ABSTRACT
We used a DNA duplex formed between the 5
'
end of a 69mer (69T) and an 11mer (OL7) as a substrate for
Bam
HI. The former oligonucleotide folds into a hairpin structure, the stem of which
contains a stretch of pyrimidines in one strand and consequently a stretch of
purines in the other strand. The oligomer 69T was used as a target for
complementary oligodeoxypyrimidines made of 10 nt (OL1), 16 nt (OL5) or 26 nt
(OL2) which can engage the same 10 pyrimidine-purine-pyrimidine triplets with the 69T hairpin stem. Although the binding site of
OL7 did not overlap that of OL1, OL2 or OL5, the
Bam
HI activity on 69T-OL7 complexes was drastically modified in the presence of these triplex-forming oligomers: OL1 abolished the cleavage by
Bam
HI whereas OL5 and OL2 strongly increased it. Using footprinting assays and
point-mutated oligonucleotides we demonstrated that these variations were due to different conformations
of the 69T-OL7 complex induced by the binding of oligomers OL1, OL2 or OL5.
Therefore, oligonucleotides can act as structural switchers, offering one
additional mode for modulating gene expression.
Antisense oligonucleotides represent an interesting means to artificially
regulate gene expression. Numerous examples are available of successful use of
complementary oligonucleotides to selectively decrease the synthesis of the
protein encoded by the target RNA (reviewed in
1
-
3
). Although they constitute powerful tools in molecular genetics and potential
interesting prototypes for therapeutic agents (
4
-
6
), these molecules are hampered with several limitations. Biochemical stability
(regular oligonucleotides are degraded by nucleases), limited uptake by live cells and availibity of the
target sequence, presently constitute the major hurdles for the development of
efficient antisense oligomers. Chemists have provided a number of
oligonucleotide analogues that display nuclease resistance and fulfill the
major requirement for antisense purposes: affinity and specificity for the
target sequence (
7
-
12
). Uptake has been significantly improved using oligonucleotides conjugated
either to polycations or to hydrophobic ligands or carriers, like liposomes,
lipoproteins or nanospheres (
13
,
14
). Target site selection is, however, still largely an empirical process.
Although RNA are single-stranded molecules, they are known to fold into ordered structures (
15
-
18
). Intramolecular folding will restrict the availability of the sequence for
binding to the complementary one, thus leading to weak antisense effects (
19
-
22
). Several RNA structures have been demonstrated to play a key role in the
modulation of gene expression. For instance, iron responsive element control
the translability and the lifetime of messages coding for proteins involved in
iron metabolism (
23
). Hairpin and pseudoknot structures are part of the signal that drives
ribosomal frame-shifting or readthrough in retroviral genomes (
24
). TAR and RRE are folded motifs required for the synthesis and the maturation
of the HIV mRNAs (
25
). The design of oligonucleotides able to bind to such regions might allow to
modulate the function mediated by these structures. A few attempts have been
made to design oligomers to disrupt RNA structures (
26
) or to bind to single-stranded regions in their vicinity (
27
-
29
).
We recently developed strategies to take into account the three-dimensional organisation of RNA targets. Antisense oligonucleotides were
either selected from pools of oligodeoxynucleotide candidates (
30
,
31
) or designed to form local triple helical complexes with hairpin motifs. In the
latter case, we have demonstrated that OL2, an oligopyrimidine 26 nt long, was
able to bind to 69T, a 69mer DNA which formed a stem-loop structure derived from the mini-exon sequence of the protozoan parasite
Leishmania amazonensis
. This led to the formation of a 69T-OL2 `double-hairpin' complex involving pyrimidine-purine-pyrimidine triplets (
32
,
33
). We report here the results of further studies performed with this 69mer DNA
hairpin. We show that the binding of OL2 and other complementary
oligonucleotides induced drastic conformation changes of the target, ultimately
leading to either an increased or a decreased activity of a restriction enzyme
on a remote site of 69T.
Oligodeoxynucleotides were obtained from Genset (Paris) and purified by
electrophoresis on a 20% polyacrylamide-7 M urea gel. 5'-end-labelling was performed with T4 polynucleotide kinase
(Ozyme) and [[gamma]-
32
P]ATP (37.7 MBeq/mmol) from Dupont according to standard procedures (
34
). The sequences of oligonucleotides used in this study are shown in Figure
1
.
The cleavage of 69T-OL7 complex by
Bam
HI (Boehringer) was performed in a 50 mM sodium acetate buffer pH 6.0 containing
10 mM MgCl
2
. The
32
P 5'-end-labelled 69T (5 * 10
-8
M) was incubated at 20oC with 0.3 U/ml
Bam
HI in the presence of 3 * 10
-5
M OL7 and of the desired antisense oligonucleotide. Samples were then analyzed
on a denaturing polyacrylamide gel. For quantitation of the cleavage yield,
autoradiographs were scanned with a densitometer.
Footprinting studies were performed essentially as described in (
35
). The
32
P 5'-end-labelled oligonucleotide was mixed with the desired
complementary sequence(s) and incubated in a 50 mM sodium acetate buffer pH 6.0
containing 10 mM MgCl
2
. Following a 30 min pre-incubation at 15oC, samples were reacted for 90 min at 25oC with diethyl pyrocarbonate (10% final concentration). After addition of 1 vol 5 mM EDTA, the mixture was extracted
with ethyl ether and the DNA was precipitated with ethanol. The samples were
then treated (30 min at 90oC) with 1 M piperidine, precipitated twice with ethanol and dissolved in
80% formamide containing bromophenol blue and xylene cyanol (0.05% each). The
oligomers were then analyzed on a polyacrylamide gel containing 7 M urea in TBE (89 mM Tris, 89 mM borate, 2 mM EDTA) buffer pH
8.3.
A similar procedure was used for footprinting with permanganate. The
oligonucleotide mixture was incubated for 15 min at 20oC with 22.8 mg/ml KMnO
4
. The sample was then treated as above except that the ethyl ether extraction
step was omitted.
The 69mer DNA 69T (Fig.
1
) used in previous studies (
32
,
33
) displays a folded structure with a 13 bp stem. The 3' strand of the stem contains 10 contiguous pyrimidines at the bottom.
Consequently, the 5' strand is made of 10 purines, adjacent to a single-stranded six purine motif (5'-AGGGAG), used as an anchor binding site for the 5' end of the 26mer OL2. This oligopyrimidine was
designed such that its 3' part can fold back in a triple-stranded structure involving 16 base triplets upon binding to the
69mer (Fig.
1
). We previously demonstrated that a 69T-OL2 `double-hairpin' complex was actually formed at pH 6.0, in the presence of
Mg
2+
ions, i.e.,
under conditions that promote the formation of a triple helix involving C-G.C
+
triplets (
1
,
36
). (The `-' and `
.
' stand for Watson-Crick and Hoogsteen hydrogen bonding, respectively.)
The single-stranded 5' terminal region of 69T is complementary to OL7, generating a
Bam
HI site (Fig.
1
). Indeed, the 69T-OL7 duplex was cleaved by
Bam
HI, ~10% being restricted after 1 h incubation under our experimental conditions
(Fig.
2
a). The addition of OL2 to the 69mer-OL7 mixture resulted in a 3-fold increase in the cleavage yield after 1 h (Fig.
2
a). There is no overlap between the binding sites for OL7 and OL2 on the target:
OL7 is complementary to nt 1-11 of 69T whereas the 6 nt long anchor of OL2 forms Watson-Crick base pairs with the purine stretch extending from nt 12 to
17. We might have expected a reduced restriction endonuclease activity in the
presence of OL2 due to the steric hindrance induced by the triple-stranded structure and the loop of the double hairpin complex, next to the
Bam
HI site. A similar trend was observed when OL4 + OL5 were substituted for OL2
(Fig.
2
a). These two oligomers correspond to the 5' end and to the 3' end of OL2, respectively. They can bind simultaneously to the
target 69mer and form a structure similar to the double hairpin except that the
anchor (5'-CTCCCT) and the Hoogsteen strand are not connected to each other.
Of note, OL5 can engage 10 triplets with the stem of 69T in the absence of OL4.
This 69T-OL5 complex is therefore equivalent to the perfect triple helix formed by
OL1 with the stem of the 69T hairpin, except that the 5' terminal part of OL5 remains unbound (Fig.
1
). Surprisingly, whereas OL5 stimulated the cleavage of the 69T-OL7 duplex by
Bam
HI more efficiently than OL2, the triple helix-forming 10mer OL1 abolished the endonucleolytic process. None of the
oligomers OL1, OL2, OL4 or OL5 displayed any effect on the
Bam
HI activity on a restriction site other than the one formed by 69T and OL7.
These results suggest that the binding of the different antisense oligomers OL1
to OL5 interfered with the association between OL7 and the 69mer DNA and/or
induced overall conformation changes of the 69T-OL7 complex.
DEPC footprinting showed that the reactivity of some purines of 69T was modified upon addition of OL7 outside its binding site; in particular we observed an
increased reactivity of A and G residues from positions 12 to 17 (Fig.
3
, lane 2). This suggested that this region of 69T is shielded from acylation by
intramolecular interactions which are disrupted by the binding of OL7. The importance of the 3' end of 69T in the above effects was demonstrated using 51T, a 3' truncated version of the 69mer target in which the 18 nt
downstream of the hairpin stem were deleted (Fig.
1
). Indeed, no effect of OL7 was seen on the DEPC footprint of 51T (not shown)
indicating that the 3' end of 69T interacts with the 12-17 region. A hypothetical folded structure with a non-perfect intramolecular duplex between the 3' and the 5' ends of 69T is shown in Figure
4
a. This gives rise to an extended double-stranded stem, which includes the six purines of the anchor binding site,
taking into account two GT pairs. Two bulges, 5 and 3 nt long, are formed on
the 3' side. Whether the G(17)-C(58) and A(16)-T(59) pairs are formed is questionable as these purines
were significantly more reactive than the three contiguous Gs at positions 13-15.
According to the model shown in Figure
4
a, the binding of OL7 to 69T would only disrupt the terminal T(11)-A(68) pair. It is unlikely that the dramatic changes in the DEPC
sensitivity of purines in the 12-17 region (Fig.
3
, lane 2) result from the opening of a single base pair. It should rather be due
to the formation of a new structure involving both OL7 and the 3' end of 69T as no such change was seen with the shortened target 51T (not
shown). Addition of OL1 to the 69T-OL7 complex resulted in a still higher modification of purines 12-17 of 69T (Fig.
3
, lane 3) while no change was observed with OL1 in the absence of OL7 (not
shown). This suggested that OL1, which forms a triple-stranded structure with the stem of the target hairpin, stabilized the 69T-OL7 complex. As the addition of OL1 abolished the cleavage of the
69T-OL7 duplex by
Bam
HI, it was tempting to speculate that the 3' end of 69T formed a non-perfect triple-stranded structure with the oligopurine-oligomyrimidine part of the 69T-OL7 complex, as schematized in Figure
4
b. Alternatively to this structure in which the third strand of the complex involves T(52),
C(58), C(65) and T(66) one could suggest triplet formation, with either T(54)
or T(55) instead of T(52). Stacking interactions with the 69T-OLl triple helix could stabilize the non-perfect triplex although the two helices have opposite polarities
(Scheme 1). Such a triple helical structure would protect the
Bam
HI site from cleavage as shown for other restriction nucleases (
37
-
39
).
Several predictions can be derived from this hypothetical OL1-69T-OL7 structure involving base triplets. First, no effect of OL1 is
expected on the rate of cleavage of the 51T-OL7 duplex. Indeed, the deletion of the last 18 nt at the 3' end of 69T abolished the inhibitory effect of OL1 (Fig.
2
b). Second, the proposed structure for the OL1-69T-OL7 complex belongs to the pyrimidine-purine-pyrimidine motif, i.e. a pyrimidine third strand
interacts with the purine strand of a double helix via the formation of
Hoogsteen hydrogen bonds. Such a triple helix is not only dependent on the
sequence, but also on pH as C-G.C
+
triplets are more stable at acidic pH (
39
). Indeed, the above effects are no longer seen at neutral pH (not shown).
Third, according to the model given in Figure
4
b, the sequence of the 52-69 region in 69T is critical for the structure. In order to prove the
role played by the nucleotides engaged in the non-perfect triplex we prepared m69T, a mutated 69T in which both C(58) and
C(65) were substituted by A (Fig.
1
). This oligomer behaved similarly to the truncated target 51T with respect to
cleavage by
Bam
HI and to DEPC footprinting (not shown). Therefore, residues C(58) and C(65) are
crucial for the formation of a structure which prevents
Bam
HI activity in the presence of OL1. This is a strong support to the model
depicted in Scheme 1.
This structure also explains the effects observed in the presence of OL2. The
formation of the 69T-OL2 double hairpin complex involves the 12-17 region in a triple strand which will move T(11) away from
A(18), likely disrupting the non-perfect triplex formed by the 3' end of 69T (Fig.
4
b). As a consequence, the
Bam
HI site of the 69T-OL7 duplex will show an increased availability. Indeed, the addition of
OL2 to a 69T + OL7 mixture resulted in an increased cleavage efficiency (Fig.
2
a). As expected, no such increase was seen with 51T (Fig.
2
b).
The above model (Fig.
4
b) did not predict the effect of OL5 on the cleavage of 69T by
Bam
HI. As this oligomer is able to form 10 triplets with 69T we might have expected
an inhibition similar to the one seen with OL1. In contrast, a considerable
increase of the cleavage rate was observed (Fig.
2
a). This may be explained either by electrostatic repulsion due to the 5' non-bound part of OL5 or by an interaction of this region with the 13-17 nt sequence of the 69mer and unfolding of the non-perfect triplex.
DEPC footprinting of 69T-OL7 complexes showed a different pattern in the presence of either OL5 or
OL1. In particular, the three Gs at positions 13-15 were partially protected with OL5 while A(16) and G(17) remained
reactive (Fig.
3
, compare lanes 3 and 5). This suggested an interaction between the 5' part of OL5 and the G(13)-G(15) region.
32
P end-labeled OL5 was probed with sodium permanganate which reacts with single-stranded thymidines. The binding of OL5 to 69T dramatically
increased the reactivity of T(12) from OL5 (Fig.
5
, arrow). The extent of modification of OL5 bound to 69T was further increased
in the presence of OL7 but was decreased by the addition of OL4. Interestingly,
A(23) of the target 69T, which faces T(12) of OL5 in the 69T-OL5 triple-stranded complex displayed a higher sensitivity to DEPC compared
with 69T alone (Fig.
3
, lane 5, arrow). No such modifications of the homologous base triplet were seen
with either OL1 or OL2, suggesting a particular conformation of the triplex
formed by OL5 which can be viewed as a kink opening the [T(46)-A(23)]
69T
[middot] [T(12)]
OL5
triplet. One might speculate that this structural change was related to a
constraint imposed by the 5' moiety of OL5 interacting will the 12-17 region of 69T. The role of this 5' part of OL5 was demonstrated by the use of mOL5, an
oligomer derived from OL5 in which four Ts were substituted for the four Cs not
involved in triplex formation with the stem of 69T (Fig.
1
). The reactivity changes described above for OL5 were no longer seen with mOL5
(not shown). Moreover, the addition of mOL5 to the 69T-OL7 complex inhibited the cleavage by
Bam
HI (Fig.
2
a). Therefore mOL5 behaved as OL1, underlining the role of the C residues from
the 5' part of OL5 to destabilize the 69T-OL7 non-perfect triplex shown in Figure
4
b. Three G residues are present in the anchor binding site on 69T (positions 13-15). It is likely that an interaction takes place between this three G
motif of the 69mer and the three Cs of OL5 (Scheme 1). Due to the polarity of
the pyrimidine third strand in triple helices, OL5 runs parallel to the purine
strand of 69T. Parallel-stranded DNA duplex has been previously described for AT sequences (
40
,
41
) involving reverse Watson-Crick interactions, and for homo base pairs at acidic pH (
42
). Recent work has shown a parallel duplex with Hoogsteen pairing in equilibrium
with a regular Watson-Crick helix (
43
). Such an unusual parallel-stranded structure between the anchor binding motif and the 5' end of OL5 could be promoted by the adjacent triple helix.
Alternatively, one cannot exclude that the 3C motif of OL5 is flipped into the
anti-parallel orientation giving rise to three Watson-Crick GC pairs, the T(5) and C(6) residues of OL5 connecting two
parts of opposite polarities. This would likely induce a distorsion which might
explain both the reactivity of A(16) and G(17) to DEPC and the `kinked' triplex
in the middle of the stem.
Synthetic oligonucleotides are now routinely used to modulate gene expression.
At the level of double-stranded DNA, triplex-forming oligomers have been shown to block transcription whereas
targeting of RNA led to the inhibition of splicing, translation or reverse
transcription (
1
). With single-stranded DNA molecules two major mechanisms accounted for the observed antisense effects: (i) direct competition between the sense RNA-antisense oligonucleotide complex and the machinery in charge of reading the genetic information encoded in the RNA (
44
) and (ii) induced degradation of the target RNA by RNases H (
45
,
46
). In the above study we showed that, upon binding, antisense oligonucleotides
can induce conformation changes of the target, subsequently interfering with
the activity of a restriction enzyme acting on a remote site. Such an effect
has been suggested to explain the inhibition of translation of ICAM-1 mRNA by an antisense oligomer targeted to the 3' untranslated region of the message (
47
). From our work it is clear that an antisense oligonucleotide can shift the
equilibrium between alternative conformations. This might perturb biological
processes like protein binding or RNA lifetime. This can take place at a long
distance from the actual binding site of the oligomer if long range
interactions are involved in secondary or tertiary RNA structures.
This work also raised questions regarding non-intended effects of antisense oligonucleotides. The opposite properties
displayed by OL1 and OL5 on
Bam
HI activity on the 69T-OL7 duplex were unanticipated. Undoubtedly, such results are not
restricted to the particular hairpin we were working with. Similar effects can account for some of
the numerous non-antisense properties reported with oligonucleotides. This also strengthen
the requirement for several control oligomers in antisense studies (
48
).
Last, our results suggest that non-perfect triplexes can be stable under some conditions, without the help of
chemically-modified bases, intercalating agents or reactive oligonucleotide
conjugates. This situation can vary from one oligonucleotide analogue to
another.
This work was performed in the frame of an East-West contract from INSERM. Additional support was obtained from the `Pôle Médicament Aquitaine'.
REFERENCES
Return