In order to assess specific functional roles of plant heat shock transcription factors (HSF) we conducted a transcriptome analysis of Arabidopsis thaliana hsfA1a/hsfA1b double knock out mutants and wild-type plants. We used Affymetrix ATH1 microarrays (representing more than 24 000 genes) and conducted hybridizations for heat-treated or non-heat-treated leaf material of the respective lines. Heat stress had a severe impact on the transcriptome of mutant and wild-type plants. Approximately 11% of all monitored genes of the wild type showed a significant effect upon heat stress treatment. The difference in heat stress-induced gene expression between mutant and wild type revealed a number of HsfA1a/1b-regulated genes. Besides several heat shock protein and other stress-related genes, we found HSFA-1a/1b-regulated genes for other functions including protein biosynthesis and processing, signalling, metabolism and transport. By screening the profiling data for genes in biochemical pathways in which known HSF targets were involved, we discovered that at each step in the pathway leading to osmolytes, the expression of genes is regulated by heat stress and in several cases by HSF. Our results document that in the immediate early phase of the heat shock response HSF-dependent gene expression is not limited to known stress genes, which are involved in protection from proteotoxic effects. HsfA1a and HsfA1b-regulated gene expression also affects other pathways and mechanisms dealing with a broader range of physiological adaptations to stress.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The heat shock (HS) response is characterized by a rapid reprogramming of gene expression, leading to a transient accumulation of heat shock proteins (HSP) that is correlated with enhanced thermotolerance. HSP are a hallmark of the environmental stress response and they are induced by diverse stressful conditions. Their biological role is believed to counteract cytotoxic effects of protein denaturation (Morimoto et al., 1994; Schöffl et al., 1998). In plants there is evidence for cross-protection against different environmental stresses and mRNA expression profiling has demonstrated that, for example, HS and drought induce different but partly overlapping sets of genes including a number of HSP and other common stress-related genes (Mittler et al., 2001; Pnueli et al., 2002; Rizhsky et al., 2002).
In all organisms the HS response is primarily regulated at the transcriptional level by heat stress transcription factors (HSF), which are activated by stress that leads to a specific binding to heat shock promoter elements (HSE) and subsequently to the transcription of these genes. HSF display a modular structure with a highly conserved N-terminal DNA-binding domain, and an adjacent bipartite oligomerization domain (HR-A/B) composed of hydrophobic heptad repeats. The DNA-binding domain of plant HSF is encoded in two parts separated by an intron, which is inserted immediately upstream of the coding part for the H2-T-H3 DNA-binding motif. The position of the intron is identical in all HSF, but the size is highly variable between different plant Hsf genes (Nover et al., 2001).
In Arabidopsis thaliana 21 different HSF genes have been identified (Nover et al., 2001) which are assigned to three major classes A, B and C (Nover et al., 1996, 2001). The basic structural differences between class A and B HSF have been conserved during evolution. Most plant HSF known to date can be classified as group A or B factors and subgroups therein, as exemplified for different HSF genes of Lycopersicon esculentum, Oryza sativa, Zea mays, Pisum sativum, Nicotiana tabacum, Medicago sativa and Glycine max (Nover et al., 2001). The HR-A/B regions of class B HSF are similar to all non-plant HSF, whereas all class A and class C HSF have an extended HR-A/B region due to an insertion of 21 (class A) and seven (class C) amino acid residues between the hydrophobic parts A and B. Class C HSF show sequence variation in the DNA-binding domain. There is only one member of this class present in the Arabidopsis genome but there is evidence for class C HSF in other plant species (Nover et al., 2001).
Another structural motif (aromatic, hydrophobic, acidic amino acids) present in the C-terminal domain, was found to be crucial for the activator function of class A HSF (Döring et al., 2000), but this motif has not been identified in class B factors. The functional consequences of the structural differences between the plant class A and B HSF are still not fully understood. In contrast to class A HSF, the transient overexpression of several class B HSF was not sufficient for activation of HS promoters in tobacco protoplasts (Czarnecka-Verner et al., 1997, 2000) and class B HSF failed to rescue the yeast hsf1 mutation (Boscheinen et al., 1997). There is evidence that certain class B HSF may act as coactivators for enhanced transcription of viral and housekeeping genes during ongoing heat stress, as shown for tomato HsfB1 (Bharti et al., 2004). Another remarkable observation, thus far limited to the plant HSF family, is the heat-induced expression of several HSF genes, which indicates a multistep mechanism for the involvement of different HSF in the HS response (Nover et al., 1996; Prändl et al., 1998). The high number of class A HSF (15 members) in Arabidopsis is still puzzling but the analysis of gene knock out mutants showed that functional diversification and genetic redundancy of HSF have evolved in plant systems (Lohmann et al., 2004).
The role of Arabidopsis class A HSF as positive regulators of HS gene expression was implicated by transgenic overexpression of HsfA1a or HsfA1b (originally described as Hsf1 and Hsf3 by Hübel and Schöffl, 1994 and Prändl et al., 1998 respectively) constructs in Arabidopsis, which resulted in a constitutive expression of HSP and enhanced levels of thermotolerance (Lee et al., 1995; Prändl et al., 1998). There is evidence that not only HSP genes are controlled by HSF, but other genes encoding key enzymes in biochemical pathways that are related to environmental responses have also been identified as targets of HSF regulation in HsfA1b-transgenic plants (Panchuk et al., 2002; Panikulangara et al., 2004).
Interestingly, loss-of-function mutants of either HsfA1a or HsfA1b genes had no detectable effects on the HS response, only hsfA1a/hsfA1b (hsf1/3) double mutants were significantly impaired in the early, transient mRNA accumulation of a number of HSP genes during the first hour of HS (Lohmann et al., 2004). This deficiency correlated with the absence of heat-induced high molecular weight HSE-binding complexes and thus HsfA1a and HsfA1b are considered key regulators not only for the fast and strong but also transient HS-induced transcription. The loss of HsfA1a/1b in double knock out plants causes a slightly compromised thermotolerance phenotype that was only detected after a severe heat pulse (Lohmann et al., 2004).
The identification of the complete set of HsfA1a/1b target genes is crucial for determining the functional roles and biological importance of HSF and for understanding the molecular mechanism involved in the generation of stress tolerance in Arabidopsis. Using the Affymetrix ATH1 microarray expression profiling, we were able to discriminate at a very early time point in the HS response (1 h HS), between directly recognized HsfA1a/1b target genes, and other HS-induced genes that may be regulated by other factors or secondary effects of HS. Our data show that only a small fraction of the large total number of HS affected genes is potentially controlled by HsfA1a/1b. This set includes several but not all of the well-known HSP genes, other HSF genes, but also a number of genes that were not known to be controlled by HSF. The putative functions of many of the new HSF target genes are related to protective environmental stress responses. These targets include several genes encoding enzymes catalysing the synthesis of metabolites and products of the Raffinose Family Oligosaccharide (RFO) pathway, which are involved in drought/desiccation tolerance. This pathway in carbohydrate metabolism thus appears to be regulated at each step by heat stress and HSF respectively.
Effects of heat shock and hsfA1a/1b mutations on gene expression
The aim of the expression profiling was to define and to apply criteria that allow discrimination between the effects of HS and HSF-dependent gene expression in Arabidopsis. Therefore we determined and analysed the expression profiles of wild type (WT) and homozygous double knock out mutant plants (abbreviated hsfA1a/1b). HsfA1a/1b plants are deficient in HsfA1a and HsfA1b activity and unable to induce the fast transient expression of mRNAs of a number of HS genes tested (Lohmann et al., 2004).
By conducting a Welch's t-test on the normalized hybridization data of the two control mRNA samples (not subjected to HS) we examined whether HsfA1a and HsfA1b exert an influence on basal gene expression. We compared the data sets of genes expressed at control temperature in WT versus hsfA1a/1b plants. Only 11 genes were differentially expressed by a factor of >2. Interestingly, nine of these genes are connected to HS and related to environmental stress responses (Table 1), two genes are as yet unidentified. These data indicate that at normal temperature (i) the expression levels of >99.95% of all genes represented on the chip are identical in WT and hsfA1a/1b plants, and (ii) the low level of basal expression of several stress genes in WT is positively affected by HsfA1a/1b.
Table 1. Differentially expressed genes at control temperature in hsfA1a/1b plants versus WT
*Numbers represent fold-changes of expression estimates of genes for control treatment in WT compared with hsfA1a/1b plants.
Stress-induced protein sti1-like protein
Putative heat shock protein
Photosystem D1 carboxy-terminal protease
The effect of HS on gene expression was analysed after 1 h heat treatment (37°C) of leaf tissue. This protocol was chosen for two reasons: (i) in WT the transient mRNA levels of HS genes (known HSP and HSF) show a maximum after 1 h HS; (ii) in hsfA1a/1b plants, compared with WT, the difference of the heat-induced mRNA levels of the same genes is maximal at this time point (Lohmann et al., 2004). It should be noted that longer duration of HS adjusted the observed mRNA in the hsfA1a/1b mutant to about the same levels as in WT plants (Lohmann et al., 2004). Hence 1 h HS treatment was the sole condition that allowed a clear discrimination of HS versus HSF-dependent expression of genes.
The data set of genes, which were differentially expressed in response to HS in the WT, was obtained by filtering on a fold-change greater than 2 and performing a Welch's t-test (confidence 95%, Multiple Testing Correction Benjamini Hochberg) on expression estimates (chips hybridized to RNA from non-stressed leaves and the chips hybridized to heat-stressed leaves) on that data set. In a similar analysis we performed chip hybridizations with RNA from the hsfA1a/1b mutant. Microarray data discussed here have been deposited with the ArrayExpress at the EMBL-EBI (http://www.ebi.ac.uk/arrayexpress/).
For a large number of genes the Affymetrix chip hybridization revealed significant changes in the mRNA levels. Using a cut-off of 2 the transcripts of a total of 2567 genes were altered in their expression in WT plants: 985 of them were elevated and 1582 were suppressed (Figures 1 and 2b). In hsfA1a/1b plants HS changed the expression of 3056 genes: 1526 were upregulated and 1530 downregulated). Although a large fraction of HS-regulated genes (2163 genes) overlaps between WT and hsfA1a/1b plants, a relatively high number of genes is affected only in the WT (142 upregulated, 262 downregulated) or in the mutant (683 upregulated, 210 downregulated) respectively (Figure 1). The complete data set of the analysis of HS and HSF dependently upregulated expression is presented in Table S1.
By plotting and comparing the expression estimates of chip replicates (Schmid et al., 2003), the results indicate satisfying reproducibility. Figure 2(a) shows that the expression estimates of most genes vary by less than a factor 2 between replicates of WT control RNAs. This is particularly the case for HS-regulated genes (dark spots).
In a second comparative analysis a t-test was applied for a data set comprising the twofold differences in RNA expression levels of heat-stressed WT versus heat-stressed hsfA1a/1b leaves. Figure 2(c) demonstrates the good reproducibility of replicates: the expression estimates of most genes (grey spots) vary by less than a factor of 2. By this criterion the differentially expressed genes are defined as potentially direct targets of HsfA1a/1b. The scattering of the expression levels of these genes (dark spots, Figure 2c) indicates the robustness of the analysis. A total of 112 genes showed differential expression levels (Figure 2d, dark spots); 105 genes showed significantly higher levels, only seven genes showed lower levels in WT compared with hsfA1a/1b plants. Thus only a relatively small fraction (4%) of the 2567 HS-affected genes can be attributed to the functions of HsfA1a/1b in WT. The majority (94%) of the 112 HSF-dependent genes (lower expression levels in hsfA1a/1b plants) requires HsfA1a/1b for the fast HS induction of expression levels in WT (see Table S2). The transcript levels of only seven (6%) HSF-dependent genes were negatively affected (expressed at a higher level in hsfA1a/1b plants) by the presence of HsfA1a/1b in WT. Eleven of the positively regulated HSF-dependent HS genes are directly linked to the HS response (HSP and HSF genes), the other 94 genes could be categorized into different functional classes including other chaperones and stress proteins, proteins of the degradation pathways, enzymes of carbohydrate metabolism, membrane transporter, transcription factors, and signalling components (Table 2). According to the present annotation approximately one-third of these genes is linked to stress-related functions, one-third are unidentified, and the putative functions of the other one-third of HSF-dependent genes are not directly linked to stress responses.
Table 2. Functional categories of HsfA1a/1b-dependent HS-genes
In order to determine whether HsfA1a/1b-dependent genes are randomly distributed among the large group of HS-affected genes in WT, we determined the ratio of HS/HSF-dependent genes in different ranking intervals (not shown).
Most of the HSF-dependent genes (84 of 88) are found within the top 500 of HS-induced genes of WT Arabidopsis. Most striking is the accumulation within the top 10 ranked genes (by fold-change in WT): 70% are regulated by HsfA1a/1b (see Tables S1 and S2). This indicates a higher probability of strongly heat-induced genes (high fold-change) being HSF-dependent targets. This is clearly demonstrated by the clustering of HsfA1a/1b-dependent genes within the area of strongly heat-induced genes in plots representing the expression estimates of WT control versus heat-treated WT (Figure 2e, dark spots).
Potential HSF-binding sites in the promoter regions of HsfA1a/1b-regulated genes
We searched for the occurrence of the consensus DNA-binding motif of HSF (HSE: nGAAnnTTCn) in the putative promoter regions (within 1000 bp upstream of the coding region) of Arabidopsis genes. HSE was present in 7611 (33%) of the 22 810 genes, which had been analysed by expression profiling. In the group of HS-regulated genes (2567 genes, see Table S1) HSE was present in the promoter regions of 958 (37%) genes, whereas 53 (47%) of the 112 HsfA1a/1b-dependent genes contained HSE elements. These data indicate that the proportion of HSE-containing putative promoter regions increases only slightly for HS-regulated genes, but almost half of the Hsf1a/1b-dependent genes contain perfect binding sites for HSF in the putative promoter region. The promoter regions of all other HsfA1a/1b-dependent genes (51 available in TAIR) contain multiple copies of variant HSE as defined by Nover et al. (2001).
Heat stress and HsfA1a/1b-dependent expression profiles of sHSP family
In order to determine the function of HsfA1a/1b in the heat-inducible expression of known HS genes, we analysed the expression of all members of the complex family of sHSP genes (Figure 3a) in Arabidopsis (Scharf et al., 2001), which were present on the chip. In WT 16 of 18 sHSP genes [17.6C-CI, 17.7-CII, 17.6A-CI, 17.4-CI, 23.6-M, 17.6-CII, 17.6B-CI, 18.5-CI(r), 17.4-CIII, 25.3-P, 15.7-CI(r), 18.1-CI, 23.5-M, 22.0-ER, 26.5-P(r), 21.7-CI(r)] showed elevated, one gene [Hsp15.4-Cl(r)] showed 10-fold reduced, and one gene [Hsp14.2-P(r)] unaffected expression upon HS. The strongest heat induction (361-fold change) was observed for Hsp17.6C-CI.
When comparing the HS-dependent expression levels of sHSP genes between WT and hsfA1a/1b plants, six genes [Hsp26.5-P(r), -15.7-CI(r), -18.1-CI, -22.0-ER, -25.3-P, -23.6-M] are expressed at a lower level in hsfA1a/1b plants (Figure 3b). These genes show within the sHSP family the second to seventh highest fold-changes of HS induction (see Table S1), which is in accordance with the overall high expression levels of HSF-dependent genes (see Figure 2e).
Comparing the mRNA levels of sHSP at normal temperature (control) between WT and hsfA1a/1b plants, 17 genes show approximately the same expression levels in both lines, but one gene (Hsp17.6A-Cl) is expressed at a significantly higher level in WT (Figure 3c). Interestingly, the mRNA level of this gene is HS-induced in WT and mutant plants without significant differences (Figure 3b). This indicates that HsfA1a/1b factors are not only involved in the HS-induction of expression of a number of genes, but also seem to affect the basal expression of certain other HS genes.
HsfA1a/1b-dependent expression of HSF genes
Previous investigations have shown that the expression of certain Arabidopsis HSF genes is constitutive (Hübel and Schöffl, 1994; Prändl et al., 1998), but others showed increased mRNA levels after HS (Prändl et al., 1998), which were dependent on HSF (Wunderlich et al., 2003), particularly on HsfA1a/1b (Lohmann et al., 2004). In the present analysis we determined the expression profiles of all 21 Arabidopsis HSF genes (Figure 4). The mRNA levels of six HSF (HsfA2, -B1, -A4a, -B2a, -B2b, -A7a) are significantly increased after HS in WT, the mRNA levels of the other HSF are not significantly different under control and HS conditions (Figure 4a). By comparing the HS-dependent expression profiles of WT versus hsfA1a/1b plants (Figure 4b) it is evident that the mRNAs of three HSF (AtHsfB1, -B2a, -A7a) are less strongly induced in hsfA1a/1b plants – the criterion for HsfA1a/1b-dependent genes. HsfA2, -A4a and -B2b, which are HS-induced HSF genes in WT, show similar expression levels in hsfA1a/1b plants and therefore are not dependent on the functions of HsfA1a/1b.
Not unexpected are the nearly unchanged mRNA levels of all HSF in WT and hsfA1a/1b plants at normal temperature (Figure 4c). At control temperature the expression levels of the three potentially HsfA1a/1b-regulated HSF genes (HsfB1, -B2a, -A7a) in hsfA1a/1b plants are not significantly different from WT. Expression profiles at normal temperature show that the mRNA levels of different HSF vary considerably. The highest levels are shown for HsfB1, which is one of the HsfA1a/1b-dependent class B HSF. On the contrary, HsfA1a and HsfA1b genes, which have been inactivated by T-DNA insertions in the hsfA1a/1b double knock out plants, are constitutively expressed at relatively low levels in WT. In hsfA1a/1b plants the chip hybridization indicates the presence of transcripts of both genes; however, due to the T-DNA insertions the mRNAs are truncated and cannot generate functional DNA-binding activities of HsfA1a and HsfA1b respectively (Lohmann et al., 2004).
In order to address the question of reliability of differences in expression levels we re-examined the mRNA levels of a number of differentially expressed genes using qRT-PCR. The HS genes tested (Hsp17.7-CII, Hsp70, Hsp83.1, Hsp101, HsfB1, HsfB2b) show clearly heat-inducible transcript levels by both methods. The HSF genes HsfA1a and HsfA1, two low-level constitutively expressed genes, show insignificant changes in mRNA levels after HS in WT. These results (Table 3) confirmed the chip data but, as expected for strongly HS-induced genes (high fold-changes of expression), there was a larger discrepancy between qRT-PCR and chip hybridization data than for genes showing lower fold-changes. The heat inducibility of mRNA is confirmed for all tested genes, however, there are discrepancies in the detection of differentially expressed HsfA1a/1b-dependent genes, between qRT-PCR (Lohmann et al., 2004) and chip hybridizations (this paper). Some of the expression differences detected by qRT-PCR were not significant in chip hybridization analysis, although fold-changes were the same. These differences may be due to methodological differences of the analyses. In the analyses of chip hybridizations we used a cut-off of 2 and a t-test for significance, whereas in qRT-PCR analysis the significance was based on the evaluation of standard deviation (Lohmann et al., 2004).
Table 3. Comparison of gene expression levels determined by chip hybridization analysis and qRT-PCR
amRNA level determined by quantitative PCR. Relative amounts were calculated and normalized with respect to Act2 mRNA (=100%). Data represent mean (n = 2).
bExpression estimates based on the chip analysis, data represent RMA normalized and averaged intensity values of the replicates relative to signal of Act2 (=100%).
HS and HSF-dependent control of osmolyte synthesis pathway
Galactinol synthase (GolS) genes are important for stress-induced RFO synthesis and generation of drought tolerance of Arabidopsis (Taji et al., 2002). RFO are potential osmo-protective substances (osmolytes). Recently, GolS1 was identified as a novel HS gene, which was induced by HS and its expression correlated with the accumulation of RFO in leaf tissue (Panikulangara et al., 2004). In our analysis GolS1 appeared as the highest scoring (fold-induction rank 10) unconventional HS gene (see Table S1). It is also a highly ranked (rank 23, see Table S2) HSF-dependent gene: the HS-induced mRNA level is strongly reduced (by a factor of 7.5) in hsfA1a/1b plants compared with WT. Of the seven galactinol synthase genes (GolS1-S7) in Arabidopsis (Taji et al., 2002), the transcript levels of five genes are unaffected by HS. Only GolS2 and GolS1 showed induction upon HS by factors of 11 and 123 respectively. The expression level of GolS1 is clearly dependent on HsfA1a/1b as indicated by the enormous reduction in transcript levels (25-fold less) in hsfA1a/1b knock out plants. The heat-induced expression of GolS1 was 22 times stronger than that of GolS2 and 100 times stronger than any other GolS gene at normal temperature.
In order to examine whether, besides GolS1, other genes/enzymes of the galactinol/RFO pathway are controlled by HS and/or HsfA1a/1b, we analysed the expression of several gene families, which encode potential biosynthetic enzymes (Figure 5). Only those genes were monitored which are, annotated by TIGR, likely to have corresponding GO function in this process (Table 4). Enzymes upstream of GolS are myo-inositol-1-phosphate-synthase (GO: 0004512) and inositol-monophosphatase (GO: 0008441), which catalyse conversion of d-glucose-6-phosphate to 1l-myo-inositol-1-phosphate and myo-inositol respectively. Of three genes representing the myo-inositol-1-phosphate-synthase family, two are heat-inducible in WT, one of them (At2g22240) is negatively affected in hsfA1a/1b mutant plants (Table 4: factor 10.7 signal ratio) and thus is a clear target of HsfA1a/1b-dependent regulation.
Table 4. HS and HsfA1a/1b-dependent genes of the galactinol/RFO pathway
aFold-change HS induction in WT; bold figures indicate significant changes.
bRatio of HS-signal intensities; bold figures indicate significant values for HsfA1a/1b-dependent changes.
Five genes were identified as potential inositol-monophosphatases, only one of them is heat-inducible (At3g02870) but its expression is not depending on HsfA1a/1b (Table 4).
The second substrate required for galactinol synthesis is UDP-d-galactose. Its formation from d-glucose-1-phosphate through conversion to UDP-d-glucose is catalysed by UDP-d-glucose-pyrophosphorylase (GO: 0003983) and respectively UDP-d-galactose by UDP-d-glucose-4′-epimerase (GO: 0003978).
Two genes were found for UDP-d-glucose-pyrophosphorylase, one of them (At5g17310) is induced by HS in WT and negatively affected in hsfA1a/1b plants (Table 4). One gene (At2g33590) encoding a predicted UDP-d-glucose-4′-epimerase represents one of the top ranked HsfA1a/1b-dependently regulated genes. The expression of five other genes for potential UDP-d-glucose-4′-epimerases (identified by GO-annotation GO: 0003978) was not changed upon HS (Table 4).
These data show that not only one single key regulator (galactinol synthase) of the RFO synthesis pathway is induced by HS/HSF. The expression of at least one member in each family of genes encoding ‘upstream enzymes’ of the pathway is induced by HS and several of these genes are regulated via HsfA1a/1b.
Expression of transcription factor families
Several types of transcription factors, in particular the WRKY transcription factor superfamily (74 members; Eulgem et al., 2000) and the MYB family (133 members; Stracke et al., 2001) are involved in stress response regulation in plants. Representatives of other families, for example scarecrow (GRAS family: Pysh et al., 1999), Dof (zinc finger C2C2-DOF family: Lijavetzky et al., 2003) and AP2/EREBP (Riechmann and Meyerowitz, 1998) were identified as HSFA1a/1b-dependent genes (Table 2). In order to identify HS and/or HSF as regulators of expression we analysed the genes of these families in comparison with two other large multigene families, bHLH (Toledo-Ortiz et al., 2002) and bZIP (Schindler et al., 1992) transcription factors (Table 5). The data show that the fractions of HS-regulated genes vary between 4% for MYB and 21% for Ap2-EREBP; the largest proportion is observed for the HSF family (28%). Interestingly, the expression of the majority of these genes is downregulated in all families, with the exception of the HSF family, where all six genes are upregulated upon HS.
Table 5. HS and HsfA1a/1b-dependent genes of selected transcription factor families
Transcription factor family
No. genes in Arabidopsis
No. genes analysed
No. HS-regulated genes in WTa
No. HsfA1a/1b-dependent genesb
aPercentage of analysed genes; ↓ upregulated upon HS; ↑ downregulated upon HS.
b↓ lower transcript level, ↑ higher transcript level in hsfA1a plants upon HS.
6 (28%); ↓0,↑6
4 (12,5%); ↓3,↑1
5 (16%); ↓3,↑2
9 (15%); ↓8,↑1
6 (6,5%); ↓5,↑1
5 (4%); ↓5↑0
27 (21%); ↓15,↑12
8 (7%); ↓7,↑1
HSF-dependent genes, which show a lower expression after HS in hsfA1a/1b plants compared with WT, were detected for WRKY, GRAS, C2C2-DOF, AP2-EREBP genes, only one in each family. By contrast the HS-regulated expression of three of six genes of the HSF family is HsfA1a/1b-dependent.
The goal of the present paper was the identification of novel HsfA1a/1b-dependent genes and pathways, which contribute, besides the well-known HSP chaperons, to the development of stress tolerance in Arabidopsis. We performed an analysis of the transcriptomes of heat-stressed cells of WT and Arabidopsis hsfA1a/1b double knock out plants. In our study it was possible to monitor a large fraction of genes of the genome of A. thaliana for very early effects of environmentally induced changes on gene expression that are controlled by the function of two related transcription factors.
The total numbers of differentially expressed genes (control versus HS) in WT (2567) and hsfA1a/1b plants (3056) are relatively high but not all up- or down-regulated genes are targets of HSF-dependent transcriptional regulation.
The comparison of expression levels between microarray and qRT-PCR quantifications confirmed that all HS genes tested show clearly heat-inducible transcript levels by both methods. The microarray expression data of the sHSP genes Hsp17.6A and Hsp18.1 are in accordance with qRT-PCR measurements presented by Lohmann et al. (2004). The highest-ranking unconventional HS gene is GolS1; its mRNA is 122-fold induced by HS in WT and it has also a high score as a HsfA1a/1b-dependent gene. The GolS1 mRNA levels of WT and hsfA1a/1b plants were confirmed by Northern hybridizations and qRT-PCR (Panikulangara et al., 2004). The general comparison of microarray and qRT-PCR quantification data indicates that there is a higher probability for genes to score as a differentially expressed gene in RT-PCR analysis than in chip hybridizations. This is a consequence of the stringent parameters, which we used to minimize the detection of false positives in chip expression profiling. Hence the number of differentially expressed genes detected by our chip hybridizations represent only a minimal but reliable set of genes regulated by HS and HSF respectively.
Microarray expression analysis of long-term HS treatment (6 h, 37°C) of Arabidopsis WT plants resulted in approximately 10-fold lower numbers (262 upregulated, 279 downregulated) of differentially expressed genes (Rizhsky et al., 2004). Applying the same cut-off (1.5-fold log2) in our analysis the number of differentially expressed genes is still two- to four-fold higher (576 upregulated, 1116 downregulated). This difference reflects in part the fact that the expression of many HS-upregulated genes is transient with a maximum after 1–2 h HS followed by strong decline (Lohmann et al., 2004). The larger number of downregulated genes in our experiments is probably the result of a transient transcriptional repression of many non-HS genes (Schöffl et al., 1987). The long-term HS may lead to an adjustment of the steady-state levels of mRNAs to control levels for a large number of transiently up- or down-regulated genes. Not only the duration of HS (1 h versus 6 h) was different between our experiments and the ones described by Rizhsky et al. (2004), but several other experimental parameters (dark versus light and cut leaves in buffer versus heat treatment in air), which may have profound effects on gene expression, were also not identical.
Heat stress affects many genes in WT and hsfA1a/1b mutant plants
In general, our analysis shows that HS has a very fast and global effect on the transcriptome: approximately 11% of all monitored genes show a different expression level after 1 h HS. The differences in the numbers of HS-affected genes of WT and hsfA1a/1b plants are manifested not only in total numbers but also in the proportions of up- and down-regulated genes (Figure 1). Whereas approximately 11% in WT and 14% of all genes in hsfA1a/1b plants are affected by HS, the ratios between down- and up-regulated genes are 1.6 and 1.0 in WT and hsfA1a/1b plants respectively. These global changes indicate that the deficiency of HsfA1a/1b causes an increase in genes that score as differentially expressed upon HS and there is a relative increase in the number of HS-upregulated genes in hsfA1a/1b plants. This increase in the number of upregulated genes is contrary to expectation. It is unlikely that the lack of HsfA1a/1b activates the upregulation of mRNA levels of many (approximately 500) genes, while the expression of only approximately 105 HS genes is downregulated in mutant plants. The downregulation of HS genes in the hsfA1a/1b mutant affects mostly those genes which show very strong expression in WT (see HSP and HSF profiles in Figures 3 and 4), which results in a higher representation of other mRNAs in the sample of the hsfA1a/1b mutant upon HS. This may explain a relative increase in the number of upregulated genes compared with WT, which does not necessarily require a real change in gene expression of 500 additional genes. However, irrespective of the relative proportions of positively and negatively affected genes, there is a strong impact of HS on the transcriptome of Arabidopsis, but the effects of HSF deficiencies are not conclusively mirrored by the numbers of HS-induced genes in mutant plants.
HsfA1a/1b factors are responsible for the regulation of several of the strongest heat-induced genes
The criterion for HsfA1a/1b-dependent expression of genes was the significant quantitative difference of heat-induced mRNAs between WT and hsfA1a/1b mutant plants. The 112 HsfA1a/1b-dependent genes represent only a small fraction (4.37%) of HS-regulated WT genes. The majority of them (105 genes) is downregulated in the mutant and thus activated by HsfA1a/1b. This confirms the classification of these HSF as transcriptional activators (Lee et al., 1995; Lohmann et al., 2004; Prändl et al., 1998). Both belong to the class A HSF, which contains members for the initiation of the HS response.
Furthermore, the data show that the majority of HsfA1a/1b dependently upregulated target genes are in the group of the most strongly induced HS genes. However, there is no correlation with the presence of HSE sequences in the promoter regions of those genes. Whereas 47% of the HsfA1a/1b-regulated genes contain HSE sequences, there is no preferential representation among high-ranked HS or HSF-regulated genes. The other HSF-regulated genes contain variant HSE, which bind HSF in vitro (Nover et al., 2001) and which may also function as HSF1a/1b-binding sites in vivo. The presence of HSE is not a sufficient criterion for predicting HS or HSF-regulated expression in plants. Perfect and slightly altered HSE sequences have been identified in the promoter regions of all 21 HSF genes of Arabidopsis (Nover et al., 2001), but our analysis shows that only the expression of six of them is regulated by HS and/or HSF. Furthermore, a genome-wide analysis of the mammalian HS response has shown that there is no strict correlation between the presence of HSE, HSF1-binding, heat-induced transcription and that independent post-transcriptional mechanisms regulate the accumulation of a significant number of HS-elevated transcripts (Trinklein et al., 2004).
Many of the high-ranking HS-induced genes have been annotated as HS/HSP genes. A total of 111 genes represented on the chip are related to HS/HSP functions, the expression of 44 of these was induced by HS and 10 of them appeared to be dependent on HsfA1a/1b. Within the family of sHSP, which comprises 18 members, the induction levels of six of the HS-inducible sHSP genes (37.5%) are clearly dependent on HsfA1a/1b. The expression of nine other sHSP genes was also lower but not significantly impaired in the hsfA1a/1b mutant. Two sHSP genes, Hsp15.4-CI(r) and Hsp14.2-P(r), which are only distantly related to the other sHSP (Scharf et al., 2001), were not activated by HS, Hsp15.4-CI(r) was downregulated. This may indicate that these two genes/proteins are related sHSP isoforms, which are not required under HS conditions in leaves.
The strong but not strict correlation between high induction levels and HsfA1a/1b-dependent expression suggests that other factors and mechanisms, which cause strong induction of gene expression, must be activated upon HS. This can be also predicted for the expression of other HS-activated but not HsfA1a/1b-dependent genes in WT.
What are the differences in the regulation of the two groups representing HsfA1a/1b-dependent and independent HS genes? We propose that different HSF cooperate in the immediate early expression of different HSF target genes. The expression of HsfA1a/1b-dependent genes rely in the early stage of the HS response exclusively on these HSF. Genes that score as HsfA1a/1b-independent in expression profiling use other as yet unknown HSF for induction of transcription. HsfA1a and/or HsfA1b may also participate in the expression of some of these genes. Owing to the strict criteria used in our expression profiling analysis, the group of HsfA1a/1b-dependent genes represents only the minimal set of HSF targets. The importance of HsfA1a/1b for the initiation of HSP expression is further supported by the gain-of-function effects of HSF overexpression in transgenic plants, which resulted in a constitutive expression of a number of HS genes tested (Lee et al., 1995; Prändl et al., 1998).
Only a small number of genes (11 genes, Table 1) showed a HSF-dependent expression at normal temperature. Surprisingly, only the basal level but not the heat-induced expression of these genes was negatively affected in the hsfA1a/1b plants. This indicates that for a small number of heat shock genes HSF are involved in basal expression but they may become replaced by other HSF upon heat stress.
HsfA1A/1B are regulators of other transcription factors
The group of HsfA1a/1b-dependent genes includes not only HSP genes, which are most strongly affected by HSF during the immediate early heat shock response. Several other genes were identified, which have other or as yet unknown functions.
Interestingly, HsfA1a/1b seem to regulate the expression of a small number of transcription factor genes and hence, seem to be also responsible for secondary changes in gene expression after HS. Most striking is the HsfA1a/1b-dependent expression of other HSF, in particular HsfB1 (HSF4), B2a and A7a, which represent three of six heat-inducible HSF genes. There is evidence by qRT-PCR experiments for HsfA1a/1b-dependent expression of HsfB1 (HSF4) and HsfB2b (HSF7), which are HS-induced in WT but to lower levels in hsfA1a/1b plants (Lohmann et al., 2004). Both HSF are members of the class B factors, which lack the HSF-typical transcription activator domains. HsfB1 mRNA induction was identified by expression profiling analysis of the wound response in Arabidopsis (Cheong et al., 2002). The identification of HsfA7a (class A) as an HSF-dependent expressed gene indicates that not only are class B factors involved in controlling delayed functions in the HS response. After long-term HS the steady-state levels of HSF mRNAs (Rizhsky et al., 2004) differ significantly from the profile after short-term treatment (this paper, Figure 4a). Of the early HS-induced HSF only the mRNAs of HsfB1, HsfA2 and in addition HsfC are present at enhanced levels after long-term HS. The alterations in the expression profiles of HSF indicate that HsfA1a/1b-dependent regulation of HSF expression is transient and that other HSF and/or other regulatory mechanisms are involved in the sustained expression of certain HSF during continuous HS.
Besides HSF genes, some other transcription factors that were previously connected to stress responses are controlled by HsfA1a/1b as well: WRKY7 is connected to disease resistance (Maleck et al., 2000), Scarecrow factors were not only involved in developmental processes but were also shown to be affected in expression by salt stress in Arabidopsis (Ueda et al., 2002) and osmotic stress in white spruce (Stasolla et al., 2003). Dof zinc finger proteins are involved in stress response (Chen et al., 1996; Kang and Singh, 2000) and in the control of seed germination (Papi et al., 2000). Members of the AP2/EREBP factor family play a variety of roles throughout the plant life cycle; they are implicated in the regulation of developmental processes and environmental stress responses (Riechmann and Meyerowitz, 1998). All DRE-binding proteins identified to date belong to the AP2/EREBP family, the factors DREB1 and DREB2 are involved in the regulation of drought, salt cold-inducible genes (Kizis et al., 2001).
The expression of only relatively few environmental stress-related transcription factor genes of large multigene families appear to be affected by HS and/or HsfA1a/1b. The majority of the HS-regulated genes of these families is downregulated upon HS and only one gene of each family appears to be regulated by HSF. This demonstrates that the alterations in gene expression in the HS response are dominated by HSF. Secondary effects, based on HSF-dependent expression of other transcription factors, seem to play only a minor role. The 21 Athsf family members represent only a very small fraction of the more than 1500 potential transcription factor genes of Arabidopsis (Riechmann et al., 2000).
HS and HsfA1b/1b-controlled pathway for Raffinose Family Oligosaccharides
A major category of HsfA1b/1b-regulated genes concern enzymes involved in carbohydrate metabolism. All enzymes required for synthesis of RFO precursors leading to galactinol and RFO synthesis are members of gene families. The expression of at least one member of each family was positively affected by HS and several of them in a HsfA1a/1b-dependent fashion. Two enzymes of this pathway were previously linked to stress responses. Myo-inositol-1-phosphate-synthase was induced by cold, drought and salt stress in Arabidopsis (Kreps et al., 2002) and involved in salt stress response and osmoprotection in the halophyte Mesembryanthemum crystallinum (Ishitani et al., 1996). We found that two of seven galactinol synthase genes, GolS1 and GolS,2 were heat-inducible, but only GolS1 is activated via HsfA1a/1b. Taji et al. (2002) showed that both genes are induced by drought and high salinity stress but the expression was obviously not regulated by the transcription factor DREB. It will be interesting to see whether HSF is involved in the induction by drought and salt stress, which would suggest that HSF integrates different signalling pathways that lead to the expression of GolS1. The HSF-dependent expression of GolS1 correlates with an increase in the level of RFO and knock out mutations of GolS1 are unable to accumulate stress-induced galactinol and raffinose levels in leaves (Panikulangara et al., 2004). Enhanced levels of galactinol and raffinose, generated in the leaves of transgenic plants by overexpression of GolS2, caused improved drought tolerance of Arabidopsis (Taji et al., 2002).
The enzymatic steps downstream of galactinol, which lead to the synthesis of raffinose and stachyose (Figure 5), could not be directly investigated, the genes and processes have not yet been identified in Arabidopsis. Based on sequence similarity to raffinose and stachyose synthases of kidney bean and cucumber, we identified a glycosyl hydrolase family (family 36) of genes, which most likely also contain putative genes for such enzymes. Within this family we found two members, which are affected by HS: At5g40390 is upregulated (threefold) and At3g57520 is downregulated (threefold). Thus it seems that the entire pathway can be regulated at different levels by HS and/or HSF, providing an example of a highly adjustable pathway in carbohydrate metabolism that can be alternatively driven by a set of stress-responsive genes. The expression/function of these genes overlaps between different environmental responses that require protection by osmo-protective solutes.
In contrast to the pathway leading to RFO, HS/HSF-dependent expression affects in many cases only single genes of complex biochemical pathways, as for example genes involved in protein biosynthesis/degradation, membrane transport, oxidative stress response and signalling. The biological function of most of these genes during HS is not known. However, some have been connected with environmental stress responses. For example, IAA2 is a transcription factor, which is involved in auxin signalling (Reed, 2001), and AtERF4 is an active negative repressor of transcription, which is induced by wounding, cold, high salinity and drought stress (Fujimoto et al., 2000). Interestingly, both genes and two other signalling connected transcription factors are upregulated in hsfA1a/1b plants. Another striking connection exists to N-myristoylation which is required for SOS3 function in plant salt tolerance (Ishitani et al., 2000). The involvement of these processes in plant protection is not unexpected but our data document that pathways and mechanisms dealing with protein synthesis and stability are direct and early targets of HSF in the HS response. There is a high probability that plants, which are exposed to heat stress in nature, will subsequently suffer drought stress. We conclude that there is an overlap of drought, high salinity and heat stress for a number of genes in different pathways, which thus seem to be important for dealing with a number of stresses.
Plants material and heat stress conditions
Arabidopsis thaliana WT (ecotype Wassilewskija) and AthsfA1a/AthsfA1b double knockout plants (abbreviated hsfA1a/1b) were used (Lohmann et al., 2004). Plants were routinely grown on soil in a dark/light cycle of 16/8 h at 20°C for 5–6 weeks (60% humidity, 5000 lux light intensity). Two batches of plants for each experiment were grown independently under the same conditions to gain two experimental replicates.
Heat shock was administered by incubating leaves collected from 35 soil-grown plants for 1 h at 37°C in section incubation buffer (SIB: 1 mm potassium sulphate, pH 6.0 and 1% sucrose) in the dark; non-heat-stressed controls were incubated in SIB at 22°C in a shaking water bath. These conditions were chosen to minimize interference by photooxidation and drought stress.
RNA isolation and labelling
For RNA isolation the pooled leaves were frozen in liquid nitrogen. Frozen tissue was stored at −80°C, and RNA was extracted with the Plant RNeasy Mini Kit (Qiagen, Hilden, Germany). Ten micrograms of total RNA was used as starting material to synthesize double-stranded cDNA using the Superscript Choice System (Invitrogen, Karlsruhe, Germany) and an oligo dT-T7 primer (Genset, Paris, France). The cDNA served as a template for synthesis of biotinylated cRNA using the BioArray High Yield Transcript Labeling Kit (Enzo, Farmingdale, NY, USA). Biotinylated cRNA was cleaned with RNeasy columns (Qiagen) according to the manufacturer's protocol, with the following modifications. First, the cRNA was passed through the column twice to increase binding. Secondly, the eluate was reapplied to the column once to increase yield. Ten micrograms of concentration-adjusted cRNA was fragmented according to the GeneChip protocol (Affymetrix, Santa Clara, CA, USA).
To monitor changes in gene expression we used photolithographically produced microarrays in which each gene is represented as a probe set with several oligonucleotide features (Affymetrix GeneChips). The experiment was carried out using the Arabidopsis ATH1 array, which represents 22 810 probe sets.
Hybridization of GeneChip arrays was performed according to the manufacturer's protocol (Affymetrix). For washing and staining, protocol EukGe-WSv4 (Affymetrix) was used. Scanning was performed with the GeneChip Scanner 3000 (Affymetrix). Sample quality was assessed by examination of 3′ to 5′ intensity ratios of certain genes.
Analysis of expression data
Data for every condition were available from duplicates of each experiment. For comparison across multiple arrays, raw data were scaled using the global intensity of all probe sets on each array. Signal intensities for each probe were estimated from .cel files by the Robust Multichip Average expression summary (Irizarry et al., 2003) using RMA Express, a standalone GUI program for Windows (and Linux). The RMA normalization method was superior in terms of having a greater sensitivity and specificity in measuring differential expression (Irizarry et al., 2003) and providing an expression measure and fold-change estimation with less variance (Cope et al., 2004). Data were imported in Genespring 6.1 (Silicon Genetics, Redwood City, CA, USA) and retransformed into linear values. The Cross Gene Error model was activated and based on replicates. Statistical analyses of data were conducted in the following way: (i) all genes, which were changing less than twofold in the expression between conditions were discarded; (ii) using Genespring's Cross-Gene Error model variances (based on replicates) a Welch's t-test was conducted to compare the different conditions and to find differentially expressed genes (in all analyses a P-value of 0.05 was selected); (iii) the Benjamini and Hochberg False Discovery Rate Multiple Testing Correction was applied to minimize the detection of false positives.
Isolation of mRNA and preparation of cDNA for real-time PCR
Poly(A)+ mRNA was isolated from intact total RNA using the Oligotex mRNA Kit (Qiagen). One-fifth of the poly(A)+ mRNA isolated from 15 μg of total RNA was converted into cDNA using the ThermoScript RT-PCR System (Invitrogen) with oligo (dT)20 primers. The amount of poly(A)+ mRNA/cDNA double-stranded products obtained after reverse transcription was measured using PicoGreen dsDNA quantitation reagent (Molecular Probes, Eugene, OR, USA) and the concentration was adjusted to 1 ng μl−1.
Real-time PCR and quantification of mRNA levels
Primer pairs for quantitative real-time PCR (qRT-PCR) were designed using Primer 3 software (http://www.genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi) and gene sequences available in Genebank. Gene-specific primers (see Table S3) were chosen in a way that resulting PCR products had approximately the same size (120 bp) and did span the 3′ regions of cDNA sequences. Real-time PCR was performed in 30 μl 1x HP Buffer (ABgene, Hamburg, Germany), reactions containing cDNA (two different concentrations were tested in triplicate reactions: 1 ng of cDNA and a dilution of 1:8), 0.75 U of ThermoStart Hot Start Taq (ABgene), 3 mm MgCl2, 267 μm each dNTPs (Peqlab, Erlangen, Germany), 0.5 μm gene-specific primers, using an iCycler iQ system (Bio-Rad, Hercules, CA, USA). Amplification of PCR products was monitored via intercalation of SYBR-Green I (Molecular Probes) at 0.1× concentration. The following program was applied: initial polymerase activation at 95°C, 15 min; 40 cycles of 94°C, 15 sec, respective annealing temperature, 40 sec, 72°C, 10 sec. All experiments were performed twice for cDNAs prepared from two batches of plants. Act2 cDNA, set as 100%, was used as a standard in all experiments.
The quality of PCR products was visually inspected by agarose gel electrophoresis. The generation of only one single band of the expected size was taken as a criterion for specificity. The identity of PCR products was confirmed by direct DNA sequencing.
We thank Dr Ulrike Zengraf (ZMBP, University Tuebingen) for critical reading and suggestions, Drs Detlef Weigel and Markus Schmid (Max-Planck-Institut für Entwicklungsbiologie Tuebingen) for providing access to the expression profiling facilities and for kind introduction to chip hybridization technique. Part of the research was funded by grants of the Deutsche Forschungsgemeinschaft (SFB446, project A2).