Isoprenoid biosynthesis in dandelion latex is enhanced by the overexpression of three key enzymes involved in the mevalonate pathway

Background Latex from the dandelion species Taraxacum brevicorniculatum contains many high-value isoprenoid end products, e.g. triterpenes and polyisoprenes such as natural rubber. The isopentenyl pyrophosphate units required as precursors for these isoprenoids are provided by the mevalonate (MVA) pathway. The key enzyme in this pathway is 3-hydroxy-methyl-glutaryl-CoA reductase (HMGR) and its activity has been thoroughly characterized in many plant species including dandelion. However, two enzymes acting upstream of HMGR have not been characterized in dandelion latex: ATP citrate lyase (ACL), which provides the acetyl-CoA utilized in the MVA pathway, and acetoacetyl-CoA thiolase (AACT), which catalyzes the first step in the pathway to produce acetoacetyl-CoA. Here we isolated ACL and AACT genes from T. brevicorniculatum latex and characterized their expression profiles. We also overexpressed the well-characterized HMGR, ACL and AACT genes from Arabidopsis thaliana in T. brevicorniculatum to determine their impact on isoprenoid end products in the latex. Results The spatial and temporal expression profiles of T. brevicorniculatum ACL and AACT revealed their pivotal role in the synthesis of precursors necessary for isoprenoid biosynthesis in latex. The overexpression of A. thaliana ACL and AACT and HMGR in T. brevicorniculatum latex resulted in the accumulation of all three enzymes, increased the corresponding enzymatic activities and ultimately increased sterol levels by ~5-fold and pentacyclic triterpene and cis-1,4-isoprene levels by ~2-fold. Remarkably high levels of the triterpene precursor squalene were also detected in the triple-transgenic lines (up to 32 mg/g root dry weight) leading to the formation of numerous lipid droplets which were observed in root cross-sections. Conclusions We could show the effective expression of up to three transgenes in T. brevicorniculatum latex which led to increased enzymatic activity and resulted in high level squalene accumulation in the dandelion roots up to an industrially relevant amount. Our data provide insight into the regulation of the MVA pathway in dandelion latex and can be used as a basis for metabolic engineering to enhance the production of isoprenoid end products in this specialized tissue. Electronic supplementary material The online version of this article (doi:10.1186/s12870-017-1036-0) contains supplementary material, which is available to authorized users.


Background
The dandelion species Taraxacum brevicorniculatum belongs to more than 12,500 plant species that produce latex, a milky sap, in specialized parenchyma cells called laticifers [1]. Among other secondary metabolites and proteins, laticifers contain a large variety of industrially valuable isoprenoid end products, including triterpenes such as α-amyrin, β-amyrin, lupeol and taraxasterol, and polyisoprenes such as natural rubber [2]. The components of the rubber transferase complex have been investigated in rubber-producing plants like the rubber tree (Hevea brasiliensis), lettuce (Lactuca sativa) and T. brevicorniculatum [3][4][5]. Labeling experiments in H. brasiliensis showed that the cytosolic mevalonate (MVA) pathway provides the precursor isopentenyl pyrophosphate (IPP) for rubber biosynthesis in the laticifers [6,7]. We therefore characterized the key enzymes of the MVA pathway and upstream reactions in T. brevicorniculatum latex [see Fig. 1].
In several previous studies, 3-hydroxy-methyl-glutaryl-CoA reductase (HMGR) has been identified as the rate-limiting enzyme in the MVA pathway [8][9][10][11][12]. Plant HMGRs are characterized by four domains: a variable cytosolic N-terminus, two endoplasmic reticulum transmembrane domains, a variable region between the transmembrane domains, and a cytosolic C-terminus containing the highly-conserved catalytic domain [13,14]. Plant HMGRs are regulated at the post-transcriptional and post-translational levels, e.g. by phosphorylation and inhibition by their own N-terminus [15][16][17][18]. T. brevicorniculatum produces three HMGRs encoded by the genes TbHMGR1, TbHMGR2 and TbHMGR3, with TbHMGR1 predominantly expressed in the latex [19]. MVA pathway engineering in plants is challenging [12,20]. The overexpression of endogenous key enzymes often fails to increase the yield of end products due to negative feedback or ratelimiting downstream steps, and the regulatory mechanisms therefore need to be characterized in more detail [12,21].
Enzymes acting upstream of HMGR in the latex might act as additional rate-limiting steps for isoprenoid biosynthesis, including ATP citrate lyase (ACL) and acetoacetyl-CoA thiolase (AACT) [see Fig. 1]. Acetyl-CoA is required in many anabolic and catabolic pathways in plants and must be synthesized in every cell compartment [22]. ACL cleaves citrate into oxaloacetate and acetyl-CoA in the cytosol. It belongs to the thiokinase superfamily, which also includes succinyl-CoA synthetase (SCS) and citrate synthase (CS) among others [23]. SCS is localized in mitochondria and catalyzes the reversible conversion of succinyl-CoA to succinate and CoA. Therefore, ATP-binding and CoA-binding domains are conserved between SCS and ACL. Arabidopsis thaliana ACL is a 500 kDa heteromeric protein comprised of ACLA and ACLB subunits. AtACLA is very similar to the β-subunit of AtSCS, whereas AtACLB is more closely related to the α-subunit of AtSCS and to AtCS [24]. Transgenic A. thaliana plants expressing AtACLA-1 antisense RNA have a severe dwarf phenotype which underlines the importance of this enzyme for anabolic processes [25]. Furthermore, the overexpression of an A. thaliana AtACL(AB) fusion protein in dandelion latex increased ACL activity and led to the accumulation of various isoprenoid end products, highlighting its ability to increase flux through the MVA pathway [26]. The effective shuttling of acetyl-CoA into the MVA pathway requires a Claisen-type condensation reaction between two acetyl-CoA molecules to form acetoacetyl-CoA, which is catalyzed by AACT. In plants, catabolic [27]. The MVA pathway requires type II thiolases, and the corresponding enzymes have been characterized in A. thaliana, H. brasiliensis and Medicago sativa. In A. thaliana, the AtAACT1 and AtAACT2 genes have different spatiotemporal expression profiles and knockout phenotypes: Ataact1 T-DNA knockout lines show normal growth and development whereas Ataact2 T-DNA knockout lines are embryonic lethal [28]. Furthermore, AtAACT2 RNAi lines are sterile and are characterized by a reduced cell size and cell number, indicating that AtAACT2 is essential among others for sterol biosynthesis via the MVA pathway, which is pivotal for membrane biosynthesis. In H. brasiliensis, HbAACT1 is predominantly expressed in the latex whereas HbAACT2 and HbAACT3 are mainly expressed in mature leaves [7]. In M. sativa, the key role of MsAACT1 in isoprenoid biosynthesis was shown by overexpression experiments, which resulted in higher thiolase activity and the accumulation of squalene [29].
The dandelion ACL and AACT genes have not been characterized in detail but may play a key role in isoprenoid biosynthesis as in other plants. We therefore identified ACL and AACT genes predominantly expressed in T. brevicorniculatum latex and characterized their spatiotemporal expression profiles. We also overexpressed the wellcharacterized A. thaliana HMGR, ACL and AACT genes in T. brevicorniculatum to increase flux through the MVA pathway predominantly in the latex. This allowed us to determine the impact of the three enzymes on the accumulation of isoprenoid end products such as triterpenes and polyisoprenes.

Plant material and cultivation conditions
T. brevicorniculatum seeds were obtained from the Botanical Garden Marburg (Germany) and were cultivated at 18°C and 20 klux with a 16-h photoperiod in controlled growth chambers or in the greenhouse. Plants were cultivated in a pre-fertilized 1:1 mixture of standard soil (ED73 Einheitserde, Fröndenberg, Germany) and garden mold (Botanical Garden Münster, Germany). They were fed every 4 weeks with a commercial fertilizer according to the manufacturer's recommendations (Hakaphos Plus, Compo GmbH, Münster, Germany). Arabidopsis thaliana ecotype Columbia (Col-0) seeds were acquired from the Nottingham Arabidopsis Stock Centre (University of Nottingham, Loughborough, United Kingdom, http://arabidopsis.info) and cultivated as stated above. Seeds of Nicotiana benthamiana laboratory isolate (LAB) [30] were obtained from the Sainsbury Laboratory (John Innes Centre, Norwich, United Kingdom) and cultivated as stated above.

RNA extraction and cDNA synthesis
Total RNA was extracted from T. brevicorniculatum root latex using the RNeasy Lipid Tissue Kit (Qiagen, Hilden, Germany), and from T. brevicorniculatum root, peduncle, leaf and flower tissues using the innuPREP RNA Mini Kit (Analytik Jena, Jena, Germany), each according to the manufacturer's instructions. Full-length cDNA was synthesized from 500 ng total RNA using PrimeScript RT Master Mix (TaKaRa, Clontech, Saint-Germain-en-Laye, France) according to the manufacturer's instructions. Fulllength cDNA sequences were obtained by standard RACE procedures. Oligonucleotide sequences are shown in Additional file 1 [see Additional file 1].

Quantitative RT-PCR (qRT-PCR)
Quantitative RT-PCR analysis was carried out as previously described [26] with slight modifications. Latex was harvested from 8-week-old roots of wild-type and transgenic T. brevicorniculatum plants for RNA isolation, cDNA synthesis and qRT-PCR using the KAPA SYBR® FAST qPCR Master Mix (Peqlab, Erlangen, Germany) in a Bio-Rad iCycler real-time PCR system. Three different T. brevicorniculatum reference genes were included in the qRT-PCR analysis: The glyceraldehyde-3-phosphate dehydrogenase (TbGAPDH) and the elongation factor-1 alpha (TbEF1α) as already described [26,33], as well as the ribosomal protein L27 (TbRP) based on an expressed sequence tag (EST) (GenBank GO664824) of the T. koksaghyz root cDNA library. The relative expression levels of each target gene were calculated using Bio-Rad CFX Manager 3.1 software (Bio-Rad Laboratories Inc., Hercules, USA). All oligonucleotide sequences for the reference, endogenous and transgene expression analysis are shown in Additional file 1. Primer efficiencies, amplification factors and usage of formamide for enhancing primer specificity are shown in Additional file 2.

SDS-PAGE and western blot analysis
Taraxacum brevicorniculatum protein was extracted and analyzed as previously described [26] with slight modifications. The rubber extraction buffer contained 100 mM Tris (pH 7.5), 350 mM sorbitol, 10 mM NaCl, 5 mM MgCl 2 , and 5 mM dithiothreitol (DTT). A rabbit antibody against A. thaliana ACLB was kindly provided by Yves Poirier (Department of Plant Molecular Biology, Université de Lausanne, Switzerland) and was diluted 1:2000 before use. A rabbit antibody against the A. thaliana HMGR1 catalytic domain was kindly provided by Hubert Schaller (Institut de Biologie Moléculaire des Plantes, Université de Strasbourg, France) and diluted 1:7500 before use. Both were detected with a secondary alkaline phosphatase (AP)-conjugated goat anti-rabbit IgG antibody diluted 1:10,000 before use. A mouse antibody against A. thaliana AACT2 was kindly provided by Basil Nikolau (Roy J. Carver Department of Biochemistry, Biophysics and Molecular Biology, Iowa State University, USA) and was diluted 1:1000 before use as previously described [28]. This was detected with a secondary APconjugated goat anti-mouse IgG antibody diluted 1:10,000 before use.

Analysis of enzyme activity
ACL activity in dandelion latex from 8-week-old roots of T. brevicorniculatum plants was determined as previously described [26]. AACT activity was determined by measuring the reduction of 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB) by free CoA as previously described [34] with slight modifications. We incubated 5 μg of crude protein extracted from the latex of 8-week-old roots of T. brevicorniculatum plants in 95 μl reaction buffer (50 mM Tris/HCl, 0.5 M EDTA, 10 mM acetyl-CoA, 10 mM DTNB, pH 8.0) and measured the reduction of DTNB at 412 nm with a Tecan-Infinite®200 Pro (Tecan Group Ltd., Männedorf, Switzerland). To measure HMGR activity, crude latex harvested from 8-weekold roots of T. brevicorniculatum plants was transferred to ice-cold rubber extraction buffer, and~40 μg total protein was used in a [3-14 C]HMG-CoA incorporation assay as previously described [32,35].

Staining root cross-sections with Nile red
Taraxacum brevicorniculatum roots were sliced with a razor blade and incubated in a Nile red staining solution (1 mg/ml Nile red in dimethylsulfoxide (DMSO) stock solution diluted 1:150 in double-distilled water) for 30 min. Stained root cross-sections were analyzed under a Leica MZ16 F fluorescence stereomicroscope (Leica Microsystems GmbH, Wetzlar, Germany) using the GFP Plants filter set (excitation filter 450-490 nm, barrier filter 500-550 nm) and visualized with Leica Application Suite X software.

Statistical analysis
Normal distributions of gene expression levels and enzymatic activities at p < 0.05 were assessed using the Kolmogorov-Smirnov test in OriginPro 2016 (OriginLab, Northhampton, USA). Normally distributed datasets were analyzed using ANOVA with the post-hoc Tukey's honest significant difference test. Datasets which did not show a normal distribution and could not be transformed to be normally distributed were analyzed using the Mann-Whitney U test.

Results and discussion
Synthesis of the full-length ACLA, ACLB and AACT cDNA sequences from T. brevicorniculatum latex RNA The T. brevicorniculatum genome has not yet been sequenced, so expressed sequence tag (EST) databases at GenBank prepared from either root material of its close relative T. koksaghyz or from T. officinale (tissue not specified at GenBank) were searched for ACLA, ACLB and AACT cDNA sequences. Two different T. koksaghyz ACLA ESTs (GenBank accession numbers DR401254 and DR402930) as well as one T. officinale ACLA EST (DY808625) were found which covered the 5′ end of ACLA open reading frames based on comparisons with the A. thaliana ACLA genes At1g10670, At1g60810 and At1g09430. One full-length cDNA sequence could be amplified by 3'RACE using T. brevicorniculatum root latex RNA as template. The 1272-bp TbACLA1 cDNA (GenBank accession number KY765686) based on the overlapping ESTs DR401254 and DR402930 encoded a protein of 423 amino acids. Furthermore, two different T. koksaghyz ESTs (DR398498 and DR399293) were found with high similarity to the A. thaliana ACLB orthologs At3g06650 and At5g4946. By performing 5'RACE using the same template one full-length TbACLB cDNA (1827 bp, GenBank accession number KY765687) encoding a protein of 608 amino acids named TbACLB1 (based on EST DR398498) was produced. Finally, T. koksaghyz ESTs DR398619 and DR401218 showed high similarity to the A. thaliana AACT genes At5g47720 and At5g48230. One full-length cDNA was produced by 3'RACE again using T. brevicorniculatum root latex RNA as the template. The 1227-bp TbAACT1 cDNA sequence (GenBank accession number KY765685) based on EST DR398619 encoded a protein of 408 amino acids.
Transcriptional regulation of the TbACLA, TbACLB and TbAACT genes To determine whether the isolated TbACLA1, TbACLB1 and TbAACT1 genes encoded the dominant isoforms in the latex, qRT-PCR analyses were performed using RNA from the latex, root, peduncle, leaf and flower tissues of 10-week-old wild-type T. brevicorniculatum plants grown under greenhouse conditions. Furthermore, qRT-PCR primer pairs were included in the analysis that were based on the additional ESTs mentioned above which are partial cDNAs encoding a second isoform of ACLA, ACLB and AACT, named TbACLA2 (based on DY808625), TbACLB2 (DR399293) and TbAACT2 (DR401218), respectively. For comparison, parallel qRT-PCR tests were carried out on the TbHMGR1 and TbHMGR2 genes, whose expression profiles have already been determined by RT-PCR [19]. TbHMGR1 plays a pivotal role in the latex MVA pathway due to its expression predominantly in laticifers, whereas TbHMGR2 expression is barely detected in latex but it is expressed more strongly than TbHMGR1 in roots and leaves [19] [see Additional file 4]. Like TbHMGR1, we found that TbACLA1, TbACLB1 and TbAACT1 were predominantly expressed in the latex [see Fig. 2a-c]. The expression of these genes detected in all other tissues may reflect the presence of small amounts of latex in all parts of the dandelion plant, but might also reflect the necessity of a basal activity of theses enzymes to synthesize important MVA pathway-derived substances for regular plant development. Remarkably, an overall lower expression could be measured for TbACLB2 and TbAACT2 in all tissues analyzed compared to TbACLB1 and TbAACT1, respectively, whereas TbACLA2 showed a similar expression level in roots and even slightly increased expression levels in peduncle, leaf and flower tissues (3-4 fold) compared to TbACLA1 indicating that TbACLA2 is an important isoform in these tissues. However, the tissue-specific qRT-PCR analyses clearly demonstrate that TbACLA1, TbACLB1 and TbAACT1 represent the dominant isoforms in latex and therefore appear to be highly relevant components of the T. brevicorniculatum latex MVA pathway and are likely to be required for the production of isoprenoid end products in this specialized tissue.
The potential co-regulated expression of TbACLA, TbACLB and TbAACT during dandelion development was investigated by cultivating T. brevicorniculatum wild-type plants outside the greenhouse from March to November. Total root material (containing the greatest amount of latex in relation to other tissues) was pooled from three individual plants per month and tested by qRT-PCR [see Fig. 2d-f]. In accordance to the tissue expression analysis, TbACLB1 and TbAACT1 comprised higher expression levels over the whole year in the dandelion root compared to TbACLB2 and TbAACT2, respectively [see Fig. 2e and f], whereas in case of TbACLA the isoform TbACLA2 seemed to play an important role at defined developmental stages such as in October and November [see Fig. 2d]. Overall, TbACLA, TbACLB and TbAACT displayed highly similar expression patterns with expression peaks at the seedling stage in March, and also in July and November indicative of tight co-regulation, as already reported for the A. thaliana ACLA and ACLB subunits [24]. Given that A. thaliana ACLA and ACLB mRNA accumulates in the epidermis and trichomes of expanding leaves as well as the epidermis of growing organs [24], ACL gene expression seems to be associated with (and upregulated during) plant growth, as supported by the developmental and growth defects observed in A. thaliana ACL-RNAi plants [25]. Therefore, ACL activity in the MVA pathway may be pivotal when large quantities of isoprenoids are required, e.g. for the synthesis of membrane sterols during development. The expression peaks in early development and in the summer were coordinated with the upregulation of TbAACT in March and July. This indicated that the TbACL and TbAACT genes are coregulated to ensure the shuttling of precursors into the MVA pathway during these developmental stages which are characterized by rapid expansion of the total root volume and consequently also the differentiation of the specialized laticiferous cells. Interestingly, TbACL and TbAACT showed an additional expression peak in November, which is when dandelion plants accumulate high levels of polyisoprenes in the roots [36]. This might explain the highly active MVA pathway in the latex and therefore the additional upregulation of TbACL for providing precursors and TbAACT as the first enzymatic step in the cytosolic isoprenoid pathway.
Gene expression levels do not necessarily reflect the levels of AACT and HMGR enzyme activity due to downstream regulatory mechanisms such as phosphorylation, feedback inhibition and external stimuli [16,17,29,38,39]. Therefore, the posttranscriptional regulation of key MVA pathway enzymes will be an important additional aspect to address in future studies concerning the engineering of isoprenoid metabolism in dandelion latex.

Overexpression of ACL, AACT and HMGR orthologs from A. thaliana in dandelion latex
Having identified and characterized the T. brevicorniculatum latex-specific isoforms of ACL and AACT, we investigated the effects of overexpressing these enzymes, using the better-characterized A. thaliana orthologs to reduce the likelihood of endogenous gene inhibition. We previously reported that overexpressing the A. thaliana ATP citrate lyase fusion construct AtACL(AB) enhanced ACL activity in T. brevicorniculatum [26]. Therefore, a transgenic T. brevicorniculatum line expressing AtA-CL(AB) from our previous study [26] was used here and served as a control line to complete the setup. The A. thaliana genome contains two paralogous AACT genes (AtAACT1 and AtAACT2) although only AtAACT2 is ubiquitously expressed [28]. The AtAACT2 cDNA sequence was therefore cloned from A. thaliana leaf RNA and used for the stable transformation of wild-type T. brevicorniculatum plants.
HMGR catalyzes the rate-limiting step in the MVA pathway, and in A. thaliana the AtHMGR1 gene is ubiquitously expressed [40]. The catalytic activity of plant HMGRs is inhibited by the N-terminal domain and by phosphorylation of the C-terminus, so we tested the activity of two different variants of AtHMGR1 in the heterologous N. benthamiana system. Athmgrc1(S408A) comprises the catalytic domain including the mutated phosphorylation site at position 408, and this was 2.5-fold more active than the wild-type catalytic domain Athmgrc1 [see Additional file 5]. Athmgrc1(S408A) was therefore used for the stable transformation of wild-type T. brevicorniculatum plants.
The potential synergistic effect of AtACL(AB), AtAACT2 and Athmgrc1(S408A) overexpression was investigated by supertransforming transgenic T. brevicorniculatum plants expressing AtACL(AB) with constructs carrying AtAACT2 and Athmgrc1(S408A). All cDNAs were driven by the rubber elongation factor (REF) promoter to ensure preferential expression in the latex [41]. The constructs prepared for stable transformation are shown in Fig. 3a.
A set of 5-10 transgenic plant lines representing each transformation experiment was characterized by either RT-PCR or qRT-PCR in the T0 generation (data not shown). We then selected one line expressing the AtA-CL(AB) fusion construct (ACL-1), two lines overexpressing AtAACT2 (AACT-3 and AACT-4), two lines expressing Athmgrc1(S408A) (HMGR-7 and HMGR-8) and two lines expressing all three enzymes (ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5) for detailed analysis in the T1 generation. Transgene expression in the latex harvested from roots of 8-week-old plants from all seven transgenic lines was confirmed by qRT-PCR [see Additional file 6 A]. The expression levels of the endogenous TbACLA, TbAACT and TbHMGR1 genes were also compared to wild-type plants using the same technique [see Additional file 6 B]. No significant differences in endogenous gene expression levels were observed between wildtype plants and the transgenic lines, indicating that the overexpression of A. thaliana genes did not affect endogenous gene expression. We also measured the levels of the corresponding endogenous and heterologous proteins. Latex protein extracts obtained from roots were analyzed by western blot using antibodies specific for each enzyme [see Fig. 3b]. The anti-ACLB antibody revealed the presence of a 66 kDa band corresponding to the endogenous subunit in all lines and the presence of a 120 kDa band corresponding to the chimeric AtACL(AB) protein solely in the transgenic lines expressing the AtACL(AB) construct (ACL-1, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5). Likewise, the anti-AtAACT2 antibody revealed two protein bands of~43 kDa and~41 kDa, with the upper band corresponding to the endogenous TbAACT protein and the lower band corresponding to the AtAACT2 protein (the latter only present in transgenic lines overexpressing AACT: AACT-3, AACT-4, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5). Finally, the anti-HMGR antibody revealed one protein band of 63 kDa representing the predominant endogenous latex isoform (TbHMGR1) and another of~45 kDa representing the catalytic domain Athmgrc1(S408A), the latter present only in transgenic lines HMGR-7, HMGR-8, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5. These data confirmed that all transgenes were expressed in the latex as anticipated and were translated into the corresponding proteins without affecting endogenous protein levels.
Upregulation of the MVA pathway in T. brevicorniculatum latex enhances the accumulation of squalene, sterols, pentacyclic triterpenes and cis-1,4-isoprene Whole roots from three 12-week-old plants per line were harvested and tested for their phenotypic and metabolic characteristics. The transgenic lines showed no overt phenotypic aberrations (representative images of one plant per line are compared to a wild-type plant in Additional file 7 A). Furthermore, there were no significant differences between the wild-type plants and transgenic lines in terms of root fresh weight or root dry weight [see Additional file 7 B and C]. The overexpression of heterologous AtACL(AB), AtAACT2 and Athmgrc1(S408A) in the latex therefore had no significant impact on the growth or development of T. brevicorniculatum. To ensure no major costs arise from this overexpression of transgenes, it would be necessary to measure the overall plant fitness by e.g. evaluating the seed production, quality and quantity.
Freeze-dried roots were used for triterpene extraction and GC/MS analysis because the quantitative analysis of triterpenes in latex tissue as described in Huber et al. [42] was not feasible for comparative analysis of wild-type and transgenic plant lines. The quantity of triterpene precursors and HMGR (α-HMGR), or Ponceau S staining after protein transfer (lower part). Numbers on the left refer to molecular weight markers. c Relative ACL, AACT and HMGR enzymatic activity in the latex of 8-week-old transgenic and wild-type plants. The values of the wild-type plants were set to 1. Bars represent standard errors of three independent transgenic plants per line. Normal distribution at p < 0.05 was assessed using the Kolmogorov-Smirnov test. Different letters indicate a significant difference (ANOVA with Tukey's honest significant difference test (p < 0.01)) was considerably higher in transgenic lines HMGR-7, HMGR-8, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5 than the other lines [see Fig. 4a]. Although the levels of the precursor 2,3-oxidosqualene were only slightly (1.7-1.9-fold) higher in the transgenic lines compared to wild-type plants, large amounts of squalene accumulated in the roots of the triple-transgenic lines (up to 32 mg/g root dry weight). Squalene accumulation was therefore investigated morphologically in cross-sections of wild-type and transgenic roots [see Additional file 8]. Nile red, which stains lipophilic molecules, revealed the presence of globules resembling lipid droplets in the root sections of all four transgenic lines. These globules were much more abundant in lines ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5 than lines HMGR-7 and HMGR-8, and differed from laticifers in terms of size and position. These data suggest that excess squalene produced in these four transgenic lines is stored in lipid droplets outside the laticifers. These findings are supported by the observation of similar lipid droplets in tobacco Fig. 4 Influence of AtACL(AB), AtAACT2 and Athmgrc1(S408A) overexpression in dandelion latex on isoprenoid end products in 12-week-old T. brevicorniculatum roots. a Quantitative analysis of triterpene precursors (squalene and 2-3-oxidosqualene), sterols (campesterol, stigmasterol and sitosterol) and pentacyclic triterpenes (taraxasterol, α-amyrin, β-amyrin and lupeol) by GC/MS in freeze-dried roots of transgenic and wildtype (Wt) T. brevicorniculatum plants. Bars represent standard errors of three independent transgenic plants per line. Normal distribution at p < 0.05 was assessed using the Kolmogorov-Smirnov test. Asterisks represent significant differences relative to wild-type (ANOVA with Tukey's honest significant difference test (p < 0.01)). b Quantitative analysis of cis-1,4-isoprene by 1 H-NMR in freeze-dried roots of transgenic and wild-type T. brevicorniculatum plants. Bars represent standard errors of three independent transgenic plants per line. Normal distribution at p < 0.05 was assessed using the Kolmogorov-Smirnov test. Asterisks represent significant differences relative to wild-type (ANOVA with Tukey's honest significant difference test (p < 0.01)). c Dolichol composition of freeze-dried root material determined by LC/MS. A mixture of dolichols (13-21 isoprene units) was used as a standard. One representative result for each transgenic line and a wild-type control is shown (Nicotiana tabacum) leaves expressing full-length or truncated HbHMGR1 [10,15].
The average quantity of squalene accumulating in the roots of the triple-transgenic lines was 31.8 mg/g dry root weight for line ACL-AACT-HMGR-4 and 26.95 mg/g dry root weight for line ACL-AACT-HMGR-5. With the exception of a genetically engineered strain of Pseudozyma sp. (70 mg/g dry cell weight (DCW)) [43], this exceeds the highest squalene yields thus far achieved in microbes such as Saccharomyces cerevisiae (18.5 mg/g DCW), Rhodopseudomonas palustris (15.8 mg/g DCW) and Schizochytrium mangrovei (1.17 mg/g DCW) [44][45][46], but is lower than in plant tissues such as Amaranthus spp. seed oil (7% w/w) and olive oil (30% w/w) [47,48]. However, the genetically engineered squalene content in transgenic T. brevicorniculatum roots by simultaneous overexpression of ACL, AACT and HMGR is within an industrially relevant range. Interestingly, the large amount of squalene present in the lipid droplets in the T. brevicorniculatum root does not appear to be toxic, which is consistent with experiments in S. cerevisiae showing that squalene accumulation is toxic only when lipid droplet formation is impaired [49]. The similar levels of 2,3-oxidosqualene in wild-type and transgenic plants despite the substantial differences in squalene levels indicate that the formation of 2,3-oxidosqualene catalyzed by squalene epoxidase might act as the main bottleneck between the MVA pathway and downstream isoprenoid biosynthesis.
We also analyzed the levels of other sterols, such as campesterol, stigmasterol and sitosterol. Overall, the quantity of sterols was higher in the roots of the transgenic lines than wild-type plants but the extent of sterol accumulation was dependent on the complement of transgenes. In the lines expressing ACL or AACT, the sterol levels were less than 2-fold higher than wild-type levels. In contrast, the sterol levels were 4.2-fold, 3.3fold, 5.2-fold and 5.1-fold higher than wild-type in transgenic lines HMGR-7, HMGR-8, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5, respectively. This shows that the overexpression of ACL, AACT and HMGR increased flux through the MVA pathway in terms of IPP units directed towards sterol biosynthesis. Transgenic T. brevicorniculatum lines expressing HMGR produced the highest quantity of sterols, confirming that HMGR is the pivotal rate-limiting enzyme in the MVA pathway. Similarly, the overexpression of HbHMGR1 in tobacco achieved a 6-fold increase in sterol levels [10] and the seed-specific expression of a truncated HbHMGR1 led to a 3.2-fold increase in seed sterol accumulation [15]. Our novel approach involved the simultaneous expression of three enzymes involved in the MVA pathway, achieving even higher sterol levels than the expression of HMGR alone.
Pentacyclic triterpenes such as taraxasterol, α-amyrin, βamyrin and lupeol are also important triterpene components of T. brevicorniculatum latex because of their potential medical applications. For example, α-amyrin from Bombax malabaricum flowers is active against a range of bacteria and fungi [50] whereas α-amyrin, β-amyrin and lupeol display anti-inflammatory activities in mice and mammalian cell lines [51][52][53]. Taraxasterol has antiallergic, anti-inflammatory and anti-carcinogenic effects [54][55][56][57]. These compounds are therefore valuable products in the biotechnology industry and efforts have already been undertaken to increase their yields, e.g. the addition of abiotic elicitors to root callus suspension cultures of T. officinale, a close relative of T. brevicorniculatum, to increase the production of taraxasterol [58]. In our transgenic lines overexpressing a single enzyme, the pentacyclic triterpene content increased by 1.5-1.7-fold compared to wild-type plants. In the triple-transgenic lines ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5, the corresponding increase was 1.9-2.2-fold, suggesting that the overexpression of three MVA pathway enzymes has a synergistic effect. Such effects could therefore enhance the industrial production of pentacyclic triterpenes, especially in combination with the metabolic engineering of downstream pathways.
Next, we set out to determine the total cis-1,4-isoprene content in T. brevicorniculatum roots by 1 H-NMR spectroscopy, which includes the analysis of polyisoprenes such as dolichols as well as high-molecular-weight long-chain poly(cis-1,4-isoprenes) that form the main constituent of natural rubber [see Fig. 4b]. All transgenic lines accumulated similar or slightly higher levels of cis-1,4-isoprene compared to wild-type plants, with lines ACL-1, ACL-AACT-HMGR-4 and ACL-AACT-HMGR-5 producing the highest levels (1.4-fold, 1.7-fold and 2.3-fold higher than wild-type, respectively). In these transgenic lines, IPP generated by the increase in flux through the MVA pathway was at least in part incorporated into cis-1,4-isoprene end products. We also investigated whether the greater abundance of precursors affected the chain length of known dolichols by liquid chromatography/mass spectrometry (LC/MS). However, no substantial differences in dolichol composition were observed between the wild-type and transgenic lines [see Fig. 4c].
The synthesis of polyisoprenes in the transgenic lines was not substantially affected by the overexpression of three key MVA pathway enzymes, which may reflect the bottleneck at the squalene epoxidase step. This step is therefore ideal for more detailed characterization in future experiments. Furthermore, the production of natural rubber in laticifers might not be dependent solely on the availability of IPP units, but also on the presence and activity of the enzymes that are directly linked to the rubber particles and thus build the rubber transferase complex [5]. These specialized end-product enzymes also represent