ABSTRACT
`Stacking hybridization reactions' wherein two or more short DNA oligomers
hybridize in a contiguous tandem orientation onto a longer complementary DNA
single strand have been employed to enhance a variety of analytical
oligonucleotide hybridization schemes. If the short oligomers anneal in perfect
head-to-tail register the resulting duplex contains a nick at every boundary
between hybridized oligomers. Alternatively, if the short oligomers do not
hybridize precisely in register, i.e. single strand regions on the longer
strand are left unbound, gaps are formed between regions where short oligomers
bind. The resulting gapped DNA duplexes are considerably less stable than their
nicked duplex analogs. Formation of base pair stacking interactions between neighboring oligomers at the nicks that do not occur in
gapped duplexes has been proposed as the source of the observed added stability. However, quantitative evidence supporting this hypothesis for DNA has not been
reported. Until now, a direct comparison of the thermodynamics of DNA nicks
versus DNA gaps has not been performed. In this communication we report such a
comparison. Analysis of optical melting experiments in a well defined molecular context enabled quantitative evaluations of the relative thermodynamic difference between nicked and gapped
DNA duplexes. Results of the analysis reveal that a nick may be energetically
favored over a gap by at least 1.4 kcal/mol and perhaps as much as 2.4
kcal/mol. The presence of a 5
' phosphate at a nick or gap fails to significantly affect their stabilities.
The term `stacking hybridization' has been used to describe the additional
stability associated with DNA hybridization reactions wherein two or more DNA
oligomers hybridize in a contiguous tandem arrangement to a longer
complementary DNA single strand (
1
,
2
). If no single strand residues on the longer DNA molecule are left unbound, a
`base stack' is formed at each nick between pairs of adjacent DNA oligomers. Relative to hybridization of the same
short oligonucleotides in a tandem but non-contiguous arrangement where single strand regions, or gaps, exist between adjacent
duplex regions, nicked DNA complexes are remarkably more difficult to
dissociate. This effect has been attributed to formation of base pair stacking
interactions at the nicks that cannot form in their gapped counterparts (
1
-
4
).
Evidence that such stacking reactions can be employed to add discriminatory advantages to deoxy-oligonucleotide hybridization- dissociation reactions has been reported (
1
,
2
). In these studies decreased dissociation rates were observed for duplexes
formed from tandemly stacked short oligomers hybridized to a longer strand, relative to the same oligomers hybridized non-contiguously, i.e. without creation of new stacks. Improved rates of nucleic acid cleavage by EDTA derivatized oligodeoxynucleoside methylphosphonates have also been attributed to added stability provided by stacking
interactions (
5
). Either alone or in combination with enzymatic steps, stacking hybridization reactions have also been utilized to improve standard sequencing reactions involving hybridizations of short primers (
6
,
7
) as well as reduce ambiguities and improve resolution in sequencing by hybridization (SBH) methods (
1
-
4
,
7
-
12
). Despite these apparently successful applications of stacking hybridization, an equilibrium thermodynamic analysis of the
observed effects has not been performed. Although several solution structural
and biochemical studies of DNA nicks have been reported (
13
-
16
), direct comparisons between the melting behavior of nicked and gapped DNA
duplexes have not been made.
We encountered this deficiency when attempting to utilize such a stacking
strategy to aid in `capturing' via hybridization, specific oligonucleotide
sequences from solution at extremely dilute concentrations (<10
6
copies in 100 [mu]l). Under these conditions we observed high hybridization efficiencies for
perfectly stacked complexes. We have performed melting analysis of several
model DNA molecules and evaluated the relative thermodynamics of nicks and gaps
within this well defined molecular context.
The DNA sequences shown in Figure
1
were synthesized on a 1 [mu]mol scale and purified by the described methods (
17
) or purchased from Oligos (Wilsonville, Oregon) and received as ion-exchange purified sodium salt. All samples were exhaustively dialyzed against ddH
2
0 (nanopure), vacuum dried and stored at -20oC. As shown in Figure
1
the hairpin forming sequences are self-complementary to the extent that each can fold upon itself to form an
intramolecular hairpin having a single strand 3' dangling- end and a T-T-U
B
-T-T loop, where U
B
is uracil biotinylated at the C5 position of the uridine ring. The hairpins shown explicitly in Figure
1
terminate with a 5'-OH group (-P). Hairpins with the same sequences, phosphorylated on the 5' end were also prepared and studied (+P, not shown).
The single strand 13 base DNA oligomer with sequence complementary to the 13
most bases of the 3' dangling ends of the hairpins was also prepared. Mixture and annealing
of the four different hairpins with the 13mer single strand resulted in
formation of four variations of nicked and gapped complexes. These comprised
the model molecular system for our studies.
The thermodynamics of DNA nicks and gaps were evaluated from melting analysis of
four novel DNA molecular constructs. Similar RNA constructs have been studied
by Turner's group (
18
-
20
). Our DNA study differs from the RNA studies in several significant ways.
First, in our complexes the hairpin stems contain 16 base pairs while the stems
of the reported RNA hairpins contain seven or eight base pairs. In addition our
hairpins have 13 base 3' dangling ends while the reported RNAs have only four or five base 5' dangling ends. Finally, since our DNA hairpins were designed to
be affixed to solid support for the purpose of capturing short oligomers from
solution, our loops contain a biotinylated uracil at the third position flanked
on both sides by two Ts. The RNA constructs contained either four base 5'-G-C-A-A-3' or five base 5'-A-A-A-A-A-3'
loops.
Representations of the DNA reactions studied are depicted in Figure
1
. Consider the nicked (`stacked') duplex that forms between the hairpin and
complementary 13 base oligomer to form a nicked duplex. This is depicted as
Reaction 1
in Figure
1
a. The influence of a phosphate at the nick was also investigated by preparing
the nicked complex with a 5' phosphorylated hairpin and the 13mer. This is indicated as (+P) in
Figure
1
but not explicitly shown. Analogous `unstacked' or gapped duplexes were prepared from hairpins whose 3' most 13 bases of the 15 base 3' dangling-ends, are complementary to 13mer strand. The resulting
complex contains a 13 base pair duplex, but the hairpin stem instead adjoins a
gap opposed by two Ts on the 5' strand. This is annealing
Reaction 2
of Figure
1
b. To prepare the phosphorylated gapped complex, the hairpin was phosphorylated at the 5' terminus, and annealed with 13mer strand [(+P) in Fig.
1
].
DNA strands were prepared for melting experiments by dissolving them in melting buffer (100 mM NaCl, 10 mM Phosphate, 1 mM EDTA, pH = 7.5).
This combination of salts amounts to a final concentration of 115 mM Na
+
. The four complexes that were studied were prepared by mixing approximately
equimolar amounts of 13mer strand with each of the hairpins described above.
These mixtures were heated to 90oC for 10 min and allowed to slowly cool to 15oC. Total strand concentrations of the mixtures were ~0.8 [mu]M. In terms of the melting experiments this corresponds to a
starting absorbance at 268 nm of ~0.35. Strand concentrations were determined from the extinction
coefficients of the constituent strands estimated by the nearest-neighbor method (
21
). For the hairpins a value of 9069 cm
-1
M
-1
/base was used. For the 13mer strand an extinction coefficient of 9097 cm
-1
M
-1
/base was employed.
Melting curves were collected on a Hewlett-Packard 8452A single beam spectrophotometer equipped with a model 89090A
temperature controller. Sample temperature was determined from the cell holder temperature corrected for differences between it and the
temperature read by a probe immersed directly in the sample. From independent calibration experiments it was determined, over the temperature range where melting curves were measured (20-90oC), the sample temperature lags 2-4oC behind that of the sample holder in a linear fashion.
Temperatures of all the collected melting data were corrected to account for
this variation. In separate experiments, absorbance of the buffer alone was
also measured as a function of temperature. These values were then subtracted
from raw absorbance versus temperature curves of the DNA samples. Sample
absorbance was monitored at 268 nm while the temperature was increased linearly
at 30oC/h. At least three melting experiments were conducted on independently prepared samples of all complexes. Absorbance versus temperature curves were also collected for the hairpins alone. Analysis of these data
in conjunction with those collected for the hairpin:oligomer complexes under identical experimental conditions, enabled extraction of the
melting transitions of the nicked and gapped, phosphorylated and unphosphorylated hairpin:oligomer complexes.
Transitions corresponding to melting of the 13mer strand from the hairpin stems
were deconvoluted from melting curves of the respective complexes using the
melting curves collected for the hairpins alone. With only minor modifications
the analysis was analogous to that previously applied to extract melting
transitions of DNA triplexes formed between a duplex substrate and third strand (
22
). Steps in the procedure were as follows. (i) Pre-transition and post-transition base lines were determined from linear fits of the
sloping regions before and after the main transition on the melting curves of
the hairpins alone [A
HP
(T) versus T]. The fraction of melted hairpin as a function of temperature was
obtained by normalizing the experimental melting curve between these upper and
lower base lines,
viz.
[theta]
HP
(T) = [A
HP
(T) - A
L
HP
(T)]/[A
U
HP
(T)-A
L
HP
(T)]
1
Where, A
HP
(T) is the experimentally acquired absorbance of the hairpin alone at 268 nm, A
L
HP
(T) is the lower linear base line on the experimental curve and A
U
HP
(T) is the upper linear base line on the experimental melting curve of the hairpin alone. (ii) Similarly linear regions were fit on the melting curves of the hairpin:oligomer complexes [A
HP-O
(T) versus T]. On these curves three linear regions were fit. These were the
upper and lower linear base lines on either side of the two main transitions on
the melting curves of the hairpin:oligomer complexes designated, A
U
HP-O
(T) and A
L
HP-O
(T), respectively. And the linear region separating the two main transitions in the
middle of the complex melting curve, designated A
M
HP-O
(T). (iii) The fraction of hairpin:oligomer complexes dissociated, f
D
(T), was obtained from the experimental A
HP-O
(T) versus T curve and the aforementioned base lines. When the concentrations of the hairpin and 13mer were equal, C
HP
o
= C
O
o
, or the 13mer was in excess, C
O
o
> C
HP
o
,
f
D
(T) = [A
HP-O
(T) - A
L
HP-O
(T)]-[A
M
HP-O
(T) - A
L
HP-O
(T) + [theta]
HP
(T) (A
U
HP
(T)-A
L
HP
(T))]
-1
2
Alternatively, when the hairpin was in excess, C
HP
o
> C
O
o
,
f
D
(T) = [A
HP-O
(T) - A
L
HP-O
(T) - [theta]
HP
(T)(A
U
HP
(T)-A
L
HP
(T))(1 - C
O
o
/ C
HP
o
)] - [A
M
HP-O
(T) - A
L
HP-O
(T) + [theta]
HP
(T) (A
U
HP
(T)- A
L
HP
(T))(C
O
o
/ C
HP
o
)]
-1
3
(iv) Accuracy of the deconvolution procedure and model analysis were confirmed
by regeneration of the actual melting curves of the hairpin:oligomer complexes from the extracted f
D
versus T curves and [theta]
HP
(T) versus T curves of the hairpins alone. (v) Extracted f
D
(T) versus T transitions collected for each type of hairpin:oligomer complex
were fit utilizing the expression derived from the sequential three-state model given in equation
5
below. Values of the transition enthalpy, [Delta]H
o
, and entropy, [Delta]S
o
, were assumed to be independent of temperature and evaluated empirically as adjustable parameters in the fits. Fits were performed using the Marquardt non-linear least squares routine as described (
23
), and fits were deemed acceptable when the evaluated adjustable parameters ([Delta]H
o
and [Delta]S
o
) did not vary more than one percent between successive iterations.
Melting data were analyzed in terms of a sequential three-state equilibrium thermodynamic model. Consider the melting reactions of various
hairpin:oligomer complexes (i.e. nicked, gapped, phosphorylated,
unphosphorylated) whose melting can be generically represented by the sequential melting equilibrium,
K
obs
K
HP-C
HP-O <-> HP + O <-> S + O
4
where HP is the fully intact dangling-ended hairpin, O is the 13mer oligomer, HP-O is the hairpin:oligomer complex and S is the melted hairpin
strand. For the constructs examined, HP-O can be a phosphorylated or unphosphorylated, nicked or gapped, hairpin:13mer oligomer complex. The model has melting proceeding in a concentration dependent manner from the first state, the fully intact, hydrogen bonded and stacked hairpin:13mer complex (HP-O), to the second state, a mixture of the dissociated 13mer single
strand and the fully intact hairpin (HP + O). This is the first transition on
the melting curves of the complexes. The observed equilibrium constant for this reaction is K
obs
= [HP] [O]/[HP-O], from which the thermodynamics of nick and gap formation can be
determined. The transition from the second to the third state (the melted
hairpin and 13mer single strand) is the concentration independent melting of
the intact intramolecular hairpin in the presence of the 13mer (HP + O) to a
coiled single strand (S + O), seen as the higher temperature melting
transitions on complex melting curves.
An analytical expression for f
D
versus T was required. Since the total strand concentration is given by C
T
= 2[HP-O] + [HP] + [O], the fraction of the total strands that reside in duplex
complexes is f
I
= 2[HP-O]/C
T
. Thus, in terms of C
T
and K
obs
, the fraction of melted duplex complex, f
D
= 1-f
I
, can be expressed as,
f
D
= K
obs
((1 + 2C
T
/K
obs
)
[1/2]
- 1))/C
T
5
Expressing K
obs
= exp[(T[Delta]S
o
- [Delta]H
o
)/RT] and employing equation
5
, [Delta]S
o
and [Delta]H
o
were evaluated as adjustable parameters in fits of the experimental f
D
versus T melting curves. From the resulting values of [Delta]S
o
and [Delta]H
o
, the transition free-energies of the melting reactions were determined, [Delta]G
Reaction
= [Delta]H
o
- T[Delta]S
o
. With accurate knowledge of the initial concentrations of the single strand oligomers and hairpins, which were by design approximately equal (to within 10%) in all experiments, results obtained for each type of
complex were directly comparable. Determination of the absolute stability of
the nicked and gapped complexes would have required melting of the linear 13mer
duplex alone which was not done. Even though melting of the various hairpin:oligomer complexes as a function of total strand concentration was not performed, with
the assumptions employed results of the analysis provided quantitative
estimates of the thermodynamics of melting the various complexes, and a basis
upon which to meaningfully compare their relative thermodynamic stabilities.
Since we were primarily interested in determining relative differences between the nicked and gapped complexes, conducting melting experiments under the same conditions at the same strand concentrations
and employing the analysis as described above was sufficient for this purpose.
To test validity and accuracy of the model assumptions and analysis procedures
used to extract the f
D
versus T transitions, melting curves of the complexes were reconstructed from
concentration weighted averages of the melting transitions of the resident
duplex (HP-O <-> HP+O) extracted through the analysis procedure, and the melting transition of the hairpin alone (HP <-> S) that was independently measured. In this analysis, the
extracted normalized f
D
versus T melting curves, the measured melting curves of the complexes normalized
to upper and lower linear base lines, [theta]
M
comp
versus T, and base line normalized melting curves of the hairpins alone, [theta]
HP
versus T [see equation
1
], were employed. The reconstructed melting curve of the complex (the total
fraction of melted base pairs as a function temperature), [theta]
R
comp
(T), can be written as,
[theta]
R
comp
(T) = (C
o
HP
[theta]
HP
(T) + C
o
D
f
D
(T))/(C
o
HP
+ C
o
D
)
6
Where C
o
HP
is the total concentration of hairpin, C
o
D
is the maximum concentration of the hairpin:oligomer complexes. Assuming that
all potential complexes actually form, C
o
D
could be determined from concentrations of the hairpin and oligomer strands C
o
HP
and C
o
O
, respectively. For example, if C
o
HP
= C
o
O
, then C
o
D
= C
o
HP
= C
o
O
. However, if C
o
HP
> C
o
O
, C
o
D
= C
o
O
. Likewise, when C
o
HP
< C
o
O
, C
o
D
= C
o
HP
. Direct comparisons of the measured and reconstructed normalized melting curves
of the complexes, [theta]
M
comp
and [theta]
R
comp
versus T provided assessments of the overall accuracy of the analysis
procedures.
The high resolution optical melting curves collected in 115 mM Na
+
for the phosphorylated and unphosphorylated nicked and gapped hairpin:13mer complexes (depicted in Fig.
1
) are displayed in Figure
2
a-d. The curves shown are the average of at least three independent
experiments. Insets are the melting transitions of the respective dangling-ended hairpins alone (in the absence of the 13mer strand) at the same
concentration under identical solvent conditions.
In effect two types of hairpin:oligomer hybridization reactions (HP-O <-> HP + O) were studied; those for the nicked duplexes (phosphorylated
and unphosphorylated), referred to as
Reaction 1
(+-P) and gapped duplexes referred to as
Reaction 2
(+-P) (Fig.
1
).
From the empirically evaluated [Delta]Ho and [Delta]So values determined from three independent experiments, the free-energy changes for the reactions shown in Figure
1
were determined. For the nicked complexes, the evaluated melting transition
free-energies (20oC) were [Delta]Go
Reaction 1
(+-P) = -18.8 +- 0.1 kcal/mol. For the gapped complexes, [Delta]Go
Reaction 2
(+-P) = -17.6 +- 0.1 kcal/mol. Therefore, the additional free-energy of formation for the nicked duplexes [
Reaction 1
(+-P)] relative to the gapped duplexes [
Reaction 2
(+-P)] is -1.2 +- 0.2 kcal/mol. Through the course of the analysis
process, contributions from hairpin melting to the evaluated free-energies of
Reactions 1
and
2
were eliminated. Therefore, the evaluated standard state free-energies for the reactions, [Delta]Go
Reaction 1
(+-P) and [Delta]Go
Reaction 2
(+-P), can presumably be partitioned into two independent terms; one for
the free-energy of the duplex, [Delta]Go
D
, and one for the nick or gap (phosphorylated or unphosphorylated), [Delta]Go
nick(+-P)
or [Delta]Go
gap(+-P)
, respectively. That is, the evaluated free-energies can be written as,
[Delta]Go
Reaction 1
(+-P) = [Delta]Go
D
+ [Delta]Go
nick(+-P)
7a
[Delta]Go
Reaction 2
(+-P) = [Delta]Go
D
+ [Delta]Go
gap(+-P)
7b
where, as stated above, [Delta]Go
D
is the standard free-energy change in melting of the 13 base pair duplex formed between the
13mer and dangling-ended hairpin. This energetic contribution which considers all factors
involved in formation of the duplex not associated explicitly with the nicks
and gaps, is assumed to depend only on the sequence and length of the duplex
and oligomer:hairpin strand concentration ratio. Since the duplex regions in
the nicked and gapped complexes are identical, [Delta]Go
D
is assumed to be equivalent for the 13 base pair duplexes in all nicked and
gapped complexes at the same concentrations of 13mer and hairpin. The energy partitioning in equations
7a
and
7b
allows a precise evaluation of the relative free-energy difference between a nick and a gap. That is, [Delta]Go
nick
and [Delta]Go
gap
represent explicitly the free-energy contributions associated with the nicked and 5'-T-T gapped complexes, not included in [Delta]Go
D
. Thus, the difference in the experimentally determined values of [Delta]Go
Reaction 1
(+-P) and [Delta]Go
Reaction 2
(+-P), provides the relative free-energy difference between a nick and gap, viz. [Delta][Delta]Go = [Delta]Go
Reaction 1
(+-P) -[Delta]Go
Reaction 2
(+-P) = -1.2 +- 0.2 kcal/mol. This value is actually considered to be a
lower limit estimate on the magnitude of the nick free-energy because it was determined relative to a gapped complex that has a 5'T-T single strand sequence adjoining the duplex. It has been
shown that such 5' single strand overhangs can stabilize an adjacent duplex structure (
17
,
24
). An estimate of the energetic contribution of the T-T adjoining sequence can be readily obtained from these published studies.
From studies of the effects of 5'-T-T overhangs on the stability of linear duplex DNAs in 1.0 M
NaCl, a value of -1.2 kcal/mol was evaluated (
24
). Studies of the effects of four base 5' dangling ends on the stability of DNA hairpins in 115 mM Na
+
determined a stabilizing contribution of ~-0.2 kcal/mol (
17
). Although evaluated in different salt environments from different molecular
systems, these estimates provide a plausible range (-0.2 to -1.2 kcal/mol) for the expected energetic contributions of the T-T gap to adjoining duplex stability. Therefore, the
estimated free-energy of stability afforded by a nick is at least -1.4 kcal/mol and can be as great as -2.4 kcal/mol.
This study was undertaken with the goal of evaluating the relative
thermodynamics of DNA nicks and gaps in order to explain observations that 3' dangling-ended DNA hairpins are remarkably efficient at capturing (via stacking hybridization) complementary
DNA oligonucleotide sequences present at extremely dilute concentrations (<10
6
copies in 100 [mu]l). As far as we are aware, our results provide the first estimates and
direct comparisons of the thermodynamics of a DNA nick and gap. In this, our
initial study, we employed four model synthetic DNA hairpin constructs to
evaluate the stacking hybridization parameters explicitly associated with formation of a phosphorylated and unphosphorylated nick by establishment of a 5'-G-C-3' stack relative to a phosphorylated and
unphosphorylated 5'-T-T-3' single strand gap. In the future similar
synthetic constructs will be employed to evaluate the relative energetics and
influence of other sequence stacks and base pair mismatches at and around DNA
nicks and gaps. The methods described here provide the analytical foundation for these future studies.
We thank Dr Natasha Broude, Professor Bernard Poiesz and Sean Higgins for
helpful comments on the manuscript. Portions of this work were supported by
N.I.H. grant GM-39471 and grants from Tm Technologies, Inc.
*To whom correspondence should be addressed at: Department of Chemistry, M/CIII,
room 4500, Univerisyt of Illinois at Chigago, 845 West Taylor Street, Chicago,
IL 60607, USA. Tel: +1 312 996 0774; Fax: +1 312 996 0431
REFERENCES
Return