ABSTRACT
The formation of triplex DNA using unmodified, purine-rich oligonucleotides (ODNs) is inhibited by physiologic levels of
potassium. Changing negative phosphodiester bonds in a triplex forming oligonucleotide (TFO) to neutral linkages causes a
small increase in triplex formation. When phosphodiester bonds in a TFO are converted to positively-charged linkages the formation of triplex DNA increases dramatically. In the absence of KCl, a 17mer TFO
containing 11 positively-charged linkages at a concentration of 0.2
[mu]
M converts essentially all of a 30 bp target duplex to a triplex. Less than 15% of the target duplex
is shifted by 2
[mu]
M of the unmodified TFO. In 130 mM KCl, triplex formation is undetectable using
the unmodified TFO, while triplex formation is nearly complete with 2
[mu]
M positively-charged TFO. With increasing potassium, TFOs containing a higher
proportion of modified linkages show enhanced triplex formation compared with
those less modified. In contrast with unmodified TFOs, triplex formation with
more heavily modified TFOs can occur in the absence of divalent cations. We
conclude that replacement of phosphodiester bonds with positively-charged phosphoramidate linkages results in more efficient triplex
formation, suggesting that these compounds may prove useful for
in vivo
applications.
There are two major strategies for inhibiting gene expression with
oligonucleotides (ODNs). One method uses antisense ODNs complementary to a
specific mRNA to form DNA:RNA hybrids. These hybrids are stabilized by Watson-Crick base pairing and are substrates for cellular ribonuclease H (RNase
H) that degrades the RNA portion of the duplex, rendering the mRNA
untranslatable (
1
).
A second ODN-based strategy for altering gene expression targets DNA by the formation
of a triple helical structure. Although the ability of nucleic acids to form a
triple helix was reported in 1957 (
2
), the use of ODNs to form these structures was first described in 1987 (
3
,
4
). Under appropriate conditions, an ODN will bind in the major groove of a DNA
duplex (
3
-
6
). The interaction of a third strand could either sterically block
transcription, prevent the sequence-specific interactions of regulatory proteins with DNA, or alter the conformation of the bound duplex. There
are two known triplex binding motifs, both involving interactions between the bases of an ODN
and the purine bases of a polypurine:polypyrimidine stretch of duplex DNA. In the pyrimidine motif, thymidine residues in the third
strand interact with adenosine residues of an A:T duplex while a protonated
cytidine in the third strand is hydrogen-bonded to the guanosine of a G:C duplex (
3
,
4
). The protonation of C residues generally requires a pH <6, thus limiting the utility of this strategy
in vivo
(
7
,
8
). The second triplex motif, one which will be utilized in this paper, involves
a purine-rich triplex forming oligonucleotide (TFO). Either thymidine or adenosine residues of the third strand bind to the adenosine of an A:T
duplex while guanosine in the third strand interacts with the guanosine of a
G:C duplex (
5
,
6
,
9
). This strategy uses bases in their uncharged state and therefore may be better
suited for studies at physiologic pH.
The formation of triplex DNA using purine-rich ODNs is inhibited by monovalent cations, particularly potassium ion (
10
-
12
), the predominant intracellular cation. At physiologic K
+
levels guanosine-rich ODNs self associate into aggregates which are stabilized by guanine
quartets (
12
-
14
). The inhibitory effect of K
+
can be partially diminished by chemical modification of ODNs. Incorporation of
the modified base 6-thioguanine into TFOs decreases the association of ODNs into quartets and
increases triplex formation in the presence of monovalent cations (
15
,
16
).
One obstacle faced when using ODNs
in vivo
is their rapid degradation by intracellular nucleases (
17
-
20
). The chemical modification of ODNs provides resistance to nucleolytic degradation (
19
,
20
), potentially increasing the overall activity of these compounds
in vivo
(
19
,
21
). The type and degree of chemical modification of ODNs is limited when
strategies require the action of cellular RNase H (
20
). In contrast, the formation of triplex structures does not require an
enzymatic activity and thus allows greater flexibility with respect to ODN
design. The modification of ODNs, however, can result in nonspecific toxicity
mediated through non-nucleic acid interactions (
22
).
This paper presents a phosphodiester modification which greatly increases the
triplex forming ability of ODNs under salt concentrations that approximate physiologic conditions. The association of two
nucleic acid strands creates a highly negatively-charged environment. This is demonstrated by the increased thermal stability of a
DNA duplex in the presence of increased concentrations of monovalent cations,
divalent cations and polyamines. The formation of a triplex structure produces
an even greater negative charge density. The replacement of the negative
phosphodiester moiety with a positively-charged phosphate analog significantly reduces or eliminates the Mg
2+
dependence of triplex formation. In addition, inhibition of triplex formation
by K
+
is greatly diminished by the incorporation of positively-charged internucleoside linkages. These modifications may allow the
production of ODNs which can be used as regulators of gene expression
in vivo
.
Unmodified ODNs were purchased from DNA International. Modified ODNs were
synthesized on an ABI model 391 DNA synthesizer using hydrogen phosphonate
chemistry (
23
). All reagents used for automated DNA synthesis were obtained from Glen
Research. To generate unmodified phosphodiester bonds, hydrogen phosphonate
diesters were oxidized for 4 min with freshly prepared 5% iodine in
THF:pyridine:water (15:2:2) and then 3 min with the same solution diluted 1:1
with 8% TEA in THF:water (43:3). Oxidative amidation of hydrogen phosphonate
diesters was performed manually using a 10% solution of either 2-methoxyethylamine or
N
,
N
-diethyl-ethylenediamine (Aldrich) in anhydrous CCl
4
as previously described (
19
). Briefly, ODNs containing both phosphodiester and phosphoramidate linkages
were synthesized in blocks. The desired number of 3' residues were first coupled and then either oxidized or oxidatively
amidated. The next block of residues were then individually condensed and subsequently oxidized or oxidatively amidated. Purification of ODNs was
performed as described previously (
19
). Following Sephadex G-25 column chromatography (Pharmacia), ODNs were dissolved in sterile water
and quantitated by UV spectroscopy. It was assumed that the modified phosphate
linkages did not significantly effect the extinction coefficients of the
individual ODNs. Several model ODNs containing various positively-charged phosphoramidate linkages have been synthesized and characterized
by HPLC, gel electrophoresis and mass spectroscopy (data not shown).
Target duplexes were formed from a 1:1 mixture of unmodified complementary 30mer ODNs which were heated to 80oC for 5 min and allowed to slowly cool to room temperature. These duplexes
were 5'-end-labeled with T4 polynucleotide kinase (Promega) and [[gamma]-
32
P]ATP (6000 Ci/mmol, Amersham). Briefly, 2 pmol DNA duplex (4 pmol of 5' ends) was incubated at 37oC for 45 min under the following conditions: 70 mM Tris pH 7.6, 10
mM MgCl
2
, 5 mM DTT, 8 pmol ATP and 3 U T4 polynucleotide kinase in a total of 10 [mu]l. The reaction was stopped by phenol-chloroform extraction. The ODNs were recovered from the aqueous phase
by ethanol precipitation and were resuspended in sterile water.
Triplex assays were done as previously described with minor modifications (
12
,
16
). Triplex formation was initiated by the addition (in order) of 5 [mu]l H
2
O, 2 [mu]l 5* buffer (100 mM Tris-HCl pH 7.5, 0-50 mM MgCl
2
), 1 [mu]l labeled DNA duplex (2-10 femtomol), 1 [mu]l yeast tRNA (1 mg/ml) and 1 [mu]l TFO. The buffer solutions and ODN concentrations were altered to examine
triplex formation under various conditions. This mixture was incubated at
ambient temperature overnight. After the addition of 1 [mu]l of gel loading buffer (0-10 mM MgCl
2
to match the assay conditions, 20 mM Tris-HCl pH 7.5, 50% glycerol and 0.05% each bromo- phenol blue and xylene cyanol), the samples were loaded onto a 15% polyacrylamide (acrylamide:bisacrylamide = 100:1) gel and
analyzed by nondenaturing gel electrophoresis. The electrophoresis buffer was TBM (90 mM Tris base, 90 mM boric acid and 10 mM MgCl
2
). Samples were loaded with a small voltage applied to the gel to quickly separate the positively-charged TFO from the negatively-charged duplex, thus avoiding prolonged exposure of each sample to
the high Mg
2+
electrophoresis buffer during the loading of subsequent samples. Electrophoresis was performed at 4oC for 4-6 h. The gel was dried under vacuum and exposed at -70oC to Kodak X-OMAT AR film using an intensifying screen. The amount of radioactivity present in the duplex
and triplex forms was determined by electronic autoradiography (InstantImager, Packard). The fraction of target duplex bound by a TFO, [theta], was calculated using the equation:
[theta] = S
triplex
/ (S
triplex
+ S
duplex
)
where S
duplex
and S
triplex
represent the electronic autoradiographic signal for the duplex and triplex bands respectively (
15
). The
K
d
for an ODN in triplex formation was determined from the concentration of the
compound which caused 1/2 of the target duplex to shift to the triplex form.
Because the TFOs used in these studies vary in both the degree and type of
charge, variations in the relative mobilities of the resulting triplexes were
expected. TFOs which are more negatively charged will produce triplexes with
mobilities closer to those of the parent duplex. As the negative charge is
either neutralized or converted to a positive charge, the resulting triplex
would be expected to migrate more slowly.
Figure
1
A shows the structure of the phosphate modifications used in this study. The methoxyethylamine derivative is uncharged. The
N
,
N
-diethyl-ethylenediamine derivative is shown uncharged but would be
protonated at pH 7.4. The modifications shown, like most phosphate
modifications, produce a chiral phosphate molecule. The resulting ODN is a
mixture of stereoisomers with 2
n
species, where
n
equals the number of phosphoramidate linkages in the ODN. It is possible that
each stereoisomer will possess distinct hybridization properties with respect
to formation of either duplex or triplex DNA.
If electrostatic repulsion plays a significant role in triplex destabilization, then removing negative charge from an ODN should enhance
triplex formation. ODNs N-1 and N-2 were compared with their unmodified counterpart, U-1, for the ability to form triplexes with duplex I. These
experiments were performed in 10 mM Mg
2+
and no K
+
, with ODN concentrations ranging from 20 nM to 2 [mu]M. A gel shift assay is shown in Figure
2
. Lane 1 represents labeled duplex I in the absence of TFO. Triplex formation
was seen with increasing concentrations of U-1 (lanes 2-4), N-1 (lanes 5-7) and N-2 (lanes 8-10). Triplex formation was more efficient
as negative phosphodiester linkages were converted to neutral phosphoramidate
derivatives. ODN U-1 at a concentration of 2 [mu]M (lane 4) shifted a small amount of duplex I. The triplex is seen as
a smear of signal just above the duplex. With decreasing negative charge, an
increasing fraction of duplex I was shifted to the more slowly migrating
triplex form (lanes 7 and 10). Two bands with mobilities slower than duplex I
are apparent and may represent different triplex conformations. The smear of
radioactivity present in lanes 2-10 most likely resulted from triplex dissociation during electrophoresis.
The diffuse nature of the shifted DNA prevented an accurate assessment of the
K
d
values in these experiments.
In 1 mM MgCl
2
and no KCl, ODNs containing several positively-charged internucleoside linkages were superior to their unmodified
counterpart in triplex formation. The formation of triplex DNA in the presence
of K
+
, however, is crucial if this technique is be useful
in vivo.
In Figure
4
, positively-charged ODNs P-1 and P-2 were compared with U-1 at two concentrations each of both MgCl
2
and KCl. The ODN concentration was 200 nM. Note that U-1 was examined at 1 and 10 mM MgCl
2
, while P-1 and P-2 were examined at more stringent conditions, 0.1 and 1 mM MgCl
2
. A trace amount of triplex (just above duplex I) can be seen with ODN U-1 at 10 mM MgCl
2
but not at 1 mM MgCl
2
. In contrast, ODNs P-1 and P-2 formed triplex DNA at both 0.1 and 1 mM MgCl
2
. In both cases there was only a slight increase in triplex at 1 mM MgCl
2
compared with 0.1 mM. The response of P-1 compared with P-2 with respect to K
+
inhibition was striking, especially at 1 mM MgCl
2
. Triplex DNA formation seen with P-2 at 1 mM MgCl
2
and 100 mM KCl (lane 12) was ~80% of that seen without KCl (lane 11). In contrast, triplex DNA formed
with P-1 at 1 mM MgCl
2
and 100 mM KCl (lane 8) was reduced to <50% of that seen without KCl (lane 7). This monovalent cation-inhibition of triplex formation is similar to that reported by other
laboratories with unmodified ODNs (
10
-
12
).
Figure
6
shows the association of P-2 and P-3 with duplex I under increasing concentrations of KCl (0-250 mM). The concentration of MgCl
2
is 1 mM and the ODN concentration was 2 [mu]M. With ODN P-2 (lanes 1-6) raising the KCl concentration from 0 to 200 mM gradually
reduced the amount of triplex DNA. Triplex formation with ODN P-3 was much less sensitive to K
+
concentration (lanes 7-12). In fact, at 250 mM KCl, nearly 90% of duplex I is in the triplex
form. Lane 13 shows duplex I in the absence of any TFOs. Increasing positively-charged internucleoside linkages from 69% in P-2 to 88% in P-3 was associated with a significant increase in triplex
formation in the presence of potassium.
Because Mg
2+
is known to stabilize traditional triplex DNA, the effect of decreasing Mg
2+
and increasing
K
+
on triplex formation was assessed using ODNs P-2 and P-3. The results presented in Figure
7
showed marked differences between the two ODNs. In the absence of K
+
, triplex formation with both P-2 (lanes 1-3) and P-3 (lanes 7-9) was essentially unaffected by Mg
2+
concentrations ranging from 0 to 1 mM. Triplex formation was >80%, even in the absence of Mg
2+
. Even though increased Mg
2+
concentration had no effect on the overall amount of triplex formed, in the absence of K
+
(lanes 7-9) a small fraction of the P-3 associated triplex changes to a more rapidly migrating form. The
appearance of this band with intermediate electrophoretic mobility suggests
that different conformations of triplex DNA may have formed. In 130 mM KCl, triplex formation with ODN P-2 becomes sensitive to Mg
2+
concentration (lanes 4-6). There was a significant reduction in triplex DNA as Mg
2+
decreased from 1 to 0 mM. Triplex formation with P-3 was less affected by Mg
2+
concentration in the presence of physiologic concentrations of KCl (lanes 10-12).
Positively-charged ODNs are significantly more effective in forming triplex DNA than
the corresponding unmodified or neutrally-modified compounds. The specificity of an ODN is a crucial issue when it
is designed to interfere with intracellular nucleic acid metabolism. Therefore, nonspecific electrostatic interactions between positively-charged ODNs and nontarget duplex DNA are a potential concern. To examine
this question the binding of P-4, the most positively charged ODN in this study, to duplex II was
investigated. Figure
8
a compares the nonspecific interaction of ODNs U-1 and P-4 with duplex II in the presence of either 1 or 10 mM MgCl
2
and in the absence of KCl. Lane 1 shows duplex II in the absence of TFO. As seen previously, a contaminating band appeared in the
region slightly above where the complexed DNA migrates. Binding of U-1 to duplex II was not detected (lanes 2-7). Nonspecific interactions between P-4 and duplex II were not detected at ODN concentrations of 20
or 200 nM (lanes 8 and 9 and 11 and 12). At a P-4 concentration of 2 [mu]M, however, a low level of binding to duplex II was observed. The complex formed
at approximately equal levels at both 10 mM Mg
2+
(lane 10) and 1 mM Mg
2+
(lane 13), indicating a lack of Mg
2+
dependence at these concentrations. To investigate these nonspecific interactions under salt concentrations more closely resembling
in vivo
conditions, the experiment was repeated in 130 mM KCl. As shown in Figure
8
b, U-1 did not bind to duplex II at either 1 or 10 mM Mg
2+
(lanes 2-7). At an ODN concentration of 2 [mu]M (lanes 10 and 13) the extent of nonspecific binding of P-4 to duplex II was greatly attenuated by the increased K
+
concentration of the triplex buffer. In both cases the degree of nonspecific
binding was too low for accurate quantitation. The decrease in nonspecific
binding seen in 130 mM KCl was most likely related to a general increase in
ionic strength, however, a specific effect of K
+
was not ruled out.
The inability of purine rich ODNs to form triplex DNA in the presence of
physiologic concentrations of K
+
is a major hurdle in the development of these compounds as regulators of gene
expression. In this paper we describe a class of positively charged ODNs that
can efficiently form triplexes in the presence of physiologic levels of
potassium.
Using compounds N-1 and N-2, we first tested the effect of removing negative charge from an
ODN on triplex formation. ODNs containing neutral phosphoramidate modifications
showed only a modest increase in affinity for duplex I. The ODNs N-1 and N-2 still carry a significant net negative charge and may be repulsed,
although less than U-1, by the negative charge density of duplex I. The requirement for
relatively high concentrations of Mg
2+
in order to form triplex DNA using the neutrally-modified ODNs supports the hypothesis that charge remains a major factor
affecting triple strand association. The shielding of negative charges by
counterions does not appear to be as effective as the introduction of positive
charge in promoting third strand association. Additionally there could be
structural differences between the ODNs of the N series and the P series (Fig.
1
A) which result in their different affinities for duplex I. This hypothesis
seems less likely, however, as the large terminal diethylamine moiety of the
N
,
N
-diethyl-ethylenediamine phosphoramidate would be expected to sterically
inhibit triplex formation more than the smaller methoxy moiety of the
methoxyethylamine phosphoramidate.
Increases in triplex formation result from either increased affinity of an ODN
for a DNA duplex or decreased tendency of an ODN toward self aggregation. The
latter would increase the effective concentration of the TFO. Any ODN
modification, whether directed at the bases, the sugar moiety or the phosphate
backbone, has the potential to effect either or both of these equilibria.
Whether phosphoramidate modifications affect guanine quartet formation, in
either a positive or negative manner, requires further study.
Replacing negative phosphodiester bonds with positively-charged phosphoramidate linkages resulted in compounds which could
efficiently form triplexes with duplex I. There was an increase in triplex
forming ability with increasing positive charge, from P-1 to P-3. This was demonstrated by: (i) a decrease in
K
d
with increasing ODN modification, (ii) a decreased sensitivity to the
inhibitory effects of monovalent cations on triplex formation, and (iii) a
diminished requirement of Mg
2+
in order to form triplex DNA. ODN P-3, for example, was shown to associate with duplex I with high affinity in
the presence of 130 mM KCl and in the absence of Mg
2+
. Because of the unexpected electrophoretic properties of the triplex formed
between P-4 and duplex I, extensive studies of conditions which affect triplex
formation with this ODN were deferred until further analysis of the
conformational properties of both the ODN itself and the triplexes it forms. As
with compounds N-1 and N-2, there may be several ways that positive modification of ODNs
enhance formation of triplex DNA. Because of the Mg
2+
independence of triplex association using P-3, it is likely that electrostatic attraction between the ODN and the
target duplex is a dominant force stabilizing the triplex. We have not,
however, examined the affect of positive modification of ODNs on guanine
quartet formation. It is possible that positively charged ODNs will prove
unable to self aggregate and that both mechanisms could act to promote triplex
formation.
Positively-charged compounds, in general, have an affinity for DNA binding and
modified ODNs should be no exception. As shown in Figure
8
A, ODN P-4 associates with a nontarget duplex (II) to form a complex which most
likely represents a nonspecific triplex. An association of duplex II with
increasing numbers of P-4 ODNs would be expected to shift the resulting complexes to forms with decreased electrophoretic mobility. When repeated in the
presence of 130 mM KCl, the nonspecific binding of P-4 to duplex II was nearly undetectable (Fig.
8
b). Thus, under conditions that more closely approach physiologic, the extent of
nonspecific triplex formation becomes minimal. Regardless of the target
sequence chosen,
in vivo
applications using these ODNs must consider the vast excess of nontarget to
target sequences present. Although ODNs with positively-charged phosphate backbones show an increased affinity for triplex formation, it is uncertain whether these compounds will interact with enough specificity to be biologically useful.
It may prove necessary to couple this phosphate modification with the recently
described 6-thioguanine base modification (
15
,
16
) to achieve the necessary combination of enhanced duplex affinity and sequence
specificity.
The authors thank P. T. Gilham (Purdue University), Jeff Murray, Jeff Segar and
Jon Klein (University of Iowa) for comments on the manuscript. This work was
done during the tenure of an established investigatorship (D.L.W.) of the
American Heart Association and supported by a grant from the NIH (D.L.W.) and
the Wyeth Pediatrics Neonatology Research Fund (J.M.D.).
REFERENCES
Return