Our previous pyrosequencing annotated two HDR homologs, AaHDR1 and AaHDR2, the former of which contained a full length of nucleotide sequences (Ma et al., 2015). Based on our sequences, a cDNA fragment of AaHDR1 consisting of 1664 bp nucleotides (GenBank #: KY288069) was cloned from leaf tissues of self-pollinated A. annua using RT-PCR. Sequence analysis showed that this fragment contained the open reading frame (ORF) consisting of 1365 bp nucleotides, which were deduced to encode 455 amino acids. An rooted phylogenetic tree was built using the nucleotide sequences of AaHDR1 and AaHDR2 from A. annua and 16 other homologs from four algae, one gymnosperm, and 13 angiosperms (one monocot species and 12 dicot species) (Figure 2A). This tree was characterized by four clades, four algae species, banana (monocot, M. acuminata), gingko (gymnosperm), and angiosperms. In angiosperms, AaHDR1 and AaHDR2 are clustered with the Tanacetum parthenium HDR homolog, the two plant species of which are in the family Asteraceae, suggesting a close phylogenetic relationship. In addition, an analysis of amino acid sequences using an online ProtParam tool¹ revealed that its theoretical molecular weight and isoelectric point of the protein were 51405.46 Dalton and 5.63, respectively. Further sequence alignment with A. thaliana and Ginkgo biloba HDR homologs using the online TargetP ver. 1.0 software² suggested that the 33 N-terminal amino acids (Figure 2B) likely form a chloroplast targeting signal peptide. This feature suggests that AaHDR1 is located in plastids. In addition, sequence alignment revealed that AaHDR1 contains four conserved cysteine residues at positons 111, 202,256, and 339 (marked with triangle shapes in Figure 2B), which have been well-characterized to be essential for the activity of HDR homologs (Grawert et al., 2004).

Semi-quantitative RT-PCR analysis showed that AaHDR1 is expressed in roots, stems and leaves of sterile seedlings grown on MS medium contained in baby jars, and open flowers, flower buds, and leaves of plants on soil in the phytotron (Figure 3A). Further real time qRT-PCR analysis showed different levels in these tissues (Figure 3A). This result was in agreement of our pyrosequencing results that AaHDR1 was expressed in different positional leaves and flowers (Ma et al., 2015).

The ORF was fused to the N-end of an enhanced green florescent protein (EGFP) for a transient expression in N. benthamiana. Confocal microscopy analysis determined that the green fluorescence was detected in chloroplasts of leaf tissues, demonstrating the plastidial localization of AaHDR1 (Figure 3B).

The ORF of AaHDR1 driven by a 2 × 35S promoter was transformed into self-pollinated A. annua. Multiple hygromycin-resistant plants were obtained from selection media containing 20 mg/L hygromycin. Two overexpression lines, namely OE-3 and OE-4, were selected for further analysis (Figure 4A). RT-PCR analysis demonstrated the elevated expression levels of AaHDR1 in OE-3 and OE-4 transgenic plants (Figure 4B). Western blot demonstrated the increase of protein in these two transgenic lines (Figure 4C). After plants were grown in pot soil for 8 weeks in phytotron, leaves were collected for artemisinin and other metabolites analysis. HPLC–MS analysis showed the significant increase of artemisinin content in leaves of these two lines than in those of wild-type plants (Figure 4D).

Untargeted GC–MS analysis annotated 85 non-polar metabolites from leaves of two transgenic and two wild-type control plants, after grown on pot for 8 weeks. These compounds and their peak values were used as data matrix for principal component analysis (PCA). The resulting two-dimensional PCA plot consisting of PC1 (33.43%) and PC2 (11.94%) showed a distinct ordinate separation between two transgenic lines and two wild-type plants in the PC1 (Figure 5A), indicating alternations of non-polar metabolite profiles by the overexpression of AaHDR1.

Of 85 non-polar metabolites detected, 15 were terpenoid molecules including eight monoterpenes and seven sesquiterpenes. The peak value of each metabolite was used to compare their levels in leaves of transgenic vs. wild-type plants. The resulting data showed that the levels of six monoterpenes were significantly higher in transgenic leaves than in wild-type ones (Figure 5B). The levels of two other monoterpenes were only slightly higher in transgenic leaves. Of seven sesquiterpenes, one is arteannuin b, the peak values of which were 1.5–1.8 folds (p-value < 0.05) higher in transgenic leaves than in wild-type ones (Figure 5C and Supplementary Figure S1). In addition, transgenic leaves produced significantly higher levels of cis-lanceol and 1-enthyl-1-methyl-cyclohexane than wild-type control. The levels of beta-farnesene and germacrene D were lightly increased in transgenic leaves. In contrast with these metabolites, the level of caryophyllene was significantly decreased in transgenic leaves.

An anti-sense binary vector of AaHDR1 (Figure 6A) was constructed and introduced into A. annua. Multiple transgenic lines were obtained, two of which, namely AS-14 and AS-15 (Figure 6B), were selected for further analysis. RT-PCR analysis showed the decreased expression levels of AaHDR1 in leaves (Figure 6C). Western blot analysis determined that the level of AaHDR1 was reduced in transgenic leaves than in wild-type ones (Figure 6D). After plants were grown on pot soil for 9 weeks (Supplementary Figure S3) in the phytotron, leaves were collected for artemisinin and other metabolites analysis. HPLC–MS analysis demonstrated that contents of artemisinin were decreased 27–33% in transgenic leaves (Figure 6E).

Untargeted GC–MS analysis annotated 82 non-polar metabolites. These metabolites were used as data matrix for PCA. The resulting two dimensional plot including PC1 (35.87) and PC2 (12.71%) showed an ordinate distinction of metabolite profiles in the PC1 (Figure 7A).

Of 82 metabolites, there were 18 terpenes, including β-amyrin and phytol (Figure 7B), six sesquiterpenes (Figure 7C), and 10 monoterpenes (Figure 7D). Peak values of these metabolites were recorded to compare their levels in transgenic vs. wild-type leaves. In comparison, the level of arteannuin B (Figure 7C and Supplementary Figure S2) was reduced approximately twofold in transgenic leaves. In addition, the levels of five other sesquiterpenes were significantly or slightly decreased in transgenic leaves. Furthermore, the levels of 10 monoterpenes were significantly reduced in transgenic leaves (Figure 7D). The levels of β-amyrin and phytol were significantly reduced in transgenic leaves (Figure 7B).

A self-pollinated A. annua variety was bred from an ecotype collected from USA (Alejos-Gonzalez et al., 2011). Progenies of plants have been grown in the phytotron for seeds as described previously (Alejos-Gonzalez et al., 2011). Wild-type seedlings of the F3 progeny grown on agar-solidified MS medium contained in baby jars in a growth chamber as described previously (Alejos-Gonzalez et al., 2013) were used for genetic transformation and gene expression profiling. Wild-type and transgenic plants grown on soil in the phytotron were used to analyze transgene expression, western blot, artemisinin analysis, and metabolic profiling. The photoperiod in the growth chamber and phytotron was 16/8 h (light/dark). The light intensities in the growth chamber and the phytotron were 50 and 200 μmol/m^-2 s^-1, respectively.

We recently reported two HDR contigs, AaHDR1 and AaHDR2, assembled from six EST libraries (including leaves and flowers) of the self-pollinated A. annua (Ma et al., 2015). The first strand cDNA libraries remained after sequencing were stored in -80°C freezer and then used to clone genes. Based on EST sequence, we designed a pair of primers: HDR-F (5′-ATG GCG TCT TTG CAG CTA ACA-3′) and HDR-R (5′-CTA CAC CAA TTG CAG GGC CTC-3′). By following our gene cloning protocol reported previously (Ma et al., 2015), RT-PCR was carried out to clone the ORF sequence of AaHDR1.

Deduced amino acid sequences were analyzed using an online ProtParam tool¹ to predict molecular properties. Amino acid sequences were also analyzed using an online TargetP ver. 1.0 software² to predict conserved active amino acid residues. The full length nucleotide sequence (1664 pb) of AaHDR1 was used for blastn search at NCBI. The blasting result included 16 homologous sequences with an e-value < 10E-8. All sequences were converted to fasta format for phylogenetic analysis using the online phylogeny.fr program³. In addition, amino acids for these homologs were released from NCBI and also converted to fasta format for sequence alignment using an online Cluster Omega program⁴.

A GATEWAY cloning technique was used to clone the AaHDR1 ORF into pENTY/D-TOPO vector (Gateway, Invitrogen, USA) by following the manufacturer’s protocol. The resulting recombinant plasmid was named pENTY/D-TOPO-HDR. Then, the ORF of AaHDR1 in the pENTY/D-TOPO-HDR vector was subsequently cloned to the destination vector pSITEII-N1-EGFP (with EGFP epitope tagging in the C-terminus) by LR reactions by following the manufacturer’s protocol. The resulting plasmid was named as pSITEII-N1-HDR/EGFP, in which the stop codon of AaHDR1 was removed and ligated to the adjacent place immediately prior to the start codon ATG of EGFP. The new plasmid was then introduced to Agrobacterium tumefaciens strain GV3101. A positive colony was obtained using 100 mg/L streptomycin for selection. The positive colony was activated for agroinfiltration of N. benthamiana leaves to analyze transient protein expression by following the protocol reported previously (Martin et al., 2009). After 30 h of infection, leaf tissues were analyzed using a confocal microscope (Carl Zeiss). The GFP fluorescence was excited at 488 nm and observed between 495 and 550 nm as reported (Wang et al., 2015).

The ORF of AaHDR1 with its terminal codon was cloned to the pENTY/D-TOPO vector (Gateway, Invitrogen) to obtain a recombinant pENTY-AaHDR1 plasmid, which was then introduced to competent cells of E. coli DH5α. Given that pENTY/D-TOPO vector can allow both sense and anti-sense orientation ligation, 10 positive individual colonies were selected to isolate plasmids for sequencing to identify sense or anti-sense insertion orientations of the AaHDR1 ORF. Sequencing showed that eight colonies contained a sense orientation of the ORF cDNA, namely pENTR-AaHDR1 and two colonies contained an anti-sense orientation the ORF cDNA, namely pENTR-antiAaHDR1. The destination vector used for development of our binary vectors was PMDC-84 (Curtis and Grossniklaus, 2003). The pENTR-AaHDR1 and pENTR-antiAaHDR1 and the destination vector PMDC-84 were digested by LR Clonase II enzyme mix (Invitrogen) by following the manufacturer’s protocol. This reaction generated recombinant binary vectors, PMDC-AaHDR1 (Figure 4A) (from PMDC-84 and pENTR-AaHDR1) and PMDC-antiAaHDR1 (from PMDC-84 and pENTR-antiAaHDR1) (Figure 6A), in which the sense and antisense orientations of AaHDR1 ORF were driven by a 2 × 35S promoter.

The overexpression and anti-sense constructs, PMDC-AaHDR1 (Figure 4A) and PMDC-antiAaHDR1 (Figure 6A), were introduced to A. tumefaciens strain LBA4404 for genetic transformation of A. annua as described previously (Xie et al., 2004; Ma et al., 2015). In summary, seeds were sterilized and germinated on an agar-solidified basal MS medium (Murashige and Skoog, 1962) and leaves at the 5, 6, and 7 nodes of one and a half month old seedlings were used as explants for infection of Agrobacterium. The infected leaf disks were transferred onto an agar-solidified selection medium, which was composed of basal MS medium, 1.0 mg/L 6-benzylaminopurine (BAP), 0.05 mg/L naphthalene-1-acetic acid (NAA), 20 mg/L hygromycin, and 200 mg/L timentin. Survival infected leaf disks were transferred to freshly prepared selection medium every 10 days of interval (subculture). After approximately 50 days of selection, multiple hygromycin-resistant adventitious shoots were obtained, excised from calli, and then inoculated on a rooting medium (half-strength MS, 0.5 mg/L NAA, 20 mg/L hygromycin, and 200 mg/L timentin). Hygromycin-resistant adventitious shoots started to grow roots to develop into plantlets after 1–2 weeks of induction. Regenerated plantlets were planted on pot soil and grown in the phytotron by following our growth protocol as described above. The transgenic plant growth in the phytotron was observed carefully to compare their development. Those transgenic plants that grew at similar developmental stage to wild-type plants (Figures 4A and 6B) were used for transgene expression, western blot, and metabolite analysis described below.

For AaHDR1 expression profile analysis in wild-type plants, 8-week old seedlings grown in baby jars were used to collect roots, leaves, and stems, while 10-week old plants grown in the phytotron were used to collect open flowers, flower buds, and leaves. Three biological replicates were collected for each tissue. For transgene expression and western blot analysis, young leaves were collected from 6-week old transgenic vs. wild-type plants grown on pot soil in the phytotron. For GC–MS, and LC-MS analysis, leaves were collected 8-week old overexpression lines and 9-week old anti-sense expression lines (Supplementary Figure S3). For each plant, leaves were collected from nodes #7-13 of and then pooled together. All samples were frozen in liquid nitrogen and stored in 80°C freezer until use.

Total RNA was isolated from 100 mg of roots, leaves and stems of seedlings grown in baby jars, and open flowers, flower buds, and leaves of plants grown in the phytotron with the RNeasy Mini Kit (Qiagen, Germantown, MD, USA) in accordance with the manufacturer’s instructions. Total RNA was then treated with DNase I (Roche, CA, USA) to remove potential genomic DNA. First-strand cDNA synthesis was carried out with 1.0 μg of total RNA Super SMART PCR cDNA synthesis kit (Clontech, Mountain View, CA, USA) according to the manufacturer’s protocol. The resultant first-strand cDNA was used as the template for PCR.

A pair of gene specific primers was designed from the 3′-untranslated region of AaHDR1 cDNA, namely AaHDR-Fx (5′-GGCATGTACCTGGCAAGAGAG-3′) and AaHDR-Rx (5′-GTCGTTTATAGCAACCAGAGCC-3′). In addition, the house keeping ACTIN gene was used as reference control and was amplified using a pair of primers: actin F (5′-AACTGGGATGACATGGAGAAGATAT-3′) and actin R (5′-TCACACTTCATGATGGAGTTGTAGG-3′). These two pairs of primers were used for both semi-quantitative and real time quantitative RT-PCR. For semi-quantitative PCR, the thermal program was composed of 5 min at 95°C, 30 cycles at 94°C for 30 s, 55°C for 40 s and 72°C for 110 s, and a 10 min extension at 72°C. PCR products were electrophoresed on a 1% agarose gel and visualized after ethidium bromide staining. Images of gels were photographed using a GelDoc EQ imager (Bio-Rad). This experiment was repeated three times. For quantitative analysis, real time qPCR was performed on a Power SYBR Green PCR Master Mix (Applied Biosystems) according to the manufacturer’s guidelines. The thermal cycles were composed of 50°C for 2 min and 95°C for 10 min, followed by 95°C for 15 s and 60°C for 1 min with 45 cycles, and then finished with 40°C. Five replicates and three replicates were performed for AaHDR1 and ACTIN, respectively. ACTIN was used internal standard for normalization. The relative expression of AaHDR1 was calculated using the ddCt algorithm.

Polypeptides consisting of CIDGGEKQFDVVDKG were designed for preparing polyclonal antibody. This peptide was synthesized and then used to develop polyclonal antibody at the GenScript Company (Piscataway, NJ 08854, USA). Total protein extraction from leaves and immunoblot analysis were performed by following a method previously reported for deoxy-D-xylulose-5-phosphate synthase analysis (Estevez et al., 2001). In brief, total protein was extracted from the leaves of A. annua, separated by SDS-PAGE, transferred to nitrocellulose membrane, and probed with rabbit anti-AaHDR1 antibodies. An anti-rabbit IgG HPR conjugate was used as a second antibody (Promega, USA). The enhanced chemiluminescence (ECL) system was used for detection in the immunoblot analysis (Thermos Scientific, Rockford, IL, USA).

We developed a protocol to extract and profile non-polar metabolites from tissues of A. annua (Ma et al., 2015). In brief, hexane was used to extract non-polar metabolites from mixed samples of leave on nodes #7-13. Three replicates were prepared for each transgenic line (anti-sense and overexpression) and wild-type plants. GC–MS was performed using a gas chromatograph 6890 coupled with 5975C MSD (Agilent Technologies, USA). A RTX-5 capillary column (30 m × 0.25 mm × 0.25 μm) was used to separate metabolites. The splitless mode was used in the inlet. The injection temperature was set at 250°C. The temperature was initially set at 60°C and then ramped to 260°C at a constant rate of 10°C/min, and held at 260°C for 25 min. Pure helium was used as the carrier gas, with a flow rate of 1 ml/min. A positive electron impact ion source (70 EV) was used to ionize compounds, and mass fragments were scanned in the range of 40–800 (m/z), with 4 min of solvent delay.

We have developed a method to use an Agilent MassHunter Mass Profiler (MHMP) and Mass Profiler Professional (MPP) software to deconvolute peaks detected by GC–MS (Ma et al., 2015). This software has been used for characterizing metabolic profiles (Chen et al., 2012; Chan et al., 2013; Hu et al., 2013; Bannur et al., 2014). The fold-change for all metabolites was obtained using log 2 values, which were calculated values that resulted from the transformation of original chromatographic peak accounts compared with each metabolite median value. The median value of each metabolite was obtained from four biological samples analyzed. The peak value of each metabolite was compared with the median value for log 2 normalization. Then, normalized data were used for PCA.

We have developed a HPLC–MS protocol for artemisinin analysis on a 2010 eV LC/UV/ESI/MS instrument (Shimadzu) (Alejos-Gonzalez et al., 2011, 2013). Mixed samples of leaves on nodes #7-13 from transgenic vs. wild-type plants were stored in freezer described above and then used to extract and estimate artemisinin contents. Three replicates were carried out for each transgenic line and wild-type plants.