The metabolite chemotype of Nicotiana benthamiana transiently expressing artemisinin biosynthetic pathway genes is a function of CYP71AV1 type and relative gene dosage


Author for correspondence:

Harro J. Bouwmeester

Tel: +31 317 489 859



  • Artemisia annua, which produces the anti-malaria compound artemisinin, occurs as high-artemisinin production (HAP) and low-artemisinin production (LAP) chemotypes. Understanding the basis of the difference between these chemotypes would assist breeding and optimising artemisinin biosynthesis.
  • Here we present a systematic comparison of artemisinin biosynthesis genes that may be involved in determining the chemotype (CYP71AV1, DBR2 and ALDH1). These genes were isolated from the two chemotypes and characterized using transient expression in planta. The enzyme activity of DBR2 and ALDH1 from the two chemotypes did not differ, but structural differences in CYP71AV1 from LAP and HAP chemotypes (AMOLAP and AMOHAP, respectively) resulted in altered enzyme activity.
  • AMOLAP displays a seven amino acids N-terminal extension compared with AMOHAP. The GFP fusion of both proteins show equal localization to the ER but AMOHAP may have reduced stability.
  • Upon transient expression in Nicotiana benthamiana, AMOLAP displayed a higher enzyme activity than AMOHAP. However, expression in combination with the other pathway genes also resulted in a qualitatively different product profile (‘chemotype’); that is, in a shift in the ratio between the unsaturated and saturated (dihydro) branch of the pathway.


Based on the content of artemisinin and its precursors, two chemotypes of Artemisia  annua, can be distinguished: the low-artemisinin production (LAP) chemotype and a high-artemisinin production (HAP) chemotype (Wallaart et al., 2000). Both chemotypes contain artemisinin and arteannuin B, but the HAP chemotype has a relatively high content of artemisinin and its presumed precursor dihydroartemisinic acid (DHAA), while the LAP chemotype has a high content of arteannuin B and its presumed precursor artemisinic acid (AA).

The artemisinin biosynthesis pathway has been largely elucidated and the genes required for production of dihydroartemisinic acid, the most likely precursor of artemisinin (ADS, CYP71AV1, DBR2 and ALDH1) have all been described (Bouwmeester et al., 1999; Teoh et al., 2006, 2009; Zhang et al., 2008; Rydén et al., 2010). Artemisinin is a sesquiterpene lactone endoperoxide, which is synthesized in the cytosol from the general isoprenoid precursors IPP and DMAPP. These are converted to FPP and the first committed step in the artemisinin biosynthetic pathway is the cyclization of FPP to amorpha-4,11-diene (AD) by amorphadiene synthase (Fig. 1; Bouwmeester et al., 1999; Mercke et al., 2000). In the subsequent step, AD is oxidized by the cytochrome P450 enzyme, CYP71AV1/AMO, to artemisinic alcohol (AAOH), artemisinic aldehyde (AAA) and artemisinic acid (AA; Fig. 1; Ro et al., 2006; Teoh et al., 2006). However, the latter mainly occurs in the LAP chemotype. In the HAP chemotype only very little of the AAA is converted to AA, as most of the AAA is converted to dihydroartemisinic aldehyde (DHAAA) by DBR2, the enzyme that reduces the exocyclic double bond of AAA (Fig. 1; Bertea et al., 2005; Zhang et al., 2008). Supposedly, DHAAA is subsequently oxidized by alcohol dehydrogenase ALDH1 to the final intermediate dihydroartemisinic acid (DHAA; Bertea et al., 2005; Zhang et al., 2008). The conversion of DHAA to artemisinin, is believed to be a nonenzymatic and spontaneous photo-oxidation reaction (Wallaart et al., 1999; Sy & Brown, 2002). Similarly, in the LAP chemotype, AA is likely spontaneously converted to arteannuin B.

Figure 1.

Artemisinin biosynthetic pathway in Artemisia annua. IPP, isopentenyl diphosphate; DMAPP, dimethylallyl diphosphate; FPP, farnesyl diphosphate; AD, amorpha-4,11-diene; AAOH, artemisinic alcohol; AAA, artemisinic aldehyde; AA, artemisinic acid; AB, arteannuin B; DHAAOH, dihydroartemisinic alcohol; DHAAA, dihydroartemisinic aldehyde; DHAA, dihydroartemisinic acid; FPS, farnesyl diphosphate synthase; HMGR, 3-hydroxy-3-methylglutaryl-CoA reductase; ADS, amorphadiene synthase; CYP71AV1, amorphadiene oxidase; DBR2, artemisinic aldehyde double-bond reductase; RED1, dihydroartemisinic aldehyde reductase 1; ALDH1, aldehyde dehydrogenase 1. Broken arrows indicate the involvement of more than one step (Nguyen et al., 2011).

Recently we reported on the (transient) reconstruction of the artemisinin biosynthetic pathway in Nicotiana benthamiana leaves, resulting in up to 39.5 mg kg−1 FW of AA (van Herpen et al., 2010). In the present work we analyse the role of DBR2, ALDH1 and CYP71AV1 in determining the ‘chemotype’ (as defined by the AA and DHAA ratio) of N. benthamiana leaves agro-infiltrated with artemisinin biosynthesis genes. Results show that the chemotype is a function of the CYP71AV1 type and relative dosage of DBR2 and ALDH1.

Materials and Methods

Cloning of the ADS + FPS + HMGR expression construct AmFH

Cloning of the AmFH expression construct which contains ADS with CoxIV mitochondrial targeting signal, mitochondrial targeted FPS and a cytosolic (truncated) HMGR under control of the CaMV35S promoter has been described previously (van Herpen et al., 2010).

Identification and cloning of CYP71AV1 from HAP and LAP chemotypes

The AMOHAP EST sequence was identified in the sequence database of an Artemisia annua L. (HAP chemotype) glandular trichome cDNA library (Bertea et al., 2006). This sequence has been deposited in the GenBank database (JQ254992). The full length sequence of AMOHAP was obtained by RACE PCR (Clontech, Mountain View, CA, USA). We note that the race PCR did not yield any sequences that indicated the presence of a AMOLAP version in our cDNA library. Subsequently, the full length coding region was amplified from A. annua trichome cDNA by PCR using Phusion polymerase (Finnzymes, Espoo, Finland) using primers 1 + 2 (Supporting Information Table S1). The full length AMOLAP coding region was cloned by primers 3 + 4. After confirmation of the correct sequences both the AMOHAP and AMOLAP coding region were isolated from pGEMT/AMOHAP and pGEMT/AMOLAP using BamHI and KpnI for cloning into the yeast expression vector pYEDP60 (Pompon et al., 1996), resulting in pYEDP60-AMOHAP and pYEDP60-AMOLAP.

For the cloning into a plant binary expression vector, the genes were first introduced into ImpactVectorC3.1 ( For this a KpnI site was introduced into C3.1, resulting in C3.1/KpnI. Hereto, two oligos (GATCCATTTCGGTACCAATTAGC and GGCCGCTAATTGGTACCGAAATG) were hybridized, kinase-treated and ligated into C3.1 digested with BamHI and NotI. The BamHI/KpnI fragment of pGEMT/AMOHAP and pGEMT/AMOLAP was isolated and ligated into the vector C3.1/KpnI between the CaMV35S promoter and Rbcs1 terminator. The resulting plasmids C3.1/AMOHAP and C3.1/AMOLAP were digested with AscI and PacI and the full gene sequence was cloned into the AscI and PacI site of the pBinPlus binary vector (van Engelen et al., 1995).

Identification and cloning of DBR2 from HAP and LAP chemotypes

We cloned a DBR2HAP from A. annua HAP chemotype (Bertea et al., 2006) and DBR2LAP was isolated from A. annua LAP chemotype from Iran (Table S2). Both DBR2 cDNA sequences were isolated by RT-PCR from cDNA constructed from RNA extracted from A. annua flowers isolated from either chemotype. For amplification of the DBR2 sequence, primers 5 + 6 were used (Table S1), of which sequences were based on the published DBR2 sequence from a HAP chemotype (Zhang et al., 2008). These primer sets were also able to amplify a DBR2 sequence from the LAP chemotype. The primers used introduce BamHI and NotI restriction sites which were used for cloning into ImpactVectorpIV1A_2.1 ( The resulting pIV1A_2.1/DBR2 was digested with AscI and PacI and the full gene was cloned into the AscI and PacI sites of the pBinPlus binary vector (van Engelen et al., 1995). DBR2 sequences identified here have been deposited in the GenBank database (KC505370, JX898526 and JX898527).

Identification and cloning of ALDH1 from HAP and LAP chemotypes

ALDH1 sequences were isolated by RT-PCR on cDNA from floral RNA isolated from the A. annua HAP and LAP chemotypes using primers 7 + 8 (Table S1). The sequences of these primers were based on the published ALDH1 sequence from a HAP chemotype (Teoh et al., 2009) and could amplify an ALDH1 sequence from both HAP and LAP cDNA. Inspection of the coding sequence revealed that there were no specific amino acid residue differences in ALDH1 from HAP and LAP (both identical to ALDH1 GenBank: FJ809784). The ALDH1 cDNA was cloned into pBinPlus binary vector (van Engelen et al., 1995).

Cloning of CYP71AV1 GFP reporter constructs

For the GFP fusion protein expression constructs (NtermAMOHAP:GFP, NtermAMOLAP:GFP, AMOHAP:GFP and AMOLAP:GFP): the N-terminal domains of AMOHAP (first 43 codons) and AMOLAP (first 50 codons) were amplified using primers 9 + 10 and 11 + 12, respectively. After digestion with BamHI and KpnI the fragments were cloned into C3.1/KpnI to form C3.1/NtermAMOHAP and C3.1/NtermAMOLAP.

The GFP coding sequence was amplified by PCR using primers 13 + 14 (Table S1) using pBin-Egfp as template ( After digesting with KpnI and NotI, GFP was subcloned into plasmid C3.1/NtermAMOHAP, C3.1/NtermAMOLAP, C3.1/AMOHAP and C3.1/AMOLAP to form 35S-NtermAMOHAP:GFP, 35S-NtermAMOLAP:GFP, 35S-AMOHAP:GFP and 35S-AMOLAP:GFP.

In order to remove the stop codon of AMOHAP and AMOLAP, and fuse them in-frame to the ATG start-codon of the GFP coding sequence in the 35S-AMOHAP:GFP and 35S-AMOLAP:GFP constructs, the 3′ends of the AMOLAP and AMOHAP were PCR amplified using Phusion DNA polymerase (Finnzymes) by primers 15 + 16, and 17 + 18, respectively.

The PCR products and plasmids 35S-AMOHAP:GFP and 35S-AMOLAP:GFP were digested with KpnI and EcoRI, followed by gel purification. The PCR products were ligated into the plasmids and transformed into Escherichia coli. Positive colonies were analysed and sequenced to confirm that inserts were correct.

Transient expression in leaves of Nicotiana benthamiana

Agro-infiltration for transient expression in leaves of Nicotiana benthamiana Domin was carried out as described (van Herpen et al., 2010). Briefly, individual Agrobacterium tumefaciens strains with different expression constructs (or empty vector as control) were co-infiltrated into N. benthamiana leaves using a syringe without needle. After 7 d of transient expression, leaves were harvested for chemical analysis. In each set of experiments the total dosage of A. tumefaciens between treatments was equalized by diluting with A. tumefaciens with empty vector where necessary. We note that leaves infiltrated with AmFH AMOLAP developed necrotic lesions, indicating that at this time certain compounds started to accumulate to toxic concentrations in the infiltrated leaves.

CYP71AV1 subcellular localization studies

For subcellular localization of the AMO:GFP fusion proteins, Arabidopsis thaliana (L.) Heynh protoplasts were isolated and transfected with expression constructs 35S-NtermAMOHAP:GFP, 35S-NtermAMOLAP:GFP, 35S-AMOHAP:GFP and 35S-AMOLAP:GFP, based on a published protocol (Yoo et al., 2007). The ER-YFP construct was used as reference for ER subcellular localization (Aker et al., 2006). After transfection, protoplasts were analysed using Carl-Zeiss Confocal Scanning Laser Microscopy, with exitation of GFP at 488 nm and YFP at 514 nm. The fluorescence was detected via a band pass filter (GFP: 505–530 nm, YFP: 535–590 nm). Chlorophyll was detected using a 650 nm long pass filter.

Analysis of nonvolatile metabolites by LC-QTOF-MS/MS

Seven days after agro-infiltration of N. benthamiana, the infiltrated leaves were harvested, immediately frozen in liquid nitrogen and then ground to a fine powder. From each infiltrated leaf 100 mg of powder was extracted in 300 μl methanol : formic acid (1000 : 1, v/v). Nonvolatile compounds from the infiltrated leaves were analysed by LC-QTOF-MS as described (van Herpen et al., 2010).

Data were processed using the protocol for untargeted metabolomics of plant tissues as described (De Vos et al., 2007; van Herpen et al., 2010). Briefly, LC-QTOF-MS data were analysed using Masslynx v4.0 (Waters) and processed using MetAlign v1.0 ( for baseline correction, noise elimination and subsequent spectral data alignment (De Vos et al., 2007). The processing parameters of MetAlign for LC-QTOF-MS data were set to analyse from scan numbers 60–2590 (corresponding to retention time 1.15–49.16 min) with a maximum amplitude of 25 000. After MetAlign processing, masses were clustered using the Multivariate Mass Spectra Reconstruction (MMSR) approach (Tikunov et al., 2005) to elucidate which mass signals originate from the same metabolite. The mass signal intensity differences between treatments were compared using Student's t-test. Mass-directed LC-QTOF-MS/MS analysis for further elucidation of metabolite identities was done on differential compounds with signal intensities > 500 ion counts per scan.

Quantification of artemisinin precursors by UPLC-MRM-MS

Targeted analysis of artemisinin precursors in agro-infiltrated Nicotiana benthamiana leaves was performed with a Waters Xevo tandem quadrupole mass spectrometer equipped with an electrospray ionization source and coupled to an Acquity UPLC system (Waters) as described (Kohlen et al., 2011) with some modifications. For details and instrument settings see Methods S1.

Analysis of volatile metabolites by GC-MS

Extracts were analysed by GC-MS using a gas chromatograph (7890A; Agilent, Amstelveen, the Netherlands) equipped with a 30-m × 0.25-mm i.d., 0.25-mm film thickness column with 5-m guard column (Zebron ZB5-MS; Phenomenex, Utrecht, the Netherlands) and a mass selective detector (model 5965c, Agilent). The GC was programmed at an initial temperature of 80°C for 1 min, with a ramp of 5°C min−1 to 235°C and then a ramp of 25°C min−1 to 280°C with a final time of 5 min. The injection port temperature was 250°C, and the He inlet pressure was controlled with electronic pressure control to achieve a constant column flow of 1.0 ml min−1. One microlitre of the extracts was injected in split mode with a split flow set at 9 ml min−1. Scanning was performed from 45 to 450 atomic mass units.

Glycosidase treatment

Viscozyme L (Sigma) was used as glycosidase treatment to hydrolyse hexose-conjugated compounds for subsequent quantification using GC-MS. Hereto, 200 mg infiltrated leaf material from each treatment was incubated in 1 ml citrate phosphate buffer, pH 5.4 containing 200 μl of Viscozyme L as previously described (van Herpen et al., 2010).

Glutathione conjugation assay

In vitro conjugation of metabolites to glutathione by glutathione transferase activity (GST) was performed as described (Liu et al., 2011). In brief, glutathione (GSH; 150 mM) in 7 μl potassium phosphate buffer (100 mM; pH 6.5), and 30 mM of artemisinin precursor (AAA, AAOH, AA, DHAAOH, DHAAA and DHAA) in 7 μl ethanol were added to 200 μl potassium phosphate buffer (100 mM; pH 6.5). The reaction was initiated by adding 7 μl of glutathione transferase (GST; 1 g l−1, in 100 mM KH2PO4 potassium phosphate buffer; pH 6.5) into the mixture. The controls were complete assay mixtures without GST enzyme or either of the substrates. After 15 min incubation at room temperature, samples were cooled to −20°C until LC-QTOF-MS analysis.


Comparison of artemisinin biosynthesis protein sequences from HAP and LAP chemotypes reveals only relevant differences for CYP71AV1

Because the difference between the HAP and LAP chemotypes must arise after the ADS step in the biosynthesis pathway, and because no expression differences were found for ADS genes in HAP and LAP chemotypes (Maes et al., 2011), here we focused on analysis of putative differences in biosynthesis genes downstream of ADS (e.g. CYP71AV1, DBR2, ALDH1) in search of an explanation for the two different chemotypes of A. annua. For this purpose CYP71AV1, DBR2 and ALDH1 were isolated from A. annua HAP and LAP chemotypes and the encoded protein sequences compared.


Analysis of the different CYP71AV1 sequences that have been deposited in GenBank shows the occurrence of two major types of CYP71AV1, encoding two proteins which differ by a seven amino acids extension at the N-terminus of the protein (Fig. 2). The long version of CYP71AV1 (which we refer to as AMOLAP) has been isolated from A. annua Tanzania (Sandeman seed), which is a LAP chemotype (Ro et al., 2006). The other long version of CYP71AV1 (which we refer to as AMOLAP.1) was isolated from a different A. annua LAP chemotype (Kim et al., 1992; S-U. Kim, pers. comm.). Two versions of the CYP71AV1 (here referred to as AMOHAP and AMOHAP.1) were cloned from two different HAP chemotypes (Bertea et al., 2006; Teoh et al., 2006). The alignment of the AMOHAP and AMOLAP variants shows that none of the other single amino acid substitutions between the different AMOLAP and AMOHAP sequences are specific to the long or the short version of CYP71AV1 and therefore likely do not play a role in determining the LAP or HAP chemotypes (Fig. 2). RACE-PCR was used to analyse multiple CYP71AV1 5′ sequences amplified from RNA isolated from a HAP chemotype and only AMOHAP 5′ sequences were found (Fig. S1). Variation in the 5′ untranslated region (nt 56–60) could be an indication that two different alleles of AMOHAP are present in this chemotype, both translating into the short AMOHAP. The RACE sequence data were consistent with the recently published sequence of the CYP71AV1 promoter cloned from an A. annua HAP chemotype (Wang et al., 2011).

Figure 2.

Alignment of the AMOLAP and AMOHAP deduced amino acid sequences. All amino acid sequences used in the alignment were retrieved from GenBank (GenBank number given in front of each sequence). Two of the sequences were from confirmed LAP chemotypes: ABB82944, referred to as AMOLAP (Ro et al., 2006) and ACF74516, referred to as AMOLAP.1 (Kim et al., 1992). Two of the sequences were from confirmed HAP chemotypes: AFP19100, referred to as AMOHAP (Bertea et al., 2006) and ABC41927, referred to as AMOHAP.1 (Teoh et al., 2006). For the other sequences deposited in GenBank the chemotype of the plant from which they were isolated was not given.


We cloned DBR2 from the A. annua HAP chemotype (here referred to as DBR2HAP.1) and the sequence we obtained was similar to the recently described DBR2 (here referred to as DBR2HAP; Zhang et al., 2008), with the exception of a one amino acid difference (Fig. 3). Using primers based on DBRHAP we isolated two variants of DBR2 from an A. annua LAP chemotype (see 'Materials and Methods'). Five of the 11 clones showed few amino acid differences with the published DBR2HAP (here referred to as DBR2LAP.1). However, in the remaining six clones (here referred to as DBR2LAP), the encoded protein sequence showed a number of amino acid residue differences with DBR2HAP, including two additional amino acids in position 295 (Fig. 3).

Figure 3.

Alignment of the DBR2LAP and DBR2HAP deduced amino acid sequences. All amino acid sequences used in the alignment were retrieved from GenBank: DBR2LAP (JX898527), DBR2LAP.1 (KC505370), DBR2HAP (ACH61780) and DBR2HAP.1 (JX898526). DBR2LAP and DBR2LAP.1 were cloned from LAP, DBR2HAP was cloned from HAP (Zhang et al., 2008) and DBR2HAP.1 was cloned from HAP (Bertea et al., 2006). Alignment shows that one variant of DBR2LAP.1 is similar to DBR2HAP with only two AA differences. The second variant of the DBR2LAP has a two amino acids insertion at the C-terminus. No difference was detected in in planta activity of DBR2HAP and DBRLAP (Figs 5b,c and Supporting Information Fig. S3b).


Cloning of ALDH1 from A. annua has been described (Teoh et al., 2009). We isolated ALDH1 from both the HAP and LAP chemotypes. Alignment of the AA-sequence showed no differences between the two proteins (data not shown).

Both AMOHAP and AMOLAP are localized to the endoplasmic reticulum

Because AMOLAP and AMOHAP only consistently differ in their N-terminal amino acid residues which supposedly encode the ER anchoring domain (Fig. 2), we investigated whether this difference causes altered subcellular targeting or difference in protein stability. Expression constructs encoding either full-length protein-GFP fusions or truncated N-terminal domain-GFP fusions of AMOLAP and AMOHAP were transiently expressed in A. thaliana protoplasts. Confocal microscopy of Arabidopsis protoplasts co-transfected with the full length AMOLAP protein fused to GFP (AMOLAP:GFP) showed co-localization of the GFP fluorescence signal with the fluorescence signal of the ER marker (ER:YFP; Fig. 4a). Similarly, full length AMOHAP protein fused to GFP (AMOHAP:GFP) also showed co-localization with the ER marker (ER:YFP; Fig. 4c). In addition, the truncated N-terminal portion of AMOLAP and AMOHAP were fused to GFP. When transfected into Arabidopsis protoplasts the NtermAMOLAP:GFP and NtermAMOHAP:GFP both showed co-localization with the ER:YFP ER marker (Fig. 4c,d). Localization experiments with 35S expression constructs may lead to artefacts such as cytosolic localization when ER import is saturated, however this was not observed in these experiments. Combined, the results demonstrate that AMOLAP and AMOHAP do not differ in subcellular targeting, as both proteins localize to the ER. Although the fluorescence signal varies between transfection assays, there were indications that AMOLAP may be more stable. For instance we do find a higher fluorescence signal for NtermAMOLAP:GFP than for the NtermAMOHAP:GFP.

Figure 4.

Subcellular localization of AMOLAP and AMOHAP. Confocal microscopy analysis of Arabidopsis protoplasts co-transfected with (a) AMOLAP:GFP + ER:YFP; (b) NtermAMOLAP:GFP + ER:YFP; (c) AMOHAP:GFP + ER:YFP; (d) NtermAMOHAP:GFP + ER:YFP. Artificial colours were given to GFP fluorescence (green), YFP fluorescence (red) and auto fluorescence of chloroplasts (purple). Merging of the pictures results in a yellow colour for GFP-YFP overlap. Bars, 10 μm.

Different product profiles in planta from AMOLAP and AMOHAP

Both the AMOLAP and the AMOHAP enzymes have been characterized in a yeast expression system and both were shown to be able to produce AAOH, AAA and AA from AD (Ro et al., 2006; Teoh et al., 2006). However, a direct comparison of variants AMOLAP and AMOHAP in the same expression system has not been performed until now. The different in planta expression studies using heterologous plant hosts (tobacco and N. benthamiana) have only been reported for the longer AMOLAP (van Herpen et al., 2010; Zhang et al., 2011). To test the effect of the seven AA extension of the AMOLAP protein we made two expression constructs, both based on the AMOHAP sequence but in one construct we introduced the seven AA extension to the protein sequence as found in AMOLAP, thus limiting the difference between the two forms to the N-terminal extension. The activity of these AMOLAP and AMOHAP genes was subsequently compared in planta by co-expression with ADS, using transient expression in N. benthamiana leaves. To achieve high concentrations of artemisinin precursor production, ADS was expressed with a mitochondrial targeting signal, and overexpression was combined with a mitochondrial targeted FPS and a truncated, cytosolic form of HMGR. ADS, FPS and HMGR were combined into a single 2A expression construct (AmFH) as described before (van Herpen et al., 2010). Each expression construct (AmFH and AMOLAP or AMOHAP) was introduced into A. tumefaciens and N. benthamiana leaves were infiltrated with AmFH AMOLAP or AmFH AMOHAP.

Particularly with AmFH AMOLAP, infiltrated leaves developed symptoms of necrosis around 7 d post infiltration (Fig. S2), suggesting the production of a toxic compound. Necrosis symptoms were stronger in leaves infiltrated with AmFH AMOLAP than with AmFH AMOHAP, suggesting that the products of both treatments may not be the same. Because necrosis started to appear after 7 d, in all our experiments the leaves were harvested at day 7 instead of day 10 after infiltration, as previously done (van Herpen et al., 2010).

Analysis of free products

In order to quantify the products from the ADS and AMOLAP/HAP transient enzyme activity in the infiltrated N. benthamiana, leaves were extracted with aqueous methanol for UPLC-MRM-MS analysis. Intriguingly, the distribution over the entire product range was different in leaves infiltrated with AmFH AMOLAP and AmFH AMOHAP (Table 1). Leaves infiltrated with AmFH AMOLAP produced predominantly AA, while in leaves infiltrated with AmFH +AMOHAP AA concentrations were 50-fold lower. However, leaves infiltrated with AmFH AMOHAP contained three-fold higher concentrations of AAOH and AAA than leaves expressing AmFH AMOLAP (Table 1).

Table 1. Unconjugated artemisinin precursors produced in Nicotiana benthamiana as identified and quantified by UPLC-MRM-MS
  1. AAOH, artemisinic alcohol; AAA, artemisinic aldehyde; AA, artemisinic acid; DHAAOH, dihydroartemisinic alcohol; DHAAA, dihydroartemisinic aldehyde; DHAA, dihydroartemisinic acid; ND, not detectable.

  2. a

    The values for DHAAA are shown in brackets as they represent peak intensities and not concentrations. Results are means ± SD of three co-infiltrated leaves.

AAOH16 709 ± 397746 946 ± 56923596 ± 124716 554 ± 3233
AAA5501 ± 148619 564 ± 6314781 ± 2463418 ± 890
AA3969 ± 139174 ± 1553 ± 16ND
DHAAOH1821 ± 576597 ± 1987 972 ± 15 01457 289 ± 9455
DHAAAa(5966 ± 1646)(8067 ± 1341)(220 347 ± 65 373)(124 787 ± 26 145)
DHAA17 ± 8ND838 ± 51725 ± 9

Also DHAAOH, DHAAA and DHAA were detected in the AmFH AMOLAP leaf samples (Table 1), suggesting the presence of an endogenous N. benthamiana enzyme with carbon double bond reducing activity (catalysing the conversion of AAA to DHAAA just as DBR2 in A. annua), and enzymes similar to A. annua RED1 (catalysing the formation of DHAAOH from DHAAA) and A. annua ALDH1 (catalysing the conversion of DHAAA to DHAA; Fig. 1). Free DHAAOH concentrations were higher in AmFH AMOLAP compared to AmFH +AMOHAP infiltrated leaves, suggesting that formation of DHAAOH is not directly related to the free DHAAA, AAOH or AAA concentrations in leaves, which were lower in AmFH +AMOLAP. DHAA was only detected in AmFH AMOLAP infiltrated leaves.

Analysis of glycosylated products

Previous results showed that most of the products of the ADS and AMOLAP activity in N. benthamiana agro-infiltration are present as glycosylated conjugates, mainly of AA (van Herpen et al., 2010). Therefore, leaf material was also analysed by LC-QTOF-MS. Nicotiana benthamiana leaves co-expressing AmFH AMOLAP indeed contain AA-12-β-diglucoside, as previously reported (van Herpen et al., 2010). However, in addition several other AA-glycoside conjugates were detected, including conjugates with additional hexose units as well as malonylated hexoses. Also for the glycosylated products the distribution over the entire product range was different between AmFH AMOLAP and AmFH AMOHAP (Fig. 5a, Table 2). Leaves infiltrated with AmFH AMOLAP produced more AA conjugates, while leaves infiltrated with AmFH AMOHAP produced more AAOH conjugates (Table 2, Fig. S3a). For both treatments several DHAAOH and DHAA conjugates with hexose and malonyl groups were also detected, but no DHAAA conjugates (Table 2). Table 2 shows the mass fragmentation profiles of the detected products and their putative identification. MS/MS analysis was used to further confirm product identity and an example of the identification of one of the DHAA-hexose conjugates is shown in Fig. S5.

Table 2. Conjugated artemisinin precursors produced in agro-infiltrated Nicotiana benthamiana leaf extracts fragmentation
Ret (min)Detected mass (Da)aMS/MS fragmentsMol form∆Mass (ppm)Putative IDIntensityb
  1. AAOH, artemisinic alcohol; AAA, artemisinic aldehyde; AA, artemisinic acid; DHAAOH, dihydroartemisinic alcohol; DHAAA, dihydroartemisinic aldehyde; DHAA, dihydroartemisinic acid; GSH, glutathione; FA, formic acid adduct; Ret (min), retention time, in minutes; Mol form, molecular formula of the metabolite; ∆Mass (ppm), deviation between the detected mass and real accurate mass, in ppm; Putative ID, putative identification of metabolite; ND, not detectable.

  2. a

    Detected mass (Da): The mass was detected in negative mode of LC-QTOF-MS.

  3. b

    Peak intensities are the mean ± SD of three agro-infiltrated leaves.

  4. c

    Hex, compound conjugated with hexose; Mal, compound conjugated with malonate; (I–III), different isobaric forms (i.e. identical accurate mass, but different retention times).

  5. d

    The ions of a number of representative GSH adducts in the negative ion mode (Dieckhaus et al., 2005).

  6. e

    AA-Hex2, The structure of artemisinic acid-12-β-diglucoside was confirmed by NMR (van Herpen et al., 2010). Nonvolatile metabolites with mass intensity > 500 in LC-QTOF-MS, which were significantly increased in leaves agro-infiltrated with AmFH AMOLAP, AmFH AMOHAP, AmFH AMOLAP DBR2, or AmFH AMOHAP DBR2 were targeted for analysis by LC-QTOF-MS/MS.

27.85543.2793381[M-Hex-H]C27H44O111.3AAOH-Hex2319 ± 573715 ± 610264 ± 511753 ± 823
28.87629.2810585[M-CO2-H],543[M-Mal-H],381[M-Mal-Hex-H]C30H46O140AAOH-Hex2-Mal (I)c1096 ± 2882239 ± 124861 ± 431378 ± 246
29.35629.2810585[M-CO2-H],543[M-Mal-H],381[M-Mal-Hex-H]C30H46O140AAOH-Hex2-Mal (II)606 ± 3724 251 ± 76627 ± 10716 215 ± 5097
29.69629.2810585[M-CO2-H],543[M-Mal-H],381[M-Mal-Hex-H]C30H46O140AAOH-Hex2-Mal (III)439 ± 1981109 ± 104ND515 ± 313
23.80542.2536272,254,210,179,143,128dC25H41N3O8S1.0AAA-GSH-H2O2462 ± 3237891 ± 65130 ± 221970 ± 395
24.02765.3181719[M-H],395[M-2Hex-H],233[M-3Hex-H]C34H54O190.1(AA-Hex3) FA21 190 ± 1730612 ± 35100 ± 80ND
27.88557.2598395[M-Hex-H],233[M-2Hex-H]C27H42O120AA-Hex2 (I)e5403 ± 1218917 ± 71121 ± 61371 ± 186
28.48557.2598395[M-Hex-H],233[M-2Hex-H]C27H42O120AA-Hex2 (II)3773 ± 882110 ± 2128 ± 15ND
29.38643.2602599[M-CO2-H],395[M-Mal-Hex-H],233[M-Mal-2Hex-H]C30H44O150.5AA-Hex2-Mal (I)16 021 ± 3202972 ± 1621652 ± 3271783 ± 139
29.781287.5282643[M-H],599[M-CO2-H],395[M-Mal-Hex-H],233[M-Mal-2Hex-H]C60H88O304.5AA-Hex2-Mal ([2M-H])2611 ± 1114ND129 ± 1382 ± 9
24.86753.3545707[M-H],545[M-Hex-H],383[M-2Hex-H],221[M-3Hex-H]C34H58O180.9(DHAAOH-Hex3) FA145 ± 4685 ± 814 248 ± 67011 573 ± 206
28.80545.2962383[M-Hex-H],221[M-2Hex-H]C27H46O111.8DHAAOH-Hex2 (I)343 ± 89194 ± 4614 321 ± 261714 968 ± 5018
30.27631.2966587[M-CO2-H],545[M-Mal-H],383[M-Hex-H],221[M-2Hex-H]C30H48O141.8DHAAOH-Hex2-Mal (I)479 ± 116266 ± 1224 475 ± 12024 394 ± 181
30.75631.2966587[M-CO2-H],545[M-Mal-H],383[M-Mal-Hex-H],221[M-Mal-2Hex-H]C30H48O140DHAAOH-Hex2-Mal(II)ND18 ± 2855 ± 156919 ± 378
23.87767.3338721[M-H],397[M-2Hex-H],235[M-3Hex-H]C34H56O191.6(DHAA-Hex3) FA1237 ± 110ND23 356 ± 1554861 ± 185
27.28559.2755397[M-Hex-H],235[M-2Hex-H]C27H44O120.3DHAA-Hex2NDND5188 ± 2039120 ± 54
26.58645.2759601[M-CO2-H],397[M-Mal-Hex-H],235[M-Mal-2Hex-H]C30H46O150DHAA-Hex2-Mal (I)1912 ± 2445539 ± 174236 ± 141645 ± 256
28.92645.2759601[M-CO2-H],397[M-Mal-Hex-H],235[M-Mal-2Hex-H]C30H46O150DHAA-Hex2-Mal (II)ND20 ± 61303 ± 52876 ± 35
29.83645.2759601[M-CO2-H],397[M-Mal-Hex-H],235[M-Mal-2Hex-H]C30H46O150DHAA-Hex2-Mal (III)3424 ± 54964 ± 8NDND
Figure 5.

Artemisinic acid (AA) and dihydroartemisinic acid (DHAA) related compounds in leaves of Nicotiana benthamiana agro-infiltrated with artemisinin biosynthesis genes as identified by LC-QTOF-MS. (a) Agro-infiltrated leaves with AmFH AMOLAP and AmFH AMOHAP. For all products identified by LC-QTOF-MS see Table 2. (b) Agro-infiltrated leaves with AmFH AMOLAP/HAP DBR2LAP/HAP. For all products identified by LC-QTOF-MS see Supporting Information Table S3. (c) Agro-infiltrated leaves with AmFH AMOLAP/HAP DBR2LAP/HAP ALDH1. For all products identified by LC-QTOF-MS see Table S3. Data shown with peak intensity which analysed by LC-QTOF-MS. Peak intensities are the mean of three agro-infiltrated leaves.

In order to quantify the concentrations of glycosylated products, samples were treated with a mix of glycosidases (Viscozyme L) and deglycosylated products were quantified using GC-MS. Note that the Viscozyme treatment only cleaves hexose conjugates but not malonylated hexose conjugates (Fig. S4). Results show that leaves infiltrated with AmFH AMOLAP contained c. 40 mg kg−1 FW of AA (consistent with the previously reported 39.5 mg kg−1 FW of AA; van Herpen et al., 2010), while the sensitivity of the GC-MS was not sufficient to detect any AA in leaves infiltrated with AmFH AMOHAP (Table 3). GC-MS analysis after Viscozyme treatment confirmed that AAOH was the major glycosylated product in leaves infiltrated with AmFH AMOHAP (as was suggested by Table 2) at 24 mg kg−1 FW.

Table 3. Artemisinin precursors in Nicotiana benthamiana agro-infiltrated with artemisinin biosynthetic pathway genes
  1. Agro-infiltrated leaves were treated with glycosidase (Viscozyme L.) and hydrolysed metabolites extracted and analysed by GC-MS.AAOH, artemisinic alcohol; AAA, artemisinic aldehyde; AA, artemisinic acid; DHAAOH, dihydroartemisinic alcohol; DHAAA, dihydroartemisinic aldehyde; DHAA, dihydroartemisinic acid; ND, not detectable.

  2. Results are means ± SD of three co-infiltrated leaves.

AAOH8.1 ± 1.624.0 ± 3.55.1 ± 0.59.4 ± 2.1
AAA1.6 ± 0.11.6 ± 0.1NDND
AA39.9 ± 9.8NDNDND
DHAAOH1.6 ± 0.22.0 ± 0.242.9 ± 14.922.8 ± 8.6
DHAAANDND4.0 ± 2.31.3 ± 0.6

No difference in DBR2 activity from HAP and LAP A. annua chemotypes

We compared the activity of the two variants of DBR2 by comparing product profiles of N. benthamiana leaves infiltrated with AmFH AMOLAP in combination with either DBR2HAP or DBR2LAP. In addition we tested the two DBR2 variants in combination with AmFH AMOHAP. Analysis of the conjugated products show that there is no difference in product profile between DBR2HAP and DBR2LAP (Figs 5b, S3b, Table S3), indicating that the two forms of DBR2 do not differ in enzymatic activity. The co-infiltration with DBR2 relieved the necrosis symptoms caused by expression of AmFH AMOLAP or AmFH AMOHAP alone (Fig. S2), suggesting that additional DBR2 enzyme activity lowered the concentration(s) of the product(s) that cause necrosis. Product analysis in leaves agro-infiltrated with AmFH AMOLAP DBR2 or AmFH AMOHAP DBR2 showed that DBR2 activity resulted in a significant increase in DHAAOH, DHAAA and DHAA concentrations (Table 1) and this is also clear from LC-QTOF-MS analysis that shows a strong increase in DHAAOH and DHAA conjugates to hexose and malonyl groups (Table 2).

The analysis of deglycosylated extracts by GC-MS confirmed that co-expression of DBR2 increased the concentrations of DHAAOH, DHAAA and DHAA at the expense of AAOH, AAA and AA concentrations (Table 3). The total yield of DHAA in leaves agro-infiltrated with AmFH AMOLAP DBR2 as released by glycosidase treatment was c. 7.3 mg kg−1 FW, while DHAA in leaves agro-infiltrated with AmFH AMOHAP DBR2 was below the level of detection by GC-MS. Combined, these results show that DBR2 further enhances the double bond reduction of the CYP71AV1 products that is also already catalysed by endogenous tobacco reductase activity. In addition, endogenous tobacco glycosyl and malonyl transferases modify these double-bond-reduced products which leads to DHAAOH and DHAA conjugates.

Increased DHAA and AA by combining ADS, AMO and DBR2 with ALDH1

As already described, no differences were found between the ALDH1 protein sequence from LAP and HAP A. annua chemotypes. To test how the addition of ALDH1 activity affects the product profile of the artemisinin HAP and LAP biosynthesis pathway, leaves were agro-infiltrated with AmFH AMOLAP DBR2 + ALDH1 or AmFH AMOHAP DBR2 + ALDH1. After 7 d leaves were extracted and products were profiled by LC-QTOF-MS. The concentrations of conjugated DHAAOH products significantly decreased when ALDH1 was added to AmFH AMOLAP DBR2 (Table S3, Fig. S3b), coinciding with a substantial increase in glycosylated AA and DHAA product concentrations (Fig. 5c). This suggests that ALDH1 may be more efficient in the conversion of AAA to AA and DHAAA to DHAA than AMOLAP and AMOHAP as already suggested by the work of Teoh (Teoh et al., 2009). Although the concentration of the presumed direct precursor of artemisinin (DHAA) was substantially increased by ALDH1 (c. 13-fold by adding ALDH1 to AmFH AMOLAP DBR2 and c. 110-fold by adding ALDH1 to AmFH AMOHAP DBR2), no artemisinin could be detected in N. benthamiana by UPLC-MRM-MS (Tables 1, 4).

Table 4. Artemisinic acid (AA) and dihydroartemisinic acid (DHAA) produced in agro-infiltrated Nicotiana benthamiana as identified and quantified by UPLC-MRM-MS
(ng kg−1 FW)AmFH + AMOLAP + DBR2 + ALDH1AmFH + AMOHAP + DBR2 + ALDH1
  1. Results are means ± SD of three co-infiltrated leaves.

AA8836 ± 17302589 ± 563
DHAA10 792 ± 3412756 ± 547

Qualitative effects on product profile by AMOLAP dosage

The comparison of the total product concentrations produced by the reconstituted pathway with AMOLAP or AMOHAP suggests that AMOLAP has a higher enzyme activity than AMOHAP, in combination with a different product profile (Fig. 5a, Table 2). This difference could not be related to different subcellular localization (Fig. 4). To test if differences in relative enzyme activity within the pathway can affect the product profile in a qualitative way, we tested the effect of different dilutions of AMOLAP in combination with the rest of the biosynthesis pathway (ADS AMOLAP DBR2HAP). This was achieved by diluting the Agrobacterium strain carrying the AMOLAP expression construct with a suspension of an Agrobacterium strain carrying an empty expression vector to keep the total Agrobacterium dosage for infiltration the same. Results show that reduction of the AMOLAP agro-infiltration dosage to 1/2 had little effect, but that dilution up to 1/10 strongly decreased DHAA-glycoside production (Figs 6a, S3c; Table S4). The AA-glycoside concentrations were also reduced but to a much lower extent.

Figure 6.

Artemisinic acid (AA) and dihydroartemisinic acid (DHAA) conjugated compounds in leaves of Nicotiana benthamiana agro-infiltrated with AMOLAP/HAP and different dosage of DBR2 and ALDH1. (a) Comparison AA and DHAA conjugated compounds in agro-infiltrated leaves with dilution of AMOLAP/HAP. For all products identified by LC-QTOF-MS see Supporting Information Table S4. (b) Comparison AA and DHAA conjugated compounds in agro-infiltrated leaves with dilution of DBR2. For all products identified by LC-QTOF-MS, see Table S5. (c) Comparison AA and DHAA conjugated compounds in agro-infiltrated leaves with dilution of ALDH1. For all products identified by LC-QTOF-MS see Table S5. (d) Comparison AA and DHAA conjugated compounds in agro-infiltrated leaves with dilution of DBR2 and ALDH1. For all products identified by LC-QTOF-MS see Table S5. Data shown with peak intensity which analysed by LC-QTOF-MS. Data represent peak intensities for each of the compounds in LC-QTOF-MS analysis.

More AAA-related glutathione-conjugate from AMOHAP than from AMOLAP

Recently we described the reconstruction of the biosynthetic pathway of costunolide in N. benthamiana (Liu et al., 2011) in which it was shown that the exocyclic carbon double bond of costunolide conjugates to glutathione (GSH). Because some of the products of ADS (AD) and AMOLAP/HAP enzyme activities also contain such an exocyclic double bond (AAA), but lack the hydroxyl or acid group used for glycosylation in AAOH and AA, we specifically looked for GSH conjugates of artemisinin intermediates in N. benthamiana leaves agro-infiltrated with the biosynthetic pathway genes. A putative GSH-conjugated compounds was detected by LC-QTOF-MS, both in negative mode (m/= 542.25) and positive mode (m/= 544.24; Fig. S6). The concentration of the GSH-conjugate (m/= 542.25) was higher in AmFH AMOHAP than in AmFH AMOHAP DBR2 agro-infiltrated leaves (Table 2). We tested the artemisinin biosynthetic pathway intermediates (AAA, AAOH, DHAAOH, AA, DHAA and DHAAA) in in vitro reactions for spontaneous or glutathione-S-transferase (GST) driven GSH conjugation. Only AAA formed an AAA–GSH conjugate similar to that extracted from agro-infiltrated leaves expressing the pathway genes (m/= 526.22; Figs S6, S7). The mass of the major GSH conjugate formed in planta is 18 Da higher, suggesting an additional two protons and one oxygen atom, which could be explained by hydroxylation of the endocyclic double bond in AAA (Fig. S6). The concentration of the putative AAA glutathione conjugate was higher in leaves infiltrated with AmFH AMOHAP than in leaves infiltrated with AmFH AMOLAP (Table 2). Dilution of AMOLAP resulted in an increase in the concentration of the AAA glutathione conjugate, although not reaching the level of the full dosage of AMOHAP (Fig. 7).

Figure 7.

Artemisinic aldehyde (AAA) conjugated compounds in leaves of Nicotiana benthamiana agro-infiltrated with dilution of AMOLAP/HAP. Data shown with peak intensity which analysed by LC-QTOF-MS. Data represent peak intensities for each of the compounds in LC-QTOF-MS analysis.

Effects on product profile by DBR2 or ALDH1 dosage

We also tested the effect of DBR2 and ALDH1 gene dosage on the DHAA : AA related product ratio by testing dilutions of DBR2 and ALDH1 in combination with AMOLAP or AMOHAP. Product analysis by LC-QTOF-MS of glycosylated products showed that with full DBR2 and ALDH1 agro-infiltration dosage the ratio of DHAA : AA was more skewed towards the DHAA branch of the pathway. However, when the DBR2 infiltration dosage was diluted 10-fold, the ratio of DHAA : AA decreased, shifting the pathway activity more towards the AA branch, with relatively little effect on the total product concentration (Figs 1, 6b).

By contrast, 10-fold dilution of ALDH1 agro-infiltration dosage resulted in a strong decrease in glycosylated AA conjugates, while the glycosylated DHAA conjugates were hardly affected (Fig. 6c), suggesting that ALDH1 has a preference for the DHAAA substrate over the AAA substrate (Fig. 1).

When both DBR2 and ALDH1 agro-infiltration dosage were diluted 10-fold, the DHAA : AA related products ratio came close to one, with slight preference for the DHAA branch of the pathway with AMOHAP and slight preference for the AA branch of the pathway with AMOLAP (Fig. 6d). In combination, these results suggest that the agro-infiltrated leaf ‘chemotype’ is determined by a combination of the AMOLAP/AMOHAP catalytic effectivity in combination with especially DBR2 gene dosage (Figs 6, S3d; Table S5).


Here we have compared the different proteins from the artemisinin biosynthesis pathway encoded by genes isolated from high- and low-artemisinin producing A. annua chemotypes. The expression concentrations of artemisinin biosynthesis genes (ADS, AMO, DBR2 and ALDH1) were recently analysed in A. annua HAP and LAP chemotypes showing that the chemotype identities of A. annua did not correlate with differences in expression level of these genes in the absence of stress (Maes et al., 2011). Nevertheless, total product yield in 5-wk-old A. annua leaves is higher in LAP than HAP chemotypes (see Table S6, Maes et al., 2011). Therefore, differences in protein activity rather than differences in gene expression level may account for differences in chemotype. In the present work, two DBR2 variants were identified (Fig. 3) but characterization of the in planta activity showed that they have equal activity (Figs 5b, S3b). No difference was found for the ALDH1 amino acid sequence from A. annua HAP and LAP chemotypes. Therefore, the only consistent difference in artemisinin biosynthesis proteins in the branched pathway after ADS is in the AMOHAP and AMOLAP from the A. annua HAP and LAP chemotypes, respectively (Fig. 2).

In the LAP chemotype CYP71AV1 is consistently seven amino acids longer than CYP71AV1 from the HAP chemotype (Fig. 2). When expressed together with ADS (+FPS HMGR) the total product yield (based on the cumulative concentrations of AAOH, AAA, AA, DHAAOH and DHAA released from conjugated products) in leaves co-infiltrated with AMOLAP was approximately twice as high as in leaves co-infiltrated with AMOHAP (Table 3). This suggests a lower enzyme activity for AMOHAP than for AMOLAP (Fig. 5a). AMOHAP and AMOLAP seem to anchor equally well to the ER membrane (Fig. 4). The observed difference in efficiency may be caused by a different stability of the two proteins. Nevertheless, dilution of AMOLAP gene dosage in agro-infiltration experiments did not fully mimic the AMOHAP phenotype (Figs 6a, S3c). In theory, the lower efficiency of AMOHAP could also be due to a less efficient interaction with cytochrome P450 reductase (CPR; Lengler et al., 2006). Alignment of the AMOLAP and AMOHAP protein sequences with other related sesquiterpene oxidases like germacrene A oxidase (GAO) from several different Asteraceae (Nguyen et al., 2010) shows that the N-terminal extension in AMOLAP is the exception. However, the CPR interaction domain does not map to the N-terminus (Sevrioukova et al., 1999). Finally, it may be that the substrate entry or release of AMOHAP is compromised.

Variable ‘chemotype’ of leaves expressing artemisinin biosynthesis genes

We defined the ‘chemotype’ of the N. benthamiana agro-infiltrated leaves based on the DHAA and AA glycoside conjugates (Fig. 5). We note that the peak intensity is a relative quantification as detection efficiency (e.g. ionization) may differ between compounds. However, comparison of the relative quantifications based on peak intensities from LC-QTOF-MS and the absolute quantification of DHAA and AA products by UPLC-MRM-MS or GC-MS (Tables 1-3) indicates a good correlation between the two analytical techniques. Expression of ADS AMOHAP results in a HAP chemotype (more DHAA than AA) and expression of ADS AMOLAP in a LAP chemotype (more AA than DHAA; Fig. 5a). However, when DBR2 is included, the chemotype for both the combination ADS AMOHAP DBR2 and ADS AMOLAP DBR2 is changed to a HAP chemotype with relatively higher DHAA concentration (Fig. 5b). In all of these combinations, the overall yield of DHAA was always lower for AMOHAP (c. 15-fold). When ALDH1 was included, the chemotype remained that of HAP, but the relative yield of the gene combinations that include AMOHAP was substantially increased and was now comparable to that of the gene combinations with AMOLAP (Fig. 5c). Because in the agro-infiltration assay in N. benthamiana addition of new genes to the pathway leads to a reduction in the relative dosage of the other genes infiltrated into the leaf, we tested whether the relative gene dosage affects the product profile. Lowering the dosage of AMOLAP with similar dosage of ADS and DBR2 resulted in a profile more closely related to that of AMOHAP, but also resulted in lower product yield of AA and DHAA conjugates (Fig. 6a). Lowering the relative dosage of DBR2 resulted in a reversion of the infiltrated leaf chemotype from HAP to LAP (Fig. 6b), while lowering the relative dosage of ALDH1 did not change the chemotype, but did decrease AA conjugates more than the DHAA conjugates (Fig. 6c). When both DBR2 and ALDH1 dosage were reduced, AMOHAP resulted in a more HAP related chemotype, while the combination with AMOLAP resulted in a more LAP related chemotype (Fig. 6d).

Nicotiana benthamiana enzyme activities both enhance and limit the DHAA ‘chemotype’

AMOHAP seems to be less efficient in the conversion to AA, presumably resulting in an early release of AAA. Indeed free AAA and an AAA glutathione conjugate were present at higher concentrations when the pathway was expressed in combination with AMOHAP than with AMOLAP (Table 1, Fig. 7). The AAA released by AMOHAP, in particular, may subsequently be substrate for double-bond reductases (endogenous from N. benthamiana or the co-expressed DBR2) to produce DHAAA and DHAAA derived products (Tables 13). The endogenous DBR2-like activity is far from saturating, as introduction of A. annua DBR2 greatly enhanced the conversion to DHAAOH, DHAAA and DHAA (Table 1). The early release of AAA by AMOHAP also reveals the activity of an endogenous N. benthamiana reductase, similar to the A. annua aldehyde reductase (RED1) that catalyses the conversion of DHAAA to DHAAOH (Bertea et al., 2005; Rydén et al., 2010). Indeed, the presence of a RED1-like activity in tobacco was previously demonstrated through feeding experiments: tobacco leaves supplied with DHAAA and AAA form DHAAOH and AAOH, respectively (Zhang et al., 2011). This suggests that in N. benthamiana the elevated pools of AAOH and DHAAOH (and glycosylated derivatives) may be the result of a reverse product flux from AAA back to AAOH and DHAAA to DHAAOH (Fig. 1). If this is the case, the affinity of AMOHAP for the AAOH substrate may be underestimated.

DHAA (conjugates) were detected in the leaves infiltrated with AmFH AMOLAP. This suggests the presence of an endogenous aldehyde dehydrogenase, similar to ALDH1 from A. annua, which can produce DHAA from DHAAA (Teoh et al., 2009). Alternatively, the low concentrations of DHAA could be the result of AMOLAP catalysing oxidation of DHAAA. Work in yeast showed that AMOLAP is far less effective in the conversion of DHAAA to DHAA than in the conversion of AAA to AA (Teoh et al., 2009). If we assume that also AMOHAP catalyses this step less effectively, expression of ALDH1 in a HAP background should further enhance the production of DHAA in the HAP chemotype. Indeed, our data show that product flow towards DHAA increased c. 110-fold when expression of AmFH AMOHAP DBR2 was combined with ALDH1 (Table 4).

Conjugating activities limit precursor pool for artemisinin production

Most of the products formed upon agro-infiltration of the artemisinin pathway genes are present in the form of hexose and/or hexose/malonyl conjugates. This explains why a previous heterologous expression study in tobacco detected only limited concentrations of free AD, AAOH, DHAAOH and no AA or DHAA (Zhang et al., 2011). The deglycosylation experiments show that the conjugated forms accumulate to concentrations up to 4–10-fold higher than the corresponding free forms (Tables 1-3). The glycosylation/conjugation is most likely a response to the production of potentially toxic compounds, as leaves expressing AmFH AMOLAP/HAP showed signs of necrosis (Fig. S2). Necrosis was stronger in leaves containing higher concentrations of free (and conjugated) AA and indeed presence of DBR2 reduced both the necrotic phenotype and the free and conjugated AA concentrations (Fig. S2, Table 1). In addition to the AA/DHAA glycosides, AAA glutathione conjugates were detected. All these conjugating activities limit accumulation of DHAA, the direct precursor of artemisinin (Fig. 8). Interestingly, in extracts from A. annua HAP flowers, no glycosides of AAOH, AA, DHAAOH and DHAA or glutathione conjugates of AAA were detected (Fig. S8). This indicates that the cells of A. annua that produce artemisinin either do not have competing glycosyl/malonyl/GSH transferase activity or, perhaps more likely, product flux within them is protected from such competing activities.

Figure 8.

Schematic representation of the artemisinin precursor biosynthetic pathway in N. benthamiana. NbDBR2-like, N. benthamiana artemisinic aldehyde double-bond reductase-like enzyme; NbRED1-like, N. benthamiana dihydroartemisinic aldehyde reductase 1-like enzyme; NbALDH1-like, N. benthamiana aldehyde dehydrogenase 1-like enzyme; NbGTase, N. benthamiana glycosyl transferase; NbMTase, N. benthamiana malonyl transferase; NbGSTase, N. benthamiana glutathione-S-transferase; Enzyme names in brackets indicates low activity of the enzyme for the corresponding step.

Concluding remarks

Our results show that the chemical profile of agro-infiltrated N. benthamiana leaves is a function of both type and relative dosage of the expression constructs. Results still do not fully explain the difference in HAP and LAP chemotypes found in A. annua. The expression level of the different biosynthesis genes do not differ between A. annua chemotypes (Maes et al., 2011) and therefore, at present, the difference in CYP71AV1 catalytic efficiency between HAP and LAP is the only identified factor that contributes to this difference in chemotype. However, other, as yet unidentified, factors in A. annua may further contribute to the chemical difference in the HAP and LAP varieties.


H-M.T. was funded by the graduate school of Experimental Plant Sciences (EPS). T.v.H. and J.B. were supported through the NWO-CW/ACTS IBOS programme (053.63.305) which was co-sponsored by Dafra Pharma, Turnhout, Belgium. We thank Ric de Vos for helpful discussions on LC-QTOF-MS/MS data, Bert Schipper for assistance in LC-QTOF-MS analysis, Desalegn Woldes Etalo, Ting Yang and Lemeng Dong for help in MS data analysis, and Qing Liu for support in the glutathione conjugation assay. We also would like to acknowledge Sajad Rashidi Manfard and Peter E. Brodelius (Linneaeus University, Sweden) for providing seeds of an A. annua LAP chemotype.