Theoretical study of a new DNA structure: the antiparallel Hoogsteen duplex.
ABSTRACT The structure of a new form of duplex DNA, the antiparallel Hoogsteen duplex, is studied in polyd(AT) sequences by means of state-of-the-art molecular dynamics simulations in aqueous solution. The structure, which was found to be stable in all of the simulations, has many similarities with the standard Watson-Crick duplex in terms of general structure, flexibility, and molecular recognition patterns. Accurate MM-PB/SA (and MM-GB/SA) analysis shows that the new structure has an effective energy similar to that of the B-type duplex, while it is slightly disfavored by intramolecular entropic considerations. Overall, MD simulations strongly suggest that the antiparallel Hoogsteen duplex is an accessible structure for a polyd(AT) sequence, which might compete under proper experimental conditions with normal B-DNA. MD simulations also suggest that chimeras containing Watson-Crick duplex and Hoogsteen antiparallel helices might coexist in a common structure, but with the differential characteristics of both type of structures preserved.
Theoretical Study of a New DNA Structure: The Antiparallel
Elena Cubero,†,⊥Nicola G. A. Abrescia,‡Juan A. Subirana,§F. Javier Luque,*,|and
Contribution from the Molecular Modeling & Bioinformatic Unit, Institut de Recerca Biome `dica,
Parc Cientı ´fic de Barcelona, Josep Samitier 1-5, Barcelona 08028, Spain, DiVision of Structural
Biology, Wellcome Trust Centre for Human Genetics, Oxford UniVersity, Oxford OX3 7BN,
United Kingdom, Departament d’Enginyeria Quı ´mica, UniVersitat Polite ´cnica de Catalunya,
AVgda Diagonal 647, Barcelona 08028, Spain, Departament de Fisicoquı ´mica, Facultat de
Farmacia, UniVersitat de Barcelona, AVgda Diagonal 643, Barcelona 08028, Spain, and
Departament de Bioquı ´mica i Biologia Molecular, Facultat de Quı ´mica, UniVersitat de
Barcelona, Martı ´ i Franque `s 1, Barcelona 08028, Spain
Received May 2, 2003; E-mail: email@example.com
Abstract: The structure of a new form of duplex DNA, the antiparallel Hoogsteen duplex, is studied in
polyd(AT) sequences by means of state-of-the-art molecular dynamics simulations in aqueous solution.
The structure, which was found to be stable in all of the simulations, has many similarities with the standard
Watson-Crick duplex in terms of general structure, flexibility, and molecular recognition patterns. Accurate
MM-PB/SA (and MM-GB/SA) analysis shows that the new structure has an effective energy similar to that
of the B-type duplex, while it is slightly disfavored by intramolecular entropic considerations. Overall, MD
simulations strongly suggest that the antiparallel Hoogsteen duplex is an accessible structure for a poly-
d(AT) sequence, which might compete under proper experimental conditions with normal B-DNA. MD
simulations also suggest that chimeras containing Watson-Crick duplex and Hoogsteen antiparallel helices
might coexist in a common structure, but with the differential characteristics of both type of structures
Under physiological conditions, DNA is mostly found as a
right-handed double helix that shows a conformation, named
B-DNA, not far from that suggested by Watson and Crick.1
However, 50 years after the discovery of the double helix, it is
clear that DNA is very polymorphic and that its equilibrium
conformation can change depending on sequence and environ-
ment.2-4Repetitive sequences are especially rich in terms of
accessible helical structures (for reviews, see refs 2, 4, and 7).
For example, poly(GC), poly(G), and poly(GA) sequences are
known to form easily left-handed Z-DNA,5four-stranded helical
structures,6and triple helices,7respectively. Other examples are
sequences containing poly(C) or poly(CT) tracks, which can
generate the i form of DNA,8,9and sequences rich in polyd(A)
tracks, which can yield parallel duplexes showing reverse
Watson-Crick pairings, parallel duplexes with Hoogsteen
pairings,10or parallel duplexes involving d(A‚A)+dimers.2The
poly(AT) sequences are probably those exhibiting the widest
range of accessible structures.11-21Thus, crystal structures of
d(AT)2show a distorted B-type conformation.11,12Other crystal
†Institut de Recerca Biome `dica.
§Universitat Polite ´cnica de Catalunya.
|Departament de Fisicoquı ´mica, Universitat de Barcelona.
⊥Departament de Bioquı ´mica i Biologia Molecular, Universitat de
(1) Watson, J. D.; Crick, F. H. C. Nature 1953, 171, 737.
(2) (a) Saenger, W. Principles of Nucleic Acid Structure; Springer-Verlag: New
York, 1984. (b) Ghosh, A.; Manju, B. Acta Crystallogr. 2003, D59, 620.
(3) Bloomfield, V. A., Crothers, D. N., Tinoco, I., Eds. Nucleic Acids:
Structures, Properties and Functions; University Science Books: Sausalito,
(4) Blackburn, G. M., Gait, M. J., Eds. Nucleic Acids in Chemistry and Biology;
IRL Press: Oxford, 1990.
(5) Wang, A. H.-J.; Quiqkey, G. J.; Kolpak, F. J.; Crawford, J. L.; van Boom,
J. H.; van der Marel, G.; Rich, A. Nature 1979, 282, 680.
(6) Laughlan, G.; Murchie, A. I. H.; Norman, D. G.; Moore, M. H.; Moody,
P. C. E.; Lilley, D. M.; Luisi, B. Science 1994, 265, 520.
(7) Robles, J.; Grandas, A.; Pedroso, E.; Luque, F. J.; Eritja, R.; Orozco, M.
Curr. Org. Chem. 2002, 6, 1333 and references therein.
(8) Gehring, K.; Leroy, J. L.; Gue ´ron, M. Nature 1993, 363, 561.
(9) Gray, D. M.; Vaughn, M.; Ratliff, R. L.; Hayes, F. N. Nucleic Acids Res.
1980, 8, 3695.
(10) (a) Cubero, E.; Luque, F. J.; Orozco, M. J. Am. Chem. Soc. 2001, 123,
12018. (b) Cubero, E.; Avin ˜o ´, A.; de la Torre, B. G.; Frieden, M.; Eritja,
R.; Luque, F. J.; Gonza ´lez, C.; Orozco, M. J. Am. Chem. Soc. 2002, 124,
(11) Viswamitra, M. A.; Kennard, P.; Jones, P. G.; Sheldrick, G. M.; Salisbury,
S.; Falvello, L.; Shakked, Z. Nature 1978, 273, 687.
(12) Viswamitra, M. A.; Shakked, Z.; Sheldrick, J. G. M.; Salisbury, S. A.;
Kennard, O. Biopolymers 1982, 21, 513.
(13) Guzikevich-Guerstein, G.; Shakked, Z. Nat. Struct. Biol. 1996, 3, 32.
(14) Radwan, M. M.; Wilson, H. R. Int. J. Biol. Macromol. 1982, 4, 145.
(15) Davis, D. R.; Baldwin, R. L. J. Mol. Biol. 1963, 6, 251.
(16) Brahms, S.; Brahms, J.; Van Holde, K. E. Proc. Natl. Acad. Sci. U.S.A.
1976, 73, 3453.
(17) (a) Abrescia, N. G. A.; Thompson, A.; Dinh, T. H.; Subirana, J. A. Proc.
Natl. Acad. Sci. U.S.A. 2002, 99, 2806. (b) Abrescia, N. G. A.; Subirana,
J. A. Acta Crystallogr. 2002, D58, 2205.
(18) Aishima, J.; Gitti, R. K.; Noah, J. E.; Gan, H. H.; Schlick, T.; Wolberger,
C. Nucleic Acids Res. 2002, 30, 5244.
(19) Lefe `vre, J. F.; Lane, A. N.; Jardetzky, O. Biochemistry 1988, 27, 1086.
(20) Schmitz, U.; Sethson, I.; Egan, W. M.; James, T. L. J. Mol. Biol. 1992,
Published on Web 11/04/2003
10.1021/ja035918f CCC: $25.00 © 2003 American Chemical Society
J. AM. CHEM. SOC. 2003, 125, 14603-14612 9 14603
structures even show a larger distortion, suggesting the existence
of a very special type of B DNA named TA-DNA.13Early fiber
diffraction data on polyd(AT) sequences suggested left-handed
models,14but this has not been confirmed by higher resolution
techniques. On the other hand, polyd(AT) fibers generated under
low hydration conditions show a D form of DNA,15while at
low temperature the C form is obtained.16Very recently,
Subirana and co-workers17have reported the high-resolution
structure of the d(ATATAT) duplex, which corresponds to a
new structural motif of DNA consisting of an antiparallel right-
handed double helix with adenosines in the syn conformation
and unexpected Hoogsteen d(A‚T) pairings. To our knowledge,
the existence of this form in aqueous solution has not been
confirmed by NMR experiments.
The possibility of different H-bond patterns for the A‚T pair
(see Figure 1) has been known since the 1960s.22Accurate
theoretical calculations23showed that in fact the Watson-Crick
pairing is not the best recognition mode for the A‚T pair and
that all of the four recognition patterns shown in Figure 1 are
accessible, at least in the gas phase. Molecular dynamics
simulations have also shown that stable helices can be built for
both duplexes10aand triplexes24using Hoogsteen A‚T pairings.
Experimentally, the Hoogsteen recognition mode is found in
different structures of DNA and RNA,25including the parallel
triplexes,7,26where it stabilizes both d(A-T‚T) and d(G-C‚C)
triads. Very interestingly, Hoogsteen pairs are common in
complexes between duplex DNA and drugs or proteins.18,27-29
In conjunction with the biological role of d(AT)nregions,27,30
these findings strongly suggest an important role for Hoogsteen
pairings in the modulation of gene expression.18,27In summary,
recent theoretical and experimental information points out that
the apparently exotic Hoogsteen interaction might be common
and have a large biological significance.
In this paper, we present a wide and systematic molecular
dynamics (MD) study of the poly(AT) duplex in aqueous
solution at low ionic strength. Both the normal Watson-Crick
(B form) and the antiparallel Hoogsteen (apH) structures17are
studied using large MD trajectories for 4-, 6-, 8-, 10-, 12-, 14-,
and 16-mer duplexes (more than 40 ns of trajectories are
presented here). The structure, dynamics, and recognition
properties of the apH helix in water are determined and
compared to those of the standard B-DNA duplexes. The relative
stability of apH and B-type duplexes is discussed using
information derived from the MD simulations. Finally, trajec-
tories of chimeras containing fragments of apH and B-DNAs
are generated to analyze the impact of Hoogsteen pairings in
the structure of a long canonical piece of DNA. The results
obtained here complement previous X-ray data by Abrescia et
al.17and provide a complete picture of a new and intriguing
structure of the DNA duplex.
To analyze with good statistical accuracy10athe structure and stability
of apH and B duplexes for poly(AT) sequences, we built starting
structures for complementary duplexes d(AT)n/2 for n ) 4, 6, 8, 10,
12, 14, and 16 using standard Arnott’s parameters for B-DNA31and
the crystal structure data for the apH-DNA.17The 14 different structures
were neutralized by adding a suitable number of Na+ions (10a) and
immersing them in rectangular boxes containing between 1000 and 3500
water molecules. All of the systems were optimized, thermalized (298
K), and equilibrated using our standard multistage protocol.24,32The
equilibrated structures were then subjected to 2 ns (4-, 6-, 8-, 14-, and
16-mer) and 5 ns (10- and 12-mer) of MD simulation at constant
temperature (298 K) and pressure (1 atm) with standard relaxation times
of 0.2 ps. Periodic boundary conditions were used to simulate a diluted
environment, and the Particle Mesh Ewald method was used to account
for long-range electrostatic effects.33Ewald tolerance of 5 × 10-6was
used in conjunction with a grid spacing of 1 Å and a four-order
interpolation scheme. SHAKE34was used to constrain all of the bonds
at their equilibrium positions, which allowed us to use a 2 fs time step
for integration of Newton equations. AMBER-9835and TIP3P36force-
fields were used to describe DNA and water. All of the MD simulations
were performed using the AMBER6.0 suite of programs.37
(21) McAtter, K.; Ellis, P. D.; Kennedy, M. A. Nucleic Acids Res. 1995, 23,
(22) Hoogsteen, K. Acta Crystallogr. 1959, 12, 822.
(23) Sponer, J.; Leszczynski, J.; Hobza, P. J. Phys. Chem. 1996, 100, 1969.
(24) Shields, G. C.; Laughton, C. A.; Orozco, M. J. Am. Chem. Soc. 1997, 119,
(25) (a) Voet, D.; Rich, A. Prog. Nucleic Acid Res. Mol. Biol. 1970, 10, 183.
(b) Leontis, N. B.; Westhof, E. Q. ReV. Biophys. 1998, 31, 399. (c) Nolten,
G. M. J.; Sijtema, N. M.; Otto, C. Biochemistry 1997, 36, 13241.
(26) (a) Pauling, L.; Corey, R. B. Proc. Natl. Acad. Sci. U.S.A. 1953, 39, 84.
(b) Felsenfeld, G.; Davis, D. R.; Rich, A. J. Am. Chem. Soc. 1957, 79,
(27) Patikoglou, G. A.; Kim, J. L.; Sun, L.; Yang, S.-H.; Kodadek, T.; Burley,
S. K. Genes DeV. 1999, 13, 3217.
(28) Rice, P. A.; Yang, S.; Mizuuchi, K.; Nash, H. A. Cell 1996, 87, 1295.
(29) (a) Wang, A. H.; Ughetto, G.; Quigley, G. J.; Hakoshima, T.; van der Marel,
G. A.; van Boom, J. H.; Rich, A. Science 1984, 225, 1115. (b) Gilbert, D.
E.; van der Marel, G. A.; van Boom, J. H.; Feigon, J. Proc. Natl. Acad.
Sci. U.S.A. 1989, 86, 3006.
(30) (a) Moreau, J.; Maschat, M. F.; Kejzlarova-Lepesant, J.; Lepesant, J.-A.;
Scherrer, K. Nature 1982, 295, 260. (b) Oosumi, T.; Garlick, B.; Belknap,
W. R. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 8886.
(31) Arnott, S.; Hukins, D. W. L. Biochem. Biophys. Res. Commun. 1972, 47,
(32) (a) Soliva, R.; Laughton, C. A.; Luque, F. J.; Orozco, M. J. Am. Chem.
Soc. 1998, 120, 11226. (b) Shields, G. C.; Laughton, C. A.; Orozco, M. J.
Am. Chem. Soc. 1998, 120, 5895.
(33) Darden, T. A.; York, D. M.; Pedersen, L. G. J. Chem. Phys. 1993, 98,
(34) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977,
(35) (a) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.;
Fergurson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman,
P. A. J. Am. Chem. Soc. 1995, 117, 5179. (b) Cheatham, T. E.; Cieplak,
P.; Kollman, P. A. J. Biomol. Struct. Dyn. 1999, 16, 845.
(36) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein,
M. L. J. Chem. Phys. 1983, 79, 926.
(37) Case, D. A.; Pearlman, D. A.; Caldwell, J. W.; Cheatham, T. E., III; Ross,
W. S.; Simmerling, C. L.; Darden, T. L.; Marz, K. M.; Stanton, R. V.;
Cheng, A. L.; Vincent, J. J.; Crowley, M.; Tsui, V.; Radmer, R. J.; Duan,
Y.; Pitera, J.; Massova, I.; Seibel, G. L.; Singh, U. C.; Weiner, P. K.;
Kollman, P. A. AMBER6; University of California: San Francisco, CA,
Figure 1. Schematic representation of the possible recognition patterns
between neutral adenine and thymine.
A R T I C L E SCubero et al.
14604 J. AM. CHEM. SOC.9VOL. 125, NO. 47, 2003
The MD simulations of apH and B duplexes were analyzed to
evaluate their relative stability. For this purpose, the free energy of
every structure was computed as shown in eq 1.38MD averages (〈 〉)
for each contribution in eq 1 were obtained after dropping the first
nanosecond of every trajectory. Eintrawas computed using the standard
AMBER-98 force-field.35The solvation contribution (Gsol) was deter-
mined using two different approaches (for a discussion of the two
techniques, see ref 39): (i) finite difference Poisson-Boltzmann (PB)
calculations as implemented in the MEAD program40for initial and
final grid spacings of 1 and 0.4 Å, and (ii) the Generalized-Born (GB)
method41as implemented in AMBER6.0. A dielectric constant of 80
was used to represent water, while dielectric constants of 1 or 2 were
used to represent the solute. Finally, the contribution to the free energy
of the system due to the intramolecular DNA entropy was determined
by using Schliter’s method42as described by Harris et al.43We should
note that test calculations using the Andricioae and Karplus method44
did not lead to any significant difference with respect to Schliter-type
calculations and were then not included. To avoid artifacts in the
averaging of the different terms, the end-base pairs were removed from
Calculations noted in eq 1 were performed for the different apH
and B-type structures. Taking advantage of the linear relationship
between the length of the oligonucleotide and the free energy (see ref
10a and below), we can then derive estimates of the relative stability
of apH and B structures with very good statistical accuracy (see Results
A description of the interaction properties of apH (and B) structures
was obtained from cMIP calculations for an O+probe molecule.24,32,45
Solvation around the structures was represented by integrating the water
population along the trajectory as explained elsewhere.24,32The analysis
of molecular flexibility was performed using the principal component
analysis (PCA) method as described by Sherer et al.46Principal
components were obtained by diagonalization of the covariance matrix
obtained after recentering all of the collected snapshots. The result of
this procedure is a set of eigenvalues and the corresponding eigenvec-
tors. Eigenvalues can be manipulated to obtain vibrational frequencies
and provide useful information on the magnitude of the flexibility of
DNA along their essential movements. The eigenvectors projected into
the Cartesian space illustrate the nature of the essential movements of
the DNA molecule. As shown previously,47we can compare the
similarity between two unitary eigenvectors using the dot product,
obtaining then a quantitative measure of the similarity between two
essential movements. The strategy can be expanded (see eq 2 and ref
47) to quantitatively compute the similarity between a reduced set of
eigenvectors, which are known to explain more than a certain amount
of variance in the trajectory.47Note that due to the orthogonality
between principal components, two identical trajectories will yield a
similarity index γ of 1, while due to the limited number of eigenvectors
considered two unrelated trajectories will have γ equal to 0.
where A and B stand for two different trajectories of equal length (for
the same or different structures), ν stands for a unit eigenvector, and n
is taken as 10, a number of modes which accounts for around 80% of
the variance for the 10- and 12-mer oligonucleotides considered here.
To reduce errors, only backbone atoms (up to C1′) were considered in
comparison between B and apH helices.
A relative similarity index (κ), which accounts for the uncertainties
intrinsic to any time-limited MD trajectory, can be obtained by
normalizing the absolute index γABby the self-similarities obtained in
the A and B trajectories, γAATand γBBT(see eq 3). Self-similarities are
easily obtained by comparing the first and last halves of the same
trajectory. Both γ and κ are 1 for two identical trajectories and 0 when
they are orthogonal. As noted in ref 47, this index runs also from 1
(identical) to 0 (unrelated) and has the advantage of being less
simulation-sensitive than the absolute similarity index (γ).
Geometrical analysis was performed by using the PTRAJ module
of the AMBER6.0 program, as well as by using in-house programs.
Helical analysis was carried out using Olson’s X3DNA program48and
Curves5.3.49Base pairs at the ends of the duplex were removed in all
of the cases. All calculations were performed at the Centre de
Supercomputacio ´ de Catalunya (CESCA) and in workstations in our
Results and Discussion
Stability of Trajectories of polyd(AT)n in B and apH
Forms. All of the trajectories of B and apH structures show
stable helical conformations (see Figure 2), as noted in the small
root-mean square deviations (rmsd) with respect to the MD-
averaged structures (see Table 1). In general, lower rmsd values
are found for the apH duplexes than for the B-like ones,
suggesting that the former sample a narrower region of the
configurational space, probably due to a greater rigidity of the
apH structure (see below). Comparison of 2 and 5 ns trajectories
demonstrates that the trajectories converge well in the 2 ns time
scale and that apparently no relevant transitions are expected
for longer simulation times. The all-atom rmsd values of the
apH trajectories with respect to the crystal structure (PDB entry
1gqu17) are small (<2 Å in all of the cases), showing that MD
simulations in aqueous solution sample regions of the configu-
rational space very close to the crystal structure. This agreement
supports the suitability of the MD simulation for the study of
the apH duplex, and the ability of X-ray data to represent a
reasonable conformation of the polyd(AT)n duplex in dilute
The B-like trajectories are closer to the B than to the A form,
but the rmsd values with respect to the canonical B form are
larger than those typically found for simulations of B-DNA.
For example, for the duplex d(CGCGAATTCGCG), the rmsd
with respect to the canonical B form found in 10-20 ns
(38) Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn, B.; Huo, S.; Chong, L.;
Lee, M.; Lee, T.; Duan, Y.; Wang, W.; Donini, O.; Cieplak, P.; Srinivasan,
J.; Case, D.; Cheatham, T. E. Acc. Chem. Res. 2000, 33, 889.
(39) Orozco, M.; Luque, F. J. Chem. ReV. 2000, 100, 4187.
(40) (a) Bashford, D.; Gerwert, K. J. Mol. Biol. 1992, 224, 473. (b) Bashford,
D. In Scientific Computing in Object-Oriented Parallel EnVironments;
Reynders, J. V. M., Tholburn, M., Eds.; Springer: Berlin, 1997; pp 233-
(41) Still, W. C.; Tempcyck, A.; Hawley, R. C.; Hendrickson. T. J. Am. Chem.
Soc. 1990, 112, 6127.
(42) Schliter, J. Chem. Phys. Lett. 1993, 215, 617.
(43) Harris, S.; Gavathiotis, E.; Searle, M. S.; Orozco, M.; Laughton, C. A. J.
Am. Chem. Soc. 2001, 123, 12658.
(44) Andricioaei, I.; Karplus, M. J. Chem. Phys. 2001, 115, 6289.
(45) Gelpı ´, J. L.; Kalko, S.; de la Cruz, X.; Barril, X.; Cirera, J.; Luque, F. J.;
Orozco, M. Proteins 2001, 45, 428.
(46) Sherer, E.; Harris, S. A.; Soliva, R.; Orozco, M.; Laughton, C. A. J. Am.
Chem. Soc. 1999, 121, 5981.
(47) (a) Rueda, M.; Kalko, S. G.; Luque, F. J.; Orozco, M. J. Am. Chem. Soc.
2003, 125, 8007. (b) Orozco, M.; Noy, A.; Pe ´rez, A.; Luque, F. J. Chem.
Soc. ReV. 2003, 32.
(48) Lu, X,-J.; Olson, W. K. X3DNA Program, Rutgers University, 2001.
(49) Lavery, R.; Sklenar, J. J. Biomol. Struct. Dyn. 1988, 6, 63.
G ≈ 〈Eintra〉 + 〈Gsol〉 - T∆Sintra) 〈Eeffec〉 - TSintra
Theoretical Study of a NewDNA Structure A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 125, NO. 47, 2003 14605
trajectories in water,50using the same force-field and simulation
protocol, is around 2.3 Å (1.9 Å if the crystal structure is used
as reference), that is, around 1 Å lower than the rmsd found
here for oligonucleotides of the same length. As previously noted
by experimentalists (see Introduction), this deviation suggests
that polyd(AT)n duplexes do not adopt a pure-canonical B
structure in solution. However, for the sake of clarity, we will
continue denoting as “B form” those structures obtained in the
B trajectories of polyd(AT)nduplexes.
The hydrogen-bond (H-bond) patterns defining the Watson-
Crick and Hoogsteen pairings are well preserved for B and apH
trajectories (see Table 2). However, the total percentage of
H-bonds in the apH trajectories is typically larger than that found
for the B trajectories. Very interestingly, the general stability
(50) Rueda, M.; Cubero, E.; Laughton, C. A.; Luque, F. J.; Orozco, M., to be
Figure 2. Representation of the MD-averaged structure of the d(AT)n/2for n ) 10 in the B (left) and antiparallel Hoogsteen conformations (right). See text
Table 1. All-Atom Root-Mean-Square Deviations (rmsd; Å) of
B-DNA and apH-DNA Trajectories with Respect to Different
Reference Structures (Standard Deviations (SD) in the Averages
Are Also Displayed)
1.0 ( 0.2
1.1 ( 0.2
1.4 ( 0.3
1.5 ( 0.2
1.7 ( 0.4
1.7 ( 0.3
2.1 ( 0.6
1.4 ( 0.2
2.1 ( 0.3
2.5 ( 0.5
2.9 ( 0.5
3.3 ( 0.7
3.4 ( 0.6
3.9 ( 0.9
1.9 ( 0.2
2.3 ( 0.2
3.4 ( 0.6
3.7 ( 0.5
4.4 ( 0.7
5.0 ( 0.6
5.0 ( 0.8
1.5 ( 0.6
1.0 ( 0.2
0.8 ( 0.2
1.3 ( 0.3
1.3 ( 0.3
1.5 ( 0.4
1.3 ( 0.3
1.5 ( 0.2
1.0 ( 0.2
1.1 ( 0.2
1.9 ( 0.5
1.5 ( 0.3
1.5 ( 0.3
1.6 ( 0.3
a5 ns trajectories, the rest are only 2 ns long.bMD-averaged structures
obtained using the last 1 or 4 ns of trajectories.cArnott’s structures (see
ref 31).dStructures built by joining two 5-mer structures in PDB entry
1gpu followed by restricted minimization.
Table 2. Occurrence (in % with Respect to the Maximum Number
of Hydrogen-Bond Interactions) of Canonical Watson-Crick (WC)
and Hoogsteen (H) Hydrogen Bonds for the Different Trajectories
structure WC H-bondsstructureHH-bonds
A R T I C L E SCubero et al.
14606 J. AM. CHEM. SOC.9VOL. 125, NO. 47, 2003
of Watson-Crick and Hoogsteen H-bonds does not preclude
the existence of reversible, partial breathing movements, like
those shown in Figure 3. These movements occur in the
nanosecond time scale and do not lead to a complete opening
of the bases, which always retain at least one nonstandard
H-bond (see Figure 3). They are, in general, related to
movements of thymine toward the major groove and are more
common in the B trajectories.
Description of the apH Structure, Interaction Properties,
and Dynamics. The analysis of the 5 ns trajectories of d(AT)5
and d(AT)6apH helices provides a complete structural picture
of medium-sized apH helices in aqueous solution, complement-
ing the X-ray information derived by Abrescia et al.17The
general shape of the apH helix is similar to that of a normal
B-type duplex (see Figure 2), a quite surprising finding
considering the dramatic differences in terms of nucleotide
conformation and H-bond recognition pattern between the two
helices. The periodicity of the apH helix is similar to that of
B-DNA (see Table 3 and ref 51), the rise is only slightly larger
(around 0.1 Å), the inclination is small and very similar to that
found in B-trajectories, and the average phase angles suggest
south and east-south puckerings, as is usual in crystal structures
of B-DNA.52The minor groove in apH trajectories is around 2
Å narrower than that found in B-like trajectories, while the major
groove of the apH structures is around 3 Å wider (see also
Figure 2). There are differences between the two types of
structures in the shortest C1′-C1′ distances (around 2 Å shorter
for apH helices) and in the ? angles, which arise from the
existence of the syn-2′-deoxyadenosines in the apH helices (the
average ? values for apH 2′-deoxyadenosines are around 50°,
in contrast to standard anti values of around -110°). Finally, it
is worth noting the similarity between the X-ray structure of
the apH helix and that found in MD simulations in solution
(see Tables 1 and 3). In fact, the only noticeable discrepancy is
found for the twist, which is 3° smaller in MD simulations,
reproducing a well-known tendency of the AMBER force-field
(see Table 3 and refs 47b and 53).
(51) The existence of syn adenines makes the helical analysis of apH helices
very difficult, even when the very flexible X3DNA program is used. The
only parameters that can be safely derived from helical analysis are those
displayed in Table 3.
(52) Drew, H. R.; Wing, R. M.; Tanako, T.; Broka, C.; Tanaka, S.; Itakura, K.;
Dickerson, R. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 2179.
(53) Cheatham, T. E.; Young, M. A. Biopolymers 2001, 56, 232.
Figure 3. Evolution of key H-bond distances (in Å) in B (10-mer) and apH (12-mer) trajectories. The existence of partial breathing movements is seen in
a temporary loss of canonical Watson-Crick or Hoogsteen hydrogen bonds, and the simultaneous gain of one additional noncanonical hydrogen bond. The
color code is as follows. Top: black, H6(A7)-O4(T14); red, N1(A7)-H3(T14); green, H2(A7)-O2(T14); and blue, H6(A7)-O2(T14). Bottom: black, O4-
(T6)-H6(A19); red, H3(T6)-N7(A19); green, O2(T6)-H8(A19); and blue, O2(T6)-H6(A19).
Table 3. Key Helical Parameters for the MD-Averaged B and apH
Helices (for 10- and 12-mer Oligonucleotides)a
aThe helical parameters of a standard B-DNA and of the structure
generated from crystal data in ref 17 are also displayed. Distances are in
Å, and angles are in deg.bMinimum C1′-C1′ distance.cCrystal structure
from 1gpu (see text).dDetermined from the shortest P-P distances across
Theoretical Study of a NewDNA StructureA R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 125, NO. 47, 2003 14607
Despite the similar shape of B and apH helices in solution,
it remains to be determined whether they exhibit similar
interaction profiles. MIP maps (Figure 4) show that, as found
for the B-DNA, the minor groove is the region with a stronger
ability to interact with small cationic probes (smaller favorable
regions are found in the major groove) in the apH structure.
The shapes of MIP (isocontours of -5 kcal/mol) maps in apH
and B helices are quite similar (see Figure 4). However, while
the MIP isocontours are continuous for the B helix, they are
not for the apH structures. This indicates that the minor groove
in apH helices is slightly less charged than that of B helices
(i.e., slightly less polar). However, the differences found in the
MIP maps are less relevant than those expected a priori from
the dramatic changes in the distribution of polar groups at the
bottom of the grooves in B and apH helices (see Figure 1).
Both B and apH helices are very well solvated in our
simulations. On average (for all of the sequences and excluding
the terminal base pairs), there are 26 ( 7 (B) and 27 ( 12
(apH) water molecules per step in direct contact (distance less
than 3.5 Å) to any heteroatom of DNA, indicating a very similar
hydration. For the B trajectories (Figure 4), small amounts of
water are located in the major groove, and most of the regions
of large water concentration (>2.5 g/cm3) are in the minor
groove, tracing clearly Dickerson’s spine of hydration.52For
the apH structure, the spine of hydration along the minor groove
is also clearly defined, with waters located in the vicinities of
O2 (thymines). Interestingly, a secondary spine of hydration
(see details in Figure 5) is obtained along the major groove,
with regions of intense hydration around N6 and N1 of adenines
and O4 of thymines. A very dense hydration region is always
located around N3 of adenines, bridging this atom to the
phosphate group. The preferred regions of hydration found in
MD simulation perfectly agree with highly ordered crystal
waters located by Abrescia et al.17in the major groove.
Interestingly, the minor groove which is occupied by nucleo-
bases in the crystal structure17is occupied by water in solution
(see Figure 4). All of these regions of large hydration correspond
also to regions where waters with long residence times (up to
200 ps) are located (see Figure 5).
PCA provides a general view of the essential motions of the
duplex in B and apH conformations. The first frequencies are
very low (around 10-20 cm-1) and correspond to bending and
Figure 4. Molecular interaction potential contours (-5 kcal/mol; top) and solvation maps (2.5 g/mL; bottom) for the B (right) and apH (left) helices. The
data for the 10-mer oligonucleotides are shown in all cases (similar profiles are obtained for the other sequences).
A R T I C L E S Cubero et al.
14608 J. AM. CHEM. SOC.9VOL. 125, NO. 47, 2003
untwisting movements for the two structures. In all cases, only
very small differences are found between the frequencies of the
first essential motions of B and apH structures. For example,
the first three frequencies of the 10-mer sequence are (in cm-1)
15, 22, and 24 for the B helix, and 13, 16, and 27 for the apH
structure. A quantitative comparison between the essential
motions of B and apH helices by the absolute and relative
similarity indexes γB/apH and κB/apH (eqs 2 and 3; Table 4)
indicates a close similarity in the dynamics of the two helices.
In summary, MD simulations point out that many of the
structural, reactive, and dynamic properties of the apH helix
are similar to those of the B helix, suggesting that proteins which
evolved to recognize the general shape of B DNA could also
recognize apH DNA. The subtle differences at the bottom of
grooves might then alter the specificity of the interactions,
opening a wide range of possible biological roles for the apH
Relative Stability of B and apH Helices. The preceding
discussion points out that the crystal structure found by Abrescia
et al.17is not a lattice-artifact, but an intrinsically stable structure
of polyd(AT)nin aqueous solution. However, to determine the
possible biological role of apH helices, it is necessary to
determine the relative stability of the apH helix versus the
standard B-like conformation. For this purpose, we computed
the free energy associated with each conformation using eq 1
(see Methods). It is worth noting that with 2-5 ns trajectories
the average effective energy, 〈Eeffec〉 ) 〈Eintra〉 + 〈Gsol〉, can be
estimated more accurately than the entropic term, which should
need longer trajectories to converge.10aThus, we have performed
two separate analyses: one quantitative for 〈Eeffec〉 and another
qualitative for ∆Sintra.
The effective energies for the B and apH helices are shown
in Table 5 (PB/SA values were used here to compute the
solvation free energy; similar results were obtained with the
GB/SA method; see below). The similarity in the 〈Eeffec〉 values
determined for the two conformations is evident. Thus, for the
10- and 12-mer oligonucleotides (those for which the longest
trajectories are available), the 〈Eeffec〉 values differ by only 0
and 3 kcal/mol, that is, less than 0.06% of the total energy value,
and lie within the estimated error of the averages (see Table 5).
A more quantitative estimate of the relative effective energy
for B and apH helices can be obtained,10ataking advantage of
the linear dependence between 〈Eeffec〉 and the length (n) of the
oligonucleotide (r2) 1.0000 for the regression line 〈Eeffect〉 )
a × (n - 2) + b in all cases). Typical errors (see Figure 6) in
the slope (a: helix-growth stabilization factor) and intercept
(b: nucleation energy) are only around 0.2-0.5 and 2-4 kcal/
mol, respectively. The regression equations (see Figure 6)
suggest that there is not a significant preference for either B or
apH conformations in aqueous solution in terms of 〈Eeffec〉 (the
same trends were found when solvation was determined using
the GB/SA model, and when PB/SA calculations were repeated
considering the internal dielectric constant equal to 2; data not
shown). This supports the equivalence in effective energy
between B and apH structures and suggests that the population
of the two helices in solution might depend on the existence of
cofactors, specific hydration waters, and entropic considerations.
Due to the similar degree of hydration in the first hydration
Figure 5. Details of regions of specific hydration in the apH helix (2.5
g/mL contours) along the major groove. Data correspond to the 10-mer
structure. Water molecules with a long residence time are located in the
areas. Residence times for waters bound to O2and O4are very large (150-
200 ps), waters bound to N3show also large residence times (around 50-
100 ps), waters around N1show residence times of 30-50 ps, and those
around N6have residence times around 20 ps.
Table 4. Similarity Indexes γ Obtained Using Eq 1 and the 10
First Principal Components for the 10- and 12-mer Duplexes in B
and apH Structuresa
aSimilarity indexes γB/apHwere computed using the last 4 ns of both
trajectories. Self-similarity indexes γB/B and γapH/apH were determined
comparing the first (1 f 3 ns) and second (3 f 5 ns) halves of the
Table 5. Effective Energy (〈Eeffec〉 in Bold), and Their Intra-
molecular and Solvation Contributions (〈Gsol〉 in Roman/〈Eintra〉 in
Italics) Computed for the Different Oligonucleotides in B and apH
-637.5 ( 0.4
-544.0 ( 0.4/-93.5 ( 0.2
-1363.0 ( 0.7
-1461.3 ( 0.6/98.2 ( 0.3
-2093.3 ( 1.6
-2647.0 ( 1.5/553.7 ( 0.5
-2819.7 ( 1.9
-4075.7 ( 1.8/1256.1 ( 0.6
-3548.9 ( 3.0
-5681.4 ( 2.9/2132.5 ( 1.0
-4269.9 ( 3.3
-7453.7 ( 3.2/3183.8 ( 1.0
-4999.7 ( 4.7
-9350.3 ( 4.4/4350.6 ( 1.4
-635.4 ( 0.4
-559.9 ( 0.4/-75.6 ( 0.2
-1364.9 ( 0.6
-1484.3 ( 0.5/119.4 ( 0.2
-2095.2 ( 1.1
-2682.3 ( 1.0/587.1 ( 0.4
-2821.9 ( 1.9
-4083.8 ( 1.9/1261.9 ( 0.6
-3548.9 ( 2.6
-5681.5 ( 2.4/2132.6 ( 0.8
-4276.5 ( 3.4
-7387.8 ( 4.1/3111.3 ( 1.3
-5001.5 ( 3.6
-9297.1 ( 3.4/4295.6 ( 1.1
aValues determined using PB/SA estimates of solvation free energy for
internal and external dielectric constants of 1 and 80. Values and their
standard errors (in parentheses) are in kcal/mol.
Theoretical Study of a NewDNA StructureA R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 125, NO. 47, 2003 14609
shell (see above), no large differences in stability are expected
to arise from interaction with specific water molecules. The
impact of entropic effects and cofactors will be discussed below.
To gain further insight into the similar effective energies of
B and apH structures, we examined their intramolecular and
solvation contributions (see eq 1 and Table 5). As found in
previous work,10athere is a second-order polynomial relationship
between the size of the duplex and its solvation and intramo-
lecular energy components. The linear relationship found for
Eeffec then stems from the cancellation of the second-order
coefficients for the solvation and intramolecular terms (see
Figure 7). In general, for medium to long oligonucleotides, the
B form is slightly favored by solvation and slightly disfavored
by intramolecular interactions with respect to the apH confor-
mation. However, the differences are small (see Table 5 and
Figure 7), and we can conclude that not only the B and apH
structures are similar in terms of the effective energies, but also
in their solvation and intramolecular contributions.
The similar solvation free energy of apH and B helices is
not surprising considering the solvation maps (Figures 4 and
5) and the number of water molecules in the first hydration
shell (see above). However, the similar intramolecular energy
is surprising considering that apH duplexes imply the existence
of syn-adenosines, which are expected to be less stable than
the anti conformation. However, NMR data in DMSO and
water54show that the free energy difference between syn and
anti conformations is small (0.4-0.6 kcal/mol for adenosine).
A similar value (0.7 kcal/mol) is also found by restricted
AMBER-99 optimizations of syn and anti adenosines at the MD-
averaged ? torsions found in our B and apH simulations. This
small energy penalty and others related to unfavorable backbone-
backbone interactions in the apH helix are compensated by the
better stacking and H-bond interactions in Hoogsteen A-T pairs.
Thus, AMBER calculations in the gas phase using the MD-
averaged structures of the Watson-Crick and Hoogsteen pairs
show interaction energies of -12.7 (WC) and -13.2 (H) kcal/
mol; that is, the Hoogsteen H-bond pattern is around 0.6 kcal/
mol more stable than the Watson-Crick scheme.55Linear
regression lines (see Figure 8) between the H-bonding energy
and the length of the oligonucleotides suggest that, in the DNA
environment, the Hoogsteen scheme is around 1.3 kcal/(mol
step) more stable than the Watson-Crick pattern. Similarly
(Figure 8), the stacking energy of the Hoogsteen pair is on
average around 1.9 kcal/(mol step) more stable than that of the
Watson-Crick pair in the DNA environment.
In summary, the MM/PB-SA (and MM/GB-SA) analysis,
despite its shortcomings, strongly suggests similar stability for
the B and apH structures in aqueous solution in terms of
intramolecular energy and solvation free energy. The relative
population of the two helices might then be determined by
entropic considerations. The structural and dynamical analysis
described above points out that the B form is more flexible than
the apH form, the former being then entropically favored. To
obtain more detailed information, entropic calculations were
(54) Stolarski, R.; Pohorille, A.; Dudycz, L.; Shugar, D. Biochim. Biophys. Acta
1980, 610, 1.
(55) These values agree very well with our own B3LYP/6-31G(d) estimates
(-12.1 and -12.6 kcal/mol, respectively), and with MP2/6-31G*(0.25)
estimates by Sponer and Hobza (-11.8 (WC) and -12.7 kcal/mol,
respectively, from ref 23).
Figure 6. Representation of the variation of the effective energy 〈Eeffec〉 )
〈Eintra〉 + 〈Gsol〉 as a function of the length of the oligonucleotide for B and
apH conformations. Top: Solvation computed using the GB/SA method.
Bottom: Solvation computed using the PB/SA approach. In both cases,
interior and exterior dielectric constants are 1 and 80. All values are in
Figure 7. Representation of the variation of the solvation free energy
(top: 〈Gsol〉) and intramolecular energy (bottom: 〈Eintra〉) as a function of
the length of the oligonucleotides for B and apH conformations (all values
are in kcal/mol). Solvation values displayed here correspond to the PB/SA
calculations with dielectric constants of 1 (interior) and 80 (exterior).
Identical profiles where obtained using other solvation models (see text
and Table 5). Parabolic fits correspond to the equation: y ) a2x2+ a1x + b.
A R T I C L E SCubero et al.
14610 J. AM. CHEM. SOC.9VOL. 125, NO. 47, 2003
performed using Schlitter’s method, as described in the Methods
section. Although accurate entropic calculations would need
longer trajectories than those performed here, the analysis of
2-5 ns trajectories can still be useful to obtain relative entropy
Entropy calculations were then performed for all of the
trajectories in the 1-2 ns period, finding linear relationships
with the length of the oligonucleotides (r2> 0.993; see Figure
9). The regression lines confirm that the B form is entropically
favored as compared to the apH helix in all cases and that the
difference slightly increases with the length of the oligonucle-
otide. For example, the regression equations indicate that for
the 10- and 12-mer oligonucleotides the B structure is favored
by 0.11 and 0.14 kcal/(mol K) with respect to the apH helix.
Similar values are obtained when entropies are calculated using
the 5 ns trajectories and extrapolated to infinite simulation time
using Harris’s method:430.15 (10-mer) and 0.21 (12-mer) kcal/
(mol K). These values suggest that at T ) 298 K the entropic
effect will be around 45 (10-mer) or 63 (12-mer) kcal/mol, a
sizable difference considering the small differences in effective
energies, which would lead to a predominance of the B form
in aqueous solution. We can expect that when the flexibility of
the DNA is restricted by the presence of drugs, proteins, or
lattice constraints, the intrinsic flexibility of the DNA is
drastically altered, modifying then the entropy balance between
B and apH helices. In the latter cases, the apH helix could
become competitive with the B structure, explaining the
existence of apH structures in the crystal17and the presence of
apH recognition modes in complexes of DNA with either
intercalators or proteins (see Introduction).
Chimeras of B and apH Helices. In long DNA fragments,
the apH structure may occur in short fragments surrounded by
the B-structure, and often the fragments will contain both d(AT)
and d(AA) steps. Accordingly, whether an apH track can be
inserted into a B canonical structure without inducing large
distortions, and whether it can exist in polyd(A) sequences, are
key issues in guessing the biological importance of apH helices.
Test calculations of poly(dA)10(data not shown, but available
upon request) show that the apH helix can also be stable: the
average effective energy (eq 1) found in 2 ns trajectories of the
apH helix (-2663 kcal/mol) was almost identical to that found
for a WC duplex of the same sequence (-2662 kcal/mol; ref
10a) and was slightly more negative than that found for parallel
stranded duplexes with reverse Watson-Crick (-2642 kcal/
mol; ref 10a) and Hoogsteen pairings (-2657 kcal/mol; ref 10a).
In summary, our MD simulations suggest that the apH helix
can exist also in poly(dA) sequences, opening then the window
of structures for which apH helices are possible.
The impact of having apH sequences inserted in B-DNA was
studied in two chimeric 11-mer sequences: d(ATATAAAATAT)
and d(ATATAAAATAT), where the steps having the Watson-
Crick recognition mode are in plain text, and those with the
apH pattern are in bold text. Unrestricted NPT trajectories
collected for 2 ns after equilibration are very stable and reveal
no strong destabilization, conformational transition, or unfolding
(see Figures 10 and 11). The backbone rmsd values with respect
to the average structures are around 1.3 Å in the two cases
(Figure 11), thus confirming the stability of the trajectories. The
apH-rich d(ATATAAAATAT) structure is closer to an apH
helix than to a B conformation (backbone rmsd values of
1.9 and 2.4 Å, respectively), while the Watson-Crick-rich
d(ATATAAAATAT) structure is intermediate between apH and
B conformations (backbone rmsd values of 1.9 and 1.8 Å,
respectively). Very interestingly, the chimeras maintain clearly
separable conformations corresponding to the regions with apH
or Watson-Crick pairings (see Figure 10). Such a structural
difference has a clear impact on the recognition properties of
the helix (see Figure 10). Thus, for the d(ATATAAAATAT)
helix, the -5 kcal/mol contour is discontinuous at the central
region, as expected for an apH helix, and contiguous in the rest,
as in a canonical B-helix (see Figure 4). On the contrary, the
reverse trend occurs for the d(ATATAAAATAT) sequence.
This finding indicates that not only apH and B helices are
Figure 8. Representation of the variation of the hydrogen bond (top) and
stacking energies (bottom) as a function of the length of the oligonucleotides
for B and apH conformations (all values are in kcal/mol).
Figure 9. Representation of the variation of the entropy (in kcal/(mol K))
as a function of the length of the oligonucleotides for B and apH
conformations (all values are in kcal/mol). All values were obtained from
1 ns samplings (see text).
Theoretical Study of a NewDNA Structure A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 125, NO. 47, 2003 14611
compatible in the same oligonucleotide, but that even when they
are contiguous each helix fragment preserves its own structural
and reactive integrity.
In summary, the preceding results show that the apH structure
is a thermodynamic stable conformation, which might compete
with the B conformation if the flexibility of the helix is altered.
However, for the sequences studied, MD simulations favor the
B form in dilute aqueous solution of lineal DNAs in the absence
of cofactors. Interestingly, the structural, dynamical, and reactive
properties of the apH helices are not extremely different from
those of the B helix. However, the small changes should allow
proteins and drugs to distinguish between these two helical
models. MD simulations suggest that apH structures can coexist
next to B-type structures, demonstrating again the extreme
plasticity of DNA. If kinetic factors are not limiting (the syn
T anti barrier is around 6.2 kcal/mol56), AT-rich regions might
adopt B and apH forms, and this possibility might be exploited
to tune the interaction with specific proteins.
Acknowledgment. We are indebted to X.-J. Lu and W. Olson
for a copy of X3DNA and help in the use of the code for a
nonstandard structure. We also thank D. Bashford for a copy
of his MEAD program. The Centre de Supercomputacio ´ de
Catalunya (CESCA) and the Spanish Ministry of Science and
Technology (SAF2002-4282, PM99-0046, and BIO2003-06848)
are acknowledged for financial support.
(56) Rhodes, L. M.; Schimmel, P. R. Biochemistry 1971, 10, 4426.
Figure 10. MIP (-5 kcal/mol) contour for the d(ATATAAAATAT) (left) and d(ATATAAAATAT) (right) structures. The apH portions are colored in
green, and the WC portions are in blue. Bold means Hoogsteen pairings.
Figure 11. Backbone root-mean-square deviations (rmsd in Å) of the
d(ATATAAAATAT) (top) and d(ATATAAAATAT) (bottom) trajectories
with respect to the following: red, the MD-averaged structures; green, the
B structure; and blue, the apH structure. Averaged rmsd values with respect
to the average, B, and apH structures are 1.24, 2.44, and 1.91 Å for the
d(ATATAAAATAT) structure, and 1.37, 1.91, and 1.80 Å for the
d(ATATAAAATAT) structure. Bold means Hoogsteen pairings.
A R T I C L E S Cubero et al.
14612 J. AM. CHEM. SOC.9VOL. 125, NO. 47, 2003