ABSTRACT
The initial genomic response to serum growth factors is the transcriptional
activation of a set of immediate-early genes. Serum-induced transcriptional activation of several of these genes
involves the formation of a ternary complex that includes the serum response
factor (SRF), a 62 kDa ternary complex factor (TCF) and a serum response
element (SRE). TCF alone does not bind the SRE of the protooncogene c-
fos,
but requires the prior assembly of the SRF-SRE binary complex for it to be recruited into a ternary complex. Here we
show that this SRF-SRE binary complex is not an obligatory prerequisite for the formation of
a serum responsive ternary complex. We demonstrate that Elk-1, which has properties of TCF, can recruit SRF into a ternary complex on elements that
do not support formation of the SRF-DNA binary complex. We also show that for two immediate-early genes,
pip92
and
nur77
, formation of the ternary complex may occur without the prior assembly of SRF-DNA binary complex. Finally, we show that the ability of different
sequences to support formation of Elk-1-SRF-DNA ternary complex
in vitro
correlates with their ability to respond to serum growth factors
in vivo
. Our results suggest that a much broader range of DNA sequences than the
consensus SRF and TCF binding sites can support ternary complex formation, and by inference, serum induction. Possible implications of these results are
discussed.
One of the primary cellular responses to the actions of polypeptide growth factors is the rapid and transient activation of a set of immediate-early genes, which number as many as 100 in mouse 3T3 fibroblasts (
1
-
3
). A subset of these genes is co-regulated with the protooncogene c-
fos
: they are transcriptionally activated within minutes of growth factor addition,
reaching a maximal level of transcription at 10-30 min and are repressed within 1 h thereafter (
1
). Among this group of genes, the
c-fos
protooncogene has been studied most extensively with regard to the mechanisms
underlying its rapid induction (
4
-
6
). Activation of c-
fos
by serum growth factors is mediated through a promoter element known as the
serum response element (SRE), a sequence comprised of a CArG box [CC(A/T)
6
GG] and a weak Ets-like binding site (
5
,
6
). The c-
fos
CArG box binds a dimer of a 67 kDa protein, serum response factor (SRF) (
7
). Induction of c-
fos
by TPA in Balb/c 3T3 (
8
) or NIH 3T3 cells (
9
) requires another protein, a 62 kDa ternary complex factor (TCF or p62
TCF
), which binds at the Ets-like binding site. TCF does not bind to the c-
fos
SRE by itself, but requires the prior formation of the SRF-SRE binary complex (
10
). At least two Ets family transcription factors, Elk-1 and SAP-1a, have properties of TCF
in vitro
(
11
,
12
). Although Elk-1 can bind independently to its own high affinity binding sites
characteristic of Ets-domain proteins and can act as a transactivator (
13
,
14
), it does not bind to the c-
fos
Ets-like site by itself. SRF interacts physically with Elk-1 (
15
,
16
) while bound upon the CArG box, thus recruiting Elk-1 to bind a neighboring Ets-like binding site to form a ternary complex. The distance between
the CArG-box and the Ets-site can vary without detracting from the ability of SRF to recruit
TCF for ternary complex formation (
17
).
To gain insight into the diversity of the regulatory mechanisms that control the
coordinate activation of immediate-early genes, we have analyzed the regulation of several such genes that
are transcriptionally activated with kinetics similar to those of c-
fos
. Among those genes studied are
cyr61
, which encodes an extracellular matrix signaling molecule (
18
);
nur77
, which encodes a transcription activator of the steroid/thyroid hormone
receptor superfamily (
19
); and
pip92
, which encodes a short-lived, proline-rich cytoplasmic protein of unknown function (
20
). Our results indicate that the TCF/SRF-mediated mechanism of immediate-early gene activation is surprisingly general, at least in
fibroblasts. We initially thought that the immediate-early gene
cyr61
was not regulated through SRF since no CArG box was found within 1.5 kb of the transcriptional start. Further analysis showed that
cyr61
is indeed regulated through a CArG box located far upstream (
21
). Likewise, initial analyses suggested that
nur77
and
pip92
might be regulated through an SRF-independent mechanism, as there were no recognizable SRF-binding sites in their promoters. Detailed analysis revealed that activation of these
genes is mediated through `mutated' CArG sequences, significantly altered from
the consensus such that they are not expected to bind strongly to SRF (
22
,
23
).
As suggested by the CArG-box consensus sequence, CC(A/T)
6
GG, it was thought that the GC base pairs were invariant for SRF binding whereas
the ATs in the middle tolerated sequence diversity inasmuch as they were ATs (
4
). This notion was reinforced by the following observations:
(i) in a PCR-based selection experiment carried out with SRF and random oligonuleotides, only sequences that fit the CArG consensus were selected even
under low-stringency conditions (
24
); (ii) SRF directly contacts GG dinucleotides within the CArG box, and mutations in those contact points have
the most severe effects on the SRF/CArG interaction (
25
); (iii) the CArG boxes found in the promoters of several immediate-early genes all conform to the CArG consensus (
25
). It was thus surprising to us that deletion and linker scanning analyses of
the
pip92
and
nur77
promoters pinpointed `CArG-like' sequences, which did not fit the consensus, as being essential for
activation (
22
,
23
). Both sequences deviate from the CArG consensus significantly:
NIH 3T3 cells were grown and transiently transfected with the reference plasmid
PGK[beta]gal and a test plasmid as described (
23
). Cells were rinsed in phosphate buffered saline (PBS) 1 day following
transfection and allowed to recover in growth medium for one day. Cells were
then switched to medium containing 0.5% calf serum. Two days later, cells were
either stimulated with 20% calf serum for 3 h or were left unstimulated
(quiescent). CAT and [beta]-galactosidase assays were performed as described (
28
). At least three sets of transfections for each construct were analyzed.
Sequences of oligonucleotides used are as follows: CE, derived from the c-
fos
promoter CArG box), has the sequence gatcCTACCGCCAACCGGAATAGTCCATATAAGGACTCg;
mCE, the
pip92
CArG-like sequence and a proximal Ets site (same as `CP' in ref.
23
). The mmCE oligonucleotide differs from mCE by a change in CArG-like site: TCTTATATG
All probes were double-stranded oligonuleotides described above, except that the
nur77
126/72 fragment was excised from the corresponding tk CAT construct (
22
). All probes were labeled by DNA polymerase Klenow fragment fill-in reaction (
28
).
In vitro
transcription-translation, NIH 3T3 cell nuclear extract preparation, binding reactions
and electrophoretic analysis were carried out as described (
23
). The efficiency of translation of Elk-1 and SRF was approximately equal as determined by
35
S-Met incorporation, followed by the SDS-PAGE analysis (data not shown).
That the
nur77
and
pip92
SREs can form ternary complexes suggested to us that perhaps ternary complex
formation need not necessarily occur by SRF recruitment of Elk-1 as in the case of c-
fos
. Specifically, we proposed that in SREs where the SRF-binding sites (CArG boxes) are weak but the Est-binding sites are strong, Elk-1 might be able to recruit SRF to form the ternary complex. To
test this hypothesis, we designed a series of oligonucleotides with
progressively altered CArG-boxes and a neighboring Ets-binding site (nomenclature for the oligonucleotides: `C', high-affinity, consensus CArG box: CC(A/T)
6
GG; `E', consensus Ets core binding site: CCGGAA; `m', a deviation from these
sequences; see Fig.
2
D). In this set of oligonucleotides, CE carries the sequence of the c-
fos
SRE whereas mCE carries the sequence of the
pip92
SRE. Thus mCE has a high-affinity Ets-site and CArG-like sequence with 1 bp deviation from the consensus; mmCE has
two significant deviation from the CArG consensus (Fig.
2
D). We compared the abilities of the mCE, CE and mmCE oligonucleotides to
interact with Elk-1 and SRF and to form the ternary complex. As expected, CE (c-
fos
SRE) binds SRF strongly; while there is no detectable binding to Elk-1, CE also supports formation of the ternary complex in the presence of
SRF and Elk-1 (Fig.
2
A). Similarly, mCE binds Elk-1 and SRF independently and simultaneously, thus forming the ternary
complex. mmCE, however, is unable to bind SRF but is able to bind Elk-1. This is expected since mmCE contains a high-affinity Est-binding site but a CArG-like site with two significant deviations from the
consensus. Strikingly, mmCE is still able to form the ternary complex in the
presence of both Elk-1 and SRF under the same conditions (Fig.
2
A). These results strongly suggest that Elk-1 is able to direct SRF to form a ternary complex. Thus, formation of the
SRF-DNA binary complex may not be an obligatory prerequisite for ternary
complex assembly.
The above results show that oligonucleotides that do not exhibit detectable
binding to SRF alone can nevertheless support the formation of the ternary
complex in the presence of Elk-1, given that a strong Ets-binding site exists in the vicinity of the low affinity SRF-binding site. To test the effect of lowering the binding
affinity of the Ets-site, we introduced a mutation into the Ets-site of mmCE to create the mmCmE probe (Fig.
2
D). Weak Elk-1-DNA binary complex could be detected for mmCmE (Fig.
2
B) only when a relatively high amount of Elk-1 was used in the EMSA (2 [mu]l of translation product used in Figure
2
B for mmCmE compared with 0.5 [mu]l used in Figure
2
A for mmCE). Moreover, in the presence of relatively high amounts of both SRF
and Elk-1, a ternary complex can form (Fig.
2
B). Thus, although mmCmE does not bind SRF alone and binds Elk-1 with relatively low affinity, it can nevertheless form a ternary complex
in the presence of both SRF and Elk-1. This result provides further support for the notion that there is a
cooperativity between TCF and SRF in formation of a ternary complex (
33
).
A subfamily of proteins, the related-to-SRF (RSRF) proteins, shares considerable sequence homology with SRF
(
34
). However, the binding sites of SRF and RSRF appear to be distinct: SRF does not bind the RSRF consensus site, CTA(A/T
4
)TAG, and RSRF does not bind the CArG box (
34
). The RSRF consensus can be regarded as a subset of CArG boxes with two
nucleotide substitutions, i.e. C
Since the above experiments were carried out with SRF and Elk-1 proteins synthesized
in vitro
, we sought to determine whether similar protein binding patterns occur with NIH
3T3 cell nuclear proteins. The same probes used in Figure
2
A and B were incubated with nuclear extracts from NIH 3T3 cells, and the
resulting complexes were resolved by non-denaturing PAGE. Specific complexes were identified by their sensitivity
to competition by a consensus CArG box or an oligonucleotide with two high-affinity Ets sites. CE, mCE and mmCE all supported the formation of binary
and ternary complexes, consistent with interactions with SRF and an Elk-1-like protein. As shown in Figure
3
, complexes I, II and III correspond to DNA binding to an Elk-1-like protein (TCF), SRF and the ternary complex, respectively. CE (c-
fos
SRE) forms complexes II and III but not complex I, as expected. mCE (
pip92
SRE) forms both complexes I and II, as well as complex III, which corresponds
to the ternary complex that can be competed by either cold CArG or Ets-binding sequences (Fig.
3
). The mmCE probe formed complex I and III, but not complex II, consistent with
the mmCE binding to Elk-1 but not SRF (Figs
2
and
3
). The mmCmE probe formed a very faint, poorly reproducible ternary complex band
in some experiments where higher amount of nuclear extract was used (data not
shown). These results correlated well with the results obtained with
in vitro
translated SRF and Elk-1 (Fig.
2
).
To investigate whether formation of ternary complex
in vitro
correlates with serum responsiveness
in vivo
, we cloned the various Ets-CArG sites used in the Figure
2
A and B upstream of the herpes simplex virus thymidine kinase (tk) basal promoter driving the CAT reporter
gene. The resulting constructs were transiently transfected into NIH 3T3 cells,
and their ability to respond transcriptionally upon serum stimulation was
quantified using CAT assays. Both CE and mCE, which form the Elk-1-SRF-DNA ternary complex with the highest efficiency among the
sites tested in this study, supported strong serum inducibility of the tk CAT
reporter gene (Fig.
4
) (
23
). By contrast, mmCE and mmCmE, which form the Elk-1-SRF-DNA ternary complex with lower efficiency than mCE or CE,
mediated lower levels of serum responsiveness of the tk CAT reporter (Fig.
4
). The tk CAT vector alone provides only a low level of CAT activity that is
unaltered by serum stimulation (data not shown) (
22
,
23
). We note that mmCE and mmCmE, which do not support the SRF-DNA binary complex in our experiments but form the Elk-1-SRF-DNA ternary complex, mediate a weak serum-induced transcriptional response. This result
provides additional evidence that the ternary complex, rather than the SRF-DNA complex alone, is the functional serum-responsive complex (
23
).
Recent studies on the transcriptional activation of immediate-early genes have focused on the importance of a ternary complex, comprised
of SRF and TCF bound to SRE, as the integral part of the serum-responsive transcriptional apparatus (
16
). Since the importance of the ternary complex was first recognized in c-
fos
, where SRF binds strongly to and TCF is recruited by SRF (
35
), TCF has been regarded as an `accessory protein' subservient to SRF (
36
).
Results presented in this report suggest that given the appropriate SRE
sequence, TCF can also recruit SRF. A simple explanation for the prominence of
SRF in c-
fos
regulation is the presence of a good CArG box that binds SRF strongly. At the
same time, the c-
fos
Ets-binding site (CAGGAT) is weak, being quite different from the CCGGAA high-affinity binding consensus (
37
). In the
pip92
and
nur77
promoter, where the Ets-binding sites (CCGGAA) are strong and the SRF-binding CArG sites are weak, SRF and Elk-1 can bind independently but together they bind
synergistically (Figs
1
and
2
A). When the CArG box is further weakened by mutation, then Elk1 can apparently
recruit SRF to form a ternary complex (Fig.
2
). These results indicate that the interaction between SRF and Elk-1 allows them to bind a broader range of DNA sequences than previously
thought.
In this report we have shown that Elk-1 can recruit SRF to Ets-CArG elements that do not form detectable SRF-DNA binary complex (Fig.
2
), a situation opposite from the one at the c-
fos
SRE, where SRF recruits TCF (
10
). In addition, we have shown that the SREs of two immediate-early genes,
pip92
and
nur77
, interact with Elk-1 and SRF individually as well as in the SRF-Elk-1-DNA ternary complex (Figs
1
and
2
A) (
23
). The cooperativity of TCF-SRF interaction (
33
) most likely allows a great sequence variability of composite Ets-CArG elements in the ternary complex.
The three DNA elements tested in this study, mmCE, mmCmE and
nur77
-274/-249, form a SRF-Elk-1-DNA ternary complex with a low efficiency (Fig.
2
). The mmCE and mmCmE elements support only a low level of serum-induced transcription (Fig.
4
).
nur77
-278/-174 promoter region, which includes -274/-249 fragment that forms a Elk-1-SRF-DNA ternary complex (Fig.
2
C), acts as a weak SRE (
22
). Such weak elements are unlikely to operate alone as SREs, but may act
together with other sites present in the promoters of immediate-early genes. In the case of
nur77
promoter, distal Ets-CArG (RSRF) element at -274/-249 may interact with an Ets-CArG element at -126/-72, which forms ternary complex with a
high efficiency (Fig.
1
) and is a strong SRE (
22
). Multiple Ets-CArG (or CArG-like) sites also exist in the promoters of immediate-early genes
zif268
(
38
),
krox20
(
39
), [beta]-actin (
40
) and
pip92
(
23
).
Our results show that an RSRF site in the
nur77
promoter together with an adjacent Ets site supports formation of the SRF-Elk-1-DNA ternary complex (Fig.
2
C). Even though both RSRF and SRF bind distinct high-affinity sites (
34
), this result shows that they can bind to common (RSRF) CArG-Ets composite sites. Thus, both RSRF and SRF may be able to interact with
Ets-site binding factors and form functional ternary complexes.
In this report we have studied an SRF-Elk-1-DNA ternary complex. However, growing evidence suggests that
SRF may interact with transcription factors other then those from the Ets
family. For example, in the promoter of interleukin-2 [alpha] receptor SRF site functions together with a NF[kappa]B binding site (
41
). Recently, Treisman and colleagues showed that in NIH 3T3 cells c-
fos
SRE can retain serum responsiveness in the absence of an Ets site; they suggest
that SRF-CArG complex is likely to interact with yet unidentified factor (
30
). This interpretation is in agreement with our results from this and previous
report (
23
) that suggest that SRF-CArG binary complex is not sufficient for serum responsiveness. These
results, taken together with numerous other studies on the role of SRF in serum-induced and muscle-specific transcription, indicate the function of SRF sites depend on
the promoter context to a large extent.
Even though a relatively large subset of immediate-early genes may be regulated by a SRF-TCF-DNA ternary complex, this does not imply uniformity of
regulation of those genes. The sequence variability of distinct Ets-CArG sites may contribute to the diversity of regulation of distinct genes
that carry these sites. For example, those genes with strong Ets and weak CArG-like sites may function in a SRF-TCF mode in one cellular background, and in a SRF-independent mode, through an interaction with any of the Ets
family members, in a different cellular background. If AP-1 and SP1 sites are present in promoters of those genes with strong Ets
sites, Ets-AP-1 (
42
) and/or Ets-SP1 (
43
) composite elements may form and function. Widely different Ets-CArG sites may act as sensors of concentrations of active SRF and Ets
family members, thereby leading to the differential regulation of genes in
whose promoters they operate.
We are grateful to Dr R. Treisman for gifts of SRF and Elk-1 cDNAs. This work was supported by a grant from the National Institutes
of Health (R01 CA52220-06). L.F.L. is an Established Investigator of the American Heart
Association.
+
Present address: Division of Developmental Biology, National Institute for
Medical Research Mill Hill, London NW7 1AA, UK
REFERENCES
Return