ABSTRACT
The base pair lifetimes and apparent dissociation constants of a 21 base DNA
hairpin and an analog possessing a disulfide cross-link bridging the 3
'
- and 5
'-terminal bases were determined by measuring imino proton exchange rates as
a function of exchange catalyst concentration and temperature. A comparison of
the lifetimes and apparent dissociation constants for corresponding base pairs
of the two hairpins indicates that the cross-link neither increases the number of base pairs involved in fraying nor
alters the lifetime, dissociation constant, or the opened structure from which
exchange occurs for the base pairs that are not frayed. The cross-link does, however, stabilize the frayed penultimate base pair of the stem
duplex. Significantly, it appears that the disulfide cross-link is more effective at preventing fraying of the penultimate base pair
than is the 5 base hairpin loop. Because this disulfide cross-link can be incorporated site specifically, and does not adversely affect
static or dynamic properties of DNA, it should prove very useful in studies of
nucleic acid structure and function.
The static structure of DNA can explain many aspects of its function and
properties (
1
). However, local base pair opening is implicated in a number of important
chemical, biological, and mechanical processes that involve DNA (
2
-
5
). Hence, defining the dynamics of base pair opening is necessary to fully
understand the physicochemical and biological properties of DNA. Information on
the opening and dynamics of individual base pairs within DNA secondary
structures can be obtained through proton exchange measurements on small
oligonucleotides (
6
).
As isolated nucleosides, the pKa
values for the N1 proton of guanosine and the N3 proton of thymidine are 9.2
and 9.6, respectively (
7
). The basic (or unprotonated) form of buffers with pKa
values approaching that of these imino protons are therefore effective exchange
catalysts (
6
,
8
). Guéron and coworkers showed that imino protons exchange from duplex DNA at
rates that are catalyst dependent and developed a theoretical framework to
extract base pair lifetimes and equilibrium constants from measured exchange
rates (
6
,
8
). For example, consider a dA:dT base pair:cpile {{{roman A} : H {roman T}} above "(closed state)"} cpile {{1 / {tau sub
0}} above leftrightarrows above roman {k sub {c l}}} cpile {{{roman {A +}} H
{roman T}} above "(open state)"} {roman {{^ K} sub D}} {back 23 {{roman =}
{back 23 {{roman {{1 / ( tau} sub 0}} {back 31 {{roman cdot} {back 15 {{roman
{{k sub {c l}} )}} {back 15 {= {back 23 {{roman "[open state]"} over roman
"[closed state]"}}}}}}}}}}}}
1
where [tau]0
is the base pair lifetime, kcl
is the closing rate constant, and
H
is the exchangeable imino proton. The thymidine imino proton will exchange from
the open state at a rate, 1/[tau]ex,open
:
A + H T " " cpile {{1 / {tau sub {e x , o p e n}}} above ->} A + H T
2
For KD
<< 1 (a stable base pair) the proton exchange time, [tau]ex
, is given by:
[tau]ex
= [tau]0
+ [tau]ex,open
/F
3
where
F
is the fraction of time the nucleotide is in the open state. The imino proton
of an isolated nucleoside will exhibit an exchange time, [tau]i
, that can be related to [tau]ex,open
by introducing a factor [alpha], where [alpha] = [tau]i
/[tau]ex,open
. The factor [alpha] reflects the relative accessibility to exchange catalyst of an imino
proton in an isolated nucleoside versus an imino proton in an opened base pair.
The fraction of nucleotide in the open state can be calculated by,
F
= (
K
D
/
K
D
+1). Equation 3 can therefore be expanded:
[tau]ex
= [tau]0
+ [tau]i
(1 +
K
D
)/[alpha]
K
D
4
Proton exchange catalyst concentration, [cat], is inversely proportional to [tau]i
, so that a plot of [tau]ex
versus 1/[cat] affords a straight line that extrapolates to [tau]0
at infinite [cat]. For an isolated nucleoside, where [tau]ex
= [tau]i
, such a plot extrapolates to zero at infinite [cat]. For
K
D
<< 1, the ratio of slopes for an isolated nucleoside and a base pair is equal to
[alpha]
K
D
, the apparent dissociation constant (
6
,
8
). When comparing base pairs of the same type (i.e., dA:dT or dG:dC) the ratios
of slopes yield the ratio of the apparent dissociation constants since the
relevant exchange time, [tau]i
, is the same for base pairs of the same type.
Base pair lifetimes on the order of seconds have been observed for highly
stabilized structures like Z-DNA (
9
), while unstable constructs such as mismatched base pairs exhibit base pair
lifetimes <1 ms (
10
). In some cases, measurement of base pair opening has been used to probe for
local DNA structural perturbations within DNA (
11
,
12
). Although the opening dynamics for several different constructs have been
measured (
13
-
22
), the opening kinetics for oligonucleotides constrained with cross-links (e.g., hairpin loops, glycol bridges, etc.) have not been reported.
Indeed, cross-linked nucleic acids are becoming increasingly important in a number of
applications to overcome the limitations imposed when using short
oligonucleotides as models of high molecular weight nucleic acids (
23
-
28
). Our laboratory has developed a general strategy for conformationally
restraining nucleic acids through the site-specific incorporation of disulfide cross-links bridging the 3' and 5'-terminal bases of duplex DNA (
29
-
32
). Previous studies have described the chemical,
in vitro
biological, and thermodynamic consequences of modifying DNA with our cross-link (
33
-
37
). Of particular importance, UV, CD and NMR experiments demonstrate that our
disulfide cross-link does not perturb the static structure of duplex DNA (
32
,
33
,
36
,
38
-
40
).
A comprehensive understanding of the structural consequences of incorporating
cross-links into nucleic acids requires knowledge of how the linker affects base
pair opening. Such information is important because base pair opening is
involved in the mechanical properties of DNA (
2
,
3
), several chemical reactions of DNA (
4
), and a variety of biological processes (
5
). To determine how our disulfide cross-link affects the dynamics of duplex DNA, the base pair lifetimes and
apparent dissociation constants of a 21 base DNA hairpin and an analog
possessing our disulfide cross-link (Fig.
1
) were determined by measuring imino proton exchange rates as a function of
exchange catalyst concentration and temperature. This system was selected for
these experiments because the structures of
1
and
2
are isomorphous as determined by NMR (
38
), and hairpin loops are the most common way to stabilize DNA/RNA duplexes (
41
-
43
). Moreover, incorporating both a loop and the disulfide modification within the
same molecule should allow the assessment of which `cross-link' is better at stabilizing DNA. We find that the disulfide cross-link in
2
does not alter the base pair opening rates or the equilibrium constants for
base pairs that are not frayed as compared with the corresponding base pairs in
1
. Significantly, it appears that the disulfide cross-link is more effective at preventing fraying than is the 5 base hairpin
loop.
Hairpins
1
and
2
were synthesized and purified as described previously (
38
). Samples were ion exchanged by equilibrating in NaCl (4 M) solution for 36 h
and desalted on a Vydac C4 column eluting with water. DNA samples were
lyophilized and dissolved in Tris buffer (5.0 mM) containing NaCl (50 mM), EDTA
(2 mM) and 10% D2
O at pH 7.9. The conductivity of each sample was measured using an Orion model
50 conductivity meter.
Tris buffer concentrations were varied by addition of a stock solution of Tris
(0.93 M) in NaCl (50 mM), EDTA (2 mM) and 10% D2
O. The pH was adjusted with NaOH (1 M) or HCl (1 M) and measured with a
microelectrode (Microelectrodes, Inc. MI-412) inside the NMR tube at 22.4oC before and after each NMR experiment. Before each experiment the
samples were heated in 95oC water for 30 s and reannealed by slow cooling to room temperature. The
samples were kept under nitrogen during the NMR experiments. The concentration
of the unprotonated, exchange catalyst form of Tris, [cat], was calculated
according to:
[cat] = [B]/(1 + 10(pKa-pH)
)
5
where [B] is the total Tris concentration, and the pKa
of the Tris was determined by titration at 22.4oC (
44
). The variation of pH with temperature was calculated according to a Tris pKa
change of -0.031 [Delta]pKa
/oC (
45
).
NMR Spectra were measured on an Bruker AMX500 operating at 500 MHz. Spectra were
analyzed using FELIX versions 2.3 and 95.0 (BIOSYM Technologies). The proton
longitudinal relaxation times were measured using the inversion recovery
technique. An I-BURP (
46
) pulse (4 ms) was used for the selective inversion of the imino protons. The
water signal was suppressed with a 1-1 read pulse (
47
). For each T1 measurement 16-20 values for the recovery delay were used.
The intensities at each delay time were measured by fitting the spectra to a set
of Lorentzian peaks. For each inversion recovery series the spectrum measured
using the longest delay time was fit first. The linewidths and peak positions
were then held constant in subsequently fitting peaks to the spectra measured
with shorter delay times. As the catalyst concentrations were increased, the
resonances became progressively exchange broadened making it impossible to
measure the relaxation times of some of the resonances at high temperatures and
exchange catalyst concentrations.
To determine T1
values, plots of peak area versus delay time were fit to equation 6:
I = I0
(1 - 2 e-(t-[tau]0)/T1)
6
where
I
is the peak area at each delay time,
t
is the delay time,
I
0
is the area of the fully recovered peak, and [tau]0
is a delay value used to compensate for incompletely inverted magnetization (
6
,
8
). The exchange times were then determined according to the equation:
1/
T
l
= 1/[tau]ex
+ 1/
T
laac
7
where
T
l
is the longitudinal relaxation time measured at a given catalyst concentration,
T
1aac
is the longitudinal relaxation time measured without any added exchange
catalyst, and [tau]ex
is the proton exchange time at the given catalyst concentration (
48
). The use of T1
measurements rather than linewidth measurements for the exchange time
calculations has been shown to help diminish the effect of increased catalyst
concentration on the dipolar relaxation rate (
49
).
The opening rates for individual protons were determined by fitting plots of [tau]ex
versus 1/[cat] to a straight line. Several data points at very low catalyst
concentration did not lie on the fitting line and were excluded from the
fitting for each imino proton. The slopes of the lines were used to determine
apparent dissociation constants in accordance with equation 3.
The base pair opening kinetics of
1
and
2
were studied by measuring selective T1
times of the imino proton resonances shown in Figure
2
as a function of exchange catalyst concentration. A representative selective T1
series is presented in Figure
3
. The imino resonances of
1
at various catalyst concentrations demonstrate the effect of increased catalyst
concentration; line broadening is clearly evident due to enhanced proton
exchange catalyzed by the protonated form of the Tris buffer (Fig.
4
). The variation of exchange time ([tau]ex
) with inverse catalyst concentration for the imino protons at 10, 20 and 30oC is shown in Figure
5
, and the extrapolated values of [tau]0
along with the ratio of the apparent dissociation constants for corresponding
base pairs of hairpins
1
and
2
are presented in Table
1
. As seen in Table
1
the [tau]0
values are similar to those measured for other B-DNA duplexes (
6
,
8
-
10
,
13
-
22
). Because the imino protons for base pairs 2 and 6 of hairpin
2
are severely overlapped, separate opening rates could not be measured.
Previous NMR experiments have demonstrated that the disulfide cross-link does not perturb the static structure of
1
relative to
2
(
38
). The goal of this study was to determine how the base pair opening rates
(i.e., the `dynamic structure') are affected by the cross-link. In the context of this work, it is important to distinguish between
two modes of base pair opening. Base pairs distant from the termini in B-form DNA appear to open individually (
6
,
8
), while the base pairs near the termini open cooperatively due to fraying (
51
). As a result of fraying, imino proton exchange times increase with distance
from the termini, and typically reach a `bulk' rate at the fourth base pair
from the end of a duplex (
50
,
52
,
53
). Since our cross-link is at the terminus of a duplex, we investigated fraying as well as
`bulk' base pair opening for each base pair of
1
and
2
.
If our cross-link is destabilizing, base pairs in the stem duplex of
2
should exhibit decreased [tau]0
and increased
K
D
relative to the corresponding base pairs of
1
. The destabilizing effect could extend throughout the stem or be limited to the
base pairs closest to the cross-link which would be indicative of increased base pair fraying. In
contrast, a stabilizing cross-link would decrease terminal fraying and would be manifested by increased [tau]0
and decreased
K
D
for the terminal residues of
2
relative to the corresponding positions of
1
. The cross-link could also leave base pair fraying unaffected which would result in
identical [tau]0
and
K
D
for all of the corresponding base pairs of
1
and
2
. Because the extent of fraying changes with temperature, it was anticipated
that the effect of the cross-link on the stability of corresponding base pairs could also change with
temperature.
Within experimental error, the lifetimes and apparent dissociation constants for
base pairs 3-5 of hairpins
1
and
2
are identical at 10, 20 and 30oC. Since both the lifetimes and dissociation constants are the same for the
corresponding base pairs in position 3-5, the closing rates, kcl
, must also be the same [see equation 1]. Furthermore because the [alpha] factors of corresponding base pairs are the same, the exchange time for
the nucleotides in the open states, [tau]ex,open,
must be identical since [alpha] = [tau]i
/[tau]ex,open
. This latter point suggests that the open states themselves are identical.
These findings, in conjunction with our earlier structural data (
38
-
40
), show that the process of base pair opening including the closed state,
transition state of opening and closing, and the opened structure from which proton exchange occurs, are the same for base pairs 3, 4
and 5 in hairpin
1
and its disulfide cross-linked analog.
Base pairs 7 in
1
and
2
are adjacent to loops and are frayed. The apparent dissociation constant at 10oC is identical for both hairpins, the only temperature at which an apparent
dissociation constant could be measured for this position. However, the melting
experiments suggest that the thermal denaturation pathway of base pair 7 of
1
and
2
is identical up to 40oC (Fig.
3
). The results here indicate that the cross-linked thymidines do not affect the opening dynamics for base pairs 3-7 in
2
.
Although the terminal base pairs in duplex DNA possess a high
K
D
relative to the other base pairs, they protect penultimate base pairs from
fraying (
51
). Simple models of fraying propose that when the terminal base pair is closed
it protects the penultimate base pair, but when the terminal base pair is open
the penultimate base pair is unprotected and exhibits a large
K
D
(
50
). Comparing the opening properties of base pair 1 of hairpins
1
and
2
therefore allows for the investigation of how well the N3 cross-linked thymidines at the bottom of hairpin
2
mimic the dA:dT base pair at the bottom of hairpin
1
in protecting the adjacent base pair.
Because there is no `closed state' available to the N3 cross-linked thymidines, protection of the base pairs in position 1 of
1
and
2
must occur by a different process relative to a dA:dT base pair. In addition,
the open states of base pair 1 of
1
and
2
must also differ significantly because the frayed open state of base pair 1 of
2
is covalently constrained. Thus, the open state accessibility of the imino
protons in position 1 of
1
and
2
may differ greatly, leading to different values of [alpha] for those imino protons. While [alpha] factors in other DNA oligomers and complexes are often presumed
to approximate unity when ammonia is used as an exchange catalyst, that
assumption cannot be made here, especially since [alpha] factors that differ by a factor of three have been measured with
different exchange catalysts (
6
,
8
). These differences make interpretation of apparent dissociation constants ([alpha]
K
D
values) difficult for base pair 1 of
1
and
2
.
Incorporation of a loop on the end of a duplex is the most common method for
stabilizing nucleic acid duplexes (
41
-
43
). Examination of the melting curves of
1
and
2
allows for a qualitative evaluation of the relative stabilizing effect of the
d(CATTT) loop and the N3 cross-linked thymidines. In hairpin
1
the imino protons of base pairs 1 and 7 broaden to a similar extent as a
function of temperature, while in hairpin
2
the imino proton of base pair 1 is broadened to a far lesser extent at 40oC. From these data we conclude that the d(CATTT) loop is less effective at
stabilizing the stem duplex than the cross-linked thymidines. Since our disulfide cross-link does not significantly alter dynamic or static structure of
duplex DNA, can be easily incorporated into oligonucleotides, is much smaller
relative to hairpin loops, and the chemical,
in vitro
biological and thermodynamic consequences of constraining DNA with this
modification are established (
26
,
29
-
40
), use of this modification in nucleic acid structure-functions studies should see increased use.
Supported by NIH Grant GM 52831. G.D.G. is the recipient of a National Arthritis
Foundation Arthritis Investigator Award, an American Cancer Society Junior
Faculty Research Award, a National Science Foundation Young Investigator Award,
a Camille Dreyfus Teacher-Scholar Award, and a Research Fellowship from the Alfred P. Sloan
Foundation.
REFERENCES
Return