Biomacromolecules 2008, 9, 2705–2711
2705
Preparation of Poly(ethylene glycol) Protected Nanoparticles with Variable Bioconjugate Ligand Density Marian E. Gindy,† Shengxiang Ji,‡ Thomas R. Hoye,‡ Athanassios Z. Panagiotopoulos,†,§ and Robert K. Prud’homme*,† Department of Chemical Engineering and Princeton Institute for the Science and Technology of Materials, Princeton University, Princeton, New Jersey 08544, and Department of Chemistry, University of Minnesota, Minneapolis, Minnesota 55455 Received February 21, 2008; Revised Manuscript Received June 11, 2008
Maleimide-functional poly(ethylene glycol)-b-poly(ε-caprolactone) nanoparticles (NPs) were prepared via the Flash NanoPrecipitation technique. Subsequent reaction with a model ligand, bovine serum albumin (BSA), was conducted using thiol-maleimide conjugation. Reaction of up to 22% of NP surface maleimide-PEG tethers was obtained, with the percent conversion being essentially independent of the ratio of maleimide-PEG to methyl-PEG over the range 30-100%, respectively. At the highest surface coverage, BSA is calculated to essentially cover the NP surface area. Reaction parameters (reaction order and docking constant) describing the extent of ligand conjugation were determined. The reaction order is applicable to the conjugation of ligands presenting free thiol functionalities, while the value of the docking constant is ligand-dependent and accounts for physical and dynamic properties of the ligand-PEG interaction. Jointly, the particle formation process, using block copolymer-directed kinetically controlled assembly and surface functionalization represent a versatile new platform for the preparation of bioconjugated NPs with accurate control of ligand density and minimal processing steps.
Introduction Nanoparticles (NPs) prepared through the self-assembly of amphiphilic block copolymers have attracted considerable attention for the improved delivery of poorly water-soluble compounds, particularly chemotherapeutic agents.1-5 The utility of NPs results from the ability to incorporate hydrophobic drugs within the particle interiors at concentrations greater than their intrinsic water solubility.6,7 In addition, the hydrophilic outer corona of the block copolymer NP imparts stability in aqueous environments and prevents nonspecific protein adsorption, effectively reducing uptake by the reticuloendothelial system and extending particle circulation in vivo.8,9 It is desirable to have the NPs specifically localized at the site of diseased tissue. Site-specific delivery can lead to reduced drug toxicity and increased therapeutic efficacy. In this capacity, particle size has been exploited for the improved localization of systemically administered carriers at solid tumors through an enhanced permeation and retention (EPR) effect.10,11 Beyond passive targeting through EPR, delivery of NPs to smaller solid tumors and metastatic cells can potentially be achieved by active targeting, whereby particle surfaces are modified with moieties directed at cell surface markers unique to these cell types.12,13 Typically, targeting ligands based on monoclonal antibodies, peptides, and proteins are used for this purpose.14-17 Diverse approaches have been described for the conjugation of targeting proteins to NP surfaces.18,19 Commonly, protein-NP conjugates are synthesized via reaction of ε-amino groups of lysine residues on the protein ligand with carboxylic acid, aldehyde, or amine-functional particles. However, the large number of amine groups typically available for reaction results in a complex pattern * To whom correspondence should be addressed. E-mail: prudhomm@ princeton.edu. † Department of Chemical Engineering, Princeton University. ‡ University of Minnesota. § Princeton Institute for the Science and Technology of Materials.
of multiple substitutions and leads to the formation of structurally and functionally heterogeneous systems.20 Alternatively, more selective attachment can be achieved via reaction with lessabundant cysteine residues on the protein ligand. Reactions based on maleimide-thiol chemistry are of particular interest, as they take place under conditions that maintain the structural, biological, and binding properties of the protein.21-26 While bioconjugated NPs have shown promise toward improved drug delivery, significant challenges and opportunities for improvement remain. First, the preparation of protein-NP conjugates through traditional NP preparation processes such as emulsion, solvent-evaporation, and dialysis methods can lead to significant variability among formulations.27 The inconsistencies are largely the consequence of experimental variability associated with conducting multistep reactions and purifications required for the isolation of uniformly-sized, stabilizer-free particles. A second major variable is the surface density of the targeting component. Modulation of ligand surface density in a controllable fashion is expected to be a significant factor contributing to the overall targeting efficiency of the NP constructs.28,29 The process and strategy we present in this paper offer significant advantages for the production of targeted NPs. Flash NanoPrecipitation with reactive block copolymers allows the formation of NPs with a minimum of processing steps, control over particle size, drug loading, and composition of the protective polymer layer. The power of the Flash NanoPrecipitation process is that the difficult chemical derivatization and purification steps can be done in homogeneous solution on the components prior to particle synthesis. Subsequently, the composition of the NP in terms of drug loading, protective PEG surface concentration, and reactive PEG surface concentration is controlled quantitatively by stoichiometry during rapid precipitation. This greatly simplifies the production of NPs with a range of compositions and loadings. If the targeting ligand is a small molecule, without secondary structure, it can be attached
10.1021/bm8002013 CCC: $40.75 2008 American Chemical Society Published on Web 08/30/2008
2706
Biomacromolecules, Vol. 9, No. 10, 2008
to the block copolymer prior to NP synthesis. If the targeting ligand is a protein or antibody with secondary structure that would be denatured by solubilization in the organic stream, then the final ligand conjugation must be done post-NP formation. If quantitative coupling chemistry is employed to conjugate the targeting protein to the reactive sites on the NP surface, then the surface ligand density can be precisely known. The purpose of this study is to quantify the extent of ligand coupling so that the relationship between the stoichiometry of the reactive block copolymer feed and the surface density of ligands can be determined. We have selected bovine serum albumin (BSA) as a model ligand as it is commonly used in biomedical applications to impart biocompatibility to material surfaces. In addition, the cationic form of the protein has recently been reported to facilitate transport across the blood-brain barrier.33 The structure of the paper is as follows. The preparation of maleimide-functional poly(ethylene glycol)-block-poly(ε-caprolactone) NPs via Flash NanoPrecipitation and their characterization are first described. Next, protein reaction with the terminal maleimide reactive groups on the PEG corona of the block copolymer particles is measured. Finally, a model of the maleimide conversion is presented that enables a predictive and quantitative approach to design ligand-modified NPs for targeting applications.
Experimental Section Materials. Bicinchoninic acid (BCA) protein assay kit was purchased from Pierce Biotechnology (Rockford, IL). All other chemicals were purchased from Sigma-Aldrich (St. Louis, MO) and, unless otherwise noted, used as received. Reagents were all of analytical grade. Water, purified by reverse osmosis, ion-exchange, and filtration (Milli-Q water) was used for NP preparation and dialysis. Preparation of Poly(ethylene glycol)-b-poly(ε-caprolactone) Copolymer. Monomethoxy-poly(ethylene glycol)-block-poly(ε-caprolactone) (5000 g/mol-b-7000 g/mol; mPEG-b-PCL) copolymers were synthesized by solution-phase, acid-catalyzed ring-opening polymerization according to the published method of Shibasaki and co-workers.34 A typical procedure is as follows. All glass vessels were heated under vacuum before used and handled in a stream of dry nitrogen. Dichloromethane (CH2Cl2) and ε-caprolactone (ε-CL) were distilled over calcium hydride directly before use. Monomethoxy-terminated poly(ethylene glycol) mPEG-OH (5 g, 1 mmol; Mn ) 5000, Mw/Mn ) 1.1), dried by azeotropic distillation of anhydrous toluene (200 mL), was used as macroinitiator for polymerization of ε-CL (6.8 mL, 61.5 mmol) in dichloromethane (200 mL). The copolymerization was initiated by addition of 2 M HCl solution in diethyl ether (1 mL, 2 mmol) and carried out at room temperature for 24 h. The reaction mixture was evaporated to 50 mL and added to an excess amount of cold n-hexane (400 mL) to precipitate the polymer. The copolymer was collected by filtration and dried in vacuo. High-resolution 1H NMR spectra were obtained using a Varian Inova 400 MHz spectrometer. Copolymer composition was calculated from the integration of signals corresponding to the R-methylene proton signal of ε-CL (-CH2-OH, δ∼2.65 ppm), methylene proton signals of the ε-CL main chain (R-, β-, γ-, δ-, ε- -CH2-, δ ∼2.31, 1.65, 1.39, 1.65, and 4.06 ppm, respectively), and the methylene proton signal of mPEG-OH (-CH2-, δ ∼3.65 ppm). The molecular weight of the PCL block was calculated from 1H NMR and the known molecular weight of the mPEG-OH macroinitiator, as reported by the manufacturer (Mn ) 5000). Gel permeation chromatography in tetrahydrofuran (THF) was used to determine the molecular weight and polydispersity index. The GPC system consisted of two 30 cm Polymer Laboratories PLgel Mixed-C columns, a Waters (Milford, MA) 515 HPLC pump, and a Waters 410 differential refractometer. The system was calibrated using monodisperse polystyrene standards (Polysciences Inc., Warrington, PA).
Gindy et al. Maleimide-poly(ethylene glycol)-block-poly(ε-caprolactone) (5000 g/molb-7000 g/mol; maleimidePEG-b-PCL) copolymer was prepared by coupling an acid chloride-terminated PCL with an excess of dihydroxyterminated poly(ethylene glycol) (HO-PEG-OH) followed conversion of the OH on the purified block copolymer to the maleimide. Characterization by 1H NMR and MALDI-TOF MS showed >90% maleimide functionality. The synthesis and characterization of the copolymer are detailed elsewhere.35 Preparation of Maleimide-Functional PEG-b-PCL Nanoparticles. NPs were prepared using the Flash NanoPrecipitation30-32 technology previously developed by our group. The process uses intense mixing of organic solvent streams containing molecularly dissolved organic solutes and amphiphilic copolymers with nonsolvents, typically water, to effect high supersaturations and kinetically controlled aggregation of hydrophobic compounds using block copolymer selfassembly. Using this technology, we have demonstrated the preparation of multicomponent block copolymer stabilized organic and inorganic particles with controlled particle size (50-300 nm), narrow particle size distributions, specific component composition, and high loading efficiencies.30-32 In the present study, the technology is utilized for the preparation of surface functional NPs for conjugation to biological ligands. A representative synthesis of maleimide-functional NPs prepared via Flash NanoPrecipitation is as follows. NPs were made with a variable ratio of nonfunctional mPEG-b-PCL (5000-b-7000 g/mol) and functional maleimidePEG-b-PCL (5000-b-7000 g/mol). For example, to a solution of mPEG-b-PCL (27 mg) in THF (5 mL) was added maleimidePEGb-PCL (27 mg). The organic ssolution was fed (12 mL/min, stream 1), along with Milli-Q water (40 mL/min, streams 2-4), into a four-stream multi-inlet vortex mixer36 (MIVM) using two digitally controlled syringe pumps (Harvard Apparatus, PHD 2000 programmable, Holliston, Massachusetts). NP suspensions were either prepared for protein conjugation, as described subsequently, or were dialyzed against Milli-Q water (2 L Milli-Q water per 15 mL NP suspension) for 24 h using a Spectra/Por dialysis bag with MWCO of 6000-8000 g/mol (Spectrum Laboratories Inc., California) and stored at 4 °C. Preparation of BSA-Nanoparticle Conjugates. Reaction of maleimide-functional PEG-b-PCL NPs with BSA was carried out as follows. NP suspensions were evaporated under vacuum at 25 °C using a rotary evaporator to remove volatile THF. The suspension was degassed by gently bubbling nitrogen for approximately 30 min. Stock solutions of BSA (10 mg/mL fraction V, lyophilized powder) in degassed 0.15 M NaCl (pH 6.2-6.5) were prepared immediately before use. To the NP suspension (5 mL), the appropriate volume of BSA stock solution was added via cannula under nitrogen pressure. A solution of degassed 0.3 M NaCl (pH 6.2-6.5) containing 0.4 mM ethylenediamine tetraacetic acid (EDTA) was used to adjust the volume to 10 mL. The reaction was performed overnight (12-15 h) at room temperature on a rotating mixer set at low speed. Unreacted BSA was separated from BSA-NP conjugates via centrifugation through a 300KDa MWCO OMEGA nanoseparation filter membrane (Pall Corporation, East Hills, NY) using a benchtop microcentrifuge rotating at 8800 g for 30 min. Washes of the retentate (repeating centrifugation at 8800 g for 30 min and resuspension with 1 mL of 0.15 M NaCl, pH 7.2) were conducted three times. Finally, the purified BSA-NPs were resuspended in 0.25 mL of 0.1 M phosphate buffered saline (pH 7.2) and stored at 4 °C. Quantification of the amount of BSA reacted to NP surfaces was preformed using the BCA assay. NP suspension samples or standards were incubated with BCA assay solutions at 60 °C for 1 h. UV-visible absorbance was measured at 565 nm using an Evolution 300 spectrometer (Thermo Electron Inc., Madison, Wisconsin). Each sample was run in duplicate together with a minimum of five series dilutions of BSA standards, also run in duplicate. The unknown concentration of protein in a given sample was determined from linear regression fit of absorbance versus concentration for BSA standard solutions. To match the effects of light scattering by NPs on the absorbance measurements,
Polyethylene Glycol Protected Nanoparticles
Biomacromolecules, Vol. 9, No. 10, 2008
2707
Figure 1. (a) Schematic representation of maleimide-functional PEG-b-PCL NP formation by Flash NanoPrecipitation in a multi-inlet vortex mixer (MIVM). The density of maleimide moieties on NP surfaces is linearly proportional to the fraction of maleimidePEG-b-PCL in the inlet stream to the MIVM. The NPs can coencapsulate organic therapeutics and inorganic imaging agents.32 (b) Conditions for reaction of NPs with free thiol-containing macromolecules.
nonreactive NPs were added to solutions for the standard curve at the estimated NP content in the sample for determination. Nanoparticle Size and Size Distributions. Particle size was measured via dynamic light scattering (DLS). The DLS instrument (Brookhaven Instruments, BI-200SM, Holtsville, NY) consisted of a double-pumped continuous NdYAG laser (Coherent Inc., wavelength 532 nm, 100 mW, Santa Clara, CA) and a photomultiplier. Measurements were performed at a 90° angle for 3 min and repeated a minimum of three times per sample. The measured time correlation functions were analyzed by autocorrelation (ALV-Laser Vertriebsgesellschaft mbH, ALV-5000/E, Langen, Germany) using the method of the cumulants, providing an average value of the intensity-average particle size and particle polydispersity index (PDI), defined as the ratio of the variance to the square of the average decay rate. Intensity-average and number-average particle size distributions were calculated using the Laplace inversion program, CONTIN. The reported NP surface areas were determined from the measured number-average particle sizes.
Results and Discussion Nanoparticle Preparation and Characterization. The synthetic pathway leading to protein-derivatized NPs is depicted schematically in Figure 1. The process can be divided into two parts. In the first step, maleimide-functional NPs were prepared via Flash NanoPrecipitation in a four-stream MIVM using a blend of nonfunctional mPEG-b-PCL (5000 g/mol-b-7000 g/mol) and functional maleimidePEG-b-PCL (5000 g/mol-b-7000 g/mol) copolymers. Particles were generated at fixed total copolymer composition (0.1 wt % in the final suspension) and varying concentration of maleimidePEG-b-PCL. DLS was used to obtain a measure of NP hydrodynamic diameters and size distributions.
Mean particle diameters of unfiltered NP suspensions are presented in Figure 2a. For each formulation, NP size is reported as intensityaverage diameter (djI) and number-average diameter (djN). The error bars represent the standard deviation in measured diameters of several experimental runs generated at each condition. As can be seen, djI of NPs composed solely of mPEG-b-PCL is 50 ( 2 nm and increases to 85 ( 3 nm for NPs prepared with a copolymer composition of 30:70 mol % maleimidePEG-b-PCL/mPEG-b-PCL. When the fraction of maleimidePEG-b-PCL is increased further, djI remains relatively unaffected. Because the intensity of scattered light is proportional to the sixth moment of diameter, the presence of a few particles of larger size can significantly impact values of djI.37 For comparison, the number-average particle diameters of the same formulations are reported. As shown in Figure 2a, for pure mPEG-b-PCL NPs, djN is 26 ( 3 nm and remains essentially constant (30 ( 5 nm) with increasing maleimidePEG-b-PCL fraction. In general, djI values are nearly double the corresponding djN values, confirming the sensitivity of the intensity average diameters to the presence of a few large particles. Corresponding intensity-average particle size distributions are shown in Figure 2b. For all formulations, the particle size PDI, calculated by the cumulant method, are in the range of 0.2-0.3. For reference, the particle size distribution of calibration-grade polystyrene (PS) latex spheres (diameter, 80 ( 3 nm; PDI, 0.11 ( 0.03) is also shown. The number-average particle size distributions of the same formulations are additionally shown in Figure 2c. In all cases, the distributions are narrow, with standard deviations, indicative of the width of the distributions, of e5 nm and essentially all particles less than 200 nm in diameter. The relatively narrow dispersity in particle size allows for sterile filtration via 0.2 µm
2708
Biomacromolecules, Vol. 9, No. 10, 2008
Gindy et al.
Figure 3. DLS spectra of maleimide-functional PEG-b-PCL NPs before reaction (dashed line), in reaction mixture with BSA following a 12 h incubation (dotted line) and after purification (solid line). Complete removal of unreacted BSA following purification is confirmed by absence of free BSA peak (near 10 nm) in the purified BSA-NP conjugate sample. The BSA-NP conjugate spectrum shows an increase in diameter compared to that of unreacted NPs, corresponding to the presence of BSA on the NP surfaces.
Figure 2. NP size and size distributions by DLS. (a) Intensity-average diameters (dj I) and number-average diameters (dj N) of NPs as a function of maleimidePEG-b-PCL mole fraction in the formulation. Symbols are from experiments, with lines as guide to the eye. The standard deviations of measured diameters for multiple experiments generated at each condition are represented by error bars. (b) Representative intensity-average particle size distributions as a function of malemidePEG-b-PCL mole fraction (indicated in legend). The size distribution of calibration grade PS latex spheres of size 80 ( 3 nm (solid line) is shown for reference. (c) Corresponding numberaverage particle size distributions of samples as in (b).
pore membrane filtration with minimal material losses. Sterile filtration is a preferred method of aseptic NP preparation, as alternative methods have been shown to result in polymer degradation.38 The control of surface density of the targeting protein is dependent on the density of reactive groups presented on the NP surfaces. In the formulations described, NPs were prepared by varying the ratio of mPEG-b-PCL to maleimidePEG-b-PCL to control the density of maleimide groups on particle surfaces. The ability to precisely control reactive group density follows directly from the mechanism of particle self-assembly via Flash NanoPrecipitation. In the Flash NanoPrecipitation process, particle formation is governed by the kinetics of block copolymer precipitation and self-assembly (τsa) upon change in the solvent quality. By utilizing reactive and nonreactive copolymers in which monomer composition and block molecular weights are matched, we ensure that τsa for the two copolymers is essentially equivalent. Consequently, when the mixing rate is sufficiently fast, homogeneous diffusion limited aggregation kinetics prevail, and the fraction of each copolymer type in a given NP will be dictated solely by its solution concentration.
Reaction of Nanoparticles with Bovine Serum Albumin. BSA possess 35 cysteine residues, 34 of which are covalently linked to form 17 intramolecular disulfide bonds, with the remaining cysteine residue present as a free thiol (SH) near the amino terminus of the molecule. Oxidation can result in the formation of intermolecular disulfide bonds and a reduction in the number of thiol groups available for reaction. Because the degree of oxidation can vary significantly among samples, we first quantified the free thiol content of BSA employed in our experiments via a thiol-disulfide interchange reaction with Ellman’s reagent.39 Titration of BSA solutions of known concentrations revealed that approximately 55 ( 2 mol % of the BSA molecules contain a SH unit available for reaction (see Supporting Information for details). This estimate is line with previous reports on commercially available BSA.40 The reaction of BSA with maleimide-functional NPs was performed under mild aqueous conditions as represented in Figure 1b. Following reaction, purification of the BSA-NP conjugates from unreacted BSA was accomplished via multiple cycles of ultrafiltration via 300KDa MWCO centrifuge membranes until essentially no protein (
