Palindrome with spacer of one nucleotide is characteristic of the cis-acting unfolded protein response element in Saccharomyces cerevisiae.

When unfolded proteins are accumulated in the endoplasmic reticulum (ER), an intracellular signaling pathway termed the unfolded protein response (UPR) is activated to induce transcription of ER-localized molecular chaperones and folding enzymes in the nucleus. In Saccharomyces cerevisiae, at least six lumenal proteins including essential Kar2p and Pdi1p are known to be regulated by the UPR. We and others recently demonstrated that the basic-leucine zipper protein Hac1p/Ern4p functions as a trans-acting factor responsible for the UPR. Hac1p binds directly to the cis-acting unfolded protein response element (UPRE) responsible for Kar2p induction. Moreover, we showed that the KAR2 UPRE contains an E box-like palindrome separated by one nucleotide (CAGCGTG) that is essential for its function. We report here that the promoter regions of each of five target proteins (Kar2p, Pdi1p, Eug1p, Fkb2p, and Lhs1p) contain a single UPRE sequence that is necessary and sufficient for induction and that binds specifically to Hac1p in vitro. All of the five functional UPRE sequences identified contain a palindromic sequence that has, in four cases, a spacer of one C nucleotide. This unique characteristic of UPRE explains why only a specific set of proteins are induced in the UPR to cope with ER stress.

Eukaryotic cells possess multiple intracellular signaling pathways from the endoplasmic reticulum (ER) 1 to the nucleus (1). One of these, the unfolded protein response (UPR), is conserved from yeast to man and is activated by accumulation of unfolded proteins in the ER under a variety of stress conditions ("ER stress"), resulting in transcriptional induction of molecular chaperones and folding enzymes localized in the ER (2)(3)(4)(5). The UPR is thought to be activated primarily to cope with deleterious effects of accumulated unfolded proteins. In addition, the possible linkage of UPR with phospholipid biosynthesis, especially with inositol metabolism, and its involvement in ER homeostasis have also been suggested (6 -9).
Budding yeast Saccharomyces cerevisiae is an excellent model system in which to study the molecular mechanism of the UPR and has led to the identification of three genes as essential components of the UPR: the IRE1/ERN1 gene encoding a transmembrane protein kinase localized in the ER (6,7), the HAC1/ERN4 gene encoding a basic-leucine zipper protein (10 -12), and the RLG1 gene encoding tRNA ligase localized in the nucleus (13).
In contrast, all the known target proteins of the UPR are localized in the ER, where they assist folding and assembly of newly synthesized secretory and transmembrane proteins as molecular chaperones or folding enzymes (14,15). In S. cerevisiae, six lumenal proteins have been shown to be regulated by the UPR: Kar2p, an Escherichia coli DnaK homolog (16,17); Lhs1p/Ssi1p/Cer1p, a member of the Hsp70 subfamily (18 -21); Scj1p, an E. coli DnaJ homolog (22); Pdi1p, protein-disulfide isomerase (6,23,24); Eug1p, a multicopy suppressor of the pdi1⌬ strain (25); and Fkb2p, peptidyl-prolyl cis-trans isomerase (26). Genes encoding each of these ER stress-inducible proteins are thought to contain an upstream activator sequence, termed the unfolded protein response element (UPRE), in their promoter region (26 -28). UPRE was originally identified as a cis-acting element necessary for transcriptional induction of Kar2p by ER stress, and the 22-base pair sequence is sufficient to confer inducibility on a heterologous promoter such as the yeast CYC1 promoter (27,28). UPRE was proposed to be a binding site of a putative transcription factor termed unfolded protein response factor (UPRF) (27). A similar sequence found in the FKB2 promoter also conferred inducibility on the CYC1-lacZ gene (26). However, whether induction of other ER stress-inducible proteins is mediated by a sequence similar to KAR2 and FKB2 UPREs remains to be investigated. Moreover, it is important to address the question of why only a limited set of proteins are transcriptionally induced in response to ER stress.
We recently conducted extensive mutational analysis of KAR2 UPRE to characterize its fine structure-function relationship (11). For convenience, we have numbered the nucleotides in UPRE, with nucleotide 1 the guanine at the 5Ј-end and nucleotide 22 the adenine at the 3Ј-end (see Fig. 1). Among the 22 nucleotides in KAR2 UPRE, nucleotides 10 -12 (CAG) and [14][15][16] were most critical for its function; point mutation of any of these nucleotides abolished the response to ER stress almost completely. The sequences CAG and GTG are reminiscent of the E box consensus (CANNTG), to which trans-acting factors containing a basic region as a DNA-binding domain would bind (29,30). However, KAR2 UPRE contained a single C residue between the half-sites (CAGCGTG) and this one-base spacing was critical for the response to ER stress. This unique spacing appeared to distinguish KAR2 UPRE from other cisacting elements recognized by basic region-containing transacting factors (11).
We and others recently identified the basic-leucine zipper protein Hac1p/Ern4p as a transcription factor responsible for the UPR in S. cerevisiae; haploid cells lacking Hac1p were unable to induce transcription of any of the target proteins tested and exhibited sensitivity to ER stress (10 -12). Hac1p was shown to bind specifically to the KAR2 UPRE using elec-trophoretic mobility shift assays (10,11). More importantly, we demonstrated that Hac1p recognizes the palindrome separated by a one-nucleotide spacer in KAR2 UPRE both in vivo and in vitro (11). Furthermore, we showed that Hac1p is involved in induction of all known target proteins (11,31). This indicated that, as in the case of KAR2 UPRE, the promoters of other target proteins might contain a palindromic sequence with a spacer of one nucleotide recognized directly by Hac1p.
Construction of Reporter Plasmids-Recombinant DNA techniques were carried out as described (35). The reporter plasmids pSCZ2 and pMCZ2 were constructed by modifying the CYC1-lacZ fusion gene of the multicopy vector pLG⌬-178 (36), a derivative of pLG670-Z (37), which is often utilized to assess activities of cis-acting elements in S. cerevisiae (38). Because the lacZ reporter gene in pLG670-Z as well as pLG⌬-178 is actually a translational fusion of lacI and lacZ (38), we transferred the CYC1 portion in pLG⌬-178 into another lacZ-containing vector in frame. Thus, inserted into pUC118 was the 0.25-kb XhoI-BamHI fragment of the CYC1-lacZ fusion gene in pLG⌬-178, which contained the entire CYC1 promoter and 5Ј-terminal several nucleotides of the CYC1coding region located upstream of the BamHI site (36,37). From the resulting plasmid, one nucleotide C between the start ATG codon and the BamHI site was deleted by site-directed mutagenesis (39) so that the ATG codon would be in frame to the lacZ-coding sequence embedded in pSEYc102 (a CEN4-ARS1-based single-copy vector) or pSEY101 (a 2-m-based multicopy vector), both containing the URA3 selectable marker (27). The 0.25-kb EcoRI (derived from the multicloning site of pUC118)-BamHI fragment of the mutagenized CYC1 gene was inserted between the EcoRI and BamHI sites of pSEYc102 and pSEY101 to create the single-copy vector pSCZ2 and multicopy vector pMCZ2, respectively. pMCZ2 possesses unique EcoRI and XhoI sites upstream of the CYC1 promoter for inserting oligonucleotides, whereas pLG⌬-178 contains only one XhoI site. Various double-stranded, synthetic oligonucleotides whose 5Ј-and 3Ј-termini are complementary to protruding termini generated by EcoRI and XhoI, respectively, were inserted between the EcoRI and XhoI sites of pMCZ2.
Comparison of ␤-galactosidase activity expressed from pLG⌬-178 and pMCZ2 suggested that the presence of the lacI portion between CYC1 and lacZ represses the expression of ␤-galactosidase. First, pMCZ2 produced considerably higher basal ␤-galactosidase (24 units) than pLG⌬-178 (1 unit) in unstressed cells. Second, the absolute units of ␤-galactosidase produced from pMCZ2 containing the wild-type UPRE-Y in cells incubated in the presence of tunicamycin for 3 h was 3-fold higher than those from pLG⌬-178 containing UPRE-Y. Third, activity of UPRE-A, a point mutant of UPRE-Y described previously (27), was 3% of that of UPRE-Y when inserted into pMCZ2, whereas it was almost negligible (0.3% of UPRE-Y) when inserted into pLG⌬-178. These results indicated that, although the extent of induction with pMCZ2 containing UPRE-Y is lower than that with pLG⌬-178 containing UPRE-Y (50ϳ60-fold versus 200ϳ300-fold induction by tunicamycin treatment for 3 h), ␤-galactosidase assay with pMCZ2 provides higher sensitivity for detecting weakly active elements than that with pLG⌬-178.
Constructs to Determine Promoter Activity-The 2.5-kb ApaI-PstI fragment containing the entire PDI1 gene in the plasmid pMTY17 (24) was cloned into pUC118, and a BamHI site was created immediately downstream of the start ATG codon (ATGGATCC---) by site-directed mutagenesis. The 0.44-kb ApaI-BamHI, 0.30-kb BstBI-BamHI, 0.21-kb SpeI-BamHI, or 0.09-kb MluI-BamHI fragment of the PDI1 promoter was inserted between the SmaI and BamHI sites of pSEYc102. Two point mutations were introduced to PDI1a UPRE by site-directed mutagenesis.
The plasmid pCT20 carrying both the EUG1 and FKB2 genes in tandem was kindly provided by Dr. T. H. Stevens (University of Ore-gon), and the 2.8-kb HindIII-SalI fragment was cloned into pUC119. The genomic LHS1 gene with franking regions was obtained by screening the yeast genomic library constructed on YEp13, a multicopy yeast vector (ATCC 37323, Ref. 40) and the 1.6-kb EcoRV fragment was cloned into pUC118. A BamHI site was created immediately downstream of the start ATG codon of EUG1, FKB2, or LHS1 (ATGGATCC---). The 0.19-kb HindIII-BamHI fragment of EUG1, 0.29-kb AccI-BamHI fragment of FKB2, or 0.27-kb BstXI-BamHI fragment of LHS1 was inserted between the SmaI and BamHI sites of pSEY101. Two point mutations were introduced to EUG1 UPRE, FKB2 UPRE, or LHS1 UPRE.
The plasmid pPS177 carrying the entire SCJ1 gene was kindly provided by Dr. P. A. Silver (Harvard Medical School). The 3.0-kb KpnI-HindIII fragment was transferred to pRS316 (a CEN6-ARSH4based single-copy vector containing the URA3 selectable marker, Ref. 41) to create pRS316-SCJ1. The 1.2-kb KpnI-SphI fragment of pPS177 was cloned into pUC118, and a BamHI site was created immediately downstream of the first or second ATG codon (ATGGATCC---). The KpnI-BamHI, StyI-BamHI, AflIII-BamHI, or HgaI-BamHI fragment was inserted between the SmaI and BamHI sites of pSEY101.
Assays-␤-Galactosidase assays and Northern blot hybridization analysis were carried out as described previously (7,11,31). Electrophoretic mobility shift assays were performed as described (11). Cell extracts were prepared from the wild-type strain (KMY1005) that had been grown in YPD medium to a mid-log phase and incubated for 1 h with tunicamycin, and proteins were fractionated by ammonium sulfate according to our previous report (31). Hac1p of 238 amino acids was translated in vitro using TNT™-coupled wheat germ extract system (Promega) and a template HAC1 DNA according to the manufacturer's instructions.

KAR2 UPRE Confers Inducibility on the CYC1-lacZ Gene
Regardless of Its Orientation-We constructed the reporter plasmids pSCZ2 (a single-copy vector) and pMCZ2 (a multicopy vector) as described under "Experimental Procedures." These plasmids contained a modified version of the CYC1-lacZ gene of pLG⌬-178 (36) and provided a sensitive and convenient assay system for examining cis-acting elements in S. cerevisiae. Using pMCZ2, we showed that the KAR2 UPRE (previously referred to as UPRE-Y, Ref. 27) contained a partial palindrome separated by one nucleotide (Fig. 1); nucleotides 10 -12 (CAG) FIG. 1. KAR2 UPRE confers inducibility by ER stress on the CYC1-lacZ gene regardless of its orientation. The nucleotide sequence of UPRE-Y (KAR2 UPRE) is shown at the top with numbering; nucleotide 1 is the guanine at the 5Ј-end, and nucleotide 22 is the adenine at the 3Ј-end. Arrows indicate the partially palindromic sequence separated by one nucleotide C (dotted). The most critical and important nucleotides for the function of UPRE are boxed and double underlined. The sec53 strain (KMY2005) was transformed with the multicopy vector pMCZ2 alone (none) or pMCZ2 containing UPRE-Y upstream of the CYC1-lacZ fusion gene in either direction (indicated by arrows). Transformants were grown at the permissive temperature of 23°C in SC(ϪUra) medium to a mid-log phase, and aliquots were incubated for an additional 90 min at 23°C (unfolded proteins in the ER Ϫ) or at the semi-nonpermissive temperature of 30°C (unfolded proteins in the ER ϩ). Total RNAs were extracted and analyzed by Northern blot hybridization using DNA probes specific for lacZ, KAR2, or yeast actin ACT1. and 14 -16 (GTG) were most critical (boxed), and nucleotides 8 (G), 13 (C), and 18 (C) were also important for UPRE activity (11). In previous experiments to dissect the UPRE, ␤-galactosidase activity was measured in cells incubated in the presence or absence of tunicamycin, which is known to elicit ER stress by inhibiting N-glycosylation of newly synthesized proteins in the ER (3,42).
Since tunicamycin treatment may cause pleiotropic defects in cellular metabolism, the response of the lacZ gene in pMCZ2 to ER stress was confirmed here at the mRNA level using the temperature-sensitive sec53 strain ( Fig. 1). At the nonpermissive temperature, sec53 cells accumulated full-length precursors of secretory proteins, which were abnormally glycosylated and malfolded in the ER due to the defect in phosphomannomutase activity (43), leading to activation of the UPR (16,17). The sec53 strain was transformed with pMCZ2 alone or pMCZ2 containing UPRE-Y. Total RNAs were isolated from transformants that had been grown at the permissive temperature of 23°C or after shifting to the semi-nonpermissive temperature of 30°C for 90 min. KAR2 mRNA transcribed from the chromosomal gene was markedly induced at 30°C in all transformants examined, because Kar2p expression was under the control of the KAR2 promoter containing a functional UPRE. In contrast, lacZ mRNA was not induced at 30°C in sec53 cells carrying pMCZ2 alone. When UPRE-Y was inserted into pMCZ2, lacZ mRNA was markedly induced at 30°C, indicating that transcription of the UPRE-CYC1-lacZ gene was induced only if unfolded proteins had accumulated in the ER regardless of the nature of the stress employed. Moreover, only slightly decreased response was observed when UPRE-Y was inserted in the opposite orientation, showing that UPRE functions as an upstream activator sequence regardless of its orientation. LHS1 UPRE is indeed in the opposite orientation to other functional UPREs (see Fig. 5).

Specific Binding of UPRE to UPRF in Cell Extracts and
Hac1p Translated in Vitro-The cellular activity for specific binding to UPRE, namely UPRF activity, was detected only in ER-stressed cells (10,31); binding to the wild-type UPRE-Y was obtained with extracts prepared from tunicamycin-treated cells ( Fig. 2A) but not from untreated cells (data not shown). This binding was specific because UPRF did not bind to a point mutant of UPRE-Y designated Tv10, the activity of which was less than 1% of UPRE-Y due to a transversion (C to A) at critical nucleotide 10 (11). The faster migrating band marked by an asterisk may represent a protein(s) recognizing sequences outside of the palindrome. Similarly, binding to UPRE-Y but not to UPRE-Tv10 was obtained with in vitro translated Hac1p ( Fig. 2A), which was demonstrated previously to be a transcription factor responsible for the UPR (10,11). When this UPRE-Y (KAR2 UPRE) was inserted into pMCZ2, 56-fold induction of ␤-galactosidase was observed after treatment of the wild-type cells with tunicamycin at 30°C for 3 h, whereas pMCZ2 alone caused marginal (2.7-fold) induction (Fig. 3). Thus, ␤-galactosidase assays with pMCZ2 as well as electrophoretic mobility shift assays using cell extracts and Hac1p translated in vitro allowed us to examine whether promoters of other ER stress-inducible proteins also included a palindromic sequence directly recognized by Hac1p. FKB2 UPRE Contains a Partial Palindrome Recognized by Hac1p-The FKB2 promoter was shown previously to contain a cis-acting element that conferred inducibility by ER stress on the CYC1-lacZ gene (26). We confirmed the observation using the pMCZ2 system; ␤-galactosidase was induced 10-fold by tunicamycin treatment, although the extent of induction was much lower than that with KAR2 UPRE (Fig. 3). When aligned with KAR2 UPRE, FKB2 UPRE was found to contain a partial palindrome separated by one nucleotide (C), although the 3Ј half was not well conserved. However, all three of the nucleotides shown to be important for KAR2 UPRE activity at 8, 13, and 18 (dotted in Fig. 3) were conserved. This FKB2 UPRE  (11). Both 5Ј-and 3Ј-termini of double-stranded probes, cohesive to EcoRI and XhoI sites, respectively, were radiolabeled using the Klenow fragment of DNA polymerase I and [␣-32 P]dATP, and protein-bound probes were separated from free probes in a 5% nondenaturing gel as described previously (11). The specific binding of UPRE-Y to a protein(s) in cell extracts (termed UPRF) or Hac1p translated in vitro is marked by a black or white arrowhead, respectively. The band marked by an asterisk, migrating faster than UPRF, seems to represent a protein(s) recognizing nucleotides outside of the palindrome in UPRE. B, the specific binding between 0.3 ng of 32 P-labeled UPRE-Y and UPRF in cell extracts (20 g of proteins, upper panel) or 0.2 l of in vitro translated Hac1p (lower panel) was competed by 50-or 250-fold molar excess of unlabeled UPRE-Y (KAR2 UPRE) or various UPRE-like sequences present in the promoters of ER stress-inducible proteins as indicated. Only specific binding is shown, and the positions of UPRF and Hac1p are marked as in A. Neither 5Ј-nor 3Ј-termini of competitor oligonucleotides were filled in, and possible formation of concatemers may explain the large molar excess of unlabeled KAR2 UPRE required for competition under these conditions. competed for the specific binding between 32 P-labeled UPRE-Y (KAR2 UPRE) and UPRF in cell extracts or Hac1p translated in vitro, albeit less efficiently than KAR2 UPRE (Fig. 2B), indicating that induction of Fkb2p by ER stress is mediated, at least in part, by the interaction of Hac1p with this FKB2 UPRE. No other potential UPRE sequences were found in the 263 nucleotides between the translational start site of FKB2 and the termination site of EUG1 known to precede FKB2 (44). Indeed, this FKB2 UPRE is necessary for the induction of Fkb2p (see below).
We showed previously that in KAR2 UPRE nucleotides 2-4 (GAA) located upstream of the partial palindrome are also important for UPRE activity; triple mutation of the sequence GAA to TCC decreased in vivo activity to approximately 5% of that of the wild-type sequence (11). However, this region was not conserved in FKB2 UPRE. We then replaced nucleotides 1-5 in FKB2 UPRE with those in KAR2 UPRE and measured activity of the mutant (designated FKB2Ј) UPRE. As shown in Fig. 3, the mutation caused a slight increase in basal activity without significantly affecting the extent of induction, suggesting that the UPRE activity is mainly determined by sequence integrity in the palindrome.
The PDI1 Promoter Contains a Functional UPRE-Among the six target proteins of the UPR so far known, only Pdi1p and Kar2p are essential for vegetative growth of the cell. The presence of a functional UPRE responsible for the induction of Pdi1p by ER stress was proposed previously by sequence comparison, but its activity has not been determined (5). We thus fused the 0.44-kb PDI1 promoter region in frame to lacZ in the single-copy vector pSEYc102 as described under "Experimental Procedures." The PDI1 promoter responded to ER stress by inducing a 6-fold increase in ␤-galactosidase expression when cells were treated with tunicamycin for 3 h (Fig. 4). The location of the previously proposed UPRE (referred to here as PDI1b UPRE, Ref. 5) is indicated in Fig. 4. Deletion of 0.14 kb between the ApaI and BstBI sites showed little effect on the promoter activity, whereas deletion of an additional 87 base pairs containing PDI1b UPRE resulted in a decrease in basal activity to nearly half of that of the 0.44-kb promoter without affecting the extent of induction. In PDI1b UPRE, as shown in Fig. 5, nucleotides 10 -12 and 14 -16 were not palindromic, although all three of the nucleotides important for UPRE ac-tivity at 8, 13, and 18 were identical to those in KAR2 and FKB2 UPREs. PDI1b UPRE hardly competed for the binding of either UPRF in cell extracts or Hac1p translated in vitro to 32 P-labeled KAR2 UPRE (Fig. 2B). When inserted into pMCZ2, PDI1b UPRE increased basal activity by 4.3-fold but affected the induction only slightly; ␤-galactosidase was induced 3.5-fold by tunicamycin treatment, whereas 2.4-fold induction was observed with the vector alone (Fig. 5). These results raise doubts about the role of PDI1b UPRE in the induction of Pdi1p. The PDI1b UPRE with the surrounding sequence may provide a binding site for some factor(s) important for basal expression of Pdi1p.
In contrast, deletion between the SpeI and MluI sites containing a putative TATA sequence abolished tunicamycin-induced expression of ␤-galactosidase almost completely (Fig. 4), suggesting the presence of a functional UPRE in this region. By sequence homology search, we identified a potential UPRE designated PDI1a UPRE, in which nucleotides 10 -12 and 14 -16 were perfectly palindromic and the half-sites were separated by one nucleotide, C (Fig. 5). When inserted into pMCZ2, PDI1a UPRE conferred inducibility by ER stress on the CYC1-lacZ gene (15-fold induction by tunicamycin). Transversion of the important nucleotide 18 (C to A) is likely to explain the lower inducibility exhibited by PDI1a UPRE than by KAR2 UPRE (see "Discussion"). PDI1a UPRE competed for the specific binding between 32 P-labeled KAR2 UPRE and  4) is double underlined. The wild-type strain (KMY1005) was transformed with pMCZ2 alone (none) or pMCZ2 containing UPRE as indicated. Transformants were grown at 30°C in SC(ϪUra) medium to a mid-log phase, and aliquots were incubated in the presence or absence of tunicamycin (TM). Samples taken after 3 h were used for ␤-galactosidase assays, and the activities presented are averages of duplicate determinations with three independent transformants. Standard deviation was less than 10% for all values shown. ⌬(%) is expressed as a percentage of KAR2 UPRE after subtracting ␤-galactosidase activities in unstressed cells (ϪTM) from those in tunicamycintreated cells (ϩTM).

FIG. 4. Analysis of the PDI1 promoter.
Schematic structure of the PDI1 promoter is presented at the top. The adenine of the first ATG codon is set as ϩ1. The locations of two UPRE-like elements (designated PDI1a and PDI1b) and a putative TATA sequence are shown by boxes. 5Ј-Deletion mutants were constructed using appropriate restriction enzymes indicated, and two point mutations were introduced into PDI1a UPRE (indicated by xx), which changed the central seven nucleotides from CACCGTG to CATCTTG, as described under "Experimental Procedures." These PDI1 promoters were fused in-frame to the lacZ-coding sequence in pSEYc102 (a CEN4-ARS1-based single-copy vector). The wild-type (ERN ϩ , KMY1005) or ern1⌬ (KMY1015) strain was transformed with each of these constructs. ␤-Galactosidase assays were carried out as in Fig. 3, and the activities are presented as means Ϯ S.D. (bars), based on duplicate determinations with three independent transformants.

Essential Features of Unfolded Protein Response Element
UPRF in cell extracts or Hac1p translated in vitro, albeit less efficiently than KAR2 UPRE (Fig. 2B). We found no other potential UPRE sequences in the region from Ϫ211 to Ϫ92.
When two point mutations were introduced into PDI1a UPRE present in the 0.44-kb promoter to inactivate both halfsites (CACCGTG was mutated to CATCTTG), induction of ␤-galactosidase by tunicamycin was markedly reduced and the level of the remaining response was very close to that of the wild-type promoter in the ern1⌬ strain (Fig. 4); the ern1⌬ strain cannot transmit the signal from the ER to activate transcription in the nucleus due to the absence of Ern1p, a transmembrane protein kinase localized in the ER and essential for signal transduction across the ER membrane (6, 7). We concluded from these results that PDI1a UPRE is necessary and sufficient for the induction of Pdi1p by ER stress.
A typical pattern of the induction of various endogenous mRNAs by tunicamycin treatment for 1 h is shown at the top of Fig. 5, and the correlation of the degree of induction with the corresponding UPRE activity will be discussed after characterization of all UPREs involved in the ER stress response (see "Discussion"). It should be noted, however, that the relative intensity of mRNA bands from different genes does not reflect the actual abundance in the cell; KAR2 and PDI1 mRNAs appeared to be much more abundant than other mRNAs, as suggested previously (18,25).
Characterization of the EUG1 and FKB2 Promoters-When the sequence proposed to be responsible for the induction of Eug1p by ER stress in the previous study (25) was inserted into pMCZ2, this putative EUG1 UPRE increased basal activity by 4.8-fold but conferred only slight inducibility (3.4-fold increase by tunicamycin versus 2.4-fold increase with the vector alone) as in the case of PDI1b UPRE (Fig. 5). However, this EUG1 UPRE competed for the specific binding between 32 P-labeled KAR2 UPRE and UPRF in cell extracts or Hac1p translated in vitro, albeit very weakly (Fig. 2B). Although the palindromic sequence in EUG1 UPRE is not typical of other functional UPREs, Hac1p seems to be capable of recognizing EUG1 UPRE (see "Discussion"). Sequences responsible for increased basal activity observed with EUG1 UPRE (Fig. 5) may be able to be separated from the palindromic sequence exerting weak UPRE activity. We found no other potential UPRE sequences in the 250 nucleotides located between the start ATG codon of Eug1p and the termination codon of the preceding open reading frame (ORF) of 372 amino acids (D9719. 22;Ref. 44). To determine the EUG1 promoter activity accurately, we used the multicopy vector pSEY101, considering the previous finding that Eug1p is at least 10-fold less abundant than Pdi1p (25). The EUG1 promoter conferred marked induction (66-fold) of ␤-galactosidase by tunicamycin (Fig. 6). However, it should be noted that it showed very low basal activity (less than 1 units) and conferred only mild induction (4 -5-fold) when inserted into the singlecopy vector pSEYc102. Mutation of two nucleotides in the EUG1 UPRE (CACGCGTG was changed to CACTCTTG) eliminated most of the response of the EUG1 promoter to ER stress. Similarly, introduction of two point mutations into the FKB2 UPRE mentioned above (CAGCGCA was mutated to CATCTCA) made the FKB2 promoter almost insensitive to ER stress (Fig. 6). Thus, the EUG1 and FKB2 promoters each contain a single UPRE which are primarily responsible for the induction of Eug1p and Fkb2p, respectively.
Identification of UPRE Responsible for the Induction of Lhs1p-The sequences proposed to be functional UPREs responsible for the induction of Lhs1p by ER stress in the previous reports (18,19) did not confer inducibility on the CYC1-lacZ gene (data not shown). Instead, we found a sequence that contained the same partial palindrome with a one-nucleotide spacer (CAGCGTG) as KAR2 UPRE. This LHS1 UPRE responded to ER stress by inducing a 29-fold increase in ␤-galactosidase expression when inserted into pMCZ2 (Fig. 5), although the orientation of the LHS1 UPRE used for in vivo assay was opposite to that present in the endogenous LHS1 promoter. In addition, this LHS1 UPRE competed for the specific binding between 32 P-labeled KAR2 UPRE and UPRF in cell extracts or Hac1p translated in vitro, albeit less efficiently than KAR2 UPRE (Fig. 2B). As Lhs1p is thought to be expressed at a much lower level than Kar2p (18), we also used the multicopy vector pSEY101 to analyze the LHS1 promoter. Twelve-fold induction of ␤-galactosidase by tunicamycin treatment was observed when the LHS1 promoter was fused in FIG. 5. Activities of UPRE-like sequences present in the promoters of ER stress-inducible proteins. Nucleotide sequences of various UPRE-like sequences are shown. Negative numbers at both sides indicate the locations of these sequences relative to the translational start sites set as ϩ1. Partially palindromic sequences between nucleotides 10 and 16 are indicated by arrows. Nucleotides identical to those important for KAR2 UPRE activity such as 2, 3, and 4 or 8, 13, and 18 are marked by double underlines or dots, respectively. A transversion known to mitigate UPRE activity is indicated by x. The wildtype strain KMY1005 was transformed with pMCZ2 alone (none) or pMCZ2 containing UPRE as indicated. ␤-Galactosidase assays were carried out as in Fig. 3, and the activities are presented as means Ϯ S.D. (bars), based on duplicate determinations with four independent transformants. At the top, a typical pattern of induction at the mRNA level is shown. Total RNAs were extracted from cells in a mid-log phase that had been treated or untreated with tunicamycin for 1 h, and analyzed by Northern blot hybridization. mRNA bands were visualized by exposing the filter to x-ray film for various times. Relative radioactivities of each band were determined using a BAS-2000 BioImaging Analyzer (Fuji Photo Film), and -fold induction is presented below. frame to lacZ (Fig. 6), and this induction was abolished almost completely by two point mutations introduced into the LHS1 UPRE redefined in this study (CAGCGTG was changed to CATCTTG). We concluded that Lhs1p is induced by ER stress through the interaction between Hac1p and this LHS1 UPRE, which functions as a cis-acting element that is necessary and sufficient for induction.
Apparent Absence of a Functional UPRE in the SCJ1 Promoter-The sequence proposed previously to be a functional UPRE responsible for the induction of Scj1p by ER stress (Ref. 22; indicated by the most downstream box f in Fig. 7) did not confer significant inducibility on the CYC1-lacZ gene when inserted into pMCZ2 (data not shown). Unlike the other promoters of ER stress-inducible proteins, an ORF of 590 amino acids (YM8261.07) was predicted immediately upstream of two putative TATA sequences for SCJ1 (see Fig. 7 and also compare the broken arrow in Fig. 9 with those in Fig. 6; see also Ref. 45). We found no other potential UPRE sequences between the termination codon of this ORF and the start ATG codon of Scj1p. By sequence homology search, two sequences (indicated by boxes a and d in Fig. 7) in which a partial palindrome was separated by one C nucleotide (CAGCGTA) and a sequence (box b) in which a perfect palindrome was separated by one C nucleotide (CAGCCTG) were found within the upstream ORF. When inserted into pMCZ2, however, these putative UPREs did not confer inducibility on the CYC1-lacZ gene, probably because two important nucleotides 8 and 18 were not conserved (see "Discussion"). Only the results with one putative UPRE (box d) are shown in Fig. 5. This putative SCJ1 UPRE increased basal activity by 2.9-fold without conferring significant inducibility. In addition, this sequence did not compete for the specific binding of 32 P-labeled KAR2 UPRE to UPRF or Hac1p (Fig. 2B). We tested a total of six UPRE-like sequences present in the upstream region (indicated by boxes a-f in Fig. 7), but none were functional in ␤-galactosidase induction (data not shown).
Next, we transferred the 3.0-kb KpnI-HindIII fragment con- 6. Analysis of the EUG1, FKB2, and LHS1 promoters. Schematic structures of the EUG1, FKB2, and LHS1 promoters. The adenine of the first ATG codon is set as ϩ1. The locations of UPRE sequences are shown by boxes. Closed boxes indicate putative TATA sequences, and broken arrows denote the locations and directions of preceding ORFs. Two point mutations were introduced to create mutant UPRE, which changed the central seven nucleotides from ACGCGTG to ACTCTTG for EUG1, CAGCGCA to CATCTCA for FKB2, and CAGCGTG to CATCTTG for LHS1. These promoters containing wildtype (wt) or mutant UPRE were fused in frame to the lacZ-coding sequence in pSEY101 (a 2-m-based multicopy vector) as described under "Experimental Procedures." The wild-type (ERN ϩ , KMY1005) or ern1⌬ (KMY1015) strain was transformed with each of these constructs. ␤-Galactosidase assays were carried out as in Fig. 3, and the activities are presented as means Ϯ S.D. (bars), based on duplicate determinations with three independent transformants. FIG. 7. Apparent absence of a functional UPRE in the SCJ1 promoter region. Figure shows nucleotide sequence of the upstream region of the SCJ1 gene and the initial portion of the coding region with its deduced amino acids. Two putative TATA sequences and two methionines are underlined and circled, respectively. The termination TAG codon of the preceding ORF is marked by three asterisks. The positions of several restriction enzyme sites are also indicated. Six boxed sequences (a-f) were inserted into pMCZ2, but the resulting constructs did not respond to ER stress by inducing ␤-galactosidase (see "Results"). Nucleotides between 8 and 18 in various UPRE-like sequences are marked by arrows, dots, and lowercase x as in Fig. 5. taining the SCJ1 gene into pRS316 (a single-copy expression vector; Ref. 41) and introduced the resulting plasmid pRS316-SCJ1 into the scj1⌬ strain. In the scj1⌬ strain carrying pRS316 alone, SCJ1 mRNA was not expressed at all as expected (Fig.  8). SCJ1 mRNA transcribed from the plasmid pRS316-SCJ1 in the scj1⌬ strain was induced by tunicamycin treatment for 1 h; the pattern was almost indistinguishable from that in the parental SCJ ϩ strain, indicating that this 3.0-kb fragment is sufficient for the induction of Scj1p by ER stress.
We then determined the SCJ1 promoter activity by fusing it to lacZ in the multicopy vector pSEY101. In marked contrast to the promoters of other ER stress-inducible proteins, ␤-galactosidase was neither expressed constitutively nor induced by tunicamycin treatment when the lacZ-coding sequence was fused immediately downstream of the initial methionine ( Fig.  9). 5Ј-Deletions showed no effect on the lack of expression of ␤-galactosidase, indicating no silencing activity in the upstream region. Interestingly, significant ␤-galactosidase activities were detected when the lacZ-coding sequence was fused immediately downstream of the second methionine, providing strong support to the previous finding (22) that Scj1p is translated from Met 28 (see "Discussion"). However, the full-length sequence and various 5Ј deletion mutants responded to ER stress very poorly inducing ␤-galactosidase by only 2-3-fold, levels much lower than those induced by EUG1 or LHS1 (compare Fig. 9 with Fig. 6), although all three showed similar induction by tunicamycin at the mRNA level (see Fig. 5). It was concluded that the promoter region alone is insufficient for transcriptinal regulation of the SCJ1 gene, and we are currently investigating the region(s) responsible for the induction of Scj1p by ER stress. DISCUSSION Recent work conducted in two laboratories established that the UPR is mediated by a specific and direct interaction between the transcription factor Hac1p and cis-acting element UPRE (10,11). Hac1p was shown to be responsible for the induction of all known target proteins (11,31). However, only the promoters of Kar2p and Fkb2p were actually known to contain a functional UPRE (26,27). The KAR2 and FKB2 UPRE sequences were so divergent (see Fig. 3) that it was difficult to assess essential features of the UPRE. Indeed, although putative UPREs were proposed previously to be present in the promoters of Pdi1p (5), Scj1p (22), Lhs1p (18,19), and Eug1p (25), most of these were hardly active as shown in this study.
Our recent extensive mutational analysis of KAR2 UPRE shed some light on this issue (11). The results revealed that KAR2 UPRE contains an E box (CANNTG)-like palindromic sequence that provides a binding site for Hac1p. Interestingly, Hac1p exhibits a strong preference for a spacer of one nucleotide between the half-sites, apparently a characteristic specific to Hac1p. The lack of spacing reduced induction to 13% of the normal level, whereas increased spacing abolished induction almost completely. Hac1p also shows a preference for a specific spacer nucleotide in the order C Ͼ G Ͼ A Ͼ T. Furthermore, CAC is preferred to CAG as a half-site (11). 2 Thus, the UPRE activity of seven nucleotides containing two half-sites with a one-nucleotide spacer (referred hereafter as "central seven nucleotides") is highest with CACCGTG and decreases in the order CACCGTG Ͼ CAGCGTG Ͼ CAGCCTG.
In this study, we demonstrated that each of the KAR2, PDI1, EUG1, FKB2, and LHS1 promoters contains a single UPRE that is necessary and sufficient for induction by ER stress. Importantly, the levels of each UPRE activity determined by in vivo (Fig. 5) and in vitro (Fig. 2B) analyses correlated roughly with those of mRNA induction (Fig. 5, top); KAR2 UPRE showed the highest level of activity, FKB2 UPRE the lowest, and PDI1, EUG1, and LHS1 UPREs showed intermediate ac- FIG. 8. The 3.0-kb KpnI-HindIII fragment containing the SCJ1 gene fully responds to ER stress. Schematic structure of the plasmid pRS316-SCJ1 is shown at the bottom, which was constructed as described under "Experimental Procedures." The position and direction of the SCJ1-coding region is indicated. The wild-type (SCJ ϩ , SEY6210) or scj1⌬ (SNY1025) strain was transformed with pRS316 alone (pRS316-V) or pRS316-SCJ1. Transformants were grown at 30°C in SC(ϪUra) medium to a mid-log phase, and aliquots were incubated in the presence (ϩ) or absence (Ϫ) of tunicamycin (TM) for 1 h. Total RNAs were extracted and analyzed by Northern blot hybridization.
FIG. 9. The SCJ1 promoter region responds poorly to ER stress. Schematic structure of the SCJ1 gene is presented at the top with several restriction enzyme sites. The adenine of the first ATG codon is set as ϩ1. Two closed boxes indicate putative TATA sequences, and the broken arrow denotes the location and direction of the preceding ORF. The full-length and various 5Ј-truncated promoter regions were fused in frame to the lacZ-coding sequence in pSEY101 from the first (Met 1 ) or second (Met 28 ) methionine as described under "Experimental Procedures." The wild-type strain (KMY1005) was transformed with each of these constructs. ␤-Galactosidase assays were carried out as in Fig. 3, and the activities are presented as means, based on duplicate determinations with three independent transformants. Standard deviation was less than 15% for all values shown. n.d., not determined.
tivities. The correlation would be even more significant if we consider the orientation of LHS1 UPRE. When inserted in the opposite orientation, KAR2 UPRE showed slightly decreased activity (Fig. 1). By analogy, the degree of LHS1 mRNA induction may be lower than that expected from the LHS1 UPRE activity obtained with in vivo assay, and therefore PDI1, EUG1, and LHS1 mRNAs may show similar levels of induction (Fig. 5). In addition, the degree of KAR2 mRNA induction might be reduced by the high basal expression level of KAR2 mRNA due to the heat shock element and the GC-rich region present in the KAR2 promoter (27) but lacking in the other five promoters.
Transcriptional regulation of Scj1p was exceptional; the promoter region apparently lacks any functional UPREs (Fig. 7) and is insufficient for induction (Fig. 9). Interestingly, ␤-galactosidase was expressed only when the lacZ-coding sequence was fused immediately downstream of Met 28 but not of Met 1 , consistent with the previous report that Scj1p is translated from the second methionine (22). If Scj1p was translated from the first methionine, the first 25 amino acids would be predicted to function as a mitochondrial targeting signal (46), whereas if the second methionine was used to initiate translation, the next ϳ20 amino acids would target Scj1p into the lumen of the ER (22). Therefore, regardless of the mechanism, the preferential starting of Scj1p translation at Met 28 would ensure its function as an ER-resident molecular chaperone. In fact, the absence of Scj1p caused slight activation of the UPR; the level of KAR2 mRNA was approximately 2-fold higher in tunicamycin-untreated scj1⌬ cells than that in untreated SCJ ϩ cells presumably because of increased amounts of unfolded proteins accumulated in the ER (Fig. 8).
All of the five functional UPREs identified contained a palindromic sequence (Fig. 5) and competed for the specific binding between 32 P-labeled KAR2 UPRE and Hac1p with various efficiencies (Fig. 2B). Based on the results of analysis of a number of active and inactive UPRE-like sequences, we can now deduce some important features required for the interaction between Hac1p and UPRE.
First, Hac1p recognizes not only the central seven nucleotides but also certain surrounding sequences. The central seven nucleotides, CACCGTG or CAGCGTG, of PDI1a or LHS1 UPRE, respectively, may be expected to confer higher or equal activity, respectively, as compared with KAR2 UPRE containing CAGCGTG as discussed earlier. However, in vivo activities of PDI1a and LHS1 UPREs were much lower than that of KAR2 UPRE (Fig. 5). Since mutations of nucleotides 1-5 in PDI1a UPRE from CCAAT to GGAAC (KAR2 type) increased in vivo activity only slightly (data not shown), as was the case of FKB2 and FKB2Ј UPREs (Fig. 3), a "natural" transversion of the important nucleotide 18 seemed to be responsible for the weak activity of PDI1a UPRE observed; in the case of KAR2 UPRE, the same transversion reduced the activity to less than 20% (11). On the other hand, nucleotides 8 and 18 in LHS1 UPRE are transitions rather than transversions of those in KAR2 UPRE, perhaps explaining why LHS1 UPRE is more active than PDI1a UPRE.
Second, at least one half-site sequence must be CAG or CAC since PDI1b UPRE is virtually inactive although all three important nucleotides 8, 13 and 18 are conserved. Third, when one of the half-site sequences diverges from CAC or CAG, the three important nucleotides 8, 13, and 18 must be conserved, because FKB2 UPRE is active, whereas SCJ1 UPRE shown in Fig. 5 as well as other UPRE-like sequences shown in Fig. 7 are virtually inactive.
Finally, the case of EUG1 UPRE is exceptional, because its half-sites are separated by two nucleotides (CACGCGTG), which is usually inactive as mentioned earlier. However, this EUG1 UPRE competed for the binding between KAR2 UPRE and Hac1p, although very weakly (Fig. 2B). The presence of two overlapping half-sites with no spacing (CACGCG and CGCGTG) flanked by two important nucleotides (at positions 8 and 18) might explain the recognition of EUG1 UPRE by Hac1p.
In addition to the one-nucleotide spacing between the E box-like half-sites, sequences outside of the central seven nucleotides are important for UPRE recognition by Hac1p. This unique characteristics may distinguish UPRE from among various cis-acting elements recognized by basic region-containing transcription factors, and are likely to explain why transcription of only molecular chaperones and folding enzymes in the ER is induced when unfolded proteins are accumulated in the ER. Without this induction system, yeast cells cannot survive under ER stress conditions (6,7,11). Furthermore, increased synthesis of the target proteins of the UPR in the absence of excess unfolded proteins in the ER is also toxic to the cell (10,31). Therefore, the UPR must be tightly regulated to meet the requirements of this organelle. Recently, we and others showed that Hac1p itself is induced by ER stress and the induction is mediated by unconventional splicing of HAC1 precursor mRNA (10,13,31,47,48). The unique features of UPRE revealed in this study suggest a basis for the specificity of the UPR; Hac1p induced under ER stress conditions may activate transcription of a limited set of proteins only necessary to cope with deleterious effects of unfolded proteins accumulated in the ER. The UPR, an intracellular signaling from the ER to the nucleus, appears to possess multiple distinguished characteristics among biological signal transduction systems.