ABSTRACT
Formamide lowers melting temperatures (
T
m) of DNAs linearly by 2.4-2.9
o
C/mole of formamide (
C
F) depending on the (G+C) composition, helix conformation and state of
hydration. The inherent cooperativity of melting is unaffected by the
denaturant.
dT
m
/dC
F for 11 plasmid domains of 0.23 < (G+C) < 0.71 generally fit to a linear dependence on (G+C)-content, which, however, is consistent with a (G+C)-independent alteration in the apparent equilibrium
constant for thermally induced
helix
<->
coil
transitions. Results indicate that formamide has a destabilizing effect on the
helical state, and that sequence-dependent variations in hydration patterns are primarily responsible for
small variations in sensitivity to the denaturant. The average unit transition
enthalpy
{DELTA H} bar sub m
exhibits a biphasic dependence on formamide concentration. The initial drop of -0.8 kcal/mol bp at low formamide concentrations is attributable to a
[delta][epsilon][lambda][tau][alpha] {{[Delta]E[Lambda]TA H} [beta][alpha][rho] [sigma][upsi][beta] [mu]}
for exchange of solvent in the vicinity of the helix: displacement by formamide
of weakly bound hydrate or counterion. The phenomenological effects are
equivalent to lowering the bulk counterion concentration. Poly(dA
@
dT) exhibits a much lower sensitivity to formamide, due to the specific pattern
of tightly bound, immobilized water bridges that buttress the helix from within
the narrow minor groove. Tracts of three (A
[middot]
T)-pairs behave normally, but tracts of six exhibit the same level of reduced
sensitivity as the polymer, suggesting a conformational shift as tracts are
elongated beyond some critical length [McCarthy,J.G. and Rich,A. (1991)
Nucleic Acids Res.
19, 3421-3429].
The addition of formamide (HCONH
2
) to aqueous buffer solutions of DNAs lowers their stability, and for this
reason this denaturant is a common additive in many low temperature studies.
Studies by McConaughy
et al.
(
1
), Record (
2
), Casey and Davidson (
3
) and Hutton (
4
) demonstrated that melting temperatures,
T
m
, for thermal denaturation of DNAs decrease linearly by approximately -0.65oC/volume fraction (* 100) formamide. Earlier studies had suggested the denaturing
ability of agents such as formamide is due to their lower ion-solvating power (
5
,
6
), and to their ability to increase the solubility of free bases (
6
). It was said these agents increase the hydrophobic character of the solvent,
thereby decreasing the activity coefficients and free energies of the bases,
favoring the denatured state (
7
-
10
). The results of these early studies seem to have been widely interpreted as
indicating that since the denatured state was favored, the effects of formamide
were largely independent of base content.
Systematic studies of the effects of formamide on DNAs of different (G+C)
content have not been carried out, or at least not reported, so the
relationships of (G+C) content, formamide concentration and stability are not
known. The goal of this study was to examine this question, with the objective
of gaining further insight into the molecular reasons for the denaturing
effects of formamide. If the denaturing effects could be shown to be the
consequence of differences in perturbations of ligand association constants for
the binding of water and ions to helix and coil states of DNA, this agent could
serve as a useful means of probing variations in local structure and hydration
of DNA.
The principal specimens used in these studies were plasmid DNAs, primarily
pN/MCS-10, -11 and -12, constructed in this laboratory by established procedures (
11
) to contain a repetitive sequence at a site in the plasmid with favorable
energetic characteristics. A 55 bp multiple cloning sequence (MCS) with
recognition sites for 10 restriction enzymes was first installed at the unique
Nru
I locus (972) of pBR322 (
12
). Repetitive tridecamers of sequence: [AAGTTGAACAAAT]
n
AAGTTG, where 57 >
n
> 16, were then inserted at the unique
Sma
I locus of the MCS. This locus is immediately adjacent to a unique
Kpn
I recognition sequence of the MCS, and 803 bp away from an
Eco
RV sequence, both sites used to linearize the plasmid in preparation for
melting.
Escherichia coli
HB101 or SUREtm cells containing plasmid were grown in 1 l ampicillin selective LB
broth, and amplified by the addition of chloramphenicol. Cells were lysed in
alkaline SDS, and the plasmid precipitated from the lysate in isopropanol.
Crude plasmid DNA was purified on two CsCl gradients.
In the preparation of oligomer repeats for insertion into pN/MCS, the residues
of
n
-base oligonucleotides were first paired with the residues of complementary
overlaps of (
n
/2)-base oligonucleotides, and then ligated using T4 DNA ligase to generate
tandemly repeating polynucleotide chains of 100-1000 bp. The DNA was size fractionated on low-melting agarose, and fragments >200 bp removed from the gel, and
made blunt with Klenow enzyme. The resulting DNA was ligated into pN/MCS at the
Sma
I locus of the MCS (
12
). In preparation for insertion of the repetitive elements, pN/MCS plasmid was
restricted at the unique
Sma
I site and digested with bacterial alkaline phosphatase to remove the terminal 5'-phosphate groups to prevent vector recircularization during the
ligation step. The phosphatase was inactivated by the addition of SDS, followed
by phenol extraction. Plasmid DNA was then precipitated and ligated to the
repetitive DNA at an insert:vector mole ratio of >10:1 using T4 DNA ligase.
Circular recombinant plasmids were introduced into HB101 cells and transformed
cells selected on agar containing ampicillin. The repetitive inserts were
checked by double-stranded sequencing over the insert region by the dideoxy chain-termination method (
13
). Checked in this fashion and by electrophoretic mobility, it was found that
the insert length,
n
, of the repeat in pN/MCS-10 is 38 (500 bp), 57 in pN/MCS-11 (747 bp), and 16 in pN/MCS-12 (214 bp).
High resolution derivative melting curves were obtained by a difference-approximation method (
14
,
15
), with a modified double-beam ratio-recording spectrophotometer (Cary). The finite difference method
involves the approximation of a derivative d
A
l(nm)
/d
T
by its differential [Delta]
A
l
/[Delta]
T
, with analytical reconstruction of the true derivative achieved as previously
described (
14
). Temperatures were ramped at 6oC/h, which has been shown to provide equilibrium melting of most domains.
Slower rates lead to a significant increase in thermal degradation,
particularly of single-stranded coil regions (
16
). For this reason, specimens were never remelted; instead, replicate
experiments were always carried out on fresh aliquots of the same or new
preparations. The distribution of (G+C) compositions and sizes of domains
responsible for subtransitions were obtained by spectral decomposition of
melting curves obtained at 260, 270 and 282 nm, where the ratios of derivative
extinction coefficients for A[middot]T and G[middot]C pairs, d[epsilon]
A[middot]T
/d[epsilon]
G[middot]C
, are 4.31, 1.00 and 0.120, respectively (
15
).
Commercial poly(dA[middot]dT) (~1 [mu]g; Sigma) of mean length >20 000 bp was added to all specimens
before melting, to serve as secondary standard of solvent conditions. The sharp
melting of this synthetic DNA is given by:
{{italic T} sub m sup {{roman {p o l y ( d A . d t )}}}} = {{1 9 . 0 7 {roman {l
o g}}} sub {1 0}} {{[ N a} sup +} ] + 8 6 . 8 7 , {size 10 _} C
1
The solvent consisted of the specified formamide concentration, plus 0.074 M
NaCl, 0.005 M Na-cacodylate and 0.2 mM Na-EDTA (pH 6.85; [Na
+
] = 0.075 M).
The formamide used in these studies was commercial Ultrapure-grade (99.9%) from USB. The
A
260
of a neat solution was <1.0, and the conductivity <100 [mu]mho/cm. Formamide has a molecular weight of 45.04 and a density of 1.3340
g/ml at 20oC, so that while the relationship between volume fraction of formamide (
V
F
/
V
total
) and concentration,
C
F
(mol/l), is linear, that with the mole fraction is non-linear, and best represented by a third-order polynomial:
X
F =
-2.048 * 10
-4
+ 5.445 * 10
-3
[%
V
F
] + 1.647 * 10
-5
[%
V
F
]
2
+ 2.909 * 10
-7
[%
V
F
]
3
2
indicating the volume of a water molecule is less than half the volume of a
formamide molecule.
Enthalpic quantities associated with conformational changes in DNA were
determined by van't Hoff analysis of transition equilibria (
17
), where derivative curves of cooperative, two-state
h
->
c
transitions of individual domains are given by:
{{{{{roman d} A} sub {2 7 0}} ( T )} over {{roman d} T}} = {{{roman DELTA} {A
sub {2 7 0}} {roman {DELTA {H sub {m , i} sup {t o t a l}} /}} {{4 {roman R} T}
sup 2}} over {{cosh sup 2} {left [ {{left ( {{roman {DELTA {H sub {m , i} sup
{t o t a l}} /}} {{2 {roman R} T T} sub roman {m sub i}}} right )} {left ( {T - {T sub {{roman m} sub roman i}}} right )}} right ]}}}
3
The precise bell-shape of curves given by this equation are defined by three parameters:
T sub roman {m sub i}
, the melting temperature, [Delta]E[Lambda]TA {H [sigma][upsi][beta] {[mu] , [iota]} [sigma][upsi][pi] {[tau] o [tau] [alpha] [lambda]}}, the total enthalpy for dissociation of the
i
th
domain at, or close to the melting temperature, and [Delta]
A
270
,the total integrated change in absorbance. If [Delta]
A
270
is normalized to unity, d
A
270
/d
T
= d[theta]/d
T
where [theta] represents the fraction of base-pairs remaining. The average enthalpy per base pair for this
domain is then given by:
{roman {{{roman DELTA} H} bar sub {m , i}}} = {roman {DELTA {H sub {m , i} sup
{t o t a l}} /}} {N sub roman i}
4
Values of [Delta]E[Lambda]TA {H [sigma][upsi][beta] {[mu] , [iota]} [sigma][upsi][pi] {[tau] o [tau] [alpha] [lambda]}}
were determined by non-linear regression to experimental curves. This was achieved by reading
digitized experimental curves obtained at high densities into vector arrays,
which were passed to a least-squares equation to minimize the weighted sum of squares,
S
, in the customary fashion:
S = {roman {sum from italic j}} {left [ {{{roman {PHI - {PHI hat}}}} over roman {sigma sub italic j}} right ]}
5
[Phi] represents values for the experimental dataset, [d
A
270
(
T
)/d
T
]
j
, and [sigma]
j
is the deviation of the
j
th
experimental point from the theoretical curve ([Phi]). Unless otherwise stated in the Results, [Delta]E[Lambda]TA {H [sigma][upsi][beta] [iota] [sigma][upsi][pi] {[tau] o [tau] [alpha] [lambda]}} were determined this way by extending the summation over the entire transition region. It is assumed that temperature values
x
j
are known exactly, and that all uncertainties occur in the
yj
values, or [d
A
270
(
T
)/d
T
]
j
.
The advantage of the van't Hoff method is its high degree of sensitivity, where
excellent results can be obtained on <50 [mu]g of precious DNA. The disadvantage is that accurate knowledge of the
states of equilibria are determined indirectly, which can be problematic in
studies of the equilibrium thermal denaturation of DNA. Denaturation is a
complex piecemeal process, with subtransitions spread over a 15-20oC range, emanating from micro-domains of a few base pairs to macro-domains of >500. All have different stabilities and many
overlap one another (
18
). Under such circumstances it is difficult to determine the states of
equilibria for individual transitions. The problem can be circumvented,
however, by constructing plasmid DNAs, such as the pN/MCS-series described above, that harbor large, discrete domains of sharply-melting oligonucleotide repetitive elements. The equilibrium
denaturation of these domains can be isolated from the background contributions
of nearby transitions by subtracting denaturation profiles of plasmids without
the insert from those that have it (
12
).
Denaturation temperatures,
T
m
, and fraction (G+C) contents (
F
GC
) of domains leading to discrete transitions were determined from high
resolution derivative curves obtained from the loss of hypochromicity at 260
and 282 nm (
cf.
Materials and Methods). Sizes of the different domains are obtained by
integration of d
A
270
/d
T
curves, as well as through correlations with statistical mechanical results of
analyses of the sequences of specimens. Illustrative 270 nm curves of plasmid
pN/MCS-12 DNA in two different solvent environments are shown in Figure
1
. The uppermost curve was obtained in the standard buffer (0.075 M Na
+
) without formamide, and shows 11 transitions, 10 of which originate from the
plasmid. The first, very sharp transition at 65.49oC is due to the melting of poly(dA[middot]dT), added to all DNA specimens as convenient marker and secondary
standard of precise solvent conditions. The melting of poly(dA[middot]dT) is particularly sensitive to extrinsic factors that affect the
stabilities of all DNAs. The half-width of this curve is only 0.067 +- 0.007oC, while
T
m
of replicate experiments vary by < +-0.02oC. This means
T
m
of plasmid transitions obtained in the identical environment as that for the
poly(dA[middot]dT) reference standard can be determined with similar precision.
Sensitivities of
T
m
to variations in base composition and ionic strengths are therefore limited
only by knowledge of the sequence of the specimen and by the accuracy with
which the buffer is prepared.
In Figure
1
b, the solvent environment has been altered by the addition of 40% formamide (
V
/
V
),
C
F
= 10.1 mol/l, although the buffer concentration is identical to that in Figure
1
a. The
T
m
of poly(dA[middot]dT) has fallen from 65.49 to 39.65oC, with no significant reduction in sharpness. Plasmid stability
has been even more affected, as can be seen from the different relationships of
the transitions for poly(dA[middot]dT) and no.1 for the repetitive insert in Figure
1
a and b.
The patterns of transitions in 0 and 40% formamide are similar, but not entirely
conserved. The split in transition no.6 in formamide, for example, represents
an obvious differential effect of the denaturant. In the absence of formamide,
three transitions are superimposed under the peak labeled no.6 in Figure
1
a (
12
). One is less affected by formamide than the other two, and so falls behind in
the pattern of Figure
1
b. Also, formamide has a greater effect on domains contributing to no.7 than to
no.8, so that the former separates farther from the latter in the lower
profile. The temperature spread of all transitions has also increased, from 15
to 17oC, suggesting a possible dependence on base composition, although no
decrease could be detected in cooperativity of melting. The half-width of the sharp melting curve of poly(dA[middot]dT) remained invariant (0.067 +- 0.005oC) over the full 0-40% range of formamide concentrations.
Figure
2
shows the dependence of the
T
m
for poly(dA[middot]dT), and of plasmid transitions no.1 and no.10 on formamide
concentration,
C
F
. Curves are linear with average standard errors of only +-0.15oC; that fall to only +-0.05oC if
T
m
s are corrected for deviations of
{italic T} sub m sup {p o l y ( d {italic A} . d {roman T} )}
from the linear function describing the behavior of that polymer. A plot of the
slopes of results such as shown in Figure
2
against the fractional (G+C) composition of domains in the plasmid is shown in
Figure
3
, together with results for several synthetic polynucleotide duplexes. The
diameters of filled circles in Figure
3
are slightly larger than estimated experimental errors for these data. Results
for plasmid domains, represented by the filled points, approximate an
underlying linear dependence on (G+C) content, with a slope of 0.453oC/l/mol formamide.
Results for synthetic polynucleotide duplexes examined in this study deviate
more widely from the expected base composition dependence. This reflects the
amplification of base sequence-dependent effects that can be achieved with biased synthetic homo- and copolymers. The slope result for poly(dA[middot]dT), -2.56 +- 0.05oC/mol/l, is 11% smaller than
anticipated from the extrapolated behavior of mixed sequence domains (-2.88); whereas slopes for alternating copolymers poly(dA-T[middot]dA-T) (-2.98) and poly(dA-C[middot]dT-G) (-2.72) are slightly larger than expected.
Poly(dA[middot]dT) is clearly less affected by formamide than either poly(dA-T[middot]dA-T) or poly(dA-C[middot]dT-G). Poly(rA[middot]dT), a hybrid duplex that adopts an A-structure, reacts differently to the
presence of formamide than either poly(dA[middot]dT) or poly(dA-T[middot]dA-T), which adopt similar B'- and B-structures, respectively. Since effects that
formamide might have on the denatured coil states of these three (A[middot]T)-containing specimens is presumably the same, these results
indicate (i) that differences between the effects of formamide on stabilities
of both quasi-random and biased sequences are primarily helix conformation-dependent, and (ii) the magnitude of these differences depend in
some way on the specific conformation of the helix, perhaps through differences
in conformation-dependent hydration patterns.
The reduced effects of formamide on poly(dA[middot]dT) are not restricted to the long (A[middot]T)
[infinity]
-tracts of the homopolymer, but extend to smaller (A[middot]T)-tracts that exceed some critical length as well. The very
short (A[middot]T)
3
-tract of the repeating insert sequence in pN/MCS-12: [AAGTTGAAC(A)
3
T]
16
AAGTTG, seems to react normally to the denaturing effects of formamide. The
transition for this insert has an observed slope of -2.74 while a slope of -2.75 is expected for a domain of 0.23 (G+C)-content. However, the slope for the (A[middot]T)
6
tract-containing repeat domain of pN/MCS-22: [AAGTTGAAC(A)
6
T]
12
AAGTTG, is only -2.73, well off the line where a value of -2.79 is expected. The difference of this insert from expected is
only 32% of the observed-expected difference seen for poly(dA[middot]dT), however the (A[middot]T)
6
-tract represents only 38% of the repeat domain, and therefore should be
affected by about this amount if the anomalously low reactivity of (A[middot]T)-tracts to formamide is limited to just tract regions.
The reduced reactivity of (A[middot]T)
6
-tracts does not depend on the position and physical environment of the
domain during melting, since the same pN/MCS-22 plasmid linearized immediately adjacent to the repeat domain by
Kpn
I has almost the same slope value, -2.71, as it does when linearized at a remote site by
Eco
RV. The anomalous reactivity to formamide is clearly attributable to (A[middot]T)-tracts at least six pairs in length.
Transition enthalpies in different formamide concentrations were determined by
analysis of two-state
helix
<->
coil
equilibria associated with the repetitive insert domain. This domain is shown
to melt in two-state fashion by several criteria when the plasmid is linearized by
Eco
RV, forcing the domain to melt as a closed loop. The experimental transition for
this domain, labeled no.1 in Figure
1
a, is shown in magnified form in Figure
4
, represented by the heavy, somewhat noisy solid line. This domain has a (G+C)
composition of only 0.23, so that its transition is almost entirely isolated
from transitions at slightly higher temperatures emanating from the parent
plasmid. Nevertheless, remnants of overlapping transitions that might affect
the analysis were eliminated by subtracting a curve of the plasmid without the
insert from the curve in Figure
1
a for the plasmid with the insert, and this is the curve shown in Figure
4
. The overall transition enthalpy, [Delta]E[Lambda]TA {H [sigma][upsi][beta] {[mu] [sigma][upsi][beta] 1} [sigma][upsi][pi] {[tau] o [tau] [alpha] [lambda]}}, was then determined, as described in Materials and Methods, by least-squares fit of equation
3
for the two-state, van't Hoff dependence of the domain equilibrium on temperature to
the experimental curve. The fitted curve is given by the smooth line in Figure
4
, punctuated by diamond-shaped symbols every tenth degree. The residuals, plotted in the lower
half of this figure, are small and totally without biases that would suggest
the two-state expression
3
might be inappropriate.
The variation of {DELTA H} bar sub {m sub 1} with
C
F
for the repetitive sequence domain is shown in Figure
5
. {DELTA H} bar sub {m sub 1} drops from 8.1 to ~7.3 kcal between 0-1.0 M formamide/l, a change of -0.8 kcal/mol-bp per 1 M formamide/l. Thereafter the enthalpy drops
only 0.05 kcal/mol-bp. Since {italic T} sub {m sub 1} drops monotonically with formamide concentration, the initial drop at low
C
F
is almost perfectly compensated by a similar drop in {DELTA S} bar sub {m sub 1}, as shown in Figure
6
, where it is assumed that {[rho]o[mu][alpha][nu] {{{[Delta]E[Lambda]TA [Sigma]} [beta][alpha][rho] [sigma][upsi][beta] {[mu] [sigma][upsi][beta] 1}} = {{[Delta]E[Lambda]TA H} [beta][alpha][rho] [sigma][upsi][beta] {[mu] [sigma][upsi][beta] 1}} /}} {T [sigma][upsi][beta] [rho]o[mu][alpha][nu] {[mu] [sigma][upsi][beta] 1}}. The thermodynamic effects associated with DNA melting can be apportioned to chemical and solvation effects, where the origin of the linear compensation is attributable to perturbations of the latter by formamide. The initial drop therefore suggests a small enthalpic loss associated with the exchange of formamide for bound water and possibly counterion to the helical state. The more gradual drop in {DELTA H} bar sub {m sub 1}
above 1 M formamide corresponds to an apparent heat capacity of only +18
cal/mol[middot]deg that probably reflects a small decrease in residual single-strand stacking enthalpy with increasing
T
m
. Although {DELTA H} bar sub m decreases over the wide range of
C
F
examined in this study, we did not find a corresponding decrease in the
cooperativity of melting.
Results obtained in this study can be summarized as follows:
(i) Formamide lowers the melting temperature by 2.4-2.9oC per mole of formamide, depending on the (G+C) composition, helix
conformation and state of hydration. The inherent cooperativity of melting, the
strong dependence of the conformational state of base pairs on those of their
neighbors from hydrogen bonding and stacking forces, is unaffected by the
denaturant. Other results indicate that the denaturant both destabilizes the
helical state and stabilizes the coil state.
(ii) The dependence on formamide concentration of the average {DELTA H} bar sub m (per bp) for the repetitive insert domain of the plasmid is biphasic. Between 0
and 1.0 M formamide, the enthalpy decreases from 8.1 to 7.3 kcal/mol-bp, but thereafter the decrease is at 1/15
th
the rate; with compensating decreases in {DELTA S} bar sub m.
(iii) d
T
m
/d
C
F
for plasmid domains exhibit a dependence on (G+C) content approximated by the
expression,
{{{d T} sub roman m} over {{d C} sub roman f}} = 0 . 4 5 3 cdot {left ( {G plus
C} right )} minus 2 . 8 8
6
The dependence is interpretable in terms of a formamide-induced shift in the phenomenological equilibrium
K
app
for
h
<->
c
equilibria of domains that vary in base composition:
cpile {{K sub {a p p}} above {{{D N A} sub n} <-> {{D N A} sub c}}}
7
(iv) While d
T
m
/d
C
F
for individual domains generally follow expression
6
, some vary slightly (+-3%) from the line, reflecting local differences in sensitivity to the
denaturant. We interpret these local differences as arising from sequence-dependent variations in the energy of hydration at selected sites.
Measurements of such small effects is possible because of the high precision (+-0.03oC) in measured
T
m
, achieved through the addition of the sharply melting poly(dA[middot]dT) as an internal reference standard. Because results were obtained at
high resolution,
T
m
of a large number of domains of different base compositions were obtained
simultaneously from the same plasmid DNA specimen in a single melting
experiment.
(v) Deviations from expression
6
are much greater for synthetic polynucleotide duplexes, where three different
synthetic (A[middot]T)-containing duplexes exhibit very different sensitivities toward
the denaturant. Since the coil states are affected equally, the helical states
must be destabilized to different extents, reflecting differences in hydration
patterns and hydration energies for different helical conformations of
synthetic duplexes.
(vi) Poly(dA[middot]dT) exhibits a significantly lower sensitivity to the denaturing effects
of formamide than predicted from
6
or from that of poly(dAT[middot]dAT) or poly(rA[middot]dT). One conformational feature that distinguishes poly(dA[middot]dT) from the other (A[middot]T)-containing duplexes is an exceptionally
narrow minor groove, sustained by a well-ordered spine of hydration. We suggest that this tightly bound layer of
hydration is not displaced by formamide, leading to a reduced sensitivity.
(vii) The degree of reduced sensitivity is not found in tracts of only two or
three (A[middot]T)-pairs, but tracts of six (A[middot]T)-pairs exhibit the same level of reduced sensitivity
as the polymer. This indicates the unusual hydration pattern and stability
associated with these tracts at melting temperatures does not take place in the
duplex until some critical length between three and six (A[middot]T) base pairs.
Results of this study are consistent with formamide destabilizing the helical
state through displacement of loosely and uniformly bound hydrate. Formamide is
a strongly associating liquid (
23
), capable of four hydrogen bonds, the same as water. It is a strong donor and a
stronger acceptor than water (
24
). Calculations indicate that formamide-water hydrogen bonds are ~20% stronger than water-water bonds (
25
), so it seems probable that some or most of its effect as a denaturant must
involve hydrogen bonding with DNA and DNA hydrate. Displacement of hydrate
leads to destabilization of the helix, presumably because of the propensity of
formamide to form hydrogen bonded networks and oligomeric chains in aqueous
solution (
24
).
Supported in part by grants from MAES (Project No. 08402) and NIH.
REFERENCES
Return