ABSTRACT
Exon 5 of the human
aprt
gene contains an oligopurine-oligopyrimidine stretch of 17 bp (5
'
-CCCTCTTCTCTCTCCT-3'
) within the coding region. (T,C)-, (G,T)- and (G,A)-containing oligonucleotides were compared for their ability to
form stable triple helices with their DNA target. (G,T) oligodeoxynucleotides,
whether parallel or antiparallel, were unable to bind to this sequence. This is
in contrast to (G,A) (purine) and (T,C) (pyrimidine) oligonucleotides, which bind to the duplex at near neutral pH. Binding was highly sequence specific, as unrelated
competitors were unable to interfere with target recognition. A major
difference between the purine and pyrimidine oligodeoxynucleotides was observed
in the kinetics of binding: the (G,A) oligonucleotide binds to its target much
faster than the (T,C) oligomer. With the purine oligonucleotide, complete binding was achieved in a matter of minutes at micromolar concentrations,
whereas several hours were required with the pyrimidine oligomer. Thus, the
general observation that triplex formation is slow with pyrimidine
oligodeoxynucleotides does not hold for (G,A) oligodeoxynucleotides. Purine and
pyrimidine oligodeoxynucleotides covalently linked to a psoralen group were
able to induce crosslinks on the double-stranded DNA target upon UV irradiation. This study provides a detailed comparison of the different types of DNA triplexes under the same experimental conditions.
Triple helices were first observed in 1957 for polyribonucleotides (
1
). These polymer interactions suggested an approach to double-stranded DNA recognition called oligonucleotide-directed triple helix formation, whereby short oligonucleotides bind
to the major groove of double helical DNA at specific sequences (
2
,
3
). Triple helix-forming oligonucleotides (TFOs) can compete with the binding of proteins (
4
,
5
) and affect transcription of a specific gene (
6
-
11
). TFOs have been used to target mutations to specific sites in selectable
genes, in order to achieve permanent and inheritable changes in gene expression
(
12
,
13
). One of the best characterized selectable genes is the
aprt
gene and well-defined selection procedures allow detection of mutagenic events. APRT-/- mutants are selectable in a medium containing a drug such as 2,6
diaminopurine, which is converted to a lethal metabolite by the APRT enzyme.
The human
aprt
gene is 2.6 kb in length, contains 5 exons and is located on chromosome 16 (
14
). In order to develop an oligonucleotide-directed mutagenesis assay, we first investigated triplex formation on a double-stranded DNA fragment that mimicked part of the
aprt
exon 5 sequence.
At least two structural classes of triple helices exist that differ in sequence
composition of the third strand and relative orientations of the phosphate-deoxyribose backbones. In the first class, the pyrimidine motif, the
third strand binds parallel to the purine strand of the duplex, forming T.A*T
and C.G*C+ triplets (
2
,
3
,
15
,
16
). The p
K
a of the imino group of cytosine, which must be protonated (
17
), is well below 7, making these triplexes pH dependent (
18
,
19
). Despite this handicap, DNA triple helices of the first motif have been
reported at neutral pH, especially when cytosines on the third strand are
replaced by 5-methylcytosines (
20
,
21
). In the second class, the purine motif, guanine-rich oligonucleotides bind to the purine strand of the duplex by reverse
Hoogsteen hydrogen bonds. Purine-rich third strands can be designed to form either C.G*G and T.A*A triplets
[(G,A) oligonucleotides] or C.G*G and T.A*T triplets [(G,T) oligonucleotides].
It should be noted that oligonucleotides containing G and T can also bind in a
parallel orientation with respect to a polypurine target sequence, forming
Hoogsteen C.G*G and T.A*T base triplets (
22
,
23
). Therefore, the orientation of (G,T)-containing oligonucleotides is sequence dependent. These triple helices
often require a high divalent cation concentration. Additionally, the third strand, which is G rich, is prone to form intra- or intermolecular structures involving the formation of G tetrads. Such structures are stabilized by physiological concentrations of monovalent cations, such as potassium, making this self-association likely
in vivo
. As a result, the ability of the TFO to bind to its intended target is reduced
(
24
-
27
). Purine oligonucleotides also have the potential to form another competing
structure, a G.A parallel duplex (
28
). Thus, the optimal choice for triplex formation under physiological conditions
is not straightforward.
This work was undertaken first to determine what target would be optimal on the
aprt
gene and then what type(s) of oligonucleotide [(T,C), (G,T) or (G,A)] would be the best choice to form a triplex with that
particular target under near physiological conditions. Four parameters were
taken into account for such a selection: the possibility of self-association, the affinity of binding, the kinetics of triplex association/dissociation and the specificity of formation. Previous reports have compared (G,A) and (G,T) oligonucleotides, but results have not demonstrated a clear or consistent trend. Articles
dealing with a comparison of all three types of oligodeoxynucleotides are rare
(
29
). In this report, we compare all three kinds of TFOs for their ability to bind an oligopurine-oligopyrimidine sequence of exon 5 in the human
aprt
gene.
Unmodified oligodeoxynucleotides were synthesized by Eurogentec (Belgium) on the 0.2 [mu]mol scale and treated as in Mergny
et al.
(
30
). All other oligodeoxynucleotides (with 5' or 3' acridine and/or psoralen substitutions) were synthesized by
Appligene (Strasbourg, France) and used without further purification.
Concentrations of all oligodeoxynucleotides were estimated by UV absorption at 30oC in a pH 8.0 buffer, using the sequence-dependent absorption coefficients given by Cantor and Warshaw (
31
). All concentrations were expressed in strand molarity. Genomic DNA was
extracted from WR10 (a human lymphoblastoid cell line, a gift of Dr Tatsumi, Japan) using the QIAamp blood kit (Quiagen
Inc.).
All experiments were performed as in Noonber
et al
. (
28
). A hysteresis phenomenon was obtained. Such behavior, which is the result of
slow association and dissociation kinetics, has already been described and
analyzed for triple helix formation (
32
).
Oligodeoxynucleotides were labeled with T4 polynucleotide kinase (New England Biolabs) and [
32
P]ATP (ICN). Radiolabeled double-stranded duplex (10 nM strand concentration) was incubated at 4oC in the presence or absence of a large excess of an unlabeled
oligodeoxynucleotide third strand (1-20 [mu]M strand concentration) in 50 mM HEPES buffer, pH 7.2, with 10 mM
MgCl
2
. After incubation for 0.3-48 h, samples were loaded on a non-denaturing 12% polyacrylamide gel. Migration lasted 2-4 h at 4oC. Gels were dried and analyzed on a phosphorimager instrument (Molecular Dynamics). For the specificity
experiments, the double-stranded competitor (purified human genomic DNA or a synthetic 36 bp
kallikrein duplex) was mixed with the specific labeled duplex prior to third
strand addition.
DNase I footprinting of the triplex site upon addition of the third strand was
performed on a 36 bp duplex. The
32
P-labeled purine strand (10 nM, 10
4
c.p.m.) was pre-incubated with its complementary strand at 37oC in a 120 [mu]l solution containing 20 mM Tris-HCl, pH 7.2, 5 mM MgCl
2
, 50 mM NaCl, 0.5 mM CaCl
2
. The TFOs were added (15 [mu]M) at 4oC. After a variable incubation time (from 0.75 to 48 h), DNase I was added (0.2 [mu]g/ml final concentration) and the digestion was stopped after 20 s by
freezing at -80oC in the presence of 5 vol. ethanol. After ethanol precipitation,
the digests were separated on a 15% denaturing polyacrylamide gel and analyzed.
UV crosslinking experiments upon addition of the psoralen third strand were
performed on a 36 bp
aprt
duplex or with the 36 bp kallikrein duplex. The
32
P-labeled purine or pyrimidine strand (both 10 nM) was pre-incubated with its complementary strand (20% molar excess) at 37oC in 50 mM HEPES buffer, pH 7.2, containing 130 mM KCl and 10
mM MgCl
2
. The TFOs were added (10 [mu]M). The mixture was heated to 50oC over 30 min and then incubated at 37oC for 90 min. UV irradiation was performed with an Orion lamp
(200W Xe Hg lamp, after removal of wavelengths below 310 nm through a glass
plate) for 10-300 s. After ethanol precipitation, the reaction products were separated
on a 15% denaturing polyacrylamide gel.
Triplex formation is mostly limited to oligopurine-oligopyrimidine stretches. Using GenBank, we have searched for sites within the human
aprt
gene that would bind to a (T,C), (G,T) or (G,A) oligonucleotide. Sequences were
scanned for those containing a stretch of at least 12 contiguous purines or 12
contiguous pyrimidines. As psoralen-linked oligonucleotides mostly induce single base pair substitutions (
12
,
13
,
33
), we decided to exclude all intronic sequences, where a point mutation is not
likely to induce a change of phenotype. Only coding sequences and exon-intron boundaries were analyzed. We have identified a site within exon 5
(positions 2702-2718, according to Broderick
et al.
;
14
) of the human
aprt
gene that is susceptible to forming a stable triplex with a TFO. The
oligopurine-oligopyrimidine region is 17 bp long and contains nine C.G and eight A.T
base pairs. The sequence of the human
aprt
coding region is 90% identical to the hamster sequence. However, a single base
pair difference in the hamster homolog disrupts the oligopurine stretch in exon
5: the homologous hamster exon has only 14 purines in a row, which could
explain why this site was not favorable for the hamster homolog (
34
).
Different types of oligodeoxynucleotides could form a triplex with this duplex.
We analyzed the interactions of several single-stranded oligodeoxynucleotides: 17TC, 5'-TCCTCTCTCTTCTTCCC; 17GA, 5'-GGGAAGAAGAGAGAGGA; 17GTa, 5'-GGGTTGTTGTGTGTGGT; 17GTp, 5'-TGGTGTGTGTTGTTGGG (see
Fig.
1
). The orientation of (T,C) and (G,A) TFOs in triple helical complexes is well
established. Therefore, the (T,C) oligodeoxynucleotide was synthesized in a
parallel orientation to the purine strand, whereas the (G,A) oligonucleotide was synthesized in the reverse orientation, i.e. antiparallel to the purine
strand. This is in contrast to (G,T) oligonucleotides, which may bind in both
orientations depending on the sequence context (
22
). Only the (G,T) oligonucleotide was synthesized in both orientations, both
parallel (17GTp, 5'-d-TGGTGTGTGTTGTTGGG-3') and antiparallel (17GTa, 5'-d-GGGTTGTTGTGTGTGGT) to the
oligopurine strand.
Many cytosine-rich pyrimidine oligodeoxynucleotides are potentially able to form an intramolecular i-motif (
30
). No structure was observed for the 17TC and 17TC
m
oligonucleotides, probably because most cytosines are dispersed in the primary
sequence.
We did not observe any self-associated structure for the (G,T) oligonucleotides (17GTa and 17GTp).
Figure
2
(open triangles) illustrates the monotonous behavior of the absorbance of the
17GTa oligodeoxynucleotide as a function of temperature. No anomaly of
migration was observed on a non-denaturing gel (not shown). The absence of self-association for such a (G,T) oligonucleotide is probably the result
of dispersion of guanines in the TFO: the largest block of continuous guanines
is only three bases long and most of the sequence is composed of (G,T) dinucleotides. This sequence is not favorable to guanine quadruplex formation.
The binding of a pyrimidine third strand into the major groove of a duplex is
usually accompanied by hypochromism at 260 nm. Thus, the process of triplex
formation can be monitored by measuring the change in absorbance as a function
of temperature. A melting temperature of 15oC was obtained for the 17TC oligonucleotide at 1 [mu]M strand concentration in a pH 6.8 buffer containing 10 mM sodium cacodylate, 10 mM MgCl
2
and 140 mM KCl. Replacement of cytosines by 5-methylcytosines stabilized the triplex by >10oC (17TC
m
,
T
m
= 28oC under the same experimental conditions). Raising the pH by only 0.4 units
(i.e. to 7.2) led to disappearance of the transition corresponding to the
triplexes for all pyrimidine TFOs (17TC, 18TC, 17TC
m
and 18TC
m
), showing that these triple helices are strongly pH dependent.
Psoralen-conjugated oligonucleotides are able to crosslink a double-stranded DNA target at a 5'-TpA site (
35
) and such a site is present close, but not adjacent, to the oligopurine
stretch. Therefore, we measured the stability of triple helices containing an
extra non-canonical base triplet (on the left side of Fig.
1
A). This extra base on the third strand would be of interest to reach a nearby
TpA site. For the pyrimidine oligonucleotide, we incorporated a G at the 3'-end of the TFO. This guanine should form an A.T*G base triplet,
which is of reasonable stability (
36
). Binding of the 18mer was compared with binding of the corresponding 17mers. A small increase in the melting temperature was observed, suggesting that the A.T*G triplet is formed (
T
m
= 18oC, [Delta]
T
m
= +3oC). Similar results were obtained when comparing 17TC
m
and 18TC
m
(data not shown).
No melting transition was observed with the 17GTa, 17GTp, 17GA and 18GA
oligodeoxynucleotides. Triple helix formation without hypochromicity has
already been described in a few cases (
37
). The absence of a spectral transition does not necessarily exclude triplex
formation. Triplex formation must be evidenced using other techniques, such as
gel retardation assays and/or footprinting experiments.
UV melting experiments suggested that pyrimidine oligonucleotides were able to form a triplex, whereas footprinting experiments suggested that (G,A) oligonucleotides formed a stable triple helix. A
third technique was thus initiated to confirm triplex formation. A retarded
band is expected on a polyacrylamide non-denaturing gel upon third strand binding. No retarded band was ever
observed with (G,T) third strands (17GTa or 17GTp), even at very high strand
concentration (30 [mu]M; data not shown). Thus, no triplex was found for this family of TFO by
three different techniques.
In contrast, a retarded band was observed with the (T,C) and (G,A) oligonucleotides (Fig.
4
A and B respectively), in agreement with triplex formation. The amount of triplex formed was measured after various
incubation times. The thermodynamic properties of triple helices involving a
pyrimidine oligonucleotide have been thoroughly investigated. Fewer experiments
have been performed to determine kinetic parameters of triplex formation. As
shown in Figure
4
C, incubation times of several hours at 4oC were required to obtain complete triplex formation with the 17TC
oligonucleotide (3 [mu]M strand concentration) in a 50 mM HEPES, pH 7.2, 10 mM MgCl
2
buffer. As expected, an increase in the third strand concentration led to
faster binding. In contrast, (G,A) oligonucleotide binding was maximum with a very short incubation time (20 min). Even if heat denatured for 20 min just prior to loading, a sample containing the (G,A) TFO and the duplex showed
partial but significant triplex re-formation (19%). The minor but significant and reproducible decrease in
binding over time (Fig.
4
C) for very long periods is probably the result of competing self-association(s) of the (G,A) third strand.
This result, as well as the footprint obtained after only 45 min, indicated that
(G,A) oligonucleotides indeed formed a triplex and that the formation was fast
compared with the (T,C) oligodeoxynucleotide. Twenty minutes were sufficient to
obtain optimal binding at 3 [mu]M strand concentration of the (G,A) TFO, whereas 20 h were necessary for the
(T,C) TFO at neutral pH. This indicates that (G,A) triplexes form at a much
faster rate than (T,C) triplexes.
Sequence-specific binding of purine as well as pyrimidine oligodeoxynucleotides has
already been established. We wanted to determine whether the specificity of our
TFOs would be sufficient to recognize a complementary target in the presence of
a very large excess of competitor. We assayed two types of competitors: (i) a
target duplex with two point substitutions; (ii) total human genomic DNA. The sequence of the mutated duplex is shown in Figure
1
B. It partially corresponds to the sequence of the polypurine-polypyrimidine stretch found in the sequence of the human kallikrein
gene. The labeled
aprt
target duplex (10 nM strand concentration) was first mixed with increasing
amounts of an unlabeled competitor duplex (the kallikrein duplex shown in Fig.
1
B). The 18TC
m
or 18GA TFOs were then added. The mixture was incubated for 15 h at 4oC then loaded on a 15% non-denaturating polyacrylamide gel at 10oC. The retarded band corresponding to the triplex was completely
unaffected by the presence of the non-specific competitor (data not shown). It is interesting to note that the
18mers still bind to their targets with high specificity. Thus, the extra base
on the TFO does not have any deleterious effect on the specificity of binding.
Even a large excess (1000 times or more) of a closely related duplex was unable
to displace a significant proportion of the third strand.
The linkage of DNA ligands, especially intercalators, has been previously
described to enhance the affinity of TFOs. Covalent linkage of an acridine
derivative at one end of the TFO was shown to increase the binding affinity (
38
,
39
). The presence of an acridine moeity at the 5'-end of a (T,C) TFO strongly increased the stability of the triplex:
at pH 7.2 or above, no significant triplex formation was evident with the 18TC
m
TFO. A small hyperchromism at 260 nm could only be obtained with a long
incubation time at 0oC before heating. In contrast, Acr-18TC
m
melts in a non-reversible fashion, with a
T
m
of 25oC at pH 7.2 measured on the dissociation curve. Concerning the (G,A) TFOs,
gel retardation experiments were performed at different strand concentrations
in 50 mM HEPES, pH 7.2, containing 10 mM MgCl
2
. Apparent
K
d
values, derived from the third strand concentration necessary to obtain 50%
triplex formation at 4oC on a non-denaturing gel, were determined to be 0.3 [mu]M for the 18GA oligonucleotide and 0.03 [mu]M for the 18GA-Acr oligonucleotide (data not shown). Thus, the presence of an
acridine group at the 3'-end of the purine TFO increases its affinity by a factor of 10.
Doubly substituted TFOs, bearing an acridine group at one end and a psoralen
group at the other end, were tested for their abilities to induce crosslinks at
a nearby, but not adjacent, site (Fig.
1
A). The psoralen group was linked to the 5'-end of the (G,A) oligodeoxynucleotide and the 3'-end of the (T,C) oligodeoxynucleotide. Both oligonucleotides included an extra base towards
the 3'-side of the polypurine target sequence, to allow psoralen
intercalation at the TpA site. Comparison of the UV melting profiles of the Acr-18TC
m
and Acr-18TC
m
-Pso TFOs demonstrated that no further gain in stability was obtained with
the doubly substituted pyrimidine TFO as compared with the acridine conjugate.
In a similar fashion, the apparent
K
d
values for the Pso-18GA-Acr (
K
d
= 20 nM) and 18GA-Acr (
K
d
= 30 nM) TFOs were almost identical, suggesting that the psoralen group did not
play an important role in stability of the purine and pyrimidine triple
helices.
The ability of the psoralen-substituted TFOs to induce crosslinking of their double-stranded DNA target upon UVA irradiation was tested by denaturing gel electrophoresis. One of the strands of
the 36 bp
aprt
duplex was radioactively labeled and the duplex was formed upon addition of the
unlabeled complementary strand. The psoralen TFO was then added and the mixture
was incubated for a variable time (between 2 and 45 min), then UV irradiated.
Figure
5
shows that both (G,A) and (T,C) TFOs could induce a significant number of
crosslinks of the target after 90 min incubation at 37oC in a pH 7.2 buffer. The presence of an acridine moeity at the other end
of the TFO was not necessary to observe crosslinks. An increase in the fraction
of slowly migrating species was nevertheless obtained with Acr-18TC
m
-Pso TFO compared with 18TC
m
-Pso, suggesting that the acridine group stabilized the interaction between
the pyrimidine TFO and its target sequence, which in turns facilitates
positioning of the psoralen group at the TpA site. No such increase in
crosslinking efficiency was observed for the Pso-18GA-Acr oligonucleotide, when compared with the Pso-18GA TFO: under these conditions, the Pso-GA TFO, without any acridine, induces a large fraction
of slowly migrating species. No crosslinks were obtained with a mutated target duplex (Fig.
5
, right), demonstrating that psoralen crosslinks were abolished by the presence of only two mismatches on the triplex sequence.
Thus, even doubly substituted TFOs (bearing one acridine and one psoralen
group) are able to discriminate between their specific target and a related
sequence.
We have tested three different types of oligonucleotides for their ability to
bind to a 17 bp oligopurine-oligopyrimidine sequence present in exon 5 of the human
aprt
gene. (G,T) oligodeoxynucleotides, whether parallel or antiparallel, were unable to bind to the target
sequence. This result is apparently surprising, as many studies have shown that
(G,T) oligonucleotides could bind to double-stranded DNA. This rather weak binding is certainly the result of the
primary sequence of the
aprt
oligopurine stretch: the corresponding third strand contains a large number of
TpG and GpT steps. As the C.G*G and T.A*T base triplets are not isomorphous,
every change of base on the TFO leads to a distortion of the third strand and
thus to an energetic penalty. A comparison with the literature shows that (G,T)
oligonucleotides which are able to form a triple helix are usually G rich
rather than T rich, leading to the formation of a majority of C.G*G triplets,
and (more importantly) these guanines are clustered, thus reducing the number
of unfavorable GpT or TpG steps in the third strand. Long T stretches favor a
parallel orientation of the (G,T) oligonucleotides, whereas long G stretches
favor an antiparallel orientation (
40
). Thus, (G,T) oligonucleotides do not constitute a universal solution to triplex formation on any oligopurine-oligopyrimidine sequence.
(G,A) oligonucleotides (i.e. purine oligonucleotides) showed high affinity and
specificity for the
aprt
sequence. Their main advantage over pyrimidine TFOs did not only rely on their
pH-independent binding, but also on their kinetics of binding. Complete
binding required several hours with an unmodified pyrimidine oligonucleotide
(30 [mu]M strand concentration) at neutral pH. Under the same experimental
conditions, the (G,A) oligonucleotide was fully bound to its target in a matter
of minutes. No triplex formation at all was obtained with the (T,C) third
strand under the same conditions (data not shown).
Having in mind the formation of triple helices
in vivo
, we tried in these
in vitro
experiments to maintain several important parameters close to physiological
conditions: the pH was kept at ~7 in all experiments. The exact pH value
in vivo
will play a decisive role in the formation of a pyrimidine triplex, as a pH
increase of 0.4 units was enough to destabilize the pyrimidine triplexes
in vitro
. The length, sequence and mismatch content of the triplex was of course defined
by the sequence context of the
aprt
gene. As a consequence, the triplexes presented in this study were not very
stable at 37oC. Relatively few studies have been performed to compare purine and
pyrimidine TFOs on the same target (
29
,
41
). The choice of an oligodeoxynucleotide for triplex formation must take into
account not only its ability to form a triplex, but also its ability to form an
undesired, competing structure, such as G-quartets, (G,A) parallel duplexes or i-DNA. Self-association of the third strand does not necessarily prevent
triplex formation. In our study, the (G,A) TFO did form a stable homoduplex,
but was still able to recognize the duplex target. (G,A) duplex formation does
interfere somehow with triplex formation and might explain why complete
saturation of the duplex was not observed at high concentrations of the third
strand. Such a competing structure may even lead to a situation where the
triple helix is more readily formed at 37 than 4oC (
28
). (G,T) oligonucleotides were excluded, as none of them had a measurable
affinity for the
aprt
sequence
in vitro
. It is thus interesting to note that the only oligonucleotide that self-associated
in vitro
[the (G,A) TFO] is the one with the greater affinity for the duplex. (T,C)
oligonucleotides can form triple helices at neutral pH, but with a limited
stability. The presence of an acridine group at one end of the pyrimidine third
strand does improve the stability of the triplex, but the melting temperature
of this TFO is still below 37oC at pH 7.2, making triplex formation with this type of TFO uncertain
in vivo
and stimulating the search for chemical modifications of the third strands (2'-O-methyl, phosphoramidates, etc.) that might improve the stability of these triplexes.
The mutagenic potential of unmodified TFOs is significant, although they do not
induce any direct covalent damage in their double-stranded targets (
42
). They probably induce mutations through a transcription-coupled repair mechanism. The mutagenic potential of unmodified (G,A) TFOs was nevertheless limited compared with
psoralen-substituted (G,A) TFOs, which induce 10 times more mutations in the
supF
gene (
33
). A psoralen moiety can easily be attached to one end of the TFO and can be
activated by UV, in order to form adducts with its DNA target (
35
). The most reactive base sequence for psoralen photoaddition is at the 5'-TpA-3' site. A TpA site is present in exon 5 of the human
aprt
gene, close to the triplex site.
To reach this putative psoralen intercalation site, we investigated TFOs having an extra base towards the 3'-end of the purine strand of the triplex (Fig.
1
A). We showed that the third strand can be extended without any energetic
penalty (the extra mismatched triplet even provides extra stability to the
triplex). Extension of a triplex region can thus be locally achieved as long as
a canonical triplex is present, to `nucleate' the formation of the longer
triple helix (
43
). We have shown that
in vitro
a pyrimidine oligodeoxynucleotide linked at its 3'-end to a psoralen group is able to induce a double-stranded crosslink on the DNA target at the TpA site. Similar
results were obtained with a purine (G,A) TFO linked at its 5'-end to psoralen. The crosslinking experiments confirmed that (G,A)
and (T,C) TFOs bind to their double-stranded DNA targets and are able to induce covalent modifications on a neighboring but not adjacent TpA site. The psoralen group, which
provides a high chemical photoreactivity, does not increase the binding
affinity of the third strand in this system, probably because the TpA site is
not directly adjacent to the triplex site. This experiment shows that the
presence of a TpA site at the triplex-duplex junction is not an absolute requisite for psoralen crosslinking; a
neighboring TpA site which is close to the duplex-triplex junction can be reached by the psoralen group, provided that one
(or more) extra base(s) is present at the corresponding end of the TFO. Such an
additional nucleotide, which cannot lead to the formation of canonical base
triplets, may still play a limited role in triplex stability and does not
prevent the TFO from being highly sequence specific.
In summary, we have shown that, in the sequence context of exon 5 of the
aprt
gene, triple helix formation is possible with (T,C) and (G,A) TFOs. A major
difference between these two classes of triple helices rests upon their
kinetics of binding to their double-stranded target. Both psoralen-linked purine and pyrimidine TFOs were able to induce crosslinks in the
aprt
sequence at a TpA site not adjacent to the duplex-triplex junction. Experiments are now being conducted in cell cultures,
using human cell lines heterozygous for the
aprt
gene, doubly substituted TFOs (having one end linked to acridine, the other to
psoralen) and UVA irradiation. On the basis of the present
in vitro
experiments, (G,A) and (T,C) oligonucleotides, but not (G,T) TFOs, are
currently under investigation to determine their abilities to induce site-directed mutagenesis in an endogenous gene.
REFERENCES
Return