Comparative transcript and alkaloid profiling in Papaver species identifies a short chain dehydrogenase/reductase involved in morphine biosynthesis


*(fax +49 345 5582 1209; e-mail


Plants of the order Ranunculales, especially members of the species Papaver, accumulate a large variety of benzylisoquinoline alkaloids with about 2500 structures, but only the opium poppy (Papaver somniferum) and Papaver setigerum are able to produce the analgesic and narcotic morphine and the antitussive codeine. In this study, we investigated the molecular basis for this exceptional biosynthetic capability by comparison of alkaloid profiles with gene expression profiles between 16 different Papaver species. Out of 2000 expressed sequence tags obtained from P. somniferum, 69 show increased expression in morphinan alkaloid-containing species. One of these cDNAs, exhibiting an expression pattern very similar to previously isolated cDNAs coding for enzymes in benzylisoquinoline biosynthesis, showed the highest amino acid identity to reductases in menthol biosynthesis. After overexpression, the protein encoded by this cDNA reduced the keto group of salutaridine yielding salutaridinol, an intermediate in morphine biosynthesis. The stereoisomer 7-epi-salutaridinol was not formed. Based on its similarities to a previously purified protein from P. somniferum with respect to the high substrate specificity, molecular mass and kinetic data, the recombinant protein was identified as salutaridine reductase (SalR; EC Unlike codeinone reductase, an enzyme acting later in the pathway that catalyses the reduction of a keto group and which belongs to the family of the aldo-keto reductases, the cDNA identified in this study as SalR belongs to the family of short chain dehydrogenases/reductases and is related to reductases in monoterpene metabolism.


The benzylisoquinoline alkaloids comprise a group of secondary metabolites with roughly 2500 structures. Some of them possess important pharmacological effects such as the vasodilator papaverine, the antimicrobial sanguinarine, the antitussive codeine or the analgesic morphine (Kutchan, 1998; Kutchan et al., in press). The biosynthesis of the benzylisoquinolines starts with the condensation of the tyrosine-derived tyramine and p-hydroxyphenylacetaldehyde by norcoclaurine synthase. The product (S)-norcoclaurine is further converted by several S-adenosylmethionine-dependent methyltransferases and one P450-monoxygenase. The enzyme (S)-norcoclaurine 6-O-methyltransferase (6-OMT) methylates the 6-hydroxy position of (S)-norcoclaurine to (S)-coclaurine, which then undergoes N-methylation by coclaurine N-methyltransferase to yield (S)-N-methylcoclaurine. The P450-dependent mono-oxygenase (S)-N-methylcoclaurine 3′-hydroxylase (CYP80B3) inserts a hydroxyl group at the 3′-position and the resulting 3′-hydroxy-N-methylcoclaurine is methylated by (S)-3′-hydroxy-N-methylcocaurine 4′-O-methyltransferase (4′-OMT) to (S)-reticuline (Kutchan, 1998; Kutchan et al., in press). Up to this step, the biosynthesis is the same for all structural types of benzylisoquinolines. The broad structural variety of the benzylisoquinolines originates from the different modifications of the central intermediate (S)-reticuline (Figure 1). The core structure can be modified by oxidations and methylations to the true benzylisoquinolines such as papaverine. Other initial reactions are C–C phenol coupling or formation of C–C bonds between the isoquinoline and the benzyl moiety leading to aporphines, pavines and isopavines. Alternatively, (S)-scoulerine is formed by oxidative C–C bond formation between the N-methyl group and carbon-2 of the benzyl moiety, giving rise to phthalideisoquinolines, protoberberines and protopines, which can be further modified to papaverrubines/rhoeadines and benzophenanthridines (Kutchan, 1998; Rönsch, 1986). Whereas all of these classes of benzylisoquinolines are formed by modification of (S)-reticuline, the pathway to the morphinans is initiated by the conversion of stereochemistry at C-1 from (S)- to (R)-reticuline, catalysed in a two-step reaction including an oxidation of (S)-reticuline by 1,2-dehydroreticuline synthase to the 1,2,-dehydroreticulinium ion and subsequent reduction of the ion to (R)-reticuline by 1,2,-dehydroreticulinium ion reductase (Figure 1). Carbon–carbon phenol coupling between C-12 and C-13 by the P450 oxygenase salutaridine synthase yields salutaridine. The stereospecific, NADPH-dependent salutaridine reductase (SalR) reduces the keto group at C-7 and 7(S)-salutaridinol is formed, which is acetylated by 7(S)-salutaridinol-O-acetyltransferase (SalAT). Elimination of the acetyl residue through formation of an oxide bridge yields thebaine, the first pentacyclic alkaloid of the morphinan type. Two demethylations from thebaine to neopinone and codeine to morphine, and the reduction of codeinone to codeine by codeinone reductase (COR) result in the formation of morphine (Kutchan, 1998; Kutchan et al., 2006). Alternatively, thebaine can first be demethylated to oripavine and further to morphinone, which is then reduced by COR to morphine.

Figure 1.

 The benzylisoquinoline pathway.
One representative compound for each benzylisoquinoline class with emphasis on morphine biosynthesis is shown. The enzymes indicated are: 1, 1,2-dehydroreticuline synthase; 2, 1,2-dehydroreticulinium ion reductase; 3, salutaridine synthase; 4, salutaridine reductase; 5, salutaridinol 7-O-acetyltransferase; 6 and 8, demethylations of unknown enzymatic mechanism; 7, codeinone reductase.

The basic benzylisoquinoline pathway up to (S)-reticuline has been elucidated at the enzymatic and the molecular biological level during the past years and for all enzymes cDNAs have been isolated from several plant species (Choi et al., 2002; Facchini and De Luca, 1994; Facchini and Park, 2003; Frick and Kutchan, 1999; Huang and Kutchan, 2000; Morishige et al., 2000; Ounaroon et al., 2003; Samanani et al., 2004; Ziegler et al., 2005). However, much less is known about the pathways downstream of (S)-reticuline. In some cases, such as for the pavines and isopavines, the precise metabolic pathways are not yet known. For the biosynthesis of protoberberines and benzophenanthridines the metabolic flow has been elucidated and enzymes have been measured and partially purified (Kutchan, 1998). In this pathway, several cDNAs could be isolated, such as the berberine bridge enzyme, (S)-scoulerine 9-O-methyltransferase, columbamine-O-methyltransferase, (S)-canadine synthase and (R,S)-reticuline 7-O-methyltransferase (Dittrich and Kutchan, 1991; Ikezawa et al., 2003; Morishige et al., 2002; Ounaroon et al., 2003; Takeshita et al., 1995). Because of its high economic importance, the morphinan branch of the benzylisoquinoline pathway has been under intensive investigation over the past years and two enzymes, SalAT and COR, have been purified and the encoding cDNAs cloned (Grothe et al., 2001; Lenz and Zenk, 1995a,b; Unterlinner et al., 1999). With the exception of the demethylation steps downstream of thebaine, all enzymes of the pathway have been characterized and partially purified (De-Eknamkul and Zenk, 1992; Gerardy and Zenk, 1992, 1993; Hirata et al., 2004).

Benzylisoquinolines are mainly produced by plants belonging to the order Ranunculales. In particular, plants of the genus Papaver, of which 70 species are described, accumulate a rich spectrum of different benzylisoquinolines (Preininger, 1986). The pharmacologically most important species is the opium poppy, Papaver somniferum. Together with Papaver bracteatum and Papaver setigerum, this is the only plant species known to date capable of biosynthesizing pentacyclic morphinan alkaloids. Whereas P. bracteatum accumulates thebaine as the end product, P. setigerum and P. somniferum are able to further metabolize this molecule to codeine and morphine. The latter plant has been subjected to extensive breeding programs, partially in order to change the alkaloid profile and, mainly, to massively increase the content of specific alkaloids. Although these breeding programs have succeeded in the generation of high-alkaloid producing opium poppy varieties, the molecular mechanisms causing these phenotypes are not known. Likewise, the reason why among all these different, but genetically closely related Papaver species, only P. bracteatum, P. somniferum and P. setigerum are capable of producing pentacyclic morphinan alkaloids is not known. One method by which to investigate the underlying mechanisms making the opium poppy so distinct from the other Papaver species is the global analysis of gene expression by array technologies. Recording gene expression profiles in various Papaver species and correlating these with the benzylisoquinoline profile could lead to the discovery of cDNAs responsible for the unique quantitative and qualitative chemotype of P. somniferum. Using this approach, the comparison between morphine-containing and morphine-free Papaver species showed a higher expression of genes involved in benzylisoquinoline biosynthesis in P. somniferum (Ziegler et al., 2005). In this paper, we identify a cDNA encoding a member of the short chain dehydrogenase/reductase family of unknown function exhibiting significantly higher expression in P. somniferum. Characterization of the recombinant protein identified it as SalR of morphine biosynthesis.


Expressed sequence tag sequencing project from P. somniferum

In order to obtain additional cDNAs for macroarray analysis, we initiated an expressed sequence tag (EST) sequencing project from 2-cm long P. somniferum stem sections adjacent to the capsule. This tissue has been previously shown to contain high amounts of benzylisoquinoline alkaloids accompanied by high expression of most of the cDNAs coding for enzymes in benzylisoquinoline biosynthesis isolated thus far (Facchini and Park, 2003; Grothe et al., 2001; Huang and Kutchan, 2000; Ounaroon et al., 2003; Unterlinner et al., 1999). In total, 2078 ESTs with minimum lengths of 500 bp were sequenced from the 5′-end producing a set of 1152 unique sequences (UniGenes; 55.4%) consisting of 865 sequences occurring only once (singletons) and 1213 sequences, which, after assembling, represented 287 different cDNAs (assemblies). Each UniGene was submitted to the non-redundant database on the NCBI server using the blastx algorithm. Using an expectation value (E value) of 10−9 as threshold, slightly more than 50% of the UniGenes either showed no significant homology (28%) or homology to hypothetical or expressed proteins of unknown function (24%) (Figure 2). Additionally, 1% of the UniGenes and 3% of all sequences coded for proteins of viral origin with the coat protein from the turnip mosaic virus as the most abundant representative.

Figure 2.

 Classification and expression analyses of expressed sequence tags.
(a) For all sequences: no homology = sequences with homology of an E value >10−9; unknown/putative = sequences with homology to uncharacterized expressed proteins or hypothetical proteins.
(b) For sequences showing homology to a database entry: PS = photosynthesis; structural = cytoskeleton, cell wall synthesis/degradation proteins; MLP = major latex proteins; protein syn/deg = cell growth, cell division, DNA and protein synthesis; stress/redox = abiotic/biotic stress/redox control; G-prot = G-proteins; signal = cellular communication/signal transduction.
(c) For sequences coding for proteins involved in metabolism: AA = amino acid metabolism; alk = alkaloid biosynthesis; phenyl = phenylpropanoid biosynthesis; isopr = isoprenoid biosynthesis.

The remaining UniGenes mainly coded for proteins involved in cell growth, cell division, DNA synthesis and protein synthesis (33%) followed by those implicated in metabolism (25%) and in stress response and redox control (12%). Most of the remaining 30% of the UniGenes coded for proteins related to processes in photosynthesis (7%) and with functions in transport and cellular communication/signal transduction (8% each). Considering all sequences, and not only the UniGene set, a similar distribution into functional categories was observed. Strong differences were only observed for proteins involved in stress response and redox control as well as for major latex proteins (MLP). MLPs constitute 1% of the UniGenes, but 5% of all sequences, which is representative of the high abundance of these proteins in laticifers (Nessler and von der Haar, 1990). The twofold difference in the representation of cDNAs coding for proteins implicated in stress responses and redox control in all sequences compared to the UniGene set is due to the high abundance of a metallothionein, with 134 sequences. Some 1.8% of the UniGenes could be assigned to secondary metabolism, including four enzymes implicated in benzylisoquinoline biosynthesis, tyrosine decarboxylase (TYDC), CYP80B3, 4′-OMT and COR. Other sequences with potential functions in secondary metabolism include 18 transcription factors, eight dehydrogenases/reductases, seven methyltransferases, six cytochrome P450-dependent mono-oxygenases, three oxygenases and two ATP-binding cassette (ABC) transporters.

Alkaloid composition of Papaver species

In a previous analysis aimed at isolating cDNAs that make P. somniferum capable of production of morphinan alkaloids, nine other Papaver species devoid of this benzylisoquinoline class were compared with three P. somniferum varieties (Ziegler et al., 2005). In the current investigation, seven additional Papaver species and three more P. somniferum varieties were included. Although there are many reports in the literature about the alkaloid composition of many Papaver species, considerable variation can occur depending on the origin of the species, the tissue extracted as well as on the growth conditions. It was, therefore, necessary to analyse the alkaloid composition of each plant species that was later used for expression analysis. A section of the stem about 2-cm long adjacent to the base of the capsule was harvested from field-grown plants and one slice of thickness 2 mm was excised and used for alkaloid extraction and analysis. A major focus of this analysis was to investigate which plants produce alkaloids of the morphinan type. In order to get a first overview, the presence of masses indicative for morphinan alkaloids was examined by electrospray Fourier transform ion cyclotron resonance mass spectrometry (ESI-FT-ICRMS). If such masses were present in the plants, HPLC-MS and HPLC-MS/MS analysis were conducted to see whether the respective alkaloid belonged to the morphinan or to another benzylisoquinoline class. In most of the P. somniferum varieties masses of [M+H]+ m/z 286, m/z 298, m/z 300, m/z 312, m/z 328 and m/z 330 could be detected, which could be identified by HPLC-MS/MS as the morphinan alkaloids morphine, codeinone/oripavine (an intermediate of an alternative biosynthetic route to morphine downstream of thebaine), codeine, thebaine, salutaridine/1,2-dehydroreticulinium ion, and salutaridinol, respectively. In the P. somniferum mutant top1, the mass for codeine could not be detected and the signal of [M+H]+ m/z 286 could not be identified as morphine (Table 1) (Millgate et al., 2004). In the P. somniferum mutant Nosca, which accumulates high amounts of the phthalideisoquinoline-type alkaloid noscapine ([M+H]+ m/z 414) in addition to morphine, codeine and thebaine, signals indicative for oripavine, salutaridine, 1,2-dehydroreticulinium ion and salutaridinol could not be detected. This can be attributed to the lower amount of morphinan alkaloids compared with the main alkaloid noscapine. Masses indicative of morphinan alkaloids also occurred in extracts of other Papaver species, but only in Papaver arenarium and P. bracteatum did LC-ESI-MS/MS analysis show characteristic fragments for a morphinan structure. These signals could be identified as the morphinan alkaloids N-demethylcodeine, thebaine and thebaine-N-oxide. The major alkaloids in other Papaver species belonged to the scoulerine-derived papaverrubines, protoberberines, protopines and rhoeadines (data not shown).

Table 1.   Alkaloid composition of Papaver species
Species[M+H]+m/z indicative of morphinansaElemental compositionError (p.p.m.)Morphinan alkaloid
  1. aMasses indicative for morphinan alkaloids are: 286.14376 (morphine), 298.14376 (oripavine, codeinone), 300.15942 (codeine), 312.15942 (thebaine), 328.15433 (salutaridine, 1,2-dehydroreticulinium ion), 330.16998 (salutaridinol).bMillgate et al., 2004.

Papaver argemoneNo  No
P. arenarium286.14388[C17H20NO3]+0.5N-demethylcodeine
P. commutatumNo  No
P. bracteatum312.15948[C19H22NO3]+0.2Thebaine
P. orientale312.15953[C19H22NO3]+0.4No
P. lateritium300.15947[C18H22NO3]+0.3No
P. pyrenaicum298.14425[C18H22NO3]+1.6No
P. atlanticum300.15890[C18H20NO3]+1.7No
P. pilosum300.15908[C18H20NO3]+1.1No
P. fugax298.14303[C18H18NO3]+2.5No
P. oreophilum328.15412[C19H22NO4]+0.6No
P. hybridumNo  No
P. glaucumNo  No.
P. somniferum    
PasoAll  All
NopaAll  All
NoscaAll, but  All, but oripavine,
 298.14376,  Codeinone, salutaridine,
 328.15433,  1,2-dehydroreticulinium
 330.16998  Ion, salutaridinol
PapaAll  All
Fool OriAll  All
toplbAll but  All but codeine,
 300.15942,  Morphine, codeinone

Gene expression analysis of Papaver species

In order to correlate the alkaloid profile with the gene expression for each individual plant, the RNA was extracted from the remainder of the stems, reverse transcribed, labelled with [α33P]dATP and hybridized to the macroarrays. The macroarrays contained the 1152 UniGenes from the EST project as well as the 849 UniGenes described in Ziegler et al. (2005) as PCR fragments, spotted in quadruplicate. Additionally, eight cDNAs coding for enzymes involved in benzylisoquinoline biosynthesis, which had already been isolated from P. somniferum, were included. In total, 54 individual plants were analysed, among them 23 P. somniferum plants from six varieties (5 × PaSo, 2 × Nopa, 5 × Nosca, 3 × Papa, 4 × Fool Ori, 4 × top1) and 31 plants from other species (4 × P. bracteatum, 3 × each Papaver glaucum and Papaver orientale, 2 × each P. arenarium, Papaver argemone, Papaver atlanticum, Papaver commutatum, Papaver fugax, Papaver hybridum, Papaver lateritium, Papaver pilosum, Papaver pyrenaicum, as well as one plant each from Papaver dubium, Papaver oreophilum and Papaver pavonium). Hierarchical clustering of the expression data exhibited six clusters showing enhanced expression of cDNAs in P. somniferum varieties compared with other Papaver species. Seven out of the eight cDNAs involved in benzylisoquinoline biosynthesis cluster together (Figure 3), such as cDNAs for the early pathway up to (S)-reticuline (6-OMT, CYP80B1, 4′-OMT) as well as cDNAs involved in the morphinan-specific branch of the pathway (SalAT, COR). In total, 69 cDNAs are present in these six clusters, of which 62% showed no homology to the database or homology to uncharacterized genes. Eight cDNAs code for proteins presumably involved in secondary metabolism, among them six P450-dependent mono-oxygenases, of which four are present in the cluster shown in Figure 3. The whole median centred data set can be viewed in the Supplementary Table S1, which can be imported into the cluster program for further analysis. The intent of this study was to isolate and characterize cDNAs putatively involved in accumulation of morphinan alkaloids. As most of the cDNAs that are already known to be involved in alkaloid biosynthesis show very similar and higher expression in morphinan-producing P. somniferum plants, we first focused on that cluster in order to choose ESTs for further characterization. The EST 16B1 showed higher expression in P. somniferum plants compared with other plants. This EST exhibited the highest amino acid sequence similarity to a putative short chain dehydrogenase/reductase (SDR) from Arabidopsis thaliana. Members of this enzyme family have already been shown to catalyse reactions in secondary metabolism, for example the stereospecific reduction of tropinone to tropine in tropane alkaloid metabolism (Nakajima et al., 1999). In morphine biosynthesis, there are three reductive candidate steps, the reduction of 1,2-dehydroreticulinium ion to (R)-reticuline, of salutaridine to salutaridinol, and from codeinone to codeine, respectively. Because of its expression profile and its homology to enzymes catalysing reductions in biosynthesis of secondary metabolites, we chose this EST for further characterization to investigate a putative function in benzylisoquinoline metabolism.

Figure 3.

 Clustered display of gene expression analysis of Papaver species.
One cluster showing increased expression in P. somniferum varieties is shown. Indicated in blue: cDNAs coding for enzymes in benzylisoquinoline biosynthesis; indicated in red: cDNA characterized in this study. The P. somniferum top1 mutant is described in Millgate et al. (2004).

Analysis of EST 16B1 and its full-length sequence

The EST 16B1 comprised 750 bp and showed the highest homology to a putative SDR from A. thaliana with an amino acid identity of 57% covering amino acid positions 120 to the C-terminus. Thus, roughly 400 nucleotides from the N-terminal sequence were missing in the Papaver EST, which were amplified by rapid amplification of 5′ complementary DNA ends (5′-RACE) from P. somniferum RNA. The assembly of the 5′-RACE sequence with the EST yielded a cDNA of 1326 bp with an open reading frame of 936-bp coding for a protein of 311 amino acids with a molecular mass of 34.1 kDa and an isoelectric point of 4.69. In accordance with the macroarray data, Northern blot analysis using the full-length sequence of EST 16B1 as a probe showed the highest expression in morphinan alkaloid-containing P. somniferum varieties, P. arenarium and P. bracteatum, whereas the mRNA level in other Papaver species was very low or not detectable (Figure 4). To investigate the possibility that differences in signal intensities could be due to low sequence homology between the P. somniferum-specific probe and the respective sequence from the other Papaver species, the homologous cDNA was amplified and sequenced from P. bracteatum. The sequence showed 96% identity on the nucleotide level to the P. somniferum cDNA. Hybridization of the same Northern blot with labelled P. bracteatum cDNA as probe showed a similar result as with the P. somniferum probe, thus confirming the differential expression of this gene between species rather than differences in hybridization kinetics.

Figure 4.

 Ribonucleic acid gel blot analysis.
Total RNA from different Papaver species was separated by RNA gel electrophoresis, blotted on a membrane and hybridized with [α32P]dATP-labelled 16B1 probe from P. somniferum (top) or P. bracteatum (bottom). The photographs of ethidium bromide visualized 28S rRNA are shown as an RNA loading control.

The full-length sequence of 16B1 exhibits several motifs characteristic for the large family of short chain dehydrogenases/reductases such as the GxxxGxG segment characteristic of the coenzyme-binding fold in dehydrogenases or the catalytic YxxxK motif (Figure 5; Jörnvall et al., 1995). Database search revealed the highest homology to an uncharacterized SDR from A. thaliana (At3g61220; 55% amino acid identity), and to three SDRs acting in monoterpene metabolism from Mentha × piperita, (−)-menthone:(+)-(3S)-neomenthol reductase (MpMNR), (−)-isopiperitenone reductase (MpISPR) and (−)-menthone:(−)-(3R)-menthol reductase (MpMMR) with more than 45% identity (Figure 5; Davis et al., 2005; Ringer et al., 2003).

Figure 5.

 Protein sequence alignment of short chain dehydrogenases.
The amino acid sequence alignments were performed using the clustalw application (Thompson et al., 1994) of MegAlign (DNASTAR Inc). The protein sequence of 16B1 isolated from P. somniferum was aligned to the sequences of an SDR from A. thaliana (At3g61220) of unknown function, to Mentha × piperita (−)isopiperitenone reductase (MpISPR; GenBank accession number AY300162), (−)-menthone:(−)-menthol reductase (MpMMR; AY288138), (−)-menthol:(+)-neomenthol reductase (MpMNR; AY288137) and to the SDR from peppermint isolated in this study with (−)-menthol:(+)-neomenthol reductase activity (M-MNR). Black shading represents identical residues for at least five sequences; grey boxes indicate similar residues for all sequences. The nucleotide-binding motif TGxxxGhG and the motif GxhDhhhNNAG stabilizing the central β-sheet are boxed. Asterisks indicate the possible YxxxK catalytic centres and possible catalytic serine residues are emphasized by a line on top of the sequence. Residues responsible for the preference for NADP(H) over NAD(H) are indicated by arrowheads.

Purification and functional characterization of the recombinant enzyme

The full-length sequence was cloned into the expression vector pQE30 containing a six-histidine N-terminal extension and over-expressed in Escherichia coli strain SG13009. The recombinant protein had a relative molecular mass of 40 kDa as determined by SDS–PAGE, which compared favourably with the calculated mass of 37.3 kDa deduced from the translation of the cDNA. The native molecular mass of between 47.1 and 50.1 kDa, determined by gel filtration on a calibrated Superdex TM75 column, revealed the protein to be a monomer, but with an increased mass compared with the result obtained from SDS–PAGE. Based on the sequence homology to SDRs, pathway intermediates that undergo a reduction during morphine biosynthesis were chosen as substrates. There are three reductive steps in the biosynthesis of morphine, the conversion of 1,2-dehydroreticulinium ion to (R)-reticuline, and the reductions of the keto group of salutaridine and codeinone to salutaridinol and codeine, respectively (Figure 1). As COR had already been cloned and the recombinant protein proven to be highly specific (Unterlinner et al., 1999), we preferably investigated the reduction from 1,2-dehydroreticulinium ion to (R)-reticuline and the reduction from salutaridine to salutaridinol. After incubation with the recombinant 16B1 protein, conversion of 1,2-dehydroreticulinium ion to reticuline could not be observed after HPLC and HPLC-MS analysis. With salutaridine as the substrate, a peak at the same retention time and identical UV spectrum as salutaridinol could be observed, whereas no signal for the stereoisomer 7-epi-salutaridinol could be detected (Figure 6d). Control incubations with extracts from bacteria that had not been induced or that had been induced but contained empty vector did not show any conversion of salutaridine (Figure 6b,c). Additionally, only the LC-MS analysis of the incubation with the recombinant 16B1 protein showed the [M+H]+ ion of salutaridinol (m/z 330) in addition to the [M+H]+ ion of the substrate salutaridine (m/z 328). Furthermore, characteristic fragments of salutaridinol at m/z 58 ([CH3CHNHCH3]+, ethylidene-methyl-ammonium ion representative of a morphinan structure), m/z 181 [M+H–CH2CHNHCH3–2MeOH–CO]+, as well as m/z 267 [M+H–CH3NH2–MeOH]+ and its further loss of CO (m/z 239) were detected by LC-MS/MS (Raith et al., 2003).

Figure 6.

 HPLC analysis of salutaridine reductase assays.
(a) Salutaridine and salutaridinol standards.
(b–d) Assays with extracts from bacteria bearing (b) the overexpression vector pQE30 without an insert and induced with 1 mm isopropyl-β-d-thiogalactopyranoside (IPTG), (c) the overexpression vector pQE30 with insert 16B1 without IPTG induction, and (d) the overexpression vector pQE30 with an insert and induced with 1 mm IPTG. Peaks at 2.5 and 3 min derive from bacterial extracts and are of unknown identity.

The recombinant protein exhibited the highest activity at a temperature between 30 and 35°C with half maxima at 20 and 40°C. The pH dependence for the conversion of salutaridine to salutaridinol by the recombinant protein revealed a pH optimum between pH 5.5 and 6.0 with a sharp increase starting from pH 4.7 and a slower decrease with half-maximal activity at pH 7.5. No activity could be detected at pH 9. A partially purified protein from P. somniferum responsible for the reduction of salutaridine had been shown to catalyse the reverse reaction from salutaridinol to salutaridine in the presence of NADP and at higher pH (Gerardy and Zenk, 1993). The incubation of the recombinant 16B1 protein with salutaridinol and NADP at pH 9.5 and analysis of the reaction products by HPLC and HPLC-MS revealed a mass signal for salutaridine ([M+H]+ at m/z 328) in addition to salutaridinol ([M+H]+ at m/z 330), showing that the protein is capable of catalysing the reverse reaction. The turnover rate of the recombinant protein for the reverse reaction was maximal at pH 9.5 with a sharp decline to zero activity at pH 10. Only 20% of the maximum activity for the reverse reaction could be observed at the optimum pH for the forward reaction at pH 6. At optimal pH conditions and the same concentration of cofactor, the reverse reaction exhibited an almost twofold higher reaction velocity than the forward reaction, whereas the Km values for the substrates were in a similar range with 30.9 μm for salutaridine and 23 μm for salutaridinol (Table 2). The kinetic data showed a twofold stronger affinity of the enzyme for NADPH in the forward reaction compared with NADP in the reverse reaction, whereas the catalytic efficiency for the substrates was higher for the reverse reaction at a constant cofactor concentration of 500 μm. The recombinant enzyme strongly preferred the phosphorylated cofactor with 20% of the activity when NADH was used instead of NADPH in the forward reaction and 5% of the activity in the reverse reaction when NADP was substituted with NAD.

Table 2.   Kinetic parameters of salutaridine reductase for the forward and reverse reaction and the cofactors
SubstrateKmcofactor (μm)Vmax cofactor (pkat/mg)Km substrate (μm)Vmax substrate (pkat/mg) Vmax/Km
  1. aNADPH as cofactor, salutaridine concentration 100 μm. bNADP as cofactor, salutaridinol concentration 100 μm.

  2. cNADPH concentration 500 μm. dNADP concentration 500 μm.

Salutaridine80 ± 9a9.4 ± 0.3a30.9 ± 13.2c5.8 ± 1.3c0.18
Salutaridinol198 ± 56b11.6 ± 1.5b23.0 ± 6.9d10.6 ± 0.8d0.46

Substrate specificity of the recombinant protein

Various benzylisoquinolines were tested as substrate and the reaction products were analysed by LC-MS (Figure 7). No reduction of the N–C-1 double bond of the 1,2-dehydroreticulinium ion to reticuline was detected. Also, the formation of norreticuline was not observed with its N-demethylated derivative nordehydroreticuline. Using codeinone as the keto substrate, no conversion to codeine could be observed. Similarly, the reverse reaction with codeine as the substrate also showed no generation of codeinone. As SDR enzymes had been shown to catalyse the stereospecific reduction of tropinone in tropane alkaloid biosynthesis, we included this compound in the investigation. The formation of tropine or pseudo-tropine could not be detected by GC-MS after the incubation of tropinone with the recombinant enzyme.

Figure 7.

 Chemical structures of substrates tested for salutaridine reductase specificity. The substrates and the expected reduced products are shown.

The 16B1 protein showed the highest similarity to an SDR of unknown function from A. thaliana followed by SDRs acting in the biosynthesis of menthol from Mentha × piperita. This prompted an examination of whether it was possible that the recombinant protein accepts monoterpenes as substrates. With menthone, no conversion to neomenthol or menthol could be detected by GC-MS, even when the same extraction and assay conditions as described by Ringer et al. (2003) were used. Additionally, it was of interest to test whether the homologous proteins could catalyse the reduction of salutaridine. The open reading frame of the Arabidopsis homolog of the (−)-menthone:(+)-neomenthol reductase from A. thaliana and from a peppermint variety obtained from a local market were therefore amplified from the corresponding RNA. The deduced amino acid sequence of the Arabidopsis protein was identical to the database entry, whereas the MNR from peppermint showed an amino acid sequence identity of 75% to the (−)-menthone:(+)-neomenthol reductase and of 67% to the (−)-menthone:(−)-menthol reductase of Mentha × piperita. After overexpression and purification, both the Arabidopsis and the peppermint proteins catalysed the conversion of (−)-menthone to (+)-neomenthol (95%) and (−)-menthol (5%). However, the specific activity of the Arabidopsis protein was 20-fold lower compared with the protein from peppermint. When the same enzyme preparations were incubated with salutaridine, conversion to salutaridinol was not observed, independent of the extraction and assay conditions.


Correlation of alkaloid profiles and gene expression profiles between Papaver species detects cDNAs of benzylisoquinoline biosynthesis in P. somniferum

The aim of the present study was the isolation and characterization of new cDNAs that might be implicated in accumulation of benzylisoquinoline in Papaver species. Most of the cDNAs coding for enzymes involved in benzylisoquinoline biosynthesis have been obtained by two approaches. The first one, a classical and very targeted approach, started with the purification of the protein from the plant followed by amino acid sequencing and cDNA cloning using degenerate primers (Grothe et al., 2001; Unterlinner et al., 1999). The second one relied on the sequence homology between members of one enzyme family, such as methyltransferases or P450 mono-oxygenases (Frick and Kutchan, 1999; Huang and Kutchan, 2000). Using primers directed against conserved protein domains, a number of cDNAs coding for the examined protein family were isolated and each recombinant protein was subsequently tested for its capability to catalyse the desired reaction. Here, a broader approach was used to isolate new cDNAs. An EST library prepared from P. somniferum RNA was sequenced in order to collect a large amount of sequence information on cDNA clones. As most of the cDNAs cannot be classified with respect to their function in benzylisoquinoline biosynthesis simply by sequence homology, the expression profile was correlated to the alkaloid profile in order to obtain hints about possible biological functions of individual cDNAs. Successes have been obtained using this method in the characterization of terpene synthases, methyl ketone synthase, cytochrome P450 mono-oxygenases and methyltransferases, where the involvement in specific biosynthetic steps could be elaborated by comparison of the expression profile with the metabolite profile (Aharoni et al., 2004; Chen et al., 2003; Fridman and Pichersky, 2005; Fridman et al., 2005; Gang et al., 2002; Guterman et al., 2002; Iijima et al., 2004; Lavid et al., 2002). In these investigations, the authors used one variety or wild type accumulating low amounts of certain metabolites and compared it with another cultivar of the same plant species producing large amounts of these metabolites. The use of different cultivars of the same plant species has the advantage that differences in gene expression because of morphological characteristics are diminished, and therefore the probability of isolating the desired cDNAs is increased. This is in contrast to the approach presented in this paper, where we compared different Papaver species, which exhibit in some cases very different morphologies. Comparing only two species could therefore result in many genes that are differentially expressed and only a very minor portion of these genes may be responsible for the different alkaloid profile of the two species. This is exemplified by the comparison of P. somniferum varieties with P. pyrenaicum, where 225 cDNAs are differentially expressed, among them typical cDNAs involved in the establishment of cell differentiation such as proteins that control cell division. The comparison of P. somniferum with P. glaucum, which show rather similar morphological characteristics, yielded 133 differentially expressed genes, which obviously will not all be responsible for the capability of P. somniferum to produce morphinan alkaloids. Combination of both data sets, however, yielded 93 differentially expressed genes. Irrespective of the different habitats of the plants, these 93 cDNAs have in common that their differential expression coincides with the occurrence of morphinans in P. somniferum. Thus, by inclusion of several more Papaver species showing alkaloid profiles differing from alkaloid profile unique to P. somniferum, the number of cDNAs could be reduced to 3% of all investigated cDNAs. Although many of them will probably be responsible for processes other than benzylisoquinoline biosynthesis, it is notable that none of these cDNAs showed homology to proteins involved in general cellular functions. It is also striking that all cDNAs coding for benzylisoquinoline biosynthesis thus far identified are located in one cluster showing higher expression in P. somniferum (i.e. OMTs, SalAT and COR; Figure 3). Whereas the increased expression of the morphinan-specific cDNAs SalAT and COR can be ascribed to morphinan accumulation, the higher expression of the enzymes upstream of the central intermediate (S)-reticuline, such as 6-OMT, 4′-OMT and CYP80B3, might be attributed to the overall higher amounts of alkaloids in P. somniferum compared with the other Papaver species. The similarity in expression of the cDNAs involved in benzylisoquinoline biosynthesis suggests a common regulator of the whole pathway. However, candidate sequences with similar expression patterns for regulators, such as transcription factors, could not be detected. Nevertheless, the presence of these cDNAs in clusters with increased expression in P. somniferum shows that by comparing alkaloid profiles with gene expression profiles between distinct species, cDNAs involved in the biosynthesis of specific alkaloids can be detected, irrespective of morphological differences. As cDNAs involved in benzylisoquinoline biosynthesis are tightly clustered in this analysis, it is assumed that other cDNAs of unknown function present in that cluster could also have some function in benzylisoquinoline biosynthesis.

EST 16B1 showing increased expression in morphinan alkaloid-containing Papaver species encodes salutaridine reductase

One of the cDNAs with increased expression in P. somniferum showed the highest amino acid similarity to the protein family of short chain dehydrogenases/reductases. These enzymes have been shown to act in the biosynthetic pathways of several secondary products, such as in terpene or tropane alkaloid metabolism (Davis et al., 2005; Nakajima et al., 1999; Ringer et al., 2003). According to increased expression in morphinan alkaloid-producing plants, it was assumed that the enzyme encoded by 16B1 might catalyse either the reduction of the 1,2-dehydroreticulinium ion to (R)-reticuline or of salutaridine to salutaridinol. Incubation of the recombinant enzyme with salutaridine as substrate and subsequent identification of the product by LC-MS/MS clearly showed that the enzyme catalyses the reduction of salutaridine to salutaridinol. Only the morphine precursor salutaridinol, not its stereoisomer 7-epi-salutaridinol, was formed, showing the high product specificity of the enzyme. Furthermore, other structurally similar benzylisoquinolines were not accepted as substrates, suggesting that the enzyme is specific for salutaridine. An enzyme reducing salutaridine to salutaridinol has been previously partially purified and characterized from P. somniferum (Gerardy and Zenk, 1993). The recombinant enzyme characterized in this study exhibits the same temperature optimum and pH optimum as the protein purified from the plant. Furthermore, the Km values for the substrate salutaridine and the cofactor NADPH are similar. Both enzymes are able to catalyse the reverse reaction from salutaridinol to salutaridine equally well, with the same pH optima at basic conditions and with similar affinities for salutaridinol. Only the Km value for the cofactor NADP is about four times higher for the recombinant protein compared to the purified native protein from P. somniferum. The size of the protein from P. somniferum was 51 kDa as judged from the elution on size exclusion chromatography columns (Gerardy and Zenk, 1993). Based on the amino acid sequence, the recombinant protein has a molecular weight of 34.1 kDa. However, on size exclusion chromatography, the recombinant protein elutes as a protein of 50 kDa, which is in accordance to the purified, native protein. The discrepancy between the calculated and the measured molecular weight for the recombinant protein is most likely attributable to the shape of the protein. The substrate as well as the product specificity of the recombinant protein and the corroboration with the previously purified, native protein from P. somniferum strongly suggests that the cDNA 16B1 codes for the enzyme SalR (EC This is further confirmed by its higher expression in plants that are able to accumulate benzylisoquinolines of the morphinan type. The stereospecific reduction of the keto group at position 7 of salutaridine to salutaridinol is the prerequisite for acetylation by SalAT, which does not accept 7-epi-salutarinol as a substrate (Lenz and Zenk, 1995a). The subsequent elimination of the acetyl group concomitant with the formation of an oxide bridge leads to thebaine, the first compound in the biosynthesis of morphine exhibiting the pentacyclic ring system. Thus, the high product stereospecificity of the recombinant enzyme fortifies the identification of the cDNA 16B1 as SalR from P. somniferum as part of the biosynthesis of morphinan alkaloids.

Salutaridine reductase belongs to the reductases of the short chain dehydrogenase/reductase family

Together with SalAT and COR, the cDNA from a third enzyme specific for morphine biosynthesis has been isolated. In contrast to the previously cloned COR, performing a similar reaction by reducing a keto group in an NADPH-dependent manner, and which belongs to the family of the aldo-keto reductases, SalR belongs to the SDR family of reductases. This is indicated by the TGxxxGhG (h for hydrophobic residues) motif starting at position 18, which has a structural role in coenzyme binding, the GxhDhhhNNAG motif at position 90 possessing a structural role in stabilizing the central β-sheet, and the YxxxK motif, which is essential for the catalysis as part of the proton transfer from the cofactor to the substrate (Filling et al., 2002; Jörnvall et al., 1995; Persson et al., 2003). Two of these motifs are present in the sequence, at positions 153 and 236 (Figure 5). The presence of serine residues at position 179, however, makes the YxxxK motif at position 236 more likely, as it has been shown that a serine residue upstream of the YxxxK motif is necessary for catalytic activity (Filling et al., 2002). The preference of SalR for NADP(H) over NAD(H) can also be deduced from the amino acid sequence by the presence of the basic residues R or K downstream of the TGxxxGhG motif at position 44 or 48, respectively, instead of acidic residues (Persson et al., 2003). Based on these amino acid motifs, SalR belongs to the classical SDR and can be grouped into the family cP2 according to Persson et al. (2003). Whereas the classical SDRs are around 250 amino acids in length, the open reading frame of SalR codes for 311 amino acids. Additionally, functional units of classical SDRs are either homotetramers or homodimers, such as the tropinone reductases (Nakajima et al., 1993). As estimated by gel permeation chromatography, SalR is active as a monomeric protein. A monomeric structure has also been shown for SDRs in terpene metabolism, such as ISPR, MNR and MMR, which exhibit the highest amino acid sequence similarity to SalR of all hitherto characterized SDRs (Davis et al., 2005; Ringer et al., 2003). The monomeric structure is ascribed to a roughly 40-residue insertion preceding the catalytic tyrosine forming an all-helix subdomain that blocks the dimerization, as shown for porcine testicular carbonyl reductase (Ghosh et al., 2001). This amino acid stretch is also responsible for the increased size over the classical multimeric SDRs.

Relationship of SalR to other reductases in plant secondary metabolism

Phylogenetic analysis reveals that SalR forms one cluster together with monomeric SDRs, such as ISPR, which reduces the ring double bond in menthol biosynthesis, as well as MNR and MMR, which reduce the keto group of (−)-menthone to (+)-neomenthol or (−)-menthol, respectively (Figure 8). Salutaridine reductase does not accept (−)-menthone as a substrate. Conversely, the MNR cloned in this study, showing 75% identity to the MNR cloned from Mentha × piperita, does not reduce salutaridine, showing the high substrate specificity of these enzymes. Interestingly, there are two additional enzymes in menthol biosynthesis catalysing reductions or dehydrogenations with low sequence similarity to this cluster. (+)-Pulegone reductase, reducing the double bond of the side chain, belongs to the family of the medium chain dehydrogenases/reductases (Ringer et al., 2003), whereas (−)-trans-isopiperitenol dehydrogenase (ISPD), catalysing the dehydrogenation from (−)-trans-isopiperitenol to (−)-isopiperitenone, is classified into the classical SDRs with a homodimeric functional unit and a subunit size of about 250 amino acids (Ringer et al., 2005). The latter enzyme, therefore, resembles tropinone reductase and secoisolariciresinol dehydrogenase, an SDR in podophyllotoxin biosynthesis (Xia et al., 2001). These enzymes form a cluster independent from the monomeric SDRs ISPR, MMR, MNR and SalR. The latter two enzymes are not able to convert tropinone to tropine. Furthermore, SalR does not catalyse a similar reaction in morphine biosynthesis, the reduction of codeinone to codeine. This step is performed by COR, which belongs to the family of the aldo-keto reductases and is more related to chalcone reductase, an enzyme of flavonoid biosynthesis (Bomati et al., 2005; Unterlinner et al., 1999). Conversely, COR does not accept salutaridine as a substrate (Unterlinner et al., 1999). Taken together, it is evident that reductions or dehydrogenations in plant secondary metabolism can be performed by enzymes of high substrate specificity that can be classified into several families. This is not surprising, taking into account the diversity of compounds that the reductase superfamily can accept as substrates (Jörnvall et al., 1999). However, as in monoterpene metabolism, where rather similar reductive steps in the biosynthesis of menthol are performed by enzymes belonging to different reductase families, similar reductions with respect to substrate and mechanism in benzylisoquinoline metabolism are catalysed by enzymes belonging to at least two reductase families, the SDR and aldo-keto reductase families. As already discussed by Ringer et al. (2003) for monoterpene metabolism, it is obvious also for benzylisoquinoline biosynthesis that sequential reductase steps did not necessarily arise by gene duplication and diversification, but rather emerged from different ancestral sources.

Figure 8.

 Neighbour-joining phylogenetic tree of selected plant proteins catalysing reductions in plant secondary metabolism.
The amino acid sequence alignments were performed using the clustalw application (Thompson et al., 1994) of MegAlign (DNASTAR Inc), and the tree was generated and visualized with the treecon (Yves van de Peer, University of Konstanz, Germany) software. Bootstrap values in per cent of 1000 trials are indicated. The source and accession numbers are: At3g61220 SDR from A. thaliana with low (−)-menthone:(+)-neomenthol reductase activity, accession NM115986; PsSalR (salutaridine reductase) from P. somniferum, DQ31621; MpISPR [(−)-isopiperitenone reductase], AY300162; MpMNR [(−)-menthone:(+)-neomenthol reductase], AAQ55959; MpMMR [(−)-menthone:(−)-menthol reductase], AAQ55960; MpPulR [(+)-pulegone reductase], AY300163; MpISPD [(−)-isopiperitenol dehydrogenase], AY641428; all from Mentha × piperita; HnTR-I+HnTRII (tropinone reductase I and II) from Hyoscyamus niger, D88156 and L20485; SDH_Fi321 (secoisolariciresinol dehydrogenase) from Forsythia × intermedia, AAK38665; PsCOR1 (codeinone reductase 1) from P. somniferum, AF108432; GmCHR (chalcone reductase) from Glycine max, X55730.

Experimental procedures

Plant material

The seeds of all the Papaver plants were obtained either from the seed stock collection of the Department of Natural Product Biotechnology of the Leibniz Institute of Plant Biochemistry, Halle, Germany or were the kind gift of Tasmanian Alkaloids Pty Ltd, Westbury, Tasmania, Australia. Papaver somniferum L. plants were grown outdoors in Saxony-Anhalt. Peppermint (Mentha) plants were obtained from a local market. Arabidopsis thaliana ecotype Columbia were cultivated in controlled chambers (Percival, CLF, Perry, IA, USA) at 70% relative humidity under short-day conditions of 8-h light, 210 μE m−2 s−1.

Sequencing project

Purified mRNA isolated from P. somniferum stems was reverse transcribed and used for construction of a directionally cloned cDNA library following the instructions for the Superscript Lambda System for cDNA synthesis and λ cloning (Gibco, Karlsrahe, Germany). The generated λZipLox cDNA library was subjected to library mass excision resulting in cDNAs directionally cloned into the plasmid pZL1 propagated in the E. coli strain DH10B. Plasmids purified from randomly chosen bacterial colonies were digested with PstI and HindIII and the insert length was monitored by agarose gel electrophoresis. All plasmids showing inserts at least 500-bp length were sequenced using the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Darmstadt, Germany) with the T7 primer. Sequencing reactions were run after removal of excess dye on the ABI Prism 310 Sequencer. After trimming for vector, the sequences were clustered and assembled using the SeqManII (DNASTAR Inc., Madison, WI, USA) application using a minimum match size of 20 and a minimum match percentage of 85.

Macroarray preparation, hybridization and evaluation

Complementary DNAs for spotting were amplified by PCR using vector-specific primers adjacent to the multiple cloning site, purified by filtration through NucleoFast 96 PCR-plates (Macherey-Nage, Düren, Germany), and concentrated to at least 100 ng μl−1. Complementary DNAs were spotted in quadruplicate on Biodyne B membranes (PALL Corporation, Pensada, FL, USA) using the Microgrid II spotter (BioRobotics, Cambridge, UK). After spotting, the cDNAs were denatured by incubating the membranes on 0.5 n NaOH/1.5 m NaCl followed by neutralization on 1 m Tris–HCl pH 7.5/1.5 m NaCl. The DNA was immobilized via UV crosslinking at 120 mJ cm−1 using a Stratalinker 1800 (Stratagene, Amsterdam, the Netherlands).

Total RNA was taken as a template for reverse transcription using oligo-dT primers. After purification by gel filtration through ProbeQuant G-50 microcolumns (Amersham Biosciences, Freiburg, Germany) the cDNAs were labelled with [α33P]-dATP using the Megaprime DNA Labelling System (Amersham Biosciences). The membranes were incubated in pre-hybridization solution (5 × SSC, 0.1% SDS, 5 × Dehnhardt solution, 125 μg ml−1 denatured salmon sperm DNA) for 4 h at 65°C and overnight hybridization at 65°C was performed by addition of the labelled cDNAs. The filters were washed three times for 15 min with 2 × SSC, 0.1% SDS at 65°C and exposed to storage phosphor screens. The signals were recorded with a STORM 860 Gel and Blot Imaging System (Amersham Biosciences). Spot intensities and background subtraction were determined using the aida array software (Raytest, Straubenhardt, Germany). The background subtracted data of the spot intensities for each experiment were exported in the Scientific Graphing and Analysis Software package origin v7.5 which permits automated handling of the data analysis and visualization by programing. The analysis is similar to the one described previously (Sreenivasulu et al., 2004), emphasizing different expressions irrespective of absolute signal intensities. However, the median centring procedures extend to all four replicate spots, i.e. the averaging is carried out after this procedure in order to prove the statistical significance of the resulting signal intensities. The median centring of arrays (Eisen et al., 1998) sets the median of the logarithmically scaled (log2) intensity distribution of all genes within an experiment to zero allowing a comparison between distinct experiments. This is followed by a median centring of these array-centred data for each gene (inclusive replicates) across the experiments. In this way, the previous centring is slightly displaced again. Thus, both median centrings are iteratively repeated until the displacement is reduced to a value of <0.001, which was established after six iterations in this work. In order to prove the significance of the different expressions (x) the averaging were performed as last step, i.e. after the iterative median centring procedure. The results differ only marginally compared with the initial averaging. Thus, the significance (t-test) of each gene expression in all experiments can be determined. Generally, the occurrence of gene expressions with maximum twofold, fourfold, eightfold or higher overexpression/underexpression obeys an exponential decaying function. In our analysis the corresponding relative portions are 84.9%, 12.6%, 2.3% and 0.29% of all gene expressions, respectively (Figure S1). Of course, the relative fraction of significant gene expression depends on the elected significance level (α). At α = 0.05 this fraction increased from 62% for expressions nearby zero (|x| ≤ 1) to ≈ 95% at higher expressions (|x| > 1). The gene expressions of the cluster shown in Figure 3 reflect the general result with 76.6% significant gene expressions. Complete linkage hierarchical clustering with an uncentred Pearson's correlation metric was performed using the program cluster, version 3.0 (developed by Michael Eisen, Stanford University and Michiel de Hoon, University of Tokyo). The treeview program (version 1.0.10) was used for visualization of the clustering results.

Analysis of alkaloids

Plant material was ground with a mortar and pestle under liquid nitrogen to a fine powder and the alkaloids were extracted with 80% (v/v) ethanol at 4°C for 30 min. After centrifugation at 14 000g for 10 min and evaporation of the solvent, the residue was taken up in methanol.

An HPLC analysis was performed using an LC 1100 series Agilent system (Agilent, Waldbronn, Germany) equipped with a Lichrospher 60 RP-select B column (250 × 4 mm, 5 μm; Merck, Darmstadt, Germany) and a solvent system consisting of A:CH3CN-H2O (2:98; v.v), B:CH3CN-H2O (98:2; v.v) each with 0.001% (v.v) phosphoric acid. The gradient was from 0% B to 60% B in 25 min with a hold for 5 min followed by an increase to 100% B in 2 min and a hold for 3 min at a flow rate of 1 ml min−1. The detection wavelength was set to 210 nm. For the analysis of the enzyme assays containing salutaridine or salutaridinol, the gradient was from 25% B to 30% B in 6 min.

The ESI-MS measurements and LC separations were carried out on a Mariner time-of-flight (TOF) mass spectrometer (Applied Biosystems) equipped with a Turbulon Spray source (PE-Sciex, Concorde, ON, Canada) using an LC1100 series Agilent system adapted to flow rates of 0.2 ml min−1. Samples were injected on a Superspher 60 RP-Select B column (125 × 2 mm, 5 μm; Merck). The following LC conditions were used: solvent A:CH3CN-H2O (2:98; v.v) and solvent B:CH3CN-H2O (98:2; v.v), 0.2% (v/v) formic acid in both solvents. The gradient increased from 0% to 46% B in 25 min, to 90% in 1 min and was held at 90% for 7 min, post-time was 5 min. The TOF mass spectrometer was operated in the positive ion mode, with nebulizer gas (N2) flow of 0.5 l min−1, curtain gas (N2) flow of 1.5 l min−1 and heater gas flow of (N2) 7 l min−1. The spray tip potential of the ion source was 5.5 kV, the heater and quadrupole temperatures were 320°C and 140°C, respectively, the nozzle potential was set to 200 V, and the detector voltage to 1.56 kV. The other settings varied depending on tuning.

The high-resolution positive ion ESI mass spectra of the Papaver extracts were obtained from a Bruker Apex III Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Bruker Daltonics, Billerica, MA, USA) equipped with an InfinityTM cell, a 7.0 T superconducting magnet (Bruker, Karlsruhe, Germany), an RF-only hexapole ion guide and an external electrospray ion source (Agilent, off-axis spray; voltages: endplate, −3.700 V; capillary, −4.200 V; capillary exit, 100 V; skimmer 1, 15.0 V; skimmer 2, 10.0 V). Nitrogen was used as the drying gas at 150°C. The sample solutions were introduced continuously via a syringe pump with a flow rate of 120 μl h−1. All data were acquired with 512 k data points and zero filled to 2048 k by averaging 32 scans.

The LC-ESI-MS and LC-ESI-MS/MS spectra of the [M+H]+ ions of alkaloids were obtained from a Finnigan MAT TSQ 7000 system (Finnegan, Waltham, MA, USA; electrospray voltage 4.5 kV; sheath gas nitrogen; capillary temperature 220°C; collision gas argon; collision pressure approximately 1.8 × 10−3 Torr; collision energy depending on structure). The MS system is coupled with a Surveyor micro-HPLC (ThermoFinnigan) and equipped with an Ultrasep ES RP18E-column (5 μm, 1 × 100 mm; SepServ, Berlin, Germany). For the HPLC a gradient system was used starting from H2O:CH3CN 85:15 [each of them containing 0.2% (v.v) acetic acid] to 10:90 within 15 min; flow rate 50 μl min−1. All mass spectra are averaged and background subtracted.

GC-MS analysis

The samples from the menthone and tropinone reductase assays were analysed on a Finnigan TraceGC gas chromatograph coupled to a Finnigan Polaris Q mass spectrometer. Separation was performed on a RTx-5w/IntegraGuard column (Restek, Bad Homburg, Germany) of 15 m × 0.25 mm internal diameter, coated with crossbond-5% diphenyl-95% dimethyl polysiloxane (0.25 μm film thickness). Helium was the carrier flow gas (flow rate 1 ml min−1) and a splitless injection (injection volume of 1 μl, injection temperature 220°C) was used. Temperature gradients were 40°C (1 min hold), 2°C min−1 to 80°C followed by 100°C min−1 to 150°C, or 40°C (1 min hold), 3°C min−1 to 100°C followed by 100°C min−1 to 150°C for menthol or tropinone samples, respectively. Mass spectrometry was performed with an interface temperature of 300°C, an ion source temperature of 200°C and an ionization potential of 70 eV. The compounds were identified by their EI spectra and comparison with a NIST database or compared with authentic standards obtained from Fluka, Taufkirchen, Germany.

Generation of full-length cDNA of 16B1

The missing 5′-end of cDNA clone 16B1 was amplified from P. somniferum stem mRNA according to the SMARTTM cDNA amplification kit (Clontech, Heidelberg, Germany) using the following gene-specific primers: 16B1r1 5′-ACCTTCTTCGGCAG-3′ for reverse transcription, and 16B1r2 5′-CAGCTCCGAAACTAGGCCACCCATTTGTTTC GATC-3′ for PCR (30 sec 94°C, 10 cycles 30 sec 94°C, 30 sec 70°C, 2 min 72°C with decreasing annealing temperature by 1°C per cycle, followed by 25 cycles 30 sec 94°C, 30 sec 60°C, 2 min 72°C and a final elongation for 10 min at 72°C). The resulting band of 900 bp was ligated into pGEMTeasy (Promega, Mannheim, Germany) and sequenced. The entire open reading frame for overexpression was obtained by RT-PCR using PfuUltraTM Hotstart DNA Polymerase (Stratagene) and the primers 16B1–5-2: 5′-ATGCCTGAAACATGTCC-3′ and 16B1–3-1: 5′-ATAACTCAAAATGCAGATAGTTCTG-3′. The PCR conditions were as described above at a constant annealing temperature of 54°C. The resulting fragment of approximately 1000 bp was ligated into pBluescriptIISK and sequenced.

Heterologous expression and enzyme purification

The open reading frame of the cDNA 16B1 was excised from the pBluescriptIISK vector with KpnI and PstI and ligated into the expression vector pQE-30 (Qiagen, Hilden, Germany). For overexpression, the recombinant plasmid was transformed into E. coli SG13009. Single isolated bacterial colonies from freshly streaked plates [grown on Luria–Bertani (LB) agar medium containing 50 μg ml−1 ampicillin and 50 μg ml−1 kanamycin] were used to inoculate 2 ml liquid cultures (LB medium containing 50 μg ml−1 ampicillin and 50 μg ml−1 kanamycin), which were grown overnight at 37°C. An aliquot (100 μl) of these cultures was used to inoculate 50 ml liquid cultures. At a cell density of 0.5 OD600, recombinant protein expression was induced with 1 mm isopropyl-β-d-thiogalactopyranoside (IPTG). After incubation at 37°C for 4 h, cells were harvested and sonicated in extraction buffer [20 mm Tris–HCl pH 7.5, 100 mm KCl, 10% (v.v) glycerol, 20 mmβ-mercaptoethanol, 10 μg ml−1 lysozyme]. After removal of the cell debris by centrifugation, the supernatant was loaded onto a cobalt affinity column (Talon, Clontech), followed by washing with extraction buffer without lysozyme. The recombinant protein was eluted by a stepwise increase in the imidazole concentration from 10 to 30 mm.

Cloning and expression of a peppermint cDNA encoding (−)-menthone reductase and of the cDNA of Arabidopsis gene At3g61220

The mRNA from A. thaliana and from peppermint leaves was reverse transcribed with an oligo-dT primer. The open reading frame of (−)-menthone reductase was amplified from peppermint leaf cDNA using primers MMNR-5–1: 5′-CACCATGGGAGATGAAGTAGTCGTC-3′ and MMNR-3–1: 5′-ATACAAGCAGAACGCTTCGTCTC-3′ as deduced from the MpMNR sequence (accession number AY288137). The open reading from the Arabidopsis homologue of 16B1 was amplified by RT-PCR with primers: Ath16B1–5-1 5′-CACCATGGCAGAGGAAACTCC-3′ and Ath16B1–3-1: 5′-GAATTC TGAAACTTGCTTGCGAC-3′ as deduced from sequence deposited in the database under the accession number AY091305. Both cDNAs were amplified using PfuUltraTM Hotstart DNA Polymerase (Stratagene) under the following PCR conditions: 30 sec 94°C, 25 cycles 30 sec 94°C, 30 sec 55°C, 2 min 72°C and a final elongation for 10 min at 72°C. The resulting fragments were blunt end ligated into pQE30, and the recombinant plasmids were transformed into E. coli SG13009. Expression of the cDNAs was performed as described for P. somniferum 16B1. The cells were harvested and sonicated in extraction buffer [50 mm 3-(N-morpholino)-2-hydroxypropanesulphonic acid (MOPSO) pH 7, 10% (v.v) glycerol, 10% (w/v) sorbitol, 10 mmβ-mercaptoethanol, 10 μg ml−1 lysozyme]. For purification of the recombinant proteins, the bacterial crude extract was adjusted to an imidazole concentration of 10 mm and loaded onto spin columns from the ProPur Mini Kit (Nunc, Roskilde, Germany). Purification of the proteins was achieved by an increase in imidazole concentration from 50 to 100 mm.

Enzyme assays and enzyme characterization

The enzyme assay reaction mixture for the reduction of substrates consisted of 150 mm potassium phosphate pH 6, 500 μm NADPH, 100 μm salutaridine, codeinone or tropinone, and up to 100 μl protein extract in a total volume of 200 μl. For the reverse reaction, 150 mm glycine buffer pH 9.5, 500 μm NADP and 100 μm salutaridinol or codeine were used. The enzyme assays for the reduction of the 1,2-dehydroreticulinium ion and 1,2-nordehydroreticuline were performed in 300 mm glycine buffer pH 8.5 with 100 μm substrate. After incubation at 30°C, the enzymatic reaction was terminated by the addition of 200 μl 1 m NaHCO3 and products extracted with 500 μl ethylacetate. The organic phase evaporated, and was taken up in methanol and analysed as described in the analysis of alkaloid section. For enzyme assays containing tropinone, the compounds were extracted with 1 ml of chloroform. The (−)-menthone reductase assays contained 40 mm potassium phosphate buffer pH 7, 10 mmβ-mercaptoethanol, 100 μm (−)-menthone, 500 μm NADPH and up to 150 μl protein extract in a total volume of 500 μl. After incubation at 30°C, the products were extracted with 1 ml hexane. The reaction products of the incubations with (−)-menthone and tropinone were analysed by GC-MS.

Saturation curves and double reciprocal plots were constructed with the fig.p program version 2.98 (Biosoft, Cambridge, UK). The influence of pH on enzyme activity was monitored in sodium citrate (pH 4–6), potassium phosphate (pH 6–8), Tris–HCl (pH 7–9), Tricine-NaOH (pH 8.5–9.5) and glycine/NaOH (pH 9–12) buffered solutions. The subunit molecular mass was determined by SDS–PAGE (12% polyacrylamide) according to Laemmli (1970). The native molecular mass was estimated by gel filtration chromatography through a Superdex 75Hiload Prep 16/60 column (Amersham Biosciences) with bovine serum albumin (67 kDa), ovalbumin (45 kDa) and α-chymotrypsinogen (25 kDa) as standards. Protein concentrations were measured according to Bradford (1976).

RNA gel blot analysis

Total RNA was isolated with TRIzol according to the manufacturer's protocol (Invitrogen, Karlsrahe, Germany). Northern blot analyses were performed as described by Ausubel (1987). The 16B1 probes were produced by amplifying fragments with the primers 16B1–5-2 and 16B1–3-1 the full-length cDNA from P. somniferum or the first strand cDNA from P. bracteatum and labelling with [α32P]dATP. Hybridization conditions were as described for macroarray hybridization.


We thank Silvia Wegener and Christine Kuhnt for excellent technical assistance and gratefully acknowledge Prof. M. H. Zenk and Dr C. Boettcher, Halle, for providing tetrahydrobenzylisoquinoline alkaloids. This work was supported by the Deutsche Forschungsgemeinschaft, Bonn, SPP 1152, Priority Program ‘Evolution of Metabolic Diversity’.

Accession numbers
Sequence data from this article have been deposited with the EMBL/GenBank data libraries under the accession numbers DQ31621 for P. somniferum SalR and DQ362936 for the Mentha cDNA with (−)-menthone:(+)-neomenthol activity. The sequences from the EST sequencing project can be accessed via the homepage of The Floral Genome Project (