Involvement of snapdragon benzaldehyde dehydrogenase in benzoic acid biosynthesis


*(fax +1 765 494 0391; e-mail


Benzoic acid (BA) is an important building block in a wide spectrum of compounds varying from primary metabolites to secondary products. Benzoic acid biosynthesis from l-phenylalanine requires shortening of the propyl side chain by two carbons, which can occur via a β-oxidative pathway or a non-β-oxidative pathway, with benzaldehyde as a key intermediate. The non-β-oxidative route requires benzaldehyde dehydrogenase (BALDH) to convert benzaldehyde to BA. Using a functional genomic approach, we identified an Antirrhinum majus (snapdragon) BALDH, which exhibits 40% identity to bacterial BALDH. Transcript profiling, biochemical characterization of the purified recombinant protein, molecular homology modeling, in vivo stable isotope labeling, and transient expression in petunia flowers reveal that BALDH is capable of oxidizing benzaldehyde to BA in vivo. GFP localization and immunogold labeling studies show that this biochemical step occurs in the mitochondria, raising a question about the role of subcellular compartmentalization in BA biosynthesis.


Benzoic acid (BA) is a weak acid with anti-microbial properties mediated in part by the inhibition of glycolysis in micro-organisms (Krebs et al., 1983). Benzoic acid is produced in many plants and animals, and is thus a natural constituent of numerous foods including milk products (Sieber et al., 1995). In plants, BA and its derivatives are important building blocks in a wide spectrum of compounds varying from primary metabolites [e.g. aromatic cytokinins (Werbrouck et al., 1996) and salicylic acid] to secondary products with pharmacological activities such as the anti-cancer agent taxol (Walker and Croteau, 2000) and the local anesthetic cocaine (Bjorklund and Leete, 1992). They also play critical roles in plant survival and reproductive success in natural ecosystems. Plant–insect and plant–plant interactions are often mediated by volatile benzenoids (Qualley and Dudareva, 2008), which are highly represented within volatile natural products (Knudsen and Gershenzon, 2006). Benzenoids emitted from many fruits contribute to the overall aroma and taste, which are valuable attributes for consumers and animal seed dispersers (Schwab et al., 2008; Negre-Zakharov et al., 2009). In addition, defensive non-volatile compounds, such as benzoyloxyglucosinolates in cruciferous plants, also contain the benzoyl moiety derived from BA (Graser et al., 2001; Kliebenstein et al., 2007).

Despite the simple structure of BA, its widespread distribution and its importance for the plant life cycle, the biochemical pathways leading to its formation remain largely unknown. Benzoic acid biosynthesis has attracted the attention of scientists for nearly half a century (Ibrahim and Towers, 1959), mainly due to its involvement in synthesis of the ubiquitous plant hormone salicylic acid (2-hydroxy BA). The limited understanding of these pathways is the result of the existence of multiple routes within a plant or between plant species, and an inability to track pathway intermediates. Recently, it has been shown that pathogen-infected Arabidopsis (Wildermuth et al., 2001) and Nicotiana benthamiana (Catinot et al., 2008) are able to synthesize salicylic acid predominantly from isochorismate without BA involvement. However, a soluble mono-oxygenase responsible for BA conversion to salicylic acid was partially purified from Tobacco mosaic virus-inoculated tobacco leaves (Leon et al., 1995), and a growing body of evidence has suggested that, under biotic (Ogawa et al., 2006) and abiotic stresses, salicylic acid is formed from BA via the phenylalanine (Phe) pathway in a variety of plants including pea (Pisum sativum), rice (Oryza sativa) and tobacco (Nicotiana tabacum) (Ogawa et al., 2005; Pan et al., 2006; Sawada et al., 2006).

Benzoic acid biosynthesis from l-Phe requires shortening of the propyl side chain by two carbons, which can occur via a β-oxidative pathway with formation of CoA–ester intermediates (Jarvis et al., 2000), a non-β-oxidative pathway with benzaldehyde as a key intermediate, or a combination of both routes (Abd El-Mawla and Beerhues, 2002; Boatright et al., 2004). These pathways share the first step, the conversion of l-Phe to trans-cinnamic acid, catalyzed by phenylalanine ammonia lyase (PAL), and diverge thereafter. The CoA-independent non-β-oxidative pathway requires the action of an aldehyde dehydrogenase (ALDH) to convert benzaldehyde to BA. ALDHs (aldehyde:NAD(P)+ oxidoreductases, EC are NAD(P)+-dependent enzymes that catalyze the irreversible oxidation of a broad spectrum of aliphatic and aromatic aldehydes to their corresponding carboxylic acids, and are involved in plant metabolism and detoxification of stress-generated aldehydes (Kirch et al., 2004). Although many plant ALDHs have been characterized, none has been shown to be involved in BA biosynthesis.

Here we have isolated and biochemically characterized benzaldehyde dehydrogenase (BALDH), the first plant enzyme involved in BA biosynthesis via the non-β-oxidative pathway. A cDNA encoding BALDH was obtained from snapdragon (Antirrhinum majus), whose flowers have been shown to emit high levels of methylbenzoate, the product of BA carboxyl methylation (Dudareva et al., 2000). In contrast to previously identified plant ALDHs, BALDH is exclusively expressed in scent-producing parts of the flower, and its transient expression in petunia flowers results in elevated levels of BA and its glucosides as well as increased emission of methylbenzoate and benzylbenzoate, which require BA and benzoyl CoA for their respective biosyntheses. The mitochondrial localization of the enzyme implies the involvement of various subcellular compartments in BA biosynthesis, an important point for understanding the various levels of regulation involved in this largely unknown pathway in plants.


Isolation and biochemical characterization of snapdragon BALDH

To determine whether biosynthesis of BA from l-Phe occurs via a non-β-oxidative pathway in snapdragon flowers, cut flowers (a total of three flowers per experiment, of 0.45 g each) were fed with 1.25, 2.5, 6.25, 12.5, 25 or 37.5 μm2H6-benzaldehyde for 4 h, with simultaneous collection of floral volatiles. Methylbenzoate detected in the volatile headspace was labeled 64.7% at the lowest amount fed, followed by 73 and 83% labeling at 2.5 and 6.25 μm of 2H6-benzaldehyde, respectively, and reached 94.7–99% saturation at higher concentrations of 2H6-benzaldehyde, confirming the conversion of exogenously supplied benzaldehyde to BA with subsequent methylation of the product to methylbenzoate. After scent collection, the 2H6-benzaldehyde solutions remaining after feeding flowers were analyzed by HPLC to determine whether the supplied benzaldehyde had been oxidized to BA non-enzymatically during these experiments. No BA was detected. Consistent with the labeling data, BALDH activity (10.52 ± 1.2 pkat mg−1 protein) was detected in crude protein extracts prepared from upper and lower petal lobes, the parts of the flower that have been shown to be primarily responsible for snapdragon scent production and emission (Dudareva et al., 2000, 2003). Thus, we searched a snapdragon EST database comprising >25 000 random clones representing 11 615 unigenes from various vegetative and floral organs (Bey et al., 2004) for a potential BALDH cDNA clone encoding a protein capable of catalyzing the formation of BA from benzaldehyde. This search revealed a clone, ama0m19 (designated here as BALDH, see below), with 40% identity to a BALDH from the bacterium Novosphingobium aromaticivorans (Peng et al., 2005). RNA blot analysis of this clone demonstrated that it is expressed predominantly in the upper and lower petal lobes (Figure 1a), the scent-producing parts of snapdragon flowers (Dudareva et al., 2000, 2003). Its expression is developmentally regulated (Figure 1b), and coincides with the developmental emission of methylbenzoate, for which BA is a direct precursor (Dudareva et al., 2000). Moreover, BALDH mRNA levels displayed rhythmic oscillations during the daily light/dark cycle (Figure 1c), which correlated with the level of BA in snapdragon flowers (Kolosova et al., 2001a), suggesting that this gene might be involved in BA biosynthesis.

Figure 1.

 Snapdragon BALDH gene expression.
Tissue-specific (a), developmental (b) and rhythmic (c) expression of BALDH mRNA. Total RNA (5 μg/lane) from floral tissues of 4-day-old snapdragon flowers and leaves (a), from flower petals of mature buds to day 12 post-anthesis (b), and from petals of 4-day-old snapdragon flowers during a 27 h day/night cycle (c) were hybridized with BALDH and re-hybridized with 18S rDNA to standardize samples (upper and lower panels, respectively). In (b), RNA gel blots were scanned using a Storm 860 phosphorimager (Molecular Dynamics,, and values were standardized for equal loading using 18S rRNA. Each point is the mean of four experiments. SE values are indicated by error bars.

A full-length BALDH cDNA was obtained by screening a snapdragon petal-specific cDNA library. It contained 2035 nucleotides with an open reading frame of 1602 nucleotides that encoded a protein of 534 amino acids with a predicted molecular mass of 58 081 kDa. The deduced amino acid sequence shows high sequence similarity (79–82%) to ALDHs from tobacco (op den Camp and Kuhlemeier, 1997), Arabidopsis (Skibbe et al., 2002), maize (Cui et al., 1996) and rice (Tsuji et al., 2003) (Figures 2 and S1), which all belong to the ALDH2 family (Kirch et al., 2004), and 74% similarity to human ALDH2 (Ni et al., 1999). BALDH protein appears to contain an N-terminal mitochondrial targeting sequence (amino acids 1–34), indicating that it represents a mitochondrial member of this family (Kirch et al., 2004).

Figure 2.

 Phylogenetic tree illustrating the relatedness of snapdragon BALDH to other plant ALDHs.
An unrooted neighbor-joining tree was created using clustal x (Conway Institute, and treeview (Page, 1996) for visualization. The distance bar is shown under the tree, and bootstrap values are given for the nodes. Only a subset of known ALDHs with the highest amino acid homology to BALDH is shown: Nicotiana tabacum (NtALDH, CAA71003), Arabidopsis thaliana (At3g48000, NP_190383; At1g23800, NP_564204), Lotus corniculatus (LcALDH, AAO72532), Vitis pseudoreticulata (VpALDH, AAZ79355), Picea sitchensis (PsALDH, ABK24267), Oryza sativa (OsALDH2a, NP_001048010; OsALDH2b, AAF73828), Zea mays (Zm ALDH, NP_001105891), Sorghum bicolor (SbALDH, BAB92019), Hordeum vulgare subsp. vulgare (HvALDH, BAB62757) and Secale cereale (ScALDH, BAB92017).

To determine the enzymatic activity of the putative BALDH, the coding region of the gene was sub-cloned into the expression vector pET-28a and expressed in Escherichia coli. The affinity-purified recombinant protein was assayed with a variety of potential substrates, including saturated aliphatic aldehydes (i.e. formaldehyde, acetaldehyde, propionaldehyde and hexanal) and aromatic aldehydes such as benzaldehyde, together with structurally related molecules that have various substitutions on the aromatic ring and/or various side-chain lengths and configurations (Figure 3). Of 18 tested aromatic aldehydes, only seven served as substrates, with the highest enzyme activity being found for benzaldehyde (0.94 ± 0.09 nkat mg−1 purified protein), followed by phenylacetaldehyde, trans-cinnamaldehyde, p-anisaldehyde and m-anisaldehyde (Figure 3). HPLC analysis confirmed that the product formed from benzaldehyde was BA (Figure S2). Similar to many other ALDHs, BALDH was capable of oxidizing aliphatic aldehydes, and increased carbon chain length resulted in a decrease in its efficiency. BALDH displayed high activities with acetaldehyde and propionaldehyde, which exceeded that of benzaldehyde by three- and 1.7-fold, respectively (Figure 3).

Figure 3.

 Relative conversion of selective aldehydes to their respective acids, catalyzed by snapdragon BALDH.
Substrate specificity was determined under benzaldehyde Vmax conditions. The activity with benzaldehyde was set to 100% and was 0.94 ± 0.09nkat mg−1 protein. Aldehydes were used at 20 μm. The values are the means of three independent experiments ±SE. Activity was undetectable with the following substrates: 4-hydroxybenzaldehyde, 2,3-dihydroxybenzaldehyde, 3,4-dihydroxybenzaldehyde, 2,4-dihydroxybenzaldehyde, 3,5-dihydroxybenzaldehyde, 2,5-dihydroxybenzaldehyde, 2,3,4-trihydroxybenzaldehyde, 2,4,6-trihydroxybenzaldehyde, 2-hydroxy-3-methoxybenzaldehyde, 4-hydroxy-3-methoxybenzaldehyde and 4-hydroxy-3-methoxycinnamaldehyde.

The molecular mass of active BALDH protein determined by gel-filtration chromatography was approximately 232.4 kDa, suggesting that the native protein exists as a tetramer containing mature subunits with a molecular mass of 54.5 kDa, as predicted by the mitoprot ii program for the processed protein (Andreoli et al., 2004). Although a tetrameric structure is common for members of the ALDH2 family in mammals and yeast (Yoshida et al., 1998), it has only been reported for two maize mitochondrial enzymes (i.e. RF2A and RF2B) out of the limited number of characterized plant ALDHs (Liu and Schnable, 2002). BALDH uses only NAD+ as a coenzyme, and is active in a broad range of pH varying from 6.0 to 9.0, with maximum activity for benzaldehyde oxidation at pH 8.0. Similar to other ALDHs, BALDH exhibits substrate inhibition at benzaldehyde concentrations ≥100 μm, and was completely inhibited by 30 μm disulfiram, an ALDH inhibitor (Kitson, 1975).

Kinetic characterization of the purified recombinant BALDH protein revealed that it has high affinity towards benzaldehyde, with an apparent Km value of 1.37 ± 0.04 μm. The apparent BALDH Km value for phenylacetaldehyde, the second most effective aromatic substrate (Figure 3), was 5.35 ± 1.24 μm, nearly four times higher than the Km value for benzaldehyde (Table 1). The apparent catalytic efficiency (kcat/Km ratio) of BALDH with benzaldehyde is almost five times higher than that with phenylacetaldehyde, suggesting that, of the aromatic aldehydes, benzaldehyde is the preferred substrate. Indeed, molecular homology modeling of the active site of BALDH indicates that benzaldehyde binding provides more favorable conditions for the nucleophilic attack and hydride transfer required for carboxylic acid formation than phenylacetaldehyde binding does (Figure S3). The apparent Km value for BALDH with acetaldehyde of 2.01 ± 0.32 μm is very similar to that with benzaldehyde; however, its catalytic efficiency is 4.8-fold greater with this aliphatic aldehyde than with benzaldehyde.

Table 1.   Kinetic parameters of BALDH
SubstrateKmm)Vmax (nkat mg−1)kcat (sec−1)kcat/Km (mm−1 sec−1)
  1. All values are means ± SE (n = 3I. Km, Michaelis constant; Vmax, maximal velocity; kcat, turnover number.

Benzaldehyde1.37 ± 0.041.32 ± 0.08 0.31 ± 0.02224 ± 10
Acetaldehyde2.01 ± 0.328.93 ± 0.472.08 ± 0.111081 ± 160
Phenylacetaldehyde5.35 ± 1.24 1.00 ± 0.11 0.23 ± 0.0347 ± 9

Validation of BALDH function in vivo

Transformation of snapdragon is cultivar-specific, and is currently not possible in Maryland True Pink, the variety used in this study. Thus, to assess the biological function of BALDH in vivo, the gene was transiently expressed in petunia flowers, which also emit methylbenzoate (Kolosova et al., 2001a; Verdonk et al., 2003). One-day-old petunia flowers were infiltrated with Agrobacterium tumefaciens carrying the snapdragon BALDH gene under the control of the petal-specific Clarkia breweri Lis promoter (Orlova et al., 2006), or with agrobacteria alone, agrobacteria carrying a vector with only the Lis promoter, and agrobacteria carrying snapdragon BALDH but lacking the Ti plasmid and unable to transfer T-DNA to plants (controls). Two days after inoculation, high levels of BALDH transcripts were detected in petunia petals relative to control flowers (Figure 4a). As all controls gave similar results across experiments, data for flowers infiltrated with agrobacteria carrying a vector with only the Lis promoter are presented. BALDH expression in petunia petals resulted in accumulation of a corresponding protein detected by Western blot analysis using antibodies generated against recombinant BALDH. Although a low level of cross-reacting protein was detected in petunia petals, BALDH protein levels (Figure 4a) and activity (P < 0.02) were significantly elevated in petals expressing snapdragon BALDH (6.2 ± 0.55 pkat mg−1 protein) relative to controls (4.3 ± 0.40 pkat mg−1 protein). Benzaldehyde was supplied to BALDH-expressing and control flowers, and floral volatiles were collected over a 12 h period during the day. BALDH expression led to a mean increase of 57% in methylbenzoate emission relative to controls (Figure 4b), as well as increases in endogenous pools of free BA and BA glucosides of 100 and 70%, respectively (Figure 4d,e), indicating that BALDH is involved in BA biosynthesisin vivo. BALDH-expressing flowers also emitted 67% more benzylbenzoate (Figure 4c) and 92% more 2-phenylethanol (data not shown), when compared with controls. Increased benzylbenzoate emission is probably the result of expanded pools of benzoyl CoA produced from BA. In petunia, benzoyl CoA:benzyl alcohol/phenylethanol benzoyltransferase (BPBT) is responsible for phenylethylbenzoate and benzylbenzoate formation, slightly favoring the latter as a product (Boatright et al., 2004; Orlova et al., 2006). In the presence of high levels of benzylalcohol formed from supplied benzaldehyde, BPBT uses benzoyl CoA predominantly for benzylbenzoate formation, thus leading to increased 2-phenylethanol emission.

Figure 4.

 Transient expression of snapdragon BALDH in petunia flowers.
The results shown are from three out of five independent experiments (Exp 1, Exp 2 and Exp 3) with at least 18 flowers per experiment.
(a) BALDH mRNA expression (upper) and protein levels (lower) in petunia flowers infiltrated with Agrobacterium carrying a vector with only the Lis promoter (1, control) or the BALDH gene under the control of the Lis promoter (2). BALDH mRNA was analyzed by RT-PCR and actin was used as an endogenous control; BALDH protein was analyzed by Western blot with purified anti-BALDH antibodies.
(b–e) Levels of emitted methylbenzoate (b) and benzylbenzoate (c), as well as internal pools of free BA (d) and BA glucosides (e), in petunia flowers infiltrated with Agrobacterium carrying the BALDH gene under the control of the Lis promoter (gray bars) relative to control flowers infiltrated with Agrobacterium carrying a vector with only the Lis promoter (white bars). In control flowers, the levels of methylbenzoate, benzylbenzoate, BA and BA glucosides were 8.4 ± 1.2 nm g FW−1 h−1, 0.34 ± 0.04 nm g FW−1 h−1, 37.1 ± 9.4 nm g FW−1 and 161.6 ± 10.3 nm g FW−1, respectively (means ± SE). Error bars indicate SE.

As kinetic analysis of the recombinant protein showed that BALDH can efficiently use acetaldehyde, we examined its possible involvement in acetaldehyde oxidation to acetate. Acetaldehyde is one of the intermediate products of ethanolic fermentation, which is highly active in germinating pollen (Mellema et al., 2002). It was previously shown that ALDH activity is necessary for pollen growth (op den Camp and Kuhlemeier, 1997), and the pollen of many plant species including Nicotiana tabacum, Nicotiana plumbaginifolia and Zea mays contains substantial amounts of acetaldehyde and ethanol (Tadege and Kuhlemeier, 1997). Addition of 30 μm disulfiram, an inhibitor of ALDHs (Kitson, 1975), prevented germination of snapdragon pollen, suggesting that, as with tobacco, pollen germination in snapdragon requires the function of an ALDH (Figure S4a–c). If BALDH is involved in acetaldehyde oxidation, it should be highly expressed during pollen germination. Thus, we used RT-PCR to detect BALDH mRNA expression in snapdragon pollen before and 24 h after germination (Figure S4a,b,d). In contrast to tobacco ALDH (op den Camp and Kuhlemeier, 1997), no BALDH transcripts were found in either case, suggesting that BALDH is not involved in detoxification of acetaldehyde in pollen. However, we currently cannot exclude the possibility that BALDH is involved in the oxidation of acetaldehyde and other aldehyde substrates that may be present in the mitochondria of petal tissue or synthesized in response to environmental stresses (Kirch et al., 2004).

BALDH subcellular localization

To determine experimentally the subcellular compartment in which benzaldehyde oxidation and BA formation take place, we investigated the intracellular localization of BALDH. The BALDH coding region was fused to a GFP reporter gene and transferred to Arabidopsis protoplasts, in which the corresponding transient GFP expression was analyzed by confocal laser scanning microscopy (Figure 5). In protoplasts expressing BALDH–GFP, GFP fluorescence was localized to small, punctuate structures (Figure 5a), which were also stained by MitoTracker Red used for visualization of mitochondria (Figure 5b). Co-localization of both GFP and MitoTracker Red (Figure 5c) confirmed BALDH mitochondrial localization. Elimination of the first 34 amino acids, which are predicted to be a mitochondrial targeting sequence, changed BALDH subcellular localization from the mitochondria to the cytosol (Figure 5e–h). The mitochondrial and cytosolic localizations of BALDH and its truncated version were identical to the positive mitochondrial (Figure 5i–l) and cytosolic (Figure 5m–p) GFP controls, respectively.

Figure 5.

 Subcellular localization of BALDH–GFP fusions in Arabidopsis leaf protoplasts.
A schematic diagram of the GFP fusion constructs is shown on the left, and the corresponding transient expression in protoplasts is shown on the right. GFP fluorescence is shown in green (a,e,i,m), chloroplast autofluorescence is shown in red (f,g,j,k,n,o) and blue (c), and MitoTracker Red-stained mitochondria are shown in red (b). Co-localization of GFP and MitoTracker Red is shown in (c). Transmission images of intact protoplasts are shown on the far right (d,h,l,p). Constructs containing FγATPase–GFP and GFP were used as mitochondrial and cytosolic markers, respectively. Scale bars = 50 μm.

Mitochondrial localization of BALDH was also confirmed by immunogold localization studies using transverse sections of lower petal lobes of 7-day-old snapdragon flowers. Substantial immunogold labeling with anti-BALDH antibodies was found within the mitochondria of the conical epidermal cells (Figure 6a,b), which are extensively involved in scent production (Kolosova et al., 2001b). No significant and specific immunolabeling was observed in sections treated with a control lacking active antibodies (Figure 6c). The control was obtained by incubating antibodies with the recombinant BALDH protein.

Figure 6.

 Immunogold localization of BALDH in snapdragon petals.
Transmission electron microscopy of conical cells of lower petal lobes of 7-day-old flowers after incubation with purified anti-BALDH antibodies (a,b) or an absorption control (c), followed by incubation with gold-conjugated goat anti-rabbit antibodies. An absorption negative control was obtained by incubation of purified antibodies with the BALDH protein followed by centrifugation at 3000 × g for 10 min at 4°C. Arrows indicate mitochondria. CW, cell wall; L, leucoplast; V, vacuole. Scale bars = 0.5 μm (a–c) and 0.1 μm [inserts in (a) and (b)].


Benzoic acid biosynthesis in plants remains an important unresolved question despite numerous attempts at its elucidation (Wildermuth, 2006). Several side chain-shortening routes have been proposed for BA formation from Phe; however, no genes or enzymes responsible for these biochemical steps have been discovered (Wildermuth, 2006). BALDH activity has been demonstrated in cell cultures of Hypericum androsaemum L., which produce a high level of xanthones for which BA is a precursor, but a specific protein responsible for this activity has not been purified and characterized (Abd El-Mawla and Beerhues, 2002). Using a functional genomics approach, we have isolated and biochemically characterized a snapdragon BALDH that can efficiently oxidize benzaldehyde to BA (Figure 3 and Figure S2). BALDH is a mitochondrial homo-tetrameric enzyme (Figures 5 and 6) that belongs to the plant ALDH2 family. Similar to other members of this family, BALDH has broad substrate specificity. In addition to benzaldehyde, it can oxidize other aromatic aldehydes (with lower efficiency) as well as aliphatic aldehydes (Figure 3). The broad substrate specificity of ALDHs makes it difficult to determine their precise physiological role in planta, as the same enzyme could potentially be involved in multiple pathways and affect many cellular functions, given that a wide range of aldehydes are produced within the cell as intermediates in the metabolism of carbohydrates, vitamins, steroids, amino acids, hormones and lipids (Yoshida et al., 1998; Vasiliou et al., 2000) and synthesized in response to environmental stresses (Kirch et al., 2004). As a result, the exact function of many plant ALDH2s remains elusive. Past predictions about their in vivo involvement in some biological processes were largely based on kinetic analysis of enzymes and expression studies (Liu and Schnable, 2002; Kirch et al., 2005). Currently, direct genetic evidence has only been provided for Arabidopsis ALDH2C4, which catalyzes the oxidation of sinapaldehyde and coniferaldehyde, thus playing a major role in the formation of both soluble and cell wall-linked ferulate esters (Nair et al., 2004).

In contrast to known plant ALDHs, which are expressed constitutively (Tsuji et al., 2003) or in a wide variety of organs (Liu and Schnable, 2002), BALDH mRNA transcripts were found exclusively in the scent-producing parts of snapdragon flowers (Figure 1). Their accumulation was positively correlated with BA levels and methylbenzoate emission over a daily light/dark cycle and flower development (Dudareva et al., 2000; Kolosova et al., 2001a), and was not activated during pollen germination (Figure S4). Transient expression of the snapdragon BALDH gene in petunia flowers resulted in elevated levels of BA and methylbenzoate emission (Figure 4a,b,d), providing direct evidence that BALDH is involved in BA biosynthesis from benzaldehyde in vivo. BALDH over-expression also led to an increase in benzylbenzoate emission (Figure 4c), supporting our previous conclusion that there is a flux from BA to benzoyl CoA in petunia flowers (Boatright et al., 2004). The increase in the endogenous pool of BA resulted in elevated levels of BA glucosides (Figure 4e). Interestingly, elevated levels of BA, BA glucosides, methylbenzoate and benzylbenzoate were observed only when petunia flowers were supplied with benzaldehyde, indicating that the endogenous pool of benzaldehyde available for BA biosynthesis is a limiting factor. Recently, using computer-assisted metabolic flux analysis combined with in vivo stable isotope labeling, we have shown that both the β-oxidative and non-β-oxidative pathways contribute to benzenoid compound formation in petunia flowers (Boatright et al., 2004; Orlova et al., 2006). Our identification and characterization of BALDH as involved in the oxidation of benzaldehyde to BA revealed that the non-β-oxidative pathway of BA biosynthesis is active in snapdragon flowers as well. To date, there is still no information about the presence and activity of the β-oxidative pathway in snapdragon. However, comparison of methylbenzoate labeling (99%) from supplied 2H6-benzaldehyde (18.5 μm per gram fresh weight) over a 4 h period in snapdragon flowers (this study) with that in petunia petals obtained previously under similar conditions (36.3%, Boatright et al., 2004) suggests that snapdragon petals have a small endogenous pool of benzaldehyde with a rapid turnover to BA via the non-β-oxidative pathway.

Phenylpropanoid metabolism has long been thought to occur in the cytoplasm. However, it was recently shown that cinnamic acid is produced in two subcellular compartments, the endoplasmic reticulum and cytosol, resulting from differentially localized PAL isoforms (Achnine et al., 2004). We previously demonstrated that the methylation of BA to methylbenzoate occurs in the cytoplasm of the epidermal cells of snapdragon flower petals (Kolosova et al., 2001b). The results presented here reveal that BA formation from benzaldehyde takes place in the mitochondria (Figures 5 and 6), suggesting that the biosynthesis and emission of methylbenzoate are coupled with BA transport out of the mitochondria to the cytosol for methylation and emission. Still in question is the subcellular compartmentalization of benzaldehyde formation. In petunia, benzaldehyde is not only a BA precursor, but, in contrast to snapdragon, is also a scent compound emitted by flowers (Boatright et al., 2004). Ectopic over-expression of the BALDH gene with and without its signal peptide will provide further information as to whether separate benzaldehyde pools exist for the formation and emission of BA in petunia flowers. Moreover, the discovery of genes and enzymes responsible for other biochemical steps in BA biosynthesis will provide important new insights about the role of subcellular compartmentalization in regulation of the benzenoid network.

Experimental procedures

Plant material

Snapdragon cv. Maryland True Pink (Ball Seed, and Petunia hybrida cv. Mitchell were grown as described previously (Dudareva et al., 2000). Labeling experiments were performed on cut snapdragon flowers, and emitted volatiles were collected via a closed-loop stripping method for 4 h during the day (Boatright et al., 2004).


All reagents were purchased from Sigma-Aldrich ( or Fluka ( unless noted otherwise.

RNA isolation and RNA gel blot analysis

BALDH mRNA levels were analyzed as described previously (Dudareva et al., 2000; Kolosova et al., 2001a). The coding region of the BALDH gene was used as a probe.

BALDH expression in Escherichia coli and purification of recombinant protein

The coding region of snapdragon BALDH was PCR-amplified using forward primer 5′- GCTAGCATGGCGGCTCACCGATTTTCATC-3′, which introduced an NheI site at the initiating ATG codon, in combination with reverse primer 5′-TCACAACCAAGCAGGATTTTTCAGC-3′, and subcloned into expression vector pET-28a(+) (Novagen, Sequencing revealed that no errors were introduced during PCR.

For functional expression, E. coli BL21 Rosetta competent cells were transformed with the resulting recombinant plasmid or with pET-28a(+) vector lacking an insert (control). Induction, harvesting and protein purification by affinity chromatography on Ni-NTA (nickel-nitrilotriacetic acid) agarose (Qiagen Inc., were performed as described previously (Kaminaga et al., 2006). Protein purity was determined to be 66% by densitometry of SDS–PAGE gels after Coomassie brilliant blue staining, and this result was used for kcat calculations. Protein concentration was determined by the Bradford method (Bradford, 1976). Native molecular mass determination for BALDH was performed as described previously (Kaminaga et al., 2006).

Enzyme assays and product identification

Dehydrogenase activity was determined by measuring the rate of increase in the fluorescence of NADH formation in a reaction mixture (1 ml) containing 100 mm sodium phosphate buffer (pH 8.0), purified BALDH protein (8 μg), 1 mm NAD+, 5 mm MgCl2 and 20 μm aldehyde substrate dissolved in 0.5% DMSO or ethanol (phenylacetaldehyde). The reaction was initiated by addition of aldehyde, and performed at 29°C, with recording of NADH formation for up to 3 min on an Aminco filter fluorometer (American Instrument Co.,

For product verification, the enzymatic reaction was performed for 10 min in a reaction mixture (0.5 ml) containing 40 μg of purified BALDH protein and 50 μm benzaldehyde. After reaction termination with 25 μl 50% trichloroacetic acid, 50 μl of the sample was analyzed by HPLC (Agilent, Separation was achieved using an XDB-C18 column (4.6 mm internal diameter × 50 mm length, pore size 1.8 μm; Agilent) maintained at 40°C, using 36% methanol and 0.1% formic acid. Benzaldehyde and BA were quantified by UV absorption at 254 and 233 nm, respectively. For derivatization, BA peaks from the reaction mixture and an authentic standard were collected and converted to methylbenzoate using acetyl chloride:methanol (2:5) at 70°C for 2 h. The derivatized samples were analyzed by GC/MS (Agilent).

Enzyme activity in snapdragon and infiltrated petunia petals was determined by measuring BA formation by HPLC. Crude protein extracts were prepared from petal tissues as described previously (Dudareva et al., 2000). After desalting on an NAP-5 column (GE Healthcare,, 43 μg of protein were used in a 200 μl reaction with 50 μm benzaldehyde for 3 h in a standard reaction mixture. After reaction termination, a 10 μl aliquot was analyzed by HPLC.

Transient BALDH expression in petunia flowers

The BALDH coding region was PCR-amplified using forward primer 5′-GCCATGGATGGCGGCTCACCGATTTTC-3′ and reverse primer 5′-GCTGCAGTCACAACCAAGCAGGATTTTTC-3′, and subcloned into the NcoI/PstI site of the pEF1-LIS binary vector (Orlova et al., 2006). Transformation was performed with Agrobacterium tumefaciens strain GV3101containing pEF1-LIS-BALDH and pEF1-LIS vector (control). As an additional control, transformation was carried out using pEF1-LIS-BALDH in an Agrobacterium strain (GV3101 A136) that lacks the Ti plasmid and is unable to transfer T-DNA to plants. Overnight Agrobacterium cultures were spun down at 3000 × g for 10 min at 4°C, and resuspended in 50 mm MES pH 5.7, 0.5% glucose, 2 mm Na3PO4 and 100 μm acetosyringone until an OD600 of 0.3 was reached. One-day-old petunia flowers were vacuum-infiltrated for 15 min and kept for 36 h in the dark in 5% sucrose. Scent collections were performed from 9 am to 9 pm from flowers fed with 5 mm benzaldehyde in 5% sucrose solution. BALDH mRNA levels were analyzed by RT-PCR (Dudareva et al., 2003) using gene-specific primers 5′-CGTTAAGAGCTGGGACAGTG-3′ (forward) and 5′-TCACAACCAAGCAGGATTTTTCAGC-3′ (reverse) for BALDH, and 5′-TGCTCCACCAGAGAGGAAATATAGTGTT-3′ (forward) and 5′-TCGTATTCTCCCTTTGAAATCCACATCT-3′ (reverse) for actin. BALDH protein levels were analyzed by Western blot (Dudareva et al., 2000). Transformation of petunia flowers with Agrobacterium strain GV3101 alone, GV3101 containing pEF1-LIS, and GV3101 A136 containing pEF1-LIS-BALDH gave similar results.

Analysis of BA internal pools

Quantification of free and glucose-bound BA from infiltrated flowers was performed as described previously (Pasqualini et al., 2003), with the exception that, after chloroform extraction, BA-containing fractions were extracted three times with 500 μl ethyl acetate. The organic layer was dried, resuspended in 50 μl MeOH per 100 mg of sample, and BA was analyzed on HPLC. To release free BA from BA glucosides, samples were treated with 0.2 units β-glucosidase (Calzyme, in 50 mm sodium acetate buffer pH 5.0 for 1 h at 37°C prior to organic solvent extraction.

Subcellular localization of BALDH

Open reading frames of BALDH and Δ1–34-BALDH were fused upstream of and in-frame with GFP in XbaI/BamHI cloning sites of the p326-SGFP vector containing the CaMV 35S promoter. For both constructs, the coding regions were PCR-amplified using forward primers 5′-TCTAGAATGGCGGCTCACCGATTTTC-3′ (for BALDH–GFP) and 5′-TCTAGAATGGGCACAGCAGCAGCAGCAG-3′ (for Δ1–34-BALDH–GFP, introduced ATG underlined), and reverse primer 5′-GGATCCCAACCAAGCAGGATTTTTCAGC-3′. Sequencing confirmed the accuracy of the fusions. Arabidopsis protoplasts were prepared and transformed as described previously (Sheen, 2002; Nagegowda et al., 2008). p326-SGFP and p326-FγATPase-SGFP plasmids were used as markers for cytosolic and mitochondrial localization, respectively. Transient expression of GFP fusion proteins was observed 16–20 h after transformation. Protoplasts transformed with BALDH–GFP were stained using 500 nm MitoTracker Red (Molecular Probes Inc., for 45 min before confocal microscopy. Excess stain was eliminated by centrifugation at 100 × g for 10 min at room temperature, and protoplasts were resuspended in 154 mm NaCl, 125 mm CaCl2, 5 mm KCl, 2 mm MES pH 7.5, and 5 mm glucose. Images were acquired using a Radiance 2100 MP Rainbow confocal/multiphoton system (Bio-Rad, on a TE2000 inverted microscope (Nikon, using a 60× oil 1.4 numerical aperture lens. MitoTracker Red, GFP and chlorophyll fluorescence images were obtained sequentially to avoid possible bleed-through. The MitoTracker Red was excited at 543 nm using a green HeNe laser, and the fluorescence between 560 and 600 nm was collected. Then GFP was excited using the 488 nm line of the four-line argon laser, and the emission was collected using a 500LP/540SP filter combination. Chlorophyll was excited using a 637 nm red diode laser, and the emission greater than 660 nm was collected.

Immunogold labeling and antibody preparation

Sections from lower petal lobes of 7-day-old snapdragon flowers were prepared for immunogold labeling using high-pressure freezing in a Wohlwend Compact 01 unit (TechnoTrade International, Sample discs were immersed in hexadecane and subjected to a light vacuum to remove excess air. Each petal disc was placed in the 0.2 mm well of an ‘A’ planchette (0.1/0.2 mm), a droplet of 0.15 m sucrose was added, and the planchette was covered with the flat side of a ‘B’ planchette (flat/0.3 mm). The planchette ‘sandwich’ was frozen and split, and the sample was transferred to a cryovial containing the substitution fluid (acetone, 8% dimethylpropanol, 0.25% anhydrous glutaraldehyde and 0.2% uranyl acetate at the temperature of liquid nitrogen). Substitution was allowed to proceed for 4 days at −80°C, before a gradual warm-up to −20°C. The samples were infiltrated with Lowicryl HM-20 resin (Polysciences, Inc.,, and the resulting blocks were polymerized under UV light at −20°C.

The purified recombinant BALDH protein was used to generate polyclonal antibodies in rabbits (Cocalico Biologicals, The antiserum with the highest antibody titer was used for purification of antibodies (Kolosova et al., 2001b). Immunogold labeling was performed as described previously (Kolosova et al., 2001b) using purified antibodies diluted 1:100. Selected electron micrographs were scanned into an Apple Macintosh G4 computer ( at a resolution of 1200 dots per inch using an Epson Expression 1600 scanner (Epson, Contrast and brightness in the resulting images were adjusted using adobe photoshop 7.0 (

Modeling and substrate docking for BALDH

The 3D BALDH model and substrate-docking modeling of the BALDH active site with benzaldehyde and phenylacetaldehyde were obtained based on the crystal structure of the human mitochondrial ALDH2 (Protein Data Bank code 1CW3B) as described previously (Brichac et al., 2007).

Pollen germination

Pollen (100 mg) from 5-day-old snapdragon flowers was germinated for 24 h at room temperature in solution containing 10% sucrose, 100 mg l−1 boric acid, 100 mg l −1potassium nitrate, 200 mg l−1 magnesium sulfate and 200 mg l−1 calcium chloride (Yang, 1986) with gentle shaking on a 55D shaker (Reliable Scientific, After germination, pollen was pelleted by gravity for 30 min, and RNA was obtained together with RNA isolated from fresh pollen and from snapdragon petals. RNA samples were treated with DNase I to eliminate contaminating DNA using a TURBO DNA-free kit (Ambion,, and used for RT-PCR. Pollen was visualized by light microscopy (Optiphot; Nikon).


We thank Dr Inhwan Hwang (Center for Plant Intracellular Trafficking, Pohang University of Science and Technology, South Korea) for GFP vectors, Dr Wilfried Schwab (Biomedical Food Technology, Technical University Munich, Germany) for the infiltration protocol, Dr Stanton Gelvin for Agrobacterium strain GV3101 A136, and Dr Karl Wood and Connie Bonham for help with acetaldehyde detection, as well as Jennifer Sturgis and Chia-Ping Huang for assistance with confocal and electron microscopy, respectively. Confocal microscopy data were acquired in the Purdue Cancer Center Analytical Cytometry Laboratories supported by the Cancer Center National Cancer Institute core grant number NIH NCI-2P30CA23168. This project was supported by the National Research Initiative of the United States Department of Agriculture Cooperative State Research, Education and Extension Service (grant number 2005-35318-16207) and the Fred Gloeckner Foundation Inc. This article is contribution number 2008-18394 from Purdue University Agricultural Experimental Station.

The GenBank accession number for the Antirrhinum majus BALDH sequence is FJ151199.