Knock-down of the MEP pathway isogene 1-deoxy-d-xylulose 5-phosphate synthase 2 inhibits formation of arbuscular mycorrhiza-induced apocarotenoids, and abolishes normal expression of mycorrhiza-specific plant marker genes


*(fax +49 345 5582 1509; e-mail


The first step of the plastidial methylerythritol phosphate (MEP) pathway is catalyzed by two isoforms of 1-deoxy-d-xylulose 5-phosphate synthase (DXS1 and DXS2). In Medicago truncatula, MtDXS1 and MtDXS2 genes exhibit completely different expression patterns. Most prominently, colonization by arbuscular mycorrhizal (AM) fungi induces the accumulation of certain apocarotenoids (cyclohexenone and mycorradicin derivatives) correlated with the expression of MtDXS2 but not of MtDXS1. To prove a distinct function of DXS2, a selective RNAi approach on MtDXS2 expression was performed in transgenic hairy roots of M. truncatula. Repression of MtDXS2 consistently led to reduced transcript levels in mycorrhizal roots, and to a concomitant reduction of AM-induced apocarotenoid accumulation. The transcript levels of MtDXS1 remained unaltered in RNAi plants, and no phenotypical changes in non-AM plants were observed. Late stages of the AM symbiosis were adversely affected, but only upon strong repression with residual MtDXS2-1 transcript levels remaining below approximately 10%. This condition resulted in a strong decrease in the transcript levels of MtPT4, an AM-specific plant phosphate transporter gene, and in a multitude of other AM-induced plant marker genes, as shown by transcriptome analysis. This was accompanied by an increased proportion of degenerating and dead arbuscules at the expense of mature ones. The data reveal a requirement for DXS2-dependent MEP pathway-based isoprenoid products to sustain mycorrhizal functionality at later stages of the symbiosis. They further validate the concept of a distinct role for DXS2 in secondary metabolism, and offer a novel tool to selectively manipulate the levels of secondary isoprenoids by targeting their precursor supply.


Isoprenoids constitute the largest and most diverse group of secondary plant metabolites, with over 40 000 compounds described (Withers and Keasling, 2007). Many isoprenoids are essential for growth and development (primary metabolism). However, most of the diversity of isoprenoid compounds has been shaped by interactions with the environment, to provide plants with secondary isoprenoids important for many ecological functions (Cheng et al., 2007). The continued supply of isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) isoprenoid precursors is secured by two independent pathways in separate cellular compartments. Whereas in the cytosol IPP is derived from mevalonic acid, the plastidial IPP is generated via the recently discovered methylerythritol phosphate (MEP) pathway (previously also known as the non-mevalonate pathway; Rohmer et al., 1993; Eisenreich et al., 2001; Rodríguez-Concepción and Boronat, 2002; Rodríguez-Concepción, 2006; Hunter, 2007). Initial investigations indicated the use of cytosolic IPP pools for the biosynthesis of sesquiterpene, triterpene, and polyterpene end products, whereas plastidial IPP pools serve for the formation of monoterpenes, diterpenes tetraterpenes (carotenoids) and the prenyl moieties of chlorophyll, plastoquinone and tocopherol (Chappell, 2002). Recently, it has been shown that the MEP pathway can also support sesquiterpene formation (Dudareva et al., 2005).

In the first step of the MEP pathway the enzyme 1-deoxy-d-xylulose 5-phosphate synthase (DXS) catalyzes the formation of 1-deoxy-d-xylulose 5-phosphate from pyruvate and glyceraldehyde 3-phosphate in a thiamin-dependent transketolase-like reaction (Figure 1). DXS genes have been characterized from several plant species (Bouvier et al., 1998; Lange et al., 1998; Lois et al., 2000). The importance of a functional DXS gene for plant survival is demonstrated by the albino phenotype of the cla1 mutant of Arabidopsis (Estevez et al., 2000; Mandel et al., 1996). The reaction catalyzed by DXS appears to be a bottleneck in plastidial IPP biosynthesis and precursor supply to downstream steps. A regulatory role for DXS was suggested in ripening tomato fruits on the basis of a strong increase in DXS transcript levels in the orange and red mature stages, associated with the massive accumulation of carotenoids (Lois et al., 2000). A rate-limiting function of DXS was shown in Arabidopsis and tomato by the transgene-mediated upregulation or downregulation of DXS transcript levels, which correlated with concomitant changes in the levels of various end products (Enfissi et al., 2005; Estevez et al., 2001). Constitutive expression of the Arabidopsis DXS (CLA1) in spike lavender resulted in a significant increase in essential oil formation (Munoz-Bertomeu et al., 2006).

Figure 1.

 Input from two separate 1-deoxy-d-xylulose 5-phosphate synthase (DXS) isozymes to isoprenoid precursor pools.
Isopentenyl diphosphate (IPP)/dimethylallyl diphosphate (DMAPP) precursor pools are fed from activities of two DXS isozymes committed to the biosynthesis of various end products with housekeeping (green) or ecological functions (red). Assignments are based on reported expression profiles and accumulation of products. References: (1) Lois et al., 2000; (2) Sharkey et al., 2005; (3) Matusova et al., 2005; (4) Lange et al., 1998; (5) Totte et al., 2003; (6) Moehs et al., 2001; (7) Walter et al., 2002. Unknown assignments are depicted in grey.
Enzyme designations: CCD, carotenoid cleavage dioxygenase; DXR, 1-deoxy-d-xylulose 5-phosphate reductosiomerase; HDR, hydroxymethylbutenyl 4-diphosphate reductase; PDS phytoene desaturase; NCED, 9-cis-epoxycarotenoid dioxygenase.

Plant genomes appear to contain single-copy nuclear genes for their MEP pathway functions (Rodríguez-Concepción, 2006). The only exception reported to date is the diversification of DXS genes described for Medicago truncatula and a few other plants. Two DXS isoforms that are highly conserved between species exhibit only about 70% sequence identity with each other in mature proteins, and the corresponding genes show differential expression (Walter et al., 2002). MtDXS1 transcripts are abundant in many developing plant tissues, but are lower in roots. On the contrary, MtDXS2 transcript levels are low in most tissues but are strongly stimulated in roots upon colonization by mycorrhizal fungi. This correlates with an accumulation of certain apocarotenoids (carotenoid cleavage products). Additional reports on a correlation of DXS2 expression with the biosynthesis of various other secondary isoprenoids led us to propose dedicated roles of DXS1 in primary metabolism (housekeeping functions), and of DXS2 in secondary isoprenoid biosynthesis (ecological functions), respectively (Walter et al., 2002; Figure 1).

To study the role of DXS2 in the formation of these apocarotenoids, we are using a symbiotic plant–microbe interaction system between arbuscular mycorrhizal (AM) fungi and plant roots. During this interaction plant roots frequently accumulate two types of apocarotenoids to high levels, particularly in the late stages of the symbiosis, which leads to a macroscopically visible yellow coloration of mycorrhizal roots (Fester et al., 2002; Jones, 1924; Klingner et al., 1995; Walter et al., 2000). The chromophor of this ‘yellow pigment’ complex is an acyclic C14 apocarotenoid polyene called mycorradicin (Figure 1), which occurs in a complex mixture of derivatives (Schliemann et al., 2006). Concomitantly, glycosylated C13 cyclohexenone apocarotenoid derivatives accumulate in mycorrhizal roots (Maier et al., 1995; Strack and Fester, 2006; Schliemann et al., 2008; Figure 1). Both types of apocarotenoids are assumed to originate from a common, but still unknown, carotenoid precursor (Akiyama, 2007; Walter et al., 2000). The location of these apocarotenoids in cells harbouring an arbuscule has been reported previously (Fester et al., 2002). Their importance and physiological functions in the AM symbiosis are still elusive (Walter et al., 2007). In contrast, other apocarotenoids of low abundance (strigolactones) have been attributed to triggering hyphal branching during the early stages of the AM symbiosis (Akiyama et al., 2005; Bouwmeester et al., 2007).

Arbuscules are highly ramified fungal organs outgrowing from fungal hyphae inside root cells. They are transient, constantly decaying and reemerging structures, with an average life time of 7–10 days (Alexander et al., 1988). Arbuscules constitute the site of nutrient exchange, particularly for phosphate uptake from the fungus to the plant (Bucher, 2007; Javot et al., 2007a). A specific phosphate transporter of M. truncatula (MtPT4), encoded by a single-copy gene, is involved in the acquisition of arbuscule-derived phosphate into the cortical host cell (Harrison et al., 2002). The AM-specifically expressed MtPT4 gene is indispensable for a functional mycorrhiza (Javot et al., 2007b), and thus constitutes a useful molecular marker to follow mycorrhizal and arbuscule functionality.

Although the reaction catalyzed by DXS is at an early stage of apocarotenoid biosynthesis, we reasoned that the cut-off of AM-specific isoprenoid precursor supply by targeting DXS2 might be a means to specifically downregulate apocarotenoid biosynthesis, and thereby to get closer to an elucidation of the function of cyclohexenone and/or mycorradicin derivates. In this paper we present results from an RNAi-mediated knock-down approach on MtDXS2 isogene expression in mycorrhizal hairy roots of M. truncatula. The resulting alterations in the plant–fungus interaction provide new insights into the importance of precursor supply for plant secondary isoprenoid metabolism, and the relevance of MEP pathway-derived products for the AM symbiosis.


Unrelated promoters and distinct exon–intron structures in MtDXS1 and MtDXS2 isogenes

To characterize the MtDXS2 isogene, genomic sequences were isolated from a phage library using the MtDXS2 cDNA probe (Walter et al., 2002). Unexpectedly, a tandem repeat of two highly similar MtDXS2 paralogues was discovered on the 17 554-bp region that was sequenced (accession AM746342). Expression analyses using real-time RT-PCR indicated that the upstream gene corresponding to the cDNA, now termed MtDXS2-1, was always expressed at a much higher level than the duplicate downstream gene (MtDXS2-2), but that the latter retained low-level AM inducibility (data not shown). Therefore, MtDXS2-1 transcript levels constitute a suitable marker for overall MtDXS2 gene activity. By comparison with the orthologous MtDXS1 isogene (see below) an intron loss became apparent, which defines 10 exons and nine introns, with intron 7 being absent in MtDXS2-1 (Figure S1a). About 2.7 kb of sequence upstream from the MtDXS2-1 coding sequence was contained on one of the genomic clones.

The sequence of the MtDXS1 isogene was obtained from the Medicago genome sequencing project. The coding sequence is located on clone mth2-68g24 (AC147364) between position 103 147 and 99 323 bp. The MtDXS1 isogene also contains 10 exons and nine introns, but intron 1 is absent when compared with MtDXS2-1 (Figure S1a). The highest differences between the two M. truncatula DXS isogenes were found in exons 1, 2 and 10 (Figure S1b). As expected from the differential expression of MtDXS1 and MtDXS2 (Walter et al., 2002), no sequence similarity in promoter regions can be recognized from an alignment of the 2500-bp upstream sequences from MtDXS1 and MtDXS2-1.

MtDXS2-1 promoter activity is located to arbuscule-containing cortex cells of mycorrhizal roots

To analyse promoter activity on a cell-specific level, a promoter–reporter gene fusion was generated by fusing a 2723-bp upstream sequence of MtDXS2-1 to the uidA (β-glucuronidase, GUS)-int reporter gene. This construct was expressed in transgenic M. truncatula hairy roots resulting in individual chimaeric plants. Plantlets with transgenic root systems were either inoculated with the AM fungus Glomus intraradices or further cultivated without inoculation. Macroscopical analysis of whole root systems harvested 55 days later showed strong GUS activity in mycorrhizal roots (Figure 2a,b). Cross sections of stained roots from non-mycorrhizal plants and from mycorrhizal plants outside of the G. intraradices-colonized regions showed weak GUS activity in cells of the vascular system (Figure 2c,d). In contrast, within the colonized regions the arbuscule-containing cells exhibited strong GUS activity (Figure 2e).

Figure 2.

 GUS reporter analysis driven by the MtDXS2-1 promoter in Medicago truncatula roots.
Histochemical localization of MtDXS2-1 promoter activity in whole roots or root sections of M. truncatula resulting from uidA-intron (GUSint) reporter gene expression.
(a) Whole roots, non-mycorrhizal.
(b) Whole roots, mycorrhizal.
(c) Root section, non-mycorrhizal.
(d) Root section, mycorrhizal, non-colonized region.
(e) Root section, mycorrhizal, colonized and arbusculated region. An arbuscule is marked by an arrow.
Scale bars: 50 μm.

Knock-down of MtDXS2 expression consistently results in reduced levels of apocarotenoids

To carry out an RNAi-mediated knock-down approach on MtDXS2, designed to not affect MtDXS1 expression, a 172-bp target sequence was chosen from the region coding for the transit peptide of MtDXS2-1. As the MtDXS2-1 and MtDXS2-2 paralogous isogenes are almost identical in these regions, both genes are expected to be downregulated. Figure 3 summarizes the results from the expression of an empty vector (EV) or the RNAi construct in transgenic hairy roots of individual mycorrhizal plants in two separate experiments, 7 or 9 weeks after inoculation with the mycorrhizal inoculum. Whereas the RNAi plants exhibited a wide range of decreased MtDXS2-1 transcript levels (Figure 3a), all of them showed significantly reduced levels of cyclohexenone and mycorradicin derivatives compared with EV control plants (Figure 3b,c). Experiment II resulted in a considerable number of transformants with stronger repression compared with experiment I (classified into a separate group of RNAi plants with less than a 10% residual level of MtDXS2-1 transcripts), but this did not consistently lead to a significant further reduction of apocarotenoid levels.

Figure 3.

 Alterations of molecular and metabolic markers in MtDXS2-repressed roots.
Individual hairy root transformants from two experiments were analyzed for (a) MtDXS2-1 transcript levels by real-time RT-PCR, (b) cyclohexenone and (c) mycorradicin derivatives by HPLC, (d) MtPT4, (e) GiBTub (Glomus intraradices ß-tubulin) and (f) MtDXS1 transcript levels by real-time RT-PCR. Mycorrhizal empty vector control roots (empty columns) are compared with mycorrhizal RNAi roots exhibiting moderate repression (mr, above 10% residual MtDXS2-1 transcript levels, grey columns) or strong repression (sr, below 10% residual MtDXS2-1 transcript levels, dark columns). Plants from experiment I were colonized for 9 weeks and consisted of six empty vector (EV) controls and 14 RNAi plants. Plants from experiment II were colonized for 7 weeks and consisted of eight EV controls, eight mr RNAi plants and eight sr RNAi plants. RT-PCR results are the mean of three technical replicates normalized for equal expression of elongation factor 1α. Cyclohexenone and mycorradicin derivative levels were determined from peak areas (PA) after photometric detection at 245 or 380 nm, respectively. Statistical treatment was performed by one-way anova followed by a Tukey’s HSD test (a, b, c); P ≤ 0.05. Means sharing the same letter are not significantly different.

MtDXS1 transcript levels and non-mycorrhizal phenotypes are unchanged upon MtDXS2 repression

Transcript levels of the MtDXS1 isogene did not show a decrease upon MtDXS2 knock-down, not even with strong repression (Figure 3f). All plants with transgenic roots harbouring the RNAi construct were inspected for any potential alterations in growth, flowering time, shoot and root biomass or other distinctive features. No differences were observed between the mycorrhizal EV control and the mycorrhizal RNAi plants of both experiments I and II. However, whereas the root content of inorganic phosphate from experiment-II RNAi samples was not altered in the moderate DXS2 repression (mr) sample group [12.1 ± 4.7 μg per g fresh weight (fw−1)] compared with EV controls (9.7 ± 3.2 μg fw−1), it was elevated in the strong repression (sr) samples (20.2 ± 5.5 μg fw−1).

To corroborate the results of experiments I and II, and to also include non-mycorrhizal samples, a second target region (RNAi-II) of 525 bp covering the 3′ end of the gene and parts of the 3′ untranslated region of MtDXS2-1 (Figure S1b) was used in experiment III. The use of both RNAi constructs led to strong repression of MtDXS2-1 transcript acccumulation, either from elevated levels in mycorrhizal roots or from low basal levels in non-mycorrhizal roots (Table 1). Again, no conspicuous changes in the phenotypes or growth parameters of plants with RNAi roots were detected under both mycorrhizal and non-mycorrhizal conditions. In all roots of experiment III, the transcript levels of isoprenoid biosynthetic genes catalyzing reactions subsequent to the first step of the MEP pathway, and leading to cyclohexenone and mycorradicin derivatives (see Figure 1), were examined. The analysis of 1-deoxy-d-xylulose 5-phosphate reductoisomerase (DXR), hydroxymethylbutenyl 4-diphosphate reductase (HDR), phytoene desaturase (PDS) and carotenoid cleavage dioxygenase (CCD1) revealed a slight downward trend in mycorrhizal RNAi roots compared with EV control roots (Table 1). Among the non-mycorrhizal samples there were no major differences between EV control and RNAi roots for all the steps analyzed, except for the RNAi target gene.

Table 1.   Transcript levels of isoprenoid biosynthetic genes catalysing steps subsequent to the first step of the methylerythritol phosphate (MEP) pathway
TCcDNARelative transcript level (EΔCt × 10−3)
EV controlRNAi IRNAi IIEV controlRNAi IRNAi II
  1. Data are from real-time RT-PCR determinations using samples from experiment III, and are presented as means ± SD of three individual transformants for each category; TC, tentative consensus from DFCI Medicago Gene Index version 8;

 1-deoxy-D-xylulose 5-phosphate synthase 2-14.9 ± 3.20.3 ± 0.30.5 ± 0.51.2 ± 0.90.1 ± 0.010.2 ± 0.2
 1-deoxy-d-xylulose 5-phosphate synthase 118.4 ± 5.610.0 ± 6.115.8 ± 9.815.8 ± 8.811.5 ± 4.113.1 ± 6.5
TC1015021-deoxy-D-xylulose 5-phosphate reductoisomerase18.6 ± 7.58.9 ± 2.46.7 ± 5.49.1 ± 4.59.3 ± 2.710.1 ± 0.9
TC1092181-hydroxy-2-methyl-butenyl 4-diphosphate reductase19.4 ± 7.98.6 ± 2.012.7 ± 12.212.0 ± 1.312.3 ± 7.09.9 ± 5.6
TC106289phytoene desaturase11.9 ± 5.56.2 ± 3.59.3 ± 4.111.1 ± 3.98.7 ± 2.26.1 ± 5.4
TC100912carotenoid cleavage dioxygenase 136.3 ± 11.122.4 ± 4.521.7 ± 16.840.8 ± 6.220.7 ± 7.69.7 ± 7.2

Only strong knock-down of MtDXS2 expression affects MtPT4 transcript levels

Potential alterations in vital mycorrhiza functions resulting from the knock-down of MtDXS2 expression were investigated by analyzing the AM-specific phosphate transporter gene MtPT4. In experiment I there were no roots below 10% residual MtDXS2-1 transcript levels, compared with the average of EV control plants. Despite significantly decreased levels of apocarotenoids, MtPT4 transcript levels were not reduced in these roots (Figure 3d). By contrast, in RNAi roots with stronger repression of MtDXS2-1 transcripts to residual levels below 10% (the group of sr RNAi roots in experiment II), the MtPT4 transcript levels were significantly reduced (Figure 3d). Similarly, when five strongly repressed mycorrhizal samples of experiment III (1–4% residual MtDXS2-1 transcripts) were compared with five EV control samples, a strong reduction in MtPT4 transcript levels was observed (Figure S2). In addition to the plant marker MtPT4, the transcript levels for G. intraradices β-tubulin, a marker for viable fungal structures, was reduced in strongly repressed plants, but not to the extent as was observed for MtPT4 (Figures 3e and S2).

Microarray analysis reveals extensive adverse effects on AM physiology in strongly repressed RNAi plants

To evaluate the extent of altered gene expression in RNAi roots, a transcriptome analysis was performed from mycorrhizal EV controls compared with the four individual strongly repressed mycorrhizal RNAi samples of experiment II. Particular attention was given to genes previously reported as specifically upregulated in M. truncatula/G. intraradices colonized roots (Hohnjec et al., 2005; Liu et al., 2007). Out of the 201 M. truncatula genes co-induced in roots to at least twofold by Glomus mosseae and G. intradices colonization (Hohnjec et al., 2005), 64 genes show significantly reduced transcript levels (P ≤ 0.05) by at least twofold in the RNAi samples (Table 2). Intriguingly, many of the genes most strongly upregulated under standard mycorrhizal conditions (Hohnjec et al., 2005) are most strongly downregulated in mycorrhizal RNAi roots, relative to mycorrhizal EV controls (Table 2). Amongst them are several well-known marker genes for arbusculated cells, such as MtSCP1 (Liu et al., 2003), MtGST1 (Wulf et al., 2003), MtGLP1 (Doll et al., 2003), MtLEC7 (Frenzel et al., 2005) and MtBCP1 (Hohnjec et al., 2005). Moreover, in the complete list of 16 470 tentative consensus sequences (TCs) represented on the array, the majority of the most strongly downregulated ones in RNAi roots (nine out of 13 TCs with array ratios between −5.92 and −3.96) belong to the AM-induced gene category (Table S1, marked in bold). This indicates a major reversion of AM-mediated gene regulation upon strong knock-down of MtDXS2 expression, and a preferential impairment of mycorrhizal but not housekeeping-related gene expression.

Table 2.   Abolishment of normal expression of arbuscular mycorrhizal (AM) marker genes in MtDXS2-repressed mycorrhizal roots deduced from microarray analyses
DFCI IDOligo IDPutative annotationRNAi versus EVHohnjec et al. (2005); Array ratioa
Array ratioP-value
  1. Array ratios of four individual strongly repressed RNAi roots compared with four samples from empty vector (EV) control roots (pooled) from experiment II are aligned with the results of Hohnjec et al. (2005) defining AM-upregulated genes. DFCI-ID, see Table 1. Oligo ID, identifier of Medicago truncatula 70-mer oligonucleotides.

  2. aLog2 expression ratios in Medicago truncatula roots colonized by Glomus intraradices.

TC107197MT014644Organ-specific protein S2−5.922.39E–066.21
TC94753MT008095Hypothetical protein−5.422.76E–065.64
TC107197MT002169Organ-specific protein S2−5.079.17E–051.98
TC105109MT012688Hypothetical protein−5.061.99E–067.21
TC100812MT002208Hypothetical protein−4.901.60E–056.42
TC101060MT014645Gamma thionin / Defensin−4.878.95E–055.34
TC101124MT011611Clavaminate synthase-like protein−4.361.14E–032.29
TC102881MT009557Trypsin-alpha amylase inhibitor−4.113.08E–064.16
TC106955MT009186Serine carboxypeptidase II−3.969.92E–082.80
TC112474MT015000Kunitz inhibitor ST1-like−3.682.46E–064.34
AW329171MT007595Hypothetical protein−3.473.13E–062.68
TC106921MT002501Nitrite transporter−3.372.13E–071.03
TC106691MT007159Hypothetical protein−3.351.03E–031.14
TC108376MT002165Serine carboxypeptidase II−3.201.20E–042.65
TC104147MT004619RNA-binding domain-containing protein−2.683.50E–052.49
TC106921MT009589Nitrite transporter−2.575.06E–081.14
TC106582MT007032Trypsin-alpha amylase inhibitor−2.519.17E–031.64
TC96854MT010383Desacetoxyvindoline 4-hydroxylase−2.432.11E–031.28
TC106656MT000153Glycosyl hydrolase family 1−2.402.74E–032.51
TC107321MT000669Hypothetical protein−2.283.48E–071.08
TC101520MT008518Leucoanthocyanidin dioxygenase−2.213.51E–061.12
TC106906MT000634Auxin-regulated protein GH3−2.113.06E–033.13
TC108664MT014872Hypothetical protein−2.092.70E–043.68
AJ498718MT000228Taxane 13-alpha-hydroxylase−1.991.65E–032.61
TC103088MT003219Taxane 10-beta-hydroxylase−1.976.14E–031.05
TC94347MT007030Cationic (heme) peroxidase−1.968.77E–041.65
TC110699MT01356753-kDa symbiosome membrane nodulin−1.942.01E–042.83
TC96704MT015993Late nodulin protein−1.773.85E–021.44
TC101075MT000961Branched-chain amino acid aminotransferase−1.691.35E–032.56
TC100805MT015420Kunitz inhibitor ST1-like−1.691.11E–024.39
TC106728MT001054Cycline-like F-box containing protein−1.651.91E–051.19
BF003720MT006929Histone H1/H5−1.645.30E–064.53
TC108853MT003012Hypothetical protein−1.635.86E–031.59
TC110192MT003230Pentatricopeptide repeat-containing protein−1.633.18E–041.11
TC105254MT005171Hypothetical protein−1.615.16E–041.12
TC95933MT004574Wound-induced protein WI12−1.504.33E–062.74
TC95491MT008228Hypothetical protein−1.501.35E–024.02
TC101133MT000873Pyruvate dehydrogenase E1 alpha subunit−1.494.97E–041.59
TC103981MT011784Homeobox-leucine zipper protein HAT7−1.473.45E–051.27
TC104204MT004625Ent-kaurene synthase A−1.469.03E–032.05
TC108966MT002011Hydroxymyristoyl ACP dehydrase−1.433.87E–051.86
TC96249MT015692Cytochrome P450−1.435.43E–041.73
TC102130MT015611Hypothetical protein−1.427.24E–042.04
TC96284MT008921Hypothetical protein−1.402.15E–035.76
TC98373MT006073Hypothetical protein−1.361.88E–064.19
TC96557MT002953Prolyl oligopeptidase−1.341.09E–031.15
TC105073MT013028Proteinase inhibitor−1.333.67E–062.97
TC109691MT002974Pectate lyase−1.317.21E–031.58
TC100835MT000450Blue copper protein−1.304.47E–021.51
TC94050MT014213Histone H2B−1.291.32E–041.77
TC95384MT002711Cytochrome P450−1.159.84E–051.44
TC97926MT006551Putative phytosulfokine LRR-type receptor kinase−1.155.13E–044.02
TC97582MT010125Hypothetical protein−1.099.84E–052.38
TC103144MT003693Gibberellin 20 oxidase 1−1.091.21E–021.29
TC100493MT000216Trypsin-alpha amylase inhibitor−1.061.78E–021.54

To validate the microarray results we carried out a real-time RT-PCR analysis on a subset of genes using samples from experiment III. All genes selected showed a strong reduction in transcript levels in strongly repressed RNAi roots relative to EV controls similar to the microarray results (Table 3). For example, TC101060, annotated as defensin, exhibits the strongest alteration in both the array and real-time RT-PCR data. Differences in the extent of downregulation for some of the other genes can be ascribed to different experimental conditions, the higher dynamic range of real-time RT-PCR analyses and the high variability of mycorrhizal colonization experiments. The strong reduction of AM marker gene transcript levels was not observed for the mr RNAi samples of experiment II (Table 3).

Table 3.   Validation of abolishment of normal arbuscular mycorrhizal (AM) marker gene expression in mycorrhizal RNAi roots strongly downregulated for MtDXS2 expression
TIGR IDOligo IDAnnotationRNAi versus EV control
Log2 expression ratio real time RT-PCRArray ratio
strong repression
Moderate repressionaStrong repressionb
  1. Selected genes shown in Table 2 were analyzed by real-time RT-PCR comparing samples with moderate repression (experiment II) or strong repression (experiment III) of MtDXS2 with mycorrhizal empty vector (EV) controls. On the right, the array data of Table 2 (strongly repressed samples of experiment II) are presented for direct comparison.

  2. aLog2 expression ratios for moderate repression of MtDXS2 were obtained by comparing the mean value of five biological replicates of moderate repression samples with the mean value of four biological replicates of EV controls from experiment II (see Figure 3).

  3. bLog2 expression ratios for strong repression of MtDXS2 were obtained by comparing the mean value of four biological replicates with strong repression with the mean value of four EV controls from experiment III (see Figure S2). Only genes with significant changes (P ≤ 0.05) between RNAi and EV control are presented. Statistical analysis was performed with anova.

  4. cArray ratios are taken from Table S1.

TC100835MT000450Blue copper-binding protein−2.60−1.30

Root colonization is unaltered, but more degenerating and dead arbuscules appear in RNAi plants

To get a comprehensive overview on total arbuscule abundance and other general parameters of fungal root colonization, ink-staining tests were carried out according to the method described by Trouvelot et al. (1986), with the individual mycorrhizal root samples of experiment II. There was no change in the overall frequency of colonization (F%), even in the most strongly repressed RNAi plants (Figure 4a). The density of mycorrhizal structures and the abundance of ink-stainable arbuscules in the mycorrhizal parts of the root (m% and a%, respectively) was also largely unaltered (Figure 4b,c). Assessments from experiment III confirmed the results presented in Figure 4a–c, and indicated again that there were no major changes in the ink-stainable fungal structures (Figure S2).

Figure 4.

 Mycorrhization parameters and distribution of arbuscule developmental stages in empty vector (EV) controls and strongly repressed (sr) RNAi plants from experiment II.
(a) Frequency of colonization, F%; (b) density of colonization in mycorrhizal root fragments, m%, and (c) arbuscule abundance in mycorrhizal root part, a%, were determined from ink-stained roots in a blind test. Column specifications and sample numbers are as in Figure 3 except that one EV control sample did not have sufficient root material, and was therefore omitted.
(d) Representative confocal images of the arbuscule developmental stages used for classification after acid-fuchsin staining. The degenerating phase can be recognized by septa forming on arbuscule branches (arrow).
(e) Proportion of four different developmental stages among the total population of arbuscules classified after staining with acid-fuchsin. Eight EV control samples (empty columns) are compared with eight sr RNAi plants. About 100 arbuscules from each sample were evaluated. *P ≤ 0.05; **P ≤0 .01, ***P ≤ 0.001.

To uncover potential alterations in arbuscule morphology, or in the proportion of the various stages of arbuscule development, in a given population, confocal images of arbuscules were evaluated following acid fuchsin staining. Four stages of arbuscule development and decay were defined (Figure 4d): (i) developing arbuscules, characterized by their small size and the restricted dichotomous branching of hyphae; (ii) mature arbuscules, exhibiting extensive branching and almost completely filling the host cortical cell – this stage already exhibits early indications of collapsing branches; (iii) degenerating arbuscules, which show more degraded branches – collapsed branches are often separated by septa from apparently intact parts of the arbuscule; (iv) fully degenerated and dead arbuscules, characterized by thick, collapsed branches and a clumpy appearance of arbuscule remains. The classification was performed with individual mycorrhizal root samples from experiment II comparing eight EV control samples with eight RNAi samples exhibiting a strong knock-down status. As deduced from a total of about 1600 evaluated arbuscules, a shift towards degenerating and dead arbuscules at the expense of mature ones occurred in RNAi roots with strongly reduced MtDXS2-1 transcript levels compared with EV controls (Figure 4e).


Elucidating the role of secondary metabolites in the interaction of plants with their environment is often a difficult task. In the work described here, we have taken a novel loss-of-function approach by trying to specifically cut off precursor supply to abundant secondary isoprenoid end products, without disturbing basic plant functions. This objective has been achieved, and is, in roots with strong MtDXS2 repression, accompanied by a significant impairment of AM-specific gene expression. Nonetheless, the role of apocarotenoids in the AM symbiosis is still elusive. Before detailing these aspects below, the many implications of the work for MEP pathway and isoprenoid biosynthesis research will be highlighted.

The previous proposal of a unique role for DXS2 genes in plant secondary isoprenoid metabolism has been based on a number of correlations of elevated DXS2 transcript levels and the appearance of secondary isoprenoids (Walter et al., 2002). In the case of the mycorrhizal cyclohexenone and mycorradicin apocarotenoid derivatives, these compounds accumulate locally in the vicinity of arbuscules (Fester et al., 2002). Our analysis of the MtDXS2-1 promoter in transgenic mycorrhizal roots of M. truncatula has now shown its activity in arbusculated cells and in cell-autonomous regulation (Figure 2). This finding extends previous conclusions based on northern blots (Walter et al., 2002), and on two whole transcriptome analyses from M. truncatula (Hohnjec et al., 2005) and Oryza sativa (Güimil et al., 2005). A correlation of DXS2 expression with arbuscule occurrence is in agreement with immunolocalization studies showing strong accumulation of the enzyme coding for the second step of the MEP pathway (DXR) in arbusculated cells (Hans et al., 2004; Lohse et al., 2006). However, MtDXS2-1 promoter activity is not restricted to the site of encounter with the AM fungus (Figure 2). There appears to be weak activity in the vascular system of roots, which is consistent with low basal levels of MtDXS2-1 transcripts in non-mycorrhizal roots (Table 1).

Next to mycorrhization, herbivory has recently been characterized as a stimulus for the inducible expression of MtDXS2 in M. truncatula (Arimura et al., 2007). In other plants many additional cues for DXS2 expression related to non-housekeeping biotic interactions have been reported, further supporting the concept of a dedicated role of DXS2. Four recent examples are volatile sesquiterpene emission in snapdragon flowers (Dudareva et al., 2005), ginkgolide diterpene accumulation in Ginkgo biloba roots (Kim et al., 2006), diterpene phytoalexin accumulation following fungal elicitor treatment in rice cell cultures (Okada et al., 2007) and terpenoid-based defensive oleoresin formation in spruce (Phillips et al., 2007). These data also highlight a hitherto unrecognized requirement for precursor supply upregulation in many instances of plant secondary isoprenoid biosynthesis.

A definitive proof for a distinct function of DXS2 in the biosynthesis of secondary metabolites with an ecological function can only come from analyses of mutants or loss-of-function transgenics. To date, no DXS2 mutant is known, but there is one species (Arabidopsis thaliana) known to be devoid of a DXS2 gene copy, but which harbours a second copy of a DXS1-type gene instead (Rodríguez-Concepción, 2006). Whether this exceptional DXS isogene status contributes to the disability of this species to form root symbioses is currently unknown. To elucidate DXS2 functions in AM, we have used a knock-down approach in a transgenic hairy root system of M. truncatula that is readily amenable to colonization by mycorrhizal fungi accompanied by DXS2 upregulation. RNAi-mediated knock-down of MtDXS2 expression reduces strongly and reproducibly the accumulation of two AM-inducible apocarotenoids in this system (Figures 3b,c and S2). Unexpectedly, this reduction did not strictly coincide with an impairment of AM-specific plant gene expression, but stronger MtDXS2 repression was required for such an effect. Only residual MtDXS2-1 transcript levels below about 10% consistently resulted in an impairment of AM-related plant gene expression (Figures 3d and S2, Tables 2, 3 and S1). This indicates that MtDXS2 is indispensable for the AM symbiosis at late stages of the interaction. The results thus constitute a validation of the proposed concept of DXS2 commitment to the biosynthesis of isoprenoids of ecological importance.

The specificity of MtDXS2 knock-down was confirmed by the determination of MtDXS1 transcript levels in all RNAi experiments described. There was no reduction of MtDXS1 transcript levels, and there was also no indication of any feedback mechanism upregulating MtDXS1 to compensate for the loss of MtDXS2 transcripts. Although there appear to be reasonable basal levels of MtDXS1 transcripts in RNAi plants, they obviously cannot substitute for the lack of MtDXS2 gene products. In non-mycorrhizal plants there was no obvious phenotypical difference between RNAi samples and EV controls, suggesting that MtDXS2 repression does not affect housekeeping functions, and only cuts off the AM-induced part of isoprenoid biosynthesis. This is supported by our microarray data, which indicate that the majority of genes strongly affected as a consequence of strong MtDXS2 repression belong to the AM-induced category (Table S1).

It is particularly noteworthy that the extensive revocation of AM-induced plant gene expression in strongly MtDXS2-repressed roots does not tie in with alterations in ink-stainable symbiotic structures (Figure 4a–c). The frequency and density of fungal structures appears to be largely unaffected, but severe alterations in the expression of AM marker genes suggest a loss of activity of these structures. Particularly for the arbuscules there is a discrepancy between total stainable arbuscules (Figures 4 and S2) and the strongly reduced transcript levels of arbuscule-related marker genes (Tables 2 and 3). Thus, the ratio of non-functional versus functional arbuscules appears to increase, indicating that arbuscules may become defective. It is tempting to speculate that non-functional arbuscules also cause the increase in root phosphate content of sr RNAi roots. This increase might result from a hold-up of phosphate flux from fungal hyphae to the host root cell, and further to the shoot, at a defective arbuscular fungal–plant interface. The idea of arbuscules becoming defective is further supported by the finding that there is a significant shift towards a greater number of older, degrading and dead arbuscules at the expense of mature ones, caused by strong DXS2 repression (Figure 4e). A total population of arbuscules with a greater number of older individuals brought about either by premature senescence or by extending the life cycle could contribute to a deterioration of AM performance.

Although the molecular expression markers suggest a loss of AM functionality upon strong repression of DXS2, there was no change in plant growth parameters. This may result from the fact that the loss of AM functionality probably only occurs at later stages of the interaction, whereas a major period of plant growth may still benefit from a functional AM. This contrasts with a permanent loss-of-function of MtPT4, which was reported to result in poor development of fungal structures and low colonization rates from the very beginning of the interaction (Javot et al., 2007b). DXS2 transcript levels regularly begin to rise only after about 3 weeks of fungal colonization (Walter et al., 2002), and only after this time point might a creeping loss of AM functionality ensue upon DXS2 repression. In accordance with this view are results from an additional RNAi experiment with a colonization period of 4 weeks. Basal levels of MtDXS2-1 transcripts and apocarotenoids in EV controls were lower than in the 7 and 9 weeks experiments. The RNAi roots showed only by a moderate reduction in both parameters. There were no significant alterations in the transcript levels of either G. intraradices β-tubulin or MtPT4 in the 4-weeks experiment (data not shown).

At this time, the causes of the strong decrease in transcript levels of MtPT4 and other mycorrhizal markers related to the strong repression of MtDXS2 are still unresolved. Under conditions of moderate repression we observed a significant decrease in apocarotenoid levels without a concomitant reduction in MtPT4 transcript levels (Figure 3). This result raises doubts as to the role of apocarotenoids in sustaining regular MtPT4 transcript levels. It must now also be considered that there might be a hitherto overlooked isoprenoid compound affected by the strong repression of MtDXS2, which has an impact on AM markers. However, a specific consequence of the absence of apocarotenoids might be the accumulation of older arbuscules. This phenomenon has now also been observed in a new knock-down approach on a carotenoid cleavage dioxygenase 1 gene target located further downstream in the pathway to apocarotenoids. It was correlated with reduced apocarotenoids, but with unaltered AM-specific markers (D.S. Floß and M.H. Walter, unpublished data). Whether a shift towards a greater number of older arbuscules beyond a critical level can eventually contribute to the drop in AM marker transcript levels remains to be investigated.

In summary, the selective manipulation of secondary isoprenoid precursor supply via targeting DXS2 may not always give a clear answer as to the function of certain end products, but it may still be an excellent tool to gain initial clues on the importance of secondary isoprenoids and on their involvement in the outcome of an interaction. In particular, this strategy works even in the absence of information on downstream enzymatic steps and potential targets. Secondary isoprenoids are often synthesized in large concentrations in particular developmental situations, or as a direct response to environmental stimuli. In many such instances isoprenoid precursor supply may be the limiting factor. Although the present work still falls short of solving the role of cyclohexenone and mycorradicin derivatives in the AM symbiosis, it shows the importance of the MEP pathway-dependent precursor supply into secondary isoprenoid biosynthesis and for the AM symbiosis. It also opens up new tools to study the regulation and function of plant secondary metabolism.

Experimental procedures

Plant material and fungal inoculation

Medicago truncatula (L.) Gaertn. var. Jemalong was grown in pots filled with expanded clay of 2- to 5-mm particle size (Original Lamstedt Ton; Fibo ExClay, in a phytochamber with a 17-h day/7-h night cycle (25°C/22°C, experiment I), with a 16-h day/8-h night cycle (22°C/18°C, experiment II), both with a 375 μm light intensity, or in a greenhouse with a 16-h supplemental light regime (experiment III and 4-weeks experiment). Plants were watered with deionized water three times a week and fertilizer (10 ml of half-strength Long Ashton medium containing 20% phosphate) was applied once a week. Six-week-old plants with transgenic hairy roots were inoculated with the AM fungus G. intraradices Schenk & Smith isolate 49 (Maier et al., 1995) by transfer to clay substrate containing 30% (v/v) inoculum. The inoculum was enriched in propagules by cultivation with mycorrhizal leek (Allium porrum cv. Elefant). The inorganic phosphate content of roots was determined as described in Taussky and Shorr (1953) after sample preparation according to the method described by Schaarschmidt et al. (2007).

Genomic library screening, plasmid constructions and transformation of Agrobacterium rhizogenes

An M. truncatula genomic library in the phage vector Lambda Fix II (Stratagene,; kindly provided by M. Harrison, Boyce Thompson Institute, was screened by standard methods using the MtDXS2 cDNA probe (Walter et al., 2002). For the localization of MtDXS2-1 promoter activity, a 35S-GUS-int construct was generated from pGUS-INT (Küster et al., 1995) using primers carrying NcoI–SpeI restriction sites (Table S2), and was inserted into pRNAi (Limpens et al., 2004). In order to create pPMtDXS2-1:GUS-int 2723 bp of the MtDXS2-1 promoter was amplified with primers including the XhoI–NcoI restriction sites (Table S2), and was inserted into p35SGUS-int by replacement of the double CaMV-35S promoter. The resulting construct was inserted with HindIII into the binary vector pRedRoot (Limpens et al., 2004).

For the MtDXS2-repression constructs, two inverted repeat constructs were created in pRNAi (Limpens et al., 2004). The target regions are located at the 5′ end (RNAi-I, 172 bp) and at the 3′ end (RNAi-II, 525 bp) of the MtDX2-1 cDNA (AJ430048; Walter et al., 2002; Figure S1). Both regions were PCR amplified using primers (Table S2), which included SpeI–AscI and BamHI–SwaI restriction sites, for insertion into pRNAi according to the method described by Limpens et al. (2004). The resulting hairpin constructs were cloned with KpnI–PacI restriction sites into the binary vector pRedRoot. All binary vectors were introduced into A. rhizogenes ArquaI (Quandt et al., 1993) by electrotransformation.

Hairy root transformation and selection for composite plants

A protocol for generating transgenic hairy roots in M. truncatula was modified from Vieweg et al. (2004) as follows. After scarification seeds were germinated on 0.8% water agar plates, and were then cultivated in the dark at 4°C for 4 days. Subsequently, the temperature was increased to 20°C for 2 days. Finally the samples were illuminated with white light for one additional day before treatment. Small volumes (approximately 50–100 μl) of suspension of A. rhizogenes harbouring the appropriate constructs were injected between three and five times into the hypocotyl of seedlings. After injections, the seedlings were planted into pots with humid expanded clay, carefully avoiding covering the injection sites. The plantlets were placed under a shadowed plastic cover for 2 days, and were cultivated in a growth chamber (20°C, 16-h light/8-h dark) for 2 weeks. When the first hairy roots became visible they were covered by a vermiculite : sand (1 : 1) mixture.

Four weeks after A. rhizogenes-injection, the first screen for the DsRED1 was performed. The plants were removed from the clay, and transgenic roots were selected by the fluorescence of DsRED1 with a fluorescence stereomicroscope (Leica MZ FLIII with DsRED1-Filter; Leica, All non-fluorescent roots were removed, and the plants bearing transgenic roots were cultivated in the growth chamber (23°C, 16-h light/8-h dark) for an additional 2 weeks. After a second screen for DsRED1, plants carrying exclusively transgenic roots were inoculated with G. intraradices and cultivated under the conditions described above. The typical frequency of transgenic hairy root formation at the injection site was about 30%.

Histochemical detection of GUS and visualization of fungal structures in roots

The histochemical assay for GUS activity was performed as described by Blume and Grierson (1997). The examination of stained 8-μm-thick cross sections embedded in polyethylene glycol (PEG) 1500 was carried out by light microscopy using a Zeiss ‘Axioplan’ microscope (Zeiss, equipped with a video camera (Fujix Digital Camera HC-300Z; Fujifillm,

An estimation of the colonization rate was performed by staining roots with 5% ink (Sheaffer Skrip jet black; Sheaffer, in 2% acetic acid according to the method described by Vierheilig et al. (1998). Fungal structures were assessed using a stereomicroscope, and the results were evaluated with the computer program ‘Mycocalc’ according to the method described by Trouvelot et al. (1986).

For the investigation of arbuscule developmental stages, root fractions were stained at room temperature (20°C) overnight with 0.01% acid fuchsin in lactoglycerol (lactic acid : glycerol : water, 1 : 1 : 1; Dickson et al., 2003). Subsequently, the samples were stored at 4°C until evaluation. For observations, stained roots were rinsed in water and transferred to slides with 100% glycerol. The fungal structures within the roots were visualized by confocal laser scanning microscopy (LSM 510 Meta; Zeiss) using the 543-nm laser line for excitation.

Extraction and detection of mycorradicin and cyclohexenone derivatives

Homogenized root material (200 mg) was extracted three times with 400 μl of 80% aq. MeOH after the addition of 50 μl ribitol (2 mg ml−1 H2O, internal standard) to yield the polar fraction. For the detection of mycorradicin derivatives, 300 μl of the supernatant was adjusted with KOH to a final 0.5 m concentration. Samples were further treated according to the method described by Fester et al. (2002), and analytical HPLC was performed according to the method described by Schliemann et al. (2008) using solvent system 1.

Real-time RT-PCR

Total RNA was prepared from roots with the Qiagen RNeasy® Plant Mini Kit (Qiagen, followed by a DNase digestion (RNase-free DNase Set; Qiagen). For first-strand synthesis, 1 μg of total RNA was converted into cDNA with M-MLV Reverse Transcriptase, Rnase H Minus, Point Mutant (Promega, according to the manufacturer’s protocol using the oligodT19 primer. The cDNA was diluted to fixed quantities (5 ng per reaction of reverse transcribed total RNA).

For the semiquantitative detection of transcripts, a SYBR Green based PCR assay (Applied Biosystems, was performed. The assay mix contained 5 ng of reverse transcribed total RNA and 100 nm primers (Table S2). The reactions were performed in an Mx 3005P QPCR system (Stratagene, with the following protocol: 95°C for 10 min followed by 50 cycles of 95°C for 30 sec, 60°C for 1 min and 72°C for 30 sec, and a subsequent standard dissociation protocol. As a control for genomic DNA contamination, 5 ng of total non-transcribed RNA was used under the same conditions as described above. Only samples with no amplification or with a weak signal accounting for <1% of the template measured in the corresponding cDNA assays were analyzed.

In order to calculate relative transcription levels, the efficiency of each PCR reaction (E) was determined by the import of the measured normalized baseline corrected fluorescence data (MxPro software; Stratagene) into the LinRegPCR computer program (Ramakers et al., 2003). The delta of threshold cycle (ΔCt) values were calculated by subtracting the Ct values of the target from the arithmetic mean Ct value of the normalizing Translation Elongation Factor 1α of M. truncatula, which was obtained from the three technical replicates. The relative transcription level (EΔCt) of each biological replicate is represented by the arithmetic mean of the three technical replicates. Where described, the expression ratios were transformed to log2 values.

Microarray analysis

Microarray analysis was performed using four independent strongly repressed RNAi plants exhibiting residual MtDXS2-1 transcript levels below 5% and four EV controls (pooled control) taken from experiment II. The total RNA was extracted using TRIzol reagent (Invitrogen, and was purified by ultrafiltration (Microcon YM-30; Millipore, RNA integrity was analysed by gel electrophoresis. Twenty micrograms of RNA was used to synthesize Cy3- and Cy5-labelled cDNA, as described by Hohnjec et al. (2005). The cDNAs were hybridized to four Mt16kOLI1Plus microarrays. These microarrays were composed of 16 470 70-mer oligonucleotide probes representing TCs of the DFCI M. truncatula Gene Index version 8 (Tellström et al., 2007). Hybridization conditions, image acquisition and data processing were performed according to the method described by Hohnjec et al. (2005). Lowess normalization and Student’s t-test statistical analyses were performed as described by Tellström et al. (2007).

Statistical analyses

Each DsRED1-expressing transgenic root system was subdivided and used for transcript analyses, metabolite analyses and estimation of the colonization rate or arbuscule morphology. All data presented in the graphs are means ± SD of biological replicates. Statistical analysis was performed by using anova followed by Tukey’s honestly significant differences (HSD) test. To fulfill the assumption of normal distribution, data were transformed by using the LOG transformation. For the analysis of arbuscule stages, the percentages of arbuscules were transformed by using ARCSIN transformation before performing anova.


We thank M. J. Harrison (Boyce Thompson Institute) for providing the M. truncatula genomic library and R. Geurts (Wageningen University) for providing pRedRoot and pRNAi. B. Saal (Universität Halle) provided help with statistical analyses. The skilful technical assistance of K. Manke is gratefully acknowledged. This work was supported in part by the Deutsche Forschungsgemeinschaft in the priority programme SPP1084 (MolMyk).

Accession number: AM746342.