Identification of QTL controlling root growth response to phosphate starvation in Arabidopsis thaliana



    1. CEA Cadarache, DSV DEVM Laboratoire de Biologie du Développement des Plantes, UMR 6191 CNRS-CEA, Aix-Marseille II, F-13108, France and
    Search for more papers by this author
    • *

      Present address: Max Planck institut for plant breeding research, Carl-von-Linné Weg 10, 50829 Köln, Germany.


    1. CEA Cadarache, DSV DEVM Laboratoire de Biologie du Développement des Plantes, UMR 6191 CNRS-CEA, Aix-Marseille II, F-13108, France and
    Search for more papers by this author

    1. Station de Génétique et d’Amélioration des Plantes, INRA Versailles, route de St Cyr, 78026 Versailles, France
    Search for more papers by this author

    1. CEA Cadarache, DSV DEVM Laboratoire de Biologie du Développement des Plantes, UMR 6191 CNRS-CEA, Aix-Marseille II, F-13108, France and
    Search for more papers by this author

    Corresponding author
    1. CEA Cadarache, DSV DEVM Laboratoire de Biologie du Développement des Plantes, UMR 6191 CNRS-CEA, Aix-Marseille II, F-13108, France and
    Search for more papers by this author

Thierry Desnos. Fax: +33 4 42 25 45 97; e-mail:


One of the responses of plants to low sources of external phosphorus (P) is to modify root architecture. In Arabidopsis thaliana plantlets grown on low P, the primary root length (PRL) is reduced whereas lateral root growth is promoted. By using the Bay-0 × Shahdara recombinant inbred line (RIL) population, we have mapped three quantitative trait loci (QTL) involved in the root growth response to low P. The Shahdara alleles at these three QTL promote the response of the primary root to low P (i.e. root length reduction). One of these QTL, LPR1, located in a 2.8 Mb region at the top of chromosome 1, explains 52% of the variance of the PRL. We also detected a single QTL associated with primary root cell elongation in response to low P which colocalizes with LPR1. LPR1 does not seem to be involved in other typical P-starvation responses such as growth and density of root hairs, excretion of acid phosphatases, anthocyanin accumulation or the transcriptional induction of the P transporter Pht1;4. LPR1 might highlight new aspects of root growth that are revealed specifically under low P conditions.


Plant growth habit and physiology are profoundly altered by low availability of environmental phosphorus (P). The main features of P starvation syndrome include an increase in the root/shoot ratio; a modification of the root architecture to explore P-richer soil horizons; an increase in root hair length and density; the induction of P transporters; a higher level of root secretion of organic acids, phosphatases and nucleases and the accumulation of anthocyanins etc. (Raghothama 1999). Some of these responses are part of an adaptive strategy of plants for better acquisition of the different forms of P in the rhizosphere and an efficient utilization of the P absorbed.

The alteration of root system architecture in response to P starvation is a long-known phenomenon. In some species such as Lupinus albus and other Proteacea, rather specialized roots are generated. They emerge along the main roots in bottlebrush-like clusters of lateral rootlets, called ‘proteoid’ roots, and exude organic acids that serve in the mobilization of the insoluble P chelated with divalent or trivalent ions (Johnson, Vance & Allan 1996). This example represents an extreme specific response to very poor soils, and the general trend is a growth stimulation of the whole root system to the detriment of the shoot. The elongation of roots reaching P-rich localized volumes of soil is stimulated (Drew 1975).

P-starved Arabidopsis exhibit a short primary root growth concomitantly associated with a stimulation of lateral root initiation and growth (Williamson et al. 2001; Linkohr et al. 2002). One can question the role of growth hormones in this root architecture response. It seems, however, that it is largely independent of auxin, ethylene (Poirier & Bucher 2002; Ma et al. 2003) and abscisic acid (ABA) (Trull et al. 1997).

The genetics of the low P-induced architectural modification are not well developed, and until now no mutant specifically altered in this response has been isolated. As for the many traits that have now been analysed in Arabidopsis, the observation of accessions of worldwide origin has revealed the natural variation of responses to P starvation. With the purpose of analysing the P acquisition efficiency of 36 accessions, Narang, Bruene & Altmann (2000) selected five accessions with contrasting responses. They noted that these lines also displayed different shoot/root mass ratios, root hair lengths and densities depending on the nature (solubility) of the P in the nutrient solution. In a work more focused on the root architecture related to P nutrition, Chevalier et al. (2003) reported the diversity of Arabidopsis accessions. With two growth traits – primary root length (PRL) and number of lateral roots – they categorized 73 accessions. Compared with a P-rich medium, the P-poor medium reduced both the PRL and the number of lateral roots for 50% of the accessions. On the other hand, 25% were not affected while the remaining lines were affected in only one trait. These data may reflect different physiological strategies that plants exhibit under P starvation. These works also show that accessions are a rich source of genetic diversity to analyse P-starvation responses. Overall, these observations suggest that a deeper understanding of the root architecture of Arabidopsis in response to low exogenous P should result from the power of quantitative trait loci (QTL) analysis.


Plant material

Accessions of A. thaliana used were: Bay-0 (Bayreuth, N57923), Columbia (Col-0 CS60000), Cape Verde Islands (Cvi, N902), Shahdara (N57924) and Wassilevskaja (WS, N5390). Progenies of seeds were provided by the NASC ( Landsberg erecta was from Lehle (Round Rock, TX, USA).

For QTL analysis, we used the Core-Pop set of 165 RILs (N57921) selected from the original 411 RIL population as an optimal set for QTL mapping experiments (Loudet et al. 2002; see details).

For the confirmation and fine analysis of LPR1, we generated near-isogenic lines (NILs) by segregating F6 RIL (194B and 194S). These are called heterogeneous inbred family (HIF) lines (Tuinstra, Ejeta & Goldsbrough 1997). LPR1 maps between molecular markers MSAT1.10 and NGA248 at the top of chromosome 1. We examined the whole collection of 411 Bay-0 × Shahdara RILs and we found that the F6 generation of RIL194 contains a residual heterozygosity at markers MSAT1.10 and NGA248 only (Loudet et al. 2002). From the progeny of this RIL, we selected the HIF couple (fixed for the Bay-0 194B allele or the Shahdara 194S allele) by genotyping individuals with markers MSAT1.10 and NGA248.

For the dissection of epistasis between LPR1, LPR2 and LPR3, RILs that carry a recombinant event between the markers flanking one of these QTL were withdrawn because the exact QTL genotype could not be asserted.

Plant growth and measurements

The in vitro nutrient solution (MS) was prepared based on the Murashige & Skoog (1962) protocol; it was diluted 10 times with 8 g L−1 agar (Sigma, catalogue number A-7002; Saint Louis, MO, USA) and 1% sucrose. For the high P condition, 500 µm NaH2PO4 was added to the medium, while for the low P condition, 5 µm NaH2PO4 and 495 µm NaCl were added to the medium. The pH was buffered at 5.6–5.8 with 3.5 mm of MES. Forty millilitres of the autoclaved medium were poured into 12 × 12 cm Petri dishes. For the iron (Fe) starvation experiments, no FeSO4 was added to the media while 300 µm FerroZine [3-(2-pyridyl)-5,6diphenyl-1,2,4-triazine sulphonate; Sigma] was added to chelate any Fe as previously described (Yi & Guerinot 1996). For sulphur (S) starvation, MgSO4 was replaced by MgCl2. Media deprived of nitrogen (N) contained 2.07 mm NaCl and 1.88 mm KCl instead of 2.07 mm NH4NO3 and 1.88 mm KNO3.

The seeds were surface sterilized for 5 min in a solution of ethanol (90%)-Bayrochlore™ (Bayrol Mundolsheim, France) (10%), and was rinsed twice with ethanol (100%). Five seeds of each line were placed on each plate. The plates were stored at 4 °C overnight and then vertically disposed in a growing chamber for 10 d (16 h photoperiod; 24.5/21 °C day/night, respectively). For QTL mapping, the plates (two per line) were randomly moved every 2 d within the same shelf of the growth chamber.

Ten-day-old plantlets were scanned (Agfa Snap Scan 1236s, Ridgefield Park, NJ, USA) and the roots were measured with Bioscan Optimas V6.10 image analyser (Edmonds, WA, USA). This software allowed us to measure the PRL (in 5 or 500 µm P) and the total length of the root system (in 500 µm only) for each plant. We were also able to count the number of visible (emerged) lateral roots for each plant (in 500 µm only).

The length of the primary root epidermis cells was measured only in experiment 1 (Exp1). A 1–2 cm piece of primary root from a 10-day-old plantlet was mounted between a lame and lamella and sealed with nail varnish (Gemey, Paris, France). Cell length was measured under the microscope (Leica-Leitz DM RXA; Wetzlar, Germany) by using a micrometer. Ten cells, located above the elongation zone, were measured per root for each RIL. Statistical analysis was then performed by using SPSS (SPSS 11.0.1), general linear model, univariate. The response of PRL and primary root cell length to P starvation was estimated with the following formula:

Response = (trait in high P − trait in low P)/trait in high P

QTL mapping and analysis

For the parental lines and the 165 RILs, the root traits of five individual plantlets were measured. The growth traits measured were PRL and the number of lateral roots for both the low and the high P plantlets, and the total length of lateral roots (ΣLRl) for the high P plantlets only. We repeated the QTL experiment in the same growth chamber on two different dates.

Broad sense heritability (h2b) was estimated for each trait by using the following equation:

h2b = VG/(VG + VE)

where VG is the variance between RILs and VE is the variance within RIL.

QTL detections were performed by using QTL Cartographer (Basten, Weir & Zeng 1995). Briefly, QTL were detected successively by interval mapping (IM) and composite interval mapping (CIM). IM (Lander & Botstein 1989) was used first to determine the putative QTL involved in the variation of the considered trait. CIM model 6 of QTL Cartographer was then performed. The closest marker to each local logarithim-of-odds (LOD) score peak was considered as a cofactor while testing the presence of a novel QTL elsewhere in the genome. The number of cofactors in our models varied from three to six, depending on the trait. The same cofactors were used for identical traits measured in the two QTL experiments. When a cofactor was also a flanking marker of the tested region, the cofactor was excluded from the model. The walking speed was 0.5 cM. As suggested by Churchill & Doerge (1994), the LOD significant threshold was estimated from several test permutation analyses. One thousand permutations of phenotypic data were analysed by using the CIM model for each trait as described earlier. On average, the experiment-wise threshold obtained (α = 5%) for all the traits that were studied corresponds to a 2.5 LOD.

The additive effects of the detected QTL were estimated from CIM results. Additive effect represents the mean effect of the replacement of the Shahdara allele by the Bay-0 allele at a particular locus. The contribution of each detected QTL to the total variance (R2) was estimated by variance component analyses. Significant QTL–QTL interactions (epistatic effect) were also added to the genetic model via the corresponding marker interactions.

Analysis of low P response

In order to quantify the anthocyanin content, the plants were grown for 20 d on MS media with 5 or 500 µm P. Anthocyanins were extracted from 20 mg of pooled rosette leaves of either HIF 194B or 194S, and five independent extractions were performed for each growth condition. Anthocyanins were extracted and quantified as described by Ticconi, Delatorre & Abel (2001).

For in vivo visualization of acidic phosphatase activity, 10-day-old plants grown on MS media with 5 or 500 µm P were transferred to MS media containing 0.008% BCIP (Trull & Deikman 1998) and 5 or 500 µm P, respectively, and were replaced in normal growth conditions for 48 h. The plants were then observed and photographed with a stereomicroscope MZ16A (Leica).

For reverse transcription-polymerase chain reaction (RT-PCR) analysis of gene expression, RNA was extracted from roots of 10-day-old plants grown on an MS medium with either 5 or 500 µm P. RNA-Easy plant extraction kit (Qiagen, Valencia, CA, USA) was used according to the manufacturer's instructions. Six micrograms of total RNA were treated with DNase1 (Roche Diagnostics, Indianapolis, IN, USA) for 15 min at 37 °C and were used for the retro-transcription reaction using the AMV reverse transcriptase (Roche Diagnostics) according to the manufacturer's instructions. Aliquots of each preparation were used as templates after dilution for the PCR. The first strand cDNA amount was adjusted in order to have semi-quantitative responses. PCR was performed with specific oligonucleotides for Pht1; 4 (F: CGATCATTCCACTTC CTTCTTCTT, R: CCACGTGGTAATAAATACGACCG. Thermal cycling consisted of an initial denaturation at 94 °C for 5 min, followed by 25 cycles of denaturation step at 94 °C for 30 s, annealing at 60 °C for 30 s and extension at 72 °C for 1 min. A final extension at 72 °C was performed for 7 min. Normalization of samples was performed by using the actin gene At3g18780 with primers ACT2F: CTTCCCTCAGCACATTCCAG and ACT2R: AACAT TGCAAAGAGTTTCAAGGT. For actin, the PCR consisted of 25 thermal cycles as described above.


Natural variation in response to low P

To gain further insight into the natural variety of root system response to P starvation in Arabidopsis, we compared six commonly used accessions grown under low P and under high P (5 and 500 µm P, respectively – see Materials and methods). Typical 10-day-old seedlings are shown in Fig. 1. A rapid examination of these seedlings shows that accessions have distinct root architecture. These accessions are also contrasted for their lateral roots. The ΣLRl for Shahdara, on average, was 14 cm whereas that of Bay-0 was 11 cm. In our growth conditions, Shahdara always exhibited the most developed lateral roots and Bay-0 was one of the accessions showing the least developed lateral roots (Fig. 1a).

Figure 1.

Natural variation in root architecture response to phosphate (P) starvation. Pictures of different Arabidopsis accessions grown under (a) high P conditions or +P and (b) low P conditions or –P (scale bars = 1 cm). Bay, Bay-0; Sha, Shahdara; Col, Columbia; Ler, Landsberg erecta; WS, Wassilevskaja; Cvi, Cape Verde Islands.

By comparison with the high P medium, the general trend on low P medium shows a reduction of root length (Fig. 1b). The range of response is wide and PRL is dramatically reduced in Shahdara (75% reduction) while PRL is less affected in Bay-0 and (20 and 19% reduction, respectively). Low P medium also restrained lateral root growth in each of these accessions (Fig. 1b).

In order to better characterize the Shahdara and Bay-0 root architectures, we plotted for each seedling the ΣLRl as a function of PRL (Fig. 2). This representation allows us to visualize the kinetics of the settling up of root architecture of 1–15-day-old seedlings. On a high P medium, the length of lateral roots of Shahdara was always higher than that of Bay-0. For instance, when the primary root reached 8 cm, the length of lateral roots was more than 20 cm for Shahdara and only 15 cm for Bay-0. On the low P medium, the root architecture of Bay-0 and Shahdara were remarkably different (Fig. 2). The PRL of Shahdara did not exceed 2 cm whereas that of Bay-0 went beyond 8 cm. A polynomial of degree 2 was adjusted on the ΣLRl plotted against PRL for each accession, with a high correlation (R2 between 0.65 and 0.94 depending on line and P supply; see Fig. 2). These contrasted growth habits of Bay-0 and Shahdara accessions, especially under low P conditions, led us to characterize the genetics of these growth responses.

Figure 2.

Graphical representation of root architecture of the Bay-0 and Shahdara accessions grown under two phosphate (P) conditions. The total length of lateral roots is plotted against the primary root length (PRL). Full symbols: Shahdara accession; open symbols: Bay-0 accession; squares: plantlets grown under P starvation (5 µm P); circles: plantlets grown in high P medium (500 µm P). Each dot represents a plant between 1 and 10 d old.

Primary and lateral root length in the Bay-0 × Shahdara RIL population

A series of 165 Bay-0 × Shahdara RILs (Loudet et al. 2002; see also Materials and methods) has been grown twice [Exp1 and experiment 2 (Exp2)] in a climatic chamber. For each experiment, the environmental conditions in the climatic chamber were set up in the same way and the RILs were grown on a low P and high P medium.

Under high P conditions, the parental lines showed a well-developed primary root and did not greatly differ from each other. Bay-0 had a PRL ranging from 6.9 to 5.2 cm and Shahdara from 7.71 to 5.36 cm in Exp1 and Exp2, respectively (data not shown).

For the RILs grown under high P, the distribution of the PRL is wide and there is a transgressive variation towards shorter as well as longer roots. In Exp1, a twofold amplitude of variation of PRL (from 5.6 to 10.9 cm) was observed among RILs (data not shown). In Exp2, on the average, the PRLs were smaller than in Exp1. This might be due to a slightly lower temperature of the growth chamber when this experiment was running. Nevertheless, during Exp2, we observed the same amplitude of lengths as in the Exp1 within the population of RILs with a twofold variation (from 4.2 to 9.2 cm) of PRL. In accordance with the results presented in Fig. 2, the PRL of the parental accessions differed considerably under low P conditions. For Shahdara, it did not reach 1 cm in both experiments, whereas for Bay-0 it was 2.3 and 3.3 cm in Exp1 and Exp2, respectively. For all the RILs, PRL was shorter when grown on low P medium than on high P medium, and it ranged from 0.2 to 4.8 cm depending on the RIL and the experiment (data not shown). The number of transgressive RILs, as well as the extent of their transgression beyond the Shahdara value, was small (data not shown). This indicates that the Shahdara parental accession already carries almost the best (higher) combination of alleles that increase this response.

Root growth depends of the number of cells produced in the root meristem and the extent of their elongation. To determine whether smaller PRL under low P conditions results from a reduction of cell extension, we measured the length of epidermis cells (Materials and methods). These measurements were performed for the two P conditions only in Exp1. In a high P environment, Shahdara and Bay-0 cell lengths were close (364 and 317 µm, respectively) and for RIL they ranged from 167 to 456 µm. Therefore, this population exhibits a large transgression in both directions for this trait. Surprisingly, in this P condition, there is a weak correlation (r = 0.21) between cell length and PRL (data not shown).

The primary root cells were significantly shorter under low P conditions than under high P conditions and ranged from 22 to 189 µm. The length of primary root cells of parental lines differed considerably (48.8 and 200 µm for Shahdara and Bay-0, respectively). In addition, either no transgression or a very small transgression was observed; Bay-0 had the longest cells and Shahdara had one of the shortest cells. Unlike what we observed under high P conditions, there was a higher correlation between cell length and PRL at low P conditions (r2 = 0.58, data not shown).

The ΣLRl was measured for all the 165 RILs grown in high P conditions for 10 d (data not shown). This variable ranged from 0.4 to 31.8 cm in Exp1 and from 0.6 to 20.9 cm in Exp2. As for PRL, the ΣLRl of almost all the RILs in Exp2 were smaller than those measured in Exp1, and a correlation is observed for this trait between the two experiments (r = 0.52). In the two experiments, the parental lines differed considerably from each other and Bay-0 always had a smaller ΣLRl than Shahdara. The transgression of ΣLRl is essentially beyond Shahdara, suggesting that Bay-0 carries one of the least favourable allelic combinations for this trait.

Mapping QTL for response to P starvation

In order to understand the root growth response to P environment, we have mapped QTL for PRL response.

In both experiments, the genetic model is composed of three QTL (LPR1, LPR2 and LPR3 on chromosomes 1, 3 and 4, respectively (Fig. 3a). LPR1 has a major effect because it accounts for more than 52% of the phenotypic variation of PRL under low P conditions with an LOD value between 17.4 and 24.7 (Table 1). LPR1 and LPR2 did not colocalize with QTL of PRL in the high P environment (data not shown), underlining that these QTL are specific to PRL under P starvation. On the other hand, LPR3 colocalizes with a QTL, affecting PRL in the high P environment (data not shown), suggesting that this QTL is involved in the PRL regardless of the P environment. Significant epistatic interactions were found between LPR1 and LPR2 and between LPR1 and LPR3 (Table 1).

Figure 3.

LOD score values from composite interval mapping (CIM) analyses of traits plotted against a linear representation of the Arabidopsis genome. Black lines represent LOD score values obtained in Exp1 and grey lines represent LOD score values obtained in Exp2. (a) Primary root growth response to phosphate (P) starvation. (b) Primary root cell growth response to P starvation.

Table 1.  Quantitative trait loci (QTL) analysis results
 QTLaChromosomeMarkerbPosition (cM)c2adLOD scorer2er2 Modelfh2grh
  • a

    The name given to a local LOD score peak contains the trait name suffixed with an order number.

  • b

    The closest left marker flanking the LOD score peak.

  • c

    The position of the QTL is expressed in cM from the first marker of the chromosome.

  • d

    The mean effect (in trait unit) of the replacement of both Shahdara alleles by Bay-0 alleles at the QTL.

  • e

    Percentage of variance explained by the QTL or by QTL–QTL interaction, when significant.

  • f

    Percentage of variance explained by the sum of the QTL or by QTL–QTL interaction detected.

  • g

    Heritability of the trait (see Materials and methods).

  • h

    Coefficient of correlation of the trait measured between the two experiments carried out.

  • LOD, logarithm-of-odds; PRL, primary root length; P, phosphate.

Response of PRL to P starvation
LPR23ATHCHIB220.8−0.10 7.913   
LPR34MSAT4.1850.8−0.08 6.6 8   
LPR1*LPR2     14   
LPR1*LPR3      2   
LPR23ATHCHIB226.0−0.14 8.712   
LPR34MSAT4.1849.8−0.10 6.3 7   
LPR1*LPR2     10   
LPR1*LPR3      0.5   
Response of primary root cell length to P starvation
Exp1LCP11MSAT1.1019.2−0.14 6.02928

The heritability of the PRL response to P environment was high (0.91, Table 1). The same QTL and epistatic interactions for PRL in the low P environment were found for the PRL response (data not shown and Table 1).

In order to better understand the epistatic interactions between QTL affecting the response to P, we grouped the RIL according to their genotype (Materials and methods). According to its QTL alleles, each of the 165 RIL has been categorized into one of eight genotypic classes. Figure 4 shows that the highest response to P occurs when the allele at LPR1 is from Shahdara (columns 1–4), regardless of the genotype at the LPR2 or LPR3. Moreover, the effects of LPR2 and LPR3 could be unravelled only when the allele at LPR1 is from Bay-0 (columns 5–8). In addition, the effect of LPR2 and LPR3 appears to be additive (compare columns 5, 6 and 7 with 8). On the other hand, one could consider that the effect of LPR1 is expressed only when Bay-0 allele is present at LPR2 and/or LPR3.

Figure 4.

Relationship between the genotype of recombinant inbred line (RIL) at the different quantitative trait loci (QTL) and the response of primary root to phosphate (P) starvation. (a) RILs were classified according to their genotype at the three QTL for response of primary root length (PRL) to P starvation. (b) Phenotype of each class of RIL from Exp1 (black symbols) and from Exp2 (grey symbols). Shahdara (triangles) and Bay-0 (circles) are plotted in their respective genotypic class. Five to 22 RILs were measured per genotypic class. Error bars are standard deviations.

The heritability of cell length under low P conditions was high (0.76). Only one QTL of root cell length in response to low P was detected, at the top of chromosome 1, which explains 29% of the phenotypic variation for this trait (LPC in Table 1). Interestingly, this QTL colocalizes with the major QTL for the PRL response to phosphate environment (LPR1, Fig. 3b). These two QTL are most probably the same QTL because the data were not independent. Therefore, the small primary root under low P results, at least partially, from the effect of LPR1 on cell elongation.

The crucial effect of LPR1 on the primary root growth response to low P

To confirm and better understand the role of LPR1 in the root response to low P, we used NIL in the form of a HIF that differ only for their allele at the LPR1 region (Materials and methods): either Bay-0 (194B) or Shahdara (194S). These two lines were grown in low P and in high P environments and the PRL of 10-day-old plantlets was measured (Fig. 5).

Figure 5.

Effect of LPR1 on primary and lateral root lengths. Pictures of the Bay-0 (Bay), Shahdara (Sha) and the NIL 194B and 194S grown for 10 d on (a) low P medium and (b) high P medium (scale bars = 1 cm). Frequency distributions (%) of the (c) primary root length (PRL) and (d) total lateral root length of the NIL 194B and 194S grown in low P for 10 d.Values from NIL 194S and 194B are represented by black and white bars, respectively. (e) Root cell length of the parental accessions (Bay-0 and Shahdara) and the NIL 194B and 194S grown for 10 d on a high P (+P) or low P (–P) medium. Bars represent the average length (and its standard error) of primary root cells (white) and lateral root cells (black). Bars labelled with different letters are significantly different at P = 0005. (f) PRL of the 194B/S grown on a low P medium. Progeny of a 194B/S heterozygote has been grown on low P medium. For each seedling, the primary root length has been measured and the genotype at LPR1 determined with the MSAT1.10 and NGA248 markers. Bars labelled with different letters are significantly different at P = 0.01. Error bar represents the standard error.

In low P, 194B looks like the parental Bay-0 accession and 194S looks like the Shahdara accession (Fig. 5a). Figure 5c shows that the PRL of 194S was less than 1 cm whereas in 194B it reached 3 cm. This result further confirms the major role of LPR1 in the variation of this trait. We tested whether LPR1 also affects ΣLRl in response to low P. When grown under low P environment, ΣLRl ranged from 0.6 to 7.9 cm for 194S and from 0.5 to 12.4 cm for 194B (Fig. 5d). In contrast to PRL, the distribution of ΣLR of the HIFs is largely overlapping. Therefore, this supports the view that in a low P environment, LPR1 is involved to a much greater extent in the growth response of primary root than lateral roots.

The cell length of primary and lateral roots was measured for the two NILs and the parental accessions (Fig. 5e). Under low P conditions, the primary root cells were much shorter in Shahdara and 194S than in Bay-0 and 194B This is in good accordance with results of root length presented in Fig. 5c. On the other hand, the cell length of lateral roots was not dependent on the LPR1 allele. Once again, this corroborates data in Fig. 5d. Altogether, the NIL analysis confirmed that LPR1 is a major QTL of primary root cell length in response to the low P environment.

The Shahdara dominant allele of LPR1 is responsible for the low P response

The two NILs 194B and 194S were studied together with the original heterozygous (194B/S) plants. Interestingly, in the low P environment, the primary root of 194B/S is slightly longer than that of 194S (Fig. 5f). In other words, the heterozygote strongly responded to P starvation. The degree of dominance (d/a, with d as the dominance deviation and a as the additivity; Tanksley 1993) of the Shahdara allele on this trait is 0.6. These data convincingly show that the Shahdara allele of LPR1 is largely dominant over the Bay-0 allele. This suggests that Bay-0 carries a non-functional or a hypomorphic allele at LPR1 that is not or only poorly able to trigger the response to low P. Thus, the short primary root in low P is most probably due to a functional allele of LPR1. Alternatively, the Bay-0 allele is inappropriately expressed.

Classical traits used to characterize the P-starvation response are not under the control of LPR1

We have shown above that the primary root growth response to P starvation is strongly compromised in plants homozygous for the Bay-0 allele at LPR1. In order to know whether this allele reduces the whole P signalling pathway, we performed a comparative analysis of 194B and 194S lines for several typical low P responses.

Among the morphological responses to P starvation is the increase in the density and length of root hair (Zhang, Lynch & Brown 2003). The 194B and 194S plantlets increased their root hair length by 33 and 54%, respectively, in low compared to high P conditions (Fig. 6a). The root hair density (defined as the number of observable root hairs per mm of root) was increased by 97% for 194S grown in low P compared to 194S grown in high P. For 194B, the density was only slightly increased (20%) (Fig. 6b). Nevertheless, the Bay-0 allele at LPR1 does not interfere with root hair growth stimulation by low P.

Figure 6.

LPR1 does not affect other typical low-phosphate (P) responses. Root hair length (a) and density (b) of the primary root from NIL grown on high P (+P) or low P (–P) medium. Bar labelled with different letters are significantly different at P = 0.01. Error bar represents the standard error. (c) Anthocyanins of 10-day-old seedlings grown on high P (+P) or low P (–P) medium were extracted and absorbance reported to the fresh weight. Bar labelled with different letters are significantly different at P = 0.01. Error bar represents the standard error. (d) in vivo BCIP revelation of acidic phosphatases. Ten-day-old 194B and 194S seedlings grown on low or high P medium were stained with BCIP before picture (scale bars = 150 µm). (e) Expression of the P transporter Pht1;4 in NIL 194B and 194S RT-PCR have been performed on RNA extracted from 10-d-old seedlings grown on high P (+P) or low P (–P) medium. PCR on genomic DNA is a control of RT-PCR specificity. The expression of actin served to control the amount of template used for PCR (Misson et al. 2004).

P starvation also enhances the anthocyanin content in shoot (Ticconi et al. 2001). Anthocyanins were extracted from 194B and 194S plantlets grown either on low or high P. When grown on similar P conditions, both NILs accumulated similar anthocyanin content (Fig. 6c). In particular, under P starvation, the anthocyanin content of both lines was 10–60 times higher compared to high P, thus demonstrating that LPR1 does not play a major role in this response.

Excretion of acidic phosphatases (APs) is also a landmark of P starvation (Trull & Deikman 1998). We used BCIP staining (Trull & Deikman 1998) to monitor in vivo the AP activity in plants cultivated either on low or high P. We observed an equivalent strong blue staining on the roots of P-starved 194B and 194S seedlings (Fig. 6d) whereas no staining could be detected with the high P seedlings. This strongly suggests that LPR1 is not crucial for the excretion of APs in response to P starvation.

Finally, we monitored the expression of Pht1;4, a gene coding for a high-affinity phosphate transporter whose transcript level is strongly increased under P-deficient conditions (Muchhal, Pardo & Raghothama 1996; Misson et al. 2004). The same expression levels were observed for 194B and 194S; in both lines, the RT-PCR product of Pht1;4 was detectable only in low P plants (Fig. 6e).

LPR1 does not affect root growth in response to N, Fe or S deficiencies

Root architecture is also modulated by the availability of other mineral nutrients including N, Fe and S (Forde & Lorenzo 2001; Lopez-Bucio, Cruz-Ramirez & Herrera-Estrella 2003). In order to investigate whether LPR1 is also involved in the control of architectural responses triggered by other nutrient deficiencies, NILs 194B and 194S were grown on separate MS media lacking N, lacking Fe and lacking S. N-starved plants began to show chlorosis symptoms 4 d after germination and developed a longer primary root compared to non-starved plants. Fe starvation resulted in a severe chlorosis and poor growth of both the shoot and roots. Finally, S deficiency also reduced growth and induced a strong development of root hairs (data not shown). For all of these growth conditions, we were unable to detect any difference between 194B and 194S. This suggests that LPR1 is not involved in a general nutrient-deficiency response (data not shown).


QTL for root growth response to low P

In the field, plants are exposed to spatially and time-heterogeneous environments. These conditions hinder the precise QTL detection of any growth trait. It is not difficult to imagine that mapping QTL for root growth in soil is an even more challenging task. In a pioneer study, Reiter et al. (1991) reported that P starvation reduces root weight of maize grown in the field. They have also mapped several QTL, one of which specifically reduces root dry weight (i.e. not shoot dry weight). As far as we know, no other QTL for root growth related to P starvation have been described in maize or any other species.

With the adoption of Arabidopsis growing in vitro, root development can be standardized. This opens the way to fine mapping and analysis of QTL controlling root growth. Examination of the Bay-0 and Shahdara accessions showed their particularly contrasted phenotype in low P.

We have carried out two independent QTL experiments, and three QTL involved in the PRL response to P starvation have been detected in each experiment. These are the first described QTL of the growth response to P starvation in Arabidopsis. We have also mapped the QTL of root growth in low P as well as in high P (M. Reymond, S. Svistoonoff, L. Nussaume & T. Desnos, unpublished results). One of the QTL of root growth in a high P medium is located on chromosome 4 and colocates with LPR3. It might correspond to PRL3, a QTL mapped using the same set of Bay-0 × Shahdara RILs and grown in N-enriched medium (Loudet et al. 2005). LPR3 displays the same allelic effect as PRL3; thus, it is possible that they correspond to the same locus. Alternatively, these two QTL are distinct but are located in the same chromosome region. No other QTL of PRL are colocated near LPR1 or LPR2 (Loudet et al. 2005 and this study). Therefore, we think these two QTL are specific in their response to the P environment. In addition, when the NILs 194B and 194S were grown on a medium depleted in N, Fe or S, they behaved similarly, suggesting that the response controlled by LPR1 is not triggered by other nutrient deficiencies. This further supports the view that LPR1 is a specific QTL of the response to low P. Although LPR1 is involved in two architectural responses to low P – PRL and length of primary root cells – it does not affect other responses to P starvation. Our results with the HIF, which differ only in the region of LPR1, demonstrate that the other typical P-starvation responses like the increase in root hair length, the accumulation of anthocyanins, the excretion of acidic phosphatases or the expression of the P-transporter gene Pht1;4 are not modulated by LPR1. In particular, plants homozygous for the Bay-0 allele at the QTL, the root growth of which is almost insensitive to low P, still activate these other low P-induced responses. Therefore, LPR1 seems to be involved in a specific low P response, resulting in a reduced primary root growth rather than a general pathway of low P sensing.

P starvation reduces cell elongation

Under P starvation, by comparison to high P conditions, the Bay-0 and Shahdara accessions behaved quite differently. Bay-0 cell length was reduced by 52% whereas that of Shahdara was reduced by 90%. A similar range of variation was observed in the RIL population for which low P reduces primary root and cell lengths. By contrast to high P conditions, there is a significant correlation (R2 = 0.33) between the primary root and the cell lengths (data not shown). This indicates that, in this RIL population, cell elongation is one of the main targets of P starvation in reducing the PRL. This result corroborates with the data presented by Sànchez-Calderón et al. (2005) and shows that, in the Columbia genetics background, P starvation inhibits root cell elongation and meristem activity.

We have detected three QTL for the primary root response to low P. One of these, LPR1, also controls the primary root cell length. This QTL explains more than 52% of the variation of PRL and around 30% of the variation of primary root cell length. Therefore, LPR1 is a major QTL controlling cell elongation of the primary root growing under P starvation. LPR1 might represent one of the loci that mediate the low P induced short root phenotype described by Sànchez-Calderón et al. (2005). For the three QTL, the Shahdara allele enhances the response to low P (reduction of root length). Because we did not find QTL for cell length colocating with LPR2 and LPR3, these last two QTL must affect growth in another way; they may control the rate of cell production in the meristem. Altogether, these data show that P starvation has several targets when reducing root elongation. These targets are cell elongation, as revealed by LPR1, and perhaps the activity of the root meristem.

Only a very few genes/QTL that control specifically the elongation of root cells have been described. Recently, Mouchel, Briggs & Hardtke (2004) mapped a major effect QTL that controls the primary root growth through cell proliferation and elongation. They have isolated the corresponding gene (BREVIX) and T-DNA mutants. These mutants do not display abnormal growth response to low P (Mouchel et al. 2004). BREVIX maps at the top of chromosome 1 but not within the MSAT1.10/NGA248 interval where LPR1 is located.

The Shahdara dominant allele of LPR1 is responsible for the low P response

The fact that Bay-0 is still able to trigger several other typical low P responses demonstrates that it is not affected in a general low P-sensing mechanism. Instead, it may be altered in a root-specific growth signal/process. The recessiveness of the Bay-0 allele of LPR1, compared to the Shahdara allele, supports the idea that Bay-0 has lost a function. Taken together, these data support the idea that LPR1 is a positive factor that contributes in inhibiting or reducing root growth in low P.

We showed that the Shahdara allele of LPR1 is largely dominant over the Bay-0 allele and is responsible for the response of the primary root to low P. This situation is in contrast with that of PDR2 for which the recessive mutant allele (pdr2) hypersensitizes root growth response to low P (Ticconi et al. 2004). Both LPR1 and PDR2 specifically affect the root. This supports the speculative view according to which these two functions operate in a common molecular process, but in opposite direction. PDR2 maps on chromosome five, thus it is not an allele of LPR1. PDR2 is supposed to function at a low P-sensitive checkpoint control during root development (Ticconi et al. 2004). In terms of molecular mechanisms, LPR1 would be a positive sensor/transductor and PDR2 a negative one. It would be interesting to test whether the function of LPR1 needs that of PDR2 in mediating the response to low P.

Natural diversity in the P-starvation response

In this work, we investigated the natural variation in root length and growth of roots in response to low P. In a small panel of accessions, we observed contrasting root growth responses. A previous study has already unravelled the natural variation in this response (Chevalier et al. 2003). However, our growth conditions gave marked differences between accessions compared to those of Chevalier et al. (2003). For example, Col exhibits a strong response to our low P conditions (5 µm P), whereas in Chevalier et al. (2003) (1 µm P) the primary root is only slightly affected. In addition, comparing Col, and WS shows that WS exhibits the strongest response in Chevalier et al. (2003), whereas in our conditions this is exhibited in Col. These discrepancies may be caused by the different ways the plantlets were grown. Firstly, our growth medium contained 1% sucrose and the mineral solution was more diluted than the one used in Chevalier et al. (2003). Secondly, Chevalier et al. (2003). their lines were germinated on a 1 mm P medium for 6 d before transferring the plantlets either to a 1 mm P or a 1 µm P medium. The root lengths were then measured 2 weeks after the transfer. In our protocol, the lines were continuously grown on a low (5 µm) and a high P (500 µm) medium, and the lengths measured at day 10. Finally, the light sources were also different. It is therefore possible that some accessions are more sensitive than the others to some aspects of these growth conditions.

The results that we present here confirm the wide range of response to low P in natural accessions of A. thaliana. Unfortunately, there are no data available on soil composition, particularly in P availability, from which these accessions originated. Therefore, it remains to be clarified whether these growth responses to low P have ecological significance. In particular, whether the poor response of accessions like Bay-0 and result from their tolerance or insensitivity to low P is yet to be determined.


We have mapped three QTL controlling the primary root response to low P. LPR1, which maps at the top of chromosome 1, is a major QTL of primary root cell elongation in response to low P, and the Shahdara allele is dominant over the Bay-0 allele. LPR1 is not involved in other classical P-starvation responses, and neither is it involved in the growth response to other mineral (N, S and Fe) deficiencies.


We thank Audrey Creff, Serge Chiarenza and Renaud Blervaque for their help in the microscope slide mounting used for the cell length measurements, Séverine Boiry and the team of the GRAP for taking care of our plants, Xenie Johnson for reading the manuscript and the NASC for providing the accession seeds.