A Papaver somniferum 10-Gene Cluster for Synthesis of the Anticancer Alkaloid Noscapine

See allHide authors and affiliations

Science  29 Jun 2012:
Vol. 336, Issue 6089, pp. 1704-1708
DOI: 10.1126/science.1220757

Alkaloid Synthetic Pathway

Noscapine, a nonaddictive alkaloid found in the opium poppy, can be used as a cough suppressant and a tubulin-binding antitumor agent. Winzer et al. (p. 1704, published online 31 May; see the Perspective by DellaPenna and O'Connor) found that a cluster of 10 genes were key to the production of noscapine. Poppies homozygous for this gene cluster produced high levels of noscapine, heterozygous poppies produced low levels of noscapine, and those poppies lacking the gene cluster produced no noscapine. Silencing individual genes in turn and analyzing the accumulation of intermediate metabolites allowed the biosynthetic pathway of noscapine to be elucidated.


Noscapine is an antitumor alkaloid from opium poppy that binds tubulin, arrests metaphase, and induces apoptosis in dividing human cells. Elucidation of the biosynthetic pathway will enable improvement in the commercial production of noscapine and related bioactive molecules. Transcriptomic analysis revealed the exclusive expression of 10 genes encoding five distinct enzyme classes in a high noscapine–producing poppy variety, HN1. Analysis of an F2 mapping population indicated that these genes are tightly linked in HN1, and bacterial artificial chromosome sequencing confirmed that they exist as a complex gene cluster for plant alkaloids. Virus-induced gene silencing resulted in accumulation of pathway intermediates, allowing gene function to be linked to noscapine synthesis and a novel biosynthetic pathway to be proposed.

Noscapine was first characterized from opium poppy, Papaver somniferum, by the distinguished French chemist Pierre Jean Robiquet in 1817 (1) but unlike codeine (which he also discovered) and several other opiates, noscapine is neither painkilling nor addictive. Noscapine has been used as a human cough suppressant for decades; its effect on the cough reflex and bronchial muscles was reported as early as 1954 (2), and the bioavailability and pharmacokinetics of its orally administered form were established in 1990 (3). The demonstration in 1998 that noscapine acts as a potent antitumor agent that binds to tubulin and affects its polymerization, thereby arresting cell division and inducing apoptosis (4), was followed by confirmation of its antitumor activity in various forms of cancer (58). Because of its history of safe use as an antitussive, rapid absorption after oral administration, and apoptosis-inducing effect on a number of cancer cell lines, noscapine may have an advantage over other tubulin-binding anticancer natural products such as the well-established taxanes (9).

Noscapine belongs to the phthalideisoquinoline subclass of the structurally diverse isoquinoline alkaloids, whereas codeine, morphine, thebaine, and oripavine belong to the morphinan subclass (10). Although the biosynthesis of morphinans has been elucidated at the molecular level (1116), our knowledge of noscapine biosynthesis has not substantially advanced since the 1960s, when isotope-labeling experiments showed that it is derived from scoulerine (17). Understanding the biochemical genetics that underlie noscapine biosynthesis should enable improved production of this important pharmaceutical and related molecules in both poppy and other systems.

We created metabolite profiles of capsule extracts from three opium poppy varieties developed in Tasmania for alkaloid production, which we designate as high morphine 1 (HM1), high thebaine 1 (HT1), and high noscapine 1 (HN1) on the basis of the most abundant alkaloid in each case (Fig. 1A). Noscapine as well as 53 other low-abundance compounds, some of which are candidate intermediates in the noscapine biosynthetic pathway, are produced in HN1; none of these compounds appear in either HM1 or HT1 (table S1). We performed Roche 454 pyrosequencing on complementary DNA libraries derived from stem and capsule tissue from all three varieties. Analysis of expressed sequence tag (EST) abundance led to the discovery of a number of previously uncharacterized genes that are expressed in the HN1 variety but are completely absent from the HM1 and HT1 EST libraries (Fig. 1B). We putatively identified the corresponding genes as three O-methyltransferases (PSMT1, PSMT2, PSMT3; fig. S1), four cytochrome P450s (CYP82X1, CYP82X2, CYP82Y1, and CYP719A21; fig. S2), an acetyltransferase (PSAT1; fig. S3), a carboxylesterase (PSCXE1; fig. S4) and a short-chain dehydrogenase/reductase (PSSDR1; fig. S5). In contrast, a number of other functionally characterized genes associated with benzylisoquinoline alkaloid synthesis—including Berberine Bridge Enzyme (BBE), Tetrahydroprotoberberine cis-N-MethylTransferase (TNMT), Salutaridine Reductase (SalR), Salutaridinol 7-O-AcetylTransferase (SalAT), and Thebaine 6-O-demethylase (T6ODM)—were expressed in all three varieties (Fig. 1B). Polymerase chain reaction (PCR) analysis revealed that the genes exclusively expressed in HN1 are present as expected in the genome of HN1 but are absent from the genomes of HM1 and HT1 (Fig. 1B and fig. S6).

Fig. 1

Identification of genes exclusively present in the genome of a noscapine-producing poppy variety, HN1. (A) Relative abundance of the major alkaloids extracted from the capsules of three commercial varieties of poppy: HM1, HT1, and HN1. M, morphine; C, codeine; T, thebaine; O, oripavine; N, noscapine; rt, retention time in seconds. (B) EST libraries from stem and capsule were generated by pyrosequencing and unique contiguous sequences assembled as described (18). The first 10 genes are represented only in EST libraries from the HN1 variety; the last five genes are present in EST libraries from all three varieties. All genes are represented at consistently higher levels in stem than in capsule, as shown in color code. PCR on genomic DNA from all three varieties revealed that the 10 HN1-specific genes are absent from the genomes of the HM1 and HT1 varieties, whereas the five other functionally characterized genes are present in all three varieties (fig. S6).

Using HN1 and HM1 as parents, we generated an F2 mapping population of 271 individuals. Genotyping of the field-grown F2 population revealed that the HN1-specific genes are tightly linked and associated with the presence of noscapine, which suggests that they occur as a gene cluster involved in noscapine biosynthesis (Fig. 2). Analysis of noscapine levels in field-grown F2 capsules revealed that individuals containing this putative gene cluster fell into two classes. The first, containing 150 individuals, had relatively low levels of noscapine; the second, containing 63 individuals, exhibited the high-noscapine trait of the parental HN1 variety. The 58 F2 individuals that lacked the putative gene cluster contained undetectable levels of noscapine (Fig. 2B). F3 family analysis confirmed that F2 individuals exhibiting the high-noscapine trait were homozygous for the gene cluster, whereas those exhibiting the low-noscapine trait were heterozygous (table S2). Noscapine levels in both the glasshouse-grown F1 population and the heterozygous F2 class were much lower than the intermediate levels expected for a semidominant trait, suggesting the involvement of some form of repression. A number of other HN1-specific metabolites were similarly decreased (table S3), suggesting global down-regulation of this branch of alkaloid metabolism when the gene cluster is in the heterozygous state. However, quantitative reverse transcription PCR analysis did not detect any transcriptional repression in the heterozygous state on PSMT1 and PSMT2, which we determine below to encode the first step and an intermediate step, respectively, in the noscapine pathway (fig. S7).

Fig. 2

Segregation analysis of noscapine content in an F2 mapping population demonstrates requirement for the noscapine gene cluster. (A) Box plot depiction of noscapine levels as percentage dry weight (DW) in glasshouse-grown parental lines HN1 and HM1 and the F1 generation. (B) The F2 generation segregates into three classes of zero, low, and high noscapine (N); F2 GC– and F2 GC+ indicate the absence and presence, respectively, of the noscapine gene cluster. Numbers in parentheses indicate number of individuals in each class. (C) Total major morphinans as percentage capsule dry weight in the F2 classes of zero, low, and high noscapine (T, thebaine; O, oripavine; C, codeine; M, morphine).

The morphinan branch of alkaloid metabolism, on the other hand, remained largely unaffected in F1 (table S4) and heterozygous F2 material, showing only a decrease in capsules producing high levels of noscapine, presumably due to substrate competition (Fig. 2C). The large step change to high noscapine in the homozygous F2 class suggests that this trait is linked to the gene cluster locus rather than spread quantitatively among other loci.

To further characterize the putative noscapine gene cluster, we prepared a bacterial artificial chromosome (BAC) library from genomic DNA isolated from HN1 and identified six overlapping BACs containing genes from the cluster. Next-generation and Sanger sequencing were used to generate a high-quality assembly of 401 kb, confirming the arrangement of the 10 genes in a cluster spanning 221 kb (Fig. 3 and table S5) (18). The homology and intron-exon structure of the CYP82 and PSMT genes suggest tandem gene duplication after genome reorganization of the progenitor genes. A similar case can be made for PSCXE1 because a second homolog, PSCXE2, is present in the region flanking the gene cluster (Fig. 3 and fig. S4), but this gene is not represented in any of our EST libraries. However, on only one occasion are members of the same gene family adjacent in the cluster, when CYP82X1 and CYP82X2 are inverted with respect to each other. CYP82Y1 is separated from CYP82X2 by a gap of 45 kb containing PSAT1 and PSMT2. PSMT3 and PSMT2 are separated by a gap of 73 kb containing CYP82Y1. Interspersed among the 10 genes are both retrotransposon and DNA transposable element (TE) sequences (Fig. 3 and table S5), which may have some function in gene rearrangement for cluster formation, as is thought to be the case for the thalianol and marneral clusters from Arabidopsis thaliana (19). A search of the PLACE database of plant cis-acting DNA elements (20) revealed a number of short motifs (four or five bases) present in the 1-kb predicted promoter regions upstream of the open reading frames (ORFs) of the 10 genes, among which the WRKY elements are noteworthy (21, 22) (table S6).

Fig. 3

The HN1 gene cluster. The structure and position of the 10 HN1-specific genes expressed in stems and capsule tissues are shown above the central black line, which represents 401 kb of genomic sequence. Exons are represented by solid gray boxes and introns by fine black lines. Arrows indicate the 5′ to 3′ orientation of each gene. Additional ORFs depicted below the central black line are as defined by the key. None of these ORFs are represented in the stem and capsule EST libraries. The location and annotation of all ORFs in the 401-kb sequence are detailed in table S5.

To functionally characterize the genes in the HN1 cluster, we performed virus-induced gene silencing (VIGS) on poppy seedlings by established methods (18). VIGS in poppy seedlings persists through to mature plant stages (23), and we therefore routinely assayed both leaf latex and capsule extract (Fig. 4). We managed to silence six of the eight genes we tested by VIGS, as determined by mRNA abundance in infected leaf tissue (fig. S8). Silencing PSMT1, which shows high homology with scoulerine-9-O-methyltransferase from Coptis japonica (24) (fig. S1A), resulted in accumulation of scoulerine in both latex and capsules, as well as low levels of reticuline in latex (Fig. 4A). We expressed the PSMT1 gene product in Saccharomyces cerevisiae and, consistent with the VIGS data, this converted scoulerine to tetrahydrocolumbamine at high efficiency (fig. S9). We therefore conclude that PSMT1 is responsible for the first committed step in the pathway to noscapine synthesis.

Fig. 4

Functional characterization using virus-induced gene silencing of six genes from the HN1 gene cluster. (A to F) Results from both leaf latex (left panels) and capsules (right panels) are consistent with each of these genes encoding enzymes involved in noscapine biosynthesis. All compounds that accumulate, apart from scoulerine, have been putatively identified on the basis of mass spectra (fig. S10). The mass-to-charge (m/z) value (M) followed by retention time in seconds (T) is shown for each compound on the horizontal axis. Metabolites showing a positive change by more than a factor of 2 (VIGS versus controls) and >0.05% total alkaloid profile are shown as percentage total metabolites. A complete list of all metabolites that were significantly changed in the VIGS experiments is shown in table S7. (G) Proposed pathway for noscapine biosynthesis based on VIGS data. Solid arrows depict steps supported by VIGS data; dashed arrows depict additional proposed steps. Italics depict those reactions that await gene assignment. The noscapine structure is numbered according to the IUPAC convention. Labeling of the 3-OH secoberbine intermediates is based on the numbering of the noscapine structure. For the secoberbine intermediates that accumulate in VIGS experiments, R1 = H or OH, R2 = H or OH, and R3 = CH2OH, CHO, or COOH (fig. S10). The proposed pathway assumes R1 = H, R2 = H, and R3 = CHO in the secoberbine intermediates. The crossed-out PSMT2 depicts silencing of this gene product resulting in narcotoline accumulation, as described in the text.

The product of PSMT1, tetrahydrocolumbamine, accumulated in latex and capsules that were silenced for CYP719A21, indicating that this gene is responsible for the second step in the pathway (Fig. 4B and fig. S10). CYP719A21 shows high homology to cytochrome P450 oxidases that act as methylenedioxy bridge-forming enzymes, and we therefore propose that CYP719A21 encodes a canadine synthase (25, 26) (fig. S2B). We propose that canadine is methylated to form N-methylcanadine, which in turn is converted to secoberbine intermediates (Fig. 4G). Consistent with this, canadine and N-methylcanadine are HN1-specific metabolites (table S1). The product of TNMT has previously been shown to specifically N-methylate protoberberine alkaloids, including canadine (27). TNMT is present and expressed in HN1, HM1, and HT1 (Fig. 1B) and does not appear to be associated with the HN1 gene cluster (Fig. 3). Three other ORFs with TNMT homology are present in the flanking region of the HN1 gene cluster, but these are not expressed in stems or capsules, further implicating TNMT as having a role in the pathway.

Silencing of a second cytochrome P450 gene, CYP82X2, resulted in accumulation of several secoberbine intermediates, some of which may represent side products to the main synthetic pathway (Fig. 4C). The fragmentation pattern of intermediate 1 (fig. S10) is consistent with the compound being narcotolinol (R1 = OH, R2 = H, and R3 = CH2OH), implying CYP82X2 hydroxylates at the R2 position. The production of these secoberbine intermediates from N-methylcanadine requires breakage of the berberine bridge and ring opening, as depicted in Fig. 4G. Silencing of the carboxylesterase gene PSCXE1 resulted in accumulation of up to 20% total alkaloid content of the acetylated compound papaveroxine (Fig. 4D and fig. S10). Synthesis of papaveroxine from secoberbine intermediates requires hydroxylation and methylation at the position equivalent to the C4′ position of noscapine as well as acetylation of the hydroxyl group at the C3 position (Fig. 4G). The accumulation of papaveroxine in material silenced for PSCXE1 implies that the corresponding enzyme removes an acetyl group from this compound to produce narcotinehemiacetal. That narcotinehemiacetal is an intermediate in the pathway to noscapine synthesis is substantiated by the fact that it accumulates upon silencing of the short-chain dehydrogenase/reductase gene PSSDR1 (Fig. 4E and fig. S10). Conversion of narcotinehemiacetal to noscapine requires dehydrogenation, and we therefore conclude that PSSDR1 is involved in this final synthetic step (Fig. 4G).

The VIGS data for PSCXE1 and PSSDR1 therefore support a biosynthetic route to noscapine that involves O-methylation of secoberbine intermediates at the position equivalent to the C4′ hydroxyl group of noscapine (Fig. 4G). However, silencing PSMT2 did not show any impact on secoberbine intermediates but instead led to accumulation of narcotoline at up to 20% total alkaloids (Fig. 4F). These results suggest that narcotoline is an end product of a desmethyl pathway that accumulates when PSMT2-mediated methylation at the 4′OH group of secoberbine intermediates is compromised. As for the noscapine pathway, the production of narcotoline via a desmethyl pathway is expected to require acetyltransferase, carboxylesterase, and dehydrogenase activities.

Our findings provide evidence for the involvement of 6 of the 10 genes from the HN1 gene cluster in noscapine biosynthesis. The remaining oxidation and acetyltransferase steps in the proposed pathway remain unaccounted for; they could be encoded by the CYP82X1, CYP82Y1, and PSAT1 genes, which remain to be characterized.

This discovery extends the involvement of gene clusters to the alkaloid class of secondary metabolites in higher plants. Noscapine has been reported in a number of Papaver species, and it will be interesting to establish whether it has evolved as a single event prior to speciation of P. somniferum or independently multiple times, as recently reported for the cyanogenic glucosides gene cluster in Lotus japonicus (28). As with the other plant gene clusters reported to date (19, 2831), donor sequences could be recruited from genes encoding related plant enzymes in a process involving gene duplication and neofunctionalization. The arrangement of the HN1 cluster suggests that genome reorganization is an ongoing process, occurring in some cases before duplication, as evidenced by the small gene families (PSCXE, PSCYP82, and PSMT), or after duplication, as evidenced by the single-copy genes (PSSDR1, PSAT1, and CYP719A21). The selective advantage to drive cluster evolution in this way could come from co-inheritance of favorable combinations of alleles and coordinate regulation of gene expression at the level of chromatin (32). This work provides a platform for the improved production of noscapine and related bioactive molecules through the molecular breeding of commercial poppy varieties or engineering of new production systems.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 to S8

References (3348)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank the laboratory and field staff at GlaxoSmithKline Australia for valuable technical assistance and J. Mitchell for administrative support. Supported by GlaxoSmithKline Australia and by the Garfield Weston Foundation for the Centre for Novel Agricultural Products. The identities of three of the P450s and the methyltransferase genes associated with noscapine production have been disclosed in PCT patent applications WO2011/161431 and WO2012/010872, respectively. The sequences of four BAC clones and genomic sequences of the 10 genes are available through the National Centre for Biotechnology Information under GenBank accession nos. JQ659009 to JQ659012 and JQ658999 to JQ659008. HN1, HM1, and HT1 are available from GSK Australia for experimental purposes, subject to a material transfer agreement.
View Abstract

Navigate This Article