Arbuscular mycorrhizal (AM) symbiosis is stimulated by phosphorus (P) limitation and contributes to P and nitrogen (N) acquisition. However, the effects of combined P and N limitation on AM formation are largely unknown.
Medicago truncatula plants were cultivated in the presence or absence of Rhizophagus irregularis (formerly Glomus intraradices) in P-limited (LP), N-limited (LN) or combined P- and N-limited (LPN) conditions, and compared with plants grown in sufficient P and N.
The highest AM formation was observed in LPN, linked to systemic signaling by the plant nutrient status. Plant free phosphate concentrations were higher in LPN than in LP, as a result of cross-talk between P and N. Transcriptome analyses suggest that LPN induces the activation of NADPH oxidases in roots, concomitant with an altered profile of plant defense genes and a coordinate increase in the expression of genes involved in the methylerythritol phosphate and isoprenoid-derived pathways, including strigolactone synthesis genes.
Taken together, these results suggest that low P and N fertilization systemically induces a physiological state of plants favorable for AM symbiosis despite their higher P status. Our findings highlight the importance of the plant nutrient status in controlling plant–fungus interaction.
Arbuscular mycorrhizas (AMs) are the most common and widespread form of plant symbiosis. It is well established that AM fungi (AMF) can improve plant acquisition of phosphorus (P) and nitrogen (N) (Smith & Smith, 2011), but also sulfur and trace elements (Clark & Zeto, 2000; Allen & Shachar-Hill, 2009; Casieri et al., 2012). AM symbiosis appears to be a highly regulated process which takes place only when both partners benefit from nutrient transfer through a reciprocal reward strategy. For example, in Medicago truncatula mutants affected in the AM-specific inorganic phosphate (Pi) transporter 4 (MtPT4) gene, arbuscules accumulated polyphosphate and prematurely degenerated (Javot et al., 2007), suggesting that Pi delivery to cortical cells was necessary for sustaining symbiosis. Conversely, carbon (C) provided by the host to the fungus stimulated fungal P/N uptake and transfer and was correlated with changes in fungal gene expression (Bücking & Shachar-Hill, 2005; Helber et al., 2011; Kiers et al., 2011; Fellbaum et al., 2012).
Although both P and N appear to be important in nutrient transfer during AM symbiosis, little is known about the interconnections between these two elements. Pi is transferred through the mycelium as polyphosphate and released in arbuscules by the action of polyphosphatases (Funamoto et al., 2007). In the extraradical hyphae, N appears to be transported as arginine (Govindarajulu et al., 2005; Tian et al., 2010) which might bind polyphosphate and therefore be coupled to Pi translocation (Jin et al., 2005). Arginine degradation in the intraradical mycelium releases ammonium which then can be transferred to the host plant (Govindarajulu et al., 2005; Tian et al., 2010).
Recently, the importance of nutrient stress and cross-talk between P and N for generating a symbiosis maintenance response in cortical cells was highlighted in M. truncatula. Indeed, in the AM symbiosis-defective Mtpt4 mutant, N limitation suppressed premature arbuscule degeneration and root colonization occurred as in the wild type (Javot et al., 2011).
Plant P status regulates AM symbiosis (Smith & Smith, 2011), and the stimulation of root colonization in low P is controlled via systemic signaling, as shown by split-root experiments (Breuillin et al., 2010; Balzergue et al., 2011). By contrast, the effect of plant N status appears to be more controversial, with low plant N stimulating mycorrhiza formation sometimes but not always (Johansen et al., 1994; Blanke et al., 2005; Olsson et al., 2005). P- and N-starved plants share common features linked to nutrient deprivation, such as reduced shoot development, and accumulation of starch and anthocyanin. Recently, however, an antagonistic cross-talk between nitrate and Pi concentrations was demonstrated in Arabidopsis, where N-limited plants accumulated more Pi in shoots (Kant et al., 2011) and the Arabidopsis NITROGEN LIMITATION ADAPTATION (NLA) gene appears to play an important role in this nitrate-dependent control of Pi homeostasis (Kant et al., 2011). However, to date, little attention has been paid to the influence of combined P and N limitation on the establishment and functioning of the symbiosis, and whether the largely unknown mechanisms linking P and N status play pivotal roles in AM formation remains to be elucidated.
Among the mechanisms that could account for the regulation of AM formation, strigolactones, carotenoid-derived signals released by plant roots, were reported to stimulate fungal branching and enhance root colonization in the early stages of symbiosis (Akiyama et al., 2005; Gomez-Roldan et al., 2008). However, as exogenous strigolactones failed to restore root colonization under high-P conditions, other signals remain to be discovered that may contribute to the control of AM formation (Balzergue et al., 2011; Foo et al., 2013). Strigolactones were reported to be produced in response to either P or N deficiency depending on the plant family (Xie & Yoneyama, 2010). In nonlegumes, strigolactone concentrations increased under nitrate deficiency (Yoneyama et al., 2012). It has been speculated that this mechanism would enhance mycorrhizal symbiosis as a contributor of N. However, to date no study has described the impact of combined N and P deficiency on strigolactone synthesis.
Reactive oxygen species (ROS) are important signaling molecules that play a key role in numerous processes in plants, including nutrient stress, root development and plant–microbe interactions. Upon perception of a pathogen, ROS generated by NADPH oxidases (called respiratory burst oxidase homologs (RBOHs)) participate in a signaling cascade leading to the activation of plant defense responses (Marino et al.,2012). Several studies have also stressed the positive role of decreased (Lohar et al., 2007) or increased ROS production mediated by specific RBOHs in rhizobial symbiosis (Marino et al., 2012; Montiel et al., 2012). As nutrient deprivation has been shown to lead to an increase of ROS production in roots (Shin et al., 2005), it is possible that NADPH oxidases could also participate in AM formation. Indeed, recently, the silencing of a Rac1 (Rho-related GTPase) gene was shown to decrease MtRBOH3/E gene expression in M. truncatula and to impair rhizobial infection while enhancing colonization by AMF (Kiirika et al., 2012).
Here we investigated the effects of combined P and N limitation on AM symbiosis using the model system M. truncatula–Rhizophagus irregularis. We compared the impact of sole N limitation (LN) or combined P and N limitation (LPN) to that of sole P limitation (LP) on AM formation to elucidate whether the plant–fungus interaction depends on cross-talk between P and N. Furthermore, the effects of nutrient limitations on gene expression were studied by transcriptomics to shed light on nutrient-linked mechanisms controlling the onset of AM formation.
Materials and Methods
Plant growth and inoculation
Seeds of Medicago truncatula Gaertner cv Jemalong A17 were scarified, imbibed at 4°C on 0.7% agar plates, germinated and transferred to sand pots and inoculated or not with Rhizophagus irregularis (BEG141) as described in Casieri et al. (2012). Plants were grown for 4 wk in phytochambers with a 16 h day (light 350 μmol m−2 s−2; 23°C): 8 h night (21°C) cycle and fertilized three times per week with modified Long Ashton solutions (Hewitt, 1966). N was supplied as NH4NO3. The final concentrations in the different solutions were 1.3 mM P and 12 mM N (sufficient P and N (SPN)), 0.13 mM P and 12 mM N (LP), 1.3 mM P and 1.2 mM N (LN), and 0.13 mM P and 1.2 mM N (LPN). Split-root experiments were performed on plants cultivated in LPN in sand for 3 wk. The root system was split into two pots containing 200 ml of sand, one of which was inoculated with R. irregularis as already described. Pots containing the inoculum were fertilized three times per week with the LPN solution while the other part of the root system was fertilized with either SPN, LP, LN or LPN solution. Plants were harvested 4 wk post-inoculation. Root staining and AM colonization were performed according to Vierheilig et al. (1998) and Trouvelot et al. (1986).
For the in vitro experiment, after scarification seeds were further sterilized as described in Andrio et al. (2013). P and N limitations were achieved by growing seedlings for 12 d in the same phytochambers as above (see previous paragraph) on modified Fahraeus medium containing either 10 mM NH4NO3 and no P (LP seedlings) or 1.3 mM P and no N (LN) or neither N nor P (LPN). Control full nutrition medium contained 1.3 mM P and 10 mM NH4NO3 (SPN seedlings).
Determination of soluble Pi concentration and N content
Soluble Pi was extracted from freeze-dried shoots or roots as described by Grunwald et al. (2009). Free Pi was determined spectrophotometrically by measuring absorbance at 650 nm using the BIOMOL Green™ Reagent (Enzo Life Sciences, Villeurbanne, France) according to the instructions of the manufacturer.
N content was determined by the method of Dumas (Allen et al., 1974) on a ThermoQuest C/N Analyzer NC 2500 (Thermo Scientific, Courtaboeuf, France) using shoot material dried for 48 h at 80°C.
Reactive oxygen species (ROS) measurements
Freshly harvested roots were immediately ground in liquid nitrogen and used for ROS determination. Hydrogen peroxide measurements were performed by chemiluminescence using luminol (3-aminophthalhydrazide) in a basic solution in the presence of potassium ferricyanide as described by Djébali et al. (2011) in a Berthold LB9507 luminometer (Berthold France, Thoiry, France).
RNA extraction and reverse transcription
RNA was extracted from plant roots using the Plant RNeasy kit (Qiagen, Courtaboeuf, France) according to the recommendations of the manufacturer with an on-column DNase treatment. For microarray hybridizations, three independent biological replicates were produced from three independent cultures. For each condition, RNA samples were obtained by pooling RNA from two to four plants.
For qPCR, RNA (0.5–1 μg) was reverse-transcribed in a final volume of 25 μl in the presence of RNasin (Promega, Courtaboeuf, France) and oligo(dT)15, with Moloney Murine Leukemia Virus (MMLV) reverse-transcriptase (Promega) as recommended by the manufacturer.
Quantitative PCR was performed on reverse-transcribed RNA in three or four independent biological replicates per condition for two or three independent plant cultures. Technical replicates showed a much lower variance than biological replicates. Quantitative PCR reactions were performed in the ABI PRISM 7900 apparatus (Applied Biosystems®, Life Technologies, St Aubin, France) in a final volume of 15 μl using Absolute SYBR green ROX Mix (Thermo Scientific), 70 nM of gene-specific primers and 5 μl of cDNA template diluted 60-fold. Primers used for qPCR are listed in Supporting Information Table S1. Control qPCR reactions included nontemplate control and nonreverse-transcribed RNAs. Reference genes used for normalization were the M. truncatula translation elongation factor MtTEF1α and ubiquitin genes (Table S1). Relative expression levels compared with the reference gene were calculated using the formula 2−△CTspecific gene–reference gene.
Statistical analyses were performed using one- or two-way ANOVA followed by Fisher's test. When variance was different between samples, nonparametric tests (Kruskal–Wallis and Mann–Whitney tests) were performed. Data were considered to be significantly different at P <0.05.
Analysis of the integrity of RNA samples, labeling and hybridization to the Affymetrix GeneChip® Medicago Genome Array (Affymetrix, Santa Clara, CA, USA), subsequent washes, staining and scanning were performed as described in Andrio et al. (2013). The data were normalized with the gcrma algorithm (Irizarry et al., 2003). To identify differentially expressed genes, we performed a standard two-group t-test that assumes equal variance between groups. The variance of the gene expression per group is a homoscedastic variance, where genes displaying extremes of variance (too small or too large) are excluded. The raw P-values were adjusted by the Bonferroni method, which controls the family-wise error rate (FWER; Ge et al., 2003). A gene is differentially expressed if the Bonferroni P-value is < 0.05.
All these steps were performed on the Affymetrix platform at URGV (Unité de Recherche en Génomique Végétale, INRA, Evry, France). The raw CEL files were imported in R software (Ihary & Gentleman, 1996) for data analysis. All raw and normalized data are available through the CATdb database (AFFY_Med_2011_09; Gagnot et al., 2008) and from the Gene Expression Omnibus (GEO) repository at the National Center for Biotechnology Information (NCBI; Barrett et al., 2007), accession number GSE 38847.
Analysis of public transcriptome data
Public transcriptome data on roots interacting with Aphanomyces euteiches (ME16, NCBI GEO (GSE20587)) or 24-h salt-stressed roots (ME7, NCBI GEO (GSE14029)) or roots treated for 3 d with diphenylene iodonium (DPI) (ME13, NCBI GEO (GSE15866)) were downloaded from PLEXdb (http://www.plexdb.org/plex.php?database=Medicago). Data were normalized with the RMA algorithm and genes significantly (corrected Benjamini–Hochberg P-value <0.05) differentially expressed (log2 (fold-change) ≥ 1 or log2 (fold-change) ≤ −1) between test and control conditions were identified using the Agilent Genespring software (Agilent Technologies, Santa Clara, CA, USA). Hierarchical clustering of data was performed using the MultiExperimentViewer software (http://www.tm4.org/mev/).
LPN leads to enhanced AM colonization
To assess the impact of N on plant AM colonization and possible mycorrhizal effects on plant growth and N content, we cultivated M. truncatula plants in four different conditions: (1) SPN, as a control for full nutrition; (2) LP, as a control for conditions favorable to AMF; (3) LN; and (4) LPN. A concentration of 130 μM P was chosen to impose a mild P limitation enabling mycorrhiza formation while not imposing too severe a P stress, which could otherwise mask the effects of N limitation, as demonstrated for sulfur (Sieh et al., 2013). Plants were grown for 4 wk on sand in the presence of either nonmycorrhizal (NM) leek (Allium porum) roots (control plants) or leek roots inoculated with R. irregularis (AM plants).
Shoot biomass decreased drastically in NM-LN and NM-LPN plants compared with NM-SPN and NM-LP plants as a result of severe N limitation (Fig. 1a). The development of chlorosis (not shown) as well as shoot N content confirmed the N limitation experienced by NM-LN and NM-LPN plants, which displayed a twofold lower total N content than NM-LP and NM-SPN plants (Fig. 1b). The ratio of shoot to root biomass, which is a marker of nutrient limitation, was significantly (P <0.05) reduced only in NM-LPN plants (Fig. 1c). In all our experiments (six independent cultures), the lowest AM colonization was displayed by AM-SPN plants, whereas the highest colonization was always obtained for AM-LPN plants. Intermediate AM colonization was obtained for AM-LP and AM-LN plants (Fig. 1d). Quantitative RT-PCR (Fig. 1e) showed that the expression of AM-specific markers (the MtPT4 and the M. truncatula Blue Copper Protein MtBCP1 genes as well as the R. irregularis Tubulin A (RiTUBA) gene, mirrored the AM colonization of roots. The comparison of NM and AM plants for all measured parameters (shoot biomass, shoot/root ratio and N content) failed to uncover any significant difference (Fig. 1) even for LPN plants, despite their higher colonization.
AM colonization in LPN is controlled systemically
To assess whether the enhancement of AM symbiosis in LPN was attributable to either a local or a systemic effect, we performed split-root experiments (Fig. 2a). We monitored AM levels in the part of the roots that locally experienced LPN (inoculated part) while plant nutrient status was controlled by the nutrient conditions applied to the other part (noninoculated part).
Supplying half of the root system with SPN, LP, LN or LPN affected shoot biomass (Fig. 2b), which decreased significantly when N or P was limiting. A greater decrease in shoot biomass was, however, observed in N-limited conditions compared with SPN or LP conditions. The highest AM colonization was observed for LPN plants for all mycorrhiza formation parameters (Fig. 2c).
Although plant biomass was significantly reduced when a low P concentration was applied to the noninoculated compartment, no difference in mycorrhiza formation parameters was observed between AM-SPN and AM-LP plants, probably because mycorrhiza formation levels were low in LP. The systemic control of mycorrhiza formation by P limitation was, however, highlighted by comparing AM formation in AM-LPN and AM-LN plants (Fig. 2b). Significantly higher arbuscule abundance and mycorrhiza colonization intensity were observed in AM-LPN plants compared with AM-LN plants, in line with the systemic P limitation imposed on LPN but not on LN plants. Interestingly, split-root experiments also revealed that both mycorrhiza formation intensity and arbuscule abundance increased in LN plants compared with SPN plants, as well as in LPN plants compared with LP plants (Fig. 2c), suggesting an N-dependent systemic regulation of AM formation. Analysis of the expression of the AM marker genes MtPT4 and MtBCP1 confirmed these observations (Fig. 2d).
LPN plants are less P-limited than LP plants
As plant P status is an important regulator of AM symbiosis, we measured the soluble Pi content of roots and shoots of AM and NM plants cultivated for 4 wk in SPN, LP, LN or LPN conditions. P limitation in NM-LP plants led to a significant decrease in soluble Pi concentration in shoots and roots (Fig. 3a,b) compared with NM-SPN plants. N limitation significantly enhanced Pi accumulation: NM-LN plants accumulated more Pi than NM-SPN plants. Similarly, in NM-LPN plants the Pi concentration was about 1.5-fold higher in shoots and roots than in NM-LP plants (Fig. 3a,b).
The expression of P status marker genes was analyzed by qPCR. The M. truncatula Pi overaccumulator (MtPHO2) gene mediating systemic P starvation signals (Branscheid et al., 2010) was reported to be more highly expressed under P-replete and repressed under P-deficient conditions, whereas the M. truncatula phospholipase D (MtPLD) gene displayed the opposite profile. NM-LP plants displayed higher expression of the MtPLD gene and lower expression of the MtPHO2 gene compared with NM-SPN plants (Fig. 3c). Furthermore, NM-LPN plants were less P-limited than NM-LP plants as they displayed higher expression of the MtPHO2 gene and lower expression of the MtPLD gene in roots than NM-LP plants, in accordance with Pi measurements.
AM-LPN plants accumulated significantly more Pi in shoots and roots than NM-LPN plants (Fig. 3a,b). The higher Pi accumulation in AM-LPN roots was correlated with a significant decrease in the expression of the MtPLD gene, while no significant increase in MtPHO2 gene expression was observed (Fig. 3c), in accordance with its down-regulation in AM roots (Branscheid et al., 2010).
Thus, both Pi measurements and MtPLD gene expression indicated a mycorrhizal effect on Pi accumulation in shoots and roots of AM-LPN plants, although no effect on shoot biomass was demonstrated. Most importantly, AM-LPN plants, which displayed a higher AM level than AM-LP plants, had a higher P status than AM-LP plants.
LN/LPN up-regulates the expression of several Pi transporter genes
The expression of M. truncatula Pi transporter genes was analyzed in roots of NM and AM plants cultivated in SPN, LP, LN or LPN. In addition to the AM-specific MtPT4 transporter gene, five other Pi transporter genes were characterized in M. truncatula as displaying differential expression in AM plants and/or during P limitation (Grunwald et al., 2009). Fig. 4 shows that no repression of Pi transporter genes was induced by AM in LP plants, probably because of the moderate P starvation used, while AM led to the repression of the MtPT1–3 genes in AM-LPN roots. Interestingly, these genes were significantly more highly expressed in NM-LPN than in NM-LP roots and MtPT1 was more highly expressed in NM-LN than in SPN roots, indicating that sole N limitation or combined P and N limitation controlled the expression of some Pi transporter genes.
Transcriptomic analysis reveals a nutrient stress transcriptome in LPN roots
To understand the more intense AM colonization of AM-LPN plants compared with AM-LP plants despite their enhanced P status, we performed a transcriptomic analysis of roots of M. truncatula plants grown for 4 wk in LP or in LPN in the presence (AM-LP and AM-LPN) or absence (NM-LP and NM-LPN) of R. irregularis. RNA from three independent replicate cultures was extracted from roots and used for hybridization with the Medicago Affymetrix GeneChip®. We determined the probesets (hereafter called genes) differentially expressed in AM roots compared with NM roots in either LP or LPN conditions (comparisons AM-LP vs NM-LP and AM-LPN vs NM-LPN); and in LPN conditions compared with LP conditions in AM and in NM roots (comparisons AM-LPN vs AM-LP and NM-LPN vs NM-LP). Applying a twofold cut-off value (log2 (fold-change) ≥ 1 or ≤ −1) and a Bonferroni-corrected P-value <0.05, altogether 2806 genes were significantly differentially expressed (‘diff genes’) in at least one of these comparisons (Tables 1, S2). Clustering the experimental data revealed that the comparisons highlighting the effects of AM on gene expression (AM-LP vs NM-LP and AM-LPN vs NM-LPN) were closely related (Pearson rank correlation coefficient ρ = 0.81) and were more distantly related (ρ = 0.23–0.53) to comparisons highlighting the impact of additional N limitation on P-limited plants (NM-LPN vs NM-LP or AM-LPN vs AM-LP), which also clustered closely together (ρ = 0.77).
Table 1. Number of significantly differentially expressed genes (diff genes) or variable and induced (varI) or variable and repressed (varR) genes for the different transcriptomic comparisons in Medicago truncatula roots
Diff: genes significantly (Bonferonni-corrected P-value<0.05) and differentially expressed (log2 (fold-change) ≥ 1 or ≤ −1).
VarI: vardiff gene displaying enhanced expression in all three biological repeats of the test conditions compared with the three replicates of control conditions and log2 (fold-change) ≥ 1.
Vardiff: genes variable but differentially expressed: they display enhanced or decreased expression in all three biological repeats of the test conditions compared with the three replicates of control conditions and log2 (fold-change) ≥ 1 or ≤ −1.
VarR: vardiff gene displaying decreased expression in all three biological repeats of the test conditions compared with the three replicates of control conditions and log2 (fold-change) ≤ −1.
AM-LP vs NM-LP
AM-LPN vs NM-LPN
NM-LPN vs NM-LP
AM-LPN vs AM-LP
Some genes could not be analyzed statistically because the variability of their expression levels was too high in one or more conditions. Analyzing these genes for enhanced or decreased expression in all three biological repeats of the test conditions compared with the three replicates of control conditions and an absolute difference in expression level ≥ 1 (‘vardiff genes’) led to an extended list of differentially expressed genes (altogether 3099 diff + vardiff genes differentially expressed in one or another comparison; Tables 1, S3). These variable genes were classified as variable but induced (varI) or variable but repressed (varR), although the significance of this differential expression could not be evaluated by the statistics used for analyzing our microarray data. These genes were mainly detected in AM-LP vs NM-LP and in AM-LPN vs NM-LPN comparisons (Table 1), probably because of the low or intermediate levels of mycorrhiza formation, and will therefore be analyzed only for these two comparisons.
We focused first on the LP conditions, which are the standard conditions for studying AM symbiosis. The comparison of genes differentially expressed between AM-LP and NM-LP roots revealed that 189 genes were induced while 15 were repressed (Table S4). The induced genes were grouped into functional categories according to mapman (http://mapman.gabipd.org/web/guest/mapman) classification (Usadel et al., 2005; Table S4). The prominent classes were those involved in secondary, hormone and lipid metabolisms, transport, protein synthesis and proteolysis, development, and regulation of transcription as noted for previous transcriptome analyses on AM roots in low P. Indeed, 56% (106 genes) to 78% (147 genes) of the genes induced in AM-LP roots were common to the sets of genes defined as AM-induced, respectively, by Gomez et al. (2009) and Hogekamp et al.(2011). If varI genes were included in this comparison, these percentages of shared genes reached, respectively, 72% (335 genes) to 84% (394 genes), including the core 15 AM marker genes (Hogekamp et al., 2011) and in particular the MtPT4 and MtBCP1 genes. This validated our transcriptome data in addition to qPCR profiling of 11–16 differentially expressed genes per comparison (Fig. S1), which showed a good correlation with our microarray analyses.
Transcriptome data confirm the different mycorrhizal, N and P status of LPN plants compared with LP plants
Many genes regulated by AM in LP were also regulated by AM in LPN: indeed, 100 out of the 189 diff genes induced in AM-LP roots were also induced in AM-LPN roots (or in the extended analysis: 409 out of 467 diff + vardiff genes). However, more genes (665 diff or 875 diff + vardiff genes; Tables 1, S4) were induced in AM-LPN roots than in AM-LP roots (189 diff genes or 467 diff + vardiff genes) and, as a result, there was a higher number of genes (359 diff genes or 526 diff + vardiff genes) common to the previously identified AM-induced genes in low P (Gomez et al., 2009), indicating a better detection of AM symbiosis-regulated genes. This included arbuscule-specific or arbuscule-induced (Hogekamp et al., 2011; Gaude et al., 2012) as well as putative fungal genes (Gomez et al., 2009; Hogekamp et al., 2011; Table S4). In addition, the expression levels of the AM-induced genes in AM-LPN roots were almost always higher than or equal to those in AM-LP plants (95% of the genes) when it was possible to statistically compare these conditions (364 diff + vardiff genes; Table S4). Therefore, our transcriptomic analyses were in line with the higher AM levels of AM-LPN plants compared with AM-LP plants.
Our physiological approach had shown that P and N regulated systemically M. truncatula root colonization level by R. irregularis. To validate the N limitation experienced by plants, we checked the expression of genes identified as differentially induced or repressed systemically by both nitrate and ammonium limitation in M. truncatula roots (Ruffel et al., 2008). Compared with NM-LP roots, 153 of these genes were differentially expressed in NM-LPN roots and 125 (80%) showed the same profile in NM-LPN vs NM-LP as in N limitation experiments (Ruffel et al., 2008; Table S5). Hence we confirmed that NM-LPN plants were N-limited compared with NM-LP plants.
Also, to extend our analysis of the P status of NM-LPN plants compared with NM-LP plants, we studied the expression of M. truncatula genes homologous to Arabidopsis genes identified as up-regulated systemically by P starvation in roots (Thibaud et al., 2010). These genes were repressed in NM-LPN roots compared with NM-LP roots (Table S6), confirming the higher P status of NM-LPN plants compared with NM-LP plants.
LPN roots display altered expression of genes involved in isoprenoid metabolism and a profile opposite to that of DPI-treated roots
We then focused on the transcriptomic data comparing NM-LPN with NM-LP roots (1607 differentially expressed genes; Table S7) to possibly uncover the processes enhancing AM symbiosis in the former conditions. The biological processes (Usadel et al., 2005) whose components were significantly differentially expressed from others (Benjamini– Hochberg-corrected P-value <0.05) were those associated with RNA and in particular regulation of transcription, secondary metabolism (phenylpropanoids, and early steps of the flavonoid pathway) and glutathione S-transferases. This was reflected in the mapman biotic stress and secondary metabolism diagrams, where many genes were differentially expressed in NM-LPN vs NM-LP roots (Fig. 5a,b).
Many genes involved in secondary metabolism were repressed in NM-LPN vs NM-LP. However, a small fraction were up-regulated (Fig. 5b). This fraction comprised genes (1-deoxy-d-xylulose 5-phosphate synthase, DXS and 1-deoxy-d-xylulose 5-phosphate reductoisomerase, DXR) of the methylerythritol phosphate (MEP) pathway leading to isoprenoid precursors in plastids as well as genes involved in MEP-derived pathways leading to carotenoids (phytoene synthase, PSY, lycopene b-cyclase, LCYB and carotene β-hydroxylase, CYP97A) and strigolactones (dwarf 27, D27 and cytochrome P450, CYP711A1; Fig. 6a, Table S7). Quantitative PCR experiments confirmed that the MtD27 gene was indeed more highly expressed in NM-LPN than in NM-LP roots. In addition, this same profile was displayed by the M. truncatula carotenoid cleavage dioxygenases 7 and 8 (MtCCD7 and MtCCD8) genes (Fig. 6b). Interestingly, genes (M. truncatula ent-copalyl diphosphate synthase, MtCPS and ent-kaurene synthase, MtKS) involved in the synthesis of gibberellins (GAs), which derive primarily from the MEP pathway, were also up-regulated whereas those involved in GA deactivation (M. truncatula gibberellin 2 oxidase, MtGA2OX) were down-regulated. Furthermore, genes (M. truncatula scarecrow, MtSCR, short root, MtSHR and scarecrow-like, MtSCL3) homologous to Arabidopsis genes involved in GA-regulated processes such as ground tissue maturation or GA-mediated root elongation (Heo et al., 2011; Zhang et al., 2011) were also up-regulated in NM-LPN roots.
As many stress-regulated genes were affected in NM-LPN vs NM-LP roots, we compared the expression profiles of genes differentially expressed in NM-LPN vs NM-LP roots with their profiles in different public microarray databases using the M. truncatula Affymetrix GeneChip® and analyzing the response to abiotic/biotic stresses in roots. This included data from roots of in vitro-grown plants interacting with A. euteiches (NCBI GEO (GSE20587)) or Phymatrichopsis omnivora (Uppalapati et al., 2009) or 24-h salt-stressed roots (NCBI GEO (GSE14029; Li et al., 2009)) or roots treated with DPI (NCBI GEO (GSE15866; Andrio et al., 2013). DPI is used as an inhibitor of NADPH oxidases which generate ROS in response to pathogens and environmental cues (Marino et al., 2012; Suzuki et al., 2011). Strikingly, an opposite profile of gene expression was observed in these different cases compared with our NM-LPN root transcriptome (Fig. 5c). This was particularly highlighted in the comparison with DPI-treated roots. Indeed, out of the 927 differentially expressed genes common to our treatment and the DPI treatment, 808 genes (87%) displayed the opposite profile in NM-LPN roots to that observed in DPI-treated roots (Table S8). This suggested that a major component determining the NM-LPN vs NM-LP root profile (50% of the 1607 differentially expressed genes) is the putative activation of NADPH oxidases that are inactivated in DPI-treated roots in vitro. To assess whether NADPH oxidase(s) could indeed be activated in LPN compared with LP roots, we measured ROS concentrations in roots from SPN, LP, LN and LPN plants. After 12 d of culture, seedlings experienced N and P limitation as assessed by the lower shoot/root biomass ratio compared with SPN plants (Fig. 7a). ROS concentrations were higher in LPN than in LP roots (Fig. 7b), in line with the possible activation of an NADPH oxidase in LPN roots.
The opposite profile of NM-LPN roots compared with salt-stressed or pathogen-interacting roots suggested that NM-LPN plants displayed an altered defense/stress status. Indeed, a core set of 51 genes (Table S9) could be defined that either were repressed (43 genes) in NM-LPN roots and induced in DPI-treated, salt-stressed roots and in roots interacting with P. omnivora or with A. euteiches; or presented an opposite profile (induced in NM-LPN and repressed in DPI-treated roots or roots undergoing abiotic/biotic stresses; eight genes). This core set of genes included glutathione S-transferases, and secondary metabolism and signaling genes.
Combined P and N limitation stimulates AM formation and suggests cross-talk between N and P
We used mild P limitation (130 μM P) as low-phosphate conditions to allow AM formation at intermediate levels compared with the control. This P concentration enabled us to detect the effects of combined P and N limitation on AM formation, which otherwise could have been hidden. Our low P concentration was also chosen to limit growth depression induced by severe P limitation that could mask additional effects of other nutrient limitation (Sieh et al., 2013).
Our results showed that AM levels increased significantly when plants were grown in AM-LPN compared with other conditions (Fig. 1d). Indeed, the mycorrhizal marker genes MtPT4 and MtBCP1 were significantly induced in LPN plants, indicating that P and N deficiencies had cumulative effects on AM formation. In addition, split-root experiments revealed that AM formation in LPN plants was under systemic control not only by P, but also by N. This supports an interplay between P and N in AM symbiosis.
The comparison of NM and AM plants for different parameters (shoot biomass, shoot/root ratio and N content; Fig. 1a–c) failed to uncover any significant difference between mycorrhizal and nonmycorrhizal plants even in the well-colonized AM-LPN roots. It has been previously reported that N acquisition by the fungal symbiont depends on the N fertilization level and that mycorrhizal plants can regulate plant N uptake with respect to the amounts of N in the soil (Azcon et al., 2008). In lettuce (Lactuca sativa) plants, the highest N uptake was observed for intermediate N concentrations (6 mM) while lower (3 mM) and higher (9 mM) fertilization levels reduced N uptake. As we used a very low fertilization level in LN and LPN conditions (1.2 mM), it is possible that fungal N uptake was too limited. The contribution of AMF to plant total N acquisition can be low (Tanaka & Yano, 2005). Experiments using compartmented growth systems in which only the fungus can access the 15N nitrogen source could help to improve evaluation of the role of AM symbiosis in the plant N economy (Hodge et al., 2001). The differential expression upon AM formation of putative N transporters genes (Table S2; Hogekamp et al., 2011; Gaude et al., 2012) sustains, nevertheless, the hypothesis that AMF do participate in N transfer to plants by altering plant N transporter gene expression.
By contrast, there was a mycorrhizal effect on free Pi contents of AM-LPN shoots and roots and of AM-LN roots (Fig. 3a,b). In the latter case this was unexpected because AM levels in AM-LN plants were low and close to those of AM-LP roots (Fig. 1d), which did not display any mycorrhizal effect on Pi. AM-LN plants appear therefore to display enhanced AM-linked Pi accumulation in roots compared with AM-LP plants. This observation is reminiscent of the rescue by N limitation of the M. truncatula Mtpt4 mutants defective in AM symbiosis (Javot et al., 2011). It is possible that N limitation mimics a functional symbiosis by enhancing AM-dependent Pi accumulation in roots and thereby triggers a symbiosis maintenance response in the Mtpt4 mutant.
NM-LN and NM-LPN plants displayed a higher soluble Pi concentration in shoots and roots compared with their non-N-limited controls (respectively, NM-SPN and NM-LP plants; Fig. 3a,b). This may, presumably, be a consequence of the strong N limitation experienced by the former plants, resulting in reduced shoot development and accumulation of excess soluble Pi in both shoots and roots. This reduced plant development would also probably explain the enhancement of AM-linked Pi accumulation described in the paragraph above in N-limited conditions.
In summary, our results show that N limitation and AM are two distinct causes of an increase in the concentrations of free Pi in shoots and roots. Our qPCR experiments indicated, however, that several Pi transporters genes (in particular MtPT1 and MtPT2) were up-regulated in N-limited plants despite their higher P status, suggesting a cross-talk between P and N in regulating these transporters. In Arabidopsis, Kant et al. (2011) reported that N limitation favored Pi accumulation in shoots and, more generally, that P and nitrate displayed an antagonistic cross-talk. The induction of Pi transporter genes in low N could therefore explain in part the antagonism between nitrate and P in M. truncatula. Interestingly, N limitation in maize (Zea mays) was suggested to lead to the up-regulation of putative Pi transporter genes, again highlighting cross-talk between P and N (Xu et al., 2011) which was also underlined recently by transcriptomics (Schluter et al., 2012).
A role for strigolactones and possibly GA in AM formation in LPN
Microarray analyses revealed that genes involved in strigolactone synthesis were up-regulated in NM-LPN plants. Strigolactones favor AM formation by stimulating hyphal branching and therefore the interaction of AMF with plant roots. In petunia (Petunia hybrida), mutants affected in strigolactone synthesis or export showed retarded colonization rates (Gomez-Roldan et al., 2008; Kretzschmar et al., 2012). Arbuscules and intraradical hyphae were, however, normal, suggesting that decreased strigolactone levels impacted the frequency of hyphal penetration and intraradical hyphal expansion rates, but not the later stages of AM symbiosis. The enhanced expression of strigolactone synthesis genes in NM-LPN roots, leading presumably to higher strigolactone production, could therefore explain the quantitatively enhanced AM levels in AM-LPN roots compared with AM-LP roots. Our observations are consistent with the reported regulation of strigolactone production by P or N in plants (Breuillin et al., 2010; Xie & Yoneyama, 2010; Yoneyama et al., 2012; Foo et al., 2013) and its systemic control by plant nutrient status (Balzergue et al., 2011).
In petunia, strigolactone deficiency was suggested to explain part of the inhibitory effects of high P on symbiosis (Breuillin et al., 2010). Interestingly, in this case, high P led to the repression not only of genes involved in strigolactone synthesis, but also of carotenoid and GA biosynthetic genes, and induced the expression of a gene involved in abscisic acid synthesis (Breuillin et al., 2010). This profile is strikingly opposite to that of LPN roots (Fig. 6a), suggesting that the latter roots mimic a P starvation state despite the higher P status of NM-LPN plants. LPN would thus induce a physiological state in roots comparable to P limitation as a result of long-term nutrient deficiency, as shown by the numerous common genes regulated by N and P starvation (Morcuende et al., 2007).
In Arabidopsis, low P was correlated with lower expression of GA biosynthesis genes and enhanced expression of a GA deactivation gene (Jiang et al., 2007). This response does not appear to be conserved in M. truncatula, as higher expression of GA biosynthetic genes was observed NM plants grown in low P (Hogekamp et al., 2011) and in LPN (our data). GA synthesis genes were shown to be up-regulated upon fungal contact (Ortu et al., 2012) and during the later stages of AM symbiosis (Hogekamp et al., 2011), supporting the hypothesis that GA could participate in regulating AM formation. GA plays an important role in controlling root development; for example, in Arabidopsis, in conjunction with the AtSCR and AtSHR genes, it controls the timing of middle cortex formation and, in conjunction with the AtSCL3 gene, it controls root elongation (Heo et al., 2011; Zhang et al., 2011). It is possible that nutrient limitation, by affecting root development, up-regulates these genes. Although the role of GA in AM symbiosis remains unclear, DELLA proteins (negative regulators of GA signaling containing a unique N-terminal DELLA domain) play an important role in plant adaptation to biotic/abiotic stresses in Arabidopsis by controlling the salicylic/jasmonic acid balance and by regulating ROS concentrations (Achard et al., 2008; Navarro et al., 2008).
NADPH oxidases in LPN could affect AM symbiosis
A striking observation was that about half of the genes differentially expressed in NM-LPN vs NM-LP roots displayed a reciprocal profile in DPI-treated roots. DPI is an inhibitor of NADPH oxidases and other flavin-containing enzymes. Although we cannot exclude the possibility that the observed profile could be a result of the inactivation of a protein different from NAPDH oxidases, published data as well as our ROS measurements (Fig. 7b) suggest that LPN could lead to the activation of NADPH oxidases in plant roots. Indeed, in Arabidopsis, nutrient deprivation (N, P or potassium (K)) was shown to lead to an increased production of ROS in roots (Shin et al., 2005) and specific NADPH oxidase genes were necessary for up-regulating genes in response to nutrient deficiency (Shin & Schachtman, 2004). In the M. truncatula genome, at least 10 genes encoding NADPH oxidases (MtRBOHA–J) and several expressed sequence tags (ESTs) corresponding to other distinct MtRBOH sequences were identified (Marino et al., 2012). Inspection of the Affymetrix GeneChip® for all the probes encoding possible RBOH genes showed that no identified RBOH gene was induced in NM-LPN vs NM-LP plants (Fig. S2). Therefore, either an unidentified MtRBOH gene absent from the Affymetrix GeneChip® is induced in NM-LPN or NM-LPN leads to the post-translational activation of an NADPH oxidase. The NADPH oxidase encoded by the MtRBOHE gene and that corresponding to TC11109, which are both constitutively and highly expressed in all our conditions, could be good candidates for post-translational activation. The activation of an NADPH oxidase in NM-LPN roots could explain the coordinated regulation of numerous genes, in particular genes involved in carotenoid and GA synthesis/deactivation and root patterning genes (Fig. 6b).
The opposite profile of NM-LPN roots compared with DPI-treated roots or roots interacting with pathogens or salt-stressed roots suggested that LPN broadly altered the defense status of plants. Most genes of the defense/stress core set (Table S9) were down-regulated in NM-LPN roots, underlining the repression of defense-related genes in NM-LPN roots. Different studies have highlighted the importance of controlling ROS produced by the host plant or by the microsymbiont partner for the establishment of an effective rhizobial symbiosis or mutualistic interaction (reviewed in Puppo et al., 2012). During rhizobial symbiosis, ROS production was linked to suppression of pathogenesis-related gene expression (Peleg-Grossman et al., 2012). For AMF, the importance of controlling host plant immunity was recently highlighted in a study by Kloppholz et al. (2011), in which a secreted effector from G. intraradices was shown to promote symbiosis by counteracting the plant immunity program. We propose that in NM-LPN roots the activation of RBOH leads to altered plant defenses and results in better root colonization by AMF than in AM-LP plants. This is consistent with the positive effect of ROS production by RBOHs on rhizobial symbiosis (Marino et al., 2012; Montiel et al., 2012) but is apparently contradictory with the fact that down-regulation of an MtRBOH gene in M. truncatula promoted initial AMF colonization of roots (Kiirika et al., 2012). This latter discrepancy could be a result of the interaction being analyzed at a later stage in our experiments. Taken together, our data suggest a complex role of ROS in controlling symbiosis and plant defense responses, depending on the stage of the interaction. Interestingly, the candidate MtRBOH3/E gene that may be activated in LPN shows a particular profile during elicitation of plant cells by yeast and during plant interaction with P. omnivora (http://mtgea.noble.org/v2/): in contrast to most other MtRBOH genes of M. truncatula which are up-regulated in these cases, the MtRBOH3/E gene is repressed, suggesting that it plays a different role from the other MtRBOH genes in regulating plant defenses. It would therefore be interesting to evaluate the role of MtRBOH3/E at different stages of AM symbiosis.
This work was supported by a grant from the Burgundy Regional Council (Grant FABER 2009-9201AAO036500681). We gratefully acknowledge the help of M. Touratier for shoot N content determination, A. Colombet and V. Monfort for providing the inoculum, L. Casieri for giving advice in plant culture and inoculation, and L. Ma, A. Baraton and E. Bayet for help with plant harvest and analyses. We thank A. Krapp and colleagues from IPM and GEAPSI for helpful discussions, as well as three anonymous reviewers and the editor for helpful comments which have improved the manuscript.