ABSTRACT
Duplex DNA recognition by oligonucleotide-directed triple helix formation is being explored as a highly specific
approach to artificial gene repression. We have identified two potential
triplex target sequences in the promoter of the human
bcl-2
gene, whose product inhibits apoptosis. Oligonucleotides designed to bind these
target sequences were tested for their binding affinities and specificities
under pseudo-physiological conditions. Electrophoretic mobility shift and dimethyl sulfate footprinting assays demonstrated that an oligonucleotide designed for simultaneous recognition of homopurine domains on alternate duplex DNA strands had the highest affinity of any oligonucleotide tested. Modifications to render this oligonucleotide nuclease-resistant did not reduce its binding affinity or specificity. In additional studies under various pH conditions, pyrimidine motif complexes at these target
sequences were found to be stable at pH 8.0, despite the presumed requirement
for protonation of oligonucleotide cytidines. In contrast, purine motif
complexes, typically considered to be pH independent, were highly destabilized
at decreasing pH values. These results indicate that a natural sequence in the
human
bcl-2
promoter can form a stable triplex with a synthetic oligonucleotide under
pseudo-physiological conditions, and suggest that triple helix formation might
provide an approach to the artificial repression of
bcl-2
transcription.
Oligonucleotide-directed triple helix formation is a highly specific strategy for
designing potential artificial gene repressors (
1
-
5
).
In vitro
experiments provide evidence that triplexes can block DNA binding proteins (
6
-
8
) and inhibit transcription initiation (
2
,
9
-
13
). Recognition of homopurine/homopyrimidine sequences involves hydrogen bonds
between oligonucleotide bases and purine bases in the major groove of duplex
DNA. Triple helix formation occurs in either of two distinct patterns, termed
the pyrimidine (Pyr) motif and purine (Pur) motif (
4
). In the pyrimidine motif, oligonucleotides bind parallel to the purine strand
of the duplex by Hoogsteen hydrogen bonding to form T[middot]A[middot]T and C+[middot]G[middot]C base triplets (
1
). In the purine motif, oligonucleotides bind antiparallel to the purine strand
of the duplex by reverse Hoogsteen hydrogen bonding to form A[middot]A[middot]T (or T[middot]A[middot]T) and G[middot]G[middot]C base triplets (
14
). We are interested in the possibility that triple helix formation might be
used to artificially regulate the expression of disease-related genes.
The product of the
bcl-2
proto-oncogene acts as a negative regulator of programmed cell death (apoptosis;
15
-
17
). Apoptosis not only provides a termination option for cells that are
dangerously damaged, but also plays a key role in normal T and B cell
development (
18
-
20
). In some tissues, the propensity toward apoptosis appears to be regulated by
the ratio of
bcl-2
and
bax
proteins. Increases in
bax
concentrations counter the apoptotic suppression by
bcl-2
(
21
). One of the most common cytogenetic abnormalities in non-Hodgkin B cell lymphomas is the translocation t(14;18)(q32;q21). This
translocation places the
bcl-2
gene at 14q21 under the control of the immunoglobulin heavy-chain gene enhancer, resulting in
bcl-2
overexpression (
22
,
23
). Studies of transgenic mice overexpressing
bcl-2
show an increase in both B and T cell survival, leading to lymphomas derived
from both cell types (
22
). In contrast,
bcl-2
-/- knockout mice undergo profound apoptotic deletion of B and T
cells shortly after birth (
24
).
Therapeutic techniques such as radiation and many chemical agents act by
inducing apoptosis. This observation suggests that approaches to reducing the
damage threshold required for induction of cell death by an apoptotic signal
may sensitize tumor cells to chemotherapeutic agents. One such approach might
be to artificially reduce
bcl-2
levels in cells by repression of
bcl-2
transcription via triple helix formation targeted to the
bcl-2
promoter region. The human
bcl-2
gene is transcribed from two promoters (P1 and P2; Fig.
1
). The major promoter (P1) lacks a TATA box, is GC rich, contains seven
consensus Sp1 binding sites, and displays multiple transcription initiation
sites (
23
). We have identified two homopurine sequences just upstream of this major
bcl-2
promoter that might serve as triplex target sites.
We wished to design oligonucleotides that might bind the
bcl-2
target sequences with high specificity and affinity under physiological
conditions. Electrophoretic mobility shift titrations were performed to
estimate oligonucleotide affinities, while dimethyl sulfate footprinting assays
were used to analyze oligonucleotide recognition and induced changes in the
structure of the duplex target upon triplex formation. We report that under pseudo-physiological conditions, an oligonucleotide designed for adjacent purine
and pyrimidine motif recognition binds homopurine domains on alternate duplex
DNA strands with high affinity. This result provides an example of a naturally-occurring target sequence for which alternate strand triple helix
formation clearly increases binding affinity relative to recognition of a
single homopurine domain. Modifications to render this high-affinity oligonucleotide nuclease-resistant did not reduce its binding affinity. While many pyrimidine
motif oligonucleotides require an acidic pH to protonate cytidine residues for
tight binding to duplex DNA, the pyrimidine motif complexes in this study were
remarkably stable at pH 8.0. In contrast, the stabilities of purine motif
complexes, often considered to be pH independent, were in this study highly
reduced at decreasing pH values. Together, these results identify
oligonucleotides that bind tightly to the human
bcl-2
promoter under pseudo-physiological conditions. These results suggest a possible strategy for
artificial repression of
bcl-2
transcription.
Oligonucleotide sequences are shown in Figure
1
B. Oligonucleotides were synthesized by phosphoramidite chemistry on an ABI
Model 380B DNA synthesizer, purified by denaturing polyacrylamide gel
electrophoresis, eluted from gel slices, and desalted by Sep-Pak C
18
cartridge chromatography (Waters). Oligonucleotides were quantitated by absorbance at 260 nm using molar extinction coefficients (M
-1
cm
-1)
of 15 400 (dA), 11 700 (dG), 7300 (dC), 5700 (dMe5C) and 8800 (dT), assuming no hypochromicity. Oligonucleotides comprising
the target duplexes were annealed as follows: 500 pmol each of oligonucleotides
A
and
B
for the Pur1 duplex or oligonucleotides
C
and
D
for the Pur2 duplex were mixed with 2 [mu]l 5 M NaCl and brought to a total volume of 42 [mu]l with H
2
O. This annealing reaction mixture was incubated at 75oC for 12 min and then gradually cooled to 25oC. Thirty pmol of the resulting oligonucleotide duplexes were
radiolabeled using the Klenow fragment of DNA polymerase I and [[alpha]-
32
P]dATP in the presence of 0.1 mM dGTP, dTTP and dCTP. The resulting labeled duplex oligonucleotides were purified by precipitation from ethanol in the
presence of ammonium acetate, and resuspended in H
2
O.
Four different binding buffers were employed in these studies: pH 5.0 (100 mM NaOAc, pH 5.0, 5 mM NaCl, 10 mM MgCl
2
); pH 7.2 (100 mM MOPS, pH 7.2, 6 mM MgCl
2
); pH 7.4 [also termed nuclear extract buffer, NEB; 20 mM Hepes, pH 7.4, 5 mM
MgCl
2
, 100 mM KCl, 10% (v/v) glycerol]; and pH 8.0 (25 mM Tris-HCl, pH 8.0, 6 mM MgCl
2
). Binding reaction mixtures contained labeled duplex (50 000 c.p.m.; ~0.1 pmol), either 1 [mu]l of 10* binding buffer or 2 [mu]l of 5* binding buffer, 1 [mu]l of 1 mg/ml yeast tRNA, 1 [mu]l of oligonucleotide (to yield the indicated
final concentration) and H
2
O in a final volume of 10 [mu]l. Reaction mixtures were incubated at 22oC for 5 h and were then supplemented with 1 [mu]l of an 80% glycerol solution containing bromophenol blue.
Reactions were analyzed by electrophoresis through 20% native polyacrylamide
gels (19:1 acrylamide:bisacrylamide) prepared in electrophoresis buffer [for pH 5.0: 100 mM NaOAc, pH 5.0, 1 mM MgCl
2
; for pH 7.2: 100 mM MOPS, pH 7.2, 6 mM MgCl
2
; for pH 8.0 (used for both pH 7.4 and 8.0 binding buffers): 100 mM Tris base,
110 mM boric acid, 2 mM EDTA, 8 mM MgCl
2
]. Electrophoresis was performed with recirculation at 4oC overnight (9 V/cm). The resulting gel was imaged and quantified by
storage phosphor technology using a Molecular Dynamics PhosphorImager.
The apparent fraction, [theta] of target duplex bound by oligonucleotide was calculated for each gel
lane using the definition:
[theta]
= S
triplex
/ (
S
triplex
+ S
duplex
)
1
where
S
triplex
and
S
duplex
represent the storage phosphor signal for triplex and duplex complexes
respectively. Values of the apparent triplex dissociation constant,
K
d
, were obtained by least squares fitting of the data to the binding isotherm:
[theta]
=
([O]
n
/
K
d
n
) / (1 + [O]
n
/
K
d
n
)
2
where [O] is the concentration of oligonucleotide, and
n
is the Hill coefficient (
25
).
Pur1 and Pur2 duplexes were each ligated into plasmid pG5E4T that had been
cleaved by
Bam
HI and
Pst
I (
11
). Clones bearing the desired insertions were confirmed by sequencing. A 356 bp
Hin
dIII-
Sac
I restriction fragment from the Pur1-containing plasmid and a 361 bp
Hin
dIII-
Sac
I restriction fragment from the Pur2-containing plasmid were then prepared and dephosphorylated with calf
intestinal alkaline phosphatase. The Pur1 and Pur2 fragments were uniquely end-labeled on strands
B
and
D
respectively, using polynucleotide kinase (see Fig.
1
). For some experiments, the Pur2 fragment was uniquely end-labeled on strand
C
using the Klenow fragment of DNA polymerase I (see Fig.
1
). Labeled fragment (50 000 c.p.m.; ~0.1 pmol) was incubated with either 1 [mu]l of 10* pH 8.0 binding buffer or 2 [mu]l of 5* nuclear extract buffer (NEB), 1 [mu]l of 1 mg/ml yeast tRNA, 1 [mu]l of 10 [mu]M oligonucleotide, and H
2
O in a final volume of 10 [mu]l. Binding reactions were incubated overnight at 22oC. Dimethyl sulfate [1 [mu]l of a 4% (v/v) aqueous solution] was added to each reaction
mixture and allowed to incubate for 30 min at 4oC. Reactions were terminated with 5 [mu]l of stop mix [1.5 M NaOAc, 7% (v/v) [beta]-mercaptoethanol and 100 [mu]g/ml yeast tRNA]. For formic acid treatment, 25 [mu]l of formic acid was allowed to incubate with reaction
mixture for 2 min at 22oC, then the reaction was terminated with 180 [mu]l of HZ stop mix (0.3 M NaOAc, 0.1 mM EDTA, 25 [mu]g/ml tRNA). Following ethanol precipitation, 100 [mu]l of 10% piperidine was added, and the samples were incubated
for 30 min at 90oC. Piperidine was removed by repeated lyophilization. The DNA was then resuspended in 5 [mu]l of formamide dye mix, heated to 90oC, electrophoresed on an 8% polyacrylamide sequencing gel [19:1
(acrylamide:bisacrylamide)] containing 7.5 M urea in 0.5* TBE buffer (50 mM Tris base, 55 mM boric acid, 1 mM EDTA), and imaged by
storage phosphor technology.
Studies of triple helix formation often employ non-physiological conditions such as acidic pH and/or low monovalent cation
concentrations. We wished to study triple helix formation under physiological
conditions at a natural target sequence. The
bcl-2
gene, whose product inhibits apoptosis, contains two potential triplex target sequences just upstream of the predominant promoter, P1 (Fig.
1
A;
26
). One target sequence, termed Pur1, is located 32 bp upstream of the 5'-most major transcription start site (position -1432; Fig.
1
A;
23
) and consists of a 25 base purine-rich sequence with three pyrimidine interruptions (Fig.
1
B). The second target sequence, termed Pur2, is located 136 bp upstream of the
first major transcription start site and consists of two adjacent homopurine
sequences; nine purines on one strand of the duplex directly adjacent to 21
purines on the opposite strand (Fig.
1
B). To facilitate electrophoretic mobility analyses of oligonucleotide binding
and to simplify cloning procedures, synthetic duplexes were designed containing
the Pur1 and Pur2 sequences (Pur1 duplex and Pur2 duplex respectively; Fig.
1
B).
The purine triple helix motif, most appropriate for guanine-rich target sites, was used to design triplex oligonucleotides to
recognize the Pur1 target sequence. Oligonucleotide
Pur1-Long
was designed to span the entire Pur1 region, with thymidine residues used to
cross interrupting pyrimidines in the target (Fig.
1
B).
Pur1-Short
, which recognizes only 19 bases of the target, was designed to avoid two of the
three interrupting pyrimidines of the target sequence (Fig.
1
B).
It has previously been shown that oligonucleotides utilizing consecutive
pyrimidine motif and purine motif sequences can simultaneously recognize purine
domains on opposite strands of duplex DNA, such as those in the Pur2 target
sequence (
27
-
31
). Target recognition is accomplished by switching triplex motifs, and thus
relative strand polarities, as the oligonucleotide crosses between purine
domains. The pyrimidine triple helix motif, appropriate for an adenine-rich target site, was used to design oligonucleotide
Pur2-Pyr
to recognize the 19 base domain of the Pur2 target sequence (Fig.
1
B).
Pur2-Pur
was designed to recognize the nine base domain of the Pur2 target using the purine motif (Fig.
1
B). These two oligonucleotides were then combined in the design of
Pur2-Cross
intended to simultaneously recognize both purine domains of the Pur2 target sequence (Fig.
1
B). Using the strategy devised by Beal and Dervan (
27
), a cytidine residue was removed from the 5'-pyr motif:pur motif-3' junction of
Pur2-Cross
to allow the oligonucleotide to more smoothly cross the major groove between the
alternate strand homopurine regions. In addition, all cytidines in
oligonucleotides for triple helix formation were modified to 5-methylcytidines to enhance binding at physiological pH (
32
). Electrophoretic mobility shift titrations allowed estimation of
oligonucleotide binding affinities for
bcl-2
duplex targets, while dimethyl sulfate (DMS) footprinting assays were used to
study details of oligonucleotide binding.
Table 1
Equilibrium dissociation constants (
K
d
s) were measured for triplexes in the
bcl-2
promoter at various pH values in the range 5.0-8.0. Typically, pyrimidine motif triplexes require cytidine protonation
and are stabilized by acidic pH, while the stabilities of purine motif
triplexes are thought to be intrinsically pH independent. Figure
2
A and B presents examples of binding experiments at pH 5.0 and 8.0 respectively,
in which labeled Pur1 and Pur2 duplexes were incubated in the presence of
increasing oligonucleotide concentrations. Figure
2
C and D depicts quantitative results from these experiments. Values of
K
d
were calculated as described in Materials and Methods and are listed in Table
1
. At pH 5.0, purine motif oligonucleotides
Pur1-Long
and
Pur1-Short
bound the Pur1 duplex target weakly with
K
d
values >>2.5 * 10
-6
M (Fig.
2
A and C). As the pH was increased, oligonucleotide binding in the purine motif increased, with
K
d
values of 2.3 * 10
-7
M and 4.7 * 10
-7
M for
Pur1-Long
and
Pur1-Short
respectively, at pH 8.0 (Fig.
2
B and D). This result contradicts the conventional view that triple helix
formation in the purine motif is pH independent. However, the basis for
suppression of purine motif triplexes at low pH remains unclear.
To verify specificity and monitor any changes in target structure upon
oligonucleotide binding to the Pur1 and Pur2 duplexes, a DMS footprinting
analysis was performed. Protection of guanine N7 from DMS modification is conferred by triple helix formation. After cloning into plasmids, targeted guanines in the Pur1 sequence were specifically, (though weakly) protected by 1 [mu]M of both
Pur1-Long
and
Pur1-Short
in pH 8.0 binding buffer (Fig.
4
A, lanes 3 and 4 respectively). Although this weak footprint indicates a relatively
low binding affinity, the high level of background cleavage of this sequence
(Fig.
4
A, lane 1) should also be noted. The slight hypermethylation of guanines just 3' of the
Pur1-Short
binding domain is a phenomenon often seen at sequences adjacent to triplex
binding sites and may reflect structural changes in the duplex at the duplex-triplex junction.
We have used the purine and pyrimidine triple helix motifs to design several
oligonucleotides to recognize two target sequences in the
bcl-2
P1 promoter. At physiological pH, oligonucleotides
Pur2-Pyr
and
Pur2-Cross
bound the Pur2 target sequence with
K
d
values near 1 * 10
-7
M, affinities ~10-fold higher than observed for oligonucleotides
Pur1-Long
and
Pur1-Short
binding to the Pur1 target sequence.
Pur2-Cross
, designed to bind alternate strand homopurine domains, had the highest affinity
of all oligonucleotides tested under pseudo-physiological conditions (
K
d
= 6.1 * 10
-8
M).
In the course of binding studies under various pH conditions, we were surprised
that purine motif oligonucleotides
Pur1-Long
and
Pur1-Short
bound the Pur1 duplex very weakly at low pH. Although there is no intrinsic pH
requirement for reverse Hoogsteen hydrogen bonding in the purine motif, this
study clearly demonstrates a strong pH dependence for triplexes at this
sequence. Triplexes involving
Pur1-Long
and
Pur1-Short
were also seen to be inhibited by physiological levels of monovalent cations, particularly K
+
(data not shown). This is in agreement with previous studies that demonstrate
purine motif triplex inhibition by monovalent cations (
37
-
39
). Such inhibition is most likely caused by competing equilibria wherein the
guanine-rich oligonucleotides become sequestered into guanine quartet aggregates (
39
). Incorporation of 6-thioguanine into a limited number of positions of purine motif
oligonucleotides has been shown to alleviate ion inhibition of triplexes (
40
,
41
) and could potentially be used as a strategy to permit
Pur1-Long
and
Pur1-Short
to form triplexes under pseudo-physiological conditions.
In contrast with the purine motif oligonucleotides, pyrimidine motif
oligonucleotides
Pur2-Pyr
and
Pur2-Cross
not only tolerated physiological ion concentrations, but were remarkably pH independent, being stable at pH 8.0. This result is notable in light of the
requirement for cytidine protonation in the pyrimidine motif. The presence of 5-methylcytidine (Me
5
C) in
Pur2-Pyr
and
Pur2-Cross
is likely to be a large contributor to this pH independence. Previous studies
have indicated that Me
5
C substitution allows certain oligonucleotide sequences to form triplexes at pH
values >7, though the mechanism of stabilization is unclear (
32
). However, the pH independence of
Pur2-Pyr
and
Pur2-Cross
is also sequence dependent, as triplexes involving other pyrimidine motif
oligonucleotides containing Me
5
C are destabilized at neutral pH (data not shown). Perhaps thymidine-rich sequences lacking consecutive cytidines are most favorable for
triplex stability >pH 7. Together with the results of previous studies (
3
-
5
,
8
), these data reinforce the view that conventional purine motif triplexes are
stabilized by profoundly guanine-rich target strands, while conventional pyrimidine motif triplexes are
stabilized by adenine-rich target strands punctuated with isolated guanine residues.
Another key result from these studies is the high affinity binding of alternate
strand homopurine domains in the
bcl-2
promoter by
Pur2-Cross
. In particular,
Pur2-Cross
recognizes the nine base homopurine domain of the Pur2 duplex to which
oligonucleotide
Pur2-Pur
could not detectably bind. Moreover, the simultaneous recognition of both
homopurine domains of the Pur2 duplex increases the affinity of
Pur2-Cross
almost 2-fold relative to the affinity of
Pur2-Pyr
under pseudo-physiological conditions. These results support previous studies in which
alternate strand triple helix formation was shown to allow binding of adjacent
homopurine domains of <10 bases when oligonucleotides could not bind either domain individually (
27
). In addition, the present study provides evidence that alternate strand
binding can increase the stability of a triple helix relative to binding a
single domain of >10 bases. A previous study of alternate strand binding to
domains >10 bases demonstrated a modest 1.4-fold increase in affinity relative to binding a single domain (
31
).
DMS hypermethylation of guanines adjacent to duplex-triplex junctions is a common observation that presumably reflects
perturbations in the duplex structure upon oligonucleotide binding (
31
,
42
). An alternative hypothesis to explain the observed hypermethylation is that
the terminal unstacked bases at triplex/duplex junctions creates a hydrophobic
microenvironment that increases the local DMS concentration. Binding of
Pur1-Short
and
Pur2-Pyr
to target duplexes in this study promoted hypermethylation of the purine strand
3' to the complex. In the case of
Pur2-Pyr
, hypermethylation of a pair of guanines was also seen on the opposite DNA
strand at the same end of the complex. Interestingly, binding of
Pur2-Cross
caused an increase in hypermethylation at junction guanines on both strands of
the target duplex. This suggests a strain or distortion in the duplex when
Pur2-Cross
traverses the major groove between alternate strands, thus causing the major
groove to be more accessible to DMS. Such a distortion could reflect an energy
expense associated with alternate strand binding (
31
). Optimized oligonucleotide designs might reduce this apparent distortion.
Together, the results of this study indicate that a natural sequence in the
human
bcl-2
promoter can form a stable triplex with a synthetic oligonucleotide under
pseudo-physiological conditions. Moreover, it is promising that this
oligonucleotide can be modified to increase nuclease resistance without a
significant decrease in binding affinity. Extension of these studies to intact
cells will require attention to several obstacles: oligonucleotide delivery to
the nucleus, availability of DNA target sites within chromatin structure, and
the unknown effects of these triplex complexes on
bcl-2
transcription.
In vitro
studies have shown that occlusion of transcription factor binding sites by
triplexes can block transcription initiation (
11
,
43
,
44
). Other studies indicate that direct overlap of protein and triplex binding
sites may not be necessary for transcriptional inactivation by triplexes (
2
,
11
,
13
). The latter results suggest that oligonucleotide binding may alter duplex DNA
structure in some manner (bending, stiffening) so as to antagonize promoter
function (
11
). It is provocative to note that both Pur1 and Pur2 target sequences are
located on or near consensus Sp1 binding sites within the P1
bcl-2
promoter. Further studies will be necessary to determine if triple-helical complexes identified in the present study will offer a feasible
approach to transcriptional inhibition of the
bcl-2
gene in living cells.
We gratefully acknowledge D. Eicher and C. Mountjoy for their excellent
technical assistance. This work was supported by NIH grant GM 47814 and a
Junior Faculty Research Award from The American Cancer Society to L.J.M. W.M.O.
is supported by a University of Nebraska Medical Center Presidential
Fellowship.
+
Present address: Department of Molecular and Cellular Biology, University of
Arizona, Tucson, AZ 85721, USA
ODN
pH
M
+
K
d
(M)
Relative affinity
b
Pur1-Long
5.0
-
>>2.5 * 10
-6
<<0.02
7.2
-
1.3 * 10
-6
0.05
8.0
-
2.3 * 10
-7
0.27
Pur1-Short
5.0
-
>>2.5 * 10
-6
<<0.02
7.2
-
1.0 * 10
-6
0.06
8.0
-
4.7 * 10
-7
0.13
Pur2-Pur
5.0
-
>>2.5 * 10
-6
<<0.02
7.2
-
>>2.5 * 10
-6
<<0.02
8.0
-
>>2.5 * 10
-6
<<0.02
Pur2-Pyr
5.0
-
4.4 * 10
-8
1.4
7.2
-
1.1 * 10
-7
0.55
7.4
100 mM K
+
1.1 * 10
-7
0.55
8.0
-
1.0 * 10
-7
0.61
Pur2-Cross
5.0
-
5.3 * 10
-8
1.2
7.2
-
8.0 * 10
-8
0.76
7.4
100 mM K
+
6.1 * 10
-8
1.0
8.0
-
7.0 * 10
-8
0.87
Pur2-Modified Cross
8.0
-
9.9 * 10
-8
0.62
REFERENCES
Return