(Received for publication, January 14, 1997, and in revised form, May 30, 1997)
From the Laboratoire de Régulation de l'Expression des Gènes Eucaryotes, CNRS, UPR 37, UFR Biomédicale des Saints-Pères, Université René Descartes, Paris V, 45 rue des Saints-Pères, 75270 Paris Cedex 06, France
Comparative transfection analysis of murine
interferon A4 and interferon A11 promoter constructs transiently
transfected in mouse L929 and human HeLa S3 cells infected with
Newcastle disease virus showed that the second positive regulatory
domain I-like domain (D motif), located between nucleotides 57 and
46 upstream of the transcription start site, contributes to the
activation of virus-induced transcription of the interferon (IFN)-A4
gene promoter by cooperating with the positive regulatory domain I-like and TG-like domains previously described. Electrophoretic mobility shift assay performed with the virus-inducible fragments containing these motifs indicated that the binding activity that we have denoted
as virus-induced factor (Génin, P., Bragança, J., Darracq, N., Doly, J., and Civas, A. (1995) Nucleic Acids Res. 23, 5055-5063) is different from interferon-stimulated gene factor 3. It
binds to the D motif but not to the virus-unresponsive form of the D motif disrupted by a G
57
C substitution. We show that
the low levels of IFN-A11 gene expression are caused essentially by the
lack of two inducible enhancer domains disrupted by the
A
78
G and the G
57
C substitutions.
These data suggest a model taking account of the differential
regulation of IFN-A gene family members. They also suggest that
virus-induced factor may correspond to the primary transcription factor
directly activated by virus that is involved in the initiation of IFN-A
gene transcription.
Type I interferons (IFN-A1 and IFN-B) belong to the group of cytokines that constitute an early response to viral infection in eukaryotic cells. Whereas IFN-B is encoded by a single gene, which is expressed essentially in fibroblastic cells, IFN-A is represented by a large family of structurally related genes, which are predominantly expressed in lymphoid human cells or murine fibroblasts and macrophages (2-6). The IFN-A genes characterized so far have been shown to be coordinately induced in virus-infected cells. However, differences in the expression of individual IFN-A mRNAs are observed in both human and mouse cells, which reflect the transcriptional activity of the corresponding promoter regions in a particular cell type (2, 7-9).
Virus-induced expression of IFN-A and IFN-B genes is mediated by
regulatory sequences located within 200 base pairs upstream of the
transcription start site of their promoters (10-14). IFN-B gene
expression is regulated by multiple factors that interact with positive
regulatory domains (PRDs) located within virus-responsive element B
(VRE-B): interferon regulatory factor-1 (IRF-1), which binds to both
the PRDI and PRDIII domains, also known as IRF elements (15-18); the
nuclear factor NF-B, which recognizes the PRDII domain (19-21); and
the ATF-2/c-Jun heterodimer, which binds to the PRDIV domain and is
required for maximal virus inducibility in mouse L929 and human HeLa
cells (22). Binding of both the NF-
B and ATF-2/c-Jun complexes is
facilitated by HMGI(Y), members of the high mobility group of proteins
that bend the DNA and allow VRE-B-binding factors to assemble in a
higher order nucleoprotein complex called an "enhanceosome," thus
promoting the virus-induced transcription of the IFN-B gene (23, 24).
Postinductional repression and maintenance of a poised state of
repression seem to be mediated by IRF-2 (and its truncated form), which
recognize the IRF elements and prevent the binding of IRF-1. Other
factors have been implicated in the repression of transcription by
their interaction with the IRF elements or the negative regulatory
domains surrounding VRE-B (18, 25-29).
Factors specifically involved in regulation of the IFN-A genes are less well defined. The virus-responsive element of the human IFN-A1 gene (VRE-A1) and the inducible element of the murine IFN-A4 gene have been shown to contain a PRDI-like site, thus involving IRF-1 in the induced expression of the IFN-A genes (30-32). A comparative study of the IFN-A4 and IFN-A6 gene promoters displaying virus inducibility has led to the detection, by EMSA, of a binding activity, AF1. It has been proposed that AF1-forming proteins cooperate with IRF-1 in the virus-induced transcription of IFN-A genes (4, 33, 34). Another factor possibly involved in the specific regulation of IFN-A gene regulation is the TG protein, which binds to the hexameric GAAATG repeats that confer virus inducibility by an IRF-1-independent pathway. The so-called "TG sequence" is adjacent to the PRDI-like motif in the human VRE-A1 and is conserved in most human or murine IFN-A (30, 35). However, the AF1-related and TG-binding proteins have yet to be characterized.
Interestingly, the deletion of the gene coding for IRF-1 did not affect
virus-inducible expression of the type I IFN genes, thus calling into
question its role (36, 37). Another member of the IRF family,
IFN-stimulated gene factor 3 (ISGF3
) has been suggested to
participate in the regulation of the IFN-B gene promoter both alone and
as a heteromeric complex (ISGF3) formed with Stat1
and Stat2 (from
the signal transducer and activator of transcription family) (38, 39,
40, 41). ISGF3 has been shown to be induced secondarily by the virus
and involved in the amplification of type I IFN gene expression
triggered initially by virus infection of cells. However, the primary
transcription factor, directly activated by the virus and mediating the
PRDI-dependent initiation of transcription of the type I
IFN genes, remains to be identified.
We have previously shown that the differential expression of IFN-A4 and
IFN-A11 genes, in NDV-induced L929 cells, was due in part to the
negative effect of the A78
G substitution that
distinguishes the inducible element of IFN-A4 (IE-A4) from the
equivalent region of the IFN-A11 gene promoter and also to the presence
of negatively acting sequences located upstream of the inducible
element of the IFN-A11 promoter (42). By comparative analysis of the
109 to
64 fragments of the IFN-A4 and IFN-A11 gene promoters, which
include their inducible elements, we identified a virus-induced binding
activity that is stimulated within 1 h of virion contact with
cells (1). This virus-induced factor (VIF) recognizes specifically the
PRDI-like domain shared by both the inducible elements and the TG-like
domain of IE-A4.
In the present study, we report that VIF also binds a second PRDI-like motif located between the IE-A4 and the TATA box of the IFN-A4 promoter. We show that this motif contributes to the high level of NDV-induced transcription of the IFN-A4 gene by cooperating with the IE-A4 subdomains and that the low levels of IFN-A11 gene expression are caused essentially by the absence of both the TG-like and second PRDI-like domain in this promoter. Our data lead us to propose a model for the differential regulation of the IFN-A gene family members and to delineate the involvement of VIF in this regulation.
The Bi series of plasmids pIF4 and
pIF11 carrying, respectively, the 211 to +19 and the
198 to +19
promoter regions of the IFN-A4 and IFN-A11 genes and the J series
deleted up to position
119 were derived from the plasmids pIF4T and
pIF11T previously described (42) by the subcloning of polymerase chain
reaction-amplified fragments. The 3
-primers were p17 for pIF4Bi or
pIF4J and p18 for pIF11Bi or pIF11J, with p3 and p4 as the 5
-primers
for the Bi and J constructs, respectively. These fragments, digested by PstI and BamHI, were subcloned into the plasmid
pBLCAT3 (44). The constructs carrying a point mutation at position
78, pIF11[V4]Bi and pIF4[V11]Bi, were obtained similarly from the
corresponding T series (42). The IFN-A11/A4 hybrid plasmid pIFH3Bi was
derived from subcloning of the fragment amplified with the primers p3 and p18 with the plasmid pIFH66T (42). The pIFH1Bi and pIFH2Bi constructs were obtained in two steps by a procedure previously described (43); the fragments amplified in the first round using, respectively, pIF11T (or pIFH66T) as the template and p3 and p10 (or p3
and p9) as primers, were used as the 5
-primer in a second polymerase
chain reaction using either p17 and the hybrid plasmid pIFH66T or p18
and the plasmid pIF11T as the 3
-primer and template, respectively. The
construct pIFH4Bi was obtained similarly from pIF11T, using primers p3
and p5 in the first step, and the amplified fragment was used with
primer p18 and plasmid pIF11T in the second polymerase chain reaction.
To construct the plasmids pIF4T-3S and pIF4T-1S or plasmids pIF4T-4S
and pIF4T-2S, either plasmid pIF4T or plasmid pIF4[V11]T was used as
a template in each round. p17 was the primer for the second step,
whereas either p6 or p7, respectively, was used as a 3
-mutagenic
primer with p1 as the 5
-primer in the first round. Similarly, to
construct pIF11T-3S and pIF11T-1S, pIF11T was used as template in both
steps with p18 in the second and p5 or p8 as the mutagenic primer used
with p2, whereas pIF11[V4]T was used as template to obtain the
plasmids pIF11T-4S and pIF11T-2S. In each case, the final products were purified, digested by PstI and BamHI, and
inserted between these sites in the pBLCAT3 polylinker. The promoter
constructs carrying internal deletions were obtained by the ligation of
two separately amplified fragments, one digested by PstI,
the other by BamHI, to
PstI-BamHI-digested pBLCAT3. Thus, pIF4
Bi and
pIF11
Bi were constructed by ligation of two fragments amplified with
p3/p21 and p19/p17 or with p3/p21 and p20/p18, respectively, in the
presence of pIF4Bi or pIF11Bi as template. The pV4, p[B]2
and p[C4]2 constructs were previously described (1). The
constructions depicted in Fig. 6A were obtained by insertion
of synthetic fragments of the IFN-A4 and IFN-A11 promoters at the
HindIII and BamHI sites upstream of the herpes
simplex virus thymidine kinase promoter fused to the CAT gene (pBLCAT2
modified by G. Schütz). The correct sequences of the
constructions presented were confirmed by DNA sequence analysis. The
sequence information for polymerase chain reaction mutagenesis is
available upon request.
Transfection and CAT Assays
Mouse L929 cells (ATCC CCL 1) were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% horse serum. HeLa S3 cells (ATCC CCL 2.2) were maintained as monolayers in the same medium supplemented with 10% newborn calf serum. Cells were transfected by the calcium phosphate method (45) and induced with NDV (La Jolla strain, New Jersey) or mock-induced as described previously (1). Measurements of CAT activity were performed on cytoplasmic extracts obtained 24 h after induction (46). The protein concentration was determined with the Coomassie G250 procedure as described previously (1). The results are derived from the average CAT values obtained in at least three independent transfection experiments, each performed in triplicate with two or three different clones.
Preparation of Nuclear ExtractsL929 cells and HeLa cells
were grown in suspension, respectively, in Dulbecco's modified
Eagle's medium and suspension-modified Eagle's medium (Life
Technologies, Inc.) with 10% fetal calf serum. Virus induction for
4 h was performed by keeping the cells in contact with NDV for
1 h in serum-free medium containing 50 µg/ml of cycloheximide
and by maintaining the cells for a further 3 h in the presence of
50 µg/ml cycloheximide and 2% fetal calf serum. IFN induction was
carried out with recombinant mouse IFN-11 (103 units/ml)
for 2 h. After washing and centrifugation, the cells were
harvested immediately, and the nuclear extracts were prepared according
to Dignam et al. (47).
The oligonucleotides BC4 and BC11
correspond to the 98 to
74 region of the IFN-A4 and IFN-A11
promoters, respectively. In the dC4 and dC11 oligomers, the module B is
replaced by C4 in BC4 or by C11 in BC11. The B2
oligonucleotide corresponds to the dimer of the
98 to
87
GAAAGTGAAAAG sequence of the IFN-A4 and IFN-A11 promoters. The
C42 and D42 represent, respectively, two copies
of the
85 to
74 GAATTGGAAAGC and
57 to
46 GAAAGGAGAAAC sequences of IFN-A4, whereas C112 and D112
refer, respectively, to the dimers of their mutated forms
GAATTGGGAAGC and CAAAGGAGAAAC present in the
IFN-A11 promoter (the substituted residues are underlined).
ISRE-ISG15 (gatcCTCGGGAAAGGGAAACCGAAACTGAAGCC) and
GIRE-FcR (ctagATTTCCCAGAAAAGGAACATGATGAATctag) are,
respectively, the binding sites for ISGF3 and Stat1 dimers (48,
49). Preparation of the probe was performed using single-stranded
oligonucleotides 5
-labeled with [
-32P]ATP (Amersham,
UK; 3000 Ci/mmol) and annealed to a slight excess of the unlabeled
complementary strand. ISRE was labeled by using the Klenow fragment of
DNA polymerase I. Binding reactions and electrophoresis were performed
as described previously (1). For competition assays, the molar excess
of unlabeled competitor was mixed with the probe prior to the addition
of the nuclear extract, and the bound DNA was quantified using a
PhosphorImager (Molecular Dynamics; 445-SI). For experiments with
antibodies, the binding reaction was preincubated for 1 h on ice
with the appropriate antisera before the addition of the labeled probe. Anti-ISGF3
(sc-496 X) and anti-Stat1
/
(sc-464 X) antibodies were obtained from Santa Cruz Biotechnology, Inc., and the
anti-phosphotyrosine 4G10 antibodies were from Euromedex, France.
We have shown
that the difference in NDV inducibility of two members of the IFN-A
gene family, namely IFN-A4 and IFN-A11, is in part due to a
A78
G substitution and to the presence of negatively
acting sequences located between
244 and
199 in the IFN-A11 gene
promoter (42, 50). Transient transfection experiments were therefore
carried out to identify the additional regulatory element(s) involved in the differential NDV-induced response conferred by these two promoters.
A comparison of the results obtained with constructions carrying
different 5-deleted fragments of these promoters showed that the
region between
198 and
119 did not affect the NDV inducibility of
the IFN-A11 promoter, whereas, as described previously by Raj et
al. (31), the deletion of the corresponding region exerted a
slight positive effect on the virus-induced transcription of IFN-A4
(Fig. 1). However, the virus inducibility
of IFN-A4 remained consistently higher than IFN-A11, even when the
promoters were devoid of any modulatory region upstream of their
inducible element (the Bi or J series of constructions). The creation
of a TG-like domain in IFN-A11 by the G
78
A mutation
increased its inducibility, but not to the expression levels conferred
by IFN-A4 (Fig. 2A). The
disruption of the TG-like domain in IFN-A4 decreased the induced
transcription levels without reducing them to those of IFN-A11. Thus,
the G
78
A substitution alone is not sufficient to
explain the difference between the virus inducibility of these
promoters. Transfection of the IFN-A11/A4 hybrid promoters showed that
the substitution of the nucleotides (TAC)A11, located from
59 to
57, by (CCG)A4 significantly increased the virus
inducibility (pIFH4Bi compared with pIF11Bi), suggesting the presence
of an additional virus-responsive sequence located between the
inducible element and the TATA box in the IFN-A4 promoter.
Alternatively, a negatively acting sequence could be present in the
corresponding region of IFN-A11. To investigate these hypotheses, the
residues
59 to
57 were deleted in the Bi series of constructs or
interchanged between the two promoters extending up to nucleotides
470 to
457. The (CCG)A4 deletion significantly
decreased the virus inducibility of IFN-A4, whereas a
(TAC)A11 deletion in IFN-A11 did not have any effect (Fig.
2B). This suggests that the CCG deletion disrupts a
virus-inducible element in the IFN-A4 promoter. Furthermore, the
substitution of (CCG)A4 by (TAC)A11 decreased
the virus inducibility of the native IFN-A4 promoter fragment 3-fold,
whereas the reciprocal substitution increased the IFN-A11 inducibility
more than 6-fold (Fig. 3). These data
confirmed the presence of a virus-inducible enhancer in the IFN-A4
promoter. The more pronounced effect of (TAC)A11 conversion
in IFN-A11 in the T series compared with the Bi series may be due to
the deletion of the negatively acting sequences located between
nucleotides
244 and
199 in this promoter because the inducibility
of pIF11Bi is already the 2-3-fold higher.
The Identified Virus-responsive Domain Is a Second PRDI-like Motif Acting in Synergy with IE-A4
Since we were able to show the 78
A/G substitution had an effect on virus inducibility, we measured the
cumulative effect of the
78 and
59 to
57 substitutions in both
promoters. The substitution of (CCG)A4 by
(TAC)A11 together with the A
78
G mutation
caused a dramatic decrease in the IFN-A4 inducibility, reducing the
transcription rate to the level conferred by the IFN-A11 promoter (Fig.
3). The values obtained, compared with those conferred by the IFN-A4
constructs carrying only the A
78
G or the
(TAC)A11 substitutions, clearly indicated a synergistic effect between the newly identified virus-responsive enhancer and the
inducible element IE-A4. Consistently, reciprocal substitutions in the
IFN-A11 promoter augment the virus-induced transcription. In this case,
the cooperative effect of the enhancer with the inducible element was
hindered, probably due again to the 2-3-fold inhibitory effect of the
negatively acting sequences located in the
244 to
199] region of
the IFN-A11 promoter.
Sequence analysis of the IFN-A4 and IFN-A11 promoters revealed that in
the 70 to
40 region, the
59 to
57 substitutions may affect
potential binding sites for the CCAAT/enhancer-binding protein
,
NF
B, and IRF family of transcription factors. Although CCAAT/enhancer-binding protein
has been involved in the regulation of several cytokine genes in response to stimulation by
lipopolysaccharide (51, 52), we did not find any evidence concerning
its potential role in IFN regulation. NF
B, which is involved in the
virus-induced expression of the IFN-B gene, does not seem to
participate in IFN-A gene regulation (53, 54). We therefore focused our
attention on the potential effect of the PRDI-like motif located
between nucleotides
57 and
46. in IFN-A4, which is disrupted by the G
57
C substitution in IFN-A11.
Comparison of the results presented in Fig. 3 indicates that the
negative effect of the G57
C substitution alone on
the virus-induced transcription of IFN-A4 was comparable with the
inhibition observed with the (TAC)A11 substitutions. The
A
78
G and G
57
C mutations were
sufficient to reduce the transcription levels of a highly inducible
500-base pair IFN-A4 promoter to those of a very weakly induced IFN-A11
promoter (pIF4T-1S, pIF4T-3S, and pIF4T-2S). Interestingly, the
C
57
G substitution alone or in combination with the
G
78
A substitution (pIF11T-1S and pIF11T-2S) exerted
a reverse effect on the IFN-A11 promoter, increasing considerably its
virus-induced transcription. Taken together, these results indicate
that the proximal PRDI-like site participates in the regulation of the IFN-A4 promoter by cooperating with the inducible element to enhance its virus-induced transcription and that the disruption of both the
TG-like and the proximal PRDI-like motifs by the A
78
G and G
57
C mutations, respectively, contributes to
the low level of IFN-A11 gene expression.
A similar series of transfection experiments performed in the human
HeLa cells indicated that the 470 to +19 fragment of IFN-A4 was
highly inducible after virus infection in comparison with the
equivalent IFN-A11 promoter (Fig. 3). Moreover, the single mutation of
either the
78 A/G or the
57 G/C and the simultaneous substitution
at both positions had the same effects on NDV-induced transcription as
those observed in L929 cells. However, no difference could be observed
in the values between these mutated constructs and the native IFN-A4 or
IFN-A11 promoter fragments due to the constitutive expression levels,
which increased concurrently with the virus-induced transcription
levels of the constructs. However, the same regulatory elements,
i.e. the TG-like domain of IE-A4 and the proximal PRDI-like
domain, apparently ensured maximal NDV-induced transcription of the
IFN-A4 gene in HeLa cells, suggesting that human cell lines also
contain factors able to interact with these domains.
We therefore compared the EMSA patterns obtained with nuclear extracts
prepared from NDV-infected or mock-induced L929 and HeLa cells using as
DNA probes different oligonucleotides corresponding to dimers of the
subdomains of the inducible element IE-A4 that have been shown
previously to confer virus inducibility to the HSV-tk promoter in L929
cells (1). The results presented in Fig.
4 indicate that two copies of the
PRDI-like and TG-like motifs of the inducible element IE-A4, denoted
B2 and C42, respectively, were also
virus-responsive in HeLa cells and that the inducibility may be
correlated with a binding activity stimulated by the virus infection,
which we have denoted VIC. The fact that this complex was the only one
detected with the inducible TG-like motif (Fig. 4, lanes
5-8) strongly suggested that VIF, the factor forming the complex
VIC, was responsible for the virus inducibility conferred by these
motifs in both the HeLa and L929 cell lines. IRF-2 activity was only
detected with the PRDI, essentially in mock-induced nuclear extracts,
yet with a reduced intensity in HeLa cells (Fig. 4, lanes 1 and 3). An explanation for this decrease could be the presence of lower amounts of this repressive factor in HeLa cells, which may also account for the higher constitutive expression levels
observed in the transfection experiments.
The Virus-induced Factor VIF and IRF-2 Recognize the Dimers of the Proximal PRDI-like Domain of the IFN-A4 Promoter
Since the
G57
C substitution decreased NDV inducibility of the
IFN-A4 promoter by disrupting the proximal PRDI-like domain, we
performed EMSA with two copies of the
57 to
46 GAAAGGAGAAAC sequence of the IFN-A4 promoter corresponding to this domain and with
its mutated form CAAAGGAGAAAC present in IFN-A11, denoted as D42 and D112, respectively. The patterns
obtained with D42 were highly similar to those with
B2: three IRF-2-containing complexes (all supershifting
with anti-IRF-2 antibodies; data not shown) and a complex that
displayed identical mobility and very similar competition patterns to
VIC (Fig. 5). IRF-2-containing complexes were easily removed by the PRDI-related oligomers used as competitors but not by the TG-like domain of IFN-A4 (Fig. 5, lanes
8-13), the latter having been shown previously to be recognized
with very low affinity by IRF-1 and IRF-2 (1). The complex
corresponding to VIC could be competed out by a 50-fold molar excess of
B2 or C42 and by a 100-fold molar excess of
D42, but not by the D112 (Fig. 5, lanes
2-13). These data indicate that the binding of VIF is affected by
the G
57
C substitution and suggest that this factor
is also involved in the NDV-induced transcription mediated by the D
element of IFN-A4.
VIF Binding Activity Correlates with the Virus Inducibility
Of the heterologous HSV-tk constructs, pBC4
(carrying the PRDI-like and TG-like domains, which correspond to the
core sequence of the inducible element IE-A4) and pBC11 (containing the
98 to
74 fragment of the IFN-A11 promoter), only pBC4 displayed virus inducibility (Fig. 6A).
This difference in response suggests that the A
78
G
substitution present in the BC11 fragment disrupts the sequence required for the binding of the factor involved in the virus-induced activation of transcription. On the other hand, the similarities between the inducibility ratios of pBC4 and pdC4 suggested that the
PRDI-like and the TG-like domains of IFN-A4 should both be recognized
with similar affinity by this factor. EMSA performed with the BC4,
BC11, and dC4 probes in the presence of virus- or mock-induced L929
nuclear extracts revealed that VIF was detected only when the DNA
fragment was virus-inducible, the unresponsive combinations displaying
no VIF binding activity (Fig. 6B). These results were
confirmed by competition experiments that also showed that VIF binding
affinity for BC4 was 4-fold higher than for BC11 (data not shown).
Thus, the virus responsiveness of the PRDI-like and TG-like domains of
the inducible element IE-A4 is correlated with the binding of VIF to
these motifs.
Since the B and D motifs behaved
also as binding sites for IRF-2, we cannot exclude the possibility that
other members of the IRF family, particularly IRF-1 or ISGF3, also
recognize these motifs to mediate the virus-induced transcription. In
fact, the interaction of ISGF3 with the virus-responsive element of the human IFN-A1 gene promoter via ISGF3
was recently reported,
suggesting the requirement of this secondarily NDV-induced factor in
the amplification of IFN-A gene expression (41). Since we showed that
VIF was different from IRF-1 and IRF-2 (1), we wondered whether any
relationship existed between VIF and ISGF3. Competition experiments
indicated that the ISRE motif of the ISG15 gene promoter (48), a ISGF3
binding site, inhibited the formation of VIC with the same efficiency
as PRDI and TG4 dimers (at a 25-fold molar excess for a 60-80%
inhibition), whereas a mutated form of ISRE that is not recognized by
ISGF3 did not affect the VIF binding, suggesting a similarity in the
composition of VIF and ISGF3 (Fig. 7,
lanes 8 and 9). We therefore attempted using
B2 and C42 as specific competitors to identify
VIC among the ISRE complexes detected in NDV-induced nuclear extracts.
Strikingly, other than the IRF-2-containing complexes inhibited only by
ISRE and B2, no complex was competed out by ISRE,
B2, or C42 with the same efficiency (data not
shown). Our inability to detect the VIF binding activity with ISRE may
be due to the heteromeric structure of this factor, which could have at
least one subunit unrecognized by the probe. When used as a competitor,
ISRE could then titrate the remaining subunit(s) shared by VIF and
ISGF3. The ISRE core sequence contains a binding site for the IRF
family members and particularly for ISGF3
; the component titrated by
competition could therefore correspond to this subunit. However,
antibodies raised against ISGF3
did not prevent the formation of VIC
(Fig. 7, lane 4), whereas they efficiently inhibited ISGF3
(data not shown), indicating that VIC did not contain ISGF3
.
Moreover, the unreactivity of VIF with anti-phosphotyrosine antibodies, which recognized the activated ISGF3 in control experiments, suggested also that VIF was different from ISGF3 (Fig. 7, lane 6). The
slight inhibition of VIF binding activity observed in the presence of anti-Stat1
/
antibodies (Fig. 7, lane 5) suggested,
however, that Stat1
or Stat1
may participate in the formation of
VIF. Stat1 has been shown to homodimerize for binding to different IFN
activation site (GAS) motifs and to heterodimerize with Stat3 for recognizing various natural palindromic Stat-binding elements of
cytokine-responsive genes or with ISGF3
and Stat2 for binding to the
ISRE sequences (55). The contribution of Stat1 homodimers in the
formation of VIF is unlikely, since they do not bind to the ISRE-ISG15
sequence (56), which efficiently competed out VIF, and since the
GIRE-Fc
R sequence containing the GAS motif, which is recognized by
Stat1 homodimers (49, 56), did not inhibit the VIF binding activity
(data not shown). These data strongly suggest that VIF is different
from ISGF3, although they may share the Stat1 subunit.
The results reported in this study, together with previously
published data (4, 30), allow insight into the modular architecture of
the IFN-A gene promoters. Thus, we suggest that the virus-responsive element of IFN-A4 may be considered to be composed of four enhancer motifs exhibiting different properties: the A motif, represented by the
103 to
93 GTAAAGAAAGT sequence, which is not virus-inducible even
in multiple copies and requires a juxtaposition with the B domain to
respond to virus induction (34); the B and C motifs, corresponding,
respectively, to the
98 to
87 GAAAGTGAAAAG and
85 to
74
GAATTGGAAAGC sequences, which are virus-responsive once multimerized or
in combination with each other (see Ref. 1 and this study); and finally
the D motif, represented by the
57 to
46 GAAAGGAGAAAC sequence,
which is identified in this report and here shown to cooperate with the
B and C domains to confer maximal NDV inducibility to the IFN-A4
promoter in L929 cells (Fig. 8).
In the L929 cell line, determination of the relative NDV-induced
expression levels between individual murine IFN-A mRNAs has revealed that IFN-A4 is the predominant species over IFN-A2, while IFN-A6 and IFN-A11 are very weakly expressed (8, 9, 54). The difference
in expression between the IFN-A4 and IFN-A6 genes has been attributed
to substitutions disrupting the A and B motifs in the IFN-A6 promoter
(31). However, this finding alone is not sufficient to explain the
lower expression levels of IFN-A2 and IFN-A11, since they show perfect
homology with IFN-A4 with respect to these motifs in their promoters.
In fact, in the 109 to
75 region, which corresponds to the
previously delimited inducible element IE-A4, the IFN-A2 and IFN-A11
promoters are distinguishable from IFN-A4 by the G
80
A and A
78
G substitutions in IFN-A2, and only by the
A
78
G substitution in IFN-A11. By transient
transfection experiments performed with the IFN-A4 and IFN-A11
promoters extending up to
470 nucleotides from the transcription
initiation site, as well as with hybrid IFN-A4/A11 promoter constructs,
we showed that the difference between the virus responsiveness of the
IFN-A4 and the IFN-A11 promoters was due essentially to the
A
78
G and G
57
C substitutions that
disrupt, respectively, the C and D motifs in IFN-A11. In fact,
abrogation of either the C or D domain affected only partially the
virus-induced transcription of IFN-A4. Interestingly, a similar
organization exists naturally in IFN-A2 where the C motif is disrupted
by the A
78
G substitution, but the D motif is intact.
In NDV-induced L929 cells, the lower levels of IFN-A2 expression, in
comparison with IFN-A4, are thus most likely due to attenuated
cooperation among the A, B, and D domains instead of maximal synergy
levels conferred by the four motifs. Differential virus-induced
expression of IFN-A genes seems then to be directly dependent on the
number of cooperating enhancer motifs within their promoter.
The similarity of the results obtained in transfection experiments in
HeLa cells with different IFN-A4 and IFN-A11 constructs and the
detection of VIF binding activity in the nuclear extracts obtained from
the NDV-induced HeLa S3 cells suggest that the model proposed in this
study may also be valid for human cell lines. The results obtained here
with the transiently expressed IFN-A4 or IFN-A11 promoter constructs
correlate well with the endogenous expression patterns of the
corresponding IFN-A genes in L929 cells. In contrast, the expression of
endogenous human IFN-A genes, as well as the CAT activity of the SV-CAT
reporter constructs containing the human IFN-A1 promoter, were
undetectable upon NDV or Sendai virus induction of HeLa cells (7, 57,
58). This discrepancy may be due to the presence of a negative
regulatory sequence located in the human IFN-A1 promoter between the
two highly conserved motifs equivalent to C and D domains of murine
IFN-A4. This negatively acting region has been suggested to be the
binding site for a repressor that affects human IFN-A gene expression,
especially in restrictive cell types such as HeLa cells (58). Within
their proximal 120 to +19 promoter fragment, the murine and human
IFN-A genes exhibit sequence divergence; the putative repressor may therefore not affect the inducibility of murine IFN-A promoters in
these cell lines.
The minimal virus-inducible combination, BC4, is the binding site for
the VIF that was previously detected with two copies of the B and C4
modules (1). In contrast, VIF is not detected with the BC11
combination, which carries the A78
G substitution in
the C domain and fails to respond to virus induction. VIF also
recognizes the dimerized D motif of IFN-A4, a motif that is shown to
participate in virus-induced transcription. Conversely, VIF does not
recognize the equivalent region of IFN-A11 disrupted by the
G
57
C substitution. The D motif also recognized by
IRF-2 may be involved, as is the B motif, in IRF-2-mediated repression
of transcription. However, the B and D motifs described as PRDI-like
domains (or IRF elements) and the C motif denoted as the TG-like domain
should be considered as VIF-binding sites. Thus, we suggest that VIF is
involved in virus-induced activation of IFN-A gene transcription by
recognizing the B, C, and D motifs in the IFN-A4 promoter, whereas it
binds only to the B and D motifs of IFN-A2 and the B motif of IFN-A11
(Fig. 8).
The fact that VIF recognizes the IRF elements and is competed out by
the ISRE motif, which contains an IRF-binding site in its inner core
sequence, suggests that this factor may contain a DNA-binding subunit
related to the IRF family. However, EMSA performed in the presence of
antibodies raised against different members of the IRF family indicated
that VIF did not correspond to or contain IRF-1 and IRF-2 and that
ISGF3 is not involved in VIF formation. Nevertheless, the
possibility still cannot be excluded that VIF contains a DNA binding
subunit related to other recently cloned members of the IRF family
(59-63), although to date there is no evidence to suggest their
involvement in virus-induced type I IFN gene expression. These factors
were shown, in fact, to recognize the ISRE motif and to exert
repressive effects on IFN-stimulated gene transcription. Recently,
binding activities stimulated by double-stranded RNA (DRAFs)
characterized in HeLa cells treated with poly(rI·rC) or infected with
a mutant of adenovirus lacking the e1a oncogene were
suggested to participate directly in the virus-activated transcription
of some IFN-stimulated genes (64, 65). A relationship between VIF and
ISRE-binding DRAF activities seems unlikely, since VIF is not detected
by the ISRE probe and the DRAFs were shown not to recognize the IRF
elements. In addition, the absence of reactivity between VIF and
anti-phosphotyrosine antibodies indicated, as did anti-ISGF3
antibodies, that VIF was different from ISGF3. Furthermore, we have
shown that in NDV-induced L929 cells, VIF, unlike ISGF3, appears in the
nuclei within 1 h of contact between the cells and the virions,
without requiring de novo protein synthesis, and that VIF
activation precedes the secondary stimulation of ISGF3 (1, 40, 66). We
therefore suggest that VIF, which we have shown to be different from
IRF-1 and ISGF3, corresponds to the primary transcription factor
directly activated by NDV and mediates initiation of the IFN-A gene
transcription. The role of IRF-1 and ISGF3 may be to maintain
transcription initially activated by VIF via interaction with the B and
D domains.
We especially thank Moshe Yaniv for critical
reading of the manuscript. We are also indebted to Nobumasa Watanabe
and Tadatsugu Taniguchi for providing the mouse anti-IRF-1 and
anti-IRF-2 antisera, John Hiscott for the human anti-IRF-1 and
anti-IRF-2 antibodies, and Keiko Ozato for the human ISGF3
antibodies.