ABSTRACT
The expression of human small nuclear U2 RNA genes is controlled by the proximal
sequence element (PSE), which determines the start site of transcription, and a
distal sequence element (DSE). The DSE contains an octamer element and three
Sp1 binding sites. The octamer, like the PSE, is essential for U2
transcription. The Sp1 sites contribute to full promoter activity by distance-dependent cooperative interactions with the transcription factors Sp1 and
Oct-1. Here we show that purified recombinant Sp1 and Oct-1 bind cooperatively to the DSE and that they physically interact
in vitro
. Furthermore, we show that Sp1 and Oct-1 interact
in vivo
using a yeast two-hybrid system. The domain of Sp1 which interacts with Oct-1 is confined to the region necessary for transcriptional
stimulation of U2 RNA transcription. This region contains the glutamine-rich activation domain B and a serine/threonine-rich part. The results demonstrate that Sp1, in addition to binding
to a number of other factors, also interacts directly with transcription factor
Oct-1.
Transcription of small nuclear RNA (snRNA) genes by RNA polymerase II is
dependent on the proximal sequence element (PSE) centred around position -55 and the distal sequence element (DSE) approximately at position -220 (for reviews see
1
-
4
). The PSE determines the start site of transcription (
1
-
4
) and may be required for 3'-end formation (
5
-
8
). A multi-subunit complex (e.g. PBP, PTF and SNAP
C
) needed for transcription of both RNA polymerase II- and polymerase III-transcribed snRNA genes binds to PSE
in vitro
(
9
-
13
). The DSE regulates the level of transcription and, like all DSEs of U snRNA
genes, the DSE of human U2 snRNA genes contains an essential octamer element,
which binds transcription factor Oct-1 and other members of the POU family of homeodomain proteins (
14
-
17
). In addition to Oct-1, transcription factor Sp1 binds to the DSE of human U2 genes
in vitro
and deletion of these sites leads to an 80% reduction in U2 transcription (
14
,
16
,
17
). Experiments in
Xenopus
oocytes suggested that DSE binding factors could function by stabilizing the
formation of transcription complexes at the U2 promoter (
18
), and in agreement with this it has been shown that Oct-1 potentiates binding of PTF to the PSE
in vitro
(
10
). A combination of transient expression analysis and
in vitro
binding studies using crude nuclear extracts have revealed distance-dependent cooperative interactions with the Sp1 and Oct-1 factors at the DSE of human U2 genes (
19
). Analysis of chimeric proteins has revealed U2 promoter-specific activation domains of Oct-1 (
20
) and we have recently shown that cooperative stimulation of human U2 snRNA
transcription requires a region of Sp1 that includes a serine/threonine-rich part in addition to the glutamine-rich activation domain B (
21
).
Here we show that purified recombinant Sp1 and Oct-1 factors bind DNA cooperatively and that they interact physically
in vitro
. We also show that Sp1 and Oct-1 interact
in vivo
in a yeast two-hybrid assay. Furthermore, the domain of Sp1 that interacts with Oct-1 was mapped to a region necessary for transcriptional stimulation,
strongly suggesting that the interaction between Sp1 and Oct-1 is important for U2 snRNA gene transcription.
The Oct-1 cDNA, PCR amplified from plasmid pBS/Oct-1 (
22
), was cloned C-terminally of the histidine-tag in pET-19b (Novagen;
23
,
24
). His/Oct-1 was expressed in
Escherichia coli
BL21 pLysS at 20oC for 3 h and the extract from 1 l of culture was passed twice over a Ni
2+
-NTA (Qiagen) resin column essentially as described (
25
). His/Oct-1 eluted with 150 mM imidazole (2.5 ml) was dialyzed into a buffer (20 mM
HEPES-KOH, pH 7.9, 20% glycerol, 0.1 M KCl, 2 mM MgCl
2
) and used directly in the experiments. A control extract was prepared from
untransformed bacteria in the same way. The GST/Sp1 expression vector, which
contains the complete Sp1 cDNA (kindly provided by Dr E.Wintersberger, Vienna),
was transformed into
E.coli
DH5[alpha]. Cells were grown, extract prepared and GST/Sp1 proteins purified on
glutathione-Sepharose (Pharmacia) as described (
26
). GST (glutathione S-transferase) was prepared in the same way. The His/Oct-1 and GST/Sp1 preparations were analyzed on SDS-PAGE and major His/Oct-1 and GST/Sp1 polypeptides of the correct sizes were
detected by Coomassie brilliant blue staining, together with minor degradation
products, and by Western blotting with anti-Oct-1 and anti-Sp1 antisera (Santa Cruz Biotechnology Inc.).
For the binding reaction 4 ng Sp1 (Promega) and 4 [mu]l His/Oct-1 extract were preincubated with 4 [mu]g poly(dI[middot]dC) (Pharmacia) for 10 min at room temperature in a buffer
containing 10 mM HEPES-KOH, pH 7.9, 25% glycerol, 50 mM KCl, 5 mM MgCl
2
, 0.6 mM DTT and 0.25% non-fat dried milk. An aliquot of 2 fmol of the
32
P-labelled probe (Sp1-10-Low Octa), containing a Sp1 binding site separated by 10 bp
from a low affinity octamer site (
19
), was thereafter added and the incubation was continued for another 20 min at
room temperature. The total reaction volume was 20 [mu]l. In the competition experiments 1.5 pmol specific competitor DNA,
containing an octamer element or two Sp1 binding sites, was added. The reaction
mixtures were separated on a 4% polyacrylamide (29:1) gel in 0.25* TEB buffer at 10 V/cm for 75 min (
27
). The saturating Sp1 binding study was done in the same way except that the
probe was incubated with or without saturating amounts of Sp1 (30 ng) and with
increasing amounts of His/Oct-1 extract. The amounts of probe and shifted complexes were quantified by
PhosphorImager analysis (Molecular Dynamics).
The purified His/Oct-1, GST/Sp1 and GST proteins and the control extract were coupled to tosyl-activated magnetic beads (Dynabeads; Dynal) essentially as described
by the manufacturer. Thereafter, the beads were incubated with bovine serum
albumin to saturation. The coupled beads were stored in PBS, 0.1% BSA at 4oC. For one interaction assay 5 [mu]l coated beads were washed twice in 100 [mu]l binding buffer (25 mM Tris-HCl, pH 7.9, 10% glycerol, 0.2 M KCl, 0.1% NP-40 and 1 mM DTT). The beads were then incubated with
the various proteins in 50 [mu]l binding buffer for 90 min at room temperature with occasional gentle shaking. After
extensive washing in the binding buffer (6 * 0.5 ml) protein loading buffer was added to the beads, the samples were boiled
and proteins were separated on SDS-PAGE. The
in vitro
transcribed/translated
35
S-labeled proteins were detected by autoradiography. Equimolar amounts of
in vitro
translated Sp1 and GAL/Sp1 fusions and of Oct-1 and the Oct-1 POU domain proteins were used in the binding reactions. Aliquots
of 30 ng HeLa cell Sp1 (Promega) were used in the binding reactions with the
His/Oct-1 beads and with the control beads. Western blotting analysis was
performed by standard methods using anti-Sp1 antisera (Santa Cruz Biotechnology Inc.) and the ECL detection kit
(Amersham). Ethidium bromide (100 [mu]g/ml) was used (
28
) to show that binding of
35
S-labelled C482 to His/Oct-1 beads was a DNA-independent protein association (data not shown). To produce
the
35
S-labelled proteins a Sp1 expression vector (amino acids 83-778) and the various GAL/Sp1 constructs (
21
), which all contain a T7 RNA polymerase promoter, were
in vitro
transcribed/translated using the TNT coupled reticulocyte lysate system
(Promega). Two major translation products were obtained from the fusion
plasmids. Since both bands were detected by Western blotting with anti-GAL4 antisera (data not shown), the shorter one is the result of abortive
translation.
35
S-Labelled Oct-1 and Oct-1 POU domain (amino acids 296-455 of Oct-1) were produced from plasmids pBGO-Oct-1 and pBGO-ATG-POU1, respectively (
29
).
The interaction trap assay, a yeast two-hybrid system, was used as described by Gyuris
et al
. (
30
) and Paroush
et al
. (
31
). Yeast strain EGY48, with an integrated
LEU2
reporter gene and upstream
LexA
operators, was transformed with pSH18-34. This plasmid contains the
LexA
op
-
lacZ
reporter. This reporter strain was then transformed with LexA/Sp1(231-485) and/or B42/Oct-1(1-369). The LexA/Sp1(231-485) plasmid was constructed by ligating the
Eco
RI fragment from GAL/Sp1 (231-485) (
21
) C-terminally of
LexA
(1-202) in pEG202. The 1.2 kbp
Eco
RI fragment from plasmid pBS/Oct-1 (
22
), encoding amino acids 1-369 of Oct-1, was cloned into the yeast expression vector pJG4-5 (
30
). This vector allows galactose-dependent expression of Oct-1(1-369) as a fusion protein with N-terminal sequences consisting of a nuclear localization
signal, a transcription activation domain (B42) and the haemaglutinin epitope
tag. Galactose-dependent LEU
+
colonies were picked and grown on glucose X-gal Ura
-
His
-
Trp
-
plates and on galactose X-gal Ura
-
His
-
Trp
-
plates. This selects for the
LexA
op
-
lacZ
reporter, LexA/Sp1(231-485) and B42/Oct-1(1-369) plasmids, respectively. Blue colonies appeared only on
X-gal plates with galactose. Cells were grown in liquid culture with glucose
and galactose medium and extracts were prepared using the reporter lysis buffer
as described by the manufacturer (Promega). [beta]-Galactosidase activity was determined in a luminometer using a
chemiluminescent [beta]-galactosidase system as described by the manufacturer (Clontech).
Electrophoretic mobility shift analysis was performed to see whether purified
recombinant Sp1 and Oct-1 factors bind to the distal sequence element of human U2 snRNA genes in a
cooperative way. His-tagged Oct-1 (His/Oct-1) purified from
E.coli
and Sp1 purified from vaccinia virus-infected HeLa cells were incubated with the probe Sp1-10-Low Octa. This probe contains one Sp1 site separated by 10
bp from a low affinity binding site for Oct-1 and has been used to demonstrate cooperative binding in crude nuclear
HeLa cell extracts (
19
). His/Oct-1 bound weakly (complex A) and Sp1 bound strongly (complex B) to the
probe, as expected (Fig.
1
A, lanes 1 and 2). Addition of both factors resulted in a prominent retarded
complex (Fig.
1
C, lane 3). This complex was not seen when Sp1- or Oct-1-specific competitor DNAs were added (lanes 4 and 5), showing
that complex C consists of templates to which both Oct-1 and Sp1 have bound. These results demonstrate cooperativity in binding to DNA, since the fraction of templates with Oct-1 bound to the low affinity site is increased in the presence of Sp1
(compare lanes 1 and 3). Mobility shift experiments with or without a
saturating concentration of Sp1 and small increments of Oct-1 concentrations (Fig.
1
B) showed that the fraction of complex C (Sp1 + Oct-1) rapidly increased compared with the fraction of complex A (Oct-1). We have also found that a GAL/Sp1 fusion (C590, see Fig.
5
) bound DNA cooperatively together with His/Oct-1 (data not shown). In this experiment the probe UAS-10-Low Octa was used (
21
). This suggests that the DNA binding domain of Sp1 is not involved. We conclude
from these experiments that purified Sp1 and Oct-1 factors bind DNA cooperatively.
Cooperativity in binding to DNA can be achieved through a direct interaction
between the two proteins or could be a result of structural changes in one of
the binding sites, induced by binding of one factor, leading to increased
binding of the second factor. To test whether the Sp1 and Oct-1 factors interact without DNA, His/Oct-1 was coupled to magnetic beads and incubated with Sp1 from HeLa
cells or with
35
S-labelled Sp1 synthesized by
in vitro
translation (Fig.
2
). Both protein preparations bound to the Oct-1 beads.
35
S-Labelled Sp1 did not bind to the control beads but Sp1 from HeLa cells
showed some binding (Fig.
2
A and B). The reverse experiment was also performed (Fig.
2
C). A GST/Sp1 protein preparation was coupled to magnetic beads and incubated
with
35
S-labelled Oct-1 or with the Oct-1 POU domain. Oct-1, but not the Oct-1 POU domain, bound to the GST/Sp1 beads. Neither
protein preparation bound to control beads coupled with GST protein (Fig.
2
C). GST/Sp1 could bind DNA cooperatively together with His/Oct-1, using the mobility shift assay described above (data not shown). We
conclude from these experiments that the Sp1 and Oct-1 factors interact directly and that this interaction contributes to the
cooperativity in binding to DNA.
Equimolar amounts of
35
S-labelled GAL/Sp1 fusion proteins were tested for their ability to interact
in vitro
with His/Oct-1 beads and with control beads as described above (Fig.
3
). Fusions C590, C482, C406, N231, 231-485 and 304-485 bound to the Oct-1 beads and not to control beads (see Fig.
5
for a map of the GAL/Sp1 fusions and a summary of the results). The GAL4 DNA
binding domain and fusions C350, C262, 263-437, 368-485 showed no binding in this assay (Fig.
3
). The binding of fusion C482 to Oct-1 was found to be resistant to 100 [mu]g/ml ethidium bromide (
28
), indicating that contaminating DNA was not involved (data not shown). Fusion
C482 bound most efficiently to Oct-1, C406 bound weakly and C350 not at all. This suggests that efficient
binding to Oct-1 requires a region between amino acids 350 and 482 of Sp1. Since fusions
231-485 and 304-485, but not 368-485, bound to Oct-1, we conclude that the shortest region of Sp1 able
to bind directly to Oct-1 contains amino acids 304-485.
We have used the interaction trap (
30
,
31
), a yeast two-hybrid system (
32
), to study the interaction between Sp1 and Oct-1. In these experiments, physical association between Sp1(231-485) fused to LexA and the N-terminal part of Oct-1 (1-369) fused to a transcriptional activation
domain (B42) was analyzed (Fig.
4
A). Transcription of the two reporter genes,
LEU2
and
lacZ
, that contain upstream LexA binding sites was measured and only yeast
transformed with both the LexA/Sp1(231-485) and B42/Oct-1(1-369) plasmids were found to be galactose-dependent LEU
+
and galactose-dependent blue on X-gal indicator plates (Fig.
4
B). We conclude from these results that Sp1 and Oct-1 interact
in vivo
. The part of Oct-1 that interacts with Sp1 contains the POU-specific domain and glutamine-rich regions (
22
). Interaction between Sp1(231-485) and a B42/Oct-1 fusion containing the complete open reading frame of Oct-1 was not detected (data not shown).
We have previously analyzed the ability of GAL/Sp1 fusion proteins to
cooperatively stimulate U2 snRNA transcription together with Oct-1, using a transient expression assay in COS7 cells (
21
). These experiments showed that a region of Sp1 containing the glutamine-rich activation domain B together with an N-terminally located serine/threonine-rich part (amino acids 231-485) was sufficient for stimulation of U2 gene
transcription. In addition, we found that a GAL/Sp1 fusion with an N-terminal truncation of this region (304-485) also stimulated transcription. Figure
5
summerizes the activation data (
21
) and the results presented in Figure
3
. The results show a correlation between transcriptional stimulatory activity
in vivo
and binding of Oct-1 and GAL/Sp1 fusions
in vitro
. Thus, the part of Sp1 that interacts with Oct-1 is located in a region necessary for transcriptional stimulation of U2
RNA transcription. Moreover, the demonstration that Sp1 and Oct-1 also interact in yeast strengthens the conclusion that this interaction
is functionally relevant.
The RNA polymerase II-dependent human U2 snRNA genes are tandemly repeated, ubiquitously
expressed and have a promoter that activates transcription to very high rates,
with a minimum of one transcript/gene/2-4 s (
33
,
34
). The DSE contains an octamer element and three Sp1 binding sites. The octamer
element, like the PSE, is essential for U2 transcription and the Sp1 sites
contribute to full promoter activity, by distance-dependent cooperative interactions with the Sp1 and Oct-1 factors (
19
).
We have previously analyzed Sp1 activating functions at the U2 snRNA promoter
and found that the glutamine-rich activation domains A and B, which both strongly stimulate a TATA box
promoter, are not sufficient for U2 gene activation. Stimulation of U2
transcription requires a region between amino acid residues 231 and 485 of Sp1,
which contains a serine/threonine-rich part in addition to glutamine-rich activation domain B (
21
). The results described here show that Sp1 contains an Oct-1 interaction domain located in the part of Sp1 required for U2 gene
transcription. Thus, in addition to interacting with factors, such as dTAF
II
110, the initiator element binding factor YY1, TBP, E1a and hTAF
II
55 (
35
-
40
), Sp1 also binds to the ubiquitously expressed transcription factor Oct-1.
The
Xenopus laevis
U2 enhancer contains Oct-1 and Sp1 sites and experiments in oocytes showed that the enhancer
promotes the formation of stable transcription complexes (
18
,
41
). It is conceivable that Sp1 and Oct-1 also participate in the formation of stable transcription complexes at
the U2 enhancer in human cells. Several multicomponent PSE binding activities
have been reported, e.g. PBP, PTF and SNAP
C
(
9
-
13
). Both PTF and SNAP
C
contain four polypeptides of similar size and so far one of the subunits has
been shown to be identical in the two complexes. However, TBP (TATA box binding
protein) was reported to be a part of SNAP
C
but only loosely associated with PTF (
42
). Interestingly, it has been found that Oct-1 potentiates binding of PTF to different PSE elements
in vitro
(
10
), suggesting a role for Oct-1 in the formation of stable transcription complexes. Potentiation of PTF
binding required only the POU domain of Oct-1 (
10
), a region not sufficient for the interaction with Sp1 (Fig.
2
C). Analysis of Oct-1 activation function at U2 snRNA and mRNA promoters showed that Oct-1 contains redundant U2 promoter-specific activation domains. These were found in the N- and C-terminal parts of Oct-1, not including the POU domain (
20
). Thus, from these experiments it seems that the POU domain is not necessary
for activation of U2 transcription, although it is possible that the Pit-1 POU domain, which replaced the Oct-1 POU domain in these experiments, fulfils this function. On the
other hand, the
in vitro
potentiation of PTF binding by Oct-1 may not be involved in transcription of all snRNA genes. It is not known
yet if the U2 promoter-specific activation domains of Oct-1 are involved in the functional interaction with the Sp1 factor
that we describe here, since those experiments were performed using a reporter
gene with multiple octamer sites and without Sp1 binding sites (
20
). Interestingly, we have found that the N-terminal 369 amino acids of Oct-1 interact with Sp1 in a yeast two-hybrid assay. We are currently investigating the role of this
part of Oct-1 in the regulation of U2 snRNA gene transcription. It is not clear
whether the function of Oct-1 is to stabilize the PTF/SNAP
C
-DNA complex and/or if Oct-1 has activation domains that contact other components of the basic
transcription machinery.
Sp1 and Oct-1 bind DNA cooperatively in crude nuclear extracts (
19
). This is not, however, dependent on other activities in the nuclear extract,
since we have demonstrated that purified recombinant Sp1 and Oct-1 factors also bind DNA cooperatively. Furthermore, we show that the two
proteins interact directly, without DNA. The Oct-1 interaction domain is located between amino acids 304 and 485 of Sp1.
Fusion C406, with amino acids 83-406 of Sp1, binds Oct-1 weakly, suggesting either that the region between 304 and 406 is
sufficient for the interaction or that the N-terminal part of this fusion also contributes to binding. Our results
suggest that the binding of Sp1 to Oct-1 is relevant for U2 transcription in the living cell, since we find a
correlation between binding and transcriptional activation. The Oct-1 interaction domain of Sp1(304-485) contains the glutamine-rich activation domain B (368-485) and a serine/threonine-rich part (see Fig.
5
). This is a different region of Sp1 to that involved in binding to the human
YY1 factor, the adenovirus E1a factor and hTAF
II
55, which all require the C-terminal part of Sp1, including the DNA binding domain (
35
,
38
,
39
). Several functions of transcription factor Sp1 at the U2 snRNA promoter could
be envisaged. The Oct-1 interaction domain of Sp1 could stabilize binding of Oct-1 to the DSE and thereby stabilize the formation of transcription
complexes at the U2 promoter. It is also possible that parts of Sp1 outside or
overlapping with the Oct-1 interaction domain make contacts with the PSE binding complex or with
other factors which are involved in transcription initiation of U2 snRNA genes.
Since the transcriptional control elements coincide with elements involved in
human U2 RNA 3'-end formation (
5
), there is a possibility that the Sp1 and Oct-1 factors participate in this process as well.
We are grateful to Roger Brent, Winship Herr, Patrick Matthias and Erhard
Wintersberger for the generous gift of plasmid constructs. We thank Cathrine
Phillips for linguistic revision and Ulf Pettersson for critical reading of the
manuscript. This work was supported by grants from the Swedish Medical Research
Council and The Göran Gustafsson Foundation.
REFERENCES
Return