The thermal stability of DNA fragments with tandem mismatches at a d(CXYG)[middot]d(CY
'
X
'G) site
The thermal stability of DNA fragments with tandem mismatches at a d(CXYG)[middot]d(CY ' X 'G) site
Song-Hua
Ke
and
Roger M.
Wartell
*
School of Biology, Georgia Institute of Technology,
Atlanta
, GA 30332,
USA
Received October 10, 1995;
Revised and Accepted December 20, 1995
ABSTRACT
Temperature-Gradient Gel Electrophoresis (TGGE) was employed to determine the thermal
stabilities of 28 DNA fragments, 373 bp long, with two adjacent mismatched base
pairs, and eight DNAs with Watson-Crick base pairs at the same positions. Heteroduplex DNAs containing two
adjacent mismatches were formed by melting and reannealing pairs of homologous
373 bp DNA fragments differing by two adjacent base pairs. Product DNAs were
separated based on their thermal stability by parallel and perpendicular TGGE.
The polyacrylamide gel contained 3.36 M urea and 19.2% formamide to lower the
DNA melting temperatures. The order of stability was determined in the sequence
context d(CXYG)[middot]d(CY
'
X
'
G) where X[middot]X
'
and Y[middot]Y
'
represent the mismatched or Watson-Crick base pairs. The identity of the mismatched bases and their stacking
interactions influence DNA stability. Mobility transition melting temperatures
(T
u
) of the DNAs with adjacent mismatches were 1.0-3.6
o
C (
+-
0.2
o
C) lower than the homoduplex DNA with the d(CCAG)[middot]d(CTGG) sequence. Two adjacent G[middot]A pairs, d(CGAG)[middot]d(CGAG), created a more stable DNA than DNAs with Watson-Crick A[middot]T pairs at the same sites. The d(GA)[middot]d(GA) sequence is estimated to be
0.4 (
+-30%) kcal/mol more stable in free energy than d(AA)[middot]d(TT) base pairs. This result confirms the unusual stability of the
d(GA)[middot]d(GA) sequence previously observed in DNA oligomers. All other DNAs with
adjacent mismatched base pairs were less stable than Watson-Crick homoduplex DNAs. Their relative stabilities followed an order
expected from previous results on single mismatches. Two homoduplex DNAs with
identical nearest neighbor sequences but different next-nearest neighbor sequences had a small but reproducible difference in T
u
value. This result indicates that sequence dependent next neighbor stacking
interactions influence DNA stability.
INTRODUCTION
Most investigations on the structure and stability of DNA duplexes containing
non-Watson-Crick base pairs have focused on isolated defects (
1
,
2
). An isolated mismatched base pair or extra base can occur through replication
errors, which, if not repaired, lead to mutations (
3
). Studies have examined DNA oligonucleotides and DNA fragments with isolated
mismatched base pairs or an extra base on one strand (a bulge) (
4
-
7
). Although the structural and thermodynamic properties of a mismatch or bulge
depends on the base(s) involved and their neighboring sequence, DNAs with a
mismatch or bulged base are universally less stable than duplexes with Watson-Crick base pairs at the same position.
The physical properties of adjacent mismatched base pairs in DNA are less well
characterized. Tandem mismatches are statistically unlikely to arise from
random replication errors, but they may occur in folded-back repetitive sequence DNA. The prevalence of repetitive sequence DNA in
eucaryotic genomes, and the potential for unusual DNA structures involving
mismatches suggest further investigation. Structural studies on adjacent G[middot]A mismatches, d(GA)[middot]d(GA), have been carried out both in the crystal state and in
solution (8-
11
). Unlike single G[middot]A mismatches, which destabilize an otherwise complimentary base paired
duplex, adjacent G[middot]A mismatches can stabilize a DNA duplex (
11
). The conformation and stability of the d(GA)[middot]d(GA) sequence is sensitive to the local context (
12
). Studies with DNA oligomers show that d(GA)[middot]d(GA) pairs can be more stable than Watson-Crick A[middot]T pairs (
13
). When flanked by a 5' pyrimidine and a 3' purine, the adjacent G[middot]A mismatches adopt an unusual conformation (
11
,
14
-
16
). H-bonding forms between the normally outward facing edges of each purine
pair. This conformation allows the adjacent G[middot]A pairs to fit well within a surrounding standard B-form duplex (
11
). Maskos
et al
. determined that a different pair of adjacent mismatched bases, d(GA)[middot]d(AA), can also form a conformation similar to that of adjacent G[middot]A mismatches (
17
).
The above observations led us to investigate the thermal stabilities of 28 DNA
fragments that contain different adjacent mismatched base pairs. We wished to
determine if adjacent mismatches other than d(GA)[middot]d(GA) exhibited unusual stability. Temperature gradient gel
electrophoresis (TGGE) was employed to determine the relative stabilities of
the DNA fragments using the parallel and perpendicular formats (
5
,
6
). In TGGE experiments, a DNA fragment migrates through a polyacrylamide gel
with a superimposed thermal gradient. A DNA migrates at a constant mobility
until the least stable melting domain unwinds. At this point, the partially
melted DNA shows a large decrease in mobility. Previous studies have shown that
TGGE (
5
,
6
,
18
) and related methods (
19
) can separate two otherwise identical DNAs differing by a single base pair
substitution or defect (mismatch or bulge) in their first melting domain.
MATERIALS AND METHODS
DNA preparation
The plasmid pUC8-31 (
5
,
20
) was used as a template in a polymerase chain reaction (PCR) to amplify a
variety of 373 bp fragments (Fig.
1
). The 373 bp fragments contain the
B.subtilis
ctc
promoter sequence and portions of the pUC8 plasmid. Base pair positions are
numbered relative to the
ctc
promoter's transcription startpoint. DNA oligonucleotides used as primers for
the PCR were from Operon Inc., Alameda, CA. Their sequences and designations
are shown in Figure
1
. The downstream primer DP15 was end-labeled with
32
P for some PCR amplifications using a polynucleotide kinase reaction (
21
).
Taq
DNA polymerase was obtained from Perkin Elmer Inc. or Promega Inc. Conditions
used in the PCR are described in reference
5
.
TGGE
TGGE was carried out using an apparatus that resembles a conventional vertical
gel electrophoresis unit (
5
,
6
,
22
). Temperature gradients either parallel or perpendicular to the electric field
are established with two aluminum heating plates sandwiching the glass plates.
Warm and cold fluids from two temperature controlled baths are circulated
through the two channels of each heating plate to establish specific
temperature gradients. A 6.5% polyacrylamide gel (Protogel, Nat. Diagnostics
Inc.) at 37.5:1 acrylamide:bisacrylamide was employed. It contained 3.36 M urea
and 19.2% (vol/vol) formamide in 0.5* TBE (0.045 M sodium borate, 0.045 M Tris and 1 mM EDTA, pH 8.1).
Formamide was deionized with mixed resin AG501-X8D (BioRad). In parallel TGGE, DNA fragments are loaded into rectangular
wells and electrophoresed from a low temperature region (top) to a high
temperature region (bottom). The fragments move as duplex molecules until they
reach a temperature which causes the least stable melting domain to unwind. The
mobility of the partially melted DNA fragments decreases sharply. Based on the
size of temperature gradients used, and the ability to distinguish band
separation distances of 1.5-2 mm, parallel TGGE can distinguish DNAs with melting temperature
differences of ~0.05oC. Ethidium bromide staining is used to visualize the DNA bands.
Perpendicular TGGE was used to obtain the mobility transition curves of the
DNAs' first melting domain. In these experiments a temperature gradient is
established perpendicular to the direction of mobility (e.g. Fig.
4
). The temperature range was selected to focus on the DNAs' first melting domain
(
6
). DNA is loaded into a single long lane across the top of the gel and
electrophoresis carried out for 14.5 h at 90 V. The pre-transition increase in DNA mobility reflects temperature dependent changes
in gel properties, e.g. pore size and/or viscosity. It is not observed in the
absence of a temperature gradient. Subtraction of this baseline corrects for
the influence of temperature on the mobility of the duplex. A DNA mobility
transition is characterized by a sigmoidal decrease in DNA mobility with
increasing temperature (see Fig.
4
). The upper slope of the transition represents continued melting beyond the
first melting domain. Photographs of the mobility curves were digitized using a
digitizer tablet (SummaSketch II). The temperature scale was established by
measuring the temperature at several positions across the gel at the end of
each run (
6
). The positions where the needle-like thermocouple probe had been inserted were observed as dark lines in
the photographs. Transition curves were smoothed by a locally weighted
regression analysis (
6
). The first melting domain's mobility transition temperature, T
u
, was defined as the temperature at the peak of the derivative of the first
domain's transition curve.
RESULTS
The eight homoduplex DNAs used in this study were generated using PCR as
indicated in Figure
1
. The DNA fragments differ from each other by one base pair or two adjacent base
pairs at sites designated -38 and -39. Pairs of DNAs differing at these positions were melted and
reannealed to produce 28 DNAs that contained two adjacent base pair mismatches.
All DNA fragments with or without mismatches ran with the same mobility in non-denaturing polyacrylamide gels (not shown). Previous studies showed that
the 373 bp homoduplex DNAs have three melting domains (
5
,
20
). The first melting domain consists of ~50 bp from the 5' end of the DNA fragment in Figure
1
. It encompasses the positions of the mismatched base pairs.
DISCUSSION
Ebel
et al
. (
13
) showed that tandem G[middot]A mismatches in a d(CGAG)[middot]d(CGAG) sequence is more stable than the d(CTAG)[middot]d(CTAG) sequence in a DNA oligomer. Our results verify
this observation in a long DNA fragment and show further that the d(CGAG)[middot]d(CGAG) sequence is more stable than the d(CAAG)[middot]d(CTTG) sequence. The unusual stability of tandem G[middot]A pairs requires a 5'Py-G-A-Pu3' sequence and
involves GNH
2
-AN7 and GN3-ANH
2
(edge to edge) hydrogen bonds and cross-strand stacking of the adjacent guanine and adenine bases (
11
-
16
,
27
).
The possibility that other tandem mismatches may create unusually stable
duplexes was suggested by the studies of Maskos
et al
. (
17
) on the oligonucleotide d(GCGAATAAGCG)
2
. This oligomer forms a duplex containing two copies of the mismatched
sequences d(CGAA)[middot]d(TAAG) and two 3' unpaired guanosines. NMR analysis indicates the adjacent G[middot]A and A[middot]A mismatches have an overall duplex structure
similar to the adjacent G[middot]A mismatches. Although the d(GA)[middot]d(AA) sequence was among the more stable tandem mismatches in
the DNAs we examined, it is not more stable than a pair of Watson-Crick A[middot]T base pairs. Table
2
shows that none of the other adjacent mismatches examined had a stability
greater than Watson-Crick A[middot]T base pairs. Other sequence contexts may give different results.
The relative stability of the tandem mismatches shows a trend with regard to
purine/pyrimidine content. Sequences with more purines tend to be more stable
than mismatches with pyrimidines. The six most stable tandem mismatches have at
least one of the most stable single mismatches in DNA, G[middot]T, G[middot]G or G[middot]A, and the six least stable tandem mismatches have at
least one of the least stable single mismatches, T[middot]C or C[middot]C (
4
,
5
). The range in stability among the tandem mismatches is ~3.4 kcal/mol (Table
2
). The stability of 11 tandem mismatches in a r(CXYG)[middot] r(CX'Y'G) context have been measured in RNA oligomers (
28
). Unfortunately the tandem RNA mismatches examined and the solvent employed
differ from the current work making a comparison difficult.
One of the unexpected outcomes of this study was evidence for sequence-dependent next neighbor base pair stacking interactions. The DNAs
generated with the primers UP2BM and UP16 (Fig.
1
) have identical nearest neighbor base pair sequences yet show a slight
difference in DNA stability. Since unwinding the first melting domain of these
DNAs does not involve strand dissociation, their difference in stability cannot
be explained by differences in strand dissociation parameters. The estimated
free energy difference between the two homoduplex DNAs with identical nearest
neighbor base pairs, 130 cal/mol, may be viewed as a limit to the accuracy of
evaluated nearest neighbor stacking interactions. Deviations between
experimental DNA melting curves and theoretical predictions using nearest
neighbor stacking parameters may be due to this second order effect (
24
). Such deviations may not be evident with large melting domains where averaging
of longer than nearest neighbor effects minimizes this effect. It may be more
apparent for shorter melting domains. The second neighbor effect was not
observed when two different DNAs were examined with two adjacent mismatches
with identical nearest neighbor sequences. For this situation, the nearest
neighbor stacking interaction model appears to be valid.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge support provided by a grant from the Georgia
Tech/MCG Biomedical Research Center. We also wish to thank Tom Maier for
technical assistance.
REFERENCES
1 Kennard,O. and Hunter,W.N. (1990) Q. Rev. Biophys.23, 327-379.
2 Patel,D.J., Shapiro,L. and Hare,D. (1987) In Lilly,D. and Eckstein,E. (eds) Nucleic Acids and Molecular Biology. Springer-Verlag, Berlin, Vol. 1, pp. 70-84.
3 Lewin,B. (1994) Genes V. Oxford University Press Inc., New York, NY.
18 Riesner,D., Henco,K., and Steger,G. (1991) InChrambach,A., Dunn,M.J. and Radola,B.J. (eds) Advances in Electrophoresis. VCH Pub., New York, Vol. 4, pp.171-250.
