AtHsp70-15-deficient Arabidopsis plants are characterized by reduced growth, a constitutive cytosolic protein response and enhanced resistance to TuMV


(fax +49-9131-8528254; e-mail


Arabidopsis thaliana contains 18 genes encoding Hsp70s. This heat shock protein superfamily is divided into two sub-families: DnaK and Hsp110/SSE. In order to functionally characterize members of the Hsp70 superfamily, loss-of-function mutants with reduced cytosolic Hsp70 expression were studied. AtHsp70-1 and AtHsp70-2 are constitutively expressed and represent the major cytosolic Hsp70 isoforms under ambient conditions. Analysis of single and double mutants did not reveal any difference compared to wild-type controls. In yeast, SSE protein has been shown to act as a nucleotide exchange factor, essential for Hsp70 function. To test whether members of the Hsp110/SSE sub-family serve essential functions in plants, two members of the sub-family, AtHsp70-14 and AtHsp70-15, were analysed. Both genes are highly homologous and constitutively expressed. Deficiency of AtHsp70-15 but not of AtHsp70-14 led to severe growth retardation. AtHsp70-15-deficient plants were smaller than wild-type and exhibited a slightly different leaf shape. Stomatal closure under ambient conditions and in response to ABA was impaired in the AtHsp70-15 transgenic plants, but ABA-dependent inhibition of germination was not affected. Heat treatment of AtHsp70-15-deficient plants resulted in drastically increased mortality, indicating that AtHsp70-15 plays an essential role during normal growth and in the heat response of Arabidopsis plants. AtHsp70-15-deficient plants are more tolerant to infection by turnip mosaic virus. Comparative transcriptome analysis revealed that AtHsp70-15-deficient plants display a constitutive stress response similar to the cytosolic protein response. Based on these results, AtHsp70-15 is likely to be a key factor in proper folding of cytosolic proteins, and may function as nucleotide exchange factor as proposed for yeast.


Molecular chaperones assist protein folding and are present in all living cells. They interact with folding intermediates to prevent non-productive interactions and aggregation of polypeptides. Several chaperones were first identified as heat shock proteins (Hsps) as a result of their strong induction during heat responses (reviewed by Kotak et al., 2007). According to their size, they were initially grouped into small and 60, 70, 90 or 100 kDa Hsps. Today, Hsp families are categorized based on homology. Members have been shown to play essential roles during various forms of cellular stress and normal growth. Adaptation to stress conditions requires inducible expression of stress-responsive genes.

Heat is a major factor leading to mis-folded and aggregated proteins. Regulation of heat- and stress-responsive genes is mainly driven by heat stress transcription factors (Hsfs). These bind to so-called heat stress-responsive elements (HSEs) found in the promoters of heat-inducible genes of eukaryotes. In contrast to yeast and humans, which carry one or a few Hsf genes (Liu et al., 1997), Hsfs are encoded by large gene families in plants. Twenty-one members were found in Arabidopsis thaliana (Nover et al., 2001). Expression of Hsfs appears to be negatively regulated by Hsp70s, which may provide a feedback regulatory mechanism to fine-tune heat responses (Lee and Schoeffl, 1996).

Hsp70s are characterized by a typical N-terminal ATPase domain of approximately 45 kDa. They are involved in folding and refolding, preventing aggregation, and transport and controlled degradation of proteins (Mayer and Bukau, 2005). To fulfil their specific functions, they interact with various co-chaperones, co-factors or nucleotide exchange factors.

Arabidopsis possesses 18 genes encoding Hsp70 homologues, including one proposed pseudogene (Lin et al., 2001). The Hsp70 protein superfamily is divided into two sub-families: DnaK and Hsp110/SSE. The DnaK sub-family has 13 members, including AtHsp70-1 to AtHsp70-5 (also named Hsc70-1 to Hsc70-3, Hsp70 and Hsp70b), which have been predicted to localize to the cytosol and nucleus, AtHsp70-6 to AtHsp70-8 (also called cpHsp70s), which localize to plastids (Su and Li, 2008), AtHsp70-9 and AtHsp70-10 (also called mtHsc70), which are mitochondrial isoforms, and AtHsp70-11 to AtHsp70-13 (also called binding protein (BiP)), which are thought to reside in the lumen of the endoplasmic reticulum (ER). Members of the DnaK sub-family are either constitutively expressed or stress-inducible (Sung et al., 2001; Zimmermann et al., 2004). The Hsp110/SSE sub-family consists of four members, AtHsp70-14 to AtHsp70-17. They exhibit homology to yeast SSE1 and human Hsp110, and have been predicted to localize to the cytosol, with the exception of AtHsp70-17, which is thought to accumulate in the ER (Lin et al., 2001). AtHsp70-14 and AtHsp70-15 were the first to be described by Storozhenko et al., (1996) and since then AtHsp70-14 was referred to as Hsp91 because of its size.

Cellular functions of Hsp70s in plants have been studied extensively, but little is known about the specific functions of single Hsp70 isoforms in vivo. Antisense inhibition of cytosolic Hsp70s revealed a function in thermotolerance of Arabidopsis thaliana (Lee and Schoeffl, 1996), confirmed by over-expression of cytosolic Hsp70-1 (Sung and Guy, 2003; Cazale et al., 2009). Furthermore, cytosolic Hsp70s have been shown to be involved in the immune response in Arabidopsis, and are required for productive potyvirus infection of tobacco plants (Noel et al., 2007; Hafren et al., 2010). Over-expression of ER-localized BiP in transgenic tobacco alleviated ER stress and led to enhanced drought and heat tolerance (Leborgne-Castel et al., 1999; Alvim et al., 2001). The highly stress-inducible Hsp70-4 has been proposed to function in protein degradation and embryogenesis (Lee et al., 2009). Knockout of plastid-localized Hsp70s revealed an essential role for chloroplast development and thermotolerance of seeds (Su and Li, 2008; Latijnhouwers et al., 2010).

Our study aimed to obtain functional data on two major cytosolic Hsp70s and the barely described Hsp110/SSE genes in Arabidopsis. We verified the cytosolic and nuclear localization of AtHsp70-1, -2, -14 and -15 by GFP fusion experiments. AtHsp70-1 and AtHsp70-2 single and double knockout mutants did not exhibit any visible growth phenotype. By analysing T-DNA insertion and amiRNA-AtHsp70-14/15 lines, however, we were able to demonstrate that AtHsp70-15 is essential for plant growth and stomatal responsiveness, and that it acts as a compatibility factor for the interaction with turnip mosaic virus (TuMV). Comparative transcriptome analysis showed a constitutive heat stress response/cytosolic protein response in AtHsp70-15-deficient plants. Based on these results and the strong homology of AtHsp70-15 to yeast SSE1, we propose that AtHsp70-15 may serve as an essential nucleotide exchange factor for cytosolic members of the DnaK sub-family in Arabidopsis.


GFP–AtHsp70-1 and -2 fusion proteins localize predominantly to the cytosol

Based on in silico analysis, the two closely linked and constitutively expressed Hsp70 isoforms encoded by AtHsp70-1 (At5g02500) and AtHsp70-2 (At5g02490) were predicted to localize to the cytosol (Lin et al., 2001). To analyse the subcellular localization of both isoforms, GFP was fused to their full-length coding sequences. The construct was placed under the control of the constitutive CaMV 35S promoter as described in Experimental Procedures, and was stably transformed into Nicotiana tabacum plants using Agrobacterium-mediated gene transfer. Of the transformants, 20 and 12 individuals with detectable GFP fluorescence (presumably in the cytosol and the nucleus) were obtained, and three of them were analysed by confocal laser scanning microscopy. As evident from the confocal laser scanning microscopy images shown in Figure 1(a,b), both isoforms localize to the cytosol and the nucleus. To exclude the possibility of proteolytic breakdown of the fusion proteins, immunoblot analysis using anti-GFP-specific antibodies was performed. As shown in Figure 1(c), GFP-specific antibodies recognize a single protein band of approximately 110 kDa in protein extracts from plants expressing GFP–AtHsp70-1 or GFP–AtHsp70-2, corresponding to the size of the fusion protein.

Figure 1.

 Expression of GFP–AtHsp70-1/2 fusion proteins and characterization of hsp70-1/2 knockout mutants.
(a, b) Confocal laser scanning microscopy images of epidermis of transgenic Nicotiana tabacum plants expressing GFP–AtHsp70-1 (a) or GFP–AtHsp70-2 (b). Cell walls were stained red with propidium iodide. Scale bars = 10 μm.
(c) Immunoblotting of heterologously expressed GFP–AtHsp70-1 (lane A) and GFP–AtHsp70-2 (lane B) was performed using an anti-GFP antibody. No free GFP was detected (arrowhead).
(d) RT-PCR analysis of the Arabidopsis T-DNA insertion lines hsp70-1 knockout (SALK_135531), hsp70-2 knockout (SALK_085076) and hsp70-1/2 knockout (SALK_087844), amplified using primers specific for Hsp70-1, Hsp70-2 and the loading control Arabidopsis Actin 1.
(e) Schematic diagram of chromosome 5 in the region of AtHsp70-1/2. The positions of the T-DNA insertions in the SALK lines used are indicated; all lie in the 2nd exon of the respective genes. Genes comprised in the deletion are indicated by broad arrows. The left borders of T-DNA insertions are indicated by curved arrows. In SALK_087844, there is a deletion of approximately 10 kb upstream of AtHsp70-2, which ends in the 2nd intron of At5g02520.

The double hsp70-1/2 knockout mutant displays a wild-type-like phenotype

In order to functionally characterize AtHsp70-1 and AtHsp70-2, we analysed T-DNA insertion lines from the SALK collection (Alonso et al., 2003). Homozygous knockout lines for AtHsp70-1 and AtHsp70-2 were selected by genomic PCR. Sequence analysis confirmed insertions in SALK_135531 and SALK_085076 located in the 2nd exons of AtHsp70-1 and AtHsp70-2, respectively (data not shown). RT-PCR analysis confirmed the absence of respective transcripts in the knockout lines (Figure 1d). Surprisingly, sequencing of the T-DNA flanking region in SALK_087844 revealed a large deletion in front of the T-DNA insertion in exon 2 of AtHsp70-2. This deletion spans six open reading frames, including AtHsp70-1 (Figure 1e). Therefore, SALK_087844 can be considered as an hsp70-1/hsp70-2 double mutant, and this was confirmed by RT-PCR analysis (Figure 1d). Neither the single nor the hsp70-1/hsp70-2 double knockout mutants showed an altered growth phenotype under ambient conditions (Figure S1a). Moreover, TuMV tolerance of the double mutant did not differ from that of wild-type plants (Figure S1b).

In the light of these observations, it seems very likely that the defects in AtHsp70-1 and AtHsp70-2 can be complemented by additional cytosolic Hsp70 isoforms. To overcome the possible redundancy among Hsp70s, we focused on the cytosolic members of the Hsp110/SSE sub-family. In yeast, SSE1 was shown to act as an essential nucleotide exchange factor required for Hsp70 activity (Raviol et al., 2006a). In a co-expression analysis, expression of AtHsp70-14/15 clustered with that of AtHsp70-1 (Swindell et al., 2007). Therefore, AtHsp70-14 and AtHsp70-15 were selected for further analysis.

AtHsp70-14 and AtHsp70-15 exhibit high sequence similarity

AtHsp70-14 (At1g79930) and AtHsp70-15 (At1g79920) are localized next to each other on the same strand of chromosome 1. According to recent annotations by the Arabidopsis Information Resource (TAIR,, both genes encode proteins with 831 amino acids, but, based on published ESTs, two slightly shorter splicing variants may also be expressed. At the nucleotide level, the isoforms show sequence identity of 69% in the untranslated 5′ and 3′ regions and 96% in the coding region. At the protein level, the isoforms are 97% identical. Due to their high sequence similarities, they are represented by only one oligonucleotide (array identifier 262054_s_at) on the Affymetrix ATH1 microarray chip. In silico expression analysis using publicly available microarray data revealed that either or both of the genes are expressed ubiquitously in nearly all organs and at all developmental stages, with highest expression levels in callus and cell suspension cultures. In addition, the transcript signal is up-regulated after heat shock (Zimmermann et al., 2004).

AtHsp70-14–GFP and AtHsp70-15–GFP fusion proteins localize predominantly to the cytosol

To analyse the subcellular localization of AtHsp70-14 and AtHsp70-15, GFP was fused to the C-terminal end of the full-length coding regions of both genes, cloned under the control of the constitutive CaMV 35S promoter, and transiently expressed in Nicotiana benthamiana. As shown in Figure 2(a,b), both genes target GFP to the cytosol and nucleus. The presence of the approximately 140 kDa fusion protein was verified by immunoblotting using a GFP-specific antibody (Figure 2c).

Figure 2.

 Transient expression of AtHsp70-14–GFP and AtHsp70-15–GFP fusion proteins.
(a, b) Confocal images of AtHsp70-14–GFP (a) and AtHsp70-15–GFP (b) after transient expression in N. benthamiana epidermal cells. Scale bars = 10 μm. Cell walls were stained with propidium iodide.
(c) Immunoblot analysis of transiently expressed AtHsp70-14–GFP and AtHsp70-15–GFP was performed using an anti-GFP antibody. In addition to the approximately 140 kDa fusion protein, there are smaller bands representing putative degradation products.

Transgenic amiRNA-AtHsp70-14/15 Arabidopsis thaliana plants show reduced Hsp70-14/-15 transcript accumulation and inhibition of growth

Based on the high sequence similarity between AtHsp70-14 and AtHsp70-15, functional redundancy of the isoforms is expected. As the isoforms are localized adjacent to each other in the genome, T-DNA double knockout lines are virtually impossible to obtain. To circumvent this problem, an artificial microRNA (amiRNA) construct targeting both isoforms was designed using the WMD2 - Web MicroRNA Designer software (Schwab et al., 2006). The resulting amiRNA was cloned under the control of the CaMV 35S promoter in vector pAF16 (Stadler et al., 2005) (Figure 3a), and used for Agrobacterium-mediated transformation of Arabidopsis thaliana Col-0. Following floral-dip transformation, transgenic seeds were selected by BASTA treatment, and growth of transgenic amiRNA-AtHsp70-14/15 plants was compared with that of a wild-type-like transformed control line (PME1-9), which was also treated with BASTA. As shown in Figure 3(b), four independent transgenic amiRNA-AtHsp70-14/15 lines (2, 4, 6 and 12) were obtained, showing 3:1 Mendelian inheritance of the transgene and displaying a reduced growth phenotype and altered leaf morphology compared to the control line. Down-regulation of AtHsp70-14/15 in BASTA-selected transgenic plants was confirmed by Northern blot analysis (Figure 3c). As the growth and wilting phenotypes of these lines were similar, line 4 was used for closer analysis.

Figure 3.

 Characterization of amiRNA-AtHsp70-14/15 plants.
(a) Schematic overview of the amiRNA-AtHsp70-14/15 T-DNA, which contains the amiRNA-AtHsp70-14/15 cassette (ami) driven by the CaMV 35S promoter (35S) and a BASTA resistance gene driven by a nos promoter between the right (RB) and left borders (LB): the amiRNA cassette encodes amiRNA-AtHsp70-14/15 with the loop formed by the amiR319a precursor backbone.
(b) Four independent 8-week-old amiRNA-AtHsp70-14/15 plants (lines 2, 4, 6 and 12, from left to right) compared with the transformed control plant (left).
(c) Northern blot analysis corresponding to the plants shown in (b), loaded with 20 μg RNA per lane and hybridized with an AtHsp70-14/15-specific probe; the ethidium bromide,-stained RNA gel is shown as a loading control.

Silencing of AtHsp70-14/15 impairs stomatal closure and accelerates wilting

As detached leaves from amiRNA-AtHsp70-14/15 plants shrivelled faster than control leaves (Figure 4a), the stomatal density and aperture under ambient conditions were analysed. Microscopic inspection of abaxial leaf surfaces of control and transgenic plants revealed that stomatal density was comparable between amiRNA-AtHsp70-14/15 and control leaves, but stomatal aperture appeared to be altered. While most stomata of control leaves closed during microscopic inspection, stomata of amiRNA-AtHsp70-14/15 plants failed to close (Figure 4a). To quantify stomatal conductance, gas-exchange measurements in intact plants were performed. As shown in Figure 4(b), water vapour conductance, a measure of stomatal aperture, of amiRNA-AtHsp70-14/15 plants was strongly enhanced compared to controls. Treatment of leaves with the phytohormone abscisic acid (ABA) is known to induce stomatal closure. To test the ABA responsiveness of AtHsp70-14/15 silenced plants, control and silenced plants were treated with 50 μm ABA. As expected, stomatal closure of control plants was mediated by ABA treatment (Figure 4b). In contrast, the stomatal conductance of amiRNA-AtHsp70-14/15 plants remained basically unchanged following ABA treatment (Figure 4b). To exclude a general failure of the silenced plants to respond to ABA, sensitivity of seed germination to ABA was tested. As described previously (Hugouvieux et al., 2001; Liu et al., 2007), germination of Arabidopsis control seeds is delayed by increasing amounts of ABA (Figure 4c). Similarly, germination of seeds of AtHsp70-14/15 silenced plants was sensitive to ABA (Figure 4c), indicating no general defect in ABA responsiveness. This was further supported by similar dose–response effects of ABA on leaf senescence in control and silenced plants (Figure S2).

Figure 4.

 Wilting phenotype and ABA responsiveness of stomatal aperture and germination.
(a) Detached leaves were incubated under ambient conditions for 3 h, leading to increased wilting of amiRNA-Hsp70-14/15 plants (right) compared to controls (left); magnified paradermal views of the abaxial epidermis with representative stomata.
(b) Water vapour conductance [g(H2O)] was determined by infra-red gas-exchange measurements of light-adapted plants. In transformed control plants, the difference between leaves 1 h after treatment with 50 μm ABA and mock-treated control leaves is significant (< 0.05), but this is not the case when comparing the same treatments of amiRNA-AtHsp70-14/15 plants. However, water conductance is much higher in amiRNA-AtHsp70-14/15 plants than in the transformed control. Mean values are shown; error bars represent the SE (= 12).
(c) Five-day-old seedlings of amiRNA-AtHsp70-14/15 and control plants cultivated on MS medium without ABA (left), with 0.5 μm ABA (middle) or with 1 μm ABA (right). Scale bars = 1 mm.

AtHsp70-14/15 silenced plants are more sensitive to heat treatment

As discussed above, AtHsp70-14/15 expression is enhanced under heat stress conditions. To analyse the heat tolerance of AtHsp70-14/15 silenced plants, 4-week-old plants (55 control and 55 silenced) were subjected to heat stress (38°C) for 5 days, followed by a recovery phase (18°C night/22°C day) of 5 days (Figure 5a). During heat stress, control plants showed paraheliotropic leaf movement and reduced growth (Figure 5b). During the recovery phase, nearly all control plants (93%) survived, without visible or only slight leaf damage. AtHsp70-14/15 silenced plants did not move their leaves during heat treatment, and only a few plants (9%) survived the recovery phase. This indicates that AtHsp70-14/15 play an essential role in cellular adaptation to heat stress.

Figure 5.

 Heat treatment of amiRNA-AtHsp70-14/15 plants.
(a) Time frame of heat treatment of 4-week-old plants.
(b) amiRNA-AtHsp70-14/15 and control plants were photographed before treatment (left), on day 3 of treatment at 38°C (middle, lateral view), and 5 days after heat treatment (right). Scale bars = 0.5 cm.

amiRNA-AtHsp70-14/15 plants are more tolerant to TuMV infection

In previous studies, we showed that potyvirus infection requires functional Hsp40 isoforms and cytosolic Hsp70s (Hofius et al., 2007; Hafren et al., 2010). In order to study the effect of reduced AtHsp70-14/15 expression during potyvirus infection, silenced Arabidopsis plants were inoculated with TuMV, and plant phenotype and virus spread were monitored. Potyviruses are known to encode potent silencing suppressors. Therefore, efficient silencing of AtHsp70-14/15 in TuMV-infected plants was verified by Northern blot analysis (Figure 6a). After 4 weeks of TuMV infection, control plants developed clear TuMV symptoms, such as reduced growth and rolled leaves, whereas AtHsp70-14/15 silenced plants developed only mild symptoms (Figure S3). To quantify virus-induced losses of biomass, Mock-inoculated plants (uninfected control) or plants inoculated with virus were harvested 3 weeks after treatment and total fresh weight was determined. This analysis revealed virus-induced losses of biomass of 37% and 13% in control and silenced plants, respectively. Virus proliferation in control and silenced plants was tested 3 weeks post-infection by Northern blot analysis (Figure 6b) and ELISA assays (Figure 6c) using virus-specific probe or antibody, respectively. In addition, the viral titre of inoculated leaves was determined by ELISA 2 weeks after inoculation. Both experiments revealed a strong reduction in virus titre. Based on ELISA assays, virus accumulation in local (inoculated) and systemic leaves of AtHsp70-14/15 silenced plants was reduced by 86% and 78%, respectively, compared to the control. These results clearly demonstrate that AtHsp70-14 and/or AtHsp70-15 are important for potyvirus infection of Arabidopsis plants.

Figure 6.

 TuMV infection of amiRNA-AtHsp70-14/15 plants.
Five-week-old amiRNA-AtHsp70-14/15 and transformed control plants were inoculated with buffer or TuMV.
(a, b) RNA from systemic leaves 3 weeks after inoculation was used for a Northern blot analysis and hybridized using probes against AtHsp70-14/15 (a) and TuMV capsid protein (b). The corresponding RNA gel (20 μg RNA per lane) stained with ethidium bromide is shown as a loading control.
(c) TuMV capsid protein titre was measured by an ELISA assay, and that of the transformed control was set to 100% (bars indicate standard errors; = 30, *< 0.05). Discs from inoculated leaves were harvested 2 weeks after inoculation (left columns) and samples of systemic leaves were taken 3 weeks after inoculation (right columns).

Athsp70-15 but not Athsp70-14 knockout plants display a similar phenotype to amiRNA-AtHsp70-14/15

The possible roles of each individual Hsp70 isoform could not be addressed using an amiRNA targeted against AtHsp70-14 and AtHsp70-15. To analyse whether single mutations in either gene result in molecular and/or phenotypic changes similar to those described for the AtHsp70-14/15 silenced plants, T-DNA insertion lines for AtHsp70-14 (SALK_082815) and AtHsp70-15 (SAIL_830_C08) were selected from mutant collections. The T-DNA insertion in both lines was verified by sequence analysis of genomic DNA. In both cases, homozygous plants were identified by PCR using DNA. Loss of transcripts was revealed by RT-PCR analysis (Figure 7d). The hsp70-14 knockout mutant was indistinguishable from wild-type, but the hsp70-15 knockout mutant showed a clear phenotype similar to that of the amiRNA-AtHsp70-14/15 plants. Athsp70-15 knockout plants showed enhanced wilting, growth reduction and enhanced virus tolerance (Figure 7a–c). Moreover, the genomic sequence of AtHsp70-14 in hsp70-15 mutants was not altered compared to wild-type, indicating that AtHsp70-15 suppression accounts for these effects, while the role of AtHsp70-14 remains unclear.

Figure 7.

 Characterization of hsp70-14 and hsp70-15 knockout mutants.
Homozygous T-DNA insertion lines for AtHsp70-14 (SALK_082815) and AtHsp70-15 (SAIL_830_C08) were analysed.
(a) hsp70-15 knockout (KO) plants are smaller than wild-type (WT) and hsp70-14 knockout plants.
(b) Detached leaves were incubated under ambient conditions for 3 h; top: before treatment; bottom: after treatment.
(c) TuMV capsid protein titre of systemic leaves 3 weeks after TuMV inoculation was measured by an ELISA assay (bars indicate standard errors; = 20, *< 0.05).
(d) RT-PCR analysis of the Arabidopsis T-DNA insertion lines using primers specific for Hsp70-14 (top), Hsp70-15 (middle) and loading control Actin 1 (bottom).

AtHsp70-15 deficiency leads to constitutive up-regulation of heat response-related transcripts

Having observed the physiological phenotypes of the amiRNA-AtHsp70-14/15 and hsp70-15 mutant plants, we wished to determine which gene functions and signalling outputs were de-regulated in these transgenic plants. Microarray analysis was performed using the Agilent Arabidopsis IV microarray chip. For each sample (wild-type, amiRNA-AtHsp70-14/15 line 4, hsp70-14 and hsp70-15 knockout plants), fully expanded leaves of ten 6-week-old plants grown on soil under normal growth conditions were pooled, with four replicates each.

Using the average linkage clustering algorithm and centred Pearson correlation in Genespring XI (Agilent Technolgies,, the data clustered well into two major groups. One group consisted of wild-type and hsp70-14 knockout plants, the other consisted of amiRNA-AtHsp70-14/15 and hsp70-15 knockout plants. Thus, the two major clusters comprise wild-type-like and Hsp70-15-deficient phenotypes, respectively (Figure S4).

Differentially regulated genes between either amiRNA-AtHsp70-14/15, hsp70-14 or hsp70-15 and control plants with >2.0-fold change and P < 0.05 were identified using a Welch test after Benjamini–Hochberg multiple test correction (Table S1). For hsp70-14 knockout plants, only eight features corresponding to seven gene loci were significantly de-regulated, supporting the assumption of a wild-type-like phenotype of hsp70-14 knockout plants. Due to the high sequence similarities between AtHsp70-14 and AtHsp70-15, their transcript signals cannot clearly be differentiated by the microarray analysis. This explains the confusing fact that features assigned as AtHsp70-14 are both up- and down-regulated in the list of de-regulated genes in hsp70-14 knockout plants (Table S1).

In contrast to the wild-type-like hsp70-14 mutants, amiRNA-AtHsp70-14/15 and hsp70-15 knockout plants clearly differed from the wild-type control by 585 and 1137 differentially expressed features, corresponding to 494 and 974 gene loci, respectively. One hundred and sixty-nine significantly up-regulated genes and 217 down-regulated genes were common in both Hsp70-15-deficient plant lines. However, approximately 98% of those features which are significantly different compared to wild-type in only one of the two Hsp70-15-deficient genotypes showed similar changes in gene expression, although values were not significantly different in both comparisons. Therefore, the transcript pattern in amiRNA-AtHsp70-14/15 and hsp70-15 knockout plants is highly congruent. For further analysis, we focused on those genes that were significantly differentially expressed in both Hsp70-15-deficient genotypes.

Functional categorization of the common significantly de-regulated genes was performed using the TAIR webtool ‘Bulk Data Retrieval’ with the function ‘annotation for biological process’ ( (Figure 8a). A high number of genes diffentially expressed in both Hsp70-15-deficient genotypes fall into the cellular function categories ‘stress response’ (15% up-regulated and 15% down-regulated) and ‘response to biotic and abiotic stimulus’ (15% up-regulated and 14% down-regulated). With 15 up-regulated features, genes associated with heat stress are amongst the most consistently regulated functional category in both Hsp70-15-deficient genotypes compared to the wild-type. Within this group, only AtHsp70-14/15 and one DnaJ protein transcript were found to be significantly down-regulated. In addition to the above-mentioned categories, genes involved in lipid binding (LTP2 and LTP3) were also found to be up-regulated. LTPs typically respond to various stresses. Interestingly, a number of transcription factors, including the defence regulators WRKY22, WRKY38, WRKY40 (WRKY46) and WRKY58 were strongly down-regulated in Hsp70-15-deficient genotypes.

Figure 8.

 Transcriptional profiling of Hsp70-15-deficient plants.
(a) Analysis of down- and up-regulated genes in Hsp70-15-deficient plants according to functional categories. Functional assignment was performed using the TAIR webtool ‘Bulk Data Retrieval’.
(b) Hierarchical cluster analysis of transcript patterns, comparing regulated genes in amiRNA-Hsp70-14/15 and hsp70-15 knockout (KO) plants to publically available data for seedlings subjected to drought, salt and heat stress from the AtGenExpress dataset (Kilian et al., 2007), data for adult leaves treated with the cytosolic protein response inducer AZC, and data for adult detached leaves after 1 h of heat stress (‘leaf’). AZC and heat stress data were from Sugio et al. (2009).

In order to analyse the similarity of the transcriptional changes observed in Hsp70-15-deficient plants to transcriptional changes induced by various stress treatments, a cluster analysis was performed using publicly available data sets for heat, drought and salt treatment of seedlings (Kilian et al., 2007; Swindell et al., 2007) and recent data for heat- and AZC-treated adult Arabidopsis leaves (Sugio et al., 2009). L-azetidine-2-carboxylic acid (AZC) is a proline analogue that induces mis-folding of proteins and is used to induce the cytosolic protein response (Sugio et al., 2009). Our hierarchical cluster analysis clearly separated heat and AZC treatment from drought and salt stress (Figure 8b). Interestingly, the transcript profiles of amiRNA-AtHsp70-14/15 and hsp70-15 knockout plants fall into the heat cluster (Figure 8b), with highest similarity to AZC treatment. Twenty-four of 149 genes up-regulated in the cytosolic protein response were also up-regulated in both Hsp70-15-deficient genotypes (Table 1).

Table 1.   Common up-regulated genes in both Hsp70-15-deficient genotypes and during the cytosolic protein response
Locus identifierAnnotationFold change versus control
amiHsp70-14/15Hsp70-15 knockoutAZC treatment
  1. The table lists the genes that were significantly up-regulated in both amiRNA-AtHsp70-14/15 and hsp70-15 knockout plants and during the cytosolic protein response induced by AZC treatment (data taken from Sugio et al., 2009). Fold change values for genes from amiRNA-AtHsp70-14/15, hsp70-15 knockout and AZC microarray analyses are given.

Heat shock proteins and factors
 At5g5144023.5 kDa mitochondrial small heat shock protein (HSP23.5-M)292533
 At3g12580HSP70 (heat shock protein 70)221423
 At1g71000DnaJ heat shock N-terminal domain-containing protein11535
 At2g2950017.6 kDa class I small heat shock protein (HSP17.6B-CI)11527
 At5g02490Heat shock cognate 70 kDa protein 2 (HSC70-2/HSP70-2)1168
 At2g26150ATHSFA2 (Arabidopsis thaliana heat shock transcription factor A2)9536
 At1g74310ATHSP101 (heat shock protein 101)7530
 At5g52640HSP81-1 (heat shock protein 81-1)4318
 At5g3767015.7 kDa class I-related small heat shock protein-like (HSP15.7-CI)3214
 At2g20560DnaJ heat shock family protein2228
 At4g12400Stress-inducible protein, putative161118
 At5g48570Peptidyl-prolyl cis/trans isomerase, putative8635
 At3g53230Cell division cycle protein 48, putative7519
 At5g12110Elongation factor 1B alpha subunit 1 (eEF1Bα1)7619
 At4g23680Major latex protein-related631
 At4g23493Unknown protein637
 At3g29810COBL2 (Cobra-like protein 2 precursor)432
 At3g09350Armadillo/β-catenin repeat family protein3313
 At3g50560Short-chain dehydrogenase/reductase (SDR) family protein333
 At1g30070SGS domain-containing protein328
 At3g13470Chaperonin, putative325
 At1g14980CPN10 (chaperonin 10)221
 At1g58170Disease resistance-responsive protein-related/dirigent protein-related223


Athsp70-1 and Athsp70-2 double mutants do not show phenotypic changes or altered stress responses

As molecular chaperones, cytosolic AtHsp70s are thought to fulfil essential roles in Arabidopsis thaliana. While constitutive over-expression of AtHsp70-1 resulted in the regeneration of stress-tolerant plants (Sung and Guy, 2003; Noel et al., 2007; Cazale et al., 2009), constitutive knockdown of Hsp70s using a full-length antisense cDNA of AtHsp70-1 was not successful. Therefore, it was reasoned that simultaneous down-regulation of more than one cytosolic DnaK-like Hsp70 is lethal, explaining the failure to obtain transgenic silenced plants (Sung and Guy, 2003). To circumvent possible lethal effects of Hsp70 silencing, Lee and Schoeffl (1996) used a heat-inducible promoter to drive expression of a tobacco Hsp70 antisense construct in Arabidopsis plants. Regenerated plants were viable but showed decreased thermotolerance, providing evidence for the importance of cytosolic Hsp70s in plant fitness (Lee and Schoeffl, 1996). In these studies, cross-silencing of several Hsp70 isoforms is likely, as AtHsp70-1 to AtHsp70-5 share more than 80% identity at the amino acid level. Silencing of single Hsp70 isoforms did not reveal any obvious growth phenotypes or altered stress responses (Noel et al., 2007; Lee et al., 2009), probably due to genetic redundancy. Likewise, hsp70-1 and hsp70-2 single or double knockout mutants did not show any growth defects (this study). Despite different but partially overlapping expression patterns, other isoforms appear able to compensate for loss of the two major cytosolic Hsp70s in Arabidopsis thaliana. This implies functional redundancy of typical cytosolic Hsp70 isoforms for normal growth.

Functional AtHsp70-15 but not AtHsp70-14 is essential for normal growth and stress responses of Arabidopsis thaliana

In contrast to silencing of individual members of the DnaK sub-family, silencing of one member of the Hsp110/SSE sub-family, AtHsp70-15, led to severe phenotypic alterations in the resulting mutants compared to control plants. This shows that AtHsp70-14, a close paralog of AtHsp70-15, cannot complement the AtHsp70-15 phenotype. The observed developmental and physiological changes in AtHsp70-15 knockout mutants were phenocopied by expression of an artificial microRNA directed against AtHsp70-14 and -15. This indicates that the phenotype of the double knockdown mutant is most likely caused by silencing of AtHsp70-15. As only 21 amino acids differ between the isoforms, with most of the exchanges being functionally conservative, this outcome was rather unexpected. For yeast SSE1, essential amino acids for binding to SSA1 (the yeast homologue of DnaK-like Hsp70s) have been determined by crystallization and mutagenesis (Schuermann et al., 2008). Most of the crucial amino acid residues in yeast SSE1 are identical or similar at the homologous positions in AtHsp70-14 and AtHsp70-15, and all amino acid residues that differ between yeast and Arabidopsis are conserved between Hsp70-14 and Hsp70-15. Another reason for the failure of AtHsp70-14 to complement the phenotype of Athsp70-15 could be different expression patterns of the two genes. Although we showed that both genes are expressed in leaves, different cell-type specific expression patterns cannot be excluded.

Moreover, no growth phenotype was detected for an independent hsp70-15 mutant allele (SALK_046257) in the absence of heat stress (Larkindale and Vierling, 2008). This obvious discrepancy could be explained by the different experimental set-ups: Larkindale and Vierling (2008) performed seedling experiments, but soil-grown adult plants were used in our study. As the described phenotype only becomes visible later during development and during stress applications, the authors of the previous paper may have overlooked phenotypic changes induced in Athsp70-15 knockout mutants.

Hsp70-14 and Hsp70-15 silenced plants display an altered stomatal response and a constitutive heat stress phenotype

Silencing of AtHsp70-15 caused significant phenotypic changes, including reduced growth, accelerated wilting due to constitutively open stomata, and increased virus tolerance. Moderate increases in heat sensitivity and decreased acquired thermotolerance of an independent hsp70-15 mutant allele (SALK_046257) were reported for seedlings (Larkindale and Vierling, 2008). Analysis of tissue-specific expression data via the Arabidopsis eFP browser ( demonstrates that AtHsp70-14/15 are strongly expressed in guard cells. Hsp70-14/15 silenced plants show defects in stomatal closure, leading to accelerated wilting of detached leaves. As heat stress induces stomatal opening (Rizhsky et al., 2004), this phenotype could be the consequence of a malfunctioning heat response. Stomatal opening is largely driven by the activity of plasma membrane H+-ATPases (Schroeder et al., 2001), which is highly regulated. One potential positive regulator of plasma membrane H+-ATPases, the proton pump interactor PPI2 (At3g15340) (Morandini et al., 2002; Anzi et al., 2008), is up-regulated in hsp70-15 knockout plants, which might explain activation of H+-ATPase activity.

Up-regulation of heat shock response-related genes in Hsp70-15-deficient plants is similar to that found in HsfA2 over-expressing plants (Nishizawa et al., 2006) and wild-type plants under heat stress treatment (Rizhsky et al., 2004; Swindell et al., 2007). HsfA2, a key regulator of heat and other stress responses (Schramm et al., 2006), is constitutively up-regulated in Hsp70-15-deficient plants. On the basis of hierarchical cluster analysis, the transcriptional response of the amiRNA-AtHsp70-14/15 and hsp70-15 mutants is most similar to the cytosolic protein response induced by the proline analogue AZC (Sugio et al., 2009). Induction of Hsp70-5 is a distinctive response in drought and late heat stress. (Schramm et al., 2008). The lack of up-regulation of the heat shock marker gene Hsp70-5 in Hsp70-15-deficient genotypes may confirm that the transgenic plants specifically display a cytosolic protein response rather than a common heat response (Aparicio et al., 2005). In addition, the heat stress regulator HsfA3 is not significantly up-regulated in Hsp70-15-deficient plants or by AZC treatment but is induced by heat (Sugio et al., 2009).

Similar results to those for Hsp70-15 suppression were obtained in functional analysis of its ER homologue BiP. Over-expression of BiP in tobacco reduced the unfolded protein response, ER and water stress, whereas its suppression caused the reverse effects (Leborgne-Castel et al., 1999; Alvim et al., 2001).

AtHsp70-15 may show nucleotide exchange factor activity

It seems likely that the constitutive heat shock response/cytosolic protein response in Hsp70-15-deficient plants and up-regulation of Hsps and Hsfs as seen in microarray data are accompanied by accumulation of unfolded proteins in the transgenic plants. This could be caused by loss of Hsp70-15 chaperone or co-chaperone activity. Assuming functional conservation between Hsp70-15 and yeast SSE1 (Raviol et al., 2006a), Hsp70-15 may serve as a nucleotide exchange factor. It has previously been revealed that atypical Hsp70s are capable of accelerating nucleotide cycling (Shaner and Morano, 2007). If Hsp70-15 is an important co-chaperone for typical cytosolic Hsp70 isoforms (AtHsp70-1 to AtHsp70-5), loss of Hsp70-15 would result in impaired chaperone function of all the interacting Hsp70s.

Increased tolerance of Hsp70-15-deficient plants towards the potyvirus TuMV may support the hypothesis of generally decreased Hsp70 function, as tobacco DnaK-like Hsp70s are known to promote potyvirus infection (Hafren et al., 2010). Silencing of an Hsp70 isoform in Nicotiana benthamiana plants impaired viral infection (Chen et al., 2008; Wang et al., 2009). Based on these facts, a co-factor function of AtHsp70-15 for typical Hsp70s is slightly preferred over a chaperone function of AtHsp70-15.

Hsp70-1 over-expressing plants are more tolerant to heat shock (Sung and Guy, 2003) but enhanced Hsp expression in Hsp70-15-deficient plants did not cause enhanced tolerance. Upon knockdown of Hsp70-15, Hsp70s were transcriptionally induced, but, due to the lack of Hsp70-15 activity, this broad Hsp70 induction did not lead to increased Hsp70 activity.

An argument against a nucleotide exchange factor function is that we could not detect complementation of the yeast sse1Δ mutant with Hsp70-14/15 (Figure S5). However, human and yeast Hsp110 show significant differences in their biochemical properties, and mammalian Hsp110 could not replace SSE1 function in vivo either (Raviol et al., 2006b).

Similarly to Hsp70-15 deficiency, loss of AtFes1A (At3g09350), which plays a role in cytosolic Hsp70 stability and abiotic stress tolerance, results in a global increase in transcription of heat responsive genes. Nucleotide exchange factor activity for AtFes1A has been proposed on the basis of homology but has not yet been detected in vitro (Zhang et al., 2010).

Experimental Procedures

Plant lines and growth conditions

Arabidopsis Col-0 plants were grown on soil under short-day conditions (8 h light, 16 h dark), approximately 60% humidity and 22°C (day) or 18/19°C (night). For flowering, they were transferred to long-day conditions (16 h light, 8 h dark).

The T-DNA insertion mutants SALK_135531 (Athsp70-1 knockout), SALK_085076 (Athsp70-2 knockout), SALK_087844 (Athsp70-1/2 knockout), SALK_082815 (Athsp70-14 knockout) (Alonso et al., 2003) and SAIL_830_C08 (Athsp70-15 knockout) (Sessions et al., 2002) were obtained from the Nottingham Arabidopsis Stock Centre ( Homozygous plants were identified by PCR, and knockout was confirmed by RT-PCR analysis.

Nicotiana tabacum and N. benthamiana plants were cultivated in the greenhouse on soil, with 16 h of supplemental light (250–300 μmol quanta m−2 sec−1), approximately 60–70% relative humidity and 18°C/22°C (night/day).

For sowing on MS medium (4.4 g L−1 Murashige and Skoog medium including vitamins, 0.5 g L−1 MES, 8 g L−1 Phytoagar (Duchefa,, pH adjusted to 5.7 using KOH) without or with 0.5 or 1 μm ABA, Arabidopsis seeds were surface-sterilized by gas treatment after mixing of NaOCl and HCl. Stratification was performed for 3 days at 7°C in the dark before transfer to a light room with 16 h light per day and constant temperature of 21°C.

Stable and transient expression of GFP fusion proteins in Nicotiana sp.

For stable expression, full-length coding sequences of AtHsp70-1 and AtHsp70-2 were amplified from Arabidopsis cDNA by PCR using primers 5′-CACCATGTCGGGTAAAGGAGAAGGAC-3′/5′-TTAGTCGACCTCCTCGATCTTAGG-3′ and 5′-CACCATGGCTGGTAAAGGAGAAGGTC-3′/5′–TTAGTCGACTTCCTCGATCTTGGG-3′, and were subsequently inserted via topoisomerase reaction into the pENTRTM/D-TOPO® as brand vector (Invitrogen, Coding sequences were then introduced into destination vector pK7WGF2 (Karimi et al., 2002) by LR reaction according to the manufacturer’s protocol to obtain N-terminal GFP fusions. The resulting constructs were introduced into Agrobacterium tumefaciens strain C58C1 (pGV2260). Stable transformation of Nicotiana tabacum (SNN) was performed by Agrobacterium-mediated transfer as described previously (Rosahl et al., 1987), and transgenic plants were selected on the basis of GFP fluorescence.

For transient expression of C-terminal AtHsp70-14/15 fusions (AtHsp70-14/15–GFP), the same cloning strategy was performed using primers 5′-CACCAACAATGAGTGTAGTCGGGTTTGATTTTG-3′/5′-GGTACTACCTTCCGCGGGATTC-3′ and 5′-CACCAACAATGAGTGTAGTTGGTTTTGATTTTG-3′/5′-GGTACTACCTTCCGCGGGATTC-3′, with final recombination in pK7FWG2 (Karimi et al., 2002). The resulting constructs were introduced into Agrobacterium tumefaciens strain C58C1 (pGV2260) and infiltrated into Nicotiana benthamiana plants as described previously (Hofius et al., 2007).

Generation of amiRNA-AtHsp70-14/15 plants

An artificial microRNA construct directed against both genes was designed using the WMD2 - Web MicroRNA Designer ( (Schwab et al., 2006) with the core sequence TTAAACATAAGACTCAACGGC. In accordance with the protocol on the website, the amiRNA backbone on vector pRS300 was modified using oligonucleotides to produce amiRNA-AtHsp70-14/15. The amiRNA-AtHsp70-14/15 cassette was cloned using XbaI and XhoI digestion into pAF16 (Stadler et al., 2005), a pGPTV-BAR-derived vector for plant over-expression harbouring a CaMV 35S promoter and a BASTA resistance cassette (Becker et al., 1992). Using Agrobacterium mediated transformation by the floral-dip-method (Clough and Bent, 1998), the T-DNA construct was used to generate amiRNA-AtHsp70-14/15 in the Col-0 ecotype background. To select for the amiRNA construct, 1–2-week-old plants were sprayed three times with 0.1% BASTA (Bayer, solution with 3-4 days in between.

A wild-type like line (PME1-9) was used as a transformed control which was created in the same way and showed no obvious silencing.

Gas-exchange measurements and ABA treatment

Infra-red gas exchange was determined using attached, fully expanded leaves of 9–10-week-old plants using a combined infra-red gas exchange/chlorophyll fluorescence imaging system (GFS 3000 and Mini-Imaging-PAM chlorophyll fluorometer; Walz, For ABA treatments, plants were sprayed with 50 μm ABA (dilution of 100 mm stock ABA solution in dimethylformamide with 0.05% Tween-20 added) or water control without ABA. One hour later, gas exchange was determined directly after transfer from the growth chamber without dark adaption. The conditions in the measuring cuvette were set to 22°C, 350 ppm CO2 and 20 000 ppm H2O, light intensity approximately 250 μE m−2 sec−1. The water vapour conductance was corrected for leaf area determined by fluorescence imaging using the integrated imaging PAM module prior to the gas-exchange measurements.

TuMV infection experiments

Four- to six-week-old plants were used for TuMV infection experiments. The three largest leaves were manual inoculated with TuMV-infected plant material ground in 5 mm sodium phosphate buffer (pH 7.2). Silicon carbide powder was added to damage the leaf surface, allowing introduction of virus-containing buffer or buffer only as negative control, and was washed away with water after a few minutes. Leaf discs were cut from inoculated leaves for local samples 2 weeks after inoculation and from younger leaves for systemic samples 3 weeks after inoculation. The virus titre was determined by ELISA assay against the TuMV capsid protein using antibodies from BIOREBA ( according to the manufacturer’s protocol, and quantified by measuring optical absorbance at 405/492 nm. In addition, 3 weeks after inoculation, rosettes were harvested for fresh weight determination and RNA isolation.

Heat treatment

For heat treatment, approximately 4-week-old BASTA-selected plants were grown on soil under short-day conditions. The air temperature was increased to 38°C for 5 days (and nights), and then returned to ambient temperature. Water was not restricted.


Stomata on the abaxial side of Arabidopsis plants were analysed using a Leica DMR light microscope ( Staining of cell walls with propidium iodide and confocal microscopy were performed as described previously (Vogel et al., 2007) using a Leica TCS SP2 or SP5 laser scanning microscope (

RNA isolation and transcriptome analyses

Total RNA was extracted by guanidinium hydrochloride extraction (Logemann et al., 1987). For Northern blot analysis, 20 μg RNA were separated on 1.5% formaldehyde-containing agarose gels and blotted onto Genescreen nylon membranes (PerkinElmer, by capillary blotting overnight. The 811 bp EcoRI/PstI fragment of AtHsp70-14 cDNA and the 867 bp TuMV capsid protein cDNA (GenBank accession number D10601.1) were radioactively labelled with [32P]dCTP using a High Prime kit (Roche, and used as probes. After hybridization at a temperature of 65°C and stringent washing of the membranes, radioactive signals were detected by exposure to X-ray films.

Microarray analysis of purified RNA was performed as described previously (Ferreira et al., 2010) on a Agilent Technologies platform using Agilent Arabidopsis V4 (Design number 21169) microarray chips ( For RNA extraction, amiRNA-AtHsp70-14/15 line 4, hsp70-14 knockout, hsp70-15 knockout and wild-type Col-0 plants were grown without BASTA selection in a randomized arrangement under normal conditions, and fully expanded leaves of 6-week-old plants were harvested at midday in pools of 10 individuals per line (four replicates each). Data have been deposited in the ArrayExpress database ( under accession number E-MEXP-3082, and were analysed using Genespring XI. Data normalization to the median of all samples was performed after log2-based transformation of the raw data, with the threshold set to 1.0 and shifting the data to the 75th percentile.

Differentially regulated genes between amiRNA-AtHsp70-14/15, hsp70-14 or hsp70-15 knockout and wild-type plants with >2.0 fold change and P < 0.05 were identified using a Welch test after Benjamini–Hochberg multiple test correction. For genes represented by more than one probe, the mean of the independent probes was calculated to give the values indicated in Table S1 (using the TAIR database as a reference). Publically available Affymetrix data sets from Sugio et al. (2009) (accession number GSE11788) and from Kilian et al. (2007) (accession numbers NASCARRAY-137, -140, -141, -146) were downloaded from and, respectively. Data from Sugio et al. (2009) were analysed as described above but without multiple testing correction, and greater than two-fold differentially expressed genes from the Affymetrix data sets (Kilian et al., 2007) were extracted by comparison with respective mock treatments. After using the translation mapping function of Genespring XI to adjust the Affymetrix- and Agilent-based datasets, clustering analysis was performed using the program Cluster version 2.1, and the results were visualized using Treeview version 1.6 (Eisen et al., 1998), according to the manual at Hierarchical clustering of genes and arrays was performed using Spearman rank correlation and complete linkage clustering.

Analysis of specific Hsp70 transcripts by RT-PCR was performed using the specific primer combinations listed in Table S2. The cDNA was produced by reverse transcription of DNAse I-treated RNA using oligo(dT)30 primers and a RevertAid™ H Minus First Strand cDNA Synthesis Kit (Fermentas, according to the manufacturer’s protocol.

Gene annotation data and microarray data

Gene annotation data were submitted to TAIR: AtHsp70-1 (At5g02500), AtHsp70-2 (At5g02490), AtHsp70-14 (At1g79930) and AtHsp70-15 (At1g79920).

Microarray data have been deposited in the ArrayExpress database ( under accession number E-MEXP-3082.


This work is part of Collaborative Research Centre 796, sub-project C2, funded by the Deutsche Forschungsgemeinschaft (DFG) Gemeinschaft. We thank Norbert Sauer (FAU Erlangen-Nuremberg) for providing the confocal laser scanning microscope, and Kathrin Paulus for the transformed control line. We thank Melanie Senning, Stephanus Ferreira and Stephen Reid for bioinformatic and technical help with microarray analysis. We also thank Bernd Bukau and Heike Rampelt (ZMBH, University Heidelberg) for providing the sse1Δ yeast strain.