Most terrestrial plants form arbuscular mycorrhiza (AM), mutualistic associations with soil fungi of the order Glomeromycota. The obligate biotrophic fungi trade mineral nutrients, mainly phosphate (Pi), for carbohydrates from the plants. Under conditions of high exogenous phosphate supply, when the plant can meet its own P requirements without the fungus, AM are suppressed, an effect which could be interpreted as an active strategy of the plant to limit carbohydrate consumption of the fungus by inhibiting its proliferation in the roots. However, the mechanisms involved in fungal inhibition are poorly understood. Here, we employ a transcriptomic approach to get insight into potential shifts in metabolic activity and symbiotic signalling, and in the defence status of plants exposed to high Pi levels. We show that in mycorrhizal roots of petunia, a similar set of symbiosis-related genes is expressed as in mycorrhizal roots of Medicago, Lotus and rice. Pi acts systemically to repress symbiotic gene expression and AM colonization in the root. In established mycorrhizal roots, Pi repressed symbiotic gene expression rapidly, whereas the inhibition of colonization followed with a lag of more than a week. Taken together, these results suggest that Pi acts by repressing essential symbiotic genes, in particular genes encoding enzymes of carotenoid and strigolactone biosynthesis, and symbiosis-associated phosphate transporters. The role of these effects in the suppression of symbiosis under high Pi conditions is discussed.
Arbuscular mycorrhiza (AM) are mutualistic symbiotic associations between most vascular land plant species, and fungi of the phylum Glomeromycota (Smith and Read, 2008), in which the plant trades carbohydrates for mineral nutrients from the fungus, in particular phosphate (Pi). The factors involved in early communication between the symbiotic partners have been elucidated in considerable detail in recent years (reviewed in Parniske, 2008). The first known chemical signal in AM is the root-borne branching factor strigolactone, which promotes hyphal branching and metabolism (Akiyama et al., 2005; Besserer et al., 2006). Conversely, an unknown diffusible fungal signal triggers the induction of symbiosis-associated genes in the root before the first physical contact is established (Kosuta et al., 2003). Once a fungal hyphopodium has been formed on the surface of the root epidermis, its position is detected by the subtending epidermal cell, which reacts with the establishment of an intracellular infection structure, the pre-penetration apparatus (PPA) that is necessary for fungal invasion (Genre et al., 2005). These cellular adaptations of the host depend on a symbiotic signalling pathway referred to as the common SYM pathway, because it is shared with root nodule symbiosis (RNS). The SYM pathway is functionally conserved between monocot and dicot species (Chen et al., 2007a, 2008; Gutjahr et al., 2008), suggesting that it evolved in early land plants, and became secondarily recruited into RNS. After penetration of the epidermis and a subsequent intercellular expansion phase, the fungus resumes intracellular colonization of cortical cells, again with the help of a PPA-related cellular accommodation structure of the plant (Genre et al., 2008). This results in the establishment of arbuscules and the associated symbiotic interface, over which nutrient exchange proceeds. Hence, the establishment of functional AM involves a series of steps which are under tight control mainly by the host plant.
The cellular adaptations during elaboration of AM are associated with pronounced physiological changes (Smith and Read, 2008), in particular the establishment of the symbiotic Pi-uptake system of the plant (Bucher, 2007). This involves the induction of symbiosis-specific Pi transporters (PTs), the expression of which is triggered by lyso-phosphatidylcholine (LPC) (Drissner et al., 2007). Symbiosis-specific PTs are localized to the periarbuscular membrane where they absorb the Pi delivered over the symbiotic interface by the fungus (Harrison et al., 2002). Induction of genes encoding H+-ATPase in cells that harbour arbuscules (Gianinazzi-Pearson et al., 2000; Krajinski et al., 2002), and the acidification of the interface (Guttenberger, 2000) are indicative of an energized nutrient uptake mechanism.
Besides positive regulatory mechanisms, the plant has means to limit colonization by AM fungi. Root systems that have already been colonized by AM fungi exhibit a reduced tendency to be infected and colonized by further symbiotic propagules (Catford et al., 2003). Interestingly, as in the case of the common SYM pathway, this effect shares common aspects with autoregulation of nodulation in legumes (Catford et al., 2003), which involves a receptor kinase in the shoot (Meixner et al., 2005; Magori and Kawaguchi, 2009). Another example of negative regulation by the plant is the suppression of AM symbiosis under high Pi levels (Menge et al., 1978; Jasper et al., 1979; Thomson et al., 1986; Amijée et al., 1989; Franken and Gnädinger, 1994; Nagy et al., 2009). As Pi is the major nutrient delivered by the fungus, the suppressive effect of Pi could be interpreted as a negative regulatory feedback mechanism of the plant host to limit carbohydrate allocation to the symbiont under conditions of optimal Pi supply. This implies that the plant has means to control and limit fungal proliferation according to its nutrient status. The mechanisms involved, however, are unknown.
Considering the successive steps in the AM interaction, a number of hypothetical scenarios could potentially account for negative regulation of AM by Pi: (i) Pi could limit the biosynthesis or secretion of strigolactone or other signals involved in pre-symbiotic interaction; (ii) Pi could reduce the expression of components of the SYM pathway or stimulate negative autoregulation; (iii) Pi could increase the defence status in the roots, hence leading to the rejection of the symbiont; (iv) the plant could limit the delivery of essential nutrients (e.g. carbohydrates) to the symbiont, thereby slowing down its growth (Olsson et al., 2006); or (v) Pi could downregulate essential components involved in downstream steps of root colonization and/or establishment of the symbiotic interface (Nagy et al., 2009). Each of these scenarios, which are not mutually exclusive, would be likely to be associated with a characteristic shift in gene expression patterns of particular marker genes.
Here, we explore the transcriptional changes associated with AM development and with elevated Pi supply in Petunia hybrida. Microarray analysis reveals that the symbiosis-associated transcriptome of petunia involves a set of highly conserved genes that overlaps to a large extent with the complement of AM-associated genes of Medicago, Lotus, and rice. Pi-dependent changes in transcript levels involved mostly the down-regulation of symbiosis-responsive genes encoding PTs, pathogenesis-related (PR) proteins, and certain proteases. Most interestingly, the suppression of genes encoding enzymes involved in carotenoid and apocarotenoid biosynthesis indicates that these pathways are generally suppressed by Pi. These results suggest that high Pi levels trigger a complex anti-symbiotic syndrome, which results in strong repression of AM fungal colonization.
Pi inhibits intraradical proliferation and arbuscule development of Glomus intraradices
To determine the Pi sensitivity of AM in petunia, plants were inoculated with G.intraradices and weekly supplemented with increasing levels of KH2PO4 between 0.1 and 10 mm. Application of 0.5 mm Pi caused a reduction of colonization to approximately 50%, and the interaction was almost completely suppressed at 10 mm (Figure 1). Inspection of the rare colonized root segments showed that in addition to the reduction in total root colonization, high Pi supply caused qualitative differences in intraradical colonization patterns (Figure 2). In general, the fungus formed only small colonies that failed to extend along the root. Instead of the thick hypha formed in control roots (Figure 2a), thin short hyphae were formed at the periphery of the colonies, where they appeared to become arrested (Figure 2b). Confocal analysis of the intracellular stages indicated that the hyphal coils in epidermal cells were not affected by Pi in an obvious way (Figure 2c,d), whereas arbuscules were malformed and less branched and appeared less dense than in controls (Figure 2e,f).
To exclude potential effects of potassium in the KH2PO4 solution, and to test the sensitivity of the interaction to unspecific salt effects (osmotic stress), several control experiments were carried out. In the first experiment, inoculated plants were supplemented with the following salt solutions at a concentration of 5 mm each: KH2PO4, NaH2PO4, K2SO4, MgSO4 and KCl (Figure S1). Root colonization was reduced only in plants supplemented with KH2PO4 or NaH2PO4 (Figure S1a). Shoot and root fresh weight was only marginally affected by the treatments (Figure S1b, white and black bars, respectively), indicating that the plants did not suffer from salt stress. Treatments with nitrate (up to 5 mm KNO3) did not alter AM, indicating that the Pi effect is specific (data not shown). High salt supply (up to 30 mm KCl and K2SO4) did not affect AM colonization significantly, documenting a pronounced robustness of the symbiosis to osmotic stress (data not shown). Taken together, these results establish Pi as the suppressive agent in KH2PO4.
Phosphate acts systemically through improved P-status in the shoot
Phosphate may act directly on fungal development in the soil, or indirectly by changing plant physiology to suppress fungal development in the root. To distinguish between these possibilities we performed split-root experiments in which plants were treated with high Pi levels on one side, and the effects on the AM interaction was assessed on the other half of the root system, which was supplemented with low Pi levels. As controls, plants with split roots were treated on both sides with either low or high Pi levels. High Pi levels exerted a systemic inhibitory effect on AM colonization in roots exposed to low Pi levels (Figure 3a). This effect was associated with a systemic inhibitory effect on the expression of the AM marker gene PhPT4 (Figure 3b) (Wegmüller et al., 2008). Interestingly, repression of AM colonization and PhPT4 expression in the roots did not correlate with P levels in the respective roots, which was not significantly altered (Figure 3c), but rather correlated negatively with shoot P levels, indicating that the symbiotic status of mycorrhizal plants may depend on the P status of the shoot. Relatively high P levels in plants treated with low Pi levels (Figure 3c, left; compare with Figure 4) can be explained with the plants having access to two pots instead of one, hence doubling the absolute Pi supply per plant. The fact that relatively small differences in shoot P levels (Figure 3c) correlated with large effects on AM colonization and gene expression (Figure 3a,b) points to a pronounced threshold effect in P sensing or response. Taken together, our split root experiments suggest that Pi acts primarily through the plant rather than directly on the fungus, although direct effects of Pi on AM fungi cannot be excluded.
Generation of a petunia microarray and experimental setup
The adverse effects of high exogenous Pi levels on AM may be associated with induction of a defence response or with repression of symbiotic functions. To distinguish between these possibilities, we interrogated the transcriptomic response of petunia roots to AM and to high Pi supply using a custom made microarray. First, a set of 45 783 EST sequences was generated from cDNA libraries derived from petunia control roots, mycorrhizal roots, and Pi-treated roots (Table 1; see Supporting information for details) and assembled into a set of 10 150 contigs and 7793 singletons. A complete list of the clustered sequences (referred to as drpoolB) is provided at http://est.molgen.mpg.de/plantDR (User: DidierReinhardt, Password: Al8JJ9mt). Distribution of the sequences to functional groups based on GO-annotation of the closest homologue of A. thaliana is shown in Figure S2. This sequence information was combined with the EST sequences raised from cuttings during adventitious root formation, and with all accessible P. hybrida and P. axillaris nucleotide sequences retrieved from public databases (see Table S1 at http://pgrc.ipk-gatersleben.de/petunia_array). These sequences, which comprise the entire known petunia transcriptome, were clustered to generate a set of 24 816 unigene sequences for the design of a custom microarray by NimbleGen (see Supporting information).
Table 1. cDNA libraries representing different petunia tissues grown under various conditions
Suffix in ID names
All libraries except for dr004 were prepared from Petunia hybrida Mitchell root tissues grown under different conditions as follows: control (30 MM KH2PO4), Pi-treated (5 mm KH2PO4), and inoculated with G. intraradices under control Pi conditions (AM). Substractive libraries were prepared by substracting mRNA from these three conditions in the indicated directions.
Normalized root cDNA library (control conditions)
Substractive root cDNA library (control minus AM)
Substractive root cDNA library (AM minus control)
Substractive root cDNA library (control minus Pi-treated)
Substractive root cDNA library (Pi-treated minus control)
Normalized petal cDNA library (Petunia axillaris)
A total of four independent experiments were carried out to determine the set of genes responding to AM at low Pi levels (30 μm KH2PO4). Two different petunia lines (W115 and W138) were used and plants were harvested 5, 7 and 8 weeks after inoculation. The rational of comparing gene expression of different petunia lines harvested at different time points of mycorrhizal development was to apply a stringent filter and to retain only genes that are robustly and consistently regulated in well established AM, irrespective of the petunia cultivar. Furthermore, early time points were not considered as Pi appeared to act at a relatively late stage of the symbiotic interaction, based on the appearance of residual AM colonization in cortical tissues (Figure 2). To establish the effects of high Pi levels on gene expression, two additional treatments were included in the two-first experiments (harvested 5 weeks after inoculation). In these treatments, inoculated and mock-inoculated plants were weekly supplemented with 5 mm KH2PO4. The following colonization levels were observed in the four experiments. Experiment 1: 74.3 ± 4.4% at low Pi levels, and 2.9 ± 1.2% at high Pi levels; experiment 2: 56.0 ± 2.0% at low Pi levels, and 1.7 ± 0.58% at high Pi levels; experiment 3: 66.3 ± 10.0%; and experiment 4: 72.3 ± 12.3%, respectively. Shoot P content of mycorrhizal plants was increased compared with controls under conditions of low Pi supply whereas high Pi supply caused generally high shoot P content, irrespective of AM inoculation (Figure 4). Total RNA was extracted from roots and used for microarray analysis according to the manufacturer’s guidelines. A complete list of all gene IDs of the array, their expression signal, and their induction ratio in all treatments is presented in Table S2.
AM-specific genes of plant and fungal origin
For comparative expression analysis, the microarray analysis tool Fire2.2 (Garcion et al., 2006) was used to extract genes with particular expression patterns. First, we identified genes of which the expression level was in the range of the background in controls and in the treatment with Pi alone (<100), and which were at least twp-fold induced in all four mycorrhizal samples. These genes were classified as AM-specific (Table S3a–c). As reliable induction ratios cannot be derived if the control expression levels are at the detection limit, expression values are shown for the AM-specific genes (Table S3a–c). IDs were assigned to functional groups according to Journet et al. (2002). Many AM-specific genes were homologous to plant AM markers identified in previous studies on M. truncatula (Journet et al., 2002; Liu et al., 2003, 2007; Wulf et al., 2003; Brechenmacher et al., 2004; Hohnjec et al., 2005; Grunwald et al., 2009), O. sativa (rice) (Güimil et al., 2005), and L. japonicus (Kistner et al., 2005; Guether et al., 2009) (Table S4). They encode PHPT4 and several other transporters (class III), proteases (class IX), glutathione-S-transferase (GST), and class III chitinase (class XII.A). An AM-specific homologue of the gene encoding the carotenoid cleavage dioxygenase 4b (CCD4b) of Chrysanthemum morifolium (Ohmiya et al., 2006) points to an involvement of apocarotenoids that are known to accumulate in mycorrhizal roots (Fester et al., 2007). For 11 IDs, that did not match a sequence in the protein database of NCBI, homologous ESTs from various plant tissues and species were identified (class XIII.A), whereas for further 25 IDs, only EST sequences from mycorrhizal roots of M. truncatula or L. japonicus were identified (class XIII.B). Hence, the latter could represent sequences of plant or fungal origin. Although the AM-specific genes were not induced above background levels by Pi alone, they were induced to low levels in mycorrhizal roots at high Pi levels (AM-Pi), presumably reflecting the residual colonization level of 2.9% and 1.7%, respectively.
Consistent with the mixed contribution of plant and fungal RNA to the mycorrhizal samples used for EST sequencing and microarray analysis, a significant number of AM-specific IDs were homologous to sequences of fungal origin (Table 3b). A further 7 AM-specific IDs showed homology to organisms other than plants and fungi (Table S3c), and 40 IDs did not match any sequence of the public databases. These genes may represent either new unknown AM-specific plant genes or fungal sequences (Table S3c).
AM-induced genes and their response to Pi
Many IDs exhibited moderate to intermediate expression levels in control roots, and induced levels in mycorrhizal roots (Table S3d). This category included, among others, genes encoding symbiotic PTs (PhPT3 and PhPT5), several ABC transporters, the aquaporin NOD26 (class III), and a number of proteases of various types (class IX). Interestingly, many sequences encoded homologues of defence-related proteins (class XII.A), including PR10, barwin-related glucanases, glutathione-S-transferase (GST), peroxidases, chitinases, and germin-like proteins (Table 2). Very few AM-inducible genes were also induced by Pi alone, indicating that the improved P status of mycorrhizal plants (Figure 4) is not reflected by Pi-inducible genes in the root. Notably, a considerable fraction of transporters (group III), proteases (group IX), and stress-related genes (group XII.A) were repressed by Pi alone and/or in mycorrhizal roots at high Pi (Table S3d).
Table 2. AM regulation of genes encoding enzymes involved in carotenoid biosynthesis and processing, in hormone biosynthesis and signalling, and in defence and cell rescue
AM, – Pi
AM, + Pi
Putative function, expression levels, and expression ratios are shown for genes which encode potential carotenoid biosynthetic or processing enzymes, and for genes encoding enzymes implicated in the biosynthesis and metabolism of ABA, GA, ethylene and strigolactone, or components in ethylene and JA signaling. Furthermore, defence-related sequences such as genes encoding PR protein homologues are listed.
Germin like protein/Rhicadhesin receptor precursor
Chitinase 1 precursor
Haem peroxidase, plant/fungal/bacterial
Haem peroxidase, plant/fungal/bacterial
Pathogenesis-related protein PR10
Wound-induced protein 1
Pathogenesis-related 10 protein PR10–2
Elicitor-inducible protein EIG-J7
Cationic peroxidase 1 precursor
Elicitor-inducible protein EIG-J7
Protein disulfide isomerase (PDI)-like protein 2
Patatin-like protein 1
Macrophage migration inhibitory factor family protein
Haemolysin-III related family protein
Hemolysin III-related family protein
Wound-induced protein 1
Basic 30 kDa endochitinase precursor chitinase
Hydrogen peroxide-induced 1
TMV induced protein 1–2
Wound-induced protein WIN2 precursor WIN2 protein
Transmembrane BAX inhibitor motif-containing protein 4
Superoxide dismutase [Fe]
Chloroplast thioredoxin f
As observed in previous studies, fewer genes were repressed by AM, and the extent of regulation was weaker than in the case of the induced genes (Table S3e). Interestingly, several genes encoding mineral nutrient transporters were repressed, in particular nitrate transporters and a zinc/iron transporter (class III), presumably reflecting the improved nutrient status in mycorrhizal roots. Notably, in contrast to the poor overlap between AM-inducible and Pi-inducible genes (see above) many AM-repressed genes were also repressed by Pi. This was particularly evident for several IDs encoding phospholipase D and SPX-proteins which are involved in P-starvation response and signalling (see below). In most cases, repression by Pi was even stronger than by AM, correlating with the P levels in the shoot (Figure 4). These results indicate that improved P status is generally associated with repression of P-starvation responsive genes.
Genes induced by Pi
We next looked for markers of defence which may become induced by Pi alone (Table S3f), or in mycorrhizal roots at high Pi (AM-Pi) (Table S3g). In general, the two lists overlapped to a large degree with slightly more genes being induced by the combined AM-Pi treatment. Most of these additional genes were AM-inducible genes, for which weak levels of induction in the AM-Pi treatment likely reflect the residual colonization level. Notably, only few genes of class XII.A (defence) were induced by Pi alone (1.8%) or by AM-Pi (1.1%), and their induction was weak (Table 2). By comparison, a relatively high number of genes (13%) among the AM-inducible genes were classified as defence markers and they were strongly induced (Table S3d; Table 2). Finally, as stated above (Table S3d), very little overlap between Pi-inducible and AM-inducible genes was observed. Taken together, our data lend little support to the hypothesis that Pi may induce defence mechanisms.
Genes repressed by Pi alone or in inoculated roots at high Pi
To explore whether essential symbiosis-related functions were affected by high Pi levels, the genes repressed by Pi alone, or by Pi in mycorrhizal roots (AM-Pi) were determined. 464 and 492 genes, respectively, were repressed by the two treatments, and the lists were largely overlapping (Table S3h,i). Many repressed genes encoded known markers of P-starvation such as, purple acid phosphatases (PAPs), phytase, RNase, PEP carboxylase, SPX domain-containing proteins, the constitutive PTs PhPT2 and PhPT7, and genes encoding enzymes of fatty acid biosynthesis (Wasaki et al., 2003; Misson et al., 2005; Hernandez et al., 2007). Interestingly, a miRNA399 homologue was induced, whereas the expression of phosphate starvation regulator PHR1 (Bari et al., 2006) was not affected by Pi. A role for miR399 in the regulation of AM has recently been postulated by Branscheid et al. (2010). The petunia homologues of the phosphate transport regulator PHO2 and of At4/Mt4 are not known and were therefore not represented on the array. Taken together, this gene expression pattern is indicative of a general repression of the P-starvation response. Consistently, genes encoding enzymes involved in the biosynthesis of sulfolipids and galactolipids, which replace phospholipids under P-deprivation (Essigmann et al., 1998; Andersson et al., 2003), were repressed, together with lipolytic enzymes involved in the recycling of phospholipids (phospholipase C, phospholipase D, glycerophosphoryl diester phosphoesterase) (Li et al., 2006).
We next considered genes the repression of which may potentially affect AM colonization. Strong gene repression by Pi was observed for the symbiotic PhPT5, like in tomato (Nagy et al., 2009), as well as for the constitutive PhPT2 and PhPT7, and for other transporters (Table S3h,i, Class III). Considering genes with a potential function in secondary metabolism and hormone pathways (class VI), a large fraction encoded enzymes involved in carotenoid production and processing (Table 2, Figure 5). In particular, enzymes of the plastidial MEP pathway and of carotenoid biosynthesis, as well as enzymes involved in biosynthesis of the diterpene-derived hormone gibberellic acid and of the strigolactones were downregulated (for review see Lu and Li, 2008). Genes encoding the ABA biosynthetic enzyme 9-cis-epoxycarotenoid dioxygenase (NCED) were induced. A homologue of jasmonic acid-inducible JA2 was slightly induced, whereas genes encoding components of ethylene biosynthesis and signal transduction were slightly repressed (Table 2).
In the group of signalling components (class X), we observed a strong repression of various kinases, phosphatases, SPX domain-containing proteins with homology to the Arabidopsis P-starvation gene At-SPX3 (Duan et al., 2008), and, notably, of SYM10 of pea (P. sativum), the orthologue of the nod factor receptor NFR5 in L. japonicus (Madsen et al., 2003; Radutoiu et al., 2003). However, no effect of Pi on the expression of the SYM gene homologues PhSYMRK, PhCASTOR, PhPOLLUX, PhCCaMK, and PhNUP133, and of homologues of other components of the nuclear pore complex were found (Table S5). The repression of PAM1, a gene required for intracellular accommodation of AM fungi in cortical cells (Feddermann et al., 2010), is consistent with the defects in arbuscule development in Pi-treated roots (Figure 2b,f). Notably, genes encoding homologues of the AM-inducible PR10 and germin-like proteins, and numerous other defence-related genes that are not affected by AM, were repressed by Pi (Table 2).
Real time PCR confirms Pi-mediated repression of AM-inducible genes
To corroborate the results obtained from array analysis, we analyzed the expression of a number of selected AM-responsive genes, and of the constitutively expressed PhPT1 and SYMRK, by quantitative real time PCR (qPCR). First, qPCR was carried out using the RNA that had already been used for microarray analysis (qPCR1 and qPCR2). The resulting data confirmed the induction in AM, but the extent of regulation was in many cases larger than deduced from microarray analysis, consistent with the larger dynamic range of qPCR analysis compared with microarray analysis (Table S6). Interestingly, not only the induction ratios were larger than anticipated from array data, but also the repression by Pi, in particular of PhPT5 and a terpene synthase, was stronger than in the case of microarray analysis. Subsequently, an independent biological replicate experiment was carried out, which confirmed the general trends (Table S6, qPCR3).
The results from qPCR analysis allowed us to quantitatively assess the degree of gene repression by Pi in the samples inoculated by G. intraradices with and without high Pi supply (Table S6). According to the reduction in colonization of approximately 25-fold, a gene whose expression is reduced about 25-fold would be expressed proportional to the colonization level, and could therefore be defined as a quantitative marker for colonization. Consequently, genes that are repressed >25-fold in at least two of the three replicate experiments (bold repression values and shaded IDs) may be repressed by Pi in a direct way, not just as a result of reduced colonization. Based on this reasoning, most of the genes are repressed directly by Pi, in particular PhPT4 and PhPT5.
Pi-related repression of AM cannot be assigned solely to strigolactone deficiency
One of the major results of the microarray experiments was that Pi represses several enzymes involved in carotenoid biosynthesis and processing. Hence, we suspected that Pi may repress AM by inhibiting the biosynthesis of the apocarotenoid strigolactone. Strigolactone exudation is known to be negatively correlated with Pi supply in red clover (Yoneyama et al., 2007) and tomato (Lopez-Raez et al., 2008). Unfortunately, the strigolactones of petunia are not known preventing the direct determination of strigolactone levels in our experiments. We therefore attempted to complement the Pi-dependent repression of AM development by exogenous supply of the synthetic strigolactone GR24 (Gomez-Roldan et al., 2008) using the strigolactone defective petunia mutant dad1 (Snowden et al., 2005) as a control. Four weeks after inoculation, wild type plants had reached an intraradical colonization level of more than 70% in the presence or absence of GR24 (Figure 6). Dad1 mutants exhibited substantially lower levels of colonization in all categories (extraradical hyphae, hyphopodia, total colonization, and arbuscules) with only 22% total intraradical colonization. All aspects of colonization in dad1 were improved by GR24 (10 nm), although only extraradical and intraradical colonization reached the significance threshold of 0.05%. In contrast, all Pi-treated plants exhibited very low levels of colonization, irrespective of an additional supply of GR24. Even 10-fold higher GR24 concentrations (100 nm) did not enhance AM fungal colonization in Pi-treated roots (Figure 6).
Repression of PT genes by Pi in established AM precedes the reduction in colonization
Compared with the relatively mild effects of downregulation of carotenoid biosynthesis on arbuscule turnover (Floss et al., 2008a,b), mutation or downregulation of AM-specific PTs resulted in pronounced reduction of intraradical AM colonization, presumably due to accelerated arbuscule senescence (Maeda et al., 2006; Javot et al., 2007). Based on these observations, we reasoned that the repression of PT genes by Pi in our experiments may represent a cause of reduced AM colonization, rather than merely its consequence. To establish the sequence of events, plants with colonized roots were treated with high Pi levels, and were subsequently tested over time for reduction in colonization and for repression of PT genes. Little effects on AM colonization were observed during the first week after Pi treatment (Figure 7a), and only after 2 weeks, a significant reduction in colonization was detected. In contrast, a reduction of PhPT5 gene expression was already evident after two days, and after 2 weeks, PhPT5 expression was almost completely repressed (Figure 7b). Similar results were observed for PhPT3 and PhPT4 (Figure S3). Hence, the inhibitory effect on PT gene expression clearly preceeded the reduction of root colonization. This finding suggests that reduced PT gene expression is more likely to be the cause than the consequence of reduced AM colonization in roots supplied with high Pi concentrations.
In our experiments, AM colonization was inhibited by solutions containing 0.5 mM or more soluble orthophosphate (Figure 1), which is in the range of Pi in the soil solution of arable soils (e.g. McDowell and Sharpley, 2001). Our split root experiments show that the effect of Pi on AM symbiosis and on gene expression is systemic (Figure 3), indicating that a systemic signal may relate phosphorus (P) status throughout the plant. However, it should be noted that Pi itself is mobile within the plant (Vierheilig et al., 2000), and tends to accumulate in the shoot where it regulates the expression of P-signalling genes (Burleigh and Harrison, 1999). Interestingly, Pi application through the leaves is sufficient to inhibit AM colonization in the roots (Sanders, 1975). These observations are compatible with a scenario in which Pi from the roots is translocated to the shoot, where a mobile signal is generated to alter the physiology of the roots (Doerner, 2008), and thereby their competence to engage into AM symbiosis.
AM symbiosis in petunia is accompanied by expression of conserved marker genes
We have chosen a transcriptomic approach to reveal the pathways that are affected by Pi. As no transcriptomic study on AM in petunia has been described to date, we first discuss the AM-related changes in gene expression, and then examine how Pi interferes with regulation of gene expression, and how this may influence the interaction between petunia and G. intraradices.
Mycorrhizal petunia roots expressed a conserved complement of AM-associated marker genes (Table S4). The best-studied AM-induced genes, in terms of regulation and function, are the symbiotic PTs, which are essential for Pi transfer from the fungus to the plant, and for effective symbiotic development (Maeda et al., 2006; Javot et al., 2007). In petunia, the symbiotic PhPT3, PhPT4, and PhPT5, representatives of the three conserved symbiotic PT genes of Solanaceae (Nagy et al., 2005; Chen et al., 2007b; Reddy et al., 2008), were induced in mycorrhizal petunia (Table S3a,d) as observed in previous reports (Wegmüller et al., 2008). A characteristic response observed in all transcriptomic studies on AM is the strong induction of genes encoding proteases of various types (Takeda et al., 2007). We have detected AM-specific and AM-induced members of subtilases, cysteine proteases, legumains, and serine carboxypeptidases (Table S3a,d). Furthermore, the petunia AM-induced transcriptome comprised the following known AM-inducible genes: ABC-transporters, aquaporin (NOD26), and various signalling components (Table S3a,d).
Several AM-responsive genes encoded homologues of PR proteins (Table 2) such as chitinase III, which has been recognized previously as AM-specific marker (Salzer et al., 2000), PR10, glutathion-S-transferases (GST), barwin-related endoglucanases, germin-like proteins, and peroxidases. For PR10 and barwin-related glucanases homologues have been identified in various monocot and dicot species, but not in Arabidopsis thaliana (data not shown). This leads us to speculate that AM-induced PR gene homologues are not induced as part of a transient defence response (that is repressed at later stages of symbiosis) but as components of a conserved symbiosis program. Hence, they may serve dedicated functions in symbiosis, which have been under positive selection during the evolution of AM, and were secondarily lost in the non-symbiotic species A. thaliana.
Transcriptional effects of high Pi on defence and on hormonal stress pathways
A central question of this study was, whether inhibition of AM by Pi is related to the induction of a defence response. However, unexpectedly, the percentage of defence-related genes among Pi-induced genes (1.8%) and AM-Pi-induced genes (1.2%) was nearly 10-fold smaller than in the set of AM-inducible genes (13%), and the few potential defence genes induced by Pi exhibited only weak induction ratios, hence, the induction of homologues of PR proteins appears to be associated with the active symbiosis rather than with its inhibition.
Hormonal pathways have been shown to influence AM development (Hause et al., 2007), with ABA and JA positively influencing colonization (Isayenkov et al., 2005; Herrera-Medina et al., 2007), and ethylene limiting infection (Penmetsa et al., 2008). Our data indicate that ABA biosynthesis and JA signalling may be induced by Pi, whereas ethylene biosynthesis and signalling was reduced (Table 2), hence, being compatible with promotion rather than reduction of AM colonization. Taken together, these results do not support the hypothesis that Pi may limit AM development by inducing defence or modifying the balance of stress hormones. Consistent with this notion, the microscopic appearance of the interaction at high Pi levels did not resemble a hypersensitive response, a hallmark of resistance reactions of plants. Rather, fungal colonization appeared to be slowed down during colonization of the root cortex (Figure 2), arguing for a gradual decline and a relatively late abortion of the interaction at the level of arbuscule development and intercellular hyphal spreading in the cortex.
Transcriptional effects of high Pi on symbiotic signalling and carbohydrate metabolism
At least seven common SYM genes are required for progression of AM infection (Parniske, 2008). A reduction of expression of any of these genes by Pi could therefore potentially result in a suppression of colonization. However, the expression of petunia common SYM gene homologues was not affected by Pi (Table S5). This is consistent with the observation that the phenotype of AM inhibition by Pi in the cortex (Figure 2) is considerably different from the phenotype of sym mutants, in which epidermal entry by AM fungi is strongly inhibited (Parniske, 2008). Similarly, the homologue of a receptor kinase which negatively regulates rhizobial and AM fungal colonization in legume roots (Meixner et al., 2005; Magori and Kawaguchi, 2009), was not affected by Pi supply (Table S5). In contrast, a homologue of the L. japonicus receptor kinase NFR5, that was repressed by Pi (Table S5), could represent an important regulator of AM symbiosis. LjNFR1 and LjNFR5 are essential components of nod factor perception and signalling (Madsen et al., 2003; Radutoiu et al., 2003). In analogy, PhNFR5 could play a role as a receptor for a fungal signal.
Pi can interfere with carbohydrate metabolism and transport (Hammond and White, 2007; Müller et al., 2007), and high Pi levels have been hypothesized to lead to reduced carbohydrate allocation to the strictly biotrophic fungal symbiont, thereby providing a plausible mechanism by which plants could limit fungal proliferation under conditions of Pi saturation (Olsson et al., 2006). However, apart from weakly regulated genes encoding sucrose-6-phosphate phosphatase, sucrose-phosphate synthase isoform C, and a fructose-bisphosphate aldolase-like protein, there was no evidence for significant changes in sugar metabolism or transport. Repression of two IDs with homology to genes encoding cell wall invertase inhibitors could potentially lead to increased invertase activity in the apoplast, however, this would not be expected to affect fungal growth (Schaarschmidt et al., 2007). Hence, our data do not support a role of sugar allocation in Pi-related suppression of AM.
Pi represses carotenoid biosynthetic pathways and AM-associated genes
Combined evidence from biochemical and genetic studies suggests that carotenoids and their derivatives (apocarotenoids), which include abscisic acid (ABA) and strigolactones, play prominent roles in AM symbiosis. Apocarotenoids such as mycorradicin accumulate in mycorrhizal roots (reviewed in Strack and Fester, 2006), and maize mutants defective in carotenoid biosynthesis exhibited decreased AM fungal colonization levels (Fester et al., 2002). A root-borne carotenoid-derived signal (Matusova et al., 2005), which stimulates AM fungal branching and metabolism (Buée et al., 2000; Besserer et al., 2006), was identified as strigolactone (Akiyama et al., 2005). Carotenoid biosynthesis occurs in the plastids via the MEP pathway involving DXS2 (Walter et al., 2007; Lu and Li, 2008; Phillips et al., 2008). An alternative potential route of 1-deoxy-d-xylulose 5-phosphate biosynthesis is provided by plastidial transketolase (Bouvier et al., 1998). Inhibition of plastidial carotenoid biosynthesis by knockdown of DXS2, and of CCD1, which encodes a carotenoid processing enzyme, has been shown to result in premature arbuscule senescence in M. truncatula (Floss et al., 2008a,b).
In our experiments, Pi repressed both, transketolase and DXS2 and numerous genes that encode enzymes catalysing subsequent steps of carotenoid biosynthesis and processing (Table 2; Figure 5). In addition, the expression of DAD1, the orthologue of MAX4 in A. thaliana and RMS1 in pea, which encode the strigolactone biosynthetic enzyme CCD8 (Gomez-Roldan et al., 2008; Umehara et al., 2008), was repressed by Pi. These results are consistent with the observed negative correlation between Pi levels and strigolactone biosynthesis (Yoneyama et al., 2007; Lopez-Raez et al., 2008). Indeed, mutation of RMS1 (Gomez-Roldan et al., 2008), or of DAD1 (Figure 6), results in reduced AM colonization. Hence, reduced strigolactone production may contribute to Pi-dependent repression of AM. However, this effect cannot be the principle bottleneck, since the exogenous supply of the synthetic strigolactone GR24 did not alleviate the repression of AM by Pi. In this context it should be noted that the Pi effect was much stronger than the symbiotic mutant phenotypes of strigolactone biosynthetic mutants (Gomez-Roldan et al., 2008; Figure 6) supporting the conclusion, that strigolactone deficiency alone cannot explain the Pi effects.
Pi repressed the constitutive levels of AM-inducible PhPT5, and abolished the induction of AM-specific PhPT4 (Table S6), potentially by reducing the responsiveness to the symbiotic signal lyso-phosphatidylcholine (Drissner et al., 2007; Nagy et al., 2009). Interestingly, when high Pi levels were supplied to plants that had already been colonized for 4 weeks (approximately 50% colonization), repression of all symbiotic PTs was observed within 2 days (Figures 7 and S3), while reduction in colonization was not observed before 2 weeks after Pi addition. Hence, reduced PT gene expression cannot be the consequence of reduced colonization. In light of the observation that mutations in the symbiotic PT genes of L. japonicus (Maeda et al., 2006) and M. truncatula (Javot et al., 2007) caused defective AM interactions, the rapid repression of all symbiotic PTs of petunia may well represent one of the causes for reduced colonization.
Besides PTs, Pi repressed several AM-inducible proteases (Table S3h,i). Interestingly, AM-inducible proteases of the subtilase type are involved in AM development in L. japonicus (Takeda et al., 2009). Hence, Pi could be envisaged to impinge on symbiosis by repression of proteases. However, it should be noted that in petunia, Pi repressed only serine carboxypeptidases, but not the other protease types (subtilase, cysteine protease, legumain). Interestingly, Pi repressed PAM1, a gene that is essential for intracellular development of AM fungi in petunia (Feddermann et al., 2010). Downregulation of the PAM1 orthologue VAPYRIN in M. truncatula, caused similar defects as mutation of PAM1 (Pumplin et al., 2010), emphasizing its conserved role in intracellular accommodation of AM fungi. Hence, the downregulation of PAM1 by Pi could potentially be responsible for some of the defects in cortical colonization of roots supplied with high Pi levels.
In conclusion, taken together, comparative transcriptomic analysis in petunia, Medicago, Lotus, and rice reveals a conserved complement of AM-regulated genes, which may serve essential functions in establishment and functioning of AM symbiosis. High exogenous Pi supply caused a strong systemic inhibition of AM colonization. We have proposed a number of hypothetical mechanisms, by which Pi could interfere with AM. An involvement of a defence response or interference with the common SYM pathway is not supported by our data, nor is Pi likely to act through modification of carbohydrate relations. Rather, Pi appears to act through repression of (apo)carotenoid biosynthetic genes and of various genes with a role in symbiosis, such as phosphate transporters, proteases and genes involved in intracellular accommodation. Future studies should further address the mechanisms involved in the multi-facetted syndrome of Pi-related inhibition of AM. This will help to develop strategies aimed at reducing its adverse effects towards symbiosis, thereby allowing reconciliation of the advantages of Pi-fertilisation with the multiple benefits of the AM interaction.
Plant and fungal material, plant treatments, and evaluation of root colonization
Petunia hybrida (lines W115 and W138) and Glomus intraradices (MUCL 43204) were grown as described (Reddy et al., 2007). For the treatments with nutrient solutions, plants were grown in pots containing a mixture of sand and soil (3:1 v/v) and watered weekly with 50 ml of the indicated solutions prepared with a basic nutrient solution (Reddy et al., 2007). Inoculations of petunia with G. intraradices were performed as described (Reddy et al., 2007). Generation of cuttings from P. hybrida W115 was done according to Ahkami et al. (2009).
Plants for split root experiments were first grown in the presence of low Pi levels (30 μm KH2PO4) for 4 weeks. Then each plant was transplanted to two pots with inoculum, whereby the root system was split between the two pots. To evaluate the dynamics of the Pi effect in colonized roots, plants were first inoculated with G. intraradices. After 4 weeks of culture with low Pi supply, the plants had reached a colonization level of approximately 50% and were treated with high Pi solution for 1–56 days. Treatments with GR24 were carried out thrice weekly for 5 weeks after inoculation. For the determination of colonization levels, roots were harvested at indicated time points and stained with trypan blue according to Reddy et al. (2007). Acid fuchsin staining was carried out following the protocol of Floss et al. (2008a), except that 0.1% acid fuchsin was used. Mycorrhizal colonization levels were determined as described (Reddy et al., 2007).
Root and petal cDNA library construction and EST sequencing
Total RNA was extracted from roots of P. hybrida (W138 and W115) and from petals of P. axillaris using the hot phenol protocol as described (Verwoerd et al., 1989). mRNA purification and concentration was performed using Dynal oligo-dT magnetic beads (Invitrogen; http://www.invitrogen.com). To optimize the random sequencing from cDNA libraries, normalized cDNA libraries were generated by equilibrating abundant and rare transcripts. Normalized libraries were constructed using the cDNA SMART-kit (Clontech; http://www.clontech.com/) and the thermostable duplex-specific nuclease (Zhu et al., 2001). The suffix dr001 in the EST names refers to the normalized library from roots grown at 30 μm KH2PO4, the suffix dr004 refers to the normalized library from petunia petals (corolla tubes) (Table 1). The preparation of subtractive cDNA libraries was carried out with the PCR-Select cDNA subtraction kit (Clontech) according to the manufacturer’s specifications (for further details see Table 1 and Supporting information). The resulting cDNAs were cloned into libraries dsr12, drs21, and drs13 using the TOPO TA cloning kit (Invitrogen) or into library dsr31 using CloneJET PCR Cloning Kit (Fermentas, http://www.fermentas.com). This material was electroporated into ElectroMAX DH10B cells (Invitrogen). Randomly picked clones from cDNA libraries were sequenced at Max-Planck-Institute for Molecular Genetics (Berlin-Dahlem, Germany) using Capillary Sequencer systems: ABI 3730 XL and GE Healthcare (formerly Amersham-Pharmacia, http://www.gelifescience.com/) MegaBace 4500 equipped with Caddy system (Watrex, http://www.megabace.net/megabace/index.html/) and Sequencing kit ABI BigDye Terminator v.3.1 for both sequencing systems.
Construction, sequencing and clustering of a normalized cDNA library of petunia cuttings
RNA was extracted from cuttings at various developmental stages as described (Ahkami et al., 2009) and poly (A) RNA was prepared using oligo(dT) cellulose Type7 according to manufacturer’s instruction (Amersham Pharmacia, Germany). Construction of the normalized cDNA library was performed as described (Lein et al., 2008), with minor modifications, using 5 μg poly(A) RNA. The cDNA library was subjected to three rounds of normalization, involving the denaturation, reassociation and removal of double-stranded cDNAs and the isolation and amplification of single-stranded cDNAs via polymerase chain reaction (PCR). After normalization, equalized cDNAs were ligated at random into the pCRblunt vector and transformed into competent E. coli cells followed by the selection of blue/white colonies. Clones were picked and sequenced by at GATC Biotech AG using Capillary Sequencer systems ABI 3730 XL (Konstanz, Germany). Approximately 4700 sequences were obtained after processing, a success rate of approximately 94%. The average reading length was 495 bp.
Microarray design, hybridization and analysis
The EST sequences generated in this study were clustered together with all sequences of P. hybrida and P. axillaris available at Genbank (15 713), TIGR (4466), and the Solanaceae genomics network SGN (5135) to generate a set of 24 816 non-redundant unique sequences (unigenes) for microarray design. The entire set of sequences used for this clustering is listed in Table 1 (available at http://pgrc.ipk-gatersleben.de/petunia_array). Design of a four-plex microarray with 72 000 features was carried out using the ArrayScribe software from NimbleGen (http://www.nimblegen.com) to generate three optimized independent probes per gene, with an average length of 36 base pairs per probe. Shorter sequences were represented by two probes. Array design, probe synthesis, hybridization, analysis, and data normalization was carried out by Nimblegen. Analysis of expression data sets was carried out with Fire2.2 (Garcion et al., 2006). Quantitative real-time polymerase chain reaction coupled to reverse transcription (qPCR) was carried out as described (Reddy et al., 2007) using the primers listed in Table S7. A complete list of the gene IDs represented on the microarray, and the corresponding expression values in controls and the induction ratios in all treatments is provided in Table S2. For further experimental details see Supporting information.
Shoot phosphorous determination
Shoot phosphorus content was determined in three individual mature leaves per data point as described (Reddy et al., 2007).
We wish to thank Martin Olivi and Sergio Svistoonoff for cloning of phosphate transporters, and Sven Bacher for help with statistical analysis. This research was supported by the Swiss National Centre of Competence in Research (NCCR) ‘Plant Survival’ and by the ‘Pakt für Forschung und Innovation’ of the Leibniz Association, Germany.