ABSTRACT
Erythropoietin (EPO) plays a key role in erythropoiesis and is expressed
predominantly in the fetal liver and in the adult kidney. The
EPO
gene is up-regulated at the transcriptional level under hypoxic/anemic conditions. We
studied the role of the 5
'
- and 3
'
-flanking sequences of the mouse
EPO
gene in its tissue-specific and hypoxia-induced expression by developing transgenic mouse lines carrying
chimeric
EPO
-
lacZ
gene constructs. Transgenic mice carrying a 6.5 kb segment of the 5
'
-sequence and most of the
EPO
gene in which
lacZ
was substituted for exon 1 (5
'
-
lacZ
-
EPO
) demonstrated induction of
lacZ
expression following hypoxia/anemia induction in both the liver and kidney of
adult mice. However, transgenic mice carrying the above construct along with
the 1.2 kb 3
'
-flanking sequence (5
'
-
lacZ
-
EPO
-3
'
) showed a high level of
lacZ
expression following hypoxia/anemia induction in adult kidney but not in adult
liver. With the aim of further understanding the role of the 3
'
-flanking sequence in tissue-specific expression of the
EPO
gene, we studied the interactions of protein factors with this 1.2 kb 3
'
region and demonstrated that multiple sets of protein factors interact tissue
specifically with a 10 bp sequence, TCAAAGATGG, located downstream of the
previously characterized 3'
hypoxia-responsive enhancer element.
Erythropoietin (EPO) is a glycoprotein hormone and the primary physiological
mediator of erythropoiesis (
1
-
3
). It binds to specific cell surface receptors on erythroid progenitor cells and
stimulates their proliferation, promotes differentiation and prevents apoptosis
(
4
,
5
). In a recent report, EPO and EPO receptor (EPOR) knockout mice exhibited
reduced primitive erythropoiesis and died around embryonic day 13, owing to failure of definitive
fetal liver erythropoiesis. Both of these knockout mice exhibited identical
phenotypes, indicating that EPO and the EPOR are crucial for definitive
erythropoiesis
in vivo
and that no other ligand or receptors can replace them (
6
). In most species the liver is the major site of EPO synthesis during fetal
life, whereas late in gestation the kidney becomes the predominant site of EPO
production and remains so throughout life (
2
,
7
,
8
).
The enhancer sequences responsible for hypoxia induction of the
EPO
gene have been localized in both the 5'- and 3'-flanking sequence of the
EPO
gene (
9
-
13
). Hypoxia-inducible protein factors have been demonstrated to interact with these
enhancer sequences (
13
-
15
). Hypoxia-inducible factor 1 (HIF1) was recently cloned and characterized as a basic
helix-loop-helix-PAS heterodimer (
16
). The possible roles of hepatic nuclear factor 4 (HNF-4) and the COUP family of proteins in tissue-specific and hypoxia-inducible expression of the
EPO
gene have also been demonstrated (
17
). Maxwell
et al.
(
12
), using transgenic mice, showed that 9.0 kb 5'- and 3.5 kb 3'-flanking sequences contain the necessary regulatory
elements for mouse
EPO
gene expression. Cooperative interaction between the 5' promoter and the 3' enhancer elements for hypoxia induction has been described before
(
14
,
18
), but the 5' and 3' regulatory elements and their cooperative interactions in tissue-specific regulation of
EPO
gene expression are not well defined.
In this study we compared the pattern of expression of the
lacZ
reporter gene under normoxic and hypoxic conditions in the liver and kidney of
transgenic mice carrying: the
lacZ
reporter linked to the 6.5 kb 5'-flanking sequence, the entire first exon and 200 bp of the 5' portion of the first intron (5'-
lacZ
); the 6.5 kb 5'-flanking sequence and most (except exon 1 and 200 bp of intron 1)
of the
EPO
gene (5'-
lacZ
-
EPO
); the 6.5 kb 5'-flanking sequence plus most of the
EPO
gene and the 1.2 kb 3'-flanking sequence (5'-
lacZ
-
EPO
-3') of the mouse
EPO
gene. We provide evidence from these transgenic mouse lines that the 1.2 kb 3'-flanking region may contain a regulatory element for suppressing
EPO
gene induction in response to hypoxia/anemia in the adult liver. Also, evidence
from our DNA-protein interaction studies show that sets of protein factors interact in
a tissue- and development-specific pattern to a 10 bp sequence (TCAAAGATGG) located ~360 bp downstream of the previously described hypoxia-responsive enhancer element (
11
) within this 1.2 kb 3'-flanking sequence.
The plasmid pSG5/EPO has been described before and was constructed by inserting
the
Eco
RI fragment of the phage clone 60a (
19
-
21
). In the 5'-
lacZ
construct the
lacZ
gene was placed downstream of the 7.0 kb
Eco
RI-
Bam
HI fragment and has been described before (
19
). The 6.2 kb
Eco
RI-
Nar
I and 3.7 kb
Bam
HI-
Eco
RI fragments from the pSG5/EPO plasmid were inserted upstream and downstream
respectively of the [beta]-
geo
gene into the pSA[beta]geo plasmid (a generous gift from Dr Philippe Soriano, Baylor College of
Medicine, Houston, TX) (
22
) to construct the 5'-
lacZ
-
EPO
plasmid. The
Nar
I and
Bam
HI sites are located 548 bp upstream and 218 bp downstream of the transcription
initiation site ATG respectively (
20
). The
Nar
I-
Bam
HI fragment, which harbors the translational initiation site of the
EPO
gene, is missing from both the 5'-
lacZ
-
EPO
and 5'-
lacZ
-
EPO
-3' constructs and uses the translational initiation site of the [beta]-
geo
gene. 5'-
lacZ
-
Epo
-3' (Fig.
1
) was constructed by inserting the 1.2 kb
Eco
RI-
Bam
HI fragment from phage 18.c at the 3'-end of the 5'-
lacZ
-
EPO
construct (
23
).
The 5'-
lacZ
, 5'-
lacZ
-
EPO
and 5'-
lacZ
-
EPO
-3' constructs were digested with
Not
I and the respective
EPO
-
lacZ
fragments were purified from the agarose gel using a Gene Clean II kit (Bio 101
Inc., La Jolla, CA). Fertilized eggs were harvested from Swiss Webster
Rockefeller (SWR) female mice after superovulation and mating with SJL mice. We
then microinjected 2-5 pl of a 30 ng/[mu]l solution of
EPO
-
lacZ
DNA fragments into the male pronucleus, as described previously (
24
). The injected eggs were incubated in Whitten's medium at 37oC and transferred on the same day into the oviduct of a CD-1 pseudopregnant foster mother. At least four transgenic mice lines
of each construct were generated and three lines were analyzed.
To identify the transgenic mice, tail DNA from 3-4-week-old pups was digested with
Hin
dIII and analyzed by Southern blot analysis using a 1 kb
Bam
HI-
Hin
cII fragment as a probe (indicated by a solid bar underneath the
EPO
gene map in Fig.
1
).
Mice were made hypoxic and anemic as described previously (
19
). Briefly, for hypoxia, mice were exposed to CO
2
for 3-4 min every 24 h for 5 days and the mice killed 10 min following the last
CO
2
exposure. Anemia was induced by first anesthetizing the mice with a mixture of
ketamine hydrochloride and xylazine and then withdrawing 0.03 ml blood via
cardiac puncture. Mice were bled again 12 h later and killed 6 h following the
third bleeding.
Liver, kidney, heart, lung, spleen, muscle, brain and intestine of transgenic
mice were stained with X-gal, as previously described (
19
).
Kidney extract (100 [mu]g) or liver extract (1000 [mu]g) were incubated with the ONPG (Sigma, St Louis, MO) substrate at 37oC for 16 h and the absorbance was measured at 420 nm as described elsewhere (
19
,
25
). Protein extracts of normal kidney and liver and bovine serum albumin (BSA)
were used as negative controls.
Adult liver, adult kidney and fetal liver nuclear extracts were prepared as
described previously (
26
,
27
). EMSA was done as described (
27
,
28
) in 25 [mu]l binding mixture containing a 5'-
32
P-end-labeled DNA fragment, 5 [mu]g poly(dI[middot]dC) (Boehringer Mannheim, Germany) and 5 [mu]g nuclear extract in 10 mM Tris-HCl, pH 7.1, 0.1 mM EDTA, 10 mM 2-mercaptoethanol, 0.1% Triton X-100, 5% glycerol, 80 mM NaCl and
3 mM MgCl
2
. This mixture was incubated at 4oC for 30 min and then loaded onto a 4% polyacrylamide gel (acrylamide:bis
30:1) in 0.5* TBE buffer. For competition experiments, 100 ng unlabeled oligoduplexes were added at the beginning of the binding reaction (
27
).
Binding reactions (50 [mu]l) were set up using the 5'-
32
P-end-labeled DNA fragment and nuclear extract as described for EMSA.
After incubation, the mixture was digested with 4 [mu]l 0.2 U/[mu]l DNase I for 3 min at 20oC. The reaction was stopped by addition of 4 [mu]l 50 mM EDTA. The mixture was then electrophoresed on a 4%
polyacrylamide gel as described for EMSA. The gel was electroblotted onto a
NA45 DEAE membrane (Schleicher & Schuell Inc., Keene, NH). Free DNA (free probe) and protein-DNA complexes (gel retarded bands) were cut out from the membrane and
eluted by incubating in a buffer containing 20 mM Tris, pH 8.0, 1 mM EDTA and
1.5 M NaCl at 65oC for 15 min. The eluted DNA was precipitated with 2 vol. ethanol and
loaded onto a sequencing gel. Equal amounts of radioactivity of free DNA and
DNA-protein complexes were loaded onto an 8% polyacrylamide sequencing gel.
The G, A, T and C sequencing reactions of the DNA fragment +567/+698 were
generated using a 5'-most primer (+567/+589) and AmpliCycle sequencing kit (Perkin Elmer-Cetus, Branchburg, NJ) and loaded alongside.
The
32
P-end-labeled +588/+616 oligoduplex and nuclear extracts from adult
kidney, adult liver or fetal liver were separately incubated and
electrophoresed as described for EMSA. Following electrophoresis, the wet gel
was exposed to 254 nm UV light for 30 min. After UV cross-linking the gel was exposed to X-ray film. The regions corresponding to the protein-DNA complexes (F1-F4) were cut out and sliced into small pieces and
incubated with 1* SDS-PAGE sample buffer (10% glycerol, 0.25% SDS, 62.5 mM Tris, pH 6.8, 2.5% [beta]-mercaptoethanol) containing 0.2 M NaCl at 37oC for 2 h and at 95oC for 2 min and filtered through a glass
wool column. The filtrate was counted for radioactivity and equal amounts of
each filtrate were loaded onto a 14% SDS-PAGE gel along with the protein molecular weight standards (Sigma, St
Louis, MO).
Figure
1
is a schematic representation of the
EPO
-
lacZ
gene constructs used for generating transgenic mice. At least three to four
transgenic lines were generated for each of the constructs and analyzed for
integration of the
EPO
-
lacZ
fragment by Southern blot analysis (data not shown). Table
1
shows the X-gal staining followed by visual examination of different tissues of
transgenic mice carrying the 5'-
lacZ
, 5'-
lacZ
-
EPO
and 5'-
lacZ
-
EPO
-3' constructs. X-Gal staining was detected only in kidneys, livers and
intestine of transgenic mice, but not in any other tissues (heart, lung, brain,
spleen and muscle) examined. However, X-gal staining was also detected in the intestine of normal non-transgenic mice, but not in any other normal tissues, suggesting
that X-gal staining of intestine is non-specific and not due to transgene expression.
The lack of induction of
lacZ
expression in response to hypoxia/anemia in the adult liver of transgenic mice
carrying the
EPO
-
lacZ
construct that contains the 1.2 kb 3'-flanking sequence suggests that this region appears to harbor a
regulatory element(s) that, through interactions with the liver-specific protein factor(s), may suppress induction of
EPO
gene expression. To test this possibility, we studied the 3'-flanking sequence downstream of the hypoxia-responsive enhancer element described earlier (
11
) for protein factor(s) interactions using nuclear extracts of adult liver,
fetal liver and adult kidney on an electrophoretic mobility shift assay (EMSA).
We identified a DNA segment, +567/+657, in the 1.2 kb 3'-flanking sequence [numbering with reference to the poly(A) site as
+1] that interacts with multiple sets of protein factors in a tissue-specific fashion. Nuclear extracts of adult liver (F1 and F2), fetal liver (F2-F4), adult kidney (F1-F3) and adult brain (F1) formed different sets of DNA-protein complexes (Fig.
4
). The F1 complex is formed at a relatively higher level than the F2 complex
with the adult liver nuclear extract. Fetal liver nuclear extract does not form
the F1 complex, but does form the F2-F4 complexes at a relatively higher level when compared with the adult
liver nuclear extract. The F1 complex in adult kidney nuclear extract is formed
at a relatively lower level than the F2 and F3 complexes.
These results suggest that in the adult liver in which
EPO
induction in response to hypoxia is relatively low, the F1 complex is formed at
a relatively higher level when compared with the F2 complex. However, in the
adult kidney and fetal liver, where
EPO
is induced in response to hypoxia, the F1 complex is formed at a relatively
lower level or not at all when compared with the F2-F4 complexes. Non-EPO-producing tissues, such as adult brain and adult heart,
formed only the F1 complex (data not shown).
To precisely define the sequences involved in the protein factor interactions,
four oligoduplexes spanning the +567/+657 region were synthesized and included
in the binding reaction for competition with the +567/+657 probe in EMSA.
Unlabeled oligoduplexes made within the +567/+657 DNA segment (+567/+589,
+588/+616, +610/+635 and +627/+653) were added in the binding reaction to check
their capacities to compete with the probes for protein factor interactions
(Fig.
5
A). EMSA using oligoduplex +610/+635 is not shown. All the electrophoretic
mobility shifted bands were competed off by the oligoduplex +588/+616, but not
by any of the other three oligoduplexes, suggesting that the sequence within
the oligoduplex +588/+616 is involved in the DNA-protein interactions.
Analysis of
lacZ
expression in transgenic mice lines carrying
EPO
-
lacZ
constructs and DNA-protein interaction studies of the 1.2 kb 3'-flanking sequence suggest that: (i) the 6.5 kb 5' sequence, the body of the
EPO
gene and 1.2 kb 3'-flanking region contain sufficient
cis
-acting sequences for tissue-specific expression of the
EPO
gene; (ii) the 1.2 kb 3'-sequence appears to possess a silencer sequence capable of
suppressing hypoxia induction of
EPO
in the adult liver; (iii) a 10 bp sequence, TCAAAGATGG, located downstream of
the previously characterized hypoxia-responsive enhancer element (
11
) and within the 1.2 kb 3' region interacts with different sets of protein factors from adult
liver, adult kidney and fetal liver nuclear extracts.
Maxwell
et al.
(
12
), using a transgenic mouse model, reported earlier that
cis
-acting sequences that regulate renal
EPO
lie within 1.5-9 kb 5' and 3.5 kb 3' of the mouse
EPO
gene. Semenza and colleagues (
10
,
29
,
30
) studied the
cis
-regulatory elements in the 5'- and 3'-flanking sequences of the human
EPO
gene by generating transgenic mice carrying 0.4, 6.0 or 14 kb of the 5'-flanking sequences and 0.7 kb of the 3'-flanking sequence. These transgenic mouse models
suggested that the
cis
-acting sequences involved in hepatic and renal expression are different.
The expression of
EPO
in the kidney appeared to be dependent on the sequences between 6.0 and 14.0 kb
5' of the human gene. A recent report by Madan
et al.
(
31
) also suggested that for appropriate tissue-specific expression of the human
EPO
gene, at least a 9.5 kb 5'-flanking sequence and an 8.5 kb 3'-flanking sequence are required.
Our transgenic mice, which carry the 6.5 kb 5'-flanking sequence, the body of the
EPO
gene and the 1.2 kb 3'-flanking sequence, express
lacZ
at a higher level following hypoxia/anemia induction in the adult kidney than
in the adult liver. However,
lacZ
is induced following hypoxia/anemia induction in both the adult kidney and
liver of transgenic mice that carry the construct without the 1.2 kb 3'-flanking region (5'-
lacZ
-
EPO
). These results suggest that the 1.2 kb 3'-flanking region of the mouse
EPO
gene appears to contain a sequence element that may be involved in suppression
of
EPO
gene induction in the adult liver in response to hypoxia/anemia.
Transgenic mice generated with additional 5'- and 3'-flanking sequences (9.0 kb 5'-flanking, the body of the gene and 3.5 kb
3'-flanking sequence) of the mouse
EPO
gene showed induction in liver following anemia induction (
12
). It is possible that sequence downstream of the 1.2 kb 3' mouse
EPO
sequence may harbor an additional enhancer element(s), which through
interaction of protein factors induced under anemic conditions can act in
concert with the 3'-proximal hypoxia-responsive enhancer element (
11
) to overcome suppression of the silencer sequence present in the 1.2 kb 3' region.
lacZ
expression in response to hypoxia and bleeding is 1.4- to 1.6-fold, 1.2- to 1.7-fold and 2.2- to 3.9-fold when compared with the normoxic level
of
lacZ
expression in the kidneys of transgenic mice carrying 5'-
lacZ
, 5'-
lacZ
-
EPO
and 5'-
lacZ
-
EPO
-3' constructs respectively. The fold activation of
lacZ
in response to hypoxia/anemia is relatively lower than that reported for
induction of endogenous
EPO
expression and intact
EPO
transgene expression (50- to 100-fold) in response to hypoxia (
32
). However, a recent report indicated that transcriptional induction of
EPO
mRNA accounts for only a 10-fold increase and post-transcriptional stabilization of
EPO
mRNA appears to be the major mechanism of hypoxic induction of EPO (
32
). It is therefore possible that the hybrid
lacZ
-
EPO
mRNA is not as stable as that of
EPO
mRNA and could account for the differences in hypoxia induction. This
discrepancy in induction could also be due to the presence of an additional
hypoxia-responsive enhancer element(s) outside the boundary of the 6.5 kb 5'- and 1.2 kb 3'-flanking sequences of the mouse
EPO
gene used for making the
EPO
-
lacZ
constructs.
By analyzing the 1.2 kb 3'-flanking sequence, we identified a DNA segment, +567/+657
[numbering with reference to the poly(A) site as +1], downstream of the
previously described hypoxia-responsive enhancer element (
11
) that forms multiple sets of protein-DNA complexes in EMSA using nuclear extracts from adult liver (F1 and F2
complexes), adult kidney (F1-F3 complexes), fetal liver (F2-F4 complexes) and adult brain (F1 complex only). The capacity of
the 10 bp
EPO
sequence to bind multiple protein factors tissue and development specifically
suggests that the relative abundance and competition of these protein factors
for binding the sequence element may play a role in
EPO
gene expression. The 52 and 56 kDa protein factors that form the F1 complex may
be responsible for silencing the
EPO
gene in the adult liver. The 36, 34 and 33 kDa proteins that form the F2-F4 complexes respectively may compete with the 52 and 56 kDa proteins and
may prevent silencing of
EPO
gene expression in the adult kidney and fetal liver. The F3 (34 kDa) and F4 (33
kDa) complexes may act as enhancers or may act by preventing the formation of
the F1 silencer complex.
The role of silencer elements has also been documented in other genes: in a
tissue-specific position effect on alcohol dehydrogenase expression in
Drosophila melanogaster
(
33
); in repression of neuronal gene transcription in non-neuronal cells through interaction of the neuron-restrictive silencer factor (NRSF) with the neuron-restrictive silencer element (NRSE) (
34
); in the role of multiple silencer elements in the regulation of the chicken
vimentin gene (
35
). A silencer and an enhancer element have been reported in the 3'-flanking region of the chicken and duck [alpha]-globin gene, similarly to the
EPO
gene (
36
). We did not find any homology of the 10 bp sequence element (TCAAAGATGG) to
any of the silencer elements reported in other genes (
33
-
36
).
The hepatoma cell lines HepG2 and Hep3B were shown to regulate EPO production in
culture in response to oxygen tension and have been successfully used in
defining the hypoxia-responsive enhancer elements (
11
). However, appropriate kidney and liver cell models are not available to test
the
cis
-element responsible for tissue- and development-specific expression of the
EPO
gene. We plan to generate transgenic mice carrying
EPO
-
lacZ
gene constructs that contain different deletions in the 1.2 kb 3'-flanking sequence, including the 10 bp sequence (which interacts
with multiple sets of protein factors both tissue and development
specifically), and containing point mutations within the 10 bp sequence to
precisely define the role of this 3'-sequence in tissue- and development-specific regulation of
EPO
gene expression.
We thank Leslie Wildrick for editing the manuscript and Garland Yee for his help
in synthesizing the oligodeoxynucleotides. This work was supported in part by
NIH grant HL21676 to EG. The animal work was supported in part by a University
of Texas M.D.Anderson Cancer Center Core Grant from the National Cancer
Institute (CA 16672).
REFERENCES
Return