A Heat-Inducible Transcription Factor, HsfA2, Is Required for Extension of Acquired Thermotolerance in Arabidopsis

The expression of heat-shock proteins (Hsps) induced by nonlethal heat treatment confers acquired thermotolerance (AT) to organisms against subsequent challenges of otherwise lethal temperature. After the stress signal is removed, AT gradually decays, with decreased Hsps during recovery. AT of sufficient duration is critical for sessile organisms such as plants to survive repeated heat stress in their environment, but little is known regarding its regulation. To identify potential regulatory components, we took a reverse-genetics approach by screening for Arabidopsis T-DNA insertion mutants that show decreased thermotolerance after a long recovery (2 d) under non-stress conditions following an acclimation heat treatment. Among the tested mutants corresponding to 48 heat-induced genes, only the HsfA2 knockout mutant showed an obvious phenotype. Following pretreatment at 37 o C, the mutant line was more sensitive to severe heat stress than the wild type after long but not short recovery periods, and this could be complemented by the introduction of a wild-type copy of the HsfA2 gene. Quantitative hypocotyl elongation assay also revealed that AT decayed faster in the absence of HsfA2 . Significant reduction in the transcript levels of several highly heat-inducible genes was observed in HsfA2 knockout plants after 4 h recovery or 2 h prolonged heat stress. Immunoblot analysis showed that Hsa32 and class I small Hsp were less abundant in the mutant than in the wild type after long recovery. Our results suggest that HsfA2 as a heat-inducible transactivator sustains the expression of Hsp genes and extends the duration of AT in Arabidopsis. in 3-d-old seedlings of the wild type (wt) and hsfA2-1 during A recovery (0 to 48 h after treatment at 37 o C for 1 h), and B prolonged HS (from 0 to 8 h at 37 o C). The control (C) samples were seedlings without heat treatment and collected at the same time as the treated samples collected at 48 h. The RT-PCR products of actin were shown as a loading control. PCR cycles Apx2 and actin were 25 and 30, respectively. Similar results were obtained from 2 biological repeats, with 1 shown here.


INTRODUCTION
Eukaryotic cells respond to elevated temperature or heat shock (HS) by inducing the transcription of genes encoding proteins such as molecular chaperones. Many of these proteins are involved in preventing or repairing the damage caused by heat stress and thus confer increased thermotolerance (Lindquist and Craig, 1988;Vierling, 1991). This phenomenon, known as HS response (HSR), is initiated by the activation of the HS transcription factor (Hsf), a conserved protein present in all eukaryotic organisms studied to date (Wu, 1995;Nover et al., 2001).
The exact number of Hsf genes differs greatly among various eukaryotic organisms. For example, Saccharomyces cerevisiae, Drosophila melanogaster and Caenorhabditis elegans each has a single Hsf gene playing an important biological role in addition to the HSR (Sorger and Pelham, 1988;Wiederrecht et al., 1988;Jedlicka et al., 1997;Hsu et al., 2003). The mammalian genome contains 3 Hsf isoforms, Hsf1, Hsf2, and Hsf4, each with a distinct biological function (Xiao et al., 1999;Bu et al., 2002;Fujimoto et al., 2004;Xing et al., 2005). In contrast to the low numbers of Hsf genes found in animals and yeasts, the model plant Arabidopsis contains 21 Hsf homologues that can be sorted into 3 classes (class A, B, and C) (Nover et al., 2001). Most class A Hsfs of Arabidopsis contain the AHA motifs essential for transcription activation, while the class B and C Hsfs do not (Kotak et al., 2004). Genetic evidence shows that HsfA1a functions as a master regulator of HSR in tomato (Mishra et al., 2002). In Arabidopsis, HsfA1a and HsfA1b play important roles in the induction of a number of Hsp genes in the early phase of HSR (Lohmann et al., 2004). A rice mutant, spl7, which is sensitive to environmental stress and exhibits a spotted leaf phenotype under elevated temperature, was shown to have a missense mutation in an Hsf gene belonging to the HsfA4 group (Yamanouchi et al., 2002).
Intriguingly, several plant Hsfs are HS-inducible genes themselves, which among eukaryotic systems is a feature unique to plants (Nover et al., 2001). For example, HsfA2 is strongly induced by HS in tomato (Scharf et al., 1998) and Arabidopsis (Busch et al., 2005). The presence of HsfA2 exclusively after HS treatment suggests its role in the transcriptional regulation of Hsp genes during prolonged heat stress or in recovery after HS (Baniwal et al., 2004), and hence its importance in the development of acquired thermotolerance (AT).
Recently, two research groups have independently reported the function of HsfA2 by characterizing the same Arabidopsis T-DNA knockout line. Li et al. (2005) found that disruption of HsfA2 caused a slight reduction in expression of some Hsp genes upon HS treatment, and that the knockout mutant displayed reduced basal and acquired thermotolerance as well as oxidative stress tolerance, while overexpression of HsfA2 enhances tolerance under these stress conditions. Schramm et al. (2006) identified a subset of genes whose expression was reduced upon HS treatment in the knockout mutant. They showed that HsfA2 bound to the promoter regions of selected target genes and activated a GUS reporter that was fused to them. However, in contrast to the observation made by Li et al. (2005), no obvious phenotype was observed for the mutant under the relatively comprehensive assay conditions (Schramm et al., 2006). The discrepancy in these studies obscures the biological function of this important component of plant HSR.
In the post-genomic era, functional study of target genes by reverse genetic approaches has become the norm, but knockout mutation often leads to no significant change in phenotype. This approach is especially difficult for the study of stressinduced genes, because plants without these genes often show no or very subtle difference from the wild type even under the stress condition that induces them.
Apparently, developing a more subtle and effective assay is essential for elucidating the biological function of HSR genes since so far only a few successful cases have been reported (Hong and Vierling, 2000;Queitsch et al., 2000;Panikulangara et al., 2004;Li et al., 2005;Charng et al., 2006;Chen et al., 2006). Recently, we have demonstrated the function of a plant-specific Hsp, Hsa32, in Arabidopsis by employing a novel assay method (Charng et al., 2006). The mutant plants lacking Hsa32 became more sensitive to severe HS than the wild type after long (48 h) but not short recovery following heat acclimation treatment, suggesting that the protein is associated with the duration of AT. AT of significant duration is presumably an important feature for plants to cope with repeated heat stress with intervening periods of milder temperature, but very little information is available on this issue so far.
Here, we report the screening of Arabidopsis T-DNA insertion mutants of 48 heat-induced genes under the same assay conditions used to characterize Hsa32; only the HsfA2 knockout plants showed a significant heat-sensitive phenotype, which could be rescued by introducing a wild-type copy of HsfA2. Physiological studies showed that HsfA2 is essential for AT after long but not short recovery. Further diagnosis by microarray and RT-PCR suggested that the heat-induced Hsf is not required for initial regulation of HSR genes but for sustaining the transcript level of

Recovery from Acclimation
To identify genes involved in duration of AT, we took a reverse genetic approach by randomly screening Arabidopsis T-DNA insertion mutants corresponding to 48 genes (Table I), which previous microarray analyses had shown to be upregulated by HS treatment (Busch et al., 2005;Charng et al., 2006). Most of the mutant lines had T-DNA inserted in the exon region of the target gene, which should have resulted in loss-of-function alleles. Homozygous lines of these mutants were isolated by PCR analysis. RT-PCR experiments confirmed that all the T-DNA insertion lines were unable to accumulate corresponding transcripts under normal and HS conditions (data not shown). Three-day-old seedlings of these mutants grown on agar plates were treated with severe HS (44 o C for 50 min) after a 2 d recovery from previous acclimation treatment (37 o C for 1 h). These conditions killed the Hsa32 knockout but not the wild-type plants (Charng et al., 2006) that were included in each plate as positive and negative controls, respectively, to ensure the effectiveness of the treatment. After 2 rounds of screening, a lethal phenotype under the test condition was repeatedly observed only for the HsfA2 T-DNA insertion mutant (Fig. 1A). We therefore carried out further studies on this putative HsfA2 mutant.
The defect of HsfA2 T-DNA insertion was associated with the homozygous line containing both disrupted HsfA2 alleles with T-DNA inserted within the second exon near the 5'-end (Supplemental Fig. 1A), whereas about 75% and 100% of the offspring of the hemizygous or azygous sister lines, respectively, showed the wildtype phenotype. These results indicated that the T-DNA insertion caused a loss-offunction mutant allele, hereafter designated as hsfA2-1. RT-PCR analysis showed that transcripts of HsfA2 were induced by HS treatment (37 o C, 1 h) in the wild type but not in the hsfA2-1 plants, whereas the expression of Hsa32 and Hsp101 control genes was not significantly affected in the mutant (Supplemental Fig. 1B). The RT-PCR results confirmed that hsfA2-1 was a null mutant of HsfA2. Since Southern blot analysis showed that multiple T-DNA insertion events occurred in the hsfA2-1 mutant line (data not shown), we performed a complementation test by transforming a wildtype copy of HsfA2 genomic DNA into the knockout mutant to ensure that the observed phenotype was not caused by disruption of other genes by T-DNA or by a secondary mutation in the genome of the mutant line. The introduced wild-type gene, covering a 387-bp potential promoter region (Supplemental Fig. 1), rescued the mutant phenotype of hsfA2-1 (Fig. 1A) and restored the heat-inducible expression of HsfA2 in independent transgenic lines of the T2 generation (Fig. 1B). The survival rate of the thermotolerance test was about the same as the resistance rate against Lphosphinothricin (Basta®, Hoechst AG), conferred by the co-transformed bar gene, in lines C112-1 and C112-2, which suggests co-segregation of the transgene and complementation capability. Both tests yielded a survival rate close to 75% (Fig. 1B), indicating that a single T-DNA insertion event may have occurred in these transgenic lines. The results of the complementation experiment confirm that disruption of HsfA2 caused the mutant phenotype observed in hsfA2-1 seedlings.
Besides the defect in AT, hsfA2-1 did not exhibit an obvious phenotypic difference in morphology, germination time and rate, growth rate, time to flowering, and seed yield as compared to the wild type under non-stress conditions, which suggests that HsfA2 is not required for normal growth and development.

HsfA2 Was Essential for the Duration but Not Induction of AT
To investigate how HsfA2 was involved in the development of AT, we compared results of a thermotolerance test between the hsfA2-1 mutant and the wild type as well as the T-DNA knockout mutants for Hsa32 (hsa32-1) and Hsp101 (hsp101), which show different types of defects in AT. Hsa32 was shown to be essential for protecting Arabidopsis seedlings against severe heat stress after long but not short recovery (Charng et al., 2006), whereas Hsp101 was shown to be required for AT developed during a short (Hong and Vierling, 2000;Queitsch et al., 2000) as well as a long recovery time (Charng et al., 2006).
In a survival rate assay, 3-day-old Arabidopsis seedlings grown on solid agar medium were first conditioned at 37 o C for 1 h, allowed to recover for various times at room temperature, challenged by a severe HS (at 44 o C for various times), then allowed to recover at room temperature again for more than 7 d. For the wild type, the conditioning treatment prevented the seedlings from being killed by the severe HS, and the strength of AT gradually declined but lasted up to 72 h ( Fig. 2A-E). The hsfA2-1 mutant did significantly differ in survival rate from the wild type if the severe HS challenge was applied following a 2 h ( Fig. 2B and C) or 24 h (data not shown) recovery, which was similar to the case for hsa32-1 seedlings. In contrast, the hsp101 plants undergoing the same treatment all died (Fig. 2C). Similarly to the hsa32-1 and hsp101 mutant plants, the hsfA2-1 plants became more sensitive to the severe HS challenge than the wild-type plants after a 2-3 d recovery ( Fig. 2D and E). However, the damage caused by the severe HS was less severe in hsfA2-1 than in hsa32-1 and hsp101 mutants. In the hsfA2-1 mutant population, a small number of the seedlings, despite being severely injured, managed to survive and eventually produce true leaves, whereas the hsa32-1 and hsp101 plants were all dead under the same conditions ( Fig.   2D and E). If the treatment time at 44 o C was increased to 50 min, all the hsfA2-1 mutant plants would die, as shown in Fig. 1. The decreased thermotolerance after long recovery in all the mutants, however, was reversible. A second conditioning treatment 2 h before the severe HS challenge after 72 h of recovery protected the mutant plants against severe injury (Fig. 2F). These results suggest that HsfA2 is required for longer duration instead of initial induction of AT.
To elucidate further details of this process, we monitored and compared the development of phenotype in the heat-sensitive mutants after the HS regime shown in Figure 2D. The severe HS challenge did not cause any immediately observable damage in the wild-type and mutant plants, but the heat-induced injury syndrome began to develop after 2 d of recovery in the mutants (Fig. 2G). The degree of injury in the hsfA2-1 plants seemed to be less severe than that in the hsp101 plants (Fig. 2G).
The hypocotyls of the hsfA2-1 plants looked healthier than those of the hsp101 mutant plants, which became severely dehydrated and wilted after 6 d of recovery. In the root, severe HS inhibited the growth of the main root in both the wild type and mutants (Fig. 2H). However, vigorous adventitious roots developed in the wild type but only began to emerge in the hsfA2-1 or did not emerge at all in the hsa32-1 and hsp101 mutants (Fig. 2H). Ion leakage analysis (measuring the extent of membrane damage) 2 h after the severe HS challenge showed no significant difference between the wild type and the mutants (Fig. 3). However, ion leakage in the wild-type plants gradually decreased after 6 h of recovery, whereas that of the mutants remained high.
We also examined the role of the other 47 heat-induced genes in AT during short recovery. Interestingly, the T-DNA insertion mutant (SALK_088054) of a putative membrane-associated protease gene showed lethal phenotype after severe HS challenge following a short but not long recovery (Supplemental Fig. 2), in contrast to the cases of hsfA2-1 and hsa32-1. These results revealed the temporal roles of these genes during the development of AT, and that the function of HsfA2 specifically affect AT at later phase.

AT Declined Faster in the Absence of HsfA2
Quantitative hypocotyl elongation assay was performed as previously described to determine the decay rate of AT (Charng et al., 2006). In this study, we measured the elongation of hypocotyls of vertically-grown etiolated seedlings after severe HS challenge (44 o C for 45 min) following conditioning treatment at 37 o C for 1 h. The recovery time between the conditioning and HS challenge was varied to reveal the kinetics of thermotolerance decay. Hypocotyls could not elongate if they were directly challenged without conditioning. Fig. 4A shows the baseline of hypocotyl elongation after various recovery times following conditioning treatment, and no significant difference was observed among the hsfA2-1, hsp101, hsa32-1, and wildtype plants in the absence of severe HS treatment. Consistent with the results of survival rate assay, hsfA2-1 plants showed a significant decrease in hypocotyl elongation after the severe HS challenge applied during long (> 24 h) but not short (< 12 h) recovery times (Fig. 4B). After challenge at 48 h of recovery, the hsfA2-1 plants still showed some growth, while the hsa32-1 and hsp101 plants showed no further growth. This observation agreed with the survival rate data (Fig. 2), which suggests that HsfA2 is essential for extending the duration but not the initial induction of AT.

hsfA2-1 Showed Slightly Increased Sensitivity to HS Without Acclimation
We also employed the hypocotyl elongation and survival rate assay to determine whether HsfA2 is required for basal thermotolerance, defined as the capability to tolerate severe heat stress without prior conditioning that induces HSR.
We measured the length of hypocotyls elongated for 3 d after the application of a short pulse of severe HS (5 min at 46-50 o C or 15-30 min at 44 o C). During the short treatment at 44 o C or above, the expression of the strongly heat-induced genes, such as Hsp25.3-P and Hsp18.1-CI was not invoked (data not shown). wild-type plants. These data suggest that Hsa32 and Hsp101 are required for basal thermotolerance, while HsfA2 is minimally involved. In addition, when grown under continual moderate heat stress (see Materials and Methods for condition), no obvious phenotypic difference was observed between the wild-type and hsfA2-1 seedlings.

Recovery and Prolonged Heat Stress
Since AtHsfA2 was previously shown to have transcription factor activity (Port et al., 2004;Schramm et al., 2006), disruption of HsfA2 may have significantly compromised the expression of certain Hsp genes during recovery, which in turn shortened the duration of AT. However, we could not exclude the possibility that the initial induction of some heat-induced genes required HsfA2, which may be present at Since the phenotype of hsfA2-1 was manifested after the severe HS challenge after 48 h or longer recovery, we used the ATH1 chip to query the transcriptome profiles of the wild type and the mutant seedlings harvested right after the treatment at 44 o C for 45 min in the HS regime indicated in Fig. 2D. Control samples were seedlings of the same age but without both the conditioning and severe HS treatment.
Microarray data revealed only a few genes showing significantly lower HS/control signal ratio in the mutant than in the wild type in 2 biological repeats. Hsp18.1-CI (previously known as Hsp18.2 (Takahashi and Komeda, 1989)) and Hsp25.3-P (previously known as Hsp21 (Osteryoung et al., 1993)) were the most affected (Supplemental Table I). The small number of affected genes was probably due to the time point chosen for the microarray study, when the transcript levels of other Hsp genes were too low to show a difference after the long recovery. Since the expression of Hsp18.1-CI and Hsp25.3-P was not induced by severe HS at 44 o C for 45 min in the wild-type and mutant plants (data not shown), abundance of the transcripts of these genes probably declined faster in hsfA2-1 than in the wild type during recovery. Thus, we examined the transcript level of Hsp18.1-CI, Hsp25.3-P, Hsa32, and Hsp101 by semi-quantitative RT-PCR during the course of recovery (0 to 48 h) after conditioning at 37 o C for 1 h (Fig. 6A). Apart from Hsp101, these genes have been shown to be the potential targets of HsfA1a/b in Arabidopsis, because in the HsfA1a and HsfA1b double knockout mutant their expression was significantly suppressed in response to HS (Busch et al., 2005). Directly after the conditioning treatment or within 2 h of recovery, the mRNA levels of all Hsps tested differed little, which was consistent with the microarray data (Supplemental Fig. 3A). However, the transcript levels of Hsp18.1, Hsp25.3-P, and Hsa32 were substantially lower in the hsfA2-1 than in the wild-type plants after 4 h of recovery, with no significant difference found for Hsp101 transcript (Fig. 6A). Western blot analysis showed that after 48 h of recovery, the levels of Hsa32 and sHsp-CI proteins of hsfA2-1 were lower than in the wild type ( Fig.   6B), consistent with the transcript levels.
To determine whether HsfA2 is also required for the duration of HSR under prolonged HS, we also examined the transcript levels of Hsp18.1-CI, Hsp25.3-P, Hsa32, and Hsp101 in seedlings subjected to continuous heat treatment at 37 o C for up to 8 h. RT-PCR analysis again revealed no significant difference in transcript levels of these genes within the first hour of HS treatment but significantly lowered levels in mutant plants after 2 h of prolonged HS (Fig. 7). However, no difference in protein levels of Hsa32 and sHsp-CI was detected in these samples (data not shown), perhaps due to the stability and/or lower saturation threshold of the amount of mRNA actively engaged in translation of these proteins under continuous heat. These data again suggest that HsfA2 is not required for the initial induction but, rather, for the transcription of certain heat-induced genes during recovery or prolonged HS.

DISCUSSION
In an attempt to identify genetic components involved in the regulation of the duration of AT, we tested T-DNA insertion mutant lines corresponding to 48 heatinduced genes (Table I) by a HS regime that applied severe heat stress after 2 d of recovery from acclimation to mild heat stress. Under this condition, the HsfA2 knockout line, hsfA2-1, exhibited a severely heat-sensitive phenotype (Fig. 1A), while all the other mutants were not substantially different from the wild type. The results suggest that all of these genes except HsfA2 were not essential for AT duration. Gene redundancy or existence of alternative pathways may explain why some of these mutants did not show a thermotolerance defect under the test conditions. Alternatively, they may be essential for tolerance against heat stress of different magnitude, in combination with other stresses, or at different stages of AT development. Indeed, the finding of a T-DNA knockout mutant of a putative membrane-associated protease gene, which is sensitive to severe HS challenge after a short but not long recovery following acclimation (Supplemental Fig. 2), suggests the existence of distinct stages of AT development during recovery involving different HSR genes. Investigation of this protease mutant is currently underway.
Further characterization of hsfA2-1 revealed that HsfA2 was dispensable in the development of AT during the early phase ( Fig. 2 and Fig. 4). It appears that the gene is essential in extending the duration of AT but not for its induction. The distinct phenotype of hsfA2-1 was in good agreement with the transcriptomic, RT-PCR and immunoblot results. According to microarray analysis, disruption of HsfA2 did not significantly affect the transcript levels of other genes under normal conditions or after 37 o C treatment for 1 h. Consistently, RT-PCR semi-quantitative analysis indicated that HS-induction of several tested Hsp genes was not affected in the absence of HsfA2 in the early phase ( Fig. 6 and 7). However, in the hsfA2-1 mutant the level of the mRNA of Hsp genes, most prominently Hsp25.3-P and Hsa32, declined faster than that in the wild type during the recovery period (Fig. 6A). The protein level of Hsa32 in hsfA2-1 plants was consistent with the transcript result but declined more slowly than the transcripts (Fig. 6B), probably because of its stability.
The time course of the changes in protein levels correlates well with the development of the mutant phenotype ( Fig. 2 and 4). Hsa32 is a novel Hsp mainly found in land plants (Liu et al., 2006a), whose molecular function is currently unknown (Liu et al., 2006b). It was previously shown to be essential for AT after long recovery (Charng et al., 2006). Therefore, the heat-sensitive phenotype of hsfA2-1 could be attributed in part to the lower Hsa32 level in the mutant than in the wild type after 48 h of recovery.
The lower transcript abundance of other Hsp genes, such as Hsp25.3-P and Hsp18.1-CI, in the mutant during recovery (Fig. 6A) may result in lower levels of encoded proteins and thus lead to a defect in thermotolerance. However, a T-DNA insertion mutant of Hsp18.1-CI (SALK_042001) did not show a heat-sensitive phenotype under the tested condition (Table I and data not shown), which may be due to gene redundancy of class I small Hsps in Arabidopsis (Scharf et al., 2001) or that this gene plays a relatively minor role in AT after long recovery. Our results were consistent with the data of the transcriptome analysis of hsfA2-1 by Schramm et al. (2006), in which the transcript levels of Hsa32, Hsp25.3-P, and Hsp18.1-CI were substantially affected after a repeated HS (42 o C for 1-3 h) treatment.
Several groups (Panchuk et al., 2002;Li et al., 2005;Miller and Mittler, 2006) have reported the connection between plant Hsf and oxidative stress. Studies on transgenic Arabidopsis overexpressing AtHsfA1b suggest that Hsf is involved in the induction of Apx2, which encodes the cytosolic H 2 O 2 scavenging enzyme, ascorbate peroxidase (Panchuk et al., 2002). In the studies of Schramm et al. (2006), the heatinduction of Apx2 is totally abolished by disruption of HsfA2. Moreover, a GUS reporter gene fused to the Apx2 promoter can be transactivated by co-transfecting with AtHsfA2 driven by CaMV35S promoter in tobacco protoplasts. These results suggest that Apx2 is a potential target of HsfA2. However, it is not clear why the initial induction of Apx2 is abolished in the mutant after treatment at 44 o C for 1 h when HsfA2 was absent in the wild-type plants (Schramm et al., 2006). In our case, the induction of Apx2 by HS treatment (37 o C for 1 h) was not affected by the absence of HsfA2, as revealed by microarray and RT-PCR analysis, although it was greatly suppressed in the mutant during recovery and prolonged HS (Supplemental Fig. 4).
Heat stress induces production of H 2 O 2 (Vacca et al., 2004;Volkov et al., 2006) that causes oxidative stress to the cell. HS-induced Apx2 may be required to remove the reactive oxygen species (ROS) by compensating the activity of thermolabile Apx1 (Panchuk et al., 2002). The H 2 O 2 level was shown to be higher in the HsfA2 knockout mutant than in the wild-type plants after HS treatment and the mutant is more sensitive to oxidative stress (Li et al., 2005), suggesting that the decreased Apx2 expression may be responsible. However, we did not observe a substantial decrease of AT in the Arabidopsis knockout mutant of Apx2 after short or long recovery ( and data not shown), which suggests that suppression of the HS-induced ROS scavenger alone does not account for the phenotype of hsfA2-1.

According to the transcriptome studies of Busch et al. (2005), 6 Hsf genes --
HsfA2, HsfA4a, HsfA7a, HsfB1, HsfB2a, and HsfB2b --are significantly upregulated in Arabidopsis thaliana leaves (ecotype Wassilewskija, 5-6 weeks old) by HS treatment (37 o C for 1 h). This result is comparable to our microarray data (Supplemental Fig. 3B), except that in our Arabidopsis sample (ecotype Col-0, 5-d seedlings), HsfA7b but not HsfA4a was significantly responsive to heat. Our results are consistent with the microarray data generated by von Koskull-Doring's group for heat-stress response of Arabidopsis (ecotype Col-0, 16-d seedlings) at 37 o C for 1 h, which was examined using the Genevestigator tool (Zimmermann et al., 2004). This  (Baniwal et al., 2004). From the HSR microarray data available to date, HsfA2 is the gene most induced by heat, which suggests a dominant role for the gene in sustaining the HS response, a role that cannot be replaced by HsfA7a and

HsfA7b. This idea is supported by the observations that the T-DNA knockout lines of
HsfA7a and HsfA7b did not show a thermotolerance defect like that of hsfA2-1 (Table   I, and data not shown).

expression of heat-inducible reporter gene when transiently co-expressed with
AtHsfA4a (Czarnecka-Verner et al., 2000). It is possible that HsfB1 acts differentially as a positive or negative transcription coactivator of some members of the Hsf family.
The T-DNA knockout line of HsfB1 did not show thermotolerance defect like hsfA2-1 under our test condition (Table I and data not shown), suggesting that HsfB1 is not essential for extending the duration of AT like HsfA2 does. This is consistent with our observation that the expression of Hsa32, Hsp25.3-P and Hsp18.1-CI was not substantially affected in the HsfB1 T-DNA knockout mutant during recovery or prolonged heat stress, unlike in the knockout mutant hsfA2-1 (unpublished data). The function of AtHsfB2a/b is currently not clear. No significant heat-sensitive phenotype was observed for the AtHsfB2b knockout line (Table I and  This observation suggests that other Hsfs, such as HsfA2, act as major transactivators during prolonged heat stress. This hypothesis is supported by the finding that in the absence of HsfA2 the expression of tested Hsp genes was substantially lower than in the wild type after 2 h of prolonged HS (Fig. 7). HsfA2 acting as a transactivator has been well characterized by previous studies (Boscheinen et al., 1997;Rojas et al., 2002;Kotak et al., 2004;Schramm et al., 2006). It is tempting to speculate that HsfA1s initiates HSR by inducing downstream genes including HsfA2, which then acts as a positive feedback regulator to sustain the expression of heat-induced genes during recovery or prolonged HS. However, it is to be noted that induction of tomato HsfA2 by HS requires HsfA1a (Mishra et al., 2002), but in Arabidopsis, induction of HsfA2 seems to be independent of HsfA1a/b (Busch et al., 2005). Whether the other two A1-type Hsfs --HsfA1d/HsfA1e --are responsible for the heat induction of HsfA2 is currently not known. Our preliminary data showed that neither the HsfA1d nor HsfA1e knockout led to alteration of the heat-induced expression of HsfA2.
Further study on a double knockout line of these two genes is underway. Given the status of HsfA2 as a heat-inducible gene and its function in sustaining the expression level of tested Hsp genes ( Fig. 6 and 7), the role of HsfA2 in extending the duration but not in induction of AT is conceivable. This inference is well supported by the immunoblot result of Schramm et al. (2006) that HsfA2 is a heatinducible protein greatly accumulated in leaves after 3 h of HS at 42 o C. Their data show no or very low abundance of HsfA2 protein without HS or after 1 h of HS, which is in good agreement with our conclusion that HsfA2 is not involved in the initial induction of HSR and AT developed at early phase. This explains why the mutant did not show heat-sensitive phenotype when a short recovery was associated with acclimation treatment both in this study as well as in that of Schramm et al. (2006). However, these observations were not in agreement with the results of Li et al. (2005), who also studied the function of HsfA2 using the same T-DNA knockout line (Salk_008978). In their study, the mutant plants displayed significantly reduced AT, but their thermotolerance phenotype was demonstrated under rather different conditions (in 7-day-old seedlings, acclimatized at 37 o C for 1 h, then challenged at 46 o C for 2 h without recovery). In addition, basal thermotolerance of the mutant was also reduced. The authors then showed that the HS (37 o C)-induced transcript levels of Hsp70 (At3g12580) and Hsp101 (At1g74310) in the HsfA2 knockout mutant were substantially reduced. These results are inconsistent with the data reported here and by Schramm et al. (2006), in which the expression of most genes, including Hsp70 and Hsp101, were not changed and no significant thermotolerance defect was observed in the mutant after short recovery or without acclimation. One possible cause of this discrepancy may be the different methods or apparatus used for thermotolerance assay. Alternatively, the discrepancy may be due to different T-DNA insertion number in the knockout mutants that we used. We did observe multiple copies of T-DNA in homozygous hsfA2-1 plants by Southern blot (data not shown), and segregation of these T-DNA insertions in its offspring may have resulted in different genetic backgrounds that lead to weaker stress tolerance in some of them.
Unfortunately, a complementation experiment was not conducted by Li et al. (2005) to exclude this possibility.
Taken together, our results suggest that HsfA2 is one of the components involved in extending the duration of AT in plants by positively regulating the expression of at least some, if not all, heat-induced genes after their initial induction by HsfA1(s).

Plant materials, transformation, and growth condition
The Arabidopsis thaliana (Col-0) HsfA2 (At2g26150) T-DNA insertion line Salk_008978 (Alonso et al., 2003) was obtained from the Arabidopsis Biological Resource Center (ABRC, Ohio State U.). The T-DNA insertion lines SALK_066374 for Hsp101 (At1g74310) and GABI-Kat 268A08 for Hsa32 (At4g21320) were obtained as previously described (Charng et al., 2006). Information on all the other T-DNA insertion lines obtained from ABRC or GABI-Kat program (Rosso et al., 2003) is given in Table I Table II. For complementation of the hsfA2-1 line, the genomic DNA of Arabidopsis wild-type (Col-0) HsfA2 was amplified by PCR with a pair of gene-specific primers (Supplemental Table II). The 2,177 bp PCR product was cloned into pCR8/GW/TOPO (Invitrogen), then sequenced to confirm no mis-sense or non-sense mutation in coding region nor any point mutation in the promoter region. The cloned HsfA2 genomic DNA was then subcloned into pBGW,0 (Karimi et al., 2005) by an in vitro recombinase method to yield the binary vector pYC112, which was then transferred into A. tumefaciens LBA4404 strain for Arabidopsis (Col-0) transformation by the "floral dip" method (Clough and Bent, 1998). Transformants were selected on 0.5x Murashige and Skoog (MS) medium plates containing 25 µg/mL Basta. T2 seeds of independent lines underwent herbicide resistance and thermotolerance tests. For propagating seeds, Arabidopsis plants in soil were grown in walk-in growth chambers at 22 o C and 14 h of light (120 µmol m -2 s -1 ).

Thermotolerance test
All the knockout lines were grown under the same condition as the wild-type plants for harvesting seeds for thermotolerance test. For the AT test on Arabidopsis seedlings, survival rate and quantitative hypocotyl elongation assay were performed as previously described (Charng et al., 2006). A modified quantitative hypocotyl elongation assay was applied to evaluate the basal thermotolerance of etiolated seedlings. The seeds were sown and grown vertically in the dark for 3 d at 24 o C, and then the plants were heated in a water bath at 44 o C for 15 to 30 min or at 46 to 50 o C for 5 min. After this heat treatment, the hypocotyl positions were labeled. The plants were kept grown vertically for 3 d at 24 o C, and then the length of hypocotyl elongation was measured. To evaluate the effect of continuously moderate heat stress, 3-d old seedlings were grown on agar plate up to 3 weeks in growth chambers with temperature set at 35 o C during the light period (14 h, 120 mmol m -2 s -1 ) and 33 o C during the dark.

Measurement of ion leakage
Three-day-old seedlings were conditioned at 37 o C for 1 h, recovered for 2 d, then heated at 44 o C for 45 min. At each time point during the recovery after 44 o C HS, 30 seedlings were collected and immersed in 5 mL of deionized water. The samples were kept at room temperature for 5 h, then ion leakage level was measured by a conductivity meter Suntex,Taiwan). The seedlings were autoclaved, then the total ion leakage was measured. The relative ion leakage was calculated as (ion leakage level/total)*100%.

RNA isolation and RT-PCR analysis
Total RNA was isolated with TRIZOL reagent (Invitrogen) from plant samples frozen in liquid nitrogen according to the manufacturer's protocol. The presence or absence of transcript of the genes listed in Table I was analyzed by RT-PCR using RNA isolated from the wild-type and mutant plants with or without HS (37 o C for 1 h).
The sequences of the primers for this purpose were shown in Supplemental Table II. Semi-quantitative RT-PCR analysis of the marker Hsp genes was performed basically as previously described (Wang et al., 2001). The sequences of the primers for RT-PCR analysis for each gene are also described in Supplemental Table II. PCR product was analyzed by agarose gel electrophoresis and staining with ethidium bromide.
Florescence images of PCR products were digitized and quantified with Image Gauge V3.12 (Fujifilm). Transcriptome profiling was performed by use of the ATH1 GeneChip array (Affymetrix, Santa Clara, CA) as previously described (Charng et al., 2006). For each single treatment, 2 biological replicates were used and 1 chip for each replicate. The microarray data were processed and analyzed by use of Spotfire DecisionSite 8.0 as previously described (Charng et al., 2006). The microarray data reported here were deposited in the Gene Expression Omnibus at the National Center for Biotechnology Information (GSE4760).

Immunoblotting
The total proteins of plant samples were extracted and quantified and then underwent immunoblot analysis for Hsp90, Hsa32 and sHsp-CI as previously described (Charng et al., 2006).     undergoing ion leakage analysis as described in Materials and Methods. Error bars represent standard deviation based on data in 3 separate duplicates. The plants with asterisks had significantly higher ion leakage than wild-type plants (P < 0.05, independent Student's t test).   without heat treatment and collected at the same time of the treated samples collected at 48 h. The expected sizes of the PCR products are indicated at the right. B, Immunoblot analysis of the protein levels of Hsp90, Hsa32, and class I small Hsps detected with corresponding antibodies. The samples were the same as described in A, except the additional samples were collected at 72 h of recovery. In each lane, 50 µg of total protein was loaded. RuBisCO large subunit stained by Amido black was shown to ensure equal loading. The control (C) samples were seedlings without heat treatment and collected at the same time of the treated samples collected at 72 h.
Similar results were obtained from 2 biological repeats, with 1 shown here. Supplemental Fig. 2 The SALK_088054 line was more sensitive than the wild type to severe HS challenge after a short but not long recovery following acclimation treatment. The phenotypes of Arabidopsis seedlings of the wild type (wt), SALK_088054, and T-DNA knockout lines of Hsa32 (hsa32-1) and HsfA2 (hsfA2-1) were revealed after treatment by different HS regimes with short or long recovery schematically shown on the right of each panel. The arrowheads indicate the end of seed imbibitions. All the seedlings were germinated, conditioned, and challenged in the same plate. The photographs were taken 7 d after the challenge and reorganized for presentation.

Supplemental Fig. 3 Transcriptome profile analysis and comparison between the
hsfA2-1 and wild-type plants. 5-d-old Arabidopsis seedlings were subjected to 37 o C for 1 h (HS) or non-treatment (control). Total RNA was purified for microarray analysis with the ATH1 chip of Affymetrix. A, The correlation of signals from 2 biological replicates of the control samples of the wild type (wt-c) or hsfA2-1 (HsfA2KO-C) and HS-treated samples of the wild type (wt-HS) or hsfA2-1 (HsfA2KO-HS). B, The expression profiles of Hsf genes derived from the data in A. Fig. 4 The expression of heat-induced Apx2 was substantially affected in hsfA2-1. Semi-quantitative RT-PCR analysis of the mRNA levels of Apx2