ABSTRACT
The high mobility group protein (HMG)-box is a DNA-binding domain found in many proteins that bind preferentially to
DNA of irregular structures in a sequence-independent manner and can bend the DNA. We show here that GST-fusion proteins of HMG domains from HMG1 and HMG2 promote a triple-stranded complex formation between DNA containing the
(GGA/TCC)
11
repeat and oligonucleotides of d(GGA)
11
probably due to G:G base pairing. The activity is to reduce association time
and requirements of Mg
2+
and oligonucleotide concentrations. The HMG box of SRY, the protein determining
male-sex differentiation, also has the activity, suggesting that it is not
restricted to the HMG-box domains derived from HMG1/2 but is common to those from other members
of the HMG-box family of proteins. Interestingly, the box-AB and box-B of HMG1 bend DNA containing the repeat, but SRY fails to
bend in a circularization assay. The difference suggests that the two
activities of association-promotion and DNA bending are distinct. These results suggest that the HMG-box domain has a novel activity of promoting the association between
GGA repeats which might be involved in higher-order architecture of chromatin.
The high mobility group proteins HMG1 and HMG2 are abundant, chromatin-associated proteins whose function is unknown. The two proteins belong to
one class of the HMG family of proteins that is a collection of small acidic
proteins extracted from chromatin with 0.35 M NaCl (
1
). Extensive studies of HMG1/2 proteins revealed that they bind preferentially
to DNA of irregular structures in a sequence-independent manner (
2
-
4
). This binding occurs through several domains called HMG boxes, which were
first identified by the comparison of HMG1 with UBF, a protein involved in
regulating the expression of genes encoding ribosomal RNA (
2
,
5
). Two homologous segments of an 80-amino acid sequence, box-A and box-B, are present in HMG1/2 (
6
,
7
). Interestingly, similar domains are found in many other proteins such as LEF-1, TCF-1, SRY and the related SOX proteins (
4
,
8
-
12
). They comprise single HMG-box domains and have the additional property of sequence-specific binding to linear DNA. The HMG-box proteins have a capacity to unwind DNA structure by
bending (
13
,
14
). Such distortions are probably important to facilitate the action of
transcription factors bound nearby (
4
,
15
,
16
). For this reason they are considered as architectural proteins in nuclei (
4
,
17
).
Homopurine/homopyrimidine sequences are abundant in eukaryotic genomes and some
of them consist of tandem repeats such as (GGA/TCC)
n
(
18
-
20
). The sequences are known to take a non-B DNA conformation which may provide some signals for a variety of DNA
transactions (
21
-
23
). We previously reported that d(GGA)
11
oligonucleotides, but not d(GAA)
11,
dimerize
in vitro
probably through guanine-guanine base pairing (
24
). The oligonucleotides also bind to double-stranded DNA containing (GGA/TCC)
11
, resulting in a triple-stranded complex (
25
). The complex is presumed to form a unique structure, consisting of a
homoduplex between the two GGA repeats and a single-stranded loop of the TCC repeat. Such a D-loop-like structure is distinct from the triplex model (
26
) applicable to the complex formed between the (GGA/TCC)
11
repeats and d(GGT)
11
oligonucleotides, because the two complexes exhibit distinct association
kinetics and DNase I footprints (
25
).
Search for binding protein by Southwestern analysis revealed that HMG1
preferentially binds to the homoduplex between d(GGA)
11
oligonucleotides. This suggested a possible involvement of this architectural
protein in the triple-stranded complex formation, although the binding may have resulted from
the binding preference of HMG1 to irregular DNA structures. The finding
prompted us to investigate effect of HMG1 or HMG-box proteins on the complex formation. We found that GST-fusion proteins containing HMG-boxes of HMG1, HMG2 and SRY enhance the association of triple-stranded complexes between DNA containing the (GGA/TCC)
11
repeat and (GGA)
11
oligonucleotides. The present paper describes details of this enhancement. The
results obtained here provide evidence for a novel activity of the HMG domain:
promotion of complexes between GGA-repeats. Biological relevance is also discussed.
Recombinant clones were constructed using pGEX2T (
27
). They are capable of directing the synthesis in
Escherichia coli
of box-AB, box-A and box-B of HMG1 and HMG2, and the HMG box of SRY. Sets of primers
carrying
Bam
HI and
Eco
RI tags were designed according to each sequence (
28
-
30
) to produce GST-fused proteins of those proteins. Primer sequences are as follows: 5'-AGGATCCAAAGGAGATCCTAAGAAGC-3' and 5'-TTGAATTCCCTTTTTCGCTGCATCAGG-3' for HMG1 box-AB; 5'-AGGATCCAAAGGAGATCCTAAGAAGC-3'
and 5'-TTGAATTCCTTGAACTTCTTTTTGGTCT-3' for HMG1 box-A; 5'-AAGGATCCAATGCACCCAAGAGGC-3' and 5'-TTGAATTCCCTTTTTCGCTGCATCAGG-3'
for HMG1 box-B; 5'-AAGGATCCATCAAGCCGCTG-GGC-3' and 5'-TT-
GAATTCCAGGACCCTTCTTTCCT3' for HMG2 box-AB; 5'-AAGGATCCATCAAGCCGCTGGGC-3' and 5'-AAGAATTCTCTTCGGAGCATTTGGAT-3' for HMG2 box
A; 5'-AAGGATCCAAATGCTCCGAAG-3' and 5'-TTGAATTCCAGGACCCTTCTTTCCT-3' for HMG2 box-B; 5'-AAGGATCCCATGGAGGGCCATGTCAAGC-3'
and 5'-TTGAAT- TCACTTTAGCCCTCCGATGAG-3' for HMG box of SRY. PCR amplification was
carried out under a standard condition using mouse P19 cell cDNA as template (
31
,
32
). The PCR products and pGEX2T DNA were both digested with
Bam
HI and
Eco
RI, ligated each other, and then subjected to transformation of DH5[alpha] cells (
32
). Inserted DNA sequences of plasmids in the transformants were determined, and
appropriate cells were selected. Each fused protein was purified from the cell
extracts according to the method of Smith and Johnson (
27
) using prepacked glutathione-Sepharose 4B columns (Pharmacia Biotech). More than 95% purity was
confirmed by staining with Coomassie Brilliant Blue R-250 (Nakarai Inc., Japan) after separation with 10% SDS-PAGE. The amino acid sequences fused to GST protein are as follows.
HMG1; box-AB, box-A and box-B contain K2-K172, K2-K89 for and N92-K172, respectively (
28
): HMG2; box-AB, box-A and box-B comprise I6-P174, I6-P91 and P94-P174, respectively (
29
): SRY has M1-K81 (
30
). Two amino acid sequences, KKLGEMWNNTA and KDIAAYRAK, are the sequences
obtained by analysis of the two peptide fragments which were derived from a DNA-protein complex when protein binding to the d(GGA)
11
homoduplex was searched with Southwestern analysis (data not shown). Sequence
analysis of the HMG2 sequence revealed difference in one amino acid from the
reported one (
29
); K72 was found in our clone instead of R72.
Native HMG1 protein was prepared from mouse Ehlich ascites tumor cells as
described (
33
).
The antiserum to HMG1 was raised in rabbits against purified GST-fusion proteins of HMG1 box-AB. For immunoinhibition, 1 [mu]l of affinity-purified polyclonal anti-HMG1 antibodies (0.4 [mu]g/[mu]l) was added to 10 [mu]l of the association mixture and
incubated overnight at 4oC.
Double-stranded DNA of 162 bp was synthesized in the presence of [[alpha]-
32
P]dCTP with PCR using the pUC118 carrying d(GGA/TCC)
11
between the
Xba
I and
Bam
HI sites. Primer sequences used are 5'-GTTTTCCCAGTCACGAC-3' (M4) and 5'-CAGGAAACAGCTATGAC-3' (RV). The labeled fragments
having (GGA/TCC)
11
(1 nM) was incubated in the absence or presence of HMG-domain proteins (6-750 nM) with d(GGA)
11
(50 or 100 nM) at 37oC for 30 min in 10 [mu]l of a buffer containing 50 mM NaCl, 10 mM Mg
2+
, 0.2 mM DTT, 20 mM HEPES (pH 7.9) and 6% glycerol, followed by cooling to 4oC in the course of 30 min by a programmable thermal cycler (ASTEC Inc.,
Japan) (
25
). The products were subjected to electrophoresis in 5% nondenaturing
polyacrylamide gels (PAGE) with and without the digestion with proteinase K at
a concentration of 50 [mu]g/ml for 2 h at 4oC. The electrophoresis was performed in 1* TBE (50 mM Tris-borate, pH 8.3, 1 mM EDTA) buffer containing 50 mM NaCl
and 10 mM Mg
2+
at 4oC. T4 gene 32 protein and cytochrome C were purchased from Boehringer
Mannheim Co., Germany.
The triple-stranded complexes were obtained as described above except for the use of
10 nM of the 162 bp fragment containing (GGA/TCC)
11
. The DNA bound with HMG1 was obtained by incubation with box-AB (10 nM DNA and 750 nM protein: a protein-DNA molar ratio of 75:1) in a buffer containing 20 mM HEPES (pH
7.9), 50 mM NaCl, 2 mM MgCl
2
0.2 mM DTT for 15 min at 20oC. The complexes and the 162 bp fragment were subjected to limited DNase I
digestion in 250 [mu]l of a buffer containing 10 mM HEPES (pH 8.0), 40 mM KCl, 5 mM MgCl
2
, 0.5 mM DTT, 2 [mu]g of sonicated calf thymus DNA and 2.5 U of DNase I (Takara, Japan). The
digestion was allowed for 5 min at 20oC and terminated by addition of 50 [mu]l of a buffer containing 10 mM Tris-HCl (pH 7.4), 20 mM EDTA, 100 [mu]g/ml tRNA and 0.1% SDS. The DNA was recovered with ethanol
precipitation after phenol treatment. Levels of the digestion were analyzed by
primer extension method using Circum Vent thermal cycle dideoxy DNA sequencing
system (New England BioLabs) except 0.2 mM each of dNTPs used. The M4 and RV
primers were 5' terminally labeled with [[gamma]-
32
P]ATP and polynucleotide kinase and were used in primer extension for analysis
of TCC- and GGA-strands, respectively. The location of modified sites was determined
by electrophoresis on a 6% polyacrylamide-8 M urea gel in parallel with sequencing ladder. The bands were
visualized by autoradiography.
Circulization assay was carried out as described (
13
).
32
P-labeled DNA fragment with
Eco
RI sticky ends was synthesized as follows; PCR was carried out in the presence
of [[alpha]-
32
P]dCTP using the pUC118 carrying (GGA/TCC)
11
as template and a set of primers: 5'-GTTTTCCCAGTCACGAC-3' and 5'-AGAATTCAGGAAACAGCTATGAC-3'. Amplified DNA was cleaved with
Eco
RI and fragments of 120 bp were purified after polyacrylamide gel electrophoresis. The fragment (1 nM) was incubated without and with GST-fusion proteins containing HMG-box in 10 [mu]l of a buffer containing 50 mM HEPES-NaOH (pH 7.5), 50 mM potassium glutamate, 10 mM Mg-acetate and 1 mM ATP for 10 min at 30oC. T4 DNA ligase (0.6 Weiss U,Takara, Japan)
was then added and incubated for 10 min followed by inactivation of the ligase
by shifting to 65oC for 15 min. Reactions were subsequently incubated with 1 U of exonuclease
III (Takara, Japan) for 45 min at 37oC to remove linear ligation products. The reaction products were extracted
with phenol and ethanol precipitated. The nuclease protected products were
recovered by centrifugation and subjected to electrophoresis on a 5%
polyacrylamide gel in Tris-borate-EDTA (pH 8.3) buffer which was then dried and autoradiographed.
Gel-mobility shift assays were carried out as described (
15
,
32
). DNA containing the SRY recognition sequence, AACAAAG, used here was produced
as follows. Two complementary sequences, 5'-GATCCGGGAGACTGAGAACAAAGCGCTCTT-3' and 5'-CTAGAAGAGCGCTTTGTTCTCAGTCTCCCG-3', were synthesized and
annealed, and then the fragment was cloned into
Bam
HI and
Xba
I of pUC118. PCR was carried out in the presence of [[alpha]-
32
P]dCTP with M4 and RV primers using the plasmid DNA as template.
Plasmids capable of directing the synthesis in
E.coli
of box-AB, box-A and box-B of HMG1 were constructed by cloning the relevant sequences
of the mouse cDNA in pGEX2T. They encoded for amino acids K2-K172, K2-K89 and N92-K172, respectively. Each GST-fusion protein was purified to near homogeneity and used
for association assay. Also, three GST-fusion proteins of HMG2 were synthesized in a similar manner. Figure
1
A shows the effect of HMG1 on the formation of a triple-stranded complex.
32
P-labeled double-stranded DNA (1 nM) containing the (GGA/TCC)
11
repeat was synthesized with PCR and incubated with a low concentration (50 nM)
of d(GGA)
11
oligonucleotides. Products in the mixture were then separated by polyacrylamide
gel electrophoresis. Only a faint band was seen in the position of gel-shifted band (lane 2). However, in the presence of a GST-fusion protein containing box-AB, box-A or box-B of HMG1, the shifted band was detected with much
more intense signals (lanes 3, 4 and 5, respectively). GST protein itself did
not show that shift (lane 6). These patterns were given by electrophoresis
after the proteinase K treatment of DNA-protein complexes. Lanes 8-12 show respective patterns without the proteinase K treatment.
Most products were retained in the origin of gel except those in the incubate
with GST protein.
Properties of HMG1 for the association were investigated in this assay using the
box-AB protein. Figure
2
A shows an analysis of time course in the presence of 10 mM Mg
2+
. Fifty percent complex formation required 40 min without the protein but was
achieved within only 10 min in the presence of box-AB (750 nM). The formation of triple-stranded complex was dependent on the Mg
2+
concentration; 2 mM Mg
2+
did not give the complex. At least 6 mM was required for 50% binding under this
condition. However, the addition of box-AB reduced the Mg
2+
requirement to 2 mM (Fig.
2
B). Therefore, the assay done under the 2 mM Mg
2+
gave more striking difference. Figure
2
C shows effect of d(GGA)
11
oligonucleotide concentrations. The protein was able to reduce the minimal
concentration required for the band shift; at the concentration of as low as
1.6 nM of d(GGA)
11
, duplex containing (GGA/TCC)
11
was little converted into the triple-stranded DNA (lane 2). Upon addition of the box-AB protein, however, the complex was formed under the same condition
(lane 8). These results demonstrated that the enhancing activity of the box AB
protein for the triple-stranded complex formation is to reduce association time and requirements
of the Mg
2+
and oligonucleotide concentrations.
Structural analysis was carried out using DNase I. The complexes between DNA
containing (GGA/TCC)
11
and d(GGA)
11
formed in the presence and absence of HMG1 were partially digested with DNase
I, and the sensitivity was measured by the extension method using
32
P-labeled primers as described in Materials and Methods. Figure
4
shows gel electrophoretic patterns of the PCR products. The complex formed in
the absence of HMG1 exhibited increased sensitivity for both strands of the
duplex and a hypersensitive site in the vicinity of the repeat region of the
GGA strand (lanes 2 and 6), consistent with our previous result (
25
). The same pattern was observed for the complex formed in the presence of HMG1
(lanes 4 and 8). This sensitivity suggests that the triple-stranded complex is not the triplex previously defined (
26
) but a D-loop-like structure (
25
). Effect of the HMG1 binding was also examined; the box-AB did not show significant protection of the repeat and flanking
sequences (lanes 3 and 7), but slightly increased sensitivity in the repeat
region of the GGA strand (lane 3).
HMG-boxes are known to have the capacity to modulate DNA structure by bending
(
13
,
14
). Since promotion of the association may reflect the bending activity, their
relationship was examined. DNA bending was measured by the T4 DNA ligase-catalyzed circularization of the short DNA fragment having the repeat in
the middle and the
Eco
RI site at both ends. Figure
5
shows the ligation products in this assay. In the presence of T4 DNA ligase
alone, the fragment formed a few broad bands (lane 2). Those bands are not
circular molecules induced by the binding because they are cut into shorter
fragments by the digestion with exonuclease III (lane 3). Upon addition of the
box-AB of HMG1, however, the ligation provided a band of closed-circular molecules resistant to the exonuclease digestion (lane 4).
DNA fragments in the band were isolated and their sizes were determined by
polyacrylamide gel electrophoresis after the digestion with
Hin
dIII and
Sma
I in the vicinity of
Eco
RI ends. The fragment containing the
Eco
RI site was observed (data not shown), indicating that major HMG1-induced ligation product is a monomer circular. At the box-AB concentration of 30 nM (a protein-DNA molar ratio of 30:1), the circle was detected but levels
decreased rapidly at the concentrations >150 nM (lanes 5 and 6). The box-B of HMG1 gave a maximal level of the product at the concentration of 750
nM (lane 9). In contrast, addition of HMG1 box-A showed much weaker activities (lanes 10-12). This is consistent with previous papers (
13
,
35
) except one (
36
). These results suggest that the bending activity of HMG1 does not parallel the
association-promoting activity shown in Figure
3
.
The SRY protein plays a primary role in male-sex determination and contains a single HMG-box (
30
,
37
). In contrast with HMG1/2, it is a regulatory protein with a limited cell-type distribution and with sequence specificity for the binding.
Therefore, effect of its HMG-box was studied. GST-fusion protein containing the HMG-box of SRY was synthesized, and its binding was first examined
in a gel mobility-shift assay (Fig.
6
A). The SRY bound to DNA fragments containing the (GGA/TCC)
11
and provided a discrete band at the concentration of 40 nM (lane 9). At 200 nM
it showed a super-shift pattern (lane 10). Similar shift patterns were observed for DNA
fragments containing the SRY-recognition sequence of AACAAAG (
9
) (lanes 4 and 5). These results indicated that the fusion protein of SRY binds
to both sequences.
The present studies with HMG1 and HMG2 indicate that these abundant nuclear
proteins are capable of promoting a triple-stranded complex formation between DNA containing the (GGA/TCC)
11
repeat and oligonucleotides of d(GGA)
11
(Fig.
1
). Association of the complex is probably due to G-G base pairing between the two GGA repeats, because d(GGA)
11
oligonucleotides form homoduplex in a parallel orientation through that pairing
(
24
). The promoting activity of HMG1 is to reduce association time and requirements
of Mg
2+
and oligonucleotide concentrations, which is demonstrated by several
experiments (Fig.
2
): (i) analysis of reaction time was done under the condition containing 10 mM
Mg
2+
where the complex was spontaneously formed. The result reveals that the binding
shortens time in the incubation needed for the complex formation even under
that condition; (ii) the Mg
2+
concentration required is decreased by the binding. The association occurs at a
Mg
2+
concentration of 2 mM, at which concentration no complex is observed without
the box-AB protein; and (iii) the required oligonucleotide concentration is also
reduced. The complex is clearly seen at 1.6 nM of d(GGA)
11
,
~2-fold excess of the double-stranded DNA, only upon addition of the box-AB protein. Such enhancement is not observed in other two
single-stranded DNA-binding proteins that can promote the annealing of complementary DNA
strands (Fig.
1
D). These results substantiate the ability of the HMG proteins to facilitate the
complex formation.
Our studies also reveal that the activity of promoting the complex is maintained
by one of the two HMG-box domains present in HMG1 and HMG2. Each of the HMG-box domains consist of ~80 amino acids (
6
,
11
,
12
). It seems likely, however, that the box-AB protein of HMG1 has an activity stronger than the box-A and the box-B (Fig.
3
A). This may be attributed to the presence of two domains as one unit. On the
other hand, native HMG1 exhibits a weaker activity (Fig.
3
B) which may be due to the presence of the acidic domain in the C-terminal end (
39
). Actually, a fusion protein of HMG1 box-BC (C indicates an acidic domain) displayed an activity weaker than that
of box-B (data not shown).
HMG box is present in many other proteins which comprise the HMG-box family of proteins (
4
,
11
,
12
). There are two subfamilies. Members with multiple HMG domains including HMG1/2
and UBF present in all cell types. They show low sequence specificity for the
binding. The others have single HMG domains and have a restricted cell-type distribution. They recognize specific sequences. SRY, a protein to
determine the male differentiation (
30
,
37
), is one of the proteins belonging to the latter group. GST-fusion protein carrying the SRY HMG-box domain can bind to DNA containing (GGA/TCC)
11
and facilitates the association between the DNA and d(GGA)
11
oligonucleotides as efficiently as the box-B of HMG1 does (Fig.
6
). This suggests that the association-enhancing activity is not restricted to the HMG-box domains derived from abundant nuclear proteins of HMG1/2 but is
common to HMG domains from other members of the HMG-box family of proteins involved in transcription and cell differentiation
as well.
The HMG-box is known to have the capacity to modulate DNA structure by bending (
13
,
14
). The bending may be an activity responsible for the association promotion.
Therefore, their relationship was examined by measuring the T4-DNA ligase-catalyzed circularization of a short DNA fragment having the repeat
(Fig.
5
). The box-AB and box-B of HMG1 exhibit circularization activities much higher than the
box-A does, consistent with previous reports (
13
,
35
). On the other hand, the box-A shows an association activity similar to or rather higher than that of
box-B. This discrepancy may be due to different detection sensitivity of the
two assays but suggests that the association promotion does not simply reflect
the bending activity. A similar difference is also observed in the case of HMG
box of SRY. The protein fails to bend DNA in our circularization assay,
consistent with a previous report (
35
), whereas it facilitates the association between the DNA containing (GGA/TCC)
11
and d(GGA)
11
as efficiently as the box-B of HMG1 does (Fig.
6
). These results suggest that the association-promoting activity and the bending activity are distinct and that the HMG-box domain has a novel property of promoting the association between
GGA repeats in addition to its abilities to bind to irregular DNA and to bent
DNA.
DNase I protection analysis has indicated that the binding of HMG1 to linear DNA
does not affect the DNase I cleavage patterns for most DNA because of the low
sequence specificity except for DNA of irregular structures (
17
,
40
,
41
). The DNA containing (GGA/TCC)
11
bound with the box-AB did not give the protection pattern, either, but showed
hypersensitivity in the repeat region of the GGA strand (Fig.
4
A). The hypersensitivity might reflect the HMG1-induced distortion or unwinding of DNA. If it is the case, the unwinding
would provide separation of the GGA- and TCC-strands which facilitates the association of two GGA-repeats by increasing their accessibility.
The finding that HMG1/2 promote the association between the (GGA/TCC) repeats
and single-stranded DNA of homologous sequence suggests that the association involves
two double-stranded DNA harboring such repeats. This implication is possible, because
the repeats have a property of adopting two single-stranded loops, as revealed by nuclease S
1
digestion experiments (
18
,
20
). We are currently testing this possibility
in vitro
and have obtained a positive result in preliminary experiments (Mishima
et al
. unpublished). The promotion activity may be regarded as one analogous to the
function of certain protein chaperons, which stabilize a polypeptide in a
conformation that is appropriate for subsequent assembly, but would be unstable
without a chaperone (
42
). This may provide another clue to the function of HMG1 and HMG2. There are
other tandem repeats that can form four-stranded DNA complexes: telomeric DNA (
43
-
45
), guanine-rich sequences (
46
) and (CA/TG)
n
repeats (
47
). HMG1 preferentially binds to the latter repeats (
17
,
47
). Those four-stranded complexes could take place with aid of HMG1/2 among many
chromosomal regions because such repeats are abundant in the nucleus (
4
9
). It is possible that those complexes act as architectural elements
constituting chromosomal domains or providing condensation (or decondensation)
of the 30 nm fiber. The abundance of HMG1 and HMG2 in a nucleus is consistent
with a general requirement for those conformations, particularly during times
of chromatin remodeling in cells undergoing DNA replication and mitosis.
Sequence-specific association between the GGA repeats might be implicated in the
initiation of the pairing of homologues during meiosis and a pairing mechanism
of recombination or repair. The former was already suggested for G-rich four-stranded DNA (
46
). The latter speculation is consistent with the presence of a unique triple-stranded structure in the recombination process that is similar to the
complex formed between (GGA/TCC)
11
and d(GGA)
11
(
48
,
49
). The finding that microsatellites comprising (GGA/TCC)
n
repeats are recombinogenic as deduced by their polymorphisms (
50
) may also support this speculation.
We thank Ohtsura Niwa, Hiroshima University, and Hiroshi Hamada, Osaka
University, for critical reading this manuscript and providing cDNA library
from mouse P19 cells, respectively. This work was supported by a grant-in-aid from the Ministry of Education, Science and Culture of Japan.
REFERENCES
Return