Experimental and computational study of BODIPY dye-labeled cavitand dynamics

Article abstract of DOI:10.1021/ja4104292

Understanding the distance distribution and dynamics between moieties attached to the walls of a resorcin[4]arene cavitand, which is switchable between an expanded kite and a contracted vase form, might enable the use of this molecular system for the study of fundamental distance-dependent interactions. Toward this goal, a combined experimental and molecular dynamics (MD) simulation study on donor/acceptor borondipyrromethene (BODIPY) dye-labeled cavitands present in the vase and kite forms was performed. Direct comparison between anisotropy decays calculated from MD simulations with experimental fluorescence anisotropy data showed excellent agreement, indicating that the simulations provide an accurate representation of the dynamics of the system. Distance distributions between the BODIPY dyes were established by comparing time-resolved Foerster resonance energy transfer experiments and MD simulations. Fluorescence intensity decay curves emulated on the basis of the MD trajectories showed good agreement with the experimental data, suggesting that the simulations present an accurate picture of the distance distributions and dynamics in this molecular system and provide an important tool for understanding the behavior of extended molecular systems and designing future applications.

Full text of DOI:10.1021/ja4104292

Article
pubs.acs.org/JACS
Experimental and Computational Study of BODIPY Dye-Labeled
Cavitand Dynamics
Igor Pochorovski,† Tim Knehans,‡ Daniel Nettels,‡ Astrid M. Muller,§ W. Bernd Schweizer,†
̈
Amedeo Caflisch,*,‡ Benjamin Schuler,*,‡ and Franco
̧
is Diederich*,†
†Laboratorium fur Organische Chemie, ETH Zurich, Honggerberg, HCI, 8093 Zurich, Switzerland
‡Biochemisches Institut, Universitat Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerland
§Beckman Institute, and Division of Chemistry and Chemical Engineering, California Institute of Technology, M/C 139-74,
Pasadena, California 91125, United States
̈
̈
̈
̈
̈
̈
̈
S
* Supporting Information
ABSTRACT: Understanding the distance distribution and
dynamics between moieties attached to the walls of a
resorcin[4]arene cavitand, which is switchable between an
expanded kite and a contracted vase form, might enable the
use of this molecular system for the study of fundamental
distance-dependent interactions. Toward this goal, a combined
experimental and molecular dynamics (MD) simulation study
on donor/acceptor borondipyrromethene (BODIPY) dye-
labeled cavitands present in the vase and kite forms was
performed. Direct comparison between anisotropy decays calculated from MD simulations with experimental fluorescence
anisotropy data showed excellent agreement, indicating that the simulations provide an accurate representation of the dynamics
of the system. Distance distributions between the BODIPY dyes were established by comparing time-resolved Forster resonance
̈
energy transfer experiments and MD simulations. Fluorescence intensity decay curves emulated on the basis of the MD
trajectories showed good agreement with the experimental data, suggesting that the simulations present an accurate picture of the
distance distributions and dynamics in this molecular system and provide an important tool for understanding the behavior of
extended molecular systems and designing future applications.
vase form (ca. 66%), although an efficiency close to 100% was
expected assuming a close dye−dye distance of ca. 1 nm. Such
distance would prevail if the cavitand arms were oriented
parallel to one another, i.e., the average opening angle of the
cavitand walls would be 0°. The low FRET efficiency was
attributed to either dynamic behavior of the cavitand or an
unfavorable orientation of the transition dipole moments of the
dyes.7c,9 To gain more insights into this problem, cavitand 1a
was resynthesized together with the analogous cavitands with
shorter phenylene−ethynylene linkers, 1b and 1c.10 As the
FRET efficiencies increased toward cavitands with shorter
linkers (in the series 1a, 1b, 1c), it was concluded that the two
arms of the cavitands are separated by a certain average opening
angle and are not aligned parallel. Thereby, an average opening
angle of 16° was inferred to explain the observed FRET
efficiencies.10
INTRODUCTION
■
Resorcin[4]arene cavitands are a fascinating switching platform
because of their ability to adopt two spatially well-defined
conformations: an expanded kite and a contracted vase.
Switching the conformational and binding properties of
cavitands has been achieved with a variety of stimuli, such as
changes in temperature,1 pH,2 metal ion concentration,3 light
irradiation,4 solvent,5 and redox state.6 Besides employing
conformational switching of resorcin[4]arene cavitands as a
means to change their binding properties, the cavitand system
could eventually be used as a platform to investigate
fundamental interactions between objects attached to the
cavitand’s walls with respect to their variable distance. The
development of such molecular machines able to controllably
perform mechanical motions involving large spatial rearrange-
ments is a long-standing goal.7 The cavitand system could be
used for this purpose if the distance distribution and dynamics
between objects connected to the cavitand can be precisely
determined in both the vase and kite conformations.
In this work, we sought to gain detailed insights into the
distance distribution and dynamics of the cavitand system in
both the vase and the kite conformations. Toward this goal, we
prepared BODIPY dye-labeled quinone-based cavitand 2
(Chart 1) that is present in the kite form (designed based on
We had set out toward this goal by preparing a donor−
acceptor borondipyrromethene (BODIPY) dye-substituted
cavitand 1a (Chart 1), present in the vase form, for Forster
̈
resonance energy transfer (FRET)8 studies.7c,9 Surprisingly, an
Received: October 16, 2013
Published: February 3, 2014
unexpectedly low FRET efficiency was observed already in the
© 2014 American Chemical Society
2441
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
Chart 1. Structures of Vase Cavitands 1a−d, Kite Cavitand 2, and Donor-Only Substituted Reference Cavitand 3
what we learned about conformational properties of
diquinone−diquinoxaline cavitands).5,6 In addition, we ex-
panded the BODIPY dye-labeled cavitand vase series 1a−c by
cavitand 1d featuring phenylene−ethynylene linkers of different
lengths. Together, systems 1a−d and 2 embody the two
conformational extremes of cavitands and are therefore ideally
suited for the projected investigation. We investigated the
BODIPY dye-labeled cavitands in a combined experimental and
theoretical study consisting of time-resolved fluorescence
spectroscopy and molecular dynamics (MD) simulations.
BODIPY donor dye-substituted cavitand 3 shown in Chart 1
served as a reference compound for fluorescence studies. MD
simulations complemented the experimental results and yielded
theoretical dye−dye distance distributions. Emulation of
fluorescence decay curves based on the distance distributions
obtained by MD simulations allowed direct comparison of
experimental and theoretical results. This work not only
provides insights into the conformational dynamics of cavitands
but also serves as a case study for the interplay between time-
resolved fluorescence spectroscopy and MD simulationsa
combination of methods that is often used to investigate the
dynamics of biological macromolecules11applied to a
relatively simple artificial small molecule system.
ratio was 1.00/0.82, corresponding to a donor-only fraction of
18%. In cavitand 1b the donor-only fraction was 11%. On the
other hand, in cavitands 1c, 1d, and 2, the donor-only fractions
were below the sensitivity limit of the NMR measurements
(∼2%). A possible explanation for higher ratios between
donor/acceptor F atoms could be partial loss of the BF2 units
of the acceptor dyes during the course of cavitand syntheses,
resulting in mixtures of fully labeled cavitands and cavitands
lacking the BF2 unit on the acceptor dyes.13 Loss of the BF2
unit in BODIPY dyes has precedence and was observed in high
acidity media14 or under strongly basic conditions,15 and the
corresponding products have been shown to be nonfluor-
escent.14 Nevertheless, this finding is unexpected and has
important implications for fluorescence studies, since the
emission from this fraction of donor-only molecules needs to
be taken into account for analysis of the fluorescence emission
data.
X-ray Analysis. The X-ray structures of the precursors of
cavitands 1a−d and 2 and diiodocavitands 416 and 517 are
shown in Figure 1. The solid-state conformational properties of
the diiodocavitands are reflected by their solution-state
properties: compound 4 crystallized in the vase form from
(CH3)2CO/CH2Cl2, while compound 5 crystallized in the kite
form from CDCl3. While in cavitand 4 the I-bearing carbon
atoms are placed at a distance of 0.86 nm to one another, in
cavitand 5 the corresponding distance is 2.39 nm, which is
almost 3 times larger.
Absorption and Steady-State Fluorescence Spectros-
copy. The absorption and steady-state fluorescence spectra of
cavitands 1a−d and 2 are depicted in Figure 2 (top and
bottom, respectively). The absorption spectra exhibit two main
absorption bands corresponding to the donor dye moieties
(λmax = 529 nm) and the acceptor dye moieties (λmax = 619
nm). The fluorescence spectra were recorded using an
excitation wavelength of λexc = 490 nm; direct excitation of
the acceptor is negligible at this wavelength.18 The emission
maxima at λmax = 542 nm (IDA) stem from the donor and at λmax
= 630 nm from the acceptor (IAD) dye moieties. The intensity
scale is referenced relative to the intensity maximum ID of
donor-only substituted cavitand 3. The FRET efficiencies can
be estimated according to eq 1
RESULTS AND DISCUSSION
■
Synthesis and Characterization. The synthesis of
cavitands 1a−c and 3 had been recently reported.10 The
synthesis and characterization of the newly prepared cavitands
1d and 2 is described in section 1 of the Supporting
Information.
Cavitands 1a−d and 2 differ in that 1a−d possess two
quinoxaline walls, while 2 is equipped with two quinone walls.
This small structural difference has a dramatic effect on
cavitand conformation:5 cavitands 1a−c are present in the vase
form (1H NMR methine proton shifts at 5.61 and 5.69
ppm),1,12 while cavitand 2 adopts the kite form (methine
protons at 3.69 and 4.35 ppm) in CDCl3 solution.
19F NMR spectroscopy employing a pulse sequence with a
30° flip angle enabled determination of the BODIPY dye
donor/acceptor ratios in cavitands 1a−d and 2 based on the
integral ratios of the respective BF2 units. In cavitand 1a this
2442
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
IDA
ID
E = 1 −
(1)
Higher FRET efficiencies are expected for shorter dye−dye
distances. Consequently, cavitand 2, which is present in the kite
conformation, exhibits the lowest FRET efficiency (64%)
among the five cavitands. On the other hand, vase cavitands
1a−d exhibit E values ranging from 84% to 97% with smaller
values for cavitands with longer arms. While these efficiencies
are higher than in cavitand 2, suggesting shorter dye−dye
distances, they are not consistent with the assumption of
parallel oriented cavitand walls found in X-ray structures, in
which case FRET efficiencies of over 99% would be expected.
This trend was originally explained with the presence of a
cavitand opening angle of ca. 16°.10 In light of the above-
mentioned finding based on NMR that samples of cavitands
with longer arms possess significant donor-only labeled
fractions (18% and 11% for cavitands 1a and 1b, respectively),
the increasing donor fluorescence intensity for cavitands with
longer arms could also be explained by residual fluorescence
stemming from these contributions. This interpretation is
supported by fluorescence lifetime measurements, where the
contribution of doubly and singly labeled molecules can be
resolved more easily.
Time-Resolved Fluorescence Spectroscopy. We re-
corded donor and acceptor fluorescence decay curves using
time-correlated single-photon counting after pulsed excitation
of the donor dye at λexc = 470 nm for each sample containing
cavitands 1a−d and 2 (Figure 3) at concentrations of ca. 10−6
M in CHCl3. For a pure sample of donor−acceptor-labeled
species with a single fixed distance, one expects to measure
single-exponential donor fluorescence decays with a mean
fluorescence lifetime τDA reduced by a factor of (1 − E) as
compared to the mean donor fluorescence lifetime τD obtained
in the absence of the acceptor. Further, the acceptor decay
Figure 1. Molecular structures of 416 (at 123 K) and 5 (at 100 K) in
the crystals. Crystals of 4 and 5 were obtained by evaporation from
(CH3)2CO/CH2Cl2 and CDCl3, respectively. Solvent molecules, n-
hexyl chains, and hydrogen atoms are omitted for clarity.
curve should exhibit an initial rise (with rate constant 1/τDA
)
due to the FRET-induced population of the acceptor excited
state and a subsequent single-exponential decay with 1/τA,
where τA is the mean fluorescence lifetime of the acceptor
dye.19 This behavior is clearly observed for the kite cavitand 2.
The corresponding donor decay curve was fitted with a single-
exponential decay convolved with the instrument response
function (IRF), yielding τDA = 1.51 ns. This value results in a
mean FRET efficiency of 64% (using τD = 4.21 ns of reference
cavitand 3 that is lacking the acceptor dye),20 which is in
excellent agreement with the steady-state fluorescence spec-
troscopy data. Note, however, that obtaining accurate distance
information from this value requires the distance distribution
and dynamics of the system to be taken into account.
The acceptor decay curves of vase cavitands 1a−d are
virtually identical and identical to the decay curve of the
acceptor in 1a directly excited with τexc = 582 nm.20 Their steep
initial rise, which occurs within the response time of the
instrument, shows that the FRET efficiencies in cavitands 1a−d
are near 100% and donor and acceptor dyes thus in very close
proximity. The fluorescence emission of the donor is thus
expected to be very weak, and the corresponding donor
fluorescence lifetimes are very short. Note, however, that
monitoring the sensitized acceptor emission has the advantage
that only signal from molecules containing both a donor and an
intact acceptor fluorophore is detected. In contrast, the
fluorescence emission from the donor contains the signal of
molecules lacking an active acceptor, whose presence we
Figure 2. Absorption (top) and fluorescence emission (bottom, c = 0.5
× 10−7 M, λexc = 490 nm, ID is the emission maximum of reference
donor dye cavitand 3) spectra of cavitands 1a−d and 2 in CDCl3.
2443
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
goal of our study was therefore to utilize these developments
and to investigate the applicability of MD simulations to
molecular and supramolecular chemistry. The most common
general force fields employed in contemporary MD literature
are the General AMBER Force Field (GAFF)24a and the
CHARMM General Force Field (CGenFF).24b For compar-
ison, we included both force fields in our study.27
The most critical aspect in modeling cavitands 1a−d and 2
are the phenylene−ethynylene linkers; small errors in their
force field parameters would propagate to larger errors in
BODIPY dye−dye distances. A common parameter in
describing the stiffness of linker units is the persistence length
Lp, whereby larger Lp values represent stiffer linkers. The
persistence length of the phenylene−ethynylene unit had been
recently experimentally determined on the basis of pulse
electron paramagnetic resonance (EPR) spectroscopy on spin-
labeled test systems.28 We subjected these phenylene−
ethynylene-containing systems (see the Supporting Informa-
tion, section 6) to MD simulations with CGenFF and GAFF
(explicit chloroform, 100 ns each) to investigate how well the
two force fields reproduce the experimentally determined Lp
value. Values of Lp = 18.4 2.2 nm for CGenFF and Lp = 27.0
2.4 nm for GAFF were obtained. Comparison to the
experimental value of Lp = 13.8 1.5 nm28 suggests that both
force fields slightly overestimate the rigidity of the oligo-
(phenylene−ethynelene) linker, but especially the results from
CGenFF provide reasonably good agreement.
The cavitands 1a−d and 2 were subjected to MD simulations
with the force fields CGenFF and GAFF (explicit chloroform,
500 ns each). The resulting BODIPY dye−dye distance
histograms are illustrated in Figure 4. For the vase cavitands
1a−d, both force fields yield histograms composed of sharp
maxima at ca. 0.5 and 1.0 nm and broader distributions
spanning from 0.5 to 1.9 nm in the case of the smallest cavitand
1c and from 0.5 to 3.0 nm in the case of the largest cavitand 1a.
While the sharp maxima can be ascribed to arrangements where
the BODIPY dyes are in direct contact, the broader
distributions are due to motions of separated dyes.
On the other hand, the kite cavitand 2 yields single-
distribution histograms spanning from 3.0 to 5.0 nm. The mean
dye−dye distances ⟨d(B···B)⟩ obtained with the two force fields
differ only marginally. The MD simulations slightly under-
estimate the average opening angle of cavitand 2, as evidenced
by comparing the C370−C420 distance (Figure 1) in the X-ray
structure of cavitand 5 (2.39 nm) with the corresponding
average C−C distance stemming from the CGenFF simulation
of cavitand 2 (2.33 nm). If this discrepancy of 0.06 nm is
propagated toward the BODIPY dyes, a small deviation of ca.
0.1 nm for the average simulated dye−dye distance can be
expected.
Figure 3. Fluorescence decay curves of the acceptor (top) and donor
(bottom) moieties of cavitands 1a−d and 2. Measurements were
performed under magic angle configuration, i.e., with the emission
polarizer set to 54.7° with respect to the excitation polarization. IRF =
Instrument response function.
quantified using NMR spectroscopy (see Synthesis and
Characterization). The residual donor fluorescence intensity
decays of cavitands 1a−d (Figure 3b) are thus dominated by
the emission from these donor-only molecules. As a result, they
exhibit decay components up to the lifetime of the isolated
donor fluorophore in the range of 4 ns21 but also additional
shorter lifetime components, presumably because of collisional
quenching with the acceptor moiety lacking the BF2 unit. Such
multiexponential behavior is typical of BODIPY dye dimers or
BODIPY dyes in confined environments (in proteins, lipids,
micelles, or glasses).22
MD Simulations. To obtain a detailed molecular picture of
the dye−dye distance distributions and dynamics underlying
the fluorescence results, we performed MD simulations with
explicit chloroform solvent for cavitands 1a−d and 2. For direct
comparison to the experimental data and benchmarking of the
MD results, we calculated observable properties such as
fluorescence intensity and fluorescence anisotropy decay curves
from the MD trajectories. Reports on MD simulations of
supramolecular systems in explicit solvents are rare.23 One
reason for this is that early force fields were designed to
simulate biomolecules in aqueous media. Recently, all-atom
general force fields24 and tools for automatic parameter
assignment24a,25 have been developed, as well as structure
and topology files for various common organic solvents.26 One
Notably, there is no overlap between the dye−dye distance
distributions of vase cavitand 1c and kite cavitand 2, which have
the same linker length. Thus, switching between both cavitand
conformations can entirely change the distance distribution
profile of moieties attached to the cavitand walls.
Distance Autocorrelation from MD Data. To determine
the time scale on which the distance dynamics of the BODIPY
dye arms relative to each other takes place, we calculated the
time-correlation functions29
⟨R(t + τ)R(t)⟩
t
g(τ) =
⟨R(t)⟩2
(2)
t
2444
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
presumably due to a slow down of the distance dynamics in
cavitand 1a caused by dye−dye contacts in the closed
conformation.
Fluorescence Anisotropy. A stringent way of testing the
accuracy of the time scales of dynamics in the simulations,
including the viscosity of the solvent, is direct comparison to
fluorescence anisotropy data. Anisotropy decays report on
rotational diffusion times of the entire cavitands as well as on
segmental rotation times of the BODIPY dye arms.30 We
measured the anisotropy decay of the acceptor dye of cavitand
1a and compared it to the expected anisotropy decay as
calculated from the corresponding MD simulations.
Fluorescence anisotropy decays were obtained by measuring
the fluorescence intensity decays IVV(t) and IVH(t) observed
after pulsed excitation of the acceptor dye (λexc = 582 nm); in
both measurements the plane of the linear polarized excitation
light was vertically oriented and the emission polarizer was set
vertically for IVV and horizontally for IVH (Figure 6, top). We
fitted the two curves globally with model curves I∥ and I⊥
I = I0(1 + 2r(t))e−t/τ
A
(3)
Figure 4. Histograms of distances between the B atoms of the donor
and acceptor BODIPY dyes of cavitands 1a−d and 2, simulated with
CGenFF (black) and GAFF (red).
from the CGenFF-simulated interdye distance data of cavitands
1a and 2 (Figure 5); R(t) is the dye−dye distance d(B···B) at
time t, ⟨...⟩t denotes the time average over t, and τ is the lag
time. The results show that the interdye distance dynamics of
both cavitands occur on a subnanosecond time scale, with the
dynamics of vase cavitand 1a being slower than those of kite
cavitand 2 by almost an order of magnitude. This finding is
Figure 6. (Top) Acceptor fluorescence intensity decay curves of
cavitand 1a measured after acceptor excitation with λexc = 582 nm. IVV
was recorded with an emission polarization filter set to 0° and IVH to
90° with respect to excitation polarization. (Bottom) Fluorescence
anisotropy decay curves constructed from CGenFF-MD trajectory of
cavitand 1a using eq 6, fitted by the model described by eq 5, and the
curve resulting from measured data inserted in eq 5.
Figure 5. Autocorrelation functions of the interdye distances derived
from CGenFF-MD simulations of cavitands 1a and 2.
2445
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
I⊥ = I0(1 − r(t))e−t/τ
A
Fluorescence Decay Curves from MD Data. To enable a
direct comparison of the simulations to the experimental data,
we emulated donor fluorescence decay curves that would be
expected on the basis of the CGenFF-simulated MD
trajectories by taking into account the dye−dye distance and
orientation time series (κ2) and compared them with the
measured decay curves. The decay curves were emulated using
the Markov chain32 model presented in Figure 7 (top), which
was shown to be the best approach for calculating fluorescence
observables.11h−k Starting from the donor excited state (D*A),
the probabilities of FRET (pFRET), donor emission (pD), and
acceptor emission (pA) were calculated for every saved
snapshot of the MD trajectory according to
(4)
τA is the mean fluorescence lifetime of the acceptor dye and was
determined to be τA = 5.24 ns.20 I∥ and I⊥ were convolved with
the IRF to yield IVV = IRF⊗I∥ and IVH = G·IRF⊗I⊥. The factor
G describes the relative difference in detection efficiencies of
vertical and horizontal polarized photons of the instrumenta-
tion.30 For our instrument, we determined G = 1.1. The
fluorescence anisotropy decay is defined as r(t) = (I∥ − I⊥)/(I∥
+ 2I⊥). We expected r(t) to decay with two decay times: one
corresponding to the overall rotational diffusion of the whole
cavitand (τM), and the other corresponding to the rotational
motion of the BODIPY dye relative to the cavitand (τrot).
Hence, the anisotropy decay was described according to31
kFRET
kD + kFRET
FRET)Δt
D
r(t) = ((r0 − r∞)e−t/τ + r∞)e−t/τ
rot
M
pFRET = (1 − e−(k +k
)
(5)
(7)
where r0 is the limiting anisotropy, which was fixed to 0.37,22a
and r∞ is the residual anisotropy. Fitting the intensity decays
kD
FRET)Δt
pD = (1 − e−(k +k
)
D
(Figure 6, top) yielded τM = 1.64 ns, τrot = 0.15 ns, and r∞
=
kD + kFRET
(8)
(9)
0.07.
p = 1 − e−k Δt
For direct comparison, we determined r(t) also from the MD
simulation of cavitand 1a. The time trajectory of the normalized
A
A
⇀
with Δt = 4 ps being the time step at which coordinates were
saved, kFRET the rate constant of energy transfer, kd = 1/τD =
0.24 ns−1, and kA = 1/τA = 0.19 ns−1 the respective fluorescence
rate constants of the donor and acceptor dyes. The rate
constant of energy transfer kFRET was calculated from
orientation vector A (t) of the acceptor dye was used to obtain
the anisotropy decay according to31
⇀
⇀
r(t) = r0⟨P (A (t′)·A (t′ + t))⟩
(6)
2
t′
Here, P2(x) = (3x2 − 1)/2 is the second Legendre polynomial
⃗
6
2
and vector A is defined as indicated in Figure 7. We fitted the
⎛
⎜
⎝
⎞
⎟
⎠
R0(κ )
R
kFRET = kD
(10)
with R being the dye−dye distance d(B···B), and R0(κ2) the
dye−dye orientation-dependent Forster radius.8 R0(κ2) was
̈
calculated according to
3
2
2
κ2
6
R0(κ ) = R0(2/3)
(11)
The Forster radius R (2/3) for κ2 = 2/3 that is valid in case of
̈
0
freely rotating, isotropically averaged dyes was determined to
be 4.91 nm. The orientation factor κ2 was calculated according
to30
κ2 = (cos θDA − 3 cos θD cos θA)2
(12)
⃗
⃗
⃗
⃗
with the angles defined by θDA = ∠(D,A),θD = ∠(D,R), and θA
⃗
⃗
⃗
⃗
= ∠(A,R). The vectors A and D represent the emission
transition dipole moment of the donor and the absorption
transition dipole moment of the acceptor dye, respectively. R is
⃗
Figure 7. (Top) Markov chain model devised to emulate donor- and
acceptor-fluorescence decay curves from simulated MD trajectories.
(Bottom) Description of vectors and angles used for calculating κ2
according to eq 12.
the connection vector between the boron atoms of the dyes
(Figure 7, bottom).
A total of 2500 different snapshots picked along regular
intervals of 0.2 ns served as starting points for the Markov
model. The model was started at state D*A and advanced by
stepping through the trajectory. For every snapshot, transition
probabilities according to eqs 7−9 were recalculated as
functions of R and κ2 (eqs 10−12). This process was stopped
if photon emission from either the donor or the acceptor
occurred or if the end of the trajectory was reached. The
donor/acceptor photon counts at their respective emission
times were summed up. To collect a statistically significant
number of photon emission events, each of the 2500 runs was
repeated 48 000 times to inject overall 120 millions of photons
into each FRET emulation. The resulting decay curves were
resulting anisotropy decay by eq 5 and obtained τM = 1.57 ns,
τrot = 0.16 ns, and r∞ = 0.08 (Figure 6, bottom). The excellent
agreement between these values and the ones obtained from
experimental anisotropy decay measurements provides good
evidence that the MD simulations accurately capture the
dynamics of the cavitands, including the effect of solvent and
viscosity. The MD simulations should thus also be able to
provide accurate insights in the combined effects of distance
distributions and dynamics on the experimentally observed
fluorescence intensity decays.
2446
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
1
convolved with the respective IRFs to allow a direct
comparison with experimental data.
The measured and emulated donor fluorescence emission
decay curves of vase cavitand 1a and kite cavitand 2 based on
CGenFF trajectories are presented in Figure 8. When
⟨E⟩ =
1 + ( (R /r)6P(r)dr)−1
∫
0
(13)
Using the distance distribution of cavitand 2 from the
simulations shifted by 0.1 nm as described above, we obtain
an average transfer efficiency, ⟨E⟩, of 68%. The fair agreement
with the experimental value of 64% indicates that dynamic
averaging is a reasonable approximation.
CONCLUSIONS
■
The distance distribution and dynamics between moieties
attached to resorcin[4]arene cavitand walls in both the kite and
the vase conformations have been studied by a combination of
experimental and theoretical methods on the basis of BODIPY
dye-substituted vase cavitands 1a−d and kite cavitand 2. In kite
cavitand 2, featuring one phenylene−ethynylene linker unit per
dye, the dye moieties adopt an average distance of ca. 4.2 nm
according to MD simulations (CGenFF). Time-resolved
fluorescence spectroscopy revealed a FRET efficiency of 64%.
In the case of vase cavitands 1a−d, steady-state and time-
resolved fluorescence spectroscopy showed increasing amounts
of donor fluorescence for cavitands with longer arms. In early
studies, this trend was explained by increasing dye−dye
distances for cavitands with longer arms, which would be
consistent with the presence of a nonzero cavitand opening
angle. In the current study, we found according to MD
simulations (CGenFF) that while the average dye−dye distance
does indeed increase from 0.8 nm in the shortest cavitand 1c to
1.2 nm in the longest cavitand 1a, these distances are still in a
range that should yield FRET efficiencies of 100%. Instead, the
increasing donor fluorescence for cavitands with longer arms
could be explained by the previously undetected presence of
cavitand fractions with inactive acceptor components lacking
the BF2 unit. Emulation of the fluorescence intensity decay
curve of cavitand 1a from simulated MD trajectories confirmed
that the experimentally observed decay curve can only be
reproduced when a donor-only-labeled cavitand fraction is
taken into account.
The dynamics of vase cavitand 1a was investigated by
fluorescence anisotropy measurements. Evaluation of both
measured and simulated fluorescence anisotropy decay curves
showed that the rotation time of a BODIPY dye arm in
cavitand 1a is ca. 0.15 ns. Autocorrelation analysis of the dye−
dye distance time series revealed that the distance dynamics of
the BODIPY dye arms takes place on the subnanosecond time
scale. Most importantly, the fluorescence lifetime decays
calculated based on the MD simulations of cavitand 2 are in
excellent agreement with the experimental data, showing that
the simulations present an accurate picture of the distance
distribution and dynamics in this molecular system. With these
results, the cavitand system can now be used as a platform to
investigate fundamental, distance-dependent interactions be-
tween objects attached to the cavitand’s walls.
Figure 8. Measured and emulated donor fluorescence emission decay
curves of vase cavitand 1a (with an implemented donor-only fraction
of 18%) and kite cavitand 2.
performing the emulation of the donor decay in cavitand 1a
without taking a donor-only-labeled fraction into account, an
extremely rapid decay is observed that basically parallels the
IRF.33 The experimentally observed donor decay curve can
only be reproduced if a donor-only labeled fraction is taken into
account. Thus, 100% FRET efficiency is indeed expected for
vase cavitands 1a−d due to the close dye−dye distance and
should not be obscured by a possibly suboptimal orientation of
transition dipole moments. The deviation between the
emulated and the measured curves stems from the fast decay
component in the experimental curve that we attribute to
dynamic collisional quenching between the fluorophores. The
effect of quenching at low dye−dye distances is not described
by Forster theory and is therefore not taken into account in the
̈
FRET emulation.11i
In contrast, the emulated donor decay curve of kite cavitand
2 is close to the experimental result, albeit with a slightly
shorter average lifetime (1.01 ns) than the measured curve
(1.51 ns), suggesting that the MD simulation underestimates
the average dye−dye distance to some extent, most probably
because of a slight deviation in the kite opening angle. Indeed,
emulation of the donor fluorescence intensity decay curve on
the basis of a simulated dye−dye distance distribution that is
shifted by +0.1 nm yielded excellent overlap with the measured
curve (Figure 8). This deviation in the average distance of ∼2%
is in the same range as the difference between the results from
the two force fields used (Figure 4) and illustrates the accuracy
of MD simulations in reflecting the structural properties of
molecular systems. With the information from the simulations,
we can also test the accuracy of simple averaging regimes
commonly used for analysis of FRET in dynamic systems.11b,34
Since for cavitand 2 both the rotational correlation times of the
dyes (Figure 6) and the distance dynamics between them
(Figure 5) are much shorter than the fluorescence lifetimes, the
system should be well approximated by the dynamic averaging
regime, for which
In the case of extended molecular and supramolecular
systems with pronounced flexibility and therefore broad
intramolecular distance distributions, the combination of
time-resolved FRET with simulations is essential to quantify
the underlying dynamics because the experimental observables
depend on both the shape of the distance distribution and the
time scale of the dynamics. In this work, we have in-depth
structurally characterized the two conformational states of
resorcin[4]arene cavitands, which opens the possibility to
utilize the cavitand system for studying intraspace interactions
2447
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
(d) Azov, V. A.; Beeby, A.; Cacciarini, M.; Cheetham, A. G.; Diederich,
F.; Frei, M.; Gimzewski, J. K.; Gramlich, V.; Hecht, B.; Jaun, B.;
Latychevskaia, T.; Lieb, A.; Lill, Y.; Marotti, F.; Schlegel, A.; Schlittler,
R. R.; Skinner, P. J.; Seiler, P.; Yamakoshi, Y. Adv. Funct. Mater. 2006,
16, 147. (e) Wu, J.; Leung, K. C.-F.; Benítez, D.; Han, J.-Y.; Cantrill, S.
J.; Fang, L.; Stoddart, J. F. Angew. Chem., Int. Ed. 2008, 47, 7470.
ranging from 1 to 7 nm (adjustable via the linker length).
Furthermore, the good agreement between MD simulations
and experimental data should further encourage application of
MD simulations to molecular chemical systems. Progress in
force-field development now allows such simulations to be
performed with high accuracy. MD simulations could serve as a
testing tool for envisioned concepts and thereby guide the
synthetic chemist toward successful implementation of his/her
ideas.
Although the dyes used in this study are not suitable for
single-molecule fluorescence detection, an approach analogous
to the one demonstrated here could be employed for single-
molecule studies. Such experiments could further enhance the
resolution of structural and dynamic heterogeneities in
molecular systems.
(8) Forster, T. Ann. Phys. (Berlin, Ger.) 1948, 2, 55.
̈
(9) Azov, V. A.; Schlegel, A.; Diederich, F. Bull. Chem. Soc. Jpn. 2006,
79, 1926.
(10) Pochorovski, I.; Breiten, B.; Schweizer, W. B.; Diederich, F.
Chem.Eur. J. 2010, 16, 12590.
(11) (a) Gustiananda, M.; Liggins, J. R.; Cummins, P. L.; Gready, J.
E. Biophys. J. 2004, 86, 2467. (b) Schuler, B.; Lipman, E. A.; Steinbach,
P. J.; Kumke, M.; Eaton, W. A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102,
2754. (c) Merchant, K. A.; Best, R. B.; Louis, J. M.; Gopich, I. V.;
Eaton, W. A. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 1528. (d) Best, R.
B.; Merchant, K. A.; Gopich, I. V.; Schuler, B.; Bax, A.; Eaton, W. A.
Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 18964. (e) Corry, B.; Jayatilaka,
D. Biophys. J. 2008, 95, 2711. (f) Unruh, J. R.; Kuczera, K.; Johnson, C.
K. J. Phys. Chem. B 2009, 113, 14381. (g) Schuetz, P.; Wuttke, R.;
Schuler, B.; Caflisch, A. J. Phys. Chem. B 2010, 114, 15227. (h) Liao, J.-
M.; Wang, Y.-T.; Chen, C.-L. Phys. Chem. Chem. Phys. 2011, 13,
10364. (i) Hoefling, M.; Lima, N.; Haenni, D.; Seidel, C. A. M.;
ASSOCIATED CONTENT
* Supporting Information
■
S
Synthetic procedures, characterization data, X-ray data,
fluorescence data, MD methods, NMR spectra. This material
is available free of charge via the Internet at http://pubs.acs.org.
Schuler, B.; Grubmuller, H. PLoS ONE 2011, 6, e19791. (j) Speelman,
̈
A. L.; Munoz-Losa, A.; Hinkle, K. L.; VanBeek, D. B.; Mennucci, B.;
̃
AUTHOR INFORMATION
Corresponding Authors
caflisch@bioc.uzh.ch
■
Krueger, B. P. J. Phys. Chem. A 2011, 115,, 3997;. (k) Hoefling, M.;
Grubmuller, H. Comput. Phys. Commun. 2013, 184, 841.
̈
(12) Azov, V. A.; Jaun, B.; Diederich, F. Helv. Chim. Acta 2004, 87,
schuler@bioc.uzh.ch
diederich@org.chem.ethz.ch
449.
(13) A more detailed discussion on the instability of the acceptor dye
is provided in section 1.4 of the Supporting Information.
(14) (a) Liras, M.; Prieto, J. B.; Pintado-Sierra, M.; Arbeloa, F. L.;
Notes
The authors declare no competing financial interest.
́
García-Moreno, I.; Costela, A.; Infantes, L.; Sastre, R.; Amat-Guerri, F.
Org. Lett. 2007, 9, 4183. (b) Banuelos, J.; Lopez Arbeloa, F.; Arbeloa,
T.; Salleres, S.; Vilas, J. L.; Amat-Guerri, F.; Liras, M.; Lopez Arbeloa, I.
J. Fluoresc. 2008, 18, 899.
́
̃
ACKNOWLEDGMENTS
■
́
This work was supported by grants from the Swiss National
Science Foundation (SNF). I.P. acknowledges the receipt of a
fellowship from the Fonds der Chemischen Industrie. Research
was partly carried out in the Laser Resource Center of the
Beckman Institute of the California Institute of Technology and
supported by the Arnold and Mabel Beckman Foundation. We
thank Prof. Dr. Gunnar Jeschke, Dr. Marc-Olivier Ebert, and
Dr. Andreas Vitalis for helpful discussions, Dimitry Kotlyar for
help with the Table of Contents artwork, and Dr. Kenno
Vanommeslaeghe for providing parametrization of the
cavitands via the Paramchem engine.
(15) (a) Crawford, S. M.; Thompson, A. Org. Lett. 2010, 12, 1424.
(b) Smithen, D. A.; Baker, A. E. G.; Offman, M.; Crawford, S. M.;
Cameron, T. S.; Thompson, A. J. Org. Chem. 2012, 77, 3439.
(16) Published in ref 10.
(17) For details, see section 2 of the Supporting Information.
(18) There are no indications that electron transfer, neither between
the two dyes nor between the dyes and the quinone walls, significantly
affects the transfer dynamics. For a more detailed discussion, see
section 3.2 of the Supporting Information.
(19) Meer, B. W. V. d.; Coker, G.; Chen, S.-Y. S. Resonance Energy
Transfer: Theory and Data; VCH Publishers, Inc.: New York, 1994.
(20) For details, see Figure S5 in section 3.1 of the Supporting
Information.
REFERENCES
■
(21) An analyis of the decay curves is provided in section 3.4 of the
Supporting Information.
(22) (a) Karolin, J.; Johansson, L. B.-Å.; Strandberg, L.; Ny, T. J. Am.
(1) (a) Moran, J. R.; Ericson, J. L.; Dalcanale, E.; Bryant, J. A.;
Knobler, C. B.; Cram, D. J. J. Am. Chem. Soc. 1991, 113, 5707.
(b) Moran, J. R.; Karbach, S.; Cram, D. J. J. Am. Chem. Soc. 1982, 104,
5826.
Chem. Soc. 1994, 116, 7801. (b) Bergstrom, F.; Hagglof, P.; Karolin, J.;
̈
̈
̈
Ny, T.; Johansson, L. B.-Å. Proc. Natl. Acad. Sci. U.S.A. 1999, 96,
12477. (c) Bergstrom, F.; Mikhalyov, I.; Hagglof, P.; Wortmann, R.;
(2) Skinner, P. J.; Cheetham, A. G.; Beeby, A.; Gramlich, V.;
Diederich, F. Helv. Chim. Acta 2001, 84, 2146.
(3) (a) Frei, M.; Marotti, F.; Diederich, F. Chem. Commun. 2004,
1362. (b) Durola, F.; Rebek, J., Jr. Angew. Chem., Int. Ed. 2010, 49,
3189.
(4) (a) Berryman, O. B.; Sather, A. C.; Lledo, A.; Rebek, J. Angew.
Chem., Int. Ed. 2011, 50, 9400. (b) Berryman, O. B.; Sather, A. C.;
Rebek, J., Jr. Chem. Commun. 2010, 47, 656.
(5) Pochorovski, I.; Boudon, C.; Gisselbrecht, J.-P.; Ebert, M.-O.;
Schweizer, W. B.; Diederich, F. Angew. Chem., Int. Ed. 2012, 51, 262.
(6) Pochorovski, I.; Ebert, M.-O.; Gisselbrecht, J.-P.; Boudon, C.;
Schweizer, W. B.; Diederich, F. J. Am. Chem. Soc. 2012, 134, 14702.
(7) (a) Barboiu, M.; Lehn, J.-M. Proc. Natl. Acad. Sci. U.S.A. 2002, 99,
5201. (b) Dietrich-Buchecker, C. O.; Jimenez-Molero, M. C.; Sartor,
V.; Sauvage, J. P. Pure Appl. Chem. 2003, 75, 1383. (c) Azov, V. A.;
Schlegel, A.; Diederich, F. Angew. Chem., Int. Ed. 2005, 44, 4635.
̈
̈
̈
Ny, T.; Johansson, L. B.-Å. J. Am. Chem. Soc. 2002, 124, 196.
(d) Mikhalyov, I.; Gretskaya, N.; Bergstrom, F.; Johansson, L. B.-Å.
̈
Phys. Chem. Chem. Phys. 2002, 4, 5663. (e) Tleugabulova, D.; Zhang,
Z.; Brennan, J. D. J. Phys. Chem. B 2002, 106, 13133.
(23) (a) Fischer, S.; Grootenhuis, P. D. J.; Groenen, L. C.; van
Hoorn, W. P.; van Veggel, F. C. J. M.; Reinhoudt, D. N.; Karplus, M. J.
Am. Chem. Soc. 1995, 117, 1611. (b) den Otter, W. K.; Briels, W. J. J.
Am. Chem. Soc. 1998, 120, 13167. (c) Tolpekina, T. V.; den Otter, W.
K.; Briels, W. J. J. Phys. Chem. B 2003, 107, 14476. (d) Casanovas, J.;
Zanuy, D.; Aleman, C. Angew. Chem., Int. Ed. 2006, 45, 1103.
(e) Ewell, J.; Gibb, B. C.; Rick, S. W. J. Phys. Chem. B 2008, 112,
10272. (f) Javor, S.; Rebek, J. J. Am. Chem. Soc. 2011, 133, 17473.
́
(g) Cezard, C.; Trivelli, X.; Aubry, F.; Djedaïni-Pilard, F.; Dupradeau,
F.-Y. Phys. Chem. Chem. Phys. 2011, 13, 15103. (h) Zhang, H.; Tan, T.;
2448
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449

Journal of the American Chemical Society
Article
Feng, W.; van der Spoel, D. J. Phys. Chem. B 2012, 116, 12684.
(i) Zheng, X.; Wang, D.; Shuai, Z.; Zhang, X. J. Phys. Chem. B 2012,
116, 823. (j) Simona, F.; Nussbaumer, A. L.; Haner, R.; Cascella, M. J.
̈
Phys. Chem. B 2013, 117, 2576.
(24) (a) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case,
D. A. J. Comput. Chem. 2004, 25, 1157. (b) Vanommeslaeghe, K.;
Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; Darian, E.;
Guvench, O.; Lopes, P.; Vorobyov, I.; Mackerell, A. D., Jr. J. Comput.
Chem. 2009, 31, 672.
(25) (a) Wang, J.; Wang, W.; Kollman, P. A.; Case, D. A. J. Mol.
Graph. Model. 2006, 25, 247. (b) Vanommeslaeghe, K.; Mackerell, A.
D., Jr. J. Chem. Inf. Model. 2012, 52, 3144. (c) Vanommeslaeghe, K.;
Raman, E. P.; Mackerell, A. D., Jr. J. Chem. Inf. Model. 2012, 52, 3155.
(26) (a) van der Spoel, D.; van Maaren, P. J.; Caleman, C.
Bioinformatics 2012, 28, 752. (b) Caleman, C.; van Maaren, P. J.;
Hong, M.; Hub, J. S.; Costa, L. T.; van der Spoel, D. J. Chem. Theory
Comput. 2012, 8, 61.
(27) For details, see section 4 of the Supporting Information.
(28) (a) Godt, A.; Schulte, M.; Zimmermann, H.; Jeschke, G. Angew.
Chem., Int. Ed. 2006, 45, 7560. (b) Jeschke, G.; Sajid, M.; Schulte, M.;
Ramezanian, N.; Volkov, A.; Zimmermann, H.; Godt, A. J. Am. Chem.
Soc. 2010, 132, 10107.
(29) Dunn, P. F. Measurement and Data Analysis for Engineering and
Science; McGraw-Hill: New York, 2005.
(30) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 2nd ed.;
Kluwer Academic/Plenum Publishers: New York, 1999.
(31) Lipari, G.; Szabo, A. Biophys. J. 1980, 30, 489.
(32) Norris, J. R. Markov Chains; University of Cambridge:
Cambridge, 1998.
(33) See Figure S9 in section 5 of the Supporting Information.
(34) (a) Schuler, B.; Muller-Spath, S.; Soranno, A.; Nettels, D.
̈
̈
Methods Mol. Biol. 2012, 896, 21. (b) Wozniak, A. K.; Schroder, G. F.;
̈
Grubmuller, H.; Seidel, C. A. M.; Oesterhelt, F. Proc. Natl. Acad. Sci.
̈
U.S.A. 2008, 105, 18337.
2449
dx.doi.org/10.1021/ja4104292 | J. Am. Chem. Soc. 2014, 136, 2441−2449