Diversity in the complexity of phosphate starvation transcriptomes among rice cultivars based on RNA-Seq profiles
Rice has developed several morphological and physiological strategies to adapt to phosphate starvation in the soil. In order to elucidate the molecular basis of response to phosphate starvation, we performed mRNA sequencing of 4 rice cultivars with variation in growth response to Pi starvation as indicated by the shoot/root dry weight ratio. Approximately 254 million sequence reads were mapped onto the IRGSP-1.0 reference rice genome sequence and an average of about 5,000 transcripts from each cultivar were found to be responsive under phosphate starvation. Comparative analysis of the RNA-Seq profiles of the 4 cultivars revealed similarities as well as distinct differences in expression of these responsive transcripts. We elucidated a set of core responsive transcripts including annotated and unannotated transcripts commonly expressed in the 4 cultivars but with different levels of expression. De novo assembly of unmapped reads to the Nipponbare genome generated a set of sequence contigs representing potential new transcripts that may be involved in tolerance to phosphate starvation. This study can be used for identification of genes and gene networks associated with environmental stress and the development of novel strategies for improving tolerance to phosphate starvation in rice and other cereal crops.
KeywordsAbiotic stress Phosphate starvation Phosphorus Transcriptome RNA-Seq Rice
Phosphorus (P) is one of the essential macronutrient for growth and productivity of cereal crops but is also one of the least available in the soils. It has therefore become a major component of inorganic fertilizers used in modern agriculture to achieve high yield of various crops. In recent years, excessive application of phosphate (Pi) has become a major concern because this non-renewable element is continuously being depleted at an alarming rate (Raghothama 1999). As plants in general rely on P for many biological functions particularly for storage and transfer of energy, which are involved in almost all metabolic processes throughout growth and development, considerable research has focused on the physiological and biochemical mechanisms of adaptation to Pi starvation (Miura et al. 2005, Jiang et al. 2007), identification of genes that control Pi stress tolerance (Rubio et al. 2001, Bari et al. 2006, Gamuyao et al. 2012), and characterization of response to Pi stress at the genome level (Wasaki et al. 2003a; Misson et al. 2005; Oono et al. 2011).
Rice adapts to Pi stress with a wide range of morphological changes such as increased root proliferation and physiological changes associated with efficient P acquisition, transport and utilization. Genes involved in enhancing P acquisition efficiency from the soil, increasing its utilization efficiency via remobilization, and translocation from shoots to roots, all of which function to compensate for the adverse effects on metabolic processes that rely on high-energy Pi compounds have been analyzed (Huang et al. 2011). The roles of transcription factors such as PHR1 (Rubio et al. 2001), WRKY75 (Devaiah et al. 2007), OsPTF1 (Yi et al. 2005), as well as genes encoding high-affinity Pi transporters (Paszkowski et al. 2002), RNases (Bariola et al. 1994), acid phosphatases (Hur et al. 2010; Wang et al. 2011), and non-protein coding gene IPS1 (Hou et al. 2005) have been the focus of studies aimed at elucidating the mechanism of tolerance to Pi stress. Additionally, genetic variation in terms of P uptake on Pi deficient soils has been pursued with the aim of identifying tolerant cultivars that can be used in breeding. Analysis of four distinct barley genotypes showed that genetic variation in P acquisition efficiency required optimization of utilization efficiency, which was correlated with the expression of low-affinity Pi transporters and IPS1 (Huang et al. 2011). In rice, several cultivars were analyzed in terms of P content on Pi deficient soils (Wissuwa and Ae 2001). The quantitative trait locus for tolerance to Pi starvation has been identified from a cross between the japonica cultivar Nipponbare with low tolerance and the indica cultivar Kasalath with high tolerance, and the QTL was eventually transferred to a near-isogenic line (NIL) of Nipponbare that eventually showed much higher P content and grain yield than the sensitive cultivar Nipponbare (Wissuwa and Ae 2001).
Recent advances in structural and functional genomics strategies led to deeper analysis of response to Pi starvation at the transcriptome level. Wasaki et al. (2003a) used a microarray platform with 8,987 ESTs to characterize the gene expression profile of Pi deficient rice roots. This study led to the characterization of OsPI1, which shares some of the properties of TPSI1/Mt4, the Pi starvation inducible gene family that plays an important role in the early stages of adaptation to low Pi availability. Using the 22K microarray platform, the gene expression level under Pi starvation between Nipponbare and NIL6-4 carrying a major QTL for Pi starvation tolerance Pup1, was analyzed (Pariasca-Tanaka et al. 2009). Genes putatively associated with root cell wall loosening and root hair extension such as xyloglucan endotrans-glycosylases/hydrolases and NAD (P) H-dependent oxidoreductase showed higher expression in roots of tolerant NIL6-4. Analysis of gene expression profiles of two indica varieties, a low P-tolerant Zhongzao 18 and not so low P-tolerant Lagrue under low P stress showed that several genes involved in glycolysis and TCA cycle were upregulated during the early stages of low P treatment in roots of Zhongzao 18 but not in root of Lagrue (Li et al. 2010). These studies clearly showed genetic variation in response to Pi starvation. However, since only a limited number of associated genes with response to Pi starvation were characterized, the entire mechanism of Pi starvation tolerance at the molecular level is barely elucidated.
We therefore embarked on RNA-Seq analysis of four Oryza sativa cultivars to characterize the variation in the transcriptomes in response to Pi starvation and to provide an overview of the regulatory mechanisms associated with Pi stress tolerance in rice and other cereal crops. We analysed the transcriptome of a japonica cultivar Nipponbare with low tolerance to Pi stress, two japonica cultivars, namely, IAC 25 and Vary Lava 701 with relatively higher tolerance, and an indica cultivar Kasalath, which is known to be highly tolerant to Pi stress.
Materials and methods
Plant materials and growth evaluation
Seeds of the japonica cultivars Nipponbare, IAC 25 and Vary Lava 701, and indica cultivar Kasalath were germinated and grown by hydroponic culture in Yoshida nutrient medium which consisted of 1.425 mM NH4NO3, 0.323 mM NaH2PO4, 0.513 mM K2SO4, 0.998 mM CaCl2, 1.643 mM MgSO4, 0.009 mM MnCl2, 0.075 mM (NH4)6Mo7O24, 0.019 mM H3BO3, 0.155 mM CuSO4, 0.036 mM FeCl3, 0.070 mM citric acid, and 0.152 mM ZnSO4 (Yoshida et al. 1976). Two-week old seedlings were subjected to Pi starvation treatment by transferring in the same nutrient medium but with the Pi concentration reduced to 0.00323 mM NaH2PO4.
The total dry weight of root and shoot samples from seedlings grown in Pi deficient medium and from untreated control were measured at regular intervals. Additionally, the dry weights under an overabundant supply of Pi were also determined for comparative purposes using root and shoot samples from seedlings grown in nutrient medium containing 3.23 mM NaH2PO4. The total P content per plant and P concentration in 1 mg plant sample from Pi deficient medium and control were measured as described previously (Oono et al. 2011). The inorganic Pi content was determined by releasing the cellular content of cells in water through repeated freeze–thaw cycle, and quantification with the molybdate assay method (Ames 1966).
The samples used for RNA preparation were collected before the onset of stress treatment (0 d), and after 10 days (10 d) and 22 days (22 d) of growth in Pi deficient medium, frozen immediately in liquid nitrogen, and stored at −80 °C until extraction.
Confirmation of expression by qRT-PCR
The expression of IPS1 and other Pi starvation responsive genes in the root and shoot samples of the 4 rice cultivars was confirmed by quantitative RT-PCR (qRT-PCR) analysis using three technical replicates from one of the three biological replicates used for RNA-Seq analysis. Frozen root and shoot samples collected at 0, 10 and 22 d of Pi starvation treatment were grounded separately. Total RNA was extracted from those samples using the RNeasy Plant Kit (Qiagen, Hilden, Germany) and treated with DNase I (Takara, Shiga, Japan). The first-strand cDNA was synthesized using the Transcriptor First Strand cDNA synthesis kit (Roche, Basel, Switzerland) according to the manufacturer’s protocol. The resulting cDNAs were amplified in the LightCycler® 480 system (Roche, Basel, Switzerland) using transcript-specific primers (Supplementary Table S1). The detection threshold cycle for each reaction was normalized using Ubiquitin1 with 5′-CCAGGACAAGATGATCTGCC-3′ and 5′-AAGAAGCTGAAGCATCCAGC-3′ as primers.
RNA-Seq analysis and identification of responsive transcripts
Total RNA from root and shoot samples was extracted and processed for construction of cDNA libraries using the TruSeq™ RNA sample preparation kit. We constructed a total of 48 cDNA libraries corresponding to root and shoot of the 4 cultivars at 0 and 22 d of Pi starvation treatment with three biological replicates for each sample. Sequencing was performed in the Illumina Genome Analyzer IIx as described previously (Oono et al. 2011). The sequence reads filtered by CASAVA (ver. 1.8) were removed using a customized Java program. Stretches of low quality bases at both sides of reads were trimmed using a customized C program (Q value <15). Adapter sequences were removed using cutadapt version 1.0 (http://code.google.com/p/cutadapt/) with default parameters. All reads were aligned to rice rRNA genes using Bowtie version 0.12.7 (http://bowtie-bio.sourceforge.net/index.shtml) with parameters (-q–threads 2–sam–un) to remove reads derived from rRNA molecules. After pre-processing the Illumina reads, the transcript structures were reconstructed using a series of programs, namely, Bowtie version 0.12.7 for short-read mapping (Langmead et al. 2009), TopHat version 1.4.1 for defining exon–intron junctions (Trapnell et al. 2009), and Cufflinks version 1.3.0 for gene structure predictions (Trapnell et al. 2010). For TopHat, the Os-Nipponbare-Reference-IRGSP-1.0 (IRGSP-1.0) pseudomolecules (http://rapdb.dna.affrc.go.jp/) were used as the reference sequences with the following options: segment-length 20, segment-mismatches 1, min-intron-length 30, max-intron 6000, max-multihits 40, no-closure-search, min-coverage-intron 30, max-coverage-intron 6000, min-segment-intron 30, max-segment-intron 6000, coverage-search, num-threads 2. All reads that could not be aligned to the IRGSP-1.0 reference genome sequence were separately analysed as described below. The expression level for each transcript was calculated as reads per kilobase of exon model per million mapped (RPKM) values based on the number of uniquely mapped reads that completely overlap with the exonic regions, using at least 2 replicates with correlation coefficient of >0.92 in each library. To detect transcripts expressed as a response to Pi starvation, G-test was performed on the read count of transcripts obtained from root and shoot at 0 and 22 d of stress treatment. The number of mapped reads on a given transcript and those on other regions for two stages were used as variables in 2 × 2 contingency tables for each test. All p-values were corrected with false discovery rate (FDR) of 0.1 % using the R package version 2.14.2 and in-house Perl scripts (Benjamini and Hochberg 1995). The resulting RNA-Seq data have been deposited to the DNA Data Bank of Japan (DDBJ) sequence read archive under the accession number DRA000685.
Venn diagram, hierarchical clustering and GO enrichment analysis
The upregulated and downregulated transcripts in the 4 cultivars were used for Venn diagram analysis using R base package version 2.14.0 and in-house Perl scripts. The commonly upregulated transcripts in root and shoot among the 4 cultivars were used for hierarchical clustering analysis. We used the heatmap.2 in the R package gplots (ver. 2.11.0) to perform clustering analyses of transcripts. The Z scores were used to compare significant changes in gene expression including fold changes. A GO term was assigned to each transcript based on the GO annotations for biological process, molecular function and cellular component in RAP-DB. GO enrichment was evaluated by Fisher’s exact test with a FDR threshold of 5 % for responsive transcripts in the biological process category which overlapped among the 4 cultivars. The results were plotted as −log10 of FDR values in a heatmap.
Transcript assembly of unaligned reads
The sequence reads from each cultivar that could not be aligned to the IRGSP-1.0 genome sequence were assembled into contigs of various k-mer sizes (k = 21 to k = 51) using various options for Velvet version 1.2.03 (parameter for velveth; -fastq -short’, for velvetg; -read_trkg yes) and Oases version 0.2.05 (defaults). The resulting contigs were merged into a final assembly with parameters ‘27 -long’ (for velveth), ‘-read_trkg yes -conserveLong yes’ (for velvetg) and ‘-merge yes’ (for oases). Redundant contigs were removed using the cd-hit-est version 4.5.4 with default parameters. To retrieve genotype specific contigs, the contigs were mapped against the IRGSP-1.0 genome sequence by Blat version v.34 resulting in the removal of contigs with more than 50 % coverage. To infer the function of contigs, a Blastx search against the NCBI RefSeq and SwissProt databases were performed using E-value of 1E − 10 as cutoff threshold. For RefSeq, only transcripts or protein records with status of ‘validated’ or ‘reviewed’ were used. Furthermore, only contigs with hit to proteins of land plant species were retained to eliminate contaminations. Lastly, contigs from IAC 25, Vary Lava 701 and Kasalath similar to Nipponbare were eliminated and the remaining contigs were identified as genotype specific. The reads used for de novo assembly were aligned back to the contigs and the RPKM values were calculated as described above.
Changes in plant growth induced by Pi starvation
Effect of Pi starvation on dry weight, total P content and P concentration relative to control of rice seedlings after 10 and 22 days in −P medium
Relative value (Pi starvation/control, control = 100)a
Vary Lava 701
Total P content of root
Total P content of shoot
P concentration of root
P concentration of shoot
The total P content of root and shoot per plant decreased from 0 d until 10 d under −P and remained at this level at 22 d in all cultivars (Supplementary Fig. S3a). In contrast, the P concentration gradually decreased from 0 d until 22 d under −P in all cultivars (Supplementary Fig. S3b). Among the 4 cultivars, relative P content of IAC 25 and Kasalath in root was higher than Nipponbare whereas relative P concentration was lower than Nipponbare (Table 1). Furthermore, although the P content of Vary Lava 701 in root was higher than Nipponbare, the P concentration was higher than Nipponbare.
RNA-Seq data sets and characterization of responsive transcripts
Mapping of RNA-Seq reads obtained from root and shoot samples of the 4 rice cultivars into the IRGSP-1.0 reference genome sequence
Vary Lava 701
Identification of core Pi starvation responsive transcripts
A total of 581 and 340 transcripts were commonly upregulated (Supplementary Table S6) and downregulated (Supplementary Table S7), respectively, in both root and shoot of the 4 cultivars. The upregulated transcripts include many Pi-related genes such as IPS1, IPS2 (Wasaki et al. 2003b; Hou et al. 2005), SPX1, SPX3 (Wang et al. 2009) and ACP (Bari et al. 2006). The downregulated transcripts include PHO2/UBC24 (Bari et al. 2006). Transcripts with no distinct functions such as Os12t0576600 (metallophosphoesterase domain containing protein), Os02t0609000, Os02t0208500 (conserved hypothetical protein), Os03t0603600 (PLC-like phosphodiesterase), Os11t0658900 (similar to lipase family protein), Os08t0280100 (similar to phytase), Os01t0128200 (similar to nuclease I) and Os04t0423400 (ABA/WDS induced protein) were also strongly upregulated. Strongly downregulated transcripts include Os12t0274700 (petunia ribulose 1,5-bisphosphate carboxylase small subunit), Os04t0380300 (kelch-type beta propeller domain containing protein), Os05t0542200 (similar to catalytic/hydrolase), Os11t0707000 (ribulose-bisphosphate carboxylase activase), Os03t0689100 (histidine acid phosphatase family protein) Os09t0246300 (conserved hypothetical protein), Os05t0105800 (hypothetical protein) and Os08t0157600 (MYB transcription factor). The expression of these genes were validated by qRT-PCR (Supplementary Fig. S5 and Supplementary Fig. S6). Although most of these genes have not been previously reported as Pi starvation responsive genes, a high level of expression in one or more cultivars may suggest specific functions associated with the response to Pi starvation.
We performed GO enrichment analysis of upregulated and downregulated transcripts in shoot and root for transcripts using GO terms in the biological process category (Supplementary Fig. S7). Enriched GO terms significantly upregulated or downregulated in all 4 cultivars may represent the core responsive transcripts in rice under −P. Twelve GO terms (ex. phosphate ion transport [GO:0006817] and glycolysis [GO:0006096]) were associated with upregulated transcripts and two GO terms (transmembrane transport [GO:0055085] and nitrogen compound metabolic process [GO:0006807]) were associated with downregulated transcripts in both root and shoot under −P. Interestingly, malate metabolic process (GO:0006108), l-phenylalanine catabolic process (GO:0006559) and flavonoid biosynthetic process (GO:0009813) were enriched in upregulated transcripts of root and downregulated transcripts of shoot. Similarly, GO terms significantly upregulated or downregulated in IAC 25, Vary Lava 701 and Kasalath may represent specific transcripts that function mainly in −P tolerant cultivars. These include transcripts for response to oxidative stress (GO:0006979) and negative regulation of apoptotic process (GO:0043066) among significantly enriched upregulated transcripts. On the other hand, transcripts for ATP biosynthetic process (GO:0006754), ATP catabolic process (GO:0006200), mannose metabolic process (GO:0006013), carbon fixation (GO:0015977), intracellular protein transport (GO:0006886) and vesicle-mediated transport (GO:0016192) were among the significantly downregulated transcripts.
Genotype specific Pi starvation responsive transcripts
Both the clustering analysis and GO enrichment analysis revealed genotype specificity of response to −P. In addition to cluster 2 transcripts in root and cluster 4 transcripts in shoot, which were significantly more upregulated in Kasalath, other clusters were also more significantly upregulated in specific genotypes. These include cluster 10 in Nipponbare shoot, cluster 5 in IAC 25 shoot and cluster 6 in Vary Lava 701 shoot (Fig. 4). Similarly, GO enrichment analysis also revealed genotype specific enriched GO terms. In Nipponbare, GO for glycolysis (GO:0006096) and defence response (GO:0006952) were enriched among the downregulated transcripts in root. In IAC 25, enriched GO for transport such as intracellular protein transport (GO:0006886) and vesicle-mediated transport (GO:0016192) among upregulated transcripts in shoot may be related to the internal translation of P and maintenance of growth activity. In Vary Lava 701, GO for tricarboxylic acid (TCA) cycle (GO:0006099), ATP hydrolysis coupled proton transport (GO:0015991), and ATP metabolic process (GO:0046034), were enriched among upregulated transcripts in root. In the tolerant cultivar Zhongzao 18, Li et al. (2010) reported that upregulation of several genes involved in the tricarboxylic acid cycle can improve the efficiency of Pi absorption under −P to produce more organic acids which are eventually released into the soil to activate the insoluble P. Kasalath showed the most number of specifically responsive transcripts. Significantly enriched GO terms include dephosphorylation (GO:0016311) and protein dephosphorylation (GO:0006470) that may function in Pi remobilization to enhance utilization efficiency in root. The GO terms for photosynthesis light harvesting, protein folding, translational elongation, translation, protein polymerization, DNA-dependent DNA replication initiation and DNA replication were enriched among downregulated transcripts in shoot resulting in growth retardation (Supplementary Fig. S2) and the repression of the synthesis of nucleic acids and proteins required for photosynthesis under −P.
Identification of Pi starvation responsive unannotated transcripts
Characterization of unaligned reads expressed under Pi starvation
An average 7.4 % of the total sequence reads from each cultivar could not be mapped to the IRGSP-1.0 genome sequence. Although most of these unaligned reads may include artifacts such as low-quality reads, sequencing errors, or sequences derived from adaptors and contaminating organisms (Oono et al. 2011), some may also represent novel transcripts that may be involved in Pi starvation. De novo transcript assembly of these unaligned reads resulted in 33,078 Nipponbare contigs, 15,971 IAC 25 contigs, 23,707 Vary Lava 701 contigs and 13,994 Kasalath contigs with average length of 484 bp. Redundant contigs among the 4 cultivars comprising 90 % of total as well as unaligned contigs from Nipponbare were presumed to be artifacts and were removed. The unaligned reads were then used for alignment to the remaining contigs from IAC 25, Vary Lava 701 and Kasalath using bowtie. To characterize these contigs more accurately, we calculated the RPKM value for each assembly and performed G-test between the control and −P treatment. As a result, we identified 144 contigs from IAC, 194 contigs from Vary Lava 701, and 162 contigs from Kasalath in either root or shoot, which were responsive under −P (Supplementary Tables S9, S10). BLASTX search in RefSeq and Swissprot databases showed homology to amino acid sequences in rice as well as other organisms. We searched the Pstol1 transcripts at Pup1 locus (Gamuyao et al. 2012) conferring the tolerance among our contigs and found full-length transcripts in Kasalath as well as the japonica cultivars IAC 25 and Vary Lava 701 (identity 100 %, coverage 100 %). The transcript showed weak upregulation in the three tolerant genotypes. This result indicates that a similar allelic composition of the Pup1 locus of IAC 25, Vary Lava and Kasalath (Chin et al. 2011), and further suggests that the contigs obtained from the different genotypes can be used for identification of genotype specificity. We also found that HOX1 (Scarpella et al. 2005), a positive regulator of root cell differentiation was upregulated in root of Kasalath and IAC 25. Additionally, DOS (Kong et al. 2006), which was shown to delay leaf senescence in rice, was also upregulated in root of Kasalath. The contigs that changed to >100-fold and <0.1-fold under −P are shown in Supplementary Tables S9 and S10, respectively. Most of the upregulated contigs were obtained from Vary Lava 701 and included transcripts associated with −P response such as inorganic pyrophosphatase 1, nucleotide pyrophosphatase/phosphodiesterase, pyrophosphate-energized vacuolar membrane proton pump, protein-tyrosine phosphatase. Most of the downregulated contigs were also obtained from Vary Lava 701. These contigs which were not identified in Nipponbare may be specifically transcribed and function only in tolerant genotypes under −P.
Identification of basal responsive transcripts under Pi starvation in rice
We investigated the dynamic expression patterns under −P treatment by identifying genes showing differential expression in the 4 rice cultivars using G-test (FDR < 0.01). Overall, approximately 20,030 (38.1 %) of 52,640 RAP-annotated transcripts showed significantly variable expression under −P treatment in at least one cultivar. This suggests that Pi starvation induce a marked systemic effect on the transcriptome of rice. Based on comparative analysis of the responsive transcripts among the 4 rice cultivars under −P, we were able to identify approximately 1,500 annotated transcripts, including many well-known Pi related genes, and several unannotated transcripts as core responsive transcripts (Figs. 3, 4, 5). Several upregulated and downregulated core genes in both root and shoot were validated by qRT-PCR (Supplementary Figs. S5, S6). We used the public microarray data (GSE6901, http://www.ncbi.nlm.nih.gov/geo/) to compare the expression of these genes to other abiotic stresses. At 2-fold or 0.5-fold cut-off and 10-fold or 0.1-fold cut-off, less than 20 % and 2 % core responsive genes, respectively, were also responsive to drought, salt and cold stress. This suggests that a large proportion of core responsive genes identified in this study may be specifically expressed in response to −P. Most of the upregulated transcripts were more strongly expressed in the tolerant indica cultivar Kasalath as well as japonica cultivars IAC 25 and Vary Lava 701 with relatively higher tolerance to −P stress than Nipponbare (Fig. 4). Existing substantial expression diversity in the core transcripts should account for the difference in response to −P between the subspecies japonica and indica as well as among the japonica cultivars. Furthermore, we have identified core transcripts expressed in both root and shoot as well as other tissue-specific core transcripts. Thus, RNA-Seq accurately measures the expression frequencies of genes over a broad dynamic range and detects previously annotated as well as unannotated transcripts that are not supported by the microarray platform in rice. For overall gene expression, we observed high correlation coefficient, suggesting a clear validation of the microarray-based gene expression profiling data with the RNA-Seq data (Oono et al. 2011). In addition, we were able to identify transcripts from IAC 25, Vary Lava 701 and Kasalath which could not be aligned to the Nipponbare genome sequence. The RNA-Seq could therefore be an efficient strategy in identifying novel transcripts particularly in cultivars with no genome sequence information.
Genotypic variation in P content, P utilization and biomass under Pi starvation
Substantial expression diversity among the 4 cultivars also exists in non-core responsive transcripts. In general, Kasalath showed a higher percentage of annotated −P responsive transcripts than those obtained from the other cultivars (Fig. 3). In all analyses, diversity in expression level was most prominent in Kasalath among the 4 cultivars. In general, indica cultivars have been shown to maintain higher relative P content and can be classified as more tolerant to −P than japonica cultivars in Pi deficient soil (Wissuwa and Ae 2001). Using P content as a measure of acquisition efficiency, IAC 25, Vary Lava 701 and Kasalath showed higher relative total P content in the roots indicating a more efficient P acquisition as compared to Nipponbare. In shoot however, Vary Lava 701 showed the highest P content among the 4 cultivars. These genotypic variations in morphology and physiological processes associated with root and shoot growth could be adaptive measures of each cultivar to enhance P acquisition under starvation. To understand the P utilization efficiency of each cultivar, we investigated the effect of −P stress treatment on inorganic P content (Supplementary Table S3). The inorganic P content relative to control was lower in tolerant cultivars as compared to Nipponbare at 10 and 22 d, except in root of IAC 25 at 22 d suggesting that tolerant cultivars tend to reduce inorganic P content to facilitate more efficient P utilization. The P utilization efficiency in root of Kasalath might have been enhanced after 22 d under −P as shown by upregulation of several genes associated to Pi starvation response (Supplementary Fig. S4) and enrichment of GO terms for dephosphorylation and protein dephosphorylation (Supplementary Fig. S7).
With an increase in root weight of Kasalath under −P, the ratio of shoot weight to root weight was decreased. The root system may have been modified to maximize Pi interception, solubilisation and acquisition under −P, the efficiency of which might be affected by the developmental stage, growth condition, and treatment. Based on this observation, it can be assumed that P acquisition from root is more important for maintaining homeostasis. Both P acquisition efficiency and P utilization efficiency are important for characterization of genotype under −P. In shoot of Kasalath, the GO terms for photosynthesis, light harvesting, translation and DNA replication were enriched among downregulated transcripts, and could therefore be correlated with growth inhibition (Supplementary Figure S1, S2). In root, P content of IAC 25 and Kasalath was higher than Nipponbare but P concentration was lower than Nipponbare (Table 1). Both P content and P concentration of Vary Lava 701 in root were higher than Nipponbare (Table 1). Although both cultivars are more tolerant under −P, a large portion of responsive transcripts as well as their expression levels also showed variation among these cultivars (Figs. 3, 4). These results indicate significant variations in P content and gene expression among the 4 cultivars. However, the regulation of these parameters is quite complex and would require more detailed analysis.
Genotype specific transcripts for tolerance to Pi starvation
Role of P1BS cis-acting element in Pi starvation signalling
The difference in transcriptome between Nipponbare and tolerant cultivars maybe associated with the level of expression among the core transcripts that were generally responsive under Pi starvation stress including many known stress related responsive transcripts. We found that most upregulated transcripts have tendency to be upregulated more in tolerant genotypes as compared with Nipponbare (Fig. 6b). However, further verification such as overexpression of some genes in Nipponbare will be necessary to establish the relationship between tolerance and transcript level. Several genes showed higher expression in roots of tolerant cultivars putatively associated with root cell wall loosening and root hair extension (Pariasca-Tanaka et al. 2009) and glycolysis and TCA cycle (Li et al. 2010) under −P. Overexpression of OsMYB2P-1 in rice enhanced tolerance to −P with greater expression of −P responsive genes such as OsIPS1, OsPAP10 and several high-affinity Pi transporters (Dai et al. 2012). In total, 27.9 % (456 transcripts) of core upregulated transcripts in root and 25.0 % (446 transcripts) of core upregulated transcripts in shoot of Nipponbare have P1BS in their promoter region. Moreover, among these P1BS-containing transcripts, 96.5 % (440 transcripts) in root and 88.3 % (394 transcripts) in shoot were more upregulated in at least one of the three tolerant cultivars as compared to Nipponbare. These suggest that a Pi-signalling mediated major system PHR1-IPS1-miR399-PHO2/UBC24 (Bari et al. 2006) and P1BS may enhance the expression of core responsive transcripts under Pi starvation. Bustos et al. (2010) analyzed the P1BS representation relative to the x-fold induction and showed a striking correlation between inducibility and P1BS content only in the 1 kb promoter regions. In addition to the P1BS system, there maybe other systems that mediate stress tolerance in rice under Pi starvation.
Changes in gene expression associated with genomic structure
The differences among cultivars in response to −P may be attributed to the differences in the genomic structure of transcripts as well as the differences in expression level. Although −P responsive transcripts derived from IAC 25, Vary Lava 701 and Kasalath were mapped onto the Nipponbare genome, the promoter regions may differ among the genotypes resulting in variation in the control of transcription and response to Pi starvation. One possible inherent factor is DNA methylation and histone modifications in the transcribed region of responsive genes. A chromatin-level regulation of −P response genes that involved the deposition of histone H2A.Z and resulting in multiple phenotypes has been demonstrated in Arabidopsis (Smith et al. 2010). Furthermore, differential epigenetic modifications that have been correlated with changes in transcript levels among hybrids and parental lines based on analysis of single nucleotide polymorphisms (SNP) of the genome sequence (He et al. 2010) could also account for the differences in response to −P. The SNP in the regulatory region of specific genes has been found to induce significant alteration of gene expression as reported in the loss of seed shattering in Nipponbare owing to the absence of abscission layer formation (Konishi et al. 2006). A single mutation that resulted in a frame-shift deletion within the Rc gene was known to induce the change in seed colour from red in wild rice to white in cultivated rice (Sweeney et al. 2007). Furthermore, minor changes in sequence during domestication of cultivated rice have also been associated with genes such as Bh4 (hull colour), PROG1 (tiller angle), sh4 (seed shattering), qSW5 (grain width) and OsC1 (leaf sheath colour and apiculus colour) (Huang et al. 2012). The evolution of morphological features has been associated with changes in the cis-regulatory sequences as induced by various biochemically functional elements and buffering action of enhancers (Meireles-Filho and Stark 2009). In the present study, various novel responsive transcripts from the 4 cultivars identified among unaligned reads (Supplementary Tables S9, S10) also suggest differences in genomic structure associated with the response to Pi-starvation. Transposon-mediated transcriptional control of neighbouring genes may also add to the complexity of the regulatory networks that can be initiated by transposon insertions that render adjacent genes stress-inducible (Naito et al. 2009). Therefore, in the process of hybridization to develop new cultivars, epigenetic modifications may have occurred resulting in differences in expression, protein activity, and target specificity from −P tolerance.
In the case of adaptation to Pi stress, the difference in response between Nipponbare and the three tolerant cultivars could be possibly associated to orthologous genes that evolved from a common ancestral gene but eventually diverged in structure and function to a certain degree. Therefore the difference in transcriptome among genotypes could also be associated to the presence of responsive transcripts totally absent in Nipponbare (Fig. 6c). These responsive transcripts include sequence reads that could not be mapped to the Nipponbare genome but showed homology to known sequences (Supplementary Tables S9, S10). In the case of Pup1, a major −P tolerance QTL located on rice chromosome 12 was initially identified in Kasalath and molecular markers evenly distributed over the fine-mapped 278-kb Pup1 region were found to differ in allele constitutions in 81 rice accessions (Chin et al. 2011). This may suggest that other −P responsive genes in the three tolerant cultivars may have alleles that are totally undetected in Nipponbare. We searched the Pstol1 (Gamuyao et al. 2012) among unaligned contigs and found full-length transcripts in Kasalath as well as the japonica cultivars IAC 25 and Vary Lava 701. Overall, we obtained a few thousand contigs from these unaligned reads in the Nipponbare genome. Homology search of these contigs revealed a wide range of possible putative functions that maybe directly or indirectly involved in response to −P among different rice cultivars. These contigs may represent genes involved in the biochemical adaptation of Pi-starved plants. The genetic basis of specific differences between Nipponbare and tolerant genotypes must be explored further based on the expression patterns, distribution of reads, and responsive contigs to elucidate the mechanisms involved in tolerance to −P. Recent studies have also shown that OsPHF1 (Pi transporter traffic facilitator) was involved in trafficking Pi transporters from endoplasmic reticulum to plasma membrane that resulted in adjustment of Pi uptake ability (Wu et al. 2013, Chen et al. 2011). Thus, in addition to transcriptional level, it would necessary to elucidate posttranscriptional regulation mechanisms for more comprehensive understanding of tolerance under −P.
In this study, we were able to characterize the diversity of transcriptomes under −P based on RNA-Seq profiles of 4 rice genotypes. Additionally, we were able to identify many annotated, unannotated and unaligned responsive transcripts for accessing, mobilization, acquisition and utilization of Pi under stress conditions. Variation in the expression of these transcripts provides an overall view on how genotypes with different levels of tolerance to Pi stress respond under −P. Genotypic differences in overcoming Pi stress could be associated with differences in the genomic structure of transcripts involved in tolerance to Pi stress, differences in expression level of core responsive transcripts, and genotype-specific genes that play significant roles in overcoming Pi stress. These results will be useful deciphering gene networks involved in −P stress and for identifying genes that could be exploited in breeding for P-efficient and high yielding cultivars under −P.
We thank Dr. F. Sakakibara, Ms. F. Aota, Ms. K. Ohtsu, and Ms. K. Yamada for technical assistance. Seeds of the 4 rice cultivars were obtained from the Genebank of the National Institute of Agrobiological Sciences. This work was supported by a grant from the Ministry of Agriculture, Forestry and Fisheries of Japan (Genomics for Agricultural Innovation, RTR-0001).
- Ames BN (1966) Assay of inorganic phosphate, total phosphate and phosphatases. In: Elizabeth F, Neufeld VG (eds) Methods in enzymology. Academic, New York, pp 115–118Google Scholar
- Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289–300Google Scholar
- Huang X, Kurata N, Wei X, Wang ZX, Wang A, Zhao Q, Zhao Y, Liu K, Lu H, Li W, Guo Y, Lu Y, Zhou C, Fan D, Weng Q, Zhu C, Huang T, Zhang L, Wang Y, Feng L, Furuumi H, Kubo T, Miyabayashi T, Yuan X, Xu Q, Dong G, Zhan Q, Li C, Fujiyama A, Toyoda A, Lu T, Feng Q, Qian Q, Li J, Han B (2012) A map of rice genome variation reveals the origin of cultivated rice. Nature 490:497–501CrossRefPubMedGoogle Scholar
- Misson J, Raghothama KG, Jain A, Jouhet J, Block MA, Bligny R, Ortet P, Creff A, Somerville S, Rolland N, Doumas P, Nacry P, Herrerra-Estrella L, Nussaume L, Thibaud MC (2005) A genome-wide transcriptional analysis using Arabidopsis thaliana Affymetrix gene chips determined plant responses to phosphate deprivation. Proc Natl Acad Sci USA 102:11934–11939CrossRefPubMedGoogle Scholar
- Wang L, Li Z, Qian W, Guo W, Gao X, Huang L, Wang H, Zhu H, Wu JW, Wang D, Liu D (2011) The Arabidopsis purple acid phosphatase AtPAP10 is predominantly associated with the root surface and plays an important role in plant tolerance to phosphate limitation. Plant Physiol 157:1283–1299CrossRefPubMedGoogle Scholar
- Yoshida S, Forno DA, Cock JH, Gomez KA (1976) Laboratory manual for physiological studies of rice, 3rd edn. International Rice Research Institute, Manila, pp 61–67Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.