ABSTRACT
The binding of TBP (TFIID) to the TATA box has been considered to direct promoter recognition and pre-initiation complex formation because it is the first event leading to
basal transcription by RNA polymerase II. Here, we analyse the binding of yeast
TBP to a consensus TATAAA box and two point mutations, TAAAAA (inactive) and
TATATA (active). Despite the fact that the TAAAAA sequence does not support
transcription
in vitro
, yeast TBP binds the three sequences showing, in this sense, only a limited
sequence specificity. However, the TBP-TAAAAA complex cannot be recognised by other basal transcription factors,
in particular by TFIIB. DNase I footprinting patterns of the TBP-TAAAAA complex are different from those observed in functional TBP-TATA box complexes, indicating that, most likely, it is a
different spatial arrangement of the TBP-DNA complex that prevents formation of the TFIIB-TBP-TAAAAA complex, also seriously impairing entry of TFIIA to
the complex. DNA deformability of the A/T-rich sequences appears to be an important determinant in the formation of a productive TBP-TATA complex. These results indicate that the transcriptional competence of A/T-rich sequences is determined not only by TBP binding, but
also by the ability of other basal transcription factors to recognise the preformed TBP-DNA complexes.
Assembly of the RNA polymerase II transcription machinery onto a TATA box-containing promoter is known to follow a well-defined pathway that ultimately leads to the recruitment of RNA
polymerase II and transcription initiation (
1
). TFIID, through its TATA box binding protein (TBP) subunit, is the first
transcription factor to interact with the TATA box. Other basal transcription
factors, TFIIB and TFIIA, sequentially interact with this initial complex in
the formation of a complete pre-initiation complex (reviewed in
2
,
3
). As TBP is the only basal transcription factor with DNA binding specificity,
it has been widely accepted that TBP binding leads to promoter commitment. TBP
binds the TATA box consensus core sequence (TATAAA) with high affinity (
K
d
= 2-4 * 10
-9
M). However, the molecular basis of this specific recognition is still poorly
understood. TBP interacts with TATA boxes in the minor groove (
4
,
5
), where it is almost impossible to discriminate between many different A/T-rich sequences. Actually, depending on the promoter and the organism, many other sequences have
been found to bind TBP and be functionally active in transcription. Recently, the crystal structures of
Arabidopsis
(
6
) and yeast (
7
) TBPs and of their co-crystals with two different TATA boxes have been obtained (
8
,
9
). The crystals have revealed a very similar structure of a saddle in both
cases. From the co-crystal structures, TBPs were shown to interact on the concave surface of
the saddle with 8 bp of DNA encompassing the core TATA box; the interaction
takes place in the minor groove and, in good agreement with previously reported
results (
10
), produces a dramatic bending of the DNA without major distortions in the TBP
structure itself. The co-crystal structures show extensive interactions with the nucleic acid
backbone, which include salt bridges, water-mediated hydrogen bonds and van der Waals contacts with the sugar groups.
Interactions with the bases also include van der Waals contacts and the
formation of a few specific hydrogen bond interactions, occurring at the very center of the TATA box
sequence, involving the symmetrical and stereochemically equivalent groups O2 of thymine and N3 of adenine. Altogether, the interactions
observed in the co-crystals explain in part the lack of discrimination between A.T and T.A base pairs.
However, it has been clearly shown both
in vitro
and
in vivo
that some A/T point mutations in the TATA core sequence have strong effects on
transcriptional rates (
11
,
12
). In particular, replacement of T(-29) in the consensus TATA box sequence by an adenine is known to abolish
transcription almost completely.
In order to gain some insight on how the transcription machinery discriminates
between wild-type (active) and mutated (inactive) TATA boxes when only A/T changes are
introduced, we have studied the interaction of TBP with a set of artificial
basal promoters in which we introduced either a consensus TATAAA box or mutated
TAAAAA and TATATA sequences. Our results show that TBP binds to all three
sequences giving rise to complexes of similar stabilities, despite the TAAAAA
sequence not being able to direct transcription
in vitro
. A different structural organisation of the TBP-TAAAAA complex impairs formation of the TFIIB-TBP-, TFIIA-TBP- and TFIIA-TFIIB-TBP-TAAAAA complexes and
appears to be responsible for the observed lack of activity.
The WT, TA
5
and (TA)
3
oligonucleotides were synthesized by the phosphoroamidite method on an Applied
Biosystems synthesizer. The WT double-stranded oligonucleotide was inserted into a G-less[180] plasmid between the
Eco
RI and
Sst
I (made blunt by T4 DNA pol digestion) sites. Subsequently, TA
5
and (TA)
3
were inserted by replacement in the WT construct using the
Eco
RI and
Sst
I sites.
Recombinant his-tagged yeast TBP from the pET14b expression vector (Novagen) was overexpressed in
Escherichia coli
strain BL21(DE3) pLys E. Two hours after induction by addition of IPTG at up to
1 mM, cells were collected and lysed by sonication in lysis buffer (0.5 M NaCl,
20 mM HEPES, pH 7.9, 1 mM EDTA, 20 mM 2-mercaptoethanol, 0.1% NP-40, 1 mM PMSF, 20% glycerol; 30 ml/l culture). The homogenate was
clarified by centrifugation (100 000
g
, 1 h, 4oC), the supernatant loaded onto a 10 ml DEAE-Sephadex A25 column and the flow-through applied to a 1 ml Ni
2+
-NTA column pre-equilibrated in lysis buffer. After extensive washing with lysis buffer and buffer D (0.1 M KCl, 20 mM HEPES, pH 7.9, 0.2 mM EDTA, 0.5 mM DTT, 20%
glycerol, 0.1 mM PMSF), the column was eluted with buffer D containing 100 mM
imidazole. All steps were carried out at 4oC. The fractions containing highly purified his-tagged yTBP were determined by SDS-PAGE electrophoresis and stored at -80oC. The his-tag peptide was removed by thrombin
digestion, as suggested by the manufacturer (Novagen). The purity of the yTBP
used was >95%.
Recombinant non-tagged human TFIIB protein was expressed in
E.coli
BL21(DE3) and purified as described (
13
). The purity as assessed by SDS-PAGE was ~95%.
Human TFIIA was purified from nuclear extracts of HeLa cells exactly as
described (
14
).
Nuclear extracts were prepared from HeLa cells and fractionated as described (
15
). In our assays, a phosphocellulose 0.5 M fraction was used as a source of general transcription factors and RNA polymerase II
because this fraction does not support polymerase II transcription on its own
(see lane 1 in Fig.
4
C).
In vitro
transcription analysis was performed using 4-8 [mu]l (~10-20 [mu]g protein) of this fraction and 1 [mu]l (~100 ng protein) of recombinant yTBP. The
amounts of the three supercoiled templates used were carefully normalised by
restriction enzyme digestion and analysis by agarose gel electrophoresis.
Transcription reactions were carried out as described (
15
). Unrelated T7-transcribed RNAs were added to the samples prior to phenol extraction and
precipitation as recovery controls. Gels were dried and exposed to X-ray-sensitive film at -80oC with intensifying screens.
Reaction mixtures contained 25 mM HEPES, pH 7.9, 45 mM KCl, 2 mM spermidine, 0.1
mM EDTA, 0.025% NP-40, 5 mM MgCl
2
, 15% glycerol, 0.5 mM DTT and 20-120 ng poly(dG[middot]dC) and 50 ng acetylated BSA. Increasing amounts of the final
yTBP preparation were incubated, for 30-40 min at 30oC, with 0.5-1 ng
32
P-labeled double-stranded oligonucleotides in a final volume of 20 [mu]l. For simple yTBP shift analysis samples were directly loaded onto 5% polyacrylamide gels containing 25 mM Tris, 190 mM glycine, 5 mM MgCl
2
, 1 mM EDTA, 10% glycerol, 0.5 mM DTT; the running buffer was the same except that DTT was replaced by 5 mM 2-mercaptoethanol and glycerol was omitted (
16
). Quantification was performed by scanning and integration of the autoradiographs using a Molecular Dynamics laser densitometer.
Oligonucleotide competition experiments were performed under exactly the same conditions by prior mixing all of the reagents and
increasing amounts of cold oligonucleotides before yTBP addition. Formation of
the TFIIB-TBP-DNA, TFIIA-TBP-DNA and TFIIA-TFIIB-TBP-DNA complexes was assayed under
the same conditions of incubation and analysed on 4% polyacrylamide gels containing 25 mM Tris, 190 mM glycine, 0.5 mM DTT; the running buffer was the same except that DTT was
omitted (
17
). Gels were dried and exposed to X-ray-sensitive film at -80oC with intensifying screens.
DNase I footprinting analysis was performed using exactly the same conditions as
described for the simple yTBP EMSA. After incubation samples were digested with 1.5 [mu]l DNase I (5 * 10
-3
U) for exactly 60 s at 30oC. Reactions were stopped by addition of 200 [mu]l 0.1 M NaCl, 50 mM Tris-HCl, pH 8, 0.5% SDS, extracted with phenol/chloroform and
ethanol precipitated. G+A ladders were obtained by chemical sequencing.
Digestion products were analysed on denaturing 20% polyacrylamide gels and
directly exposed to X-ray-sensitive film at -80oC with intensifying screens. Quantitative analysis of
the results was performed by densitometric scanning of the autoradiographs
using a Molecular Dynamics laser densitometer. The intensities of the bands in
the protected region were integrated and referred to the intensity observed for
the same region in naked DNA of the corresponding sequence. The same region was considered for all three sequences, despite protection on TA
5
being weaker on the 5'-half.
The WT, TA
5
and (TA)
3
sequences were subcloned as blunt-end inserts at the unique
Sal
I site (made blunt) of pBend 2 (
18
) and checked by sequencing. Each construct was digested separately with
Bgl
II,
Pvu
II and
Bam
HI to produce a 175 bp band having the insert at 134, 87 and 35 bp from one of
the ends respectively and analysed on 30 cm long 7% polyacrylamide-TBE gels run at 150 V for 18 h at 4oC. Bands were visualized by ethidium bromide staining and photographed under UV light.
In order to study how TBP recognises TATA boxes, we have designed a set of
artificial basal promoters in which we have introduced either a consensus TATAAA box, a TAAAAA sequence [in which T(-29) of the consensus TATA box was replaced by an A], or a TATATA sequence
[in which A(-27) was replaced by a T] (abbreviated as WT, TA
5
and (TA)
3
respectively in Fig.
1
A). These sequences were fused to a G-less cassette and tested in an
in vitro
transcription assay dependent on added TBP (Fig.
1
B). As expected (
12
), TA
5
does not support basal transcription, whereas WT and (TA)
3
direct almost equivalent levels of transcription at the three template
concentrations tested when yeast TBP is added. TA
5
cannot direct transcription even in the presence of TFIID or in crude HeLa
nuclear extracts (results not shown). Thus, the T(-29) -> A change in TA
5
renders the promoter totally inactive in transcription, in agreement with
results obtained
in vivo
in yeast cells (
11
).
The interaction of yTBP with the three different A/T sequences indicated above
was studied by EMSA. As shown in Figure
2
A, the formation of specific TBP-DNA complexes was detected with all three sequences. Moreover, similar
amounts of complex were detected with the WT and TA
5
sequences (Fig.
2
A and C), indicating that the complexes formed with these sequences have a
similar relative stability. On the other hand, the amount of complex formed
with the (TA)
3
sequence is significantly higher (Fig.
2
A and C). Identical results were obtained using human TBP (not shown). These
results were further confirmed by oligonucleotide competition assays. In these
experiments binding of yTBP to the WT sequence was competed equally well by WT
or TA
5
and much better by (TA)
3
(Fig.
2
B). Complementary experiments using TA
5
or (TA)
3
against all three cold oligonucleotides gave the same results, whereas large
amounts of non-specific competitors [poly(dG[middot]dC)] had no effect (not shown).
The efficiency of formation of the next two intermediates in the assembly of a productive pre-initiation complex, the TFIIA-TBP-DNA and TFIIB-TBP-DNA complexes (
20
), has been determined for the three sequences described above.
In the case of the TA
5
sequence, formation of the TFIIB-TBP-DNA complex is very inefficient, while WT and (TA)
3
yielded very stable, roughly equivalent ternary complexes (Fig.
4
A). Remarkably, an initially equivalent binding to both WT and TA
5
results in very different efficiencies of TFIIB-TBP-DNA complex formation; on the other hand, clearly different
efficiencies of formation of the TBP-DNA complex on WT and (TA)
3
result in equivalent TFIIB-TBP-DNA complex formation. A similar situation was observed when
formation of the TFIIA-TBP-DNA complex was investigated. Figure
4
B shows that formation of the TFIIA-TBP-DNA complex is less efficient, by at least an order of magnitude,
on TA
5
than on WT or (TA)
3
(which are similar to each other). These results indicate that though TBP can
bind to the TA
5
sequence, the resulting TBP-TA
5
complex is not efficiently recognised by TFIIB or TFIIA, providing a much
better correlation with the transcriptional efficiencies of the sequences as
basal promoters (Fig.
1
B). Nevertheless, on the TA
5
sequence the relative efficiency of formation of the TFIIA-TBP-TA
5
complex is significantly higher than for the TFIIB-TBP-TA
5
complex (compare Fig.
4
A and B). These results indicate that loading of TFIIA onto the TBP-TA
5
complex is less strongly affected than TFIIB loading. TFIIA is known to
stabilise the interaction of TBP with the TATA box. It is therefore conceivable
that it might help stabilise the interaction of TFIIB with TBP on TA
5
and eventually bring it back to a normal level. This possibility was assessed by
studying formation of the TFIIA-TFIIB-TBP-DNA complex on the three sequences used. As shown in Figure
4
C, formation of this quaternary complex is seriously compromised in the case of
the TA
5
sequence. Again, formation of this complex on WT and (TA)
3
is very similar and efficient, as expected. However, formation of the TFIIB-TFIIA-TBP-DNA complex on the TA
5
sequence is still noticeable, though less efficient, indicating that TFIIB
loading is facilitated in the presence of TFIIA (compare Fig.
4
A and C). This result and the fact that the
in vitro
transcription system we use is basically devoid of TFIIA prompted us to check
whether, in the presence of TFIIA, TA
5
might recover some activity in transcription. Figure
4
D shows that addition of increasing amounts of TFIIA, TFIIB or both does not
result in increased TA
5
transcription, which is never above background in any case. On the other hand,
WT shows normal levels of transcription under all conditions except in the
presence of large amounts of TFIIA, where it is lower (Fig.
4
D, lane 4). Identical results were obtained using a complete nuclear HeLa
extract, heat inactivated to destroy endogenous TBP activity and complemented
as above (not shown). Therefore, the TA
5
sequence is transcriptionally inactive even in the presence of large amounts of
TFIIA and/or TFIIB.
The results presented above indicate that the lack of transcription from the
promoter having the TA
5
sequence should not only be attributed to a difference in its affinity for TBP,
but also to the inability of other basal transcription factors, in particular
TFIIB, to recognise the preformed TBP-TA
5
complex. As TBP interacts free in solution with both TFIIB and TFIIA (reviewed
in
3
), the lower efficiency observed in formation of the early steps of the pre-initiation complex on the TA
5
sequence is likely to reflect structural peculiarities of the initial TBP-TA
5
interaction. Actually, DNase I footprints of the TBP-TA
5
complex show some important differences with respect to the footprints obtained
with the WT and (TA)
3
sequences (Fig.
3
). A classical TBP footprint can be observed over the -34 to -21 region on the top strand and over -37 to -19 on the bottom strand with both WT and (TA)
3
probes (Fig.
3
A and B). However, in the case of the TA
5
sequence a striking lack of protection was observed at residues A(-29) and A(-30) (Fig.
3
A, indicated by asterisks in Fig.
3
D), situated in a central position on the top strand. As a consequence, the
footprint is split into two halves. Protection on the 5'-half is slightly weaker than on the 3'-half (shown schematically as a gradient of grey in
Fig.
3
D). Note that the central residues on the bottom strand are protected. The
addition of TFIIB does not result in any major change in the footprints (data
not shown;
21
,
22
). These results strongly suggest that the structure of the TBP-TA
5
complex is different.
DNA deformability is likely to be an important determinant of the interaction of
TBP with DNA because binding of TBP to a TATA box sequence induces a major conformational change in the DNA which results in the formation of two sharp kinks at either end of the TATA
sequence. Between the kinks the DNA is smoothly curved towards the major groove
and partially underwound. Depending upon the precise nucleotide sequence, A/T-rich sequences might have different structural characteristics (
23
) and, in particular, homoadenine runs are known to be intrinsically curved (
24
). Therefore, the differential structural organisation of the TBP-TA
5
complex could have arisen from the particular conformational properties of the
DNA substrate. Taking this possibility into consideration, a circular
permutation analysis of the three sequences studied here was performed. The TA
5
sequence was shown to have clear differences in electrophoretic mobility depending on the internal position of the TA
5
sequence (Fig.
5
, lanes 5-7). In comparison, (TA)
3
does not show any difference and WT shows only a minor change (compare lanes 8-10 and 2-4 respectively). These data show that the TA
5
sequence is intrinsically curved with an estimated angle of 19o.
An increasing number of DNA binding proteins with very little in common are
being found to interact with DNA through its minor groove. These include fairly
abundant nuclear proteins such as HMG1/2 (
25
,
26
), HMGI/Y (
27
), transcription factors of the HMG box family (
28
,
29
), basal transcriptional factors such as TBP (
4
,
5
), recombination proteins (
30
,
31
), etc. However, the molecular bases of the specific interactions of proteins
with DNA in the minor groove are poorly understood and, in fact, many of them
have loose sequence specificities. The reason for that has to be found in the
lack of information about sequence composition in the minor groove. This is
particularly evident in the case of A/T-rich sequences, where the thymine O2 and adenine N3 groups that can
participate in hydrogen bonding are stereochemically equivalent, making A.T and T.A base pairs difficult to discriminate. In this respect, TBP constitutes a particularly interesting and
well-studied case. TBP binds the TATA box consensus core motif TATAAA with high
affinity, but its sequence preferences are not yet completely understood (
8
,
9
). The protein has been shown to promote transcription from a large variety of
A/T-rich sequences, but some mutations of the TATA box consensus sequence,
especially a T -> A change at position -29, abolish transcription both
in vitro
and
in vivo
(
11
,
12
). Here we have addressed the question of how the transcription machinery
discriminates between active and inactive TATA sequences and at what level this
discrimination takes place. From our results, we conclude that there is no
direct correlation between TBP binding to TATA boxes and basal promoter
activity (Fig.
1
). TBP binds both the transcriptionally inactive TA
5
sequence and the active WT or (TA)
3
sequences, the affinity for the TA
5
sequence being only 5-fold lower than for WT and (TA)
3
(Fig.
3
). This slight decrease in affinity is unlikely to explain, on its own, the
complete lack of transcriptional activity of the TA
5
sequence. Indeed, similar differences in affinity have been observed between
otherwise functional TATA sequences, such as the AdMLP and the yeast CYC1-52 TATA boxes (
19
). Furthermore, the stabilities of the TBP-TA
5
and TBP-WT complexes are very similar, as observed by EMSA (Fig.
2
). The results reported here suggest that discrimination between functional and
non-functional promoters takes place at the level of formation of the next
intermediates. The TBP-TA
5
complex cannot efficiently interact with either TFIIB or TFIIA. In addition,
loading of RAP30 (the TFIIF small subunit), which is responsible for the
recruitment of RNA polymerase II, is undetectable on the TA
5
sequence, while it is normal on WT and (TA)
3
sequences (not shown). As a consequence of these poor interactions, formation
of the pre-initiation complex on the TA
5
sequence is blocked. Others (
32
) have reported a larger difference in stability of the complex formed by TBP
with a TA
6
sequence when compared with the TATA box of the AdML promoter. This difference
can be explained by the completely different sequence context in which our
study was performed and also by the fact that the AdMLP TATA box is one of the
strongest minimal promoters known. In this respect, note that TBP shows significant affinities for a broad range of different DNA sequences, varying from
2-4 * 10
-9
M for TATA boxes to 5 * 10
-6
M for non-specific average DNA sequences (
19
,
33
).
The inability of TFIIB and other basal transcription factors to recognise the
TBP-TA
5
complex likely arises from the different structure of the complex. Several
results indicate that the structural organisation of the TBP-TA
5
complex differs from the structure of the complexes formed with
transcriptionally active sequences. On the one hand, the DNase I footprinting
patterns of the TBP-TA
5
complex show important differences when compared with the patterns of complexes formed with the WT and (TA)
3
sequences (Fig.
3
). Basically, the central residues and the 5'-side are more sensitive to DNase I cleavage in the TBP-TA
5
complex than in the TBP-WT or TBP-(TA)
3
complexes. Consistent with this interpretation, the 5'-side also shows a higher reactivity with hydroxyl radicals in the
TBP-TA
5
complex than in the complexes formed with the functional sequences (not shown).
Finally, it was shown that a core version of yTBP hardly bends a closely
related TA
6
sequence (
32
). Altogether these results strongly indicate that the structure of the TBP-TA
5
complex is different. From our results it is difficult to determine the precise structural organisation of the TBP-TA
5
complex. However, a fundamental difference between the three sequences studied here relates to the intrinsically curved character of the TA
5
sequence (Fig.
5
). The curved character of the sequence is expected to interfere with the usual
path of the DNA underneath the protein saddle, so that binding of TBP to the TA
5
sequence may not be able to induce the conformational change in the DNA that
leads to formation of a productive complex. In fact, the high level of DNase I
digestion of the 5'-side in the TBP-TA
5
complex strongly suggests that the TBP-TA
5
interaction occurs principally on the 3'-half of the sequence, likely using one half of the TBP bipartite
structure (
34
). The weaker interaction with the 5'-side can likely be explained as the result of the unfavourable
angle generated by the DNA sequence itself around that position. These two
possible modes of interaction of TBP with A/T-rich sequences are schematically presented in Figure
6
.
We thank Danny Reinberg, Jean-Marc Egly and Jack Greenblatt for TFIIB, yTBP and RAP30 cDNAs respectively, Sam Gunderson and Marco Bianchi for the G-less[180] and pBend2 plasmids and Ramón Eritja for oligonucleotide synthesis. We are grateful to
Gemma Moll and Jenny Colom for excellent technical support and all other
members of the Department for scientific discussion and support. P.C. was
supported by a doctoral fellowship from the Ministerio de Educación y Ciencia and J.B. by a CSIC post-doctoral contract. The financial support of grants PB93-102 from the Spanish DGICYT and PL-932217 from the CEC is gratefully acknowledged. This
work was carried out within the framework of the `Centre de Referéncia en Biotecnologia' of the Generalitat de Catalunya.
+
Present address: Abteilung Molekulare Entwicklungsbiologie, MPI für Biophysikalische Chemie, D-37018 Göttingen, Germany
REFERENCES
Return