Biodegradable poly(ethylene glycol)–poly(ε-carprolactone) polymeric micelles with different tailored topological amphiphilies for doxorubicin (DOX) drug delivery

Abstract

Three poly(ethylene glycol)–poly(ε-carprolactone) (PEG–PCL) copolymers with different topologies but identical molar ratio between PEG to PCL were designed. These copolymers, namely, diblock (L-PEG–PCL), triblock (B-PEG–PCL), and star shaped (S-PEG–PCL) copolymers, were extensively characterized by ¹H Nuclear Magnetic Resonance (¹H NMR), Fourier Transform Infrared Spectroscopy (FTIR), and Gel Permeation Chromatography (GPC) analyses. The effect of topology on crystallization was investigated by X-Ray Diffraction (XRD), Differential Scanning Calorimetry (DSC). Results showed that the diblock copolymer possessed the highest crystallinity, followed by triblock and star shaped copolymers. Although the topology did not affect the self-assembly behavior of copolymers, the effects on size, size distribution, drug loading content, and drug release rate of the polymeric micelles were observed. The micelles self-assembled from linear diblock copolymer achieved higher cellular uptake and lower half maximal inhibitory concentration (IC₅₀). These findings are favorable to establish the foundation to further design proper structure of amphiphilic copolymers as nanocarrier systems for efficient anticancer therapy.

---

1. Introduction

Polymeric micelles have been considered as excellent candidates in the fabrication of drug delivery systems for their widely recognized merits such as enhancement of drug solubility in water, amelioration of side effects, prolongation of circulation time and improvement of drug bioavailability.^1–4 As we know, amphiphilic copolymers can self-assemble into core–shell structured micelles in aqueous medium, where the cores are usually constructed with the hydrophobic polymers and the shells are hydrophilic polymers.^5,6 Poly(ε-caprolactone) (PCL) has been frequently used to form the core of self-assembled amphiphilic copolymers owing to its desirable hydrophobicity, high aggregation propensity, and low glass transition temperature that ensures a partial rubbery core. Poly(ethylene glycol) (PEG) is the most widely used hydrophilic polymer to impart structure stability and stealth properties to reduce non-specific protein adsorption, thus reducing clearance of nanocarriers by monocyte phagocytic system (MPS) cells.^7,8 Herein, PEG–PCL copolymers are currently investigated as carriers in different drug delivery systems, and the most promising application remains in the field of anticancer drug delivery.

In recent years, much interest has been focused on designing different architectures of PEG–PCL copolymers, which can be categorized as linear, graft, block, branched, cross-linked, star-shaped, dendron/dendrimer topologies.^9,10 Among them, the di-block and tri-block polymers are the most common polymers used in drug delivery system. Interestingly, the area of block copolymers has been explored with new macromolecules derived of miktoarm architecture, where a miktoarm star copolymer is formed by different species of non-linear chains in form of star, polymers with these characteristics in its geometric order provide the possibility of new potential properties to conventional linear block copolymers. In this way, some research groups have reported macromolecules with miktoarm star copolymer architecture, as for example: AB₂ and AB₄ where A = poly(ε-caprolactone) (PCL) and B = polystyrene,¹¹ A₂B and A₂BA₂ where A = poly(n-hexyl isocyanate) and B = poly(ethylene glycol) (PEG),¹² or ABC star terpolymer where A = PEG, B = poly(2-vinylpyridine) and C = PCL.¹³ In particular, the unique properties of the structures that assembled from these polymers in solution have led to the application in drug delivery. As miktoarm polymers are branched macromolecules consisting of polymeric arms emanating from their core, this architecture offers great opportunities for the tailoring of nanocarriers, as the chemistry of both the core and the arms can be finely tuned in order to trigger their controlled self-assembly in aqueous media.¹⁴ These unique properties of miktoarm polymers are actively investigated, particularly in small molecule encapsulation and release.^14–16 Obviously, miktoarm star copolymers presented different properties compared to conventional linear block copolymers, many studies have focuse on the relation between polymer architecture and their properties.

Bogdanov and co-workers investigated the isothermal crystallization of three different types of PEG–PCL incompatible block copolymers in detail using DSC. The molecular weight of the PEG blocks they used was 4000 or 5000 and the PEG weight fraction was varied from 18 to 22 wt%. Results showed that the diblock copolymers exhibited the highest crystallization among those three copolymers.¹⁷ He et al. synthesized ABA type (PLA–PEG–PLA) and BAB type (PEG–PLA–PEG) triblock copolymers and compared their drug release properties and stealth particle characteristics.¹⁸ The morphology, stability, thermal characteristics and release profile of ABA and BAB type copolymer nanoparticles based on PCL and PEG chain segments were also investigated by Zamani and co-workers.⁶ They found that the architectures of copolymers had a profound effect on their micellar properties. Moreover, Yin, et al. synthesized a series of AB₂ type 3-miktoarm star copolymers using PEG as the A arm and PLLA as the two B arms and found that the 3-miktoarm polymers have superior vesicle-forming abilities compared to the linear diblock counterparts.¹⁶ Many star polymers with different architectures which can shape various morphologies were also reported.^19–21 From these reported studies, it was obviously found that the extensive researches have been concentrated on synthesis and applications of different novel architectures. Although most of the works focused on the creation of new spatial structure, change of the compositions, intelligence of the functional group and the chain length to investigate those factors influence on their properties, the information on the influence of the motion of the molecular and biological processes exerted by architecture are relatively limited and unsystematically.

Prompted by these limitations, here we designed PEG–PCL amphiphilic copolymers with different structures and the copolymers were used to investigate the effect of topologies of PEG–PCL copolymers on their micellar properties and anticancer efficiency as shown in Scheme 1. A comprehensive investigation on different topologies of three type copolymers with same molar ratio and molecular weight, namely L-PEG–PCL, B-PEG–PCL and S-PEG–PCL, was conducted. The crystallization of three types of PEG–PCL copolymers were investigated by XRD, FTIR and DSC analysis. The effect of topologies of copolymers on their micellar properties such as CMC, particle size and size distribution, drug loading capacity and in vitro drug release profiles, as well as anticancer efficiency were studied.

Scheme 1 Schematic illustration of the self-assembly and drug release of polymeric micelles.

2. Materials and methods

2.1. Materials

Methylated poly(ethylene glycol) (mPEG2k, M_w = 2000 g mol⁻¹), polyethylene glycol (PEG, M_w = 2000 g mol⁻¹), stannous octoate (Sn(Oct)₂), glycerol and ethylene oxide were purchased from Sigma-Aldrich Co. and used as received. ε-Caprolactone (CL) (Sigma-Aldrich Co.) was dehydrated by CaH₂ and purified via vacuum distillation. Doxorubicin hydrochloride (DOX·HCl) (Zhejiang Hisun Pharmaceutical) was dissolved into water and the pH adjusted to 9.6 to prepare doxorubicin.²² Benzyl bromide (BnBr) was bought from Changzhou Xinhua Active Material Institute. Dulbecco's Modified Eagle's Medium (DMEM), fetal bovine serum (FBS), trypsin solution, 100 × penicillin–streptomycin solution were obtained from HyClone Inc. and used for cytotoxicity assay. N,N-Dimethylformamide (DMF) (Kermel Co., Tianjin, China) and tetrahydrofuran (THF) (Kelong Chemical Co.) were purified before used. All the other solvents were purchased from Kelong Chemical Co. (Chengdu, China) and used as received.

2.2. Characterizations

¹H NMR spectra were collected on a Bruker 400 MHz spectrometer using CDCl₃ as solvent and tetramethylsilane (TMS) (0.5%) as internal standard. FTIR spectra were recorded on a FTIR spectrometer (Thermo Fisher Scientific Nicolet 8700) within the range of 4000–400 cm⁻¹ and the samples were prepared using the KBr disk method under the irradiation of sodium lamp. Gel permeation chromatography (GPC) measurement was carried out on HLC-8320 chromotography with THF as eluent (0.6 mL min⁻¹) at 35 °C and respect to polystyrene standards. The thermal properties of copolymers were determined by differential scanning calorimetry (DSC, Q2000 TA Instruments). About 5 mg samples in hermetically closed aluminum pans were subjected to a secondary heating program from −80 to 100 °C at a heating rate of 10 °C min⁻¹ and a cooling rate of 10 °C min⁻¹. Powder X-ray diffractometry (XRD) patterns were obtained on an X'Pert pro MPD X-ray diffractometer at 25 °C using CuK^α (λ = 0.154 nm) as the target material with the diffraction angle of 5–50°. Dynamic light scattering (DLS) tests were performed on Malvern Zatasizer Nano ZS at 25 °C. The samples were dispersed in deionized water with the concentration of 1 mg mL⁻¹ and each sample was filtered through a 450 nm syringe filter before analysis. The scattering angle is 90° and the coefficient of refraction is 1.330. Each sample tested three times. The atomic force microscopy (AFM) pictures were received on a MFP-3D-BIO instrument (Oxford Instruments company) using tapping mode. The AFM samples were prepared by casting a dilute micellar solution (0.1 mg mL⁻¹) on a mica substrate and dried in vacuum. The scan angle was maintained at 0°, and the images were captured in the trace direction with a scan rate of 1 Hz. Cellular uptake was examined by confocal laser scanning microscopy (CLSM, Leica TCP SP5). DOX was 10 μg mL⁻¹ and excited at 480 nm with the emission at 590 nm. The flow cytometry was tested on a BD FACS Calibur flow cytometer (Beckton Dickinson) and analyzed approximately for 1 × 10⁴ cells from each sample.

2.3. Synthesis of phenyl-PEG-1,3-propanediol

A mixture of glycerol (18.4 g, 0.2 mol), benzaldehyde (21.2 g, 0.2 mol) and concentrated sulfuric acid (0.02 mL) were stirred vigorously at room temperature (RT) for 4 h under the protection of nitrogen. The received water was removed under reduced pressure at 40 °C. The residues were cooled to room temperature, added in ethyl ether (25 mL), then washed with 10% Na₂CO₃ and saturated NaCl aqueous solutions (aq) for three times. The organic phase was dried with anhydrous MgSO₄ overnight. The filtrate was concentrated and recrystallized three times at −20 °C to produce white crystal 2-phenyl-1,3-dioxan-5-ol (intermediate 1).²³

2-Phenyl-1,3-dioxan-5-PEG2k (intermediate 2) was synthesized via an anionic ring-opening polymerization modified from the approach previously reported.^24,25 Under the nitrogen atmosphere, potassium (0.19 g, 5 mmol) was added in three-neck flask and then naphthalene (0.64 g, 5 mmol) was added in dropwise. After a vigorous stirring of 2 h, intermediate 1 (0.9 g, 5 mmol) was injected into the mixture and cold ethylene oxide (17 mL) was added immediately. Hydrochloric acid (1 M, 5 mL) was used as a terminator when the reaction was preceded for 48 h at 0 °C. The mixture was stirred vigorously for another 5 h. After THF was removed, the residues were dissolved in CHCl₃, washed with deionized water and NaCl (aq) for three times alternately. The organic layer was dried by MgSO₄ overnight. The filtrate was concentrated and recrystallized in excess cold ethyl ether. The powder (2-phenyl-1,3-dioxan-5-PEG2k, intermediate 2) was vacuum-dried at room temperature overnight.

Intermediate 2 (4 g, 2 mmol) was dissolved in DMF (45 mL), NaH (0.19 g, 8 mmol) was added in the solution. The mixture was stirred for 4 h and then BnBr (1.0 g, 6 mmol) dissolved in DMF was injected in the mixture. The reaction was preceded for 48 h at room temperature. The DMF was removed completely under reduced pressure and CDCl₃ was added. The solution was filtered and the filtrate was washed by saturated NaCl (aq) for 3 times. The pure product (2-phenyl-1,3-dioxan-5-PEG2k-benzyl, intermediate 3) was obtained after the organic phase was removed in reduced pressure.

The intermediate 3 (3 g, 1.5 mmol) was dissolved in methanol (30 mL). After 1 M HCl (aq) (30 mL) was dropped in the system, the mixture was refluxed for 6 h and then concentrated. Dichloromethane was used to extract the organic phase. After washed with deionized water for three times, the organic layer was dried by MgSO₄ and precipitated in cold ethyl ether to produce phenyl-PEG2k-1,3-propanediol (intermediate 4).

2.4. Synthesis of different topologies of poly(ethylene glycol)–poly(ε-caprolactone) (PEG–PCL) copolymers

Different types of PEG–PCL copolymers were synthesized via ring-opening polymerization.²⁶ Phenyl-PEG2k-1,3-propanediol, mPEG2k, PEG2k were used as macroinitiators and Sn(Oct)₂ as the catalyst. The synthesis of linear diblock copolymer (L-PEG2k–PCL4k) was used as an example. mPEG2k (2 g, 1 mmol) was added in a round-bottom flask with a vacuum apparatus. The flask was heated to 110 °C for 2 h with the purpose of removing the residual water completely. When the flask was cooled to room temperature, CL (4 g, 35 mmol) and Sn(Oct)₂ (0.1 wt%) were added quickly. The flask was sealed and immersed in an oil bath at 130 °C for 48 h. The compound was dissolved in dichloromethane and precipitated in excess cold ethyl ether for three times. The purified white powder was obtained when the precipitate was vacuum-dried. The tri-block copolymer (B-PEG2k–PCL4k), star shaped copolymer (S-PEG2k–PCL4k) were synthesized as the same procedure above mentioned.

2.5. Critical micelle concentration (CMC)

The CMC of the copolymers were determined by fluorescence spectrophotometer (F-7000, Hitachi High-Technologies Co. Japan) using pyrene as a probe. A series of blank copolymers with different concentrations (0.01–500 μg mL⁻¹) containing pyrene at the fixed concentration of 6.0 × 10⁻⁷ mol L⁻¹ were prepared. The excitation fluorescence spectra from 300 to 360 nm were tested with the emission wavelength fixed at 390 nm. The fluorescence absorbance values of I₃₃₈ and I₃₃₅ at excitation fluorescence spectra were recorded. The CMC value was the turning point in the logarithmic graph of the ratio of I₃₃₈ and I₃₃₅ to the samples' concentration.

2.6. Preparation of DOX loaded micelles

The PEG–PCL amphiphilic macromolecules (20 mg) and DOX (5 mg) were dissolved in DMF (2 mL) and stirred for 4 h, the solution was added dropwise into pure water (20 mL) under vigorous ultrasonic agitation for 30 min. The solution was dialyzed (Spectra/Por MWCO = 2000) against deionized water at 4 °C for 36 h after stirred overnight. The outer phase was replaced with fresh deionized water every 4 h. The mixture was centrifuged and lyophilized (3500 rpm, 5 min) to obtain the DOX loaded micelles. The concentration of DOX was tested using a UV-Vis measurement (excitation at 485 nm) (Fig. S1 and Tables S1 and S2 in the ESI†) with the calibration curve of DOX/DMF solutions. Drug loading content (DLC) and drug loading efficiency (DLE) were calculated according to the following formulas:

DLC (wt%) = (weight of DOX in micelles/weight of DOX loaded micelles) × 100%

DLE (wt%) = (weight of DOX in micelles/weight of DOX in feeding) × 100%

2.7. In vitro drug release test

The release of DOX loaded micelles was evaluated at 37 °C in buffer solution (pH 7.4 and pH 5.5) with 1 mL of dialysis solution (adjust the DOX concentration to 30 μg mL⁻¹) against 25 mL of the PBS medium under sink condition (the saturated solubility of DOX in PBS was determined to be about 20 μg mL⁻¹).²⁷ The 1 mL of fresh dialysis solution of DOX loaded micelles was put in dialysis membrane tubing (Spectra/Por MWCO = 1000). The tubings were immersed in vials containing 25 mL of PBS solution and put in a shaking bed at 37 °C with the shaking rate of 150 rpm. 1 mL of the buffer solution from the release medium was taken out at designated time intervals and the same volume of fresh medium was added into the vials. The amount of released DOX was detected by a fluorescence spectrometer with the excitation wavelength at 480 nm and emission wavelength at 550 nm (Fig. S3 and Table S3 in the ESI†). The release experiments were conducted in triplicate and the results were expressed as mean value with standard deviation (SD).

2.8. In vitro cytotoxicity test

The cytotoxicity of blank micelles was tested by CCK-8 assay against NIH 3T3 fibroblasts, C2C12 cells and HepG2 liver cancer cells. The cells mentioned above were inoculated into 96-well plates separately (4 × 10⁴ cells per mL) in 100 μL DMEM medium which was supplemented with 10% FBS and 1% penicillin–streptomycin. The cells were cultured at 37 °C in a humid atmosphere of 5% CO₂ for 24 h. After the medium was removed, four different concentrations of blank micelles in DMEM media were added. The cells were incubated for another 48 h. The culture medium was removed and the wells were washed with PBS (pH 7.4) for three times. Afterwards, the fresh medium containing 10 μL of CCK-8 (5 mg mL⁻¹) solution was added to each well. After incubated for another 2 h, the absorbance was measured at the wavelength of 450 nm.

2.9. In vitro anticancer activity

The anticancer activity of DOX loaded nanoparticles was evaluated with HepG2 cells. HepG2 cells were harvested and seeded in 96-well plates (4 × 10⁴ cells per mL) in 100 μL medium for 24 h. DOX·HCl and DOX loaded micelles with different concentrations was added to the medium-removed 96-well plates and incubated for 48 h. The cell viability was measured by CCK-8 assay.

2.10. Cellular uptake assay

Both flow cytometry and confocal laser scanning microscopy (CLSM) were used to examine the cellular uptake and the distribution of drug loaded nanoparticles in cells. To study the cellular uptake, HepG2 cells at a logarithm phase (1 × 10⁴ cells per mL) were seeded in glass dishes (diameter = 35 mm). DOX·HCl and DOX loaded micelles ([DOX] = 10 μg mL⁻¹) were added into the dishes for 1 and 4 h incubation, the medium was removed. The cells were washed by PBS for three times, followed by adding 200 μL fresh PBS into each dish. The cells were imaged by CLSM (TCP SP5, Leica, Germany). The fluorescence of DOX was excited at wavelength of 480 nm with emission at 590 nm.

For the flow cytometry test, HepG2 cells (1 × 10⁶ cells per well) were seeded in 6-well plates and incubated for 24 h. The cells were treated with DOX·HCl and DOX loaded micelles at the same concentration (10 μg mL⁻¹) for 1 and 4 h. After that, the cells were rinsed with PBS, trypsinized and resuspended in 1 mL PBS after concentration (1000 rpm, 5 min). The fluorescence intensity was measured (excitation 485 nm/emission 590 nm) on a BD FACS Calibur flow cytometer (Beckton Dickinson).

2.11. Statistical analysis

The Student's T-test was employed to assess the statistical significance and it was performed using GraphPad Prism 5 Software. The results were presented as means ± SD. Statistical significance was set at *p < 0.05, and extreme significance was set at **p < 0.01.

3. Results and discussion

3.1. Characterization of PEG–PCL copolymers with different topologies

As the structures of copolymers affect their micellar properties,^28–30 copolymers with different topologies, namely block, triblock and star shaped copolymers, were synthesized (as shown in Scheme 2) and used to study the effect of architecture on micellar properties and antitumor efficiency. Three different topological structures of poly(ethylene glycol)–poly(ε-caprolactone) with nearly identical molar ratio between PEG and PCL was synthesized. Those amphiphilic copolymers self-assembled into micelles with hydrophobic PCL segments as the core and hydrophilic PEG blocks as the shell.

Scheme 2 Synthetic routes of PEG–PCL copolymers with different structures.

The structures of those copolymers were characterized by ¹H NMR, and the spectra were presented in Fig. 1. First, poly(ethylene glycol) with expected terminal group was synthesized and the spectra (intermediate 1, 2, 3 and 4) were shown in Fig. S3–S5 in the ESI.† As the primary and secondary hydroxyl group in glycerol possessed the identical activity, the mixture products of intermediate 1 (I, II, III, IV) were obtained as shown in Fig. S3 in the ESI† and the corresponding proton signals of intermediate 1 were all detected in Fig. S3.† The mass spectrum in Fig. S4† further demonstrated the successful synthesis of intermediate 1 (+c ESI-MS, m/z: 181.01 [M + H], 202.98 [M + Na], calculated m/z: 180.08). The intermediate 2 was synthesized via an anionic ring-opening polymerization (AROP) of ethylene oxide using unprotected hydroxyl of intermediate 1 as the initiator and the ¹H NMR spectrum was shown in Fig. S5a.† After the terminal hydroxyl group of intermediate 2 was protected by benzyl, the double signals attributed to cyclobenzene (C₆H₅–) were observed at 7.3–7.8 ppm. Meanwhile, a single peak ascribed to benzyloxy (C₆H₅CH₂O–, 9) at 4.6 ppm appeared (Fig. S5b†). After the intermediate 3 was acidized, the signal at δ = 5.2 ppm assigned to C₆H₅CH(O₂)– (4) disappeared, whereas the peak ascribed to cyclobenzene was halved and the signal corresponding to proton of OH (14) appeared at δ = 3.8 (Fig. S5c†), indicating the intermediate 4 was successfully obtained.

Fig. 1 The ¹H NMR spectra (400 MHz, CDCl₃) of S-PEG2k–PCL4k (a), B-PEG2k–PCL4k (b) and L-PEG2k–PCL4k (c) copolymers.

After the macroinitiators (poly(ethylene glycol) with expected terminal groups) were obtained, the designed copolymers with different spatial structures were synthesized by ring opening polymerization. The ¹H NMR spectra in Fig. 1 clearly confirmed the successful synthesis of di-, tri- and star shape copolymers. The typical peaks attributed to PCL and PEG segments were all observed in Fig. 1a–c, which was accordant with previous reports.^26,31 The characteristic chemical shifts at δ = 3.6 to 3.7 ppm were assigned to PEG segment. The signals of 3, 4, 5, 6 attributed to PCL blocks at the chemical shifts δ = 2.3, 1.6, 1.4, 4.1 ppm were all observed in Fig. 1. The molecular weight of PCL blocks were calculated from the integral area ratio of characteristic peaks between PEG (multiple, –CH₂CH₂O–, δ = 3.6–3.7) and PCL (triplet, –CH₂OOC–, δ = 4.1) by using the known molecular weight of PEG. As shown in Table 1, the compositions of copolymer and molecular weights determined by ¹H NMR spectra were very close to that determined by the ratio of monomer to the initiator, suggesting that the molecular weight and composition of those copolymers can be easily controlled by feeding dose.

Table 1 The molecular weights and thermal properties of PEG–PCL copolymers with different structures

Sample	Molecular weight			Thermal properties
	^1H NMR	GPC		PEG segment		PCL segment
	Mn	Mn	Mw/Mn	ΔH (J g^−1)	Tm (°C)	ΔH (J g^−1)	Tm (°C)
L-PEG2k–PCL4k	5870	8188	1.69	17.01	46.79	42.29	51.18, 55.38
B-PEG2k–PCL4k	5942	9074	1.41	9.65	37.83	41.38	49.84, 54.06
S-PEG2k–PCL4k	6087	13 915	1.66	14.82	45.12	37.14	54.20

---

FTIR was also employed to characterize the structure of copolymers. As shown in Fig. 2, the absorption bands at 2866 cm⁻¹ belonged to the characteristic C–H stretching vibration of the repeated –OCH₂CH₂ units of PEG backbone. The peaks at 2943 cm⁻¹ could be attributed to the C–H absorption of ester bonds in the repeated units of PCL. More importantly, the characteristic stretching vibration absorbance of –C=O in ester bond at around 1725 cm⁻¹ and –CH₂OCH₂ at 1105 to 1243 cm⁻¹ were all detected, indicating the successful synthesis of the designed copolymers. It is known that FTIR can not only be used to characterize the functional group of the copolymer qualitatively, it can also be applied to confirm the composition ratio of the compound. Therefore, as shown in Fig. 2, the corrected areas of C–H stretching vibrations of ether bond in PEG at 2886 cm⁻¹ (S2) and in PCL at 2943 cm⁻¹ (S1) were calculated. The area ratio of S1 to S2 is 2.03 for S-PEG2k–PCL4k, 1.83 for B-PEG2k–PCL4k, 1.91 for L-PEG2k–PCL4k, which corresponded to the theoretical value 1.95. That further confirmed the designed copolymers were successfully obtained.

Fig. 2 FTIR spectra of L-PEG2k–PCL4k, B-PEG2k–PCL4k and S-PEG2k–PCL4k copolymers.

The molecular weight and molecular weight distribution of the synthetic copolymers were monitored by Gel Permeation Chromotography (GPC) as shown in Table 1 and Fig. S6.† A unimodal peak with high molecular weight of copolymers was detected in the GPC spectra, indicating no or negligible PEG or PCL oligomers existed. Nevertheless, unlike the ¹H NMR, the molecular weights tested by GPC were all larger than the designed molecular weights, especially for star-copolymers. That was because the molecular weight measured by GPC was a relative molecular weight, and the star shaped copolymer has a bigger steric hindrance than linear or block copolymers due to GPC adopt size-exclusion method. Combined with ¹H NMR, FTIR and GPC results, the copolymers with different spatial structures were successful synthesized.

3.2. Crystallization of PEG–PCL copolymers with different topologies

The effect of topologies of copolymers on their crystallization was studied by XRD, FTIR and DSC and the results were shown in Fig. 3 and Table 1. In the spectra of XRD, strong peaks at 2θ = 21.7° and 24° attributed to PCL blocks and peaks at 2θ = 19.4° and 23.7° assigned to PEG blocks were clearly observed in Fig. 3A, which was in good accordance with other studies reported.³² However, there were some differences in the intensity of those peaks. When the PEG constituent was in the central block (B-PEG–PCL), the crystallization of PEG in B-PEG2k–PCL4k copolymer was strongly restrained, behaving as a weak diffraction peak at 19.4°, while two separate clear crystal peaks attributed to PCL and PEG segments were observed in the copolymers of L-PEG2k–PCL4k and S-PEG2k–PCL4k. Moreover, the linear copolymer L-PEG2k–PCL4k showed the highest intensity for the peak at 19.4°, possibly because of the weaker restriction in PCL segments compared to the other two copolymers. The relative crystallinity of PCL segments was calculated by the area under the peak at 21.7°, and the results showed that the crystallization of those copolymers decreased in the sequence of L-PEG2k–PCL4k, B-PEG2k–PCL4k, S-PEG2k–PCL4k.

Fig. 3 The XRD spectra (A), enlarged segment of C=O stretching vibration in FTIR spectrum (B) and second heating curve of DSC (C).

It is well known that there are two peaks for the C=O vibration of semi-crystalline PCL, one at 1726 cm⁻¹ and the other at 1736 cm⁻¹, corresponding to the crystalline and amorphous PCL phase,^33,34 respectively. As shown in Fig. 3B, the absorption peak at 1726 cm⁻¹ attributed to the crystalline PCL phase was the strongest in linear copolymer (L-PEG2k–PCL4k), yet the weakest was observed in star copolymer (S-PEG2k–PCL4k). The results were in good agreement with the XRD results. DSC results in Fig. 3C and Table 1 exhibited that the highest crystallinity of both PEG and PCL segments was observed in L-PEG–PCL copolymers due to the weakest spatial hindrance in the linear copolymer.¹⁷ In addition, PCL segments in linear L-PEG–PCL were located and combined in one end, leading to a longer PCL block than the other two copolymers. Therefore, PCL blocks are more prone to be stacked tightly, thus resulted in higher crystallinity.^35,36

3.3. Characterization of polymeric micelles

Critical micellar concentration (CMC) is an effective parameter to show the stability of micelles and a low critical value is desired. In this study, micelle formation was monitored by using pyrene as a hydrophobic probe. From the plot of fluorescence intensity ratio of I₃₃₈/I₃₃₅ versus log C of the copolymer, the CMC values were obtained, which were compiled in Table 4 and Fig. S7.† The low CMCs (10⁻³ g L⁻¹) in the table suggested that the micelles were highly thermodynamically stable and could be potentially used as candidates for drug delivery. Among them, star-shaped copolymer micelles exhibited a little lower CMC, indicating star-shaped copolymers can easily form micelles, which was in accordance with other reports that the star shape copolymers facilitated the micellization as the unimer state of a star shape copolymer with many arms resembled the micellar state, which resulted in easier formation of micelles.³⁷

In order to achieve the EPR effect and longevity during systemic circulation, the size of micelles should be appropriate for permeating through the microvasculature of solid tumors and evading detection by the reticulo-endothelial system.^38,39 The morphology, size and size distribution of the micelles with or without DOX loading were determined by AFM and DLS.^40,41 The results were shown in Fig. 4 and Table 2. All three types of copolymers can self-assemble into mono-dispersed nanoparticles with sizes below or around 100 nm and spherical shapes. The sizes obtained from AFM were larger due to the aggregation of NPs in dry state and the AFM tip broadening effect. The mean sizes of L-PEG2k–PCL4k, B-PEG2k–PCL4k, S-PEG2k–PCL4k were 43, 74 and 53 nm, respectively. The size of polymeric micelles self-assembled by L-PEG–PCL copolymer was the smallest among those three polymeric micelles possibly because of high crystallinity and the topological structure. In the process of micellar formation, the PEG chains stayed in the outer phase to form the shell and the PCL chains stacked together to form the inner core. As elaborated in Fig. 5, due to linear structure of L-PEG–PCL, the PCL segment will be intertwined tightly, exhibit the smallest space volume of inner core, thus resulted in the smallest micellar size. While in the formation of B-micelles, the hydrophilic PEG chain in the middle must be folded to make sure the hydrophobic PCL chain flock together, then forming an angle in the two PCL segments. Thus a bigger cavity was produced and the size of B-PEG2k–PCL4k was the largest. Also, the highest drug loading content for B-micelles were further provided the evidence. In the S-PEG2k–PCL4k micelle, although an angle was also existed in the two PCL segments, the PEG segment was unfolded, led to less repulsive force occurs thus resulted in smaller cavity than B-PEG2k–PCL4k micelles. After drug loaded, the size of copolymer micelles measured by DLS were increased to 55, 115, 73 nm for L-, B- and S-shape copolymer micelles, respectively, which was similar to the research previously reported.^42,43 Moreover, the size increase values of DOX-loaded B-PEG–PCL micelles became bigger compared to other two copolymer micelles, possibly due to that the two PCL chains were separately connected on both ends of PEG, which increased the chain flexibility. The increased flexibility promoted the hydrophobic core became looser during the micellization, thus resulted in larger hydrophobic cores. This hypothesis was supported by crystallinity results as shown in Table 1. However, although the S-shaped copolymer exhibited the lowest crystallinity, the micellar size was smaller than B-shaped copolymer micelles. This was because the two hydrophobic PCL segments were both connected in one point, resulting in the restricted movement of PCL segments. The results above revealed that both of the crystallinity and structure of copolymers had impacts on the micellar size. Generally speaking, larger hydrophobic core can afford encapsulation of more hydrophobic drugs, behaving as a higher drug loading content. This rule was fitted for the DLC results as shown in Table 4. B-PEG–PCL copolymer micelles possessed the highest drug loading content and drug loading efficiency, whereas the linear copolymer micelles (L-PEG2k–PCL4k) exhibited the lowest DLC because of the highest crystallinity of PCL segment as shown in Table 1. It was reported that the drug was likely to be encapsulated into the amorphous PCL phase.⁴⁴

Fig. 4 Size tested by DLS and the AFM images of NPs with or without DOX loading. (A) blank NPs and (B) DOX loaded NPs measured by DLS. (C–E) AFM images of blank NPs (1) and DOX loaded NPs (2) of L- (C), B- (D) and S-PEG2k–PCL4k (E) copolymers.

Table 2 The size of blank and DOX loaded copolymer micelles

Sample	Blank NPs^a		DOX loaded NPs^a
	Number mean (nm)	PDI	Number mean (nm)	PDI
a Measured by DLS. The results were expressed as mean ± SD (n = 3).
L-PEG2k–PCL4k	43 ± 6	0.345 ± 0.049	55 ± 7	0.284 ± 0.008
B-PEG2k–PCL4k	74 ± 12	0.331 ± 0.054	115 ± 20	0.315 ± 0.02
S-PEG2k–PCL4k	53 ± 11	0.258 ± 0.036	73 ± 1	0.235 ± 0.024

---

Fig. 5 The molecular stack and micellar formation process self-assembled by different topologies copolymers.

In addition, we tested the size and zeta potential of the DOX loaded micelles in DMEM to evaluate the micellar behavior in the cell culture medium (Table 3). Compared to fresh micellar solution, the micelles in the cell culture medium maintains their original mean size while the zeta potential all decreased. As we know, the DMEM High Glucose cell media contains glucose, glutamine, sodium pyruvate and sodium bicarbonate. That is to say, the ionic concentration in the DMEM is higher than the reverse-osmosis water. As the literature pointed out⁴⁵ that the higher of the ions in the solution resulted in a thinner electrical double layer, exhibiting as lower zeta potential. More importantly, although the zeta potential of those micelles decreased, the decrease degree of zeta potential were in same tendency. Moreover, zeta potential of those micelles were all negative both in water or cell media. That meant there is little effect on their further used in the cell media.

Table 3 The size and zeta potential of DOX loaded micelles in DMEM and their zeta potential in water^a

Sample	Number mean (nm)	PDI	Zeta potential (mV)
			In DMEM	In water
a The results were expressed as mean ± SD (n = 3).
DOX/L-PEG2k–PCL4k	56 ± 12	0.390 ± 0.010	−2.72 ± 0.75	−10.20 ± 0.30
DOX/B-PEG2k–PCL4k	125 ± 7	0.200 ± 0.031	−0.98 ± 0.13	−7.44 ± 0.08
DOX/S-PEG2k–PCL4k	75 ± 4	0.378 ± 0.015	−1.22 ± 0.20	−10.83 ± 0.79

---

3.4. In vitro drug release profiles

The in vitro drug release profiles of drug loaded copolymer micelles were investigated in different pH media (pH 7.4 and 5.5) and the results were shown in Fig. 6. The release profiles included two components: a rapid initial burst release and a slower release phase. The rapid initial release might be attributed to the drug located in the cores closely to the interface between the core and shell of micelles as well as the large surface areas because of their minimal diameter. After the initial burst release, DOX was slowly released in the following time. Among them, star-shaped copolymer micelles showed faster drug release than the other two copolymer micelles systems in pH 7.4 medium. The difference in release rate among those copolymer micelles could be attributed to their crystallinity (Table 1 and Fig. 3). L- and B-PEG–PCL copolymer could form highly packed crystal in micellar cores, preventing the drug from being dispersed homogeneously inside the core, whereas S-PEG–PCL copolymer possessed the lowest crystallinity, which resulted in drug distributed homogenously inside the core, leading to the formation of more hydrophilic channels.⁴⁶ Those hydrophilic channels can prompt the drug release from the core of micelles. However, when the released media was in acidic environment (pH 5.5), the drug release rates were all improved. That was because the acidic environment could protonate the amino group of DOX and transformed the drug from hydrophobic to hydrophilic, thus made the trapped drug rapidly diffused out from the core of micelles, which is desirable for enhancing the cancer killing effects of DOX. Moreover, as protonated amino group of DOX override the effect of structure of copolymers on drug release, as shown in Fig. 6B, the released rates of three copolymer micelles were comparable.

Fig. 6 The release profiles of drug loaded nanoparticles in PBS media of pH 7.4 (A) and pH 5.5 (B). The results were expressed as mean ± SD (n = 3). Statistical significance was set at *p < 0.05, and extreme significance was set at **p < 0.01 in red (B- versus S-) and in black (L- versus S-).

3.5. In vitro cytotoxicity of copolymer micelles

For the purpose of evaluating the cytotoxicity, three blank micelles were incubated with 3T3 fibroblasts, C2C12 mouse myoblast cells and HepG2 cancer cells for 48 h. Fig. 7 showed that the cell viabilities of three cell lines were higher than 90% and some of them over 100% even the micelles concentration was as high as 600 μg mL⁻¹. It implied that the micelles were non-cytotoxic to normal and cancer cells.

Fig. 7 The cytotoxicity of blank micelles incubated with 3T3 cells (A), C2C12 cells (B) and HepG2 cells (C) and (D) DOX·HCl and DOX loaded micelles against HepG2 cells for 48 h. The results were expressed as mean ± SD (n = 4).

After drug was loaded, the cell viability of DOX loaded micelles is significantly lower than that of blank micelles after treated with HepG2 for 48 h via CCK-8 assay (Fig. 7D), indicating higher cytotoxicity was obtained by encapsulating DOX into micelles for improvement of its solubility. Moreover, both DOX·HCl and DOX loaded micelles exhibited a dose-dependent behavior. For three drug loaded micelles, cell viability decreased dramatically with the DOX concentrations increasing. As shown in Table 4, the IC₅₀ values of free DOX·HCl, DOX loaded L-PEG2k–PCL4k, B-PEG2k–PCL4k, S-PEG2k–PCL4k micelles were 0.047, 1.46, 1.78, 2.03 μg mL⁻¹, respectively. This phenomenon appeared to correspond well to the cellular uptake efficiency as follow discussed. Obviously, DOX·HCl expressed the lowest IC₅₀ as DOX·HCl was passive diffused into HepG2 cells and it immediately accumulated in the nuclei,⁴⁷ and then exerted its function. Although the DOX·HCl exhibited better anticancer efficiency, the unexpected systemic toxicity and inevitable side effects were always induced due to DOX·HCl diffused into cells without selectivity, which restricted their clinical application.^48,49 But, the encapsulation of DOX into polymeric micelles can avoid side effects. The formed nanoscale particles could accumulate in tumor tissues via the enhanced permeation and retention (EPR) effect owing to the appropriate size, which reduced the poisonous side effect and meanwhile promoted the utilization ratio of the drugs.^50–52 As we expected, the DOX loaded L-PEG–PCL micelles exhibited the highest anticancer efficiency among three DOX loaded micelles. It was because the smallest particle sizes which could foam cells more easily and advance the cellular uptake efficiency, it was also confirmed by CLSM and FCM results and accordingly promote the inhibition effect of DOX on the proliferation of cancer cells significantly. However, the poorest anticancer efficiency was emerged in DOX loaded S-PEG–PCL micelles. Combing the analysis of size, release profile and DLC results, it was not difficult to explain this phenomenon. The DOX loaded S-PEG–PCL micelles exhibited faster release profiles at pH 7.4 and the size and DLC were not much preferable, thus resulting in a less dosage of DOX was in the target site when arriving at the tumor cells.

Table 4 The CMC, drug loading capability and IC₅₀ of copolymer micelles

Sample	CMC^a (μg mL^−1)	DLC (%)	DLE (%)	IC50^b (μg mL^−1)
a The critical micelle concentration was measured by pyrene.b The half maximal inhibitory concentration.
L-PEG2k–PCL4k	0.833	5.4	23.6	1.46
B-PEG2k–PCL4k	0.821	6.0	26.5	1.78
S-PEG2k–PCL4k	0.801	5.6	24.7	2.03

---

3.6. Cellular internalization of polymeric micelles

As the cellular internalization is the key step for drug loaded micelles to fulfill their anticancer activity, the effect of copolymer micelles self-assembled by copolymers with different topologies on the internalization efficiency was compared. Confocal laser microscopy and flow cytometry⁵³ were used to study the cellular uptake of those copolymer micelles and the results were shown in Fig. 8. From the images of CLSM, we can see that the cells treated with DOX·HCl showed the highest red fluorescence and most red florescence was located in the nuclei even the incubation time was 1 h. The red fluorescence became stronger and stronger as the time progressed (4 h). The law of increased fluorescence intensity was the same as FCM results. Unlike DOX·HCl, the red fluorescence of drug loaded micelles were detected in cytoplasm due to the different pathway of cellular internalization. The free DOX·HCl was internalized into cells by diffusion for its water-soluble property and the internalization of passive diffusion was much faster than that of endocytosis process. On the other hand, the free DOX·HCl would penetrate into the tumor cells quickly without any selectivity and sustained-release property, thus resulting in stronger red fluorescence in nuclei. The DOX loaded micelles was internalized into cells mainly via endocytosis process and thus localized in endocytic compartments, such as endosomes and lysosomes. With increasing incubation time, the red fluorescence intensity became stronger as shown in Fig. 8. When the incubation time was short (1 h), the red fluorescence intensity measured by CLSM and FCM showed no significant difference (Fig. 8I, III and V), while lower intracellular DOX fluorescence intensity was observed in the cells treated with DOX loaded S-PEG2k–PCL4k micelles at 4 h (Fig. 8II, IV and V), the results were in good agreement with IC₅₀ results as shown in Fig. 7D and Table 4. However, the internalization efficiency of DOX loaded L-PEG–PCL and B-PEG–PCL copolymer micelles was seemed to conflict with IC₅₀ results, where lower IC₅₀ of DOX loaded L-PEG–PCL micelles was observed compared to B-PEG–PCL micelles (Fig. 7D and Table 4). One possible explanation was that the incubation time for CLSM was only 4 h whereas that for IC₅₀ was up to 48 h, thus resulting in the internalization efficiency was not up to the best condition. Also, many studies reported that the fluorescence intensity increased with the increase of incubation time.^54,55 If the incubation time was up to 48 h, the internalization efficiency of L-PEG–PCL micelles was progressed increased than that of B-PEG–PCL micelles.

Fig. 8 The cellular uptake results against HepG2 cells treated with DOX/L-PEG2k–PCL4k (A), DOX/B-PEG2k–PCL4k (B), DOX/S-PEG2k–PCL4k (C), DOX·HCl (D). (I) and (II) were the confocal microscopy images captured for 1 h and 4 h, respectively. For each panel, the images from top to bottom (number 1, 2, 3) indicated cells in bright field, with DOX fluorescence, and overlays of the two channels. (III) and (IV) were the flow cytometry results of 1 h (III) and 4 h (IV) with the same concentration of DOX (10 μg mL⁻¹). (V) was the histogram of the mean fluorescence intensity and the results were expressed as mean ± SD. Statistical significance was set at *p < 0.05, and extreme significance was set at **p < 0.01.

4. Conclusions

In this study, three amphiphilic copolymers, L-PEG–PCL, B-PEG–PCL and S-PEG–PCL with different topologies were successfully synthesized. Those copolymers could self-assemble into micelles with low critical micellar concentrations. The micelles were in spherical shapes and nano-scale particle sizes (less than 100 nm) with narrow size distributions. The topologies of the copolymer were found to affect the crystallinity, particle size, encapsulation efficiency, and in vitro drug release characteristics. B-PEG–PCL copolymer micelles possessed larger particle size, higher DOX encapsulation efficiency, and slower drug release rate. In contrast, L-PEG–PCL copolymer micelles exhibited faster internalization and better anticancer efficiency. In summary, these findings can provide a new insight to rational design of amphiphilic copolymers for anticancer drug delivery system.

Acknowledgements

This work was funded by Natural Science Foundation of China (No. 51403137), Doctoral Fund of Ministry of Education of China (No. 20130181110038), Young teachers' scientific research foundation of Sichuan University (2014SCU11015).

Notes and references

1. I. Bala, S. Hariharan and M. R. Kumar, Critical Reviews™ in Therapeutic Drug Carrier Systems, 2004, p. 21.
2. Y. Dong and S. S. Feng, Biomaterials, 2004, 25, 2843–2849.
3. G. He, L. L. Ma, J. Pan and S. Venkatraman, Int. J. Pharm., 2007, 334, 48–55.
4. R. Singh and J. W. Lillard, Exp. Mol. Pathol., 2009, 86, 215–223.
5. D. Chen and M. Jiang, Acc. Chem. Res., 2005, 38, 494–502.
6. S. Zamani and S. Khoee, Polymer, 2012, 53, 5723–5736.
7. X. Gao, B. Wang, X. Wei, W. Rao, F. Ai, F. Zhao, K. Men, B. Yang, X. Liu and M. Huang, Int. J. Nanomed., 2013, 8, 971.
8. S. Zhou, X. Deng and H. Yang, Biomaterials, 2003, 24, 3563–3570.
9. L. Y. Qiu and Y. H. Bae, Pharm. Res., 2006, 23, 1–30.
10. V. P. Torchilin, J. Controlled Release, 2001, 73, 137–172.
11. S. Nakagawa, K. I. Kadena, T. Ishizone, S. Nojima, T. Shimizu, K. Yamaguchi and S. Nakahama, Macromolecules, 2012, 45, 1892–1900.
12. M. S. Rahman, M. Changez, J. W. Yoo, C. H. Lee, S. Samal and J. S. Lee, Macromolecules, 2008, 41, 7029–7032.
13. K. V. Butsele, C. A. Fustin, J. F. Gohy, R. Jérôme and C. Jérôme, Langmuir, 2008, 25, 107–111.
14. G. M. Soliman, R. Sharma, A. O. Choi, S. K. Varshney, F. M. Winnik, A. K. Kakkar and D. Maysinger, Biomaterials, 2010, 31, 8382–8392.
15. T. K. Georgiou, L. A. Phylactou and C. S. Patrickios, Biomacromolecules, 2006, 7, 3505–3512.
16. H. Yin, S. W. Kang and Y. H. Bae, Macromolecules, 2009, 42, 7456–7464.
17. B. Bogdanov, A. Vidts, A. Van Den Buicke, R. Verbeeck and E. Schacht, Polymer, 1998, 39, 1631–1636.
18. G. He, L. L. Ma, J. Pan and S. Venkatraman, Int. J. Pharm., 2007, 334, 48–55.
19. R. Francis, D. Taton, J. L. Logan, P. Masse, Y. Gnanou and R. S. Duran, Macromolecules, 2003, 36, 8253–8259.
20. H. Gao and K. Matyjaszewski, Prog. Polym. Sci., 2009, 34, 317–350.
21. S. Venkataraman, J. L. Hedrick, Z. Y. Ong, C. Yang, P. L. R. Ee, P. T. Hammond and Y. Y. Yang, Adv. Drug Delivery Rev., 2011, 63, 1228–1246.
22. J. Y. Yhee, S. Lee and K. Kim, Nanoscale, 2014, 6, 13383–13390.
23. J. Deutsch, A. Martin and H. Lieske, J. Catal., 2007, 245, 428–435.
24. H. Hsieh and R. P. Quirk, Anionic polymerization: principles and practical applications, CRC Press, 1996.
25. V. Rejsek, D. Sauvanier, C. Billouard, P. Desbois, A. Deffieux and S. Carlotti, Macromolecules, 2007, 40, 6510–6514.
26. M. J. Huang, M. L. Gou, Z. Y. Qian, M. Dai, X. Y. Li, M. Cao, K. Wang, J. Zhao, J. L. Yang and Y. Lu, J. Biomed. Mater. Res., Part A, 2008, 86, 979–986.
27. D. L. Feng and Y. Z. Du, Int. J. Nanomed., 2013, 8, 3141–3150.
28. K. Mortensen, J. Phys.: Condens. Matter, 1996, 8, A103.
29. S. Pispas, N. Hadjichristidis, I. Potemkin and A. Khokhlov, Macromolecules, 2000, 33, 1741–1746.
30. Y. X. Wang, J. L. Robertson, W. B. Spillman Jr and R. O. Claus, Pharm. Res., 2004, 21, 1362–1373.
31. I. G. Shin, S. Y. Kim, Y. M. Lee, C. S. Cho and Y. K. Sung, J. Controlled Release, 1998, 51, 1–11.
32. J. Guan, M. S. Sacks, E. J. Beckman and W. R. Wagner, Biomaterials, 2004, 25, 85–96.
33. X. Shuai, F. E. Porbeni, M. Wei, T. Bullions and A. E. Tonelli, Macromolecules, 2002, 35, 2401–2405.
34. X. Shuai, F. E. Porbeni, M. Wei, I. D. Shin and A. E. Tonelli, Macromolecules, 2001, 34, 7355–7361.
35. L. Glavas, P. Olsén, K. Odelius and A. C. Albertsson, Biomacromolecules, 2013, 14, 4150–4156.
36. W. Zhu, W. Xie, X. Tong and Z. Shen, Eur. Polym. J., 2007, 43, 3522–3530.
37. K. H. Kim, G. H. Cui, H. J. Lim, J. Huh, C. H. Ahn and W. H. Jo, Macromol. Chem. Phys., 2004, 205, 1684–1692.
38. A. Balducci, Y. Wen, Y. Zhang, B. M. Helfer, T. K. Hitchens, W. S. Meng, A. K. Wesa and J. M. Janjic, OncoImmunology, 2013, 2, e23034.
39. Y. Wen and W. S. Meng, J. Pharm. Sci. Innovation, 2014, 9, 158–173.
40. C. Engman, Y. Wen, W. S. Meng, R. Bottino, M. Trucco and N. Giannoukakis, Clin. Immunol., 2015, 160, 103–123.
41. Y. Wen, S. L. Roudebush, G. A. Buckholtz, T. R. Goehring, N. Giannoukakis, E. S. Gawalt and W. S. Meng, Biomaterials, 2014, 35, 5196–5205.
42. Y. Lei, Y. Lai, Y. Li, S. Li, G. Cheng, D. Li, H. Li, B. He and Z. Gu, Int. J. Pharm., 2013, 453, 579–586.
43. Y. Liang, X. Peng, Y. Chen, X. Deng, W. Gao, J. Cao, J. Chang, Z. Gu and B. He, RSC Adv., 2015, 5, 59014–59021.
44. X. Shuai, H. Ai, N. Nasongkla, S. Kim and J. Gao, J. Controlled Release, 2004, 98, 415–426.
45. N. Wang, C. Hsu, L. Zhu, S. Tseng and J.-P. Hsu, J. Colloid Interface Sci., 2013, 407, 22–28.
46. S. K. Agrawal, N. Sanabria, D. Long, J. M. Coburn, G. N. Tew and S. R. Bhatia, J. Controlled Release, 2006, 112, 64–71.
47. H. S. Yoo, J. E. Oh, K. H. Lee and T. G. Park, Pharm. Res., 1999, 16, 1114–1118.
48. T. Nakanishi, S. Fukushima, K. Okamoto, M. Suzuki, Y. Matsumura, M. Yokoyama, T. Okano, Y. Sakurai and K. Kataoka, J. Controlled Release, 2001, 74, 295–302.
49. M. K. Yu, Y. Y. Jeong, J. Park, S. Park, J. W. Kim, J. J. Min, K. Kim and S. Jon, Angew. Chem., Int. Ed., 2008, 47, 5362–5365.
50. H. Maeda, J. Controlled Release, 2012, 164, 138–144.
51. M. Li, Z. Tang, S. Lv, W. Song, H. Hong, X. Jing, Y. Zhang and X. Chen, Biomaterials, 2014, 35, 3851–3864.
52. M. E. Davis and D. M. Shin, Nat. Rev. Drug Discovery, 2008, 7, 771–782.
53. Y. Wen, W. Liu, C. Bagia, S. Zhang, M. Bai, J. M. Janjic, N. Giannoukakis, E. S. Gawalt and W. S. Meng, Acta Biomater., 2014, 10, 4759–4767.
54. T. Su, X. Peng, J. Cao, J. Chang, R. Liu, Z. Gu and B. He, RSC Adv., 2015, 5, 13157–13165.
55. H. Yuan, J. Miao, Y.-Z. Du, J. You, F. Q. Hu and S. Zeng, Int. J. Pharm., 2008, 348, 137–145.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra06040d

---