Dissection of the phytohormonal regulation of trichome formation and biosynthesis of the antimalarial compound artemisinin in Artemisia annua plants


Author for correspondence:
Alain Goossens
Tel: +32 9 3313800
Email: alain.goossens@psb.vib-ugent.be


  • Biosynthesis of the sesquiterpene lactone and potent antimalarial drug artemisinin occurs in glandular trichomes of Artemisia annua plants and is subjected to a strict network of developmental and other regulatory cues.
  • The effects of three hormones, jasmonate, gibberellin and cytokinin, were studied at the structural and molecular levels in two different A. annua chemotypes by microscopic analysis of gland development, and by targeted metabolite and transcript profiling. Furthermore, a genome-wide cDNA-amplified fragment length polymorphism (AFLP)-based transcriptome profiling was carried out of jasmonate-elicited leaves at different developmental stages.
  • Although cytokinin and gibberellin positively affected at least one aspect of gland formation, these two hormones did not stimulate artemisinin biosynthesis. Only jasmonate simultaneously promoted gland formation and coordinated transcriptional activation of biosynthetic gene expression, which ultimately led to increased sesquiterpenoid accumulation with chemotype-dependent effects on the distinct pathway branches. Transcriptome profiling revealed a trichome-specific fatty acyl- coenzyme A reductase, trichome-specific fatty acyl-CoA reductase 1 (TFAR1), the expression of which correlates with trichome development and sesquiterpenoid biosynthesis.
  • TFAR1 is potentially involved in cuticular wax formation during glandular trichome expansion in leaves and flowers of A. annua plants. Analysis of phytohormone-modulated transcriptional regulons provides clues to dissect the concerted regulation of metabolism and development of plant trichomes.


Artemisinin, a sesquiterpene lactone found in Artemisia annua (sweet wormwood) plants, has been used in Chinese medicine for centuries. Currently, artemisinin and its semisynthetic derivatives are extensively used in the treatment of malaria, mostly in combination therapies (Haynes, 2006), and have gained additional interest because of their potential in the treatment of several cancers and viral diseases (Efferth, 2007; Efferth et al., 2008). Despite promising advances towards the fermentative production of artemisinin precursors by the expression of biosynthetic genes in microbial hosts (Ro et al., 2006; Chang et al., 2007; Arsenault et al., 2008; Zhang et al., 2008), engineering of A. annua plants for increased artemisinin production still remains of great interest (Covello, 2008; Graham et al., 2010).

The first committed step in artemisinin biosynthesis (Fig. 1) is the cyclization of farnesyl diphosphate (FDP) to generate amorpha-4,11-diene, catalysed by amorpha-4,11-diene synthase (ADS) (Mercke et al., 2000; Wallaart et al., 2001). Subsequent oxidation at the C12 position, mediated by the cytochrome P450 enzyme CYP71AV1, leads to artemisinic alcohol (Ro et al., 2006; Teoh et al., 2006). While arteannuin B has been suggested as a late precursor in artemisinin biosynthesis (Sangwan et al., 1993; Zeng et al., 2008), evidence now favours a route from artemisinic alcohol via dihydroartemisinic acid (Bertea et al., 2005; Covello et al., 2007; Covello, 2008). This route is supported by the cloning and characterization of double bond reductase 2 (DBR2), which reduces the Δ11(13) double bond of artemisinic aldehyde, but not of arteannuin B (Zhang et al., 2008), and the cloning of aldehyde dehydrogenase 1 (ALDH1), which catalyses the oxidation of artemisinic and dihydroartemisinic aldehydes (Teoh et al., 2009). The conversion of dihydroartemisinic acid to artemisinin, and of artemisinic acid to arteannuin B, has been suggested to occur via enzyme-independent reactions (Sy & Brown, 2002; Brown & Sy, 2004, 2007). Recently, a broad substrate oxidoreductase (RED1) with high affinity for dihydroartemisinic aldehyde and monoterpenes was identified that may have a negative impact on the flux to artemisinin biosynthesis (Rydén et al., 2010).

Figure 1.

 Biosynthetic pathway of artemisinin and arteannuin B. Dashed arrows indicate steps for which no enzymes have been identified.

The production of artemisinin occurs in specialized 10-celled biseriate glandular trichomes present on the leaves, stems and inflorescences of A. annua plants (Duke et al., 1994; Van Nieuwerburgh et al., 2006). All of these biosynthetic enzymes have been shown to be highly expressed in these particular trichomes (Bertea et al., 2005; Teoh et al., 2006, 2009; Zhang et al., 2008), most probably exclusively in the two outer apical cells (Olsson et al., 2009). Large differences in artemisinin content have been reported, depending on variety, season, cultivation condition and plant developmental stage (Ferreira et al., 1995; Wallaart et al., 2000; Delabays et al., 2001; Lommen et al., 2007; Davies et al., 2009; Yang et al., 2009). For instance, artemisinin concentrations are higher in leaves that are formed later in development than in leaves formed early in the plant’s development; this difference has been attributed to a higher trichome density and a higher capacity per trichome in the upper leaves (Lommen et al., 2006). Other leaf traits, such as perimeter, area and architecture, have recently been proposed as excellent targets for increasing artemisinin production (Graham et al., 2010). However, to date, the regulatory mechanisms that control artemisinin biosynthesis and the formation of the specialized cells in which it takes place are poorly characterized molecularly.

Very recently, a WRKY transcription factor, responsive to the phytohormone jasmonate, has been characterized that regulates the expression of the ADS gene (Ma et al., 2009). Another phytohormone, salicylic acid, affects artemisinin biosynthesis as well, presumably by enhancing the conversion of precursor pools and by up-regulating biosynthetic gene expression (Pu et al., 2009). Phytohormones are key molecules that are involved in practically any aspect of a plant’s development and its adaptation to the environment. First, the ability of jasmonates, in particular, but also of other phytohormones, to act as elicitors of plant secondary metabolism has been extensively proven. Across the plant kingdom, jasmonates are capable of inducing multiple secondary metabolic pathways, synthesizing molecules of an incredible structural variety and of various biochemical origin (Zhao et al., 2005; Pauwels et al., 2009). Secondly, jasmonates, but also cytokinins, gibberellins and brassinosteroids, are known to modulate epidermal differentiation programmes, resulting in increased trichome densities, ectopic trichome formation or aberrant trichome morphologies, as demonstrated in Arabidopsis (Arabidopsis thaliana) and tomato (Solanum lycopersicum) (Perazza et al., 1998; Traw & Bergelson, 2003; Li et al., 2004; Boughton et al., 2005; Gan et al., 2007; Maes et al., 2008; Lattarulo Campos et al., 2009; Yoshida et al., 2009; Maes & Goossens, 2010).

The observation that in model plant species phytohormones can modulate both the onset of secondary metabolism and trichome initiation at the transcriptional level (Maes et al., 2008; Pauwels et al., 2009) provides an excellent rationale to dissect the phenotypic responses of two different A. annua cultivars to different phytohormones, in order to unravel the regulatory mechanisms underlying these important traits and eventually identify the gene products involved. The effect of the three phytohormones mostly renowned for their stimulating effect on trichome development, namely cytokinin (6-benzylaminopurine (BAP)), gibberellin (GA3) and jasmonic acid (JA), was assessed at the developmental, metabolic and transcriptional levels in A. annua. In parallel, genome-wide cDNA-amplified fragment length polymorphism (AFLP)-based transcript profiling of JA-elicited leaves at two different developmental stages revealed transcriptional regulons consisting of genes with expression patterns that correlate with trichome development and metabolism, including a gene encoding a fatty acyl-coenzyme A (acyl-CoA) reductase.

Materials and Methods

Plant material, maintenance, and phytohormonal treatment

Two different Artemisia annua L. cultivars were used throughout the experiments. Seeds from the ‘2/39’ (HAP, for high-artemisinin producer) and the ‘Meise’ (LAP, for low-artemisinin producer) cultivars were provided by Pedro Melillo de Magalhães (State University of Campinas, Brazil) and the National Botanic Garden of Belgium (Meise, Belgium), respectively. Seeds of both cultivars were germinated and grown in soil under normal, controlled conditions (21°C; 12 h day : 12 h night regime), until they had formed the first four leaves. From then on, seedlings were treated every 2 d with three different phytohormone solutions, containing either 100 μM JA, 50 μM BAP or 100 μM GA3 (all from Sigma-Aldrich). For each treatment, 2.5 ml of the hormone or mock (water) solutions were applied to the soil and 2.5 ml were sprayed on the leaves of each plant. Treatment was carried out during five consecutive weeks and samples were harvested every week.

Measurement of trichome density and glandular trichome cross-sectional area

After 2 wk of phytohormonal treatment, the eighth leaf (from the bottom) was removed from each plant, mounted on scanning electron microscopy (SEM) stubs with double-sided sticky carbon tape, and examined in a tabletop SEM (TM-1000) under an accelerating voltage of 5 kV. Leaves of at least six different plants per treatment were scanned and digital micrographs recorded. Trichome density was determined for the abaxial leaf epidermis of at least 100 mm2 per leaf and spanning different areas by using the digital micrographs and the gridlines in Photoshop 7 (http://www.adobe.com). Of the almost elliptical cross-sectional area of the glandular trichomes, the major (A) and minor (B) axes of the ellipse were measured with the program ImageJ (version 1.37; http://rsb.info.nih.gov/ij/). The area was calculated according to the formula, area = ABπ/4.

Metabolite profiling

For each treatment and at each time point, leaves of three whole plants were pooled and weighed, and two pools were used for further analysis. Sesquiterpenoids and precursors were released from the glandular trichomes by a 1 min chloroform extraction and measured by reversed-phase high-pressure liquid chromatography (HPLC) electrospray quadrupole time of flight tandem MS essentially as described in Van Nieuwerburgh et al. (2006).

Quantitative real-time-polymerase chain reaction (qRT-PCR)

RNA was extracted with Plant Reagent (Invitrogen) and poly(dT) cDNA was prepared from 1 μg of total RNA with Superscript II reverse transcriptase (Invitrogen). qRT-PCR was carried out with SYBR Green QPCR Master Mix (Stratagene, La Jolla, CA, USA) and gene-specific primers were designed with Primer3 (http://biotools.umassmed.edu/bioapps/primer3_www.cgi). Expression levels were normalized to those of the ACTIN control gene (GenBank accession number EU531837). For phytohormone-treated samples, three biological repeats were analysed, each consisting of pooled leaves from three plants, harvested 2 wk after treatment. For measurement of organ-specific expression, sampling was done as described previously (Zhang et al., 2008).

cDNA-AFLP-based transcript profiling

For cDNA-AFLP analysis, another elicitation experiment with the LAP cultivar was set up, in which samples were harvested 1, 2, 4, 8, 24 and 48 h after JA or mock treatment. Leaves were harvested from the lower and the upper parts of the plant, and processed separately. Total leaf RNA was extracted as for the qRT-PCR analysis. cDNA-AFLP-based transcript profiling was done essentially as described previously (Rischer et al., 2006; Vuylsteke et al., 2007).

Full-length cDNA-cloning of trichome-specific fatty acyl-CoA reductase 1 (TFAR1)

Total RNA was isolated from the glandular trichomes of A. annua line 2/39. Single-stranded cDNA was synthesized as described previously (Zhang et al., 2008). The 5′- and 3′-end sequences of TFAR1 were recovered by a rapid amplification of cDNA ends (RACE)-PCR strategy based on the cDNA-AFLP tag sequence AA387. cDNA encoding the TFAR1 open reading frame was obtained through 33 cycles of PCR with the glandular trichome cDNA as the template with an annealing temperature of 58°C, Taq DNA polymerase (Invitrogen) and gene-specific oligonucleotide primers 5′-GGATCCATGATGGAGTTGGGTAGAATTG-3′ and 5′-CTCGAGTCATTTAAACTCACGGTGCCTT-AG-3′, with the BamHI and XhoI restriction sites underlined and the start and stop codons in bold. The PCR product was initially cloned into the vector pCR2.1-TOPO (Invitrogen), to give the plasmid pCR2.1-TFAR1, then subcloned into the yeast expression vector pESC-Leu (Stratagene) with the BamHI and XhoI restriction sites and yielding the plasmid pESC-Leu-TFAR1.

Functional analysis of TFAR1 in yeast

The plasmid pESC-Leu-TFAR1 and empty vector pESC-Leu were introduced separately into the yeast (Saccharomyces cerevisiae) strain WAT11 (Teoh et al., 2006). Yeast cultures (10 ml) were grown overnight at 30°C in Leu dropout liquid medium (Clontech, Mountain View, CA, USA) containing 2% (w/v) glucose. After 24 h in an orbital shaker maintained at 250 rpm, cultures were collected by centrifugation, washed three times in sterile deionized water, and resuspended to an A600nm of 0.8 in 10 ml Leu dropout liquid medium containing 2% (w/v) galactose supplemented with a mixture of very-long-chain fatty acids (100 μl of a solution containing 10% (v/v) tergitol and 0.1 g l−1 of each of the free fatty acids 20:1, 22:1, 24:1, 26:0 and 30:0 (Nu-chek Prep, Elysian, MN, USA)). After incubation at 30°C for 24 h, the cell pellets were collected by centrifugation (8000 g for 10 min).

The yeast pellets were prepared and analysed by gas chromatography (GC)-MS as described (Gagnéet al., 2009), except that the final residue was dissolved in 40 μl of a derivatization solution containing dichloromethane, N,O-bis-(trimethylsilyl)acetamide (Sigma-Aldrich) and pyridine (Sigma-Aldrich) (1 : 1 : 1) before GC-MS analysis. Fatty alcohol reference standards (24:0-OH and 26:0-OH; Nu-chek Prep) were dissolved in the same derivatization solution before GC-MS analysis.


Phytohormones promote the formation of both glandular and filamentous trichomes on leaves of different A. annua cultivars

Leaves of A. annua possess not only glandular trichomes, in which artemisinin is produced, but also nonglandular, multicellular filamentous trichomes. Measurements based on tabletop SEM revealed a large variation in trichome density within the two A. annua cultivars used here, designated HAP and LAP for the 2/39 (from Brazil) and Meise (from Belgium) cultivars, and representing high- and low-artemisinin cultivars, respectively. A higher trichome density, with 14.2-fold more filamentous and 1.4-fold more glandular trichomes, was observed in control (mock treated) leaves of HAP plants than in control LAP plants (Fig. 2a).

Figure 2.

 Effect of phytohormonal treatment on leaf trichome development. (a) Trichome density. Numbers in the ordinate indicate the glandular (GT) and filamentous (FT) trichomes mm−2 in Artemisia annua low-artemisinin producer (LAP, open bars) and high-artemisinin producer (HAP, closed bars) cultivars. (b) Gland size. Numbers in the ordinate indicate the mean cross-sectional area of the gland in LAP and HAP cultivars. Error bars are SE (= 6). Statistical significance was determined by Student’s t-test (**, < 0.01; *, < 0.05). (c) Representative SEM images of mock-treated and hormone-elicited LAP leaves. Bar, 1 mm (top); 300 μm (bottom). JA, jasmonic acid; BAP, 6-benzylaminopurine; GA3, gibberellin.

To investigate the effect of JA, BAP and GA3 on the regulation of the two trichome types, an experimental method designed in Arabidopsis to score the effects of these phytohormones (Maes et al., 2008) was adapted to A. annua plants. Two general observations could be drawn from this elicitation experiment. First, the two different trichome types were clearly differentially regulated. Whereas all phytohormones increased the density of filamentous trichomes, only BAP and JA, but not GA3, stimulated the glandular trichome formation (Fig. 2a,c). Secondly, the phytohormone-triggered trichome promotion trends were identical in the two cultivars, although quantitative differences in induction levels were discernible (Fig. 2a,c). For the glandular trichomes, JA and BAP treatments resulted in a 4.3- and 5.5-fold higher trichome density in LAP plants, whereas only a 1.6- and 2.5-fold induction could be observed in HAP plants, respectively. Similarly, filamentous trichome densities increased 23.0-, 22.6- and 7.6-fold in LAP plants and only 1.4-, 2.5- and 1.5-fold in HAP plants after elicitation with JA, BAP and GA3, respectively. Hence, although the ‘basal’ trichome density in LAP leaves was lower than that in HAP leaves, the plasticity with regard to trichome formation seemed higher in the LAP cultivar.

Phytohormones differentially affect glandular trichome size

Besides affecting trichome initiation, previous studies have indicated that phytohormones can also interfere with trichome maturation and size (Maes et al., 2008). To assess whether glandular trichome size was also affected in elicited A. annua plants, we measured the cross-sectional area of glandular trichomes, as a representative parameter. Notably, in addition to gland density differences (Fig. 2a), gland size also differed between the two cultivars under control (mock-treated) conditions: trichomes of HAP plants were 26.5% larger than those of LAP plants (Fig. 2b).

In contrast to the effects on trichome density, the phytohormonal effects on gland size were markedly distinct in the two cultivars (Fig. 2b,c). JA and GA3 both increased the size of glandular trichomes in LAP plants, by 65.6 and 45.6%, respectively, but not in HAP plants. Only after BAP treatment was a similar pattern observed in the two cultivars, namely a clear decrease of 44.0% in trichome size in the two lines. Hence, for both size and density of glandular trichomes, the LAP plants appeared to possess a greater plasticity than the HAP plants. The lack of positive effects on HAP gland size might reflect a certain degree of saturation in this cultivar.

Jasmonate, but not gibberellin or cytokinin, promotes sesquiterpenoid accumulation in glandular trichome exudates

To analyse how the phytohormonal treatment altered the artemisinin production, we measured the concentrations of a range of sesquiterpenoids that included artemisinic acid, dihydroartemisinic acid, artemisinic alcohol, dihydroartemisinic alcohol, artemisinic aldehyde, dihydroartemisinic aldehyde, artemisinin and arteannuin B (Fig. 1) in the two cultivars over a time period of 5 wk (Fig. 3 and Supporting Information Fig. S1). As for the trichome parameters, both cultivars had different ‘basal’ sesquiterpenoid concentrations. The HAP cultivar represents a high-artemisinin chemotype, whereas the LAP cultivar can be considered as a low-artemisinin chemotype. Also dihydroartemisinic acid, artemisinic aldehyde and dihydroartemisinic aldehyde accumulated more in HAP plants, but, conversely, concentrations of arteannuin B and artemisinic acid were notably lower in HAP than in LAP plants (Figs 3, S1). Overall, the HAP cultivar was characterized by high dihydroartemisinic acid : artemisinic acid and artemisinin : arteannuin B ratios (c. 50 : 1 and 54 : 1 after 5 wk of mock treatment, respectively), whereas LAP plants showed an opposite chemotype (ratios of c. 1 : 19 and 1 : 4 after 5 wk of mock treatment, respectively).

Figure 3.

 Metabolite profiling of elicited Artemisia annua plants. Accumulation of artemisinic acid, dihydroartemisinic acid, artemisinin and arteannuin B in leaves of low-artemisinin producer (LAP) and high-artemisinin producer (HAP) cultivars measured 1–5 wk after mock treatment (CON) or treatment with jasmonic acid (JA), 6-benzylaminopurine (BAP) and gibberellin (GA3). Error bars are SE (= 2). Measurements were repeated twice with similar results.

Of the three phytohormones, only JA was capable of eliciting pronounced increases in sesquiterpenoid accumulation, but with different specificities on the flux through the different pathway branches in the two chemotypes (Figs 3, S1). The analysis of the individual compounds revealed a clear impact of JA treatment on artemisinic acid, artemisinin and arteannuin B in LAP plants, on the one hand, and on artemisinic aldehyde, dihydroartemisinic aldehyde, dihydroartemisinic acid and artemisinin in HAP plants, on the other hand. No significant induction occurred in any of the two A. annua chemotypes for artemisinic alcohol or dihydroartemisinic alcohol. In all cases, a maximum elicitation effect was obtained after 3 wk of treatment, with two notable exceptions, namely the end products artemisinin and arteannuin B in LAP plants. In fact, the accumulation after 5 wk of JA treatment for both artemisinin and arteannuin B was higher in LAP plants than in HAP plants, further underscoring the pronounced plasticity of the LAP trichome machinery and their biosynthetic capacities. Furthermore, even though the absolute amount of arteannuin B remained higher than that of artemisinin in JA-elicited LAP plants, the artemisinin: arteannuin B ratio did increase relative to mock-treated plants (c. 1 : 2 vs 1 : 4, respectively). In HAP plants, JA-elicitation clearly promoted the flux to artemisinin only, with an artemisinin : arteannuin B ratio of c. 116 : 1 in JA-elicited plants vs 54 : 1 in mock-treated plants.

No stimulation of artemisinin or arteannuin B biosynthesis, or of any of their precursors, was observed after BAP or GA3 treatment in either of the two cultivars (Figs 3, S1). On the contrary, the overall tendency pointed rather to a slightly negative effect of both hormones.

Jasmonate induces expression of artemisinin biosynthesis genes in a coordinated manner

To characterize the phytohormonal effects on the regulation of artemisinin biosynthesis, we evaluated the transcriptional regulation of the known biosynthetic genes (Fig. 1) by qRT-PCR. As the maximal metabolite accumulation was observed 3 wk after elicitation, leaves sampled at the preceding time point, that is, after 2 wk of treatment, were chosen for expression analysis.

Expression of all artemisinin biosynthetic genes significantly increased after JA treatment (Fig. 4), which is in clear agreement with the data obtained by metabolite profiling. Expression of ADS, CYP71AV1, DBR2 and the cytochrome P450 reductase (CPR), of which the corresponding proteins catalyse the formation of amorpha-4,11-diene and its ultimate conversion to dihydroartemisinic acid, was induced in both cultivars. The expression of the upstream gene FDP synthase (FDS), of which the corresponding enzyme catalyses the formation of the central sesquiterpenoid precursor FDP, increased only in the HAP cultivar, and that of ALDH1, of which the gene product catalyses the oxidation of artemisinic and dihydroartemisinic aldehydes, increased only in the LAP cultivar. Notably, the increase in DBR2 expression was more pronounced in the HAP cultivar (five- to sevenfold in LAP plants vs 50- to 70-fold in HAP plants), correlating with the specific and marked induction of dihydroartemisinic aldehyde and dihydroartemisinic acid metabolite levels in HAP plants.

Figure 4.

 Regulation of artemisinin biosynthetic gene expression by jasmonic acid (JA) elicitation in Artemisia annua low-artemisinin producer (LAP) and high-artemisinin producer (HAP) leaves. Expression of farnesyl diphosphate synthase (FDS) (a), amorpha-4,11-diene synthase (ADS) (b), CYP71AV1 (c), double bond reductase 2 (DBR2) (d), aldehyde dehydrogenase 1 (ALDH1) (e) and cytochrome P450 reductase (CPR) (f) verified by quantitative real-time-polymerase chain reaction (qRT-PCR) analysis and normalized to that of the ACTIN control gene (GenBank accession number EU531837). Error bars are SE (= 3).

Although no pronounced changes in metabolite levels after phytohormonal induction with BAP and GA3 were observed, some of the artemisinin biosynthetic genes showed altered expression levels in either LAP or HAP plants (Fig. S2). However, in contrast to the concerted up-regulation observed after JA elicitation, the patterns were not consistent after either BAP or GA3 treatments in either of the two cultivars. Most importantly, for both cultivars and both BAP and GA3 treatments, at least one of the known artemisinin biosynthesis genes was down-regulated, which correlated with the overall metabolite profiles of plants treated with BAP or GA3. For instance, in HAP plants, ADS and FDS were down-regulated after BAP and GA3 treatment, respectively, and both BAP and GA3 down-regulated ADS and CYP71AV1 in LAP plants. These observations suggest that BAP and GA3 could not stimulate sesquiterpenoid biosynthesis within the glands, in spite of their positive effect on trichome initiation and size, respectively.

Jasmonate-modulated reprogramming of the A. annua leaf transcriptome

To characterize in depth the transcriptional response of A. annua plants to phytohormones that affect glandular trichome development and metabolism and to reveal novel genes potentially involved in these processes, a genome-wide cDNA-AFLP-based transcriptome analysis was launched. Because LAP plants showed the largest plasticity for all the parameters tested, that is, sesquiterpenoid accumulation, gland density and gland size, this cultivar was chosen as the model system. Similarly, because it was the only phytohormone that positively affected all parameters, JA was selected as the preferred elicitor for a new transcript profiling experiment, in which we focused on early elicitation events, from 0 to 48 h after JA elicitation. Furthermore, the transcriptional response in A. annua leaves at two different leaf developmental stages was compared. These stages corresponded physically with leaves from the bottom and top parts of the same plant that were harvested over 48 h after JA treatment and were designated ‘lower’ and ‘upper’ leaves, respectively. More importantly, these two stages reflected leaves that were either already (or nearly) fully developed (‘lower’) vs leaves that were still developing and thus still actively forming trichomes (‘upper’).

In total, the expression of c. 12 000 gene tags was visualized. Clustering and sequence analysis of the 493 differentially expressed transcript tags for which a unique sequence with a BLAST hit in the public databases was obtained led to the identification of clusters with gene tags corresponding to gene products with yet unknown function or putatively involved in protein synthesis and fate, transcription, signal transduction, transport, photosynthesis, energy and metabolism, and cell organization and defence (Table S1).

The transcriptional response of the two leaf stages to JA treatment differed remarkably, demonstrating that the context within which the JA signal was perceived was crucial for the shaping of the transcriptional response (Fig. 5; Table S1). Here, we further concentrated on a cluster of 174 genes, potentially encoding metabolic enzymes of which the expression was stimulated by JA (Fig. 5). Among this cluster, tags corresponding to FDS and 1-deoxyxylulose 5-phosphate synthase (DXS), as well as CYP71AV1, were identified. The former two encode enzymes known to be involved in the biosynthesis of the necessary FDP precursor (Schramek et al., 2009), while the latter gene is specifically expressed in trichomes and dedicated to artemisinin biosynthesis (Teoh et al., 2006). Interestingly, and particularly relevant to artemisinin biosynthesis and gland formation, all three of these known artemisinin biosynthesis genes were only (CYP71AV1) or primarily (DXS and FDP) up-regulated in the upper leaves in response to JA treatment, indicating the importance of the developmental stage for the plants’ capacity to synthesize artemisinin (Fig. 5).

Figure 5.

 cDNA-amplified fragment length polymorphism (AFLP) transcript profiling of jasmonic acid (JA)-elicited Artemisia annua low-artemisinin producer (LAP) plants. Average linkage hierarchical clustering of LAP cDNA-AFLP tags with high similarity to genes encoding metabolic enzymes. The treatments and time points (in h) are indicated at the top and the gene annotations on the right. Yellow and blue boxes correspond to increased and reduced transcript accumulation, respectively, relative to the average accumulation level of all samples. Left, full average linkage hierarchical clustering of all 174 enzyme tags. Right, expanded view of the pink part of the tree, corresponding to a cluster of 57 genes, the expression of which is induced by JA in the upper leaves primarily, and in which the artemisinin biosynthesis genes 1-deoxyxylulose 5-phosphate synthase (DXS), farnesyl diphosphate synthase (FDS) and CYP71AV1 as well as trichome-specific fatty acyl-CoA reductase 1 (TFAR1) are present.

In addition, a large number of genes potentially encoding enzymes involved in other secondary metabolic pathways, in particular flavonoid biosynthesis, were identified. Many of these genes had an expression pattern similar to that of artemisinin biosynthesis genes, that is, a differential JA-induction in upper vs lower leaves, pointing to a coordinated transcriptional activation of several secondary metabolic pathways within JA-elicited A. annua leaves and/or trichomes (Fig. 5; Table S1).

Identification of a novel trichome-specific gene encoding an alcohol-forming very-long-chain fatty acyl-CoA reductase

Three gene tags, AA064, AA387 and AA707, were investigated that had been annotated as a putative β-ocimene synthase, an acyl-CoA reductase and an 2-alkenal reductase, respectively, and that were all induced by JA treatment. AA387 and AA707 were induced in the upper leaves only and clustered with the three known artemisinin biosynthesis genes, whereas AA064 was induced in leaves from both developmental stages and belonged to another cluster (Fig. 5).

First, we assessed whether JA induction of these three genes was conserved among different A. annua cultivars. Expression analysis in LAP and HAP plants of the first elicitation series confirmed independently that all three genes were inducible by JA in LAP plants and indicated that a similar JA-mediated transcriptional induction also occurs in other A. annua cultivars, at least in the HAP line (Fig. S3). Second, because the cDNA-AFLP profiling was done with RNA from whole leaves only, we also analysed the tissue-specific expression of these three genes in more detail. All of them were expressed in flower buds, leaves and flower bud trichomes, but negligibly in roots (Fig. 6). Only one gene, corresponding to tag AA387, showed a true trichome-specific expression pattern, like that of the known artemisinin biosynthesis genes such as ADS, DBR2, CYP71AV1 and ALDH1 (Fig. 6).

Figure 6.

 Tissue-specific expression of jasmonic acid (JA)-elicited genes in Artemisia annua. Tissue-specific expression of genes encoding amorpha-4,11-diene synthase (ADS) (a), CYP71AV1 (b), double bond reductase 2 (DBR2) (c), aldehyde dehydrogenase 1 (ALDH1) (d), cytochrome P450 reductase (CPR) (e), the annotated β-ocimene synthase (f), acyl-CoA reductase (trichome-specific fatty acyl-CoA reductase 1 (TFAR1)) (g), and 2-alkenal reductase (h), verified by quantitative real-time-polymerase chain reaction (qRT-PCR) analysis. Error bars are SE (= 3).

Considering that the expression of AA387 closely matched that of the established artemisinin biosynthetic genes, we cloned the full-length cDNA corresponding to AA387 and functionally characterized it. Full-length AA387, hereafter referred to as TFAR1, was predicted to be 493 amino acids long, with a molecular weight of 55 914. TFAR1 showed the highest similarity to the family of fatty acyl-CoA reductases (FARs), which includes the jojoba (Simmondsia chinensis) FAR involved in seed wax ester biosynthesis and the Arabidopsis CER4 involved in cuticular wax biosynthesis (Metz et al., 2000; Rowland et al., 2006; Doan et al., 2009) (Fig. S4). Expression of TFAR1 in yeast (S. cerevisiae), supplemented with very-long-chain fatty acids, resulted in the gene-dependent production of 24:0 and 26:0 primary alcohols (Figs 7, S5), indicating that TFAR1 encodes a functional fatty acyl-CoA reductase.

Figure 7.

 Heterologous expression of trichome-specific fatty acyl-CoA reductase 1 (TFAR1) in yeast. GC-MS total ion current traces of derivatized lipid extracts from induced yeast cultures, supplemented with very-long-chain fatty acids containing the negative control empty vector pESC-Leu (a) and TFAR1 expression vector pESC-Leu-TFAR1 (b) compared with a standard mixture of fatty alcohols (c). TIC, total ion count.


Despite the pharmaceutical importance of plant-derived artemisinin, relatively little is known about the regulation of its biosynthesis in A. annua plants, and the development of the specialized producer organ, the glandular trichome, in which biosynthesis exclusively occurs. Here, we studied the regulation of these fundamental processes by investigating the plants’ dynamic response to phytohormonal cues, important modulators of development and metabolism.

Phytohormones differentially affect gland formation and sesquiterpenoid accumulation

Two different A. annua cultivars, characterized by different artemisinin : arteannuin B ratios and trichome parameters, were profiled and three phytohormones, GA3, BAP and JA, that had been shown previously to promote trichome initiation in Arabidopsis were used (Maes et al., 2008). Arabidopsis has only unicellular filamentous trichomes, whereas A. annua has filamentous and glandular trichomes, both of the multicellular type. In Asterids and Rosids, the lineages to which A. annua and Arabidopsis belong, respectively, filamentous trichomes have been postulated to be analogous rather than homologous structures and to develop through different transcriptional regulatory networks (Serna & Martin, 2006). Nonetheless, the phytohormonal elicitation patterns for the filamentous trichomes of A. annua closely matched those observed for Arabidopsis trichomes. Indeed, all three hormones clearly promoted hairy trichome formation, in the two A. annua cultivars tested, suggesting that the formation of at least these filamentous structures obeys similar regulatory cues, although not necessarily by means of similar molecular components.

By contrast, the patterns for promotion of glandular trichome density and size were distinct and showed great variability between the A. annua cultivars. First, JA and BAP, but not GA3, could stimulate gland formation. Second, only after JA treatment, an increase in gland density was accompanied with an increase in gland size, possibly reflecting an advanced maturation and/or increased biosynthetic activity. By contrast, BAP markedly reduced A. annua gland size, as was similarly observed in BAP-treated Arabidopsis trichomes. This characteristic had been attributed to the general capacity of cytokinins to keep cells in a division state, thereby stimulating formation, but preventing full maturation, of, for instance, trichome cells (Maes et al., 2008). Third, the plasticity of trichome capacities differed strikingly between the two cultivars; the cultivar with the lowest ‘basal’ density and size of trichomes (LAP) was far more susceptible to JA elicitation and ultimately displayed more and larger glands than the elicited cultivar with a higher ‘basal gland score’ (HAP). Although it cannot be excluded that this observation could be (partly) the result of a different ability to uptake phytohormones, this important trait could be taken into consideration when planning research to increase our fundamental understanding of gland formation or artemisinin biosynthesis or when designing A. annua breeding programmes for commercial purposes.

Sesquiterpenoid accumulation was stimulated markedly in JA-elicited plants only, and not in plants treated with BAP or GA3, indicating that an increase in gland density was not necessarily accompanied by an increase in metabolite production. Expression analysis corroborated that this effect could be attributed, at least in part, to the fact that JA is the only phytohormone capable of generating a coordinated transcriptional activation of the biosynthetic genes involved. It is worth noting that our profiling results are consistent with a branched pathway that leads to two related major sesquiterpenoids – artemisinin and arteannuin B – with artemisinic aldehyde as the last common intermediate. Based on the different JA-mediated elicitation and accumulation trends in the two chemotypes, the accumulation of dihydroartemisinic acid appears to be associated with the production of artemisinin, whereas that of artemisinic acid appears to be associated with the production of arteannuin B (Fig. 1). The existence of and relative flux through the two branches of the pathway might provide a molecular basis for the high- or low-artemisinin character of particular A. annua chemotypes. For instance, in LAP plants, the arteannuin B route seems the most competent, whereas the biosynthetic machinery of HAP plants clearly promotes the flux towards artemisinin (Fig. 3).

A transcriptomics-based screening strategy reveals genes potentially involved in the development and metabolism of glandular trichomes

Because transcriptional coregulation is an important hallmark of genes involved in secondary metabolite pathways in general and, as shown here, also in A. annua sesquiterpenoid biosynthesis, a transcript profiling-based experimental strategy might well lead to the identification of novel genes encoding enzymes that act, for instance, in arteannuin B biosynthesis, or in metabolic pathways that support or compete with artemisinin biosynthesis in A. annua glands. Careful design of such a transcript profiling exercise might also yield novel genes involved in the actual formation of the artemisinin producer organs, the glands. Indeed, although no information is yet available on the identity of the molecular players that steer the formation of trichomes of the glandular type from any plant species, at least some phytohormone-modulated regulatory cues seem to be conserved across the plant kingdom (Maes & Goossens, 2010; this study). Furthermore, from research conducted in Arabidopsis, it became clear that the phytohormone-mediated promotion of unicellular, filamentous trichome formation can be triggered by the transcriptional activation of the components of the established trichome initiation complex (Maes et al., 2008).

Therefore, a cDNA-AFLP-based transcript profiling analysis was launched on JA-elicited plants of the cultivar with the lowest basal metabolite and gland scores, but that showed the highest plasticity and, thus, might inherently display more pronounced transcriptional responses to a particular stimulus. Accordingly, in this analysis known trichome-specific artemisinin biosynthesis genes were identified, as well as novel genes encoding enzymes catalysing as yet unknown metabolic conversions or potential regulatory proteins steering trichome formation and metabolism (Table S1). Furthermore, by profiling the JA response in leaves at distinct developmental stages, we demonstrated that the developmental time frame within which the plant or its leaves are elicited, represents a key parameter, which illustrates the importance of the ‘context’ for the perception of the jasmonate signal and the eventual physiological responses. Notably, the transcriptional up-regulation of the artemisinin biosynthetic genes was also observed primarily in developing leaves, indicating that the developmental stage is not only important for the formation of the producer organs themselves but also for their capacity to synthesize artemisinin.

Based on this transcriptome analysis, a new functional FAR, designated TFAR1 and closely related to the Arabidopsis CER4 involved in cuticular wax biosynthesis (Rowland et al., 2006), has been characterized. Like CER4, TFAR1 catalyses the formation of C24 and C26 fatty alcohols, presumably from fatty acyl-CoAs. Notably, CER4 expression in Arabidopsis stems and leaves is detected exclusively in epidermal cells and is, in rosette leaves, even confined to trichome cell types (Rowland et al., 2006). Similar trichome-specific expression patterns have also been observed for CER2 and YORE-YORE1 (YRE1)/WAX2, both encoding enzymes catalysing other steps in cuticular wax production in Arabidopsis (Xia et al., 1997; Kurata et al., 2003). Moreover, several Arabidopsis mutants, defective in loci encoding proteins with either catalytic, regulatory or transport roles in cuticular wax biosynthesis, display smaller (yre1; Kurata et al., 2003), distorted (desperado; Panikashvili et al., 2007), fewer (fiddlehead; Yephremov et al., 1999) or more (shine; Aharoni et al., 2004) trichomes than the wild type, indicating that cuticular wax production is required for normal trichome development. Hence, taking into account its activity, its analogy with the Arabidopsis CER4, and its JA-inducible, and trichome- and ‘developing leaf’-specific expression pattern, we postulate that TFAR1 is involved in cuticular wax formation during glandular trichome expansion in leaves and flowers of A. annua plants. Intriguingly, the phytohormone-regulated and the development- and organ-specific expression pattern of TFAR1 very closely matches that of key artemisinin biosynthetic genes, such as ADS and CYP71AV1, which might point to the existence of a common regulatory programme for both the metabolic content and the structure of the gland. In particular, jasmonates seem to be capable of impinging on this programme, as supported by the observation that only JA can stimulate jointly gland initiation, gland size, sesquiterpenoid biosynthesis and, last but not least, expression of the genes involved.

These results open exciting perspectives for the exhaustive transcriptome mining that is in progress in various laboratories (Wang et al., 2009; Dai et al., 2010; Graham et al., 2010; this work) or that will be undertaken in the near future. Such gene discovery programmes will help to fill a major gap in our understanding of plant development: undoubtedly the resulting transcriptome data will reveal genes encoding as yet unknown regulatory proteins that determine the formation of glandular trichomes. Ultimately, such genes will be invaluable, not only for successful metabolic engineering of artemisinin biosynthesis in A. annua, but also for numerous other high-value compounds produced in the glandular trichomes of other plants.


The authors thank Martine De Cock for help in preparing the manuscript. This work was supported by grants from Ghent University (VARL9104), the Vlaamse Inter universitaire Raad (VLADOC-B/09269/02; predoctoral fellowship to L.M.), and the National Research Council of Canada’s Plants for Health and Wellness Program.