Isolation and Characterization of the Mouse Heme Oxygenase-1 Gene DISTAL 5’ SEQUENCES ARE REQUIRED FOR INDUCTION BY HEME OR HEAVY METALS*

Mouse genomic fragments encoding heme oxygen-An HO-l/CAT fusion gene, pMHOOcat, containing the ase-1 (HO-1) were isolated from a recombinant A library HO-1 proximal promoter and the 9-kbp upstream frag-by in situ plaque hybridization. The mouse HO-1 gene, ment, is not inducible in C6 cells, but is activated 11- and approximately 7 kilobase pairs (kbp) in length, is orga-12-fold by Cd2+ and heme, respectively, in Hepa cells, nized into 5 exons and 4 introns. The primary structure indicating tissue-specific variation in the activation of the exons and 1287 base pairs (bp) of the 5’-flanking mechanism. The active, inducer-dependent element(s) is region was determined. The deduced amino acid se-located within a 268-bp fragment previously shown to quence of the mouse HO-1 gene is identical to that of p32, mediate 12-0-tetradecanoylphorbol-13-acetate induc- initially identified as a stress-induced protein in mouse tion of the mouse HO-1 gene (Alam, J., and Zhining, D. BA.LBd3T3 cells. Asingle, major transcription initiation (1902) Chromosomal in-site

site is utilized for constitutive and hemeor metal-in-tegration of the HO-1ICAT constructs is required for duced expression of the HO-1 gene in mouse hepatoma heme-or Cd2+-mediated activation in Hepa cells, sug-(Hepa) cells. The transcriptional activity of the 6'-flank-gesting a role for chromatin organization in the inducing region was examined by transient expression assays tion process. using the chloramphenicol acetyltransferase gene as the reporter gene. Basal promoter activity in several cell lines Was localized to within 149 bP Of the upstream Microsomal heme oxygenase (EC 1. 14 C/EBP and c"yc"sF' Heavy the rate-limiting step in the conversion of potentially toxic tivate HO-1 gene transcription and the rat gene contains a putative metal element (Miiller, M., heme to bile pigments with conservation of the iron atom. Fur-6802) that is comp~ete~y in the mouse gene. that carbon monoxide derived from the enzymatic degradation Transient expression analyses, however, indicate that Of heme may fUnction as a messenger.
this sequence, which contains a core heptanucleotide, ' h o isozymes of heme oxygenase, HO-1 and HO-2, have been TMACTC, identical to that of the strongest metal r e m identified to date in all animals tested. The HO isozymes can be latory element of the mouse metallothionein-1 gene, is clearly distinguished by several biochemical and biophysical not responsive to Cd2+ or Zn2+. Stable transfection of parameters: apparent molecular weight, temperature versus constructs containing the entire mouse HO-1 gene and enzymatic activity profile and kinetics of reactivity with hemavarious portions of the 5'-flanking region into rat C6 tin; differ in abundance in various tissues; and are immunoglioma cells and simultaneous, quantitative analysis of logically distinct. The most striking difference observed, howthe mouse and rat HO-1 indicate that distal 5' ever, is the dissimilarity in the regulation of HO-1 and HO-2 SeQuenCeS, between Positions -3.6 and - 12.6 kbP, are reexpression. In general, HO-2 is constitutively expressed wired for induction Of mouse HO-1 gene transcription whereas the expression of HO-1 is dramatically induced in by both heme and heavy metals. A S-7-fo1d difference in response to the substrate heme and a variety of stress-associthe levels of induction between stably integrated and ated agents including heavy metals, hyperthermia, and transiently expressed mouse HO-1 gene constructs is obirradiation (reviex:,ed in Ref. 2). served in this cell line.
chi, H., and Shibahma, s. (1987) J. BioZ. Chem A.) and DK-37463 A. s. ). The Costs Of publi-hepatoma (Hepa) cells and that this induction is controlled cation of this article were defrayed in part by the payment of page in accordance with 18 (5) and UV irradiation (6) also activate transcription of the HO-1 gene. Furthermore, induction of the rat HO-1 gene by hyperthermia is apparently mediated by a heat shock response element (HSE) located within the promoter of this gene (7,8 To better define the mechanisms involved in the regulation of HO-1 gene transcription, we have isolated the gene encoding the mouse HO-1 enzyme. Previous studies utilizing transient expression assays were unable to demonstrate heme-mediated induction of chimeric genes containing either 549 bp of the rat or 1.4 kbp of the human HO-1 gene 5'-flanking regions (8,9). However, by stable integration of constructs containing the entire mouse HO-1 gene and various portions of the 5"flanking region into rat C6 glioma cells, we now show that distal 5' sequences, between -3.5 and - 12.5 kbp, are required for transcriptional activation by both heme and heavy metals. This 9-kbp fragment mediates inducer-dependent activation of the heterologous CAT gene in Hepa cells. The cis-acting sequences are located within a 268-bp fragment previously shown to mediate 12-0-tetradecanoylphorbol-13-acetate stimulation of the mouse HO-1 gene (10). Furthermore, optimal induction of the transfected mouse HO-1 gene or the HO-l/CAT fusion genes requires stable integration into chromosomal DNA, implying a potential role for chromatin structure in the transcriptional activation process.

EXPERIMENTAL PROCEDURES
Materials-Restriction endonucleases and other DNAfRNA modifying enzymes were purchased from either Life Technologies, Inc. or Sigma. DNA sequence analysis was carried out in part using the Sequenase Version 2.0 Kit (U. S. Biochemical Corp.); all radiolabeled nucleotides were obtained from DuPont NEN. Mesoheme and hemin were purchased from Porphyrin Products. Hemopexin was isolated from rabbit serum, and mesoheme-hemopexin complexes were prepared, purified, and characterized as previously described (11). All enzymes and reagents for CAT and luciferase assays were purchased from Sigma. All other chemicals were reagent grade.
Isolation a n d Sequencing of the Mouse HO-1 Gene-A recombinant A library, constructed by cloning partially digested and size selected MboI fragments of mouse genomic DNA into the replacement vector EMBL 4, was kindly provided by Dr. J. Casey. The library was screened by in situ plaque hybridization (12) using a 5' fragment of the rat HO-1 cDNA (13), and hybridization-positive phage clones were isolated by repeated plaque purification. A DNA was prepared, and subfragments were cloned into the phagemids pUC118/119. Sequence analysis was camed out by both dideoxynucleotide chain-termination (14) and chemical (15) methods and was determined on both strands of a given fragment. Preparation and Analysis of RNA-Total RNA from cultured cells was prepared by the acid guanidinium thiocyanate-phenol-chloroform extraction method (16). The 5' end of the mouse HO-1 mRNA was mapped by S1 endonuclease and primer extension analyses using standard techniques (17,18). The DNA probes for each assay were prepared in the following manner. A dodecanucleotide (5"TCGGGCT-GTGGACGCTCCAT-3', complementary to the sense strand of the HO gene from residues +129 to +148 (residue +1 corresponds to the transcription start site)) was labeled at the 5' end with [y-32PlATP using T4 polynucleotide kinase, hybridized to denatured plasmid DNA containing the appropriate HO-1 genomic sequence, and extended using the Klenow fragment and unlabeled deoxynucleoside triphosphates. The extended product was digested, either with XhoI (to generate a primer extension probe 79 bases in length) or with ApaI (to generate an S1 endonuclease probe 257 bases in length), and resolved on a polyacrylamidehrea gel. The end-labeled, single-stranded probes were subsequently eluted from the gel slices. RNase protection assays were carried out as previously described (19). The in vitro generated RNA probe used for these assays is 330 bases in length and contains sequences complementary to residues -60 to +211 ofthe mouse HO-1 gene and 59 bases of vector-derived sequence. Appropriately protected probe species were quantified using a PhosphorImager (Molecular Dynamics, Mountainview, CA).
The mouse HO-1 gene expression plasmids used to transfect C6 glioma cells were constructed as follows. The 3.5-kbp BamHVEcoRI fragment (containing exons 4 and 5) of AMHO4-l (see Fig. 1) was cloned into pUC119. The 6.5-kbp BamHIIBamHI fragment (containing exons 1 3 ) of AMHO4-1 was cloned into the BamHI site of the previous construct to generate pMHO4a. Plasmid pMH04a was partially digested with BamHI, the ends were filled in with the Klenow fragment, and the DNA was recircularized to yield pMH04 in which the BamHI recognition sequence between exons 3 and 4 has been destroyed. Plasmids pMH04B1 and pMH04B9 were constructed by cloning the 2.2-kbp (-1288 to -3.5 kbp) and 9.0-kbp (-3.5 to -12.5 kbp) BamHVBamHI fragments of AMHO3-1, respectively, into pMH04. The 11.5-kbp (-3.5 to -15 kbp) BamHIIBamHI fragment of AMH02-1 was cloned into pMH04 to yield pMHO4B11. The orientation of the inserts was determined by restriction endonuclease mapping.
An EcoRVNcoI (blunt ended) fragment (residues -861 to -589) was cloned between the EcoRVSmaI sites of pUC119 to generate plasmid pMH0R.N-1. This construct contains both of the putative metal regulatory element (MRE) of the mouse HO-1 gene. Plasmids pMHOMREx4 and pMTMREx2 contain four tandem copies of the putative mouse HO-1 MREa and two tandem copies of the mouse MT-1 MREd, respectively, cloned in the BamHI site upstream of the minimal HO-1 promoter in the plasmid pMHOlcatA-33. The complementary oligonucleotides used to prepare these constructs are: 5"GATCCGGGTGCT-GCACTCCATGAG-3' and 5'-GATCCTCATGGAGTGCAGCACCCG-3' (HO-1 MREa) and 5'-GATCCGAGCTCTGCACTCCGCCCG-3' and 5'-Cell Culture, Dansfection, and Enzyme Assays-Mouse fibroblast L929 cells, rat C6 glioma cells, and rat hepatoma H4IIEC3 cells were maintained in Dulbecco's modified Eagle's medium containing 0.45% glucose and supplemented with 10% fetal bovine serum. Mouse Hepa cells were cultured in the same medium containing 2% fetal bovine serum. Stable transfectants were generated by the calcium phosphate precipitation technique (23). C6 or Hepa cells (-1-2 x 106/10-cm plate) were exposed to the DNA-CaP04 precipitate for 16 h. (The DNAmixture was comprised of 10 pg of the mouse HO-1 gene expression plasmid or HO-1/CAT fusion constructs and 1 pg of pSV2neo (24).) The precipitate was removed and the cells were cultured in complete medium. Geneticin (G418 sulfate) was added 24 h later to a final concentration of 400 pg/ml, and resistant colonies were selected over a 2-week period. The colonies (between 20 and 50/plate) were pooled and propagated in the presence of G418. Transient transfection was also camed out by the calcium phosphate precipitation method. CAT activity was determined by the procedure of Nordeen et al. (25), and luciferase assays were carried out as previously described (26)

RESULTS AND DISCUSSION
Isolation of A Clones-Approximately 5 x lo6 clones from a mouse genomic A library were screened by in situ plaque hybridization using a 5' fragment of a full-length rat HO-1 cDNA (13) as the probe. Assuming that all phage particles are recombinant and an average insert size of 15 kbp, this number of plaques represents approximately 2.5 genome equivalents. After successive purification of hybridization positive plaques, three independent clones were isolated. Dot blot analysis of A lysates using subfragments of the rat HO-1 cDNA as hybridization probes indicated that one of the clones, AMHO4-1, con-  TCTTCTCTTC TCTTCTCTTC TCTTCTCTTC TCTTCTCTTC TCTTCTCTTC TCTTCTCTTC TCTTCTCTTC   TCTTCTCTTC TCTTCTCTTC TCTTCTCTTT TTTTCTCTCT CTCTCCCTTT TTTTTT,GTAA CAGACTTGCC   -1  tained the entire gene, whereas clones AMHO2-1 and AMHO3-1 contained primarily 5"flanking sequences. Structural Organization and Sequence Analysis of the Mouse HO-1 Gene-A partial restriction endonuclease map of the mouse HO-1 gene and its flanking sequences is presented in Fig. lA. An equivalent map was obtained by Southern blot analysis of genomic DNA from mouse liver or mouse L929 fibroblast cells (data not shown). All analyses indicate that HO-1 is encoded by a single copy gene. The mouse HO-1 gene, approximately 7 kbp in length, is organized into five exons interrupted by four introns and, in this respect, is equivalent to the rat (7) and human (9) HO-1 genes.

0 m T A T GATTTATCCC CTTACAGGCA ~GGcCCGCCT C]CGGGCTGGAITGTTGCAAC$ GCAGCGAGAA C G C C G G C w ] G G C CAC~CACGTGJA CCCGCGTAdT T d G G G C T G G C G G G C A G CTGCTCGCTC
Cloned subfragments of the insert in AMHO4-1 containing exons, exodintron junctions andor 5'-flanking region were identified and subjected to DNA sequence analysis. The combined nucleotide sequence of the exons (data not shown) is identical to that of the cDNA (27) encoding a mouse protein, p32, initially identified as a 32-kDa polypeptide rapidly induced by the phorbol ester, TPA, and chemical carcinogens, metal salts, and heat shock in BALB/c 3T3 cells (28, 29). Sequence similarity between the cDNAs for p32 and rat HO-1 suggested that the former is the mouse homolog of the rat   The position and phase of the introns with respect to the coding sequence are completely conserved between the mouse, rat and human genes. As summarized in Table I, all the exon/ intron junctions obey the GT/AG rule (34) with the exception of the 5"splice donor site of intron 1 where the second position is occupied by a C residue. Interestingly, this residue is conserved in intron 1 of the mouse, rat and human genes. Deviation from the GT dinucleotide of the 5"splice site is extremely rare, with only 19 substitutions observed at the second position in 3,294 introns from plants to primates (35). The conservation of the GC dinucleotide in intron 1 of the three HO-1 genes, therefore, suggests a potential role for this junction in the regulation of HO-1 gene expression. In this regard, it is noteworthy that a mutation from GT to GC in the human pro-a 2(I) collagen gene causes exon skipping (36), suggesting that the efficiency of splicing with a GC donor site is low, at least in the context of this collagen gene. The nucleotide sequence of the 5"flanking region of the mouse HO-1 gene is presented in Fig. 1B and discussed in more detail below.
Mapping of the Dunscription Initiation Site-The transcription initiation site of the mouse HO-1 gene was determined by S1 endonuclease and primer extension analyses (Fig. 2). Only one major start site is observed with both analyses using RNA from CdC12-treated Hepa cells. A smaller protected fragment, only occasionally detected with S1 endonuclease mapping, is probably due to overdigestion of the RNADNA hybrid. The prematurely terminated primer extension product (which is more abundant in the experiment depicted in Fig. 3 and is not inducer-regulated) is also an experimental artefact caused by self-priming and extension of the probe (data not shown). The proposed transcription initiation sites of the mouse HO-1 and rat HO-1 gene (7) are at analogous residues. The start site of the uninduced mouse HO-1 transcript in Hepa cells is identical to that of transcripts induced by heme, heme-hemopexin, ZnClz, or CdClZ (Fig. 3). The induction of the HO-1 transcript by Zn2+ is noteworthy. We recently reported (37) that incubation of Hepa cells with 10 p~ ZnC12 increases the steady-state level of MT transcripts 20-fold without altering the level of HO-1 mRNA. As shown here, the HO-1 gene is clearly not refractory to induction by Zn2+, but requires a much higher concentration of this metal ion. In contrast, a given concentration of CdC12 (1-10 p~) induces MT and HO-1 transcripts to similar levels in Hepa cells.2 A second, inducer-regulated primer extension product (EP2, Fig. 3) is observed upon prolonged autoradiography. The nature of this product is unknown but appears to represent a minor transcription initiation site, corresponding to residue position - 56. Structural and Functional Analysis of the Proximal 5'-FZanking Region-The DNA sequences upstream of the proposed transcription initiation site of the mouse and rat HO-1 genes (7) are highly conserved, with the notable exception of an insertion of a polypyrimidine tract, comprised mainly of the J. Alam  repeated pentamer TCTCT, in the mouse gene. Homopurine/ homopyrimidine sequences are present in the promoter regions of several genes including the human c-Ki-ras proto-oncogene (38) and the Drosophila heat shock genes, hsp70 and hsp26 (39). These domains are proposed to function as transcription effectors largely on the observation that they interact with sequence specific DNA-binding proteins. Excluding the polypyrimidine tract, the mouse and rat HO-1 gene 5"flanking regions exhibit 86% sequence similarity.
Muller and colleagues (7) have identified within the 5'-flanking region of the rat HO-1 gene a single sequence motif similar to the consensus HSE: at least three copies of alternating G M I T C blocks separated by 2 nucleotides (40). The analogous sequence, HSEa, and two other potential HSEs, are present in the mouse HO-1 gene 5"flanking region (Fig. 1B). None of these putative HSEs, however, contains the three intact GAA/ I T C blocks, which may explain the weak response of the mouse HO-1 gene to hyperthermia (3). On the other hand, the human HO-1 gene promoter contains, by the above definition, an intact HSE but is unresponsive or only minimally responsive to hyperthermia in several cell lines (9, 41). In vivo, the rat HO-1 gene is clearly activated by hyperthermia (42) but similar analyses have not been carried out in other animals.
To examine the transcriptional activity of the 5'4anking sequence, a BarnHVXhoI fragment (-1287 to +73) was cloned upstream of the Escherichia coli CAT gene. The resulting construct, pMHOlcat, was transiently transfected into cultured cells and transcriptional activity of the test DNA was deduced from CAT activity in cellular extracts. Plasmid pMHOlcat is expressed in all cells tested (Fig. 4). Deletion of the putative MREs does not affect basal expression of the chimeric gene unlike that observed with most MT promoters where the MREs are interdigitated with basal expression elements (43). Deletion of the putative HSEs and the polypyrimidine tract also has no effect on basal activity and full activity is observed with only 149 bp of the 5'-flanking sequence. Comparison of this region with the rat and human HO-1 promoters reveals several segments of sequence conservation, including elements that resemble consensus binding sites for the transcription factors AP-1 (441, C/EBP (43, AP-4 (46), and c-Myc (47) (see Fig. 1B) . Removal of the putative AP-1 site reduces basal expression between 2-and 4-fold. This AP-1 motif, TGACTCT, (complementary to that in Fig. 0 1 B ) deviates from the consensus sequence, TGA(G/C)TCA, at the final position. While purified c-Jun homodimers do not bind to this site (lo), various Jun homodimers and/or Fos:Jun heterodimers may interact with this sequence in the context of the nuclear milieu. The c-Myc consensus recognition sequence, CACGTG, also overlaps with the binding site for the heme oxygenase transcription factor, probably identical to or a variant of USF, that is proposed to regulate basal expression of the rat and human HO-1 genes (48,49).
The rat HO-1 gene contains a putative MRE identified by sequence similarity to the consensus MRE of MT genes (7). This element, MREa, is completely conserved in the mouse HO-1 gene, and a second element (MREb, TGCTCTC) that deviates only slightly from the core segment of the consensus MRE, TGC(A/G)CNC, is also present in the mouse and the rat HO-1 genes. The role of these elements in induction of HO-1 gene transcription was tested by examining the effects of heavy metals on the expression of the fusion gene, pMHOlcat. In all cells tested, the expression of pMHOlcat was unresponsive to CdCl2 (up to 10 w) or ZnC12 (up to 100 w), while the mouse MT-1 promoter was maximally activated between 6-and 9-fold with 2.5 w CdC12 ( Fig. 5A and data not shown). Furthermore, a fragment containing the HO MREs did not competitively inhibit the activation of the MT-1 promoter, even when present at an almost 100-fold molar excess (Fig. 5B). In fact, both the basal and Cd2+-induced expression of pmMT-lcat was increased 2-fold in the presence of the competitor. The HO MREa core sequence, TGCACTC, is identical to the core segment of the mouse MT-1 MREd, the element which individually responds most strongly to heavy metals (50). %o tandem copies of MT-1 MREd cloned upstream of a minimal promoter (pMT-MREx2) were readily activated by Cd2+ , whereas a construct containing two (data not shown) or four tandem copies of the HO-1 MREa (pMHOMREx4) remained inactive (Fig. 5C). This result suggests that the core portion of the consensus MRE, while necessary (51), is insufficient for transcriptional activation and that adjacent sequences, possibly the GC-rich flanking region typically observed in the functional elements of the MT genes, are important in the induction process. Taken together these experiments clearly indicate that the putative rat and mouse HO-1 MREs are not functional metal-responsive elements.

Induction of the Mouse HO-1 Gene Constructs in C6 Cells-
Heavy metals are unable to activate pMHOlcat possibly because the cis-acting element(s) required for induction are not present within the 5"flanking region used to construct the fusion gene; they may reside further upstream, in the body of the gene and/or even in the 3"flanking region. Alternatively, or in addition, chromatin structure may influence the activation process and optimal transcriptional activation may require chromosomal integration of the fusion gene.
In light of these possibilities, non-reporter gene expression plasmids containing various segments of the 5"flanking region cloned upstream of the structural region of the mouse HO-1 gene were constructed. These constructs were then stably integrated into rat C6 cells, and the expression of the transfected mouse HO-1 genes and the endogenous rat HO-1 gene were quantified by RNase protection assays. The probe used for these assays protects 151 bases of correctly initiated mouse mid, 1 p g f pRSVluc, and an appropriate amount of pBSSKfor 4 h. The cells were exposed to 10% glycerol for 1 min and incubated in complete medium for 40 h. CAT activity in each extract was normalized to luciferase activity in the same extract. Each data point represents the average value from five to seven independent transfections.
HO-1 mRNA and 211 bases of unspliced, pre-mFtNA. In addition, this probe protects a series of fragments, approximately 45-49 bases in length, of the rat HO-1 mRNA. The data presented in Figures 6 and 7 clearly demonstrate that distal 5' sequences are required for induction of the mouse HO-1 gene. The construct containing 5"flanking sequences from -3.5 to -15 kb, pMH04B11, is activated approximately 20-fold by both Cd2+ (20.2 2 2.4, n = 4) or heme (18.1 2 2.3, n = 4) (Fig. 6, lunes d-f). The fold induction by these agents is similar to that observed for the endogenous rat HO-1 gene: 22.5 * 6.6, n = 12 (Cd2+) and 21.0 2 7.0, n = 12 (heme). A notable characteristic of HO-1 gene regulation is that on-going protein synthesis is required for induction by various agents (37) and both the basal and inducible expression of pMH04B11 as well as of the endogenous rat HO-1 gene are almost completely abrogated by cycloheximide (Fig. 6, lunes g-i).
Deletion analysis (Fig. 7A) has localized the Cd2+-and hemeresponsive elements to between positions -3.5 and -12.5 kb (pMH04B9, lunes u-c) of the mouse HO-1 gene as constructs containing 1287 bp (pMH04, lunes g-i) or 3.5 kb (pMH04B1, lunes d-f) of the 5"flanking region are not inducible by either agent. We have also examined the level of induction of transiently transfected mouse HO-1 genes. Transiently transfected pMH04B11 (Fig. 7B, lunes u-c) and pMH04B9 (lunes d-f) are only minimally induced by either Cd2+ (average from both constructs = 3.9-fold, n = 6) or heme (2.7-fold, n = 6) compared to the 20-fold activation observed with chromosomally integrated pH04B11 and pMH04B9. No consistent induction is observed with transiently transfected pMH04 or pMH04B1.
Induction of HO-1 /CAT Fusion Genes in Hepu Cells--To determine if the distal 5' region can activate a heterologous gene, the inducer-responsive 9-kbp fragment was cloned upstream of the HO-1 5'-proximal sequences in the plasmid pMHOlcat to generate the construct pMHO9cat. CAT activity is increased 11.0-fold in response to CdClz (10 p~) and 11.9-fold by heme (10 p~) in Hepa cells stably transfected with pMHO9cat whereas the parent plasmid, pMHOlcat, is unresponsive to either agent (Table 11). Previous studies in our laboratory have identified a 268-bp fragment, SX2, located approximately 4 kbp upstream of the transcription initiation site, that functions as a basal level enhancer and mediates TPA induction of the mouse HO-1 gene (10). The data in Table I1 demonstrates that the SX2 fragment also functions as a heme-and Cd2+-dependent enhancer in stably transfected Hepa cells but not in transiently transfected cells. The basal activity of stably integrated pMHO9cat is considerably greater than that of pMHOlcat+SX2 in Hepa cells. This variation may partly be a consequence of differences in copy number and/or integration sites. We have, however, identified a fragment located between positions -9 and -10 kb that functions as a potent basal level Full repression function may also require distal 5' sequences as the basal expression of pMH04, which would contain the putative downstream repressor element, is greater than that of pMH04B9 or pMH04B11. Hepa cells may contain a second cell-specific repressor that interacts with sequences in the 5'flanking region causing repression of pMHOlcat in Hepa cells but not in C6 cells. The cell-specific repressor hypothesis, although quite complicated, would explain not only the differences between the expression patterns of transfected constructs in Hepa and C6 cells presented here but also the following result from preliminary studies: deletion of most of the mouse HO-1 gene proximal promoter elements (to position -33, see Fig. 41, which would presumably obviate the need for the downstream repressor element, allows SX2 enhancer-dependent, inducible activation of HO-l/CAT fusion genes in C6 cells. Chromosomal integration of the fusion constructs is required for this activation. w e l l et ul. (52) recently reported that a human HO-1 promoter (-1416 to +24)/CAT fusion gene, a construct analogous to pMHOlcat, is not induced by a variety of agents after transient transfection into HeLa cells. Interestingly, these investigators did observe a highly variable, but significant (2-&fold) induction of 5'-deletion mutants of the parent construct containing 283 or 121 bp of the proximal promoter. This induction, however, was strictly dependent on the presence of the SV40 enhancer, leading to the suggestion that distal inducible enhancer elements in addition to the proximal promoter sequences are required for optimal induction of the human HO-1 gene. Our results with the SX2 enhancer fragment are consistent with this conclusion. Interestingly, the SX2 fragment, like the SV40 enhancer, contains high affinity binding sites for both the AP-1 (10) and CEBP (data not shown) family of transcription factors. The role of these and other putative cis-acting elements in transcriptional activation of the mouse HO-1 gene by various inducers is currently under investigation.
Nascimento and colleagues (53) have also observed that ul- RNase T1 (2 pg/ml), the digestion products were electrophoresed on a 8% polyacrylamide, 8 M urea gel, and the gel was exposed to x-ray film for 4 h. The mature mouse HO-1 mRNA (rnHO), the mouse HO-1 unspliced, pre-mRNA (pre-rnHO), and the rat HO-1 mRNA ( r H 0 ) are marked. The size, in bases, of the marker fragments (HaeIII digestion products of plasmid Bluescript IISK-) is indicated. agents. Indeed, data obtained in our laboratory indicate that the inducible enhancer activity of the SX2 fragment in Hepa cells is dependent on an intact mouse HO-1 proximal promoter. One important region maps between positions -33 and -63 bp, which coincidently contains the c-MycAJSF binding sites. 3 The most striking feature of the expression of pMHO9cat and pMHOlcat+SX2 in Hepa cells is the complete inability of heme or cadmium to activate these fusion genes in transiently transfected cells. This dependence of transcription activation on the chromosomal environment is even more dramatic than that observed with glucocorticoid induction of the mouse mammary tumor virus (MMTV) long terminal repeat where the degree of induction of the stable MMTV promoter is 34-fold greater than that from an analogous transient construct (54). In the latter system (and in others (reviewed in Refs. 55 and 56)), the nucleosome structure is proposed to repress the activity of the MMTV long terminal repeat largely by exclusion of specific transcription factors from their cognate binding sites. One function proposed for the glucocorticoid receptor is the remodeling of the chromatin structure to permit access of transcriptional activators, such as NF-1, to the MMTV promoter. We suggest that the induction of the stably integrated HO-1 genes by heme and heavy metals may also occur by a similar antirepression mechanism. Transiently introduced constructs would be predicted to exist predominately in a de-repressed state, thus limiting the potential for induction.
Equally consistent with the present data, as mentioned previously, is a mechanism in which the expression of the HO-1 gene is repressed by a sequence-specific regulatory proteinb) with the added provision that these factors are present in limited quantity. Transcriptional activation by inducing agents would proceed, in part, by inactivation of repression. Introduc-

Inducibility of transiently transfected and stably integrated mouse HO-1 /CAT fusion genes
C6 or Hepa cells were plated (5 x lo5 cells/60-mm plate) 16 h prior to transfection. CaP0,-mediated transfection (5 pg of CAT plasmid + 1 pg of pRSVLuc) was carried out for 6 h, and the cells were shocked with 10% glycerol for 1 min and incubated in complete medium for 16 h. Stably transfected cells were plated (1 x 106/60-mm plate) 16 h prior to treatment. Cells were treated with vehicle, 10 p~ CdC12, or 10 p~ heme in Hepes-buffered (pH 7.2), serum-free medium for 3 h and subsequently incubated in complete medium for 5 h. Extracts equivalent to 3 pg of protein ("Stable"), or aliquots containing equivalent amounts of luciferase activity (Transient") were used to measure CAT activity. The reaction was camed out for 2 h (Stable) or 16 h (Transient). Each value represents the average of two independent experiments or the average f S.D. of four to five experiments. The fold induction is given in parentheses and was calculated after subtracting the background CAT activity observed with the promoterless CAT vector, pSKcat. tion of a large number of DNA molecules per cell, as generally attained during transient transfection, would readily titrate the repressor protein, permitting unregulated expression of most of the transfected genes. Because only a small fraction of the internalized DNA is integrated into the cellular chromosome, the copy number of stably transfected mouse HO-1 gene constructs would be sufficiently small to allow repression and subsequent activation by the inducing agents. Since a small but consistent induction of transiently transfected pMH04B11 and pMH04B9 is obsemed, anti-repression mechanisms are unlikely to fully account for induction of the HO-1 gene by heme and heavy metals.

Relative CAT activity
If, as proposed, transcriptional activation of the HO-1 gene occurs largely by de-repression, then the regulation of the HO-1 gene by heavy metals differs markedly from that of MT genes, where the MREs function as positive regulatory elements. This mechanistic difference was predicted based on our previous results (37) demonstrating a striking divergence between the effects of cycloheximide on Cd2+-mediated stimulation of HO-1 and MT-1 gene expression in mouse Hepa cells: abrogation of HO-1 induction uersus superinduction of MT-1 expression. In support of this idea, we note that the SX2 fragment contains no sequences similar to the consensus MRE and the fact that this enhancer mediates transcriptional activation of the mouse HO-1 gene by a diverse group of agents diminishes the possibility of the existence of inducer-specific, cis-acting elements, such as MREs. Rather, we propose, in agreement with other investigators (2, 57), that the action of most of the inducers of HO-1 gene expression converges at a common second messenger(s) and/or generates a common state (e.g. oxidative stress) which serve as the proximate signals for activation (or deactivation) of the appropriate transcription factors.