- Email: [email protected]
An Investigation of the Factors Controlling the Adsorption of Protein Antigens to Anionic PLG Microparticles
An Investigation of the Factors Controlling the Adsorption of Protein Antigens to Anionic PLG Microparticles JAMES CHESKO, JINA KAZZAZ, MILDRED UGOZZOLI, DEREK T. O’HAGAN, MANMOHAN SINGH Vaccine Delivery Group, Chiron Corporation, 4560 Horton St., Emeryville, California 94608
Received 18 March 2005; accepted 15 July 2005 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20472
ABSTRACT: This work examines physico-chemical properties influencing protein adsorption to anionic PLG microparticles and demonstrates the ability to bind and release vaccine antigens over a range of loads, pH values, and ionic strengths. Poly(lactide-co-glycolide) microparticles were synthesized by a w/o/w emulsification method in the presence of the anionic surfactant DSS (dioctyl sodium sulfosuccinate). Ovalbumin (OVA), carbonic anhydrase (CAN), lysozyme (LYZ), lactic acid dehydrogenase, bovine serum albumin (BSA), an HIV envelope glyocoprotein, and a Neisseria meningitidis B protein were adsorbed to the PLG microparticles, with binding efficiency, initial release and zeta potentials measured. Protein (antigen) binding to PLG microparticles was influenced by both electrostatic interaction and other mechanisms such as van der Waals forces. The protein binding capacity was directly proportional to the available surface area and may have a practical upper limit imposed by the formation of a complete protein monolayer as suggested by AFM images. The protein affinity for the PLG surface depended strongly on the isoelectric point (pI) and electrostatic forces, but also showed contributions from nonCoulombic interactions. Protein antigens were adsorbed on anionic PLG microparticles with varying degrees of efficiency under different conditions such as pH and ionic strength. Observable changes in zeta potentials and morphology suggest the formation of a surface monolayer. Antigen binding and release occur through a combination of electrostatic and van der Waals interactions occurring at the polymer-solution interface. ß 2005 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 94:2510–2519, 2005
Keywords: microparticles; antigen; vaccine delivery; polylactide-co-glycolide; protein adsorption; PLG; anionic particles; electrostatic; microscopy
INTRODUCTION Next generation vaccines will be comprised mainly of recombinant protein antigens that are often poorly immunogenic. These vaccines will need optimal antigen delivery systems with or without an additional immunopotentiator to generate potent immune responses.1,2 Emulsions, micro/nanoparticles, aluminum salt adjuvants,
Correspondence to: James Chesko (Telephone: 510-9233896; Fax: 510-923-2586; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 94, 2510–2519 (2005) ß 2005 Wiley-Liss, Inc. and the American Pharmacists Association
2510
and iscoms have all been used as delivery systems to boost immune responses to vaccine antigens. Enhanced immune responses are induced through a variety of mechanisms, including increased persistence of antigen at the site of injection, improved targeting to antigen presenting cells, or enhanced protection of antigen linear, and conformational epitopes against degradation.2–4 We have previously described studies which showed that anionic poly(lactide-co-glycolide) (PLG) microparticles were effective delivery systems for adsorbed vaccine antigens, including recombinant proteins from Neisseria meningitidis type B (MB)5 and human immunodeficiency virus
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
(HIV).6 The PLG polymer used in these formulations has a long history of safe use in humans.7–9 Work with the traditional aluminum salt adjuvants has elucidated rules for predicting protein adsorption,10 encouraging adjuvant-antigen interactions,11 and highlighted the significance of surface electrostatic forces in antigen adsorption and release.12 Anionic PLG microparticles13,14 represent an alternative approach to aluminum salt adjuvants,15 or the MF5916 emulsion adjuvant. To assess the strengths and limitations of anionic PLG microparticles as a delivery system for diverse antigens, we have measured the physico-chemical properties such as charge and ionic strength that have been implicated as important to the description of binding and release of a range of proteins from polymer surfaces.17–20 In addition to vaccine relevant protein antigens, model proteins with a wide range of isoelectric points (pI) including ovalbumin (pI 4.6), carbonic anhydrase (pI 6.0), and lysozyme (pI 10.7) were evaluated, to represent proteins with diverse electrostatic properties. Evaluation of the physicochemical properties of anionic PLG microparticles, including size, charge, and surface morphology may provide information that can be used to guide formulation strategies for diverse proteins. The complex behavior of protein on solid surfaces has been studied by various techniques21,22 to suggest the formation of structures including multilayers. Using atomic force microscopy to image the microparticles before and after protein adsorption, we visualized the surface morphology to determine if it was consistent with our hypothesis of the formation of an adsorbed protein monolayer. Furthermore, we investigated the physico-chemical properties influencing protein adsorption to microparticles and demonstrated their ability to bind and release diverse protein antigens over a range of loads, pH, and ionic strengths.
EXPERIMENTAL Materials RG503 and RG502H, poly(D,L-lactide-co-glycolide) 50:50 lactide to glycolide co-polymers were obtained from Boehringer Ingelheim (Petersburg, VA). Dioctylsulfosuccinate (DSS), lysozyme (LYZ), carbonic anhydrase (CAN), bovine serum albumin (BSA), ovalbumin (OVA), lactate dehydrogenase (LADH), USP grade mannitol, sucrose, and tre-
2511
halose came from Sigma-Aldrich Chemical (St. Louis, MO) Chinese Hamster Ovary derived recombinant gp120dV2 was synthesized and purified in house (Chiron Vaccine Research in Emeryville, CA) Escherichia coli (E. Coli)-derived recombinant meningococcal proteins (MB1 and MB2) were obtained from Chiron Vaccines, Siena (Siena, Italy) and were isolated and purified as previously described.23
Microparticle Preparation Microparticles were prepared by a solvent evaporation technique. A 10 mL methylene chloride solution of 6% w/v polymer with 2.5 mL PBS was homogenized using a 10 mm probe, (UltraTurrax, T25 IKA-Labortechnik, Germany) forming water in oil emulsion which was then added to 50 mL of distilled water containing 6 mg/mL DSS and homogenized at very high speed using a homogenizer with a 20-mm probe (ES-15 Omni International, GA) for 25 min in an ice-water bath. This resulted in water in oil in water emulsion, which was stirred at 1000 rpm for 12 h at room temperature as the methylene chloride was allowed to evaporate.
Emulsion Characterization The size distribution of the microparticles was determined using a particle size analyzer (Master sizer, Malvern Instruments, UK). The electrokinetic mobility of the PLG microparticles was measured in PBS buffer on a Malvern ZetaSizer (Malvern Instruments, UK) and the zeta potential (roughly equivalent to particle surface charge) determined. Model proteins with a range of pI were chosen to span a range of buffer adsorption conditions: lysozyme with pI 10.7, lactic acid dehydrogenase with pI 6.8, carbonic anhydrase with pI 6.0 and ovalbumin with pI 4.6. The PLG content of the suspension was measured by aliquoting 1 mL of the suspension into preweighed vials, which were lyophilized and weighed again and the average net weight change was used as PLG content/1 mL suspension. Atomic force microscopy was employed in the noncontact mode using a Digital Instruments BioScope (Digital Instruments, Santa Barbara, CA). The polymer suspension was imaged before and after exposure to protein (1% weight protein to polymer, overnight rocking at 48C, then sample and control lyophilized overnight). The powder
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2512
CHESKO ET AL.
samples were dispersed on a flat, adhesive surface, with representative sampling of numerous (>10) particle populations both within the preparation exposed to protein and the control not exposed to protein.
2 mL of 5% SDS-0.2 M sodium hydroxide solution at room temperature and measuring protein concentration by BCA and HPLC-SEC.
RESULTS AND DISCUSSION Protein Adsorption and Release Microparticles with protein adsorbed for release studies were made by combining a suspension containing 100 mg of PLG with 1 mg protein in 10 mL total volume of 10 mM of the appropriate buffer and left overnight on a lab rocker (aliquot mixer, Miles Laboratories) at 48C. Particles were incubated with the proteins OVA, BSA, CAN, LYS, MB1, MB2, and gp120 with a variety of buffer conditions and target loads ranging from 0 to 10% w/w polymer. Buffers used for pH 5 and 5.5 were histidine and citrate, for pH 7 phosphate, and pH 9 borate. Samples to be used for protein release studies were lyophilized after overnight adsorption (16–18 h). To determine the amount of adsorbed protein, the microparticles were separated from the incubation medium before the final lyophilization step by centrifugation, and the pellet washed three times with distilled water, then lyophilized. The loading level of protein adsorbed to the microparticles was determined by dissolving 10 mg of the microparticles in 2 mL of 5% SDS-0.2 M sodium hydroxide solution at room temperature followed by the bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL) using BSA concentration standards. To confirm that the BSA standard was appropriate, representative samples were analyzed for protein integrity and concentration by size exclusion chromatography (HPLC-SEC method, Tosoh-BioSep SW3000XL 4.6 mm 30 cm column, in PBS) on the supernatant fraction with 90%þ mass balance achieved. Uncertainties in concentration determinations were estimated by combining variances from replicates (intra batch), different microparticle (inter batch) and serial dilution multiple determinations from the BCA determination. To determine the amount of protein released over a 2 h time period was measured as follows. Twenty-five milligrams of lyophilized microparticles following protein adsorption was incubated with 1 mL of the appropriate buffer at 378C for 2 h. The amount of protein present in the supernatant was determined by BCA. To determine the % of 2-h release, the total protein load was measured by dissolving 10 mg of lyophilized microparticles in
Microparticles were prepared with a typical mean size of 1 mm (size distribution of 0.4–1.9 mm, with polydispersity about 50%). Atomic force imaging and scanning electron micrographs24 taken before and after protein adsorption showed these microparticles to be spherical in shape, with a smooth outer surface (Fig. 1). The AFM phase image showed greater contrast and detail than the topographic profile. Morphological differences measured between the untreated and protein treated samples were attributed to the presence and interactions of the protein on the polymer surface. The phase contrast AFM images show notable differences in surface features, with the addition
Figure 1. AFM topography (left) and phase images (right) of dry PLG microparticles before (above) and after (below) overnight gp120dV2 protein adsorption. Microparticles were air dried and imaged under ambient conditions.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
of protein resulting in nearly contiguous islands about 50 nm wide and 5 nm in height that are attributed to clusters of proteins approximately one molecular layer thick. The PLG/DSS microparticles had a zeta potential of 55 mV in 10 mM pH 5.0 citrate buffer, while PLG/DSS particles with 1% w/w adsorbed gp120dV2 protein had a zeta potential in the same buffer of 8 mV, indicating that charge neutralization at the protein/polymer interface had occurred as a result of protein adsorption (Tab. 1). The proteins studied show a general trend of increasing (less negative) zeta potential as additional protein was adsorbed, as would be predicted for a potentiometric titration of the anionic microparticles with the protein. In the cases where the protein carried a net negative charge, the adsorption was generally inefficient and the Stern layer structure may involve inclusion of positive counter ions and solvation molecules. Pseudotherms Show Site Saturation Behavior for Surface Adsorption and the Formation of a Langmuir Monolayer Figure 2 shows adsorption pseudotherms for the model proteins lysozyme, bovine serum albumin, and carbonic anhydrase in addition to vaccine antigens, HIV-envelope protein gp120dV2 and two Meningococcus B proteins MB1 and MB2 on the PLG microparticles. As expected for an adsorption process, binding efficiency decreases with increasing protein input and saturated as the available adsorption sites become completely occupied. This curve profile has been consistently reproduced for the numerous proteins that we have studied, and empirically finds an asymptote between 1 and 2 % w/w protein to polymer. For some species such as carbonic anhydrase and the two MenB proteins, the asymptotic saturation is less distinct. A simplistic treatment of the protein
2513
adsorption data is explained quite adequately by a linear Langmuir plot25 of a ratio of aqueous protein concentration divided by surface adsorbed protein versus aqueous protein concentration, supporting the description that a monolayer of adsorbed protein is formed (Fig. 3). The PLG surface fully covered with a protein monolayer has a reduced affinity to adsorb additional protein as compared to the polymer/surfactant interface with sites unoccupied by protein. For proteins adsorbing close to their isoelectric point, binding of the primary monolayer appears weaker, perhaps due to reduced electrostatic forces, but subsequent layer formation is more facile as a result of reduced charge repulsion. Protein Adsorption Capacity May be Predicted from the Microparticle Surface Area The existence of an upper limit (maximum binding capacity) for adsorption is consistent with the idea that a monolayer (finite, defined surface area available for adsorption) may exist. We may calculate an estimated value for this loading level by comparing geometric dimensions of the microparticles and the protein lysozyme by a geometric calculation (see Tab. 2) that predicts approximately a 1.2% protein to polymer ratio for 1 mm PLG microparticles. Calculations with albumin (not shown) yield a similar load value (1.5% w/w), suggesting that for globular proteins that retain their shape, an upper loading limit of 1–2% w/w will represent the accessible PLG surface area for adsorption of a protein monolayer. Accommodation and aggregation of protein could alter this value, but the data in Figure 1 suggests that this trend is often well observed. To test the available surface area model, PLG microparticles with constant lactide to glycolide (50% Lactide:50% Glycolide) ratio were synthesized with a range of sizes. The surfactant (DSS)
Table 1. Surface Charge on PLG Microparticles before and after Protein Adsorption
PLG particles PLG þ BSA PLG þ OVA PLG þ CAN PLG þ LYS PLG þ gp120dV2
x, mV, pH 5.0 (Citrate)
x, mV, pH 7.0 (PBS)
x, mV, pH 9.0 (Borate)
Max Dx, mV
% Binding at Max Dx
55 2 40 2 49 2 25 2 9 2 8 2
28 2 23 2 23 2 26 2 24 2 29 2
82 2 81 2 82 2 80 2 25 2 37 2
þ15 3 þ 53 þ30 3 þ57 3 þ47 3
48 5 15 3 87 3 100 2 98 2
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
4/3pr3r ¼ 4p(0.5 104 cm)3 (1.5 g/cm3) ¼ 7.9 1013 g
N/A
4pr ¼ 4(p)(0.5 104 cm)2 ¼ 3.1 108 cm2
Number of protein molecules per microparticle Mass per particle, 1% w/w loading Ratio, weight of protein to polymer
79.10 79.10 37.90 A P43212-tetragonal, Z ¼ 8 ¼ 7.9 1014 cm2 3.1 108 cm2/7.9 1014 cm2 ¼ 392000 (lying flat) or ¼ 818000 (on end) (392000)(14296 g/NA) ¼ 9.3 1015 g 9.3 1015 g/7.9 1013 g ¼ 1.2%
Protein (Lysozyme)—Ref. 28
2
Surface area, 4pr
Figure 3. Classic linear Langmuir adsorption plot for a typical protein (Lysozyme) on PLG microparticles.
2
concentration was kept constant at 0.05% w/w PLG, but differences in polymer viscosity (polymer concentration, molecular length, end groups), shear force, and emulsification time resulted in a range of mean diameters from 0.42 to 1.18 mm. The particle surfaces had virtually identical surface properties such as zeta potential, and their compositional similarity and surface area differences made them good candidates to measure asymptotic protein adsorption. The adsorption
Property
Figure 2. Adsorption Isotherms for various proteins on anionic PLG microparticles under different buffer conditions.
Polymer Microparticle
CHESKO ET AL.
Table 2. Estimated Protein Load Adsorbed on PLG Surface Based Upon Geometric Considerations and Formation of a Monolayer
2514
ADSORPTION OF PROTEIN ANTIGENS
2515
scales with the available surface area (i.e., protein adsorption sites).
Electrostatic Interactions Are a Significant Driving Force for Adsorption to PLG Microparticles
Figure 4. Adsorption of Lysozyme on PLG microparticles of identical weight but different surface area.
characteristics of lysozyme on these 50:50 L/G PLG microparticles with roughly a factor of three difference in surface area (inversely proportional to mean diameter for a fixed solid content or total particle weight) is shown in Figure 4. The noticeable increase in protein adsorption efficiency and capacity for the smaller particles follows the available surface area nearly quantitatively (a geometric ratio of 1.00:1.71:2.87 gives absorption capacities of 1.00:1.67:3.08). The slight discrepancies at higher protein concentrations are attributable to a limited formation of multilayer surface structures that have been identified by other investigators.17 The pronounced drop in adsorption efficiency reflects the large decrease in affinity of the relatively uniform protein monolayer to adsorb addition protein molecules. The implications of the surface area limiting behavior are potentially significant. If much smaller particles such as nanoparticles could be synthesized with desirable properties such as aggregation stability, the ability to deliver higher protein loads per weight of polymer may be realized. We have observed the same phenomenon with PLG microparticles made of different sizes due to variations in surfactant concentration.26 To illustrate, 100 nm PLG nanoparticles would be predicted to adsorb about 10% w/w protein to polymer, since the protein adsorption capacity
The binding interactions of proteins to adjuvants such as aluminum salt adjuvants are often governed by forces including electrostatic interactions between the ionic salts (aluminum phosphate and hydroxyls) and the protein.27 To facilitate binding to both acidic and basic proteins, the phosphate content of aluminum hydroxy phosphate must be adjusted to shift the isoelectric point to a favorable position. In contrast to aluminum salt adjuvants, PLG microparticles have surface functional groups, such as aliphatic chains and ester linkages, in addition to ionizable groups such as carboxyls and hydroxyls, that will carry a net negative charge at physiological pH, which are confirmed by the zeta potential measurements. If electrostatic interactions dominate the adsorption characteristics of proteins to microparticles, a direct comparison of the protein isoelectric point (pI) and the buffer solution pH will explain all variations observed. Looking at Table 3, we see that for the acidic protein ovalbumin (pI ¼ 4.6), a buffer close to this pI is needed for protein adsorption to occur. From the equilibrium equation HA ! A þ Hþ Ka ¼ ½A ½Hþ =½HA; For pH controlled by a large excess of buffering species, i.e., [Hþ] & 10pH(buffer) The ratio of dissociated to undissociated protein species is proportional to the 10th power exponent of the difference between buffer pH and protein pI ½A =½HA ¼ 10ðpIpHÞ The data in Table 3 suggests that the amount of protein released rapidly (initial release, following lyophilization) is complementary to the adsorption affinity. Moreover, the amount of protein adsorbed at a given pH tends to follow the fraction of net charge neutral molecules (zwitterions). The driving force of solvation is undergoing rapid changes through a solubility minima at this pH, contributing to the net driving force precipitating the protein from solution.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2516
CHESKO ET AL.
Table 3. Effect of Buffers and Salt Concentration on Adsorption of Proteins
Protein Ovalbumin
Lysozyme
Carbonic Anhydrase
Lysozyme
pH (0.05)
Buffer (10 mM)
Adsorption Efficiency (%)
5.0 7.0 9.0 5.0 7.0 9.0 5.0 7.0 9.0 5.0 5.0 5.0
Histidine Phosphate Borate Citrate Phosphate Borate Histidine Phosphate Borate Citrate þ 60 mM NaCl Citrate þ 250 mM NaCl Citrate þ 800 mM NaCl
63 9 15 4 12 68 3 100 1 95 1 87 2 20 2 22 57 3 37 2 16 2
Additional Forces Such as van der Waals Interactions Contribute to Adsorption Behavior Proteins are heteropolymeric species, with an ensemble of acidic residues (Aspartic and Glutamic Acid), near-neutral residues (Histidine), and basic residues (Lysine, Arginine) that equilibrate to the buffer pH by a combination of accepting and donating protons, folding, and adopting a structure whose internal microenvironments are thermodynamically favorable and kinetically accessible. It is expected that polar and charged residues will show a propensity to migrate to the solvent accessible surface of the protein to interact strongly with surrounding water molecules. From Table 3 we see for the acidic protein ovalbumin that the amount of protein adsorbed closely tracks the fraction of near-neutral ovalbumin at the buffered pH of interest, that is, electrostatic attraction is not driving the adsorption process. The pI of ovalbumin4,6 is lower than the two buffer conditions tested, so the electrostatic interaction between the PLG microparticles and the ionized protein will be repulsive due to the common, net negative charge that both species carry. Because some protein adsorption does occur, nonColoumbic forces, such as van der Waals interactions are strong enough to overcome the charge repulsion. Since it is generally accepted that the solubility of proteins in aqueous solution goes through a minimum at the isoelectric point, resulting primarily from the loss of favorable interactions between water molecules and the ions they solvate, binding interactions occurring at the protein-polymer interface are expected to compete most effectively near the pI of the protein.
% (Neutral and Positively Charged % 2 h Release Protein) to Total In vitro 29 0.3 0.0 100 99.9 98.0 90 10 0.1 100 100 100
51 7 89 5 99 2 65 3 40 2 52 30 4 85 3 98 25 72 5 86 4 93 2
For the basic protein lysozyme (pI of 10.7) the charge effect is opposite, following the change of sign in charge from positive (i.e., attractive with respect to the PLG) at middle pH values to uncharged in high pH solutions. Since the electrostatic interaction is attractive due to unlike signs of the protein (positive) and PLG (negative), the charged lysozyme species are in excess and bind with high affinity to the microparticle surface. This is confirmed by the large positive changes measured in Table 1 following adsorption (PLG microparticle zeta potential). The apparent increase in binding affinity at higher pH may be a result of greater negative charge density on the PLG particles that compensates for the slightly reduced protein charge. This effect appears to have a dramatic impact on the initial release measurement, which shows the release a larger fraction of protein than for ovalbumin. We would expect this trend since the primary interaction facilitating binding is now electrostatic. Lysozyme is a relatively small (17 kDa) rigid protein, with an electrostatically charged outer surface, that will have strong Coulomb interactions at the microparticle interface, with less opportunity for structural accommodation and hydrophobic interactions. The effect of salt concentration on adsorption is very strong, as shown by the dramatic decrease in lysozyme binding with increasing solution ionic strength (Tab. 3) The higher salt concentration screens the charges more effectively, decreasing the strength of the Coulomb interaction (and Debye length) significantly, and reduces the electrostatic driving force for adsorption. For a protein of intermediate isoelectric point such as Carbonic Anhydrase (pI 6.0) the contribu-
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
2517
tions from electrostatic forces seem less dominant. At pH 7 the protein is primarily anionic or neutral in charge, yet 20% adsorption is observed. At lower pH (5.0) the adsorption grows to almost 90%, closely following the percentage of net cationic molecules. The amount of protein released in 2 h is complementary to the fraction adsorbed. The lower adsorption efficiency of CAN at small loading levels and its gradual saturation curve suggest that it may be capable of dimerizing or selfassociating with a comparable affinity to the PLGprotein interaction. Electrostatic Interactions May be Directly Measured and Correlate with Molecular Adsorption Events Measurement of surface charge on PLG before and after protein adsorption is a direct observation of the electrostatic attractive forces present during the binding process. The change in zeta potential following the adsorption of protein is indicative of charge neutralization (e.g., salt bridge formation) at the protein-polymer interface. We observed a direct correlation of the change in zeta potential on the microparticles to the binding efficiency of the protein, demonstrating that electrostatics drive the bound to free equilibrium strongly towards adsorption onto the microparticles (Tab. 1). It is also evident that the proteins can overcome repulsive electrostatic interactions and still adsorb, but to a smaller extent than near either the isoelectric point or at a lower pH where the protein carries a net positive (attractive) charge. The plot of the logarithm of the adsorbate surface concentration divided by the liquid phase equilibrium concentration versus the change in zeta potential (Fig. 5) shows a rough linear correlation over physiological pH values (5 pH 9) and across a set of eight diverse proteins. The slope of the curve gives the electrostatic contribution to the variance in binding, while the intercept measures the affinity at the point of zero charge. Each individual protein shows a tight linear dependence, suggesting that electrostatics play a dominant role in driving adsorption, but additional factors such as nonColoumbic (London, van der Waals, hydrophobic interaction) forces also play a role. In conclusion, by varying buffer, pH, and ionic strength the electrostatic forces driving protein adsorption may be altered to achieve substantial surface binding to PLG microparticles. Measure-
Figure 5. Correlation between adsorption efficiency and change in microparticle surface charge for eight proteins over a range of physiologically relevant pH conditions (5 pH 9).
ments of surface morphology by AFM and surface charge (zeta potential) support the assertion that a relatively uniform monolayer of protein is formed. The adsorption capacity of the particles can be estimated from geometric consideration of interfacial surface areas of the particles and protein molecules. Near the protein isoelectric point the strength of the electrostatic interaction greatly diminishes and other forces that we will refer to as van der Waals attraction becomes more dominant, though their contribution may also be significant at other pH values. The surface adsorption and desorption process has favorable characteristics for allowing the efficient binding, transport and release of protein antigens, while preserving immunogenic structural elements of these molecules.24
SUMMARY Proteins can be efficiently adsorbed on anionic PLG microparticles through controlling conditions such as solution pH and ionic strength. Significant changes in the zeta potential and microscopic morphology confirm that both model and vaccine relevant proteins are adsorbed at the surface through electrostatic and van der Waals type interactions to form a relatively uniform monolayer. This mechanism for delivering protein antigens in vaccines may provide a favorable microparticle delivery system that efficiently carries recombinant proteins into antigen presenting cells while retaining structural features that confer immunogenicity.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2518
CHESKO ET AL.
ACKNOWLEDGMENTS Atomic force microscopy images were obtained with the valuable assistance of David Sampson of Veeco’s Metrology Application Group. We wish to acknowledge the help of Indresh Srivastava and Elaine Kan for providing the HIV-env (gp120dV2) protein and with many helpful discussions about its properties. We also acknowledge the laboratory help of Elawati Soenawan in assisting with formulation characterization.
REFERENCES 1. O’Hagan D, Rappuoli R. 2004. Novel approaches to vaccine delivery. Pharm Res 9:1519–1530. 2. Singh M, O’Hagan DT. 2002. Recent advances in vaccine adjuvants. Pharm Res 19:715–728. 3. Gupta RK, Siber GR. 1995. Adjuvants for human vaccines–current status, problems and future prospects. Vaccine 13:1263–1276. 4. O’Hagan DT, Valiante NM. 2003. Recent advances in the discovery and delivery of vaccine adjuvants. Nat Rev Drug Discov 2:727–735. 5. Singh M, Kazzaz J, Chesko J, Soenawan E, Ugozzoli M, Giuliani M, Pizza M, Rappuoli R, O’Hagan DT. 2004. Anionic microparticles are a potent delivery system for recombinant antigens from Neisseria meningitidis serotype B. J Pharm Sci 93:273–282. 6. Kazzaz J, Neidleman J, Singh M, Ott G, O’Hagan DT. 2000. Novel anionic microparticles are a potent adjuvant for the induction of cytotoxic T lymphocytes against recombinant p55 gag from HIV-1. J Control Release 67:347–356. 7. Emerich DF, Tracy MA, Ward KL, Figueiredo M, Qian R, Henschel C, Bartus RT. 1999. Biocompatibility of poly (DL-lactide-co-glycolide) microspheres implanted into the brain. Cell Transplant 8:47–58. 8. Jones AJ, Putney S, Johnson OL, Cleland JL. 1997. Recombinant human growth hormone poly(lacticco-glycolic acid) microsphere formulation development. Adv Drug Deliv Rev 28:71–84. 9. Thomas TT, Kohane DS, Wang A, Langer R. 2004. Microparticulate formulations for the controlled release of interleukin-2. J Pharm Sci 93:1100– 1109. 10. Seeber SJ, White JL, Hem SL. 1991. Predicting the adsorption of proteins by aluminium-containing adjuvants. Vaccine 9:201–203. 11. Callahan PM, Shorter AL, Hem SL. 1991. The importance of surface charge in the optimization of antigen-adjuvant interactions. Pharm Res 8:851– 858.
12. Chang MF, White JL, Nail SL, Hem SL. 1997. Role of the electrostatic attractive force in the adsorption of proteins by aluminum hydroxide adjuvant. PDA J Pharm Sci Technol 51: 25 –29. 13. Cleland JL. 1995. Design and production of singleimmunization vaccines using polylactide polyglycolide microsphere systems. Pharm Biotechnol 6:439–462. 14. O’Hagan DT. 1997. Prospects for the development of new and improved vaccines through the use of microencapsulation technology. In: Levine MM, Woodrow GC, Kaper JB, Cobon GS, editors. New generation vaccines, edn. New York: Marcel Dekker, Inc. pp 215–228. 15. Vogel FR, Hem SL, editors. 2003. Vaccines, 4 edn. New York: Sanders. p 1 ff. 16. Singh M, Carlson JR, Briones M, Ugozzoli M, Kazzaz J, Barackman J, Ott G, O’Hagan D. 1998. A comparison of biodegradable microparticles and MF59 as systemic adjuvants for recombinant gD from HSV-2. Vaccine 16:1822–1827. 17. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces I. Adsorption isotherms. Effects of charge, ionic strength and temperature. J Colloidal and Interface Science 66:257–265. 18. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces II. Hydrogen ion titrations. J Colloidal and Interface Science 66:266–276. 19. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces III. Electrophoresis. J Colloidal and Interface Science 66:277–284. 20. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces V. Microcalorimetry. J Colloidal and Interface Science 66:295–302. 21. Van Norde W, MacRitchie F, Nowicka G, Lyklema J. 1986. Protein adsorption at solid-liquid interfaces: reversibility and conformation aspects. J Colloid Interface Sci 112:447–456. 22. Horbett TA, Brash JL, editors. 1995. Proteins at Interfaces II, Fundamentals and Applications. ACS Symposium Series 60, chap. 1–38. 23. Pizza M, Scarlato V, Masignani V, Giuliani MM, Arico B, Comanducci M, Jennings GT, Baldi L, Bartolini E, Capecchi B, Galeotti CL, Luzzi E, Manetti R, Marchetti E, Mora M, Nuti S, Ratti G, Santini L, Savino S, Scarselli M, Storni E, Zuo P, Broeker M, Hundt E, Knapp B, Blair E, Mason T, Tettelin H, Hood DW, Jeffries AC, Saunders NJ, Granoff DM, Venter JC, Moxon ER, Grandi
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
G, Rappuoli R. 2000. Identification of vaccine candidates against serogroup B meningococcus by whole-genome sequencing. Science 287:1816– 1820. 24. Singh M, Chesko J, Kazzaz J, Ugozzoli M, Kan E, Srivastava I, O’Hagan D. 2004. Adsorption of a novel recombinant glycoprotein from HIV (Env gp120dV2SF162) to anionic PLG microparticles retains the structural integrity of the protein, while encapsulation in PLG microparticles does not. Pharm Res 21:2148–2152.
2519
25. Martin A. 1993. Physical Pharmacy, 4 edn. Philadelphia: Lea and Febiger. p 1 ff. 26. Singh-Jasuja H, Hilf N, Arnold-Schild D, Schild H. 2001. The role of heat shock proteins and their receptors in the activation of the immune system. Biol Chem 382:629–636. 27. Iyer S, Robinett RS, HogenEsch H, Hem SL. 2004. Mechanism of adsorption of hepatitis B surface antigen by aluminum hydroxide adjuvant. Vaccine 22:1475–1479.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
Received 18 March 2005; accepted 15 July 2005 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20472
ABSTRACT: This work examines physico-chemical properties influencing protein adsorption to anionic PLG microparticles and demonstrates the ability to bind and release vaccine antigens over a range of loads, pH values, and ionic strengths. Poly(lactide-co-glycolide) microparticles were synthesized by a w/o/w emulsification method in the presence of the anionic surfactant DSS (dioctyl sodium sulfosuccinate). Ovalbumin (OVA), carbonic anhydrase (CAN), lysozyme (LYZ), lactic acid dehydrogenase, bovine serum albumin (BSA), an HIV envelope glyocoprotein, and a Neisseria meningitidis B protein were adsorbed to the PLG microparticles, with binding efficiency, initial release and zeta potentials measured. Protein (antigen) binding to PLG microparticles was influenced by both electrostatic interaction and other mechanisms such as van der Waals forces. The protein binding capacity was directly proportional to the available surface area and may have a practical upper limit imposed by the formation of a complete protein monolayer as suggested by AFM images. The protein affinity for the PLG surface depended strongly on the isoelectric point (pI) and electrostatic forces, but also showed contributions from nonCoulombic interactions. Protein antigens were adsorbed on anionic PLG microparticles with varying degrees of efficiency under different conditions such as pH and ionic strength. Observable changes in zeta potentials and morphology suggest the formation of a surface monolayer. Antigen binding and release occur through a combination of electrostatic and van der Waals interactions occurring at the polymer-solution interface. ß 2005 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 94:2510–2519, 2005
Keywords: microparticles; antigen; vaccine delivery; polylactide-co-glycolide; protein adsorption; PLG; anionic particles; electrostatic; microscopy
INTRODUCTION Next generation vaccines will be comprised mainly of recombinant protein antigens that are often poorly immunogenic. These vaccines will need optimal antigen delivery systems with or without an additional immunopotentiator to generate potent immune responses.1,2 Emulsions, micro/nanoparticles, aluminum salt adjuvants,
Correspondence to: James Chesko (Telephone: 510-9233896; Fax: 510-923-2586; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 94, 2510–2519 (2005) ß 2005 Wiley-Liss, Inc. and the American Pharmacists Association
2510
and iscoms have all been used as delivery systems to boost immune responses to vaccine antigens. Enhanced immune responses are induced through a variety of mechanisms, including increased persistence of antigen at the site of injection, improved targeting to antigen presenting cells, or enhanced protection of antigen linear, and conformational epitopes against degradation.2–4 We have previously described studies which showed that anionic poly(lactide-co-glycolide) (PLG) microparticles were effective delivery systems for adsorbed vaccine antigens, including recombinant proteins from Neisseria meningitidis type B (MB)5 and human immunodeficiency virus
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
(HIV).6 The PLG polymer used in these formulations has a long history of safe use in humans.7–9 Work with the traditional aluminum salt adjuvants has elucidated rules for predicting protein adsorption,10 encouraging adjuvant-antigen interactions,11 and highlighted the significance of surface electrostatic forces in antigen adsorption and release.12 Anionic PLG microparticles13,14 represent an alternative approach to aluminum salt adjuvants,15 or the MF5916 emulsion adjuvant. To assess the strengths and limitations of anionic PLG microparticles as a delivery system for diverse antigens, we have measured the physico-chemical properties such as charge and ionic strength that have been implicated as important to the description of binding and release of a range of proteins from polymer surfaces.17–20 In addition to vaccine relevant protein antigens, model proteins with a wide range of isoelectric points (pI) including ovalbumin (pI 4.6), carbonic anhydrase (pI 6.0), and lysozyme (pI 10.7) were evaluated, to represent proteins with diverse electrostatic properties. Evaluation of the physicochemical properties of anionic PLG microparticles, including size, charge, and surface morphology may provide information that can be used to guide formulation strategies for diverse proteins. The complex behavior of protein on solid surfaces has been studied by various techniques21,22 to suggest the formation of structures including multilayers. Using atomic force microscopy to image the microparticles before and after protein adsorption, we visualized the surface morphology to determine if it was consistent with our hypothesis of the formation of an adsorbed protein monolayer. Furthermore, we investigated the physico-chemical properties influencing protein adsorption to microparticles and demonstrated their ability to bind and release diverse protein antigens over a range of loads, pH, and ionic strengths.
EXPERIMENTAL Materials RG503 and RG502H, poly(D,L-lactide-co-glycolide) 50:50 lactide to glycolide co-polymers were obtained from Boehringer Ingelheim (Petersburg, VA). Dioctylsulfosuccinate (DSS), lysozyme (LYZ), carbonic anhydrase (CAN), bovine serum albumin (BSA), ovalbumin (OVA), lactate dehydrogenase (LADH), USP grade mannitol, sucrose, and tre-
2511
halose came from Sigma-Aldrich Chemical (St. Louis, MO) Chinese Hamster Ovary derived recombinant gp120dV2 was synthesized and purified in house (Chiron Vaccine Research in Emeryville, CA) Escherichia coli (E. Coli)-derived recombinant meningococcal proteins (MB1 and MB2) were obtained from Chiron Vaccines, Siena (Siena, Italy) and were isolated and purified as previously described.23
Microparticle Preparation Microparticles were prepared by a solvent evaporation technique. A 10 mL methylene chloride solution of 6% w/v polymer with 2.5 mL PBS was homogenized using a 10 mm probe, (UltraTurrax, T25 IKA-Labortechnik, Germany) forming water in oil emulsion which was then added to 50 mL of distilled water containing 6 mg/mL DSS and homogenized at very high speed using a homogenizer with a 20-mm probe (ES-15 Omni International, GA) for 25 min in an ice-water bath. This resulted in water in oil in water emulsion, which was stirred at 1000 rpm for 12 h at room temperature as the methylene chloride was allowed to evaporate.
Emulsion Characterization The size distribution of the microparticles was determined using a particle size analyzer (Master sizer, Malvern Instruments, UK). The electrokinetic mobility of the PLG microparticles was measured in PBS buffer on a Malvern ZetaSizer (Malvern Instruments, UK) and the zeta potential (roughly equivalent to particle surface charge) determined. Model proteins with a range of pI were chosen to span a range of buffer adsorption conditions: lysozyme with pI 10.7, lactic acid dehydrogenase with pI 6.8, carbonic anhydrase with pI 6.0 and ovalbumin with pI 4.6. The PLG content of the suspension was measured by aliquoting 1 mL of the suspension into preweighed vials, which were lyophilized and weighed again and the average net weight change was used as PLG content/1 mL suspension. Atomic force microscopy was employed in the noncontact mode using a Digital Instruments BioScope (Digital Instruments, Santa Barbara, CA). The polymer suspension was imaged before and after exposure to protein (1% weight protein to polymer, overnight rocking at 48C, then sample and control lyophilized overnight). The powder
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2512
CHESKO ET AL.
samples were dispersed on a flat, adhesive surface, with representative sampling of numerous (>10) particle populations both within the preparation exposed to protein and the control not exposed to protein.
2 mL of 5% SDS-0.2 M sodium hydroxide solution at room temperature and measuring protein concentration by BCA and HPLC-SEC.
RESULTS AND DISCUSSION Protein Adsorption and Release Microparticles with protein adsorbed for release studies were made by combining a suspension containing 100 mg of PLG with 1 mg protein in 10 mL total volume of 10 mM of the appropriate buffer and left overnight on a lab rocker (aliquot mixer, Miles Laboratories) at 48C. Particles were incubated with the proteins OVA, BSA, CAN, LYS, MB1, MB2, and gp120 with a variety of buffer conditions and target loads ranging from 0 to 10% w/w polymer. Buffers used for pH 5 and 5.5 were histidine and citrate, for pH 7 phosphate, and pH 9 borate. Samples to be used for protein release studies were lyophilized after overnight adsorption (16–18 h). To determine the amount of adsorbed protein, the microparticles were separated from the incubation medium before the final lyophilization step by centrifugation, and the pellet washed three times with distilled water, then lyophilized. The loading level of protein adsorbed to the microparticles was determined by dissolving 10 mg of the microparticles in 2 mL of 5% SDS-0.2 M sodium hydroxide solution at room temperature followed by the bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL) using BSA concentration standards. To confirm that the BSA standard was appropriate, representative samples were analyzed for protein integrity and concentration by size exclusion chromatography (HPLC-SEC method, Tosoh-BioSep SW3000XL 4.6 mm 30 cm column, in PBS) on the supernatant fraction with 90%þ mass balance achieved. Uncertainties in concentration determinations were estimated by combining variances from replicates (intra batch), different microparticle (inter batch) and serial dilution multiple determinations from the BCA determination. To determine the amount of protein released over a 2 h time period was measured as follows. Twenty-five milligrams of lyophilized microparticles following protein adsorption was incubated with 1 mL of the appropriate buffer at 378C for 2 h. The amount of protein present in the supernatant was determined by BCA. To determine the % of 2-h release, the total protein load was measured by dissolving 10 mg of lyophilized microparticles in
Microparticles were prepared with a typical mean size of 1 mm (size distribution of 0.4–1.9 mm, with polydispersity about 50%). Atomic force imaging and scanning electron micrographs24 taken before and after protein adsorption showed these microparticles to be spherical in shape, with a smooth outer surface (Fig. 1). The AFM phase image showed greater contrast and detail than the topographic profile. Morphological differences measured between the untreated and protein treated samples were attributed to the presence and interactions of the protein on the polymer surface. The phase contrast AFM images show notable differences in surface features, with the addition
Figure 1. AFM topography (left) and phase images (right) of dry PLG microparticles before (above) and after (below) overnight gp120dV2 protein adsorption. Microparticles were air dried and imaged under ambient conditions.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
of protein resulting in nearly contiguous islands about 50 nm wide and 5 nm in height that are attributed to clusters of proteins approximately one molecular layer thick. The PLG/DSS microparticles had a zeta potential of 55 mV in 10 mM pH 5.0 citrate buffer, while PLG/DSS particles with 1% w/w adsorbed gp120dV2 protein had a zeta potential in the same buffer of 8 mV, indicating that charge neutralization at the protein/polymer interface had occurred as a result of protein adsorption (Tab. 1). The proteins studied show a general trend of increasing (less negative) zeta potential as additional protein was adsorbed, as would be predicted for a potentiometric titration of the anionic microparticles with the protein. In the cases where the protein carried a net negative charge, the adsorption was generally inefficient and the Stern layer structure may involve inclusion of positive counter ions and solvation molecules. Pseudotherms Show Site Saturation Behavior for Surface Adsorption and the Formation of a Langmuir Monolayer Figure 2 shows adsorption pseudotherms for the model proteins lysozyme, bovine serum albumin, and carbonic anhydrase in addition to vaccine antigens, HIV-envelope protein gp120dV2 and two Meningococcus B proteins MB1 and MB2 on the PLG microparticles. As expected for an adsorption process, binding efficiency decreases with increasing protein input and saturated as the available adsorption sites become completely occupied. This curve profile has been consistently reproduced for the numerous proteins that we have studied, and empirically finds an asymptote between 1 and 2 % w/w protein to polymer. For some species such as carbonic anhydrase and the two MenB proteins, the asymptotic saturation is less distinct. A simplistic treatment of the protein
2513
adsorption data is explained quite adequately by a linear Langmuir plot25 of a ratio of aqueous protein concentration divided by surface adsorbed protein versus aqueous protein concentration, supporting the description that a monolayer of adsorbed protein is formed (Fig. 3). The PLG surface fully covered with a protein monolayer has a reduced affinity to adsorb additional protein as compared to the polymer/surfactant interface with sites unoccupied by protein. For proteins adsorbing close to their isoelectric point, binding of the primary monolayer appears weaker, perhaps due to reduced electrostatic forces, but subsequent layer formation is more facile as a result of reduced charge repulsion. Protein Adsorption Capacity May be Predicted from the Microparticle Surface Area The existence of an upper limit (maximum binding capacity) for adsorption is consistent with the idea that a monolayer (finite, defined surface area available for adsorption) may exist. We may calculate an estimated value for this loading level by comparing geometric dimensions of the microparticles and the protein lysozyme by a geometric calculation (see Tab. 2) that predicts approximately a 1.2% protein to polymer ratio for 1 mm PLG microparticles. Calculations with albumin (not shown) yield a similar load value (1.5% w/w), suggesting that for globular proteins that retain their shape, an upper loading limit of 1–2% w/w will represent the accessible PLG surface area for adsorption of a protein monolayer. Accommodation and aggregation of protein could alter this value, but the data in Figure 1 suggests that this trend is often well observed. To test the available surface area model, PLG microparticles with constant lactide to glycolide (50% Lactide:50% Glycolide) ratio were synthesized with a range of sizes. The surfactant (DSS)
Table 1. Surface Charge on PLG Microparticles before and after Protein Adsorption
PLG particles PLG þ BSA PLG þ OVA PLG þ CAN PLG þ LYS PLG þ gp120dV2
x, mV, pH 5.0 (Citrate)
x, mV, pH 7.0 (PBS)
x, mV, pH 9.0 (Borate)
Max Dx, mV
% Binding at Max Dx
55 2 40 2 49 2 25 2 9 2 8 2
28 2 23 2 23 2 26 2 24 2 29 2
82 2 81 2 82 2 80 2 25 2 37 2
þ15 3 þ 53 þ30 3 þ57 3 þ47 3
48 5 15 3 87 3 100 2 98 2
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
4/3pr3r ¼ 4p(0.5 104 cm)3 (1.5 g/cm3) ¼ 7.9 1013 g
N/A
4pr ¼ 4(p)(0.5 104 cm)2 ¼ 3.1 108 cm2
Number of protein molecules per microparticle Mass per particle, 1% w/w loading Ratio, weight of protein to polymer
79.10 79.10 37.90 A P43212-tetragonal, Z ¼ 8 ¼ 7.9 1014 cm2 3.1 108 cm2/7.9 1014 cm2 ¼ 392000 (lying flat) or ¼ 818000 (on end) (392000)(14296 g/NA) ¼ 9.3 1015 g 9.3 1015 g/7.9 1013 g ¼ 1.2%
Protein (Lysozyme)—Ref. 28
2
Surface area, 4pr
Figure 3. Classic linear Langmuir adsorption plot for a typical protein (Lysozyme) on PLG microparticles.
2
concentration was kept constant at 0.05% w/w PLG, but differences in polymer viscosity (polymer concentration, molecular length, end groups), shear force, and emulsification time resulted in a range of mean diameters from 0.42 to 1.18 mm. The particle surfaces had virtually identical surface properties such as zeta potential, and their compositional similarity and surface area differences made them good candidates to measure asymptotic protein adsorption. The adsorption
Property
Figure 2. Adsorption Isotherms for various proteins on anionic PLG microparticles under different buffer conditions.
Polymer Microparticle
CHESKO ET AL.
Table 2. Estimated Protein Load Adsorbed on PLG Surface Based Upon Geometric Considerations and Formation of a Monolayer
2514
ADSORPTION OF PROTEIN ANTIGENS
2515
scales with the available surface area (i.e., protein adsorption sites).
Electrostatic Interactions Are a Significant Driving Force for Adsorption to PLG Microparticles
Figure 4. Adsorption of Lysozyme on PLG microparticles of identical weight but different surface area.
characteristics of lysozyme on these 50:50 L/G PLG microparticles with roughly a factor of three difference in surface area (inversely proportional to mean diameter for a fixed solid content or total particle weight) is shown in Figure 4. The noticeable increase in protein adsorption efficiency and capacity for the smaller particles follows the available surface area nearly quantitatively (a geometric ratio of 1.00:1.71:2.87 gives absorption capacities of 1.00:1.67:3.08). The slight discrepancies at higher protein concentrations are attributable to a limited formation of multilayer surface structures that have been identified by other investigators.17 The pronounced drop in adsorption efficiency reflects the large decrease in affinity of the relatively uniform protein monolayer to adsorb addition protein molecules. The implications of the surface area limiting behavior are potentially significant. If much smaller particles such as nanoparticles could be synthesized with desirable properties such as aggregation stability, the ability to deliver higher protein loads per weight of polymer may be realized. We have observed the same phenomenon with PLG microparticles made of different sizes due to variations in surfactant concentration.26 To illustrate, 100 nm PLG nanoparticles would be predicted to adsorb about 10% w/w protein to polymer, since the protein adsorption capacity
The binding interactions of proteins to adjuvants such as aluminum salt adjuvants are often governed by forces including electrostatic interactions between the ionic salts (aluminum phosphate and hydroxyls) and the protein.27 To facilitate binding to both acidic and basic proteins, the phosphate content of aluminum hydroxy phosphate must be adjusted to shift the isoelectric point to a favorable position. In contrast to aluminum salt adjuvants, PLG microparticles have surface functional groups, such as aliphatic chains and ester linkages, in addition to ionizable groups such as carboxyls and hydroxyls, that will carry a net negative charge at physiological pH, which are confirmed by the zeta potential measurements. If electrostatic interactions dominate the adsorption characteristics of proteins to microparticles, a direct comparison of the protein isoelectric point (pI) and the buffer solution pH will explain all variations observed. Looking at Table 3, we see that for the acidic protein ovalbumin (pI ¼ 4.6), a buffer close to this pI is needed for protein adsorption to occur. From the equilibrium equation HA ! A þ Hþ Ka ¼ ½A ½Hþ =½HA; For pH controlled by a large excess of buffering species, i.e., [Hþ] & 10pH(buffer) The ratio of dissociated to undissociated protein species is proportional to the 10th power exponent of the difference between buffer pH and protein pI ½A =½HA ¼ 10ðpIpHÞ The data in Table 3 suggests that the amount of protein released rapidly (initial release, following lyophilization) is complementary to the adsorption affinity. Moreover, the amount of protein adsorbed at a given pH tends to follow the fraction of net charge neutral molecules (zwitterions). The driving force of solvation is undergoing rapid changes through a solubility minima at this pH, contributing to the net driving force precipitating the protein from solution.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2516
CHESKO ET AL.
Table 3. Effect of Buffers and Salt Concentration on Adsorption of Proteins
Protein Ovalbumin
Lysozyme
Carbonic Anhydrase
Lysozyme
pH (0.05)
Buffer (10 mM)
Adsorption Efficiency (%)
5.0 7.0 9.0 5.0 7.0 9.0 5.0 7.0 9.0 5.0 5.0 5.0
Histidine Phosphate Borate Citrate Phosphate Borate Histidine Phosphate Borate Citrate þ 60 mM NaCl Citrate þ 250 mM NaCl Citrate þ 800 mM NaCl
63 9 15 4 12 68 3 100 1 95 1 87 2 20 2 22 57 3 37 2 16 2
Additional Forces Such as van der Waals Interactions Contribute to Adsorption Behavior Proteins are heteropolymeric species, with an ensemble of acidic residues (Aspartic and Glutamic Acid), near-neutral residues (Histidine), and basic residues (Lysine, Arginine) that equilibrate to the buffer pH by a combination of accepting and donating protons, folding, and adopting a structure whose internal microenvironments are thermodynamically favorable and kinetically accessible. It is expected that polar and charged residues will show a propensity to migrate to the solvent accessible surface of the protein to interact strongly with surrounding water molecules. From Table 3 we see for the acidic protein ovalbumin that the amount of protein adsorbed closely tracks the fraction of near-neutral ovalbumin at the buffered pH of interest, that is, electrostatic attraction is not driving the adsorption process. The pI of ovalbumin4,6 is lower than the two buffer conditions tested, so the electrostatic interaction between the PLG microparticles and the ionized protein will be repulsive due to the common, net negative charge that both species carry. Because some protein adsorption does occur, nonColoumbic forces, such as van der Waals interactions are strong enough to overcome the charge repulsion. Since it is generally accepted that the solubility of proteins in aqueous solution goes through a minimum at the isoelectric point, resulting primarily from the loss of favorable interactions between water molecules and the ions they solvate, binding interactions occurring at the protein-polymer interface are expected to compete most effectively near the pI of the protein.
% (Neutral and Positively Charged % 2 h Release Protein) to Total In vitro 29 0.3 0.0 100 99.9 98.0 90 10 0.1 100 100 100
51 7 89 5 99 2 65 3 40 2 52 30 4 85 3 98 25 72 5 86 4 93 2
For the basic protein lysozyme (pI of 10.7) the charge effect is opposite, following the change of sign in charge from positive (i.e., attractive with respect to the PLG) at middle pH values to uncharged in high pH solutions. Since the electrostatic interaction is attractive due to unlike signs of the protein (positive) and PLG (negative), the charged lysozyme species are in excess and bind with high affinity to the microparticle surface. This is confirmed by the large positive changes measured in Table 1 following adsorption (PLG microparticle zeta potential). The apparent increase in binding affinity at higher pH may be a result of greater negative charge density on the PLG particles that compensates for the slightly reduced protein charge. This effect appears to have a dramatic impact on the initial release measurement, which shows the release a larger fraction of protein than for ovalbumin. We would expect this trend since the primary interaction facilitating binding is now electrostatic. Lysozyme is a relatively small (17 kDa) rigid protein, with an electrostatically charged outer surface, that will have strong Coulomb interactions at the microparticle interface, with less opportunity for structural accommodation and hydrophobic interactions. The effect of salt concentration on adsorption is very strong, as shown by the dramatic decrease in lysozyme binding with increasing solution ionic strength (Tab. 3) The higher salt concentration screens the charges more effectively, decreasing the strength of the Coulomb interaction (and Debye length) significantly, and reduces the electrostatic driving force for adsorption. For a protein of intermediate isoelectric point such as Carbonic Anhydrase (pI 6.0) the contribu-
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
2517
tions from electrostatic forces seem less dominant. At pH 7 the protein is primarily anionic or neutral in charge, yet 20% adsorption is observed. At lower pH (5.0) the adsorption grows to almost 90%, closely following the percentage of net cationic molecules. The amount of protein released in 2 h is complementary to the fraction adsorbed. The lower adsorption efficiency of CAN at small loading levels and its gradual saturation curve suggest that it may be capable of dimerizing or selfassociating with a comparable affinity to the PLGprotein interaction. Electrostatic Interactions May be Directly Measured and Correlate with Molecular Adsorption Events Measurement of surface charge on PLG before and after protein adsorption is a direct observation of the electrostatic attractive forces present during the binding process. The change in zeta potential following the adsorption of protein is indicative of charge neutralization (e.g., salt bridge formation) at the protein-polymer interface. We observed a direct correlation of the change in zeta potential on the microparticles to the binding efficiency of the protein, demonstrating that electrostatics drive the bound to free equilibrium strongly towards adsorption onto the microparticles (Tab. 1). It is also evident that the proteins can overcome repulsive electrostatic interactions and still adsorb, but to a smaller extent than near either the isoelectric point or at a lower pH where the protein carries a net positive (attractive) charge. The plot of the logarithm of the adsorbate surface concentration divided by the liquid phase equilibrium concentration versus the change in zeta potential (Fig. 5) shows a rough linear correlation over physiological pH values (5 pH 9) and across a set of eight diverse proteins. The slope of the curve gives the electrostatic contribution to the variance in binding, while the intercept measures the affinity at the point of zero charge. Each individual protein shows a tight linear dependence, suggesting that electrostatics play a dominant role in driving adsorption, but additional factors such as nonColoumbic (London, van der Waals, hydrophobic interaction) forces also play a role. In conclusion, by varying buffer, pH, and ionic strength the electrostatic forces driving protein adsorption may be altered to achieve substantial surface binding to PLG microparticles. Measure-
Figure 5. Correlation between adsorption efficiency and change in microparticle surface charge for eight proteins over a range of physiologically relevant pH conditions (5 pH 9).
ments of surface morphology by AFM and surface charge (zeta potential) support the assertion that a relatively uniform monolayer of protein is formed. The adsorption capacity of the particles can be estimated from geometric consideration of interfacial surface areas of the particles and protein molecules. Near the protein isoelectric point the strength of the electrostatic interaction greatly diminishes and other forces that we will refer to as van der Waals attraction becomes more dominant, though their contribution may also be significant at other pH values. The surface adsorption and desorption process has favorable characteristics for allowing the efficient binding, transport and release of protein antigens, while preserving immunogenic structural elements of these molecules.24
SUMMARY Proteins can be efficiently adsorbed on anionic PLG microparticles through controlling conditions such as solution pH and ionic strength. Significant changes in the zeta potential and microscopic morphology confirm that both model and vaccine relevant proteins are adsorbed at the surface through electrostatic and van der Waals type interactions to form a relatively uniform monolayer. This mechanism for delivering protein antigens in vaccines may provide a favorable microparticle delivery system that efficiently carries recombinant proteins into antigen presenting cells while retaining structural features that confer immunogenicity.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
2518
CHESKO ET AL.
ACKNOWLEDGMENTS Atomic force microscopy images were obtained with the valuable assistance of David Sampson of Veeco’s Metrology Application Group. We wish to acknowledge the help of Indresh Srivastava and Elaine Kan for providing the HIV-env (gp120dV2) protein and with many helpful discussions about its properties. We also acknowledge the laboratory help of Elawati Soenawan in assisting with formulation characterization.
REFERENCES 1. O’Hagan D, Rappuoli R. 2004. Novel approaches to vaccine delivery. Pharm Res 9:1519–1530. 2. Singh M, O’Hagan DT. 2002. Recent advances in vaccine adjuvants. Pharm Res 19:715–728. 3. Gupta RK, Siber GR. 1995. Adjuvants for human vaccines–current status, problems and future prospects. Vaccine 13:1263–1276. 4. O’Hagan DT, Valiante NM. 2003. Recent advances in the discovery and delivery of vaccine adjuvants. Nat Rev Drug Discov 2:727–735. 5. Singh M, Kazzaz J, Chesko J, Soenawan E, Ugozzoli M, Giuliani M, Pizza M, Rappuoli R, O’Hagan DT. 2004. Anionic microparticles are a potent delivery system for recombinant antigens from Neisseria meningitidis serotype B. J Pharm Sci 93:273–282. 6. Kazzaz J, Neidleman J, Singh M, Ott G, O’Hagan DT. 2000. Novel anionic microparticles are a potent adjuvant for the induction of cytotoxic T lymphocytes against recombinant p55 gag from HIV-1. J Control Release 67:347–356. 7. Emerich DF, Tracy MA, Ward KL, Figueiredo M, Qian R, Henschel C, Bartus RT. 1999. Biocompatibility of poly (DL-lactide-co-glycolide) microspheres implanted into the brain. Cell Transplant 8:47–58. 8. Jones AJ, Putney S, Johnson OL, Cleland JL. 1997. Recombinant human growth hormone poly(lacticco-glycolic acid) microsphere formulation development. Adv Drug Deliv Rev 28:71–84. 9. Thomas TT, Kohane DS, Wang A, Langer R. 2004. Microparticulate formulations for the controlled release of interleukin-2. J Pharm Sci 93:1100– 1109. 10. Seeber SJ, White JL, Hem SL. 1991. Predicting the adsorption of proteins by aluminium-containing adjuvants. Vaccine 9:201–203. 11. Callahan PM, Shorter AL, Hem SL. 1991. The importance of surface charge in the optimization of antigen-adjuvant interactions. Pharm Res 8:851– 858.
12. Chang MF, White JL, Nail SL, Hem SL. 1997. Role of the electrostatic attractive force in the adsorption of proteins by aluminum hydroxide adjuvant. PDA J Pharm Sci Technol 51: 25 –29. 13. Cleland JL. 1995. Design and production of singleimmunization vaccines using polylactide polyglycolide microsphere systems. Pharm Biotechnol 6:439–462. 14. O’Hagan DT. 1997. Prospects for the development of new and improved vaccines through the use of microencapsulation technology. In: Levine MM, Woodrow GC, Kaper JB, Cobon GS, editors. New generation vaccines, edn. New York: Marcel Dekker, Inc. pp 215–228. 15. Vogel FR, Hem SL, editors. 2003. Vaccines, 4 edn. New York: Sanders. p 1 ff. 16. Singh M, Carlson JR, Briones M, Ugozzoli M, Kazzaz J, Barackman J, Ott G, O’Hagan D. 1998. A comparison of biodegradable microparticles and MF59 as systemic adjuvants for recombinant gD from HSV-2. Vaccine 16:1822–1827. 17. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces I. Adsorption isotherms. Effects of charge, ionic strength and temperature. J Colloidal and Interface Science 66:257–265. 18. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces II. Hydrogen ion titrations. J Colloidal and Interface Science 66:266–276. 19. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces III. Electrophoresis. J Colloidal and Interface Science 66:277–284. 20. Van Norde W, Lyklema J. 1978. The adsorption of human plasma albumin and bovine pancreas ribonuclease at negatively charged polystyrene surfaces V. Microcalorimetry. J Colloidal and Interface Science 66:295–302. 21. Van Norde W, MacRitchie F, Nowicka G, Lyklema J. 1986. Protein adsorption at solid-liquid interfaces: reversibility and conformation aspects. J Colloid Interface Sci 112:447–456. 22. Horbett TA, Brash JL, editors. 1995. Proteins at Interfaces II, Fundamentals and Applications. ACS Symposium Series 60, chap. 1–38. 23. Pizza M, Scarlato V, Masignani V, Giuliani MM, Arico B, Comanducci M, Jennings GT, Baldi L, Bartolini E, Capecchi B, Galeotti CL, Luzzi E, Manetti R, Marchetti E, Mora M, Nuti S, Ratti G, Santini L, Savino S, Scarselli M, Storni E, Zuo P, Broeker M, Hundt E, Knapp B, Blair E, Mason T, Tettelin H, Hood DW, Jeffries AC, Saunders NJ, Granoff DM, Venter JC, Moxon ER, Grandi
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005
ADSORPTION OF PROTEIN ANTIGENS
G, Rappuoli R. 2000. Identification of vaccine candidates against serogroup B meningococcus by whole-genome sequencing. Science 287:1816– 1820. 24. Singh M, Chesko J, Kazzaz J, Ugozzoli M, Kan E, Srivastava I, O’Hagan D. 2004. Adsorption of a novel recombinant glycoprotein from HIV (Env gp120dV2SF162) to anionic PLG microparticles retains the structural integrity of the protein, while encapsulation in PLG microparticles does not. Pharm Res 21:2148–2152.
2519
25. Martin A. 1993. Physical Pharmacy, 4 edn. Philadelphia: Lea and Febiger. p 1 ff. 26. Singh-Jasuja H, Hilf N, Arnold-Schild D, Schild H. 2001. The role of heat shock proteins and their receptors in the activation of the immune system. Biol Chem 382:629–636. 27. Iyer S, Robinett RS, HogenEsch H, Hem SL. 2004. Mechanism of adsorption of hepatitis B surface antigen by aluminum hydroxide adjuvant. Vaccine 22:1475–1479.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 11, NOVEMBER 2005