DNA binding specificity of proteins derived from alternatively spliced mouse p53
mRNAs
DNA binding specificity of proteins derived from alternatively spliced mouse p53 mRNAs
Zoe
Miner
and
Molly
Kulesz-Martin*
Roswell Park Cancer Institute, Department of Experimental Therapeutics, GCDC
Room 403, Elm and Carlton Streets,
Buffalo
, NY 14263,
USA
Received January 3, 1997;
Accepted February 18, 1997
ABSTRACT
The mouse p53 gene generates two alternative splice products encoding p53 protein and a naturally occurring protein (p53as) with changes at the C-terminus. In p53as the negative regulatory region for DNA binding and
PAb421 antibody binding site are replaced, and p53as is constitutively active
for sequence-specific DNA binding. Using the technique of randomized synthetic
oligonucleotide in cyclic amplification and selection of targets, we have found
that p53as and p53 proteins have the same DNA binding specificities but that
these specificities frequently diverge from the consensus of two copies of
PuPuPuCATGPyPyPy. The importance of tetranucleotide CATG was confirmed but
there was a less rigorous requirement for patterns of flanking or intervening
sequences. In particular, the three purines upstream and three pyrimidines downstream of CATG are not required for p53 or p53as binding, 29 or more intervening nucleotides are tolerated, and one CATG is sufficient where adjacent nucleotides contain a region of
homology with certain previously reported non-consensus p53 binding sequences. These results suggested further definition of the non-consensus motifs, and database searches with these uncovered
additional candidate genes for p53 protein binding. We conclude that p53as and
perhaps other activated forms of p53 exert their effects on the same genes and
that differential activities of p53 protein forms are not due to inherently
different sequence selectivities of DNA binding.
INTRODUCTION
The tumor suppressor gene p53 encodes a protein with multiple functional
domains, including an N-terminal region involved in transcriptional activation, a central hydrophobic sequence-specific DNA binding region and a C-terminal non-specific DNA binding region which can negatively
regulate specific DNA binding. Most DNA binding studies of p53 require
interaction with or modulation of a C-terminal inhibitory domain in p53. This is accomplished by addition of a C-terminal-specific antibody PAb421, which activates and shifts DNA-p53 complexes in electrophoretic mobility shift assays
(EMSAs). The major form of p53 also can be activated by modifications of the C-terminus including phosphorylation, interaction with dnaK or truncation of
the last 30 amino acids (
1
-
3
) or by non-specific binding of single-stranded DNA (
4
-
8
).
p53 functions as a transcription factor affecting the expression of genes
involved in DNA repair, control of the cell cycle and apoptosis. DNA sequences
which bind p53 and mediate its transcriptional regulation have been identified
in the upstream promoter regions and introns of specific genes (
9
-
17
) and by
in vitro
studies using either genomic DNA fragments (
18
-
20
) or randomized oligonucleotides (
21
,
22
). These studies have delineated a consensus p53 binding sequence consisting of two copies of 5'-PuPuPuC(A/T)(T/A)GPyPyPy-3' (
19
,
22
) which has been refined to two repeats of 5'-PuGPuCATGPyCPy-3' (
21
).
Our laboratory has detected an alternatively spliced form of wild type p53 (designated p53as) in normal mouse cells and tissues (
23
,
24
) and shown that
in vitro
produced p53as constitutively binds to a p53 consensus sequence in gel shift experiments (
6
,
25
,
26
). p53as has 17 different amino acids at the C-terminus and is nine amino acids shorter than p53, resulting in losses of
the PAb421 epitope, two of three nuclear localization signals, a 5.8S rRNA
binding site and a casein kinase II phosphorylation site. Although the N-terminus and the central domain, which is responsible for sequence-specific interaction with DNA, are identical in the two proteins,
other reports suggest that the C-terminus may be capable of modulating the specific DNA binding of p53 (
1
,
2
). p53as is preferentially associated with the G
2
phase of the cell cycle (
24
) while the p53 protein is associated with a G
1
arrest thought to be mediated by transcriptional activation of downstream genes
(
27
,
28
). In addition, p53as responds to DNA damage with different kinetics compared to
the major p53 form (Wu,Y. and Kulesz-Martin,M., unpublished).
These differences in properties and cell cycle expression of p53 and p53as
suggest different functional specialization in cells. They could be explained
by differences in the regulation of sequence-specific DNA binding due to the differences in their C-termini. Alternatively, p53as could have different DNA sequence
specificity than p53 and exert its effects through differential transcriptional
modulation of downstream genes. In order to test the hypothesis that the DNA
sequence specificity of p53as and p53 are different, with implications for
widening the search for specific p53as target genes, we sought p53as binding
motifs using randomized synthetic oligonucleotides and the technique of cyclic amplification and selection of targets (CASTing) (
29
,
30
).
MATERIALS AND METHODS
Antibodies
Mouse monoclonal BC4-17 antibody and rabbit polyclonal ApAs antibody were generated against the
unique 17 amino acids of p53as (
24
) and are available from Oncogene Research Products as Ab-9 and Ab-10 respectively. PAb421 (Oncogene Research Products) recognizes an
epitope between amino acids 370 and 378 in the major form of p53 and absent in
p53as.
In vitro
transcription and translation
Plasmids and conditions used in these reactions were described previously (
25
).
CASTing
This procedure was essentially carried out as described (
29
,
30
) according to the following steps.
Polymerization of double-stranded DNA.
A 96mer ssDNA was synthesized (Biopolymer Laboratory, Roswell Park Cancer
Institute) beginning with an 18 nt primer site (5'-ATACCAGCTTATTCAATT-3') followed by 60 random bases and ending with an 18 nt
primer site (5'-AGATAGTAAGTGCAATCT-3') (
31
). Five [mu]g of this oligonucleotide were made double-stranded by annealing a primer complementary to the 3' flanking site (5'-AGATTGCACTTACTATCT-3') in a 1:1 molar ratio followed by
primer extension with 5 U
Taq
DNA polymerase in a standard reaction mixture (100 [mu]l containing 10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl, 0.001% gelatin and 200 [mu]M of each deoxynucleoside triphosphate) at 94oC for 15 min, 60oC for 1 min and 72oC for 30 min.
Binding of p53as-specific antibody to magnetic beads.
Twenty
mg of sheep anti-mouse IgG coated magnetic beads (Dynabeads 280, Dynal Inc.) were washed in
5 ml PBS/0.1% BSA solution three times and resuspended at 30 mg/ml in the same solution. A 100 [mu]l sample of the washed beads was added to 0.5 mg BC4-17 antibody and rocked at 4oC for 2.5 h. The beads were washed four times by adding 1 ml PBS/0.1% BSA, incubating at 4oC for 30 min, and recovering the beads with a magnet. Final resuspension
was in PBS/0.1% BSA at a concentration of 30 mg beads/ml.
Molecular cloning of PCR products
DNA binding assay
Fifteen [mu]l reactions were carried out on ice for 30 min and contained 3 [mu]l of an
in vitro
protein reaction, 10 [mu]g poly[d(I-C)], 0.1 mg/ml BSA and 0.1 [mu]l of a labeled PCR reaction (~100 000 c.p.m.) or 20 000-70 000 c.p.m. of an end-labeled DNA fragment in a buffer composed of 20
mM Tris-HCl pH 7.2, 80 mM NaCl, 1 mM EDTA, 5 mM DTT, 4% glycerol and 0.1% Triton
X-100. Supershifting of protein-DNA complexes was carried out using 100 ng PAb421 or 200 ng ApAs.
Reactions products were run on a 4% non-denaturing polyacrylamide gel in 0.5* TBE at 4oC, the gel dried and products visualized by autoradiography. A
p53 consensus DNA binding sequence is used as a positive control (p53con) (
19
,
25
). A mutated p53 consensus DNA binding sequence was used for competition as has
been previously described (
19
,
25
).
RESULTS
p53as-binding sequences obtained by CASTing
Aliquots of PCR reactions from six successive PCR cycles were observed by
agarose gel electrophoresis as shown in Figure
1
. The expected length of the bound DNA fragment was 96 nt and is indicated by
the arrow. Increasing amounts of higher molecular weight products visible
immediately above the putative specific product are commonly observed in
successive cycles using this technique, and have been attributed to futile
cycling due to limiting reagents (
29
,
30
). The lowest molecular weight products are primer dimers.
Figure
2
depicts gel shift assays carried out with aliquots of labeled PCR reactions
from cycles 3, 4 and 5 (as in Fig.
1
) and
in vitro
translated p53as protein. The arrow indicates the shifted band containing p53as
present in each cycle, which migrates similarly to the p53as bound to a
consensus p53 binding sequence shown in lane 4. The specificity of this band
was determined by competition with cold probe DNA (abrogation of band by wild
type p53 binding consensus sequence, lane 7, but not with mutated sequence,
lane 8) and by supershifting with ApAs antibody (lane 9). The faster migrating
band visible in lanes 4 and 6 was non-specific as it was not supershifted in the presence of ApAs.
An aliquot of cycle 5 was used to generate clones of these putative p53as
binding sequences for verification of p53as binding by gel shift assays,
comparison of binding of p53as and p53, and sequence analysis of binding
motifs. Plasmid DNA was isolated from 64 transformants and used in PCR
reactions containing [[alpha]-
32
P]dCTP. The labeled PCR products were used in gel shift assays with
in vitro
translated p53as. Fourteen of the 64 clones were verified positive for p53as
binding. Their randomized 60 bp inserts were sequenced on both strands and
results are presented in Table
1
. Flanking primer sequences were present in all clones and are not shown. Three
clones were represented twice (28/24, 59/42 and 64/38). Common artifacts when
cloning PCR products were observed in several clones but are unlikely to affect
the analysis: several clones had two or more copies of the primers at either or
both ends, and clones 28/24 and 36 had two randomized 60 bp segments separated
by multiple primers, as indicated in Table
1
. The 11 unique cloned sequences were verified for binding to p53as and also
were found to bind p53. Figure
3
A depicts an autoradiogram of gel binding assays containing the 11 unique
CASTing sequences and p53as protein in the presence or absence of ApAs. An
arrow indicates the specific protein-DNA complex which is abrogated or supershifted upon the addition of ApAs (as previously reported ref.
25
). Slight differences in mobility of bands reflect differences in the labeled
probes. The mobilities of non-specific bands are not affected by the addition of ApAs. Figure
3
B shows the same labeled CASTing sequences in gel shift assays containing the
regular form of p53 in the presence or absence of PAb421. As expected, little
of the
in vitro
translated p53 is in the active form and able to bind DNA (arrowhead) unless
PAb421 is present for activation and supershifting (arrow). The presence of
doublets in the supershifted complexes has been occasionally observed and is
reticulocyte lysate dependent. It also may reflect the ability of these
relatively long DNA probes to bind more than one p53 tetramer.
Underlined sequences are motifs shared with known p53 binding sequences shown in
Table 3. CATG sequences are in bold.R/Y, + indicates three purines upstream and three pyrimidine downstream of
CATGs; +/- indicates some, but not all, of the upstream or downstream sequences
conform to three purines upstream and three pyrimidines downstream of CATGs; -
indicates none of the upstream or downstream nucleotides conform to three
purines upstream and three pyrimidines downstream of CATGs.F, 3 nt flanking CATGs hybridize.I, intervening sequences (between CATGs) hybridize.#I, number of intervening nucleotides between CATG sequences.*, homology with MgBH6 (ref. 20); see Tables 3 and 4.
All 11 of the non-duplicated p53as binding sequences analyzed contained tetranucleotide CATG
and eight of the 11 contained two or more CATG motifs (shown in bold print in
Table
1
), as found in the p53 refined consensus sequence ('5'-PuGPuCATGPyCPy-3') (
21
). Only two of these, designated 13 and 64/38, fit the consensus sequence with
respect to three purines followed by a CATG followed by three pyrimidines. In
both these cases complementarities also were created by the flanking sequences
and within the intervening 6 nt (those between the two CATGs), a property
shared by two published consensus sequences (
19
,
22
). Neither 13 nor 64/38 fit the refined consensus requirement for `GPu' immediately upstream and `PyC' immediately downstream of the CATG (
21
). Two sequences, 6 and 14, did adhere to this requirement if only the 3 nt
upstream of the first CATG and the 3 nt downstream of the last CATG are
considered (indicated +/- in column R/Y, Table
1
). Three sequences, 18, 21 and 62, contained only one CATG although two of
these, 18 and 62, also had CTTG or CTAG respectively. This is consistent with
the less stringent consensus sequence (5'-PuPuPuC(A/T) (T/A)GPyPyPy-3') (
19
,
22
) but, unlike CATG itself, is not a perfect hybrid with the upstream CATG.
Sequence 21 had only one CATG and did not have an additional C(A/T)(T/A)G.
Furthermore, eight of the 11 sequences deviated from the consensus in that the
CATG was either not repeated, or the intervening nucleotides numbered <6 or >19. Therefore we designed synthetic oligonucleotides to test these points
further.
Synthetic DNA sequence tested in gel shift assays with p53 and pS3as proteins
translated
in vitro
The flanking sequences used in all oligonucleotides were described previously
(21). p53 was tested in the presence or absence of PAb421 and bound only in its
presence. WAF-1 has the wild type intervening sequence but not the flanking sequences.
mWAF-1 has changes in intervening and flanking sequences compared with WAF-1 (see Table 3). CCCGGG has only one CATG sequence upstream of a
heptanucleotide found in other published p53 binding sequences (see Table 3).
CATGs are shown in bold; underlined sequences are motifs shared with CASTing
sequences.
R/Y, + indicates three purines upstream and three pyrimidine downstream of
CATGs; +/- indicates some, but not all, of the upstream or downstream sequences
conform to three purines upstream and three pyrimidines downstream of CATGs.
F, 3 nt flanking CATGs hybridize.
I, intervening sequences (between CATGs) hybridize.
#I, number of intervening nucleotides between CATG sequences.
*, homology with MgBH6 (ref. 20).Several oligonucleotides and their complement sequences were synthesized,
annealed and used in DNA binding assays with
in vitro
translated p53as or p53. The sequences used and the results of these assays are
shown in Table
2
. Each oligonucleotide contains a central test sequence flanked by a common
unrelated sequence (
21
). p53as binding corresponded to p53 binding (in the presence of PAb421), so
both are shown together.
One copy of 5'-PuGPuCATGPyCPy-3' was insufficient to confer p53as binding, but two
copies of CATG with five (mWAF-1) or six (WAF-1) intervening nucleotides was sufficient for binding even without
internal symmetry or consensus flanking sequences. Binding was not observed
when either 0, 1 or 3 intervening nt were present, indicating that the CATGs
separated by 0-3 nt in CASTing sequences 6, 14, 28/24, 36 and 59/42 were not sufficient
for the observed p53as-p53 binding and that regions within the longer stretches between CATGs
(ranging from 23 to 88 nt) or sequences other than CATG (sequence 59/42) must
be involved. DNA sequences containing two copies of 5'-PuGPuCATGPyCPy-3' in which the CATG sequences are separated by 6 or 29
nt (2CATG6 and 2CATG29) were bound by both proteins, supporting the results
with CASTing sequences 6, 14, 28/24 and 36 where the number of intervening
nucleotides was either <5 or >23.
The results with synthetic probes did not answer the question of whether two
CATG motifs separated by >29 nt are sufficient for p53as-p53 binding. In addition, the probe designated CCCGGG for the sequence
adjacent to the only CATG in sequence 62 [and also found in the p53 binding
sites of p53con (
22
) and the cyclin G promoter (
16
)] was not bound by either protein, indicating that other sequences are
necessary for p53as-p53 binding. In the case of sequence 62 the other motif may be the CTAG
present 41 bp downstream of the CATG. The need for additional unidentified
sequence(s) for binding must be invoked for the sequences 21, which has only
one CATG, and 59/42, which has two CATG sequences separated by 3 nt and no
additional C(A/T)(T/A)G motifs. Therefore, we examined the sequences retrieved
by CASTing for homology with other reported p53 binding sequences presented for reference in Table
3
.
Summary of motifs and sequences in which each is found
*, extended to TTGGC
T
in 21, 59/42 and MgBH6.
Sequence comparison with other known p53 binding sequences
All of the sequences except CASTing sequence 6 (which had three CATGs separated
by 23 and 3 nt) contained one or more copies of regions found in reported p53
binding sequences. Of the sequences which were tested, all bound to p53as as
indicated in Table
3
. The homologous regions from these reported sequences and the sequences
retrieved by CASTing with p53as protein are underlined in Table
3
and summarized in Table
4
. As noted above, we were particularly interested in whether other non-consensus p53 binding motifs were present in the six out of 11 sequences
in which CATG was unique (sequences 18, 21 and 62) or in which two CATGs were
separated by <5 (59/42) or >29 (28/24, 36) nt. It was striking that five of these six
sequences (all but sequence 62) contained a region of homology to the non-consensus sequence, MgBH6 (
20
) (see the column headed * in Table
1
). Cyclin G and RGC also contain regions of homology to MgBH6 (see Table
3
). As detailed previously, the p53 binding sequence MgBH6 was obtained from mouse fibroblast DNA and significant sequence homology to known genes was not found at the time. Furthermore, the
entire region shown in that report was protected by p53 protein produced in
Escherichia coli
in DNA footprinting assays and conferred activation by p53 of a reporter gene
in transcription assays in cells (
20
).
Affinity of p53 and p53as for DNA binding
Two of the synthesized DNAs that were positive for binding, 2CATG6 and 2CATG29,
were examined for differences in affinity between
in vitro
translated p53 and p53as. Gel shift assays that used 2 ng
32
P-labeled consensus sequence (see Materials and Methods) probe were
subjected to increasing amounts of 2CATG6 or 2CATG29 unlabeled probe. Figure
4
shows the autoradiogram from one of several such experiments. The affinity of
both proteins for 2CATG6 was greater than for 2CATG29 since 2 ng cold 2CATG6
probe was able to compete all binding of the labeled consensus sequence probe
while 10 ng cold 2CATG29 did not completely compete binding to the labeled
probe for either protein.
DISCUSSION
Previous studies have shown that p53 binds two repeats of 5'-PuPuPuC(A/T)(T/A)GPyPyPy-3' and that the spacing between these repeats ranged from 0 to 13 nt (
19
), which is equivalent to 6-19 nt between CATGs. The independent retrieval of p53as binding sequences
confirms certain features of published p53 DNA binding and indicates a lack of
strict requirement for others as follows: (i) confirms one or more CATG motifs;
(ii) no strict requirement either for specific sequence or for secondary
structure of the internal nucleotides between two copies of CATG; (iii) lack of
requirement for three purines preceding the CATG or three pyrimidines following
the CATG; (iv) no requirement for flanking nucleotide triplets to contain
particular nucleotides or secondary structure; and (v) intervening space
between two CATGs can contain 29 nt or more. p53as and p53 proteins showed the
same sequence specificities and the affinities of binding to sequences with
CATGs separated by 6 or 29 intervening nt was the same for both proteins.
All of the CASTing sequences contained one or more copies of CATG. Two of the
retrieved sequences, 13 and 64/38, fit the consensus sequence (
19
,
22
), and three additional sequences, 6, 14 and 63, fit a refined consensus (
21
) with respect to the placement of G 2 nt upstream and C 2 nt downstream of a
CATG. However two (
6
,
14
) of these five sequences with similarities to the consensus had more than the
reported 19 nt between the CATG motifs and no additional reported p53 binding
motifs, supporting the results with synthetic oligos and the conclusion that <= 29 bp between CATGs are tolerated. Of the six additional sequences (18, 21, 28/24, 36, 59/42, 62) which did
not fit the consensus, having only one CATG or <5 or >19 nt between two CATG motifs, five contained a region of homology with
the p53 binding sequence MgBH6 (
20
). The remaining sequence, 62, contained a CATGCCCGGG sequence and a downstream
CTAG with additional intervening potential for secondary structure. Both p53as
and p53 bound a sequence from the cyclin G promoter (Wu,Y., unpublished) which
contains one copy each of TGCCCGGG (found in p53con) and TTGGC (found in MgBH6)
but does not contain CATG (
16
). The sequences TGCCT, TGTCC and TGCCC either are adjacent to or overlap a
C(A/T)(T/A)G consensus in retrieved sequences and in all but one of the
reported consensus sequences noted, making their significance ambiguous. The
exception, 50-2, contains no overlaps with a C(A/T)(T/A)G consensus, but does contain
one CATG upstream of two copies of TGCCT. While not tested by gel binding
assays, the p53as protein transcriptionally activated a reporter via the 50-2 sequence (
32
) shown in Table
3
(Huang,H. and Kulesz-Martin,M., unpublished).
The affinity of p53as and p53 for CATG separated by 29 nt was within the range
observed for other published p53 binding sites such as RGC (
18
) and WAF-1 (
13
). However, one study reports that DNA binding sequences that contained >10 nt
between the CATG motif were transcriptionally inactive when tested by
in vivo
assays in yeast (
33
) suggesting that
in vitro
DNA binding does not necessarily convey
in vivo
function.
The sequences retrieved by CASTing for p53as binding further defined p53 binding
motifs and suggest additional combinations for database searches of potential
p53 target sites. We performed FASTA searches of the GenBank and EMBL databases
using several combinations of the motifs found in the p53as retrieved sequences
or shared with known p53 binding sequences. The identical consensus-like sequence (GAA
CATG
TCCGGA
CATG
TTC) independently retrieved in sequences 13 and 64/38 was not found as a perfect
match to any known sequence; however, homologous sequences were found in many
genes including the 5' promoter region of AP-1 (accession no. L16546, a rat homologue of human multidrug
resistance gene), and the 3' region of mRNA from the human ICAM gene (accession J03132). An inverted
repeat, CGGTT-TTGGC, of a segment (TTGGC) of the published sequence MgBH6 (
20
) which is also present in 21, 28/24 and 59/42 was found in intron 6 of the
thrombospondin gene (accession no. J05605), the human CMV enhancer region
(accession no. M64944) and in the coding region of a human homologue of dnaJ
(accession no. D13388). Certain of these genes are transcriptionally modulated
by p53 or their family member proteins are known to bind p53. We speculate that
searches of combinations of motifs more specifically defined by p53as binding
in the current studies is warranted, followed by testing of any naturally
occurring regions for p53 or p53as binding, transcriptional activity of
reporter genes in mammalian cells, and p53as- or p53-dependent modulation of endogenous gene expression.
We conclude from these results that the DNA binding sequences of p53as and p53
are the same. However, since p53as is constitutively active for DNA binding,
and p53 oligomerization with p53as blocks that binding (
25
), the p53:p53as ratio may be important in regulating expression of
transcriptional target genes. We conclude further that any differences in the
function of p53 and p53as proteins must result not from inherent differences in
DNA sequence specificity of binding, but from differential expression of the
proteins or from differential regulation of their function by modifications at
the C-terminus, either by post-translational modifications of amino acids present in p53 but not in
p53as and/or by differential binding of p53-associated proteins.
ACKNOWLEDGEMENTS
We are grateful to Jean Seo for her help in beginning this project and Dr Bruce
Dolnick for helpful suggestions. This work was supported by a US Army
Department of Defense Medical Research Grant DAMD17-94-J-4341 and by Roswell Park Cancer Institute NIH Core Grant CA16056.
