ABSTRACT
In contrast to shorter homologs which only form a single-stranded nucleic acid
[alpha]
-helix in acid solution at [Na
+
]
<=
0.02 M Na
+
, d(A-G)
20,30
form
in addition
a parallel-stranded duplex with (A
+
[middot]A
+
) and (G[middot]G) base pairs and interstrand dA
+...
PO
2
-
ionic and dA
+
NH
2
...
O=P H-bonds. Under conditions where duplex prevails over
[alpha]
-helix, the contribution of the base-backbone interactions to stability varies directly with [H
+
] and inversely with [Na
+
], just as in poly(A
+
[middot]A
+
). These duplexes are characterized by intense circular dichroism and a large cooperative thermally-induced hyperchromic transition that is dependent on oligomer concentration. Dimethylsulfate reactivity of the dG residues indicates G[middot]G and therefore dA
+
[middot]dA
+
rather than dA
+
[middot]G base pairs. At much higher ionic strength (Na
+
>=
0.2 M) the protonated base-backbone interactions are so weakened that duplex stability becomes
increasingly dependent upon H-bonded base pairing and stacking and almost independent of pH. Between pH 6 and 8 this duplex structure is devoid of protonated dA residues and shows positive dependence of
T
m on ionic strength similar to that of DNA.
Repeating homopurine-homopyrimidine sequences adopt a variety of unique structures to relieve superhelical stress (
1
,
2
). In upstream gene control elements such sequences also serve as sites of
protein binding not to the DNA duplex itself, but rather to some conformational
variant of one of the strands (
3
,
4
). It is probably for this reason that a number of studies have focused on the
alternating homopurine sequence d(A-G)
n
. This sequence forms a broad spectrum of conformations under different pH and
ionic conditions, including parallel-stranded duplexes (
5
,
6
), a hairpin duplex (
7
-
9
), a two-hairpin tetraplex (
7
,
10
) and a novel type of single-stranded helix (
11
-
14
). The latter conformation, which has been observed for d(A-G)
6
and d(A-G)
10
in an acidic milieu at low ionic strength, is stabilized not by helically wound
stacks of bases or base pairs, but by an unusual combination of ionic and hydrogen (H-) bonds between dA
+
residues and the phosphodiester backbone such that the dA residues do not overlap their dG nearest neighbors.
This structure, referred to as the (nucleic acid) [alpha]-helix, occurs below pH 6 in 0.01 M Na
+
, reaches maximum stability at pH 4 and is characterized by intense circular
dichroism and minor hypochromicity. The p
K
a
for the acid-induced transition of d(A-G)
10
to d(A
+
-G)
10
in 0.01 M Na
+
is 5.3 at 25oC and increases with lower temperature and lower Na
+
concentration. The ionic interactions that maintain the [alpha]-helix are characteristically very sensitive to cation concentration because shielding of the
phosphate groups weakens the ionic and associated H-bonds that stabilize the structure.
Increasing the length of d(A-G)
n
from
n
= 6 to
n
= 10 raises the p
K
a
of the coil to [alpha]-helix transition (
12
). If this trend were to continue as the number of d(A-G) repeats is increased, it is conceivable that the longer lengths found
in mammalian gene control elements would have p
K
a
values within the physiological pH range. To determine if this is so, we
examined the solution properties of the structures formed by d(A-G)
20
and d(A-G)
30
under varying conditions of pH and ionic strength. We thereby identified two
related double-stranded helical structures formed by these oligomers but not by d(A-G)
10
. Between 0.001 and 0.01 M Na
+
below pH 6.0, both oligomers form an acid-dependent parallel-stranded duplex with A
+
[middot]A
+
and G[middot]G base pairs. Nevertheless, this structure is stabilized primarily by
ionic and H-bonds between the protonated dA residues and the backbone phosphates of
opposing strands, since its stability increases upon lowering either ionic
strength or the dielectric constant of the medium. Under slightly more acidic
conditions, these duplexes are in equilibrium with the single-stranded [alpha]-helix previously described for d(A
+
-G)
6,10
(
11
-
14
). At Na
+
>= 0.21 M above pH 3.5, the ionic and associated H-bonds between dA
+
and the backbone in the linear duplexes of d(A-G)
20,30
are suppressed. While the basic duplex structure is retained, it is now
stabilized principally by helically wound and stacked A[middot]A and G[middot]G base pairs, for the counterion shielding now reduces
interstrand backbone repulsion. As a consequence, the strong pH-dependence of stability evident at low ionic strength essentially disappears and the
duplex even exists above neutral pH. As noted, in these higher Na
+
concentrations the single-stranded [alpha]-helix does not occur at any pH or strand length.
d(A-G)
20
and d(A-G)
30
were synthesized by the phosphoramidite method, deprotected, purified by 8 or
6% denaturing PAGE and their bands eluted and desalted by reverse-phase chromatography (
11
). Oligomer concentrations (
c
) were determined spectrophotometrically and are reported on a residue basis.
Purity was confirmed by denaturing PAGE of 5'-
32
P-labeled oligomers. [epsilon]
260
per residue of d(A-G)
20
and d(A-G)
30
were assumed to be 9500 in distilled water, the same as for d(A-G)
10
and poly[d(A-G)] (
15
). The deoxyoligomers 5'-CACCTGACTCCTGTGGAGAAGTCTGCCGTTACTGCCCTGTG-3' (41mer) and 5'-CTGACTCCTGTGGAGAAGTCTGCCGTTACTGCCCT-3' (35mer) and their
complementary strands were synthesized by the phosphoramidite method and purified as above.
Aqueous buffers between pH 3.5 and 6.0 +- 0.05 were prepared by titrating sodium acetate with acetic acid to a final Na
+
concentration of 0.01 M. Aliquots of these buffers were also diluted appropriately with H
2
O or near-saturated NaCl to achieve 0.001 M and 0.21 M Na
+
and their pH values redetermined. pH measurements were made with a Radiometer pH meter 26 and a glass pH microelectrode standardized with an appropriate buffer at room temperature and at 4oC.
Samples containing 8.4 * 10
-5
M residues of d(A-G)
20
and d(A-G)
30
in buffer were heated to 40-50oC, cooled slowly and incubated for 10 min at the desired temperature
prior to measurement of CD. For melting profiles, CD spectra from 320 to 220 nm
between 2 and 60oC were recorded every 6oC on a computer-driven AVIV 62DS CD spectrometer equipped with a
thermoelectrically-controlled cell holder. Digitized data obtained every 1 nm were corrected
for baseline at the ambient temperature and smoothed by a least-squares polynomial fit up to the third order. CD spectra per mole of
monomer are plotted as [Delta][epsilon] in units of l/mol/cm. Profile reproducibility was excellent after
several weeks storage of samples at -20oC.
Absorption spectra and thermal melting profiles were measured with a computer-driven AVIV 14DS spectrophotometer equipped with a thermoelectrically-controlled cell holder. For melting experiments, data were taken as
spectra measured between 320 and 220 nm at 1 nm intervals every 2oC and converted to melting profiles at desired wavelengths.
T
m
was determined as the maximum of the differentials (d
A
/d
T
versus
T
) of the profiles.
Samples of d(A-G)
20
and
d(A-G)
30
at pH 5.0 in 0.001, 0.01 and 0.21 M Na
+
and varying in concentration from 2.6 * 10
-6
to 2.1 * 10
-3
M (residues) were heated [duplex standards to 85oC, d(A-G)
20
and d(A-G)
30
to 50oC], cooled slowly to 25oC and then equilibrated at 4oC overnight. The duplex and triplex size markers, each 1 * 10
-4
M (residues), were similarly treated. d(A-G)
20
, d(A-G)
30
and the 41mer were each 5'-
32
P-end-labeled and purified (
13
) and an equal amount of each was used to label appropriate samples. At 4oC, 4 [mu]l aliquots were added to 1 [mu]l loading dye [15% Ficoll type 400 (Pharmacia), 0.25% bromophenol blue, 0.25% xylene cyanol in H
2
O], mixed briefly and loaded onto a pre-electrophoresed 12% native polyacrylamide gel. Gel and circulating
electrophoresis buffers contained 1 mM EDTA in addition to 0.01 or 0.21 M Na
+
, all titrated to the desired pH with acetic acid. Gels were run at 25-28 V/cm at 4oC with recirculation. For autoradiography, X-ray film was exposed to wet gels for 8-12 h at 4oC.
Reaction mixtures containing 1 * 10
-4
M residues of 5'-
32
P-labeled d(A-G)
20
or d(A-G)
30
in 0.001, 0.01 or 0.21 M Na
+
buffered at pH 5.0 and 0.5% DMS (added as a 10% solution in ethanol) in a total
volume of 20 [mu]l were incubated for 40 min at 4oC. Samples of d(A-G)
20
[middot]2[d(C-T)
10
], d(A-G)
30
[middot]3[d(C-T)
10
] and 2[d(C-T)
10
]
:
d(A-G)
20
[middot]2[d(C-T)
10
] (1 * 10
-4
M residues) containing [
32
P]d(A-G)
n
oligomer were similarly treated. After incubation, 5 [mu]l stop reagent (1.5 M Na
+
, pH 5.2, 1 M [beta]-mercaptoethanol, 250 [mu]g/ml tRNA) at 0oC were added and the oligomer was precipitated with
ethanol, redissolved in 0.3 M Na
+
, pH 5.2, ethanol precipitated twice more, dried and digested with 1 M
piperidine in H
2
O for 30 min at 85oC. Samples were then twice evaporated and washed with H
2
O, evaporated and mixed with loading dye (9 [mu]l deionized 98% formamide, 10 mM EDTA, pH 8.0, 0.25% bromophenol blue in H
2
O), heated to 90oC and analyzed on a 12% denaturing gel subjected to pre-electrophoresis. The percentage of full-length oligomer remaining at each time point was determined
after autoradiography of wet gels and scintillation counting of cut bands of
full-length oligomer and of digestion products. The results are corrected for baseline piperidine cleavage and expressed as percentage oligomer cleaved.
Samples in a total volume of 30 [mu]l contained 1 * 10
-4
M 5'-
32
P-labeled d(A-G)
20
in 0.001, 0.01 or 0.21 M Na
+
at pH 5.0 and d(A-G)
20
[middot]2[d(C-T)
10
] in 0.21 M Na
+
at pH 5.0 were treated with 0.017 U S1 nuclease at 4oC. Reaction was stopped by transferring 5 [mu]l aliquots at 1, 2, 5, 10, 15 and 30 min to 4 [mu]l loading dye (70% formamide, 0.1% bromophenol blue, 57 mM EDTA, pH
7.5) and freezing. Frozen samples were heated at 90oC for 2 min prior to loading on a pre-electrophoresed 12% denaturing gel and electrophoresis at 45 V/cm.
The percent full-length oligomer remaining at each time point was determined after
autoradiography of wet gels and scintillation counting of cut bands of full-length oligomer and of digestion products.
CD spectra of d(A-G)
20
and d(A-G)
30
in 0.001 M Na
+
(very low) and 0.01 M Na
+
(low) were measured between pH 3.5 and 6.0 at 3 and 25oC. Both oligomers display strong pH-dependence of CD intensity under these conditions, with the intense
circular dichroism at lower pH characteristic of a helical structure. A
representative plot of [Delta][epsilon] versus [lambda]
nm
for d(A-G)
30
in 0.01 M Na
+
illustrates this dependence (Fig.
1
A). The very significant decrease in intensity between pH 5.0 and 5.5 shows that
a strongly chiral structure undergoes a major conformational transition as pH
rises in this range. The CD-monitored titration curves in Figure
2
show that the p
K
a
values for both lengths of oligomer drop on going from 3 to 25oC and as Na
+
is raised from 0.001 to 0.01 M.
On increasing the ionic strength to 0.21 M Na
+
(moderate), the acid-dependence of duplex formation is greatly diminished. Thus, titration
monitored by CD does not reveal a pH-dependent transition (Fig.
1
B). The small decrease and redshift of [Delta][epsilon]
max
represented by the family of CD spectra with three isosbestic points in Figure
1
B probably reflects a decrease in the degree of protonation of the dA residues
as the pH is raised. But this decrease is not attended by a conformational
transition, as is evident in 0.01 M Na
+
(Fig.
1
A). The contrasting pH-dependencies in low and moderate Na
+
are suggestive of different sources of helix stability, as are the contrasting
ionic strength dependencies of stability (see below). Nevertheless, the pH-independent duplexes in 0.21 M Na
+
do undergo cooperative melting with large CD and UV absorbance changes (data
not shown). These duplexes show only slight
T
m
maxima at pH 4.5 in both cases and complete pH-independence above pH 5.5 (Table
1
), where dA residues are no longer protonated. Consistent with such behavior,
this structure shows
positive
ionic strength dependence of melting at pH 6 not unlike that for DNA duplexes
of similar length, with d
T
m
/dlog[Na
+
] = ~18oC.
Table 1
Figure
7
shows CD spectra measured on samples of d(A-G)
30
under different conditions that are believed to stabilize the three
conformations discussed here for d(A-G)
n
, i.e., the [alpha]-helix (spectrum 1), the acid-dependent putative duplex (spectrum 2) and the pH-independent putative duplex (spectrum 3). Although all three structures are characterized by intense
circular dichroism, it is apparent that their CD spectra are unique.
To further characterize the interactions that contribute predominantly to stabilization of the acid-dependent and pH-independent duplexes, the effect of ethanol on
T
m
was examined. Lowering the dielectric constant is known to stabilize
electrostatic interactions, but to weaken base stacking due to enhanced base
solvation. On going from 0 to 5% ethanol in 0.01 M Na
+
at pH 5.0, the
T
m
of d(A-G)
20
increases
from 21.8 to 24oC, confirming that ionic interaction between protonated dA residues and the
PO
2
-
of the backbone phosphate moieties contribute significantly to stabilization of the acid-dependent duplex. In contrast,
T
m
values
decrease
for the pH-independent duplex formed by d(A-G)
20
in 0.21 M Na
+
at pH 5.5, from 43.8 to 42.8 and then to 41.4oC in the presence 0, 5 and 10% ethanol respectively. This effect is as
observed for Watson-Crick duplexes, which have no interstrand ionic attractions.
Indications that the acid-dependent and pH-independent structures are multistranded were first obtained from the observation of a
concentration-dependence of
T
m
values provided by UV melting profiles. For d(A-G)
20
at pH 4.5 in 0.01 M Na
+
and at pH 5.0 in 0.21 M Na
+
, 10-fold increases in oligomer concentration raise the
T
m
value from 23.8 to 26oC and from 45.1 to 47.8oC respectively, and a similar concentration dependence of thermal
stability was observed for d(A-G)
30
. Strandedness was more directly examined by native gel-mobility assays. In running buffer with 0.01 M Na
+
at pH 5.0, d(A-G)
20
incubated in 0.001 and 0.01 M Na
+
migrates essentially the same as the 41 bp Watson-Crick duplex size marker (Fig.
8
) and d(A-G)
30
migrates as expected for a 60 bp duplex, i.e., with proportionately slower
mobility than the 41mer duplex marker. Electrophoretic mobilities of d(A-G)
20
and d(A-G)
30
were also examined over a 1000-fold oligomer concentration range, from 5 * 10
-6
to 1 * 10
-3
M in 0.01 M Na
+
, pH 5.0. At very low oligomer concentration, trace amounts of single strands of
d(A-G)
20
but not of d(A-G)
30
are present, showing that there is an equilibrium between the single-strand [alpha]-helix and the duplex formed by the shorter oligomer.
In attempting to deduce H-bonding schemes for the two types of duplex, we considered antiparallel versus parallel strand orientation and homo versus hetero base pairs (Fig.
9
). When both members of the pair are in antiparallel strands, one of the two
bases in A[middot]A and G[middot]G base pairs must be an unfavored tautomer (pairs 1 and 2) (
17
), but this is energetically intolerable when such pairs are present in the
duplex in high frequency. The hetero pair G
anti
[middot]A
anti
(
18
) may be dismissed for the acid-dependent duplex because protonation of A is prohibited by the involvement
of N1 of A as a hydrogen acceptor (pair 3); and while such a scheme is
conceivable for the neutral
anti
[middot]
anti
pair, the smooth and instantaneous transition of acid duplex to neutral duplex
makes it unlikely. For energetic reasons, a full complement of G
anti
[middot]A
syn
pairs with the A residues protonated or neutral (pair 4) (
8
,
19
) seems unlikely. In fact, this pairing scheme has been shown to occur for d(A-G)
n
only when forced in a hairpin duplex stabilized by a long run of flanking
Watson-Crick pairs (
8
,
9
). Finally, protonated A
anti
[middot]G
syn
pairs (pair 5) (
20
) are ruled out by the unavailability of N7 of dG residues for reaction with DMS
(see below).
To distinguish between hairpin and linear structures, 5'-
32
P-labeled d(A-G)
20
at pH 5.0 in 0.001, 0.01 and 0.21 M Na
+
was digested with S1 nuclease. A Watson-Crick d(A-G)
20
[middot]2[d(C-T)
10
] duplex in 0.21 M Na
+
at pH 5.0 served as a duplex standard. Electrophoresis on a denaturing gel
revealed very small amounts of the
n
- 1,
n
- 2 and corresponding mono- and dinucleotide products, presumably due to fraying of the ends
of the duplexes, but
none of the intermediate size fragments
which would result from random cleavage of a single-stranded structure or the half-molecule fragments expected from cleavage of a hairpin turn (cf.
7
). This general resistance of both types of d(A-G)
20
duplex to endonuclease cleavage is like that of the Watson-Crick duplex and shows that the structures formed in very low, low and
moderate Na
+
do not contain hairpin turns (Fig.
11
). Moreover, the efficacy of the enzyme for cleaving single strands preferentially at pH 5.0 at all three Na
+
concentrations was confirmed using a non-repetitive single-stranded 35mer and its Watson-Crick duplex (data not shown).
The pH-dependence of the helical structure at 0.001 and 0.01 M Na
+
indicates that protonation of most dA residues is essential to the stability of
the acid-dependent d(A
+
-G)
20,30
duplexes, since cooperative melting and hypochromism disappear near pH 5.5. In contrast, the structure in moderate salt depends little on protonation of dA residues for stability, its structure extending well into the neutral pH
range with almost no diminution in
T
m
value.
What must distinguish the duplexes stabilized below 0.02 M Na
+
and above 0.2 M Na
+
is the relative significance of the base-backbone interactions. At low ionic strength and pH, these interactions
are the dominant cohesive ones; but with more effective charge shielding and
reduction of the fraction of dA residues that are protonated at pH values much
above their intrinsic p
K
a
, the dA
+
-backbone interactions are substantially diminished and weakened.
Consequently, while the two types of duplex share the same interbase H-bonding, the ionic and H-bonds between dA
+
and the backbone, which are associated because of the hybrid character of PO
2
-
, probably alter the helical twist of the two strands, so that the structure of
the grooves in the two cases must be different. This must be what largely
accounts for the difference between the CD spectra of the two duplex forms at
pH 5.0 shown in Figure
7
.
It is interesting that poly(A
+
[middot]A
+
), which has the same base-backbone interactions as the d(A
+
[middot]A
+
) base pairs in the d(A
+
-G)
20,30
duplexes, does not survive at neutral pH (
23
). If it did, the interbase A[middot]A H-bonding would be as in the d(A-G)
20,30
linear duplexes, stable at neutrality in moderate to high salt (data not
shown). We take this difference in pH-sensitivity to mean that the neutral duplexes of d(A-G)
20,30
derive their unique stabilization from the alternating presence of the G[middot]G base pairs, which NMR and model building studies (
6
) have shown to stack especially well with their cross-strand nearest-neighbor dA residues.
It is also worth noting that at the higher cation concentrations where the pH-independent duplex is stable, the ionic bond-stabilized [alpha]-helix does not form at even the most optimal acidic pH.
This emphasizes the similarity between the base-backbone interactions of the nucleic acid [alpha]-helix and the acid-dependent duplex, as well as the differences between
the two types of helix. Although the single-stranded structure lacks the stabilization that comes from stacking of
base pairs, stacking that becomes increasingly advantageous for the duplex with
greater strand length, the single-stranded structure is preferred entropically for very short strands,
because of the much greater ease of intra- than intermolecular nucleation.
The biological relevance of the d(A-G)
n
sequences in mammalian genomes is not illuminated by the occurrence of the pH-independent parallel duplex with A[middot]A and G[middot]G base pairs. However, it does suggest, as we have indeed
found, that irregular sequences of dA and dG residues form duplexes, presumably
with irregular sequences of A[middot]A and G[middot]G base pairs. This must be why some homopurine sequences do not
serve effectively as third strands for triplex formation.
This work was supported by NIH grant GM 42936 to J.R.F. and a Fellowship from
Oncor Inc. to N.G.D.
*To whom correspondence should be addressed. Tel: +1 609 258 3927; Fax: +1 609
258 6730; Email: jrfresco@princeton.edu
pH (+- 0.1)
T
m
(oC +- 0.1)
d[(A-G)
20
[middot](A-G)
20
]
d[(A-G)
30
[middot](A-G)
30
]
4.0
46.0
51.7
4.5
47.1
52.0
5.0
45.0
50.0
5.5
43.5
47.9
6.0
43.1
47.4
7.0
43.0
47.3
7.5
43.3
47.4
8.0
43.4
47.2
REFERENCES
Return