Compartmentalized Metabolic Engineering for Artemisinin Biosynthesis and Effective Malaria Treatment by Oral Delivery of Plant Cells

Abstract

Artemisinin is highly effective against drug-resistant malarial parasites, which affects nearly half of the global population and kills >500 000 people each year. The primary cost of artemisinin is the very expensive process used to extract and purify the drug from Artemisia annua. Elimination of this apparently unnecessary step will make this potent antimalarial drug affordable to the global population living in endemic regions. Here we reported the oral delivery of a non-protein drug artemisinin biosynthesized (~0.8 mg/g dry weight) at clinically meaningful levels in tobacco by engineering two metabolic pathways targeted to three different cellular compartments (chloroplast, nucleus, and mitochondria). The doubly transgenic lines showed a three-fold enhancement of isopentenyl pyrophosphate, and targeting AACPR, DBR2, and CYP71AV1 to chloroplasts resulted in higher expression and an efficient photo-oxidation of dihydroartemisinic acid to artemisinin. Partially purified extracts from the leaves of transgenic tobacco plants inhibited in vitro growth progression of Plasmodium falciparum-infected red blood cells. Oral feeding of whole intact plant cells bioencapsulating the artemisinin reduced the parasitemia levels in challenged mice in comparison with commercial drug. Such novel synergistic approaches should facilitate low-cost production and delivery of artemisinin and other drugs through metabolic engineering of edible plants.

INTRODUCTION

Artemisinin-based combination therapies, commonly referred to as ACTs, are the frontline treatments in the fight against malaria (Dalrymple, 2013). Besides the antiplasmodial activity of artemisinin and its semi-synthetic derivative artesunate, the drug has demonstrated cytotoxicity against cancer cells and schistosomiasis (Efferth, 2006; Utzinger et al., 2007; Efferth et al., 2011). A notable broad-spectrum activity of artesunate has been reported against human cytomegalovirus, which has become resistant to traditional drugs such as ganciclovir (Shapira et al., 2008). However, these therapies are not readily available to patients in endemic regions due to the low yield of artemisinin from the native plant. Artemisinin is biosynthesized mainly in the glandular trichomes on leaves. Three key enzymes of artemisinin biosynthesis were exclusively expressed in apical cells of glandular trichomes but not in the subapical chloroplast-containing cells, although all of these cells did express the farnesyl pyrophosphate synthase (Olsson et al., 2009). Artemisinin production in aerial parts of the plant is sensitive to variation in genetic backgrounds, cultivation conditions, and harvesting periods. Its biosynthesis is stimulated by biotic and abiotic stress (Ferreira et al., 1995; Davies et al., 2009, 2011). The yield of artemisinin (~0.5% on dry weight basis) from Artemisia annua is relatively low when cultivated under Indian climatic conditions (Brisibe and Chukwurah, 2014), which limits its commercial production and value. Even the high-yield variety “CIM-Arogya,” developed for marker-assisted breeding, has maximum reported yields of ~1% in Indian conditions (Khanuja et al., 2008). Endeavors to increase artemisinin levels in A. annua by elicitor or hormone treatments (Smith et al., 1997; Baldi and Dixit, 2008) have met with limited success, and present efforts rely mostly on marker-assisted breeding of high-yield varieties (Graham et al., 2010). Low yields of artemisinin from crop-based production systems and the high cost of chemical synthesis, make these approaches impractical for commercial production (Enserink, 2005; White, 2008). Further to these issues, yearly variations in the availability of starting raw materials and unpredictable demand for this antimalarial agent greatly affect the price of artemisinin (Kindermans et al., 2007). Thus, alternative heterologous production platforms should be explored to address these issues. Attempts have been made to produce its precursor, amorphadiene, via metabolic engineering of Escherichia coli with the mevalonate (MEV) pathway and the amorphadiene synthase gene (ADS). Unfortunately, transgene expression inhibited E. coli growth, resulting in very low amorphadiene levels (Martin et al., 2003). The semi-synthetic approach in yeast yielded a high level of artemisinic acid, a precursor of artemisinin (Paddon et al., 2013), but complete biosynthesis of artemisinin was not accomplished. Expression of artemisinin is difficult in microbial hosts, mainly due to lack of plant cytochrome P450. The successive oxidation of amorphadiene to artemisinic alcohol is catalyzed by a cytochrome P450 oxidase gene (CYP71AV1). Artemisinin biosynthesis is favored through the dihydroartemisinic acid (DHAA) biosynthetic pathway using double bond reductase 2 (DBR2), which reduces the Δ11(13) double bond of artemisinic aldehyde (Covello, 2008; Zhang et al., 2008). The conversion of DHAA to artemisinin is a non-enzymatic photochemical oxidation process (Brown and Sy, 2004).

Tobacco became our platform of choice because artemisinin is a plant product and other terpenoids have successfully been expressed (Ikram et al., 2015). Tobacco is a large plant that can be harvested several times a year and can be grown as a perennial. Each plant produces nearly a million seeds per season, so production acreage can be rapidly scaled up for the low costs production of pharmaceuticals (Arlen et al., 2007). The constitutive expression of the ADS gene in Nicotiana tabacum has been achieved (Wallaart et al., 2001) but very low levels of amorphadiene biosynthesis were realized (<2 ng/g of fresh weight). Recently, the complete biosynthesis of artemisinin was reported in N. tabacum via nuclear transformation by expressing CYP71AV1 and cytochrome P450 reductase (CPR), along with ADS, DBR2, and a deregulated form of HMGR (Farhi et al., 2011a). However, artemisinin levels were found to be low in these transgenic lines (6.8 μg/g dry weight), which may be attributed to a limiting pool of isopentenyl pyrophosphate (IPP).

Subcellular targeting of ADS to mitochondria significantly improved the accumulation of amorphadiene (Wu et al., 2006; van Herpen et al., 2010; Farhi et al., 2011b). These studies showed that farnesyl pyrophosphate (FPP) (a precursor to amorphadiene) is readily available in different cellular compartments and could be utilized for amorphadiene and artemisinin biosynthesis. Chloroplasts offer an ideal platform for accumulating large amounts of IPP and overcome substrate limitations apparent in the cytosol. Homoplastomic plants engineered with the MEV pathway accumulate higher levels of mevalonate than untransformed plants (Kumar et al., 2012). Previously, we attempted to achieve complete artemisinic acid biosynthesis in tobacco chloroplasts by introducing both MEV and artemisinic acid pathways (~12 genes). Unfortunately, this approach resulted in a low yield of artemisinic acid and stunted the growth of transgenic plants (Saxena et al., 2014), possibly because of suppression of one of the added pathways or precursor limitations.

To overcome these limitations, we developed a novel strategy for complete artemisinin biosynthesis in tobacco through combining expression in both chloroplast and nuclear genomes (Figure 1). The stable doubly transgenic (DT) plants produced adequate levels of artemisinin, and drug functionality was studied on the malarial parasite Plasmodium using in vitro and in vivo assays.

Figure 1. Schematic Representation of Artemisinin Biosynthesis by Sequential Metabolic Engineering of Chloroplast and Nuclear Genomes.

Six genes, AACT, HMGS, HMGRt, MVK, PMK, and PMD (in red font and inside a dotted rectangle, Supplemental Figure 1A) encoding yeast MEV pathway were integrated into the chloroplast genome using a biolistic approach to generate an elevated IPP pool. Six genes (ADS, CYP71AV1, AACPR, DBR2, IDI, and FPP) of the artemisinin biosynthetic pathway were integrated into the nuclear genome of homoplastomic plants (Supplemental Figure 1B) using vector pNuc-cox-ART. Subcellular targeting of DBR2, AACPR, and CYP71AV1 were mediated by a chloroplast transit peptide (Figure 2A). Dihydroartemisinic acid (DHAA) converts itself to artemisinin via photochemical oxidation, a non-enzymatic reaction. AACT, acetoacetyl CoA thiolase; HMGS, HMGCoA synthase; HMGRt, C-terminal truncated HMGCoA reductase; MVK, mevalonate kinase; PMK, phosphomevalonate kinase; PMD, mevalonate pyrophosphate decarboxylase; DXP, 1-Deoxy-D-xylulose 5-phosphate; MEP, 2-C-methyl-D-erythritol 4-phosphate; IPP, isopentenyl diphosphate; IDI, IPP isomerase; DMAPP, dimethylallyl pyrophosphate; ADS, amorphadiene synthase; FPP, farnesyl diphosphate; CYP71AV1, amorpha-4,11-diene C-12 oxidase; AACPR, Artemisia annua cytochrome P450 reductase; DBR2, double bond reductase2.

RESULTS

Confirmation of Plant Transformation and Molecular Analysis

Homoplastomic plants expressing the mevalonate pathway grew similarly to wild-type plants in the greenhouse, without observable phenotypic abnormality (Supplemental Figure 1C). The functionality of the yeast mevalonate biosynthetic pathway in T-MEV plants was confirmed by treatment of transplastomic plants with selection medium containing 100 μM fosmidomycin, a specific inhibitor of 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR). These transplastomic plants thrived (Supplemental Figure 1D, right), while the wild-type plants had stunted growth and bleached leaves (Supplemental Figure 1D, left), confirming that IPP was produced in chloroplast by the genetically engineered mevalonate pathway (Supplemental Figure 1D, right).

Twelve independent nuclear transgenic lines were regenerated on double-selection media when T-MEV was re-transformed via Agrobacterium tumefaciens (strain EHA105) harboring a binary vector pNuc-Cox-Art containing DHAA biosynthetic pathway genes (Figure 2A). PCR analysis using ADS and DBR2 gene-specific primers confirmed the integration of transgenes into the nuclear genome in all 12 DT plants. All transgenic lines showed 539-bp and 441-bp amplicons for ADS and DBR2, respectively (Figure 2B). No PCR amplification product was observed in wild-type plants. All DT plants grew normally and set flowers and fertile seeds (Figure 2C). No morphological abnormality was observed in DT plants with respect to growth pattern, leaf size, or plant height when compared with wild-type plants. However, there was a decrease in the total chlorophyll and carotenoid content in all DT lines (DT1–DT4), with DT2 and DT4 showing the lowest pigment levels (Supplemental Figure 6A). Pigment levels were unaffected in the T-MEV plants. Maximum quantum yield (F_v/F_m), i.e., photosynthetic performance, was marginally lower in all DT lines (except DT3 and T-MEV) (Supplemental Figure 6B). The effective quantum yield, or Y(II) and Y(NPQ), or quantum yield of regulated energy dissipation, was lower for DT3 and T-MEV plants (Supplemental Figure 6C) than for the other DT plants or the wild type. As predicted, only T1 seeds from DT plants were able to grow on double-selection RMOP medium containing hygromycin (50 mg/l) and spectinomycin (500 mg/l) (Figure 2D).

Figure 2. Nuclear Genome Transformation of Homoplastomic Plant.

(A) Physical map of the DHAA biosynthetic pathway vector pNuc-cox-Art.

(B) PCR products amplified using primers ADS-F and ADS-R (marked by purple in A) and DBR2-F and DBR2-R (marked by blue in A) from 12 doubly transgenic (DT1–DT12) plants yielded a 539-bp and 441-bp amplicon, respectively. No PCR products were observed in wild-type (WT) plants.

(C) DT plants transferred to a greenhouse grew normally like WT plants and set seeds.

(D) Seeds of matured DT plants germinated and seedlings grew well on double-selection medium containing hygromycin (50 mg/l) and spectinomycin (500 mg/l). T-MEV and WT plants lacking essential transgenes conferring resistance to both antibiotics could not germinate on double-selection medium (see also Supplemental Figure 1).

(E) Four transgenic lines namely DT1–DT4, with elevated artemisinin and DHAA contents were examined for copy number. Genomic DNA (~15 μg) from DT1–DT4 and WT plants digested with EcoRV were analyzed by Southern blot using DBR2 as probe (shown as blue line in vector A).

(F) Total RNA was isolated from DT1–DT4 and resulting cDNAs were amplified using gene-specific RT–PCR primers of ADS, DBR2, AACPR, and CYP71AV1.

The highest yield of artemisinin was observed in four lines, named DT1 to DT4, ranging from 0.3 to 0.8 mg/g dry weight. All four lines (DT1–DT4) contained a single integrated copy of the DHAA biosynthetic transgene cassette (Figure 2E). Amplicon signals of 219 bp for ADS, 281 bp for DBR2, 190 bp for CYP71AV1, and 275 bp for AACPR were observed in greenhouse-grown DT1–DT4 plants (Figure 2F), confirming nuclear expression.

Production of Artemisinin, DHAA, IPP/Dimethylallyl Pyrophosphate, and Amorphadiene in Transformed Plants

All doubly transgenic (DT) lines accumulated artemisinin except DT9 and DT10 (Supplemental Table 3). Maximum biosynthesis was observed in DT1, DT2, DT3, and DT4. These lines accumulated artemisinin due to the non-enzymatic reaction of photo-oxidation of DHAA (Figure 3B), while no such accumulation was observed in wild-type tobacco plants (Figure 3A). The levels of artemisinin in the four transgenic lines varied between 0.3 and 0.8 mg/g dry weight (n = 3). The artemisinin content was highest in DT4 (Figure 3B and 3C) plants, about 0.8 mg/g dry weight (n = 3). The product ion (MS2) spectra of artemisinin in DT4 plant is mentioned in Supplemental Figure 2B. Such variation in transgene expression among nuclear transgenic lines is expected and occurs due to positional effects, gene silencing, and other regulatory controls.

Figure 3. LC–MS/MS Analysis of Artemisinin Biosynthesized in Double-Transformed Plants.

(A) Chromatographic profile of wild-type plant extract, analyzed in MRM mode for artemisinin detection.

(B) Extracted ion chromatogram for artemisinin–MRM transition (m/z 283–209) standard peak shown in red. The artemisinin peak of transgenic (DT4) shown in blue is overlaid with standard peak (red) at ~5.25 min. No such corresponding peak for artemisinin was observed in extracts of wild-type (A).

(C) Artemisinin yield quantified in DTs by MultiQuant (ABSciex, Singapore) (see also Supplemental Table 3). Standard curve was plotted and the R² coefficient was 0.998. ANOVA followed by Duncan’s multiple range test demonstrated that the mean artemisinin levels were significantly different among the lines (P < 0.01). Means followed by the same alphabets are not significantly different by Duncan’s multiple range test at P < 0.01, n = 3. Data are represented as mean ± SD. N.D., not determined; WT, wild-type.

We analyzed the synthesis of two intermediates of the artemisinin biosynthetic pathway, namely DHAA and IPP, in the transgenic lines DT1–DT4. A peak corresponding to the retention time of DHAA was observed in both a DHAA standard (Figure 4A) and an extract of A. annua (Figure 4C). Interestingly, an additional peak at 5.2 min was observed in the DTs (Figure 4B), which matched the chromatographic profile of A. annua (Figure 4C) used as a control. However, this additional peak was not observed in the authentic standard of DHAA (Figure 4A). Maximum DHAA accumulation observed was in DT4, up to 150 μg/g dry weight (Figure 4B).

Figure 4. LC–MS/MS Analysis of Dihydroartemisinic Acid Biosynthesized in DT Plants.

(A–C) Extracted ion chromatogram for dihydroartemisinic acid (DHAA)–MRM transitions for two fragments m/z 237–163 and m/z 237–107. (A) Standard DHAA peak at retention time ~11 min. (B) DHAA in DT4. (C) Similar peak was observed in Artemisia annua extract. An unknown peak was seen at retention time 5.25 min in both DT4 and A. annua samples. The red and blue lines denote the MRM transitions (163 and 107) for the parent ion m/z 237 DHAA.

Intermediate metabolite IPP synthesis in the DTs was compared with the wild-type plants, using an authentic standard of IPP (Supplemental Figure 3B) by liquid chromatography–tandem mass spectrometry (LC–MS/MS). The DT plants accumulated more IPP (Figure 5C–5E) than the wild type (Figure 5A) and T-MEV plants (Figure 5B). The mean amount of IPP ranged from 0.1 to 2.5 μg/g fresh weight in DTs, compared with only ~0.07 μg/g fresh weight in the wild type (n = 3, Figure 5G). The intermediate volatile metabolite amorphadiene was detected in DTs by gas chromatography (GC)–MS by comparing the mass spectra with the standard (Figure 6A and 6B). The transgenic line DT4 (Figure 6A, upper chromatogram) had the highest amorphadiene levels, up to 60 μg/g dry weight. However, absolute quantification of this compound is difficult due to the volatile nature of amorphadiene (Beale et al., 2006; van Herpen et al., 2010).

Figure 5. Comparative LC–MS/MS Analysis of Intermediate Isoprenoid IPP-Metabolite Levels in Controls (Wild-Type) and Chloroplast Transgenic with Mevalonate Pathway (MEV) and Double-Transformed Plants (DT1–DT4).

(A–G) Extracted ion chromatogram of MRM transition (m/z 245–159) for IPP (see also Supplemental Figure 3) in (A) wild-type, (B) MEV, and (C–F) DT1–DT4. (G) IPP yield quantified in wild-type (WT), MEV, and DTs by MultiQuant (ABSciex, Singapore). A standard curve was plotted and the R² coefficient was 0.998. Analysis of variance (ANOVA) followed by Duncan’s multiple range test demonstrated that the mean IPP levels were significantly different among the lines (P < 0.01). Means followed by the same alphabets are not significantly different by Duncan’s multiple range test at P < 0.01, n = 3. Data are represented as mean ± SD.

Figure 6. Accumulation of Amorphadiene in DT4.

(A) Total ion chromatograms from ethyl acetate extract of DT4 plant (upper) and amorphadiene standard (lower). Peak corresponding to amorphadiene was observed at retention time ~23 min in the chromatogram of DT4 and amorphadiene standard.

(B) Mass spectra of amorphadiene obtained from the peak in DT4 plant at retention time ~23 min in GC profile.

In Vitro Anti-parasitic Activity of Artemisinin Produced by DTs

Partially purified transgenic plant extracts of DT lines inhibited the growth of parasites after 24 and 48 h (Figure 7A). Few or no new ring-stage parasites, and dead parasites with constricted and shrunken morphology, were observed (Figure 7A). In comparison, parasite cultures treated with either DMSO or with an extract from wild-type plants grew normally. The merozoites released from schizonts further invaded fresh red blood cells (RBCs) and formed new rings by 48 h (Figure 7A). Extract from DT4, which contained the highest artemisinin level, was most effective, with IC₅₀ ~79 nM (Figure 7B; see also Supplemental Figure 4).

Figure 7. In Vitro and In Vivo Efficacy of Artemisinin Biosynthesized in Dual Metabolically Engineered Plants.

(A) In vitro mortality assay of artemisinin from double transformed (DTs) on RBCs infected with Plasmodium falciparum 3D7. Giemsa staining of RBCs infected with P. falciparum 3D7. No mortality on parasite growth was seen in DMSO, wild-type (WT), and schizonts at the end of 48 h matured to release the ring-stage parasites. DT1–DT4 extracts controlled parasite growth in 24–48 h assay.

(B) IC₅₀ values were plotted and calculated using linear regression analysis (see also Supplemental Figure 4).

(C) Balb/C mice infected with Plasmodium berghei were orally fed for nine consecutive days (3–11) with dried DT4 leaves (dose equivalent to 24 mg/kg), control wild-type (WT), and pure drug (Art. drug; equivalent to 24 mg/kg). Arrow in the graph represents the start of oral feeding on the third day post infection. The error bars represent mean parasitemia ± SD.

Efficacy of Artemisinin-Producing Plant Extracts in the P. berghei Mouse Model

The in vivo anti-malarial efficacy of DT plant extracts was analyzed using the Plasmodium berghei murine malaria model. Infected mice were fed with dried intact plant cells using a gavage needle, and parasitemia was evaluated every day until day 15. Parasitemia profiles of mice treated with transgenic plants were compared with mice fed with pure artemisinin drug or wild-type plant extracts. Similar progression in parasitemia was observed in mice fed with pure artemisinin as well as those fed with the wild-type plant extract (Figure 7C), the control group. However, mice fed with crushed DT leaves showed a considerable delay in progression of parasitemia, most notable in mice fed with DT4 on the 10th day post infection. A significant reduction in parasitemia from ~2.5% to ~1.6% was observed when compared with mice fed pure artemisinin (Figure 7C). In addition, mice fed with DT4 leaf material did not show any sign of malarial sickness during the first week of infection, while mice fed with pure artemisinin or wild-type plants had malarial symptoms. Together, these results show that DT leaf material considerably delayed the development of parasitemia in mice and that this delay was longer than that induced by pure artemisinin.

DISCUSSION

In our previous study (Saxena et al., 2014), expression of all the 12 transgenes into the tobacco chloroplast genome caused stunted growth, which highlighted the need to spread the biosynthetic load to more than one metabolic compartment. We circumvented these defects by using a balanced compartmentalization approach to transgene expression using three cellular compartments (cytosol, chloroplast, and mitochondria), and obtained DT plants with normal phenotypes and up to 0.8 mg artemisinin per gram dry weight leaf. A similar effort using combinatorial super transformation of transplastomic lines engineered with the artemisinic acid pathway resulted in an accumulation of 120 μg/g fresh weight of artemisinic acid (Fuentes et al., 2016). Low artemisinin accumulation in tobacco in earlier reports was thought to be either due to a high level of glycosylation by endogenous glycosyl transferases (van Herpen et al., 2010; Ting et al., 2013) or unfavorable cytosolic environmental conditions preventing the formation of late-stage artemisinin precursors (Zhang et al., 2011). The transgenes AACPR, CYP71AV1, and DBR2, beyond ADS, were targeted to the chloroplast in our study because it is a glycosylation-free compartment. Thus, chloroplasts offer ideal sites for accumulation of complex transgene products that may otherwise be harmful if accumulated in the cytosol (Bogorad, 2000). The use of intense light during greenhouse cultivation and harvesting of DT lines may have favored the bioconversion of DHAA to artemisinin by photo-oxidation (Farhi et al., 2011b). Also, the excess pool of IPP/dimethylallyl pyrophosphate (DMAPP), produced in the DT plants (via T-MEV pathway and IDI gene expressed in chloroplast and nuclear genome, respectively) may have helped produce the higher end product. We hypothesize that compartmentalization, subcellular targeting of artemisinin biosynthetic enzymes to chloroplasts, and a surplus amount of endogenous IPP (Figure 5B) are key to producing much higher levels of artemisinin than previously reported (Supplemental Table 2).

All terpenoids, including artemisinin, are synthesized from two isoprenoid precursors, IPP and DMAPP. However, the synthesis of IPP by the MEV pathway is tightly regulated due to its need to supply diverse metabolic processes (Goldstein and Brown, 1990). The limited, wild-type, IPP pool is unlikely to permit DHAA accumulation and would result in the little biosynthesis of artemisinin. LC–MS/MS measures the total amount of IPP and DMAPP as both the metabolites are allylic isomers with the same parent mass, i.e., 245 Da (Li and Sharkey, 2013); thus, in our analysis the total amount of IPP and DMAPP was quantified (Figure 5A–5G). The IDI gene (IPP/DMAPP isomerase) was introduced into the nuclear genome, which may have helped to maintain a favorable equilibrium between IPP and DMAPP, and normal physiological functions in eukaryotes (Koyama et al., 1983; Lützow and Beyer, 1988). Moreover, reduced IDI activity affects isoprenoid biosynthesis in the plastids of tobacco leaves and results in depletion of photosynthetic pigments (Page et al., 2004). Amorphadiene is the first committed precursor in the artemisinin biosynthetic pathway. Previous reports suggest that fusion of sesquiterpene synthases with cox4 results in elevated levels of sesquiterpenoid production because of the considerable pool of FPP found in mitochondria (Kappers et al., 2005; Farhi et al., 2011b). Therefore, to enhance amorphadiene synthesis, the ADS gene was fused with mitochondrial signal peptide (cox4), and this did result in higher amorphadiene biosynthesis (Figure 6A).

Previously, we have reported the expression of individual transgenes related to the MEV pathway in chloroplasts (Kumar et al., 2012). The RT–PCR study was conducted for mRNA analysis for determining the expression of other essential transgenes (DBR2, ADS, AACPR, and CYP71AV1). The weak correlations between mRNA and protein expression are attributed to various post-transcriptional processes that complicate accurate estimates of quantities of the corresponding mRNAs destined for translation (Gry et al., 2009; Vogel and Marcotte, 2012). Furthermore, it may be difficult to decipher accumulation of individual metabolites in chloroplasts using the metabolomics approach, as native MEV/MEP (methylerythritol-4-phosphate) pathways also function in plants.

Studies on A. annua high-artemisinin-producing and low-artemisinin-producing chemotypes suggest different biosynthetic routes beyond artemisinic aldehyde intermediate (Bertea et al., 2005; Covello, 2008). The two chemotypes had different CYP71AV1 enzyme activities (Ting et al., 2013), and the activity of the Dbr2 promoter also was implicated (Yang et al., 2015). We chose a strong constitutive promoter, the 35S promoter of CaMV with double enhancer, to drive the expression of Dbr2 gene and thereby generate excess DBR2 transcripts, to favor artemisinic aldehyde flux toward DHAA and eventually artemisinin biosynthesis.

There were no adverse effects on the growth or morphology of DT plants producing artemisinin (Supplemental Figure 6A), and the minor decrease in pigments had no observable effect on the whole plants. Nonetheless, the pigment decrease in DT2 and DT4 may have been caused by the consumption of IPP by the artemisinin pathway. The marginally lower Y(NPQ), in DT3, most likely explains the reduced Y(II) value (Supplemental Figure 6C). In general, a high Y(NPQ) value indicates that plants are more adapted to light stress and more efficiently direct it toward photosynthesis (Gao et al., 2011).

In our case, the co-expression of two pathways did not imbalance plant metabolism to a sufficient degree to cause gross aberrant effects. This is in contrast to a recent report, which found extensive off-target effects induced by overexpression, in tobacco, of FPP synthase and squalene synthase in the cytosol, chloroplast, or both. Their large changes were attributed to metabolite-mediated anterograde and/or retrograde signaling regardless of the level of transgene expression or end product, due to imbalanced metabolic pools (Pasoreck et al., 2016). These negative impacts were not related to the number of genes expressed or the level of expression. However, it is possible that our co-expression strategy to increase the pool IPP and the physiological regulator IDI provided the observed metabolic balance not observed by Pasoreck et al. (2016). IPP does appear to pass between the chloroplastic and cytosolic compartments independent of the synthetic site (Bick and Lange, 2003; Dudareva et al., 2005; Kumar et al., 2012). This is supported by our study on the levels of squalene in untransformed (wild-type), chloroplast transformed (T-MEV), and chloroplast and nuclear transformed (DT) plants. GC–MS/MS analysis showed a significant increase in the overall yield of squalene in DT and T-MEV plants compared with wild-type plants (Supplemental Figure 7), which could be due to a unidirectional transport of metabolites across the chloroplast envelope membrane or to direct effects of excess FPP. However, low levels of squalene were present in the high-artemisinin-producing DT2 and DT4 (Table 1), indicating a preferential consumption of IPP and FPP by the artemisinin pathway in these plants.

Table 1.

Metabolite Profile in Tobacco Lines (DT1–DT4) Engineered with DHAA Biosynthetic Pathway.

Metabolites	WT	T-MEV	DT1	DT2	DT3	DT4
Artemisinin	ND	ND	331 ± 71	773 ± 27	567 ± 76	877 ± 40
Dihydroartemisinic acid	ND	ND	62 ± 3	91 ± 6	82 ± 3	150 ± 10
Amorphadiene	ND	ND	10 ± 5	53 ± 8	32 ± 5	60 ± 5
IPP/DMAPP	80 ± 6	114 ± 12	175 ± 13	228 ± 5	186 ± 13	253 ± 20
Squalene	2.77 ± 0.1	10 ± 0.5	27 ± 0.5	22 ± 1.1	26 ± 0.4	21 ± 0.1

Values are mean ± SD. ND, not detected. Yield of all metabolites is mentioned in μg/g dry weight except for IPP/DMAPP, which is expressed as ng/g fresh weight, and squalene as μg/g fresh weight. All values have been rounded to the nearest integer.

The primary cost of artemisinin is the very expensive process used to extract and purify the drug from A. annua. Our report indicates that this is an unnecessary step and that artemisinin can be produced in edible leaves. We provide a proof-of-concept demonstration of the functionality of tobacco-biosynthesized artemisinin in DT lines. This finding suggests that artemisinin can become accessible and affordable to the large global population living in malarial regions. Malaria remains a global health problem with nearly 3.2 billion people at risk (almost half of the world’s population) (http://www.who.int/mediacentre/factsheets/fs094/en/). Results demonstrated through in vitro (Figure 7A) and in vivo animal studies (Figure 7A and 7B) are critical for functional evaluation of foreign products expressed in plants.

Earlier reports had shown that none of the substances in tobacco leaves, including nicotine, reduced the parasitemia load when wild-type tobacco leaves were fed orally to mice (Davoodi-Semiromi et al., 2010). Mice fed orally with extracts from wild-type and commercially available pure drug resulted in similar parasite loads. This was probably due to the poor bioavailability of the pure (commercial) artemisinin, which is rapidly degraded by hepatic and intestinal cytochrome enzymes (Svensson and Ashton, 1999). The enhanced efficacy of artemisinin delivered via intact whole-plant DT4 leaves compared with pure (commercial) artemisinin is probably due to the bioencapsulation of this bioactive compound. A previous report on the oral feeding of A. annua (Weathers et al., 2014; Elfawal et al., 2015) suggested that higher efficacy of orally fed A. annua leaves could be due to the protection of flavonoids or other antioxidants present in plants or animal feed. We have previously shown that plant cells protect biopharmaceuticals from acids and enzymes in the stomach via bioencapsulation. The gut microbes release these drugs in the gut lumen by the action of commensal microbes, where they are absorbed and delivered to the blood circulation system (Limaye et al., 2006; Kwon et al., 2013; Kohli et al., 2014; Sherman et al., 2014; Shil et al., 2014). Human digestive enzymes are incapable of breaking down glycosidic bonds in plant cell wall carbohydrates, and this process is performed by the gut microflora, which has evolved to break down every component of plant cell wall (Flint et al., 2008; Martens et al., 2011). Similarly, when intact plant tissue and cells containing drugs reach the gut, commensal microbes gradually digest plant cell walls and release drugs relatively slowly. Furthermore, DT plants biosynthesized both DHAA and artemisinin. DHAA has been shown to have higher bioavailability when compared with pure artemisinin and may play a critical role in the treatment of malaria (Li et al., 1998). However, for clinical translation, this concept should be transferred to proven edible plants for oral delivery studies (Kwon and Daniell, 2016).

Several other biopharmaceuticals have been made in tobacco leaves for the treatment of infectious diseases such as tuberculosis (Lakshmi et al., 2013), cholera (Davoodi-Semiromi et al., 2010), anthrax (Watson et al., 2004; Koya et al., 2005), and plague (Arlen et al., 2008). These could only be tested under field conditions under USDA-APHIS approval (Arlen et al., 2007). Chloroplast engineering is not regulated by USDA-APHIS because this approach does not use plant pathogens or pathogen regulatory sequences (Chan and Daniell, 2015; Kwon and Daniell, 2015). Therefore, DT lines, as created in our study, may have significant regulatory advantages for further development.

In conclusion, we have achieved biosynthesis of the potent anti-malarial drug, artemisinin, in tobacco plants at enhanced levels (~0.8 mg/g dry weight) by introducing two mega-biosynthetic pathways separately into the chloroplast and nuclear genomes, without perturbing the plant health. This is the first report of sequential chloroplast and nuclear genome transformation successfully expressing two mega-biosynthetic pathways and accomplishing complete artemisinin biosynthesis in an alternative host plant.

Most importantly, we report the stability of biosynthesized artemisinin bioencapsulated in plant cells upon oral delivery. The functionality of tobacco DT leaf artemisinin (fed as crushed leaves) was demonstrated by in vitro and in vivo antimalarial efficacy studies and was considerably better than pure artemisinin. This could significantly increase drug effectiveness and lower the cost of malaria treatment. Malaria disproportionately affects low socio-economic areas and children, the most susceptible to this disease, die in their millions every year. A similar approach could be extended to edible crops, such as lettuce, for artemisinin biosynthesis, and delivering the drug orally may provide a low-cost, environmentally friendly, and supplementary source of bioencapsulated artemisinin, eliminating expensive extraction and purification processes.

METHODS

Construction of Plant Transformation Vectors

The tobacco-specific chloroplast transformation vector (pCT-MEV) was designed as reported by Kumar et al. (2012). The nuclear transformation vector pCox-Nuc-Art was designed in silico, and contains the genes IDI (IPP/DMAPP isomerase; gi|4633512), FPP synthase (gi|216582) from E. coli, a modified amorpha-4,11-diene synthase gene (Anthony et al., 2009), amorpha-4,11-diene C-12 oxidase (CYP71AV1, gi|82548247), the A. annua cytochrome P450 reductase (AACPR, gi|83854016), and the double bond reductase-2 (DBR2; gi|494114616). Supplemental Figure 5 contains the entire sequence of vector pCox-Nuc-Art. For efficient expression, each gene was placed under the control of an individual promoter. The APX constitutive promoter (Park et al., 2010) drives the expression of IDI. CYP71AV1 and DBR2 were controlled by double enhancer CaMV 35S promoter (Guilley et al., 1982) while FPP and AACPR were driven by the CsVMV promoter. The hpt II gene, conferring resistance to hygromycin, was placed under the control of CaMV 35S promoter. The DHAA transgene cassette was released from the pUC57 plasmid by double digestion and ligated into the binary vector pCAMBIA1300 (kindly provided by Dr. Narendra Tuteja, ICGEB, New Delhi). Positive clones, confirmed by restriction digestion using enzymes BamHI and HindIII (NEB, USA), were electroporated into Agrobacterium strain EHA105 for nuclear transformation of the T-MEV plants. Transformants were selected on Luria–Bertani plates containing 50 mg/l kanamycin and 10 mg/l rifampicin. In addition, the ADS gene was targeted to mitochondria by use of cox4 mitochondrial signal peptide. AACPR and CYP71AV1 genes were targeted to chloroplasts using a chloroplast transit peptide (gi|20023|emb|X02353.1| Tobacco gene for ribulose 1,5-bisphosphate carboxylase small subunit) and the DBR2 gene was targeted to chloroplasts by the rbcS1 transit peptide.

Transformation and Selection of Tissues

Chloroplast transformation was performed in Nicotiana tabacum cv. Petit Havana SR1 leaves with pCT-MEV vector using S550d gold particles according to manufacturer’s instructions (Seashell Technology, CA, USA). Transgenic tissues were selected on RMOP medium containing spectinomycin (500 mg/l). Homoplasmy in the transformed plants (T-MEV) was obtained on antibiotic-containing selection medium. For nuclear transformation, homoplastomic tobacco plants (containing the yeast mevalonate pathway) were re-transformed with Agrobacterium tumefaciens (strain EHA105) containing a pCox-Nuc-Art vector. The doubly transformed tissues were selected on RMOP medium containing 50 mg/l hygromycin and 500 mg/l spectinomycin.

Molecular Analysis of Double Transformants

To test integration of the DHAA biosynthetic cassette, we tested doubly transformed tobacco plants for transgene integration by PCR using gene-specific primers. Primers for ADS and DBR2 are listed in Supplemental Table 1 and were synthesized according to their gene sequence (Sigma, Bengaluru, India).

Southern Blot Analysis of Double Transformants

Prior to nuclear transformation, homoplasmy of the engineered MEV pathway in the T-MEV plants was confirmed by Southern blot. A hybridization fragment, corresponding to 5.5 kb in the T-MEV plants, confirmed the homoplastomic status of transgenic plants (Supplemental Figure 1B). The hybridization fragment, corresponding to 1.4 kb, represented the native chloroplast genome of wild-type control plants (Supplemental Figure 1B). Transgene copy number in greenhouse-grown, DT plants carrying the pCox-Nuc-Art vector was determined using Southern blots following the DHAA and artemisinin biosynthesis study. Genomic DNA (~15 μg), was isolated from transgenic plants DT1–DT4 and control wild-type plants using a modified CTAB method (Porebski et al., 1997). It was digested overnight with EcoRV and separated on a 0.8% agarose gel. It was further transferred to Hybond N⁺ membranes (GE Healthcare, Buckinghamshire, UK). A PCR-amplified DBR2 gene (441 bp) was used as a probe (Figure 2A). Probe labeling of DNA blots was carried out using the AlkPhos direct-labeling kit, and signals were detected with the CDP-Star detection reagent (GE Healthcare).

RNA Transcript Analysis

Total RNA was isolated from wild-type and DT1–DT4 plants using the RNeasy Plant Mini Kit (Qiagen) according to the vendor’s instructions. Total RNA was treated with DNase (Qiagen) to remove any remaining genomic DNA contamination. The cDNA synthesis from RNA was carried using the SuperScript III First-strand cDNA synthesis kit (Invitrogen, Carlsbad, CA, USA). The resulting cDNA was amplified using gene-specific primers (Supplemental Table 1) to examine the mRNA transcripts of the transgenes introduced into the nuclear genome of DT plants.

Analysis of Artemisinin and Intermediate Metabolites

Artemisinin and DHAA levels in leaves of greenhouse-grown DT plants were determined by LC–MS/MS (Suberu et al., 2013), using an Agilent 1200 series reverse-phase C-18 column (4.6 × 100 mm × 3.5 μm) at a flow rate of 0.8 ml/min coupled with a 3200 Q TRAP linear ion trap mass spectrometer (Applied Biosystems, ABSCIEX, Singapore), equipped with a TurboIon spray source. Data were acquired and evaluated by Analyst software (version 1.6, ABSCIEX, Singapore). The MS was operated in positive mode and in MRM (Multiple Reaction Monitoring) mode. Declustering potential (DP), collision energy (CE), and collision cell exit potential (CXP) of each transition were optimized. The major MS/MS fragmentation patterns of both artemisinin (Supplemental Figure 2A) and DHAA were determined. For optimization of MS-dependent parameters, artemisinin (Sigma, Bengaluru, India) and DHAA (Apin Chemicals, Oxon, UK) standards in acetonitrile were injected separately using a gas-tight syringe (Harvard Apparatus, Holliston, USA). Quantification was done using MultiQuant software (Colangelo et al., 2013) and the R² value was 0.9987.

Analysis of IPP and DMAPP

Wild-type and DT leaves were lyophilized. Approximately 1 g of each sample was crushed in liquid nitrogen and extracted on ice, for 30 min in methanol (5 ml) in a water bath sonicator (Branson Ultrasonic, Danbury, USA). An IPP (Sigma, Bengaluru, India) standard was diluted with methanol to optimize MS-dependent parameters. The MS was operated in negative ion and MRM mode. Pure IPP was infused at a flow rate of 10 μl/min by a gas-tight syringe (Harvard Apparatus, Holliston, USA). The major MS/MS fragment for IPP (Supplemental Figure 3A), DP, CE, and CXP for each transition were optimized. MRM mode was used to quantify IPP in each sample using MultiQuant software (Colangelo et al., 2013). Metabolites were separated using an Agilent 1200 series reverse-phase C-18 column (4.6 × 100 mm × 3.5 μm) at a flow rate of 0.8 ml/min.

GC–MS Analysis of Amorphadiene

Leaves from the greenhouse-grown wild-type and DT plants (μ1 g) were crushed in liquid nitrogen and extracted with ethyl acetate (2 ml) for 30 min in a water bath sonicator (Branson Ultrasonic, Danbury, USA). The extracts were concentrated by evaporation under nitrogen. A 1-μl aliquot from the extract was analyzed by GC–MS, using an Agilent 7890A with an HP-5 column (30 m × 0.25 mm internal diameter, 0.25 μm film thickness) and a 7000C triple quadrupole GC–MS/MS spectrometer. The MS was operated in electron ionization mode and scanning mode (50–500 m/z). Amorphadiene was detected by comparing the retention time and mass spectra with an authentic standard (kindly provided by Amyris Biotechnology, Emeryville, CA, USA).

In Vitro Functionality of Artemisinin against P. falciparum

Malarial parasite in vitro assays were conducted to test the functionality of DT tobacco-produced artemisinin. A Plasmodium falciparum 3D7 culture was synchronized to the ring stage by two consecutive sorbitol treatments separated by 4 h, then cultured for 40 h and allowed to re-invade RBCs and cultured in RPMI-1640 medium (Gibco BRL, Grand Island, USA). Growth inhibition of the P. falciparum 3D7 cultures was determined by the SYBR Green assay (Smilkstein et al., 2004). In brief, a parasite culture with an initial parasitemia of 1%, in 2% hematocrit, was treated with different doses of DT tobacco artemisinin in 96 well plates, each well containing 100 μl of culture of parasite-infected RBCs. DMSO and commercially sourced artemisinin (Sigma-Aldrich, Bengaluru, India) served as negative and positive controls, respectively. Plates containing DT plant extracts and parasites were incubated at 37°C for 48 h. Parasite growth progression was monitored by preparing Giemsa-stained blood smears after 24 h and 48 h. After each time point, 100 μl of lysis buffer (20 mM Tris, 5 mM EDTA, 0.008% saponin, 0.08% Triton X) containing 1×SYBR Green, was added to the cultures and incubated at 37°C for 1 h. Fluorescence was read using a PerkinElmer Victor X3 Multilabel Plate Reader (PerkinElmer, Waltham, MA, USA) at 485/530 nm. The obtained values were plotted, and the IC₅₀ of artemisinin was calculated using linear regression analysis.

In Vivo Functionality of Artemisinin in the Murine Model

Cryopreserved blood from Plasmodium-infected mice was revived by re-injecting 200 μl of blood intraperitoneally into Balb/C mice. Giemsa-stained blood smears were made to determine percentages of parasitemia. After 1 week a mouse was euthanized, and blood was collected for subsequent experimentation. Four-week-old Balb/C mice were injected with 10⁵ RBCs infected with P. berghei and percentage of parasitemia was determined from Giemsa-stained blood smears prepared from the tail region. Mice infected with P. berghei were divided into three groups (four mice in each group). One group of mice was orally given a single dose (24 mg/kg) of pure artemisinin drug (Sigma-Aldrich, Bengaluru, India), dissolved in 60 μl of DMSO and reconstituted in water, for nine consecutive days from the third day post infection; the second group of mice was fed oven-dried leaves from DT4 (500 mg), crushed into fine powder with a homogenizer, mixed with 1 ml of water, and fed by gavage needle (18 gauge, curved, 2 inches, and 2.25 mm ball diameter); the third group was fed dried leaves from wild-type plants, prepared as were the DT leaves. All preparations were freshly prepared before the feeding assay and all treatments were performed in parallel. Treated parasite cultures were also analyzed microscopically.

Statistical Analysis

The differences of artemisinin and IPP produced in wild-type, T-MEV, and doubly transgenic lines (DT1–DT4) were tested by one-way ANOVA with Duncan’s multiple range test, followed by a post hoc analysis (Tallarida and Murray, 1987). Analysis of all samples was done in triplicate (n = 3).