bs_bs_banner
Plant, Cell and Environment (2014) 37, 2024–2035
doi: 10.1111/pce.12274
Original Article
A step towards understanding plant responses to multiple environmental stresses: a genome-wide study Nasser Sewelam1,2, Yoshimi Oshima1, Nobutaka Mitsuda1 & Masaru Ohme-Takagi1,3 1
Bioproduction Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8566, Japan, 2Botany Department, Faculty of Science, Tanta University, 31527 Tanta, Egypt and 3Institute for Environmental Science and Technology (IEST), Saitama University, Sakura-ku, Saitama 338-8570, Japan
ABSTRACT In natural habitats, especially in arid areas, plants are often simultaneously exposed to multiple abiotic stresses, such as salt, osmotic and heat stresses. However, most analyses of gene expression in stress responses examine individual stresses. In this report, we compare gene expression in individual and combined stresses. We show that combined stress treatments with salt, mannitol and heat induce a unique pattern of gene expression that is not a simple merge of the individual stress responses. Under multiple stress conditions, expression of most heat and salt stress-responsive genes increased to levels similar to or higher than those measured in single stress conditions, but osmotic stress-responsive genes increased to lower levels. Genes up-regulated to higher levels under multiple stress condition than single stress conditions include genes for heat shock proteins, heat shock regulators and late embryogenesis abundant proteins (LEAs), which protect other proteins from damage caused by stresses, suggesting their importance in multiple stress condition. Based on this analysis, we identify candidate genes for engineering crop plants tolerant to multiple stresses. Key-words: Arabidopsis; abiotic stress; heat; high salinity; microarray; multiple stresses; osmotic stress.
INTRODUCTION Abiotic stresses reduce crop productivity by an estimated 50% worldwide (Bray et al. 2000; reviewed by Kreps et al. 2002). Transcriptome analysis of plants under abiotic stresses can identify the key genes that coordinate and orchestrate plant response to these stresses, and help unravel the signalling events that determine plant stress tolerance in the natural environment. A number of global transcriptome analyses have examined various environmental stresses (biotic and abiotic), mainly in Arabidopsis thaliana. Schenk et al. (2000) analysed the expression of 2375 selected genes in Arabidopsis infected with fungus or treated with defence-related molecules. They identified a group of 169 mRNAs that were coordinately expressed in response to different treatments. Similarly, Seki Correspondence: M. Ohme-Takagi. e-mail: [email protected] 2024
et al. (2002) identified 22 genes induced in common by cold, drought and NaCl treatments. Kreps et al. (2002) showed that 118 genes were up- or down-regulated by salt, osmotic and cold stresses, and Kilian et al. (2007) reported that 59 genes were induced in Arabidopsis shoots by cold, drought and UV-B light. Matsui et al. (2008) showed that 1275 new transcriptional units, which were not defined in the standard database, were up-regulated by drought, cold or salt stresses, or by abscisic acid (ABA) treatment. The genes identified as co-expressed are postulated to play important roles in the investigated abiotic stresses, possibly in crosstalk between signalling pathways, or in common responses. Hence, these co-expressed genes represent promising candidates for engineering crop plants tolerant to multiple stresses. However, in natural conditions, stresses often occur together and the responses of these candidate genes in plants simultaneously subjected to multiple stresses remain unclear. Indeed, the effect of one stress may be synergistic with or antagonistic to the effect of other stresses, as individual stress factors modulate the plant’s responses to other stresses. For instance, cold stress antagonizes the effect of osmotic stress on the induction of the dehydration-responsive gene RD29A (Xiong et al. 1999). In addition, salicylic acid (SA) suppresses ABA-mediated abiotic stress responses, and ABA-mediated responses to abiotic stresses such as salt treatment antagonize SA-mediated systemic-acquired resistance (Yasuda et al. 2008). Similarly, ABA antagonizes jasmonate-ethylene signalling pathways and modulates defence gene expression and disease resistance in Arabidopsis (Anderson et al. 2004). Many of the genes induced by auxin are down-regulated by wounding stress (Cheong et al. 2002). In natural habitats, plants are usually subjected to more complex combinations of stress factors. These stress factors may include extreme temperatures, drought, salinity, heavy metals, UV light, pathogens and others. Especially in arid and semi-arid areas, which represent about one-third of the earth’s land (Huang et al. 2010), high evaporation rates lead to water scarcity and accumulation of salts in the surface layers of the soil. Accordingly, plants in these environments experience multiple stresses – high temperature, drought and salinity – all at once. Few studies have considered the effect of a combination of stress factors on gene expression on a whole-genome basis. For example, the combined effects of drought and heat on © 2014 John Wiley & Sons Ltd
Plant responses to multiple abiotic stresses gene expression have been investigated in tobacco (Rizhsky et al. 2002) and Arabidopsis (Rizhsky et al. 2004). Another study compared the transcriptome responses of six pairs of stresses, including cold, heat, excess light, salt and flagellin treatments (Rasmussen et al. 2013). This study found that salt and heat generally produce opposite responses. Prasch & Sonnewald (2013) measured gene expression by microarray in plants subjected to the triple stress of heat, drought and virus. They showed that heat is the major stress factor and that the combination of the abiotic stresses largely abolished the defence response induced by virus treatment. Here, we take a further step towards understanding plant responses to complex environmental conditions. Using microarrays, we examined the global changes in gene expression of Arabidopsis plants subjected to individual and combined salt, osmotic and heat stress treatments, the components of canonical arid environments. Our analysis showed that plant response to multiple stress is not a simple additive effect of the individual treatments; rather, the simultaneous application of multiple stresses produces a unique pattern of gene expression, distinct from those observed in previous studies.
heat treatment, plants were treated with salt and mannitol for 12 h and then transferred to a growth cabinet pre-adjusted to 35 °C for 4 h (Fig. 1b). To replicate field conditions as much as possible, the Arabidopsis plants were treated with salt and mannitol for 1 h before the light was turned off, kept in the dark for 8 h and then 3 h after turning the light back on the plant trays were transferred to a 35 °C growth cabinet for 4 h to give a total of 16 h treatment (Fig. 1b).
RNA extraction, microarray experiment and data analysis Rosette leaves were harvested and grounded in liquid nitrogen. Total RNA was extracted using the RNeasy plant kit (Qiagen, Hilden, Germany) for microarray and real-time RT-PCR experiments. The microarray experiments were performed using Agilent Arabidopsis (V4, 4 × 44k; Santa Clara, CA, USA) microarrays according to the manufacturer’s instructions. Three biological replicates were tested with a one-colour method. In each case, 1 μg of total RNA was used as a starting material. Spot signal values were calculated with the Feature Extraction version 9.1 software supplied by Agilent. We defined the QC value as 1 when a spot passed the ‘FeatNonUnifOL’ filter and 2 when the spot further passed the ‘FeatPopnOL’ filter, and the detection value as 1 when a spot passed the ‘IsPosAndSignif’ filter and 2 when the spot further passed the ‘IsWellAboveBG’ filter. All signal values were divided by the median value of spots with QC = 2 and detection value = 2. Spot-to-gene conversion was accomplished based on a table provided by The Arabidopsis Information Resourse (TAIR) (ftp://ftp.arabidopsis.org/home/ tair/Microarrays/Agilent/agilent_array_elements-2010-12-20 .txt). For genes corresponding to two or more probes, the average values were used. We further applied quantile normalization using all microarray data analysed in our group to make the signal distribution of all experiments the same. The P-value of each gene, compared with data obtained from control plants, was calculated by Dunnett’s test. To reduce
MATERIALS AND METHODS Growth and treatments of plants Wild type Col-0 A. thaliana seeds were sown on soil, and stratified at 4 °C for 2 d. Ten-day-old seedlings were transplanted to new soil. After another 10 d, the plants were treated as follows. For salt stress treatment, plants were watered with 150 or 300 mm NaCl 16 h before sampling. For osmotic stress treatment, plants were watered with 200 or 400 mm mannitol 16 h before sampling. For heat treatment, plant trays were transferred to a growth cabinet pre-adjusted to 35 °C for 4 h before sampling. For multiple stress treatments, plants were watered with a solution containing final concentrations of 150 mm NaCl and 200 mm mannitol, or 300 mm NaCl and 400 mm mannitol. To combine this with
(a)
2025
Sampling
( b) 1 h light
7 h dark
8 h light Control
Salt stress (NaCl 150 or 300 m M ) OsmoƟc stress (mannitol 200 or 400 m M ) Control
Heat 35 °C
NaCl 150 mM
Heat stress 35 °C, 4 h
NaCl 300 m M NaCl 150 m M + Mannitol 200 m M +
Mannitol 200 m M
Mannitol 400 mM
MulƟple low
Heat stress 35 °C, 4 h
MulƟple high
Figure 1. Stress treatments of plants. (a) Phenotypes of wild-type Arabidopsis plants treated with individual and multiple stresses. Photos were taken 5 d after treatment. ‘Multiple low’ represents 150 mm NaCl, 200 mm mannitol and 35 °C. ‘Multiple high’ represents 300 mm NaCl, 400 mm mannitol and 35 °C. (b) A schematic diagram explaining the treatment and sampling procedures. © 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
2026
N. Sewelam et al.
type I family-wise errors, we adjusted the P-values using the Benjamini–Hochberg method (Benjamini & Hochberg 1995) and selected differentially expressed genes (DEGs) as the genes up-/down-regulated in at least one condition to more/ less than ±onefold in ln value with adjusted P-value less than 0.05. All our microarray data were deposited in NCBI Gene Expression Omnibus (GEO) under the accession number GSE39956. Principal component analysis (PCA), proportion tests and binomial tests were performed by ‘prcomp’, ‘prop’ and ‘binom’, respectively, of the R statistical suit (R Development Core Team 2008). Hierarchical and k-means clustering was performed by Cluster 3.0 software (de Hoon et al. 2004). Distance of gene expression was measured by the ‘uncentred correlation’ value in the Cluster 3.0 software.The pre-defined number of clusters in k-means clustering analysis was determined to minimize the number of similar clusters but to ensure that each cluster has clear characteristics by changing the pre-defined number from 10 to 20.
Real-time RT-PCR qRT-PCR was performed using the specific primers listed in Supporting Information Table S1. First-strand cDNA was synthesized using the PrimeScript RT reagent kit (TakaraBio Inc., Ohtsu, Japan). For qRT-PCR, the PP2AA3 gene (AT1G13320) was used as an internal control and was amplified with the primer pair: forward 5′-GTTCCAAACTCT TACCTGCGGTAA-3′ and reverse 5′-TACTCTCCAGT GCCTGTCTTCA-3′. All experiments were executed in biological triplicate.
RESULTS AND DISCUSSION Experimental design to mimic arid conditions in the field To compare gene expression in plants treated with individual stresses and those treated with combined salinity, osmotic and heat stress, we treated plants with 150 mm NaCl, 200 mm mannitol and 35 °C heat individually or together. Arabidopsis plants were treated with NaCl and/or mannitol for 16 h before sampling. Heat stress was added to the treatment 4 h before sampling (Fig. 1b). We selected 16 h as the optimal timing for salt and mannitol treatments because plant responses to these treatments peak between 12 and 24 h. Similarly, we selected 4 h for heat stress because heat responses peak 2–4 h after treatment at 38 °C (Kilian et al. 2007). Multiple stress conditions of salinity, osmotic and heat stress were designed to mimic arid areas in the summer, where high temperatures produce high evaporation rates, leading to water scarcity and accumulation of salts in the surface layers of the soil. We also selected our conditions to accurately represent stress responses, rather than more severe effects such as cell death. High stress conditions, such as 250–300 mm NaCl and 300–400 mm mannitol, are often used for analyses of global gene expression (Kreps et al. 2002; Seki et al. 2002; Kilian et al. 2007; Matsui et al. 2008; Zeller et al. 2009). However, we
did not analyse plants treated with a combination of 300 mm NaCl, 400 mm mannitol and 35 °C heat because these conditions were lethal to the plants (Fig. 1a).
Plants under multiple stresses show distinctive transcriptomic responses We used microarrays to examine gene expression in plants treated with six different stress conditions, including the multiple stress condition. We used real-time PCR to confirm the accuracy of the microarray data and found the data from both were highly consistent (Supporting Information Table S2). PCA of the microarray data indicated that the low and high salt responses and low and high osmotic stress responses have similar transcriptome profiles (Fig. 2a). However, we found that the heat stress response differed from salt and osmotic stresses; indeed, this difference highlighted the similarity between salt and osmotic stresses (Fig. 2a). We found that the transcript profile of the multiple stress condition also differed from all the individual stresses, suggesting its distinctive nature (Fig. 2a). Correlation coefficient (r) analysis also showed the similarity between low and high concentrations of salt and osmotic stress. The correlation between high and low salt (0.5) was lower than between high and low osmotic stresses (0.64), as indicated by the PCA (Fig. 2a,b). By contrast, the heat and other stresses showed relatively low correlation coefficients, supporting the isolation of heat from other stresses (Fig. 2b). However, in contrast to the PCA analysis, the correlation analysis found a high correlation between the heat and multiple stress treatments (r = 0.74), higher than the correlation between low and high osmotic stresses (r = 0.64). We investigated whether the similarities and differences between single and multiple stress conditions are consistent with previously published work focusing on multiple stresses. Because we employed the Agilent microarray, we compared our data mainly to the data of Prasch & Sonnewald (2013) using the Agilent microarray, rather than the data by Rasmussen et al. (2013) using the NimbleGen (Madison, WI) microarray. We performed PCA and correlation coefficient analysis using the microarray data for mild heat, drought, heat and drought stress, and severe heat (Prasch & Sonnewald 2013), because we expected drought stress to induce a similar response to osmotic stress with mannitol. However, we found that the PCA analysis highlighted the difference between the experiments performed by Prasch et al. and our experiments, rather than the similarity/difference between each stress (Supporting Information Fig. S1a). Correlation coefficients between each transcriptome also showed low correlation between our data and their data, except for the similarity between our multiple stress and heat and their severe heat conditions (0.48 and 0.42, respectively) (Supporting Information Fig. S1b,c). However, we did find similar induction of expression of representative genes responsive to heat, osmotic stress and hyperosmotic salinity, as defined by Gene Ontology (GO), in our experiment and Prasch et al. (Supporting Information Fig. S2), suggesting that we can
© 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
Plant responses to multiple abiotic stresses (a)
2027
(c)
PC2
High osmo High salt PC1 –0.44
0.4
Low osmo
–0.4
–0.38
(22%)
(b)
0.74 0.33 0.24 0.5 High salt
0.24
0.32 Mult iple
0.6
0.43
60
37
108
560 15
(36%) 192
105 287 319
Low osmo
High osmo
200 High osmo
0.64
Heat
High osmo
0.49
Heat
Multiple
331 475 244
0.6 0.36
496
(59%) (66%)
0.5 Low salt
61
Low osmo
Low osmo
0.24
(42%)
56 (44%)
Heat
Multiple
560 42
–0.8
High salt
Low salt
High salt
385
27
Multiple
–0.6
232 192
(53%) 22
165
5
Low salt
Heat
0.28
469 74
15
(49%)
–0.4
Multiple
(51%) 162
1 19
0 –0.36 –0.2
Up-regulated Multiple
541
(25%)
0.2
Low salt
–0.42
Down-regulated Multiple
0.6
(d)
Multiple
547 420
(52%) (57%) Low salt
Heat
High osmo
High salt
Low osmo
Figure 2. Relationships between transcriptome responses under various stress treatments. (a) The results of principal component analysis. (b) Correlation coefficients of gene expression between transcriptomes under various stress treatments. The number on each edge indicates the correlation coefficient between two samples. (c) Venn diagram of genes down-regulated or up-regulated by each stress. The number in parenthesis shown with upper or right circle indicates percentage overlap of the genes down-regulated or up-regulated by multiple stress condition with genes down-regulated or up-regulated by single stress. The number in parenthesis shown with lower or left circle indicates percentage overlap of the genes down-regulated or up-regulated by single stress with genes down-regulated or up-regulated by multiple stress. (d) Relative cumulative multiple stress inducible log-fold changes of 967 genes, which were induced by multiple stress condition, in each single stress.
© 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
2028
N. Sewelam et al.
compare these transcriptomes for these genes and both experiments reflect the genuine biological response, at least in part. To examine the difference and similarity of all these stresses in more detail, we focused on DEGs that changed more than 2.71828-fold (exponential) with an adjusted P-value < 0.05. The multiple stress treatment induced 967 genes and 51, 42 and 57% of these genes were also induced by salt, osmotic and heat stresses, respectively (Fig. 2c). The multiple stress treatments also repressed 719 genes and 25, 22 and 66% of these genes were also repressed by the salt, osmotic and heat stresses, respectively (Fig. 2c). These results suggested that heat stress made a higher contribution to the transcriptome of the multiple stress treatment, consistent with the correlation analysis. Rasmussen et al. (2013) reported the contributions of two different stresses to the response under combined stress using ‘cumulative log-fold changes of the 500 most significantly responding transcripts in the single stress experiments when the particular stress is combined with another stress in a double stress experiment’.We employed a similar approach to investigate the contributions of single stresses to the response to multiple stress and found a higher contribution of heat stress to the response to multiple stress, compared with low salt and low osmotic conditions (Fig. 2d). However, because high salt and high osmotic potential induced similar cumulative log-fold changes to heat stress, the contribution of each stress to the transcriptome response to multiple stress may be more comparable in actual field conditions. These results differed from the results of Rasmussen et al. (2013), in which salt stress made a higher contribution than heat stress in response to combined salt and heat stress.
(a)
Osmo
Salt 185
439
400
250 283
58
Multiple 452 1234 833 134
Heat
Osmo
Salt 84
149
146
70 64
64
Multiple 608 Heat
619
566 153
(b)
The response to multiple stress condition does not reflect a simple merge of the single stress responses a 3467 DEGs
The Venn diagram analysis of DEGs indicated that many of the genes induced by the single stress treatment were not fully induced by multiple stress treatment. For example, among the 2067 genes that are induced by at least one single stress treatment, only 833 (41%) are also induced by multiple stress treatment (Fig. 3a). Furthermore, among the 1185 genes that are repressed by at least one single stress treatment, only 566 (48%) are also repressed by multiple stress treatment. This indicates that under multiple stress conditions, the plant tends to repress genes that should be repressed, but can fail to activate genes that should be activated (P < 3.514e-05 by proportion test). These responses may reflect an adaptation allowing the plant to cope with many stresses, but with limited resources. We compared our results with the results of Prasch & Sonnewald (2013) and found that, in their experiment, 1873 genes of 2418 (77.5%) that were induced in either drought or heat stress were also induced in drought and heat double stress, whereas 1084 of 1299 genes (83.5%) that were repressed in either stress were also repressed in the double stress, suggesting that Prasch et al.’s data show a similar but weaker tendency (P < 1.589e-05 by proportion test), compared with our triple stress results (Fig. 3a and Supporting Information Fig. S3).
b
c
>148.416 >54.599 >20.086 >7.389 >2.718
d
2
Low salt High salt Low osmo Highosmo Heat Multiple
N. Sewelam et al.
Low salt High salt Low osmo Highosmo Heat Multiple
2030
Group 6 170 genes
Low salt High salt Low osmo Highosmo Heat Multiple
1.5
d
Plant, Cell and Environment (2014) 37, 2024–2035
doi: 10.1111/pce.12274
Original Article
A step towards understanding plant responses to multiple environmental stresses: a genome-wide study Nasser Sewelam1,2, Yoshimi Oshima1, Nobutaka Mitsuda1 & Masaru Ohme-Takagi1,3 1
Bioproduction Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8566, Japan, 2Botany Department, Faculty of Science, Tanta University, 31527 Tanta, Egypt and 3Institute for Environmental Science and Technology (IEST), Saitama University, Sakura-ku, Saitama 338-8570, Japan
ABSTRACT In natural habitats, especially in arid areas, plants are often simultaneously exposed to multiple abiotic stresses, such as salt, osmotic and heat stresses. However, most analyses of gene expression in stress responses examine individual stresses. In this report, we compare gene expression in individual and combined stresses. We show that combined stress treatments with salt, mannitol and heat induce a unique pattern of gene expression that is not a simple merge of the individual stress responses. Under multiple stress conditions, expression of most heat and salt stress-responsive genes increased to levels similar to or higher than those measured in single stress conditions, but osmotic stress-responsive genes increased to lower levels. Genes up-regulated to higher levels under multiple stress condition than single stress conditions include genes for heat shock proteins, heat shock regulators and late embryogenesis abundant proteins (LEAs), which protect other proteins from damage caused by stresses, suggesting their importance in multiple stress condition. Based on this analysis, we identify candidate genes for engineering crop plants tolerant to multiple stresses. Key-words: Arabidopsis; abiotic stress; heat; high salinity; microarray; multiple stresses; osmotic stress.
INTRODUCTION Abiotic stresses reduce crop productivity by an estimated 50% worldwide (Bray et al. 2000; reviewed by Kreps et al. 2002). Transcriptome analysis of plants under abiotic stresses can identify the key genes that coordinate and orchestrate plant response to these stresses, and help unravel the signalling events that determine plant stress tolerance in the natural environment. A number of global transcriptome analyses have examined various environmental stresses (biotic and abiotic), mainly in Arabidopsis thaliana. Schenk et al. (2000) analysed the expression of 2375 selected genes in Arabidopsis infected with fungus or treated with defence-related molecules. They identified a group of 169 mRNAs that were coordinately expressed in response to different treatments. Similarly, Seki Correspondence: M. Ohme-Takagi. e-mail: [email protected] 2024
et al. (2002) identified 22 genes induced in common by cold, drought and NaCl treatments. Kreps et al. (2002) showed that 118 genes were up- or down-regulated by salt, osmotic and cold stresses, and Kilian et al. (2007) reported that 59 genes were induced in Arabidopsis shoots by cold, drought and UV-B light. Matsui et al. (2008) showed that 1275 new transcriptional units, which were not defined in the standard database, were up-regulated by drought, cold or salt stresses, or by abscisic acid (ABA) treatment. The genes identified as co-expressed are postulated to play important roles in the investigated abiotic stresses, possibly in crosstalk between signalling pathways, or in common responses. Hence, these co-expressed genes represent promising candidates for engineering crop plants tolerant to multiple stresses. However, in natural conditions, stresses often occur together and the responses of these candidate genes in plants simultaneously subjected to multiple stresses remain unclear. Indeed, the effect of one stress may be synergistic with or antagonistic to the effect of other stresses, as individual stress factors modulate the plant’s responses to other stresses. For instance, cold stress antagonizes the effect of osmotic stress on the induction of the dehydration-responsive gene RD29A (Xiong et al. 1999). In addition, salicylic acid (SA) suppresses ABA-mediated abiotic stress responses, and ABA-mediated responses to abiotic stresses such as salt treatment antagonize SA-mediated systemic-acquired resistance (Yasuda et al. 2008). Similarly, ABA antagonizes jasmonate-ethylene signalling pathways and modulates defence gene expression and disease resistance in Arabidopsis (Anderson et al. 2004). Many of the genes induced by auxin are down-regulated by wounding stress (Cheong et al. 2002). In natural habitats, plants are usually subjected to more complex combinations of stress factors. These stress factors may include extreme temperatures, drought, salinity, heavy metals, UV light, pathogens and others. Especially in arid and semi-arid areas, which represent about one-third of the earth’s land (Huang et al. 2010), high evaporation rates lead to water scarcity and accumulation of salts in the surface layers of the soil. Accordingly, plants in these environments experience multiple stresses – high temperature, drought and salinity – all at once. Few studies have considered the effect of a combination of stress factors on gene expression on a whole-genome basis. For example, the combined effects of drought and heat on © 2014 John Wiley & Sons Ltd
Plant responses to multiple abiotic stresses gene expression have been investigated in tobacco (Rizhsky et al. 2002) and Arabidopsis (Rizhsky et al. 2004). Another study compared the transcriptome responses of six pairs of stresses, including cold, heat, excess light, salt and flagellin treatments (Rasmussen et al. 2013). This study found that salt and heat generally produce opposite responses. Prasch & Sonnewald (2013) measured gene expression by microarray in plants subjected to the triple stress of heat, drought and virus. They showed that heat is the major stress factor and that the combination of the abiotic stresses largely abolished the defence response induced by virus treatment. Here, we take a further step towards understanding plant responses to complex environmental conditions. Using microarrays, we examined the global changes in gene expression of Arabidopsis plants subjected to individual and combined salt, osmotic and heat stress treatments, the components of canonical arid environments. Our analysis showed that plant response to multiple stress is not a simple additive effect of the individual treatments; rather, the simultaneous application of multiple stresses produces a unique pattern of gene expression, distinct from those observed in previous studies.
heat treatment, plants were treated with salt and mannitol for 12 h and then transferred to a growth cabinet pre-adjusted to 35 °C for 4 h (Fig. 1b). To replicate field conditions as much as possible, the Arabidopsis plants were treated with salt and mannitol for 1 h before the light was turned off, kept in the dark for 8 h and then 3 h after turning the light back on the plant trays were transferred to a 35 °C growth cabinet for 4 h to give a total of 16 h treatment (Fig. 1b).
RNA extraction, microarray experiment and data analysis Rosette leaves were harvested and grounded in liquid nitrogen. Total RNA was extracted using the RNeasy plant kit (Qiagen, Hilden, Germany) for microarray and real-time RT-PCR experiments. The microarray experiments were performed using Agilent Arabidopsis (V4, 4 × 44k; Santa Clara, CA, USA) microarrays according to the manufacturer’s instructions. Three biological replicates were tested with a one-colour method. In each case, 1 μg of total RNA was used as a starting material. Spot signal values were calculated with the Feature Extraction version 9.1 software supplied by Agilent. We defined the QC value as 1 when a spot passed the ‘FeatNonUnifOL’ filter and 2 when the spot further passed the ‘FeatPopnOL’ filter, and the detection value as 1 when a spot passed the ‘IsPosAndSignif’ filter and 2 when the spot further passed the ‘IsWellAboveBG’ filter. All signal values were divided by the median value of spots with QC = 2 and detection value = 2. Spot-to-gene conversion was accomplished based on a table provided by The Arabidopsis Information Resourse (TAIR) (ftp://ftp.arabidopsis.org/home/ tair/Microarrays/Agilent/agilent_array_elements-2010-12-20 .txt). For genes corresponding to two or more probes, the average values were used. We further applied quantile normalization using all microarray data analysed in our group to make the signal distribution of all experiments the same. The P-value of each gene, compared with data obtained from control plants, was calculated by Dunnett’s test. To reduce
MATERIALS AND METHODS Growth and treatments of plants Wild type Col-0 A. thaliana seeds were sown on soil, and stratified at 4 °C for 2 d. Ten-day-old seedlings were transplanted to new soil. After another 10 d, the plants were treated as follows. For salt stress treatment, plants were watered with 150 or 300 mm NaCl 16 h before sampling. For osmotic stress treatment, plants were watered with 200 or 400 mm mannitol 16 h before sampling. For heat treatment, plant trays were transferred to a growth cabinet pre-adjusted to 35 °C for 4 h before sampling. For multiple stress treatments, plants were watered with a solution containing final concentrations of 150 mm NaCl and 200 mm mannitol, or 300 mm NaCl and 400 mm mannitol. To combine this with
(a)
2025
Sampling
( b) 1 h light
7 h dark
8 h light Control
Salt stress (NaCl 150 or 300 m M ) OsmoƟc stress (mannitol 200 or 400 m M ) Control
Heat 35 °C
NaCl 150 mM
Heat stress 35 °C, 4 h
NaCl 300 m M NaCl 150 m M + Mannitol 200 m M +
Mannitol 200 m M
Mannitol 400 mM
MulƟple low
Heat stress 35 °C, 4 h
MulƟple high
Figure 1. Stress treatments of plants. (a) Phenotypes of wild-type Arabidopsis plants treated with individual and multiple stresses. Photos were taken 5 d after treatment. ‘Multiple low’ represents 150 mm NaCl, 200 mm mannitol and 35 °C. ‘Multiple high’ represents 300 mm NaCl, 400 mm mannitol and 35 °C. (b) A schematic diagram explaining the treatment and sampling procedures. © 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
2026
N. Sewelam et al.
type I family-wise errors, we adjusted the P-values using the Benjamini–Hochberg method (Benjamini & Hochberg 1995) and selected differentially expressed genes (DEGs) as the genes up-/down-regulated in at least one condition to more/ less than ±onefold in ln value with adjusted P-value less than 0.05. All our microarray data were deposited in NCBI Gene Expression Omnibus (GEO) under the accession number GSE39956. Principal component analysis (PCA), proportion tests and binomial tests were performed by ‘prcomp’, ‘prop’ and ‘binom’, respectively, of the R statistical suit (R Development Core Team 2008). Hierarchical and k-means clustering was performed by Cluster 3.0 software (de Hoon et al. 2004). Distance of gene expression was measured by the ‘uncentred correlation’ value in the Cluster 3.0 software.The pre-defined number of clusters in k-means clustering analysis was determined to minimize the number of similar clusters but to ensure that each cluster has clear characteristics by changing the pre-defined number from 10 to 20.
Real-time RT-PCR qRT-PCR was performed using the specific primers listed in Supporting Information Table S1. First-strand cDNA was synthesized using the PrimeScript RT reagent kit (TakaraBio Inc., Ohtsu, Japan). For qRT-PCR, the PP2AA3 gene (AT1G13320) was used as an internal control and was amplified with the primer pair: forward 5′-GTTCCAAACTCT TACCTGCGGTAA-3′ and reverse 5′-TACTCTCCAGT GCCTGTCTTCA-3′. All experiments were executed in biological triplicate.
RESULTS AND DISCUSSION Experimental design to mimic arid conditions in the field To compare gene expression in plants treated with individual stresses and those treated with combined salinity, osmotic and heat stress, we treated plants with 150 mm NaCl, 200 mm mannitol and 35 °C heat individually or together. Arabidopsis plants were treated with NaCl and/or mannitol for 16 h before sampling. Heat stress was added to the treatment 4 h before sampling (Fig. 1b). We selected 16 h as the optimal timing for salt and mannitol treatments because plant responses to these treatments peak between 12 and 24 h. Similarly, we selected 4 h for heat stress because heat responses peak 2–4 h after treatment at 38 °C (Kilian et al. 2007). Multiple stress conditions of salinity, osmotic and heat stress were designed to mimic arid areas in the summer, where high temperatures produce high evaporation rates, leading to water scarcity and accumulation of salts in the surface layers of the soil. We also selected our conditions to accurately represent stress responses, rather than more severe effects such as cell death. High stress conditions, such as 250–300 mm NaCl and 300–400 mm mannitol, are often used for analyses of global gene expression (Kreps et al. 2002; Seki et al. 2002; Kilian et al. 2007; Matsui et al. 2008; Zeller et al. 2009). However, we
did not analyse plants treated with a combination of 300 mm NaCl, 400 mm mannitol and 35 °C heat because these conditions were lethal to the plants (Fig. 1a).
Plants under multiple stresses show distinctive transcriptomic responses We used microarrays to examine gene expression in plants treated with six different stress conditions, including the multiple stress condition. We used real-time PCR to confirm the accuracy of the microarray data and found the data from both were highly consistent (Supporting Information Table S2). PCA of the microarray data indicated that the low and high salt responses and low and high osmotic stress responses have similar transcriptome profiles (Fig. 2a). However, we found that the heat stress response differed from salt and osmotic stresses; indeed, this difference highlighted the similarity between salt and osmotic stresses (Fig. 2a). We found that the transcript profile of the multiple stress condition also differed from all the individual stresses, suggesting its distinctive nature (Fig. 2a). Correlation coefficient (r) analysis also showed the similarity between low and high concentrations of salt and osmotic stress. The correlation between high and low salt (0.5) was lower than between high and low osmotic stresses (0.64), as indicated by the PCA (Fig. 2a,b). By contrast, the heat and other stresses showed relatively low correlation coefficients, supporting the isolation of heat from other stresses (Fig. 2b). However, in contrast to the PCA analysis, the correlation analysis found a high correlation between the heat and multiple stress treatments (r = 0.74), higher than the correlation between low and high osmotic stresses (r = 0.64). We investigated whether the similarities and differences between single and multiple stress conditions are consistent with previously published work focusing on multiple stresses. Because we employed the Agilent microarray, we compared our data mainly to the data of Prasch & Sonnewald (2013) using the Agilent microarray, rather than the data by Rasmussen et al. (2013) using the NimbleGen (Madison, WI) microarray. We performed PCA and correlation coefficient analysis using the microarray data for mild heat, drought, heat and drought stress, and severe heat (Prasch & Sonnewald 2013), because we expected drought stress to induce a similar response to osmotic stress with mannitol. However, we found that the PCA analysis highlighted the difference between the experiments performed by Prasch et al. and our experiments, rather than the similarity/difference between each stress (Supporting Information Fig. S1a). Correlation coefficients between each transcriptome also showed low correlation between our data and their data, except for the similarity between our multiple stress and heat and their severe heat conditions (0.48 and 0.42, respectively) (Supporting Information Fig. S1b,c). However, we did find similar induction of expression of representative genes responsive to heat, osmotic stress and hyperosmotic salinity, as defined by Gene Ontology (GO), in our experiment and Prasch et al. (Supporting Information Fig. S2), suggesting that we can
© 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
Plant responses to multiple abiotic stresses (a)
2027
(c)
PC2
High osmo High salt PC1 –0.44
0.4
Low osmo
–0.4
–0.38
(22%)
(b)
0.74 0.33 0.24 0.5 High salt
0.24
0.32 Mult iple
0.6
0.43
60
37
108
560 15
(36%) 192
105 287 319
Low osmo
High osmo
200 High osmo
0.64
Heat
High osmo
0.49
Heat
Multiple
331 475 244
0.6 0.36
496
(59%) (66%)
0.5 Low salt
61
Low osmo
Low osmo
0.24
(42%)
56 (44%)
Heat
Multiple
560 42
–0.8
High salt
Low salt
High salt
385
27
Multiple
–0.6
232 192
(53%) 22
165
5
Low salt
Heat
0.28
469 74
15
(49%)
–0.4
Multiple
(51%) 162
1 19
0 –0.36 –0.2
Up-regulated Multiple
541
(25%)
0.2
Low salt
–0.42
Down-regulated Multiple
0.6
(d)
Multiple
547 420
(52%) (57%) Low salt
Heat
High osmo
High salt
Low osmo
Figure 2. Relationships between transcriptome responses under various stress treatments. (a) The results of principal component analysis. (b) Correlation coefficients of gene expression between transcriptomes under various stress treatments. The number on each edge indicates the correlation coefficient between two samples. (c) Venn diagram of genes down-regulated or up-regulated by each stress. The number in parenthesis shown with upper or right circle indicates percentage overlap of the genes down-regulated or up-regulated by multiple stress condition with genes down-regulated or up-regulated by single stress. The number in parenthesis shown with lower or left circle indicates percentage overlap of the genes down-regulated or up-regulated by single stress with genes down-regulated or up-regulated by multiple stress. (d) Relative cumulative multiple stress inducible log-fold changes of 967 genes, which were induced by multiple stress condition, in each single stress.
© 2014 John Wiley & Sons Ltd, Plant, Cell and Environment, 37, 2024–2035
2028
N. Sewelam et al.
compare these transcriptomes for these genes and both experiments reflect the genuine biological response, at least in part. To examine the difference and similarity of all these stresses in more detail, we focused on DEGs that changed more than 2.71828-fold (exponential) with an adjusted P-value < 0.05. The multiple stress treatment induced 967 genes and 51, 42 and 57% of these genes were also induced by salt, osmotic and heat stresses, respectively (Fig. 2c). The multiple stress treatments also repressed 719 genes and 25, 22 and 66% of these genes were also repressed by the salt, osmotic and heat stresses, respectively (Fig. 2c). These results suggested that heat stress made a higher contribution to the transcriptome of the multiple stress treatment, consistent with the correlation analysis. Rasmussen et al. (2013) reported the contributions of two different stresses to the response under combined stress using ‘cumulative log-fold changes of the 500 most significantly responding transcripts in the single stress experiments when the particular stress is combined with another stress in a double stress experiment’.We employed a similar approach to investigate the contributions of single stresses to the response to multiple stress and found a higher contribution of heat stress to the response to multiple stress, compared with low salt and low osmotic conditions (Fig. 2d). However, because high salt and high osmotic potential induced similar cumulative log-fold changes to heat stress, the contribution of each stress to the transcriptome response to multiple stress may be more comparable in actual field conditions. These results differed from the results of Rasmussen et al. (2013), in which salt stress made a higher contribution than heat stress in response to combined salt and heat stress.
(a)
Osmo
Salt 185
439
400
250 283
58
Multiple 452 1234 833 134
Heat
Osmo
Salt 84
149
146
70 64
64
Multiple 608 Heat
619
566 153
(b)
The response to multiple stress condition does not reflect a simple merge of the single stress responses a 3467 DEGs
The Venn diagram analysis of DEGs indicated that many of the genes induced by the single stress treatment were not fully induced by multiple stress treatment. For example, among the 2067 genes that are induced by at least one single stress treatment, only 833 (41%) are also induced by multiple stress treatment (Fig. 3a). Furthermore, among the 1185 genes that are repressed by at least one single stress treatment, only 566 (48%) are also repressed by multiple stress treatment. This indicates that under multiple stress conditions, the plant tends to repress genes that should be repressed, but can fail to activate genes that should be activated (P < 3.514e-05 by proportion test). These responses may reflect an adaptation allowing the plant to cope with many stresses, but with limited resources. We compared our results with the results of Prasch & Sonnewald (2013) and found that, in their experiment, 1873 genes of 2418 (77.5%) that were induced in either drought or heat stress were also induced in drought and heat double stress, whereas 1084 of 1299 genes (83.5%) that were repressed in either stress were also repressed in the double stress, suggesting that Prasch et al.’s data show a similar but weaker tendency (P < 1.589e-05 by proportion test), compared with our triple stress results (Fig. 3a and Supporting Information Fig. S3).
b
c
>148.416 >54.599 >20.086 >7.389 >2.718
d
2
Low salt High salt Low osmo Highosmo Heat Multiple
N. Sewelam et al.
Low salt High salt Low osmo Highosmo Heat Multiple
2030
Group 6 170 genes
Low salt High salt Low osmo Highosmo Heat Multiple
1.5
d
