Insights into templated supramolecular polymerization: Binding of naphthalene derivatives to ssDNA templates of different lengths

Article abstract of DOI:10.1021/ja808075h

We report on two diaminotriazine-equipped naphthalene derivatives that bind reversibly to a single-stranded DNA template or "tape-measure molecule" via hydrogen bonding, yielding monodisperse double-stranded DNA hybrids with one strand consisting of a sup

Full text of DOI:10.1021/ja808075h

Published on Web 12/29/2008
Insights into Templated Supramolecular Polymerization:
Binding of Naphthalene Derivatives to ssDNA Templates of
Different Lengths
Pim G. A. Janssen,†,§ Sara Jabbari-Farouji,‡,§,| Mathieu Surin, Xavier Vila,†
Jeroen C. Gielen,# Tom F. A. de Greef,†,§ Matthijn R. J. Vos,† Paul H. H. Bomans,†
Nico A. J. M. Sommerdijk,† Peter C. M. Christianen,# Philippe Lecle`re,†,
Roberto Lazzaroni, Paul van der Schoot,‡ E. W. Meijer,*,†,§ and
Albertus P. H. J. Schenning*,†,§
Laboratory for Macromolecular and Organic Chemistry, Group Theoretical and Polymer
Physics, and Institute for Complex Molecular Systems, EindhoVen UniVersity of Technology,
P.O. Box 513, 5600 MB EindhoVen, The Netherlands, Dutch Polymer Institute,
P.O. Box 902, 5600 AX EindhoVen, The Netherlands, Laboratory for Chemistry of NoVel
Materials, UniVersity of Mons-Hainaut, B-7000 Mons, Belgium, and High Field Magnet
Laboratory, Institute of Molecules and Materials, Radboud UniVersity Nijmegen,
ToernooiVeld 7, 6525 ED Nijmegen, The Netherlands
Received October 14, 2008; E-mail: e.w.meijer@tue.nl; a.p.h.j.schenning@tue.nl
Abstract: We report on two diaminotriazine-equipped naphthalene derivatives that bind reversibly to a
single-stranded DNA template or “tape-measure molecule” via hydrogen bonding, yielding monodisperse
double-stranded DNA hybrids with one strand consisting of a supramolecular naphthalene backbone. These
assemblies have been investigated extensively, both experimentally and theoretically. The structure and
the templated self-assembly process of the complex have been characterized with UV-vis spectroscopy,
circular dichroism spectroscopy, molecular dynamics simulations, cryo-transmission electron microscopy,
1
liquid atomic force microscopy, electrospray ionization mass spectrometry, light scattering, and H NMR
and infrared spectroscopy. We have found that the DNA hybrid complexes have a right-handed helical
arrangement stabilized by π-π interactions and hydrogen bonds. The hydrophilic hydroxyl group at the
end of the ethylene glycol of the guest molecule suppressed both the nontemplated self-assembly of the
naphthalene guest molecules and the further aggregation of the entire DNA hybrid complex. Through
the use of a theoretical mass-action model for the templated self-assembly, the host-guest and guest-guest
interaction energies were estimated by fitting to the spectroscopic data. The differently estimated values of
the interaction energies and thermodynamic parameters vary within experimental error, showing the self-
consistency of the model. From the obtained correlation between the positions of the guest molecules
bound on the template, we have obtained a qualitative theoretical picture of the way in which the guests
are physically distributed on the templates. For short templates, the templates are filled one-by-one, even
at moderate fractions of bound sites. For larger templates, the templates first have alternating sequences
of filled and empty sections, after which, at large fractions of bound sites, virtually all of the binding sites
for all template lengths are filled.
for the assembly of coat proteins of the tobacco mosaic virus,1
while its sequence can also be transcribed into proteins via the
Introduction
In nature, templates with specific binding sites are used to
efficiently form assemblies and polymers with defined sizes or
sequence. For instance, RNA acts as a “tape-measure” template
specific recognition of codons.2 Nature’s concept has inspired
many researchers to exploit templated polymerization3,4 as a
tool for controlling the size and sequence of synthetic polymers
viaa“bottom-up”approach.5Asaresult,viruses,6biopolymers,3,4f,g
and synthetic polymers4b,c have been used as templates for the
self-assembly of a variety of biological, organic, and inorganic
building blocks. For example, synthetic poly(trimethylene
iminium)4b and oligo(phenylene ethynylene)4c templates have
† Laboratory for Macromolecular and Organic Chemistry, Eindhoven
University of Technology.
‡ Group Theoretical and Polymer Physics, Eindhoven University of
Technology.
§ Institute for Complex Molecular Systems, Eindhoven University of
Technology.
(1) (a) Special Issue: Tobacco Mosaic Virus: Pioneering Research for a
Century. Philos. Trans. R. Soc London, Ser. B 1999, 354, 517-685.
(b) Kittel, C. Am. J. Phys. 1969, 37, 917–920.
| Dutch Polymer Institute.
Laboratory for Chemistry of Novel Materials, University of Mons-
Hainaut.
(2) Stryer, L. Biochemistry, 4th revised ed.; Freeman: New York, 1996;
p 1100.
# Institute of Molecules and Materials, Radboud University Nijmegen.
9
1222 J. AM. CHEM. SOC. 2009, 131, 1222–1231
10.1021/ja808075h CCC: $40.75
2009 American Chemical Society

Insights into Templated Supramolecular Polymerization
A R T I C L E S
been used to organize porphyrins between two strands and to
bind amphiphilic peptides, respectively. Biopolymers, such as
RNA and DNA, have also been used as templates to mimic
their replication nonenzymatically3b and to create non-natural
polymers with a predefined sequence.3c Similarly, the DNA-
templated self-assembly of nanoparticles,7 chromophores,8 and
lipids8 has been reported. Apart from being templates, DNA
and functionalized nucleotides9 can be used as building blocks
to create predefined complex nanosized structures via sticky-
end cohesion.10
For the exploitation of templated polymerizations, it is
essential to develop a deeper understanding of the binding
interaction between the host template and the guest molecules
and that between the guest molecules.3,4,11 For example, to
obtain a so-called all-or-nothing process in which complete
templates are filled one by one, the guest-guest and host-guest
interactions cannot be arbitrary. When the host-guest interaction
is strong and the guest-guest interaction weak, the guest
molecules are uniformly dispersed over all of the available
template binding sites and therefore do not entirely fill any of
the templates, except of course when the overall template
coverage is very high. In the other extreme, a strong guest-guest
interaction could lead to a competition between templated and
nontemplated self-assembly and to very polydisperse aggregates
that are potentially larger on average than the template size.
We have previously shown that two water-soluble guest
report on the templated self-assembly of a newly designed
naphthalene guest molecule G2 on oligothymine templates of
various lengths (denoted as dTq, where q is the number of
thymines), yielding objects whose size is controlled by the length
of the template (Scheme 1). We have redesigned the naphthalene
guest molecule G1 to suppress the nontemplated self-assembly
by replacing the methyl-terminated ethylene oxide with a
hydroxyl-terminated ethylene oxide, yielding guest molecule
G2.13 The self-assembled structures are characterized in detail
with molecular dynamics (MD) simulations, dynamic light
scattering (DLS), and microscopy. Furthermore, we have
developed a theoretical model for guests adsorbing to a linear
template with a finite size in order to gain insight into the
templated supramolecular polymerization process. With this
model, we were able to analyze our experimental findings and
obtain binding energies for both the host-guest and guest-guest
interactions. With the obtained guest-guest interaction energy,
we can determine the correlation length, i.e., the length of
correlated template binding sites, which predicts the way that
the guests are distributed on the template as a function of the
template length.14
Results and Discussion
Nontemplated Self-Assembly. We first compare the nontem-
plated self-assembly of G2 with the earlier reported data for
molecules,
a naphthalene derivative G1 and an oligo
(6) Examples where viruses act as templates: (a) Huang, Y.; Chiang, C.-
Y.; Lee, S. K.; Gao, Y.; Hu, E. L.; De Yoreo, J.; Belcher, A. M. Nano
Lett. 2005, 5, 1429–1434. (b) Mao, C.; Solis, D. J.; Reiss, B. D.;
Kottmann, S. T.; Sweeney, R. Y.; Hayhurst, A.; Georgiou, G.; Iverson,
B.; Belcher, A. M. Science 2004, 303, 213–217. (c) Sotiropoulou, S.;
Sierra-Sastre, Y.; Mark, S. S.; Batt, C. A. Chem. Mater. 2008, 20,
821–834.
(p-phenylene)vinylene diaminotriazine derivative, under ap-
propriate conditions bind to a complementary oligothymine
strand (dT40 having 40 thymines), as evidenced by optical and
electrospray ionization (ESI) studies (Scheme 1).12 We observed
that at high concentrations of the guests, nontemplated self-
assembly interferes with the templated self-assembly. We here
(7) Examples of DNA as a structural scaffold to organize particles: (a)
Pinto, Y. Y.; Le, J. D.; Seeman, N. C.; Musier-Forsyth, K.; Taton,
T. A.; Kiehl, R. A. Nano Lett. 2005, 5, 2399–2402. (b) Tang, Z.;
Kotov, N. A. AdV. Mater. 2005, 17, 951–962. (c) Nykypanchuk, D.;
Maye, M. M.; van der Lelie, D.; Gang, O. Nature 2008, 451, 549–
552. (d) Zhang, J.; Liu, Y.; Ke, Y.; Yan, H. Nano Lett. 2006, 6, 248–
251. (e) Gothelf, K. V.; LaBean, T. H. Org. Biomol. Chem. 2005, 3,
4023–4037. (f) Park, S. H.; Yin, P.; Liu, Y.; Reif, J. H.; LaBean, T. H.;
Yan, H. Nano Lett. 2005, 5, 729–733.
(3) Examples and reviews of templated polymerizations: (a) Gons, J.;
Vorenkamp, E. J.; Challa, G. J. Polym. Sci., Part A: Polym. Chem.
1975, 13, 1699–1709. (b) Inoue, T.; Orgel, L. E. Science 1983, 219,
859–862. (c) Inoue, T.; Joyce, G. F.; Grzeskowiak, K.; Orgel, L. E.;
Brown, J. M.; Reese, C. B. J. Mol. Biol. 1984, 178, 669–676. (d)
Kleiner, R. E.; Brudno, Y.; Birnbaum, M. E.; Liu, D. R. J. Am. Chem.
Soc. 2008, 130, 4646–4659. (e) Saito, R. Polymer 2008, 49, 2625–
2631. (f) van Hest, J. C. M. Nat. Chem. Biol. 2008, 4, 272–273. (g)
Clapper, J. D.; Sievens-Figueroa, L.; Guymon, C. A. Chem. Mater.
2008, 20, 768–781.
(8) (a) Iwaura, R.; Hoeben, F. J. M.; Masuda, M.; Schenning, A. P. H. J.;
Meijer, E. W.; Shimizu, T. J. Am. Chem. Soc. 2006, 128, 13298–
13304. (b) Iwaura, R.; Yoshida, K.; Masuda, M.; Ohnishi-Kameyama,
M.; Yoshida, M.; Shimizu, T. Angew. Chem., Int. Ed. 2003, 42, 1009–
1012. (c) Iwaura, R.; Kikkawa, Y.; Ohnishi-Kameyama, M.; Shimizu,
T. Org. Biomol. Chem. 2007, 5, 3450–3455.
(4) Examples of templated supramolecular polymerizations: (a) Lindsey,
J. S. New J. Chem. 1991, 15, 153–180. (b) Sugimoto, T.; Suzuki, T.;
Shinkai, S.; Sada, K. J. Am. Chem. Soc. 2007, 129, 270–271. (c) Bull,
S. R.; Palmer, L. C.; Fry, N. J.; Greenfield, M. A.; Messmore, B. W.;
Meade, T. J.; Stupp, S. I. J. Am. Chem. Soc. 2008, 130, 2742–2743.
(d) Ohkawa, H.; Ligthart, G. B. W. L.; Sijbesma, R. P.; Meijer, E. W.
Macromolecules 2007, 40, 1453–1459. (e) Hoogenboom, R.; Fournier,
D.; Schubert, U. S. Chem. Commun. 2008, 155–162. (f) de la Escosura,
A.; Verwegen, M.; Sikkema, F. D.; Comellas-Aragones, M.; Kirilyuk,
A.; Rasing, T.; Nolte, R. J. M.; Cornelissen, J. J. L. M. Chem.
Commun. 2008, 1542–1544. (g) Xu, Y.; Ye, J.; Liu, H.; Cheng, E.;
Yang, Y.; Wang, W.; Zhao, M.; Zhou, D.; Liu, D.; Fang, R. Chem.
Commun. 2008, 49–51. (h) Benvin, A. L.; Creeger, Y.; Fisher, G. W.;
Ballou, B.; Waggoner, A. S.; Armitage, B. A. J. Am. Chem. Soc. 2007,
129, 2025–2034. (i) Hunter, C. A.; Tomas, S. J. Am. Chem. Soc. 2006,
128, 8975–8979.
(9) Examples of chromophores covalently linked to DNA: (a) Krueger,
A. T.; Lu, H.; Lee, A. H. F.; Kool, E. T. Acc. Chem. Res. 2007, 40,
141–150. (b) Kool, E. T. Acc. Chem. Res. 2002, 35, 936–943. (c)
Lewis, F. D.; Zhang, L.; Zuo, X. J. Am. Chem. Soc. 2005, 127, 10002–
10003. (d) Balaz, M.; Holmes, A. E.; Benedetti, M.; Rodriguez, P. C.;
Berova, N.; Nakanishi, K.; Proni, G. J. Am. Chem. Soc. 2005, 127,
4172–4173. (e) Kashida, H.; Asanuma, H.; Komiyama, M. Angew.
Chem., Int. Ed. 2004, 43, 6522–6525.
(10) Examples of 2D and 3D DNA structures: (a) Seeman, N. C. Methods
Mol. Biol. 2005, 303, 143–166. (b) Chen, J.; Seeman, N. C. Nature
1991, 350, 631–633. (c) Rothemund, P. W. K. Nature 2006, 440, 297–
302. (d) He, Y.; Ye, T.; Su, M.; Zhang, C.; Ribbe, A. E.; Jiang, W.;
Mao, C. Nature 2008, 452, 198–201.
(11) Recent reviews of template-directed reactions: (a) Pianowski, Z. L.;
Winssinger, N. Chem. Soc. ReV. 2008, 37, 1330–1336. (b) MacGillivray,
L. R. J. Org. Chem. 2008, 73, 3311–3317. (c) Meyer, C. D.; Joiner,
C. S.; Stoddart, J. F. Chem. Soc. ReV. 2007, 36, 1705–1723. (d)
Silverman, A. P.; Kool, E. T. Chem. ReV. 2006, 106, 3775–3789.
(12) (a) Janssen, P. G. A.; Vandenbergh, J.; van Dongen, J. L. J.; Meijer,
E. W.; Schenning, A. P. H. J. J. Am. Chem. Soc. 2007, 129, 6078–
6079. (b) Janssen, P. G. A.; van Dongen, J. L. J.; Meijer, E. W.;
Schenning, A. P. H. J. Chem.sEur. J. [Online early access]. DOI:
10.1002/chem.200801506. Published Online: Nov 28, 2008.
(13) Kuehne, R.; Ebert, R. U.; Kleint, F.; Schmidt, G.; Schueuermann, G.
Chemosphere 1995, 30, 2061–2077.
(5) Recent examples and reviews of bottom-up approaches: (a) Service,
R. F. Science 2005, 309, 95. (b) Schenning, A. P. H. J. Synth. Met.
2004, 147, 43–48. (c) Hoeben, F. J. M.; Jonkheijm, P.; Meijer, E. W.;
Schenning, A. P. H. J. Chem. ReV. 2005, 105, 1491–1546. (d) Ikkala,
O.; ten Brinke, G. Chem. Commun. 2004, 2131–2137. (e) Ajayaghosh,
A.; Praveen, V. K. Acc. Chem. Res. 2007, 40, 644–656. (f) Ligthart,
G. B. W. L.; Ohkawa, H.; Sijbesma, R. P.; Meijer, E. W. J. Am. Chem.
Soc. 2005, 127, 810–811. (g) Knoben, W.; Besseling, N. A. M.; Stuart,
M. A. C. Macromolecules 2006, 39, 2643–2653. (h) Jonkheijm, P.;
van der Schoot, P.; Schenning, A. P. H. J.; Meijer, E. W. Science
2006, 313, 80–83. (i) Smulders, M. M. J.; Schenning, A. P. H. J.;
Meijer, E. W. J. Am. Chem. Soc. 2008, 130, 606–611.
(14) Zimm, B. H.; Bragg, J. K. J. Chem. Phys. 1959, 31, 526–535.
9
J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009 1223

A R T I C L E S
Janssen et al.
Figure 1. (a) UV spectra of a 1 mM G2 solution at 263 and 283 K in water. The inset shows the fraction of aggregated G2 measured at 269 nm and the
fit based on the previously developed model for cooperative self-assembly.5h (b) Cryo-TEM image of a 1 mM G1 solution in water vitrified at 283 K.
Figure 2. (a) Normalized CD spectra of G-dT40 mixtures in water. (b) Normalized UV-vis spectra of G-dT40 mixtures in water. (c) Determination of
Tp and Tte for [dT40] ) 6.25 µM and [G2] ) 0.5 mM. The CD intensity and absorption were measured at λ ) 340 nm. (d) Te and Tte as functions of the
guest concentration determined via temperature-dependent UV-vis measurements on G and 1:40 dT40-G mixtures, respectively, at 269 nm.
Scheme 1. Schematic Representation of the ssDNA-Templated Self-Assembly and the Molecular Structures of the Host Template dTq and
the Guests G1 and G2
G112 by monitoring the hypochromicity upon cooling for
different concentrations (Figures 1a and 2d).5h,i Similar to G1
at a concentration of 1 mM,12 G2 can aggregate at low
temperatures, as evidenced by the hypochromicities at 269 and
320 nm and the scattering above 370 nm (Figure 1a). When
the sample is cooled while the UV absorption is continuously
monitored, a sharp exponential transition is observed that is
indicative of a cooperative self-assembly process.5h,i We call
this transition temperature the elongation temperature, Te, at
which the self-assembly starts.5h,i Compared to G1, G2 exhibits
the same kind of cooperative self-assembly, albeit at a Te value
that is 10 °C lower at the same concentration of dissolved
material. The enthalpy of elongation, ∆He, which is calculated
from the slope of a plot of the natural logarithm of the guest
concentration [G] versus 1/Te, is very similar for G2 (-28 kJ/
mol) and G1 (-27 kJ/mol).15 Cryo-transmission electron
9
1224 J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009

Insights into Templated Supramolecular Polymerization
A R T I C L E S
microscopy (cryo-TEM) images of a G1 solution with a 1 mM
concentration (Figure 1b) revealed micrometer-long laterally
interacting fibers of G1 with a width of 3 nm, which corresponds
to the length of one G1 molecule. A similar solution of G2
does not show fibers, since G2 is molecularly dissolved at this
temperature and concentration (Figure 1a).
Templated Self-Assembly. When G2 was mixed with dT40
in a 40:1 ratio in water at a concentration at which G2 is
molecularly dissolved, a hypochromicity, red-shift, and Cotton
effect in the naphthalene absorption region similar to those
reported previously for the right-handed G1-dT40 complex
were observed (Figure 2a,b).16 Furthermore, the absorption onset
for the G1-dT40 mixture is red-shifted compared with that of
the G2-dT40 mixture, which can possibly be attributed to
scattering due to large particles in the G1-dT40 solution (see
the discussion of the light-scattering experiments, below). To
compare the reversible templated self-assembly of G1 and G2,
the assembly process was monitored by temperature-dependent
UV-vis and circular dichroism (CD) measurements starting
from high temperatures at which both components are molecu-
larly dissolved. As the sample is cooled while its absorbance is
measured at 340 nm, both the formation of the Cotton effect
and the red shift can be monitored as a function of the
temperature (Figure 2c).17 This yields a cooperative buildup5h,i
of the helically bound fraction θCD determined by CD, which
overlaps with the fraction of bound sites θUV obtained from the
increase in UV absorption, i.e., θUV ≈ θCD. This implies that
binding almost immediately gives rise to a helical arrangement
of the complex formed by single-stranded DNA (ssDNA) and
bound guest molecules.
1
Figure 3. (a) H NMR spectra at various temperatures for a dT40-G2
mixture with [dT40] ) 9 µM and [G2] ) 1.8 mM. (b) ATR-IR spectra of
dried films on glass obtained from solutions of dT40, G1, and 1:40
dT40-G1.
Characterization of the dT40-G Complexes. We previously
characterized the dT40-G1 complex with UV-vis and CD
spectroscopy.12 Here, the dT40-G1 and dT40-G2 complexes
1
were characterized in more detail using cryo-TEM, H NMR
and infrared spectroscopy, light scattering, and atomic force
microscopy (AFM). To see whether hydrogen bonds in the
dT40-G2 complex could directly be detected, 1H NMR spectra
using Watergate 3-9-13 solvent suppression were obtained
in 9:1 H2O/D2O at G2 and dT40 concentrations of 1.8 mM
and 9 µM, respectively (Figure 3a).12,18,19 The imino protons
of the thymine bases of dT40 in water can only be detected
when they are hydrogen-bonded.20 At 60 °C, the rapidly
exchanging imino protons of the thymines were not detected.
When the sample was cooled to 25 °C, the imino-proton signal
appeared at 10.5 ppm, indicating hydrogen bonding. This signal
shifted further downfield to 10.7 ppm upon cooling to 4 °C.
The fact that this signal was sharp indicates that the complexes
are still mobile and that they are neither polydisperse nor very
large.31 When the sample was heated back to 60 °C, the imino-
proton signal disappeared again, showing the reversibility of
the self-assembly process. The NMR data show that G2 binds
to dT40 at 25 °C, which is in agreement with the optical data
(Figure 2d). Hydrogen bonding was further evidenced by
attenuated total reflection infrared (ATR-IR) spectroscopy on
films obtained by drying a dT40-G solution in D2O (Figure
3b), which revealed that the CdO stretch was shifted from 1656
cm-1 (for dT40) to 1666 and 1702 cm-1, which is typical for
hydrogen-bonded thymines.8b
Characteristic of each mixture are its templated polymeriza-
tion temperature Tp (see below), the temperature at which half
of the template binding sites are occupied by guest monomers,
and the apparent elongation temperature Tte, the temperature at
which the straight line through the temperature-dependent data
points with the maximum slope crosses the temperature axis
(see Figure 2c).17 Tet for the templated assembly is comparable
to the elongation temperature Te for nontemplated cooperative
self-assembly discussed above.12a The quantity 1/Tet shows a
linear relation with the logarithm of the guest concentration, as
does 1/Te (Figure 2d). Comparing G1 with G2 shows that the
removal of the methyl group not only decreased the naphthalene
self-assembly temperature Te but also slightly increased the
dT40-G self-assembly temperature Tte (Figure 2d). The con-
centration range where the nontemplated self-assembly of G2
interferes with the templated self-assembly therefore lies at
higher concentrations than that for G1.
To visualize the templated self-assemblies, cryo-TEM was
performed.21 In the case of G2-dT40, small objects with
diameters of 4-5 nm22 and lengths varying between 5 and 30
nm were observed (Figure 4a). For the G1-dT40 mixture
(Figure 4b), similar diameters of 4-5 nm were observed, but
here the objects appeared to be more elongated and laterally
(15) In principle, ∆He can be determined by fitting the temperature-
dependent data with the model for cooperative self-assembly. However,
we found that because of a second process, the calculated enthalpy of
the temperature-dependent measurements increases with the concentra-
tion and deviates from the enthalpy found via the concentration-
dependent Te measurement. Possibly the lateral interactions present
between strands of G (see the cryo-TEM image) give rise to an
additional decrease in absorption intensity and therefore a higher
enthalpy.
(16) Previous titration experiments have shown that one G1 binds to one
thymine (see ref 12a). We have found similar results for G2 (Figure
7a).
(17) In this experiment, a nonstoichiometric ratio was used in order to
determine Tp more accurately. It should be noted that Tte does not
change with the ratio between the guest and the thymine while Tp
shows a linear trend with the logarithm of this ratio. Tte depends on
the guest concentration, while Tp also depends on the template
concentration. See the Supporting Information.
(18) The G1-dT40 mixture at this concentration was not suitable for
detecting hydrogen bonds because the mixture precipitated during the
measurements and thus gave only very broad signals at temperatures
where aggregates are formed.
(19) Previously, we have shown with ESI-MS that G1 binds to oligo-
thymines via hydrogen bonding (see ref 12b). We obtained similar
results for G2; see the Supporting Information.
(20) Cahen, P.; Luhmer, M.; Fontaine, C.; Morat, C.; Reisse, J.; Bartik, K.
Biophys. J. 2000, 78, 1059–1069.
(21) It should be noted that for concentrations and temperatures where the
complexes are not formed, no objects could be observed with cryo-
TEM. Furthermore, at room temperature and at these concentrations
of G1 and G2, the molecules were molecularly dissolved, and
therefore, no structures were observed with cryo-TEM.
(22) Similar dimensions have been observed with AFM in the liquid cell.
See the Supporting Information and: Surin, M.; Janssen, P. G. A.;
Lazzaroni, R.; Meijer, E. W.; Schenning, A. P. H. J. AdV. Mater., in
press; DOI: 10.1002/adma.200801701.
9
J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009 1225

A R T I C L E S
Janssen et al.
Figure 4. Cryo-TEM micrographs of (a) dT40-G2 ([G2] ) 0.8 mM and [dT40] ) 20 µM) and (b) dT40-G1 ([G1] ) 0.8 mM and [dT40] ) 20 µM)
vitrified at 288 K in MilliQ water. (c) Intensity-weighted size distributions for dT40-G2 and dT40-G1 mixtures ([G] ) 0.8 mM and [dT40] ) 10 µM)
at 294 K obtained using DLS at 90°. (d) Decay rate (Γ) as a function of q2 determined via angle-dependent DLS and the corresponding linear fit for the
mixture with [G2] ) 0.8 mM and [dT40] ) 10 µM at 294 K.
aggregated, supporting the hydrodynamic radius (Rh) value of
40 nm estimated from the DLS measurements (see below). The
measured diameter of 4-5 nm agrees well with the calculated
diameter obtained from the molecular dynamics simulations,
which revealed helical complexes with a diameter of 4.5 nm
and a length of 14 nm (see below). The measured lengths of
the G2-dT40 complexes varied around the expected dT40
length of 14 nm, which could be due to a random orientation
as a result of the vitrification process and overlap of aggregates
due to the sample thickness of at least 100 nm.23 Extended
aggregation or incomplete filling of templates could also cause
such an effect (see below). AFM images of G1-dT40 mixtures
in a liquid cell show objects with heights of ∼2.7 nm and widths
with tip convolution24 of 20-35 nm, similar to what was found
for dTp-G2 mixtures in a liquid cell.26
The complexes were further characterized by light-scattering
measurements. DLS at 90° revealed an intensity-weighted
average Rh value of 40 nm for the G1-dT40 complex, while
the G2-dT40 complex has an Rh value of 4 nm (Figure 4c).
The latter value agrees with the 3.9 ( 0.1 nm value of Rh found
with angle-dependent DLS (Figure 4d). The size of the
G1-dT40 complex is much larger than the expected 5 nm,
suggesting that the assembled structures further aggregate into
larger assemblies, as previously seen with cryo-TEM (Figure
4b). The existence of larger particles also explains the apparent
red-shift in the absorption onset for G1-dT40 due to scattering
from the G1-dT40 complexes, which are larger than the
G2-dT40 complexes. The Rh values of the dT40-G2 ag-
gregates are close to the Rh value of 5 nm calculated for a single
dT40-G2 complex assuming a cylindrical structure.25 The
small difference might be due to the fact that the complex is
not a perfectly filled cylinder, which would result in an
experimentally determined radius that is smaller than the
calculated Rh for a perfectly filled cylinder. The static light
scattering (SLS) experiments on the dT40-G2 aggregates
showed that the intensity was independent of the angle,
indicating that the sample consists predominately of small
particles.26 The scattering intensity at smaller scattering vectors
was not accurate enough to determine the radius of gyration,
and therefore, the aspect ratio of the assemblies could not be
determined. A temperature-dependent SLS intensity measure-
(25) For the calculation of Rh, we used D0 ) kBT/6πηRh ) kBT(ln F + 0.3
+ 0.4738/F + 0.4167/F2-0.3394/F3)/6πηb′, in which D0 is the
diffusion coefficient, η is the coefficient of viscosity, b′ ) L/2, and F
) L/(2r), where L ) 14 nm is the length from the MD simulations
and r ) (4.5 nm)/2 ) 2.25 nm is the radius from the MD simulations.
See: (a) Tirado, M. M.; Garcia de la Torre, J. J. Chem. Phys. 1979,
71, 2581–2587. (b) Santos, N. C.; Castanho, M. A. R. B. Biophys. J.
1996, 71, 1641–1650.
(23) Vos, M. R. J.; Bomans, P. H. H.; de Haas, F.; Frederik, P. M.; Jansen,
J. A.; Nolte, R. J. M.; Sommerdijk, N. A. J. M. J. Am. Chem. Soc.
2007, 129, 11894–11895.
(24) Bustamante, C.; Vesenka, J.; Tang, C. L.; Rees, W.; Guthold, M.;
Keller, R. Biochemistry 1992, 31, 22–26.
(26) See the Supporting Information.
9
1226 J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009

Insights into Templated Supramolecular Polymerization
A R T I C L E S
ment revealed a reversible transition at 310 K, above which
the intensity dropped to the intensity of a dT40 solution.27 The
transition temperature observed with SLS corresponds well with
what was also observed in the temperature-dependent UV and
CD data.26 The light-scattering experiments show that there is
little or no higher-order aggregation occurring for these
dT40-G2 aggregates, as observed with the cryo-TEM experi-
ments. Thus, removing the methyl group not only suppresses
the nontemplated self-assembly of the guests without decreasing
the binding affinity to dT40 but also prevents further aggregation
of the dT40-G complexes.
Molecular Modeling. In order to gain more insight into the
supramolecular organization in the DNA hybrid assemblies,
molecular mechanics (MM) calculations and molecular dynam-
ics (MD) simulations were performed.28 An MM calculation
shows that the lowest-energy conformation of a diaminotriazine
moiety binding to the thymine base consists of the expected
three H bonds, while the adenine-thymine (A-T) base pair
and the diaminotriazine-diaminotriazine pair both have two H
bonds.29
To characterize the structure of the G1-dT40 complex, MD
simulations with the CHARMm force field26,30 were performed
using the generalized Born implicit water solvent model. We
started the MD simulations with the complex obtained from
MM calculations on a G1-dT40 complex built up from a right-
handed B-helix DNA structure. Two types of structure were
considered with the implicit water solvent model: (1) one
without counterions bound to the phosphate groups along the
oligonucleotide backbone, in which the global structure carries
a total charge of -40e due to the phosphate backbone; and (2)
one with 40 Na+ counterions, each bound to a phosphate group
along the DNA backbone, in which the global system is
therefore neutral.28 These latter two extremes were considered
for comparison and because the dT40 aqueous solutions were
measured both in buffered solution and in ultrapure water. It
should be noted that even in ultrapure water, counterions are
bound to the phosphate.
In the MD simulation, the structure without the counterions
tends to stretch because of Coulomb repulsion between the
phosphate groups, but this is balanced by π stacking of the
naphthalenes, maintaining a 14.5 nm long right-handed helix
with three turns, a pitch of 13-14 G1 units, and a rotation per
“base pair” of ∼27° (Figure 5a). In this conformation, adjacent
naphthalenes overlap and are parallel to each other (Figure 5a)
but perpendicular to the strand axis. The ethylene oxide side
chains remain extended, interacting with each other only slightly.
The MD simulations performed on dT40-G1 with Na+
counterions show that the dT40 right-handed helix is stabilized
in a B-helix-like right-handed conformation with four turns, a
pitch of 10 G1 units, and a rotation per “base pair” of 36°; the
Figure 5. CPK models of snapshots from an MD simulation optimized
with MM of the G1-dT40 complex (a) without and (b) with Na+
counterions (mauve) along the ssDNA backbone. The bottom figures show
stacks of four G1 units inside the assemblies in (a) and (b), illustrating the
packing of the naphthalene units in the corresponding models above. (c)
Evolution of the number of H bonds during the last 500 ps of the MD run
for the model with counterions.
length of the stack varies around 13.3 nm (Figure 5b).
Remarkably, adjacent naphthalenes do not or only slightly
overlap (Figure 5b) and are not all parallel. The thymines are
much closer than in the previous case without Na+ counterions,
giving rise to multiple H bonds between each diaminotriazine
and adjacent thymines as a result of base flipping. The number
of intermolecular H bonds in the system (Figure 5c) fluctuates
between 100 and 120 during the MD run (simulation at 263 K)
with an average of ∼110 H bonds, illustrating that this structure
is clearly stabilized by H bonds between neighboring G1
molecules and thymines rather than by the poor π stacking
between the naphthalenes.26 By subtraction of the enthalpy of
(27) It should be noted that the scattering intensity of a dT40 solution is
almost the same as that of ultrapure water. Therefore, it does not allow
for a reliable calculation of Rh.
(28) A related study on two different guests binding to dT40 or dT20 model
stacks with Na+ counterions has been published. See ref 22.
(29) Although the absolute values have no meaning, the comparison of
the hydrogen bonding energies, as estimated in vacuum with the
Dreiding force field (see: Mayo, S. L.; Olafson, B. D.; Goddard, W. A.
J. Phys. Chem. 1990, 94, 8897–909. ), give a clue about the relative
amplitudes of the interactions: for G-T, A-T and G-G pairs, this
value amounts to-45,-29, and-26 kJ/mol, respectively.
(30) CHARMm Force Field, version 22; Accelrys: San Diego, CA, 2006;
Momany, F. A.; Rone, R. J. Comput. Chem. 1992, 13, 888–900.
(31) Saenger, W. Principles of Nucleic Acid Structure; Springer-Verlag:
New York, 1984; p 556.
9
J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009 1227

A R T I C L E S
Janssen et al.
Figure 6. (a) Titration of G1 into dTq (q ) 5, 10, 20, 40) measured via the CD intensity at 327 nm at 263 K. The lines have been added to guide the eye.
The dotted line indicates 2 equiv of G1 with respect to thymine. (b, c) CD spectra of (b) G1-dTq and (c) G2-dTq mixtures at 268 K. For G1, the host
binding site concentration was [H] ) 0.25 mM and [G1] ) 0.5 mM. For G2, [H] ) 0.2 mM and [G2] ) 0.4 mM. (d, e) Fraction of occupied sites θ as a
function of temperature for different dTq template lengths q for (d) G2 and (e) G1. The stoichiometric guest-host binding site ratio was [G]/[H] ) 2, with
[H] ) 0.2 mM for G2 and [H] ) 0.25 mM for G1. The absorption was measured at 269 nm for G1 and 340 nm for G2. The red lines show the fits to the
model. (f) Tp and Tte as functions of q for G2. The stoichiometric ratio was [G]/[H] ) 2 and [H] ) 0.2 mM. The red lines are fits to the functions Tp(q) )
Tp(∞) + Ap/q and Tte(q) ) Tet(∞) + Ae/q, where Ap and Ae are constants.
Figure 7. (a) Titration of G2 into dT40: θCD as a function of η ) [G2]/[H] for dT40-G2 mixtures with [H] ) 0.08 mM at λ ) 269 nm and T ) 268 K.
(b) The polymerization temperature Tp of G2-dT40 mixtures as a function of ln [G2] with η fixed at 2 results in a straight line. (c) The polymerization
temperature Tp of G2-dT40 mixtures as a function of ln(2 - 1/η) with [G] fixed at 0.5 mM results in a straight line. The red lines are fits to the model.
each single molecule from the total enthalpy of the assembly,
the enthalpy change per monomer due to binding, ∆h, in the
presence of Na+ was estimated with CHARMm to be -54 kJ/
mol. In both cases considered here, the simulations are consistent
with the existence of right-handed helices, in agreement with
the CD measurements. The radius of gyration, as estimated from
the molecular models, is significantly different for the two types
of structure considered: 4.47 nm without Na+ (Figure 5a) and
4.09 nm with Na+ (Figure 5b). The latter is in agreement with
the hydrodynamic radius estimated from the DLS experiments
(Rh ) 3.9 ( 0.1 nm).
Different Template Lengths. Having analyzed the self-
assembly of dT40 and G in detail, we next investigated the
influence of the length of the template on the self-assembly
process, with the final goal of obtaining control over the size
of the assemblies. Titration studies of G1 and G2 with dTq (q
) 5, 10, 20, 40) measured with UV-vis26 and CD spectroscopy
at 263 K (Figures 6a and 7a) show that saturation of the signal
occurs at [G]/[H] ratios greater than 1, where [H] is the host
binding site concentration. For smaller template sizes, the
binding of guest molecules is less efficient, and more than
stoichiometric amounts of guest molecules are needed to get a
high degree of binding. This suggests that guest-guest interac-
tions occurring on the template are important for stabilizing the
complexes (see below). At least 2 equiv of guests must be
present in order to compare the Cotton effects of the G-dTq
complexes (Figure 6a). All of the G-dTq mixtures have
similarly shaped Cotton effects at 263 K, but the intensity
decreases upon going to smaller templates (Figure 6b,c),
suggesting that the CD intensity depends on the template length.
The stabilizing effect of larger templates is further evidenced
by temperature-dependent UV measurements (Figure 6d,e). For
larger templates, a larger total amount of guest-guest interaction
is present for each template, thereby increasing the stability of
the stack. The steep decline in Tte for small q shows that a certain
size of the template is needed to stabilize the hybrid complex
(Figure 6f). This is also known to be true for the helix-coil
transition in biopolymers and in fact also for the formation of
a double strand in DNA, for which at least four bases are needed
to have favorable hybridization.26,31 End effects quite generally
9
1228 J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009

Insights into Templated Supramolecular Polymerization
A R T I C L E S
Scheme 2. Schematic Drawing of the Ising Chain Model for
The fraction of occupied sites, θ, is a function of g, ε, q,
temperature, and the template and guest concentrations. θ is
experimentally determined via the normalized change in UV
absorption, a measure of guest binding that is assumed to be
proportional to θ. It should be noted that the normalized change
in UV-vis absorption is proportional to the fraction of bound
sites on the template, θUV, while the change in CD intensity is
Templated Supramolecular Polymerizationa
a The guest molecules (blue balls) adsorb onto a template having q
binding sites, with a negative binding free energy, g. If two guests bind at
neighboring sites on the template, they interact according to a stacking free
energy, ε, which is also negative for the cooperative case. Each bound
molecule also represents chemical work that enters the model through the
chemical potential of the guest molecules in the solution.
proportional to the helical fraction of the occupied sites, θCD
,
which in principle can be smaller than θUV. In our case, we do
not see a significant difference between θCD and θUV (Figure
2c). The experiments show that templated supramolecular
polymerization occurs only below a certain transition temper-
ature, Tte. For definiteness, the transition temperature from
nonoccupied templates to fully occupied templates, Tp, is defined
as the temperature at which half of the template binding sites
destabilize ordered configurations in polymeric objects because
groups near the ends have fewer interactions with neighbors.
This makes chain ends more statistically disordered than the
central parts.
1
are occupied by guest monomers [i.e., θ(Tp) ≡ /2] and can be
determined directly from the temperature-dependent measure-
ments of θ (Figure 2c).
Model for Templated Self-Assembly. To determine how the
guests are distributed on the different template lengths, we
developed a mathematical model that can be used to fit the
experimental data for guests binding to a finite-sized template.
We use a two-parameter model that is based on the one-
dimensional (1D) Ising model for ferromagnetism and is similar
in spirit to the binding model put forward by McGhee and von
Hippel32 as well as to the Zimm-Bragg model for the helical
transition in biopolymers.14 It provides a coarse-grained descrip-
tion of the process of binding guest molecules to a template of
finite size in terms of g, the free energy for binding between a
guest and a binding site on a host template, and ε, the free energy
for stacking of two bound monomers (Scheme 2). It should be
noted that we use one energy for the guest-guest interaction
and another for the host-guest interaction in our 1D Ising
model. This assumption seems appropriate to describe the
experimental data (see below). If a guest molecule is bound to
a template site, its free energy is lowered by the binding free
energy g, and if two neighboring sites are occupied, the
free energy of the system is additionally altered by the stacking
free energy ε, which is negative for cooperative guest binding
and positive for anticooperative guest binding. The interplay
between the free energies ε and g determines the distribution
of bound guests across the template molecule, i.e., the way the
templates are filled as function of the fraction of occupied
binding sites, θ. When all of the template binding sites are
occupied, θ ≡ 1. When the stacking free energy ε is much
smaller than the thermal energy kBT, the guest molecules are
equally dispersed over the entire template. A relatively high
stacking free energy ε makes the assembly process more
cooperative in the sense that binding of monomers next to each
other is favored. The fractions of empty and filled templates
are then set by mass action, i.e., the concentration and the
guest-template stoichiometry. A brief description of the model
and its validation with the experimental data is given below.
Fitting the data with this model also provides an estimate of
the enthalpy and entropy of binding that make up the binding
free energy. The model does not account for additional processes
such as two or more guest-guest or host-guest interaction
energies, self-aggregation, or higher-order aggregation. The
template dTq is modeled as a quasi-1D string with q binding
sites. Each binding site represents a thymine of dTq and can
be either empty or occupied by one guest molecule.
We tested and validated whether our model is applicable for
ssDNA-templated self-assembly by analyzing independent types
of experiments: length-, temperature-, concentration-, and
host-guest-ratio-dependent measurements (Figures 6 and 7).
A detailed description of the model and the fitting procedures
can be found in the Supporting Information. We first tested the
influence of the length of the template with temperature-
dependent measurements. The fraction of occupied sites, θ, as
a function of temperature shows an increase in Tp with the
template length q for both G1 and G2 (Figure 6d,e). Decreasing
the template size allows the size of the aggregates to be
controlled, but lower temperatures are needed to get a high
coverage (large θ). Analysis of our model shows that 100%
binding can be achieved only when η g 1, where η ) [G]/[H]
is the stoichiometric ratio and [H] is the thymine host binding
site concentration; in other words, an excess of guest monomers
over host binding sites is required. Therefore, we used an excess
of guests (η ) 2) to make sure that we obtained full coverage
at low enough temperatures. Fitting the data yielded ∆h(Tp) ≈
-19kBTp and ε ≈ -5.5kBTp for G2 and ∆h(Tp) ≈ -22kBTp and
ε ≈ -5.1kBTp for G1, independent of the template length q,
within experimental error, corresponding to values of -47, -14,
-55, and -13 kJ/mol, respectively, at Tp ) 300 K. The very
close values of binding enthalpy and stacking free energy for
the two types of guests shows the similarity in the cooperative
nature of the interactions of the guests with the template, and
thus, a similar structure can be expected for all q. The
experimental enthalpy of binding for G1, ∆h(Tp) ≈ -22kBTp
) -55 kJ/mol, is remarkably close to the value found with the
MD simulations for the system with counterions, ∆h(263 K) ≈
-54 kJ/mol.
The model furthermore predicts that the polymerization
temperature Tp is inversely proportional to q, the length of the
template. The obtained value of Tp as defined above as a function
of q is in good agreement with the model, as evidenced by the
good fit to the functional form of Tp(q) ) Tp(∞) + Ap/q, even
down to q ) 5 (Figure 6f), where Ap is a constant. Also, Tte has
the same functional dependence on q as Tp, albeit with a different
constant Ae, as expected theoretically. In principle, q of a host
of unknown length can be determined via the transition
temperatures of a calibrated set of lengths and host and guest
concentrations. In the case of a calibration with Tte, this can be
done irrespective of the host concentration, since the value of
Tte depends only on the guest concentration and the length q.26
(32) McGhee, J. D.; von Hippel, P. H. J. Mol. Biol. 1974, 86, 469–489.
9
J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009 1229

A R T I C L E S
Janssen et al.
Table 1. Summary of the Average Fitting Parameters (in units of
kBTp), As Determined by Comparison with Our Experimental Data,
Showing the Similarity of the Parameters for G1 and G2
templates in the solution are either filled or empty: mixed forms
are not favored because of a lack of combinatorial entropy on
account of the fact that there are in that case fewer mutually
distinguishable distributions of bound guest molecules on the
template. On the other hand, for template lengths that are longer
than the bare correlation length, i.e., for q > ꢀ0, the templates
have alternating filled and bound sections with lengths on the
order ꢀ0 at Tp: the combinatorial factor can benefit from a large
number of entropically favorable fragmentations on a long chain.
Of course, the fraction and the actual correlation length decrease
below the transition temperature and increase above it.
guest monomer
ε
g
∆h(Tp)
Tp∆S(Tp)
G1
G2
-5.1 ( 0.3
-5.3 ( 0.3
-10 ( 1
-9 ( 1
-22 ( 2
-20 ( 2
-12 ( 2
-11 ( 2
For the titration of G2 into a dT40 template solution at a
fixed temperature, the change in CD intensity was used to
determine θ (Figure 7a). Again, the model fit very well with
the data, and the binding and stacking free energy values were
extracted (g ≈ -9kBTp and ε ≈ -5.1kBTp, respectively),
revealing a cooperative templated self-assembly process. The
stacking free energy value obtained here is comparable to the
value extracted from temperature-dependent UV absorption
curve. Combining the binding free energy, g ≈ -9kBTp, with
the enthalpy change for binding to the template, ∆h(Tp) ≈
-19kBTp, gives an estimate of Tp∆S(Tp) ≈ -10kBTp for the
entropy loss due to the organization of monomers on the
template as a result of binding. This entropy loss upon binding
is therefore quite appreciable. We speculate that this must also
be attributed in part to the loss of conformational freedom of
the backbone ssDNA chain.
The bare correlation length as defined above can be shown
1
to obey the relation ꢀ0 ≈ /2σ-1/2, where σ ) eε is a measure of
the cooperativity of the templated polymerization transition and
ε is in units of thermal energy kBTp.14,35 For example, if the
polymerization transition becomes sharper, this is caused by a
stronger cooperative guest-guest interaction; ε becomes more
negative, and hence, σ becomes smaller, i.e., more cooperative,
implying that ꢀ0 becomes larger. Because we obtained ε at Tp
(Table 1), we can calculate the correlation length exactly at the
polymerization transition temperature Tp: the ε value gives σ
≈ 4 × 10-3, which in turn yields the correlation length ꢀ0 ≈ 8
for G2. This means that according to our model, the average
length of bound sections at Tp is ∼8.36 For G2 binding to dTq,
we therefore expect the small templates dT5 and dT10 to be
either filled or empty at temperatures close to Tp. For the long
templates dT20, dT30, and dT40 close to Tp, we expect to get
alternating sequences of filled and empty sections upon binding,
with domains having a length on the order of the correlation
length ꢀ0.37 For a complete qualitative theoretical picture of the
dominant configurations of the template for G2, we constructed
a phase diagram based on the fitting results of the model, in
which we indicated the fraction of occupied template binding
sites θ (contour lines) as a function of the length of the template
(vertical axis) and a measure of the concentration or temperature
(horizontal axis); this diagram is shown in Figure 8.38
In the model, we can also describe the dependence of Tp on
the host and guest concentrations for large enough templates.
This allows for a direct determination of the enthalpy change
per monomer upon binding to the template, ∆h, from the
measured Tp at different concentrations with a fixed η ratio
(Figure 7b) or at different ratios at a fixed guest concentration
(Figure 7c).26,33,34 The slope (R) of the plot of Tp versus the
natural logarithm of the guest concentration, ln [G2], was used
2
to estimate the enthalpy change via the relation R ) -kTp /
∆h(Tp), yielding ∆h(Tp) ≈ -21.5kBTp. Similarly, the slope (R)
of the plot of Tp versus ln(2 - 1/η) yielded ∆h(Tp) ≈ -18kBTp.
Both values are in agreement with the enthalpy value obtained
from fitting θ as function of temperature, ∆h(Tp) ≈ -19kBTp.
We can conclude that our coarse-grained 1D Ising model is
able to describe the nature of the self-assembly process and
shows good self-consistency and good agreement with the
experimental data (Table 1). Therefore, we now use the
experimentally obtained guest-guest interaction ε to get a
qualitative, theoretical picture of how the guests are distributed
over the templates by considering the correlation length, which
is defined as the number of correlated template binding sites:
in our case, the correlation length is a measure of the number
of template sites filled one after the other. The correlation length
ꢀ ) ꢀ (q, T, ...) depends not only on the fraction of bound sites
but also on the interaction between the guests on the template.
For our purposes, it is useful to focus on ꢀ0, the value of the
correlation length of bound guest molecules in the idealized
case of an infinitely long template (q f ∞), for conditions
exactly at the polymerization transition temperature, so T ) Tp.
We might call this the “bare”, underlying correlation length,
which acts as a reference value and explains the existence of
two regimes for how the guests distribute over the template.
Our model predicts that for the template lengths shorter than
this correlation length, i.e., for q < ꢀ0, the assembly process is
in essence all-or-nothing at Tp. This means that at Tp these
Conclusion
We have characterized the structure and templated self-
assembly process of two naphthalene derivatives binding to
ssDNA templates of different lengths using a combined
experimental and theoretical approach. The supramolecular stack
of guest molecules binds to the ssDNA via hydrogen bonding
and is held together by π-π and hydrophobic interactions. The
removal of the terminal methyl group of the ethylene oxide tail
of the guest suppresses nontemplated aggregation but also
(35) Goldenfeld, N. Lectures on Phase Transitions and the Renormalization
Group; Addison-Wesley: Boston, 1992; p 394.
(36) The correlation length ꢀ0 here is defined exactly at the polymerization
transition for the idealized case of an infinitely long chain. For template
sizes q > ꢀ0, ꢀ0 ≈ ꢀ is reasonably accurate. For template sizes q <
ꢀ0, ꢀ decreases with q.
(37) For these long template lengths, a predominance of completely filled
templates is present only at high template coverages of, say, θ > 0.9,
where the actual correlation length ꢀ approaches infinity.
(38) The theoretical model provides us with much more information about
the details of the system than only gross averages, which can be of
great value for the practical applications of templated self-assemblies.
In principle, our course-grained model can be used to calculate the
way that the guest monomers are distributed over the templates
quantitatively as a function of q and θ via an implicit set of equations
(see the Supporting Information). For ease of calculation and clarity,
we chose to directly construct a qualitative picture of the dominant
configurations of the occupation of the templates by relating the
binding energies ε and g with q and θ, similar to what was done by
Zimm and Bragg14 for the helix-coil transition in polypeptides.
(33) It should be noted that changing the ratio at a fixed guest concentration
hardly affects Tte. See the Supporting Information.
(34) In principle, Tp can be estimated directly from the data without the
necessity of fitting the data to the coarse-grained model.
9
1230 J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009

Insights into Templated Supramolecular Polymerization
A R T I C L E S
concentration of guest and template. The obtained guest-guest
and host-guest interaction parameters provide qualitative
information on the size distribution and the efficiency of
covering whole templates, yielding a correlation length of 8 at
the polymerization temperature. This indicates that template
sizes below this length will be filled one-by-one at the transition
temperature. Ideally, the correlation length could be further
improved by increasing the attractive interactions between the
guest molecules. Our study, in which an ssDNA was used as a
template for the self-assembly of a water-soluble functional
molecule equipped with an appropriate hydrogen-bonding
moiety and a model description was included, provides insights
into how tape-measure molecules can be used to create
nanoscale objects of uniform length. These insights are of great
importance for templated supramolecular polymerizations in
both synthetic chemistry and biology.
Figure 8. Configuration-state diagram of the template for G2 as a function
of the template length q and the mass-action variable -g - ε + ln φfg,
which is actually a chemical potential and a measure of the concentration
of the guest molecules or the temperature of the solution. In the mass-
action variable, the stacking free energy ε and the binding free energy g
are in units of thermal energy kBTp, and φfg is the concentration of free
guest molecules. Shown are results calculated for G2 and stoichiometric
ratios η g 2, obtained from fitting the data with the model. The two contours
of constant θ are the (arbitrarily chosen) boundaries of the transition region
from sparely bound to densely bound guest molecules. The black dotted
line shows q ) ꢀ0. It should be noted that the width of the transition region
varies as the template length q changes. For very long templates (q . ꢀ0),
the width is proportional to σ1/2 at Tp. For short templates (q < ꢀ0), the
width is proportional to σ-1/q at Tp.
Acknowledgment. The authors acknowledge X. Lou for the
MALDI-TOF spectra, M. van Genderen for the scientific discussion,
K. Pieterse for the artwork, and the EURYI scheme for financial
support. The work in Mons was supported by the Belgian Federal
Science Policy Office (PAI 6/27) and the Belgian National Fund
for Scientific Research (FNRS/FRFC). M.S. and Ph.L. are Charge´
de Recherches and Chercheur Qualifie´, respectively, of the F.R.S.-
FNRS (Belgium). A portion of this work (by S.J.-F.) was part of
the Research Programme of the Dutch Polymer Institute (DPI),
Eindhoven, The Netherlands, as Project 610.
prevents further aggregation of the templated complexes. The
experimental findings on the binding of the guest to the ssDNA
template show good agreement with our self-consistent model
for templated self-assembly as well as with molecular dynamics
simulations. For long enough templates, the most important
parameters (host-guest and guest-guest interaction energies)
for templated self-assembly can be obtained from just two sets
of experiments: a guest titration and a concentration- or
stoichiometric-ratio-dependent measurement of the transition
temperature. Together with the enthalpy of binding, a prediction
can be made for the transition temperature for any given
Supporting Information Available: Complete ref 5b; full
experimental section, including general methods, materials, the
synthetic route and details of G2, and general sample preparation
methods; temperature-dependent CD, DLS, and UV measure-
ments, the SLS experiment, ESI-MS spectra, AFM micrographs,
and cryo-TEM micrographs; and details concerning the MD
simulations and the model for templated polymerization. This
material is available free of charge via the Internet at http://
pubs.acs.org.
JA808075H
9
J. AM. CHEM. SOC. VOL. 131, NO. 3, 2009 1231