pH-Sensitive Vesicles Formed by Amphiphilic Grafted Copolymers

Aug 11, 2016 - ... Grafted Copolymers with Tunable Membrane Permeability for Drug ... and drug loading/release, and provides bottom-up insights toward...
1 downloads 0 Views 6MB Size
Download PDF
Article pubs.acs.org/Macromolecules

pH-Sensitive Vesicles Formed by Amphiphilic Grafted Copolymers with Tunable Membrane Permeability for Drug Loading/Release: A Multiscale Simulation Study Zhonglin Luo,†,‡ Yan Li,† Biaobing Wang,† and Jianwen Jiang*,‡ †

School of Material Science and Engineering, Jiangsu Collaborative Innovation Center of Photovolatic Science and Engineering, Changzhou University, Changzhou, Jiangsu 213164, China ‡ Department of Chemical and Biomolecular Engineering, National University of Singapore, 117576 Singapore S Supporting Information *

ABSTRACT: By synergizing molecular dynamics and dissipative particle dynamics simulations, we investigate the assembly of amphiphilic grafted copolymers into vesicles and the loading/release of doxorubicin hydrochloride (DOX·HCl). The copolymers, PAE-g-PEGLA, comprise pH-sensitive poly(β-amino ester) grafted with hydrophilic poly(ethylene glycol) and hydrophobic poly(D,L-lactide). The vesicle formation is revealed to follow an aggregation−rearrangement mechanism, in which small clusters first form, then rearrange, and finally merge into bilayer-structured vesicles. The vesicle interior size and membrane thickness are substantially affected by the exchange quantity and frequency between tetrahydrofuran and water. At pH = 7, DOX·HCl is loaded into the vesicle interior, and the loading efficiency increases with increasing polymer concentration. At pH < 7, PAE blocks are protonated and hydrophilic, which causes the structure transition of membrane thus tuning membrane permeability for DOX·HCl release. When PLA blocks become longer, vesicle stability is enhanced and DOX·HCl release is suppressed. To mimic controlled release, a mixture of two copolymers is proposed, which form hybrid vesicles and lead to a moderate release rate of DOX·HCl. After multiple sequential pH variations between acidic and neutral circulatory environment, DOX·HCl is gradually released from the hybrid vesicles. This multiscale simulation study identifies the key factors governing vesicle formation and drug loading/release, and provides bottom-up insights toward the design and optimization of new amphiphilic polymers for high-efficacy drug delivery.

1. INTRODUCTION In aqueous solutions, amphiphilic polymers can assemble into various morphologies such as micelles, rods, toroids, vesicles, and tubes.1,2 Among these, vesicles (or called polymersomes), consisting of an aqueous interior surrounded by a membrane, are particularly interesting because of their excellent performance as drug carriers,3−7 bioreactors,8,9 and artificial biological cells.10,11 Polymersomes are structurally similar to natural phospholipid liposomes, but they can be formed by a wide variety of synthetic polymers. Their mechanical stability, membrane permeability, and other properties can be rationally designed and tuned for specific applications.12,13 If used as drug carriers, the membrane permeability and controlled rate of drug release are crucial factors, and they can be modulated through external stimulation, e.g., by varying pH, temperature, and composition or by incorporating light, reduction, and chemical agents.12,14 Practically, pH-sensitive polymersomes are especially intriguing because of naturally occurring pH difference between tissues. The extracellular pH of most tumor tissues is ∼ 6.8, lower than that of normal cells (pH ∼ 7.4);15 such pH difference is caused by the elevated level of lactic acid produced during tumor metabolism. The membrane of a pH-sensitive polymersome can undergo structure transition upon trans© 2016 American Chemical Society

ferring from normal tissue to extracellular tumor tissue, thus leading to tunable permeability to control drug delivery. A number of pH-sensitive polymersomes have been reported. It is recognized that the incorporation of ionizable groups into a copolymer can cause structure change by pH stimulus and form tunable membrane. Via dissociation of a boronic acid copolymer embedded in a polymersome membrane, Kim et al. found that the membrane permeability could be tuned by hydroxide ions and sugar stimulus.8 From a copolymer comprising acrylic acid and distearin acrylate, Chiu and coworkers produced a polymersome consisting of open channels at pH = 8.0 but closed channels at pH = 5.0.11 These polymersomes were synthesized by nonbiodegradable and nonbiocompatible copolymers. Recently, Kim and Lee reported biodegradable and biocompatible copolymers (PAE-g-PEGLA), i.e., poly(β-amino ester) grafted with hydrophilic poly(ethylene glycol) and hydrophobic polylactide. PAE-g-PEGLA could assemble into polymersomes by the solvent exchange method, first dissolved in tetrahydrofuran (THF) and then exchanged Received: June 6, 2016 Revised: July 26, 2016 Published: August 11, 2016 6084

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

polymersomes were formed by exchanging THF and H2O. To explore this formation process, we also simulate the effects of THF/H2O exchange rate on the dynamics and morphology of polymersome formation. It was demonstrated recently by both experiment and simulation that solvent exchange rate plays a key role in polymersome formation. At a fast rate, a bilayerclosing pathway was observed; at a slow rate, however, polymersome was formed via the diffusion of hydrophilic blocks into the center of irregular aggregates, which is similar to nucleation−growth mechanism.31

with water. The polymersomes were found to possess pHdependent membrane permeability modulated by the packing state of PLA and affected by the ionization of PAE backbone. Calcein was stably loaded into the polymersomes at physiological pH = 7.4 and then released under acidified endosomal pH = 4.5−6.5. After hydrolysis, degradable and nontoxic compounds (lactic acid, PEG, and β-amino acid) were produced by PAE-g-PEGLA; consequently, the grafted copolymers are promising candidates for drug delivery.16 To develop new polymersomes for high-efficacy drug delivery, it is indispensable to fundamentally understand their formation process and mechanism. Indeed, the formation pathway plays an important role in drug loading.17 Along with many experiments attempted toward this end,18−21 theoretical and simulation studies have been reported as they are helpful in examining microscopic formation process and providing bottom-up insights. From Brownian dynamics simulation, polymersomes were proposed to form via the bending and closing of a flat bilayer (bilayer-closing mechanism), which grew up from micelles, rods, disks, or bowels.22 Three contributions were identified to compete during polymersome formation: the interfacial interaction between hydrophobic blocks and outside solution, the stretching of hydrophobic blocks, and the repulsive interaction among hydrophilic chains. Although experimentally proved to be useful, this mechanism is not applicable in all cases.23 Another formation pathway was revealed by mesoscopic field-based simulation, in which polymersomes were found to spontaneously form via direct evolution of micelles and rearrangement of hydrophilic blocks (nucleation−growth mechanism).24,25 A recent study using dissipative particle dynamics (DPD) method suggested that polymersome formation routes could be modulated by adjusting hydrophobic/hydrophilic ratio, Flory−Huggins interaction parameters, and polymer concentration.26 The above-mentioned studies on the formation process and mechanism of polymersomes were largely for diblock or triblock copolymers, and there is scarce endeavor for grafted copolymers. In backbone-selective solvents, grafted copolymers can form richer complex morphologies. Besides unilamellar vesicles, hierarchical vesicles such as onion-like multilayered were experimentally observed for grafted copolymers.27−29 The effects of grafting density, side-chain length, interaction parameter, and polymer concentration on morphological outcomes were examined.28 However, these morphologies are thermodynamically controlled, and their dynamic formation process remains elusive. Several interesting questions are associated with grafted copolymers: Is the pathway of polymersome formation different from that of block copolymers? Can the morphologies be dynamically regulated by the preparation method? How do the polymersomes affect drug loading and release? These questions are fundamentally interesting, and quantitative understanding is critical for the development of new grafted copolymers for drug delivery. Here, we report a multiscale simulation study to investigate the assembly process and mechanism of grafted copolymer PAE-g-PEGLA synthesized by Kim and Lee16 as well as the loading/release of doxorubicin hydrochloride (DOX·HCl) in PAE-g-PEGLA polymersomes. As one of the most important medications needed in basic health system, DOX is commonly used to treat various cancers such as hematological malignancies carcinomas and soft tissue sarcomas. It is administered intravenously as a hydrochloride salt, DOX· HCl.30 In the experiment by Kim and Lee, PAE-g-PEGLA

2. METHODOLOGY The multiscale simulation methodology adopted here synergizes both molecular dynamics (MD) and DPD techniques. With a fully atomistic detail, MD simulations were used to calculate the Flory−Huggins interaction parameters between binary components; on this basis, DPD simulations were applied, at a coarse-grained level, to investigate the assembly process of PAE-g-PEGLA with various block lengths and the loading/release of DOX·HCl upon changing pH. Practically, assembly process and drug loading/release occur on a longtime scale, usually on the order of several microseconds or longer. Fully atomistic MD simulations are prohibitively timeconsuming to mimic such phenomena. In this context, mesoscale DPD simulations are commonly adopted in which the degrees of freedom are reduced to accelerate simulations. 2.1. MD Simulations. Figure 1a illustrates the atomistic structures of PAE-g-PEGLA and DOX·HCl. The Flory− Huggins interaction parameters χij among polymer blocks

Figure 1. (a) Atomistic structures of PAE-g-PEGLA and DOX·HCl. (b) Coarse-grained models of PAE14-g-P(EG8)(LAx)13, DOX·HCl, H2O, and THF. 6085

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

convergence threshold of 10−4 kcal mol−1 and a force convergence of 0.005 kcal mol−1 Å−1. The van der Waals interactions were calculated with a cutoff of 12.5 Å, a spline width of 1 Å, and a buffer width of 0.5 Å, while the electrostatic interactions were estimated by the Ewald summation with an accuracy of 0.001 kcal mol−1. After minimization, three configurations with the lowest energies were chosen for equilibration. Initially, 40 isothermal and isochoric (NVT) annealing circles from 300 to 800 K and then back to 300 K (at 300, 550, 800, 550, and 300 K intervals) were conducted; then 5 ns isothermal and isobaric (NPT) MD simulations were carried out at 298 K using the Nośe thermostat and 1 bar using the Andersen barostat for pure components. The final 1.5 ns was used to calculate the equilibrium density and potential energy. For binary components, the configurations were built with a density estimated from the volume fractions of two pure components assuming no volume change after mixing. Then, 5 ns NVT MD simulation was performed at 298 K after 40 NVT annealing circles, and the final 1.5 ns was used to calculate potential energy. 2.2. DPD Simulations. A key feature of DPD method is momentum conservation between dissipative particles. Compared with MD method, it is able to simulate much longer hydrodynamic time and length scales. A system in DPD method is represented by a set of soft beads. The forces acting on a bead i can be described by37

(PEG, PLA, PAE, and PAEH, which is the protonated state of PAE), DOX·HCl, and solvents (H2O and THF) were calculated by MD simulations χij =

ΔEmix Vr RTϕϕ V i j

(1)

where R is gas constant and T is temperature; ϕi are ϕj are the volume fractions of components i and j, respectively; V is total volume and Vr is a reference volume; ΔEmix is the energy of mixing ΔEmix = Eij − (Ei + Ej)

(2)

where Eij is the potential energy of a binary mixture; Ei and Ej are those of corresponding pure components i and j, respectively. This approach was revealed to be more reliable than the traditional method based on solubility parameters32,33 and has been used by us to investigate drug loading/releasing in diblock copolymer PAE−PEG.34 Table 1 lists all the pure and Table 1. Pure and Binary Components Examined by MD Simulations component

repeat units

no. of molecules

density (g/cm3)

H2O THF DOX·HCl PAEa PAEHa PEGa PLAa DOX·HCl/H2O PAE/H2O PAEH/H2O PLA/H2O DOX·HCl/THF PAE/THF PEG/THF PLA/THF PAE/DOX·HCl PAEH/DOX·HCl PEG/DOX·HCl PLA/DOX·HCl PEG/PAE PLA/PAE PEG/PAEH PLA/PAEH PLA/PEG

1 1 1 2 2 11 8

900 250 56 40 40 50 45 12/2700 12/2700 12/2700 12/2700 12/500 12/500 12/500 12/500 30/10 30/10 50/10 45/10 20/20 20/20 20/20 20/20 20/20

0.958 0.854 1.331 1.094 1.083 1.084 1.162 1.000 0.968 0.981 0.978 0.918 0.891 0.882 0.901 1.141 1.128 1.124 1.188 1.090 1.127 1.083 1.119 1.127

fi =

∑ (FCij + FijD + FijR ) + f Si + f iA j≠i

(3)

FCij

where is the conservative repulsive force accounting for excluded volume effect, FDij is the dissipative force for viscous drag between moving beads, and FRij is the random force representing stochastic impulse. Both FDij and FRij act together as a thermostat for the beads. The remaining terms fSi and fAi are the spring forces controlling bond stretching and bending, respectively. All the forces are short-ranged within a cutoff rcut. Specifically, the conservative force FCij is given by ⎧ ⎪ aij(1 − rij/ rcut) (rij < rcut) FCij = ⎨ ⎪ (rij > rcut) ⎩0

(4)

where aij is a repulsive parameter between two beads. By assuming aii = ajj, the interaction parameters aij can be estimated from Flory−Huggins parameter χij through38 aij = aii + 3.50χij

(5)

where χij can be mapped from a long polymer chain to a short DPD chain or from several small molecules to one DPD bead. In this context, eq 5 bridges the gap between atomistic MD and mesoscale DPD methods. In this study, one repeat unit of PAE, as shown in Figure 1a, was grouped into one DPD bead. To meet the requirement that all the beads should have identical volume, other compounds were grouped accordingly, as listed in Table S1. For example, 6 repeat units of PEG and 4 repeat units of PLA were grouped into one PEG and one PLA bead, respectively. Experimental 1H NMR spectra demonstrated that the number-averaged molecular weights of PAE and PEG were 3822 and 2100, and the grafting ratio of PEG/PLA was 1/13.16 Therefore, the coarse-grained models of grafted copolymers were represented by PAE14-g-P(EG8)(LAx)13, as shown in Figure 1b. The numbers 14, 8, and x indicate the block lengths of PAE,

a

To avoid the formation of artificial hydrogen bonds, PAE, PEG, and PLA oligomers were end-capped by methyl groups.

binary components in this study examined by MD simulations. In principle, χij depends on the concentrations of components i and j; however, such dependence was not explored here due to expensive simulations involved. All the components were described by the COMPASS force field.35 Each system was constructed by the Amorphous Cell in Materials Studio.36 PAE, PEG, and PLA oligomers were end-capped by methyl groups. In an acidic environment, PAE is protonated at its tertiary amine groups (PAEH); thus, Cl− ions were added in PAEHcontaining systems for electrical neutrality. To eliminate unfavorable contacts, the initial configurations were subjected to 10 000 steps of energy minimization with an energy 6086

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

PAE and PLA are hydrophobic as experimentally observed.16 DOX·HCl and PAEH contain protonated tertiary amines, which have strong interaction with H2O, but they comprise mostly hydrophobic groups; therefore, the volume of single H2O molecule was used as Vr in eq 1 to estimate χDOX·HCl/H2O and χPAEH/H2O. For THF-containing pairs, the estimated χDOX·HCl/THF, χPAE/THF, χPEG/THF, and χPLA/THF are −3.56, −0.02, 0.73, and −1.38, respectively, indicating THF is a cosolvent for all components. For DOX·HCl-containing pairs, both DOX·HCl and PAEH possess protonated amine groups, and the estimated χPAEH/DOX·HCl is 0.06. Detailed analysis indicates that the hydroxyl and tertiary amine groups in DOX· HCl have attractive interactions with ethylene oxide groups in PEG (data not shown), and the calculated χPEG/DOX·HCl is −0.51. PAE and PLA are immiscible with DOX·HCl; thus, both χPAE/DOX·HCl = 2.90 and χPLA/DOX·HCl = 9.48 are positive. For blocks in PAE-g-PEGLA, χPEG/PAE = 1.65 and χPLA/PAE = 0.90, which imply the interactions in PEG/PAE and PLA/PAE pairs are slightly repulsive. We further calculated the solubility parameters δ from Ecoh /V , where Ecoh is the cohesive energy equal to the energy difference between vacuum and bulk phase.33 Table S2 lists the solubility parameters δ for PEG, PLA, PAE, and PAEH as well as their van der Waals (δvdW) and electrostatic (δelec) terms. We should note that δ for PEG is 20.93, slightly different from 21.81 in our previous study.34 This is because PEG oligomers are end-capped by methyl groups in this study, while they were previously capped by hydroxyl groups. For all the four blocks, the values of δ are close. According to the “like dissolves like” principle, these blocks are miscible. Indeed, PLA and PEG were experimentally observed to be miscible at an amorphous state.43 This is also consistent with the interaction parameter χPEG/PLA of −0.23. For PAEH, δvdW = 15.32 is smaller than that for other blocks; however, the reverse is true for δele = 12.43. Overall, the δ for PAEH is close to those δ for other blocks. 3.2. Solvent Exchange and Vesicle Formation. From the Flory−Huggins parameters in Table 2, the conservative force constants aij were calculated from eq 5, as listed in Table 3. These aij were used in DPD simulations to examine vesicle

PEG, and PLA, respectively, while 13 means the number of grafted PLA chains. The value x can vary from 1 to 4, referring to the molecular weight of PLA from 288 to 1152. At acidic pH, PAEH joined with Cl− were considered to have the same size as PAE. Moreover, 16 H2O molecules, 3 THF molecules, and 0.66 DOX·HCl molecules were grouped separately into one DPD bead. After these groupings, the DPD simulations were conducted using DL_MESO.39 The cutoff rcut was set as a unit length, the bead mass m as a unit mass, and the thermal energy kBT as a unit energy. All the beads had the same volume as that of one PAE repeat unit, about 4.8 × 102 Å3 (estimated from the molecular weight and density). Thus, the physical length of rcut was 3 3 × Vbead = 11.3 Å at ρ = 3, the average mass was 300 g/mol, and the DPD reduced time was t = rcut m /kBT = 0.013 ns at 298 K. The thermal energy kBT = 1 was used to maintain the default values for the dissipation parameter of 4.5 and the spring constant of 4.0. No dihedral angle constraint was included in the DPD simulations. For each system, the simulation box size was 70 × 70 × 70 containing 1.029 × 106 beads, corresponding to 79.1 × 79.1 × 79.1 nm3. The time step adopted was 0.05, and the equilibrium duration was 10 000 steps (i.e., 6.5 ns) for each THF/H2O exchange, unless otherwise stated, or 2 000 000 steps (i.e., 1.3 μs) for drug release at pH < 7.

3. RESULTS AND DISCUSSION 3.1. Flory−Huggins Interaction Parameters and Miscibility. Table 2 lists the Flory−Huggins interaction Table 2. Flory−Huggins Interaction Parameters χij Calculated by MD Simulations

H2O THF DOX· HCl PAE PAEH PEG PLA a

H2O

THF

DOX· HCl

0 2.0a −0.98

0 −3.56

0

9.40 −1.34 0.3b 8.51

−0.02 0.73 −1.38

2.90 0.06 −0.51 9.48

PAE

PAEH

PEG

PLA

0 −0.09 3.97

0 −0.23

0

0 1.65 0.90

Table 3. Conservative Force Constants aij Used in DPD Simulations

b

From ref 41. From ref 42.

parameters χij for binary components. In experiment, nanoscale aggregates were observed for THF/H2O mixtures with a low amount of THF;40 thus, χTHF/H2O = 2.041 was used in the entire concentration range of THF. For PEG/H2O, χPEG/H2O = 0.30 was adopted from the literature. 42 For other binary components, χij were calculated using eqs 1 and 2. It worthwhile to point out that the χij for PAE/H2O, PAEH/ H2O, PAE/PEG, and PAEH/PEG differ from the values reported previously.34 This is due to different chemical structures: PAE and PAEH here contain aliphatic chains, while the previous ones had six-member rings. Consequently, PAE in the previous study was more hydrophobic (unfavorable interaction with H2O), as also reflected by the larger value of χPAE/H2O (28.67 versus 9.40). To estimate χPAE/H2O and χPLA/H2O, Vr in eq 1 was set as the molar volume of one AE repeat unit. The χPAE/H2O and χPLA/H2O were estimated to be 9.40 and 8.51, respectively, suggesting

H2O THF DOX· HCl PAE PAEH PEG PLA

H2O

THF

DOX· HCl

25 32.0 21.57

25 12.54

25

57.9 20.31 26.05 54.79

24.93 27.56 20.17

35.15 25.21 23.22 58.18

PAE

PAEH

PEG

PLA

25 24.89 38.90

25 24.20

25

25 30.78 28.15

formation as well as drug loading/release. In Kim and Lee’s experiment, vesicles were produced by dissolving PAE-gPEGLA copolymer in THF and then exchanging with H2O.16 To explore if vesicles would be formed by simply dissolving the copolymer in H2O, the assembly of PAE14-g-P(EG8)(LA3)13 was simulated in H2O. This approach is commonly used in experiments to prepare micelles.44 As shown in Figure S1, spherical irregular aggregates are formed by PAE14-g-P(EG8)(LA3)13 in H2O with a volume fraction of ϕp = 4%, and there is 6087

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

attributed to the increase of membrane thickness; nevertheless, the interior remains no discernible change (Figures 2d,f). This reveals a longer teq (a lower frequency of solvent exchange) facilitates the formation of vesicles with thicker membrane. To examine the effect of exchange quantity vex, Figure S3 illustrates the morphologies with various vex = 2.5%, 3.3%, 5%, and 10%, and a constant equilibrium time teq = 6.5 ns. With increasing vex from 2.5% to 3.3%, the vesicles tend to be larger particularly in the interior, whereas the membrane thickness keeps nearly a constant. Further increasing vex to 5% and 10%, however, the interior sharply shrinks and the vesicle size significantly decreases. These results imply that an increase of THF/H2O exchange quantity vex would favor the formation of vesicles with a smaller interior. Figures 3a−g show the morphologies of PAE14-g-P(EG8)(LA3)13 during various stages of THF/H2O exchange with vex = 2.5% and teq = 6.5 ns. With increasing ϕH2O from 60% to 80%, the polymer chains first aggregate irregularly (Figure 3b), then rearrange (Figures 3c,d), and finally form vesicles possessing a large interior and a thin membrane (Figure 3e). Upon further exchange (ϕH2O = 82.5% and 85%), the vesicles are broken (Figure 3f) and shrunk to form new vesicles with a smaller interior and a thicker membrane (Figure 3g). At ϕH2O = 100%, i.e., when THF is completely exchanged by H2O, the dynamic assembly is illustrated in Figures 3h−k. Initially, the vesicles with a large interior are observed; along with time, the vesicles are shrunk with a smaller interior and thicker membrane. These simulation results reveal an aggregation−rearrangement mechanism for vesicle formation, also suggest that the exchange volume vex = 2.5% should be further reduced to allow the formation of vesicles with a large interior. Consequently, vex = 0.05% and 0.025% were applied in simulations for THF/H2O exchange when there existed 5−10% and 5% THF in the system, respectively. This very small quantity of solvent exchange is similar to the “dialysis” in experiment.16 Figure S4 shows the dynamic fusion of PAE14-g-P(EG8)(LA3)13 at ϕH2O = 95.675% (i.e. 4.325% THF present). Two stable vesicles are observed until 2.6 ns; afterward, they fuse and merge into a large vesicle that remain stable for a long time. This integrated strategy for THF/H2O exchange, i.e., vex = 2.5% for the first 90% THF and followed by the “dialysis” for the last 10% THF, was adopted for the simulations discussed below. While the effects of exchange volume vex and equilibrium time teq on the vesicle formation of PAE14-g-P(EG8)(LA3)13 are presented above at a constant volume fraction of polymer (ϕp = 4%), it is also instructive to understand how vesicle formation is affected by ϕp. The typical vesicles formed by PAE14-gP(EG8)(LA3)13 at different ϕp are demonstrated in Figure S5. At a low ϕp (e.g., 1%), the vesicle is small but with a thick membrane. Upon increasing ϕp to 3%, both the vesicle size and membrane thickness increase. From ϕp = 3% to 3.5% to 4%, however, the membrane thickness decreases whereas the vesicle size increases. Further increasing ϕp from 4% to 8%, the vesicle radius increases from 15.5 to 22.5 nm, while the membrane thickness remains nearly a constant between 4.8 and 5.0 nm. These variations are quantitatively plotted in Figure 4a. The average radius of gyration for PAE backbone is 2.5 nm; as visualized, the membrane is thus a bilayer structure. The predicted membrane thickness agrees fairly well with experimentally reported 2−3 nm.16 It is worthwhile to note that in simulation of polymer assembly, the assembled structure might be affected by system size. For our case, the vesicle size

no evidence of vesicle formation. The aggregates appear to possess a core/shell structure; the shell contains primarily hydrophilic PEG blocks, whereas the core is composed of randomly arranged PLA, PAE, and even PEG blocks. This is due to the geometric constraint of grafted copolymer chains and the weak repulsive interactions of PEG with PLA/PAE. In this context, the aggregates are not micelles, which commonly possess a hydrophobic core and a hydrophilic shell. Figure S2 shows the dynamic assembly of PAE14-g-P(EG8)(LA3)13 in H2O. At the very early state, small spherical clusters are rapidly formed; they grow up, become relatively larger, and finally merge into aggregates. For PAE14-g-P(EG8)(LAx)13 series (x = 1−4), the ratio of hydrophilic to hydrophobic blocks is 0.30− 0.12, under which micelles and vesicles are usually observed for linear amphiphiles.26,45 However, only irregular aggregates are formed here for PAE-g-PEGLA, suggesting that the grafted copolymers with rich hydrophobic branches possess higher hydrophobicity than linear counterparts. To mimic the formation of PAE-g-PEGLA vesicles experimentally observed by Kim and Lee,16 solvent exchange between THF and H2O was simulated. PAE-g-PEGLA was initially dissolved in THF, which was gradually exchanged by H2O each with a quantify (volume) of vex. After each step, the system was equilibrated with a time duration of teq; then, the next step was performed until no THF existed in the system (i.e., only H2O present as a solvent). The parameters vex and teq were used to adjust the exchange quantity at each step and the frequency between THF and H2O, thus regulating vesicle formation. Figure 2 shows the morphologies of PAE14-g-

Figure 2. Morphologies of 4% PAE14-g-P(EG8)(LA3)13 formed after THF/H2O exchange with vex = 2.5% and various teq: (a) 0.325, (b) 0.65, (c) 3.25, and (d) 6.5 ns. (e) and (f) are the section views of (c) and (d). PEG, PAE, and PLA are in cyan, blue, and red, respectively. H2O is not shown for clarity.

P(EG8)(LA3)13 at ϕp = 4% formed after THF/H2O exchange with vex = 2.5% and various teq = 0.325, 0.65, 3.25, and 6.5 ns. There were 40 exchanges needed to completely exchange THF to H2O, and the morphologies shown were equilibrated in 100% H2O. With teq = 0.325 and 0.65 ns, spherical aggregates similar to Figure S1 are observed (Figures 2a,b). As teq is prolonged, the number of clusters drops whereas the cluster size increases. Interestingly, vesicles are clearly seen with teq = 3.25 ns (Figures 2c,e). PAE and PLA blocks reside within the bilayer membrane, which is surrounded by PEG at the membrane interface exposed to H2O. When teq is further prolonged to 6.5 ns, the vesicle size appears to be larger 6088

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

Figure 3. Morphologies of 4% PAE14-g-P(EG8)(LA3)13 during various stages of THF/H2O exchange with vex = 2.5% and teq = 6.5 ns. ϕH2O: (a) 0%, (b) 60%, (c) 72.5%, (d) section view of (c), (e) 80%, (f) 82.5%, and (g) 85%. Dynamic assembly at ϕH2O = 100%: (h) 0.65 ns, (i) 1.3 ns, (j) 3.25 ns, and (k) 6.5 ns. PEG, PAE, and PLA are in cyan, blue, and red, respectively. H2O and THF are not shown.

Figure 4. (a) Interior radius and membrane thickness versus ϕp. (b) Density profiles of H2O and copolymer blocks at ϕp = 8% versus the distance from vesicle center.

and loose aggregates are observed. With increasing ϕH2O to 70%, many small aqueous compartments are formed in the aggregates, and these compartments fuse to form a tubelike vesicle at ϕH2O = 75%. Further increasing ϕH2O to 90%, a neck is formed and finally evolved to a large vesicle at ϕH2O = 100% when THF is completely exchanged. Again, the aggregation− rearrangement process for the copolymer, along with DOX· HCl loading, is evidenced. Nevertheless, DOX·HCl is not fully loaded with the unloaded DOX·HCl remaining in the solution. Approximately, the loading efficiency is 9%, and ϕDOX·HCl in the vesicle interior is 0.36%, slightly lower than the feed ϕDOX·HCl = 0.4%. This implies that the vesicle is not capable to enrich DOX·HCl during its formation process, although DOX·HCl can be encapsulated in the interior. Figure S6 further shows the loading efficiency and ϕDOX·HCl in vesicle interior versus ϕp. Upon increasing ϕp from 4% to 8%, the loading efficiency increases from 2.4% to 9%. This is because, as illustrated in Figure 4, the radius of interior increases with ϕp; thus, more

might vary if the system is extremely large; however, this is beyond our current computational capability. Figure 4b further elucidates the density profiles of H2O and copolymer blocks versus the distance from vesicle center. Apparently, H2O molecules stay exclusively outside of the membrane (i.e., the interior and aqueous solution). Both hydrophobic PAE and PLA blocks are largely populated in the membrane with a peak at the center, whereas hydrophilic PEG blocks reside mostly at the membrane/H2O interface. As the membrane is bilayer structured, thus two peaks are seen for PEG blocks. 3.3. Loading and Release of DOX·HCl. To examine the loading of DOX·HCl, a mixture of PAE14-g-P(EG8)(LA3)13 and DOX·HCl was simulated. The volume fractions of polymer and drug in the system were ϕp = 8% and ϕDOX·HCl = 0.4%, respectively. During the assembly, THF/H2O exchange was performed with vex = 2.5% and followed by the dialysis method mentioned above. Figure 5 shows the morphologies of PAE14-gP(EG8)(LA3)13 along with DOX·HCl. At ϕH2O = 60%, irregular 6089

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

Figure 5. Morphologies of 8% PAE14-g-P(EG8)(LA3)13 along with 0.4% DOX·HCl loading during various stages of THF/H2O exchange with vex = 2.5% and dialysis method. ϕH2O: (a) 0%, (b) 60%, (c) 70%, (d) 75%, (e) 90%, and (f) 100%. PEG, PAE, and PLA are in cyan, blue, and red, respectively; DOX·HCl is in green; H2O and THF are not shown.

Figure 6. Morphologies of 8% PAE14-g-P(EG8)(LAx)13 along with DOX·HCl release at pH < 7. (a) PAE14-g-P(EG8)(LA1)13, (b) PAE14-gP(EG8)(LA2)13, (c) PAE14-g-P(EG8)(LA3)13, and (d) PAE14-g-P(EG8)(LA4)13. (e)−(h) are the section views of (a)−(d). PEG, PAEH, and PLA are in cyan, blue, and red, respectively; DOX·HCl is in green; H2O and THF are not shown.

bump on the membrane is observed after 1.3 μs, and the interior size decreases (Figures 6b,f); along with release time, the bump grows up, suggesting the vesicle is not stable (Video S2). When x = 3, most DOX·HCl molecules cannot be released within 1.3 μs, and a subsphaeroidal vesicle is formed (Video S3 and Figures 6c,g). Finally, when x = 4, a spherical vesicle is formed with only a few DOX·HCl molecules released (Video S4 and Figures 6d,h). Therefore, with increasing the length x of PLA blocks, vesicle stability is enhanced, and the drug release rate is suppressed. Moreover, the membrane structures at neural and acidic solutions are substantially different. At pH = 7, hydrophobic PAE backbones and PLA blocks are inside the membrane. At pH < 7, however, PAE blocks are protonated to PAEH, become hydrophilic, and separate from PLA blocks; consequently, PAEH reside at the membrane/H2O interface. Such a phenomenon is clearly seen in Figures 6g and 6h for PAE14-g-P(EG8)(LA3)13 and PAE14-g-P(EG8)(LA4)13, respec-

drug molecules can be encapsulated. Meanwhile, ϕDOX·HCl slightly increases from 0.3% to 0.36%. Overall, the loading efficiency in the grafted copolymer under study is low, similar to the case observed in diblock copolymers.17 The morphologies of PAE14-g-P(EG8)(LA3)13 along with DOX·HCl in Figure 5 are considered at pH = 7. Upon reducing pH < 7, PAE blocks are protonated into PAEH and the membrane of vesicle undergoes structure transition; consequently, the membrane is more permeable and drug loaded can be released. Figure 6 shows the morphologies of PAE14-gP(EG8)(LAx)13 along with DOX·HCl release after 1.3 μs. The length of PLA blocks is tuned (x = 1, 2, 3, and 4) while maintaining ϕp = 8%, and the system is at pH < 7. Videos S1− S4 in the Supporting Information provide the detailed dynamic release processes. When x = 1, all DOX·HCl molecules are released, and finally a spherical irregular aggregate is formed by (Video S1 and Figures 6a,e). When x = 2, vesicle with a thick 6090

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

Figure 7. Density profiles of PAE14-g-P(EG8)(LA3)13 and PAE14-g-P(EG8)(LA4)13 at pH < 7.

Figure 8. Percentages of drug release at pH < 7 and of drug leak at pH = 7 versus time.

Figure 9. Dynamic release of DOX·HCl from hybrid vesicles formed by a mixture of 4% PAE14-g-P(EG8)(LA2)13 and 4% PAE14-g-P(EG8)(LA4)13 at pH < 7.

x = 3 and 4. To mimic controlled release with a moderate rate, a mixture of 4% PAE14-g-P(EG8)(LA2)13 and 4% PAE14-gP(EG8)(LA4)13 is used to form hybrid vesicles and examined for DOX·HCl loading and release. In the hybrid vesicles, DOX· HCl is released gradually, and about 66% is released within 1.3 μs. For comparison, the percentages of drug leak from vesicle at pH = 7 are also plotted in Figure 8. When x = 1, 32% DOX· HCl is leaked within 1.3 μs though the vesicle is stable. With increasing x to 2, 3, and 4, the percentage of leak reduces to 18%, 9.5%, and 7.7%, respectively. In the hybrid vesicles at pH = 7, most DOX·HCl molecules are retained with 11% leak within 1.3 μs. Apparently, the percentages of drug release at pH < 7 are higher than those of drug leak at pH = 7. Kim and Lee reported that calcein encapsulated in PAE-g-PEGLA vesicles could be completely released within 15 min at pH = 4.5−6.5.16 The experimentally measured time scale of calcein release is significantly longer than that of DOX·HCl in the current simulation study. In other words, the experimental time scale of drug release is not well captured by the simulation. This is

tively. To quantify, their density profiles are plotted in Figure 7. Only PLA blocks exhibit a peak in the membrane, while PAEH and PEG blocks are more populated at the membrane/H2O interface. This is remarkably different from the profiles in Figure 4b, where PAE blocks are mostly located in the membrane. Furthermore, with increasing x from 3 to 4, the density of PLA in the membrane increases from 2.3 to 3.0, whereas the density of H2O decreases from 0.3 to almost 0. Compared with PAE14-g-P(EG8)(LA3)13, PAE14-g-P(EG8)(LA4)13 contains more hydrophobic PLA blocks and the membrane is less permeable. Consequently, as seen in Figures 6c,d, fewer DOX·HCl molecules are released from PAE14-gP(EG8)(LA4)13 vesicle. For DOX·HCl release from PAE14-g-P(EG8)(LAx)13 vesicle at pH < 7 (Figure 6), Figure 8 quantitatively shows the percentages of release versus time. When x = 1 and 2, the release rate is very rapid (particularly x = 1), and DOX·HCl is almost released completely within 1.3 μs; however, the rate is quite slow, and only 17% and 4% DOX·HCl are released when 6091

DOI: 10.1021/acs.macromol.6b01211 Macromolecules 2016, 49, 6084−6094

Article

Macromolecules

Figure 10. Section views and density profiles of hybrid vesicles formed by a mixture of 4% PAE14-g-P(EG8)(LA2)13 and 4% PAE14-g-P(EG8)(LA4)13 at pH < 7 and pH = 7, respectively. The distance is from vesicle center. In the insets, PEG, PAE (PAEH) for x = 2 and 4, and PLA are in cyan, orange, blue, and gray, respectively; H2O and THF are not shown.

sequential pH variation between acidic (