A combined experimental and computational study of the substituent effect on micellar behavior of γ-substituted thermoresponsive amphiphilic poly(ε-caprolactone)s

Article abstract of DOI:10.1021/ma400855z

The effect of the core substituent structure on the micellar behavior of thermoresponsive amphiphilic poly(ε-caprolactone) diblock copolymer micelles was investigated through a combination of experimental and computational methods. The polycaprolactone (P

Full text of DOI:10.1021/ma400855z

Article
pubs.acs.org/Macromolecules
A Combined Experimental and Computational Study of the
Substituent Effect on Micellar Behavior of γ‑Substituted
Thermoresponsive Amphiphilic Poly(ε-caprolactone)s
Jing Hao,‡,§ Yixing Cheng,†,§ R. J. K. Udayana Ranatunga,‡ Suchithra Senevirathne,‡ Michael C. Biewer,‡
Steven O. Nielsen,‡ Qian Wang,† and Mihaela C. Stefan*,‡
†Department of Chemistry and Biochemistry, University of South Carolina, Columbia, South Carolina 29208, United States
‡Department of Chemistry, University of Texas at Dallas, Richardson, Texas 75080, United States
S
* Supporting Information
ABSTRACT: The effect of the core substituent structure on
the micellar behavior of thermoresponsive amphiphilic poly(ε-
caprolactone) diblock copolymer micelles was investigated
through a combination of experimental and computational
methods. The polycaprolactone (PCL) amphiphilic block
copolymers used in this study consisted of a hydrophilic
poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}
block, which also endowed the polymer with thermores-
ponsiveness, and various hydrophobic poly(γ-alkoxy-ε-caprolactone) blocks. Five different substituents have been attached to the
γ-position of the ε-caprolactone of the hydrophobic block, namely octyloxy, ethylhexyloxy, ethoxy, benzyloxy, and
cyclohexylmethoxy, which self-assembled in aqueous media to generate the core of the micelles. All five synthesized diblock
copolymers formed micelles in water and displayed thermoresponsive behavior with lower critical solution temperature (LCST)
in the range of 36−39 °C. The impact of different substituents on the micelle properties such as size, stability, and phase
transition behavior was investigated. Drug loading and release properties were also studied by employing doxorubicin (DOX) as
payload. Molecular dynamics modeling was used to predict the variation of particle size, free volume, and drug loading capacity.
The drug loading capacity predicted from molecular dynamics simulation was found to be comparable with the experimental
data, which suggests that molecular dynamic simulations may be a useful tool to provide valuable selection criteria for the
engineering of polymeric micelles with tunable size and drug loading capacity.
(OEG) polymers have emerged as a new class of thermores-
ponsive polymers.21,22 The phase transition behavior of the
OEG-functionalized polymers have been shown to be
insensitive to ionic strength, concentration, etc., and thus are
superior to N-substituted acrylamide polymers, like PNIPAM.21
In addition, polymeric micelles from OEG grafted polymers
have shown controllable drug release and improved cellular
uptake in response to temperature increase.23 In our previous
studies, we have reported the synthesis of tri(ethylene glycol)-
substituted amphiphilic polycaprolactone diblock copolymer
(PMEEECL-b-POCTCL) and its self-assembled micelle.14 The
drug loading, cyctotoxicity, and thermo-induced drug release
behavior of this polymeric micelle system have also been
studied.24 The obtained experimental results indicated that this
polyester-based amphiphilic diblock copolymer is an ideal
polymeric micelle system for controlled drug release, thus
deserving further investigation.
INTRODUCTION
■
Polymeric micelles are core−shell structures formed by self-
assembly of amphiphilic block copolymers.1−5 The hydro-
phobic core acts as a microreservoir for the encapsulation of
drugs, while the hydrophilic shell acts as a corona which
protects the micelles from protein adsorption and cellular
adhesion. Among the polymeric micelle systems under clinical
study, micelles formed from amphiphilic aliphatic polyesters
have shown great promise due to their superior biocompati-
bility and biodegradability. Moreover, due to the flexibility and
feasibility of chemical modification of polyesters,6−8 the
properties of these micelles can be tailored by introducing
functional groups to the core or shell to optimize delivery
efficacy and maximize the therapeutic effect.9 For instance,
micellar cores have been conjugated with functional groups,
drug molecules, or cross-linked to enhance drug loading
capacity, micelle stability, and controlled release proper-
ties.10−13 Moreover, micellar shells have been engineered to
achieve active targeting, enhanced cellular uptake, and stimuli-
responsive drug release properties.14−17
Micelle core engineering represents a promising method-
ology to optimize various properties of a micelle as a drug
Imparting thermoresponsive properties to the micelles allows
the micelle to release encapsulated molecules in a controlled
manner upon temperature change.18−20 Oligo(ethylene glycol)
Received: April 25, 2013
Revised: May 30, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
vector.9 Rationally designed functionalities of the micelle core
are expected to enhance the overall performance of micelle
nanocarriers by tuning the drug loading capacity, stability, and
“smartness”.10−12,25,26 Previously, we employed an octyloxy
substituent attached to the γ-position of the ε-caprolactone to
form the micelle core. However, only moderate drug loading
and stability were achieved. To further optimize this class of
amphiphilic thermoresponsive polycaprolactone block copoly-
mers as drug carriers and systematically study the substituent
effect on the micellar assembly, we synthesized five
polycaprolactone amphiphilic block copolymers with different
hydrophobic substituents. Octyloxy, ethylhexyloxy, ethoxy,
benzyloxy, and cyclohexylmethoxy were used as substituents
on the core-forming block. The substituents were chosen to
compare linear vs branched aliphatic substituents, long vs short
aliphatic substituents, and aromatic vs nonaromatic ring
substituents. These functionalities on the core segment were
expected to interact intra- and intermolecularly with the
encapsulated drug molecules by noncovalent interactions,
such as hydrophobic, π−π stacking, and hydrogen bonding.
Doxorubicin (DOX) was employed as a model drug as its
interactions with polymer chains govern the encapsulation
behavior by polymeric micelles.12 In summary, the substituent
effect on thermal-induced phase transition, thermodynamic and
kinetic stability, drug loading, and thermo-induced drug release
of DOX were investigated for this library of polycaprolactone
amphiphilic block copolymers.
rate = 1.0 mL/min, injector volume = 100 μL, detector temperature =
30 °C, and column temperature = 35 °C. All the polymers samples
were dissolved in THF, and the solutions were filtered through PTFE
filters (0.45 μm) prior to injection.
General Procedure for the Synthesis of Amphiphilic Diblock
Copolymers P1−P5. All the monomers were dried by azeotropic
distillation from toluene before the reaction. Dried γ-2-[2-(2-
methoxyethoxy)ethoxy]ethoxy-ε-caprolactone (0.5 g, 1.8 × 10−3
mol) was transferred into a flame-dried 10 mL Schlenk flask under a
nitrogen atmosphere. Stock solutions of Sn(Oct)2 (0.016 g, 3.6 × 10−5
mol) in hexane and benzyl alcohol (0.004g, 3.6 × 10−5 mol) in hexane
were added to the Schlenk flask under a nitrogen atmosphere. The
reaction mixture was deoxygenated by three consecutive freeze−
pump−thaw cycles, and the vacuum of the last cycle was canceled with
nitrogen. The reaction flask was heated in a thermostated oil bath at
110 °C for 4 h. At this time a sample was collected to determine the
monomer conversion and molecular weight by 1H NMR analysis.
Deoxygenated monomers (M1−M5) (1.8 × 10−3 mol) were added to
the reaction flask under a nitrogen atmosphere, and the reaction was
left overnight at 110 °C.
Poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}-b-
1
poly(γ-octyloxy-ε-caprolactone) (P1). H NMR (500 MHz, CDCl3):
δH 0.89 (t, 3H), 1.27 (m, 10H), 1.54 (m, 2H), 1.80 (m, 8H), 2.38 (m,
4H), 3.38 (m, 6H), 3.60 (m, 13H), 4.15 (m, 4H) (the polymer
contains 53.5% PMEEE).
Poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}-b-
1
poly[γ-(2-ethylhexyloxy)-ε-caprolactone] (P2). H NMR (500 MHz,
CDCl3): δH 0.87 (m, 6H), 1.27 (m, 9H), 1.80 (m, 8H), 2.37 (m, 4H),
3.38 (m, 3H), 3.60 (m, 16H), 4.16 (b, 4H) (the polymer contains
54.1% PMEEE).
Poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}-b-
poly(γ-ethoxy-ε-caprolactone) (P3). 1H NMR (500 MHz, CDCl3): δH
1.17 (t, 3H), 1.8 (m, 8H), 2.39 (m, 4H), 3.38 (s, 3H), 3.47 (m, 3H),
3.59 (m, 13H), 4.17 (t, 4H) (the polymer contains 62.8% PMEEE).
Poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}-b-
poly(γ-benzyloxy-ε-caprolactone) (P4). 1H NMR (500 MHz, CDCl3):
δH 1.80 (m, 8H), 2.37 (m, 4H), 3.38 (s, 3H), 3.59 (m, 14H), 4.15 (b,
4H), 4.47 (m, 2H), 7.30 (b, 5H) (the polymer contains 54.4%
PMEEE).
Poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-caprolactone}-b-
poly(γ-cyclohexylmethoxy-ε-caprolactone) (P5). 1H NMR (500 MHz,
CDCl3): δH 0.93 (m, 2H), 1.24 (m, 3H), 1.81 (m, 14H), 2.40 (m,
4H), 3.22 (m, 2H), 3.40 (s, 3H), 3.62 (m, 14H), 4.18 (m, 4H) (the
polymer contains 53.4% PMEEE).
Preparation of Polymeric Micelles and Drug Encapsulation.
Each polymer (20 mg) was dissolved in 1 mL of THF, and 40 μL of
the polymer THF solution was added dropwise to 4 mL of deionized
water under vigorous agitation. THF was removed by dialyzing the
mixture against deionized water for 1 day (MWCO = 3 kDa). The
micelle suspension was filtered through a 0.45 μm filter before
characterization. Doxorubicin (DOX) was encapsulated in polymeric
micelles as a model hydrophobic guest molecule. DOX was first
treated with 3 equiv of triethylamine in DMSO. The neutralized DOX
solution was mixed with each polymer THF solution at a mass ratio of
1:10. The polymer/DOX mixture (40 μL) was added dropwise to 4
mL of deionized water under vigorous agitation. The DOX-loaded
micelle suspension was dialyzed against deionized water for 1 day and
then filtered with a 0.45 μm filter before characterization. Equivalent
DOX concentration in the micelles was determined by fitting readout
absorbance of DOX loaded micelles at 485 nm to a pre-established
standard curve of DOX. Loading efficiency (LE) and loading capacity
(LC) of all polymeric micelles, assuming there was no loss of polymer
during sample preparation, were calculated according to the following
equations:
In addition to exploring the substituent effect experimentally,
we were also interested in whether the micellar behavior can be
predicted by molecular dynamics (MD) simulation. While
molecular dynamics has been applied to study micellation
behavior,27,28 this is the first report of using MD methodology
to study the substituent effect of polymeric micelles. MD
simulations were performed for micelles with the same polymer
backbone and similar functional hydrophobic substituents as
the experimentally synthesized polymers. The drug loading
behavior was predicted by using phenol as a DOX alternative.
The free volume of the micellar core was calculated for both
drug loaded and unloaded micelles. The results indicated that
the interaction between the drug and polymer is more
important than the void volume of the micelle cores in
determining the drug loading capacity (DLC).
MATERIALS AND METHODS
■
Materials and Characterization. All commercial chemicals were
purchased from Aldrich Chemical Co., Inc., and were used without
further purification unless otherwise noted. Stannous(II) 2-ethyl-
hexanoate was purified by vacuum distillation prior to use. γ-Octyloxy-
ε-caprolactone and γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-capro-
lactone monomers were synthesized according to the previously
reported procedure.14
1H NMR spectra of the synthesized monomers and polymers were
1
recorded on a Bruker 500 MHz spectrometer at 30 °C in CDCl3. H
NMR data are reported in parts per million as chemical shift relative to
tetramethylsilane (TMS) as the internal standard. GC/MS was
performed on an Agilent 6890-5973 GC-MS workstation. The
following conditions were used for all GC/MS analyses: injector and
detector temperature, 250 °C; initial temperature, 70 °C; temperature
ramp, 10 °C/min; final temperature, 280 °C. Molecular weights of the
synthesized polymers were measured by size exclusion chromatog-
raphy (SEC) analysis on a Viscotek VE 3580 system equipped with
ViscoGEL columns (GMHHR-M), connected to a refractive index
(RI) detector. GPC solvent/sample module (GPCmax) was used with
HPLC grade THF as the eluent, and calibration was based on
polystyrene standards. Running conditions for SEC analysis were flow
weight of encapsulated DOX
LE (wt %) =
LC (wt %) =
× 100%
× 100%
weight of total DOX
weight of encapsulated DOX
weight of polymer
B
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
Scheme 1. Synthesis of Monomers (M1−M5) and Diblock Copolymers (P1−P5)
weight of encapsulated DOX/MW of DOX
weight of polymer/(DP × mass of repeating unit)
resolved fluorescence spectra of Nile Red loaded micelles were
recorded on a fluorospectrometer at predetermined time intervals
(excitation at 570 nm). The release percentage was represented by the
normalization of fluorescence intensity at 620 nm to the fluorescence
intensity of time zero.
LC (mol %) =
× 100%
Micelle Characterization. Particle size, size distribution, and zeta
potential of blank and DOX-loaded micelles were studied by dynamic
light scattering (DLS) performed on a Zetasizer Nano ZS (Malvern
Instrument) using the filtered suspension without further dilution. All
measurements were carried out at 25 °C. Optical transmittance of
polymeric micelles in response to temperature change was monitored.
Briefly, a polymeric micelle (2−4 mg/mL) suspension was subjected
to gradual heating followed by transmittance measurement by UV−vis
at 600 nm. The lower critical solution temperature (LCST) of the
micelle solution was defined as the temperature at which the
transmittance decreased to 50% of its original value. The critical
micellar concentration (CMC) was determined by fluorescence
spectroscopy using pyrene as a fluorescent probe. Pyrene loaded
micelles were prepared at various polymer concentrations (0.1−10−4
g/L) while the pyrene concentration was kept constant at 3.3 × 10−6
M. Fluorescence excitation spectra (emission at 390 nm) were
recorded on a FP-6200 spectrofluorometer. The morphology of blank
and DOX-loaded micelles was investigated by transmission electron
microscopy (TEM, Hitachi H-8000). TEM samples were prepared by
dipping the micelle suspension onto the grid. Extra suspension was
blotted by Kimwipe. After air-drying, the sample grid was stained with
2% uranyl acetate and viewed at 200 kV accelerating voltage.
Molecular Dynamics Simulations. The SDK coarse-grained
(CG) force field was used for molecular dynamics computer
simulations.30−32 This force field incorporates 3−4 heavy atoms and
their associated hydrogen atoms into a single CG bead. This
representation is used to achieve time and length scales necessary to
simulate aqueous polymer micelles. Simulations were carried out using
the LAMMPS code maintained by Sandia National Laboratories.33
Simulations were run using a multistep rRESPA integrator,33 where an
outer time step of 10 fs was used for the nonbonded interactions and a
time step of 2.5 fs was used for the bonded interactions. A Nose−
Hoover thermostat (at 300 K) and barostat (at 1 atm)34 were used to
generate system configurations from the isothermal−isobaric (NPT)
ensemble.
Simulation of Micelles. Initial coordinates for monomer units
were created by hand and repeated as required to obtain single
polymer chains consisting of a polycaprolactone backbone with 20
hydrophilic and hydrophobic pendant groups. Simulations were
carried out on polymer chains that have similar structure to those
synthesized. Polymers P2 and P4 were simulated with the structures
given above. For P1, the hydrophobic pendent group was changed
from −C8H17 to −C9H20, while for P3 the hydrophobic pendent
simulated was −C3H7 instead of −C2H5. However, polymer P5, which
has a cyclohexyl pendent group, cannot currently be represented using
the SDK force field and hence was omitted from the simulation study.
Single chains were minimized in vacuum, and the resulting structure
was replicated and placed randomly on the surface of a sphere such
that the hydrophobic groups form the core. One hundred chains were
used for each micelle, and the structure was then solvated using CG
water and relaxed for ∼10 ns with a cavitation potential acting on
water to stabilize the micelle during equilibration. Subsequently, the
cavitation potential was released, and the micelle was equilibrated for a
further 20 ns prior to data collection. Coordinates of all beads
(including water) were written every 50 ps, while the solute
coordinates were written every 10 ps.
In Vitro Micelle Stability. The in vitro kinetic stability of micelles
composed of various polymers was measured using Forster resonance
̈
energy transfer (FRET)-based technology.29 To prepare FRET
micelles, the FRET pair 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocar-
bocyanine perchlorate (DiI), 3,3′-dioctadecyloxacarbocyanine per-
chlorate (DiO) was mixed with the polymer (1:100, w/w), and the
mixture was used to prepare FRET micelles using the aforementioned
procedure. The FRET micelle suspension were suspended in 30%
FBS. The micelle suspension was placed on an orbital shaker at room
temperature. At predesigned time intervals, the fluorescence spectra of
all FRET micelles were recorded using a fluorospectrometer. The
excitation was 484 nm, which is the excitation maximum of DiO.
Emission spectra were collected from 490 to 600 nm. The kinetic
Simulations were run for at least 50 ns in the case of micelles absent
of drug analogues. For the simulation of drug loaded micelles, micelles
were created around a spherical globule of 725 phenol molecules, and
simulations were carried out for ∼100 ns to track the escape of phenol
from within the core region to the solvent.
Calculation of Bead Distributions. For each frame of the solute
trajectories, the center-of-mass (COM) location of the core is
calculated. The distance of each hydrophobic bead and hydrophilic
bead from the core COM is tabulated into a histogram with a bin
width of 0.1 Å. The final normalized histograms are presented as
stability of the micelles was quantified by the FRET ratio I565/(I501
I565). I501 and I565 are the fluorescence intensity at 501 and 565 nm,
respectively.
In Vitro Drug Release. Nile Red was chosen as a model
hydrophobic guest to study the substituent effect on the thermo-
induced release behavior of various polymeric micelles. Nile Red (1:10,
w/w) containing micelles were prepared in PBS using a similar
procedure as previously described.24 The Nile Red loaded micelles
were either incubated at room temperature or at 43 °C, which is higher
than the LCST of all polymers as previously determined. The time-
+
C
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
a
Table 1. Molecular Characteristics for Polymers P1−P5
MEEEm/(M1−M5)n
a
b
polymer
feed (m:n)
expt (m:n)
Mn (g/mol)(NMR)
Mn (g/mol)(SEC)
PDI
P1
P2
P3
P4
P5
50:50
50:50
50:50
50:50
50:50
54:53
48:46
47:48
50:52
53:56
27 840
24 490
20 660
25 350
27 390
10 540
10 330
8 870
1.81
1.61
1.85
1.77
2.02
10 850
10 510
a
1
Obtained from H NMR analysis by multiplying the degree of polymerization (DP) of the block copolymer with the molecular weight of the
monomer repeating unit (DP was estimated from the integration of the methylene benzyl protons vs the methylene protons adjacent to the oxygen
in the polymer backbone); m indicates the DP of the hydrophilic block, and n indicates the DP of the hydrophobic block.
probabilities P(r) as a function of radial distance r. The same analysis
is carried out for phenol molecules in the case of the drug loaded
micelles.
P1 without any encapsulation is shown in Figure 1B. The image
indicates that micelles are well-dispersed, and no apparent
Calculation of Void Volume Fraction. For each frame of the
trajectory, 1000 random placed probe points were sampled from
within the core of the micelle, and the number of points that did not
contact any CG bead were noted. A probe point was considered to be
in contact if the distance between the point and the closest bead was
smaller than the sum of the probe radius and that bead’s van der Waals
radius. The fraction of probe points within free volume to the total
number of probe points is taken as the void volume. This procedure
was repeated for probe radii from 0.0 to 4.0 Å.
Analysis of the Retention of Drug Analogue. The number of
drug molecules within 5.0 Å of any hydrophobic pendent bead was
recorded as a function of time. This analysis was found to effectively
represent the total number of hydrophobic molecules within the core
of the micelle. The ability of micelles to retain drug molecules is
related to the drug loading capacity.
RESULTS AND DISCUSSION
■
Synthesis and Characterization of Diblock Copolymer
with Different Side Chains. Syntheses of amphiphilic diblock
copolymers were performed by sequential monomer addition
(Scheme 1). γ-2-[2-(2-Methoxyethoxy)ethoxy]ethoxy-ε-capro-
lactone (MEEE-CL) was used as the hydrophilic monomer,
while γ-octyloxy-ε-caprolactone (M1), γ-(2-ethylhexyloxy)-ε-
caprolactone (M2), γ-ethoxy-ε-caprolactone (M3), γ-benzyl-
oxy-ε-caprolactone (M4), and γ-cyclohexylmethoxy-ε-caprolac-
tone (M5) were used as hydrophobic monomers. The
polymerizations were performed in bulk at 110 °C at a feeding
ratio of MEEE-CL to alkoxy-CL to catalyst to initiator of
50:50:1:1 to ensure that all resulting polymers have comparable
composition. Sn(Oct)2 was used as catalyst, and benzyl alcohol
was used as initiator for the ROP of MEEE-CL monomer to
generate poly{γ-2-[2-(2-methoxyethoxy)ethoxy]ethoxy-ε-cap-
rolactone}. The sequential addition of γ-alkoxy-ε-caprolactone
monomer generated the amphiphilic diblock copolymers
(Table 1).
Preparation and Characterization of Blank and DOX-
Loaded Micelles. Amphiphilic block copolymers self-assemble
in aqueous media into micelles whose morphology is
dependent on the solvent, molecular weight, and composition
of the constituent blocks.35 Our previous study showed that
synthesized polymers with similar structure and composition
could be assembled into spherical micelles.36 To study the
substituent effect on the self-assembled micelles, dynamic light
scattering (DLS) and transmission electron microscopy (TEM)
were employed to investigate the size and morphology of the
resulting micelles. All these diblock copolymers were shown to
self-assemble into spherical micelles in aqueous solution as
predicted. A typical TEM image of air-dried micelles made from
Figure 1. Size distribution of polymeric micelles P1−P5 and TEM
images of polymeric micelle P1 before (A, B) and after DOX loading
(C, D).
aggregation was observed. The estimated size of the micelles
from TEM is ∼30 nm. Diblock copolymers P2, P3, P4, and P5
were also observed to form spherical micelles (data not shown).
The hydrodynamic volume and zeta potential of all micelles
were determined by DLS analysis. As shown in Figure 1A and
Table 2, the size of micelles was on the order of tens of
nanometers. P1 and P5 micelles had the largest hydrodynamic
size, most likely due to the bulky substituents which could
result in an increase of the volume of the hydrophobic core.
The polydispersity index (PDI) of these two polymeric micelles
was also higher, suggesting that the large substituent may
contribute to a broadened distribution. Micelles of P3 were
found to be the smallest in average size, which could be
attributed to the small ethoxy substituent.26 P2 and P4 micelle
sizes are in the intermediate range which was not expected, as
the branched substituents were expected to create more free
volume and give larger micelles.
The hydrophobic core of micelles can sequestrate and
solubilize hydrophobic therapeutic or imaging agents, thus
changing the pharmacokinetics and bioavailability.37 Doxor-
ubicin (DOX) was used as a model drug to study the drug
loading of the amphiphilic diblock copolymer series. As shown
in Figure 1D, the spherical morphology was preserved after
D
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
Table 2. Physicochemical Properties of Blank and DOX-Loaded Polymeric Micelles Based on Polymers P1−P5
polymer
P1
CMC (g/L)
LCST (°C)
size (nm)
PDI
zeta potential (mV)
DLC (wt %)
DLC (mol %)
LE (wt %)
6.04 × 10−3
4.84 × 10−3
8.95 × 10−3
2.48 × 10−3
1.92 × 10−3
37.8
37.5
36.2
37.2
39.4
45.0 9.1
37.3 7.4
26.7 2.4
35.4 4.0
47.6 5.8
49.7 17.6
49.2 8.5
31.2 2.5
50.2 10.8
53.1 15.5
0.33
0.26
0.17
0.24
0.38
0.42
0.27
0.51
0.46
0.47
−11.3 0.6
−13.6 0.5
−16.2 0.5
−11.9 0.4
−14.8 0.7
−12.0 0.1
−13.3 0.3
−15.4 0.5
−12.7 0.6
−16.0 0.3
NA
NA
NA
P2
NA
NA
NA
P3
NA
NA
NA
P4
NA
NA
NA
P5
NA
NA
NA
P1 + DOX
P2 + DOX
P3 + DOX
P4 + DOX
P5 + DOX
2.47
1.59
2.05
2.35
1.41
63.3
35.8
39.0
54.8
35.5
24.7
15.9
20.5
23.5
14.1
drug loading, and this behavior was observed for all other
investigated diblock copolymers (data not shown).
largest transmittance drop from 100% to 16%, which is similar
to the thermo-transition behavior of PMEEE homopolymer.
This is also due to the ethoxy substitution of the core segment
which has the highest hydrophilicity among the polymers. The
half-transmittance drop point for each polymer was taken as the
LCST. It has been observed that LCST is dependent on
molecular weight, and a lower LCST was observed for
oligo(ethylene glycol)-substituted polymers with a lower
molecular weight.39 Even though the LCST is largely
dependent on the balance of hydrophobic/hydrophilic blocks,
it can be assumed that the differences among the LCST values
of the polymers could be attributed to a combination of the
hydrophobicity of the core-segment block and the molecular
weights.
Substituent Effect on Thermodynamic and Kinetic
Stability. The thermodynamic stability is indicated by the
critical micellar concentration (CMC) which can be used to
estimate the potential of micelles to disassemble. Lower CMC
values indicate a higher thermodynamic stability of the micelles.
The CMC of micelles was quantified using pyrene as a
fluorescent probe. The peak shifted from 335 to 339 nm upon
the excitation of pyrene which indicates the incorporation of
hydrophobic pyrene molecules in the micelles. The intensity
ratio (I339/I335) was plotted against the logarithm of polymer
concentration to estimate the CMC value (Figure 3). While the
The size of the micelles increased after DOX loading as
demonstrated by the DLS analysis shown in Figure 1C and
Table 2. This is consistent with what has been observed from
previous studies.26,38 In addition, all the micelles showed
negative surface charge which is desirable because it can
prevent dissociation of the micelles caused by serum protein
adsorption while being administered in blood plasma.
The drug loading content (DLC) and drug loading efficiency
(DLE) were determined by UV−vis spectroscopy. As can be
seen from Table 2, the weight percent DLC are relatively low
for all five diblock copolymers. However, trends can be
estimated based on mole percent DLC. Copolymers P1 and P4
have higher drug content than P2, P3, and P5. For polymer P4,
the enhanced π−π interaction between the polymer and DOX
molecules could account for the increased DLC. However,
despite P1 and P2 having the same number of carbons and
similar hydrophobicity, polymer P1 with the linear alkyl
substituent had a higher drug loading capacity than P2 with a
branched alkyl substituent.
Thermal-Responsive Behavior of the Micelles. The
phase transition of diblock copolymers P1−P5 in aqueous
solution is attributed to a change in the hydrophilic/
hydrophobic equilibrium of the polymers with respect to
their interaction through hydrogen bonding between the
polymer and water molecules.
LCST was determined by measuring the transmittance of
aqueous micellar solutions upon heating. All five amphiphilic
diblock copolymers exhibited LCSTs in the range of 36−39 °C.
(Figure 2 and Table 2) We have demonstrated in our previous
paper that the homopolymer PMEEE had a LCST of 47 °C,
while incorporating a hydrophobic block the LCST was
decreased to 37 °C.14 A similar effect was observed for
polymers P1−P5, and these polymers had a different thermo-
transition range upon heating. Among all polymers, P3 had the
Figure 3. Dependence of intensity ratio I339/I335 (from pyrene
excitation spectra) as a function of polymer concentration for P1−P5.
[Py] = 3.3 × 10−6 M, λem = 390 nm.
CMC values (Table 2) of these polymers are on the same order
of magnitude (10−3 mg/mL) they displayed a variation as a
function of hydrophobic substituent. Micelles of P3, which have
ethoxy substituents on the hydrophobic block, exhibited the
highest CMC value. It can be observed that all polymers with
branched substituents have lower CMCs than the linear
substituents which could be due to the enhanced hydrophobic
Figure 2. Temperature dependence of transmittance of P1−P5 block
copolymers in aqueous solution (0.2 wt % polymer concentration).
E
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
interactions among the branched side chains. The decrease in
CMC for polymer P4 with benzyloxy functional groups can
also be due to increased π−π interactions between polymer
chains in the core.40 The result suggests that not only the
length of hydrophobic segment and hydrophobicity of the
backbone polymer26,41 but also the physical properties of the
substituent on the core-forming segment influence the
thermodynamic stability of the micelles. More hydrophobic
substituents lead to an increase in both intermolecular and
intramolecular hydrophobic−hydrophobic interactions, making
the core more rigid upon dilution after administration.
release of the two dyes. The FRET ratio I565/(I501 + I565), where
I501 is the emission intensity of DiO and I565 is the emission
intensity of DiI, was used to quantify the extent and rate of
dissociation. A faster decrease in the FRET ratio also indicates
faster dissociation of micelles. For each of polymer micelles, the
FRET ratios were normalized to the ratio at time zero and
plotted against incubation time, as shown in Figure 4 (bottom).
The results imply that micelles of P3 showed the fastest
decrease of FRET ratio and reached a plateau only after ∼10 h
of incubation, which indicates that it is the most unstable
micelle against serum protein. Micelles composed of P2 and P5
exhibited the slowest FRET ratio decrease and reached a
plateau after 60 h, suggesting that the substituents on P2 and
P5 help stabilize the micelle and result in higher kinetic
stability. The substituents on P1 and P4 have an intermediate
stabilizing effect based on the intermediate decrease in FRET
ratio.
Substituent Effect on Thermal-Induced Release of
Cargo. Thermoresponsivity is one of the most used stimuli
that have been employed to achieve stimuli-responsive
properties of biomaterials. Thermoresponsive polymer based
nanocarriers for drug delivery aim to program temporal drug
release in response to the temperature change in the
microenvironment. Here we examined the impact of different
substituents on the drug release profile in response to an
elevation of the local temperature. Nile Red fluorescent probe
was encapsulated in the polymeric micelles to estimate the drug
release profile. Nile Red is known to have minimal fluorescence
in an aqueous environment, whereas its fluorescence increases
in a hydrophobic microenvironment such as the micelle core.42
Micelles with encapsulated Nile Red were incubated either at
elevated temperature (43 °C) (above LCST) or at room
temperature (below LCST). The emission spectra (excitation
at 570 nm) were recorded at various time intervals. The
fluorescence intensity at 620 nm at each time interval was
normalized to time zero which was used to represent the
release percentage of Nile Red.
In addition to the thermodynamic stability, Forster
̈
resonance energy transfer (FRET) efficiency between FRET
pair molecules (DiO/DiI) is used to estimate the kinetic
stability, which correlates to the micelle dissociation rate.24 In
the initial micelle, both FRET donors and acceptors are
encapsulated inside the micelle core and stay in close proximity;
thus, the efficiency of FRET should be high. However, if the
micelles start to disassemble, donors and acceptors will be
released and then diffuse further apart which will decrease the
FRET efficiency between the FRET pairs. To test the stability
of micelles with different hydrophobic substituents in the core,
the FRET pair loaded micelles with equal concentrations were
treated with 30% fetal bovine serum (FBS) under mild
agitation. The time-resolved fluorescence emission spectra
(excitation: 484 nm) of the micelle were recorded at certain
time intervals. As shown in Figure 4 (top), increase of DiI
signal and quenching of DiO signal were observed for all
micelles, indicating the successful formation of micelles and
entrapment of FRET pairs. While incubating in FBS containing
buffer, the decrease in DiI signal and the increase in DiO signal
were observed to different extents which is due to the
disruption of micelles by serum proteins and the subsequent
Figure 5. Cumulative Nile Red release from P1−P5 micelles at room
temperature (below LCST) and 43 °C (above LCST) in PBS. All
measurements were performed in triplicate.
As shown in Figure 5, all micelles showed considerable
increase of Nile Red release at 43 °C (higher than the LCST of
polymers) when compared to micelles incubated at room
temperature (lower than the LCST of polymers), suggesting
that the deformation of the hydrophilic outer shell resulted in
the release of loaded cargo. Moreover, these micelles with
different substituents show different release rates of Nile Red.
Nile Red in P2 and P5 micelles exhibited fastest release,
Figure 4. Time-resolved fluorescence spectra of FRET pairs (DiI and
DiO) encapsulated in P1−P5 micelles (top). Stability of FRET
micelles (P1−P5) in the presence of 30% FBS (bottom).
F
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
possibly due to the molecular motion of their bulky
substituents in the amorphous core which then excluded the
cargo from the core. In comparison, the release rate from the
P1 micelle with a linear alkoxy substituent was not as efficient
as the P2 micelle with a branched alkoxy substituent.
Additionally, due to the enhanced interaction between the
benzyl group on P4 and Nile Red, the release of Nile Red was
diminished. The ethoxy group on P3 was found to lead to the
slowest release of Nile Red from the core. These results
suggested that substituents on the core-forming segment do
play a role in regulating the release of cargo in addition to the
physical properties of the polymer backbone. For P2 and P5,
we assumed that molecular motion of bulky substituents
expedites the diffusion of the cargo out of the collapsed micelle
under elevated temperature, while smaller substituents lead to a
slower release of cargo due to a less crowded core. These
results shed more light on the rational engineering of the
micelle core to achieve programmable release.
Figure 7. Micelles of polymers P1−P4 after 50 ns of simulation. Each
micelle is a 100 chain aggregate of A20B20-type polymers. Cutouts are
used to show the internal structure of the micelles. Coloring as in
Figure 6.
Molecular Dynamics Simulations Relating to Micelle
Size, Free Volume, and Drug Loading Capacity.
Molecular dynamics simulations were used to gain insight
into the behavior and interaction of polymers and drugs. A
coarse-grained representation was used for efficiency (Figure
6). Four polymers (P1−P4) were modeled resembling the
Figure 6. Coarse-grained molecular representation used to model
polymers. The beads are colored such that white represents backbone
beads, green represents hydrophobic beads, and blue represent
hydrophilic beads. Left and right images represent the van der
Waals and stick representations for polymer P1, respectively.
experimentally studied species. The modeled hydrophobic
substituents represent the change in hydrophobic chain length:
linear chain vs branched chain and aromatic chain vs
nonaromatic chain. All four modeled diblock copolymers
formed stable spherical micelles in molecular dynamics
simulations with the expected structure of a compact
hydrophobic core and a swollen hydrophilic corona (Figure 7).
The P3 micelle showed the most ellipticity, as it can be
observed from Figure 7. This effect was attributed to the size
mismatch between the hydrophilic and hydrophobic blocks,
leading to the core deforming from a spherical shape with
increased surface area. The sizes of the simulated micelles
(Figure 8) are smaller than the experimental values which is
due to the polymer chain length used in simulations (20 repeat
units vs ∼50 for the synthesized polymers) and the number of
chains within a micelle (100 chains/micelle). Furthermore, in
experiments polymer chains within a micelle are in dynamic
equilibrium with chains in solution, and chains can migrate
between micelles until an equilibrium distribution is achieved.
In the case of simulations, the number of chains for each
Figure 8. Plots of distance distribution of hydrophilic beads (A)
(shown in dashed lines) and hydrophobic beads (B) (shown in solid
lines) from the center of simulated micelle cores. The upper plot
shows the distribution for the pristine micelles while the lower plot
shows those for phenol (listed as drug in the plots) loaded micelles.
micelle remains unchanged. Even with the differences in micelle
size between experimental and computational studies, we
expect the trends in structural properties and behavior of the
micelles to follow the real systems.
Bead Distributions. The distance distribution plots for the
investigated diblock copolymer structures are given in Figure 8
for both drug-loaded and unloaded micelles. The comparison of
distributions shows only slight changes in the core size and the
overall size of the micelles.
For unloaded micelles, the P3 micelle has the broadest
distribution, while micelles of P1 and P4 are relatively compact.
This may be a result of the shape anisotropy of the micelle of
polymer P3. The overall micelle size from DLS measurements
G
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
are related to the hydrodynamic radius of the micelles, which
reflect the hydrophilic corona. The trends shown in simulations
are not consistent with the DLS measurements, which could be
attributed to the constraint of having a fixed number of chains
within a simulated micelle. Moreover, both experimental data
and simulation data indicate small variations in the micelle size
for polymers P1−P4.
Void Volume Measurements. The void volume of the
polymer micelles was expected to have implications in
predicting the drug loading capacity because the void volume
fraction was thought to be indicative of the space available to
accommodate the hydrophobic drug molecules. The void
volume was analyzed for diblock copolymers P1−P4 for a
range of probe sizes (see Figure 9). The plots follow an
Figure 10. Polymer micelle P1 loaded with phenol (colored red), the
DOX analogue. A cutout is used to show the internal structure of the
micelle.
Figure 9. Void volume fraction for polymeric micelles as a function of
probe size.
intuitive trend, namely, micelles from diblock copolymers with
branched hydrophobic substituent are the least well-packed and
consequently have the highest void volume fraction. By
contrast, the micelles with linear and benzyl substituents
showed better packing, with the benzyl chains showing slightly
better packing than the linear alkyl chains (Figure 9).
Figure 11. Prediction of drug loading capacity on A20B20-type
polymers by using nondegenerative drug (diffusion) release.
Drug Loading and Retention. The prediction of drug
loading using only void volume neglects two essential factors:
(1) the interactions between the drug molecules and polymers
and (2) the rearrangement of the polymer chains upon drug
loading. We therefore introduced a DOX analogue to better
estimate the drug loading capacity. Micelles were created with a
high ∼10% w/w loading of hydrophobic phenol molecules
(Figure 10). Phenol was chosen as the DOX analogue due to its
similar octanol:water partitioning behavior. Simulations were
run by initially creating a micelle structure around a globule of
phenol. Because the number of phenol molecules loaded is
initially too high for the core to retain, and due to the partial
solubility of phenol in water, some molecules of phenol diffused
from the core to the solvent. By tracking the number of phenol
molecules retained inside the core, we were able to estimate the
drug loading capacity of the polymer micelles. The plot of
phenol retention vs time showed the decrease in drug retention
within the simulation time frame of 100 ns (Figure 11). This
data allows us to estimate the relative drug loading capacity of
the micelles as P3 ∼ P2 < P4 < P1 (Figure 11). This is an
important result because the trend shown here is different from
what we expected from the void volume measurements which
indicated that polymers with branched chains would have the
highest loading capacity. However, the simulations for drug
loaded micelles show that the branched and short chain groups
are the least effective at retaining the drug molecules, which is
possibly due to the inefficient packing of the phenol molecules
with the core substituents and weak interactions between the
drug analogue and the polymer chains. The benzyl and octyl
groups have a high loading; π−π interactions may be partly
responsible for the better performance of the aromatic group.
This trend of drug loading behavior was further supported by
the probability bead distributions (Figure 8). Based on the
overlapping intensity and areas of the phenol molecule
distribution with the hydrophobic block distribution, a similar
trend of P3 < P2 < P4 < P1 can be estimated. Note that the
number of drug molecules found between r and r + dr from the
micelle core center is 4πr2P(r) dr, so that a drug distribution
peaked at a larger r value contains many more drug molecules.
Overall, the molecular dynamics simulation data agreed with
the experimental findings that the straight linear chain octyloxy
and the aromatic benzyloxy offer the best drug loading capacity.
The analysis of void volume does not follow this trend,
indicating that the interactions of the drug with the polymers
play an important role in the loading of drugs into these block
copolymer micelles.
Thermal Analysis. The glass transition temperatures (Tg) of
the polymers P1−P5 were determined by DSC analysis (Table
3 and Supporting Information). All the copolymers displayed
two Tg values corresponding to each block. The Tg for the
common hydrophilic poly{γ-2-[2-(2-methoxyethoxy)ethoxy]-
ethoxy-ε-caprolactone}block was −82 °C for all the diblock
H
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
Table 3. Glass Transition Temperatures of the Diblock
Copolymers P1−P5
ACKNOWLEDGMENTS
■
a
Steven O. Nielsen and Udayana Ranatunga thank Dr. Sharon
Loverde for her assistance and also acknowledge Texas
Advanced Computing Center (TACC) at The University of
Texas at Austin for providing HPC resources that have
contributed to the research results reported within this paper.
Mihaela C. Stefan acknowledges financial support from NSF
(DMR-0956116 and CHE-1126177) and Welch Foundation
(AT-1740). Qian Wang acknowledges the financial support
from NSF CAREER CHE-0748690 and the Camille Dreyfus
Teacher Scholar Award.
polymer
Tg1 (°C)
Tg2 (°C)
polymer
Tg1 (°C)
Tg2 (°C)
P1
P2
P3
−82
−82
−82
−70
−68
−64
P4
P5
−82
−82
−51
−63
a
Tg1 = glass transition temperature of the hydrophilic block; Tg2
glass transition temperature of the hydrophobic block.
=
copolymers. Considering the second Tg for the diblock
copolymers P1−P5, which corresponds to the hydrophobic
block, the polymer P4 containing poly(γ-benzyloxy-ε-capro-
lactone) had the highest Tg of −51 °C, which is due to the
presence of the bulky phenyloxy group. By contrast, the diblock
copolymer P1 containing the hydrophobic poly(γ-octyloxy-ε-
caprolactone) had the lowest Tg of −70 °C which was
attributed to the relatively flexible octyloxy substituents. No
melting temperature transitions were detected on the DSC
thermograms which confirmed the amorphous nature of
copolymers P1−P5.
REFERENCES
■
(1) Rosler, A.; Vandermeulen, G. W. M.; Klok, H.-A. Adv. Drug
Delivery Rev. 2012, 64 (Suppl.), 270−279.
(2) Kataoka, K.; Harada, A.; Nagasaki, Y. Adv. Drug Delivery Rev.
2012, 64 (Suppl.), 37−48.
(3) Tyrrell, Z. L.; Shen, Y.; Radosz, M. Prog. Polym. Sci. 2010, 35,
1128−1143.
(4) Gaucher, G.; Dufresne, M.-H.; Sant, V. P.; Kang, N.; Maysinger,
D.; Leroux, J.-C. J. Controlled Release 2005, 109, 169−188.
(5) Abetz, V.; Gohy, J.-F. In Block Copolymers II; Springer: Berlin,
2005; Vol. 190, pp 65−136.
(6) Jer
1076.
(7) Philippe, L.; Raphael, R.; Step
́
o
̂
me, C.; Lecomte, P. Adv. Drug Delivery Rev. 2008, 60, 1056−
CONCLUSION
■
Five γ-substituted caprolactone hydrophobic monomers were
synthesized and used to generate amphiphilic thermosensitive
polycaprolactone diblock copolymers. The five polymers (P1−
P5) were synthesized by ring-opening polymerization with
Sn(Oct)2 catalyst. The synthesized amphiphilic diblock
copolymers formed micelles in an aqueous environment and
maintained the thermoresponsive behavior imparted by the
PMEEE block. The synthesized micelles were found to be
thermodynamically and kinetically stable and showed thermo-
controlled release behavior upon temperature increase above
LCST. The drug loading capacity (DLC) of the micelles were
estimated to probe the difference generated by the variation of
the substituents on the hydrophobic block. The DLC was
estimated both experimentally by using UV−vis spectroscopy
and computationally (using phenol as the DOX analogue) by
molecular dynamics modeling. Molecular dynamics simulations
showed a similar trend in DLC based on the substituent change
as the experiments. These findings are expected to provide an
easy computational method to predict drug loading before the
synthesis of amphiphilic diblock copolymers and could become
a valuable tool for the engineering of micellar drug delivery
systems.
́
hanie, S.; Jutta, R.; Kathy Van, B.;
̈
Christine, J.; Robert, J. Macromol. Symp. 2006, 240, 157−165.
(8) Albertsson, A.-C.; Varma, I. K. Biomacromolecules 2003, 4, 1466−
1486.
(9) Xiong, X.-B.; Falamarzian, A.; Garg, S. M.; Lavasanifar, A. J.
Controlled Release 2011, 155, 248−261.
(10) Xiong, X.-B.; Ma, Z.; Lai, R.; Lavasanifar, A. Biomaterials 2010,
31, 757−768.
̈
̈
(11) Xiong, X.-B.; UludaAY, H.; Lavasanifar, A. Biomaterials 2009,
30, 242−253.
(12) Yang, C.; Ebrahim Attia, A. B.; Tan, J. P. K.; Ke, X.; Gao, S.;
Hedrick, J. L.; Yang, Y.-Y. Biomaterials 2012, 33, 2971−2979.
(13) Garg, S. M.; Xiong, X.-B.; Lu, C.; Lavasanifar, A. Macromolecules
2011, 44, 2058−2066.
(14) Hao, J.; Servello, J.; Sista, P.; Biewer, M. C.; Stefan, M. C. J.
Mater. Chem. 2011, 21, 10623−10628.
(15) Su, R.-J.; Yang, H.-W.; Leu, Y.-L.; Hua, M.-Y.; Lee, R.-S. React.
Funct. Polym. 2012, 72, 36−44.
(16) Dai, W.; Huang, H.; Du, Z.; Lang, M. Polym. Degrad. Stab. 2008,
93, 2089−2095.
(17) Darcos, V.; El Habnouni, S.; Nottelet, B.; El Ghzaoui, A.;
Coudane, J. Polym. Chem. 2010, 1, 280−282.
(18) Ward, M. A.; Georgiou, T. K. Polymers 2012, 3, 1215−1242.
(19) Bajpai, A. K.; Shukla, S. K.; Bhanu, S.; Kankane, S. Prog. Polym.
Sci. 2008, 33, 1088−1118.
ASSOCIATED CONTENT
(20) Ganta, S.; Devalapally, H.; Shahiwala, A.; Amiji, M. J. Controlled
Release 2008, 126, 187−204.
■
S
* Supporting Information
̈
(21) Lutz, J.-F.; Akdemir, O.; Hoth, A. J. Am. Chem. Soc. 2006, 128,
Experimental procedures for the synthesis of caprolactone
monomers, 1H and 13C NMR spectra. This material is available
free of charge via the Internet at http://pubs.acs.org.
13046−13047.
(22) Lutz, J.-F.; Hoth, A. Macromolecules 2005, 39, 893−896.
(23) Kim, S. H.; Tan, J. P. K.; Fukushima, K.; Nederberg, F.; Yang, Y.
Y.; Waymouth, R. M.; Hedrick, J. L. Biomaterials 2011, 32, 5505−
5514.
AUTHOR INFORMATION
■
(24) Cheng, Y.; Hao, J.; Lee, L. A.; Biewer, M. C.; Wang, Q.; Stefan,
M. C. Biomacromolecules 2012, 13, 2163−2173.
(25) Mahmud, A.; Patel, S.; Molavi, O.; Choi, P.; Samuel, J.;
Lavasanifar, A. Biomacromolecules 2009, 10, 471−478.
(26) Zhao, X.; Poon, Z.; Engler, A. C.; Bonner, D. K.; Hammond, P.
T. Biomacromolecules 2012, 13, 1315−1322.
Corresponding Author
*E-mail: mci071000@utdallas.edu (M.C.S.).
Author Contributions
§J.H. and Y.C. contributed equally.
Notes
(27) Patel, S. K.; Lavasanifar, A.; Choi, P. Biomaterials 2010, 31,
1780−1786.
The authors declare no competing financial interest.
I
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules
Article
(28) Patel, S. K.; Lavasanifar, A.; Choi, P. Biomaterials 2010, 31,
345−357.
(29) Lu, J.; Owen, S. C.; Shoichet, M. S. Macromolecules 2011, 44,
6002−6008.
(30) Shinoda, W.; DeVane, R.; Klein, M. L. Mol. Simul. 2007, 33,
27−36.
(31) Shinoda, W.; DeVane, R.; Klein, M. L. Soft Matter 2008, 4,
2454−2462.
(32) Loverde, S. M.; Klein, M. L.; Discher, D. E. Adv. Mater. 2011,
24, 3823−3830.
(33) Plimpton, S. J. Comput. Phys. 1995, 117, 1−19.
(34) Shinoda, W.; Shiga, M.; Mikami, M. Phys. Rev. B 2004, 69,
134103.
(35) Kim, J. K.; Yang, S. Y.; Lee, Y.; Kim, Y. Prog. Polym. Sci. 2010,
35, 1325−1349.
(36) Cheng, Y.; Hao, J.; Lee, L. A.; Biewer, M. C.; Wang, Q.; Stefan,
M. C. Biomacromolecules 2012, 13, 2163−2173.
(37) Kataoka, K.; Harada, A.; Nagasaki, Y. Adv. Drug Delivery Rev.
(38) Yan, J.; Ye, Z.; Chen, M.; Liu, Z.; Xiao, Y.; Zhang, Y.; Zhou, Y.;
Tan, W.; Lang, M. Biomacromolecules 2011, 12, 2562−2572.
(39) Han, S.; Hagiwara, M.; Ishizone, T. Macromolecules 2003, 36,
8312−8319.
(40) Carstens, M. G.; Bevernage, J. J. L.; van Nostrum, C. F.; van
Steenbergen, M. J.; Flesch, F. M.; Verrijk, R.; de Leede, L. G. J.;
Crommelin, D. J. A.; Hennink, W. E. Macromolecules 2006, 40, 116−
122.
(41) Owen, S. C.; Chan, D. P. Y.; Shoichet, M. S. Nano Today 2012,
7, 53−65.
(42) Jackson, A. W.; Fulton, D. A. Macromolecules 2012, 45, 2699−
2708.
J
dx.doi.org/10.1021/ma400855z | Macromolecules XXXX, XXX, XXX−XXX