Dissection of local and systemic transcriptional responses to phosphate starvation in Arabidopsis


  • Marie-Christine Thibaud,

    Corresponding author
    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author
  • Jean-François Arrighi,

    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author
    • Present address: Laboratoire des Interactions Plantes-Microorganismes, UMR 441/2594 INRA/CNRS, F-31326 Castanet-Tolosan, France.

  • Vincent Bayle,

    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author
    • Present address: Laboratoire de Reproduction et Développement des Plantes, Ecole Normale Supérieure de Lyon, IFR 128 Biosciences Lyon-Gerland UMR5667, 46 allée d’Italie 69364 Lyon cedex 07, France.

  • Serge Chiarenza,

    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author
  • Audrey Creff,

    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author
  • Regla Bustos,

    1. Department of Plant Molecular Genetics, Centro Nacional de Biotecnologia, Consejo Superior de Investigaciones Cientificas (CSIC), Campus de Cantoblanco, 28049 Madrid, Spain
    Search for more papers by this author
  • Javier Paz-Ares,

    1. Department of Plant Molecular Genetics, Centro Nacional de Biotecnologia, Consejo Superior de Investigaciones Cientificas (CSIC), Campus de Cantoblanco, 28049 Madrid, Spain
    Search for more papers by this author
  • Yves Poirier,

    1. Département de Biologie Moléculaire Végétale, Biophore, Université de Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author
  • Laurent Nussaume

    1. CEA, DSV, IBEB, Lab Biol Develop Plantes, CNRS, UMR 6191 Biol Veget & Microbiol Environ, Aix-Marseille Université, Saint-Paul-lez-Durance, F-13108, France
    Search for more papers by this author

For correspondence (fax +33 (0) 442 254597; e-mail mcthibaud@cea.fr or laurent.nussaume@cea.fr).


Phosphate is a crucial and often limiting nutrient for plant growth. To obtain inorganic phosphate (Pi), which is very insoluble, and is heterogeneously distributed in the soil, plants have evolved a complex network of morphological and biochemical processes. These processes are controlled by a regulatory system triggered by Pi concentration, not only present in the medium (external Pi), but also inside plant cells (internal Pi). A ‘split-root’ assay was performed to mimic a heterogeneous environment, after which a transcriptomic analysis identified groups of genes either locally or systemically regulated by Pi starvation at the transcriptional level. These groups revealed coordinated regulations for various functions associated with Pi starvation (including Pi uptake, Pi recovery, lipid metabolism, and metal uptake), and distinct roles for members in gene families. Genetic tools and physiological analyses revealed that genes that are locally regulated appear to be modulated mostly by root development independently of the internal Pi content. By contrast, internal Pi was essential to promote the activation of systemic regulation. Reducing the flow of Pi had no effect on the systemic response, suggesting that a secondary signal, independent of Pi, could be involved in the response. Furthermore, our results display a direct role for the transcription factor PHR1, as genes systemically controlled by low Pi have promoters enriched with P1BS motif (PHR1-binding sequences). These data detail various regulatory systems regarding Pi starvation responses (systemic versus local, and internal versus external Pi), and provide tools to analyze and classify the effects of Pi starvation on plant physiology.


Inorganic phosphate (Pi) represents the main source of phosphorus for plants, yet this crucial macroelement is often found in very low concentrations, and is insoluble and unevenly distributed in soil. To counter these constraints, plants have developed various strategies to detect Pi distribution in their environment and to adapt their physiology to variations in Pi concentration.

It was previously shown that a local supply of Pi can simulate soil heterogeneity, demonstrating the capacity of plants to stimulate root development in the Pi-enriched area (Drew, 1975; Linkohr et al., 2002). Furthermore, numerous physiological traits of root growth are associated with a local response to Pi deficiency, including: a reduction in the size of the root cells (Reymond et al., 2006); an increase in the length of the root hair (Bates and Lynch, 1996); a cell cycle arrest of the primary root meristem, promoted by a sensing mechanism located in the root tip (Ticconi et al., 2004; Svistoonoff et al., 2007; Arnaud et al., 2010); and an altered development affecting secondary roots (Linkohr et al., 2002; Lopez-Bucio et al., 2002; Reymond et al., 2006).

In addition, there is also an integration of the global Pi status of the plant involving a systemic response (also known as a long-distance response). For example, it has been demonstrated with a split-root experimental design that communication takes place between the root parts of a single plant submitted to uneven distribution of Pi concentrations (Liu et al., 1998; Burleigh and Harrison, 1999; Franco-Zorrilla et al., 2005). This was revealed by the downregulation of Mt4 and several high-affinity Pi transporters in the phosphate-starved part of the root of Medicago (Burleigh and Harrison, 1999), Solanum lycopersicum (tomato; Liu et al., 1998) and Arabidopsis (Franco-Zorrilla et al., 2005), respectively.

The primary signal involved in long-distance signaling has been directly linked to Pi present in the medium and/or the plant cell, as evidenced by a non-metabolizable form of Pi, such as phosphite (Ticconi et al., 2001; Varadarajan et al., 2002), which suppresses phosphate starvation traits. This effect has also been shown by phosphate-sequestering intracellular metabolites, which induce Pi starvation-responsive ribonuclease genes (Köck et al., 1998). Recent genetic evidence corroborates these previous studies, with the identification of phf1, a mutant of the Pi transporter traffic facilitator 1 gene. In this mutant, the activity in several high-affinity phosphate transporters is abolished, resulting in a constitutive low level of intracellular Pi. Such mutant plants express many Pi starvation genes constitutively, even when grown on a phosphate-rich medium (Gonzalez et al., 2005). The factors involved in the long-distance signaling pathway are controversial, but they seem to require components other than Pi. Genetic approaches have further identified some elements of the pathway. For instance, the pho2 mutant, which was selected on the basis of Pi accumulation in leaves (Delhaize and Randall, 1995), is affected in the production of an E2 ubiquitin conjugase enzyme. This enzyme was found to modulate the expression of genes such as High-affinity transporters (PHT1;8 and PHT1;9) and members of the IPS1/At4 family. A complex regulatory system has been identified involving the degradation of PHO2 mRNA through the action of microRNA399 (mir399). This mir399, regulated by At4 (Franco-Zorrilla et al., 2007), moves through the phloem, and might act as a long-distance signal (Aung et al., 2006; Bari et al., 2006). Although these elements appear to be involved in the systemic regulation of genes that respond to Pi starvation, only a limited number of targets have been investigated, as it is unknown which genes are systemically regulated by phosphate deficiency at the whole-genome level. Nevertheless, transcriptional control is a key level of regulation for the response to Pi starvation, and one of its major actors has been identified as the Myb transcription factor PHR1 (Rubio et al., 2001). Recently, exhaustive genome-wide analyses on Arabidopsis thaliana (Misson et al., 2005; Morcuende et al., 2007) have identified numerous genes regulated by Pi starvation; however, neither of these studies was able to distinguish between local and systemic regulations. In the present work, we have addressed this issue by performing a split-root experiment, in which two parts of the same root were supplied with independent media that differ in their Pi content. RNA extracted from the different root parts was then analyzed with Affymetrix ATH1 DNA chips for a full genome analysis. Such experiments gave a global view of the transcripts that are regulated locally or systemically by Pi supply, and lead to a detailed characterization of the roles of external versus internal Pi, and of elements involved in the Pi transduction pathway.


Transcriptomic analysis of genes locally or systemically regulated by phosphate starvation

Arabidopsis plants were grown hydroponically. The roots were divided into two halves and placed in separate media compartments, such that one half received phosphate, while the other half did not receive any phosphate (Figure 1a). For control plants, both compartments were filled with the same solution [either 500 μm Pi (cP500) or no Pi (cP0)] to mimic plants growing in a homogeneous medium. A global transcriptomic analysis (using ATH1 affymetrix chips) was then performed with mRNA extracted from these roots (Figure 1a).

Figure 1.

 Root development in the split-root experiments.
(a) In 2-month-old mature plants, photographs are taken 7 days after transfer to a liquid medium supplied with phosphate (500 μm) or no phosphate (above), and Pi content in the roots (below); control plants are named cP500 and cP0, and split experiments are named spP500 and spP0.
(b) In vitro root growth after transfer of 7-day-old plantlets following 5 days in compartmentalized dishes; black arrows show root tips when transferred, whereas white arrowheads show root tips after 5 days in the split condition.

The controls were used to select transcripts, which present an at least two-fold change in expression between the two conditions (cP500 and cP0) after 2 days of incubation. We identified 412 and 351 genes significantly up- or downregulated by Pi starvation, respectively (Student’s t-test, P < 0.05) (Figure 2a; Table S1).

Figure 2.

 Transcript analysis in roots of Arabidopsis plantlets submitted to a split-root experiment.
(a) Systemic or local regulation of genes induced (left panel) or repressed (right panel) by inorganic phosphate (Pi) starvation. In black: representative expression level of genes in the control conditions (cP500 in high level of Pi; cP0 in low level of Pi). In blue: representative expression level of genes in the split conditions (spP500 and spP0 correspond to the half root system in high Pi or low Pi respectively). IS (systemic) and IL (local) for the induced (I) genes in the control conditions and RS (systemic) and RL (local) for the repressed genes (R). The a, b and c classification denotes variations in systemic response: a, strong (80–100%); b, mild (50–80%); c, weak (20–50%).
(b) Distribution and function of locally or systemically regulated genes, as detailed in Tables 2 and 3 (the number of genes in the corresponding class is given in).

The levels for the selected transcripts were determined for both halves of the root system immersed in the two distinct culture media (high and low Pi, designated as spP500 and spP0, respectively). Transcripts were classified according to their expression level in spP500 or spP0: (i) locally regulated genes were designated as transcripts that present a similar expression level compared with cP500 or cP0 controls (Figure 2a; Table S1), whereas (ii) transcripts that displayed significantly altered accumulation of spP0 and spP500 versus their respective controls (cP0 and cP500) were termed systemically regulated genes (Figure 2a; Table S1). The ATH1 array data were validated by a selection of genes tested by RT-qPCR (Figure S1a), and also by using various transgenic lines that express reporter genes fused to the promoter of genes identified by the present analysis (PHT1;4, IRT1, ACP5, PLDz2) (Figure S1b).

Local regulation

We identified 301 and 240 genes that were either induced or repressed by Pi starvation, respectively. These genes had similar values when compared with their respective controls (classes IL and RL; Figure 2a), indicating that a large number of the genes modulated by Pi starvation are locally regulated (73 and 68%, respectively).

Systemic regulation

We distinguished subclasses of systemically regulated genes based on the level of deregulation observed, directly reflecting a gradation in the response at the transcription level. In class IS1, the systemic effect only modulated genes in the low Pi compartment (spP0). Within IS1, we differentiated between three subclasses (Figure 2a), based on the degree of repression of gene expression in low Pi (spP0) as compared with the control (cP0). For class IS1a, eight out of nine genes show a significant modification of gene expression. The effect of a systemic regulation in class IS1a (Figure 2a; Table S1) is very clear: the induction of these genes when plants are grown in homogenous Pi-starved medium (cP0) is almost completely lost (80–100%) in the Pi-deprived compartment (spP0). Classes IS1b and IS1c (Figure 2a; Table S1) exhibit a similar trend compared with class ISa, but with milder and less significant effects: transcript levels are reduced 50–80% (ISb) and 20–50% (IS1c), relative to control plants grown in low Pi (cP0). In parallel, repressed genes were also classified according to their response in high Pi (spP500), as compared with the expression level in the control (cP500) (Figure 2a; Table S1).

In addition, we also observed genes that were deregulated in both compartments (classes IS2 and RS2). In these two classes (Figure 2a), the transcripts of selected genes appear to present intermediary values when compared with levels measured in cP500 or cP0. These values are quite similar in both spP500 and spP0 compartments (Table S1).

Cis-regulatory sequences

Numerous cis-regulatory sequences (P1BS, PHO, PHO-like, W-box or Pi-responsive) have been previously described as controlling the response to Pi starvation (Mukatira et al., 2001; Rubio et al., 2001; Hammond et al., 2003; Devaiah et al., 2007). Their putative presence in the promoter (2-kb upstream ATG; Table S1) of the selected genes was investigated (Table S2). Only the P1BS box, the binding site for the PHR1 MYB transcription factor (Rubio et al., 2001), was correlated with the different IS classes (Table 1), and was clearly over-represented in the IS1a and b classes (89 and 84%, respectively). These in silico data are corroborated by a published transcriptomic analysis of phr1 and phr1 phl1 mutants (Bustos et al., 2010). PHR1 belongs to a wide multigenic family, and PHL1 is its closest homolog in Arabidopsis. This analysis of phr1 and phr1 phl1 transcripts revealed a perfect correlation of the genes in the IS1a and b classes, which contain the P1BS box in their promoter. On the other hand, classes IS1c and IS2 show a reduced but still significant correlation for markers in these mutants (Table 1). As expected for redundant genes, this correlation is stronger in the phr1 phl1 double mutant (100% for IS1 a and b; 87% for IS1c and 48% for IS2, respectively; Table 1), than when compared with the phr1 mutant (100% for 1S1a; 70% for IS1b; 82% for IS1c and 34% for IS2). It should be noted, however, that the single phr1 mutation alone affects a large majority of the systemically regulated genes containing P1BS boxes (IS classes). This emphasizes the role of PHR1 as a major element of the Pi signal transduction pathway.

Table 1.   Distribution of the PHR1 binding site (GNATATNC) in the promoter of the genes classified, and the proportion of these genes regulated in the phr1 mutant and the phr1 phl1 double mutant
 Distribution in the classes (%)
  1. aPHR1 binding site (34.7% in the full genome). Values in bold reveal enrichment in the class compared to the entire genome.

phr1 phl189844024151317121525

Local and systemic classes cluster functions modulated by phosphate starvation

Distinct functions appear systemically or locally regulated in response to Pi starvation. Genes related to Pi recovery or recycling (for instance High-affinity Pi transporters, phosphatases, enzymes involved in phospholipid remobilization, sulfo- and galactolipid synthesis and nucleases) have been previously identified as strongly induced by Pi starvation conditions (Misson et al., 2005). Interestingly, we found these same genes to be systemically downregulated in the spP0 compartment (Figure 2b; Tables 2 and S1). This is consistent with the result that they cluster mainly in the IS1a and b classes, where they represent 43% of the known genes. As reported here (Table 2), many genes involved directly in Pi uptake or recycling, such as high-affinity phosphate transporters (PHT1;8, PHT1;4 and PHT1;5) and phosphatases (AT1g73010, PAP12, PAP22 and PAP17), are therefore present in the list of genes regulated systemically. Several important genes such as the Pi transporters PHT1;1 and PHT1;7 were not analyzed in the present study, because of their absence on the ATH1 array. In addition to genes directly involved in Pi uptake or recycling, plants can remobilize Pi from various components (nucleic acids and phospholipids) to cope with Pi starvation. Genes associated with these processes are also systemically regulated (Table 2). Therefore, many genes involved in the conversion of phospholipids into galacto- or sulfo lipids are located in the IS1 classes (SQD1, SQD2, MGD2, MGD3, DGD2, PLDz2 and NPC4/NPC5). Their induction in the split-root experiment (spP0 compared with spP500; Table 2) is still relevant, yet attenuated compared with the control experiment (cP0 compared with cP500; Table 2).

Table 2.   Selection of systemically-regulated genes with specific functions as described in Figure 2(b)
ClassGene nameGene IDFold change
  1. Fold changes (ratio of normalized values) are given for the control (cP0/cP500 or cP500/cp0 for induced or repressed genes, respectively) and the split treatment (spP0/spP500 or spP500/spP0 for induced or repressed genes, respectively).

 Pi recyclingSQD2AT5G012204.501.63
Inositol-1(or 4)-monophosphataseAT3G028702.331.12
 Pi recoveryPHT1;8AT1G208602.891.20
 Pi signaling and sensingPHO1-H1AT1G687402.970.82
 Pi recyclingMGD3AT2G1181020.194.66
Protein similar to phosphataseAT1G7301019.295.03
Glycerophosphoryl diester phosphodiesteraseAT3G020407.633.20
 Pi signaling and sensingSPX3AT2G4513047.997.51
 Pi recoveryPHT1;5AT2G3283010.363.22
 Pi recyclingACP5 (PAP17)AT3G177904.582.14
3(2),5-bisphosphate nucleotidaseAT5G640002.992.31
 Metal homeostasisZinc transporterAT3G588107.561.45
Ferric-chelate reductaseAT1G230203.651.22
 Response to metalCation efflux family protein/metal tolerance proteinAT3G580604.730.90
Iron-responsive transporterAT5G035702.801.21
 Metal bindingOxidoreductase, 2OG-Fe(II) oxygenase family proteinAT3G1290027.562.44
ATPase E1-E2 type family protein/heavy-metal-associated domain-containing proteinAT4G301206.141.67
 Metal homeostasisIRT1AT4G1969083.585.13
Copper transporterAT3G4690020.003.27
 Pi sensingPHOI-H2, EXS family protein/ERD1/XPR1/SYG1 family proteinAT2G032602.891.54
 Response to metalBasic helix-loop-helix (bHLH) family proteinAT2G281609.193.38
 Metal bindingZn ion binding/protein bindingAT1G7477010.332.00
Zinc finger (C3HC4-type RING finger) family proteinAT1G747608.262.01
Oxidoreductase, 2OG-Fe(II) oxygenase family proteinAT1G493902.111.40
 Metal bindingZinc finger (C3HC4-type RING finger) family proteinAT1G722004.712.82
Zinc finger (C3HC4-type RING finger) family proteinAT2G350002.491.76
 Metal homeostasisPhytochelatin synthetase-relatedAT1G097905.581.20

A second important point concerns a few elements that have been previously associated with the plant response to Pi starvation. Notably, we found that three SPX genes (SPX1, SPX2 and SPX3) modulated by Pi deficiency (Duan et al., 2008) are also systemically regulated (class IS1b; Tables 2 and S1). These genes appear to be part of the Pi-signaling pathway that involves SIZ1/PHR1 (Duan et al., 2008), and share SPX domains with proteins previously identified in yeast as important elements of Pi sensing (Lenburg and O’Shea, 1996). PHR1, an important component of the plant response to Pi starvation, is not present in the list of genes modulated in our split-root experiment. This observation confirms previous studies showing that the expression of this gene is not regulated at the transcriptional level (Rubio et al., 2001), and may require additional components such as SIZ1 (Miura et al., 2005). Other crucial elements involved in the control of Pi homeostasis are members of the PHO1 family (Wang et al., 2004). Surprisingly, three members identified by this study exhibit very different behaviors. They can be found associated with IS1a (PHO1-H1), IL (PHO1-H7) and RS1b (PHO1-H2) classes (Table S1), whereas the expression of PHO1 (At3g23430) remains unchanged in our experiment. Finally, PHO2, which has been described as participating in the systemic response (Aung et al., 2006; Bari et al., 2006), was only slightly repressed in our experiment (but not significantly).

The third hallmark of the systemic regulation is the strong downregulation of genes involved in metal uptake or transport in the spP500 compartment (compared with cP500). These genes, downregulated in low Pi, are classified mainly in the RS1a and b classes (32% of the known genes; Figure 2b; Tables 2 and S1). They are mostly involved in the homeostasis of metals such as iron and copper, zinc or cadmium (IRT2, IRT1, NAS1, NAS2, ferric chelate reductase, nicotianamide synthase, etc.). Additional genes in the metal uptake/transport category can be found in the RS1c and RS2 groups, whereas a few genes involved in metal storage (Ferritin 1) and detoxification (FRD3, MATE efflux proteins, heavy metal-associated proteins) are induced in low Pi (cP0), and are locally regulated (IL class; Tables 3 and S1).

Table 3.   Selection of locally-regulated genes with specific functions as described in Figure 2(b)
ClassFC > 105 < FC < 102 < FC < 5
  1. Genes are classified according to their level of induction or repression (FC > 10, 5 < FC < 10 or 2 < FC < 5; FC, fold change). For induced genes, FC = cP0/cP500; for repressed genes, FC = cP500/cP0.

 Hormone related
  Ethylene response factors At3G23240At1G28370
  Ethylene-responsive element binding factor  At5G47220
  Cytochrome P450  At4G39950
  Anthranilate synthase component I-1 precursor  At5G05730
  Auxin-responsive family protein  At5G35735
  Unknown (response to ethylene)  At5G64900
 Metal related
  Metal storage FER1  At5g01600
  ABC transporter – like protein At3G55090 
  Heavy-metal-associated domain At5G52750, At5G52670At5G52720, At1G55780, At2G35730
  Magnesium transporter CorA-like  At5G09710
  MATE efflux family protein (FRD3 and others) At1G71140At5G44050, At1G33110, At3G08040
  Matrix metalloproteinase  At1G70170
  Putative metalloproteinase  At1G24140
 Stress related
  Anionic peroxidase At1G14540At1G14550
  Cytochrome P450 (and -like and putative)At4G31940, At2G30770, At2G30750At3G26830At4G22710, At3G14620, At5G58860, At3G26210, At5G23190, At3G14650, At4G37310, At1G55940, At2G24180
  Disease resistance (and -like and putative)  At2G34930, At1G72940, At5G40170, At1G33590
  Endochitinase-like protein  At3G47540
  Glutathione-S-transferase (and putative) At1G69920, At2G29470, At3G09270, At2G29460At1G74590, At1G69930, At1G02920, At4G02520, At2G29440
  Harpin-induced protein-like  At5G06320
  Osmotin-like protein (OSM34)  At4G11650
  Oxidoreductase family protein  At5G38900
  Peroxidases (and putative)At5G06730, At5G19880, At3G49120 At4G36430, At1G71695, At1G68850, At2G18140, At4G08770, At5G14130, At5G64110
  Protease inhibitor (ATDR4)  At1G73330
  Putative lipoxygenase  At1G72520
  Secretory protein, putative At2G15220 
  Thaumatin-like protein (PR-5-like)  At1G75030
  Trypsin inhibitor propeptide  At1G73260
  Unknown (response to JA)  At1G17380
  Unknown (response to stress)  At5G13220, At5G20270
 Transcription factors
  AP2 domain-containing  At5G13330
  F-box family protein  At5G52120
  Homeobox-leucine zipper family protein  At5G66700
  Homeodomain (ATHB-7)  At2G46680
  MADS-box protein  At1G17310
  Myb family  At3G23250, At1G74080, At1G17950, At5G49620
  Myb-related  At1G48000
  NAM (and related) At1G52890At5G39610, At3G04070, At1G02220, At2G33480
  WRKY family  At2G46400, At3G01970, At3G56400, At2G40750, At5G13080, At1G29860, At2G46130, At5G49520
  Zinc finger (C2H2 type) family protein  At1G55110
 Growth, development
  Actin 11 (ACT11)  At3G12110
  Acyl CoA reductase  At4G33790
  Armadillo  At3G54870
  Cytokinin-related  At3g63110, At1g13420
  Dynein light chain type-1 family  At1G52250
  Expansin At1G62980, At1G12560 
  Hydroxyproline-rich glycoprotein family At5G19800 
  Kinesin motor protein-related  At1G09170, At3G49650
  Leucine-rich repeat family
  Protein/extensin family protein (LRX1)
  Pectinesterase family protein At5G04960At3G10710, At5G53370
  Proline-rich extensin-like family protein At5G35190, At5G49080/At2G24980/At5G06630, At3G54580, At3G54590At5G06640, At5G49080/At4G08400/At4G08410, At5G49080/At4G08410/At5G06630/At2G24980, At1G23720, At3G28550
  Proline-rich family protein At3G62680 
  Proline-rich family protein contains
  Proline-rich extensin domains
  Putative actin-depolymerizing factor  At4G00680
  Putative growth regulator protein  At4G38390
  Putative leucine-rich repeat
  Transmembrane protein kinase
  Putative SEC14 cytosolic factor At4G34580 
  Similarity to NADPH-cytochrome P450 reductase  At1G13590
  Universal stress protein (USP) family protein At2G03720 
  Unk (developmental process) At4G40090 
  Xyloglucan:xyloglucosyl transferase At5G57530/At5G57540, At4G25820At4G28850, At1G10550, At2G06850

Among the genes induced and locally regulated by Pi starvation is an over-representation of genes associated with stress- and hormone-related responses (Table 3), including biosynthesis and response to ethylene, stress-related markers (P450, defense response genes) and oxidative processes (peroxidase, oxidoreductase, GST). In this category of genes, we revealed a high level of induction for some of the genes coding for cytochrome P450 and peroxidases (fold change, FC > 10; Table 3). Specific transcription factors linked to development (Myb, MADS) or stress (WRKY) are abundant in the IL class (Tables 3 and S1). In addition to this, in the Pi-rich medium, we found (RL class; Tables 3 and S1B) an induction of an important set of genes involved in cell wall synthesis (extensins, hydrolase, expansin, etc.) or cell activity and growth (microtubule organization, secretion, etc.), and genes involved in cytokinin production (IPT3) or response to cytokinin accumulation (AT1G13420).

Role of Pi internalization in systemic or local regulation

Local regulation.  In our split-root experiments, root growth was promoted in the high-Pi compartment, whereas roots displayed a typical Pi-starved phenotype in the low-Pi compartment. The age of the plant had no impact on the observed phenotype: young plantlets (2 weeks old; Figure 1b) grown in vitro on agar as well as more mature plants (2 months old) grown in hydroponic conditions (Figure 1a) showed the same phenotype. Importantly, the split-root experiments are presented here after 7 days of growth to clearly distinguish these phenotypes, although root growth modification (meristem growth arrest) was observed within 2 days of growth.

Nevertheless, we cannot rule out the possibility that the root architecture responds to a local internal concentration of Pi, as there is still internal phosphate depletion in roots growing in spP0 conditions (Figure 1a). To address the specific role of internal versus external Pi, we produced plants in which the internal and external levels of phosphate were completely disconnected. A new experimental feeding protocol was developed to enhance Pi accumulation in roots, irregardless of the Pi concentration in their growth medium, by feeding leaves a high-Pi solution. We found that an increase in Pi concentration measured in the root was proportional to the level of Pi supplied on leaves, and could even reach levels usually observed in plants grown in a Pi-rich medium (Figures 3 and 4a). This is effectively demonstrated with 33P-labeled Pi, which is able to reach the entire root tissue, including the root tip (Figure S2a). Interestingly, feeding leaves with a high-Pi solution does not alleviate root growth arrest for plants grown on low-Pi media (Figures 3 and 4). Thus, despite the high accumulation of Pi in roots, they still exhibit a Pi-starved phenotype characterized by a rapid arrest of root growth (Figures 3 and 4). By contrast, the long-root phenotype associated with a Pi-rich medium does not require the presence of Pi inside the roots. The phf1 mutant (Gonzalez et al., 2005), which affects Pi uptake, displays a long root system in Pi-rich medium, despite an internal Pi level identical to that in starved plants (Figure 3a,b). A similar result is obtained with the lpr1 lpr2 double mutant, which is altered in two multicopper oxidases required for the root response to phosphate starvation (Svistoonoff et al., 2007). In low Pi conditions, these plants exhibit a long root phenotype despite a low Pi content (Figure 3a,b). Previous reports have drawn a direct link between the control of root growth and the local Pi concentration present in the culture medium (Drew, 1975; Linkohr et al., 2002; Svistoonoff et al., 2007), but could not rule out the possible influence of internal Pi concentration (Williamson et al., 2001). Our experiments described above not only confirm the role of external Pi level on root architecture, but also clearly demonstrate that this trait is disconnected from the internal Pi content.

Figure 3.

 Local and systemic regulations of the phenotypic and molecular modifications promoted by phosphate starvation in Arabidopsis. Root development (a), root inorganic phosphate (Pi) content modulation (b) and quantitative PCR results for locally-regulated genes (c) on a high (+P) or low (−P) level of Pi in wild-type (WT) plants or mutants (phf1 and lpr1 lpr2 double mutant).
(a) WT plants are grown in one of four conditions: high Pi (+P), low Pi (−P), low Pi supplied with Pi on the leaf (100 mm Pi), or low Pi and low Fe (−P−Fe). The phf1 mutant was grown in +P, and the lpr1 lpr2 double mutant was grown in +P or −P. Scale bar: 1 cm.
(b, c) Mean Pi content ± SD and mean expression level (relative) ± SD are calculated from triplicates of three independent experiments.

Figure 4.

 Local and systemic regulation of the phenotypic and molecular modifications promoted by phosphate starvation in Arabidopsis. Effect of the modulation of inorganic phosphate Pi content in the root: (a) root phenotype of the plants supplied with Pi on the leaves; intracellular Pi content in the roots and quantitative PCR results for systemically regulated genes; (b) expression modulation of downregulated genes under low Pi conditions.
Plants were grown in high Pi (+P) or low Pi (−P) conditions, and were supplied with Pi on the leaves (1, 10 or 100 mm Pi). Mean Pi content ± SD and mean expression level (relative) ± SD are calculated from triplicates of three independent experiments. Scale bar: 1 cm.

To investigate the effect of internal versus external Pi on transcript levels, we analysed the expression of genes that are locally regulated by Pi starvation, but not directly related to developmental processes. Here, we selected several genes in the IL and RL classes related to transport or metabolism. In addition to the genetic tools (phf1, lpr1 lpr2) or physiological experiments (Pi feeding on leaves) described above to manipulate the Pi status of plants, we used plants grown in a low Pi and low iron medium (Figure 3). Under these conditions Arabidopsis plants exhibited a long-root phenotype, despite the absence of phosphate, as the presence of iron is required for root growth limitation on a low Pi medium (Svistoonoff et al., 2007). According to our analysis by quantitative PCR (Real-time qPCR; Figure 3c) and reporter genes (Figure S3), the expression of all selected genes is fully independent of Pi internal status (Figure 3b), whereas gene expression is connected directly to the development of the roots. Together, these results suggest that local regulation of gene expression is modulated mostly by root development, and is independent of internal Pi content.

Systemic regulation.  As expected, all genes induced by Pi starvation and associated with the systemic response to Pi deficiency were also downregulated by the application of Pi to the leaves (Figure 4a). We analysed genes from the various IS1 classes, with the addition of IPS1 (Burleigh and Harrison, 1997, 1999; Rubio et al., 2001), an important element of the Pi regulation process not present in the Affymetrix chip. The internal root Pi content is modulated by the level of Pi supplied to the leaves, whereas the genes react strongly even with the lowest application (1 mm), the root Pi content is only partially complemented (Figure 4a). Nevertheless, some discrepancies can still be observed between the sensitivity of the various markers in response to the treatment. For example, PHO1-H1 exhibits a slightly atypical response. It reacts to Pi treatment, but never reaches the basal level observed in high-Pi medium (Figure 4a). Furthermore, we tested genes downregulated by Pi-starvation (RS classes), and observed no reversion of the expression (Figure 4b), even with the highest level of Pi supplied on the leaves (100 mm), suggesting that the effect of Pi is indirect in this case.

Together, these results confirm that Pi internalization is required to promote the activation of the systemic response of IS1 genes. The question of whether or not Pi is directly involved in the triggering of the systemic signal in the plant has been addressed by using a PHO1-underexpressor line named B1 (Rouached, unpublished data). Like the pho1 loss-of-function mutant (Poirier et al., 1991; Hamburger et al., 2002), the B1 line is impaired in its loading of Pi into the xylem. This leads to Pi accumulation in the roots, whereas the leaves are depleted of Pi (Figure 5a). Fortuitously, the development of B1 plants is similar to that of the wild type (WT), which in practice allowed us to perform a split-root experiment that would be difficult to set up with the extremely dwarf pho1 mutant. In spite of this strong alteration of Pi translocation and accumulation between the P500 and the P0 halves of the root, all the IS1 markers assayed in the B1 line behave identically to the WT control (Figure 5b). This result indicates that the systemic signal that propagates through the plant is mostly independent of the Pi concentration in the leaves.

Figure 5.

 Split-root analysis of a PHO1-underexpressor line (B1): inorganic phosphate (Pi) content (a) and expression of systemically regulated genes (IS1 class) in roots (b) according to treatment: cP500 (white bars), cP0 (black bars), spP500 (light-gray bars), spP0 (dark-gray bars). Pi content and transcript level are the means (±SD) of three replicates, each one having been analyzed three times.
(a) Pi content in rosettes is measured in the controls cP500 (white bars), cP0 (black bars) and in the ‘split experiment’ (black to white gradation).
(b) Real-time quantitative PCR values are normalized to the control in low Pi conditions.

Recent work has established that the ubiquitin-conjugating E2 enzyme (UBC24) PHO2 is involved in systemic Pi signaling (Poirier et al., 1991; Aung et al., 2006; Bari et al., 2006). Using the pho2 mutant (Delhaize and Randall, 1995), we analysed the expression of a set of systemically regulated genes (class IS1; Figure 6a) in a split-root experiment, which revealed either a complete (MGDG3, SPX3, SQD2, PAP22) or partial (PHT1;4, PHO1-H1) deregulation. In the pho2 mutant, the expression level was similar in the control (cP0) and the split-root sample (spP0). These results confirm the disruption of long-distance signaling, even if in the P0 split roots (spP0) the level of free Pi in the pho2 mutant and the WT is similar (Figure 6b).

Figure 6.

 Split-root analysis in the pho2 mutant of the expression of systemically regulated (IS1 class) genes (a) and inorganic phosphate (Pi) content (b) in roots, according to treatment (see captions in Figure 5).
Real-time quantitative PCR values are normalized to the control in low Pi conditions. Pi content and transcript level are the means (±SD) of three replicates, each one having been analyzed three times.


This study, based on a split-root experimental design, clearly highlights the occurrence of different patterns of regulation. It shows that systemic regulation is most probably mediated through the plant by signals other than Pi, and that part of local regulation is primarily determined by changes in root growth mediated by external Pi. Our full-genome transcriptomic analysis gives further insight into the distinct and specific behavior of genes that belong to multigenic families, in addition to the impact of Pi on metal homeostasis and root development.

Systemic versus local regulation

Previous studies analysed a limited number of targets, which offered a very restricted view of the transcriptomic modifications that are systemically or locally regulated by Pi starvation. This work now provides a global view of this process, and reveals new targets associated with either systemic (such as lipid metabolism) or local regulation (stress and hormone-related). The use of ATH1 Affymetrix chips offers access to an extensive number of transcripts (22 810 probes), and highlights that most of the genes related to Pi homeostasis (transporters, Pi recycling and signaling) are induced and systemically regulated by low levels of Pi (IS1 class). In particular, we found that genes involved in phospholipid remobilization and synthesis of galactolipids and sulfolipids are systemically regulated, revealing the capacity of the plant to respond to its global Pi status (Nussaume et al., 2011).

Moreover, this analysis addresses Pi transduction pathways, allowing us to differentiate the response profile depending on factors involved in the regulation of this process (as illustrated here by PHR1 and PHL1). Nevertheless, it should be noted that PHR1 and PHL1 are not modulated at the transcriptional level by Pi starvation, suggesting a supplemental level of regulation between the transcription factors and the target genes. SIZ1 has already been identified and is implicated in this process. It is an ubiquitin-like modifier (SUMO) E3 ligase, described as modulating PHR1 expression (Miura et al., 2005). But, like PHR1 and PHL1, SIZ1 does not appear to be regulated at the transcriptional level, and therefore could not have been identified in the present study.

Furthermore, the use of DNA chips allowed us to distinguish between gene members among multigenic families (such as phosphatases, transporters, transcription factors, etc.). Despite similar biological activities, these members behave differently according to the established classes (locally, systemically, induced or repressed). This draws focus to a new level of complexity, which has been previously overlooked in the study of Pi starvation. It also highlights a pitfall of using biochemical tests to assay gene expression (such as phophatase assay), which can only reveal global activity and cannot distinguish between the members of the same family acting either locally or systemically.

The deregulation of IS1 genes observed in the pho2 and phr1 phl1 mutants reinforces the central role for the PHR1, PHL1, PHO2 proteins (Rubio et al., 2001; Lin et al., 2008) in long-distance signaling. Although internal Pi triggers this systemic regulation, it suggests that a molecule other than Pi is involved in mediating systemic signaling. The PHO2 regulator, mir399, moves through the phloem from the shoot to the roots, making it an optimal candidate for this long-distance signal (Aung et al., 2006; Bari et al., 2006). However, the nature of the ascending signal from root to shoot remains unknown.

Pi and metal crosstalk

A strong impact of Pi concentration on metal availability is clearly illustrated by the behavior of metal transporters (RS1 class) and genes involved in detoxification (IL class). Genes involved in the uptake of iron (IRT1 and IRT2) and other metals (Zn, Cu) are repressed in low levels of Pi, and are systemically deregulated in high levels of Pi (where expression in spP500 is lower than in cP500). This reveals that the plant responds to metal availability in both the whole plant (systemic response) and in the medium (local response). The metal bioavailability in the medium is increased in low Pi as a result of the reduced interaction with Pi in Pi-deprived medium. This is in agreement with the higher accumulation of metals in plants (e.g. +400% of iron) observed in low Pi (Hirsch et al., 2006), and could be associated with a reduced interaction with Pi in a Pi-deprived medium. Nevertheless, the systemic regulation of some genes related to metal uptake or homeostasis suggests a response to the global metal status of the plant. Importantly, reducing Pi concentration in the medium could reduce the pH (Narang et al., 2000), and consequently impact metal solubility in soils. To circumvent this possible effect of Pi starvation, the culture medium was buffered at pH 5.8 in the present experiment. In any case, we cannot rule out the occurrence of direct crosstalk between phosphate and various metal signaling pathways, as reported for the growth of root hairs (Schmidt and Schikora, 2001).

Pi starvation and root development

Locally regulated genes are mostly controlled by external Pi and are correlated with root growth, indicating that the primary effect of external Pi is the alteration of root development. LPR1 (Svistoonoff et al., 2007) and PDR2 (Ticconi et al., 2004) are involved in the control of the root meristem activity when Pi is limiting, but the function of these genes remains unknown (Desnos, 2008). This view is in agreement with previous data that indicate that cell proliferation controls the magnitude of the Pi starvation response in Arabidopsis (Lai et al., 2007). Many other traits of Pi starvation that affect root development have been described, and these affect cell size and root hair development (Bates and Lynch, 1996; Reymond et al., 2006). In our experiments, these traits were clearly established after 5 or 7 days, but are not evident at 2 days of growth. A prominent role for ethylene in the control of root growth and cell division has been reported previously (Ma et al., 2003; Zhang et al., 2003). Indeed, many enzymes of ethylene biosynthesis and responses are clearly identified here as locally induced by Pi starvation (ACC synthase, ACC oxidase, ERF1, ERF2 and various ethylene-responsive element binding factors). This suggests that the existence of a crosstalk between phosphate and ethylene involves a step of transcriptional control. A few elements involving auxin are also identified, and are probably related to the major role of this hormone in driving root architecture.

All together, these results provide new insight into Pi starvation effects and Pi signaling. They also offer a rational way to dissect the various complex transduction pathways triggered by such an abiotic stress.

Experimental procedures

Plant material

For global transcriptomic analysis, Arabidopsis plants were grown in sand and supplied with full nutrients for 6–7 weeks in a growth chamber in controlled conditions (21/18°C day/night temperature, 60% humidity, 8-h photoperiod). For the split-root experiments (Gansel et al., 2001), the root system was washed in water to remove sand, separated into two parts and then immersed in two containers (300 ml each), containing either the same medium [MS/10 at pH 5.8 (Misson et al., 2004) supplied with 500 μm Pi (cP500) or without Pi (cP0)] or different media (500 μm Pi and no Pi; spP500 and spP0, respectively), for 2 days. Three plants were pooled for each condition and the experiment was performed in triplicate. Among different lines expressing various levels of PHO1 (Stefanovic et al., 2007), the B1 line exhibits 20% of PHO1 transcript level and a low Pi level in the leaves, despite a growth rate similar to the wild type.

For root development observations, plants were grown hydroponically for 7 days as described above. In vitro plants were grown on agar-containing MS/10 medium for 7 days (Misson et al., 2004), followed by transfer on compartmentalized plates for 5 days before root development or GUS and GFP observations.

Leaf Pi-supply experiments

To produce plants that combine a high internal Pi content with growth on low Pi medium, seedlings were plated for 5 days on MS/10 agar containing 500 μm Pi. Plants were then transferred to new agar plates containing 500 μm Pi (+P), 5 μm Pi (−P) or 5 μm Pi and 2 μm Fe (−P−Fe). For leaf Pi supply, partitioned Petri dishes were used. The lower part was fully filled whereas the upper part was only half-filled with −P medium and covered with a slice of Whatman paper. Plants were transferred in such a way that the roots were placed on the lower part and leaves in the upper compartment. For 7 days, plants received the following daily treatment: plates were placed in a horizontal position and a solution containing various concentrations of Pi (0, 1, 10 and 100 mm) was applied to the leaves (2–100 μl per plantlet). At 3 h after treatment, the remaining drops on the leaves were carefully removed and the plates were replaced in their vertical position in the culture chamber. Root length was measured using ImageJ software.

Pi content and Pi downloading

Measurement of the free cellular Pi content was performed as previously described (Misson et al., 2004). To visualize 33P downloading, the seedlings were placed on partitioned Petri dishes filled with −P medium and a solution containing 1 mm Pi, and 1.5 μCi ml−1 33PO4 was applied to the leaves for 24 h. The roots were then exposed against a Kodak storage phosphor screen for 24 h before scanning (storm 840; Molecular Dynamics; http://www.gelifesciences.com) (Misson et al., 2004).

GUS and GFP reporter gene analysis

The 2628-bp sequence upstream of the PHT1;4 start codon was PCR amplified using oligonucleotides (forward, 5′—CACCAATGGGAGAAACAAGAGTAGCCAC-3′, and reverse, 5′-CTGCCTATGAACAAGGACAGAGTTG-3′), and cloned into pENTR vector using a pENTR directional TOPO cloning kit (Invitrogen, http://www.invitrogen.com). The fragment was transfered into a gateway vector pBGWFS7 (Karimi et al., 2002) to create a transcriptional fusion with GFP and GUS reporter genes. Plant transformants were generated by vacuum infiltration of inflorescence with Agrobacterium, as previously described (Bechtold et al., 1993).

Gene expression in transgenic lines containing promoter-GUS [STP13, (Schofield et al., 2009), RNS1 (Hillwig et al., 2008), CHX17 (Cellier et al., 2004), ACP5 (del Pozo et al., 1999), IRT1 (Vert et al., 2003) and PLDz2 (Cruz-Ramirez et al., 2006)] or GFP [SUC2 (Imlau et al., 1999) and PHT1;4] fusions to selected genes was examined as described by Misson et al. (2004) and Ticconi et al. (2009).

RNA extraction and gene expression analysis

Total RNA from roots was extracted in phenol-chloroform (Nussaume et al., 1991), and purified with Dnase (DNAfree; Ambion, http://www.ambion.com). cRNA was prepared for hybridization on ATH1 affymetrix chips (Affymetrix platform; Curie Institut, Paris, France) according to the manufacturer’s protocol (‘one-cycle target labelling and control reagents’). Results were analyzed with gcos (the Affymetrix Gene Chip Operating System) and Genespring v7.3 (Agilent, http://www.agilent.com). The data are deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress, accession number E-MEXP-2601).

qPCR analyses were performed after retro-transcription (kit from GE Healthcare, http://www.gehealthcare.com) and amplification (primer efficiency factors were measured for each gene and GapC was used as a reference gene; ABI7000; Applied Biosystems, http://www.appliedbiosystems.com). Primer sequences are given in Table S3.


Array hybridizations were supported in part by RNG (Réseau national des génopoles, Evry, France), and we thank Dr B. Albaud for performing hybridizations (Affymetrix platform; Curie Institut, Paris, France). We thank Dr W. Scheible (Potsdam, Germany) and Dr T. Desnos (Cadarache, France) for seeds of the pho2 mutant and lpr1 lpr2 double mutant. We acknowledge Dr N. Sauer (Erlangen, Germany), Dr S.J. Rothstein (Ontario, Canada), Dr G.C. Macintosh (Ames, IA, USA), F. Casse and C. Curie (Montpellier, France), L. Herrera-Estrella (Guanajuato, Mexico) and V. Rubio (Madrid, Spain) for providing promoter fusion lines. We are grateful to Brandon Loveall for his advice on English language. This work was supported partly by the Commissariat à l’Energie Atomique (CEA, France) and ANR plant KBBE (FOSSI), and by grants from CEA (VB and J-FA) and Région Provence-Alpes-Cote d’Azur (VB).