(Received for publication, August 13, 1996, and in revised form, January 10, 1997)
From the University of Maryland Cancer Center, Department of Microbiology & Immunology, Program in Oncology, Molecular and Cellular Biology Program, University of Maryland School of Medicine, Baltimore, MD 21201
We have previously demonstrated that
up-regulation of STAT1 protein by all-trans-retinoic acid
(RA) in interferon (IFN)-unresponsive cells permits growth inhibition
by IFNs. Here, we show that the promoter of STAT1 directly responds to
retinoic acid treatment. Sequence and functional analysis of the murine
STAT1 promoter have identified a direct repeat motif that serves as a
retinoic acid response element. Mutagenesis of this element resulted in a loss of response to RA. This element is activated by RA receptors ,
, and
. In vivo, RA receptor
and retinoid X
receptor
preferentially interacted with this element. Thus, these
data define a molecular basis for the synergy between IFNs and
retinoids in tumor growth inhibition.
Interferons (IFNs)1 regulate cellular
antiviral, antitumoral, and immunological responses via specific gene
induction (1). Ligand-bound IFN receptors induce the tyrosine
phosphorylation of cytoplasmic STAT (signal transducers and activators
of transcription) proteins employing Janus kinases (JAKs). Activated
STATs then migrate to the nucleus and bind to specific response
elements to induce the expression of IFN-stimulated genes (2).
IFNs-/
induce cellular genes via an IFN-stimulated response
element, which binds several transacting factors of which ISGF-3 is the primary regulator (2). This factor consists of a 48-kDa DNA binding
protein (ISGF3
) and three tyrosine phosphorylated proteins, STAT1
, STAT1
, and STAT2 (2). Two Janus kinases, tyk2 and JAK1,
are essential for the activation of STATs in response to IFN-
/
.
Ligand-stimulated IFN-
receptor, employing Janus kinases JAK1 and
JAK2, induces phosphorylation of STAT1 but not STAT2. STAT1 then
migrates to nucleus, binds to
-IFN-activated site, and induces gene
expression (2). Cell mutants lacking STAT1 and JAK1 fail to respond to
IFN-
/
and IFN-
(3, 4). Thus, JAK1 and STAT1 are common
signaling components for all IFNs.
All-trans-retinoic acid (RA), a potent biological response
modifying metabolite of vitamin A, induces cellular gene expression utilizing nuclear receptors that also act as transcription factors (5).
These receptors bind to retinoic acid response elements (RARE) to
activate gene transcription. Retinoic acid receptors (RAR) and retinoid
X receptors (RXR) are the two major mediators of retinoid actions (5).
Three major forms of RARs and RXRs, ,
, and
, and the
corresponding subtypes are known. Their expression is variable
depending on cell type, organ, and status of cell differentiation. The
preferred ligands for RAR and RXRs are RA and 9-cis-RA
respectively, although RA at high concentrations activates RXR (5).
RXRs form heteromeric complexes with RARs, bind to RAREs, and stimulate
specific gene transcription (5). Heteromeric complexes of RXRs formed
with other nuclear receptors such as those of thyroid hormone (TR),
vitamin D3, and peroxisome proliferator activator also
induce ligand-specific gene expression (6-10). Thus, RXRs are
important cofactors for nuclear receptor-mediated gene regulation.
Several studies have indicated that IFN and RA in combination produce additive or synergistic antitumoral effects in vivo and in vitro (11). It is not known how these two disparate ligands and their corresponding signal transduction systems cross-talk in the mediation of antitumoral activity. We have shown previously that these effects are in part mediated by an increase in the level of IFN-stimulated gene factors upon RA treatment, thus allowing their functional activation by IFN (12). In particular, STAT-1 protein is induced in IFN-resistant tumor cells treated with RA (13). In the present investigation, we have identified the mechanism by which RA up-regulates the expression of STAT1.
All cell lines were cultured in
media supplemented with charcoal-treated, dialyzed fetal bovine serum.
Murine embryonic tumor cell lines F9 and P19 and monkey kidney cell
line COS-7 were cultured in Dulbecco's modified Eagle's medium. Human
RAR, -
, and -
and mouse RXR
expression vectors were
provided by Pierre Chambon, IGBMC, France. Mouse RXR
expression
vector was described previously (8). Human thyroid hormone receptor
beta (TR
) regulated by RSV-LTR was provided by Edwards Park,
University of Tennessee. Rabbit anti-STAT1 antibodies were provided by
Chris Schindler, Columbia University. Mouse anti-JAK-1 antibody was
from Transduction Laboratories. Purified baculovirus-expressed RAR
and RXR
, mouse monoclonal antibody against RXR
(H2RIIBP), and
rabbit anti-RAR
antibody were gifts from Keiko Ozato, National
Institutes of Health. Rabbit anti RAR
, RAR
, and RXR
antibodies
were purchased from Santa Cruz Biotechnology. All these antibodies
cross-reacted with cognate proteins from mouse and human sources. Mouse
IFN-
was from Boehringer Mannheim. Purest preparations of
all-trans-retinoic acid (99%), 9-cis-retinoic
acid (high performance liquid chromatography pure) and 3,3
,5
triiodo-L-thyronine (T3 or thyroid hormone)
were purchased from Sigma and were reconstituted in
ethanol. The oligonucleotides used in this study were shown in
Table I. These were used for electrophoretic mobility
shift assays (EMSA) and cloning into TK-luciferase. Probes for EMSA
were prepared by filling in the ends with Klenow fragment in the
presence of [
-32P]dCTP or by end labeling with
[
-32P]ATP. For site-directed mutagenesis,
5
-GATAAATCAGGGTGATAA-3
(RM1) and
5
-TAGAGTTCAGTGTGATAA-3
(RM3) oligos were
used.
|
Northern blot, run off
transcription, and Western blot analyses were performed as described
elsewhere (12). Cells (2 × 105) were electroporated
with 5 µg of luciferase reporter construct, 0.5 µg of expression
vectors of retinoid receptors, and 2 µg of -actin-
-galactosidase plasmid, and luciferase assays were
performed (14).
-Galactosidase assays were performed to normalize
for variations in transfection efficiencies. Stable transfection of F9
cells with STAT1 cDNA, cloned in mammalian expression pCXN2, was
performed as described previously (13). This plasmid also carried a
G418 resistance marker for selection in mammalian cells. The resultant
drug-resistant colonies (~150) were pooled and used in the
experiments. COS-7 cell extracts, expressing individual retinoid
receptors for EMSA, were prepared as described (15).
An adult BALB/c mouse liver genomic library (Clontech) in EMBL3 phage
vector (5 × 106 plaque-forming units) was screened
with a 32P-labeled human STAT1 cDNA (16), and 6 clones
were identified. All these clones contained the same 18-kb insert as
analyzed by restriction digestion and Southern blotting. A 4.5-kb
XhoI fragment was detected when probed with a 270-bp
fragment representing the 5 end of the cDNA. This fragment was
cloned into pGL3-basic vector (Promega). pGL3-TK was constructed by
excising the TK promoter (109 bp) from pTK-CAT vector and cloning into
the pGL3-basic vector. Deletion and point mutations were constructed
(17) using a polymerase chain reaction-based kit (Stratagene) and a
reporter construct VKL-7 as template. All constructs were confirmed by
sequencing.
We have previously shown that
treatment of F9 embryonal carcinoma cells with RA enhances the ISG
transcription (12) due to an activation of ISGF-3 in dF9
(RA-differentiated) but not in F9 (undifferentiated) cells. To further
understand the basis for the failure of IFN-induced transcription, we
performed EMSAs to detect the activation of STAT1 by IFN- in F9 and
dF9 cells. As shown in Fig. 1, nuclear extracts from
IFN-
-treated dF9 but not F9 cells (compare lane 1 with
2) were able to form a complex with a
32P-labeled palindromic IFN-
response element (pIRE)
(18). Identity of this factor as STAT1 was established by a specific
antibody that neutralized the formation of the complex (lane
5). Similar activation of STAT1 was also observed in RA-treated
P19 (dP19) but not in undifferentiated P19 cells (Fig. 1, lanes
3 and 4). Mixing of F9 cell extracts with those of dF9
cells did not prevent the binding of STAT1 to pIRE (lane
6).
Since STAT1 was activated by IFN- after RA-treatment, we wanted to
further distinguish whether such activation was due to an increase in
the levels of signal transducing JAKs or transacting factors. To test
these possibilities, F9 cells were stably transfected with pCXN2
expression vector or the same vector that expressed the STAT1 cDNA.
The resultant cell transfectants were treated with IFN-
, and their
nuclear extracts were assayed for pIRE binding of STAT1 using EMSA
(Fig. 2A). Activation of STAT1 was not
detected in untreated cells (lanes 1 and 3). In
the F9 transfectants that carried the expression vector pCXN2 alone
(lanes 1 and 2), IFN-
failed to induce STAT1
binding to pIRE (lane 2). However, IFN-
robustly
activated STAT1 in cells transfected with the cDNA (Fig. 2A, lane 4). To further prove that the
transfected STAT1 was functional, luciferase reporter assays were
performed. As anticipated, no induction of luciferase activity was
detected in the F9 cells that stably expressed the pCXN2 vector alone
(Fig. 2B, bar 2). In STAT1 transfected cells
(bars 3 and 4), IFN-
readily induced the
expression of pIRE-luciferase reporter gene (Fig. 2B,
bar 4). Thus, F9 cells possessed the necessary receptors and
protein kinases for the activation of IFN-
-initiated JAK-STAT
pathway, except STAT1. Over expression of STAT1 or treatment with RA
restored IFN-
responses in these cells. Consistent with this, JAK1
levels were not affected by RA-treatment in an immunoprecipitation
assay (Fig. 2C). Western blot analyses (Fig. 2D)
revealed no detectable STAT1 protein in undifferentiated F9 cells.
However, it was strongly induced by RA-treatment (lanes 2 and 3). Thus, STAT1 appeared to be a target of
retinoid-mediated regulation.
RA Induces the Expression of STAT1 Gene
Since STAT1 protein
was induced by RA, we next determined whether such enhancement was due
to an induction of STAT1 gene expression. Northern blot analysis (Fig.
3A) did not reveal detectable STAT1 mRNA
(lanes 1 and 2) in untreated F9 cells or cells
that received ethanol (the solvent in which RA was prepared). Treatment
of cells with 1 and 10 µM RA strongly induced the
mRNAs of STAT1 and -
(lanes 3 and 4).
The probe detected both the mRNAs because they were derived from
the same gene (16). All the cells expressed similar levels of
-actin
mRNA irrespective of treatments. To examine whether the induction
of STAT1 by RA was due to de novo transcription, nuclear run
off transcription assays (Fig. 3B) were performed. No
detectable transcription of STAT1 was observed in untreated cells.
However, RA induced the transcription by 24 h, which increased
with prolonged treatment (48 h). All these cells expressed normal
levels of glyceraldehyde-3-phosphate dehydrogenase transcripts,
indicating that lack of STAT1 gene expression in F9 cells was not due
to a global transcriptional blockade. Further, RA did not induce the
expression of STAT2 mRNA (Fig. 3C). Induction of STAT1
mRNA by RA was observed in an IFN-unresponsive MCF-7 breast
carcinoma cell line and an acute promyelocytic leukemia cell line (data
not shown). Thus, STAT1 mRNA is up-regulated by RA in multiple cell
types.
Identification of a RARE in STAT1 Promoter
To examine whether
the STAT1 promoter was directly regulated by RA, a mouse genomic clone
was isolated using human STAT1 cDNA (16) as a probe. A 1.85-kb
HindIII fragment, containing 700 bp of upstream sequence,
the first exon and intron (1.15 kb), was cloned into pGL-3 basic
vector. The resultant reporter construct, VKL-2, responded to RA in
COS-7 cells upon cotransfection with RAR (Fig.
4A). Since COS-7 cells lacked these nuclear
receptors, the induction was due to the cotransfected receptor.
Deletion of a 965-bp sequence consisting of first exon and a
substantial portion of intron from VKL-2 (construct VKL-4) had no
effect on RA inducibility (Fig. 4B). Following treatment
with RA (1 µM), a strong up-regulation of luciferase
activity was noted in cells that were cotransfected with RAR
.
Furthermore, two other nuclear receptors, RXR
and TR
, along with
their ligands 9-cis-RA and T3 had no effect on
gene expression (Fig. 4B). Two other constructs, VKL-6 and
VKL-7, which contained up to
950- and
670-bp upstream elements of
STAT1 promoter, respectively, were also strongly induced by RA in these
cells (Fig. 4A). Thus, the cloned fragments contained necessary elements for a specific induction of STAT1 promoter by RA.
Primer extension analyses identified the start site at 12 nucleotides
upstream of ATG codon (data not presented). We then tested the effects
of other members of the RAR family on STAT1 promoter. COS-7 cells were
transfected with RAR
, -
, and -
expression vectors along with
luciferase reporter VKL-4 (Fig. 4C). Although all the
members of the RAR family induced the reporter gene expression, RAR
was a slightly better activator than the others. Luciferase gene did
not express upon transfection of VKL-7 in F9 cells. RA treatment for
24 h caused a 12-fold increase in luciferase activity (Fig.
4D). It was further enhanced with longer treatment (48 h).
Sequence analysis of the promoter (Fig. 5) revealed a
direct repeat element, at 467 bp, that could potentially serve as
RARE. More significantly, this sequence had closer resemblance to H2RII of the MHC class I gene (8), than to the other RAREs. Unlike the
previously described retinoid response elements (5), the STAT1-RARE had
a near perfect repeat sequence of GGGTCAGGGTGA with no spacer
nucleotides (Fig. 5). Further, GGG residues were found in both the
half-sites in place of AGG of most retinoid responsive half-sites. At
122 and
338 positions, TATA-like elements were present.
Mutational Analysis of the Retinoic Acid Response Element
Since RA stimulated STAT1 and sequence analysis
identified a RARE in the promoter, we next examined whether this
element alone was sufficient for RA inducibility. Using a PCR-based
approach (17), we constructed a deletion mutant that lacked the direct repeat element in VKL-7 reporter. The wild-type construct (VKL-7) but
not the mutant (M) responded to RA in COS-7 cells when cotransfected with RAR
(Fig. 6A, bars 2-5).
Using the same PCR approach, we generated point mutants with RM1 and
RM2 oligonucleotides in the native promoter (VKL-7). In RM1, GGG
residues of the left half-site were replaced with AAA. RM3 bore
mutations in the central G residue of the GGG bases of both the
half-sites (see "Materials and Methods"). Transfection of these two
point mutants into COS-7 cells along with RAR
did not significantly
induce the promoter (Fig. 6B, bars 3-6). Thus,
mutations in the GGG residues of either half-site abolished RA
inducibility.
We next determined whether fusion of the synthetic STAT1-RARE conferred RA inducibility to a neutral promoter driving the expression of luciferase gene. COS-7 cells were transfected with the reporter constructs bearing either a wild-type or a mutant response element upstream of herpes simplex viral TK gene promoter, along with RARs, and were treated with RA. Insertion of two copies of a wild-type element (RW2X) produced a slightly higher activity than the one with a single copy (RW1X) in COS-7 cells (Fig. 6C, compare bar 4 with 6). Mutants that lacked the GGG residues in either half-site were unresponsive to RA (bars 8, 10, and 12). This observation was consistent with the data obtained with the same point mutants in the context of native promoter (Fig. 6B). Furthermore, mutation of a central G residue of both the half-sites (RM3) caused a similar loss of response. Therefore, the direct repeat element of STAT1 promoter was sufficient for induction by RA.
Binding of Transcriptional Factors to the RA Response ElementTo identify the transcription factors that mediate these
effects, we performed EMSA. In these experiments, untreated and
RA-treated F9 nuclear extracts were incubated with a
32P-labeled synthetic RW (STAT1-RARE) to detect specific
DNA binding (6). A slow mobility complex, A, appeared in RA-treated F9 cells (Fig. 7A, lanes
2-5) whose binding was enhanced with prolonged RA
treatment. Binding of this complex to RW was eliminated upon preincubation of these nuclear extracts with the wild type (RW) oligonucleotide (lanes 6-8) but not by a mutant (RM4)
oligonucleotide (lanes 9-11). Mutant oligos RM1, RM2, and
RM3 also failed to compete out the binding (data not presented).
Formation of complex-A was inhibited by preincubation of extracts with
polyclonal antibodies specific for RAR and RXR
(lanes
12 and 13) but not by those raised against RAR
,
RAR
, and RXR
(lanes 14-16). Binding of this complex
to RW was not eliminated by preincubation of cell extracts with DR-5
(19) and H2RII oligos at 5 or 25 × concentrations (Fig.
7B, lanes 2, 3, 6, and 7). However, at
high concentrations (100 and 500 ×), DR-5 competed with the labeled RW
(lanes 4 and 5), indicating a preferential
formation of complex-A with the latter. H2RII failed to compete out the
RW complex-A (lanes 6-9). RW also formed another complex,
B, with untreated F9 cell extracts whose binding disappeared with RA
treatment of cells (Fig. 7A, see lanes 1-3).
RW binding factors in F9 cells exhibited interesting properties
compared with those that bound to consensus RARE and DR-5. RW binding
of complex-A was RA inducible (Fig. 7C, lanes
2-5), while DR-5 binding factors were constitutive (Fig.
7C, lanes 6-9). Formation of complex-A was not
detected with untreated F9 cell extracts (Fig. 7A,
lane 1). The DR-5 binding factors from untreated F9 cell
extracts (lane 6) were recognized by antibodies against RAR and RAR
(data not presented). Under the same conditions, complexes were barely detectable with H2RII probe (lanes
10-13). Interestingly, complex-B was formed only with RW upon
incubation with untreated F9 extracts but not with DR-5 or H2RII.
Binding of this complex was eliminated upon preincubation of cell
extracts with RW but not RM3, RM4, DR-5, or H2RII oligos (data not
presented). Furthermore, complex-B was not recognized by antibodies
specific for RAR
, RAR
, RAR
, RXR
, or RXR
(Fig.
7D). Thus, it appeared to be a novel factor.
Since the above studies indicated
the binding of RAR and RXR
complexes to STAT1-RARE, we next
determined whether this combination of receptors would augment
RA-dependent gene expression from the STAT1 promoter. VKL-7
reporter vector was cotransfected with expression vectors for either
RAR
or RXR
or the combination into COS-7 cells. RAR
, but not
RXR
, alone induced gene expression (Fig. 8A, bars 2 and 3) upon
RA (1 µM) treatment. However, the combination of these
two receptors caused stronger expression (bar 4). To further
test whether such induction was a result of binding of these receptors
to RW, EMSA was performed with whole cell extracts of COS-7 cells
transfected with the individual expression vectors. RAR
bound to RW,
which was further enhanced when combined with extracts from RXR
expressing cells. (See Fig. 8B, lanes 3 and 5). RXR
alone bound very weakly to this element
(lane 4). Pre-incubation of these extracts with cognate
antibodies eliminated the binding of these complexes (data not
presented). Interestingly, purified RAR
(expressed in a baculovirus
vector) alone failed to form such a complex with RW, although it bound
to DR-5 efficiently (data not shown). These data suggest that
additional cellular factors may be necessary for the formation of RW
binding complexes.
Dependence of IFN- responses in F9 cells on RA-treatment
suggests two possibilities. RA may enrich the levels or activities of
IFN-receptor components and the associated Janus kinases or of
IFN-regulated transcription factors. Since transfection of STAT1 alone
restores the IFN-
-inducible gene expression, modulation of IFN
responses by RA may not involve an enhancement of JAKs or other
receptor components. Consistent with this, RA did not increase IFN
receptors density or avidity in unresponsive cells (20). Furthermore,
RA did not induce the tyrosine phosphorylation of JAK1 (data not
shown). Thus, RA modulation of IFN responses occurs at the level of
STAT1 gene expression in these cells.
Northern and nuclear run off transcription studies have shown that RA
specifically induces transcription of STAT1 gene (Fig. 3,
A-C). RA does not affect STAT2 expression. Further, STAT2
promoter does not possess a RARE (21). Analysis of the STAT1 promoter (Figs. 4, 5, 6) identified a RARE. This element has several unique properties: i) It is a direct repeat element with no spacing between the repeats. ii) It is not stimulated by RXR in the presence of
9-cis-RA or high concentration of RA that activates RXRs.
iii) It preferentially binds the RAR
and RXR
heteromers, and such binding is increased with prolonged treatment of F9 cells with RA.
Under the same conditions RA does not alter the binding of RAR·RXR
complexes to DR-5. iv) In untreated F9 cells, a unique factor
(complex-B) binds to STAT1-RARE (Fig. 7, A and C)
but not to DR-5 (19) or H2RII (8). The STAT1-RARE requires GGG residues in both the half sites, since mutagenesis of these in any half site of
the direct repeat element abrogates the RA responses. Its specificity
for RAR
and RXR
is quite intriguing because a similar
oligonucleotide with 5-bp spacing between the direct repeat elements
(RW-DR5) has not formed the same complex (data not presented). These
data indicate the importance of close apposition of the repeat elements
for specific binding. Although the direct repeat elements with two or
five nucleotide spacing have been shown to respond to RA in several
genes (5), the gene for human medium chain acyl coenzyme A
dehydrogenase has a unique element with eight nucleotide spacing and
another one with no spacing (22). The human oxytocin and mouse laminin
B1 gene promoters also contain RA responsive elements with 13 and 47 nucleotide spacing (23-25). The STAT1-RARE appears to belong to this
class of unique nuclear receptor response elements where spacing
between the half-sites is not a primary determinant of retinoid
responsiveness (26, 27).
Preferential binding of RAR and RXR
heteromers to STAT1-RARE may
permit a selective regulation of STAT1 gene by RA. This notion is
supported by the observation that despite the abundance of DR-5 binding
factors in F9 cells, they did not bind to RW (Fig. 7C). The
DR-5 binding factors were recognized by specific antibodies against
RAR
and RAR
(data not presented). The observations that regulation of transcription by nuclear receptors is dependent on the
promoter context, response element orientation, and DNA binding domain
interface (19, 28-35) are also suggestive of such preferential
interactions. The facts that RAR
gene is inducible by RA (36, 37)
and that the binding of RAR
to RW is increased upon exposure of F9
cells to RA also support the notion of preferential binding of
RAR
·RXR
to STAT1-RARE. Consistent with these observations, cotransfection and EMSA assays with individual receptors in COS-7 cells
resulted in a stronger induction of gene expression (Fig. 8). Since
whole cell extracts containing RAR
, but not the purified RAR
,
form specific complexes with RW, it appears that additional cellular
factors are necessary for the binding to occur. Given the unique
organization of the repeat elements in STAT1-RARE, such factors may
play a crucial role in the formation of high affinity complexes.
Further investigation is necessary to address these issues. Lastly, the
F9 cellular factor (complex-B in Fig. 7A) that binds to
STAT1-RARE, may be a negative regulator since it disappears with
RA-treatment and is absent during STAT1 transcription. An important
outcome of our study is that it demonstrates for the first time a
molecular basis for the modulation of IFN action by RA. This
observation indirectly connects the retinoids to cell cycle regulation.
For example, recent studies indicate that IFNs inhibit cell growth
using STAT1 (38) and induce the expression of p21/WAF/Cip-1 gene, whose
product inhibits the cyclin-dependent kinases (39).
Elevation of STAT1 levels may thus permit a robust activation of STAT1
by IFNs, leading to growth arrest.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) U80267[GenBank].
The authors thank all the colleagues who have generously provided several valuable reagents used in this study, Sara B. Mannino for technical assistance at the early stages of this work, Rama Kudaravalli for help in transfection experiments, and Daniel Lindner and Ernest Borden for a critical reading of this manuscript.