Emulsion fabrication of magnetic mesoporous carbonated hydroxyapatite microspheres for treatment of bone infection

Abstract

Hydroxyapatite is widely used for bone filling materials because of its compositional similarities to bone mineral and excellent biocompatibility, but it does not possess good osteoinductivity and bactericidal properties. Herein, magnetic mesoporous carbonated hydroxyapatite microspheres (MEHMs) have been fabricated using CaCO₃/Fe₃O₄ microspheres as sacrificial templates. The cetyltrimethylammonium bromide/Na₂HPO₄ solution/cyclohexane/n-butanol emulsion system serves as a microreactor, in which the CaCO₃/Fe₃O₄ microspheres are converted to MEHMs via a dissolution–precipitation reaction. MEHMs with low crystallinity exhibit the hierarchical nanostructures constructed by nanoplates as building blocks with mesopores and macropores, which give them high drug loading–release properties. The controlled release of gentamicin significantly minimizes bacterial adhesion and prevents biofilm formation against S. epidermidis. The Fe₃O₄ nanoparticles are dispersed within the microspheres, which makes the MEHMs magnetic. The biocompatibility and osteoinductivity of the MEHMs have been investigated using human bone marrow stromal cells (hBMSCs) as cell models. The Fe₃O₄ nanoparticles in the MEHMs not only stimulate cell adhesion and proliferation, but also promote the osteogenic differentiation of hBMSCs. The excellent biocompatibility, osteoinductivity, drug loading–release properties and bactericidal properties give the MEHMs great potential as bone filling materials to treat bone infection.

---

1. Introduction

The reconstruction of bone defects in patients suffering from arthritis, osteoporosis, osteonecrosis, bone fracture, bone tumors and trauma are still complicated challenges in the field of orthopaedic surgery.¹ Many biomaterials, such as calcium phosphates, bioactive glasses and biopolymers, have been fabricated for bone repair and regeneration.^2–4 In clinical practice, the incidence of bone infection remains a serious unresolved problem.^5–7 When a bone becomes infected, the soft, inner part (bone marrow) often swells, resulting in the death of parts of the bone.⁷ Based on the above problems, ideal bone filling materials should not only possess excellent biocompatibility, bioactivity, osteoconductivity and osteoinductivity, but also should have other functions such as drug delivery properties and bactericidal properties.^8,9

Hydroxyapatite shows compositional similarities to bone mineral,^10,11 and the corresponding synthetic materials have been widely used for bone tissue engineering because of their excellent biocompatibility and osteoconduction.^12,13 Stoichiometric hydroxyapatite is known to be the most stable form of calcium phosphate phases in aqueous solution, resulting in low biodegradability and bioactivity.¹⁴ In order to address the drawbacks, several strategies have been developed, including lowering the crystallinity of hydroxyapatite and introducing carbonate ions into the crystal lattice.¹⁵ Although carbonated hydroxyapatite (CHA) possesses good osteoconduction, bioactivity and biocompatibility, it does not have osteoinductivity.¹⁶ Interestingly, a magnetic field may stimulate the proliferation and differentiation of osteoblasts, promote the expression of growth factors, and accelerate bone healing for a long time.^17,18 It is reasonable to speculate that the incorporation of magnetic particles into bone filling materials or scaffolds can improve the osteoinductivity of biomaterials. Among the magnetic materials, Fe₃O₄ nanoparticles have been extensively exploited for biomedical applications, including drug delivery, contrast agents for magnetic resonance imaging and magnetic hyperthermia, because of their excellent biocompatibility and unique magnetism.²¹ Hence, Fe₃O₄/hydroxyapatite composite materials have attracted much interest.^19–21

Post-surgery implant-associated infection remains a threat in the field of orthopaedic surgery,^5,6 and can be addressed by local delivery of antibiotics.²² Antibiotics are administered to defect regions and delivered locally in a controlled manner over time.²³ However, the conventional CHA particles have limited drug loading capacities (DLCs) because of their low surface area. This challenge is overcome by the fabrication of mesoporous hydroxyapatite.^10–13 The mesoporous structure and large surface area make it possible to incorporate high dosages of drugs into the mesopores and release them at a controlled rate.^24–26 Up to now, mesoporous hydroxyapatite particles have been fabricated by a templating method.^27,28 However, the monodisperse magnetic mesoporous carbonated hydroxyapatite microspheres (MEHMs) still present a fundamental challenge, because CHA has a complicated hexagonal crystal structure in the space group P6₃/m.

Herein, MEHMs with low crystallinity were successfully produced by the emulsion method, using CaCO₃/Fe₃O₄ microspheres as sacrificial templates. MEHMs possess mesoporous structure, magnetic properties, biocompatibility, osteoinductivity and drug delivery properties, so they can be used as drug delivery systems for the treatment of bone infection. The main aims are to fabricate multifunctional MEHMs, and investigate their biocompatibility, osteoinductivity, drug delivery properties and bactericidal properties.

2. Experimental

2.1. Preparation of MEHMs

3.73 g ferrous chloride tetrahydrate (FeCl₂·4H₂O) and 8.15 g ferric trichloride hexahydrate (FeCl₃·6H₂O) was dissolved in 100 mL deionized water at room temperature. With magnetic stirring, a 0.50 mol L⁻¹ ammonia solution was added dropwise to the mixed solution until the pH value was approximately 9.0, yielding a black suspension. The suspension was then stirred at 60 °C for 45 min. The Fe₃O₄ nanoparticles were washed with deionized water, collected by magnetic separation, and dried in an oven at 60 °C.

The nacre of Corbicula fluminea was collected from the Zhejiang province in China. The shell of C. fluminea was cleaned of macroscopic impurities in tap water using a brush, and the nacre was separated from the shell by shaving off the outer layers, including the periostracum and prismatic layer. The nacre was demineralized in a 1.0 mol L⁻¹ HCl solution overnight, and diluted with deionized water until the concentration of calcium ions was 0.25 mol L⁻¹. After centrifugation and separation from the HCl-insoluble material, the pH-value of the organic calcium chloride (CaCl₂) solution was adjusted up to seven by adding a few drops of 1.0 mol L⁻¹ NaOH. This yielded a calcium chloride solution, and the organic materials in the solution were termed HCl-soluble nacre materials.²³ The Fe₃O₄ nanoparticles (0.2 g) were added into the calcium chloride solution with HCl-soluble nacre materials, followed by ultrasonic treatment for 10 min. Subsequently, the rapid-mixing reaction was performed by pouring the Na₂CO₃ solution into 100 mL of the above mixed solution. The obtained CaCO₃/Fe₃O₄ microspheres were washed with deionized water and then dried in an oven at 60 °C.

7.29 g cetyltrimethylammonium bromide (CTAB) was dissolved into 40 mL cyclohexane and 5.48 mL n-butanol. After stirring, 2.15 g disodium hydrogen phosphate (Na₂HPO₄·12H₂O) in 7.2 mL deionized water was added into the above solution, followed by the addition of 0.60 g CaCO₃/Fe₃O₄ microspheres. The mixture was stirred at 50 °C for 24 h. The obtained MEHMs were washed with deionized water, and then dried in a convection oven at 60 °C for 48 h. Mesoporous carbonated hydroxyapatite microspheres (EHMs) were fabricated under the same conditions but without the addition of Fe₃O₄ nanoparticles. As is well known, hydroxyapatite is the most widely used bone filling material in orthopaedic clinical practice. Therefore, the conventional hydroxyapatite particles (HAPs) were used as control samples, and were purchased from Sinopharm Chemical Reagent Co.

2.2. Characterization

The microstructures of the samples were investigated using scanning electron microscopy (SEM, MX2600, CamScan) and transmission electron microscopy (TEM, CM200/FEG, Philips). N₂ adsorption–desorption isotherms were measured with an automatic surface area and porosity analyzer (AUTOSORB-1-C, Quantachrome) at −203.85 °C. The pore size distributions were derived from the adsorption branches of the isotherms using density functional theory (DFT). The crystalline phases of the samples were examined by X-ray powder diffraction (XRD, D/max-III C, Japan) using Cu-Kα radiation. The relative crystallinity of the samples was calculated using MDI JADE 5.0 software. The crystallinity of the samples was evaluated according to the formula: crystallinity = (X/Y) × 100%, where X is the net area of the diffracted peaks and Y is the net area of the diffracted peaks + background area.²⁹ Fourier transform infrared spectra (FTIR, VECTOR22, BRUKER) were collected at room temperature using the KBr pellet technique, working in the wavenumber range 4000–400 cm⁻¹ at a resolution of 2 cm⁻¹ (number of scans ∼ 60).

2.3. In vitro drug loading and release study

Gentamicin solutions at different concentrations (0.072 g L⁻¹, 0.144 g L⁻¹, 0.216 g L⁻¹) were prepared by the addition of gentamicin into deionized water. The MEHMs, EHMs and HAPs (1.8 g) were added to 25 mL of the above gentamicin solutions, respectively, and then were stirred at 37 °C for 24 h. After being centrifuged, the gentamicin-loaded carries were dried in a vacuum oven at 50 °C. The gentamicin-loaded MEHMs, gentamicin-loaded EHMs and gentamicin-loaded HAPs obtained by soaking carriers in the different concentrations of gentamicin (0.072 g L⁻¹, 0.144 g L⁻¹, 0.216 g L⁻¹) were represented as MEHMs01, MEHMs02, MEHMs03, EHMs01, EHMs02, EHMs03, HAPs01, HAPs02, and HAPs03, respectively. The loading efficiency of gentamicin was calculated according to the following equation:where η is the loading efficiency, C_o is the drug concentration in the solution before loading, and C_r is the drug concentration in the solution after loading.

For the in vitro release test, the gentamicin-loaded carriers were compacted into a disk (0.2 g, d = 10 mm) at a pressure of 4 MPa. Then the disks were immersed in 500 μL phosphate buffer saline (PBS) in a 24-well plate at room temperature with orbital shaking at 70 rpm. PBS is a water-based salt buffer solution, commonly used in biological and pharmacological research. PBS is widely used for gentamicin loading and release experiments in the literature. The gentamicin-release medium (200 μL) was extracted for determining the drug concentrations at given time intervals, and replaced with the same volume of fresh PBS. The drug concentrations in the gentamicin-release medium were analyzed using a colorimetric assay described elsewhere.³⁰

2.4. Bacterial culture and bactericidal effect assay

A bacterial cell line of biofilm-producing S. epidermidis (ATCC35984) was obtained in freeze-dried form from the American Type Culture Collection. The cells were propagated on an agar plate for 3 days before transferring them to Luria broth for seeding. Before bacterial seeding, bacteria was withdrawn from the plates using a sterile 10 μL loop and inoculated in a polystyrene tube with 3 mL Luria broth. The tube was agitated for approximately 16 h on a shaker at 250 rpm and 37 °C. Bacteria concentration was assessed via optical density. For this purpose, the Luria broth–bacteria solution was diluted in different ratios and the transmittance was measured using a spectrophotometer. According to McFarland standards, the concentration of bacteria solution with 30% transmittance is 900 million bacteria per mL. Further dilutions were performed until the Luria broth had a final concentration of 10 million bacteria per mL. After culturing the bacteria on the HAPs and MEHMs with or without loading gentamicin for 24 h, the samples were gently washed three times with PBS to remove non-adherent bacteria. The biofilms were fixed in 2.5% glutaraldehyde for 2 h at 4 °C, then washed three times with cacodylate buffer and dehydrated through a series of graded ethanol solutions (25, 50, 75, 95 and 100%). The samples were subsequently freeze-dried, sputter coated with gold, and observed using SEM (Joel JSM-6310LV, JEOL Ltd, Tokyo, Japan). Similarly, the discs were removed at 24 h and were gently washed three times with PBS. Subsequently, the discs were stained in a new 24-well plate with 300 μL combination dye (LIVE/DEAD Baclight bacteria viability kits, Molecular Probes, L13152) and were subsequently analyzed with a confocal laser scanning microscope (Leica TCS SP2; Leica Microsystems, Heidelberg, Germany). The viable and nonviable cells can be distinguished under a fluorescence microscope, because the viable bacteria with intact cell membranes appear fluorescent green, whereas nonviable bacteria with damaged membranes appear fluorescent red.

2.5. Proliferation and morphology of hBMSCs

The hBMSCs were isolated and expanded using a modification of standard methods, as described previously.^31,32 The study was approved by the Ethic Committee of the Ninth People's Hospital of Shanghai Jiao Tong University. Cells were grown in complete Alpha Minimum Essential Medium (α-MEM; GIBCO, Grand Island, NY, USA), supplemented with 10% fetal bovine serum (FBS; Hyclone, Tauranga, New Zealand) and antibiotics (penicillin 100 U mL⁻¹, streptomycin 100 μg mL⁻¹; Hyclone, Logan, UT, USA) in a 37 °C humidified atmosphere with 5% CO₂. Cells at a passage from P2 to P4 were used for these experiments.

The MTT assay was carried out according to the manufacturer's instructions (Sigma-Aldrich). Briefly, hBMSCs were seeded on the HAPs, EHMs and MEHMs at a density of 1 × 10⁴ cells per sample in a 24-well plate. At each time point, MTT solution was added to each well, and the plates were incubated for 3–4 h. Subsequently, dimethyl sulfoxide (DMSO, Sigma-Aldrich) was added to the wells for 5 min, and 200 μL solution of each well was added into a new 96-well plate. The optical density (OD) value was quantified by measuring the absorption at 570 nm using an automated plate reader (PerkinElmer).

The cytoskeleton of hBMSCs on the HAPs, EHMs and MEHMs was observed using double fluorescence staining. After culturing the cells for 24 h, the samples were gently washed with PBS and maintained in 4% paraformaldehyde for 15 min, followed by immersing in 0.1% Triton X-100 solution for 15 min. TRITC phalloidin was used to stain the actin filaments of cells shown as red fluorescent light, and 4′,6-diamidino-2-phenylindole (DAPI) was used to stain the nucleus of cells shown as blue fluorescent light. The cytoskeleton of hBMSCs was observed under laser scanning confocal microscopy (LSCM, LEICA TCS SP2) and fluorescence microscopy (FM).

2.6. Osteogenic differentiation of hBMSCs

hBMSCs were seeded on the HAPs, EHMs and MEHMs at a density of 1 × 10⁵ cells per sample in a 24-well plate. After 3 days culture in the basic culture medium, the cells were induced to osteogenic differentiation in the osteogenic medium (consisting of basic culture medium supplemented with 10 mM β-glycerophosphate(β-GP), 10⁻⁸ M dexamethasone and 50 μM L-ascorbic acid (all from Sigma)). The medium was changed every 3 days. At each time-point, the cells were separated from the samples with trypsin and transferred to a new 24-well plate. The ALP staining and ALP activity were tested after the cells were attached to the plate, therefore the result of the ALP activity assay could reflect the innate ALP activity of the hBMSCs. ALP staining was performed as described in the manufacturer's instructions (Rainbow, Shanghai, China). Meanwhile, ALP activity was determined at 405 nm using p-nitrophenylphosphate (pNPP) (Sigma-Aldrich) as the substrate. A 50 μL sample was mixed with 50 μL pNPP (1 mg mL⁻¹) in 1 M diethanolamine buffer containing 0.5 mM MgCl₂ at pH 9.8, and this was incubated at 37 °C for 15 min on a bench shaker. The reaction was completely inhibited by the addition of 25 μL of 3 N NaOH per 100 μL reaction mixture. The enzyme activity was quantified by absorbance measurements at 405 nm in an ELISA reader (TECAN Safire2TM, TECAN Group Ltd, Beijing, China).

3. Results and discussion

3.1. Phase structure and magnetism of MEHMs

The biological apatite has poor crystallinity, which is thought to arise from the incorporation of impurities, such as carbonate, sodium and magnesium ions.³³ In order to improve the biocompatibility, bioactivity and biodegradation of biomaterials, MEHMs with low crystallinity, nonstoichiometric composition, crystalline disorder and carbonate ions substituted in the apatite crystal lattice have been fabricated by the emulsion method. Generally, the conventional hydroxyapatite nanocrystals fabricated by the hydrothermal method, chemical precipitation and solid-state methods exhibit a rod-like structure with a high crystallinity, because they possess a hexagonal structure in the space group P6₃/m.³⁴ Herein, the HAPs purchased from Sinopharm Chemical Reagent Co., Ltd are used as the control. Fig. S1† indicates that the morphology of the HAPs is rod-like, and their length and diameter are about 160 nm and 30 nm, respectively. The XRD patterns of the MEHMs, EHMs and HAPs are shown in Fig. 1. They all exhibit the characteristic diffraction peaks, which are exclusively indexed to the structure of hydroxyapatite (JCPDS no. 35-0180). The broad peaks in the XRD patterns suggest that the MEHMs and EHMs possess low crystallinity; these are 82.2% and 78.4%, respectively. On the contrary, the XRD pattern of the HAPs shows sharp and narrow diffraction peaks, indicating the good crystallinity of 96.2%. Since the hydroxyapatite in the MEHMs possesses poor crystallinity, we can speculate reasonably that the MEHMs have good biodegradability and resorption ability. The in vitro degradability of the MEHMs has been carried out by soaking the microspheres in PBS. The Ca²⁺ ions are released from the microspheres after soaking in PBS, suggesting the excellent in vitro degradability of the MEHMs (Fig. S2†). The FTIR spectra of the MEHMs, EHMs and HAPs are revealed in Fig. 2. The intense absorption peak at 1030 cm⁻¹ is ascribed to the stretching vibration (v₃) of the phosphate (PO₄³⁻) groups, and the peaks at 563, 604 cm⁻¹ are ascribed to the bending vibration (v₄) of the phosphate (PO₄³⁻) groups.³⁵ The characteristic absorption band due to HPO₄²⁻ at around 1130 cm⁻¹ indicates that both the MEHMs and EHMs are calcium deficient apatite.³⁶ As compared with the HAPs, the MEHMs and EHMs have stronger adsorption peaks of B-type CO₃²⁻ substitution at 1420/1480 cm⁻¹ (v₃).³⁷

Fig. 1 XRD pattern of MEHMs, EHMs and HAPs.

Fig. 2 FTIR spectra of MEHMs, EHMs and HAPs.

The characteristic peaks due to Fe₃O₄ are not detected by the XRD pattern in Fig. 1 because of the low percentages and small particle size. The physical properties measurement system of Quantum Design indicates that the saturation magnetization strengths (Ms) of the Fe₃O₄ particles and MEHMs are 62.4 emu g⁻¹ and 4.03 emu g⁻¹, respectively. Based on the ratio of saturation magnetization strength between the MHMs and magnetite, the estimated percentage of Fe₃O₄ in MEHMs is ∼6.4%.³⁸

3.2. Morphology and mesoporous structure of MEHMs

The SEM and TEM images in Fig. 3a and 4a show that the MEHMs exhibit a spherical shape with a particle size of ∼5 μm. Interestingly, these multifunctional microspheres possess a flower-like structure, which is constructed using nanoplates as building blocks with a thickness of ∼40 nm and width of 100–400 nm. These nanoplates interweave together forming open macropores (Fig. 3c). The high-magnification TEM image in Fig. 4c indicates the light-shaded spots, suggesting that the nanoplates are composed of smaller plate-like nanocrystals. The particle size of the nanocrystals is 4–6 nm (Fig. 4c), which is in good agreement with the calculated results from the XRD pattern. According to the Scherrer's equation, the crystallite size of the MEHMs along the c-axis is about 4.5 nm. The aggregates of the nanocrystals result in the formation of the mesopores among them. The corresponding ED pattern shows visible diffraction rings due to the apatite structure (Fig. 4b, inset). In addition, Ca, P, C, O, Na and Fe are detected in the EDS spectrum, as shown in Fig. 3d. The Ca, P, C and O elements are ascribed to the CHA. The Na element may be ascribed to part substitution of the Ca²⁺ ions in the apatite crystal lattice or the adsorption Na⁺ ions from the Na₂HPO₄ solution. The Fe element is due to the presence of Fe₃O₄ nanoparticles within the microspheres.

Fig. 3 (a–c) SEM image and (d) EDS spectrum of MEHMs.

Fig. 4 (a–c) TEM image and (d) nitrogen adsorption–desorption isotherm of MEHMs. The inset in (b) shows the ED pattern, and the inset in (d) shows the DFT pore size distribution.

The nitrogen adsorption–desorption isotherm and corresponding DFT pore size distribution of the MEHMs are shown in Fig. 4d. According to the International Union of Pure and Applied Chemistry, the MEHMs are identified as a type H3 hysteresis loop derived from particle aggregates with slit-shaped pores (Fig. 4d). The mesopores among the nanocrystals are confirmed by the TEM image (Fig. 4c). The pore size is distributed around 3.97 nm, which is calculated from the isotherms using the DFT method (Fig. 4d, inset). The steep increase of nitrogen adsorption at P/P_o 0.9–1.0 suggests the presence of macropores, which is consistent with the SEM images (Fig. 3c). In addition, the non-magnetic EHMs also have a mesoporous structure, and the corresponding pore size is mainly distributed around 4.1 nm. The BET surface areas of the MEHMs and EHMs are 50.8 m² g⁻¹ and 55.5 m² g⁻¹, respectively. In addition, the pore volumes of the MEHMs and EHMs are 0.122 cm³ g⁻¹ and 0.151 cm³ g⁻¹, respectively. Since the HAPs do not possess a mesoporous structure, their BET surface area is only 8.1 m² g⁻¹.

3.3. Formation mechanism of MEHMs

CHA with low crystallinity exists widely in the biological apatite, which has greater solubility, biodegradability, and bioactivity than stoichiometric hydroxyapatite. In our previous work, magnetic mesoporous carbonated apatite microspheres were fabricated hydrothermally using the CaCO₃/Fe₃O₄ microspheres as sacrificial templates.²³ However, their crystallinities are high because the transformation reactions are carried out under hydrothermal conditions.²³ In order to decrease the crystallinity of CHA, we have developed the emulsion method to fabricate the MEHMs at low temperature. The CTAB/Na₂HPO₄ solution/cyclohexane/n-butanol emulsion system serves as microreactors to control the morphology of the MEHMs. For the emulsion system, CTAB is used as a surfactant, the cyclohexane as the oil phase, Na₂HPO₄ solution as the water phase, and n-butanol as the cosurfactant. After the CaCO₃/Fe₃O₄ microspheres are added into the emulsion system, these microspheres are covered with the CTAB with the hydrophobic group toward the oil phase and the hydrophilic groups toward the microspheres. As the CaCO₃/Fe₃O₄ microspheres collide with the emulsion drops, the Na₂HPO₄ solution in the emulsion drops is permeated on the microsphere surfaces. The formation of the MEHMs includes the following stages: (i) the dissolution reaction of CaCO₃ occurs, and the Ca²⁺ and CO₃²⁻ ions are released from the microspheres; (ii) the Ca²⁺ and CO₃²⁻ ions react with the PO₄³⁻, HPO₄²⁻ and OH⁻ ions to form the CHA. At the same time, the Fe₃O₄ nanoparticles are dispersed within the microspheres. Notably, the emulsion system plays an important role in the formation of the monodisperse microspheres. If the dissolution–precipitation reaction does not take place in the emulsion system, the morphologies of the products are irregular.

3.4. Drug loading–release property of MEHMs

The MEHMs have great potential as drug delivery systems because of the large specific surface and mesoporous structure. The drug loading–release property of the MEHMs has been investigated using the HAPs as the control and gentamicin as the drug model, as shown in Fig. 5. Gentamicin is widely used as an antibiotic to prevent implant-associated infection because of its advantages such as low cost, broad antibacterial spectrum of action, low rate of primarily resistant pathogens and good stability.³⁹ The MEHMs and HAPs are loaded with gentamicin by immersing the drug carriers in the gentamicin solutions of different concentrations. The loading efficiency is the ratio of the loaded amount of gentamicin to the initial amount of gentamicin before loading. Fig. 5a shows that the drug loading efficiency of the MEHMs ranges from 68.8% to 73.7% in the different gentamicin solutions, while that of the HAPs ranges from 15.2% to 26.6%. As compared with the HAPs, the higher drug loading efficiency of the MEHMs is mainly attributed to the hierarchical porous structures. The mesoporous structure increases the specific surface areas (Fig. 4d), so more drugs are adsorbed on the microspheres. At the same time, the macropores among the nanoplates not only decrease the diffusion limitations for drug delivery, but also provide space for drug loading. Moreover, since the BET surface area of the MEHMs is greater than that of the HAPs, the more OH and PO₄ units exist on the surfaces of the MEHMs than on the surfaces of the HAPs. These functional groups can provide these carriers with great affinity towards drug molecules.⁴⁰ Notably, the drug loading efficiency of the MEHMs does not decrease obviously with increasing concentration of gentamicin because the MEHMs possess high DLCs. In our previous work, the DLCs of the EHMs and MEHMs were determined by using vancomycin as a drug model; these are 1.95 wt% (1.95 g vancomycin is loaded per gram of carrier) and 1.91 wt%, respectively.⁴¹ However, the DLCs of the HAPs are only 0.053 wt% because of the low surface area and lack of porous structure.

Fig. 5 (a) Drug loading efficiencies of the MEHMs and HAPs in the gentamicin solutions with different drug concentrations. (b) The in vitro drug release profiles of gentamicin-loaded MEHMs and gentamicin-loaded HAPs. The data are represented as the mean ± standard deviation; n = 3.

For local antibiotic delivery systems, the active drug release time should be within a certain range of 3–5 days, in order to avoid development of side effects such as antibiotic resistance and renal toxicity. The in vitro drug release tests are carried out by immersing the gentamicin-loaded MEHMs and gentamicin-loaded HAPs in PBS at room temperature with orbital shaking at 70 rpm, as shown in Fig. 5b. The gentamicin-loaded MEHMs and gentamicin-loaded HAPs have obviously different in vitro drug release trends. The gentamicin-loaded HAPs, including HAPs01, HAPs02 and HAPs03, display a burst-release effect. More than 90% of the loaded drugs are released within the first 6 h, and then the release equilibrium is reached on further increasing the release time. Interestingly, the gentamicin-loaded MEHMs exhibit the slow and controlled release of gentamicin with effective drug release up to 6 days. At 3 h and 144 h, the values of the gentamicin release percentages are about 25% and 70%, respectively. The gentamicin-loaded MEHMs exhibit controlled drug release kinetics and the drug release phase is similar to the common length of prophylactic use of antibiotics (3–5 days) in current clinical practice.^42–44 The excellent drug release properties achieved in the MEHMs is attributed to the mesoporous and macroporous structures in the microspheres. The hierarchically porous structures support the gentamicin to penetrate the microspheres and release in a controlled manner. Fig. 5b indicates that MEHMs01, MEHMs02 and MEHMs03 have different cumulative release ratios of gentamicin, which may be related to the original concentration of the gentamicin solutions. The driving force to load gentamicin in the drug carriers is the concentration gradient. As the MEHMs are soaked in the gentamicin solutions, the drugs may enter into the microspheres more deeply with increasing drug concentration. The higher the concentration of the gentamicin solution, the more drugs penetrate in the MEHMs. During the drug release process, the drugs in the microspheres release more slowly than those on the microspheres, and the cumulative release ratio of gentamicin decreases in the following order: MEHMs01 > MEHMs02 > MEHMs03. Therefore, the MEHMs are more desirable as controlled antibiotic delivery systems than the conventional HAPs.

3.5. Bactericidal property of MEHMs

Implant-associated infections are one of the most serious complications in orthopaedic surgery,^45,46 and are mostly due to the biofilm mode of bacteria growth.^47,48 The formation of an infectious biofilm is initiated by transport and adhesion of bacteria to the surface of implants.^49,50 After that, adhering bacteria may overgrow on the surface, leading to the formation of a biofilm.⁵¹ The formation of the biofilm increases the resistance of the bacteria to antibiotics. Gentamicin is an aminoglycoside antibiotic and has been widely used in clinical practice. Gentamicin has a concentration-dependent antibacterial activity,⁵² suggesting that a high concentration of gentamicin at the bone–implant interface would be essential to eradicate bacterial infections. Although bone infection may be caused by different types of bacteria, gentamicin may exhibit antibacterial ability toward common pathogens of bone infection, as long as the eluting drug concentration is higher than the minimum inhibiting concentration (MIC), thanks to its broad antibacterial spectrum of action, low rate of primarily resistant pathogens and good stability. The MEHMs possess excellent drug loading–release properties because of their hierarchically porous structures (Fig. 4 and 5). In this work, MEHMs with different doses of gentamicin were used as drug delivery systems to investigate the bacteria adhesion and biofilm formation, and the HAPs loaded with the same drug dose served as the control. To eliminate the effect of burst release of drugs, the gentamicin-loaded carriers were immersed in 500 μL PBS in a 24-well plate at room temperature with orbital shaking at 70 rpm for 3 h and then rinsed twice with PBS prior to seeding of bacteria. After 24 h incubation at 37 °C, non-adherent bacteria were removed by gently washing the plates three times in PBS. Fig. 6 shows the CLSM images of bacteria colonies on the surfaces of the MEHMs and HAPs by using the dead/live Baclight bacteria viability kits. The viable cells with intact cell membranes stain fluorescent green, while nonviable cells with damaged membranes stain fluorescent red. A high intense level of green fluorescence on both the pure MEHMs and HAPs surfaces is detected, suggesting a high level of biofilm formation. After the MEHMs are loaded with gentamicin, the vast majority of bacteria are killed by the gentamicin released from the drug carrier, and fewer bacteria colonies are observed on the surface. It is noted that the bacterial adhesion decreases gradually with increasing amount of loaded gentamicin (Fig. 6). As compared with the gentamicin-loaded HAPs, the gentamicin-loaded MEHMs have much fewer and smaller bacteria colonies on the surfaces.

Fig. 6 Projected top views of biofilm formation of ATCC35984 on the surfaces of (a) HAPs, (b) HAPs01, (c) HAPs02, (d) HAPs03, (e) MEHMs, (f) MEHMs01, (g) MEHMs02, and (h) MEHMs03 by using CLSM after staining with the Baclight dead/live stain. Bacteria were stained with green fluorescent SYTO 9 and red fluorescent propidium iodide, which causes live cells to appear green and dead cells to appear red under CLSM. The projection colours are based on the average voxel intensity through the depth of the biofilm.

SEM images of bacteria adhering to the surfaces of the HAPs and MEHMs with or without gentamicin are shown in Fig. 7. There are many multiple bacterial colonies on both the surfaces of both the pure HAPs and MEHMs. Interestingly, only a few small, single bacterial colonies are observed on the gentamicin-loaded MEHM surfaces, and they decrease with increasing amount of loaded gentamicin. The results of the CLSM images and SEM images indicate that the gentamicin-loaded MEHMs are effective in minimizing bacterial adhesion and preventing biofilm formation of S. epidermis. The higher bactericidal properties of the gentamicin-loaded MEHMs compared to the gentamicin-loaded HAPs are attributed to their improved drug loading–release properties. The MEHMs possess hierarchically porous structures and large surface areas, so they have high drug loading efficiency and drug release properties (Fig. 5). The controlled release of gentamicin significantly minimizes bacterial adhesion and prevents biofilm formation against S. epidermidis (Fig. 6 and 7). However, only a small amount of gentamicin is loaded on the HAPs because of the non-porous structure. Moreover, the gentamicin adsorbed on the HAPs surfaces releases rapidly in the first 3 h before seeding of the bacteria (Fig. 5b). Based on the above reasons, only a few drugs are released from the gentamicin-loaded HAPs (Fig. 5b), and thus many bacterial colonies are observed on their surfaces (Fig. 6 and 7).

Fig. 7 SEM images of bacteria adhering to the surfaces of different biomaterials: (a) HAPs, (b) HAPs01, (c) HAPs02, (d) HAPs03, (e) MEHMs, (f) MEHMs01, (g) MEHMs02, and (h) MEHMs03.

3.6. Biocompatibility of MEHMs

Ideal bone filling materials for the treatment of bone infection should not only possess good drug loading–release properties and bactericidal properties, but also must have good biocompatibility with the surrounding cells to promote satisfactory osteointegration between the biomaterials and the bone tissue. Previous work has shown that the HAPs exhibit high bioactivity and biocompatibility because of the compositional similarities to the minerals of natural bones.^34,53 The MEHMs are different from the HAPs in the chemical components and porous structure. In the present work, the effects of the hierarchical porous structures and Fe₃O₄ nanoparticles on the biocompatibility of the MEHMs have been investigated by using the HAPs and EHMs as the control groups. The in vitro cytotoxicities of the HAPs, EHMs and MEHMs are assessed by MTT assay using the hBMSCs as cell models, as shown in Fig. 8. The number of viable cells on the HAPs is noticeably lower than that on both the EHMs and MEHMs after culturing the hBMSCs for 1 day and 2 days (P < 0.05). The initial cell adhesion and spreading are accelerated significantly on the EHMs and MEHMs, which is due to the hierarchically porous structures and nanostructures. The mesopores within the hydroxyapatite plates and macropores among the plates increase the surface areas of the MEHMs and EHMs, which may store nutrients for cells and improve the initial cell adhesion and spreading. This result is in good agreement with our previous reports. More hBMSCs are detected on the mesoporous bioactive glass than those on the nonporous bioactive glass after culturing the cells for 1 day and 2 days.⁵⁴ Moreover, the initial cell adhesion and spreading may be related to the hierarchical nanostructures of the MEHMs and EHMs, because the topographical cues at the nanoscale level increase the roughness.^55,56 Interestingly, Fig. 8 shows that the viable cell number on the MEHMs is highest among the three groups after culturing the hBMSCs for 4–7 days. The accelerated proliferation of hBMSCs is attributed to the Fe₃O₄ nanoparticles in the MEHMs. Recent research has reported that the static magnetic field has the power not only to stimulate bone formation, but also to regulate its orientation in both in vitro and in vivo models.^17,57 Based on the above results, we can infer that the intrinsic nanoscale magnetic field provided by the incorporated Fe₃O₄ nanoparticles may have a stimulating effect on the cell adhesion and proliferation, which is consistent with the previous reports that magnetic hydroxyapatite scaffolds have a positive influence on the adhesion and proliferation of ROS 17/2.8 and MC3T3-E1 cells.^58,59 However, the interaction mechanism of the Fe₃O₄ nanoparticles on cell proliferation remains to be studied further.

Fig. 8 MTT assay results of hBMSC growth on the MEHMs, EHMs and HAPs after different numbers of days. The data are represented as the mean ± standard deviation; n = 3.

Fig. 9 shows the LSCM and FM images of the hBMSCs cultured on the MEHMs for 24 h. An actin cytoskeleton and focal adhesion staining kit is used to map the orientation of actin filaments with TRITC phalloidin and label nuclei with 4′,6-diamidino-2-phenylindole (DAPI). The hBMSCs cultured on the MEHMs exhibit a rearranged cytoskeleton with better-developed stress actin fibers and stronger actin intensity. Moreover, the long red bundles of stress fibers and cell–cell contact are observed in Fig. 9, which suggests excellent cell adhesion and spreading. No evidence of any major deleterious or cytotoxic responses is detected for the MEHMs, as confirmed by the MTT results (Fig. 8).

Fig. 9 (a and d) Cytoskeletal staining, (b and e) DAPI staining and (c and f) emerged graph of hBMSCs on the surface of the MEHMs under (a–c) CLSM and (d–f) FM. The cells were stained with blue and red fluorescence, the actin filaments were stained as red fluorescent light and the nuclei were stained as blue fluorescent light.

3.7. Osteoinductivity of MEHMs

Good osteoinductivity is a preferable property of the bone filling material, which can promote the bone regeneration and remodelling at the bone defect site. Several pieces of research have indicated that the magnetic field may stimulate the proliferation and differentiation of osteoblasts, promote the expression of growth factors such as bone morphological protein, and thus accelerate the healing process of bone fracture.^60–62 Among magnetic materials, Fe₃O₄ nanoparticles have great potential in their application in bone tissue engineering because they are highly biocompatible and have unique magnetic properties. The Fe₃O₄ nanoparticles become superparamagnetic when the diameter is less than 20 nm.⁶³ Each nanoparticle can provide a magnetic field at the nanoscale, as a single magnetic domain. The magnetic field produced by the Fe₃O₄ nanoparticles is roughly around 0.1 T.⁶⁴ In this study, Fe₃O₄ nanoparticles have been successfully integrated into MEHMs. The osteoinductivity of MEHMs has been explored with osteogenic differentiation of hBMSCs. At each time-point, cells are tested for ALP staining and ALP activity. ALP is widely recognized as a marker for osteoblast differentiation. Typical ALP staining of hBMSCs after culturing in osteogenic medium for 7 days (Fig. 10) shows that the ALP staining intensity of hBMSCs on MEHMs is stronger than that on HAPs or EHMs. This result is confirmed by ALP activity assay, as shown in Fig. 11. The ALP activity of hBMSCs on all three groups shows a similar time-dependent increase, and cells on the MEHMs had a higher ALP activity than cells on the HAPs or HHMs at day 3, 7, and 14 (P < 0.05). No significant difference in ALP activity is observed at day 21 (P > 0.05). The results of ALP staining and ALP activity indicate that the osteogenic differentiation of hBMSCs is enhanced on the MEHMs. The possible mechanism to stimulate cell differentiation by magnetic fields is the increased rearrangement of membrane phospholipids. The rearrangement of phospholipids might activate membrane-associated molecules such as cell adhesion molecules, resulting in enhanced cell differentiation.¹⁷ Recently, Meng et al. have developed super-paramagnetic responsive nanofibrous scaffolds under a static magnetic field to enhance osteogenesis for bone repair in vivo.⁵⁷

Fig. 10 Typical ALP staining results of hBMSCs after cultivation of 7 days for different samples: (a) HAPs; (b) EHMs and (c) MEHMs.

Fig. 11 The relative ALP activity results at different days for the MEHMs, EHMs and HAPs.

4. Conclusion

Multifunctional MEHMs have been fabricated by the emulsion method according to the following stages: (i) co-precipitation of CaCO₃/Fe₃O₄ microspheres; and (ii) transformation of the magnetic calcium carbonate microspheres to the MEHMs via the dissolution–precipitation reaction. MEHMs with low crystallinity possess hierarchical porous structures and large surface areas, so they are suitable to be used as drug delivery systems. As compared with the HAPs, the MEHMs possess greater drug loading–release properties for gentamicin. The controlled release of gentamicin from the gentamicin-loaded MEHMs significantly minimizes bacterial adhesion and prevents biofilm formation against S. epidermidis. The Fe₃O₄ nanoparticles in the MEHMs cause them to be magnetic, which not only stimulates the cell adhesion and proliferation, but also promotes the osteogenic differentiation of hBMSCs. The excellent biocompatibility, osteoinductivity, drug loading–release properties and bactericidal properties suggest that the MEHMs have great potential as bone filling materials to treat bone infection. To better investigate the anti-infection ability of the MEHMs, our future studies would be involved with fabricating a three dimensional scaffold using the MEHMs as raw material, and evaluating the in vivo biocompatibility, biodegradability, osteoinductivity and anti-infection properties.

Acknowledgements

This research was supported by Key Disciplines of Shanghai Municipal Education Commission (no. J50206), Natural Science Foundation of China (no 51002095, 51372152 and 30973038), Science and Technology Commission of Shanghai Municipality (no. 12JC1405600), Program of Shanghai Normal University (no DZL124, DCL201303), Innovation Foundation of Shanghai Education Committee (no. 14ZZ124), and State Key Laboratory for Modification of Chemical Fibers and Polymer Materials, Dong Hua University.

Notes and references

1. S. Bose and S. Tarafder, Acta Biomater., 2012, 8, 1401–1421.
2. E. Champion, Acta Biomater., 2013, 9, 5855–5875.
3. S. S. Ray, Acc. Chem. Res., 2012, 45, 1710–1720.
4. M. A. Woodruff, C. Lange, J. Reichert, A. Berner, F. L. Chen, P. Fratzl, J. T. Schantz and D. W. Hutmacher, Mater. Today, 2012, 15, 430–435.
5. K. F. Huo, X. M. Zhang, H. R. Wang, L. Z. Zhao, X. Y. Liu and P. K. Chu, Biomaterials, 2013, 34, 3467–3478.
6. Y. Liu, Z. Zheng, J. N. Zara, C. Y. Hsu, D. E. Soofer, K. S. Lee, R. K. Siu, L. S. Miller, X. L. Zhang, D. Carpenter, C. L. Wang, K. Ting and C. Soo, Biomaterials, 2012, 33, 8745–8756.
7. K. C. Popat, M. Eltgroth, T. J. Latempa, C. A. Grimes and T. A. Desai, Biomaterials, 2007, 28, 4880–4888.
8. K. Osawa, Y. Okubo, K. Nakao, N. Koyama and K. Bessho, J. Gene Med., 2010, 12, 937–944.
9. J. D. Boerckel, Y. M. Kolambkar, K. M. Dupont, B. A. Uhrig, E. A. Phelps, H. Y. Stevens, A. J. Garcia and R. E. Guldberg, Biomaterials, 2011, 32, 5241–5251.
10. C. Rey, Biomaterials, 1990, 11, 3–5.
11. W. H. Yang, X. F. Xi, J. F. Li and K. Y. Cai, Asian J. Chem., 2013, 25, 3673–3678.
12. B. S. Kim, J. S. Kim, Y. S. Chung, Y. W. Sin, K. H. Ryu, J. Lee and H. K. You, J. Biomed. Mater. Res., Part A, 2013, 101, 1550–1558.
13. C. Mochizuki, H. Hara, I. Takano, T. Hayakawa and M. Sato, Mater. Sci. Eng., C, 2013, 33, 951–958.
14. C. P. Klein, J. M. de Blieck-Hogervorst, J. G. Wolke and K. de Groot, Biomaterials, 1990, 11, 509–512.
15. A. Rupani, L. A. Hidalgo-Bastida, F. Rutten, A. Dent, I. Turner and S. Cartmell, J. Biomed. Mater. Res., Part A, 2012, 100A, 1089–1096.
16. R. Z. LeGeros, Clin. Orthop. Relat. Res., 2002, 395, 81–98.
17. H. Kotani, H. Kawaguchi, T. Shimoaka, M. Iwasaka, S. Ueno, H. Ozawa, K. Nakamura and K. Hoshi, J. Bone Miner. Res., 2002, 17, 1814–1821.
18. J. H. Jansen, O. P. van der Jagt, B. J. Punt, J. A. Verhaar, J. P. van Leeuwen, H. Weinans and H. Jahr, BMC Musculoskeletal Disord., 2010, 11, 188.
19. X. B. Zeng, H. Hu, L. Q. Xie, F. Lan, W. Jiang, Y. Wu and Z. W. Gu, Int. J. Nanomed., 2012, 7, 3365–3378.
20. N. Bock, A. Riminucci, C. Dionigi, A. Russo, A. Tampieri, E. Landi, V. A. Goranov, M. Marcacci and V. Dediu, Acta Biomater., 2010, 6, 786–796.
21. Y. W. Jun, J. W. Seo and J. Cheon, Acc. Chem. Res., 2008, 41, 179–189.
22. S. Thanyaphoo and J. Kaewsrichan, J. Pharm. Sci., 2012, 101, 2870–2882.
23. Y. P. Guo, L. H. Guo, Y. B. Yao, C. Q. Ning and Y. J. Guo, Chem. Commun., 2011, 47, 12215–12217.
24. K. L. Lin, L. Chen, P. Y. Liu, Z. Y. Zou, M. L. Zhang, Y. H. Shen, Y. Q. Qiao, X. Y. Liu and J. Chang, CrystEngComm, 2013, 15, 2999–3008.
25. Z. H. Li, Z. Liu, M. L. Yin, X. J. Yang, Q. H. Yuan, J. S. Ren and X. G. Qu, Biomacromolecules, 2012, 13, 4257–4263.
26. K. L. Lin, P. Y. Liu, L. Wei, Z. Y. Zou, W. B. Zhang, Y. Qian, Y. H. Shen and J. Chang, Chem. Eng. J., 2013, 222, 49–59.
27. J. Moeller-Siegert, J. Parmentier, K. Anselme and C. Vix-Guterl, J. Mater. Sci., 2013, 48, 3722–3730.
28. N. Pramanik and T. Imae, Langmuir, 2012, 28, 14018–14027.
29. P. K. Ray and P. E. Montague, J. Appl. Polym. Sci., 1977, 21, 1267–1272.
30. P. Frutos Cabanillas, E. Diez Pena, J. M. Barrales-Rienda and G. Frutos, Int. J. Pharm., 2000, 209, 15–26.
31. M. F. Pittenger, A. M. Mackay, S. C. Beck, R. K. Jaiswal, R. Douglas, J. D. Mosca, M. A. Moorman, D. W. Simonetti, S. Craig and D. R. Marshak, Science, 1999, 284, 143–147.
32. H. Sun, C. Wu, K. Dai, J. Chang and T. Tang, Biomaterials, 2006, 27, 5651–5657.
33. M. J. Olszta, X. Cheng, S. S. Jee, R. Kumar, Y. Y. Kim, M. J. Kaufman, E. P. Douglas and L. B. Gower, Mater. Sci. Eng., R, 2007, 58, 77–116.
34. M. Sadat-Shojai, M.-T. Khorasani, E. Dinpanah-Khoshdargi and A. Jamshidi, Acta Biomater., 2013, 9, 7591–7621.
35. C. B. Baddiel and E. E. Berry, Spectrochim. Acta, 1966, 22, 1407–1416.
36. D. Walsh, T. Furuzono and J. Tanaka, Biomaterials, 2001, 22, 1205–1212.
37. R. M. Wilson, S. E. P. Dowker and J. C. Elliott, Biomaterials, 2006, 27, 4682–4692.
38. O. Bretcanu, S. Spriano, E. Verné, M. Cöisson, P. Tiberto and P. Allia, Acta Biomater., 2005, 1, 421–429.
39. M. R. Virto, P. Frutos, S. Torrado and G. Frutos, Biomaterials, 2003, 24, 79–87.
40. Y. P. Guo, Y. B. Yao, Y. J. Guo and C. Q. Ning, Microporous Mesoporous Mater., 2012, 155, 245–251.
41. T. Long, Y. P. Guo, Y. Z. Liu and Z. A. Zhu, RSC Adv., 2013, 3, 24169–24176.
42. A. Savarese, M. L. Nanni, C. Pasquali and A. C. Egidio, Chir. Organi Mov., 1999, 84, 247–251.
43. F. de Lalla, J. Chemother., 2000, 12, 23–27.
44. A. D. Hanssen and D. R. Osmon, Clin. Orthop. Relat. Res., 1999, 369, 124–138.
45. R. O. Darouiche, N. Engl. J. Med., 2004, 350, 1422–1429.
46. M. Kazemzadeh-Narbat, J. Kindrachuk, K. Duan, H. Jenssen, R. E. W. Hancock and R. Wang, Biomaterials, 2010, 31, 9519–9526.
47. J. G. Hendriks, J. R. van Horn, H. C. van der Mei and H. J. Busscher, Biomaterials, 2004, 25, 545–556.
48. T. F. Mah and G. A. O'Toole, Trends Microbiol., 2001, 9, 34–39.
49. A. G. Gristina and J. Kolkin, J. Bone Jt. Surg., Am. Vol., 1983, 65, 128–134.
50. B. Sugarman and E. J. Young, Infectious Disease Clinics of North America, 1989, 3, 187–198.
51. B. Gottenbos, H. C. van der Mei and H. J. Busscher, J. Biomed. Mater. Res., 2000, 50, 208–214.
52. M. K. Lacy, D. P. Nicolau, C. H. Nightingale and R. Quintiliani, Clin. Infect. Dis., 1998, 27, 23–27.
53. W. Chang, X. Mu, X. Zhu, G. Ma, C. Li, F. Xu and J. Nie, Mater. Sci. Eng., C, 2013, 33, 4369–4376.
54. Y. Li, Y. Z. Liu, T. Long, X. B. Yu, T. T. Tang, K. R. Dai, B. Tian, Y. P. Guo and Z. A. Zhu, J. Mater. Sci.: Mater. Med., 2013, 24, 1951–1961.
55. M. J. Dalby, D. McCloy, M. Robertson, C. D. W. Wilkinson and R. O. C. Oreffo, Biomaterials, 2006, 27, 1306–1315.
56. M. J. Dalby, D. McCloy, M. Robertson, H. Agheli, D. Sutherland, S. Affrossman and R. O. C. Oreffo, Biomaterials, 2006, 27, 2980–2987.
57. J. Meng, B. Xiao, Y. Zhang, J. Liu, H. Xue, J. Lei, H. Kong, Y. Huang, Z. Jin, N. Gu and H. Xu, Sci. Rep., 2013, 3, 2655.
58. Y. Wu, W. Jiang, X. Wen, B. He, X. Zeng, G. Wang and Z. Gu, Biomed. Mater., 2010, 5, 015001.
59. X. B. Zeng, H. Hu, L. Q. Xie, F. Lan, W. Jiang, Y. Wu and Z. W. Gu, Int. J. Nanomed., 2012, 7, 3365–3378.
60. C. A. Bassett, M. Schink-Ascani and S. M. Lewis, Clin. Orthop. Relat. Res., 1989, 246, 172–185.
61. M. T. Santini, G. Rainaldi, A. Ferrante, P. L. Indovina, P. Vecchia and G. Donelli, Bioelectromagnetics, 2003, 24, 327–338.
62. K. J. McLeod and L. Collazo, Radiat. Res., 2000, 153, 706–714.
63. S. Sun, H. Zeng, D. B. Robinson, S. Raoux, P. M. Rice, S. X. Wang and G. Li, J. Am. Chem. Soc., 2004, 126, 273–279.
64. Y. P. Guo, Y. Zhou, D. C. Jia and Q. C. Meng, Acta Biomater., 2008, 4, 923–931.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3ra45896b

‡ T. Long and Y. P. Guo contributed equally to this work.

---