ABSTRACT
We have reported that an upstream stimulatory factor (USF) binding site is
functional in transcription of the heme oxygenase-1 gene. In this study, we examined the role of USF in the induced state.
By transient expression analyses with the chloramphenicol acetyltransferase
gene, we found that the USF binding site plays an important role in the
induction of rat heme oxygenase-1 by cadmium, but not by hemin. To elucidate the role of USF, we prepared
USF-rich nuclear extracts from control and cadmium-treated rat liver. On electrophoretic mobility shift assay using
control nuclear proteins, one slowly migrating band was detected, whereas using
nuclear proteins of cadmium-treated rat liver, two fast migrating bands were detected. The molecular masses of the two subunits of USF prepared from cadmium-treated rat liver were
~34 kDa as determined by UV cross-linking and subsequent SDS-PAGE, while the two subunits of native USF were 43 kDa and 44 kDa.
DNase I footprinting analysis revealed that both the nuclear proteins bound to
the same region including the USF binding site. We therefore suppose that cadmium causes some structural changes in the two proteins of USF
and that the altered USF participates in the effective initiation of
transcription of the rat heme oxygenase-1 gene.
Microsomal heme oxygenase catalyzes the oxidative degradation of heme to
biliverdin, carbon monoxide and iron, in cooperation with NADPH cytochrome P-450 reductase, which functions as an electron donor, and biliverdin is
subsequently reduced to bilirubin by biliverdin reductase in the soluble cell
fraction (
1
). Heme oxygenase has two isozymes, an inducible enzyme referred to as heme
oxygenase-1 and a constitutive enzyme named heme oxygenase-2 (
2
). Heme oxygenase-1, a 33 kDa protein (
3
,
4
), is mainly distributed in reticuloendothelial cell-rich tissues such as spleen and liver (
1
). In 1970, Tenhunen
et al.
(
5
) found that the activity of rat liver heme oxygenase-1 was increased by injection with hemin or hemoglobin. This is unique, because substrate-mediated induction in animals is quite rare. Since then there have been many
reports describing how a number of non-heme substances, such as heavy metals (
6
), some organic compounds (
7
,
8
), endotoxins (
9
), interleukins (
10
-
12
), UV light (
13
) and heat shock (
14
), significantly induce heme oxygenase-1. The physiological significance of the inducibility of heme oxygenase-1 has remained obscure. However, the most likely explanation is that
heme oxygenase-1 works as a defense mechanism against oxidative stress, because
biliverdin and bilirubin may function as potent scavengers of oxygen radicals (
15
). In addition, carbon monoxide is now suggested to have a function in neurotransmission and vascular tone regulation (
16
-
18
).
Previously, Sato
et al.
(
19
,
20
) reported that binding of the upstream stimulatory factor (USF) consisting of 44 kDa and 43 kDa proteins (
21
-
23
) to a USF binding site of the heme oxygenase-1 gene, located at -51 to -35 bp from the transcription initiation site, was essential
for basal expression of this enzyme; the USF binding site was first found in
the adenovirus-2 major late promoter gene (
24
-
26
), its central motif being CACGTG (
24
). Among the above-described inducers, we have paid especial attention to hemin and cadmium,
since hemin is a substrate and cadmium is one of the most potent inducers (
27
). So, in this study, we intended to clarify the role of USF in the induced
state and found that USF participated in the induction of heme oxygenase-1 by cadmium but not by hemin. Moreover, we found that cadmium caused an
alteration in USF. The present paper is concerned with an investigation of
these problems.
The following compounds were used: restriction endonucleases and T4
polynucleotide kinase from Toyobo; DNase I from Boehringer; [[gamma]-
32
P]ATP, [[alpha]-
32
P]dCTP and chloramphenicol D-threo-[dichloroacetyl-1,2-
14
C] from New England Nuclear; heparin-agarose from Sigma; poly(dI-dC)[middot]poly(dI-dC), Sephadex G-25 and a molecular mass determination kit from
Pharmacia. Rabbit polyclonal antibodies raised against a synthetic peptide
corresponding to amino acids 291-310 of the 43 kDa protein of human USF were from Santa Cruz
Biotechnology. These antibodies react with both the 43 kDa and 44 kDa proteins
of humans, mice and rats. Non-specific rabbit IgG was from Jackson Immuno Research Laboratories. The
synthetic oligonucleotides 5'-CGGCCACCACGTGACTCGAG-3' and 5'-CTCGAGTCACGTGGTGGCCG-3', which contain the USF
binding site, were products of Sawady Technology and were used as synthetic 20
bp DNA after being annealed. DNA fragments of the rat heme oxygenase-1 gene containing the USF binding site were prepared from plasmid SpRHO8 (
19
). For labeling the 5'-end of DNA fragments, [[gamma]-
32
P]ATP and T4 polynucleotide kinase were used.
The SV40 polyadenylation sequence of pSV2cat (
28
) excised with
Hin
cII and
Bam
HI was ligated into
Sma
I and
Bam
HI sites of pUC19, because the presence of a polyadenylation site just before the promoter region of the chloramphenicol acetyltransferase (CAT) gene was reported to decrease non-specific expression of CAT (
29
). Thus the obtained plasmid was cut with
Hin
dIII and
Pvu
II. The gap between the
Hin
dIII and
Pvu
II sites was filled in with a
Hin
dIII-
Bam
HI (blunt) fragment of pSV2cat, which contains the CAT gene, polyadenylation
site and small t intron from SV40. This plasmid, carrying no promoter sequence,
was named pUC00cat and was used as a negative control for the CAT assay.
pUC00cat was cut with
Sal
I, both ends were filled in and lastly it was digested with
Hin
dIII. The SV40 promoter region of pSV2cat cut out with
Pvu
II and
Hin
dIII and was ligated into
Sal
I (blunt) and
Hin
dIII sites of pUC00cat. This plasmid, carrying the whole sequence of the SV40
promoter, was referred to as pUCcat and was used as a positive control for the
CAT assay. The expression plasmids including the 5' upstream region of the rat heme oxygenase-1 gene were constructed as follows. SpRHO8 (
19
), which contains the -748 to +2116 fragment of rat heme oxygenase-1, was linearized at the
Bam
HI site (+373) and the deletion construct was generated with exonuclease III and
mung bean nuclease. Dideoxy sequencing showed that the deletion mutant
contained the region between -748 and +53. The deletion mutant was digested with
Nde
I and the
Nde
I-deletion fragment (-137 to +53) was filled in at both ends. The fragment was cloned
into
Hin
dIII (blunt) and
Bam
HI (blunt) sites of pUC00cat to yield pUCNcat. In the same way, a
Xho
I-deletion fragment (-40 to +53) was inserted into the same sites of pUC00cat to yield
pUCXcat. A
Eco
72I-
Xho
I fragment (-44 to -40) was dissected from pUCNcat, filled in and self-ligated to generate pUCEcat.
Rat C6 glioma cells were maintained in Dulbecco's modified Eagle's medium
containing 10% fetal calf serum under humidified air containing 5% CO
2
at 37oC. Glioma cells (1-2 * 10
6
per 9 cm plate) were transiently transfected with plasmids by the calcium
phosphate method (
30
) with slight modification. The DNA mixture was composed of 15 [mu]g expression plasmid and 5 [mu]g [beta]-galactosidase expression vector pCH110 (Pharmacia). After 6
h the cells were treated with 15% glycerol for 1 min, cultured in serum-containing medium for 40 h and then incubated in serum-free medium containing the respective inducers of heme oxygenase-1 except for hydrogen peroxide. Three hours later, the cells
were washed twice with phosphate-buffered saline and cultured in serum-containing medium for 4 h. Hydrogen peroxide treatment was carried
out in serum-containing medium for 1 h, with a recovery period of 6 h. The
concentrations of inducers used were: CdCl
2
, 100 [mu]M; CoCl
2
, 500 [mu]M; hemin, 20 [mu]M; hydrogen peroxide, 600 [mu]M. Cell extracts prepared by three cycles of freezing and thawing
followed by centrifugation were assayed for CAT (
30
) and [beta]-galactosidase (
30
) activities. The radioactivity of acetylated chloramphenicol on thin layer
plates was quantified with an imaging analyzer BAS 2000 (Fuji Film). The CAT activity was normalized to [beta]-galactosidase activity (
30
).
Rat glioma cells were cultivated in serum-free medium containing the respective inducers of heme oxygenase-1 except for hydrogen peroxide. After 1 h, the cells were washed
twice with phosphate-buffered saline, followed by cultivation in serum-containing medium for 3 h. Hydrogen peroxide treatment was carried
out in serum-containing medium for 1 h, with a recovery period of 3 h. Total RNA,
prepared by the method of Chirgwin
et al.
(
31
), was separated on a denaturing formaldehyde-agarose (1%) gel, transferred to a nylon membrane (Magnagraph), fixed to
the membrane and hybridized with
32
P-labeled probes. The probe used was an
Eco
RI-
Hin
dIII fragment (+88 to +970) of rat heme oxygenase-1 cDNA (
32
) labeled with [[alpha]-
32
P]dCTP by the random priming method (
33
). The same filter was also subjected to rehybridization with a
32
P-labeled rat glyceraldehyde-3-phosphate dehydrogenase cDNA probe (
34
). The filter was autoradiographed with X-ray film. The radioactivity was determined with the imaging analyzer and
an ATTO Densitograph (ATTO).
Wistar rats (~250-300 g) were treated i.p. with saline with or without CdCl
2
(14 mg/kg body wt) (
27
). Two hours later they were killed by decapitation and their liver was excised.
Blood was washed out with cold saline injected through the portal vein. Crude
nuclear extracts prepared by the method of Dignam
et al.
(
35
) were treated with an equal volume of saturated ammonium sulfate solution at pH
7.0 and stood on ice overnight. The precipitates collected by centrifugation
were dissolved in 20 mM HEPES-KOH buffer, pH 7.9, containing 0.1 M KCl, 0.2 mM EDTA, 1 mM
dithiothreitol and 15% glycerol and dialyzed against the same buffer. The clear
supernatant obtained by centrifugation was applied to a Sephadex G-25 column equilibrated with TM buffer (50 mM Tris-HCl, pH 7.9, 12.5 mM MgCl
2
, 1 mM EDTA, 1 mM dithiothreitol and 20% glycerol) containing 0.1 M KCl (
19
). The fraction containing proteins was then applied to a heparin-agarose column equilibrated with the same buffer as that used for
Sephadex G-25 chromatography. The column was washed with TM buffer containing 0.3 M
KCl and then proteins bound to the heparin-agarose were eluted with TM buffer containing 0.6 M KCl and the resulting
eluate was used as the USF fraction.
Reaction mixture containing 20 mM HEPES-KOH, pH 7.9, 40 mM KCl, 4% Ficoll, 6 mM MgCl
2
, labeled DNA fragments (10 fmol, 10 000 c.p.m.), 1.5 [mu]g poly(dI-dC)[middot]poly(dI-dC) and the USF fraction (5 [mu]g protein) in a final volume of 20 [mu]l was incubated for 20 min at 25oC. Then the mixture was loaded onto a 4%
polyacrylamide gel containing 50 mM Tris and 380 mM glycine (
36
) and subjected to electrophoresis for 60-90 min at 10 V/cm.
The
Nde
I-
Nco
I DNA fragment, end-labeled at the
Nde
I site (10-20 fmol, 30 000 c.p.m.) was incubated with the USF fraction (0-50 [mu]g protein) in 30 [mu]l 20 mM HEPES-KOH, pH 7.9, 50 mM KCl, 1 mM dithiothreitol, 2 mM
MgCl
2
, 5% glycerol and 5 [mu]g poly(dI-dC)[middot]poly(dI-dC) for 20 min at 25oC and partially digested with 5 [mu]l DNase I (0.04 U in 20 mM MgCl
2
and 10 mM CaCl
2
) for 1 min at 25oC. The digestion was stopped by addition of 75 [mu]l 12.5 mM EDTA and the mixture was extracted with phenol/CHCl
3
. Samples were then separated in an 8% polyacrylamide-8 M urea sequencing gel (
37
)
USF factions (10 [mu]g protein each) were incubated with the labeled synthetic 20 bp DNA (10-20 fmol, 30 000 c.p.m.) in 30 [mu]l 20 mM HEPES-KOH, pH 7.9, 40 mM KCl, 4% Ficoll, 6 mM MgCl
2
and 5 [mu]g poly(dI-dC)[middot]poly(dI-dC) for 20 min at 25oC and irradiated with a UV lamp (maximum emission
wavelength 254 nm, maximum intensity 730 [mu]W/cm
2
at a distance of 15 cm from the UV source) for 30 min (
38
,
39
). Then cross-linked samples were subjected to electrophoresis in a 12% SDS-polyacrylamide gel (
40
).
Recently, Nascimento
et al.
(
41
) reported that UVA irradiation, an inducer of heme oxygenase-1, caused modification of USF. Therefore, we next investigated if cadmium
caused a similar modification or not. The USF fraction prepared from control
rat liver was mixed with the end-labeled 49 bp
Pvu
II-
Pvu
II (-59 to -11) fragment containing the USF binding site and subjected to an
electrophoretic mobility shift assay. As shown in Figure
4
, a distinct band attributable to a DNA-protein complex was detected (lane 4) and on addition of a 500-fold molar excess of synthetic 20 bp DNA containing the core
sequence (CACGAG) of the USF binding site, the DNA-protein complex band disappeared completely (lane 5). These observations
indicate that the USF in control nuclei bound to the USF binding site of the
rat heme oxygenase-1 gene, consistent with the previous finding (
19
). The bands near the bottom of the gel were due to unbound DNA fragments,
because they were also observed on electrophoresis of labeled 49 bp DNA alone
(lane 1). In contrast, two distinct, faster migrating bands were observed after
incubating the DNA fragment with the USF fraction from cadmium-treated rat liver (lane 2). Interestingly, neither band was seen in the
presence of a 500-fold excess of unlabeled synthetic 20 bp DNA (lane 3). In a similar
experiment in which a control USF fraction treated exogenously with cadmium was
used, the two fast migrating bands did not appear and only one band
corresponding to the band in lane 4 was detected (data not shown). This
suggests that cadmium does not affect the USF directly.
For determination of the molecular masses of native and altered USF, mixtures of
the labeled synthetic 20 bp DNA (~12 kDa) and each USF fraction were subjected to UV irradiation and then
examined by SDS-PAGE (Fig.
7
A). On electrophoresis, the cross-linked sample of the DNA and the control USF fraction gave two broad
bands, B-2 and B-3 (lane 3), whereas without UV irradiation no bands were seen (data
not shown). The presence of a 500-fold excess of unlabeled 20 bp DNA at the time of UV irradiation
completely abolished formation of the complexes of labeled DNA and USF (lane
4). The molecular masses of B-2 and B-3 were estimated to be 99 and 55 kDa respectively, by comparison of
the migration distances of the centers of the broad bands with those of marker
proteins. By subtracting the 12 kDa of the synthetic oligonucleotide from these
observed values, we estimated that the molecular masses of the proteins in the
B-2 and B-3 bands were 87 and 43 kDa respectively. The USF fraction from
cadmium-treated rat liver gave only one, very broad band with a molecular mass of ~46 kDa (lane 1). A value of ~34 kDa for the molecular masses of the two proteins was
estimated by a similar calculation.
So far, the cadmium-responsive
cis
elements of the human and mouse heme oxygenase-1 gene have been reported by Takeda
et al.
(
45
) and Alam
et al.
(
46
-
49
) respectively. Takeda
et al.
identified the 10 bp
cis-
acting element of the human heme oxygenase-1 gene, located ~4 kbp upstream from the transcription initiation site, as responsible
for cadmium-mediated induction, but not for hemin-, cobalt- or zinc-mediated induction (
45
). On the other hand, Alam
et al.
(
47
) reported that activation of the mouse heme oxygenase-1 gene by 12-
O
-tetradecanoylphorbol-13-acetate, heme and cadmium is mediated by a 268 bp fragment located ~4 kbp upstream of the transcription initiation site (
46
-
48
). In addition, Alam
et al.
(
49
) found that a 161 bp fragment located ~10 kbp upstream of the transcription initiation site was a second
regulatory region for cadmium- and heme-mediated induction. Although the relation between these cadmium-responsive elements and USF is unknown, the present study demonstrates that the
cis
USF binding site plays an important role in cadmium-mediated induction of rat heme oxygenase-1.
USF, a helix-loop-helix transcription factor that binds specifically to the USF binding site, consists of 43 and 44 kDa proteins (
21
-
23
), although molecular masses of 33.5 kDa for the former (
50
) and 36.9 kDa for the latter (
22
) have been calculated from the nucleotide sequences of their genes. The present
study shows that cadmium administration causes structural change in USF
resulting in an increase in the mobilities of the two proteins in a SDS-polyacrylamide gel and the loss of their ability to dimerize, although
the actual alteration remains unclear. However, they were still able to bind to
the USF binding site, though with a lower affinity than native USF, as
discussed later. The fact that only the two fast migrating bands were detected
(Fig.
4
) suggests that native USF was not present in cadmium-treated rat liver and that each altered protein bound to the USF binding
site as a monomer, to form a DNA-altered 44 kDa protein or DNA-altered 43 kDa protein complex. We observed that treatment of both
the USF fractions with calf intestinal phosphatase or [lambda] phosphatase did not alter their electrophoretic mobilities from those
seen in Figure
4
(data not shown). This indicates that phosphorylation and dephosphorylation are
not involved in the suspected structural changes. In addition, we examined
whether cadmium increased the amount of USF in rat glioma cells by Northern
blot analysis, but we could find no differences between cadmium-treated and untreated cells (not shown), supporting our view that this
metal causes structural changes in USF.
On the other hand, Gregor
et al.
(
50
) reported the interesting finding that truncated human USF retaining the helix-loop-helix domain but lacking a part of the C-terminal leucine repeat region could bind to the USF binding
site but failed to dimerize. Their report raises another possibility, that a
protease that is activated by cadmium treatment may cut both the proteins at
positions near the C-terminal side within the leucine repeat region in such a way that the
resulting truncated peptides still react well with the antibodies. However,
this possibility may be unlikely, because the antibodies used here recognize a
C-terminal sequence consisting of only 20 amino acids and because the
difference between the migration distances on a SDS-polyacrylamide gel of native and altered USF correspond to ~8 kDa.
The present footprinting analysis showed that both the USF fractions protected
the same sequence, whereas USF modified by UV irradiation protected only the
upstream half of the region protected by native USF (
41
). The extent of protection by 50 [mu]g of USF fraction prepared from cadmium-treated rat liver was similar to that by 30 [mu]g of USF fraction of control rat liver (Fig.
6
). This suggests that native USF binds more tightly than altered USF, assuming that
equal amounts of USF are present in these USF fractions. This view was
consistent with the finding that on treatment with specific antibodies, the
band attributable to the DNA-native USF complex did not disappear completely, whereas the bands of the complex of DNA-altered USF did (Fig.
5
).
When the complex of labeled DNA and native USF was subjected to UV cross-linking, two cross-linked forms appeared (Fig.
7
). Dostatni
et al.
(
51
) reported that even when a complex of DNA and dimer proteins was cross-linked by UV irradiation, two types of complex, DNA-monomer and DNA-dimer, were formed. The former corresponds to linking of one
protein to one side of the palindrome and the latter to cross-linking of two proteins to the symmetrical site. Thus, as illustrated in
Figure
7
B, we assume that bands B-2, B-3 and B-1 correspond to a complex of DNA and heterodimer, a mixture of complexes of DNA and 44 kDa protein and DNA and 43 kDa protein and a mixture of the
complexes of DNA and the 34 kDa protein resulting from 44 kDa protein and DNA
and the 34 kDa protein resulting from 43 kDa protein respectively. These
observations, together with the results in Figure
4
, suggest that native USF binds to the USF binding site as a heterodimer and
support the view that USF binds to the USF binding site as a dimer (
21
,
22
,
50
), although there is a report that it binds to DNA as a monomer (
52
).
There have been some reports that USF functions in the basal expression of some
eukaryotic genes, such as metallothionein I (
53
,
54
), rat [gamma]-fibrinogen (
55
,
56
) and liver-specific rat pyruvate kinase (
57
,
58
). The present report is probably the first one describing USF working in the
induced state. Nascimento
et al.
(
41
) reported a similar alteration of USF in UVA-irradiated human skin fibroblasts, but they did not demonstrate its role
in expression of the heme oxygenase-1 gene. In general, loss of dimerization ability of a transcription factor
and decrease in its binding ability to DNA result in a decrease in its
activity. However, our studies suggest that the altered USF possessing these
properties is more functional than native USF. Hence, we suppose that the
altered form of USF would show greater cooperativity with RNA polymerase II and
basic transcription factors necessary for the initiation of transcription of
the rat heme oxygenase-1 gene.
This work was supported in part by Grants-in-Aid 03670137, 06680603 and 06780575 for Scientific Research from the
Ministry of Education, Science and Culture of Japan.
REFERENCES
Return