A novel non-conventional heat shock element regulates expression of MDJ1 encoding a DnaJ homolog in Saccharomyces cerevisiae.

The heat shock factor (HSF) is a pivotal transcriptional factor that regulates the expression of genes encoding heat shock proteins (HSPs) via heat shock elements (HSEs). nGAAnnTTCnnGAAn functions as the minimum consensus HSE (cHSE) in vivo. Here we show that the expression of Saccharomyces cerevisiae MDJ1 encoding a mitochondrial DnaJ homolog is regulated by HSF via a novel non-consensus HSE (ncHSE(MDJ1)), which consists of three separated pentameric nGAAn motifs, nTTCn-(11 bp)-nGAAn-(5 bp)-nGAAn. This is the first evidence to show that the immediate contact of nGAAn motifs is dispensable for regulation by HSF in vivo. ncHSE(MDJ1) confers different heat shock responses versus cHSE and, unlike cHSE, definitively requires a carboxyl-terminal activation domain of HSF in the expression. ncHSE(MDJ1)-like elements are found in promoter regions of some other DnaJ-related genes. The highly conserved HSF/HSE system suggests that similar ncHSEs may be used for the expression of HSP genes in other eukaryotes including humans.

The heat shock factor (HSF) is a pivotal transcriptional factor that regulates the expression of genes encoding heat shock proteins (HSPs) via heat shock elements (HSEs). nGAAnnTTCnnGAAn functions as the minimum consensus HSE (cHSE) in vivo. Here we show that the expression of Saccharomyces cerevisiae MDJ1 encoding a mitochondrial DnaJ homolog is regulated by HSF via a novel non-consensus HSE (ncHSE MDJ1 ), which consists of three separated pentameric nGAAn motifs, nTTCn-(11 bp)-nGAAn-(5 bp)-nGAAn. This is the first evidence to show that the immediate contact of nGAAn motifs is dispensable for regulation by HSF in vivo. ncH-SE MDJ1 confers different heat shock responses versus cHSE and, unlike cHSE, definitively requires a carboxylterminal activation domain of HSF in the expression. ncHSE MDJ1 -like elements are found in promoter regions of some other DnaJ-related genes. The highly conserved HSF/HSE system suggests that similar ncHSEs may be used for the expression of HSP genes in other eukaryotes including humans.
All organisms possess a highly conserved system that responds to elevated temperatures by transcriptionally inducing genes encoding heat shock proteins (HSPs) 1 to deal with heat stress. The induction requires heat shock transcription factor (HSF) and cis-heat shock elements (HSEs) (1). HSF has two conserved domains, a helix-turn-helix DNA binding domain and a coiled-coil hydrophobic repeat domain needed for trimer or higher order multimer formation (2,3). HSFs of the budding yeasts (Saccharomyces cerevisiae and Kluyveromyces lactis) uniquely possess two activation domains in the amino terminus (AAD) and carboxyl terminus (CAD) (4,5), whereas those of fission yeast (Shizosaccharomyces pombe), Drosophilla melanogaster, and vertebrates have CADs alone. HSF is moderately phosphorylated under non-stress conditions and is further activated by hyperphosphorylation upon heat shock to induce the expression of HSPs (6 -8). In fission yeast, D. melanogaster, and mammals, the binding activity of HSFs to HSEs is dra-matically stimulated by heat shock as HSF monomers are converted to trimers (9 -11). In contrast, in the budding yeast, HSF constitutively binds to HSE, but the binding activity seems to increase after heat shock-induced hyperphosphorylation (7,(12)(13)(14)(15)(16).
The HSE is composed of several contiguous inverted repeats of the 5-base pair sequence nGAAn (where n is any nucleotide) (1,17,18). The number of the pentameric units in HSE varies, but at least three units are thought to be the minimum required for heat regulation in vivo. Namely, nGAAnnTTC-nnGAAn is the minimum consensus HSE (cHSE) (19,20). However, an HSE can tolerate and still function with a 5-bp insertion between two repeating units if the spacing and orientation of the pentameric elements are maintained, e.g. nGAAn-(5-bp)-nGAAn (21). The most characterized endogenous non-consensus HSE (ncHSE) is that of CUP1 in S. cerevisiae, nTTCnnGAAn-(5-bp)-nGAGn denoted ncHSE CUP1 (22)(23)(24). Similar elements regulate HSP82 and HSC82 (12,(25)(26)(27). It is unclear whether the immediate contact of at least two pentameric motifs is needed for heat induction in vivo.
In higher organisms possessing multiple HSF genes, HSF isoforms appear to perform different biological functions and regulate distinct target genes via differential binding preferences for specific HSE architectures (6, 28 -36). By contrast, yeasts and D. melanogaster have a single essential HSF gene (7,(37)(38)(39)(40)(41). Interestingly, the induction maximum of CUP1 and cHSE-driven SSA1 is observed at different high temperatures (22)(23)(24)42). In addition, it is known that the binding preference of HSF for HSE variants changes under different physiological conditions (23,24,43,44). Thus, it is probable that the HSFregulated gene expression is modulated by different HSE architecture in these organisms.
In S. cerevisiae, there are various nuclear-encoded genes for mitochondrial HSPs (mtHSPs) as follows: DnaK homolog Ssc1, DnaJ homolog Mdj1, and chaperonin complex Hsp60⅐Hsp10. They are involved in protein import, protein folding and assembly, and proteolysis in mitochondria (45)(46)(47)(48). Although some of the mtHSP genes (HSP60 and HSP10) but not all (SSC1 and MDJ1) have obvious HSEs in their promoters, all of them are heat-inducible (49 -52). Here we show that MDJ1 is regulated by HSF via a novel ncHSE.

EXPERIMENTAL PROCEDURES
Culture Media, Strains, and Plasmids-The media used include glucose-based rich medium (YPD) and synthetic medium (SD). YPGalR and SGalR are identical to YPD and SD with the exception that they contain 1% galactose and 1% raffinose instead of 2% glucose. The strains used are listed in Table I. The promoter regions of HSP genes were amplified by PCR using an Expand High Fidelity PCR system (Roche Diagnostics, Mannheim, Germany): MDJ1-(Ϫ403 to ϩ3) (translation start site is ϩ1), MDJ1-(Ϫ273 to ϩ3), MDJ1-(Ϫ239 to ϩ3), * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
␤-Galactosidase Assay-Cells were cultivated in appropriate SDuracil medium to select for URA3-marked plasmids overnight at 30°C, diluted to an optical density of ϳ0.2 at 600 nm (A 600 ) in SD-uracil and cultivated for another 5 h at 23 or 30°C. For heat shock, an aliquot of the cells cultivated at 23°C were transferred to 39°C. ␤-Galactosidase activity in the extract of chloroform/SDS-permeabilized cells was assayed and expressed in Miller units (55). Each value is expressed as the mean Ϯ S.D. of duplicate determinations of three independent yeast transformants.
Electrophoretic Mobility Shift Assay-Protein extracts were prepared from 100 ml of logarithmic phase cell cultures of strain SCU166 in the YPGalR at 30°C. The cells harvested by centrifugation were washed once with distilled water and suspended in an equal pellet volume of breakage buffer (20 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 10% (v/v) glycerol, 0.3 M KCl, 10 mM NaF, 0.2 mM NaVO 4 , 2 mM 2-mercaptoethanol, 1 g/ml pepstatinA, 2 g/liter leupeptin, and 1 mM phenylmethylsulfonyl fluoride) at 4°C. An equal volume of glass beads was added, and the cells were disrupted by vortexing. The extracts were clarified by 15 min of centrifugation at 15,000 rpm, and the supernatants were used. Double-stranded synthetic oligonucleotide ncHSE MDJ1 (5Ј-CATTATTTTTCTTTACATCCTGTGGAACTCTATGGAA-3Ј) and cHSE HSP60 (5Ј-TCGTGGAATTTTCCAGAAAACCA-3Ј) were radiolabeled using the T4 polynucleotide kinase and [␥-32 P]dATP (3000 Ci/ mmol) and purified by ethanol precipitation and dissolved in TE buffer (10 mM Tris-HCl, pH 8.0, and 1 mM EDTA). The binding buffer consisted of 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 2.5 mM MgCl 2 , 0.03% (w/v) bovine serum albumin, 10% glycerol, and 0.1% Nonidet P-40. Binding reactions were initiated by the addition of protein extracts (50 -60 g of protein), 0.25 pmol of radiolabeled probe, 1 g of poly(dI⅐dC) (Amersham Biosciences), and 1 g of denatured salmon sperm DNA. For the competition assay, an unlabeled 10-or 100-fold amount of the competitor oligonucleotide was added. For supershifts, 3.2 g of anti-HA antibody was added. After incubation for 60 min at 4°C in a final volume of 15 l, samples were loaded onto non-denatured 5% polyacrylamide gels and electrophoresed at room temperature in 0.5 ϫ Tris borate EDTA buffer. The gels were then dried and exposed to x-ray film.

Promoters of MDJ1 and SSC1 Confer
Heat Inducibility in a HSF-dependent Manner-It has been shown by Northern blot analysis that mRNA level of MDJ1 is increased by heat shock (52) despite no obvious HSE. To see transcriptional heat inducibility of the MDJ1 promoter, the promoter activity was assayed using a construct MDJ1 promoter fused to lacZ (MDJ1-lacZ). This construct showed heat induction similar to SSA3-lacZ as the positive control, although it already had a higher basal expression at 23 and 30°C than SSA3-lacZ (Fig.  1A). Ssc1 protein is also heat-inducible despite no obvious HSE (50). SSC1 was shown to be heat-inducible by SSC1-lacZ although to a lesser extent than that of MDJ1 promoter. The fold induction of MDJ1 and SSC1 was smaller compared with SSA3, which is well known to be remarkably heat-inducible. The apparent decreased induction is probably because of the relatively high basal expression levels of MDJ1 and SSC1, which presumably reflect their essentiality for mitochondrial functions even under non-stress conditions (52,58). hsf1 R206S , which is a constitutively active allele (43), boosted the basal transcriptional activities of MDJ1 and SSC1 promoters ( Fig. 1B and data not shown), indicating that they are regulated by HSF. In addition, in a hsf1 V203A mutant, which shows a poor heat inducibility (43), the heat induction of MDJ1 was compromised. MDJ1 showed more remarkable heat inducibility than SSC1, so we chose MDJ1 for further study.
Identification of the Element Responsible for Heat Inducibility in the MDJ1 Promoter-The MDJ1 promoter region from Ϫ403 bp (translation start site is ϩ1) used in Fig. 1 contains two stress response elements (STREs) (Fig. 2). STRE also confers heat inducibility to genes by redundant transcriptional factors Msn2 and Msn4. MDJ1 is induced by ethanol, NaCl, and sorbitol in Msn2/4-dependent manners (59). However, heat induction still occurred even when the two STREs were deleted (Fig. 2, ⌬1). Furthermore, a msn2⌬ msn4D double mutant showed similar heat inducibility of MDJ1-lacZ (data not shown). Thus, these STREs were not needed for heat inducibility, but if involved, their contribution was marginal.
To dissect the function of the MDJ1 promoter, a series of progressive 5Ј deletion clones was made (Fig. 2). Both basal and heat-induced levels gradually decreased when the sequence upstream of Ϫ239 was deleted (clones ⌬1-5). In contrast, further removal of the region from Ϫ239 to Ϫ203 increased both levels (clones ⌬5-7). Further elimination of the region from Ϫ203 to Ϫ184 again decreased both levels (clones ⌬7-9). These findings suggest that the sequences from Ϫ403 to Ϫ239 and from Ϫ203 to Ϫ184 contain upstream-activating sequences, and that the sequence from Ϫ239 to Ϫ203 contains an upstream-repressing sequence. However, all of these deletion clones still retain heat inducibility, indicating that these regions contain no element responsible for the heat induction. Importantly, a deletion of the region containing Ϫ173 TCC Ϫ171 abolished basal level (clone ⌬10). A further deletion of the region containing Ϫ156 GAA Ϫ154 entirely abrogated heat induction (clone ⌬11). These results indicate that these regions are necessary for heat inducibility.
A Novel Non-canonical HSE of MDJ1, ncHSE MDJ -The involvement of HSF in MDJ1 expression suggests that Ϫ173 TCC Ϫ171 and Ϫ156 GAA Ϫ154 are putative HSF binding sites. A putative third Ϫ147 GAA Ϫ145 exists 7 bp downstream of Ϫ156 GAA Ϫ154 . This gap length is the same as the lengths found in ncHSE CUP1 . Thus, Ϫ174 TTTCTTTACATCCTGTG-GAACTCTATGGAAA Ϫ144 ( Ϫ173 TCC Ϫ171 , Ϫ156 GAA Ϫ154 , and Ϫ147 GAA Ϫ145 are referred to as units A to C, respectively, and the gaps between units A and B and units B and C are denoted as gap1 and gap2, respectively) serves as a putative HSE. This assumption was tested by inserting an oligonucleotide corresponding to Ϫ174 to Ϫ144 of the MDJ1 promoter into a CYC1-lacZ reporter construct. As hypothesized, the basal expression level of CYC1-lacZ driven by this element was significantly (reproducibly but not remarkably) enhanced by hsf1 R206S (Fig.  3A). Therefore, it was named ncHSE MDJ1 . On the other hand, a loss of a two-component response regulator homolog Skn7, which has a DNA-binding domain similar to that of Hsf1 and induces some HSP genes in response to oxidative stress cooperatively with Hsf1 (60), showed no obvious effect on expression of ncHSE MDJ1 -CYC1-lacZ (data not shown), indicating that its expression is Skn7-independent. ncHSE MDJ1 also conferred heat inducibility to CYC1-lacZ (Fig. 3B). Thus, this element is necessary and sufficient for heat induction. The conversions of GAA of units B and C to non-functional mutations GAG completely abolished heat inducibility (M2 and M3). In contrast, the conversion of TCC of unit A to CTC did not affect heat induction (M1). The lack of affect of this mutation is most probably attributed to the generation of a new TTC site 2 bp upstream TTTTC to TTCTC, because an additional mutation introduced in the region TTTTC to CTCTC entirely abrogated heat inducibility (M4). These results demonstrate that all three pentamers are essential for heat inducibility and show that ncHSE MDJ1 functions as an HSE. Conversely, when the GAA-gap-2-GAA was altered to cHSE (GAACTTTCTGGAA), the heat inducibility was reinforced (M5), indicating that this gap weakens heat inducibility. However, this gap length is very important for heat inducibility, because even if only one bp of this region was added or removed, the heat inducibility was completely abrogated (M6 and M7). By contrast, M1 shows that the gap1 length is flexible in terms of heat inducibility.
This novel ncHSE architecture ncHSE MDJ1 is quite distinct even from ncHSE CUP1 (Table II). Three pentameric motifs of ncHSE MDJ1 are separated from each other (nTTCn-(gap)-nGAAn-(gap)-nGAAn) in a different manner than those of the HSEs previously described, cHSE (GAAnnTTCnnGAA) or ncH-SE CUP1 (nTTnnGGAn-(gap)-nGGAn). It demonstrates that no contact of pentameric motifs is needed. The gap1 length (13 bp) is extraordinary, because the presence of a ncHSE with such a long gap was not expected. Furthermore, a ncHSE with 15 bp of gap2 length was still functional (Fig. 3B, M1).
Binding of HSF to ncHSE MDJ in Vitro-To assess the binding ability of HSF to ncHSE MDJ1 , Electrophoretic mobility shift assay was carried out with ncHSE MDJ1 as a probe and with protein extracts from yeast cells expressing 2HA-tagged Hsf1 (2HA-Hsf1). Similar to cHSE, a specific DNA-protein complex was found using ncHSE MDJ1 (Fig. 4, lanes 2 and 8). These complexes were converted to slower migrating forms (supershift) when anti-HA antibody was co-incubated prior to electrophoresis (data not shown), indicating that these complexes contain HSF. However, ncHSE MDJ1 was a less effective competitor than cHSE to cHSE (lanes 3-6) and cHSE to ncH-
FIG. 2. Deletion analysis of the MDJ1 promoter. lacZ was expressed under the control of the sequentially deleted constructs of the MDJ1 upstream region. Wild-type strains (KMY1005) harboring these constructs were transferred from 23 to 37°C for 1 h. Translation start site is ϩ1. SE MDJ1 (lanes 9 -12). Thus, ncHSE MDJ1 can bind to HSF in vitro although with weaker affinity than cHSE. ncHSE MDJ1 and cHSE Confer Different Heat Responses-The different architecture between ncHSE MDJ1 and cHSE suggests that they may confer different heat responses to gene expression. This appears to be the case. Whereas they showed maximum induction at 39°C, even at 41°C, ncHSE MDJ1 still maintained heat inducibility (40% heat induction at 39°C) as compared with cHSE (22%) (Fig. 5A). Additionally, heat response conferred by cHSE was transient, whereas the heat response mediated by ncHSE MDJ1 was rather sustained (Fig. 5B).
Hsf1 CAD Is Essential for ncHSE MDJ1 -driven Gene Expression-AAD, but not CAD, of Hsf1 is essential for viability and for basal expression of HSP genes (5) (see also Fig. 6A). However, CAD is critical for heat induction of CUP1, HSP82, and HSC82 but not of cHSE-driven HSP genes (22)(23)(24)42), suggesting that CAD is specifically required for ncHSEs with gaps. This is also true for ncHSE MDJ1 . The heat induction of cHSEdriven reporter still occurred in the strain lacking CAD of Hsf1 (hsf1⌬CAD), albeit to reduced levels (Fig. 6B). On the contrary, both basal and heat-induced expression of ncHSE MDJ1 -driven reporter was drastically reduced by the loss of CAD. These results demonstrate that CAD is essential for ncHSE MDJ1mediated gene expression. This finding is consistent with the observations that CAD is required for sustained responses to heat shock (5) and that ncHSE MDJ1 conferred sustained heat response (Fig. 6B).
Glc7 Differentially Regulates cHSE-directed and ncH-SE MDJ1 -directed Gene Expression-HSF activity is regulated by its phosphorylation status. Loss-of-function mutations of a type 1 serine/threonine protein phosphatase complex Glc7⅐Gac1 (Glc7 and Gac1 are catalytic and regulatory subunits, respectively) such as glc7-1 (61) attenuate heat induc-   ibility of CUP1 but not a cHSE-driven SSA4 gene (62). Although not directly shown, these results suggest that HSF in a specific phosphorylation status differentially regulates different HSE architecture. We tested and confirmed this idea using our constructs (Fig. 7). For the cHSE-driven reporter, heatinduced but not basal expression increased in the glc7-1 strain, whereas basal but not heat-induced expression increased for the ncHSE MDJ1 -driven reporter. Interestingly, the effect of glc7-1 on ncHSE MDJ1 -driven reporter seems to be opposite that of CUP1 expression (62), which might be explained by the different HSE architectures. However, we cannot formally exclude the possibility that another transcriptional factor, which is also regulated by Glc7⅐Gac1, may bind to the long gap of ncHSE MDJ1 element and influence reporter expression.
ncHSE MDJ1 -like Elements Are Found in Promoter Regions of Other DnaJ Homolog Genes-Computer-assisted analysis revealed that there are ncHSE MDJ1 -like elements in promoter regions of many genes. 2 Most interestingly, ncHSE MDJ1 -like elements were found in the promoter regions of two genes for DnaJ homologs, cytosolic Ydj1, and as-yet-uncharacterized Ynl077w (Table II). Northern analysis showed that YDJ1 is heat-induced (63). We also found that YDJ1-lacZ and YNL077W-lacZ were also transcriptionally heat-inducible (data not shown). From similarities in the DNA architecture, it is most probable that these ncHSE MDJ1 -like elements generally confer heat inducibility. In contrast, DnaJ-related SIS1 is heat-induced via a cHSE (64). Other DnaJ-related genes, SCJ1, XDJ1, JEM1, SEC63, YNL227C, HLJ1, CAJ1, ZUO1, DJP1, YJL162C, YFR041C, and YJR097W possess no obvious HSE-like elements. Thus, some but not all DnaJ-related genes appear to be driven via ncHSE MDJ1 -like elements. On the other hand, a region containing several nGAAn motifs was found in the SSC1 promoter (Table II), but this architecture is different from those of cHSE or ncHSE. Similar heat induction was observed in SSC1-(Ϫ178 to ϩ3)-lacZ and SSC1-lacZ (data not shown), indicating that the element responsible for heat induction is contained within Ϫ145 to Ϫ1.

DISCUSSION
A single HSF trimer binds to the one copy of the HSE element (17,65), whereas a pair of trimers interacts with the ncHSE cooperatively in the presence of more than three GAA motifs (66). Although there is no direct evidence, we assume that a single HSF trimer binds to ncHSE MDJ1 , because the mobility of the HSF⅐ncHSE MDJ1 complex was nearly the same as that of HSF⅐cHSE complex (Fig. 4). The 73 amino acids between the DNA-binding and the trimerization domain in Hsf1 are suggested to be a flexible linker (67). We believe that this flexible linker would be long enough to allow each trimerization domain to come close and form a trimeric coiled-coil structure. The hsf1⌬CAD strain almost completely lost heat induction of ncHSE MDJ1 -directed but not cHSE-directed reporter, indicating that CAD function depends on cis-element FIG. 5. ncHSE MDJ1 and cHSE confer different heat shock responses. ␤-Galactosidase activity in wild-type strain (KMY1005) with ncHSE MDJ1 -CYC1-lacZ or cHSE-CYC1-lacZ was determined. A, cells preincubated at 23°C were exposed to 37, 39, 41, or 43°C for 40 min. B, cells preincubated at 23°C were exposed to 39°C for various times. architecture. We propose the following model to explain these observations. The AADs of three HSF molecules can associate with each other in the HSF⅐cHSE complex but not in the HSF⅐ncHSE complex. In contrast, three CADs probably are able to associate with each other in both complexes (Fig. 8). At least either clustered AADs or clustered CADs are required for HSF-mediated gene expression. Therefore, Hsf1⌬CAD can still transactivate gene expression via cHSE but not ncHSE MDJ1 .
We found that the expression of mtHSP genes examined (MDJ1, SSC1, HSP10, and HSP60) were all regulated by HSF. 2 (this study and Ushimaru et al., unpublished data). Mitochondria are thought to have evolved from a free-living eubacterium endosymbionts. At present, however, all of these mtHSP genes have transferred from the mitochondrial genome to the nuclear genome during evolution. Subsequently, mtHSP genes have been put under the HSF/HSE system of the host cell. However, the adoption of this system is not self-evident, because S. cerevisiae possesses other distinct heat response systems, i.e. Ras-protein kinase A, calcineurin, and protein kinase C pathways. The calcineurin and PKC pathways appear to play restricted roles in cation homeostasis and reinforcement of membrane rigidity (68 -73). In contrast, the HSF/HSE system performs general roles in the protection against heat stress. Most HSP genes are regulated by the HSF/HSE system, which is activated by the accumulation of unfolded proteins in the cytosol, e.g. after heat shock (74). Because upon heat shock, unfolded proteins are accumulated in all intracellular compartments including in mitochondria, this simple HSF/HSE heatresponsible system might be sufficient for the regulation of mtHSP genes. Some genes regulated by HSF are also controlled by Msn2/4 (75). These two systems are differentially used under various stress conditions (59,75). In the case of MDJ1, the HSF/HSE and the Msn2/4/STRE systems are responsible for induction by heat (this study), and by ethanol, NaCl, and sorbitol (59), respectively.
Mdj1 is a partner of Ssc1 in the mitochondrial matrix. Jac1 is another J-type chaperone working together with another HSP70 protein Ssq1 in the mitochondrial matrix (76 -79). However, neither SSQ1 nor JAC1 has any obvious HSEs. On the other hand, many of the ncHSE MDJ1 -like elements are found in the promoters of other genes, indicating that they have general roles. From the conservation of the HSF/HSE system among eukaryotes, this study suggests that ncHSE MDJ1 -like elements may be involved in the expression of HSP gene in other organisms including humans. For example, one candidate is human HSP90 (GenBank TM accession number J04988), which has two ncHSE MDJ1 -like elements in its promoter region, Ϫ1062 TTC-(24 bp)-GAAgtgggcgGAA Ϫ1033 and Ϫ648 GAAactgctgGAA-(24 bp)-TTC Ϫ919 (transcription start site is ϩ1), although their gap1 length is rather longer.