The biosynthesis of methyl anthranilate, the volatile compound responsible for the distinctive ‘foxy’ aroma and flavor of the Washington Concord grape (Vitis labrusca), involves an alcohol acyltransferase that catalyzes the formation of methyl anthranilate from anthraniloyl-coenzyme A (CoA) and methanol. Although methanol is a poor substrate in comparison with the co-substrate, high levels of this acyltransferase (0.5% of the total protein) combined with relatively high levels of this alcohol make this reaction possible in grapes. This 449 amino acid protein belongs to the BAHD family of acyltransferases, having 58% identity with the benzoyl CoA:benzyl alcohol benzoyl transferase from Clarkia. Both native and recombinant enzymes can use a broad range of acyl-CoAs and alcohols as substrates. The ability of Concord grape alcohol acyltransferase to accept a range of different CoA esters and alcohols suggests this to be a versatile ester-forming enzyme, similar to those of other fruits that than can produce a range of fruit esters based on the supply of appropriate substrates. Expression is coordinately regulated, with transcript, protein and enzyme activities coinciding with the accumulation of methyl anthranilate that occurs after the initiation of berry ripening. The majority of acyltransferase protein in grape tissues is localized to the outer fruit mesocarp, a result consistent with the fact that methyl anthranilate is released to the external environment throughout the ripening process. Wine grapes (Vitis vinifera) that accumulate neither anthranilate nor methyl anthranilate do not express this enzyme activity nor do they accumulate this protein.
The flavor of fresh produce such as fruits is one of the most important external characteristics determining their quality (Berger, 1991). Our assessment of the flavor of fruit is greatly influenced by volatile compounds in the fruit aroma mixture. The aroma of a particular fruit can often be characterized by the presence of a few compounds having a large impact on the odor, such as the γ- and δ-lactones of peaches (Do et al., 1969), the 3-methylbutyl acetate of bananas (Berger, 1991) and the 4-(4-hydroxyphenyl)-butan-2-one of raspberries (Larsen and Poll, 1990). The genus Vitis comprises over 50 species that include the commercially important Vitis vinifera and Vitis labrusca (Concord grapes) that are widely grown for production of wine and grape juice, respectively. The aroma of different grape species consists of many volatile organic compounds, including methyl anthranilate that is responsible for the characteristic aroma of Concord grapes (Power and Chestnut, 1921; Fuleki, 1972) and is the main component giving character to the odor of the very popular Welch's grape juice. Several volatile odor-active compounds (anthranilate and its methyl ester, furaneol and its methyl ester) implicated in the ‘foxy’ odor character of Concord grapes have been shown to accumulate at the onset of ripening (Shure and Acree, 1994, 1995). Other species of American grapes also contain anthranilic acid esters, but at levels much lower than those of Concord grapes. Methyl anthranilate also occurs in the essential oils of other plant species, such as neroli, ylang ylang, bergamot and jasmine, as well as in floral honey (Nozal et al., 2001). It is one of the key odorants of the flavor of Chinese jasmine green tea (Sugimoto et al., 2002) and also of orange blossom (Shaw and Wilson, 1980).
Methyl anthranilate was one of the first artificial flavor compounds to be described, although little is known about how it was discovered. The particular aroma properties of methyl anthranilate have made it a valuable component of the fragrance of perfumes and various cosmetics and it is the chief grape flavor compound in food, used extensively in the flavoring of soft drinks and of powdered drinks. Methyl anthranilate is used as a flavor enhancer and/or mask in various other products such as oral over-the-counter and prescription pharmaceuticals. This compound is an important industrial chemical, and synthetic forms of it are widely used as a bird and goose repellent for the protection of lawns, turf and crops because of its strong avian-specific repellent properties (Gibbs and Ng, 1977). Methyl anthranilate has also been used commercially as an oxidation inhibitor, as a sunscreen agent and as an intermediate for the synthesis of a broad range of chemicals, dyes and pharmaceuticals.
In nature, methyl anthranilate plays both defensive and attractive roles enhancing plant fitness and survival. This volatile aroma compound gives post-veraison Concord grape berries a floral, sweet and warm smell that may attract certain animals to eat the fruits in order to disperse the seed. Remarkably, in the case of the African army ant a trail pheromone composed of a two-component cocktail of methyl anthranilate and methyl nicotinate is stored within the post-pygidial gland and these two substances are released to mark foraging trails that direct workers to food sources (Oldham et al., 1994). Nevertheless, the smell of methyl anthranilate in wine is an undesirable character of Concord grapes, and there is no evidence of this trait in cultivated European wine grapes (V. vinifera).
Despite the high ecological and economic value of methyl anthranilate, little is known about how it is synthesized in plants. This paper describes the identification and purification of anthraniloyl-coenzyme A (CoA):methanol acyltransferase (AMAT) that is responsible for the formation of methyl anthranilate in Washington Concord grapes. The cloning, sequencing and expression of biochemically active AMAT in Escherichia coli showed that it belongs to the plant alcohol acyltransferase BAHD [benzylalcohol acetyltransferase, anthocyanin-O-hydroxycinnamoyltransferase, athranilate-N-hydroxy-cinnamoyl/benzoyltransferase and deacetylvindoline acetyltransferase (St-Pierre and De Luca, 2000)] family. The purified enzyme from Concord grapes and recombinant enzyme from E. coli catalyzed the conversion of anthraniloyl-CoA and methanol into methyl anthranilate. The ability of AMAT to accept a variety of CoA and alcohol substrates classifies it as an ester-forming acyltransferase that could be responsible for making a range of additional fruit esters of grapes, including the abundant methyl and ethyl butanoate esters of Concord grapes. Further studies showed that expression of AMAT in Concord grapes is triggered developmentally in response to veraison that correlates with the appearance of high levels of anthranilic acid and methanol as the grapes ripen.
Maturation curve of concord grapes
Concord grapes at different stages of maturation (from weeks 2 to 16 post-flowering, when the berries become very ripe) were analyzed for changes in total soluble solids content (concentration of molecules, predominantly sugars, in grapes that are truly soluble in an aqueous sample, termed the degree Brix value as determined by refractometry) and pH (Kennedy, 2002). The sharp increase in both degree Brix value and pH that occurred from weeks 9 to 11 post-flowering was accompanied with a gradual change in berry size and berry color from green to pink to purple, together with the typical appearance of the Concord grape aroma. This ripening stage, known as veraison, represents the beginning of fundamental physiological changes in berries that include the biosynthesis and accumulation of sugars, anthocyanins, aroma compounds and increased pH that is accompanied by rapid tissue softening.
Identification of volatile compounds in the grape juice by gas chromatography–mass spectroscopy
Volatile components of grape juice from three different grape varieties were extracted and analyzed by gas chromatography–mass spectroscopy (GC-MS) (Figure 1). Seven major volatile compounds were identified, including ethyl butanoate (peak 3), methyl-3-hydroxy butanoate (peak 6), ethyl-3-hydroxy butanoate (peak 7) and methyl anthranilate (peak 9) that are highly characteristic of Concord grapes, and small amounts of ethyl butanoate (peak 3) could be found in gamay but not the muscat varietal of V. vinifera (Figure 1). In Concord grape juice, butanoate esters are the most abundant volatiles, accounting for 63% of the total detected, while only 9% is accounted for by methyl anthranilate. Nevertheless, methyl anthranilate is the major character-determining compound, probably because of its low odor threshold values, whereas esters such as ethyl butyrate and ethyl 3-hydroxy butanoate contribute to the odor. The other three major volatile compounds, butyl acetate (1), 4-hydroxy-4-methyl-2-pentanone (4) and trans-2-hexenal (5), were detected in all three grape varieties analyzed. In addition, hexanal (2) was found in gamay and muscat varieties but not in Concord grapes. Linalool (8) was found in muscat grapes, consistent with the fact that muscat odor is mainly composed of the monoterpenes linalool and geraniol (Park et al., 1991) (Figure 1).
The relatively simple GC-MS profile of grape juice, with only four to seven major peaks detected, is quite different from the volatile profile of other fruits. For example, 14 esters, seven alcohols, five ketones, three aldehydes, three terpenic hydrocarbons and two acids were identified in the juice of yellow passion fruit (Jordan et al., 2002). A possible reason for the small number of volatiles detected may be related to the extraction solvent used. Methyl t-butyl ether (MTBE), used in our extractions, has been very effective in extracting volatiles from essential oils of plants such as basil (Gang et al., 2001;Iijima et al., 2004), but it may not be the best solvent for extracting low concentrations of aroma compounds from grape tissues that contain large amounts of water.
Anthranilic acid, methyl anthranilate and methanol accumulation during berry development
Methods that have been reported recently as suitable for analysis of fruit volatiles include solid phase microextraction (SPME) that is used to release volatile compounds directly from aqueous samples (Perez et al., 2002) and the solvent-assisted flavor evaporation (SAFE) technique (Supriyadi et al., 2003). Since GC-MS is not suitable for analysis of non-volatile anthranilic acid, solid phase extraction (SPE) was used to extract both anthranilic acid and methyl anthranilate from grape juice; these were then analyzed by HPLC. Both anthranilates in grape juice were detected by fluorescence detection and were identified by comparing their retention times with those of pure standards (Figure 2a) as well as by co-injection of juice samples with pure standards. While neither anthranilic acid nor methyl anthranilate was detected in berries at weeks 8 and 9, both compounds began to appear after week 10, after which they increased throughout the rest of the ripening period (Figure 2b). The concentration of anthranilic acid increased steadily and reached a maximum of 40.2 μm at week 16, whereas the concentration of methyl anthranilate reached a maximum of 41.1 μm at week 14 and decreased slightly (range 28.8–35.1 μm) during weeks 15 and 16. These measurements, however, do not take into account the continual release of volatile methyl anthranilate into the atmosphere during grape maturation. No anthranilic acid or methyl anthranilate was found in the juice at week 16 in either the muscat or the gamay varieties (Figure 2b) or throughout grape maturation (data not shown).
The accumulation profile of volatiles (Figure 2b) was similar to that reported previously (Shure and Acree, 1994), but the peak of methyl anthranilate [about 40 μm, equivalent to 40 μg gram fresh weight (gFW)] was almost 2000 times higher than previously reported (20 ng gFW−1). Although geographical or seasonal variations are factors that affect the production of fruit aroma, the method of sample preparation described here is a major improvement that could have resulted in the higher measurements that were recorded. Compared with the traditional method of extraction using organic solvent, solid phase extraction (SPE) is fast and easy to perform, with good recovery efficiency, and it only requires a small amount of starting material.
It is well known that many fruits (Frenkel et al., 1998), including grapes (Barnavon et al., 2001; Lee et al., 1979), will accumulate methanol as they ripen via the action of pectin methyl-esterase on methyl-esterified wall-associated pectins. The methanol content of Concord grapes was measured at weekly intervals between weeks 8 and 16 after flowering, as described in Klavons and Bennett (1986), and varied between 0.2 and 0.7 mm methanol throughout the pre-veraison and post-veraison period (data not shown). The methanol content of Concord grapes was consistent with previous results obtained for several varieties of grapes (Lee et al., 1979) as well as for other fruits like tomatoes (Frenkel et al., 1998).
Discovery of AMAT activity in Concord grape berries
The similarity of anthranilic acid to benzoic acid led to initial assays to identify an S-adenosyl-l-methionine-based O-methyltransferase (OMT)-like benzoic acid OMT (Murfitt et al., 2000) or salicylic acid OMT (Ross et al., 1999) in Concord grapes that is capable of converting anthranilic acid to methyl anthranilate (data not shown). Our inability to detect an OMT responsible for this reaction in Concord grape tissues led to the search an unusual alcohol acyltransferase-mediated biosynthesis of methyl anthranilate from anthraniloyl-CoA and methanol. When desalted crude extract was incubated with anthraniloyl-CoA (AA) and methanol, the newly formed methyl anthranilate (MA) was easily detected by HPLC (Figure 3a compared with Figure 3d, elution profile of AA and MA standards), whereas no methyl anthranilate could be detected in assays with boiled crude extract (Figure 3b). When similar extracts from V. vinifera grapes, such as gamay, were assayed, no AMAT activity was detectable (Figure 3c).
Enzyme assays of different Concord grape tissues showed that AMAT is almost exclusively found in berry mesocarp, while no activity was detected in berry skin (data not shown). Studies at different stages of berry development (weeks 8 to 16) showed that like anthranilic acid and methyl anthranilate, AMAT activity started to appear at week 10 when the berries began to change color (Figure 3e). This activity increased throughout berry development to reach a maximum level in very mature berries at week 16. In contrast, no AMAT activity was found in mature grapes of two V. vinifera representatives, gamay and muscat (Figure 3e), or throughout earlier stages of grape maturation (data not shown).
Purification of the AMAT enzyme and cDNA molecular cloning
Berry mesocarp was isolated and extracted from Concord grapes at weeks 14 to 16 in order to purify AMAT enzyme activity by ammonium sulfate fractionation, DEAE anion exchange (DE53), SP-Sepharose cation exchange and Sephacryl S-100 gel filtration column chromatography; this increased the specific activity of AMAT to 443 fkat mg−1 protein and gave a 200-fold purification over crude extract (Table 1) (Figure 4, lanes 2–6). The relative abundance of AMAT in grape mesocarp (0.5% of the total soluble protein) together with its relatively high isoelectric point (theoretical pI = 8.56) permitted the use of cation exchange chromatography that is rarely used in this context in the final purification procedure. Chromatography on Sephacryl S-100 provided a highly purified fraction as shown by SDS-PAGE, which contained one major protein band, with an apparent molecular mass of 50 kDa that is diagnostic of the size of other plant acyltransferases, including the alcohol acyltransferases from several fruit species (Figure 4, lanes 6 and 7). Further analysis of purified AMAT by isoelectric focusing combined with SDS-PAGE resolved it as a single spot giving an isoelectric point consistent with the theoretical pI (data not shown). Two other minor proteins of 40 kDa were also present that could be visualized by silver staining (Figure 4, lane 6) but not by Coomassie blue staining (Figure 4, lane 7). The presence and amount of the 50 kDa protein correlated closely with the pattern of AMAT activity in various fractions eluting from the final gel filtration column, but not with the other two minor bands. In fact, peptide sequencing of one of the minor proteins matched it to V. vinifera polygalacturonase-inhibiting protein (data not shown). These chromatographic procedures were also used to purify recombinant AMAT protein from E. coli to give highly active preparations that converted anthraniloyl-CoA and methanol into methyl anthranilate (Figure 4).
Table 1. Purification of AMAT to homogeneity from Concord grapes
Total protein (mg)
Total activity (fkat)
Specific activity (fkat mg−1)
Although N-terminal sequencing of the 50 kDa protein (Figure 4, lane 7) was unsuccessful, pure AMAT was digested with trypsin to yield four internal peptide sequences that matched closely with benzoyl-CoA benzyl alcohol acyltransferase (BEBT) from Clarkia and tobacco (D'Auria et al., 2002). Degenerate oligonucleotide primers from two peptide sequences and oligo-dT primers were used in reverse transcription (RT) and polymerase chain reactions (PCR) to yield a full-length cDNA clone that contained the four peptide sequences obtained from electrospray ionization tandem mass spectrometry (ESI-MS/MS) analysis. This 1840-nucleotide cDNA clone encoded a putative open reading frame (ORF) of 449 amino acids with the first methionine codon at positions 117–119. The four internal peptides matched exactly the putative AMAT ORF from Concord grapes. The calculated molecular weight of the protein encoded by the open reading frame was 50.2 kDa.
A database search for proteins similar to this deduced AMAT revealed that V. vinifera also contained a putative alcohol acyltransferase (AAT) with 81% identity to AMAT whose theoretical pI was 7.5, a full unit below that of AMAT from Concord grape. It is well understood that sequence similarity to a gene with known function does not necessarily correlate with the biochemical role of the product of this AMAT-like gene from V. vinifera and this latter gene should be functionally expressed to determine its biological role. In addition, AMAT had 58–63% sequence identity to several plant acyltransferases, including a wound-inducible Clarkia BEBT involved in the formation of benzyl benzoate floral volatiles, the Arabidopsis gene CHAT, involved in producing green leaf volatiles (D'Auria et al., 2002), and the melon AAT1 that is responsible for formation of aroma volatile esters during melon ripening (Yahyaoui et al., 2002). The alignment of several plant AATs also indicated that AMAT belongs to the BAHD superfamily of acyltransferases (St-Pierre and De Luca, 2000). Members of this acyltransferase family all contain the two conserved motifs HTM(S/A)D(A/G) and DFGWG(K/E)(P/A) that are suggested to be involved in substrate binding and catalytic function. In tobacco a hypersensitive related protein, hsr201, induced by phytopathogenic bacteria (Czernic et al., 1996), was recently found to have activity similar to Clarkia BEBT (D'Auria, 2002). In petunia, a similar protein was also shown to be responsible for the production of the floral scent compounds benzyl benzoate and phenylethyl benzoate (Boatright et al., 2004), depending on the alcohol provided. The purified AMAT showed a substantial level of BEBT activity, even though Concord grapes do not accumulate benzyl benzoate. These closely related genes appear to be responsible for the formation of different plant volatile esters in leaves, flowers and fruits.
Functional expression of an AMAT cDNA clone in E. coli and biochemical characterization of AMAT protein
To verify that the isolated cDNA clone encodes AMAT, its coding region was mobilized into a pCRT7/CT-TOPO TA vector as a non-fusion protein and was expressed in E. coli for functional analysis. Lysates of cells carrying the complete ORF of 449 codons had substantial AMAT activity after isopropyl-beta-d-thiogalactopyranoside (IPTG) induction, while cells carrying the same plasmids but with AMAT cDNA in a reversed orientation, or without AMAT cDNA, did not have any AMAT activity.
The AMAT enzyme was purified to homogeneity from the crude E. coli extract with the same protocol that was used for the native enzyme isolated from grape mesocarp. Both plant (Figure 4, lanes 6 and 7) and recombinant AMAT (Figure 4, lane 8) were biochemically active (Figure 4, compare panels a–c) and had an apparent molecular mass of 50 kDa as determined by Sephacryl S-100 gel filtration and by SDS-PAGE, confirming that the active enzyme behaves as a monomer. Biochemical characterization of purified plant and recombinant E. coli AMAT showed they had very similar activities toward the different substrates tested (Table 2). Their Kms for anthraniloyl-CoA were 2.51 and 4.78 μm, respectively when methanol was used as the alcohol substrate, suggesting that anthraniloyl-CoA is a good substrate for AMAT in vivo. The Km for methanol was 14.97 mm for plant protein and 31.8 mm for recombinant protein, consistent with the range of Km values obtained for alcohol substrates with other alcohol acyltransferases. Melon AAT had a Km of 8.0 mm for 1-butanol and 1.4 mm for 1-hexanol (Yahyaoui et al., 2002), whereas strawberry SAAT had a Km of 46.1 mm for 1-butanol, 8.9 mm for 1-hexanol and 5.7 mm for 1-octanol (Aharoni et al., 2000). The Kcat values of AMAT for methanol were 0.0035 and 0.0058 sec−1 for native and recombinant protein, respectively. The high Km of AMAT for methanol compared with anthraniloyl-CoA could greatly decrease the overall catalytic efficiency of this reaction in Concord grapes. The average methanol concentration in grape mesocarp between 8 and 16 weeks after flowering was 0.5 mm, although the concentration at the site of methyl anthranilate biosynthesis in the cell remains to be determined. In addition the abundance of active AMAT protein in grape mesocarp (0.5% of total protein) could also compensate for the low affinity of this enzyme for the methanol co-substrate. Although AMAT protein was only 58% identical to that of the Clarkia BEBT, this enzyme was also active with benzoyl-CoA and benzyl alcohol substrates. When benzyl alcohol was used as the alcohol substrate, its Km values for benzoyl-CoA in native (11.80 μm) and recombinant protein (8.42 μm) were similar to those of Clarkia BEBT (20.5 μm) (D'Auria et al., 2002) (Table 3). Although benzoyl-CoA and anthraniloyl-CoA were good substrates for AMAT with Kms in the μm range, acetyl-CoA was not and the Km for acetyl-CoA could not be measured accurately.
Table 2. Kinetic parameters of AMAT
N (nmol−1 sec−1)
R (nmol−1 sec−1)
N = native purified protein from mature grape; R = recombinant purified protein from E. coli.
Benzoyl-CoA (with benzyl alcohol)
Anthraniloyl-CoA (with methanol)
Methanol (with anthraniloyl-CoA)
Table 3. Relative activity of AMAT with a variety of substrate
aAMAT activities with acetyl-CoA (0.005 pkat μg−1 protein), benzoyl-CoA (0.034 pkat μg−1 protein) and anthranoyl-CoA (2.85 pkat μg−1 protein) with methanol were arbitrarily set as 100.
Similar to Clarkia BEBT, AMAT also used a variety of alcohol substrates, displaying the highest enzyme activities with benzyl alcohol, butanol and octanol when assayed with both benzoyl-CoA and anthraniloyl-CoA. While Clarkia BEBT displayed no activity with ethanol, AMAT was active with both ethanol and methanol. However, the lack of enzyme activity of Clarkia BEBT with methanol and ethanol may be a result of the low concentrations (1 mm) of alcohol (D'Auria et al., 2002) used in BEBT assays compared with the 10 mm concentration of alcohol required for AMAT enzyme assays (Table 3).
Anthraniloyl-CoA:methanol acyltransferase activity was also stable over a relatively broad pH range, with no apparent loss of activity when assayed between pH 7.0 and 9.0, since 65 and 75% of maximal enzyme activities were observed, respectively, when AMAT was assayed at pH 6.5 and 9.5. In addition, the enzyme was stable for 30 min at 35°C, but lost 80% of its activity after incubation at 45°C for 5 min. The monovalent and divalent cations K+, Ca2+, Mg2+ and Mn2+ could increase enzyme activity by 50–70%, whereas heavy metals such as Cu2+ and Zn2+ completely inactivated AMAT.
Expression of AMAT transcripts, protein and enzyme activity during ripening of Concord grapes
Total RNA was isolated from Concord berry mesocarp at each stage of development of the berry and RT-PCR was used to test if AMAT gene expression was correlated with appearance and accumulation of methyl anthranilate. Specific primers for RT-PCR based on AMAT coding and 3′ untranslated regions showed that AMAT transcripts only began to appear coordinately at week 10 with the appearance of methyl anthranilate and increased continuously throughout grape ripening (Figure 5).
Protein extracts from different stages of berry development were separated by SDS-PAGE followed by electrophoretic transfer to polyvinylidene fluoride (PVDF) membranes. The AMAT protein was detected using a polyclonal antibody made against Clarkia BEBT and was quantified by comparison to known amounts of purified AMAT protein. Very low levels of AMAT protein were first detected at week 10, after which higher levels began to accumulate. Densitometric scanning of immunoblots showed that levels of AMAT protein increased coordinately with metabolites (Figure 2b), enzyme activity (Figure 3e) and transcript levels (Figure 5) throughout the ripening period, until berries were very mature at week 16 (Figure 5). Grapes from week 14, when AMAT activity and protein were maximal (Figure 5) were also used to determine the spatial distribution of AMAT within the fruit. Protein extracts from different parts of the grape were prepared for immunoblot analysis and showed that more than 90% of AMAT protein could be localized to the outer mesocarp (11.5–14.5 mm from the grape center) whereas only 4% of AMAT protein was found in either the skin or in the 7–11.5 mm mesocarp layer, respectively and only 2% of AMAT protein was detected in the 0–7 mm mesocarp layer (data not shown). The proximity of AMAT protein to the outside of the grape is consistent with the fact that methyl anthranilate is released to the external environment throughout the ripening process.
Analysis of other V. labrusca varieties for AMAT protein and methyl anthranilate production
Vitis vinifera grapes such as gamay and muscat do not accumulate methyl anthranilate, their crude protein extracts did not have AMAT activity and no protein was detected that reacted to the Clarkia BEBT antibody (Figure 6). Immunoblot analysis of two other V. labrusca varieties, both of which are completely North American in origin, Blue Star and Athens, showed similar levels of AMAT protein and enzyme activity to those of Concord grapes. However, these two varieties had 70% lower levels of anthranilic acid and methyl anthranilate than those found in Concord grapes. These data suggest that while AMAT catalyzes the last step of formation of methyl anthranilate, the production of methyl anthranilate also appears to be regulated by additional factors such as substrate supply, although this remains to be demonstrated.
AMAT is an alcohol acyltransferase belonging to the BAHD family
Methyl anthranilate is a member of the group of volatile esters that include methyl esters of various kinds and constitutes one of the largest and most important groups of fruit aroma compounds. Such esters constitute the characteristic aroma of cantaloupe (Cucumis melo L.), apple (Malus domestica), pear (Pyrus communis), banana (Musa sapientum) and strawberry (Fragaria ananassa). The formation of these esters in fruit aroma compounds is mediated by AAT, which belong to the well-known and versatile BAHD acyltransferase family. Here we report an anthraniloyl-CoA:methanol acyltransferase that produces methyl anthranilate from anthraniloyl-CoA and methanol in V. labrusca (Concord grape).
The sequence alignments and phylogenetic analysis (Figure 7) of AMAT in comparison with several plant alcohol acyltransferases that have been characterized at the biochemical level show that AMAT is closely related (58–63% sequence identity) to Clarkia BEBT that produces benzyl benzoate floral volatiles, to benzenoid-forming petunia BPBT and to a BEBT that is induced in tobacco by mechanical wounding. Anthraniloyl-CoA:methanol acyltransferase is also closely associated with Arabidopsis CHAT that is involved in making green leaf volatiles (D'Auria, 2002) and melon AAT that makes aroma volatile esters during melon ripening (Yahyaoui et al., 2002). In V. vinifera, a putative AAT1 that is 81% identical to AMAT was identified whose biological role needs to be determined.
Two mechanisms for the generation of methyl esters in plants
The striking structural similarity of methyl anthranilate, methyl benzoate and methyl salicylate raises questions about why these esters are made by completely different enzymatic mechanisms. Methyl benzoate, the major volatile scent component of snapdragon (Antirrhinum majus) flowers, is formed from benzoic acid by S-adenosyl-l-methionine:benzoic acid carboxyl methyltransferase (BAMT) (Dudareva et al., 2000; Murfitt et al., 2000). Methyl salicylate, a component of the floral scent of Clarkia breweri, is formed from salicylic acid by S-adenosyl-l-methionine:salicylic acid carboxyl methyltransferase (SAMT) (Dudareva et al., 1998; Ross et al., 1999). Both BAMT and SAMT belong to a structurally related group of methyltransferases known as the SABATH family (D'Auria et al., 2003); three genes for this family –SAMT, BAMT and theobromine synthase (Kato et al., 1999, 2000; Ogawa et al., 2001) have been isolated and characterized.
The AMAT member of the BAHD family and those proteins coded by the SABATH gene family are an interesting example of convergent evolution (Pichersky and Gang, 2000), in which enzymes from two unrelated families make the same class of methyl ester volatiles. This similarity extends to the multifunctional nature of both classes of enzymes as a consequence of their maintenance of broad substrate specificities. This trait presumably allows the organism to adapt rapidly to a changing external environment by facilitating the ‘adaptation chemistry’ without the need to evolve a new enzyme or gene (Schwab, 2003). However, methyltransferases are only capable of transferring methyl groups from S-adenosyl-l-methionine to a range of methyl acceptors, whereas some members of the BAHD family can accept a range of CoA esters and a variety of alcohol substrates that lead to a variety of different end products. The recently reported crystal structure of Clarkia SAMT coded by the SABATH gene family (Zubieta et al., 2003), and the recent structure published for the BAHD family member Rauvolfia vinorine synthase (Ma et al., 2005), will provide valuable structural tools for the functional characterization and comparative analysis of members of each family.
Methyl anthranilate production is regulated by enzyme and substrate levels
The pattern of AMAT enzyme activity, the appearance of AMAT protein and AMAT gene expression is coordinately regulated during the ripening process in Concord grapes in a way which precisely follows the pattern of accumulation of both anthranilic acid and methyl anthranilate. This fruit- and ripening-specific regulation is consistent with the expression patterns also found for AAT1 and SAAT in relation to fruit ester formation in melons and strawberries, respectively (Aharoni et al., 2000; Yahyaoui et al., 2002). However, in grapes, the biochemical role played by AMAT in the production of methyl anthranilate appears to be accompanied by uncharacterized coordinated mechanisms that must supply the anthranilic acid/anthraniloyl-CoA and methanol precursors required to create the methyl anthranilate pathway, as suggested by the 70% lower level of both anthranilic acid and methyl anthranilate occurring in two other V. labrusca hybrid varieties known as Athens and Blue Star. In addition, the relatively high levels of anthranilic acid present in Concord grapes suggest that the production of the CoA ester of this metabolite may be critical for defining the rate of production of methyl anthranilate.
AMAT-based biosynthesis of methyl and ethyl esters of 3-hydroxybutanoate
The capability of AMAT and other AATs to produce esters from a wide range of alcohols and acyl-CoAs also shows how changes in substrate supply could lead to the evolution of new aroma-producing compounds (Aharoni et al., 2000; Beekwilder et al., 2004; Yahyaoui et al., 2002). In Concord grapes the methyl and ethyl esters of 3-hydroxybutanoate constitute 30 and 26%, respectively, of the total volatiles. It is possible that these methyl and ethyl esters are also formed via the same AMAT enzyme reaction, since AMAT is also able to convert butanoyl-CoA and methanol or ethanol to the respective esters (data not shown). In this context, in vitro enzyme assays with AMAT showed that ethanol is a better substrate than methanol, and since maturing Concord grapes contain both methanol and ethanol it is intriguing why there is no detectable level of ethyl anthranilate in the ripening fruit.
The source of methanol and anthranilic acid for making methyl anthranilate
The accumulation of methyl anthranilate during grape ripening is accompanied by the accumulation of both anthranilic acid and methanol that are derived, respectively, from the biosynthesis/breakdown of tryptophan and the release of methyl groups from the pectin methylesterase (PME)-mediated hydrolysis of pectin polymers (Lee et al., 1979). It remains to be established if the surplus anthranilic acid used to make methyl anthranilate is actually derived from de novo biosynthesis through increased anthranilate synthase activity, or if it is a breakdown product of tryptophan metabolism. It has been reported that the amounts of tryptophan increase significantly during grape maturation to be converted into aroma compounds that cause an ‘untypical ageing off-flavor’ in V. vinifera wines (Hoenicke et al., 2001). Although it has not been studied, similar reactions to supply the intermediates for methyl anthranilate biosynthesis may also be occurring in Concord grapes.
The action of PME is a ubiquitous reaction in plants that specifically catalyzes methyl ester hydrolysis of carboxymethyl galacturonic acid units of pectin to release methanol. The regulation of methanol production by PME in fruits has been supported by genetic and biochemical evidence in tomatoes (Frenkel et al., 1998) and in grapes (Barnavon et al., 2001). Measurement of PME activities in Concord grapes and several other grape varieties showed that in general V. labrusca varieties have higher pectin content and also PME activity than those of V. vinifera. Of these, Concord grapes have the highest PME activity. Pectin methylesterase activity increased continuously throughout grape maturation and reached a maximum at harvest time (Lee et al., 1979). The high levels of pectin in grape berries, the timing of PME activity during Concord grape ripening and the abundance of AMAT protein (Table 1; approximately 0.5% of the total soluble protein), together with the average levels of methanol found in berries in weeks 8 to 16 (0.5 mm), are coincident with the increasing levels of anthranilic acid plus methyl anthranilate and increased AMAT gene expression described in this report.
In summary, the unique substrate requirements of the AMAT enzyme described in this study for the production of methyl anthranilate in Concord grapes, together with data obtained from other fruit species, suggest that highly similar but very versatile biological mechanisms are responsible for the production of fruit volatile esters. This study suggests that slight changes in substrate availability could dramatically alter the volatile esters produced by the organism, based on the ability of enzymes coded by members of the gene family to accommodate a variety of CoA ester and alcohol substrates. Future studies will explore if environmental conditions may trigger slight changes in the supply of various CoA and alcohol substrates to produce modified flavor/aroma profiles in Concord and even in V. vinifera wine grapes as such compounds play important roles in defining the quality parameters of wine (Aubry et al., 1997).
Grape berries were harvested weekly in the 2003 season, starting from the second week after flowering to week 16, after which the completely mature berry crop was harvested. Concord grapes (Vitis labrusca cv. Concord) were harvested from the G. H. Wiley vineyard (St Catharines, ON, Canada). Berries of V. labrusca (Blue Star and Athens) and V. vinifera (gamay and muscat) were harvested from the University of Guelph Grape Research Station, Vineland, Ontario. Fresh berries were used for sugar determination (degree Brix) and pH measurement; the rest of the berries were frozen and stored at −80°C for future use.
Berry total soluble solids determination and pH measurement
Berries were randomly picked from the middle of clusters of three to five grapes at each stage of grape maturation and were homogenized by hand squeezing the grapes in a plastic bag to produce grape juice. When frozen berries were used, they were thawed at room temperature for 30 min before juice extraction. Total soluble solids content (degree Brix) was determined with a hand-held Brix refractometer and pH was measured with a pH meter.
Volatile extraction and GC/MS analysis
Concord, gamay and muscat grapes (10 g) were extracted to obtain approximately 7 ml of juice and were mixed with 2 ml of MTBE before centrifugation for 10 min at 1000 g at 4°C to obtain the MTBE fraction. This MTBE extraction procedure was repeated twice and the pooled MTBE fractions (volatile fraction) were dehydrated with anhydrous MgSO4 before being concentrated under gentle nitrogen gas flow to 150 μl for GC-MS analysis.
A Shimadzu QP-5000 system (Shimadzu, Columbia, MD, USA) equipped with a Shimadzu GC-17 gas chromatograph was used for GC-MS analysis of the volatile compounds. A CP-5 column (30 m × 0.32 mm internal diameter × 1 μm film thickness; Alltech Associates, Deerfield, IL, USA) was used for separation of these compounds. Ultrapure helium was used as the carrier gas at a rate of 1.3 ml min−1. Two microliters from each sample was injected using a Shimadzu AOC-20i autoinjector with split mode (1:50). The GC conditions were set for 2 min at a 50°C initial temperature, followed by a 10°C min−1 temperature gradient until it reached 275°C. The injector and detector temperatures were set at 250 and 270°C, respectively. Eluted volatile peaks were identified by their retention times and by their mass fragmentation patterns compared with standard compounds, or their masses were compared with compounds in a mass spectra database library (Nist 62; Shimadzu).
Anthranilic acid and methyl anthranilate extraction and HPLC analysis
Grape extracts (1.5 ml) were centrifuged at 10 000 g at room temperature for 5 min using a microcentrifuge. One milliliter of each supernatant was loaded into an Envi-C18 column (1 ml bed volume; Supelco, Toronto, ON, Canada) that had been pre-conditioned with 2 ml methanol followed by 2 ml H2O. After washing the column with 1 ml H2O, bound substances were eluted in 1 ml methanol. Using this protocol the extraction efficiencies of anthranilic acid and methyl anthranilate standards were 44.3 and 48.8%, respectively, and this information was used in the final calculations to determine the levels of these metabolites in grape extracts. The methanol eluate was analyzed by HPLC on a Nova Pak C18 column (Waters, Milford, MA, USA) and detected with a Waters model 2475 multiwavelength fluorescence detector with the excitation and emission wavelengths set at 340 and 400 nm, respectively. The column was eluted with solvents A (100% methanol) and B (H2O with 0.1% acetic acid) using the following conditions at a flow rate of 1 ml min−1: 0–3 min, 10–90% A; 3–10 min, 90% A; 10–13 min, 90–10% A; 13–18 min, 10% A. The retention times for anthranilic acid and methyl anthranilate were 8.6 and 9.8 min, respectively. For calibration of the standard curve, 2, 4, 6, 8 or 10 μl pure anthranilic acid or methyl anthranilate standard (at a concentration of 10 μm) was injected into this column and eluted by the same procedure, and the calibration curves were obtained by plotting the amounts of standards applied versus the peak areas.
Extraction of methanol and analysis
Fifty grams of berries were blended in 200 ml water and the homogenate was filtered through one layer of Miracloth. The berry debris was washed once with 50 ml of water, filtered and the filtrates were combined. The filtered homogenate was distilled and 100 ml distillate was collected. The methanol concentration in the distillate was measured by using alcohol oxidase (Klavons and Bennett, 1986) which converts methanol into formaldehyde. The formaldehyde then reacts with acetylacetone to yield a product that is measured at 412 nm.
Enzymatic synthesis of anthraniloyl-CoA
Anthraniloyl-CoA was synthesized and purified according to a method described in Beuerle and Pichersky (2002) to produce a powder. While in solution, the product showed an absorbance spectrum similar to that of CoA, with a peak absorbance at 260 nm. When the product was hydrolyzed in the presence of 1 n NaOH, anthranilic acid was released, as detected by HPLC and fluorescence detection. Electron spray ionization mass spectrometry (ESI-MS) analysis confirmed that the synthesized product was anthraniloyl-CoA, with an estimated molecular mass of 886 Da. Purified anthraniloyl-CoA was dissolved in H2O and its concentration was calculated using an A260 CoA standard curve.
AMAT enzyme extraction and assay
Concord grape berries were homogenized using a pre-chilled mortar and pestle in the presence of 3× ice-cold extraction buffer [w/v, 250 mm Tris-HCl, pH 8.0, 5 mm Na2S2O5, 10% glycerol, 1 mm DTT, 1 mM phenylmethyl sulfonyl fluoride and 10% insoluble polyvinylpolypyrrolidone (PVPP)]. The slurry was filtered through one layer of Miracloth and the filtrate was centrifuged at 7250 g for 10 min at 4°C. The supernatant was desalted with a PD-10 desalting column (GE Healthcare, Montreal, Que, Canada) to yield the crude extract.
The enzyme assay mixture was prepared by adding the following to a 1.5 ml microcentrifuge tube: 20 μl 5× assay buffer (250 mm Tris-HCl, pH 7.5), 20 μl crude extract, 10 μl 10 mm methanol, 10 μl 100 μm anthraniloyl-CoA and 40 μl H2O to bring the final volume to 100 μl. The reaction was carried out at room temperature for 20 min, and stopped by adding 5 μl 1 n HCl. The reaction mix was diluted 10 times with 900 μl H2O and was filtered through a 0.45 μm membrane (Waters, Toronto, ON, Canada). Filtered samples (50 μl) were analyzed by HPLC, using an isocratic method with 70% solution A (100% methanol) and 30% solution B (H2O with 0.1% acetic acid) at a flow rate of 1 ml min−1. The retention time for methyl anthranilate was 3.7 min. The BEBT activity of AMAT was assayed as described by D'Auria et al. (2003) with benzoyl-CoA and benzoyl alcohol as substrates.
AMAT protein purification
After removing the skins and seeds, 90 g of ripe week 16 Concord berries were homogenized in a blender together with 3× ice-cold extraction buffer. After filtration through Miracloth, the filtrate was sequentially treated with ammonium sulfate and centrifuged at 11 950 g for 15 min to isolate the protein precipitating between 35 and 70%. This 70% ammonium sulfate pellet was dissolved in 10 ml of buffer A (50 mm Tris-HCl, pH 7.5, 10% glycerol, 1 mm DTT) and desalted using a PD-10 column (GE Healthcare). The desalted fraction (14 ml) was diluted 1:1 with buffer A and loaded onto a DEAE cellulose column (10 ml of DE-53; Whatman, Fairfield, NJ, USA) that was pre-equilibrated with buffer A. The DE-53 flow-through (28 ml) that contained AMAT activity was collected and loaded onto a SP-Sepharose column (5 ml bed volume) pre-equilibrated with buffer A. The column was washed with 20 ml of buffer A, and bound proteins were eluted with a 40 ml linear gradient of NaCl from 0 to 400 mm in buffer A (1 ml min−1). Fractions of 1 ml were collected and were assayed for AMAT activity. The fractions with the highest AMAT activity (180–190 mm NaCl) were pooled and loaded into a Sephacryl S-100 gel filtration column (GE Healthcare) pre-equilibrated with buffer B (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 1 mm DTT, 10% glycerol), and eluted with buffer B with a flow rate of 1 ml min−1. After skipping the first 40 ml, 1.5 ml fractions were collected and assayed for AMAT activity. The protein content and purity of the AMAT-containing fractions were examined by SDS-PAGE gel electrophoresis followed by silver staining. The fractions with the highest AMAT activity also had the largest amount of AMAT protein and showed the highest degree of purity. The active fractions were combined (6 ml) and concentrated to 500 μl using an Ultrafree-15 filter device (Millipore, Mississauga, ON, Canada). The concentrated protein sample was used for protein sequencing and for enzyme characterization.
Protein sequence determination
The concentrated AMAT protein was submitted to electrophoresis on a 10% acrylamide SDS-PAGE gel. After staining the protein gel lightly with Coomassie brilliant blue R250, the major band of 50 kDa and one minor band of 40 kDa were excised from the gel to be digested with trypsin. The extracted peptides were subject to ESI-MS/MS analysis in the Proteomics Core Facility (Southwest Environmental Health Sciences Center and Arizona Cancer Center, University of Arizona, Tucson, AZ, USA) as described in Iijima et al. (2004).
Alternatively, the SDS-PAGE gel was blotted onto a PVDF membrane using the CAPS (3-[cyclohexylamino] 1-propane-sulfonic acid) buffer system (10 mm CAPS, pH 11, 10% methanol) and semi-dry blotting apparatus (Bio-Rad, Mississauga, ON, Canada). The transfer was set at 25 V, 100 mm for 60 min. After transfer, the PVDF membrane was rinsed in water for 5 min, briefly stained with Ponceau S (0.2% solution in 1% acetic acid) for 1–2 min and rinsed with deionized water. The stained 50 kDa band was removed using a sharp razor blade, transferred to a microcentrifuge tube and rinsed extensively with deionized water. Mass spectrometric analysis of this 50 kDa protein was carried out in the Eastern Quebec Proteomics Center (Sainte-Foy, Quebec, Canada).
AMAT cDNA cloning and characterization
Several degenerate oligonucleotides were designed based on the peptide sequences obtained by mass spectrometry. A sense 17-mer oligonucleotide for the amino acid sequence VADLMV (positions 351 to 356 from the N-terminus) 5′-GT(ACGT)GC(ACGT)GA(CT)(CT)T(ACGT)ATGGT-3′ and oligo d(T) primer were used to amplify a fragment of 650 bp, using cDNA made from RNA of week 16 Concord berries. This fragment included the 300 bp nucleotides of the AMAT cDNA sequence containing the putative C-terminal open reading frame and 350 bp nucleotides of the 3′ untranslated region. Another sense 18-mer oligonucleotide for the amino acid sequence CGGFIFA (positions 155 to 161 from the N-terminus) 5′-TG(ACGT)GG(ACGT)GG(ACGT)TT(CT)AT(ACT)TT(CT)GC-3′ and two antisense primers based on the amplified fragment sequence 5′-GACCATGAGATCCGCTAC-3′ and 5′-GCGTGACATCTGAAACAG-3′ were used to obtain another 830 bp nucleotide coding region. Finally, the N-terminal of the AMAT sequence was obtained by 5′ rapid amplification of cDNA ends (5′-RACE) using the primer set 5′-GTGGAGCCAGGGACATCATACAGAA-3′ and 5′-CTCAAAGCAGGGACACATTGGCTGAA-3′. The full length AMAT cDNA clone of 1350 bp nucleotides was amplified using the primers 5′-AATGGCATCACCGTCGTCTCCTC-3′ and 5′-GAGCATGGATGTAATTAACAGCTC-3′ and cloned into the pCRT7/CT-TOPO TA vector.
Recombinant protein expression and purification
The pCRT7/CT-TOPO plasmid containing the AMAT coding region was introduced into E. coli BL21 cells to produce recombinant AMAT protein as described previously (Wang and Pichersky, 1998), except that E. coli cells were extracted in buffer A. The soluble recombinant AMAT protein was purified using the same method employed for the purification of AMAT from Concord berries.
AMAT kinetic analysis
Kinetic studies were performed after conditions were established to ensure that the enzymatic reaction velocity was linear during the incubation period. To determine the Km of anthraniloyl-CoA and methanol, one substrate concentration was fixed at a saturated level and the concentration of the substrate to be measured was varied. For the measurement of Km for benzoyl-CoA, benzoyl alcohol was kept at a concentration of 1 mm and different concentrations of radioactive benzoyl-CoA were used. Lineweaver–Burk plots were constructed to obtain the Km and Vmax, and the Kcat was calculated accordingly. Final results are an average of at least two independent assays.
Effect of temperature on AMAT activity
The AMAT protein was incubated at different temperatures, ranging from 4 to 65°C for 5 min and then chilled on ice. Protein samples incubated at each temperature were then used for the enzyme assays. Results represent an average of at least two independent assays.
pH optimum of AMAT activity
The optimum pH for AMAT activity was determined using assay buffer (Tris-HCl buffer system) with pH ranging from 6.0 to 9.5. Results represent an average of at least two independent assays.
Enzyme assays were performed with one of the following cations present in the assay mix at a final concentration of 5 mm: Na+, K+, Ca2+, Cu2+, Mg2+, Mn2+ and EDTA. Results represent an average of at least two independent assays.
Determination of molecular weight
Fractions with peak AMAT activity from SP-Sepharose were pooled and loaded onto a Sephacryl S-100 gel filtration unit together with bovine serum albumin protein carrier (2 mg). Column fractions were assayed for AMAT activity and the elution profile was compared with the elution of standard molecular weight markers.
AMAT gene expression by RT-PCR
Ribonucleic acid was isolated from Concord berries using the Concert Plant RNA Reagent (Invitrogen, Carlsbad, CA, USA). One microgram of total RNA was used to produce cDNAs using oligo d(T) primer with the Takara RNA PCR kit according to the manufacturer's instructions (Fisher Scientific, Toronto, ON, Canada). Two sets of gene-specific primers were generated to analyze the expression pattern of AMAT [390 bp at the N-terminal of the coding region using 5′-AATGGCATCACCGTCGTCTCCTC-3′ (forward) 5′-GTGGAGCCAGGGACATCATACAGAA-3′ (reverse); 240 bp C-terminal untranslated region using 5′-TGGAAAGGTTTGAACAGGAG-3′ (forward) and 5′-CAAGGCATGTTTATACTTGC-3′ (reverse)]. Reverse transcriptase PCR products were resolved on a 1% agarose gel containing ethidium bromide (EtBr), and quantified by measuring the fluorescence intensity of the PCR products with a FLA-3000 fluorescent image analyzer. Analysis of the products obtained with different amounts of cDNA (0.25, 0.5, 0.75, 1 μl) in PCR reactions showed a linear increase in formation of reaction product. Using 0.5 μl of cDNA as a template, the PCR products of different amplification cycles were analyzed and were shown to increase linearly between PCR cycles 21 to 30. Thus 0.5 μl of cDNA and 25 PCR cycles were chosen as the optimal conditions for further studies. To ensure that an equal amount of RNA from each stage was used, and also for calibration purposes, 1 μg of total RNA from each stage were resolved on 1% (w/v) agarose-formaldehyde gels containing EtBr. The total fluorescence intensity of 18S and 28S from each RNA sample was quantified and used to normalize RNA levels and to control for discrepancies in spectrophotometer concentration readings. An RT-PCR reaction with RNA from week 16 berries, but without reverse transcriptase, was carried out as a negative control. Concord grape genomic DNA was also used as a template for RT-PCR in order to test for contamination with genomic DNA in the RNA sample.
Immunological determination of levels of AMAT protein in grape tissues
Crude extracts were prepared for Western blotting as described above, except that no insoluble PVPP was used in the extraction buffer. Samples (24 μl of each extract, equivalent to 10 mg fresh weight of mesocarp) were submitted to 10% SDS-PAGE along with known amount of purified AMAT, and proteins were transferred to PVDF membranes using Bio-Rad semi-dry membrane transfer apparatus. Proteins were transferred at room temperature in transfer buffer (25 mm Tris-HCl, pH 8.3–8.4, 200 mm glycine and 15% methanol) at 25 V for 1 h.
Rabbit antiserum produced against Clarkia BEBT protein was affinity purified against pure native AMAT protein and this was used as the primary antibody after a 500-fold dilution. Antibody-bound AMAT protein was then detected with a chemifluorescence assay using the ECL Plus Western blotting detection reagent (GE Healthcare) and the fluorescent signals were quantified using a FLA-3000 fluorescent image analyzer.
This research was funded in part by an NSERC Discovery grant to VDL. VDL holds a Canada Research Chair in Plant Biotechnology. We thank Dr Yoko Iijima for her help with GC-MS analysis, Dr David Gang for help on protein sequencing, Dr Till Beuerle for advice on the synthesis of anthraniloyl CoA, Dr Pichersky for his advice and generous gifts of BEBT antibody and BZL clone and Mr Sean Prager and Dr Jun Murata for advice on phylogenetic tree construction. We also thank G & H Wiley Ltd, St Catharines, Ontario and Dr Helen Fisher of the University of Guelph Grape Research Station, Vineland, Ontario for the supply of grape samples.
The gene sequence described in this report has been submitted to GenBank database and has been assigned the following accession number: AY705388.