Transcriptomic analysis of metabolic changes by phosphorus stress in rice plant roots


Correspondence: Jun Wasaki, Tel. & Fax: + 81 11 706 3845; e-mail:


As most soil phosphates exist as insoluble inorganic phosphate and organic phosphates, higher plants have developed several strategies for adaptation to low phosphorus (P). These include the secretion of acid phosphatase and organic acids, induction of the inorganic phosphate (Pi) transporter and the substitution of some enzyme activities as alternative pathways to increase P utilization efficiency. It has been proposed that plants also have a ‘pho regulon’ system, as observed in yeast and Escherichia coli; however, the detail of the regulation system for gene expression on P status is still unclear in plants. To investigate the alteration of gene expression of rice roots grown under P-deficient conditions, a transcriptomic analysis was conducted using a cDNA microarray on rice. Based on the changes of gene expression under a –P treatment, the up-regulation of some genes due to P deficiency was confirmed . Some new important metabolic changes are suggested, namely: (1) acceleration of carbon supply for organic acid synthesis through glycolysis; (2) alteration of lipid metabolism; (3) rearrangement of compounds for cell wall; and (4) changes of gene expression related to the response for metallic elements such as Al, Fe and Zn.


As most of phosphorus (P) compounds exist as insoluble inorganic phosphate or organic phosphate, the acquisition ratio of P is lower than other essential elements. With the increase in world population, an efficient use of P resources is required for sustainable food production. To achieve this aim, the strategies for adaptation to low P need to be well known.

When plants are grown under P-deficient conditions, their roots secrete acid phosphatase (APase) and organic acids to release inorganic phosphate (Pi) from organic compounds and insoluble inorganic phosphate compounds, respectively (Gardner, Barber & Parbey 1983; Tadano & Sakai 1991). Several related genes have been isolated and characterized such as LASAP2 encoding S-APase of white lupin (Wasaki et al. 2000). It has also been reported that gene expression of high affinity Pi transporter was increased in roots of P-deficient plants (Mucchal, Pardo & Raghothama 1996; Mucchal & Raghothama 1999; Liu et al. 2001; Kai et al. 2002). The efficient utilization of absorbed P in plant tissues is also known. Many papers describe the bypass pathway contributing to save Pi rather than producing energy (Duff et al. 1989; Duff, Plaxton & Lefebvre 1991). Production of ribonucleases also increases under P deficiency, and its function is to mobilize Pi from internal RNA pools (Nürnberger et al. 1990; Löffler et al. 1992; Green 1994; Bosse & Köck 1998). Goldstein, Baretlein & Danon (1989) proposed that plant has a ‘pho regulon’, as observed in yeast (Bergman et al. 1986) and Escherichia coli (Torriani & Ludtke 1985), which may regulate the expression of these low P adaptation-related genes. However, the detail of the regulation mechanism of the gene expression system responding on P status has not yet been explained.

In recent years, genomic studies have been undertaken. For example, the genome sequence of Arabidopsis thaliana has been completed (The Arabidopsis Genome Initiative 2000), and the draft sequence of the rice genome determined (Goff et al. 2002; Yu et al. 2002). Further research involving the genome sequence includes the development of post-genomic studies that include transcriptomic analysis using cDNA arrays. Recently, several articles describing the results of transcriptomic analysis for inorganic nutrient metabolisms of higher plants have been published (Wang et al. 2000; Thimm et al. 2001; Wang, Gravin & Kochian 2001, 2002; Negishi et al. 2002). These analyses have contributed notably to the understanding of how changes of metabolism respond on nutrient status .

In the present study, we used the cDNA microarray of rice made by the microarray project in Japan (Kishimoto et al. 2002). The all-inclusive expression analysis was performed to investigate the changes of gene expression under P-deficient conditions and to understand the function of the low P adaptation mechanism of plant.


Plant material

Rice (Oryza sativa L. ssp. japonica cv. Michikogane) seeds were sterilized with 70% ethanol and germinated for 36 h in 1 mm CaCl2 with aeration in dark conditions. Germinated plants were transferred to pots containing 1 mm CaCl2 and grown for 4 d to reduce the level of stored phosphorus in the seed. They were then transferred to a nutrient solution and treated with or without P. All culture was conducted in a growth chamber at 28 °C, 24 h light, approximately 130 µmol photons m−2 s−1 at leaf level and 50% relative humidity. The composition of the nutrient solution was as follows: 0.83 mm NH4NO3, 0 (–P solution) or 32 µm (+P solution) NaH2PO4, 0.38 mm KCl, 0.19 mm K2SO4, 0.75 mm CaCl2, 0.82 mm MgSO4, 36 µm Fe (III)-EDTA, 9.1 µm MnSO4, 46 µm H3BO3, 3.1 µm ZnSO4, 0.16 µm CuSO4 and 7.4 nm (NH4)6Mo7O24. The pH was adjusted to 5.2 using 0.1 m NaOH and the solution was changed every 24 h. Half of the plants grown on the +P solution were transferred to the –P solution after 8 d and all plants were collected the next day. The treatments were defined as +P treatment, short-term –P treatment (transferred from the +P solution after 8 d) and long-term –P treatment (continuously grown in the –P solution for 9d). The fresh weights of shoots and roots were measured, then the samples were frozen immediately using liquid nitrogen and stored at −80 °C.

Measurement of total P content

The shoots were freeze-dried and ground, then approximately 50 mg of each sample was digested with H2SO4–H2O2. The P content in the digested solution was measured using the method of Saheki, Takeda & Shimizu (1985).

RNA extraction and labelling

Total RNAs were extracted by sodium dodecyl sulphate (SDS)–phenol method (Palmiter 1974) and purified by CsCl gradient ultracentrifuge method (Sambrook, Fritsch & Maniatis 1989). Forty micrograms of purified RNAs were used per array and set as the targets for microarrays. These were reverse-transcribed and labelled using Superscript II (Gibco BRL, Rockville, MD, USA) with Cy5-dCTP (Amersham Bioscience, Piscataway, NJ, USA). Unlabelled primers and dyes were removed with GFP columns (Amersham Bioscience).

Microarray analysis

Rice cDNA microarrays were prepared by the microarray project in Japan (Kishimoto et al. 2002), with a system developed by Amersham Bioscience. The arrays contained 8987 cDNA clones from several organs of rice (Oryza sativa L. ssp. japonica cv. Nipponbare) and were prepared on two slide glasses. Half of the cDNAs were spotted on each slide as duplicates, and hybridization was performed at 60 °C for 4 h using a humidified chamber and an ExpressHyb solution (Clontech, Palo Alto, CA, USA). After hybridization, the glasses were washed and dried. The signal intensity of each spot was obtained using an array scanner (FLA8000; Fuji Film, Tokyo, Japan) and the scanned signals were analysed using Array Gauge software from Fuji Film.

The values gained from each spot were standardized to remove influences among slide glasses caused by unevenness of hybridization and differences in labelling efficiency. Since the distribution of the logarithmic value for each spot was similar to normal distribution, the deviation values from all signal intensities were calculated using logarithmic values. Each spotted field consisted of 384 spots and was regarded as a population for normalization. Changes of the expression mass were estimated by the differences of the deviation values between treatments. If the deviation value in the –P treatment increased or decreased by more than five among all replications, they were defined as ‘up-regulated genes’ or ‘down-regulated genes’, respectively. The normalization and statistical analyses were performed using Microsoft Excel 2001 (Microsoft, Redmond, WA, USA) on personal computers (Apple Computer, Cupertino, CA, USA). Each EST was annotated by the superior known gene identified in a Blast search, and the significance of annotation was estimated by whether the e-value was under 1.0e−10.

Investigation of mRNA accumulation using quantitative real-time polymerase chain reaction

Isolated RNA was treated with DNase (RT grade; Nippon Gene, Tokyo, Japan) to digest the contaminated genomic DNA, then reverse-transcribed using the 1st Strand cDNA Synthesis Kit for reverse transcription (RT)-polymerase chain reaction (PCR) (AMV) (Roche Diagnostics, Basel, Switzerland). The first stranded cDNAs were used as templates for quantitative real-time PCR using the Light Cycler system (Roche Diagnostics). The TaqStart antibody (Clontech) was used for repression of non-specific amplification and the RAc1 gene was selected as the control gene because it was reported that expression of this gene is relatively stable among actin isoforms in rice (McElroy et al. 1990). Calibration curves were drawn with two replications based on the specific plasmids.


The P concentration in leaves was markedly decreased after the long-term –P treatment [7.1 ± 0.05 mg P g−1 dry weight (DW) in the +P treatment, 2.0 ± 0.20 mg P g−1 DW in the long-term –P treatment]. In the long-term –P treatment, the fresh weight of shoots decreased significantly (160 ± 1.5 mg plant−1 in the +P treatment, 107 ± 1.7 mg plant−1 in the long-term –P treatment), whereas the root fresh weight was similar among the treatments (111 ± 2.7 mg plant−1 in the +P treatment, 110 ± 4.0 mg plant−1 in the long-term –P treatment).

The genes up- and down-regulated by the –P treatment are listed in Table 1. There were 15 up-regulated genes in the short-term (24 h) and 86 in the long-term (9 d) –P treatment, whereas there were 23 and 97 down-regulated genes in the two treatments, respectively. Nine of 15 genes up-regulated, and 14 of 23 genes down-regulated in the short-term –P treatment showed similar regulation to the long-term –P treatment. Element no. 3526 showed the most significant increase of its transcription in the long-term –P treatment. Its function was unknown, and was designated OsPI1 (Oryza sativa phosphate-limitation inducible gene 1; Wasaki et al. 2003). The functions of 46 (47%) of the down-regulated genes were known, whereas those of 63 (73%) were known for the up-regulated genes.

Figure 1.

Figure 1.

List of up- and down-regulated genes in -P treatment. (a) Up-regulated genes by short and/or long term -P treatment; (b) Down-regulated genes by short and/or long term -P treatment

Figure 1.

Figure 1.

List of up- and down-regulated genes in -P treatment. (a) Up-regulated genes by short and/or long term -P treatment; (b) Down-regulated genes by short and/or long term -P treatment

To evaluate the validity of microarray analysis, the mRNA level of two genes encoding high affinity Pi transporters was quantified by RT-PCR. One of them was up-regulated, and the level of another decreased (Fig. 1a). The results of quantitative real-time PCR indicated similar patterns, that is, significant increase in no. 6593 and slight decrease in no. 7574 (Fig. 1b). High affinity Pi transporter genes were isolated and characterized in several plants, such as Catharanthus roseus (Kai et al. 1997), Arabidopsis thaliana (Mucchal et al. 1996; Smith et al. 1997; Mudge et al. 2002), Lycopersicon esculentum (Mucchal & Raghothama 1999), Nicotiana tobaccum (Kai et al. 2002) and Lupinus albus (Liu et al. 2001), some of which have been suggested to play a role in high Pi absorption efficiency in roots under P-deficient conditions. It was also known that some Pi transporters are more constitutive in root tissues(Mucchal & Raghothama 1999; Liu et al. 2001; Mudge et al. 2002).

Figure 1.

Verification of the results obtained from microarray analysis. (a) Change of the deviation value of Pi transporter homologues analysed by microarray. Open and closed squares indicate Pi transporter homologues spotted on the element numbered 6593 and 7574, respectively. (b) Relative mRNA accumulation of Pi transporter homologues analysed by quantitative real time PCR. The relative amount of mRNA in +P shoots was defined as 1.0. (c) Change of the deviation value of actin homologues analysed by microarray. □, ▪, ○, •, ▵, ▴, ▿, and ▾ indicate actin homologues spotted on the element numbered 628, 3149, 5502, 5820, 6468, 6718, 7145, and 8738, respectively. (d) Relative mRNA accumulation of RAc1 analysed by quantitative real time PCR. The relative amount of mRNA in +P shoots was defined as 1.0.

Microarray analysis suggested that the slope of the increase of actin mRNA was similar in all isoforms (Fig. 1c). The reproducibility of the result was confirmed by quantitative real time PCR analysis about one of actin isoforms, designated as RAc1 (Fig. 1d). It was reported that the RNase activities in plant cells increased under P limitation (Nürnberger et al. 1990; Löffler et al. 1992; Green 1994; Bosse & Köck 1998) and, thus, RNA pool may act as a phosphorus reservoir in the cell. As the result of a decrease of total RNA pool, important mRNA molecules, such as actin, may increase relatively. The decreases of expression of some photosynthesis-related genes are also considered to present a similar phenomenon, because it is considered that these genes are not essential for root function.

del Pozo et al. (1999) reported that the expression of type 5 acid phosphatase (APase) in shoots and roots of Arabidopsis increases under low P conditions. A homologue of type 5 APase spotted on the microarray also increased in our assay (Table 1a, element no. 1597), indicating that the homologue played a role in the efficient use of P in rice plants, similar to the Arabidopsis gene. However, another homologue of type 5 APase decreased in the long-term –P treatment (Table 1b, element no. 8064). Further studies focused on the localization and enzymatic properties are required to clarify their function.

Genes up-regulated in the long-term –P treatment contained pyrophosphatase (PPiase) homologues (Table 1a, element no. 301, 3001). Pyrophosphate (PPi) is not only hydrolysed to two inorganic phosphate molecules, but also has a high-energy phosphate ester bond, which can replace ATP as an energy source. Palma, Blumwald & Plaxton (2000) reported that the activity and protein mass of vacuolar type H+-PPiase increases under P-deficient conditions. In this study, both vacuolar H+-ATPase (element no. 301) and H+-PPiase (element no. 3001) were up-regulated, suggesting that their functions were related to the active transport between the vacuole and cytosol. The cDNA microarray used in this study does not include all genes of rice. Nevertheless, alteration of numerous genes related to carbon skeleton metabolism was indicated by the array analysis. When the alteration fits on the metabolism map, important information is found (Fig. 2).

Figure 2.

Enhancement of scheme of the carbon skeleton metabolism under the low P condition. Red and blue letters indicate the up- and down-regulated genes.

Several genes related to glycolysis increased their expression level under the –P condition (Table 1a, yellow, Fig. 2); this provides a larger amount of carbon sources for the tricarboxylic acid (TCA) cycle. Uhde-Stone et al. (2003a) also reported that some glycolytic pathway related genes were induced in –P proteoid roots, which were formed in white lupin lateral roots under the P-deficient conditions. Furthermore, because the level of OsAMT1, which encodes the ammonia transporter, and glutamine synthase homologues were down-regulated in the –P roots (Table 1b, yellow, Fig. 2), it seems that the carbon flow for amino acid synthesis is repressed. Many plants secrete organic acids under P-deficient conditions to use the sparingly soluble inorganic phosphate compounds from roots (Gardner et al. 1983). It has been reported that rice roots also secrete citrate (Kirk, Santos & Findenegg 1999). The necessity of carbon supply by the acceleration of glycolysis seems appropriate  for  organic  acid  exudation.  In  the  long-term –P treatment, some up- or down-regulated genes are related to lipid metabolism (Table 1, pink). It is interesting that a rice homologue of the sqdX gene, required for sulfolipid synthesis, was up-regulated. The sqdX gene was isolated from a cyanobacterium with sqdA, sqdB, sqdC and sqdD, and consists of an operon for sulfoquinovosyl diacylglycerol (SQDG) synthesis (Benning & Somerville 1992a ,b; Rossak et al. 1995; Güler et al. 1996). It was predicted that sqdX product catalyses the SQDG synthesizing reaction from UDP-sulfoquinovose and diacylglycerol, because of its homology with glycosyltransferase (Güler, Essigmann & Benning 2000). In higher plants, SQD1, homologous with sqdB of cyanobacteria, was isolated from Arabidopsis (Essigmann et al. 1998). P deficiency enhances SQD1 expression and SQDG accumulation in leaves (Essigmann et al. 1998). As SQDG has the ability to substitute for phospholipids, it was suggested that the increase of SQGD synthesis is available for the efficient use of P in the membrane (Essigmann et al. 1998). Gniazdowska, Szal & Rychter (1999) reported that the phospholipids of kidney bean roots were reduced by approximately half under a –P condition, although the total lipid content per protein of plasma membrane fraction was not changed. It was understandable that non-phospholipids containing SQGD were newly synthesized under P-deficient conditions and substituted for phospholipids.

Sulfoquinovose, a saccharide containing S, was synthesized from UDP-glucose and used as a substrate for SQDG. UDP-galactose-4-epimerase was up-regulated and UDP-glucose-4-epimerase was down-regulated in the microarray analysis (Table 1, pink). In other words, the equilibrium of UDP-galactose and UDP-glucose was inclined to increase UDP-glucose, which might contribute to supply the substrate for sulfoquinovose. UDP-glucose is produced by catalysis of UDP-glucose pyrophosphorylase and sucrose synthase. Ciereszko et al. (2001) showed that stimulation of UDP-glucose pyrophosphorylase under P-deficient conditions was transcriptially regulated. Sucrose synthase activity increased more than twice in –P roots of bean (Ciereszko, Zambrzycka & Rychter 1998). These facts coincide with the inclination of UDP-glucose increase in this study.

A possibility should be considered that the increased UDP-glucose was used as a substrate for rearrangement of cell wall components. Up- and down-regulation of polysaccharides synthesis or degradation of related genes , such as cellulose synthase, callose synthase, β-glucanases, and UDP-glucose 6-dehydrogenase was shown in this study (Table 1). A change of secondary metabolisms was suggested. A homologue for the PEP/Pi translocator was up-regulated (Table 1a, blue). The PEP/Pi translocator is an antiporter on plastid membranes and the transported PEP into the plastids could be used as a substrate of shikimate kinase (Hausler et al. 2000), whose homologue was up-regulated (Table 1a, blue). The Trp synthesis pathway was assumed to be repressed by the down-regulation of anthranilate phosphoribosyltransferase homologue (Table 1b, blue). On the other hand, members of the Tyr/Phe synthesis pathway were not influenced. Phe is one of the substrates for anthocyanin synthesis, and the accumulation of anthocyanin is known under P starvation (Bariola, MacIntosh & Green 1999). Another research using macroarray of white lupin revealed up-regulation of lignin synthesis-related genes in –P proteoid roots of white lupin (Uhde-Stone et al. 2003b). An increase of the lignin content may relate to the rearrangement of cell wall components.

Generally, P is over-accumulated under Zn-deficient and P-sufficient conditions (Huang et al. 2000). Huang et al. (2000) explained that this accumulation is caused by over-expression of a high affinity Pi transporter, which is up-regulated in Zn-deficient barley independently of the P status. It was considered that Zn has a role in regulating the P starvation inducible molecular mechanism with some molecules, such as RezA of rice (the function is not still clarified; accession No. AAA87049), which was down-regulated in our long-term –P treatment (Table 1b, green).

When Poaceae plants are grown under Fe deficiency, their roots exude phytosiderophore to obtain Fe from insoluble forms (Takagi 1976; Takagi, Nomoto & Takemoto 1984). In this study, homologues for nicotianamine aminotransferase and yellow stripe1 (transporter of Fe(III)-phytosiderophore in maize; Curie et al. 2001) encoding a transporter of Fe (III)-phytosiderophore in maize were down-regulated in the long-term –P treatment (Table 1b, green). It is possible that Fe and Pi will form insoluble compounds in the +P nutrient solution. This indicates that the +P plants would initially fail under a slight iron deficiency, then their phytosiderophore system would respond.

Wang et al. (2001, 2002) made a macroarray spotted 1280 mineral nutrient-related cDNAs of tomato and analysed the responses on P, K and Fe deficiencies. They identified some genes previously not associated with P, K and Fe nutrition, and suggested the co-ordination and co-regulation of uptake of these mineral nutrients. Although there is a qualitative difference between our study and theirs in aspects of the period of treatments and light condition, P responsible genes associated with N, Zn, Fe and Al were also identified in this study. To clarify the whole molecular mechanism of plant strategy to adapt to P deficiency, it will be necessary to consider the cross talk among responses to different elements.


This study was supported by a grant from the Ministry of Agriculture, Forestry and Fisheries of Japan (Rice Genome Project MA-2111).