanin acyl transferase. Despite cDNA sequences being available from other plant species for enzymes with exactly the same substrate specificity as the first two acyl transferases (Suzuki et al., 2001; Yonekura-Sakakibara et al., 2000), none of the predicted Arabidopsis BAHD proteins aligned closely with either the hydroxycinnamoyl CoA:anthocyanin 3-O-glucoside-6′′-O-acyltransferase from P. frutescens (Pf3AT; Yonekura-Sakakibara et al., 2000) or the malonyl CoA:anthocyanin 5-O-glucoside-6′′′-O-malonyltransferase (Ss5MaT1) from S. splendens (Suzuki et al., 2001, 2004b) (Figure 1c and Figure S2). However, all but one of the known anthocyanin acyl transferases aligned in a single super-clade with significant bootstrap support. Within this clade were three sub-clades, one of which contained the known anthocyanin acyltransferases and two which contained predicted BAHD enzymes from Arabidopsis, one comprising just two very similar proteins and the other comprising eight proteins. The failure of the Arabidopsis proteins to align closely with any of the known anthocyanin acyl transferases may be because Arabidopsis belongs to the Rosid tribe whereas all the characterized anthocyanin acyl transferases come from species belonging to the Asterid tribe. Whatever the cause, the lack of close homology meant that primary sequence alone could not provide enough information for functional assignment.

Figure 1.

  Anthocyanins of Arabidopsis and phylogenetic analysis of the anthocyanin acyltransferase superclade of proteins. (a) Structure of anthocyanins that are accumulated in wild-type and mutant/transgenic Arabidopsis.
(b) HPLC spectra of the major anthocyanins in wild-type Arabidopsis
(c) Phylogenetic analysis of annotated anthocyanin acyltransferases genes in Arabidopsis with anthocyanin acyltransferases in other species. The following gene sequences were used for the analysis (accession numbers in parentheses): Glossy2 (CAA61258), Dv3MaT (AF489108), Sc3MaT (AAO38058), Dm3MaT1 (AAQ63615), Dm3MaT2 (AAQ63616), Pf3AT (BBA93475), Ss3AT (AY395719), SsAT201 (AY383735), Pf5MaT (AAL50565), Ss5MaT1 (AAL50566), Gt5AT (BBA74428), NtMAT1 (AB176525), Lp3MaT1 (AAS77404), Vh3MaT1 (AAS77402), At1g03940 (NP_171890), At3g29590 (NP_189600), At3g29635 (NP_189605), At3g29670 (AAP04017), At3g29680 (NP_189610), At5g39050 (NP_568561), At5g39080 (AAP49522), At5g39090 (AAP49516), At5g61160 (NP_200924). At1g03495 contains an error in the original sequence, so the full-length cDNA was isolated, sequenced and used instead (see Figure S3b). The amino acid sequences were aligned using the clustalw (version 1.83) program (Thompson et al., 1994). Phylogenetic analysis was performed using the phylip program (version 3.63) with whole-protein alignment. A distance matrix method employing the Jones–Taylor–Thornton model (Perkins and Jones, 1992) was used to compare the sequences, and a tree was built using the neighbour-joining clustering method (Saitou and Nei, 1987). One thousand bootstrapped data sets were used to calculate the confidence of each tree clade.

Five BAHD genes have expression patterns that coordinate with the production of acylated anthocyanins

To delimit candidate genes likely to encode anthocyanin acyltransferases, we treated Arabidopsis seedlings with two stresses known to induce anthocyanin production strongly. Plants were grown separately on high concentrations of sucrose (high Suc) and low concentrations of phosphate (low P).

Young seedlings of Arabidopsis grown on control agar plates produced very low levels of anthocyanin (Figure 2a), whereas both high-Suc and low-P treatments greatly increased anthocyanin production in 9-day-old seedlings. The total amounts of anthocyanins produced in high-Suc and low-P treatments were similar (increased by 10-fold on a DW basis), and the anthocyanins induced by the two stresses were qualitatively similar (Figure 2b). Treatment by combination of these two stresses resulted in even higher anthocyanin production. Detailed analysis by LC/MS showed that the majority of anthocyanins were acylated, with the biggest peak (A11) containing all three acyl groups: p-coumaroyl, malonyl and sinapoyl.

Figure 2.

 Induction of anthocyanin production and expression of At1g03940, At1g03495, At3g29590, At3g29670 and At3g29680 by various stresses.
(a) Effects of low-P (1 mm) and high-Suc (5%) conditions on anthocyanin accumulation. The data represent the mean values (±SD) of three biologically independent experiments.
(b) HPLC analysis of anthocyanins after low-P and high-Suc treatments. Arabidopsis seedlings were grown on the indicated media for 9 days. Peaks labelled with an asterisk indicate cis isomers of peaks with the same name.
(c) mRNA accumulation for At1g03940, At1g03495, At3g29590, At3g29670 and At3g29680 in low-P- and high-Suc-stressed Arabidopsis seedlings. Seedlings were grown on the indicated media for 7 days, RNA was extracted, cDNA obtained, and real-time RT-PCR was carried out using gene-specific primers. The data represent the mean values (± SD) of three biologically independent experiments.
(d) Expression of At1g03940, At1g03495 and At3g29590 in various tissues of Arabidopsis. RNA was extracted from flowers (F), leaves (L), stems (St), siliques (Si) and roots (R) of mature 6-week-old plants. Real-time RT-PCR reactions were performed with primers specific for At1g03940, At1g03495, At3g29590 and Ef1α (used as a constitutive control). The data represent the mean values (± SD) of three biologically independent experiments.

The expression levels of all Arabidopsis BAHD genes in the anthocyanin acyl transferase super-clade were determined by RT-PCR, and the results showed that five genes out of 10 were substantially upregulated (> 4-fold) by both stresses (Figure 2c and Figure S3). Of these five, two (At1g03940 and At3g29590) were also significantly upregulated in their expression in response to high-level expression of the anthocyanin regulatory factor PAP1 in the activation-tagged, high-pigment line PAP1-D (Borevitz et al., 2000; Tohge et al., 2005). A third gene (At1g03495) was annotated as a pseudogene in the TAIR database ( When the cDNA encoding this gene was synthesized by RT-PCR and sequenced (Figure S4a), we found that it encoded a protein with the normal length of a BAHD acyltransferase (Figure S4b) and showed significant sequence identity to At1g03940. Comparing our sequence with the sequence in the database showed that the sequence in the database had a deletion within the open reading frame of this gene that made it appear to encode a truncated protein. This error has been reported to the TAIR database and corrections will be made in the next genome release (TAIR7) in early 2007. We concluded from our results that At1g03495 might encode a functional protein. Consequently, these three genes (At1g03940, At1g03495 and At3g29590) whose expression was very closely associated with anthocyanin production were characterized further.

To determine the tissues in which the three genes, At1g03940, At1g03495 and At3g29590, are expressed, total RNA was isolated from flowers, leaves, stems, siliques and roots of A. thaliana. cDNAs were prepared and used for real-time RT-PCR analyses, and the results are summarized in Figure 2(d). At1g03940 showed high transcript levels in flower, leaf and root, while the expression levels in stem and silique were slightly lower. At1g03495 showed the highest transcript levels in leaves and the lowest levels in siliques and roots. Interestingly, expression of At1g03940 was more responsive to stress from high-Suc treatment than expression of the very similar gene At1g03495. However, At1g03495 expression responded more to low-P treatment. At3g29590 showed similar changes in transcript levels to both stresses. Overall, the transcript levels of At3g29590 were lower than those of At1g03940 and At1g03495, and it had a much higher expression level in flowers than in other tissues.

Expression of At3g29590 in Escherichia coli and assay of the recombinant At3g29590 enzyme

To determine the biochemical function of the product of At3g29590, total RNA was extracted from Arabidopsis flowers and the full-length cDNA was obtained by RT-PCR. The cDNA sequence showed that the structure of the At3g29590 gene predicted by the genome annotation in the TAIR database was correct, consisting of a single exon with no introns. The entire At3g29590 cDNA was then recombined into a Gateway-compatible E. coli expression vector, pSTAG-GW (Edwards et al., 1999), under the control of the T7 promoter. This construct was designed to express At3g29590 with an S-TAG fused to the N-terminal end of the recombinant protein. The recombinant protein was expressed in E. coli, and production of the fusion protein was confirmed by sodium dodecyl sulphate–polyacrylamide-gel electrophoresis and S-TAG Western blot. One of the advantages of using S-TAG fusions is that the kinetics of the activity of the fusion proteins can be determined in crude extracts of the recombinant enzymes (Edwards et al., 1999). We confirmed that the presence of the S-TAG did not affect enzyme activity by cleavage with thrombin (Figure S6). Enzyme assays were performed using a range of possible acyl CoA and flavonoid substrates, and the kinetics of the recombinant At3g29590 enzyme are summarized in Table 2.

Table 2.   Kinetics of recombinant At3g29590 in Escherichia coli
 Relative activity (%)akcat (sec−1)Kmm)
  1. aSpecific activities were determined under the conditions described in Experimental procedures. Acyl donors and acceptors with final concentrations of 60 and 120 μm, respectively, were used when determining the Km for the other substrate type. The specific activity with malonyl CoA and cyanidin 3,5-diglucoside (120 nkat mg−1) was taken to be 100%. The relative activities were determined from product peak integrals, assuming that the extinction coefficient of the product was the same as that of the substrate.

  2. bThe reactions were carried out with cyanidin 3,5-diglucoside as the acyl acceptor.

  3. cThe reactions were carried out with malonyl CoA as the acyl donor.

  4. dNA, no activity.

Acyl donorb
 Malonyl CoA1006.4 ± 0.85.5 ± 0.7
 Acetyl CoANAd  
 Methylmalonyl CoANAd  
 Succinyl CoANAd  
 p-Coumaroyl CoANAd  
Acyl acceptorc
 Cyanidin 3,5-diglucoside1006.4 ± 0.86.6 ± 0.5
 Pelargonidin 3,5-diglucoside 1075.9 ± 0.74.5 ± 0.7
 Peonidin 3,5-diglucoside1266.6 ± 0.76.9 ± 0.6

The recombinant enzyme showed malonyl CoA anthocyanin acyl transferase activity exclusively, with no activity with p-coumaroyl CoA or other aromatic acyl CoA donors. It effectively catalysed the malonyl CoA-dependent malonyl transfer to cyanidin 3,5-diglucoside to produce cyanidin 3-O-glucoside 5-O-malonylglucoside as the only product, with a kcat of 6.4 ± 0.8 sec−1, a Km of 5.5 ± 0.7 μm for malonyl CoA and a Km of 6.6 ± 0.5 μm for cyanidin 3,5-diglucoside. In addition to cyanidin 3,5-diglucoside, the enzyme was also able to use pelargonidin 3,5-diglucoside and peonidin 3,5-diglucoside as substrates with similar affinities as for cyanidin 3,5-diglucoside (Km values of 4.5 ± 0.7 and 6.9 ± 0.6 μm, respectively) when malonyl CoA (60 μm) was used as the acyl donor. In contrast to its ability to use a wide range of anthocyanins as substrates, the recombinant enzyme showed very strict specificity for acyl donors, and no activity could be detected when acetyl CoA, methylmalonyl CoA or succinyl CoA were used as acyl donors. The activity of the recombinant enzyme was enhanced by 10 μmβ-mercaptoethanol (2ME) (120%), whereas 5 mm ethylenediaminetetra-acetate (EDTA) had no significant inhibitory effect on enzyme activity. The Km values and kcat values for At3g29590 were in the same range as reported for other anthocyanin malonyltransferases.

The specificity of the enzyme encoded by At3g29590 for a range of anthocynin substrates was tested further. We were unable to prepare large enough amounts of these substrates for detailed kinetic analysis, but we were able to obtain enough to determine the relative activity of the enzyme with these anthocyanins at a concentration of 30 μm. The results are shown in Table 3. The activity of the At3g29590 enzyme was, to a large extent, independent of modifications to the glucose at the 3 position of the anthocyanin 3,5-diglucoside (Table 3). It showed almost the same activity on cyanidin 3-glucoside 5-glucoside, cyanidin 3-coumaroylglucoside 5-glucoside and cyanidin 3-coumaroylsambubioside 5-glucoside (A3) when coumaroyl CoA was used as the acyl donor. Similar activities were obtained when delphinidin 3-coumaroylrutinoside 5-glucoside and petunidin 3-coumaroylrutinoside 5-glucoside were used as acyl acceptors. The activity dropped by about 50% when cyanidin 3-(sinapoyl) (coumaroy)l sambubioside 5-glucoside (A7) was used as a substrate. However, very little or no activity could be detected with cyanidin 3-(glucosyl-coumaroyl) sambubioside 5-glucoside (A6) or cyanidin 3-(sinapoyl)-(glucosyl-coumaroyl) sambubioside 5-O-glucoside (A10) (both have further glycosylation on the coumaroyl group) as the acyl acceptors.

Table 3.   Relative activities of At3g29590 (At5MaT) on various anthocyanins with modified 3-glucosidesa
AnthocyaninRelative activity (%)
  1. aEnzyme reactions were carried out as described in Experimental procedures using malonyl CoA as the acyl donor. Acyl donors and acceptors with final concentrations of 30 μm were used for the reactions. The relative activities were determined from product peak integrals, assuming that the extinction coefficient of the product was the same as that of the substrate, using cyanidin 3,5-diglucoside as a standard.

  2. bNA, no activity.

Cyanidin 3,5-diglucoside100 ± 10.9
Cyanidin 3-coumaroylglucoside 5-glucoside104 ± 12.1
Cyanidin 3-O-(xylosyl) 6′′-O-p-coumaroyl glucoside 5-O-glucoside (A3)94.6 ± 8.9
Delphinidin 3-coumaroylrutinoside 5-glucoside88.5 ± 10.2
Petunidin 3-coumaroylrutinoside 5-glucoside74.9 ± 7.3
Cyanidin 3-O-[(sinapoyl) xylosyl 6′′-O-p-coumaroyl glucoside] 5-O- glucoside (A7)48.1 ± 3.5
Cyanidin 3-O-[(xylosyl) 6′′-O- (p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-glucoside (A6)4.2 ± 0.5
Cyanidin 3-O-(sinapoyl) xylosyl 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-glucoside (A10)NAb

Knockout of At3g29590 confirmed its catalytic function in vivo

To confirm the function of At3g29590 (henceforth referred to as At5MaT) in vivo, a dSpm insertion mutant of this gene was obtained. The SM line (SM_3_35619) contained a dSpm insertion in the single exon of At3g29590 (Figure 3a). Homozygotes were identified by PCR screening, and cDNA from leaves was analysed for expression of At5MaT. No transcript was detected in the insertion line, confirming that the insertion eliminated production of the At5MaT transcript (Figure 3b). Both wild-type and homozygous At5MaT mutant seeds were sown on low P and high Suc medium, and anthocyanins and flavonoids were analysed 9 days after germination. The total anthocyanin content was unaffected in the homozygous mutant compared to wild-type plants, although the composition of the anthocyanins that accumulated was greatly changed (Figure 3c). When the HPLC profile of anthocyanins from the knockout mutant was compared with the profile from the wild-type, all the major peaks were shifted to the left (with decreased retention time). Detailed study by LC/MS showed that, in the At5MaT mutant, all the anthocyanins with malonyl groups attached to the 5-glucoside were lost. For example, cyanidin 3-O-[2′′-O-(6′′′-O-(sinapoyl) xylosyl) 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-(6′′′′-O-malonyl) glucoside (A11), the major anthocyanin in wild-type Arabidopsis, was replaced by cyanidin 3-O-[2′′-O-(6′′′-O-(sinapoyl) xylosyl) 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-glucoside (A10) in the knockout mutant. All the other malonylated anthocyanins (A5, A8 and A9) were lost and were replaced by the 5-glucosides A3, A6 and A7, respectively. The UV spectra of the malonylated anthocyanins were the same as those of the corresponding non-malonylated ones (Figure 3d). These results indicate that At3g29590 functions as an anthocyanin 5-glucoside malonyl CoA acyl transferase in vivo.

Figure 3.

 dSpm insertion mutant of At5MaT (At3g29590) and analysis of the effects of the mutation on anthocyanins in Arabidopsis seedlings.
(a) Schematic of the structure of the dSpm insertion in At3g29590.
(b) RT-PCR analysis of transcription of At3g29590 in wild-type (WT) and homozygous mutants (SM_3_35619, with a dSpm insertion in At3g29590). RT-PCR with primers for elongation factor 1α (Ef1α) was used to judge the equality of the concentration of the cDNA templates used in each treatment.
(c) HPLC analysis of the anthocyanins in WT and the At5MaT mutant line SM_3_35619 grown under low-P and high-Suc stress. Arabidopsis seedlings were grown on medium with both low P and high Suc for 9 days. Peaks labelled with an asterisk indicate cis isomers of peaks with the same name.
(d) UV spectra of peak A10 (solid line) and A11 (dotted line) in Arabidopsis. No significant spectral shift was observed in the anthocyanin without the malonyl substitution (A10) compared to that with the malonyl substitution (A11).

Taken together; these results showed that At3g29590 functions as an anthocyanin 5-glucoside malonyltransferase (At5MaT) both in vitro and in vivo.

Recombinant expression of At1g03940 and At1g03495 in E. coli and biochemical assay of the At3AT1 and At3AT2 recombinant enzymes

To investigate the functions of At1g03940 and At1g03495, the cDNAs encoding these genes were amplified and recombined into the expression vector pSTAG-GW to allow expression of recombinant fusion proteins under regulation of the T7 promoter in E. coli. The soluble fractions of the E. coli extracts were subjected to an anthocyanin acyltransferase assay using a range of substrates. The reaction mixtures were analysed by reverse-phase HPLC, and compounds were identified by LC/MS, retention time and UV spectra. The results are summarized in Table 4. The presence of the S-TAG did not affect the activities of the enzymes (Figure S6). The recombinant At1g03940 protein was active with three of the hydroxycinnamoyl CoA donors tested, and the greatest activity was achieved when caffeoyl CoA was used as an acyl donor. The Km values for each individual CoA were 3.9 ± 0.5, 4.2 ± 0.6 and 2.2 ± 0.2 μm, for coumaroyl CoA, feruloyl CoA and caffeoyl CoA, respectively. While activity with coumaroyl CoA, feruloyl CoA and caffeoyl CoA acyl donors showed Michaelis–Menton kinetics, activity declined with high levels of sinapoyl CoA (above 10 μm), suggesting that, at high levels, this acyl donor may inhibit acyl transferase activity (Figure S7). Consequently we could not calculate reliable Km values for this substrate. In contrast to its similar specificity for various acyl donors, recombinant At1g03940 protein showed selectivity for the various acyl acceptors used in the tests. The recombinant enzyme was much more active on anthocyanin 3-glucosides than on anthocyanin 3,5-diglucoside, with a 20-fold increase in Km on cyanidin 3,5-diglucoside compared to cyanidin 3-glucoside. No activity was detected when cyanidin 3-rutinoside was used as an acyl acceptor. This supports the specificity of the enzyme for acyl transfer to the 6′′ position of the 3-O-glucose, because this is linked to rhamnose in the 3-O-rutinoside. In addition to anthocyanin 3-glucoside, this enzyme was also able to use flavonol 3-glucosides as the acyl acceptor, although its Km values for flavonol 3-glucosides were about sixfold (for quercetin 3-glucoside) and ninefold (for kaempferol 3-glucoside) higher than for the anthocyanin 3-glucosides. Surprisingly, this enzyme showed relatively high affinity for kaempferol 7-glucoside (Km 9.4 ± 0.6 μm) although its kcat with this acyl acceptor was about one-third of that for the anthocyanins. The At1g03495 recombinant enzyme showed very similar properties to recombinant At1g03940 protein (Table 4). Considering their high sequence identity (98% for DNA sequence and 97% for protein sequence), the products of these two genes may be regarded as functionally equivalent in vitro. Given their very similar expression patterns, the At1g03940 and At1g03495 proteins may also be regarded as functionally redundant.

Table 4.   Kinetics of recombinant At3AT1 and At3AT2 in Escherichia coli
 Relative activity (%)akcat (sec−1)Kmm)
  1. aThe specific activities were determined under the conditions described in Experimental procedures. Acyl donors and acceptors at final concentrations of 30 and 60 μm, respectively, were used when determining the Km for the other substrate type. The specific activity with p-coumaroyl CoA and cyanidin 3-glucoside (141 nkat mg−1) was taken to be 100%. The relative activities were determined from product peak integrals, assuming that the extinction coefficient of the product was the same as that of the substrate.

  2. bThe reactions were carried out with cyanidin 3,5-diglucoside as an acyl acceptor.

  3. cThe reactions were carried out with malonyl CoA as an acyl donor.

  4. dNA, no activity.

  5. eND, not determined.

  6. fThe low relative activity with sinapoyl CoA reflects the inhibition of At3AT1 and At3AT2 activity by sinapoyl CoA at concentrations above 10 μm (see Figure S7).

For At3AT1
Acyl donorb
 p-Coumaroyl CoA1007.6 ± 0.73.9 ± 0.5
 Feruloyl CoA1208.5 ± 1.14.2 ± 0.6
 Caffeoyl CoA1339.3 ± 1.22.2 ± 0.2
 Sinapoyl CoA0.96fNDeNDe
Acyl acceptorc
 Cyanidin 3-glucoside1007.6 ± 0.77.0 ± 0.5
 Pelargonidin 3-glucoside 1198.0 ± 0.77.3 ± 0.8
 Malvidin 3-glucoside1047.8 ± 0.96.9±0.8
 Quercetin 3-glucoside10.31.2 ± 0.245.7 ± 7.6
 Kaempferol 3-glucoside9.01.1 ± 0.164.8 ± 5.0
 Kaempferol 7-glucoside36.12.6 ± 0.39.4 ± 1.0
 Cyanidin 3-rutinosideNAdNDeNDe
 Cyanidin 3,5-diglucoside0.03<0.001141 ± 18
 Pelargonidin 3,5-diglucoside 0.03<0.001116 ± 14
For At3AT2
Acyl donor
 p-Coumaroyl CoA1006.9 ± 0.86.6 ± 0.9
 Feruloyl CoA1197.7 ± 0.99.4 ± 1.6
 Caffeoyl CoA1308.5 ± 1.24.9 ± 0.7
 Sinapoyl CoA0.7fNDeNDe
Acyl acceptor
 Cyanidin 3-glucoside1006.9 ± 0.810.3 ± 1.1
 Pelargonidin 3-glucoside 1077.5 ± 0.69.9 ± 1.3
 Malvidin 3-glucoside946.3 ± 0.57.6 ± 1.2
 Quercetin 3-glucoside8.60.9 ± 0.167.9 ± 10.2
 Kaempferol 3-glucoside7.30.8 ± 0.182.6 ± 13.7
 Kaempferol 7-glucoside26.42.0 ± 0.211.6 ± 1.2
 Cyanidin 3-rutinosideNAdNDeNDe
 Cyanidin 3,5-diglucoside0.04<0.001186±21
 Pelargonidin 3,5-diglucoside 0.04<0.001152

Expression of At3AT1 and At3AT2 in tobacco, and modification of anthocyanins in tobacco flowers

The in vitro assays of At1g03940 (At3AT1) and At1g03495 (At3AT2) showed that the genes encoded anthocyanidin 3-O-glucoside acyl CoA transferase enzymes that were catalytically equivalent in vitro, and we sought to confirm their functions in vivo. However, two independent knockout lines of At1g03940 and two independent knockout lines of At1g03495, where the corresponding transcripts were abolished, exhibited exactly the same anthocyanin profile as the wild-type plants (see Figure S5). As the two genes are closely linked, it proved difficult to create double knockout mutant lines. To establish function in vivo, an alternative strategy was adopted. We expressed each gene in a host plant that does not normally produce aromatic acylated anthocyanins. Tobacco has been reported to produce unsubstituted anthocyanin 3-O-rutinosides in its flowers (Timberlake and Bridle, 1975), the structures of which are shown in Figure 4(a). We confirmed this by extracting the anthocyanins from Nicotiana tabacum var. Samsun flowers and analysing them by LC/MS/MS. Wild-type N. tabacum var. Samsum flowers contained cyanidin 3-O-rutinoside (kerocyanin) as the dominant anthocyanin (Figure 4c; A13). This was confirmed by the UV/visible spectrum during HPLC which was indicative of a cyanidin derivative (λmax = 520 nm), a component with the predicted MS behaviour of cyanidin rutinoside (molecular ion at m/= 595 and 287 for MS and MS/MS, respectively). A13 ran with the same retention time as the authentic kerocyanin (standard). In addition to this major peak, the wild-type flowers also contained a small amount of pelargonidin 3-rutinoside (Figure 4c; A12). This compound was also identified by its UV spectrum, MS and comparison to the retention time of the standard.

Figure 4.

 Over-expression of At1g03940 and At1g03495 in tobacco and LC/MS analyses of anthocyanin in transgenic tobacco flowers.
(a) Structure of anthocyanins accumulated in the wild-type and transgenic tobacco flowers.
(b) RT-PCR analysis of transcription of At3AT1 (At1g03940) and At3AT2 (At1g03495) in wild-type (WT), At3AT1-over-expressing (line 75-4) and At3AT2-over-expressing (line 104-2) tobacco. RT-PCR with primers for ubiquitin 2 (ub2) was used to judge the equality of the concentration of the cDNA templates used in each treatment.
(c) HPLC analysis of anthocyanins in wild-type (WT), At3AT1-over-expressing (line 75-4) and At3AT2-over-expressing (line 104-2) tobacco flowers. Peaks labelled with an asterisk indicate cis isomers of peaks with the same name.
(d) UV spectra of peak A12 (solid line) compared to A14 (dotted line) from transgenic tobacco flowers over-expressing At3AT1 or At3AT2.
(e) UV spectra of peak A13 (solid line) compared to A15 (dotted line) from transgenic tobacco flowers over-expressing At3AT1 or At3AT2.
(f) MS for peak A14 from transgenic tobacco flowers.
(g) MS for peak A15 from transgenic tobacco flowers.
(h) Flowers of wild-type (left) and an At3AT1-over-expressing tobacco line (right).
(I) Flowers of wild-type (left) and an At3AT2-over-expressing tobacco line (right).

Consequently, we created transgenic tobacco (var. Samsun) plants expressing either of the two genes under the control of a double 35S promoter. Fifteen independent kanamycin-resistant tobacco plants (T0) for each gene were transferred to soil. The expression of transgenes was assayed by RT-PCR (Figure 4b), and expression was detected in ten and nine of the At3AT1 and At3AT2 transgenic lines, respectively. The anthocyanins in the flowers of the transgenic lines were then analysed and compared to those of wild-type by LC/MS/MS.

At3AT1 and At3AT2 transgenic flowers showed the same results with respect to their LC/MS/MS profiles. Cyanidin 3-rutinoside (A13) was still the dominant anthocyanin in flowers of the At3AT1 and At3AT2 over-expression lines, but some new anthocyanins were also present. There were several newly formed peaks that were more hydrophobic than cyanidin 3-O-rutinoside and pelargonidin 3-O-rutinoside (Figure 4c). One major new peak (A15) had an m/z value of 595 for MS1, and further fragmented to m/z 287 for MS/MS as predicted for cyanidin 3-O-(coumaroyl) glucoside (Figure 4g). It was also detected by an additional spectral maximum at 315 nm, and a shift in absorption maximum from 520 to 526 nm, which indicated an additional coumaroyl group (Figure 4e). These data showed that although the m/z was still 595, this was not simply another cyanidin derivative of glucose and rhamnose, with the sugars linked in different positions to the cyanidin 3-O-rutinoside. In addition, the UV spectrum, and the greatly enhanced hydrophobicity (which indicated loss of a hydrophilic sugar and replacement by a hydrophobic group), showed that it could not be pelargonidin/delphinidin with two glucose/two rhamnose units. Taken together, the data establish that the new anthocyanin peak at 42.9 min present in transgenic flowers is most likely cyanidin 3-O-(coumaroyl) glucoside. A second smaller peak (A14) was detected only in the transgenic lines. This peak had an m/z value of 579 and further fragmented to 271 as predicted for pelargonidin 3-O-(coumaroyl) glucoside (Figure 4f). It had an additional spectral maximum at 315 nm and a shift in absorption maximum from 501 to 507 nm (Figure 4b). This peak was identified as most likely to be pelargonidin 3-O-(coumaroyl) glucoside.

These results showed that At1g03940 and At1g03495 encode enzymes with anthocyanidin 3-O-glucoside coumaroyl CoA transferase activity in vivo. We deduced that the coumaroyl group was transferred to the 6′′ position of the 3-O-glucoside because both enzymes appeared to operate in competition with the endogenous rhamnosyl transferase in flowers of the transgenic lines, and rhamnose is attached to the 6′′ position of the 3-O-glucose in rutinosides.

Changes in the spectral properties and stability of anthocyanins caused by acylation

Analysis of the absorption spectra of anthocyanins from the transgenic tobacco lines showed that coumaroylation by At3AT1 and At3AT2 could shift the maximum absorbance of the anthocyanins (by 6 nm towards the blue part of the spectrum for cyanidin 3-O-(coumaroyl) glucoside and for pelargonidin 3-O-(coumaroyl) glucoside compared to cyanidin 3-O-rutinoside and pelargonidin 3-O-rutinoside, respectively). The shifts in the absorption maxima observed for the coumaroylation of cyanidin 3-glucoside and pelargonidin 3-glucoside are equivalent to the spectral shift reported for delphinidin 3-(caffeoyl) glucoside 5-glucoside compared to delphinidin 3,5-diglucoside, the product of Pf3AT activity from P. frutescens (Yonekura-Sakakibara et al., 2000). Despite the measured shift, we were unable to detect any major change in the hue of the transgenic tobacco flowers by eye (Figure 4h,i), although the colour of the transgenic flowers appeared slightly more intense than that of controls.

To investigate the effects of acylation on the properties of anthocyanins further, the stability of acylated anthocyanins was compared with the stability of non-acylated ones. This was performed with anthocyanins with various patterns of glycosylation and acylation in neutral aqueous solutions. The stability was evaluated on the basis of the half-life (t½) which was defined as the time to reach 50% residual colour. As shown in Figure 5(a) and Table 5, both coumaroylated and malonylated anthocyanins showed increased stability at pH 7.0 compared to their non-acylated counterparts. The highest stability was obtained for cyanidin 3-(coumaroyl) glucoside (t½ > 48 h), while the lowest stability was shown by cyanidin 3,5-diglucoside (t½ = 1.5 h). The greater stability of cyanidin 3-(coumaroyl) glucoside compared to cyanidin 3-rutinoside (t½ = 24 h) and cyanidin 3-glucoside suggested that coumaroylation is more effective at enhancing stability of anthocyanins than rhamnosylation. However, the addition of a second glucose at the 5 position of the anthocyanins severely decreased their stability in solution. It has been suggested that anthocyanins with aromatic acyl substitutions are more stable than those with aliphatic acyl substitutions (Stintzing and Carle, 2004). As the malonylation and coumaroylation catalysed by the BAHD enzymes characterized from Arabidopsis in this study transferred their acyl groups to different sugar residues on the anthocyanins, the effects of aromatic substitution compared to aliphatic substitution on absorption properties and stability in solution could not be compared directly.

Figure 5.

 Stability of various anthocyanins at neutral pH.
(a) Stability of purified anthocyanins at neutral pH as shown by the decline in absorption at the absorption maximum. Cyanidin 3-glucoside (□), cyanidin 3-rutinoside (•), cyanidin 3,5-diglucoside (Δ), cyanidin 3-glucoside 5-malonylglucoside (bsl00066) and cyanidin 3-(coumaroyl) glucoside (bsl00001) in aqueous solution at 25°C (50 μm of each anthocyanin in 20 μm phosphate buffer at pH 7.0).
(b) Stability of anthocyanins extracted from tobacco flowers in aqueous solution at 25°C (50 μm of each anthocyanin in 20 μm phosphate buffer at pH 7.0). Anthocyanin extractions of flowers from wild-type (•) and T1 generation plants of At3AT1 (bsl00066) and At3AT2 (bsl00001) over-expression tobacco lines.
(c) Stabilities of anthocyanins extracted from Arabidopsis seedlings in aqueous solution at 25°C (50 μm of each anthocyanin in 20 μm phosphate buffer at pH 7.0). Anthocyanin extractions from wild-type (□) and At5MaT knockout mutant (bsl00001) Arabidopsis seedlings.
The data represent the mean values (± SD) of two independent experiments.

Table 5.   Effect of acylation on the stability (defined as half-life, t½) of anthocyaninsa
AnthocyaninHalf-life (t½) (h)
  1. aExperiments were carried out as described in Experimental procedures and in the legend to Figure 5.

Purified anthocyanins
 Cyanidin 3,5-diglucoside1.5 ± 0.1
 Cyanidin 3-glucoside 5-malonylglucoside5.5 ± 0.5
 Cyanidin 3-glucoside10.0 ± 0.8
 Cyanidin 3-rutinoside24.0 ± 2.0
 Cyanidin 3-(coumaroyl) glucoside> 48.0
Plant tissue extracts
 Wild-type tobacco flower extract28.0 ± 3.0
 At3AT1 over-expression tobacco flower extract> 48.0
 At3AT2 over-expression tobacco flower extract44.0 ± 4.5
 Wild-type Arabidopsis seedling extract32.0 ± 3.0
 At5MaT mutant Arabidopsis seedling extract20.0 ± 1.8

The stabilities of acylated and non-acylated anthocyanins in crude extracts of plant tissues were also investigated. Anthocyanins in crude extracts of flowers from the At3AT1- and At3AT2-over-expressing tobacco lines were more stable than those from the wild-type plants at all time intervals (Figure 5b and Table 5), demonstrating that coumaroylation improves anthocyanin stability in complex mixtures as well as in purified form. The effects of malonylation on stability of anthocyanins in crude extracts were more complex (Figure 5c and Table 5). The stability of anthocyanins in extracts from wild-type and At5MaT knockout Arabidopsis seedlings was the same during the first 10 h of the assay, but after 24 h the anthocyanins in the extracts from the wild-type seedlings were significantly more stable than those in extracts from the knockout mutant seedlings, which lacked malonylation of the 5-glucoside of the anthocyanins. The increases in anthocyanin stabilities detected in vivo were moderate compared to those detected in vitro. This might reflect the complexity of the in vivo system where the mixing of anthocyanins with other co-pigments such as flavonols may also affect stability.


With the completion of sequencing of the whole genome, there is growing interest in using Arabidopsis as a model system to investigate secondary metabolism in plants (D’Auria and Gershenzon, 2005). Sequence annotation has identified a number of gene families that are known to be involved in secondary metabolism, such as acyltransferases (D’Auria, 2006; Milkowski and Strack, 2004), terpene synthases (Aubourg et al., 2002; Chen et al., 2003b) and glucosyltransferases (Bowles, 2002; Li et al., 2001). Despite the large number of genes suggested by annotation, only a few of them have been characterized, and a few have been identified by functional genomic approaches (Chen et al., 2003a; Tohge et al., 2005). By combining analysis of gene expression linked to conditions that induce the accumulation of acylated anthocyanins (high-Suc, low-P or ectopic expression of PAP1) with phylogenetic analysis of the BAHD family of acyltransferases, we have been able to identify two genes encoding coumaroyl CoA cyanidin 3-O-glucose transferase (3AT) and one encoding malonyl CoA cyanidin 3,5-diglucoside transferase activity (5MaT). In our survey of candidate anthocyanin acyltransferase genes, the genes that were annotated as anthocyanin acyltransferases reacted differently to stresses that induced the production of acylated anthocyanins. This allowed us to select a more limited number of candidate genes from the anthocyanin acyltransferase super-clade for detailed investigation. In addition to the three genes that have been characterized in this paper, the transcript levels of another two genes (At3g29670 and At3g29680) were also increased substantially (> 4-fold) by both stresses. Considering their expression patterns and the nature of the acylation of the anthocyanins induced by high-Suc or low-P stresses, these genes are candidates for encoding sinapoyl CoA anthocyanin 3-sambubioside transferase activity. In contrast to these genes, some genes within the anthocyanin acyl transferase super-clade (such as At5g39080 and At5g39090) showed no response in their expression to either stress, and one gene (At5g61160) was severely repressed by high-Suc conditions. It seems unlikely that these genes encode anthocyanin acyltransferases even though they align in the same clade as bona fide anthocyanin acyl transferases. The versatility of the family in general could mean that the acyl acceptors for these enzymes are other flavonoids, or perhaps even other phenolics.

From the phylogenetic analysis, it can also be seen that none of the Arabidopsis acyltransferases aligned with proteins with identical specificity from other species. At3g29590 (At5MaT) was not even in the same sub-clade of malonyl CoA acyltransferases as those from other species. This suggests that these anthocyanin acyl transferases evolved their catalytic specificity independently (see also (Tanaka and Brugliera, 2006). It also means that, although phylogenies can give clues to function, in families of proteins with very versatile catalytic specificities, function needs to be established on a gene-by-gene basis for different species, as suggested by D’Auria (2006).

The anthocyanin acyltransferases from Arabidopsis also have distinct catalytic properties to those anthocyanin acyltransferases characterized from other plant species. Unlike other malonyltransferases, At5MaT has very narrow selectivity for its acyl donor, accepting nothing except malonyl CoA. These distinct catalytic properties support the suggestion that anthocyanin 5-malonyl acyltransferases in the Rosids and Asterids have evolved independently. At5MaT has rather broad specificity for its acyl acceptor, and can use anthocyanins modified by acylation or by further glycosylation on the 3-glucoside, including cyanidin 3-(sinapoyl) (coumaroyl) sambubioside 5-glucoside (A7). However, it was unable to act on cyanidin 3-O-(sinapoyl) (glucosyl)-coumaroyl sambubioside 5-O-glucoside (A10), where the coumaroyl group had been further modified by addition of an extra glucose residue, suggesting that the glycosylation of the coumaroyl group usually occurs after malonylation of the 5-glucoside.

To date, only two other hydroxycinnamoyl CoA anthocyanin acyltransferases have been identified. One is hydroxycinnamoyl CoA:anthocyanin 5-glucoside-6′′′-O-acyltransferase (Gt5AT) from G. triflora (Fujiwara et al., 1998), and the other is hydroxycinnamoyl CoA:anthocyanin 3-O-glucoside-6′′-O-acyltransferase (Pf3AT) from P. frutescens (Yonekura-Sakakibara et al., 2000). Gt5AT showed higher affinity towards delphinidin 3,5-diglucoside than towards cyanidin 3,5-diglucoside and higher affinity towards caffeoyl CoA than towards coumaroyl CoA (Fujiwara et al., 1997). Although At3AT1 and At3AT2 have similar kcat values to Pf3AT, their Km values for both acyl donors and acceptors are lower than those for Pf3AT. One additional feature of the activity of At3AT1 and At3AT2 is that they are both severely inhibited by sinapoyl CoA when its concentration is higher than 10 μm. That could explain why no sinapoylated anthocyanidin 3-glucosides are found in Arabidopsis, despite sinapoyl CoA being available as an acyl donor [as evidenced by the fact that the most abundant anthocyanin in Arabidopsis leaves is cyanidin 3-O-[2′′-O-(6′′′-O-(sinapoyl) xylosyl) 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-(6′′′′-O-malonyl) glucoside (A11)].

Another unique feature of At3AT1 and At3AT2 is that the enzymes can act not only on anthocyanins but also on flavonols. Their Km values for kaempferol 3-glucoside (K3G) and quercetin 3-glucoside (Q3G) are higher than for anthocyanins and their kcat values are lower, suggesting that it is unlikely that these enzymes use flavonol glycosides as acyl acceptors in vivo. There are, as yet, no reports of acylated flavonol 3-glucosides in Arabidopsis. Surprisingly, both enzymes also have very high affinity for another flavonol glycoside, kaempferol 7-glucoside (K7G) , although their kcat values with this acceptor were about one-third of those for anthocyanins. Whether these enzymes can act as K7G acyltransferases in vivo would depend on the availability of the kaempferol 7-O-glucoside in Arabidopsis.

At3AT1 and At3AT2 had much higher Km and much lower kcat values for anthocyanin 3,5-diglucoside compared to anthocyanin 3-glucoside, showing that the acylation at 3 position of the cyanidin 3-O-glucoside must occur before glucosyl transfer to the 5 position of the anthocyanin and the subsequent acylation of this glucose. Therefore, the specificity of the anthocyanin-modifying enzymes means that modifications must occur in a defined order rather than in a network of reactions. This is unlike Pf3AT from P. fructescens, which has similar activities on anthocyanidin 3-glucosides and anthocyanidin 3,5 diglucosides, implying that, in Perilla, glycosylation and aromatic acylation reactions may occur via a metabolic grid (Yamazaki et al., 1999; Yonekura-Sakakibara et al., 2000). This distinction again highlights the probable independent evolution of these two enzymes with identical biological functions. At5MaT can use a variety of anthocyanin acyl acceptors, suggesting that addition of the malonyl group may occur at several different stages following glucosylation of the 5 position of the anthocyanin. However, addition of another glucose residue to the coumaroyl group of cyanidin 3-(coumaroyl) sambubioside 5-glucoside resulted in a very significant decrease in the activity of At5MaT, suggesting that this final glucosylation must be one of the last in the anthocyanin biosynthetic pathway in Arabidopsis.

When the At3AT1 and At3AT2 enzymes were expressed in tobacco, both could use the cyanidin or pelargonidin 3-glucosides (intermediate metabolites in the endogenous anthocyanin biosynthesis pathway) to make new end products. Although incorporation of a coumaroyl group to form a novel anthocyanin in this species did cause a shift of maximum absorption to the blue side by about 6 nm, no difference in flower colour between wild-type and At3AT1- and At3AT2-expressing flowers was obvious by eye. This might be because the cyanidin 3-rutinoside was still the dominant anthocyanin in the transgenic flowers; the molar ratio of cyanidin 3-(coumaroyl) glucoside to cyanidin 3-rutinoside was between 5 and 25% in various T0 plants (20-45% in various T1 lines). This suggested that the endogenous rhamnosyl transferase was more efficient at using cyanidin 3-O-glucoside as a substrate than the introduced At3AT1 and At3AT2 enzymes were. In contrast, the amount of novel pelargonidin 3-(coumaroyl) glucoside was greater than the amount of original pelargonidin 3-rutinoside in the transgenic lines, suggesting that the relative selectivity of At3AT1 and At3AT2 for pelargonidin 3-glucoside compared to cyanidin 3-glucoside was higher than for the endogenous rhamnosyl transferase. It has been suggested that anthocyanin acyl transferases could be used to modify flower colour by genetic engineering (Tanaka et al., 2005) To achieve this successfully, the specificity of the gene products for the acyl donors and acceptors will need to be taken into consideration, along with knowledge of likely competing anthocyanin-modifying activities present in the target species to be engineered. It is also possible that anthocyanin acyl transferases could be used to modify anthocyanins to provide more stable natural food colourants (Francis, 1989). The efficacy of acylation in increasing pigment stability was demonstrated very clearly by the activity of the anthocyanin acyl transferases from Arabidopsis, both in vitro and in crude extracts from plant tissues.

Experimental procedures

Plant materials and treatments

For the analysis of gene expression patterns, A. thaliana (ecotype Columbia) plants were grown in soil at 23°C under a 16 h light/8 h dark cycle for up to 5 weeks. Various tissues (flowers, leaves, stems, siliques and roots) were then harvested for analysis.

For the stress treatments, Arabidopsis seeds were sown on various media: control, low concentration of phosphorus (LP), control plus high concentration of sucrose (HSuc), and high concentration of phosphorus plus high concentration of sucrose (LPHSuc). The control medium contained 3 mm KNO3, 2 mm Ca(NO3)2, 0.5 mm MgSO4, 25 μm KCl, 12.5 μm H3BO3, 1 μm MnSO4, 1 μm ZnSO4, 0.25 μm CuSO4, 0.25 μm (NH4)6Mo7O24, 25 μm Fe-EDTA, 0.55 mm myoinositol, 2.5 mm MES, 1 mm NH4H2PO4, 29.2 mm sucrose and 7 g l−1 agarose (Sigma Chemical Co.; (Ma et al., 2001). The agar concentration used here contains approximately 1.0 ± 0.5 μm phosphorus in the final medium, determined by the phosphomolybdate method (Watanabe and Olsen, 1965). LP medium was the above medium with 1 mm NH4H2PO4 replaced by 1 mm (NH4)2SO4. The HSuc and LPHSuc media were control and LP media supplemented with 146 mm sucrose. The various media were titrated to pH 5.7 with KOH before agar was added. After incubating at 4°C in the dark for 6 days, seeds were then placed under a 16 h light/8 h dark cycle at 23°C. Half of the seedlings were harvested 7 days after germination for RNA extraction. Anthocyanins were determined at day 9 by spectrometry and HPLC for the remaining half of the seedlings.

Transgenic and control (N. tabacum var. Samsum) tobacco plants were grown in soil at 25°C under a 16 h light/8 h dark cycle, and flowers from both control and transgenic plants were taken three months after sowing for transgene expression and anthocyanin measurements.


Acetyl CoA, malonyl CoA, methylmalonyl CoA and succinyl CoA were obtained from Sigma. p-Coumaroyl CoA, feruloyl CoA, caffeoyl CoA and sinapoyl CoA were synthesized as described by Niggeweg et al. (2004). Cyanidin 3-coumaroylglucoside 5-glucoside was synthesized using At3AT1 and cyanidin 3,5 diglucoside as an acyl acceptor. Cyanidin 3-O-[2′′-O-(xylosyl) 6′′-O-(p-coumaroyl) glucoside] 5-O-glucoside (A3), cyanidin 3-O-[2′′-O-(xylosyl) 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-glucoside (A6), cyanidin 3-O-[2′′-O-(2′′′-O-(sinapoyl) xylosyl) 6′′-O-p-(coumaroyl) glucoside] 5-O-glucoside (A7) and cyanidin 3-O-[2′′-O-(2′′′-O-(sinapoyl) xylosyl) 6′′-O-(p-O-(glucosyl)-p-coumaroyl) glucoside] 5-O-glucoside (A10) were purified from low-P- and high-Suc-stressed At5MaT knockout Arabidopsis seedlings. Delphinidin 3-coumaroylrutinoside 5-glucoside and petunidin 3-coumaroylrutinoside 5-glucoside were purified from tomato fruit (Solanum lycopersicum). All other flavonoids used in this study were purchased from Apin Chemicals Ltd ( or Extrasynthese (

Quantitative RT-PCR

Total RNA was obtained using an RNeasy Plant Mini Kit (Qiagen; First-strand cDNA was synthesized using the adaptor oligoDT17 primer (Frohman et al., 1988) (Sigma, and SuperScript III (Invitrogen, from 5 μg of total RNA. Quantitative real-time RT-PCR was carried out using a SYBR Green I fluorescence-based assay kit (DyNAmo SYBR Green qPCR kit, Finnzymes, Specific primers were designed for At1g03940 (5′-CAGAGCCACTTTTACATTGAGC-3′ and 5′-TCATCCTTGTCTTCCTCGTTG-3′), At1g03495 (5′-TGGTCAGAGCCACTTTTACATTG-3′ and 5′-GACTTCGTCCTTGGCCTCTG-3′) and At3g29590 (5′-AGCCACGCTCCTCCACTATC-3′ and 5′-ACGGCATCTTTGTCGTCAGG-3′). All PCRs were performed using an Opticon 2 Real Time PCR machine (MJ Research, for 10 min at 95°C and then 40 cycles consisting of 15 sec at 95°C, 30 sec at 60°C and 20 sec at 72°C, followed by 10 min at 72°C. All quantifications were normalized to the elongation factor 1α (ef1α) cDNA fragment amplified under the same conditions using primers described previously (Czechowski et al., 2004). Each real-time assay was tested by a dissociation protocol to ensure that each amplicon was a single product. The data were analysed by the method described by Pfaffl (2001).

Evaluation of the dSpm insertion mutant of At3g29590

A dSpm insertion mutant of At3g29590 in A. thaliana (SM_3_35619) was obtained from the genome lab of the John Innes Centre (Norwich, UK). Pieces of leaves (6 mm in diameter) were excised from 4–6-week-old plants. Genomic DNA was extracted using the method described by Edwards et al. (1991) in a final volume of 100 μl, and 2 μl aliquots were used for subsequent PCR. dSpm lines were genotyped by amplifying the genomic DNA with either the right genomic PCR primer (RP) and a dSpm left border primer (BP, 5′-TACGAATAAGAGCGTCCATTTTAGAGT-3′), or two genomic PCR primers. The genomic primers used for At3g29590 were LP (5′-ACGACAAAGATGCCGTGTATC-3′) and RP (5′-CAACCTCAACACCACCACTTC-3′). The dSpm insertion site was determined by sequencing the PCR product using the BP oligo nucleotides. Total RNA from the wild-type and mutant plants was extracted using an RNeasy Plant Mini Kit (Qiagen), and cDNA was synthesized using SuperScript II RNase H reverse transcriptase according to the manufacturer’s instructions (Invitrogen). The lack of At3g29590 transcript was confirmed by RT-PCR using primers 5′-AGCCACGCTCCTCCACTATC-3′ and 5′-ACGGCATCTTTGTCGTCAGG-3′. The Ef1α gene was amplified as a control using primers described previously (Czechowski et al., 2004).

In vitro assay of recombinant proteins

The S-TAG expression vector pJAM1061 (Edwards et al., 1999) was cut with EcoRV and ligated with the Gateway™ vector conversion cassette [reading frame B; Invitrogen] to create a Gateway-compatible expression vector, pJAM1784. Full-length cDNAs of At3g29590, At1g03940 and At1g03495 were introduced into the Gateway™ system individually according to the manufacturer’s instructions. Full-length genes were amplified from A. thaliana (Columbia ecotype) cDNA using the following primers (with attB1/attB2 sites at the 5′ end of each primer): At1g03940, 5′-attB1-CCATGGTGGCTCATCTTCAACC-3′ and 5′-attB2-CTACGTTGCGAATTTCTTGATCC-3′; At1g03495, 5′-attB1-ACACCATGGCGGCTCAACTTC-3′ and 5′-attB2-CTACGTTGCGAATTTCTTGATCC-3′; At3g29590, 5′-attB1-CAGTAATGGTGAATTTCAACTCAGC-3′ and 5′-attB2-CATTTTTATTCCAACCCGATGGAG-3′. The entry clones (pDONR207-29590, pDONR207-03940 and pDONR207-03495) were obtained through recombination of the PCR products with pDONR207. After sequencing to confirm that the cDNA clones were error-free, the entry clones were then introduced into pJAM1784 to produce expression vectors pJAM1784-03940, pJAM1784-03495 and pJAM1784-29590. Recombinant proteins with an N-terminal S-TAG were expressed in E. coli BL-21 Rosetta cells transformed with the expression vectors. Cells were grown at 37°C with shaking at 250 rpm until the OD600 was 0.6. isopropyl β-D-1-thiogalactopyranoside was added to a concentration of 0.4 mm. Cells were then cultured at 30°C with shaking at 350 rpm for 4 h before harvesting. Harvested cells were resuspended in extraction buffer (20 μm potassium phosphate buffer, pH 7.0, with 1 mm EDTA), broken using a French press (three repeats), centrifuged at 38742 g for 30 min at 4°C, and then filtered through a 0.45 μm filter. The concentration of the fusion protein was determined using an S-TAG Rapid Assay Kit (Novagen, In the recombinant enzyme activity assays, the standard reaction mixture (100 μl) consisted of 20 mm potassium phosphate buffer (pH 7.0), 60 μm acyl donor, 120 μm acyl acceptor, 1 mm EDTA and enzyme (plus 10 mmβ-mercaptoethanol in the assay of At5MaT(At3g29590). After incubation at 30°C for 10 min, the reaction was terminated by adding 200 μl of ice-cold 0.5% v/v trifluoroacetic acid. The reaction mixture was then filtered with a 0.2 μm spin filter before being used for HPLC/LC/MS analysis.

To test the effect of the S-TAG on the activity of the recombinant proteins, the recombinant S-TAG fusion proteins At3AT1, At3AT2 and At5MaT were incubated with thrombin using a thrombin kit (Novagen) according to the manufacturer’s protocol. The activities of cut BAHD enzymes (with thrombin) and uncut BAHD enzymes (without thrombin) were then determined.

Construction of binary vectors, tobacco transformation, and confirmation of transgenic plants

The Gateway-compatible binary vector pJAM1502 was produced by cloning a double 35S CaMV promoter sequence and the CaMV terminator from pJIT60 (Guerineau and Mullineaux, 1993), cut with KpnI and XhoI, into the KpnI and SalI sites of pBin19. A Gateway™ destination cassette was then cloned into the SmaI site between the CaMV 35S promoter and the CaMV terminator sequences. Entry clones pDONR207-03940 and pDONR207-03495 containing the full-length sequences of At3AT1 and At3AT2 were introduced into pJAM1502 individually by the LR recombination reaction to produce expression vectors pJAM1502-03940 and pJAM1502-03495. The expression vectors were transformed using Agrobacterium tumefaciens (LBA4404) into N. tabacum var. Samsum using the leaf disc transformation method (Horsch et al., 1985). RNA was extracted from the flowers of the kanamycin-resistant plants, cDNA was prepared, and RT-PCR was carried out using the same primers as used for quantitative RT-PCR. The tobacco ubiquitin gene was amplified as a control using primers 5′-GTAGCTGAGGGGAGGAATGCAGA-3′ and 5′-CGCAACCTAGAAACCACCACG -3′.

Anthocyanin extraction and analysis

Plant material was harvested, ground in liquid nitrogen to produce a fine powder, and then immediately freeze-dried, and stored at -80°C until use. Anthocyanins and flavonoids were extracted with 40 μl 70% MeOH per mg DW of tissue, extracts were centrifuged at 16,000×g , and supernatants were filter spun using 0.2 μm filters before being used for analysis. Typically 20 μl of the filtered extracts were used for analysis.

HPLC was carried out on a Waters HPLC system using a 250 × 4.6 mm internal diameter Spherisorb® 5 μm C18 column (Waters, This was equilibrated with 96% solvent A (water/acetonitrile/formic acid, 87:3:10) and 4% solvent B (water/acetonitrile/formic acid, 40:50:10), and eluted with a gradient of increasing solvent B at a flow rate of 1 ml min−1. Gradient conditions were as follows: time 0, 96% A, 4% B; 20 min, 80% A, 20% B; 35 min, 60% A, 40% B; 40 min, 40% A, 60% B; 45 min, 10% A, 90% B; 55 min, 96% A, 4% B. The eluant was monitored with a diode array detector over the range 200–600 nm. Total anthocyanin contents were calculated from their peak areas at 520 nm in the chromatograms and compared with an external standard of cyanidin 3-glucoside as previously described (Miguel et al., 2004).

Structures were confirmed by LC/MS using a Thermo Finnigan Surveyor HPLC system (Thermo Scientific, equipped with a diode array (PDA) detector and a Deca XPplus ion trap mass spectrometer (Thermo Scientific). Anthocyanins were separated on a 3 μm, 100 × 2 mm Luna C18(2) column (Phenomenex, at 30°C and a flow rate of 230 μl min−1 using a linear gradient from 2% to 70% methanol vs. H2O/formic acid (99.9:0.1). Detection was by UV absorption at 520 ± 10 nm and positive ion electrospray (ESI) MS. Spray chamber conditions were 50 units sheath gas, 5 units auxiliary gas, 5.2 kV spray voltage and 350°C capillary temperature. General screening included data-dependent MS2 and MS3 scan events (35% collision energy, isolation width of 4 m/z) using dynamic exclusion. Anthocyanins of interest were targeted explicitly with an isolation width of 3 m/z, and, where necessary, reducing the value of Q from its default of 0.25 to as low as 0.2 to ensure that cyanidin fell within the detectable mass range following fragmentation.

The enzymatic products of At5MaT [cyanidin 3-O-glucoside 5-O-(6′′′-O-malonyl)glucoside] and of At3AT1 and At3AT2 [cyanidin (pelargonidin) 3-O-(6′′-O-p-coumaroyl)glucoside] were confirmed further by comparison with the products of Ss5MaT1 from S. splendens (Suzuki et al., 2001) and Pf3AT from P. frutescens (Yonekura-Sakakibara et al., 2000).

Stability tests of anthocyanins

Each anthocyanin was dissolved into potassium phosphate buffer (20 μm, pH 7.0) at a final concentration of 50 μm. For crude extracts from plant tissues, anthocyanin concentrations equivalent to 50 μm cyanidin 3-glucoside were used. The visible (VIS) spectra (400-700 nm) were measured automatically at appropriate time intervals. Based on the absorbance at the VISmax for each spectrum, the residual colour was calculated as a percentage of the initial absorbance. The stability of the anthocyanin could be evaluated on the basis of the half-life (t½), defined as the time to reach 50% residual colour.


J.L, C.F. and A.P. acknowledge support from the AgriFood Committee of Biotechnology and Biological Sciences Research Council (BBSRC) Project. A.J.M., C.M., K.L., L.H. and P.B. are supported by core strategic grants from BBSRC to the Institute of Food Research and the John Innes Centre. M.Y. and K.S. acknowledge grants from the Ministry of Education, Culture, Sports, Science and Technology, Japan, and from Core Research for Evolutional Science and Technology (CREST) of the Japan Science and Technology Agency (JST). Ms Chika Hirose and Ms Miyuki Ogawa are thanked for excellent technical assistance.