ABSTRACT
Nucleic acids containing tracts of contiguous guanines tend to self-associate into four-stranded (quadruplex) structures, based on reciprocal non-Watson-Crick (G*G*G*G) hydrogen bonds. The quadruplex
structure is induced/stabilized by monovalent cations, particularly potassium.
Using circular dichroism, we have determined that the induction/stabilization of quadruplex structure by K
+
is specifically counteracted by low concentrations of Mn
2+
(4-10 mM), Co
2+
(0.3-2 mM) or Ni
2+
(0.3-0.8 mM). G-Tract-containing single strands are also capable of sequence-specific non-Watson-Crick interaction with d(G[middot]C)-tract-containing (target) sequences within
double-stranded DNA. The assembly of these G*G[middot]C-based triple helical structures is supported by magnesium,
but is potently inhibited by potassium due to sequestration of the G-tract single strand into quadruplex structure. We have used DNase I protection assays
to demonstrate that competition between quadruplex self-association and triplex assembly is altered in the presence of Mn
2+
, Co
2+
or Ni
2+
. By specifically counteracting the induction/stabilization of quadruplex
structure by potassium, these divalent transition metal cations allow triplex formation in
the presence of K
+
and shift the position of equilibrium so that a very high proportion of triplex
target sites are bound. Thus, variation of the cation environment can
differentially promote the assembly of multistranded nucleic acid structural
alternatives.
Nucleic acid sequences containing tracts of contiguous guanines tend to self-associate into a four-stranded structure (quadruplex;
1
-
9
). Within the quadruplex, the guanine residues of the four strands are arranged
in a square planar array with reciprocal non-Watson-Crick HN-1*O-6 and H
2
N-2*N-7 hydrogen bonds (a total of eight hydrogen bonds per G-quartet). The G*G*G*G quadruplex structure is inordinately (
T
m
typically >90oC) and differentially (K
+
>> Na
+
) stabilized by monovalent cations that are positioned within the quadruplex helical core, between adjacent G-quartets, coordinated to carbonyl oxygens. The relationship of cationic
radius to the size of the quadruplex helical core is responsible for the differential influence of electrostatically nearly equivalent alkali metal cations on quadruplex structure.
Single-stranded guanine-rich nucleic acids can also bind double-stranded DNA, forming a triple helical structure (
10
-
17
). The target for spontaneous triplex formation must generally contain an asymmetrical distribution of purines on one strand and pyrimidines on the
other (a pur[middot]pyr region), with major groove dimensions capable of accommodating a third strand (
18
-
19
). The purine*purine[middot]pyrimidine (pur*pur[middot]pyr) triple helical structure is based predominately on antiparallel G*G[middot]C interactions, with hydrogen bonds between HN-1 of the third strand and N-7 of the target and between H
2
N-2 of the third strand and O-6 of the target (
20
-
21
). Formation of the purine*purine[middot]pyrimidine triplex requires the presence of divalent magnesium counterions (
22
-
25
), though the precise mode of phosphate coordination has not been defined.
Sequences potentially capable of quadruplex and/or triplex formation have been
found in critical regulatory regions of the genome and evidence supporting a
natural functional or possible therapeutic significance of multistranded
conformations
in vivo
continues to accumulate. We have chosen to utilize the divalent transition
metal cations, which exhibit a higher affinity for phosphate relative to the
alkali and alkaline earth metals (
26
) and additionally possess the capability of coordinating, either directly or
indirectly, with nucleophilic atoms of the heterocyclic bases (
26
-
31
), to further probe the nature of the G*G-based multiple-stranded assemblies. Here, the differential influence of the cation
environment on competing quadruplex and triplex assemblies and equilibria are
presented.
The oligonucleotides were prepared by automated phosphoramidite synthesis, eluted through reverse phase chromatography and analyzed by
polyacrylamide electrophoresis as described (
32
). Stock solutions were lyophilized from methanol/H
2
O and stored in distilled, deionized water. The sequences utilized were as follows:
G-tract oligonucleotides
dist-14a, 5'-CGGGGGGGCGGGGC-3',
prox-F, 5'-CGGGGCGGGGGGAGC-3';
mixed sequence controls
18mer, 5'-CGAGACATGGCAGGGCAA-3',
10mer, 5'-CGAGCAGTCC-3'.
All major conclusions were confirmed using both G-tract oligonucleotide sequences.
CD spectra (210-340 nm) of oligodeoxyribonucleotides (4.8 [mu]M strand concentration) were obtained with a JASCO 500A spectropolarimeter (sensitivity 2-5 mo/cm). Oligonucleotides were allowed to equilibrate in 20
mM Tris-HCl, pH 7.2, 10 mM total divalent cation concentration (see figure
legends), and initial CD spectra were obtained. UV absorbance spectra were
obtained concomitantly. Upon addition of KCl (final concentration 5-90 mM), a series of CD spectra were obtained at timed intervals. The change in
molar ellipticity at [theta]
max
(259-261 nm), expressed either in absolute terms or as a proportion of the
total change over a particular time period, was taken as a measure of the
increase in relative proportion of oligonucleotide molecules involved in
quadruplex structure.
The human dihydrofolate reductase (
dhfr
) promoter sequence (-112 to +56) containing the two purine[middot]pyrimidine asymmetrical regions (targets for triplex formation)
was 3'-
32
P-end-labeled on either the purine-rich or pyrimidine-rich strand and isolated by native polyacrylamide gel
electrophoresis as described (
33
). Co-incubation of the
dhfr
promoter fragment (200 000 c.p.m./ reaction, ~40 nM) with one of the G-tract oligonucleotides (32-40 [mu]M) was carried out in 20 mM Tris-HCl, pH 7.2, 10 mM MgCl
2
for 45 min at room temperature (
22
), resulting in >90% intermolecular triplex formation (
32
). Alkali metal cations (supplied as the chloride salt) were variably included
as stated in the text and figure legends. The Mg
2+
and transition metal composition were also varied as stated, but the total
divalent cation concentration was maintained at 10 mM. Unless otherwise stated,
the oligonucleotide was the final component added and was thus exposed
simultaneously to the monovalent and divalent cations and the double helical
target.
Following incubation, samples were subjected to limited digestion with DNase I
(45 s on ice, 2-20 U/ml, determined empirically and dependent upon the divalent cation
composition) and the products analyzed by electrophoresis on an 8% denaturing polyacrylamide gel. Autoradiographic data were quantitated by laser
densitometric analysis. The intensities of bands (within the linear range)
within the target sequence were used as a measure of triplex formation.
Reference bands above and below the target which were unaffected by triplex formation were utilized to control for degree of digestion and sample recovery. All densitometric parameters
(e.g. area scanned, peak width) were standardized and results derived from quadruple integrations.
The oligodeoxyribonucleotides used for these studies each contain two tracts of
contiguous guanines. As such, these sequences resemble the telomeric repeats which have served as the prototype for
the study of G*G*G*G quadruplexes (
4
-
8
,
34
); the length of the G-tracts is equal to or greater than that found in the telomeric repeats and
the isolation of G-tracts from each other or from the terminus of the strand is limited to
only one base. The oligonucleotide sequences correspond with inverse
orientation to a G-tract pattern found in the human dihydrofolate reductase promoter (see
below).
The CD spectra of either of these oligonucleotides in deionized, distilled water
or in 20 mM Tris-HCl, pH 7.2, 10 mM MgCl
2
(Fig.
1
A, solid line = 0 min) exhibits the characteristic pattern of the parallel intermolecular quadruplex (
35
-
37
) with a large positive ellipticity at ~260 nm and a smaller negative peak at ~240 nm. Thus, even in the absence of monovalent cations, these
oligonucleotides exhibit a certain degree of native quadruplex character. Hardin
et al
. (
35
) have shown that the relative height of the positive CD peak at 260 nm can be
correlated with the prevalence of quadruplex structure. Over time, the
ellipticity of dist-14a or prox-F varies <= 10%. However, upon addition of 30 mM KCl, a marked, progressive
increase in the magnitude of the CD is observed (discernable within 2 min; Fig.
1
A, broken lines). Note that the wavelength of maximal ellipticity shifts
slightly from 259 to 261 nm upon addition of K
+
and that an isosbestic point is present at 248-249 nm, indicating that the transition involves only two species. Thus,
exposure of the oligonucleotide to potassium causes a rapid induction and/or stabilization of quadruplex structure.
For the next set of experiments, Mg
2+
was eliminated or else partially replaced by a divalent transition metal cation
while keeping the total divalent cation concentration constant at 10 mM. An
essentially identical alteration of the CD spectrum and rate of increase in molar ellipticity at 260 nm was obtained with >= 30 mM K
+
in the presence or absence of Mg
2+
(data not shown). However, in the presence of 3-4 mM Mn
2+
, 2 mM Co
2+
or 0.5 mM Ni
2+
the rapid increase in molar ellipticity otherwise observed upon addition of K
+
was severly blunted (by 50->90%; Fig.
2
A and B). Neither Mn
2+
, Co
2+
nor Ni
2+
had any immediate appreciable influence on the native CD pattern of either of the G-tract oligonucleotides. Thus, without otherwise altering the secondary structure of the oligonucleotide, these divalent transition metal cations effectively counteract the induction/stabilization of
quadruplex structure by potassium.
The human
dhfr
core promoter sequence (
38
) contains two closely spaced and very similar regions of purine[middot]pyrimidine (pur[middot]pyr) asymmetry, which consist of (on the purine-rich strand) tracts of contiguous guanines with
interspersed individual A or C residues. These two pur[middot]pyr regions represent potential targets for assembly of intermolecular
triple helical structures and the G-tract patterns of the oligonucleotides dist-14a and prox-F correlate in an antiparallel orientation with the G-tract patterns of the distal and proximal pur[middot]pyr regions. We have in fact demonstrated specific binding of these single-stranded oligonucleotides to their respective double-stranded target sequences on a
restriction fragment of the
dhfr
promoter in a simple buffer consisting only of 20 mM Tris-HCl, pH 7.2, 10 mM MgCl
2
(
32
,
39
). With an excess of oligonucleotide, intermolecular triplex assembly proceeds
rapidly at ambient temperature such that essentially 100% of the specific target site is occupied within 10 min. However, we observe a potent concentration-dependent negative influence of potassium cations on formation of the
triple helical structure. For the experiment shown in Figure
3
A, concentrations of KCl from 0 to 60 mM were included from the onset of an
incubation of the oligonucleotide prox-F with the
dhfr
promoter fragment under conditions otherwise conducive to triplex formation (20 mM Tris-HCl, pH 7.2, 10 mM MgCl
2
, room temperature, 45 min, oligonucleotide excess). Without KCl, prox-F produces a high degree of DNase protection over its intended target
within the proximal pur[middot]pyr region, as well as a minor footprint resulting from mismatched
binding to the homologous distal pur[middot]pyr sequence (lane 1). An increase in KCl concentration is accompanied by a decrease in the proportion of
dhfr
promoter molecules to which the oligonucleotide third strand is bound (lanes 2-5). A considerable loss of occupation of the proximal (specific) target
site by prox-F is observed with >= 30 mM KCl. The mismatched binding of prox-F to the distal target is almost completely inhibited with only
5 mM KCl.
With simultaneous exposure of the oligonucleotide to K
+
as well as Mg
2+
and the specific double helical target, an initial kinetic `race', a
competition between assembly of the triplex (binding of the G-tract oligonucleotide to its double-stranded target sequence) and assembly of the quadruplex (self-association of the G-tract oligonucleotide) is observed. Once assembled, both
the triplex and the K
+
-associated quadruplex are very stable structures (
4
,
21
,
24
-
25
,
35
,
43
; CD thermal denaturation data not shown). Thus an apparent equilibrium is
rapidly established (in <45 min). In the absence of K
+
and with an excess of oligonucleotide over target, triplex formation may be
driven to near completion, however, in the presence of K
+
, the quadruplex is favored to the exclusion of the triplex. Thus factors differentially affecting the progress of these divergent, competing intermolecular assemblies will determine the resulting relative proportion of triplex or quadruplex products.
A series of experiments was performed in which triplex formation in the presence
of K
+
was measured while Mg
2+
was replaced to a variable degree by a divalent transition metal cation (total
divalent cation concentration held constant at 10 mM). The oligonucleotide was
added last to the incubation and thus simultaneously exposed to the monovalent
and divalent cations as well as its double helical target. Assay of the binding
of oligonucleotide prox-F to the
dhfr
promoter fragment in the presence of 30 mM K
+
and variable Mg
2+
:Mn
2+
composition is shown in Figure
4
A. With increasing Mn
2+
content (2-10 mM), the DNase I footprint of prox-F at its target (which disappeared in Fig.
3
A) reappears and becomes increasingly more pronounced, indicative of triplex formation on a considerable proportion of the
dhfr
promoter molecules. Concentrations as low as 300 [mu]M Ni
2+
or Co
2+
significantly enhanced triplex formation in the presence of K
+
(Fig.
4
B). These alterations of the cationic environment effectively counteract the inhibition of intermolecular pur*pur[middot]pyr triplex formation by K
+
, permitting the oligonucleotide third strand to bind to its specific target on
75-90% of
dhfr
promoter molecules otherwise unbound. These results correlate well with the
ability of these divalent transition metal cations to counteract the
induction/stabilization of quadruplex structure by K
+
. The biphasic nature of each of these titrations is apparently indicative of an
additional mode of cation coordination which is detrimental to triplex
formation and is reminiscent of the potential for destabilization of duplex DNA by transition metal cations (
44
-
46
).
These results are influenced by the choice of experimental protocol
(oligonucleotide added last). If the oligonucleotide is pre-incubated with the effective divalent cation composition prior to exposure
to the monovalent cation and double helical target, K
+
-resistant triple helical assembly is enhanced. Conversely, if the transition metal cation is initially withheld from the incubation, its immediate effectiveness in fostering the triple helical association (in competition with K
+
-induced self-association of the G-tract single strand which has already taken place) is
drastically reduced (see below).
The titration of triplex assembly versus increasing K
+
concentration (as in Fig.
3
A) is also altered by the transition metal cations (Fig.
4
C). In the presence of either 5 mM Mn
2+
or 2 mM Co
2+
, ~80, 75 and 60% of specific target sites were occupied by the third strand
despite 60, 90 or 120 mM K
+
. This represents at minimum a 4- to 5-fold decrease in sensitivity of triplex formation to K
+
relative to that observed with 10 mM Mg
2+
alone and relatively efficient triple helical association at near physiological
K
+
concentration.
In order to investigate the influence of the cationic environment on the
establishment of a true equilibrium between quadruplex and triplex species,
three series of samples were used, differing only in divalent cation
composition (10 mM Mg
2+
, 10 mM Mn
2+
or 8 mM Mg
2+
/2 mM Co
2+
; Fig.
5
). In half of the samples of each series, the oligonucleotide was exposed to 30
mM K
+
prior to the divalent cations and the target sequence. For the other half, the
oligonucleotide was exposed to the divalent cations and the target sequence
first, after which the K
+
was added. Thus, in one case, opportunity is presented for initial quadruplex
assembly, while in the other case, the triplex is formed first.
The G-tract oligonucleotides exhibit a certain degree of native quadruplex
character in aqueous solution, however, a significant proportion of molecules
remain in the free single-stranded form. Upon exposure to monovalent alkali metal cations,
particularly K
+
, an increasing proportion of the oligonucleotide population is recruited into
quadruplex structure. The mechanism by which a net increase in quadruplex
structure is mediated by K
+
potentially involves (i) the facilitation of
de novo
association of oligonucleotide single strands and/or (ii) additional stabilization (decreased rate of
dissociation) of the existing quadruplex population. At 1.3 Å, the cationic radius of K
+
is apparently optimally suited for occupation of the quadruplex helical core
and coordination to carbonyl oxygen atoms (
6
). The divalent transition metal cations Mn
2+
, Co
2+
and Ni
2+
potently counteract the effect of K
+
on the G-tract oligonucleotides: the rate at which oligonucleotide is sequestered into quadruplex structure by K
+
is slowed considerably and the K
+
-induced shift in equilibrium toward the quadruplex is lessened. These
results are consistent with a direct involvement of these divalent cations in
opposing the mechanism of action of the monovalent cation. This could
conceptually involve either of two mechanisms (which are not necessarily
mutually exclusive): (i) the transition metal may chelate or coordinate
directly to nucleophilic atoms (i.e. N-7 and O-6) of the guanine bases of single- stranded oligonucleotide molecules (
27
,
30
-
31
,
47
-
49
) and prevent formation of the Hoogsteen hydrogen bonds essential for quadruplex
assembly; (ii) the transition metal may compete directly with potassium for
occupation of the quadruplex helical core, possibly for coordination to O-6 atoms, and exert a less stabilizing or even destabilizing effect on the
existing quadruplex structure. At 0.6-0.8 Å, the radii of Mn
2+
, Co
2+
and Ni
2+
are much smaller than K
+
and therefore a poorer `fit' within the quadruplex helical core. Further
experimentation will be required to define more precisely the nature of these transition metal-nucleic acid interactions.
Both the parallel quadruplex and the antiparallel purine*purine[middot] pyrimidine triple helix are based on non-Watson-Crick guanine* guanine interactions and a G-tract-containing sequence may be simultaneously
capable of involvement in either structure. These two non-canonical nucleic acid structures are fundamentally distinct in several
ways, including: (i) Hoogsteen (quadruplex) versus reverse Hoogsteen (triplex)
hydrogen bonding; (ii) strand composition [(single strand * 4) versus (single strand plus double-stranded target)]; (iii) geometrical configuration, which certainly results in disparate modes/sites of counterion
coordination. The data presented here have demonstrated that the cationic environment
influences competition between quadruplex and triplex assemblies (Fig.
6
). The same divalent cationic compositions which counteract the induction/stabilization of quadruplex structure by K
+
also abrogate the inhibition of triplex formation by K
+
. Furthermore, the equilibrium between the competing triplex and quadruplex
associations is significantly altered by the cationic environment, such that
the occupation of a high proportion of triplex target sites in the presence of
K
+
becomes possible.
Naturally occurring G-tract-containing sequences in the eukaryotic genome have frequently been implicated in important molecular
biological functions, such as regulation of gene expression, replication, recombination and chromosomal condensation (
50
-
53
). It has been proposed that the potential of these sequences for adoption of non-Watson-Crick structures, such as the quadruplex or the triplex, may be
functionally relevant
in vivo
[e.g. human [gamma]-globin 5'-flanking region, associated with hereditary
persistence of fetal hemoglobin (
54
); Chinese hamster ovary origin of replication (
55
); human c-
myc
promoter (
56
); d(CGG)
n
repeats, associated with the fragile X syndrome (
57
); human insulin gene-linked polymorphic region (
58
); telomeric repeats, associated with integrity of chromosomal termini,
chromosomal pairing at meiosis, cellular senescence and immortality of
transformed cells (
59
-
61
)]. Naturally occurring triplex binding proteins have been identified (
62
); eukaryotic chromatin is heterogeneously stained by anti-triplex antibodies (
63
-
64
); intramolecular triplex structures have been detected in
E.coli
. (
65
-
66
). Miller and colleagues have putatively identified at least two naturally
occurring untranslated transcripts which may function through intermolecular
triplex formation with genomic target sequences (D.M.Miller, unpublished data). Naturally occurring polypeptides which bind quadruplex structures (
67
-
69
), promote quadruplex formation (
37
,
70
) or cleave DNA specifically near quadruplex structures (
71
) have been reported. Quadruplex formation has been implicated in the life-cycle of the HIV virus (
72
-
73
) and may be involved in developmental regulation (
74
). Thus several lines of evidence point to the existence and functional
significance of quadruplex and triplex nucleic acid structures
in vivo
. An understanding of the factors which influence the assembly of such multi-stranded nucleic acid structures is crucial to discerning the associated
mechanisms of molecular biological regulation.
The data presented here have implicated the existence of a distinct mode of
cation coordination through which nucleic acid reactivity is significantly
altered, making possible the differential promotion of one non-Watson-Crick G*G interaction over another. The potential for quadruplex or
triplex formation
in vivo
would undoubtedly be influenced by resident cationic moieties or putative accessory proteins. Although total cellular concentrations of Mn
2+
, Co
2+
and Ni
2+
are very low (submicromolar), it is conceivable that compartmentalization,
synergistic interactions (
75
) or presentation by accessory molecules could facilitate their interaction with
nucleic acids under appropriate conditions. Alternatively, the effects of the
transition metal cations
in vitro
might well be duplicated naturally within the cell (via high positive charge, not coordination to base) by oligovalent polyamines or a specialized basic
polypeptide domain (
76
-
78
).
The results presented here are also relevant for proposed oligonucleotide-based therapeutic approaches, involving the exogenous pharmacological
administration of nucleic acid molecules (some of which would contain G-tracts) intended for antisense, triplex or aptamer strategy (
25
,
79
-
85
). The ability of such an oligonucleotide effector molecule to reach and bind its intended target will be influenced by
the specific cationic environments to which it is exposed within the cell. The
inhibition of intermolecular pur*pur[middot]pyr triplex assembly by moderate concentrations of potassium appears to
represent a considerable obstacle to the natural or practical therapeutic
utility of this interaction within a physiological (i.e. ~140 mM K
+
) environment (
86
-
88
), however, an effective cationic species may mitigate this obstacle
in vivo
. The divalent transition cations which we have utilized in this work are labile metal centers, with relatively short-lived interactions with nucleic acids. This laboratory is currently investigating the
feasibility of site-specific incorporation of inert (covalent) metal centers into synthetic
oligonucleotide structures. The data presented here have certainly suggested
that such an approach might allow the permanent addition of physical/chemical attributes favorable for triplex formation. In related approaches, the incorporation of
positive charges in the backbone of synthetic oligonucleotides and the use of a
6-thio-substituted guanine derivative have been investigated as a means to disfavor quadruplex self-association (
89
-
91
). Positively charged ligands have been shown to specifically stabilize triplex structures (
92
-
93
). Interestingly, the interactions of the various cationic environments on other modified oligonucleotide structures proposed for therapeutic use (methylphosphonate, phosphorothioate, 2'-
O
-methyl, etc.) largely remain to be determined.
The authors wish to express their grateful appreciation to Dr J.Lebowitz for
allowing them the extensive use of his laboratory's laser densitometer. This
work was supported by National Institutes of Health grants K11 CA01581 (SB),
CA42664 and CA54380 (DM) and a VA Merit Review Award (DM).
*To whom correspondence should be addressed at : University of Alabama at
Birmingham, Comprehensive Cancer Center, 1824 6th Avenue South, Room 508,
Birmingham, AL 35294, USA. Tel: +1 205 934 1977; Fax: +1 205 975 6911; Email:
dmiller@ms.ccc.uab.edu
REFERENCES
Return