Naphthalene- and perylene-based linkers for the stabilization of hairpin triplexes

Article abstract of DOI:10.1021/ja0001714

Planar perylene- and naphthalene-based diimide linkers can be employed to tether the Watson-Crick and the Hoogsteen strands of a DNA triplex, thus providing conjugates capable of targeting single-stranded nucleic acids with the formation of hairpin triple

Full text of DOI:10.1021/ja0001714

VOLUME 122, NUMBER 25
J UNE 28, 2000
© Copyright 2000 by the
American Chemical Society
Naphthalene- and Perylene-Based Linkers for the Stabilization of
Hairpin Triplexes
Susan Bevers, Susan Schutte, and Larry W. McLaughlin*
Contribution from the Department of Chemistry, Merkert Chemistry Center, Boston College,
2609 Beacon Street, Chestnut Hill, Massachusetts 02467
ReceiVed January 18, 2000
Abstract: Planar perylene- and naphthalene-based diimide linkers can be employed to tether the Watson-
Crick and the Hoogsteen strands of a DNA triplex, thus providing conjugates capable of targeting single-
stranded nucleic acids with the formation of hairpin triplexes. The planar linkers are designed to bridge the
terminal base triplet of the three-stranded complex and provide base-stacking interactions with all three residues.
Sixteen complexes have been prepared, eight with each linker, half with RNA (R) targets and half with DNA
(D) targets. The conjugate sequences are composed of two strands of DNA, two of 2′-O-methyl RNA (M), or
one of each. In comparison to similar complexes formed with a hexa(ethylene glycol) linker, the planar linkers
enhance the TM values for the complexes by as much as 28 °C with ∆G values indicating as much as 12.3
kcal/mol of stabilization relative to the simple glycol linker. All sixteen complexes have been characterized by
TM measurements and ∆G determinations. That π-stacking interactions are present between the linkers, and
the nucleobases can be inferred from the quenching of the perylene fluorescence upon complex formation,
and the observation of an absorbance vs temperature transition for the naphthalene-based linker at 383 nm and
for the perylene-based linker monitored at 537 nm.
Introduction
tribute to an increased lifetime for the antisense or antigene
complexes designed for gene-specific regulation of protein
expression.
Conventional DNA triplexes are formed from a target duplex
and a third single-stranded sequence.1-3 Single-stranded se-
quences themselves can also be targeted by the triplex motif
with sequences that can hybridize to the target by Watson-
Crick base pairs and Hoogsteen base pairs.4 In such complexes
the two hybridizing arms are tethered either by a short
nonbinding stretch of nucleotides,5 or by a simple linker.4,6
Ligands have been used in numerous studies to target
macromolecules such as proteins and nucleic acids. Site-specific
binding by a ligand permits it to compete effectively with
biological targets and thereby interfere with specific biological
processes; nucleic acid therapeutics based upon an antisense or
an antigene approach are examples of such targeted interference.
A number of small-molecule ligands are known to bind to
nucleic acid structures either by intercalation, groove binding,
or other less well-characterized formats. In many cases such
binding interactions function to over-stabilize duplex, triplex,
or tetraplex structures, and such over-stabilization may con-
(1) Cooney, M.; Czernuszewicz, G.; Postel, E. H.; Flint, S. J.; Hogan,
M. E. Science 1988, 241, 456-459.
(2) LeDoan, T.; Perrouault, L.; Praseuth, D.; Habhoub, N.; Decout, J.
L.; Thoung, N. T.; Lhomme, J.; Helene, C. Nucleic Acids Res. 1987, 15,
7749-7760.
(3) Moser, H. E.; Dervan, P. B. Science 1987, 238, 645-650.
(4) Giovannangeli, C.; Montenaygarestier, T.; Rougee, M.; Chassignol,
M.; Thuong, N. T.; Helene, C. J. Am. Chem. Soc. 1991, 113, 7775-7777.
* To whom correspondence should be addressed: Telephone: (617) 552-
3622. Fax: (617) 552-2705. E-mail. larry.mclaughlin@bc.edu.
10.1021/ja0001714 CCC: $19.00 © 2000 American Chemical Society
Published on Web 06/03/2000

5906 J. Am. Chem. Soc., Vol. 122, No. 25, 2000
BeVers et al.
When both termini of the hybridizing sequence are connected,
circular oligonucleotides result7 that can very effectively bind
single-stranded target sequences.8 Enhanced stabilization of
related complexes also results when two independently hybrid-
izing sequences are tethered by a short linker.3-5
intercalated into a DNA duplex and preferred to intercalate into
steps containing at least one G:C base pair.27
Earlier work with perylene derivatives has largely concen-
trated on spectral and fluorescence characterizations of the
molecule.28-30 Spectral characteristics of the naphthalene and
perylene chromophores have been compared.31,32 Red pig-
mentary polyimides have been synthesized from N,N′-diamino-
3,4,9,10-perylenetetracarboxylic acid bisimide.33 The N,N′-bis-
[3,3′-(dimethylamino)propylamine]-3,4,9,10-perylene tetracar-
boxylic diimide (a dicationic dye) has been used for precipitation
and quantitation of DNA,34 and recently a related perylene
tetracarboxylic diimide-based ligand was shown to bind G-
tetraplex and to inhibit human telomerase.35
In preliminary communications we have reported that tethered
derivatives of naphthalene and perylene diimides can function
to stabilize DNA duplexes36 and triplexes.36-38 In the present
work we report on the synthesis and stability of sixteen different
hairpin triplexes containing either tethered naphthalene or
perylene conjugates, as well as the fluorescence characteristics
of perylene-based DNA conjugates.
Enhanced triplex stabilization can also be achieved through
additional binding interactions from a ligand that is covalently
tethered to one of the macromolecular members of the DNA
triplex.9-12 Such covalent attachment usually results in signifi-
cant proximity effects and provides, in part, entropic-based
stabilization of the complex. A variety of ligands have been
tethered to nucleic acids, and these include intercalators9,13-15
and groove-binding agents,10,11,16 as well as metal complexes.17-19
Although the nature of the linker tethering the ligand to the
nucleic acid sequence varies considerably, the site of attachment
is usually one of the two nucleic acid termini, a nucleobase
residue, or the backbone of the oligomer.
Previous studies have shown that naphthalene diimide chro-
mophores function as effective intercalators20,21 even polyinter-
calators for DNA. For example, L-lysine has been derivatized
at it’s ꢀ-amino group, by attachment of a 1,8-naphthalimide
moiety; this conjugate functions as a site specific DNA-cleaving
agent.22,23 Naphthalene diimide derivatives have additionally
been studied for properties related to their redox activity.24,25
A 1,4,5,8-naphthalene tetracarboxylic diimide chromophore
attached in a head-to-tail arrangement by peptide linkers
was suggested to be fully intercalated into DNA through the
major groove.26 Binding studies of a N-[3-(dimethylamino)-
propylamine]-1,8-naphthalene dicarboxylic imide and N,N′-bis-
[3,3′-(dimethylamino)propylamine]-naphthalene-1,4,5,8-tetracar-
boxylic diimide indicated that both of these compounds
Experimental Section
NMR spectra were determined on Varian spectrophotometers. Mass
spectra were obtained using FAB ionization from the Mass Spectrom-
etry Laboratory, School of Chemical Sciences, University of Illinois,
Urbana, IL. MALDI-TOF mass spectral analyses of the oligonucleotides
were performed by Dr. D. J. Fu, Sequenome, San Diego, CA. Rotary
evaporations were performed under reduced pressure with Buchi
systems. Thin-layer chromatography (TLC) was performed on Silica
Gel 60 F254 precoated on aluminum sheets (EM Separations Technol-
ogy), in CH3OH:CH2Cl2 (1:19) unless otherwise specified. All phos-
phitylation reactions were monitored with aluminum oxide IBF TLC
sheets from JT Baker Inc. Flash column chromatography was performed
on Silica Gel 60, Particle size 0.040-0.063 (EM Separations Technol-
ogy), using a CH3OH gradient in CH2Cl2 unless otherwise specified.
CH2Cl2 and diisopropyl ethylamine were dried by heating over calcium
hydride under reflux for several hours followed by distillation. Other
anhydrous solvents and starting materials were purchased from Aldrich
Chemical Co. and used without further purification unless otherwise
specified. UV scans and absorbances were obtained using a Beckman
DU 640 spectrophotometer. Oligodeoxynucleotides were synthesized
on an Applied Biosystems 381A DNA Synthesizer. 2′-Deoxynucleotide
phosphoramidites and 3′-terminal nucleoside-controlled pore glass
support (CPG) were purchased from Glenn Research (Sterling, VA).
The hexaethylene glycol linker, prepared as a 4,4′-dimethoxytrityl
â-cyanoethyl phosphoramidite derivative was synthesized as described
elsewhere.36 High performance liquid chromatography (HPLC) was
performed on a Beckman HPLC system using C-18 reversed-phase
columns (ODS-Hypersil, 5 mm particle size, 120A pore) with detection
at 260 nm. Oligonucleotides were desalted with Econo-Pac 10DG
(5) Wang, S. H.; Booher, M. A.; Kool, E. T. Biochemistry 1994, 33,
4639-4644.
(6) Rumney, S.; Kool, E. T. J. Am. Chem. Soc. 1995, 117, 5635-5646.
(7) Kool, E. T. J. Am. Chem. Soc. 1991, 113, 6265-6266.
(8) Prakash, G.; Kool, E. T. J. Chem. Soc., Chem. Commun. 1991, 1161-
1163.
(9) Collier, D. A.; Thuong, N. T.; Helene, C. J. Am. Chem. Soc. 1991,
113, 1457-1458.
(10) Lukhtanov, E. A.; Podyminogin, M. A.; Kutyavin, I. V.; Meyer, R.
B.; Gamper, H. B. Nucleic Acids Res. 1996, 24, 683-687.
(11) Robles, J.; McLaughlin, L. W. J. Am. Chem. Soc. 1997, 119, 6014-
6021.
(12) Silver, G. C.; Sun, J. S.; Nguyen, C. H.; Boutorine, A. S.; Bisagni,
E.; Helene, C. J. Am. Chem. Soc. 1997, 119, 263-268.
(13) Sun, J.-S.; Francois, J. C; Montenay-Garestier, T.; Saison-Beh-
moaras, T.; Roig, v.; Thuong, N. T.; Helene, C. Proc. Natl. Acad. Sci. U.S.A.
1989, 88, 6023-6027.
(14) Stonehouse, T. J.; Fox, K. R. Biochim. Biophys. Acta 1994, 1218,
322-330.
(15) Silver, G. C.; Sun, J. s.; Nguyen, C. H.; Boutorine, A. S.; Helene,
C. J. Am. Chem. Soc. 1997, 119, 263-268.
(16) Rajur, S. B.; Robles, J.; Wiederholt, K.; Kuimelis, R. W.; McLaugh-
lin, L. W. J. Org. Chem. 1997, 62, 523-529.
(27) Liu, Z. R.; Hecker, K. H.; Rill, R. L. J. Biomol. Struct. Dyn. 1996,
14, 331-339.
(17) Lewis, F. D.; Helvoigt, S. A.; Letsinger, R. L. J. Chem. Soc., Chem.
Commun. 1999, 327-328.
(28) Icli, S.; Icil, H. Spectrosc. Lett. 1994, 27, 323-332.
(29) Bisht, P. B.; Fukuda, K.; Hirayama, S. J. Phys. Chem. B 1997, 101,
8054-8058.
(18) Wiederholt, K.; McLaughlin, L. W. Nucleic Acids Res. 1999, 27,
2487-2493.
(30) Langhals, H.; Jona, W. Angew. Chem., Int. Ed. 1998, 37, 952-
955.
(19) Connors, P. J.; Tzalis, D.; Dunnick, A. L.; Tor, Y. Z. Inorg. Chem.
1998, 37, 1121-1123.
(31) Adachi, M.; Murata, Y.; Nakamura, S. J. Phys. Chem. 1995, 99,
14240-14246.
(32) Ghassemi, H.; Zhu, J. H. J. Polym. Sci., Part B: Polym. Phys. 1995,
33, 1633-1639.
(33) Ghassemi, H.; Hay, A. S. Macromolecules 1994, 27, 4410-4412.
(34) Liu, Z. R.; Rill, R. L. Anal. Biochem. 1996, 236, 139-145.
(35) Fedoroff, O. Y.; Salazar, M.; Han, H.; Chemeris, V. V.; Kerwin, S.
M.; Hurley, L. H. Biochemistry 1998, 37, 12367-12374.
(36) Bevers, S.; Odea, T. P.; McLaughlin, L. W. J. Am. Chem. Soc. 1998,
120, 11004-11005.
(37) Gianolio, D. A.; McLaughlin, L. W. J. Am. Chem. Soc. 1999, 121,
6334-6335.
(38) Gianolio, D.; Segismundo, J.; McLaughlin, L. W. Nucleic Acids
Res. 2000, 28, 2128-2134.
(20) Gabbay, E. J.; DeStefano, R.; Baxter, C. S. Biochem. Biophys. Res.
Commun. 1973, 51, 1083-1089.
(21) Yen, S. F.; Gabbay, E. J.; Wilson, W. D. Biochemistry 1982, 21,
2070-2076.
(22) Saito, I.; Takayama, M.; Sugiyama, H.; Nakatani, K. J. Am. Chem.
Soc. 1995, 117, 6406-6407.
(23) Saito, I.; Takayama, M. J. Am. Chem. Soc. 1995, 117, 5590-5591.
(24) Lewis, F. D.; Zhang, Y. F.; Liu, X. Y.; Xu, N.; Letsinger, R. L. J.
Phys. Chem. B 1999, 103, 2570-2578.
(25) Matsugo, S.; Kawanishi, S.; Yamamoto, K.; Sugiyama, H.; Matsuura,
T.; Saito, I. Angew. Chem., Int. Ed. Engl. 1991, 30, 1351-1353.
(26) Lokey, R. S.; Kwok, Y.; Guelev, V.; Pursell, C. J.; Hurley, L. H.;
Iverson, B. L. J. Am. Chem. Soc. 1997, 119, 7202-7210.

Naphthalene and Perylene Hairpin Triplexes
J. Am. Chem. Soc., Vol. 122, No. 25, 2000 5907
disposable chromatography Columns (Bio Rad, Hercules, CA). TM
measurements were performed on an AVIV Spectrophotometer, model
14DS UV/vis.
overnight. TLC analysis indicated the absence of starting material and
formation of new products. Pyridine was removed in vacuo. The mixture
was dissolved in dichloromethane (50 mL) and was washed with an
aqueous solution of saturated sodium bicarbonate (2 × 50 mL). The
organic layer was dried over anhydrous sodium sulfate. The mixture
was then vacuum filtered, and the filter was washed with dichloro-
methane. The product was purified by flash chromatography to yield
3.5 g (16 mmol) of 4 as a brown oil (36% yield). Rf (MeOH:CH2Cl2,
1:9) ) 0.80. 1H NMR (CDCl3): δ ) 0.00 (s, 6H, SiCH3), 0.83 (s, 9H,
-CH3), 2.77 (m, 2H, -CH2-), 3.44 (t, 2H, -CH2-), 3.70 (m, 4H,
-CH2-) ppm. 13C NMR (CDCl3): δ ) 3.3, 27.0, 34.4, 34.5, 50.4,
71.3, 81.0, 81.8 ppm. HRMS: calcd for C10H25NO2Si (M + H+)
220.1733, found 220.1734.
N,N′-Bis-2-[(2-O-tert-butyldimethylsilylethoxy)ethyl]-3,4,9,10-
perylenetetracarboxylic diimide (5). To 2.0 g (5.1 mmol) perelyne-
tetracarboxylicdianhydride suspended in anhydrous pyridine (25 mL)
was added 3.4 g (15 mmol) 2-(2-O-tert-butyldimethylsilylethoxy)-
ethylamine (4) and 1.1 g (5.1 mmol) zinc acetate dihydrate. The reaction
mixture was refluxed overnight. TLC analysis indicated the absence
of starting material and the formation of products. Pyridine was removed
in vacuo. The mixture was suspended in chloroform and loaded onto
a large silica column. The product was purified by flash chromatography
to yield 3.4 g (4.4 mmol) of 5 as a dark red solid (86% yield). Rf
(methanol:dichloromethane, 1:9) ) 0.90. 1H NMR (CDCl3): δ ) 0.00
(s, 12H, SiCH3), 0.83 (s, 18H, -CH3), 3.61 (t, 4H, -CH2-), 3.68 (t,
4H, -CH2-), 3.83 (t, 4H, -CH2-), 4.44 (t, 4H, -CH2-), 8.60 (q,
8H of Ar H) ppm. 13C NMR (CDCl3): δ ) -4.2, 19.7, 27.0, 40.5,
63.8, 69.0, 73.3, 123.6, 123.9, 126.8, 129.0, 131.8, 134.7, 163.9 ppm.
UV (chloroform): λmax 524, 489, 457, 265, 243. HRMS: calcd for
C44H54N2O8Si2 (M + H+) 795.3490, found 795.3500.
N,N′-Bis 2-(2-hydroxyethoxy)ethyl-1,4,5,8-naphthalenetetracar-
boxylic diimide (1). To 1.0 g (3.7 mmol) 1,4,5,8-naphthalenetetracar-
boxylic dianhydride suspended in anhydrous pyridine (25 mL) was
added 1.6 g (15 mmol) 2-(2-aminoethoxy)ethanol and 1.0 g (4.6 mmol)
zinc acetate dihydrate. The mixture was refluxed overnight. TLC
analysis indicated the absence of starting material and the presence of
a new product. Pyridine was removed in vacuo. The residue was
suspended in dichloromethane, and a small amount of methanol was
added until the solids dissolved. Silica gel was added to the solution,
and the solvents were removed in vacuo. The silica-absorbed material
was poured onto a short column of silica gel dissolved in dichloro-
methane and eluted with 0-15% methanol. The fastest moving band
corresponded to the diimide product, and 1.16 g (2.6 mmol) of 1 was
obtained (70% yield). Rf: 0.14. 1H NMR (DMSO-d6) δ ) 3.67 (m, 8H,
-CH2-), 4.22 (t, 4H, -CH2-) 4.56 (t, 4H, -CH2), 8.60 (s,4H of Ar
H) ppm. 13C NMR (DMSO-d6) δ 49.4, 69.3, 81.2, 134.8, 135.1, 138.9,
171.5 ppm. UV (dichloromethane:methanol, 9/1): λmax 377, 359, 241
nm. HRMS: calcd for C22H22N2O8 (M + H+) 443.1454, found
443.1454.
N-2-[(2-O-4,4′-dimethoxytritylethoxy)ethyl]-N′-2-(2-hydroxyethoxy)-
ethyl- 1,4,5,8-naphthalenetetracarboxylic diimide (2). To 1.1 g (2.5
mmol) of N,N′-bis-2-(2-hydroxyethoxy)ethyl-1,4,5,8-naphthalenetetra-
carboxylic diimide (1) dissolved in anhydrous pyridine (25 mL) was
added 0.89 g (2.6 mmol) 4,4′-dimethoxytrityl chloride. The reaction
mixture was stirred for 3 h. TLC analysis at this time indicated the
disappearance of 1 and the presence of two products. The reaction was
stopped by adding several drops of methanol followed by the removal
of solvents in vacuo. The mixture was dissolved in dichloromethane
(50 mL) and was washed with an aqueous solution of saturated sodium
bicarbonate (2 × 50 mL). The organic layer was dried over anhydrous
sodium sulfate. The mixture was then vacuum filtered, and the filter
was washed with dichloromethane. The mono(dimethoxytrityl) product
was purified by flash chromatography using 1% triethylamine/dichlo-
romethane and a methanol gradient. After evaporation of appropriate
fractions, this material was precipitated into cold hexane to remove
residual triethylamine. The precipitate was filtered and washed with
cold hexane. The product was rinsed from the filter with dichloro-
methane and was concentrated to a yellow foam to yield 0.61 g (0.82
mmol) of 2 (32% yield). Rf ) 0.37. A faster moving compound was
also obtained, and it appeared to be the bis(dimethoxytrityl) product.
Rf ) 0.71. 1H NMR (CDCl3) δ ) 3.18 (t, 2H, -CH2-), 3.66 (m, 6H,
-CH2-), 3.75 (s, 6H, OCH3), 3.88 (t, 4H, -CH2-), 4.49 (m, 4H,
-CH2-), 6.69 (d, 4H, Ar H), 7.1-7.4 (m, 9H, Ar H), 8.7 (s, 4H, Ar
H) ppm. 13C NMR (CDCl3) δ 48.5, 48.6, 63.8, 70.4, 71.6, 76.5, 76.8,
79.1, 80.9, 94.4, 121.5, 135.0, 135.2, 135.2, 135.3, 136.3, 136.8, 138.6,
139.7, 144.8, 153.4, 166.8, 171.5, 171.8 ppm. UV (dichloromethane:
methanol, 9/1): λmax 379, 359, 241 nm. HRMS: calcd for C43H40N2O10
(M + H+) 744.2683, found 744.2682.
N-2-[(2-O-tert-Butyldimethylsilylethoxy)ethyl]-N′-2-(2-hydroxy-
ethoxy)ethyl-3,4,9,10,-perylenetetracarboxylic diimide (6). To 3.4
g (4.2 mmol) N,N′-bis-2-[(2-O-tert-butyldimethylsilylethoxy)ethyl]-
3,4,9,10-perylenetetracarboxylic diimide (5) dissolved in anhydrous
pyridine (50 mL) was added HF/pyridine (70%) until approximately
half of the starting material had lost one silyl-protecting group as
monitored by TLC analysis (loss of both silyl-protecting groups resulted
in an insoluble precipitate). Pyridine was removed in vacuo. The mixture
was resuspended in chloroform and was purified by flash chromatog-
raphy to yield 0.69 g (1.0 mmol) of 6 as a dark red solid (24% yield).
1
Rf (methanol:dichloromethane, 1:9) ) 0.43. H NMR (CDCl3): δ )
0.00 (s, 6H, SiCH3), 0.83 (s, 9H, -CH3), 3.60-3.80 (m, 8H,
-CH2-), 3.90 (m, 4H, -CH2-), 4.44 (m, 4H, -CH2-), 8.0-9.0 (m,
8H of Ar H) ppm. UV (chloroform): λmax 525, 489, 458, 263, 241.
HRMS: calcd for C38H40N2O8Si (M + H+), 681.2632, found 681.2634.
N-2-[(2-O-4,4′-Dimethoxytritylethoxy)ethyl]-N′-2-[(2-O-tert-bu-
tyldimethyl-silylethoxy)ethyl]-3,4,9,10,-perylenetetracarboxylic di-
imide (7). To 0.67 g (1.1 mmol) N-2-[(2-O-tert-butyldimethylsilylethoxy)-
ethyl]-N′-2-(2-hydroxyethoxy)ethyl-3,4,9,10,-perylenetetracarboxylic
diimide (6) dissolved in anhydrous pyridine (10 mL) was added 0.58
g (1.7 mmol) 4, 4′-dimethoxytrityl cholride. The reaction was stirred
for 3 h. TLC analysis indicated the absence of starting material and
the formation of a new product. The reaction was stopped with 1 mL
of methanol. The solvents were removed in vacuo. The mixture was
dissolved in chloroform to which 2 drops of triethylamine had been
added. The mixture was purified by flash chromatography in 1%
triethylamine/chloroform with a methanol gradient to yield 0.75 g (0.76
mmol) of 7 as a dark red solid (69% yield). Rf (methanol:dichloro-
N-2-[(2-O-4,4′-dimethoxytritylethoxy)ethyl]-N′-2-{[2-O-(2-cyano-
ethyl diisopropyl phosphino)ethoxy]ethyl}-1,4,5,8-naphthalene-
tetracarboxylic diimide (3). To 30 mg (0.040 mmol) of N-2-[(2-O-
4,4′-dimethoxytritylethoxy)ethyl]-N′-2-(2-hydroxyethoxy)ethyl-1,4,5,8-
naphthalenetetracarboxylic diimide (2) dissolved in anhydrous di-
chloromethane (1 mL at 0 °C) was added 0.036 mL (0.20 mmol)
diisopropylethylamine followed by 0.018 mL (0.080 mmol) 2-cyano-
ethyl diisopropylchlorophosphoramidite. The reaction was slowly
brought to 25 °C and stirred for 30 min. TLC analysis indicated the
absence of 2 and the presence of a new product. The reaction was
stopped with 2-3 drops of methanol, and the phosphitylated product
was precipitated in cold hexane (0 °C). The product was filtered, washed
with cold hexane, and rinsed from the filter with dichloromethane. The
product was concentrated to yield 0.026 g (0.028 mmol) of 3 as a yellow
foam (70% yield). Rf ) 0.90. 31P NMR (CDCl3, H3PO4 external
reference) 148 ppm.
1
methane, 1:9) ) 0.87. H NMR (CDCl3): δ ) 0.00 (s, 6H, SiCH3),
0.83 (s, 9H, -CH3), 3.60-3.80 (m, 8H, -CH2-), 3.80 (s, 6H, OCH3),
3.90 (m, 4H, -CH2-), 4.44 (m, 4H, -CH2-), 6.72 (d, 2H, ArH), 6.83
(d, 2H, ArH), 7.1-7.4 (m, 9H, ArH), 8.25 (m, 4H, Ar H), 8.45 (m,
4H, Ar H) ppm. 13C NMR (CDCl3): δ ) -4.1, 19.7, 27.0, 40.7, 56.2,
56.3, 63.9, 64.3, 69.0, 69.3, 71.5, 73.3, 86.5, 86.8, 114.0, 114.2, 123.5,
123.5, 123.8, 123.8, 127.6, 128.7, 128.8, 128.9, 129.3, 130.2, 131.1,
131.7, 131.7, 134.5, 137.4, 137.4, 146.1, 159.3, 163.9 ppm. UV
(chloroform): λmax 525, 489, 458, 263, 243. MS: calcd for C38H39N2O8-
Si (M+), 982, found 619 (M+ - DMT-OCH2CH2O-), 807 (M+
-
2-(2-O-tert-Butyldimethylsilylethoxy)ethylamine (4). To 5.0 g (48
mmol) of 2-(2-aminoethoxy)ethanol dissolved in anhydrous pyridine
(50 mL) was added 7.2 g (48 mmol) tert-butyldimethylsilyl chloride
and 4.9 g (72 mmol) of imidazole. The reaction mixture was stirred
tBDMS-OCH2CH2O-), 444 (M+ - DMT-OCH2CH2O and -
tBDMS-OCH2CH2O-), 303 (DMT).
N-2-[(2-O-4,4′-Dimethoxytritylethoxy)ethyl]-N′-2-(2-hydroxy-

5908 J. Am. Chem. Soc., Vol. 122, No. 25, 2000
BeVers et al.
ethoxy)ethyl-3,4,9,10,-perylenetetracarboxylic diimide (8). To 0.75
g (0.82 mmol) N-2-[(2-O-4,4′-dimethoxytritylethoxy)ethyl]-N′-2-[(2-
O-tert-butyldimethylsilylethoxy)ethyl] 3,4,9,10,-perylenetetracarboxylic
diimide (7) dissolved in anhydrous pyridine (20 mL) was titrated HF/
pyridine (70%) until the starting material had been transformed as
monitored by TLC analysis. Pyridine was removed in vacuo. The
mixture was resuspended in chloroform and was purified by flash
chromatography in in 1% triethylamine/chloroform with a methanol
gradient to yield 0.36 g (0.41 mmol) of 8 as a dark red solid (50%
yield). Rf (methanol:dichloromethane, 1:19) ) 0.25. 1H NMR
(CDCl3): δ ) 3.10 (m, 2H, CH2) 3.19 (t, 2H, CH2) 3.68 (m, 4H,
-CH2-), 3.78 (s, 6H, OCH3), 3.92 (t, 4H, -CH2-), 4.50 (m, 4H,
-CH2-), 6.72 (d, 2H, ArH), 6.83 (d, 4H, ArH), 7.1-7.4 (m, 9H, ArH),
8.45 (m, 4H of Ar H), 8.60 (m, 4H of Ar H) ppm. 13C NMR (CDCl3):
δ ) 40.7, 40.9, 56.2, 63.0, 64.3, 69.3, 69.4, 71.5, 73.5, 86.8, 123.5,
123.6, 123.7, 123.8, 126.3, 127.6, 128.7, 129.3, 129.6, 129.7, 131.1,
131.8, 131.9, 134.4, 134.74 137.4, 146.0, 159.3, 163.9, 164.3 ppm.
UV (chloroform): λmax 525, 489, 458, 263, 243. HRMS: calcd for
C53H44N2O10 (M + H+) 869.3074, found 869.3071.
N-2-[(2-O-4,4′-Dimethoxytritylethoxy)ethyl]-N′-2-{[2-O-(2-cyano-
ethyl diisopropylchlorophosphino)ethoxy]ethyl}-3,4,9,10,-perylene-
tetracarboxylic diimide (9). To 40 mg (0.055 mmol) N-2-[(2-O-4,4′-
dimethoxytritylethoxy)ethyl]-N′-2-(2-hydroxyethoxy)ethyl-3,4,9,10-
perylenetetracarboxylic diimide 8 dissolved in anhydrous dichloromethane
(1 mL at 0 °C) was added 0.048 mL (0.28 mmol) diisopropylethylamine
followed by 0.025 mL (0.11 mmol) 2-cyanoethyl diisopropylchloro-
phosphoramidite. The reaction was slowly brought to 25 °C and stirred
for 30 min. TLC analysis indicated the absence of starting material
and the presence of a new product. The reaction was stopped with 2-3
drops of methanol, and the phosphitylated product was precipitated
into cold hexane (0 °C). The product was filtered, washed with cold
hexane and rinsed from the filter with dichloromethane. The product
was concentrated to yield 0.035 g (0.036 mmol) of 9 as a red solid
(65% yield). Rf ) 0.67. 31P NMR (CDCl3, H3PO4 external reference)
149 ppm.
Synthesis of Oligonucleotides Containing Linkers. DNA sequences
were assembled using standard phosphoramidite protocols. The hexa-
ethylene glycol and naphthalene linkers were incorporated on the DNA
synthesizer by increasing the coupling time of the corresponding
phosphoramidite from 30 s to 10 min. Oligonucleotides containing the
perylene linker were assembled on the synthesizer up to the site of the
linker. The final 4,4′-dimethoxytrityl (DMT)-protecting group was
removed from the 5′-hydroxyl of the base 3′ to the linker and at this
point the synthesis was interrupted. The CPG containing the oligo-
nucleotide was removed from the synthesizer, dried in vacuo, and
poured into a small flask. To the CPG was added 30 µmol of the
perylene linker phosphoramidite (9) dissolved in 0.50 mL of anhydrous
dichloromethane and 0.50 mL of 0.5 M tetrazole in acetonitrile. The
reaction was mixed in an argon atmosphere for 1 h. The CPG was
then filtered and rinsed with dichloromethane to remove unincorporated
phosphoramidite. The CPG was replaced in the column which was
reattached to the synthesizer and the synthesis was resumed at the
capping step. This procedure was necessary owing to the moderate
solubility characteristics of perylene derivative in acetonitrile. The base
5′ to the perylene linker was dissolved in anhydrous dichloromethane
and incorporated by increasing the coupling time from 30 s to 10 min.
The remaining bases were added on by standard methods with
phosphoramidites dissolved in acetonitrile.
The terminal DMT groups were removed from the oligonucleotides
containing the naphthalene and perylene linkers. The sequences were
deprotected in concentrated methanolic ammonia at 55 °C overnight
(perylene-based conjugates were typically deprotected in concentrated
aqueous ammonia at 55 °C overnight). The oligonucleotides were
purified by denaturing polyacrylamide gel electrophoresis using 20%
polyacrylamide/1% bisacrylamide/7 M urea gels in 89 mM tris-borate
buffer containing 2 mM Na2EDTA, pH 8 (1× TBE buffer). The bands
were visualized by UV shadowing (and by the obvious colored
properties of the linkers) and excised. The pure oligonucleotides were
electroeluted in 0.5× TBE buffer using a Schleicher and Schuell (Keene,
NH) Elutrap. In contrast, after deprotection in concentrated aqueous
ammonia at 55 °C overnight, the 5′-terminal DMT group was not
cleaved from oligonucleotides containing the hexaethylene glycol linker.
These oligonucleotides were purified by HPLC in 50 mM triethyl-
ammonium acetate buffer (pH 7) with a gradient of acetonitrile. The
DMT group was then removed by treatment with acetic acid for 30
min. All oligonucleotides were desalted on columns of Sephadex G-10
prior to use.
TM Values. TM values were obtained for duplexes or complexes
containing a 1:1 mixture of oligonucleotide-linker-oligonucleotide and
single stranded target each at various concentrations from 0.5 to 20
µM in 50 mM NaCl, 10 mM MgCl2 and 10 mM buffer: (pH 6.4 and
7.0) or TM values in the tables are reported for oligonucleotide
concentrations of 1 µM. Solutions were heated in 1 °C steps, and
absorbances were recorded after temperature stabilization using an
AVIV 14DS spectrophotometer. Absorbance and temperature readings
were plotted using Igor Pro software. TM values were determined from
first-order derivatives as well as graphically from absorbance vs
temperature plots. Differences greater than (1 °C were not observed
when using the two procedures. Histeresis experiments were performed
by cooling the solutions in 1 °C steps with corresponding absorbance
readings. Thermodynamic parameters were obtained using a van’t Hoff
analysis from plots of 1/TM vs log C for the concentration range of 0.5
to 20 µM.39,40
Fluorescence Data. Fluorescence spectra were obtained for single-
stranded conjugates or triplex conjugates at concentrations of 1 µM in
50 mM NaCl, 10 mM MgCl2 and 10 mM PIPES, pH 6.4 at ambient
temperature. Excitation occurred at 500 nm and emission was observed
in the range of 520-570 nm.
Results
The design of the hairpin triplexes relied upon a DNA or
RNA sequence that binds to the single-stranded target by both
Watson-Crick (W-C) hydrogen bonding as well as by Hoogs-
teen hydrogen bonding.4,6 This format requires that the probe
sequence bind in an antiparallel manner for the W-C interac-
tions and then make a 180° turn in order to bind in a parallel
manner to the Hoogsteen face of the target strand. In previous
studies this turn is facilitated by the use of five T residues
(Figure 1a) or by the replacement of the T residues by simple
linker such as hexa(ethylene glycol).4-6 We opted to design a
simple linker tethering the two different binding sequences that
could itself contribute to enhancing complex stability. The
design recognized that at the hairpin turn, the triplex structure
provides a terminal base triplet that is largely planar in nature
and could be envisaged as a target for hydrophobic base stacking
or intercalative interactions by an appropriate planar linker
(Figure 1b). To take advantage of such targeting we have
examined both naphthalene diimide and perylene diimide
derivatives as potential planar hydrophobic linkers. These simple
aromatic molecules were more readily available than others
reported in the literature12 to be specific for triplex binding.
The design of the nucleic acid conjugates employed either a
DNA or an RNA target sequence, a DNA/RNA strand for
Watson-Crick base pairing, an internally tethered naphthalene
or perylene linker followed by a second DNA/RNA strand for
Hoogsteen base pairing (Figure 1c).
Synthesis. The linkers were prepared as DMT-protected
phosphoramidites such that they could be used in conventional
solid-phase DNA synthesis. The synthesis of the various
perylene intermediates was complicated by low solubility in
most solvents. A significant improvement resulted when at least
one of the tBDMS- or DMT-protecting groups remained
attached to at least one of the linker arms throughout the
synthesis. Solubility of the perylene chromophore was no better
(39) Petersheim, M.; Turner, D. Biochemistry 1983, 22, 256-263.
(40) Marky, L. A.; Breslauer, K. J. Biopolymers 1987, 26, 1601-1620.

Naphthalene and Perylene Hairpin Triplexes
J. Am. Chem. Soc., Vol. 122, No. 25, 2000 5909
Figure 1. (a) Hairpin triplex with five dT residues tethering the two pyrimidine binding sequences. (b) Complex illustrated in (a) with a non-
nucleoside linker tethering to two binding sequences. (c) A DNA-naphthalene diimide conjugate. (d) A DNA-perylene diimide conjugate.
Scheme 1
remaining tBDMS-protecting group) could be obtained with a
limited HF/pyridine reaction. The mono-tBDMS-protected
product (6) could then be converted to the DMT-protected
derivative 7. The second tBDMS-protecting group was then be
removed, and the resulting product (8) converted to the
phosphoramidite derivative (9).
The naphthalene linker was prepared in a similar but more
straightforward manner. After incorporation of the two tBDMS
linkers, the tBDMS-protecting groups were both removed, and
the product was converted successively to the mono-DMT
derivative and then to the corresponding phosphoramidite.
Sequences of DNA or 2′-O-methyl RNA, or appropriate
chimeric sequences were prepared in the conventional manner
on CPG solid-phase supports. At the site of the linker, the
naphthalene-based phosphoramidite was incorporated in much
that same manner as a nucleoside derivative. The perylene
derivative continued to exhibit poorer solubility characteristics,
and to avoid clogging the lines of the DNA synthesizer, the
coupling was performed with the CPG in a small flask. After
this coupling reaction, further elongation in the normal manner
was also not effective. The first coupling after incorporation of
the perylene diimide needed to be performed in a 50:50
dichloromethane:acetonitrile solution. Presumably the solubility
of the tethered perylene was problematic in acetonitrile. After
the coupling reaction, and the removal of the excess linker, the
support became yellow (naphthalene) or burgundy (perylene)
in color. After completion of the synthesis, the CPG beads were
treated with concentrated methanolic ammonia (naphthalene)
or concentrated aqueous ammonia (perylene) overnight. After
isolation by PAGE, each complex was analyzed by HPLC with
detection at 260 nm and either 383 nm (naphthalene) or 537
nm (perylene). All of the conjugates eluted as single peaks under
both conditions of detection.
in aqueous solutions (quite likely worse), but after incorporation
into sequences of DNA, the polyanionic nature of the DNA
resulted in a significant improvement in aqueous solubility of
the conjugates.
We were concerned about the stability of the diimide linkages
and after isolation subjected the conjugates to conditions of
concentrated aqueous ammonia at 50 °C. The perylene conju-
gates were essentially insensitive to these conditions, but the
naphthalene conjugates were readily degraded. After ap-
proximately 4 h under these conditions, little intact conjugate
remained. Owing to this sensitivity, we performed the depro-
tection of the naphthalene linkers in methanolic ammonia.
A suspension of 3,4,9,10-perylene tetracarboxylicdianhydride
(Scheme 1) was treated with 2-(2-O-tert-butyldimethylsilyl-
ethoxy)-ethylamine (4) at elevated temperature in the presence
of zinc acetate to generate the target diimide 5. In principle it
was only necessary to remove the tBDMS-protecting groups
and convert this compound to the DMT-protected phosphor-
amidite. However, upon removal of the tBDMS-protecting
groups (TBAF) an insoluble solid precipitated. Similar results
were obtained in attempting to do a limited deprotection with
TBAF. An appropriately soluble product 6 (containing one
Effects of Naphthalene and Perylene Linkers on Hairpin
Triplex Stability. In our initial contribution36 we reported only
on the TM values for triplexes composed entirely of DNA

5910 J. Am. Chem. Soc., Vol. 122, No. 25, 2000
BeVers et al.
Figure 2. Eight hairpin triplexes formed with a DNA (D) target (upper
grouping) or an RNA (R) target (lower grouping) and binding sequences
that are both DNA (D), both 2′-O-methyl RNA (M) or one strand of
each.
sequences and either linker. These represent perhaps the simplest
of the possible complexes. Although the hairpin triplexes are
composed of only two sequences, there are still three strands
of nucleic acid involved in the complex. The nonconjugated
sequence can be considered as the “target” strand, and the
conjugate binds to the target with two strands, one a Watson-
Crick strand (antiparallel) and the second a Hoogsteen strand
(parallel). It is then possible to study two groups of complexes,
those that have a DNA target strand (D) and those that have an
RNA target strand (R) (Figure 2). Within those two groups eight
complexes were generated (four with the naphthalene linker and
four with the perylene linker), one contains DNA for both the
Watson-Crick and Hoogsteen strands, while a second contains
RNA (in this case the 2′-O-methyl analogue) for both strands.
Additional complexes were formed with one DNA strand and
one 2′-O-methyl RNA strand (M). In total eight conjugated
triplexes were prepared for each linker (Figure 2).
The triplexes studied contained a target strand of 19 residues
and the polypurine target site occupied the central 9 residues.
This target sequence contained two G-C base pairs and seven
A-T base pairs. We employed T (or U) in the third strand for
A-T base pair targets, and C for G-C base pair targets. In the
latter case protonation is necessary to permit bidentate binding
to the target G, so we obtained TM values at pH values of 6.4
and 7.0. In all cases a single thermally induced transition was
observed for the conjugated triplexes (Figure 3a), and the
cooperativity observed suggested that the transition could be
approximated by a two state model. While triplexes formed from
a target duplex and a third single strand commonly result in
two transitions (triplex f duplex and duplex f random coil)
hairpin triplexes formed from a single strand target and a hairpin
second/third strand more commonly result in a single transi-
tion.4,6,7 In the present case this observation likely results from
the fact that the Watson-Crick and Hoogsteen strands are
tethered to one another so denaturation of both strands occurs
Figure 3. (a) Absorbance vs temperature plot (at 260 nm) for the
D-D-D hairpin triplex containing a perylene diimide-based linker.
Inset shows hysteresis for the heating and cooling experiments. (b) The
same temperature vs absorbance plot as in (a) monitored at 537 nm.
as a single event. The observed transitions were monitored in
both “heating” and “cooling” experiments with only a moderate
hysteresis effect present (Figure 3a, inset). We additionally
monitored the transitions at 383 nm (naphthalene) or 537 nm
(perylene) and observed very similar transitions with midpoints
essentially identical with those monitored at 260 nm (see Figure
3b). These data further suggested that the denaturation of these
conjugate triplexes could be approximated by a two state model.
For the conditions of pH 6.4 we employed a van’t Hoff
analysis29,30 to obtain free energy parameters for the transitions
of sixteen different complexes.
The first group of eight complexes described (Table 1) are
those containing a DNA target strand. For comparative purposes
we have also prepared a hairpin triplex with a simple hexa-
(ethylene glycol) linker. The TM values for the all DNA triplex
formed with this nonplanar simple linker were 24 and 31 °C,
at pH values of 7.0 and 6.4, respectively, and the ∆G37
determined at pH 6.4 was -7.7 kcal/mol. When the hexa-
(ethylene glycol) linker was replaced with naphthalene diimide
[and two tri(ethylene glycol) chains] a 6 kcal/mol change in
∆G37 was observed. With the perylene diimide present, a further
1.3 kcal/mol of stabilization results, some 7.3 kcal/mol more
stable than the complex containing the simple glycol linker. As
the nature of the backbones (DNA vs RNA) of the Watson-
Crick and Hoogsteen strands were varied, notable changes in
stability were observed. The most dramatic difference occurred
when both strands were 2′-O-methyl RNA. This complex, with
a TM increase of 28 °C, exhibited 21 kcal/mol of stabilization,
some 13.3 kcal/mol more stable that the complex containing
the simple glycol linker. The mixed DNA/RNA complexes

Naphthalene and Perylene Hairpin Triplexes
J. Am. Chem. Soc., Vol. 122, No. 25, 2000 5911
Table 1. Stability of Triplexes Containing Naphthalene- and
Table 2. Stability of Triplexes Containing Naphthalene- and
Perylene-Based Linkers with a DNA Target Stranda
Perylene-Based Linkers with an RNA Target Stranda
a TM values were obtained for duplexes or complexes containing a
1:1 mixture of oligonucleotide-linker-oligonucleotide and single-
stranded target in PIPES at the noted pH at oligonucleotide concentra-
tions of 1 µM. Solutions were heated in 1 °C steps, and absorbances
were recorded after temperature stabilization using an AVIV 14DS
spectrophotometer. Absorbance and temperature readings were plotted
using Igor Pro software. TM values were determined from first-order
derivatives as well as graphically from absorbance vs temperature plots.
TM values are the average of at least two determinations reproducible
to (1 °C. ∆G37 values were calculated for the data at pH 6.4 from
1/Tm vs log C plots. b D ) oligodeoxynucleotide, M ) 2′-o-methyl-
oligoribonucleotide.
a See footnote a of Table 1. b See footnote b of Table 2. c No
transition detected. d Transition begins <0 °C, TM is approximated.
In general the larger perylene linker was more effective than
the naphthalene in enhancing the stabilization of the three
stranded complexes. The largest difference between the two
planar linkers was 5.6 kcal/mol observed for the complex formed
from a DNA target and a conjugate containing two RNA strands
(Table 1). By comparison, all of the complexes containing an
RNA target had very similar ∆G values regardless of the linker
employed.
exhibited 17.7 and 10.3 kcal/mol of complex stabilization, the
more stable complex occurring when both strands of the target
duplex were DNA and the Hoogsteen strand was RNA.
The second group of eight complexes (Table 2) are those
containing an RNA target strand. In these cases the triplex was
much less stable. For the complex containing a hexa(ethylene
glycol) linker, no transition was observed at pH 7.0, and at pH
6.4 a very early transition was observed with an estimated TM
of about 10 °C. However, this transition was not sufficiently
distinct to permit thermodynamic characterization. The stabiliz-
ing effect of the planar linkers resulted in much more stable
complexes that could be effectively characterized. In the best
case (the all RNA triplex) the TM value was increased by roughly
36 °C with 10.8 kcal/mol of stabilization. By comparison, when
two DNA strands were used with the RNA target, a more
Fluorescence Properties. Perylene derivatives have well-
documented fluorescence properties that were useful in char-
acterizing some of the conjugated triplex properties. Neither
the perylene tetraacetic dianhydride, nor any of the linker-only
derivatives prepared, could be used to obtain usable fluorescence
spectra in aqueous solution. Either the solubility of these
derivatives was so poor that no significant amounts of material
were solubilized, or material in solution underwent self-
aggregation and resulted in self-quenching of any fluorescence
emission properties. However, upon conjugation to the charged
nucleic acids, the DNA/RNA-perylene linkers could be dis-
solved in water and fluorescence spectra were obtained. With
an excitation wavelength of 500 nm, the perylene conjugates
exhibited an emission maximum at 546 nm. The quantum yields
for various conjugates exhibited significant variations (Figure
4a). The conjugate containing two RNA strands (as the 2′-O-
methyl derivatives) resulted in the highest quantum yield, while
that containing two DNA strands the lowest (reduced some
6-fold). The conjugate containing one RNA and one DNA strand
was intermediate in effect.
moderate increase in TM of about 12 °C was observed (∆G37
)
-7.8 kcal/mol). With both linkers, those complexes containing
at least one RNA strand were more effective in stabilizing the
triplex with the RNA target than the conjugate containing two
DNA strands.

5912 J. Am. Chem. Soc., Vol. 122, No. 25, 2000
BeVers et al.
Figure 4. Fluorescence spectra of various perylene conjugates obtained at concentrations of 1 µM and pH 6.4. (a) Fluorescence emission spectra
for (i) the 5′-RNA-perylene-RNA conjugate, (ii) the 5′-RNA-perylene-DNA conjugate, (iii) the 5′-DNA-perylene-DNA conjugate, all obtained
at concentrations of 1 µM. (b) Fluorescence emission spectra for (i) the 5′-DNA-perylene-DNA conjugate, (ii) the 5′-DNA-perylene-DNA
conjugate with the DNA 19-nucleotide target, and (iii) the 5′-DNA-perylene-DNA conjugate with the purine-only 9-nucleotide target, all obtained
at concentrations of 1 µM.
Upon complexation to the target strand the emission of the
perylene linker in the all DNA complex was quenched to some
14% of its initial value (Figure 4b). Some variation in the
quenching effects were observed for the other seven complexes
(Figure 5). During the course of these studies we also prepared
a smaller triplex complex, one composed of the shortest possible
target, a 9-mer. The TM values obtained for the all DNA complex
with this 9-mer were similar to those obtained with the 19-mer
target, but the fluorescence effects were quite different. With
this complex almost complete quenching of the emission signal
was observed. Both of these experiments, with either the 19-
mer or the 9-mer target, are consistent with energy transfer
between the perylene and stacked base residues of the triplexes.
Quenching of the perylene excited state by aromatic pi systems
has been observed previously.41
Discussion
The goal of this study was to target the base triplet at the
terminus of the three-stranded complex formed from a hairpin
triplex as a site to enhance hydrophobic base-stacking interac-
tions and thereby provide additional overall stabilization to the
complex. The design of both linkers employed was based upon
aromatic diimides, one a derivative of naphthalene and the
second originating from perylene. The synthetic scheme for the
former was relatively straightforward, while that for the latter
(41) Icli, S.; Icil, H.; Gurol, I. Turk. J. Chem. 1997, 21, 363-368.

Naphthalene and Perylene Hairpin Triplexes
J. Am. Chem. Soc., Vol. 122, No. 25, 2000 5913
Figure 5. Bar graph illustration of the relative fluorescence quantum yields for the simple uncomplexed DNA/RNA conjugates containing a
perylene-based linker (black) in comparison with the corresponding triplexes (gray) obtained by titrating the perylene-based conjugates with one
equivalent of the DNA or RNA target strand. For example, D-R*D represents the Watson-Crick*Hoogsteen complex with an RNA (R) target and
two DNA (D) strands from the perylene conjugate.
was complicated by the poor solubility characteristics of the
perylenes in general. However, appropriate selection of protec-
tion groups permitted us to convert the perylene diimide into
the desired phosphoramidite, and appropriate choice of coupling
procedures permitted the preparation of the complexes in
sufficient quantities. Isolation by HPLC methods of the perylene-
based conjugates was complicated by the hydrophobic nature
of the linkers, but PAGE isolation eliminated these complica-
tions and was an effective alternative.
Base-stacking interactions between the planar linkers and the
nucleic acid triplexes are further indicated in that (i) absorbance
vs temperature plots obtained at 383 or 537 nm for duplexes
containing the naphthalene- or perylene-based linkers, respec-
tively, also exhibited identical cooperative increases in hyper-
chromicity with a midpoint in the transitions that corresponded
to those obtained at 260 nm (see Figure 3a and 3b). (ii)
Additionally, absorbance vs temperature plots obtained in
solutions containing 10-30% ethanol, resulted in TM reductions
from 3 to 7 °C as we described in our initial contribution.36
While these experiments suggest that stacking interactions
between the linkers and the nucleic acids bases are present,
ethanol can also destabilize nucleic acid complexes in the
absence of any aromatic linkers by interfering with base-stacking
interactions.
TM and ∆G values varied for the complexes, and were
dependent on the nature of the DNA or RNA composition of
the complexes. For the DNA target, with both conjugate strands
containing 2′-O-methyl RNA (i.e., M-D*M ) Watson-
Crick*Hoogsteen strands), the TM was increased by as much as
12 °C for the naphthalene linker and as much as 16 °C for the
perylene linker (relative to conjugates in which one or both
strands are DNA). ∆G values for the M-D*M complex
containing the naphthalene- and perylene-based linkers were
-15.4 and -21.0 kcal‚mol-1, respectively. By comparison, ∆G
values for the M-D*D complexes were -9.9 (naphthalene
linker) and -10.3 (perylene linker) kcal‚mol-1. The TM and ∆G
data therefore suggest that the complexes in which both probes
are 2′-O-methyl RNA are the most stable. One mixed probe
strand containing the naphthalene linker (D-D*M) had a TM
similar to that of the M-D*M complex with the same linker.
This trend has been observed previously where it was reported42
that R-D*R and D-D*R followed by D-D*D were the most
stable complexes formed. In other studies, and in the presence
of 2′-O-methyl RNA, the following relative stabilities: M-D*M
(42) Roberts, R. W.; Crothers, D. M. Science 1992, 258, 1463-1466.
The perylene-based linker is more likely to be effective in
triplex stabilization since modeling studies suggest that it can
potentially provide more effective stacking interactions with all
three base residues of the terminal base triplet (Figure 6). This
expectation was observed for those complexes involving a DNA
target. Less differentiation was present with complexes contain-
ing an RNA target suggesting the possibility that in these latter
complexes the perylene linker was not positioned in such a
manner to maximize stacking interactions. With the described
19-mer the linker must “cross” the target strand at some point
in order for both Watson-Crick and Hoogsteen interactions to
take place. With the described complexes there is the possibility
that the first residue extending beyond the triplex complex may
be stacked onto either the naphthalene- or perylene-based linker
at the point in which the linker crosses the target strand in such
a manner that the linker would effectively be intercalated
between the last base triplet and the first single-stranded base
of the target. In the complex illustrated here the first single-
stranded base beyond the last triplet is a cytosine residue, which
typically has the poorest of base-stacking interactions of the
four common bases. By comparison, adenine has demonstrated
the strongest base-stacking interactions. As we reported in our
initial contribution,36 when the target sequence was changed to
effect a substitution of dA for dC at this site, an additional 3
°C increase in TM was observed. This result suggests a complex
with the naphthalene or perylene derivative sandwiched between
the terminal base triplet and the next single-stranded base.

5914 J. Am. Chem. Soc., Vol. 122, No. 25, 2000
BeVers et al.
able to stabilize complexes containing the DNA target than the
naphthalene linker. We also note that the TM values at pH values
of 6.4 and 7.0 do not differ much for the RNA target series.
Since we were unable to obtain effective transitions for those
complexes containing the hexa(ethylene glycol) linker, the
presence of the planar linker appears to be a dominating force
in the formation of this series of complexes.
For complexes with a RNA target the results are less dramatic.
The TM values obtained for the D-R*D complex were 22 and
28 °C for the naphthalene- and perylene-based linkers respec-
tively (the lowest TM values observed in this study). ∆G37 values
for the DR*D complexes were -7.8 and -7.9 kcal/mol (the
lowest ∆G37 values observed in this study). A general rule for
DNA/RNA mixed triplexes seems to be that if the duplex
contains the R purine strand, only the R third strand will bind.
TM values for the M-R*M complexes were 17-20 °C higher
than those for the D-R*D complexes, consistent with this
observation. Again, the ∆G37 values correlated well with the
TM values (2.2-2.9 kcal‚mol-1 more negative than those for
the D-R*D complexes), which indicates that the all-RNA probe
is significantly more stabilizing when the target is RNA than
the all-DNA probe. The mixed probes demonstrated intermediate
values of TM and ∆G. The order of stability previously reported
from most stable to least is R-R*R > R-R*D > D-R*R,
D-R*D; and if 2′-O-methyl made up one or two of the
pyrimidine strands, the order was M-R*M > M-R*D >
D-R*M > D-R*D.35 This pattern seems similar to the pattern
observed in these experiments; however, in that previous work,
circular oligonucleotides were used in which the loop portions
were penta-nucleotides.
In comparing the naphthalene-based and perylene-based linker
complexes containing the RNA probe, again in each case the
TM values for the complexes containing a perylene-based linker
are only slightly higher than those for the corresponding complex
with a naphthalene-based linker. The corresponding ∆G37 values
are also quite similar regardless of the linker. These results
suggest that with the RNA target the perylene linker may be
less optimally positioned, or that the terminal base triplet is less
coplanar, such that effective stacking interactions with the larger
perylene-based conjugate are not maximized.
Figure 6. Illustration of the approximate relationships of the naph-
thalene linker (upper) and perylene linker (lower) linkers to the terminal
base triplet of the hairpin triplex.
> D-D*M > D-D*D > M-D*D were reported.43 Other
studies44 have suggested much different relative orders, and these
differences may be due to variation in the sequences employed
or in the experimental conditions. Previous studies determined
the ∆G values for similar complexes containing the hexa-
(ethylene glycol) linker.4 For the hexa(ethylene glycol) sequence
illustrated in Table 1 (with an all DNA conjugate) the ∆G37
was determined to be -7.6 kcal‚mol-1. This value is comparable
to that previously determined for a similar complex and is
consistent with our finding that complexes containing this simple
glycol linker are less stable than those containing the naphtha-
lene- and perylene-based linkers.
Tables 1 and 2 also indicate that the TM values for complexes
containing a perylene-based linker are 4-8 °C higher than the
corresponding complex with a naphthalene-based linker (at pH
6.4). The corresponding ∆G values correlate well with these
data in that for each type of complex the ∆G values were slightly
more negative for the complexes containing the perylene-based
linkers. Complex formation appears to be enthalpy-driven in
all cases, but the differences in stability between the naphthalene
and perylene complexes for the DNA target arise in part from
a slightly less unfavorable entropy term for the perylene
complexes. The data indicate that the perylene linker is better
Nearly all effective DNA intercalators are not only planar in
geometry but also contain a positive charge. The former property
permits the agent to “slide” between base pairs, while the latter
may be important for charge-charge interactions with neighbor-
ing phosphates and necessary to lock the intercalator into
position. Other types of planar molecules lacking a positive
charge, most notably psoralen and the diol-expoxide derivatives
of benzopyrene, have not been observed in fully intercalated
states until they are covalently tethered to the duplex. After
alkylation, the intercalated mode of binding appears to pre-
dominate.45,46 Such ligands can be considered to be transient
intercalators in the absence of covalent bond formation. The
planar linkers described in this study also lack any positive
charge; however, the presence of significant binding interactions
with the tethered Watson-Crick and Hoogsten pyrimidine
strands will certainly position the planar intercalating linkers
in position to be effectively stacked on the terminal base triplet,
and intercalated between the triplet and the next base of the
single-stranded target. The energetics of binding by selected
positively charged intercalators has been determined, and for
(45) Tomic, M. T.; Wemmer, D. E.; Kim, S. H. Science 1987, 238,
1722-1725.
(46) Cosman, M.; Delossantos, C.; Fiala, R.; Hingerty, B. E.; Singh, S.
B.; Ibanez, V.; Margulis, L. A.; Live, D.; Geacintov, N. E.; Broyde, S.;
Patel, D. J. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 1914-1918.
(43) Wang, S.; Kool, E. T. Nucleic Acids Res. 1995, 23, 1157-1164.
(44) Han, H.; Dervan, P. B. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 3806-
3810.

Naphthalene and Perylene Hairpin Triplexes
J. Am. Chem. Soc., Vol. 122, No. 25, 2000 5915
two quite different agents, ethidium bromide and daunomycin,
the free energies for intecalative binding are essentially identical
at -9 kcal/mol.47 With the present complexes, the difference
between the hexa(ethylene glycol) linker and the planar linkers
can be roughly compared. And although it is unlikely that the
energetic effects are strictly additive, this comparison (for the
D-D*D complex) suggests that the intercalating linkers provide
roughly 6 kcal/mol (naphthalene) and 7.3 kcal/mol of stabiliza-
tion energy. These values are probably not unreasonably
different from the ethidium bromide and daunomycin values,
given that those agents when intercalated experience π-π
interactions to both planar surfaces, while with the present
complexes the π-π interactions largely occur to one of the
aromatic faces.
The large increase in complex stability between the control
complex containing the hexa(ethylene glycol) linker and those
containing the naphthalene and perylene linkers appears to be
largely due to the stacking interactions afforded by the latter
planar derivatives. However, other aspects may also contribute
to this enhanced stability. The planar linkers have fewer rotatable
bonds, relative to hexa(ethylene glycol) so that fewer unfavor-
able conformations may be present for the planar linkers.
Additionally, while the planar linkers are designed to interact
favorably with the target strand, the simple hexa(ethylene glycol)
linkers offers no favorable interactions, and may in fact be
presented with unfavorable steric challenges as it must cross
the target strand in its role of tethering the Watson-Crick and
Hoogsteen binding oligonucleotides. Other parameters clearly
affect complex stability, the most obvious being the mixture of
DNA/RNA within the complexes. Regardless of the target strand
(DNA or RNA) the most stable complexes in each series resulted
when both the Watson-Crick and Hoogsteen strands were RNA
or 2′-O-methyl RNA (M-D*M and M-R*M complexes). By
comparison, the least stable complexes in each series were
generally those in which a Watson-Crick heteroduplex was
formed with a third strand of DNA (M-D*D and D-R*D
complexes). Although the structures of these various triple
helices have eluded formal explanation by X-ray analyses, it is
clear that base-base interactions alone cannot explain these
differences.
sive for the 19-mer target (Figures 4b and 5). Although
quenching was observed in all cases, significant variations were
present. Within the RNA target series, the smallest quenching
effects were observed for the M-R*D complex (46% of
conjugate), and this is also the least stable triplex (see Table
2). The most dramatic quenching effects were observed for the
all-M-R*M triplex (18% of conjugate); this complex is also
the most stable of the four triplexes formed from RNA target
strands. The remaining two triplexes have intermediate quench-
ing effects and exhibit intermediate ∆G values (compare Figure
1 with Figure 5). The results with the DNA target are less
conclusive. The smallest quenching effects were observed for
the M-D*D triplex (48% of conjugate), and this complex also
exhibited the lowest ∆G value for the DNA target series (Table
1). However, the most significant fluorescence quenching was
observed for the D-D*D and D-D*M triplexes (14 and 17%
of conjugate, respectively), while the most thermodynamically
stable triplex is the M-D*M complex (with 25% fluorescence
relative to the conjugate). This inconsistency may simply reflect
some preference for stacking on the DNA-DNA Watson-Crick
duplex that is only present in these two complexes. With the
DNA-DNA duplex present, the fluorophore stacks onto a dA-
dT base pair. Since the quantum yield for the single-stranded
conjugates is most dramatically reduced for the DNA-linker-
DNA sequence, the presence of the dT residues likely stack
more effectively with the linker and result in more dramatic
quenching whether in simply the single-stranded conjugate or
the corresponding three-stranded complex. Such an effect would
skew the relationship between triplex stability, which is the result
of both the nature of the three nucleic acid strands and of the
stacked ligand, while the fluorescence quenching data only
reflects the nature of the interaction of the perylene with,
presumably the terminal base triplet.
While variations in the fluorescence quenching effects are
likely to reflect some slight differences in the nature of the
hydrophobic stacking interactions between the fluorophore and
the planar aromatic bases, the most likely structure consistent
with the observed data is that in which the planar chromophore
stacks on top of the terminal base triplet of the three-stranded
complex (see Figure 6). That the complex with the shorter 9-mer
target is largely non-fluorescent is likely a result of dimerization
about the perylene linkers occurring between two triplexes. The
additional perylene-perylene stacking as the result of dimer-
ization likely contributes to additional self-quenching effects.
The quantum yields from the fluorescence experiments for
the perylene-based conjugates are dependent upon the nucleic
acid composition. The all-RNA conjugate is likely the most polar
of the conjugates, owing to the presence of the 2′-OH on each
sugar residue and the absence of the methyl group on the uracils.
In the RNA-DNA conjugate, the uracils of the DNA strand
are methylated as thymines. The presence of the methyl groups
and the absence of the 2-OH at each site results in a more
hydrophobic sequence, and one that is likely more capable of
stacking or otherwise interacting with the hydrophobic perylene
linker. Any kind of base-stacking event will likely result in
energy transfer, and as observed, the conjugate with two DNA
strands is much less fluorescent than that with two RNA strands.
The quenching observed upon complex formation is most
dramatic for the minimum 9-nucleotide target, but still impres-
Conclusions
The use of planar naphthalene and perylene diimide linkers
to bridge the Watson-Crick and Hoogsteen hydrogen bonding
strands in a hairpin triplex imparts significant additional stability
to the three-stranded complex. These types of conjugates can
be used effectively to target single-stranded nucleic acids and
could be effective modulators of genetic expression by com-
plexing to single-stranded mRNA or single-stranded portions
of DNA actively involved in transcriptional processes.
Acknowledgment. This work was supported by a grant from
(47) Breslauer, K. J.; Remeta, D.; Chou, W. Y.; Ferrante, R.; Curry, J.;
Zaunczkowski, D.; Snyder, J. G.; Marky, L. A. Proc. Natl. Acad. Sci. U.S.A.
1987, 84, 8922-8926.
the NIH (GM 53201).
JA0001714