Functional & Integrative Genomics

, Volume 5, Issue 2, pp 104–116

Expression profiling of rice segregating for drought tolerance QTLs using a rice genome array

  • Samuel P. Hazen
  • M. Safiullah Pathan
  • Alma Sanchez
  • Ivan Baxter
  • Molly Dunn
  • Bram Estes
  • Hur-Song Chang
  • Tong Zhu
  • Joel A. Kreps
  • Henry T. Nguyen
Original Paper

DOI: 10.1007/s10142-004-0126-x

Cite this article as:
Hazen, S.P., Pathan, M.S., Sanchez, A. et al. Funct Integr Genomics (2005) 5: 104. doi:10.1007/s10142-004-0126-x

Abstract

Plants alter their gene expression patterns in response to drought. Sometimes these transcriptional changes are successful adaptations leading to tolerance, while in other instances the plant ultimately fails to adapt to the stress and is labeled as sensitive to that condition. We measured the expression of approximately half of the genes in rice (∼21,000) in phenotypically divergent accessions and their transgressive segregants to associate stress-regulated gene expression changes with quantitative trait loci (QTLs) for osmotic adjustment (OA, a trait associated with drought tolerance). Among the parental lines, a total of 662 transcripts were differentially expressed. Only 12 genes were induced in the low OA parent, CT9993, at moderate dehydration stress levels while over 200 genes were induced in the high OA parent, IR62266. The high and low OA parents had almost entirely different transcriptional responses to dehydration stress suggesting a complete absence of an appropriate response rather than a slower response in CT9993. Sixty-nine genes were up-regulated in all the high OA lines and nine of those genes were not induced in any of the low OA lines. The annotation of four of those genes, sucrose synthase, a pore protein, a heat shock and an LEA protein, suggests a role in maintaining high OA and membrane stability. Of the 3,954-probe sets that correspond to the QTL intervals, very few had a differential expression pattern between the high OA and low OA lines that suggest a role leading to the phenotypic variation. However, several promising candidates were identified for each of the five QTL including a snRNP auxiliary factor, a LEA protein, a protein phosphatase 2C and a Sar1 homolog.

Keywords

Rice Osmotic adjustment Drought tolerance Expression profiling Rice genome array 

Introduction

Plants cope with dehydration stress through latent phenotypic characteristics and active responses to the environment. Those properties can result in drought escape, avoidance, or drought tolerance. For example, a plant may escape stress by completing its life cycle before the particular stress occurs. Other forms of latent stress resistance known as mechanisms of avoidance are root traits such as root thickness, root penetration ability through compacted soil layers, and root depth and mass (Price et al. 2002). These phenotypic traits do not require stress conditions to come about. In contrast, adaptive traits such as solute accumulation resulting in an osmotic adjustment (OA) and dehydration tolerance arise in response to water deficit (Ingram and Bartels 1996). The response is initiated when a plant perceives dehydration stress through an unknown mechanism (Xiong et al. 2002). That sensory cue and subsequent secondary signaling results in signal transduction within the plant leading to widespread changes in gene expression (Kreps et al. 2002; Ozturk et al. 2002; Rabbani et al. 2003; Seki et al. 2002).

Three basic types of transcriptional changes can occur in response to dehydration stress. A plant suffers the effects of drought physiologically through decreased hydraulic conductance and loss of cell turgor resulting in diminished photosynthesis and growth and development (Lu and Neumann 1999; O’Toole and Cruz 1980). Subsequently, the transcriptional control of those metabolic processes changes as well (Bray 2002). Second, the cellular disequilibria brought about by disrupting normal metabolism causes the accumulation of reactive oxygen intermediates resulting in defensive changes in transcription (Mittler 2002; Ramanjulu and Bartels 2002; Rizhsky et al. 2002). A third effect is a transcriptional change that bestows the ability to endure dehydration through a physio-chemical change in cell structure or water potential (Ingram and Bartels 1996; Tripathy et al. 2000; Zhang et al. 1999). As a result of solute concentration gradients and osmosis, drought causes the movement of water out of the cell resulting in loss of turgor. Some plants have the ability to tolerate dehydration or maintain turgor pressure through an osmotic adjustment via the active accumulation of solutes. Rice has the capacity for both drought escape and several forms of resistance through avoidance and tolerance (Lilley and Ludlow 1996; Mackill et al. 1999; Price et al. 2002; Zhang et al. 1999). The phenotypic properties of rice range from exceptionally susceptible to robustly tolerant to dehydration. Several quantitative trait loci (QTL) for OA and root traits have been identified in rice and other crops (Ali et al. 2000; Ray et al. 1996; Zhang et al. 2001; Zheng et al. 2000). The rice accessions IR62266-42-6-2 and CT9993-5-10-1-M are phenotypically different for OA. They differ for five osmotic adjustment QTLs, two of which are syntenic with other cereal drought stress QTLs (Zhang et al. 2001). A major OA QTL on rice chromosome 8 was found to be the same as one detected in the same genomic region in a different cross (Lilley et al. 1996). Comparative mapping indicated that this QTL was mapped to a corresponding region of chromosome 7 in wheat that carries a major gene controlling OA (Blum et al. 1999; Morgan and Tan 1996). Similarly, there is a genomic region for drought response conserved between rice chromosome 3 and maize chromosome 1 (Zhang et al. 2001). These results not only imply a common mechanism to drought tolerance among cultivated grasses but also these genomic regions contain genes or clusters of genes that confer drought tolerance that might be directly applied across species for improvement of drought resistance in cereal crops. Considering plants, including rice, exhibit massive changes in gene expression in response to dehydration stress (Kreps et al. 2002; Ozturk et al. 2002; Seki et al. 2002) and there are differences in drought tolerance among rice accessions, the QTL may be a manifestation of differences in transcriptional response to environmental stress.

We investigated the transcript profile of parental and double haploid rice accessions in order to (1) determine the general changes in gene expression that are induced in response to dehydration, (2) determine the differences between high and low osmotic adjusting parental lines, and (3) measure the expression pattern of a transgressive segregants derived from a cross between these lines, and identify candidate genes based on previous QTL mapping efforts.

Materials and methods

Plant material and stress treatment

A double haploid population was derived from the cross CT9993-5-10-1-M (abbreviated as CT9993), an Oryza sativa spp. japonica upland accession with a low osmotic adjustment capacity, and IR62266-42-6-2 (abbreviated as IR62266), an O. sativa spp. indica accession with high osmotic adjustment capacity (Zhang et al. 2001). The two parental lines and double haploid lines with the highest osmotic adjustment (HOADH02, HOADH65, HOADH98) and the lowest osmotic adjustment (LOADH47 and LOADH120) were used. Plants were grown in the greenhouse in a randomized complete block design with three replications consisting of three pots per replication. Plant growth and water-stressed treatments for this experiment followed the OA measurement protocol described by Babu et al. (1999) and used by Zhang et al. (2001). Two rice seedlings were grown in 5-gallon plastic pots filled with 6 kg potting mix sold as Ball Growing Mix (Ball Seed Co, 622 West Chicago, Ill.). Water was applied every alternate day and fertilizer (Miracle-Gro) was applied with water every 2 weeks. The uniform conditions among pots, established by Babu et al. (1999) to be optimal for rice cultivation, minimized the amount of variation among pot soil water potential differences. The last irrigation was administered 55 days after transplanting. Gradual water deficit treatment was imposed to represent natural drought conditions. Fully expanded second young rice leaf tissues were collected at full irrigation [∼95% leaf relative water content (RWC)], moderate dehydration stress (MDS), ∼80% leaf RWC, and at severe dehydration stress (SDS), ∼65% leaf RWC. The first RWC measurement was taken the morning following the day of last irrigation. RWC is an estimate of plant water status in terms of cellular hydration, which is in part a function of leaf water potential and OA (http://www.plantstress.com/files/RWC.htm). Considering two plants with the same leaf water potential can have different OA, we chose RWC as an estimate of OA. In addition to RWC, we visually monitored signs of leaf rolling and leaf wilting. Leaf tissue was collected from pots of different replications as soon as RWC decreased to ∼80% or ∼65%. Leaf tissue with ∼80% leaf RWC was collected between 6 and 12 days after last irrigation and leaf tissue with ∼65% RWC was collected 10–23 days after last irrigation. A set of plants was kept well watered during the stress period and leaf tissue was collected at different time points. All samples were collected prior to flowering at the tillering and stem elongation growth stage as described by the International Rice Research Institute (http://www.knowledgebank.irri.org/RP/growthStages/growthStages.htm). Four to six plants were repeatedly sampled at both moderate and severe stress per replicate. Average greenhouse day conditions were 32°C, 43% humidity, 13-h day length, and 825 μE m−2 s−1 light intensity. Average greenhouse night temperature was 27°C.

Expression profiling

Probe preparation, hybridization, and normalization procedures were conducted as previously described (Zhu et al. 2003). Total RNA was isolated from tissue pooled from approximately 18 plants that received the same treatment across three replicates in the randomized complete block design described in the previous section. Briefly, total RNA was isolated from 300 μl volume of ground, liquid-nitrogen freeze-dried tissue using the RNAwiz reagent (Ambion, Austin, Tex.). Double-strand cDNAs were synthesized from total RNA using oligo dT(24) primer containing 5′-T7 RNA polymerase promoter sequence and SuperScript II reverse transcriptase (Invitrogen, Carlsbad, Calif.). Double-stranded cDNAs were purified following second strand synthesis. Biotinylated complementary RNAs were transcribed in vitro from synthesized cDNA by T7 RNA polymerase (ENZO Biochem, New York, N.Y.). Hybridizations with a labeled cRNA were conducted with rice GeneChip microarray (Affymetrix, Santa Clara, Calif.). The GeneChip contained probe sets for ∼21,000 rice genes designed from computational predictions of the O. sativa spp. japonica genome sequence (Goff et al. 2002) and expressed sequence tag and protein sequence. The sequence and expression values for 3,551 probe sets with a twofold-change in transcript abundance following stress treatment are presented in the Electronic Supplementary Material Tables 1 and 2. The fold-change and limit fold-change (LFC) analysis are presented in Supplementary Table 3. Probe intensity was calculated as previously described (Zhu et al. 2003). Expression data were normalized as in Zhu et al. (2003) with an average target intensity of 100.

We used a modified LFC (Mutch et al. 2002) approach to account for the observed association of variance of gene expression as a function of absolute expression, i.e., higher variation was observed at lower expression levels. This correction decreases the number of changes in gene expression that will be detected due to this experimentally derived variation. Each genotype was analyzed separately and then the results were combined to calculate the global LFC cutoff. First, the data are sorted into bins using the following procedure. Genes were sorted and divided into bins based on their minimum expression values (MEVs), i.e., the lowest expression value detected on any of the arrays for that genotype. Genes with all values at or below 50 were discarded, while genes with at least one value above 50 but a MEV equal to or below the floor were assigned to the same bin. The remaining genes were divided into bins initially ten expression values wide. If a bin contained less than three genes, the standard bin width was doubled. Next, the properties of each bin were used to calculate the fold-change threshold. A gene in a bin with large variance has a stricter cut-off than a gene in a bin with low variance. The ratio of the maximum and MEVs was calculated for each gene within a bin for a given genotype. The average and standard deviation of this ratio for all the genes in each bin was calculated. In a normal distribution, a cutoff set at the sum of the average and one standard deviation will select the top 16% of ratios. The value of the 16% cutoff of each bin was plotted against the median MEV for the bin. In contrast to Mutch et al. (2002), who used a simple decay function \({\left( {y = a + {\left( {b/x} \right)}} \right)}\), we found that a double exponential equation provided the best fit for our data. To create a global limit fold-change curve, we averaged the limit fold-change cutoffs of each genotype in the following manner: First, the parameters for the fit of each genotype were used to create simulated curves. Second, an average LFC value for each expression level was calculated from the average of the curves. Third, the average data was refit and the parameters from this global average fit were used to calculate the LFC cutoff for each gene. A change in a gene was considered significant if the ratio of expression values between one of the drought stressed time points and the unstressed time point for that genotype was greater than the LFC value calculated from the minimum expression value. An example of the difference between LFC and twofold cut-off is shown in Fig. 1 comparing IR62266 well-watered versus severe stressed plants. The cut-off is meant to be inclusive. We recognize that biologically relevant changes in transcription may have been excluded and irrelevant changes included. A sample was hybridized to two different genome arrays manufactured on the same silicon wafer to assess the reproducibility of the rice microarray (Fig. S1, Zhu et al. 2003). In the same study, the congruencies of expression values of multiple probe sets were confirmed by quantitative RT-PCR (Fig. S2, Zhu et al. 2003). Similarly, using quantitative RT-PCR we measured the transcript abundance of three genes: OS009133 (3-hydroxybutyryl-coa dehydrogenase), OS012021 (protein phosphatase 2C), and OS015052 (PP2A regulatory subunit). For all three genes, the logarithm (base 2) of optical density adjusted RT-PCR values were highly correlated with logarithm (base 2) array values (r =0.88, 0.86, and 0.89, respectively) across 19 samples.
Fig. 1

Comparison of the limit fold-change (LFC) and twofold cut-off threshold for determining a significant change in gene expression. Plot of the ratio of well watered versus severe water stress IR62266 plants and minimum expression level. Genes between the LFC (solid black) and twofold cut-off (gray) lines to the right (n =4) of the intersect qualify as significant according to LFC, but do not satisfy a twofold cutoff. Those to the left of the intersect (n =198) do not qualify as significantly modulated using the LFC cutoff

The sequence of rice was compared to currently classified Arabidopsis protein sequences using FASTX and FASTX i. Where a match in the Arabidopsis genome was identified, transcripts were assigned to functional categories based on the Gene Ontology Annotation used to describe the roles of genes and gene products (http://www.arabidopsis.org/info/ontologies/go/).

QTL interval open reading frames

Candidate genes corresponding to the QTL intervals were identified through a comparative analysis of the IR62266/CT9993 genetic map (Zhang et al. 2001), the Rice Genome Research Project genetic map (http://rgp.dna.affrc.go.jp/) and the draft sequence of the rice genome, which was annotated using several gene prediction algorithms (Goff et al. 2002). Markers flanking the QTLs were placed on an integrated genetic/physical map by assigning mapped cDNA sequence by BLAST to the sequenced contigs placed on the Rice Genome Project genetic map. Predicted open reading frames within the QTL intervals were extracted for five osmotic adjustment QTLs (Zhang et al. 2001).

Results

General changes in gene expression in response to dehydration

We measured the expression profile of phenotypically divergent rice accessions and their transgressive segregants using a high-density oligonucleotide array. To increase the accuracy of our measurements, for each treatment genotype combination, leaf tissue from approximately 18 rice plants was pooled from three replicates consisting of three pots with two plants each in a randomized complete block design controlled greenhouse experiment. Through an analysis of the inherent qualities of the expression profiles, we established a moving fold-change threshold, the LFC, that considers both expression levels and fold-change for the identification of differentially expressed genes (Fig. 1). There was an appreciable difference in overall transcriptional responses of the high OA and low OA parental lines experiencing dehydration stress. Between the parental lines, IR62266 and CT9993, a total of 662 transcripts exhibited a change in expression in response to dehydration stress. A majority of those, 413, increased in abundance and 274 transcripts diminished. Expression of 25 genes was both induced and repressed depending on the accession and stage of dehydration under consideration. Differentially expressed transcripts with a match identified by protein sequence similarity to the well-annotated Arabidopsis genome were classified according to the Arabidopsis Gene Ontology Consortium protein functional categories (Table 1; http://www.arabidopsis.org/info/ontologies/go/). While many categories were similarly represented by up- and down-regulated genes, there was a greater occurrence of up-regulated rather than down-regulated changes for the following categories: cell organization and biogenesis, protein biosynthesis, protein metabolism, response to abiotic stimulus and stress, transcription, and transport (Table 1). The only process to have a notably greater number of genes down-regulated than induced was photosynthesis.
Table 1

Biological process categorization (according to the Gene Ontology Consortium scheme) of transcripts with significant changes in gene expression in IR62266 and CT9993 in response to dehydration stress

Induced

Repressed

Biological process

9

5

Biosynthesis

10

11

Carbohydrate metabolism

3

3

Catabolism

2

0

Cell communication

3

0

Cell cycle

2

1

Cell death

1

0

Cell differentiation

5

2

Cell homeostasis

19

3

Cell organization and biogenesis

1

0

Cell-cell signaling

2

3

Cellular process

7

1

Development

14

10

Electron transport

4

3

Embryonic development

7

6

Energy pathways

2

0

Lipid metabolism

48

37

Metabolism

4

0

Morphogenesis

5

8

Nucleobase, nucleoside, nucleotide and nucleic acid metabolism

3

16

Photosynthesis

73

10

Protein biosynthesis

33

14

Protein metabolism

28

8

Response to abiotic stimulus

1

5

Response to biotic stimulus

19

7

Response to endogenous stimulus

3

3

Response to external stimulus

1

0

Response to extracellular stimulus

31

8

Response to stress

2

2

Secondary metabolism

7

6

Signal transduction

19

8

Transcription

32

17

Transport

19

7

Biological process unknown

Difference in response between IR62266 and CT9993

The phenotypic response to dehydration stress of the two parental lines, IR62266 and CT9993, was very different (Table 2). At severe drought stress (SDS, ∼65% leaf RWC) only IR62266 was capable of a strong osmotic adjustment. CT9993 was not and is considered sensitive to dehydration stress. In addition to differences in osmotic adjustment, the two parental accessions also had a very different transcriptional response to dehydration stress (Table 3). Changes in gene expression from well-watered (WW, ∼95% leaf RWC) plants to moderate water stress (MDS, ∼80% leaf RWC) and SDS were measured as the number of genes that exceed the LFC threshold. More than 200 genes were induced in IR62266 and only 12 in CT9993 at MDS. The only instance where there was a greater number of genes differentially expressed in CT9993 than IR62266 was from MDS to SDS. Twenty genes were induced, none of which were the same as any of the IR62266 induced genes. Only 68 genes were differentially expressed in the low osmotic adjusting line, CT9993, from well-watered to SDS, whereas 557 changes in expression were observed in IR62266. Most of the transcriptional changes that occurred were not shared between the two parental lines.
Table 2

Phenotypic and genotypic characterization of mapping parents and five transgressive segregant double haploid lines (+ IR62266 allele, − CT9993 allele, Nd data not available; RWC relative water content, OA osmotic adjustment)

Accession

RWC

OA (Mpa)

oa1.1a

oa2.1

oa3.1

oa8.1

oa9.1

Well-watered (%)

Severe-stress (%)

Interval molecular markers

ME2_12-RG140

RM263-R3393

EM17_1-C63

G2163-R1394A

E14_6-ME4_13

IR62266

96

64

0.78

+

+

+

+

+

+

+

+

+

+

CT9993

95

63

0.47

HOADH 65

96

65

0.98

+

+

+

+

+

+

+

Nd

HOADH 98

94

64

0.96

+

+

+

+

+

+

+

+

HOADH 02

95

64

0.91

Nd

+

+

+

+

+

+

+

+

+

LOADH 47

95

66

0.24

+

Nd

LOADH 120

96

61

0.29

+

+

+

a Zhang et al. (2001)

Table 3

Number of genes with fold-change in gene expression exceeding the limit fold-change (LFC) threshold among and within groups of drought tolerant and sensitive rice accessions. Parentheses denote sample size where comparisons among multiple accessions were made

Accession(s)

Moderate dehydration stress

Severe dehydration stressa

Severe dehydration stressb

Up-regulated

Down-regulated

Up-regulated

Down-regulated

Up-regulated

Down-regulated

IR62266

221

86

8

85

340

217

CT9993

12

17

20

69

5

63

HOADH 02

Nd

Nd

Nd

Nd

544

441

HOADH 65

303

367

298

168

549

457

HOADH 98

Nd

Nd

Nd

Nd

338

443

LOADH 47

32

27

193

253

236

281

LOADH 120

248

108

70

42

308

192

IR62266 not CT9993

218

85

8

40

340

167

CT9993 not IR62266

9

16

20

24

5

13

IR62266 and CT9993

3

1

0

45

0

50

HOA lines

103 (2)

60 (2)

2 (2)

32 (2)

69 (4)c

140 (4)

HOA lines only

39 (2)

25 (2)

0 (2)

2 (2)

9 (4)

19 (4)

LOA line

0 (3)

0 (3)

0 (3)

2 (3)

0 (3)

17 (3)

LOA lines only

0 (3)

0 (3)

0 (3)

0 (3)

0 (3)

0 (3)

All lines

0 (5)

0 (5)

0 (5)

0 (5)

0 (7)

16 (7)

a Number of changes in gene expression from moderate to severe dehydration stress

b Number of changes in gene expression from well watered to severe dehydration stress

c See Table 4 for details

Double haploid transgressive segregant response to dehydration stress

The lines with the most extreme OA phenotype were selected from the double haploid population derived from a cross between IR62266 and CT9993. Three high osmotic adjusting double haploid (HOADH) lines 65, 98, and 02 had an average OA of 0.95 Mpa, a greater value than the high OA parent, IR62266 (0.78 Mpa; Table 2). Two low OA double haploid (LOADH) lines 47 and 120 had a very low average response (0.27 Mpa) to dehydration stress relative to the low OA parent, CT9993 (0.47 Mpa). A greater number of transcriptional changes, both up and down, occurred in the HOADHs than occurred in IR62266 (Table 3). While fewer genes were differentially expressed in the LOADHs than in the HOADHs, there were many more changes than occurred in CT9993. None of the induced genes were common to all the non-OA lines and therefore none of the seven lines shared a single induced gene in common across all treatments. A handful of genes (n =16) were commonly repressed in all accessions when comparing WW and SDS plants. The HOADHs had several common responses, both induced and repressed (Table 3). There were many differentially expressed genes in common among the high OA lines. The 69 genes with increased expression at SDS in all the high OA lines are described in Table 4. Nine of those genes were induced in all the high OA lines and not in any of the low OA lines. The annotation of several of those genes suggests they may have a mechanistic effect on high OA: sucrose synthase (OS021052.1_at), proline rich protein (OS006304.1_at), cysteine protease (OS001772_at), LEA (OS001162_at), and three proteins with unknown function.
Table 4

Fold-change expression values of the differentially regulated transcripts (69) in all the high osmotic adjustment (OA) lines. The nine probe sets in bold were unique to the high OA lines (WW well-watered, SDS severe dehydration stress, QTL quantitative trait loci)

Probe set

Fold-change from WW to SDS

Description, putative function

QTLa

IR62266

HOADH02

HOADH65

HOADH98

OS005816_at

9.6

18.5

5.2

4.2

ABA-responsive protein

 

OS002619_s_at

3.2

7.0

4.8

3.4

Alcohol dehydrogenase 1

 

OS000924_r_at

2.8

3.6

5.1

6.4

Beta-tubulin

 

OS000404_at

3.3

18.9

5.9

3.4

Bowman Birk trypsin inhibitor

 

OS009755.1_at

2.8

2.6

2.7

2.7

Cell death suppressor protein

 

OS014161_at

11.4

9.5

7.5

8.2

Chloroplast precursor, thioredoxin M-type

 

OS005636_at

5.2

9.5

7.8

5.3

Chloroplast, chaperonin

 

OS021798_at

4.9

3.1

2.9

2.9

Chloroplast, putative ketol-acid reductoisomerase

 

OS001772_at

3.4

4.6

2.5

3.7

Cysteine protease

 

OS005344_at

2.5

6.3

9.4

7.3

Cysteine protease

oa9.1

OS012699_at

3.0

7.6

8.8

9.5

Cysteine protease

 

OS001562.1_r_at

53.4

65.2

31.9

36.0

Dehydrin rab

 

OS001730.1_f_at

30.3

41.7

20.2

18.0

Dehydrin rab

 

OS008919.1_f_at

7.5

20.4

7.8

4.6

Dehydrin rab

 

OS011886_at

3.3

2.8

2.8

2.4

Dehydroascorbate reductase, GSH-dependent

 

OS001918.1_at

4.3

8.3

5.8

2.9

Dehydrogenase Adh1, alcohol

 

OS007190.1_at

4.3

7.1

4.3

3.7

EFA27 for EF hand, abscisic acid

 

OS008316.1_at

5.9

11.1

7.8

4.2

Ferritin

 

OS009208.1_f_at

6.3

10.2

6.5

3.7

Ferritin

 

OS016732.1_f_at

6.0

10.2

6.2

3.9

Ferritin

 

OS001113.1_s_at

3.1

3.6

6.2

2.4

Glyceraldehyde 3-phosphate dehydrogenase, cytosolic

 

OS000170_f_at

7.6

5.7

10.7

8.0

Heat shock protein

oa3.1

OS000506_f_at

2.5

5.2

3.5

3.0

Heat shock protein

oa3.1

OS011512_f_at

6.0

15.1

9.1

8.1

Heat shock protein

 

OS017210_at

9.9

2.7

5.1

6.1

Heat shock protein

 

OS007368_at

6.4

7.3

3.6

2.6

Heat shock, dnak-type molecular chaperone

 

OS012130.1_at

4.3

5.9

2.9

2.6

Homogentisate 1,2-dioxygenase

 

OS005931.1_at

2.9

5.2

4.9

2.8

Hypothetical protein

 

OS_ORF012710_at

7.6

2.6

3.0

3.6

Hypothetical protein

 

OS009845_f_at

2.5

4.4

5.7

4.8

Initiation factor

 

OS001162_at

5.2

7.9

3.5

3.0

Late embryogenesis abundant protein

oa1.1

OS001203_r_at

2.6

2.5

7.9

4.8

Lipid transfer

oa9.1

OS001783_at

5.7

7.8

8.3

5.7

Lipid transfer

 

OS002762_f_at

3.6

10.7

3.7

2.5

Lipid transfer

 

OS001718.1_i_at

13.2

34.9

9.9

5.6

Low temperature and salt responsive protein LTI6A

 

OS002491_at

5.4

5.4

3.7

3.1

Membrane protein, low-temperature induced

 

OS015687_at

2.8

4.0

2.5

3.0

Mitochondrial carrier, putative

 

OS000990.1_at

2.5

7.3

5.2

4.4

Mitochondrial precursor, aconitate hydratase

 

OS000632_at

4.5

5.3

5.3

2.9

Mitochondrial precursor, putative

 

OS023339.1_at

3.7

2.6

3.6

2.5

N-carbamyl-L-amino acid amidohydrolase

 

OS000167.1_at

5.1

8.4

4.1

5.0

Osr40g2 protein

 

OS000659.1_at

4.0

6.9

2.9

2.6

Osr40g3 protein

 

OS000788_f_at

3.9

9.3

3.7

5.5

Oxalate oxidase

 

OS003792_f_at

6.1

12.1

4.7

8.5

Oxalate oxidase

 

OS005623_f_at

4.2

11.7

2.5

6.5

Oxalate oxidase

 

OS003534_at

4.4

5.4

2.8

3.1

Polyadenylate-binding protein

 

OS009470_f_at

34.3

43.6

11.6

11.3

Pore protein

oa3.1

OS006304.1_at

3.5

2.5

2.8

2.7

Proline-rich protein, putative

 

OS001518_at

2.9

3.9

4.3

5.1

Putative chloroplast outer envelope 86-like protein

 

OS016193_at

12.0

6.4

4.5

3.7

Pyrroline-5-carboxylate synthetase (Delta 1)

 

OS018781_s_at

6.8

3.6

5.8

4.4

Pyrroline-5-carboxylate synthetase (Delta l)

 

OS006309.1_at

5.4

3.1

3.7

2.9

Pyrroline-5-carboxylate synthetase, probable

 

OS001415_at

8.8

79.8

19.1

17.4

Seed maturation-associated

 

OS003725_s_at

3.8

5.5

3.8

4.6

Senescence-associated

 

OS014087_at

2.9

3.3

4.2

3.3

Sodium sulfate or dicarboxylate transporter

 

OS021052.1_at

12.7

12.3

11.1

9.9

Sucrose synthase

oa3.1

OS000704.1_s_at

4.2

9.2

4.9

3.4

Superoxide dismutase

 

OS000509_at

3.5

4.2

3.3

2.9

Superoxide dismutase

 

OS004289.1_at

3.9

9.8

5.5

3.0

Superoxide dismutase

 

OS004666_at

2.7

13.7

33.9

10.3

Universal stress protein 1

 

OS010810_f_at

3.5

20.0

35.4

11.2

Universal stress protein 1

 

OS012557_f_at

4.1

16.3

34.6

9.7

Universal stress protein 1

 

OS002033.1_at

6.0

6.5

2.6

3.4

Unknown protein

 

OS003449_at

2.9

3.9

2.9

2.7

Unknown protein

 

OS004732_r_at

2.9

4.1

4.7

2.6

Unknown protein

 

OS005531_at

5.6

8.5

2.9

3.4

Unknown protein

 

OS008295.1_at

4.9

4.0

3.7

4.4

Unknown protein

 

OS010047.1_at

3.8

9.8

4.7

3.4

Unknown protein

 

OS013091.1_at

3.1

2.5

2.6

2.8

Xylulose kinase

 

a Genes predicted to be within an OA QTL interval

Expression profile of QTL interval genes

For each of the five OA QTL intervals, predicted genes were extracted from the rice genome sequence. One cM of the QTL intervals contains an average of 182 genes and 1.3 Mb of DNA (Table 5). The RGP map, which accounts for considerably more recombination events, has approximately eightfold fewer genes per cM. The number of genes predicted to be in each interval ranged from 891 (oa2.1) to 1,636 (oa9.1). Probes on the array represented a majority of the predicted genes found in each interval. Of the many genes that correspond to the QTL intervals that are also represented on the array (n =3,572), very few changed transcriptionally in ether high OA or low OA lines to suggest a critical role leading to the phenotypic variation (Table 5). Two of the unknown proteins found in oa1.1 and oa9.1 are among the genes up-regulated in all the high OA lines only. Several promising candidates based on expression pattern and gene annotation were identified for each QTL including a trans-acting transcriptional protein, an LEA protein, a pore protein, sucrose synthase, heat shock protein, a ribosomal protein, and an unknown protein (Table 6).
Table 5

Description of QTL intervals and corresponding ORFs and number of differentially regulated genes (WW well-watered, MDS moderate dehydration stress, SDS severe dehydration stress)

QTLs

oa1.1

oa2.1

oa3.1

oa8.1

oa9.1

Interval

RG104-RG109 (RG811)

RM263-R3393

EM17_1 (RGP)-CDO20

R1394A-G2132

RZ698-RZ792

Genetic distance (cM)

RGPa

40.7

45.6

28.7

26.5

68.5

IR62266/CT9993b

5.0

13.0

5.0

5.0

4.0

Physical distance (Mbp)

7.0

7.4

6.6

6.5

14.0

No. of predicted ORFs

1,200

891

1,126

924

1,636

No. of array probe sets

729

533

588

488

953

Accession(s)

HOA lines

WW to MDS

Induced

3

0

5

2

3

Repressed

1

0

1

0

1

WW to SDS

Induced

1

0

4

0

2

Repressed

2

0

3

0

1

LOA lines

WW to MDS

Induced

0

0

0

0

0

Repressed

0

0

0

0

0

WW to SDS

Induced

0

0

0

0

0

Repressed

0

0

0

0

0

IR62266

WW to MDS

Induced

0

1

7

4

5

Repressed

1

0

1

0

1

WW to SDS

Induced

11

4

12

7

10

Repressed

3

2

4

0

2

CT9993

WW to MDS

Induced

0

0

0

0

0

Repressed

1

1

0

0

1

WW to SDS

Induced

0

0

0

0

0

Repressed

0

0

1

0

0

Table 6

Transcripts with changes in gene expression among the predicted ORFs in the osmotic adjustment QTL intervals (WW well-watered, MDS medium dehydration stress, SDS severe dehydration stress, Nd no data, Ns not significant)

QTL

Probe set

Accession

Fold-change from WW to:

ID and similarity

MDS

SDS

oa1.1

OS011620_at

IR62266

3.7

5.2

gi|1170745|sp|P46518| late embryogenesis abundant protein

HOADH02

Nd

7.9

HOADH65

6.2

3.5

HOADH98

Nd

2.6

LOADH120

2.33

Ns

OS003033.1_at

IR62266

3.1

2.7

gi|9049456|dbj|BAA99421.1| hypothetical protein (Oryza sativa)

HOADH02

2.7

4.2

HOADH65

3.7

Ns

HOADN98

Ns

2.9

LOADH120

Ns

2.3

OS005294_S_at

IR62266

2.7

2.6

gi|1314860|gb|AAA99827.1| Sar1 homolog

HOADH02

Nd

4.5

HOADH65

3.6

Ns

HOADH98

Nd

Ns

OS001005_AT

IR62266

−3.9

−10.6

Similar to gi|3075488|gb|AAC14566.1| chlorophyll a/b-binding protein (Oryza sativa)

HOADH02

Nd

−9.3

HOADH65

−21.7

Ns

HOADH98

Nd

−2.6

LOADH120

−3.6

Ns

oa2.1

OS012941_at

IR62266

Ns

3.1

gi|124135|sp|P28284|ICP0_HSV2H trans-acting transcriptional protein putative retro element

HOADH02

Ns

Ns

HOADH65

Nd

2.6

HOADH98

Nd

2.8

OS011742_at

IR62266

3.0

2.1

Similar to gi|1881585|gb|AAB49425.1| remorin (Solanum tuberosum)

HOADH02

5.0

Ns

LOADH47

Ns

2.4

OS002156_f_at

IR62266

2.6

Ns

gi|1168548|sp|P46897|ath7_arath homeobox-leucine zipper protein

HOADH02

Ns

5.6

oa3.1

OS009470_f_at

IR62266

30.6

34.3

gi|2244974|emb|CAB10395.1| pore protein homolog (Arabidopsis thaliana)

HOADH02

Nd

43.6

HOADH65

43.6

11.6

HOADH98

Nd

11.3

LOADH47

Ns

7.6

LOADH120

18.6

7.1

OS021052.1_at

IR62266

5.4

12.7

gi|3915037|sp|O24301|SUS2_pea sucrose synthase 2

HOADH02

Nd

12.3

HOADH65

12.9

11.1

HOADH98

Nd

9.9

LOADH47

1.1

4.5

LOADH120

6.5

8.8

OS000506_f_at

IR62266

2.8

2.5

gi|1815664|gb|AAC78394.1| low molecular mass heat shock protein Oshsp17.7 (Oryza sativa)

HOADH02

Nd

4.1

HOADH65

5.2

3.5

HOADH98

Nd

3.0

LOADH47

Ns

5.6

OS001275_I_at

IR62266

−3.5

−5.3

gi|6919945|sp|P74367|PS11_SYNY3 photosystem II 11 kDa protein

HOADH02

Nd

−9.8

HOADH65

−10.1

−10.4

HOADH98

Nd

−5.9

LOADH120

−3.5

−3.2

OS012616.1_at

IR62266

Ns

2.5

gi|3108209|gb|AAC17220.1| eukaryotic cap-binding protein (Arabidopsis thaliana)

HOADH02

Ns

2.9

oa8.1

OS011880_at

IR62266

4.8

2.6

gi|2150002|gb|AAB58719.1| ribonuclease (Hordeum vulgare)

HOADH02

Nd

11.5

HOADH65

7.1

3.7

OS012021_at

IR62266

4.7

4.9

Similar to gi|633028|dbj|BAA07287.1| protein phosphatase 2C (Arabidopsis thaliana)

HOADH02

Nd

9.6

HOADH65

6.7

5.4

LOADH47

3.3

2.3

oa9.1

OS000194_at

IR62266

3.8

4.4

gi|3850818|emb|CAA77133.1| U2 snRNP auxiliary factor, small subunit (Oryza sativa)

HOADH02

Nd

2.4

HOADH65

4.8

4.0

OS012283_at

IR62266

2.3

2.9

gi|6686285 contains ER lumen protein retaining receptor PF|00810 domain. (Arabidopsis thaliana)

HOADH02

Nd

3.4

HOADH65

2.8

3.0

OS018926_at

IR62266

Ns

2.5

gi|1173139|sp|P46969|rpe_yeast ribulose-phosphate 3-epimerase

HOADH65

3.4

2.2

HOADH98

Nd

2.8

Relationship between the transcriptome and the proteome

In a similar experiment by Salekdeh et al. (2002), 14 proteins with differences in abundance in dehydration stressed CT9993 and IR62266 were identified by proteomic analysis. Eleven of the corresponding genes had complementary probes on the rice array. None of them were induced in all the osmotic adjusting accessions (Table 4). The actin-depolymerizing factor protein (S30934) was very highly expressed in IR62266 and several double haploids during dehydration stress and also in cv. Nipponbare leaves and roots when plants were treated with ABA, wounding, salt, drought, or cold stress (Syngenta Biotechnology, unpublished data). We were able to predict from rice genomic and cDNA sequences the full-length rice gene sequence for all 14 proteins identified. One of the 14 proteins identified in the study is found in an OA QTL interval. The S-like RNase homolog (AY061961) maps to the oa9.1 locus. That gene did not exhibit a significant change in gene expression in either parent.

Discussion

In this study, we measured large-scale changes in gene expression of two rice accessions and their progeny in response to dehydration stress. Approximately 10% of the genome surveyed exhibited a significant change in transcription. These results are very similar to another study that measured transcription response in dehydration stressed Arabidopsis using a full-length cDNA microarray (Seki et al. 2002). Of the ∼7,000 genes 356 were induced or repressed. Similarly, a change in transcription of 12% of the Arabidopsis genome was detected in response to hyperosmotic stress (Kreps et al. 2002). In an analysis of dehydration stressed barley, nearly 15% of all transcripts profiled were differentially expressed (Ozturk et al. 2002). A greater proportion of barley genes may have been induced due to a sampling bias of 1,463 cDNAs derived from libraries of dehydration stressed plants compared to ∼21,000 rice gene probes derived from the rice genome sequence. One of the unique aspects to the study reported herein is the analysis of two phenotypically divergent rice accessions, IR62266 and CT9993. While combined, ∼10% of the genome was differentially expressed in response to dehydration stress, and the two accessions had a very different transcript profile. For example, from well-watered to SDS conditions 340 and 5 genes were up-regulated in IR62266 and CT9993, respectively. The transcription profile of two rice accessions in response to salinity, one considered sensitive while the other tolerant to saline conditions, was similar but the susceptible accession exhibited a much slower response over time (Kawasaki et al. 2001). On the contrary, the low OA parent, CT9993, had far fewer changes in gene expression and had a very different response relative to IR62266. Thus, there was a difference in both the magnitude and the nature of the response to dehydration stress. The two studies differed in array platform as well as how the water stress was administered. If the effect is genetic, it appears that the low OA parent has no transcriptional response at all to dehydration suggesting an absence of an environment-sensing mechanism or failure to initiate the signal transduction that leads to an osmotic adjustment.

The metabolic changes that occur in plants in response to dehydration stress are described in several excellent reviews (e.g., Bray 2002; Ramanjulu and Bartels 2002; Zhu 2002). We measured changes in gene transcription that suggest many genes classified as having a photosynthetic role were down-regulated and few genes with that predicted function were up-regulated. These results support the notion that dehydration-stressed plants are physiologically challenged by dehydration and loss of turgor (reviewed by Zhu 2002). There was extensive up-regulation of genes predicted to encode antioxidants. Of the 69 genes induced in all high OA lines, several appear to have a role in antioxidant defense mechanisms: three superoxide dismutases, three oxalate oxidases, and a dehydroascorbate reductase. Many other genes that were up-regulated are believed to play a role in maintaining cell structure and water potential such as LEA proteins, dehydrin rab, lipid transferases, ferritin, sucrose synthase, and heat-shock proteins. Many of the genes identified in this study are predictably regulated in response to dehydration stress; however, the allelic variation that differentiates a high OA line and a low OA remains unclear.

QTL are in large part a manifestation of DNA sequence polymorphism that cause altered gene expression (Cong et al. 2002; Wang et al. 1999) or encode a dysfunctional protein (El Assal et al. 2001; Yano et al. 2000). Certain types of QTLs are in theory detectable using large-scale expression analysis. The fw2.2 allele causing large fruit size in tomato reaches peak levels of expression 1 week later than plants carrying the small-fruited allele (Cong et al. 2002). At a locus controlling apical dominance in maize, the maize allele of teosinte branched 1 is expressed at much higher levels than the teosinte allele (Wang et al. 1999). The sequence polymorphism for both genes, fw2.2 and teosinte branched 1, are a result of differential regulation caused by sequence polymorphism in non-coding regions of the gene. Several other examples of cloned QTLs have revealed phenotypic variation caused by posttranscriptional variation (El Assal et al. 2001; Fridman et al. 2000). The allelic variants of the Hd1 locus in rice, which is a major QTL controlling response to photoperiod, are expressed at similar levels. The functional difference between the alleles is a 43-bp deletion in the coding region resulting in increased photoperiod sensitivity (Yano et al. 2000). Detecting short sequence polymorphisms down to a single base pair such as those found at the CRY2 and PHYA loci that result in variable light sensitivity and flowering time requires strict hybridization stringency and that the polymorphic region of the gene is present on the array (El Assal et al. 2001; Maloof et al. 2001). These types of sequence polymorphism have been detected in yeast and Arabidopsis by hybridizing genomic DNA to a high density oligonucleotide array (Borevitz et al. 2003; Winzeler et al. 1998). This approach was used to fine map and clone a high temperature growth QTL in yeast (Steinmetz et al. 2002). Other recent reports describe treating gene expression as a quantitative trait within a recombinant population (e.g., Brem et al. 2002; Schadt et al. 2003). Hybridization of RNA from recombinant genetic material has successfully identified a QTL affecting glucose and fatty acid metabolism in rats (Aitman et al. 1999). The transcript was differentially expressed between spontaneously hypersensitive and normal rats as well as between congenic strains with and without the hypersensitive allele introgressed. Similar to this study, the combined use of mouse congenic strains and gene expression profiling has derived several candidates for multiple QTLs (Eaves et al. 2002). Through expression analysis of divergent and segregating lines and scanning DNA sequence within the OA QTL intervals (Zhang et al. 2001), we were able to identify several promising candidates genes.

Two transcription factors found in the oa2.1 locus were expressed at very high levels in high OA lines but not in low OA lines. The first (OS012941_at) shares similarity to a herpes virus trans-acting transcriptional protein ICP0 and a putative rice retro element. A gene that encodes a homeodomain leucine zipper (OS002156_f_at) is found in the same interval and is induced in all OA lines at MDS. The oa3.1 interval contains several genes that are up regulated in HOA lines and encode known players in abiotic stress response in plants including a pore protein (OS009470_f_at), sucrose synthase (OS021052.1_at), and a heat-shock protein (OS000506_f_at). The interval also contains a gene that encodes a eukaryotic cap-binding protein (OS012616.1_at), one of which has recently been implicated in the modulation of the ABA response pathway in Arabidopsis (Hugouvieux et al. 2001). A gene for a protein phosphatase 2C (OS012021_at) at the oa8.1 locus is significantly induced in high OA lines and at lower levels in LOADH47. Protein phosphatase 2Cs have been implicated in several instances (e.g., Meyer et al. 1994; Miyazaki et al. 1999) to be involved in the signal-transduction of the abiotic stress response. Three genes found at oa9.1 were exclusively induced in high OA lines: a U2 snRNP auxiliary factor (OS000194_at), ER lumen protein-retaining receptor (OS012283), and an epimerase (OS018926_at). Allelic variation at these loci may be the genetic cause of the phenotypic differences measured between low and high OA lines. A substantial number of genes are under consideration due to the current resolution of the QTL map. By chance, we expect to find genes in the QTL intervals that are differentially expressed that are not the direct cause of the quantitative variation between parents. Higher resolution mapping will reduce the number of genes under consideration and therefore decrease the frequency of false positives. It is possible that the allelic variation was not detected using the oligo array because expression was not measured when the causal differences were apparent, the genes are not represented on the array, or the variation is not a product of a difference in gene expression. However, these sequences co-segregate with OA phenotype, have annotation that suggests an active role, and have expression profiles that further corroborate their candidacy and warrants further investigation.

Copyright information

© Springer-Verlag 2004

Authors and Affiliations

  • Samuel P. Hazen
    • 1
    • 2
  • M. Safiullah Pathan
    • 3
    • 4
  • Alma Sanchez
    • 3
  • Ivan Baxter
    • 2
  • Molly Dunn
    • 1
    • 5
  • Bram Estes
    • 1
    • 5
  • Hur-Song Chang
    • 1
    • 6
  • Tong Zhu
    • 1
    • 5
  • Joel A. Kreps
    • 1
    • 6
  • Henry T. Nguyen
    • 3
    • 4
  1. 1.Torrey Mesa Research InstituteSyngentaSan DiegoUSA
  2. 2.The Scripps Research InstituteLa JollaUSA
  3. 3.Department of Plant and Soil SciencesTexas Tech UniversityLubbockUSA
  4. 4.Department of AgronomyUniversity of MissouriColumbiaUSA
  5. 5.Syngenta Biotechnology Inc.Research Triangle ParkUSA
  6. 6.Diversa CorporationSan DiegoUSA

Personalised recommendations