Analysis of the Enzymatic Formation of Citral in the Glands of Sweet Basil

Archives of Biochemistry and Biophysics 448 (2006) 141–149 www.elsevier.com/locate/yabbi

Analysis of the enzymatic formation of citral in the glands of sweet basil

Yoko Iijima, Guodong Wang, Eyal Fridman, Eran Pichersky ¤

Department of Molecular, Cellular and Developmental Biology, University of Michigan, Ann Arbor, MI 48109-1048, USA

Received 26 April 2005, and in revised form 21 July 2005 Available online 19 August 2005

Abstract

Basil glands of the Sweet Dani cultivar contain high levels of citral, a mixture of geranial and its cis-isomer neral, as well as low levels of geraniol and nerol. We have previously reported the identiWcation of a cDNA from Sweet Dani that encodes an enzyme responsible for the formation of geraniol from geranyl diphosphate in the glands, and that these glands cannot synthesize nerol directly from geranyl diphosphate. Here, we report the identiWcation of two basil cDNAs encoding NADP+-dependent dehydrogenases that can use geraniol as the substrate. One cDNA, designated CAD1, represents a gene whose expression is highly speciWc to gland cells of all three basil cultivars examined, regardless of their citral content, and encodes an enzyme with high sequence similarity to known cinnamyl alcohol dehydrogenases (CADs). The enzyme encoded by CAD1 reversibly oxidizes geraniol to produce geranial (which reversibly isomerizes to neral via keto–enol tautomerization) at half the eYciency compared with its activity with cinnamyl alcohol. CAD1 does not use nerol and neral as substrates. A second cDNA, designated GEDH1, encodes an enzyme with sequence similarity to CAD1 that is capable of reversibly oxidizing geraniol and nerol in equal eYciency, and prolonged incubation of geraniol with GEDH1 in vitro produces not only geranial and neral, but also nerol. GEDH1 is also active, although at a lower eYciency, with cinnamyl alcohol. However, GEDH1 is expressed at low levels in glands of all cultivars compared with its expression in leaves. These and additional data presented indicate that basil glands may contain additional dehydrogenases capable of oxidizing geraniol. © 2005 Elsevier Inc. All rights reserved.

Keywords: Citral; Geraniol; Nerol; Geraniol dehydrogenase; Cinnamyl alcohol dehydrogenase; Sweet basil; Secondary compounds; Plant biochemistry; EST databases; Evolution

Citral, the name given to the mixture of the monoter- Previously, we showed that sweet basil cultivar Sweet pene aldehydes geranial and neral, imparts a strong “lem- Dani is particularly rich in citral, which is localized in ony” scent and is known to be emitted or accumulated in the oil sac attached to the peltate glands on the surface such herbs as lemon grass [1], ginger [2], and some varie- of the leaves [11]. In glands of this variety, geranial and ties of sweet basil [3,4]. Citral is a valuable Xavor and scent neral are the main terpene constituents, while small reagent that is heavily used in the food and perfume amounts of geraniol and nerol are also observed. We industries [5,6]. Several previous investigations have have previously shown that Sweet Dani glands synthe- reported that citral is synthesized from geraniol or nerol size geraniol from geranyl diphosphate (GPP)1 in a by an alcohol dehydrogenase [7,8] or alcohol oxidase reaction catalyzed by geraniol synthase (GES). [9,10], but an enzyme capable of catalyzing such a reaction has not yet been puriWed and characterized. 1 Abbreviations used: GPP, geranyl diphosphate; GES, geraniol synthase; SD, Sweet Dani; GEDH, geraniol dehydrogenase; GC, gas chro- * Corresponding author. Fax: +1 734 647 0884. matography; EST, expressed sequence tag; CAD, cinnamyl alcohol E-mail address: [email protected] (E. Pichersky). dehydrogenase; SAD, sinapyl alcohol dehydrogenase.

Furthermore, these glands do not express a terpene syn- IPTG with 10 M ZnCl2 and grown at 18 °C for 18 h. thase that can convert GPP to nerol [11]. Cells were then harvested by centrifugation, resuspended Here, we report the isolation of two basil cDNAs in lysis buVer, and sonicated as previously described [12]. encoding dehydrogenases capable of oxidizing geraniol to geranial, which then undergoes tautomerization to PuriWcation of heterologously expressed GEDH1 neral, yielding the mixture of geranial and neral (i.e., citral). We functionally expressed these cDNAs in Esche- All puriWcation steps were carried out at 4 °C unless richia coli and biochemically characterized the enzymes stated otherwise. The supernatant of the bacterial lysate to determine their kinetic parameters and cofactor (20 mL) after induction of GEDH1 protein in E. coli was requirements. Moreover, we examined the expression loaded onto a DEAE–cellulose column (8 mL of DE53; level of the genes encoding these enzymes in Sweet Dani Whatman, Clifton, NJ) pre-equilibrated with a buVer glands and leaves as well as in glands and leaves of two containing 100 mM Tris–HCl, pH 7.5, 20% glycerol [v/v], other basil cultivars that do not produce citral. The anal- and 10 mM -mercaptoethanol. After the sample was ysis of the combined data indicates that while basil loaded, the column was washed with 30 mL of the pre- plants have at least one dehydrogenase that is highly equilibration buVer. GEDH1 protein did not bind to speciWc for geraniol, this enzyme is expressed at low lev- DE53 as most of the recombinant protein eluted in the els in the glands as compared to its expression in non- wash fractions. Next, 150 L of this enzyme solution was gland leaf tissue. On the other hand, a gland-speciWc loaded onto a size-exclusion column (10 £ 300 mm) dehydrogenase that has a higher speciWcity to cinnamyl packed with Superose 12 (Pharmacia Biotech), and the alcohol than to geraniol nonetheless appears to make an active fractions were isocratically eluted with 100 mM important contribution to the production of citral in the KCl in BuVer A at 0.5 mL/min. Fractions (0.5 mL each) glands. were collected and protein purity was estimated by SDS– PAGE, followed by Coomassie brilliant blue staining or silver staining. The fraction with the highest degree of Materials and methods purity was used for further characterization. Protein concentrations were measured by the Bradford method Plant material or by staining intensity on SDS–PAGE gel compared with bovine serum albumin concentration standards. Plants of the three basil cultivars Sweet Dani (SD), EMX, and SW were grown in the green house under PuriWcation of heterologously expressed CAD1 controlled illumination as described in previous reports [11,12]. His-tagged basil CAD1 cDNA was expressed in E. coli as described above for the GEDH1 cDNA and Glands’ crude enzyme extraction puriWed using Ni–NTA aYnity columns as previously described [13]. Glands were isolated following the procedures previously described by Gang et al. [16]. Crude enzyme prepa- Sequence analysis rations from these glands were prepared as previously described [11]. Alignment of multiple protein sequences was performed using the ClustalX program [14]. Sequence relat- Isolation of cDNAs and expression in E. coli edness by the neighbor-joining method was determined using the protocol included in the ClustalX package. The The construction of EST databases from the peltate phylogenic tree was drawn using the TREEVIEW pro- glands of cultivars SD, EMX, and SW was previously gram (http://taxonomy.zoology.gla.ac.uk/rod/treeview. reported [12]. BLAST searches revealed numerous ESTs html) [15]. with sequence similarity to alcohol dehydrogenases. Putative ADH cDNAs were assembled into contigs. Dehydrogenase enzyme assays Full-length cDNA from each contig (obtained by 5Ј RACE when necessary) was cloned into the pCRT7/CT- Oxidative dehydrogenase activity was assayed by TOPO TA vector (Invitrogen, Carlsbad, CA) and incubating 5 L of the enzyme sample in a Wnal volume expressed in the E. coli expression system. CAD1 cDNA of 100 L buVer containing 100 mM glycine–NaOH, pH was also ligated into the expression vector pET28-(a) to 9.5, 1 mM NADP+, and 1 mM substrate. Assays were produce a protein with an N-terminal His tag. These carried out at 25 °C for 15–30 min and time courses were plasmids were transformed into Codon Plus cells (Invit- measured by monitoring the formation of the generated rogen). E. coli cultures carrying alcohol dehydrogenase NADPH (or NADH) in a spectrophotometer at 340 nm. Y expression constructs were induced by adding 0.5 mM The molar extinction coe cient ( 340) used for NADPH Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149 143

3 ¡1 was 6.22 £ 10 M . DMSO was added instead of sub- A 300 strate to the control reaction. Substrate speciWcity was NADP+ NAD+ examined using identical conditions as just described 250 with each alcohol. 200 Product conWrmation 150 To chemically identify the products formed in the enzyme-catalyzed reactions, the same enzyme assay 100 described above was performed but with 10-fold higher 50

concentrations of enzyme. After 10, 30, and 60 min incu- GEDH activity (pkat / mg protein) bations at 25 °C, 200 L of pentane was added to the 0 tube, vortexed brieXy, and centrifuged to separate the Small Medium Large Glands Glands Glands phases. The pentane layer was directly placed into a GC leaves leaves leaves SD EMX SW vial for GC–MS analysis. A Shimadzu QP-5000 system (Shimadzu, Columbia, B geraniol geranial MD) equipped with a Shimadzu GC-17 gas chromato- neral nerol graph was used for GC–MS analysis of volatile compounds. Separations were performed on a EC-5 column (30 m £ 0.32 mm i.d. £ 1 m Wlm thickness, Alltech Asso- ciates, DeerWeld, IL). Samples (2 L) were injected by the Shimadzu AOC-17 Autoinjector. Injection and detector temperatures were set at 220 and 250 °C, respectively. 5.0 7.5 10.0 12.5 15.0 17.5 The initial temperature of the column was set to 55 °C (min) for 2 min and the temperature gradient was performed Fig. 1. Geraniol dehydrogenase (GEDH) activity in leaves and glands by increasing the temperature by 2 °C/min to a Wnal tem- of basil. (A) Comparison of NADP+-dependent and NAD+-dependent GEDH activities in the leaves of basil cultivar Sweet Dani (SD) at perature of 220 °C. Ultrapure helium was used as the V X ¡1 di erent stages of leaf development, and comparison of GEDH activi- carrier gas at a ow rate of 1.3 mL min . All other con- ties in glands of SD, EMX, and SW. (B) Products generated by crude ditions were the same as previously reported [11]. Com- protein extract of SD glands with geraniol as the substrate. (Top panel) pounds separated on the column were identiWed by Gas chromatogram of volatiles extracted from leaves of basil cultivar comparing their retention time and mass fragmentation Sweet Dani as a reference; (middle panel) GC separation of the prod- + patterns with those of authentic standards. ucts of the reaction containing geraniol and NADP ; (bottom panel) GC separation products of the reaction containing geraniol and NAD+. RNA blot analysis other cultivars which do not synthesize geraniol and cit- RNA blot analysis was performed as previously ral [12]. The NADP+-dependent GEDH-speciWc activity described [12]. in the peltate glands of cultivar Sweet Dani was 262 pkat/mg protein (Fig. 1A), which was three times higher than the highest speciWc activity of GEDH Results observed in whole leaves. Glands of the cultivar SW had similar NADP+-dependent GEDH-speciWc activity, NADP+-dependent geraniol dehydrogenase activity in while EMX glands contained 2.3-fold less GEDH-spe- leaves and glands of basil ciWc activity (Fig. 1A). NAD+-dependent GEDH-speciWc activities in the glands of all three cultivars were always Crude enzyme extracts were prepared from small lower than NADP+-dependent GEDH-speciWc activities (0.5–1.5 cm), medium (1.5–3 cm), and large (3–4 cm) (ranging from 16 to 26% of the activities with NADP+). leaves of the Sweet Dani cultivar, and their geraniol The products from the oxidation of geraniol by dehydrogenase (GEDH) activities were measured crude enzyme extract of Sweet Dani glands with (Fig. 1A). At every stage of leaf development, geraniol NADP+ or NAD+ cofactors added were analyzed by dehydrogenase (GEDH)-speciWc activity levels with gas chromatography (GC) (Fig. 1B). The reaction NADP+ were 4.3 times higher than with NAD+. GEDH products with NADP+ present in the reaction included activity levels were highest in young leaves and neral, geranial, as well as nerol (Fig. 1B, middle panel). decreased by 60% in older leaves. On the other hand, signiWcant amounts of the geraniol Since young leaves have more glands per unit area substrate remained when NAD+ was added as a cofac- [16], we examined the levels of GEDH activity in isolated tor implying the GEDH activity was speciWc for glands of SD cultivar as well as in the glands of two NADP+ (Fig. 1, bottom panel). 144 Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149

Isolation of cDNAs encoding dehydrogenases capable of (Fig. 1A), and also because the frequency of a given oxidizing geraniol EST in a database is not always reXective of the true levels of gene expression, all Wve types of genes were We have previously constructed expressed sequence considered as potentially encoding GEDH. Full-length tag (EST) databases from the peltate glands of basil cDNAs were obtained for all of the Wve contigs, the cultivars SD, EMX, and SW [12]. Exhaustive BLAST open reading frames were cloned into an E. coli expres- searches of these databases identiWed Wve types of sion vector and transformed to E. coli cells, and the cDNAs encoding proteins with sequence homology to resultant crude protein extracts tested for geraniol known alcohol dehydrogenases (Table 1). The abun- dehydrogenase activity. dance of these Wve types diVered among them and there Three of these putative ADHs (ADH-like1, ADH- was also a diVerence in abundance within each type in like2, and ADH-like3, Fig. 2) showed no activity with the EST databases of the three cultivars. However, geraniol. Only a few copies of cDNAs for ADH-like2 because GEDH activity was found in all three cultivars and ADH-like3 were present in the SD EST database, and none in the other two EST databases. For ADH- Table 1 like1, a larger number of cDNAs were found in the Frequency of Wve cDNAs encoding alcohol dehydrogenase homologs EMX and SD EST databases, and none in the SW in the EST databases of three basil cultivars (number of ESTs/1000) EST database. Because of the lack of activity of all SD SW EMX three of these ADH-like proteins when using geraniol CAD1 11.92.9as a substrate, these sequences were not investigated GEDH1 0 0 0.73 further. ADH-like1 8.2 0 13.3 Another cDNA encoded a protein with high sequence ADH-like2 10 0identity (>68%) to previously identiWed cinnamyl alco- ADH-like3 0.34 0 0 hol dehydrogenases (CADs) (Figs. 2 and 3), which are

10-Hydroxy GEDH (Catharanthus roseus) CAD1 (Ocimum basilicum) GEDH1 (Ocimum basilicum) CAD (Populus deltoides) CAD (Arabidopsis thaliana) 10-Hydroxy GEDH (Camptotheca acuminata) CAD (Fragaria x ananassa) CAD (Zea mays) SAD (Populus tremuloides)

CAD (Pinus taeda)

Allyl ADH (Nicotiana tabacum)

ADH3 (Oryza sativa) ADH3 (Zea mays) Pulegone reductase (Mentha x piperita) ADH3 (Pisum sativum)

ADH-like3 (Ocimum basilicum) ADH2 (Lycopersicon esculentum)

ADH1 (Petunia x hybrida)

ADH2 (Oryza sativa) ADH1 (Zea mays)

ADH-like1 (Ocimum basilicum) ADH-like2 (Ocimum basilicum)

(-)-isopiperitenone reductase (Mentha x piperita) 0.1

Fig. 2. Relatedness of basil CAD1 and GEDH1 to other plant alcohol dehydrogenases. All CAD sequences shown here, with the exception of the Fragaria CAD, have been experimentally demonstrated to possess CAD activity. The tree was constructed using the nearest neighbor-joining method. The GenBank accession numbers for the sequences analyzed in this Wgure are as follows: basil GEDH1 (AY879284), basil CAD1 (AY879285), basil ADH-like1 (AY879286), basil ADH-like2 (AY872287), and basil ADH-like3 (AY879288), Zea mays ADH1 (P00333), Petunia x hybrida ADH1 (P25141), Lycopersicon esculentum ADH2 (P28032), Oryza sativa ADH2 (AAF34412), Pisum sativum ADH3 (P80572), Z. mays ADH3 (P93629), O. sativa ADH3 (P93436), Arabidopsis thaliana CAD5 (AAP59435), Fragaria x ananassa CAD (AAK28509), Populus deltoides CAD (P31657), Z. mays CAD (O24562), Pinus taeda CAD (P41637), C. roseus 10-hydroxyGEDH (AAQ55962), Camptotheca acuminate 10-hydroxy- GEDH (AAQ20892), Nicotiana tabacum allyl ADH (BBA89423), Mentha x piperita pulegone reductase (AAQ75423), Mentha x piperita (¡)-isopiperitenone reductase (AAK58693), and P. tremuloides SAD (AAK58693). Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149 145

Fig. 3. Comparison of the amino acid sequences of basil CAD1 (ObaCAD1) and GEDH1 (ObaGEDH1) with the three most similar protein sequences to GEDH1 registered in GenBank, a protein from Fragaria x ananassa annotated as cinnamyl alcohol dehydrogenase (FanCAD), sinapyl alcohol dehydrogenase from P. tremuloides (PtrSAD), and a protein from C. roseus annotated as 10-hydroxy geraniol dehydrogenase (Cro10- HGEDH). Accession numbers for these sequences are given in legend to Fig. 2. Residues shown in white letters on black background are those identical in at least three of these proteins. known to be cytosolically localized [17]. This protein alcohol or geraniol when NADP+ was used as the cofac- possessed GEDH activity with geraniol at a 56% level tor in the oxidative direction, but only oxidized cinnamyl compared with its activity using cinnamyl alcohol, and alcohol, and not geraniol, when NAD+ was used as the was therefore designated as basil CAD1. Lastly, one cofactor. We then tested the puriWed CAD1 for activity cDNA encoded a protein (Figs. 2 and 3) whose activity with geraniol, cinnamyl alcohol, and a number of other with geraniol was 2-fold higher than with cinnamyl related substrates that included other terpene alcohols as alcohol. This protein, which we designated GEDH1, well as terpene analogues, saturated and unsaturated had 70% identity to a Catharanthus roseus protein straight-chain alcohols, and phenylpropanoid alcohols, annotated as 10-hydroxygeraniol oxidoreductase in using NADP+ as the cofactor (Table 2). The activity of GenBank as well as 66% identity with a sinapyl alcohol CAD1 with cinnamyl alcohol was 1.8-fold higher than dehydrogenase from Populus tremuloides whose activity with geraniol. It exhibited activity with a small subset of has been shown experimentally [18] and 65% identity the compounds tested, with hexanol, phenylethanol, and with a protein annotated as cinnamyl alcohol phenylpropanol showing similar reactivity to geraniol dehydrogenase from Fragaria x ananassa but whose but the remaining compounds, such as 3,7-dimethylocta- activity has not yet been demonstrated [19] (Fig. 3). It nol and cis-3-hexenol, being less reactive. Notably, was 49% identical to basil CAD1. No transit or signal CAD1 could not use nerol as a substrate. peptides could be discerned, nor any other targeting sig- The Km values of CAD1 for cinnamyl alcohol and nals when the protein sequence was analyzed by TargetP geraniol were 46 and 72 M, respectively, and the (http://www.cbs.dtu.dk/services/TargetP). enzyme maintained similar turnover rates for both substrates (Table 3). The pH optimum for CAD1 reactions Biochemical properties of puriWed recombinant CAD1 in the oxidative direction was 8.9 and the enzyme had >80% activity of the maximum in the pH range of 8.5– + + + The basil CAD1 enzyme produced in E. coli with an 9.4. The monovalent cations Na , K , and NH4 and the N-terminal hexahistidine peptide tag was puriWed by divalent cations Mn2+ and Zn2+ did not show any eVect aYnity chromatography on a Ni–NTA column. The on CAD1 activity. However, Ca2+, Mg2+, and Cu2+ each puriWed enzyme possessed activity using either cinnamyl inhibited the reaction by 60% at 5 mM concentrations. 146 Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149

Table 2 Comparison of substrate speciWcities of dehydrogenases in leaf and gland protein crude extracts and of puriWed recombinant GEDH1 and CAD1 Substrate GEDH1 CAD1 Glands crude Leaf crude % § SE % § SE % § SE % § SE Terpene alcohols Geraniol 100a 100b 100c 100d Nerol 101 § 4.4 0 131.0 § 16.4 112 § 2.6 Citronellol 16.6 § 5.3 0 60.3 § 7.9 44.7 § 3.5 Menthol 0 0 10.4 § 3.1 8.0 § 3.6 Carveol 0 0 35.9 § 4.0 10.3 § 2.5 Dihydrocarveol 0 0 14.8 § 3.0 8.6 § 2.5 3,7-Dimethyl-octanol 0 29.0 § 6.9 0 0 3-Methyl-3-buten-1-ol 6.5 § 3.2 0 8.7 § 1.5 10.4 § 2.6 3-Methyl-2-buten-1-ol 22.7 § 6.2 0 12.7 § 5.2 17.4 § 2.2 Aliphatic alcohols Ethanol 0 0 7.4 § 0.3 6.1 § 3.3 Butanol 0 4.1 § 3.4 11.7 § 0.1 7.4 § 4.7 Isobutanol 0 0 0 8.4 § 4.6 Hexanol 31.6 § 8.9 114 § 3.5 56.2 § 12.2 33.1 § 6.2 cis-3-Hexenol 13.3 § 5.4 62.6 § 5.2 21.0 § 5.1 13.8 § 1.3 Phenolics Benzyl alcohol 5.5 § 2.2 0 24.6 § 6.3 18.4 § 0.9 Phenylethanol 10.2 § 1.2 17.9 § 2.1 0 18.7 § 3.8 Phenylpropanol 22.1 § 7.2 95.4 § 2.3 20.6 § 2.5 22.9 § 3.7 Cinnamyl alcohol 53.0 § 3.1 178 § 12.5 133.1 § 11.5 58.5 § 1.8 Coniferyl alcohol 0 11.6 § 0.5 0 0 a Representing 19,840 pkat/mg protein. b Representing 7413 pkat/mg protein. c Representing 262 pkat/mg protein. d Representing 85 pkat/mg protein.

Table 3 that observed with geraniol. Finally, the other substrates Kinetic parameters of CAD1 and GEDH1 tested were oxidized at even lower rates, or not at all. ¡1 Substrate Km ( M) Vmax (pkat/ g protein) Kcat (s ) Kcat/Km The Km values of GEDH1 for geraniol and nerol were CAD1 very similar (30 and 37 M, respectively), but the Km Geraniol 72 § 13 9.8 § 20.3§ 0.08 0.0055 value for cinnamyl alcohol, 655 M, was more than 20- Cinnamyl 46 § 713§ 10.5§ 0.04 0.011 fold higher than geraniol and nerol (Table 3). The pH alcohol optimum for the GEDH1 reaction was 9.5 and the GEDH1 enzyme maintained >72% of its maximal activity over Geraniol 30 § 513§ 11.0§ 0.1 0.033 the pH range of 8.5–10.0. The monovalent cations Na+, Nerol 37 § 819§ 21.5§ 0.2 0.040 + + 2+ 2+ Cinnamyl 655 § 45 23 § 11.8§ 0.1 0.0027 K , and NH4 and the divalent cations, Ca and Mg , 2+ alcohol did not aVect GEDH activity in vitro. However, Zn inhibited the reaction by 66% at 5 mM.

Comparison of GEDH1 and CAD1 substrate speciWcities Biochemical properties of puriWed recombinant GEDH1 with NADP+-dependent dehydrogenase activities in leaves and glands The basil GEDH1 enzyme produced in E. coli was puriWed by DEAE–cellulose anion-exchange chromatog- Because both puriWed GEDH1 and CAD1 reacted raphy and gel Wltration chromatography on an FPLC with several substrates yet their range of substrates and system. The puriWed enzyme did not possess any activity the relative activities with each of these substrates were with NAD+ as a coenzyme with any alcohol substrate distinct, we tested crude protein extracts of leaves and tested. Assays of the NADP+-dependent dehydrogenase glands of SD cultivar for NADP+-dependent dehydroge- activity of the puriWed GEDH1 with the set of substrates nase activity with the same set of substrates used to test Wrst tested with CAD1 indicated that GEDH1 maintains recombinant CAD1 and GEDH1 (Table 2). Notably, the limited but not absolute substrate speciWcity (Table 2). crude extract of the glands exhibited 1.3-fold higher Nerol was an equally good substrate as geraniol, and cinnamyl alcohol dehydrogenase activity than geraniol GEDH1 oxidized cinnamyl alcohol at 50% the rate as dehydrogenase activity, yet it also had activity with nerol Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149 147 at a similar level to that with cinnamyl alcohol. In con- trast, the crude extract from whole leaves (which 100 CAD1 included glands) had much less dehydrogenase activity with cinnamyl alcohol as compared to its activity with 80 geraniol and nerol. The crude extracts of glands and/or 60 leaves were also capable of oxidizing several other compounds (e.g., carveol and menthol) which are not sub- 40 strates for either CAD1 or GEDH1. 20

Time-course analysis of geraniol oxidation by GEDH1 and by CAD1 100 GEDH1 GC analyses of the reaction products of geraniol oxidation by GEDH1 (Fig. 4) indicated that as the reaction 80 progressed, in addition to geranial and neral, nerol also 60 began to accumulate. Oxidation of geraniol by CAD1 yielded geranial and neral, but very little nerol, consis- 40 tent with our observation that nerol and neral are poor 20 substrates for CAD1.

Gland Leaf Gland Leaf Gland Leaf SD SW EMX A Standards neral geraniol nerol geranial Fig. 5. Analysis of the relative expression levels of the genes encoding B GEDH1, no NADP+, 60 min CAD1 and GEDH1 in glands and leaves of three basil cultivars. Slot blots with gland and leaf RNA samples were performed in duplicates. For each gene transcript, relative levels of transcripts are shown with C GEDH1 + NADP+, 10 min the sample having the highest levels set arbitrarily at 100%. Levels of expression of CAD1 and GEDH1 cannot be compared to one another.

D GEDH1 + NADP+, 30 min

Analysis of the expression of the genes encoding CAD1 E GEDH1 + NADP+, 60 min and GEDH1

F CAD1, no NADP+, 60 min The levels of CAD1 and GEDH1 transcripts were examined in glands of the three cultivars and compared to the levels in whole leaves (containing glands) G CAD1 + NADP+, 10 min (Fig. 5). CAD1 transcript levels in glands of SD and SW were similar to each other and both were 2.5-fold lower than the levels of CAD1 in the glands of EMX. H CAD1 + NADP+, 30 min CAD1 transcript levels in each cultivar were always several fold higher in glands than in leaves. On the other hand, GEDH1 transcript levels in glands were I CAD1 + NADP+, 60 min found at negligible levels compared to GEDH1 transcript levels in whole leaves.

12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 (min)

Fig. 4. Time-course analysis of the products of a reaction initially con- Discussion taining geraniol, NADP+, and puriWed GEDH1 and CAD1. (A) Gas chromatographic separation of authentic standards of nerol, neral, geraniol, and geranial. (B) Products after incubation of geraniol with Geraniol oxidation in basil glands may involve multiple GEDH1 without NADP+ for 60 min. (C–E) Products after incubation dehydrogenases of geraniol with GEDH1 with NADP+ for 10 min (C), 30 min (D), and 60 min (E). (F) Products after incubation of geraniol with CAD1 with- In an initial bioinformatics screen of EST databases out NADP+ for 60 min. (G–I) Products after incubation of geraniol V + of glands of three di erent basil cultivars for a with CAD1 with NADP for 10 min (G), 30 min (H), and 60 min (I). W V The reactions were carried out in 100 L volume of 100 mM glycine– potential geraniol dehydrogenase, ve di erent types of W NaOH buVer (pH 9.5) containing 1 mM geraniol, 1 mM NADP+, and ADH-related sequences were identi ed. Subsequent 1.25 g of enzyme. examination of in vitro activities of the proteins encoded 148 Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149 by these Wve sequences, produced through heterologous the lignin biosynthetic pathway, and they have hydroxyl expression in E. coli, identiWed one dehydrogenase pos- and methoxyl functionalities on the benzene ring). How- sessing catalytic eYciency >10-fold higher with geraniol ever, the Arabidopsis genome, and other plant genomes (and nerol) than with cinnamyl alcohol. Therefore, this which have been investigated in some detail, contain a enzyme was designated GEDH1. A second candidate family of CAD-like sequences that are often annotated turned out to be a bona Wde CAD that also possessed as “CAD” genes but only a portion of which can be robust activity with geraniol as a substrate, with a turn- shown to be involved in lignin biosynthesis and to act on over rate for geraniol that was lower by only 3.3-fold hydroxycinnamyl or sinapyl alcohols, based on expres- than GEDH1 and a catalytic eYciency 6-fold lower than sion proWling and biochemical characterization [22,23]. for GEDH1. The other three ADH-like sequences, Basil CAD1 is more highly expressed in the glands of all which are more distally related to basil GEDH1 and three basil cultivars examined compared to leaves CAD1 than the latter two are to each other, had no (Fig. 5); it is likely to be involved in phenylpropene bio- activity with geraniol as a substrate. synthesis, which appears to require the reduction of Surprisingly, however, expression analysis of the hydroxycinnamyl aldehydes to hydroxycinnamyl alco- GEDH1 and CAD1 genes indicated that CAD1 is hols [16], and its activity with geraniol may be fortuitous. expressed at higher levels in the glands compared to the The involvement of CAD1 in geraniol oxidation rest of the basil leaf, while GEDH1 is expressed more explains the presence of GEDH activity in basil cultivars abundantly in leaf cells other than glands (Fig. 5). While that do not produce citral. On the other hand, the much the expression levels of CAD1 and GEDH1 could not be higher levels of basil GEDH1 in leaves, which do not directly compared to each other, analysis of the abun- appear to synthesize geraniol, as compared to glands dance of speciWc ESTs in the database (Table 1) supports remain obscure. Although this enzyme has high aYnity the conclusion that CAD1 is more highly expressed than for geraniol, it is possible that it is involved in the oxida- GEDH1 in the glands of all three cultivars. And while tion of other alcohols. GEDH1 shows high similarity to transcript levels do not necessarily correlate with levels putative 10-hydroxygeraniol dehydrogenases encoded of enzymatic activity, the dehydrogenase activity proWle by cDNAs isolated from C. roseus and Camptotheca measured in the glands suggests a predominance of acuminata, although no biochemical data have been CAD1 activity over GEDH1 activity, since SD gland published regarding these two GenBank entries (Acces- crude extract displays higher levels of activity with cinn- sion Nos. AAQ55962 and AAQ20892, respectively). 10- amyl alcohol than with geraniol. However, if we assume Hydroxygeraniol is an important intermediate in the that the majority of the geraniol-oxidizing activity pres- biosynthesis of iridoids, and 10-hydroxygeraniol dehy- ent in SD gland crude extract is due to CAD1, than it is drogenases were puriWed from catmint [24] and Rau- diYcult to explain the high levels of nerol-oxidizing wolWa serpentina [25], respectively. Both of these enzymes activity in the glands (Table 2), since nerol is not a sub- were reported to work also on nerol and geraniol. How- strate for CAD1. It is therefore likely that in addition to ever, no protein sequences were reported for the puriWed GEDH1, at least one more dehydrogenase that is capa- enzymes. Since basil GEDH1 shows high similarity to ble of oxidizing nerol is present in the glands. It is possi- putative 10-hydroxygeraniol dehydrogenases, it is possible that one or more of the three ADH-like enzymes ble that it also has activity with 10-hydroxygeraniol. whose ESTs we have attempted to express in E. coli are However, there is no evidence for the production of 10- active with nerol (and perhaps geraniol as well) in planta hydroxygeraniol in basil leaves and we have not yet but not in E. coli. Additional heterologous expression tested basil GEDH1 for such activity. experiments in yeast or plants may resolve this question. Other ADHs capable of oxidizing geraniol may also be Formation of nerol in basil glands occurs via the reduction present in the glands, but currently we have not identi- of neral in a reaction catalyzed by GEDH Wed any candidate ESTs in the databases of the glands of the three basil cultivars (which total >7000 ESTs) that Nerol, the cis-isomer of geraniol, has been found in could possibly encode such ADHs. many plants but an enzyme that catalyzes the formation Bona Wde CAD, and a related enzyme sinapyl alcohol of nerol directly from GPP has not yet been reported. A dehydrogenase (SAD) that is also closely related to basil similar situation exists in basil: although nerol is found GEDH1 (Figs. 2 and 3), are involved in the lignin bio- in basil glands of Sweet Dani cultivar alongside geraniol synthetic pathway [18,20,21]. Cinnamyl alcohol, 4- (Fig. 1), we had previously identiWed GPP-dependent hydroxycinnamyl alcohol (one of the true substrates/ geraniol synthase activity but no GPP-dependent nerol products of CADs in vivo), and sinapyl alcohol are all synthase in these glands [11]. In the present study, we primary alcohols as is geraniol, and they share with gera- demonstrate that incubation of geraniol with GEDH1 niol the additional property of having a double bond results in the formation of both geranial and neral, as between C2 and C3 (4-hydroxycinnamyl alcohol and well as nerol (Fig. 4), with the proportion of neral and sinapyl alcohol are phenylpropanoid intermediates in particularly nerol increasing over time. This observation Y. Iijima et al. / Archives of Biochemistry and Biophysics 448 (2006) 141–149 149

and by the National Research Initiative of the USDA H2 H C C Cooperative State Research, Education, and Extension OH O Service, Grant No. 2001-35318-10006. GEDH1

CAD1 GES Geraniol Geranial References H C OH [1] N. Singh-Sangwan, R.S. Sangwan, R. Luthra, R.S. Thakur, Planta H2 C Med. 59 (1993) 168–170. OPP [2] M. Miyazawa, H. Kameoka, Agric. Biol. Chem. 52 (1988) 2961–2963. [3] R.J. Grayer, G.C. Kite, F.J. Goldstone, S.E. Bryan, A. Paton, E. Putievsky, Phytochemistry 43 (1996) 1033–1039. HC [4] J.E. Simon, M.R. Morales, W.B. Phippen, R.F. Vieira, Z. Hao, in: GPP OH J. Janick (Ed.), Perspectives on New Crops and New Uses, ASHS X Press, Alexandria, VA, 1999, pp. 499–505. [5] F. Dawson, Naval Stores Review (March/April) (1994) 6–12. [6] L. Somogyi, Chem. Ind. (1996) 170–173. GEDH1 [7] R.S. Sangwan, N. Singhsangwan, R. Luthra, J. Plant Physiol. 142 (1993) 129–134. H2C HC OH O [8] Y. Sekiwa-Iijima, Y. Aizawa, K. Kubota, J. Agric. Food Chem. 49 (2001) 5902–5906. [9] D.V. Banthorpe, E. Cardemil, M.C. Contreras, Phytochemistry 15 Nerol Neral (1976) 391–394. [10] V.H. Potty, J.H. Bruemmer, Phytochemistry 9 (1970) 1003–1007. Fig. 6. A proposed mechanism for the production of geraniol, geranial, [11] Y. Iijima, D.R. Gang, E. Fridman, E. Lewinsohn, E. Pichersky, neral, and nerol from geranyl diphosphate (GPP) in basil glands. Plant Physiol. 134 (2004) 370–379. CAD1, cinnamyl alcohol dehydrogenase; GEDH1, geraniol dehydro- [12] Y. Iijima, R. Davidovich-Rikanati, E. Fridman, D.R. Gang, E. Bar, genase; GES, geraniol synthase. The enolic tautomers of geranial and E. Lewinsohn, E. Pichersky, Plant Physiol. 136 (2004) 2736–3724. neral, which form the link between geranial and neral, are shown [13] D. Tholl, C.M. Kish, I. Orlova, D. Sherman, J. Gershenzon, E. inside the brackets. Only hydrogen atoms bound to C1 are shown. Pichersky, N. Dudareva, Plant Cell 16 (2004) 977–992. [14] J.D. Thompson, T.J. Gibson, F. Plewniak, F. Jeanmougin, D.G. suggests that in basil glands neral is produced through Higgins, Nucleic Acids Res. 24 (1997) 4876–4882. V [15] R.D.M. Page, Comput. Appl. Biosci. 12 (1996) 357–358. the keto–enol tautomerization of geranial, which a ords [16] D.R. Gang, J.H. Wang, N. Dudareva, K.H. Nam, J.E. Simon, a facile trans to cis rotation of the C2–C3 bond resulting E. Lewinsohn, E. Pichersky, Plant Physiol. 125 (2001) in the observed cis conWguration found in neral and 539–555. nerol. Nerol is likely generated through the reduction of [17] J. Nakashima, T. Awano, K. Takabe, M. Fujita, H. Saiki, Plant neral in a reaction catalyzed by GEDH1, using reduced Cell Physiol. 38 (1997) 113–123. [18] L. Li, X.F. Cheng, J. Leshkevich, T. Umezawa, S.A. Harding, V.L. NADPH (Fig. 6). While in vivo very little nerol accumu- Chiang, Plant Cell 13 (2001) 1567–1586. lates relative to neral, most likely because the neral (and [19] R. Blanco-Portales, N. Medina-Escobar, J.A. Lopez-Raez, J.A. geranial) are secreted out of the gland cell and into the Gonzalez-Reyes, J.M. Villalba, E. Moyano, J.L. Caballero, sac, in the in vitro assays the NADPH concentration J. Munoz-Blanco, J. Exp. Bot. 53 (2002) 1723–1734. builds over time due to the oxidation of geraniol. This [20] A.M. Anterola, N.G. Lewis, Phytochemistry 61 (2002) 221–294. [21] A.M. Boudet, C. Lapierre, J. Grima-Pettenati, New Phytol. 129 reduced cofactor can then be used to reduce neral to (1995) 203–226. nerol (as well as geranial back to geraniol). [22] S.J. Kim, M.R. Kim, D.L. Bedgar, S.G. Moinuddin, C.L. Cardenas, L.B. Davin, C. Kang, N.G. Lewis, Proc. Natl. Acad. Sci. USA 101 (2004) 1455–1460. Acknowledgments [23] E.K. Bomati, J.P. Noel, Plant Cell 17 (2005) 1598–1611. [24] D.L. Hallahan, J.M. West, R.M. Wallsgrove, D.W.M. Smiley, G.W. Dawson, J.A. Pickett, J.G.C. Hamilton, Arch. Biochem. Biophys. We thank Drs. Efraim Lewinsohn and Joseph P. Noel 318 (1995) 105–112. for valuable comments on the manuscript. This research [25] H. Ikeda, N. Esaki, S. Nakai, K. Hashimoto, S. Uesato, K. Soda, was supported by USDA-BARD Grant IS-3332-02C T. Fujita, J. Biochem. 109 (1991) 341–347.