Discrete and polymeric self-assembled dendrimers:
Hydrogen bond-mediated assembly with
high stability and high fidelity
Perry S. Corbin, Laurence J. Lawless, Zhanting Li, Yuguo Ma, Melissa J. Witmer, and Steven C. Zimmerman*
Department of Chemistry, University of Illinois, 600 South Mathews Avenue, Urbana, IL 61801
Edited by Jack Halpern, University of Chicago, Chicago, IL, and approved January 24, 2002 (received for review November 30, 2001)
t the most fundamental level, the readout of stored information in biochemical systems requires both a code and the
machinery for expressing the code (1). Distinction at the molecular
level is critical in such processes and in its simplest form involves a
molecular recognition event. The code may be, for example, the
primary sequence of a polypeptide, which through a series of
recognition events folds into a specific secondary or tertiary structure, or it may be a linear array of recognition sites that are
sequentially processed through a distinctive engagement at each
site. However, a distinctive chemical process alone (e.g., DNA base
pairing) is not sufficient. It must be coupled with a biological
process (e.g., replication) to be considered authentic information
retrieval.
Inspired by nature’s ability to create extraordinarily complex
systems from comparatively simple information codes, and with
an eye toward creating nanoscale devices (2), chemists have
sought to develop small molecules capable of self-assembling
into larger structures (3-10). Molecular recognition sites within
these small molecules carry the code that guide the assembly.
Thus, the information retrieval process is expressed through the
formation of the self-assembled structure, which can further
manifest itself through the bulk properties of the material
formed (11, 12). Despite the many successful examples that have
appeared over the past decade, the power of this biomimetic
self-assembly strategy has not been fully realized largely because
of the limited number of recognition motifs available, and, in
particular, their relatively low stability (4, 13). Moreover, nearly
all self-assembling systems reported to date use a single type of
distinction. Developing systems where bi- or polyinstructional
codes guide the assembly is a special challenge because it
requires fidelity in the individual recognition events.
We recently reported (14) that ureidodeazapterin 1 dimerizes
very strongly (Kdimer ⬎ 107 M⫺1) in chloroform-d by means of its
self-complementary, DDAA array, which is present in both the
1(H) and 3(H)-protomeric forms. Heterocycle 1 is inexpensively
www.pnas.org兾cgi兾doi兾10.1073兾pnas.062641199
Materials and Methods
Compounds. All compounds described herein gave NMR, matrixassisted laser desorption ionization (MALDI)-time of flight MS,
and UV-visible spectra in accord with their structures, and each key
compound gave a passing elemental analysis. Compounds 3 (15), 9
a–c (16, 17), 10 a–c (16, 17), 13 a–c (16, 17), 14a (18), 14 b and c
(Y.M., S. V. Kolotuchin, and S.C.Z., unpublished work), 15 (16, 17),
3,3,3-triphenylpropylamine (19), 6-bromo-5-deazapterin (20), and
Gn-Br dendrons a–c (16, 17) were prepared by known methods.
Size standards 16 a and b were prepared by using the same
precursors as for 15 and by analogous methods.
4-Pyren-1-yl-Butyric Acid Pent-4-ynyl Ester. To a suspension of
1-pyrenebutyric acid (1.00 g, 3.50 mmol) and 4-pentyn-1-ol (0.30 g,
3.57 mmol) in CH2Cl2 (100 ml) was added dicyclohexylcarbodiimide (0.72 g, 3.50 mmol). The solution was stirred at room temperature (RT) for 3 h. Work-up and column chromatography
(CH2Cl2兾hexane 3:2) afforded the title compound (0.93 g, 75%) as
a white solid: mp 60–62°C; 1H NMR (CDCl3) ␦ 7.86–8.35 (m, 9H),
4.24 (t, 2H), 3.42 (t, 2H), 2.48 (t, 2H), 2.32 (m, 2H), 2.23 (m, 2H),
1.89 (m, 2H), 1.56 (t, 1H); MS (fast atom bombardment, FAB): 354
(M⫹); Anal. Calcd for C25H22O2: C, 84.72; H, 6.26; Found: C, 84.35;
H, 6.36.
Compound 4. To the solution of the N-butyl urea of 6-bromo-5deazapterin (0.17 g, 0.50 mmol) (prepared by using butyl isocyanate) and 4-pyren-1-yl-butyric acid pent-4-ynyl ester (0.18 g, 0.50
mmol) in acetonitrile (30 ml) were added triethylamine (1.0 ml),
CuI (10.0 mg, 10%), and Pd(PPh3)4 (30.0 mg, 5%). The mixture was
heated under reflux for 20 h and then cooled to RT. The solvent was
removed under reduced pressure, and the residue was triturated in
CH2Cl2 (100 ml). The organic phase was washed with aqueous
hydrochloride solution (1 M, 20 ml), sodium carbonate solution (1
M, 20 ml), water (2 ⫻ 20 ml), and brine (20 ml), and dried over
MgSO4. After the solvent was removed under reduced pressure, the
residue was subjected to column chromatography (CH2Cl2兾
methanol 50:1) to afford the corresponding alkyne (0.19 g, 65%) as
a pale solid: mp 83–85°C; 1H NMR (DMSO-d6) ␦ 11.96 (s, 1H), 9.72
(s, 1H), 8.68 (s, 1H), 7.85–8.35 (m, 10H), 7.53 (s, 1H,), 4.19 (t, 2H),
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: MALDI, matrix-assisted laser desorption ionization; RT, room temperature;
FAB, fast atom bombardment; THF, tetrahydrofuran; SEC, size-exclusion chromatography.
*To whom reprint requests should be addressed. E-mail: sczimmer@uiuc.edu.
PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5099 –5104
CHEMISTRY
A
and conveniently prepared (14) by reacting butyl isocyanate with 3,
which in turn is made in two steps from 2,4-diamino-6hydroxypyrimidine (2) and 1,1,3,3-tetramethoxypropane (15) (Fig.
1). Herein, we describe structural studies on the 3(H) dimer of 1 and
quantitative complexation studies using fluorescence spectroscopy
on a pyrene derivative. We further show that a ditopic unit based
on 1 can assemble dendrimers into discrete aggregates in a generation-dependent fashion and that its AADD units pair with high
fidelity in the presence of a competing ADD䡠DAA code.
SPECIAL FEATURE
Hydrogen bond-mediated self-assembly is a powerful strategy for
creating nanoscale structures. However, little is known about the
fidelity of assembly processes that must occur when similar and
potentially competing hydrogen-bonding motifs are present. Furthermore, there is a continuing need for new modules and strategies that
can amplify the relatively weak strength of a hydrogen bond to give
more stable assemblies. Herein we report quantitative complexation
studies on a ureidodeazapterin-based module revealing an unprecedented stability for dimers of its self-complementary acceptoracceptor-donor-donor (AADD) array. Linking two such units together
with a semirigid spacer that carries a first-, second-, or third-generation Fréchet-type dendron affords a ditopic structure programmed to
self assemble. The specific structure that is formed depends both on
the size of the dendron and the solvent, but all of the assemblies have
exceptionally high stability. The largest discrete nanoscale assembly
is a hexamer with a molecular mass of about 17.8 kDa. It is stabilized
by 30 hydrogen bonds, including six AADD䡠DDAA contacts. The
hexamer forms and is indefinitely stable in the presence of a hexamer
containing six ADD䡠DAA hydrogen-bonding arrays.
7.43 (t, 2H), 7.30 (d, 4H), 6.74 (d, 2H), 6.69 (t, 1H), 5.13 (s, 2H),
5.03 (s, 4H), 1.36 (s, 36H); 13C NMR ␦ 171.36, 160.53, 159.12,
151.10, 137.81, 135.53, 132.56, 122.83, 122.36, 122.30, 120.91,
106.41, 101.77, 71.06, 70.58, 34.85, 31.43; IR (KBr, cm⫺1) 2,146
(N3), 1,698 (CAO); Anal. Calcd for C45H54N6O5: C, 70.64; H,
7.12; N, 10.96. Found: C, 70.60; H, 7.06; N, 10.57.
Compound 6b. Using the same procedure as for 6a afforded 447
mg (85%) of the title compound, which was determined to be
approximately 95% pure by 1H NMR spectroscopy and was used
in the next step without further purification: 1H NMR ␦ 8.29 (t,
1H); 7.88 (d, 2H); 7.43 (t, 4H), 7.31 (d, 8H), 6.76 (d, 4H), 6.71
(d, 2H), 6.68 (t, 2H), 6.66 (t, 1H), 5.12 (s, 2H), 5.03 (br s, 12H),
1.36 (s, 72H); IR (KBr, cm⫺1) 2,149 (N3), 1,702 (CAO); MS
(MALDI) 1,379.10 (rearrangement product)⫹.
Fig. 1.
Synthesis of 1, its protomeric forms, and dimers.
3.30 (t, 2H), 3.22 (t, 2H), 2.52 (m, 4H), 2.03 (m, 2H), 1.75 (m, 2H),
1.45 (m, 2H), 1.28 (m, 2H), 0.85 (t, 3H); MS (FAB): 614 (M ⫹ H)⫹;
high-resolution MS (FAB): Calcd for C37H35N5O2: 614.2764.
Found: 614.2764. Anal. Calcd for C37H35N5O2䡠H2O: C, 70.34; H,
5.92; N, 11.09. Found: C, 70.67; H, 6.27; N, 11.11. The alkyne (0.12
g, 0.20 mmol) was dissolved in chloroform (100 ml) and hydrogenated (18 psi H2) over 10% Pd-C at RT for 24 h. After work-up,
compound 4 (60 mg, 60%) was obtained as a pale solid: mp
68–70°C; 1H NMR (DMSO-d6) ␦ 11.95 (s, 1H), 9.57 (s, 1H), 8.64
(s, 1H), 7.85–8.35 (m, 10H), 7.53 (s, 1H), 3.95 (t, 2H), 3.32 (m, 2H),
3.20 (m, 2H), 2.56 (m, 4H), 2.00 (m, 2H), 1.78 (m, 4H), 1.46 (m, 2H),
1.25 (m, 4H), 0.85 (t, 3H); MS (FAB): 617 (M ⫹ H)⫹; Anal. Calcd
for C37H39N5O2䡠H2O: C, 69.90; H, 6.51; N, 11.02. Found: C, 70.01;
H, 6.25; N, 10.97.
Compound 5. To a solution of 3,3,3-triphenylpropylamine (0.29 g,
1.00 mmol) in 100 ml of toluene was added a solution of 20%
phosgene (2 ml) in toluene at RT. The solution was stirred for
24 h and the solvent was removed under reduced pressure. The
oily residue was dissolved in tetrahydrofuran (THF) (50 ml) and
3 (0.24 g, 1.50 mmol) added. The mixture was heated under
reflux for 48 h and cooled to RT. The solvent was then removed
and the residue was purified by column chromatography
(CHCl3兾methanol, 100:1), to give 5 (71 mg) in 15% yield as a
white solid: mp 210–212°C; 1H NMR (CDCl3) ␦ 13.71–13.52 (m,
1H), 11.82–9.74 (m, 2H), 8.42–8.81 (m, 2H), 7.85 (s, 1H), 7.23
(m, 15H), 3.55 (t, 2H), 2.92 (t, 2H); MS (FAB): 476 (M ⫹ H)⫹;
high-resolution MS (FAB): Calcd for C29H26N5O2: 476.2088.
Found: 476.2087. White needles suitable for x-ray analysis were
obtained by slow evaporation of a chloroform-methanol solution
open to the atmosphere.
Compound 6a. A solution of sodium azide (31 mg, 0.48 mmol) in
1 ml of water was added to a solution of the activated ester 11a
(145 mg, 0.160 mmol) in 10 ml of acetone. The resulting mixture
was stirred at RT for 4 h or until no starting material remained
by TLC. The suspension was poured over ice, and the solid that
formed was collected by vacuum filtration and washed with cold
water. The solid was dissolved in 50 ml of CH2Cl2 and dried over
sodium sulfate. Solvent was removed in vacuo (Danger! No
heating as azides may be explosive) to give 70 mg of the product,
which was purified by column chromatography (Rf ⫽ 0.67, 1:1
CH2Cl2兾petroleum ether) to give 40 mg (81%) of the title
compound as a white solid: 1H NMR ␦ 8.28 (t, 1H), 7.89 (d, 2H),
5100 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.062641199
Compound 6c. Oxalyl chloride (40 ␮l, 0.459 mmol) was added slowly
with a syringe to a stirred solution of diacid 10c (300 mg, 0.113
mmol; dried overnight in vacuum dessicator) in dry THF (4.5 ml)
and dry N,N-dimethylformamide (9.0 ␮l). Stirring was continued at
RT for 100 min, then at 50°C for 15 min and RT for 1 h.
Trimethylsilyl azide (90 ␮l, 0.678 mmol) was added slowly with a
syringe and the mixture was stirred at RT for 1.5 h. The mixture was
evaporated on the rotary evaporator (Danger! No heating as azides
may be explosive) then dried in vacuo for 1 h. The resulting pale
yellow foam was taken up in CH2Cl2兾petroleum ether (1:1, 10 ml)
and filtered. The filtrate was purified by flash chromatography
eluting with CH2Cl2兾petroleum ether (1:1). Evaporation on the
rotary evaporator of fractions containing pure product (see above),
followed by further evaporation in vacuo, afforded 6c (154 mg,
50.3%) as a white foam, which was stored under an atmosphere of
N2 in the freezer: mp (decomp.) 85°C; 1H NMR (CDCl3) ␦ 8.24 (t,
1H), 7.82 (d, 2H), 7.40 (t, 8H), 7.27 (d, 16H), 6.73 (d, 8H), 6.70 (d,
4H), 6.67 (d, 2H), 6.64 (m, 5H), 6.60 (t, 2H), 5.07 (s, 2H), 5.00 and
4.99 (2 s, 28H), 1.32 (s, 144H); IR (film) 2,963 (COH), 2,868
(COH), 2,143 (N3), 1,697 (CAO), 1,595 cm⫺1; MS (MALDI)
2,672.51 (rearrangement product ⫹ Na)⫹; Anal. Calcd for
C177H222N6O17: C, 78.57; H, 8.27; N, 3.11. Found: C, 78.30; H, 8.02;
N, 2.98.
Compound 7a. A solution of diazide 6a (600 mg, 0.790 mmol) in 10
ml of toluene was heated at reflux for 2.5 h or until the reaction was
complete, as indicated by the disappearance of the N3 stretch at
2,146 cm⫺1 and the appearance of the NCO stretch at 2,261 cm⫺1
in IR spectra. The resulting solution was cooled to RT, and solvent
was removed in vacuo to give approximately 560 mg (99%) of crude
diisocyanate, which was ⬎95% pure and used immediately in the
next step without further purification: 1H NMR ␦ 7.44 (t, 2H), 7.30
(d, 4H), 6.69 (m, 3H), 6.56 (d, 2H), 6.48 (t, 1H), 5.03 (s, 4H), 5.00
(s, 2H), 1.36 (s, 36H); 13C NMR ␦ 160.52, 160.07, 151.09, 138.10,
135.52, 135.24, 125.13, 122.36, 122.30, 113.99, 109.22, 106.23, 101.60,
71.04, 70.32, 34.84, 31.42; IR (KBr, cm⫺1) 2,962 (COH), 2,261
(NACAO). A suspension of the isocyanate (550 mg, 0.79 mmol)
and 3 (300 mg, 1.85 mmol) in 90 ml of THF was heated at reflux
for 48 h. The resulting suspension was filtered through a fine frit to
remove residual pyrimidinone starting material, and solvent was
removed in vacuo. The crude product was purified by column
chromatography (0–5% methanol兾CH2Cl2) and reprecipitated
from isopropanol兾toluene to give 390 mg (49%) of the title
compound as a white powder: mp ⬎265°C (dec); 1H NMR (⬇60%
DMSO-d6兾CDCl3, 50°C) ␦ 11.63 (br s, 2H), 10.20 (br s, 1H), 9.58
(br s, 1H), 8.75 (s, 2H), 8.35 (m, 1H), 7.29 (br s, 2H), 7.23 (s, 4H),
7.20 (s, 1H), 7.03 (m, 2H), 6.93 (br s, 1H), 6.70 (m, 2H), 6.60 (s, 1H),
5.00 (m, 14H), 1.28 (s, 36H); IR (KBr, cm⫺1) 3,200–3,300 (broad
NH band), 1,705 (CAO), 1,625 (CAO); UV ␭max (CH2Cl2, nm)
285, 313; MS (MALDI) 1,027.4 (M ⫹ H)⫹; Anal. Calcd for
C59H66N10O7: C, 68.99; H, 6.48; N, 13.64. Found: C, 68.70; H, 6.53;
N, 13.31.
Corbin et al.
Compound 7b. Following the same procedure as described for 7a
afforded 139 mg (29%) of 7b as a light yellow powder: mp
⬎260°C (dec); 1H NMR ␦ 13.39 (br m, 2H), 12.6–11 (br m, 2H),
9.0–8.0 (br m, 2H), 7.4–7.1 (m, 13H), 6.9–6.2 (m, 12H), 5.2–4.6
(m, 14H), 1.23 (br m, 72H); IR (nujol, cm⫺1) 3,100–3,300 (broad
NH), 1,702 (CAO), 1,623 (CAO); MS (MALDI) 1,678.0 (M ⫹
H)⫹, 1,700.3 (M ⫹ Na)⫹; Anal. Calcd for C103H122N10O11: C,
73.81; H, 7.34; N, 8.36. Found: 73.78; H, 7.33; N, 8.66.
Compound 7c. Following the same procedure as described for 7a
N-hydroxy succinimide (403 mg, 3.50 mmol), and dicyclohexylcarbodiimide (722 mg, 3.50 mmol) was stirred at RT for 10 h or until
no starting material remained by TLC. The resulting suspension was
filtered through a fine frit to remove dicyclohexyl urea (DCU), and
solvent was removed in vacuo. The solid was redissolved in approximately 50 ml of CH2Cl2 and filtered a second time to remove
residual DCU. Solvent was removed in vacuo, and the crude
product was reprecipitated from isopropanol to give 1.19 g (76%)
of the title compound as a white powder: mp 177–179°C; 1H NMR
␦ 8.53 (t, 1H), 8.03 (d, 2H), 7.43 (t, 2H), 7.31 (d, 4H), 6.75 (d, 2H),
6.70 (t, 1H), 5.15 (s, 2H), 5.04 (s, 4H), 2.94 (br s, 8H), 1.36 (s, 36H);
13C NMR ␦ 168.79, 160.58, 160.55, 159.22, 151.05, 137.41, 135.55,
127.40, 124.64, 122.59, 122.31, 106.43, 102.11, 71.05, 70.89, 34.84,
31.43, 25.61; IR (KBr, cm⫺1) 1,776 (CAO), 1,746 (CAO); MS
(MALDI) 901.97 (M ⫹ H)⫹, 925.44 (M ⫹ Na)⫹. Anal. Calcd for
C53H62N2O11: C, 70.49; H, 6.92; N, 3.10. Found: C, 70.22; H, 6.92;
N, 3.28.
Compound 11b. Using the same procedure described for 11a afforded 60 mg (53%) of product, which was determined to be ⬎95%
pure by 1H NMR spectroscopy: mp 143–145°C; 1H NMR ␦ 8.52
(t, 1H), 8.03 (d, 2H), 7.42 (t, 4H), 7.29 (d, 8H), 6.76 (d, 4H), 6.73
(d, 2H,), 6.67 (t, 2H), 6.66 (t, 1H), 5.14 (s, 2H), 5.03 (s, 12H), 2.90
(s, 8H), 1.35 (s, 72H); 13C NMR ␦ 168.76, 160.57, 160.43, 160.35,
159.14, 151.04, 138.86, 137.56, 135.66, 127.41, 124.70, 122.64, 122.34,
122.29, 106.46, 102.14, 101.59, 71.02, 70.85, 70.21, 34.84, 31.44,
25.60; MS (MALDI) 1,573.3 (M ⫹ Na)⫹.
X-Ray Analysis of Compound 5. A 0.24 ⫻ 0.12 ⫻ 0.03-mm colorless
crystal was obtained from CHCl3-methanol, MF ⫽ C29H25N5O2,
M 475.54, triclinic P-1, a ⫽ 9.613(2), b ⫽ 14.394(4), and c ⫽
18.872(5) Å, ␣ ⫽ 75.323(5)°, ␤ ⫽ 84.653(6)°, and ␥ ⫽ 70.763(5)°,
V ⫽ 2,384.9(10) Å3, ␮(M␱ K␣) ⫽ 0.086 mm⫺1, Z ⫽ 4, ␳calc ⫽
1.324 g䡠cm⫺3, F (000) ⫽ 1,000. Data were collected with a Bruker
SMART兾CCD diffractometer. A total of 14,598 reflections were
collected with 5,002 independent ones [R(int) ⫽ 0.1937] used for
refinement. Structure was solved by direct methods. Final R [I
⬎2␴(I)], R1 ⫽ 0.0805, wR2 ⫽ 0.1595, R (all data) ⫽ R1 ⫽ 0.2638,
wR2 ⫽ 0.2371 with data兾restraints兾parameters ⫽ 5,002兾0兾649.
Goodness of fit on F2 ⫽ 0.928.
Results and Discussion
Previous 1H NMR studies in apolar organic solvents indicated
that 1 forms three dimers (1䡠1, 1⬘䡠1⬘, and 1䡠1⬘) of comparable
stability (14). There were no changes in the 1H NMR spectra of
1 when diluting samples in chloroform-d to the limits of detection
at 750 MHz (12 ␮M). This observation provided the lower limit
Corbin et al.
Structure of compounds 4 and 5.
on the dimerization constant indicated above (i.e., Kdimer ⬎ 107
M⫺1). To more accurately establish the strength of dimerization,
compound 4 (Fig. 2) was synthesized and studied by fluorescence
spectroscopy. This method, recently used by Sijbesma and
colleagues (21) to measure the strong dimerization of a module
similar to 1, takes advantage of the excimer signal of the pyrene
dimer that exhibits an emission band at 500–600 nm that is well
separated from that of the monomers. In chloroform saturated
with water, the Kdimer of 4 was 3.0 (⫾0.4) ⫻ 107 M⫺1; whereas in
freshly opened chloroform ([water] ⬇17 mM) Kdimer ⫽ 8.5
(⫾1.6) ⫻ 107 M⫺1. In chloroform carefully dried over calcium
chloride, no decrease in the emission intensity of 4 was observed
from 10⫺6 to 10⫺9 M. Assuming that a 10% change in fluorescence intensity would be detected a lower limit can be placed on
the dimerization constant: Kdimer ⬎5 ⫻ 108 M⫺1.
Numerous attempts to crystallize 1 and other N-alkyl analogs
were unsuccessful. Given the well-known crystallinity of triphenylmethyl (trityl) groups, and their ability to render heterocyclic compounds crystalline that otherwise form powders (19),
compound 5 (Fig. 2) was synthesized from 3 and 3,3,3triphenylpropylisocyanate. Crystals of 5 suitable for x-ray analysis were grown by slow evaporation of a methanol-chloroform
solution of 5. The solid-state structure shows a homodimer of the
N(3H) protomer held together by four intermolecular hydrogen
bonds (Fig. 3). The individual components are further rigidified
by an intramolecular hydrogen bond from N3-H to O-4.
The exceptionally high dimerization constant of 1 and the
established pairing motif makes it an ideal candidate for nanoscale
construction by using hydrogen bond-mediated self-assembly. To
explore this possibility, two units of module 3 were linked by using
6 to give ditopic compounds 7 a–c (Figs. 4 and 5). Thus, alkylation
of 8 with first-, second-, and third-generation dendritic bromides
Fig. 3. X-ray structure of 5 showing a dimeric hydrogen-bonding motif. See
Materials and Methods for details.
PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5101
CHEMISTRY
Compound 11a. A solution of diacid 10a (1.23 g, 1.73 mmol),
Fig. 2.
SPECIAL FEATURE
afforded 59.4 mg (57.2%) of 7c as a white powder: 1H NMR (500
MHz, d9-NMP, referenced at ␦ 2.96 for residual NCH3 signal) ␦
12.08 (s, 2H), 10.67 (s, 2H), 9.11 (br s, 2H), 9.10 (br s, 2H), 8.68
(d, 2H), 8.02 (br s, 2H), 7.85–7.28 (m, 24H), 7.22–6.61 (m, 23H),
6.47 (s, 1H), 5.40 and 5.37 (2 br s, 30H), 1.44 (br s, 144H); IR
(film, cm⫺1) 3,100–3,300 (br NH), 1,702 (CAO), 1,623 (CAO);
UV ␭max (toluene, nm) 313; MS (MALDI): 2,972.7 (M ⫹ H)⫹,
3,012.5 (M ⫹ K)⫹. Anal. Calcd for C191H234N10O19: C, 77.14; H,
7.93; N, 4.71. Found: C, 77.24; H, 7.91; N, 4.68.
Fig. 4.
Structure of compounds 7, 13, and 14.
(Gn-Br, n ⫽ 1–3) provided the corresponding isophthalate esters 9
a–c, which were hydrolyzed to diacids 10 a–c. Conversion of 10 a
and b to acyl azides 6 a and b proceeded through succinimide esters
11 a and b, whereas the conversion of 10c to 6c was most
conveniently effected through the intermediacy of acid chloride
12c. Treatment of 6 a–c with 3 gave 7 a–c in moderate yield. The
Fréchet-type dendrons (22) solublize the heterocyclic unit, but
more importantly allow direct comparison to previously reported
self-assembling dendrimers 13 and 14 (16–18).
Fig. 5. Synthesis of 7. Key: Gn ⫽ dendrons a--c in Fig. 4. Suc ⫽ N-succinimide
(a) Gn-Br, K2CO3, acetone, 18-cr-6, reflux, (9a: 72%; 9b: 82%; 9c: 50%). (b) KOH,
THF, MeOH, H2O, reflux (10a: 98%; 10b: 81%; 10c: 80%). (c) For 10 a and b:
HOSuc, dicyclohexylcarbodiimide, dioxane (11a: 76%; 11b: 53%). For 10c:
(COCl)2, THF, RT350°C, dimethylformamide (cat). (d) For 11 a and b: NaN3 (aq),
acetone, THF (6a: 81%; 6b: 85%). For 12: TMSN3, RT (50%, two steps). (e)
PhCH3, reflux; 3, THF, reflux (7a: 49%; 7b: 29%; 7c: 57%).
5102 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.062641199
Fig. 6. Scheme showing different conformations of spacer unit of 7 and its
affect on the assembly process.
The ureidodeazapterin units of 7 may adopt both N(3H) and
N(1H) protomeric forms; however, both hetero- and homodimers of this unit maintain the same spatial arrangement of
the R-substituents (see 1䡠1 and 1⬘䡠1⬘ in Fig. 1). Thus, this
variability does not provide information that can be retrieved in
assembly. In contrast, the spacer unit may adopt symmetrical (7)
and nonsymmetrical conformations (7⬘), and this information is
likely to be expressed by the formation of either polymeric or
cyclic aggregates (Fig. 6).
We as well as others have shown that size-exclusion chromatography (SEC) using appropriate synthetic size standards can be a
powerful method for determining the structure of solution aggregates (2, 16-18). Thus, the self-assembly of 7 a–c was studied by
SEC, with most of the experiments focused on the first- and
third-generation dendrons (i.e., 7 a and c). Despite the potential for
forming many different aggregates, 7a and 7c gave sharp, symmetrical peaks (Fig. 7) as did 7b (data not shown). The polydispersity
index for 7a and 7c, was calculated to be 1.07 and 1.03, respectively,
suggesting that these two compounds form discrete aggregates in
toluene. Open aggregates such as (7)n would exhibit concentrationdependent sizes. The polystyrene equivalent molecular weights of
such aggregates can be calculated by using the Carothers equation
if an isodesmic model is assumed wherein each Kn is identical and
approximately equal to the Kdimer (4). Thus, concentrationindependent SEC retention can be taken as prima facie evidence
that the discrete aggregates are closed (i.e., cyclic). As shown in Fig.
8, the SEC-derived molecular weight values for both 7a and 7c
minimally changed across a nearly 104-fold dilution range. If open
aggregates were present, the molecular weights would have changed
significantly, e.g., from ⬍5,000 to ⬎80,000 for 7a.
The structure of each closed aggregate was inferred from its
size, which in turn was determined with both polystyrene and
Corbin et al.
Fig. 7. (A) Overlay of SEC traces of 7c and 9c.
Toluene, 1 ml兾min, double Waters Ultrastyragel
HR3 column (molecular weight 500 –30,000). (B)
Overlay of SEC traces of 7a and 16a, same conditions, single column.
Fig. 8. Change in Mn as a function of concentration (at SEC injector).
Compound 7a (Œ), toluene; 7c (}), THF, (■), toluene.
Corbin et al.
CHEMISTRY
VPO method. The findings can be explained by a preference for
the bent conformation 7⬘, which minimizes alignment of the urea
carbonyl dipoles, thereby preorganizing the system for cyclic
assembly. Although a head-to-tail pairing of the conformationally heterotopic units gives hexamer (7⬘)6 (see Fig. 6), molecular
modeling suggests that there are better optimized hydrogen
bonds in larger aggregates that contain some head-to-head
contacts. That 7b and 7c are unable to form such aggregates can
be clearly seen by noting that 17 has three dendrons that are
forced into the interior of the structure. Modeling indicates that
SPECIAL FEATURE
synthetic size standards—the latter including dendrimers 15 and
16 (Fig. 9) and previously reported hexameric aggregates (13
a–c)6 and (14a)6 (16–18). As shown in Table 1, the polystyrenederived Mn values were consistent with formation of hexamer
from 7b and 7c [see (7⬘)6] and formation of a dodecamer such
as 17 (Fig. 10) from 7a. More importantly the retention times of
7 a–c in comparison to that of known hexamers 13 a–c and 14 a–c
and dendrimer standards 9 a–c, 15 b and c, and 16 a and b were
appropriate given their dimensions obtained by molecular modeling. For example, 7c elutes well before its monomeric analog
9c (Fig. 7A) and somewhat earlier than hexamer (14c)6, but after
hexamer (13c)6. The average modeled diameter of the (7c)6
heterocyclic core is 30.6 Å, which is between the core diameters
of (13c)6 (41.1 Å) and (14c)6 (22.7 Å). Likewise, 7b eluted
between 13b and 14b, and with a retention volume close to but
before 16b. However, 7a elutes much earlier than corresponding
unimolecular control 16a (Fig. 7B). Instead, the elution times for
the aggregate of 7a and the control 15b were nearly identical.
Likewise, a single relatively sharp peak was observed when
coinjecting the two components. The modeled sizes of (7a)12
(i.e., 17) and 15b are nearly identical. Attempts to grow crystals
of 7 a–c or image their assemblies by using microscopy have, thus
far, been unsuccessful.
There are several striking features of the aforementioned
assembly process. Most notable is the preference for cyclic
aggregates and the unexpectedly large aggregate formed from
7a. The latter finding was confirmed by vapor pressure osmometry (VPO); the benzil-calibrated Mn ⫽ 16,500 is likely within
error of the SEC molecular weight given the uncertainties in the
Fig. 9.
Structures of SEC size standards 15 and 16, R ⫽ Gn in Fig. 4.
PNAS 兩 April 16, 2002 兩 vol. 99 兩 no. 8 兩 5103
Table 1. Calculated and experimental molecular weights
Molecular weight (MW)
Compound
Calcd
MALDI*
SEC†
MW (n-mer)
7a
1,026.5
1,027.4‡
13,072
1,674.9
2,971.8
2,681.7
3,176.8
783.3
2,727.7
9,144.9
16,932.0
4,794.0
8,688.3
1,678.0‡
2,972.7‡
2,706.4§
3,200.0§
784.4‡
2,751.0§
9,168.5§
13,004
17,777
3,217
20,674
7,668
11,500
12,362
20,791
7,045
12,300
6,163 (6-mer)
12,326 (12-mer)
10,057 (6-mer)
17,844 (6-mer)
7b
7c
9c
13c
14a
14c
15b
15c
16a
16b
4,818.0§
8,718.0§
19,061 (6-mer)
4,698 (6-mer)
16,366 (6-mer)
*m兾z from MALDI spectra.
†Polystyrene-derived M with toluene eluent, (Waters Ultrastyragel HR3 coln
umn, MW range 500 –30,000) UV and refractive index detection.
‡(M ⫹ H)⫹ signal.
§(M ⫹ Na⫹) signal.
the first-generation dendrons from three units of 7a may be just
accommodated inside 17. If this working hypothesis is correct, a
more polar solvent should stabilize the linear conformation,
which would favor polymeric aggregates. Strikingly, two peaks
were observed in the SEC traces of 7a in dilute (⬍10⫺4 M) THF
solutions. One peak corresponded to the same species seen in
toluene (e.g., 12-mer), whereas the other had a very high
polystyrene-based molecular weight (⬎85,000) consistent with
formation of polymer (7)n (12, 17, 23).
Another notable feature of the assemblies formed by 7 a–c is
their exceptional stability—in particular the hexamer formed from
7c. Whereas 13 a–c and 14 b and c are monomeric in moderately
competitive solvents such as THF and 14a dissociates at high
dilution, 7c does not dissociate in THF even upon dilution to 0.6
␮M—the limit of detection by SEC (Fig. 8). Only in highly polar
solvents such as N-methylpyrrolidinone does the equilibrium shift
to monomer. The high stability of the aggregates clearly originates
in the high stability of the ureidodeazapterin dimers. For comparison, the benzoic acid dimer used to assemble 13 exhibits Kdimer
⬇102 M⫺1 in chloroform; whereas, the DDA䡠AAD hydrogenbonding motif used by 14 has a Kassoc ⬇104–105 M⫺1 (4, 13). Thus,
both contacts exhibit interaction constants that are orders of
magnitude lower than the Kdimer ⬇108 M⫺1 of 1䡠1.
Can the high stability of the ureidodeazpterine dimer translate
into high fidelity in the assembly process? To answer this question
approximately equal volumes of toluene solutions of 7c and 14a
were mixed (about 1 mM final concentration) and periodically
analyzed by SEC over the course of 53 days. Throughout the period,
two discrete peaks were observed corresponding to (7c)6 and
(14a)6. The two hydrogen-bonding arrays within 14, AAD and
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
Eigen, M. & De Maeyer, L. (1966) Naturwissenschaften 53, 50–57.
Whitesides, G. M., Mathias, J. P. & Seto, C. T. (1991) Science 254, 1312–1319.
Lehn, J.-M. (1988) Angew. Chem. Int. Ed. 27, 89–112.
Zimmerman, S. C. & Corbin, P. S. (2000) Struct. Bond. 96, 63–94.
Leininger, S., Prins, L. J., Timmerman, P. & Reinhoudt, D. N. (1998) Pure Appl.
Chem. 70, 1459–1468.
Conn, M. M. & Rebek, J., Jr. (1997) Chem. Rev. 97, 1647–1668.
Fan, E., Vicent, C., Geib, S. J. & Hamilton, A. D. (1994) Chem. Mater. 6, 1113–1117.
Raymo, F. M. & Stoddart, J. F. (1999) Chem. Rev. 99, 1643–1663.
Fujita, M. & Ogura, K. (1996) Coord. Chem. Rev. 148, 249–264.
Olenyuk, B. & Stang, P. J. (2000) Chem. Rev. 100, 853–907.
Zimmerman, N., Moore, J. S. & Zimmerman, S. C. (1998) Chem. Ind. (London) 15,
604–610.
Sijbesma, R. P., Beijer, F. H., Brusveld, L., Folmer, B. J. B., Hirschberg, K., Lange,
R. F. M., Lowe, J. K. L. & Meijer, E. W. (1997) Science 278, 1601–1604.
Murray, T. J. & Zimmerman, S. C. (1992) J. Am. Chem. Soc. 114, 4010–4011.
5104 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.062641199
Fig. 10.
Structure of dodecameric assembly 17.
DDA, are complementary to three of the four adjacent donoracceptor sites found in the AADD array of 7, yet no mixing occurs.
To determine whether this observation originates in a kinetic or
thermodynamic effect, solid 7c and 14a were ground together and
then dissolved in toluene. Additionally, a mixture of 7c and 14a in
N-methylpyrrolidinone was evaporated and redissolved in toluene.
Both of these solutions were analyzed over time by SEC and found
to contain only the separate hexamer peaks corresponding to (7c)6
and (14a)6. Thus, mispairings that lead to kinetically trapped
intermediates do not form; the coded information within these two
molecules is correctly expressed in their assembly.
Conclusion
The very strong hydrogen bond-mediated dimerization of the
deazapterin units in 7 allows it to assemble into aggregates of
exceptionally high stability. The results described here demonstrate how the subtle interplay of steric effects, solvent, and
conformation controls the self-assembly process. As future
systems are developed with multiple information codes it will be
essential that the retrieval process be accurate. That the complementary pairings studied here take place faithfully in the
presence of very similar hydrogen-bonding motifs bodes well for
the development of such systems.
We thank Scott Wilson for assistance with the x-ray analysis. This work
was supported by National Institutes of Health Grant GM39782.
14. Corbin, P. S. & Zimmerman, S. C. (1998) J. Am. Chem. Soc. 120, 9710–9711.
15. Pfleiderer, M. & Pfleiderer, W. (1992) Heterocycles 33, 905–929.
16. Zimmerman, S. C., Zeng, F., Reichert, D. E. C. & Kolotuchin, S. V. (1996) Science
271, 1095–1098.
17. Zeng, F., Zimmerman, S. C., Kolotuchin, S. V., Reichert, D. E. C. & Ma, Y. (2002)
Tetrahedron 58, 825–843.
18. Zimmerman, S. C. & Kolotuchin, S. V. (1998) J. Am. Chem. Soc. 120, 9092–9093.
19. Petersen, P., Wu, W., Fenlon, E. E., Kim, S. & Zimmerman, S. C. (1996) Bioorg. Med.
Chem. 4, 1107–1112.
20. Taylor, E. C., Otiv, S. R. & Durucasu, I. (1993) Heterocycles 36, 1883–1895.
21. Söntjens, S. H. M., Sijbesma, R. P., van Genderen, M. H. P. & Meijer, E. W. (2000)
J. Am. Chem. Soc. 122, 7487–7493.
22. Hawker, C. J. & Fréchet, J. M. J. (1990) J. Am. Chem. Soc. 112, 7638–7647.
23. Castellano, R. K., Rudkevich, D. M. & Rebek, J., Jr. (1997) Proc. Natl. Acad. Sci. USA
94, 7132–7137.
Corbin et al.