Optimization of interstrand hydrophobic packing interactions within unnatural DNA base pairs

Article abstract of DOI:10.1021/ja047291m

As part of an effort to expand the genetic alphabet, we have evaluated a large number of predominantly hydrophobic unnatural base pairs. We now report the synthesis and stability of unnatural base pairs formed between simple phenyl rings modified at diffe

Full text of DOI:10.1021/ja047291m

Published on Web 10/19/2004
Optimization of Interstrand Hydrophobic Packing Interactions
within Unnatural DNA Base Pairs
Shigeo Matsuda and Floyd E. Romesberg*
Contribution from the Department of Chemistry, The Scripps Research Institute,
10550 North Torrey Pines Road, La Jolla, California 92037
Received May 9, 2004; E-mail: floyd@scripps.edu
Abstract: As part of an effort to expand the genetic alphabet, we have evaluated a large number of
predominantly hydrophobic unnatural base pairs. We now report the synthesis and stability of unnatural
base pairs formed between simple phenyl rings modified at different positions with methyl groups.
Surprisingly, several of the unnatural base pairs are virtually as stable as a natural base pair in the same
sequence context. The results show that neither hydrogen-bonding nor large aromatic surface area are
required for base pair stability within duplex DNA and that interstrand interactions between small aromatic
rings may be optimized for both stability and selectivity. These smaller nucleobases are not expected to
induce the distortions in duplex DNA or at the primer terminus that seem to limit replication of larger unnatural
base pairs, and they therefore represent a promising approach to the expansion of the genetic alphabet.
1. Introduction
In an effort to develop a third base pair to expand the genetic
alphabet, we have synthesized and characterized a large number
of nucleotides bearing unnatural nucleobase analogues.1-10 Just
as with their natural counterparts, these unnatural nucleotides
must pair stably and selectively within duplex DNA. In the case
of natural DNA, thermal stability and selectivity result from a
combination of intrastrand packing and interstrand hydrogen
bonding (H-bonding) interactions. Correspondingly, thermal
stability and selectivity of DNA containing unnatural base pairs
may be mediated by either intra- or interstrand interactions.
Nucleobase H-bonding interactions have been modified11-15 and
optimized,16,17 and their intrastrand packing has been improved
Figure 1. Isocarbostyril (left), naphthyl (middle), and indol (right) scaffolds.
by increasing their surface area.18,19 Increased surface area has
been achieved by the addition of alkyl, alkynyl, or aryl
substituents at the C7 and C5 positions of purines and
pyrimidines, respectively.18-22 Similarly, we have evaluated a
variety of large aromatic nucleobase analogues, based on the
isocarbostyril, naphthyl, and indole scaffolds (Figure 1). (Al-
though many of the unnatural nucleobases are not actually basic,
we refer to them as nucleobase analogues for simplicity.)
Generally, these nucleobases form unnatural self-pairs (pairs
formed between two identical unnatural nucleotides) and
heteropairs (pairs formed between different unnatual nucleotides)
with reasonable stability and good selectivity, relative to the
natural base pairs.1-10 Either self-pairs or heteropairs may be
used for expansion of the genetic code.
Efficient replication by a DNA polymerase is also required
of an unnatural base pair. Unfortunately, we have generally
found that nucleobases with large hydrophobic surface area are
poor substrates for replication.1-10 While the unnatural tri-
phosphates tend to be efficiently inserted opposite their designed
partner in the template, the nascent unnatural primer terminus
(1) Berger, M.; Ogawa, A. K.; McMinn, D. L.; Wu, Y.; Schultz, P. G.;
Romesberg, F. E. Angew. Chem., Int. Ed. 2000, 39, 2940-2942.
(2) Berger, M.; Luzzi, S. D.; Henry, A. A.; Romesberg, F. E. J. Am. Chem.
Soc. 2002, 124, 1222-1226.
(3) Henry, A. A.; Yu, C.; Romesberg, F. E. J. Am. Chem. Soc. 2003, 125,
9638-9646.
(4) Matsuda, S.; Henry, A. A.; Schultz, P. G.; Romesberg, F. E. J. Am. Chem.
Soc. 2003, 125, 6134-6139.
(5) McMinn, D. L.; Ogawa, A. K.; Wu, Y.; Liu, J.; Schultz, P. G.; Romesberg,
F. E. J. Am. Chem. Soc. 1999, 121, 11585-11586.
(6) Ogawa, A. K.; Wu, Y.; McMinn, D. L.; Liu, J.; Schultz, P. G.; Romesberg,
F. E. J. Am. Chem. Soc. 2000, 122, 3274-3287.
(7) Ogawa, A. K.; Wu, Y.; Berger, M.; Schultz, P. G.; Romesberg, F. E. J.
Am. Chem. Soc. 2000, 122, 8803-8804.
(8) Tae, E. L.; Wu, Y. Q.; Xia, G.; Schultz, P. G.; Romesberg, F. E. J. Am.
Chem. Soc. 2001, 123, 7439-7440.
(9) Wu, Y.; Ogawa, A. K.; Berger, M.; McMinn, D. L.; Schultz, P. G.;
Romesberg, F. E. J. Am. Chem. Soc. 2000, 122, 7621-7632.
(10) Yu, C.; Henry, A. A.; Romesberg, F. E.; Schultz, P. G. Angew. Chem., Int.
Ed. 2002, 41, 3841-3844.
(11) Switzer, C.; Moroney, S. E.; Benner, S. A. J. Am. Chem. Soc. 1989, 111,
8322-8323.
(16) Lin, K.-Y.; Matteucci, M. D. J. Am. Chem. Soc. 1998, 120, 8531-8532.
(17) Flanagan, W. M.; Wolf, J. J.; Olsen, P.; Grant, D.; Lin, K. Y.; Wagner, R.
W.; Matteucci, M. D. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 3515-3518.
(18) Kool, E. T.; Morales, J. C.; Guckian, K. M. Angew. Chem., Int. Ed. 2000,
39, 990-1009.
(12) Piccirilli, J. A.; Krauch, T.; Moroney, S. E.; Benner, S. A. Nature 1990,
343, 33-37.
(13) Switzer, C. Y.; Moroney, S. E.; Benner, S. A. Biochemistry 1993, 32, 2,
10489-10496.
(19) Kool, E. T. Chem. ReV. 1997, 97, 1473-1487.
(14) Horlacher, J.; Hottiger, M.; Podust, V. N.; Hu¨bscher, U.; Benner, S. A.
Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 6329-6333.
(15) Lutz, M. J.; Held, H. A.; Hottiger, M.; Hu¨bscher, U.; Benner, S. A. Nucleic
Acids Res. 1996, 24, 1308-1313.
(20) Ahmadian, M.; Zhang, P.; Bergstrom, D. E. Nucleic Acids Res. 1998, 26,
3127-3135.
(21) He, J.; Seela, F. Nucleic Acids Res. 2002, 30, 5485-5496.
(22) Seela, F.; Zulauf, M. HelV. Chim. Acta 1999, 82, 1878-1898.
9
10.1021/ja047291m CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 14419-14427
14419

A R T I C L E S
Matsuda and Romesberg
sufficient for controlling molecular recognition within duplex
DNA without H-bonding or extended aromatic surface area.
2. Results
2.1. Nucleobase Analogues. The synthesis and characteriza-
tion of the unnatural nucleoside is described in the Supporting
Information. Briefly, nucleobase-sugar coupling relied on
Heck-type coupling reactions28-31 between the silyl-protected
glycal32 and each methyl-substituted iodobenzene or addition
of the aryllithium reagent to the disiloxane-protected 2-deoxy-
D-ribono-1,4-lactone.33 In all cases, the phosphoramidite used
in automated DNA synthesis was obtained by standard dimethoxy-
tritylation and phosphitylation. Oligonucleotides were synthe-
sized on an Applied Biosystems 392 DNA synthesizer using
standard solid-phase â-cyanoethyl phosphoramidite chemistry.
After deprotection under basic conditions, the crude oligonucleo-
tides were purified by polyacrylamide gel electrophoresis.
2.2. Preliminary Structural Characterization of DNA
Containing Unnatural Self-Pairs. To determine whether the
unnatural self-pairs perturbed the overall structure of a short
DNA duplex, each unnatural nucleoside was incorporated, at
position X, into the complementary oligonucleotides 5′-d(GCG-
TACXCATGCG)-3′ and 5′-d(CGCATGXGTACGC)-3′. The
circular dichroism (CD) spectrum, from 220 to 360 nm, was
recorded for each unnatural duplex, as well as for the fully
natural duplexes containing dA:dT or dG:dC (Figure 3). The
λmax and λmin of all of the spectra are virtually identical,
indicating that the unnatural base pairs do not significantly
distort the B-form conformation of the DNA duplex. The
amplitudes of the negative feature for each duplex containing a
self-pair are similar and intermediate between the duplex
containing a dA:dT and dG:dC base pair. There is more variation
in the amplitude of the positive feature. While the varying
ellipticity is difficult to interpret, we note that it at least roughly
follows the stability of the self-pairs and thus may reflect
differences in packing with neighboring base pairs.
Figure 2. Unnatural nucleobases used in this study.
is then a poor substrate for continued extension. One explanation
for the poor extension rates may be the absence of heteroatoms
in both the major and the minor grooves. These atoms would
contribute to the physical properties of the nucleobases, includ-
ing dipole moment and polarizability.23-27 Indeed, we have
recently demonstrated that heteroatom substitution may be used
to improve unnatural base pair extension rates.3,4,10 It may also
be that the large aromatic surface area itself contributes to
inefficient extension. Steric repulsion or interstrand intercalation
between the unnatural bases may distort the primer terminus
and disfavor extension. In fact, preliminary NMR studies suggest
that the isocarbostyril-based unnatural nucleobases do interstrand
intercalate and appear to adopt a geometry that is favorable for
packing but inappropriate for polymerase-catalyzed primer
extension (Wemmer, D., unpublished data). We hypothesize that
small nucleobases, having an aromatic surface area that is too
limited to allow interstrand intercalation and having suitably
positioned heteroatoms, will form unnatural base pairs that are
thermally stable and selective and also efficiently synthesized
and extended by a DNA polymerase.
To generate a more detailed structural model, we energy-
minimized the same duplex containing the DM5 self-pair
(Figure 4) using the AMBER* force field,34,35 implemented with
the MacroModel program (version 6.0).36 The resulting structure
predicts that the DNA double helix can accommodate the DM5
self-pair without major backbone or stacking distortions (Figure
4a). The torsion angle about the exocyclic C4′-C5′ bond, critical
for positioning of the 5′ phosphate group, relative to the sugar
and the nucleobase, is ∼63° for both unnatural nucleotides and
is between 62 and 65 °C for each flanking natural nucleotide.
The intrastrand phosphate distance (6.6-7.0 Å) is typical of
B-form DNA.37
To test this hypothesis, we first asked whether packing
interactions between benzene-based nucleobase analogues that
are too small to intercalate might be optimized for stability and
selectivity within duplex DNA. Herein, we report the synthesis
and thermodynamic characterization of 10 novel nucleotides
bearing methyl-derivatized benzene rings which, when combined
with two previously reported analogues,6 provide a systematic
analysis of heteropair and self-pair stability (Figure 2). All of
the unnatural base pairs have good selectivity, preferring to pair
with one another rather than pair with a natural base, and several
are also reasonably stable relative to a natural base pair. These
studies demonstrate that hydrophobic and packing forces are
(28) Farr, R. N.; Outten, R. A.; Cheng, J. C. Y.; Daves, G. D. Organometallics
1990, 9, 3151-3156.
(29) Zhang, H. C.; Daves, G. D. J. Org. Chem. 1992, 57, 4690-4696.
(30) Hsieh, H. P.; McLaughlin, L. W. J. Org. Chem. 1995, 60, 5356-5359.
(31) Zhang, H. C.; Brakta, M.; Daves, G. D. Nucleosides Nucleotides 1995, 14,
105-116.
(23) Bergmann, E. D.; Weiler-Feilchenfeld, H. In The Dipole Moments of
Purines; Bergmann, E. D., Pullman, B., Eds.; The Israel Academy of
Sciences and Humanities: Jerusalem, 1971; pp 21-28.
(24) Lister, J. H. The Purines. Supplement 1; John Wiley and Sons: New York,
1996; Vol. 54.
(25) Hurst, D. T. An Introduction to the Chemistry and Biochemistry of
Pyrimidines, Purines and Pteridines; Page Bros.: Norwich, Great Britain,
1980.
(26) Gilchrist, T. L. Heterocyclic Chemistry; Longman Scientific & Technical:
Essex, 1985.
(32) Cameron, M. A.; Cush, S. B.; Hammer, R. P. J. Org. Chem. 1997, 62,
9065-9069.
(33) Wichai, U.; Woski, S. A. Org. Lett. 1999, 1, 1173-1175.
(34) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.;
Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P.
A. J. Am. Chem. Soc. 1995, 117, 5179-5197.
(35) Ferguson, D. M.; Kollman, P. A. J. Comput. Chem. 1991, 12, 620-626.
(36) Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton, M.;
Caufield, C.; Chang, G.; Hendrickson, T.; Still, W. C. J. Comput. Chem.
1990, 11, 440-467.
(27) Bugg, C. E. In Solid-State Packing Patterns of Purine Bases; Bergmann,
E. D., Pullman, B., Eds.; The Israel Academy of Sciences and Humani-
ties: Jerusalem, 1971; pp 178-204.
(37) Saenger, W. Principles of Nucleic Acid Structure; Springer-Verlag: New
York, 1984.
9
14420 J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004

Optimization of Hydrophobic Packing in DNA
A R T I C L E S
Figure 3. CD spectra of DNA duplexes 5’-d(GCGTACXCATGCG)-3’/5’-d(CGCATGXGTACGC)-3’ containing self-pairs or natural base pairs at position
X. The spectra of the fully natural duplexes are indicated. In the inset are listed the base pairs from top to bottom corresponding to decreasing amplitude of
the CD signal at 280 nm.
The O4′-C1′-C1-C2 dihedral angles of the DM5 nucleo-
tides are -148° (for -GDM5G-) and -113° (for -CDM5C-).
We will refer to this conformation as anti by analogy to natural
nucleotides. The anti conformation positions the ortho-methyl
group in the minor groove as opposed to over the deoxyribose
ring. We speculate that all of the ortho-substituted unnatural
nucleotides adopt such anti-orientations to minimize steric
interactions between the methyl group and the sugar ring. This
conformation positions the methyl substituents of both DM5
nucleobases appropriately to pack with one another. As a result,
the two DM5 analogues appear to pack well with each other,
with interbase distances approaching the sum of the van der
Waals radii (Figure 4b). However, the methyl groups are not
equally packed within the duplex (Figure 4b). For the DM5
nucleobase flanked by two pyrimidines, only the para-methyl
group is well packed by the flanking bases. In contrast, for the
DM5 nucleobase flanked by two purines, both the ortho- and
the para-methyl groups appear to be well packed, due to the
increased aromatic surface area of the neighboring bases.
Together, the CD and modeling studies suggest that the
unnatural base pairs are accommodated in a B-form DNA duplex
with each unnatural nucleotide adopting an orientation that
minimizes steric interactions with the sugar and optimizes
methyl group packing with the pairing nucleobase. This
structural model is used to interpret the thermodynamic data
described in the following section.
at the positions labeled X and Y. This sequence context was
chosen to examine the effect of sequence context by comparing
the effects of flanking pyrimidines (position X) and flanking
purines (position Y). The melting temperature (Tm) of each
duplex was determined by thermal denaturation experiments
(Table 1). The term “stability” is used here only to refer to Tm
values, unless otherwise indicated. The unnatural nucleotides
were examined both as self-pairs and as heteropairs. For
comparison, the Tm for a duplex containing a natural base pair
(X:Y ) dA:dT) is 59.2 °C.
Generally, the least substituted unnatural base pairs were the
least stable. Accordingly, the BEN self-pair was the least stable,
with a Tm of only 52.8 °C. The heteropairs formed between
BEN and a monosubstituted derivative (MM1, MM2, or MM3)
were only slightly more stable, and independent of sequence
context, with a constant Tm of approximately 53 °C. The
heteropairs formed between BEN and more highly substituted
analogues were generally more stable, and in these cases the
stability did depend on the sequence context. With BEN at
position X (flanked by pyrimidines) and the more substituted
analogue at position Y (flanked by purines), the pairs showed
stability that increased with substitution, culminating in the
BEN:TM2 pair, only 3.7 °C less stable than a dA:dT base pair.
Each pair was slightly less stable in the opposite sequence
context (Y ) BEN) in which the more substituted analogue
was flanked by pyrimidines. The sequence dependence suggests
that, with these pairs, intrastrand packing is more important than
interstrand packing.
2.3. Stability of Unnatural Base Pairs. To evaluate the
thermodynamic stability and selectivity of the unnatural base
pairs in duplex DNA, all unnatural and natural nucleosides were
incorporated into the complementary oligonucleotides 5′-
d(GCGTACXCATGCG)-3′ and 5′-d(CGCATGYGTACGC)-3′
The self-pairs formed among the monosubstituted analogues,
MM1, MM2, and MM3, are more stable than the BEN:BEN
self-pair, by 1.4, 0.5, and 1.3 °C, respectively. MM2 forms the
9
J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004 14421

A R T I C L E S
Matsuda and Romesberg
Figure 4. Energy-minimized model structure of the duplex containing the DM5 self-pair. (a) View from major groove (left) and from minor groove (right).
(b) Intrastrand stacking of the DM5 (shown in green) self-pair (left). The DM5 (gray) self-pair shown as a CPK model (right).
Table 1. Tm Values (°C) for Duplex Containing Unnatural Base Pairsa
X
Y
BEN
MM1
MM2
MM3
DM
DM2
TM
DM3
DM4
DM5
TM2
TMB
BEN
MM1
MM2
MM3
DM
52.8
53.1
53.0
53.2
53.2
53.0
54.1
54.7
54.5
54.6
55.5
55.1
53.3
54.2
54.1
54.3
54.0
54.1
54.1
55.2
55.2
55.3
56.1
55.0
53.0
53.1
53.3
53.2
54.0
54.0
53.2
54.2
54.2
55.0
55.1
54.2
53.0
54.1
54.0
54.1
54.0
54.0
54.0
55.1
55.1
55.2
57.2
56.1
53.3
53.2
54.0
54.2
53.7
54.3
54.2
55.0
55.1
55.2
55.1
54.1
53.1
54.2
54.1
54.0
54.1
54.5
54.2
55.7
54.6
55.6
56.5
55.2
53.1
54.3
54.2
54.1
54.2
55.1
55.2
56.1
56.1
56.1
58.0
57.2
54.2
55.0
54.2
55.1
55.2
55.1
55.1
55.7
56.1
58.1
57.0
56.0
53.1
54.1
54.3
55.1
54.0
54.1
56.1
56.2
56.1
56.1
57.0
56.0
53.7
55.1
55.0
55.2
54.1
55.6
56.2
55.7
56.6
57.3
57.5
57.1
54.1
54.2
54.3
56.2
55.2
55.1
57.1
56.2
57.0
57.0
57.2
57.1
54.0
54.3
54.0
54.0
54.2
55.3
56.0
55.0
56.1
56.2
57.2
57.2
DM2
TM
DM3
DM4
DM5
TM2
TMB
a Determined in 10 mM PIPES, 10 mM MgCl2, 100 mM NaCl, pH 7.0. For reference, Tm ) 59.2 °C for X ) dA, Y ) dT.
least stable self-pair, and it also forms less stable heteropairs
when flanked by pyrimidines. However, the stabilities of all
nine of these self-pairs and heteropairs differ by only ∼1 °C.
When a monosubstituted analogue is paired opposite an
analogue with greater methyl substitution, the base pair is
stabilized. These pairs also display a sequence dependence that
is similar to that observed with the BEN pairs: the pairs tend
to be more stable when the more substituted analogue is flanked
by purines. This again indicates that stacking between the
unnatural analogue and the flanking bases makes an important
contribution to the stability of the unnatural base pair. However,
these more substituted pairs also begin to show a dependence
9
14422 J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004

Optimization of Hydrophobic Packing in DNA
A R T I C L E S
Table 2. Tm Values (°C) for Duplex Containing Self-Pairs or
on the specific substitution pattern. For example, when paired
opposite DM5, each of the monosubstituted analogues forms
pairs with virtually identical stability (∼55 °C), but the pairs
formed with the most highly substituted TM2 or TMB differ
by nearly 2 °C. This indicates that specific interstand interactions
between the unnatural analogues also contribute to stability.
Mispairsa
X
Y
Tm
(
°
C)
X
Y
Tm (°C)
BEN
BEN
C
G
T
A
MM2
C
G
T
A
DM
C
G
T
A
DM3
C
G
T
A
DM5
C
G
T
A
TM2
C
G
T
A
52.8
45.0
49.1
47.6
49.2
53.3
45.0
49.0
48.2
49.2
53.7
43.8
48.7
48.2
48.5
55.7
46.0
49.0
48.1
49.0
57.3
47.2
50.0
49.0
49.2
57.2
47.0
50.1
49.1
50.2
MM1
MM1
C
G
T
A
MM3
C
G
T
A
DM2
C
G
T
A
DM4
C
G
T
A
TM
C
G
T
A
TMB
C
G
T
A
54.2
45.0
49.0
46.1
50.1
54.1
46.2
50.0
48.1
50.1
54.5
44.0
49.1
48.1
50.2
56.1
47.3
50.0
49.1
50.1
55.2
44.7
50.0
49.2
51.7
57.2
45.2
49.2
49.0
50.3
The stabilities of the self-pairs formed between disubstituted
analogues are more variable. The DM3, DM4, and DM5 self-
pairs are more stable (Tm ) 55.7-57.3 °C) than the DM and
DM2 self-pairs (Tm ≈ 54 °C). This is also generally true of the
corresponding heteropairs: those formed between DM3, DM4,
and DM5 are only 2-3 °C less stable than a dA:dT in the same
sequence context. The reduced stability of the pairs involving
DM and DM2 may be understood in the context of the structural
model described above. Assuming that each nucleobase ana-
logue is oriented similarly to DM5 in the model, it seems
unlikely that the two methyl groups of DM and DM2 could
simultaneously pack with the pairing base. The increased
dependence of stability on substitution pattern suggests that, in
addition to intrastrand packing, interstrand packing interactions
contribute to molecular recognition within these moderately
substituted unnatural nucleobases.
MM2
DM
MM3
DM2
DM4
TM
DM3
DM5
TM2
The unnatural pairs involving the more highly substituted
analogues are generally quite stable, typically only 3 °C less
stable than a dA:dT in the same sequence context. The Tm values
for the TM and TM2 self-pairs are 55.26 and 57.2 °C,
respectively. This again demonstrates the importance of the
substitution pattern at the putative interstrand interface (i.e., all
three methyl groups can simultaneously pack at the interface
only with TM2). The TMB self-pair shows no increase in
stability relative to the TM2 self-pair, again indicating that
methyl groups are most stabilizing when they are positioned to
pack with the other analogue. Further supporting the importance
of interstrand interactions, heteropair stability varied widely
depending on the specific substitution pattern of the unnatural
nucleobases. The Tm values range between 53.1 and 58.0 °C.
The stabilities of the self-pairs and heteropairs formed with TM,
TM2, and TMB indicate that these nucleobases generally have
sufficient size to contribute to an interstrand interface that is
sufficiently well packed to stabilize the unnatural base pairs.
TMB
a Determined in 10 mM PIPES, 10 mM MgCl2, 100 mM NaCl, pH 7.0.
The stabilities of all possible mispairs in one sequence context
(Y ) dG, dA, dC, or dT) were determined and are reported in
Table 2. All of the mispairs are significantly less stable than
even the least stable unnatural pair. This demonstrates the
specificity inherent to these hydrophobic nucleobases, which
is likely to be caused by forced desolvation of the natural bases.
However, desolvation is not sufficient to explain the relative
stabilities of the mispairs. For example, in the case of BEN,
mispairs with dG and dA are the most stable, followed by dT
and then dC. The relative stabilities do not reflect hydrophobic-
ity, as dA and dT are the most hydrophobic.38 Instead, it seems
likely that intrastrand stacking underlies the stability of the
purine mispairs, just as it contributes to the stability of natural
base pairs.37 Thus, similar to the unnatural pairs involving BEN,
the stabilities of its mispairs appear to depend on intrastrand
stacking of the pairing base. Nonetheless, even the most stable
mispairs, BEN:dA and BEN:dG, are almost 4 °C less stable
than the BEN self-pair.
Similarly, mispairs between the monosubstituted analogues
(MM1, MM2, and MM3) and dC are least stable, and those
involving dG and dA are most stable. The stabilities were similar
to those with BEN, implying an absence of specific interactions.
The only exceptions were the MM1:dA and MM1:dT mispairs,
which are slightly more and less stable, respectively, than the
same mispairs with MM2 or MM3. Assuming that the ortho-
methyl group is positioned in the minor groove (see above),
this may be the result of a stabilizing (dA) or a destabilizing
Interestingly, the two most stable base pairs are not the most
substituted, again demonstrating that the interstrand interactions
become specific with a sufficiently packed interface. Duplex
DNA containing the TM:TM2 and DM3:DM5 pairs have Tm
values of 58.0 and 58.1 °C, respectively. These unnatural base
pairs are only ∼1.2 °C less stable than a dA:dT pair in the same
sequence context. Both pairs are formed by nucleobases that
have an ortho-methyl substituent. These methyl groups may
pack with each other in the minor groove, as it is assumed that
the nucleobases adopt an orientation in duplex DNA that
minimizes repulsion between any ortho-methyl group and the
sugar (Figure 4). Both pairs also contain a single meta-
substituted nucleobase, which is likely oriented toward the
pairing nucleobase. The stability of DM3:DM5 is especially
surprising, considering that these nucleobases have only two
methyl substituents. This unnatural base pair may be optimized
for interstrand packing.
2.4. Stabilities of Mispairs between Unnatural and Natural
Bases. It is important that the unnatural nucleobases are
thermally selective against mispairing with the natural bases.
(38) Shih, P.; Pedersen, L. G.; Gibbs, P. R.; Wolfenden, R. J. Mol. Biol. 1998,
280, 421-430.
9
J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004 14423

A R T I C L E S
Matsuda and Romesberg
Table 3. Tm Values (°C) of Duplex with Multiple DM5 Self-Pairsa
(dT) interaction in the minor groove. Similar interactions have
previously been evoked to explain the selective recognition of
dA in duplex DNA, and during polymerase-mediated base pair
synthesis.7 Most importantly, all mispairs were significantly less
stable than the “correct” unnatural pairs.
Mispairs between the disubstituted analogues (DM, DM2,
DM3, DM4, or DM5) and dC are all less stable than their
mispairs with the other natural bases. In addition, the mispair
stability varied by 3.5 °C, depending on the substitution pattern
of the unnatural base. This greater range in stability results
largely from the smaller Tm values of the DM:dC and DM2:dC
mispairs. These were the least stable mispairs of all, including
the mispairs with BEN or with the monosubstituted analogues.
This may again result from DM and DM2 being unable to
present both methyl substituents to the pairing base, as was
discussed above in relation to the unnatural pairs involving these
two analogues. However, the second methyl group not only fails
to stabilize the mispair, but actually destabilizes it by ∼1 °C
(compare MM1:dC to DM2:dC and MM2:dC to DM:dC). This
may result from the size and hydrophilicity of dC exacerbating
the poor packing of DM and DM2. The stabilities of the
mispairs between a disubstituted analogue and dG, dA, or dT
were all greater and less variable than the mispairs with dC.
Again, all of the mispairs between an unnatural and a natural
base are less stable than the “correct” unnatural base pairs.
a Determined in 10 mM PIPES, 10 mM MgCl2, 100 mM NaCl, pH 7.0.
b As compared to duplex 1. c As compared to duplex 2. d As compared to
duplex 9.
Mispairs involving the three most highly substituted ana-
logues, TM, TM2, and TMB, are also least stable in the case
of dC. However, unlike the less-substituted analogues, the
stabilities of the mispairs with dA are generally most stable,
although the effect is small with TM2. This may result from
specific minor groove interactions (TM, TM2, and TMB all
have a methyl group that is expected to be oriented into the
minor groove) and/or from desolvation. Interestingly, unlike the
unnatural base pairs discussed above, the stability of the mispairs
does not increase with increasing substitution; mispair stability
does not exceed ∼50 °C. Preferential stabilization of the
unnatural pairs, such as TM:TM2 and DM3:DM5, results in
thermal selectivities in excess of 8 °C, which compare favorably
with those seen among the natural base pairs. Even with this
remarkably simple framework, hydrophobic and van der Waals
forces are sufficient to endow the unnatural base pairs with
stability and thermal selectivity
corresponding dC:dC, DM5:dC, or dC:DM5 mispairs failed to
form duplex DNA under all conditions. We conclude that the
self-pair will be accommodated in an arbitrary sequence context
with reasonable stability and selectivity.
To further analyze BEN and DM5 self-pair formation, the
free energy change on duplex formation was calculated and
partitioned into enthalpic and entropic components by analysis
of the melting curves at different DNA concentrations39 (Table
5 and Supporting Information). The ∆G298K° values of duplex
formation are consistent with those expected from the Tm values.
The data imply that the reduced stability of the duplexes
containing the self-pairs does not result from entropic effects,
but rather from enthalpic effects. The BEN and DM5 self-pairs
both show a less favorable enthalpy change upon duplex
formation, relative to a natural dA:dT pair in the same sequence
context. This may result from the absence of H-bonds and/or
the significantly reduced surface area of the nucleobase ana-
logues. In contrast, self-pair formation is accompanied by a more
favorable (less negative) entropy change. Interestingly, the
entropy change is also responsible for the increased stability of
the DM5 self-pair relative to the BEN self-pair. This implies
that the classical hydrophobic effect contributes significantly
to the stability of the unnatural base pairs (see below).
2.5. Further Characterization of the DM5 Self-Pair. As
discussed above, several of the unnatural base pairs, including
the DM5 self-pair, are remarkably stable and selective in the
above sequence context. We were interested in determining the
generality of these characteristics and thus determined the Tm
values of DNA containing the DM5 self-pairs in a variety of
different sequence contexts (Table 3). The sequences included
multiple self-pairs separated by one to five natural pairs, or runs
of up to three contiguous self-pairs. While destabilization varied
between 1.9 and 5.1 °C per unnatural pair, relative to a dA:dT
pair, the melting curves of each duplex showed a well-defined
two-state transition (Supporting Information), and the CD
spectrum of each indicates a typical B-form duplex (Figure 5).
The duplex containing a triplet run is of particular interest. In
this case, the DNA duplex is only destabilized by 3.3 °C per
self-pair relative to an all-natural sequence, and even in this
case the self-pairs remain strongly selective against mispairing
with the natural bases (Table 4); the sequences containing the
3. Discussion
The storage and replication of biological information is based
on complementary interactions between the natural nucleobases,
dA with dT and dG with dC. Expansion of the genetic alphabet
by the introduction of a third base pair would enable a wide
variety of biotechnology applications based on improved or
(39) Marky, L. A.; Breslauer, K. J. Biopolymers 1987, 26, 1601-1620.
9
14424 J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004

Optimization of Hydrophobic Packing in DNA
A R T I C L E S
Figure 5. CD spectra of DNA duplexes containing multiple DM5 self-pairs. See text and Table 3 for details.
Table 4. Tm Values (°C) of Duplex with Multiple DM5 Self-Pairs
Table 5. Thermodynamic Parameters for Duplex Formation
and Mispairsa
∆
H
°
∆
S
°
∆G298K°
1
1
1
Tm
(
°
C)
(kcal mol-
)
(cal K-1 mol-
)
(kcal mol-
)
X ) A, Y ) T
X ) Y ) BEN
X ) Y ) DM5
59.2
52.8
57.3
-95.1
-91.6
-83.1
-259.7
-254.5
-225.5
-17.7
-15.7
-16.0
A wide variety of stable unnatural base pairs were identified;
however, they were generally not good substrates for DNA
polymerases. Some were efficiently synthesized (by incorpora-
tion of the unnatural triphosphate opposite the unnatural base
in the template), but generally the nascent unnatural pair was
not a good substrate for continued primer extension. We have
found that continued synthesis may be made more efficient by
introducing heteroatoms such as nitrogen, oxygen, or sulfur.4,10
However, preliminary structural studies indicate that pairs
formed between two large aromatic nucleobase analogues are
prone to interstrand intercalate, and we suspect that a similar
structural distortion at the primer terminus might inhibit
continued synthesis. It appeared that these first generation
unnatural base pairs achieved stability at the cost of replicability.
We have thus been interested in examining second generation
base pairs that are both modified with heteroatoms and also
too small to interstrand intercalate. However, it was unclear
whether this last design criterion could be accomplished while
simultaneously maintaining reasonable base pair stability and
thermal selectivity. To systematically evaluate this issue, we
synthesized and characterized a series of unnatural nucleosides
bearing simple methyl-substituted benzene rings as nucleobase
analogues.
a Determined in 10 mM PIPES, 10 mM MgCl2, 100 mM NaCl, pH 7.0.
b Cooperative melting behavior was not detected.
altered hybridization or replication properties, and also the
encoding of additional information for both in vitro and in vivo
applications. We have examined a wide variety of candidate
unnatural base pairs, including both self-pairs formed between
identical unnatural nucleotides and heteropairs formed between
two different unnatural nucleotides, whose interstrand inter-
actions are largely based on hydrophobic and van der Waals
forces. At the time of our first publication, no predominantly
hydrophobic base pairs had ever been found to stabilize duplex
formation. Thus, much of our initial efforts focused on the
design of unnatural nucleotides having a large aromatic surface
area to ensure base pair stability via intrastrand stacking
interactions.
3.1. Stability and Selectivity of the Unnatural Base Pairs.
Without substitution, the unnatural base pair formed between
two benzene rings is less stable than a typical mispair among
the natural bases (i.e., the Tm for a dT:dG mispair in the same
sequence context is 53.3 °C). This demonstrates that, as
9
J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004 14425

A R T I C L E S
Matsuda and Romesberg
expected, two simple benzene rings are not able to efficiently
fill the space occupied in the natural case by a purine-
pyrimidine pair. However, even the least stable unnatural pairs
are thermally selective, by almost 4 °C, against mispairing with
the natural bases. This selectivity is comparable to that observed
among the natural bases. With increasing substitution, the
unnatural base pairs become progressively more stable. Remark-
ably, the most stable base pairs, TM:TM2 and DM3:DM5, are
only ∼1.2 °C less stable than a dA:dT pair in the same sequence
context. In addition, several more of the duplexes are only
slightly less stable than natural duplex DNA. While the unnatural
pairs generally become more stable with substitution, the
mispairs show little increase in stability, resulting in thermal
selectivities of greater than 8 °C. This selectivity is actually
greater than that typically observed among natural nucleobases.
It is thus apparent that even base pairs with both reduced surface
area and no ability to H-bond can form stable and selective
pairs within duplex DNA.
Efforts to develop an unnatural base pair require that the base
pair be accommodated in DNA in a sequence-independent
manner. We thus characterized the stability of the DM5 self-
pair in a variety of sequence contexts, including those with
multiple self-pairs. While each duplex was synergistically
destabilized (the destabilization resulting from multiple substitu-
tions is not simply additive), each sequence still forms a B-form
duplex by CD spectroscopy, demonstrates a well-behaved two-
state transition during melting, and is significantly more stable
than the corresponding duplexes containing mispairs. This
implies, at least for DM5, that most sequence contexts should
selectively accommodate the self-pair.
3.2. Origins of Base Pair Stability. It is interesting to
speculate about how nucleobase analogues with such limited
surface area, and no ability to form H-bonds, can pair with
reasonable stability and thermal selectivity. In general, dipole
and dipole-induced dipole are all known to contribute to stable
intrastrand stacking in natural DNA.37 However, because the
unnatural analogues examined in this study have neither large
permanent dipole moments nor significant polarizability, it
seems unlikely that these electrostatic interactions contribute
significantly to their pairing upon duplex formation. The
hydrophobic effect is also thought to contribute to base stacking
interactions in natural DNA.37 In this case, the driving force
would result from the unnatural base having greater solvent
exposure in single-stranded DNA, and thus greater solvent
ordering (solvent surrounding hydrophobic moieties such as the
unnatural nucleobase is thought to become structured in a
fashion similar to ice clathrate formation37). Duplex formation
reduces the exposed hydrophobic surface area and thus decreases
the water ordering. This decrease in order increases the entropy
of the system, which should be apparent in the thermodynamic
data, if the hydrophobic effect contributes significantly to
unnatural base pair formation.40-43 When the entropy change
associated with formation of a dA:dT pair is compared to that
of BEN and DM5 self-pairs, it is apparent that unnatural base
pair formation is associated with a significantly more positive
entropy change. In addition, the larger effect associated with
the DM5 self-pair is consistent with a dominant contribution
from the hydrophobic effect, as in this case duplex formation
buries more hydrophobic surface area than with BEN. We thus
suggest that pairing of the unnatural nucleobases is mediated
predominantly by the hydrophobic effect.
In addition to the hydrophobic effect, specific intra- and
interstrand packing interactions appear to also contribute to
unnatural base pair stability. A structural model, based on CD
spectroscopy and molecular modeling, suggests that the un-
natural base pairs pack within an undistorted B-form duplex
and orient any ortho-substituent into the minor groove. This
structural model allows for the identification of the substituents
of one nucleobase analogue that are likely to be well packed
by flanking bases, and thus contribute to intrastrand packing,
or oriented toward the other nucleobases, and are thus likely to
contribute to “edge-edge” or interstrand packing. The contribu-
tion of these interactions is expected to be made manifest as a
dependence of the Tm values on the sequence (intrastrand
packing) or on the specific substituent pattern of the nucleobase
analogues (interstrand packing). The stability of the pairs with
BEN is illustrative. When BEN is positioned between purines
and the pairing base (packed by pyrimidines) is varied, the Tm
values change by less than 1 °C. However, when the pairing
base is packed by two purines, substitution results in Tm values
that vary by almost 3 °C. An origin of this sequence dependence
is suggested by the structural model of the DM5 self-pair (Figure
4). The model predicts that flanking pyrimidines are of
insufficient size to efficiently pack with multiple methyl group
substituents of the unnatural nucleobases. Thus, the favorable
hydrophobic packing may not be fully realized when the
unnatural nucleotide is flanked by a dC or a dT. This sequence
dependence and specific substitution pattern independence of
the pairs with BEN implies that the interface between the
unnatural nucleobases is not sufficiently developed to mediate
specific interstrand interactions.
When both pairing analogues are methyl-substituted, there
remains sequence dependence, especially when there is a large
discrepancy in the level of substitution; the pairs are generally
more stable when the more substituted base is flanked by
purines. However, the stability of the more highly substituted
unnatural base pairs also depends on the specific substitution
pattern. This is apparent from the variable stabilities of the
heteropairs formed between the di-, tri-, and tetra-substituted
analogues. The dependence on specific interactions is also
demonstrated by the stabilities of the self-pairs, which are by
definition sequence independent. For example, in the case of
the disubstituted analogues, the stabilities of the five self-pairs
vary by 3.6 °C. In addition, specific interactions must underlie
the stabilities of the TM:TM2 and DM3:DM5 pairs, which are
each significantly more stable than other pairs, including those
with more substituted analogues. Thus, at least when the
interstrand interface is developed with multiple methyl groups,
the stability and thermal selectivity appear to result from a
combination of the classical hydrophobic effect, intrastrand
packing, and specific interactions between the nucleobase
analogues. A well-packed interstrand interface that contributes
to unnatural base pair stability is supported by the energy-
minimized DM5 self-pair model structure, where the two DM5
surfaces approach van der Waals contacts. Despite their reduced
(40) Schweitzer, B. A.; Kool, E. T. J. Am. Chem. Soc. 1995, 117, 1863-1872.
(41) Guckian, K. M.; Morales, J. C.; Kool, E. T. J. Org. Chem. 1998, 63, 9652-
9656.
(42) Brotschi, C.; Haberli, A.; Leumann, C. J. Angew. Chem., Int. Ed. 2001,
40, 3012-3014.
(43) Brotschi, C.; Leumann, C. J. Angew. Chem., Int. Ed. 2003, 42, 1655-
1658.
9
14426 J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004

Optimization of Hydrophobic Packing in DNA
A R T I C L E S
surface area, the interstrand interactions of these unnatural base
pairs may be optimized for molecular recognition by the
judicious placement of methyl groups.
on unnatural base pair replication.45 We are now focused on
identifying combinations of methyl group substitutions and
heteroatom derivatizations that will impart these smaller un-
natural nucleobases with further improvements in stability and
replication.
The stability of the duplex containing three adjacent unnatural
base pairs merits additional comment. A significant component
of duplex stability in this case must derive from both inter- and
intrastrand packing of the unnatural nucleobases with each other.
Recently, Shionoya and colleagues elegantly demonstrated that
a run of five hydroxypyridone self-pairs may mediate duplex
formation via Cu2+ ligation.44 While duplex stability and
selectivity were not reported, it seems likely that, in addition
to interstrand interactions mediated by bidentate metal ligation,
interstrand packing interactions between the unnatural base also
contribute to stability in this case. Thus, favorable intrastrand
packing among small predominantly hydrophobic moieties in
DNA may be general, and, when combined with suitable
interstrand interactions, for example, those mediated by metal
chelation or by hydrophobic packing, runs of unnatural base
pairs may form extended, stable duplexes.
3.3. Conclusions and Implications for the Expansion of
the Genetic Alphabet. It is well known that increased aromatic
surface area may stabilize duplex formation. However, no
studies involving the systematic optimization of interstrand
packing interactions had been reported, so surface area and
H-bonding requirements remained unclear. We have now shown
that unnatural nucleotides bearing simple methyl-substituted
benzene rings as nucleobase analogues may form base pairs
that are virtually as stable as natural pairs, although they possess
neither H-bonds nor large aromatic surface area. In fact, these
unnatural base pairs are stable and selective despite having
significantly less surface area than the natural base pairs. The
molecular recognition of these base analogues may be tuned
and optimized by the judicious substitution of methyl groups.
These smaller nucleobases are not expected to induce distortions
in duplex DNA, including those at the primer terminus that seem
to limit replication of the larger unnatural base pairs. It seems
likely that these nucleobase analogues will serve as scaffolds
for the development of unnatural base pairs with further
improved properties. For example, heteroatom substitution
patterns should exist that further stabilize the unnatural pairs
by dipole and induced-dipole interactions. In fact, fluorine
substitution has recently been shown to have a pronounced effect
4. Experimental Section
Oligonucleotide Synthesis. Oligonucleotides were prepared by the
â-cyanoethylphosphoramidite method on controlled pore glass supports
(1 µmol) by using an Applied Biosystems, Inc. 392 DNA/RNA
synthesizer as the standard method. After automated synthesis, the
oligomers were cleaved from the support by concentrated aqueous
ammonia for 1 h, deprotected by heating at 55 °C for 12 h, and purified
by polyacrylamide gel electrophoresis. The concentration of olig-
odeoxynucleotides was determined by UV.
Circular Dichroism Measurements. CD experiments were per-
formed with an Aviv model 61 DS spectropolarimeter equipped with
a Peltier thermoelectric temperature control unit (3 µM strand concen-
tration, 10 mM PIPES buffer, pH 7.0, 100 mM NaCl, and 10 mM
MgCl2). The data were collected using a 1 cm path length quartz cuvette
with scanning from 360 to 220 nm, a time constant of 3 s, and a
wavelength step size of 0.5 nm, at 25 °C. CD data were transformed
into molar ellipticity in the units of deg cm2/dm of monomer subunits.
Duplex Melting Temperature Measurements. UV melting experi-
ments were carried out by means of a Cary 200 Bio UV-visible
spectrometer. The absorbance of the sample (3 µM strand concentration,
10 mM PIPES buffer, pH 7.0, 100 mM NaCl, and 10 mM MgCl2) was
monitored at 260 nm from 21 to 80 °C at a heating rate of 0.5 °C per
min. Melting temperatures were determined from the derivative method
using the Cary Win UV thermal application software.
Modeling Studies. Proposed duplex structures containing the DM5
self base pair were obtained by molecular modeling of the 13-mer
duplex 5′-d(GCGTACXCATGCG)-3′/5′-d(CGCATGYGTACGC)-3′ (X
) Y ) DM5). The initial structure of the duplex was obtained by
manually docking DM5 to the sugar C1′ position. Energy minimization
of the duplex was carried out by using an Amber* force field with
GB/SA solvation treatment of water. Each self-pair orientation
(anti:anti, anti:syn, syn:anti, and syn:syn) was evaluated.
Acknowledgment. Funding was provided by the NIH (GM
60005) and the JSPS Research Fellowships for Young Scientists
(S.M.).
Supporting Information Available: Details of the synthetic
procedures, characterizations, melting curves, and plots used
for the determination of thermodynamic parameters. This
material is available free of charge via the Internet at
http://pubs.acs.org.
(44) Tanaka, K.; Tengeiji, A.; Kato, T.; Toyama, N.; Shionoya, M. Science 2003,
299, 1212-1213.
(45) Henry, A. A.; Olsen, A. G.; Matsuda, S.; Yu, C.; Geierstanger, B. H.;
Romesberg, F. E. J. Am. Chem. Soc. 2004, 126, 6923-6931.
JA047291M
9
J. AM. CHEM. SOC. VOL. 126, NO. 44, 2004 14427