Asymmetrical Flow Field-Flow Fractionation and Multiangle Light

Sungwon Lee, Yanhui Zhang, and Henry S. White , C. Chad Harrell and Charles R. .... Christoph Johann , Stephan Elsenberg , Ulrich Roesch , Diana C. Ra...
0 downloads 0 Views 219KB Size
Anal. Chem. 2004, 76, 1909-1920

Asymmetrical Flow Field-Flow Fractionation and Multiangle Light Scattering for Analysis of Gelatin Nanoparticle Drug Carrier Systems Wolfgang Fraunhofer,‡,† Gerhard Winter,§ and Conrad Coester*,§

Abbott GmbH & Co. KG, Department Pharmaceutical Development, Knollstrasse, 67008 Ludwigshafen, Germany, and Department of Pharmacy, Pharmaceutical Technology and Biopharmaceutics, Ludwig-Maximilians-University, Center for Drug Research, 81377 Munich, Germany.

The physicochemical properties of nanosized colloidal drug carrier systems are of great influence on drug efficacy. Consequently, a broad spectrum of analytical techniques is applied for comprehensive drug carrier characterization. It is the primary objective of this paper to present asymmetrical flow field-flow fractionation (AF4), coupled online with multiangle light scattering detection, for the characterization of gelatin nanoparticles. Size and size distribution of drug-loaded and unloaded nanoparticles were determined, and data were correlated with results of state-of-the-art methods, such as scanning electron microscopy and photon correlation spectroscopy. Moreover, the AF4 fractionation of gelatin nanoparticulate carriers from a protein model drug is demonstrated for the first time, proposing a feasible way to assess the amount of loaded drug in situ without sample preparation. This hypothesis was set into practice by monitoring the drug loading of nanoparticles with oligonucleotide payloads. In this realm, various fractions of gelatin bulk material were analyzed via AF4 and size-exclusion highpressure liquid chromatography. Mass distributions and high-molecular-weight fraction ratios of the gelatin samples varied, depending on the separation method applied. In general, the AF4 method demonstrated the ability to comprehensively characterize polymeric gelatin bulk material as well as drug-loaded and unloaded nanoparticles in terms of size, size distribution, molecular weight, and loading efficiency. The highly interdisciplinary field of nanoparticles encompasses chemistry, material research, molecular biotechnology, immunology and medicinal and engineering sciences.1 In the pharmaceutical industry, nanoparticles prepared from a variety of natural or synthetic polymers have attracted much attention because of the target-oriented advanced drug delivery features of these systems.2 * Corresponding author, colloidal drug carrier section. Phone:+49-(0)-89-218077042. Fax: +49-(0)-89-2180-77020. E-mail: [email protected]. † Corresponding author, analytical section. Phone:+49-(0)-621-589-1210. Fax: +49-(0)-621-589-1212. E-mail: [email protected]. ‡ Abbott GmbH & Co. KG. § Ludwig-Maximilians-University. (1) Niemeyer, C. M. Angew. Chem., Int. Ed. 2001, 40, 4128-4185. (2) Farrugia, C. A.; Groves, M. J. J. Pharm. Pharmacol. 1999, 51, 643-649. 10.1021/ac0353031 CCC: $27.50 Published on Web 02/24/2004

© 2004 American Chemical Society

Especially in applications such as gene therapy, the design of ideal polymeric carriers, those able to target specific cell types, is highly desired. The research field of nanoparticles is expected to rapidly evolve and gain increasing importance for medicine and pharmaceutics in the next years.3 Via nanoparticles, the sustained parenteral delivery of DNA4 and oral gene delivery5 were, as well, successfully approached as the transfection of stem cells only a few months ago.6 Nanoparticles based on the natural biopolymer gelatin are very promising, because they are readily available, possess a relatively low antigenicity,7 and enable high transfection rates.8 The parenteral use of gelatin derivatives over many years provides a sound safety basis for future applications. There are numerous ways for nanoparticle application. Consequently, the implementation of new analytical methods in nanoparticle characterization struggles to keep pace. Fundamental knowledge of nanoparticle dimensions is of utmost importance as the influence of the size on nanoparticle efficiency is still much discussed. For particle size determination, two main techniques are used: electron microscopy and photon correlation spectroscopy (PCS), often referred to as dynamic light scattering (DLS) or quasi-elastic light scattering (QELS). The intensive utilization of scanning electron microscopy (SEM) may be ascribed to both the high resolution and the access to physicochemical parameters, which include particle size, porosity, and morphology.4,9,10 However, the conditions precedent to SEM applications are sample conductivity and tolerance of vacuum ambience. Sometimes the presence of surfactants in the preparation may inhibit nanoparticle characterization via SEM due to the formation of a smooth, camouflaging coating on the particle surfaces.11 The exceeding magnifications of transmission electron (3) Langer, R. Nature 1998, 392, 5-10. (4) Cohen, H.; Levy, R. J.; Gao, J.; Fishbein, I.; Kousaev, V.; Sosnovsky, S.; Slomkowski, S.; Golomb, G. Gene Ther. 2000, 7, 1896-1905. (5) Roy, K.; Mao, H.-Q.; Huang, S.-K.; Leong, K. W. Nat. Med. 1999, 5, 387391. (6) Corsi, K.; Chellat, F.; Yahia, L’H.; Fernandes, J. C. Biomaterials 2003, 24, 1255-1264. (7) Schwick, H. G.; Heide, K. Bibl. Haematol. 1969, 33, 111-125. (8) Truong-Lee, V. L.; Walsh, S. M.; Schweibert, E.; Mao, H.-Q.; Guggino, W. B.; August, J. T.; Leong, K. W. Arch. Biochem. Biophys. 1999, 361, 47-56. (9) Gref, R.; Domb, A.; Qellec, P.; Blunk, T.; Mueller, R. H.; Verbavatz, J. M.; Langer, R. Adv. Drug Delivery Rev. 1995, 16, 215-233. (10) Mu, L.; Feng, S. S. J. Controlled Release 2003, 86, 33-48. (11) Kreuter, J. Int. J. Pharm. 1983, 14, 43-58.

Analytical Chemistry, Vol. 76, No. 7, April 1, 2004 1909

microscopy (TEM), the second microscopic method implemented for nanoparticle characterization, enable accurate particle surface and morphology analysis.12,13 Inherent drawbacks of both methods are random sampling (instead of an overall sample analysis) and time-consuming sample preparation and measurement procedures. In reviewing the literature dealing with nanoparticle characterization, one will come across PCS as the dominating sizing technique. Due to its noninvasive and nondestructive performance, PCS evades artifacts. Dissolved and undissolved matter can be sized within minutes in a reproducible way. The major obstacle in order to achieve veritable results is due to the underlying principles: since PCS measures the effective z average of the diffusion coefficientsthis coefficient being proportional to the reciprocal particle radius for spherical shapessthe sizes derived are influenced by the presence of dust or agglomerated fractions present in the sample. Furthermore, a number of assumptions inherent in data analysis also affect particle distributions. Many of these issues were addressed in a recent work, in which SEC was coupled with both static and dynamic light scattering, and PCS was found to be only applicable for high molecular weight polymers with a higher refractive index increment.14 However, PCS for reason is to be considered the preferred choice for nanoparticle sizing, provided that size distributions are narrow. Other sizing techniques play a minor role in nanoparticle characterization. The detection limit of g1 µm bars light obscuration from application. Equally, nanoparticle size determination via the Coulter principle can be considered as critical. The latest generation of Coulter Counters scaled the detection limit down to a few hundred nanometers under optimal conditions; however, the prevalent dimensions of polymeric nanoparticles used in medicinal and pharmaceutical sciences range from 50 to 400 nm. The upcoming technique of scanning force microscopy (SFM), also known as atomic force microscopy (AFM), was successfully used in visualizing nanoparticles,15 but the main area of AFM application lies in the investigation of surface morphology of membranes, molecular unbinding,, and adhesion forces, not size determination. It is the aim of this study to demonstrate the ability of asymmetrical flow field-flow fractionation (AF4) to provide an overall characterization of gelatin nanoparticles, running the gamut from gelatin bulk material analysis to the monitoring of nanoparticle drug-loading processes. For the first time, gelatin nanoparticles will be characterized via AF4/MALS, and focusing on a more general application, nanoparticles will be fractionated from pharmaceutical drugs, such as proteins and oligonucleotides. Additionally, the utilization of AF4 in the course of establishing a novel method for the accurate and convenient assessment of the loading efficiency of nanoparticle drug carrier systems will be presented. Asymmetrical Flow Field-Flow Fractionation. The nanoparticle analysis by AF4/MALS in our laboratories was based on the following strategy: light scattering is the state-of-the-art method for determining particle sizes, assuming that all particles have exactly an identical size (i.e., a group of equally sized particles (12) Li, M.; Schnablegger, H.; Mann, S. Nature 1999, 402, 393-395. (13) Chui, Z.; Mumper, R. J. Bioconjugate Chem. 2002, 13, 1319-1327. (14) Liu, Y.; Bo, S.; Zhu, Y., Zhang, W. Polymer 2003, 44, 7209-7220. (15) Dong, R.; Yu, L. E. Environ. Sci. Technol. 2003, 37, 2813-2819.

1910

Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

produces a scattering pattern identically to that of a single particle, except that it is amplified proportionally to the number of particles present in the illuminated scattering volume; otherwise, referring to conditions in which the particles present are not of equal size, MALS yields the z-average radius of the distribution). An inherent AF4 feature is the ability to separate dissolved and undissolved substances within the sample volume, typically spanning dimensions from the lowest nanometer range up to several micrometers. Consequently, AF4 enables accurate characterization of both nanoparticles and nanoparticle bulk material, since the fractionation process provides a narrow dispersity of the particle population being present in the detector. Thereby, the prerequisite of static or so-called classic light scattering is met.16 The underlying principles of AF4 have been reviewed elsewhere.17-22 Briefly, the sample species are fractionated within a separation channel, wherein a laminar flow, with parabolic flow profile, transports them along the channel toward the channel outlet after injection of the sample into the channel near the channel inlet. The injection of the sample into the channel is performed through a separate port near the channel inlet. Concomitantly, a field of separation is exerted on the sample by a second flow profile perpendicular to the laminar flow. The resulting cross-flow volume leaves the channel via an ultrafiltration membrane covering the bottom of the channel. Due to differences in hydrodynamic diameters and in diffusion coefficients, smaller sample species are eluted prior to larger ones. The correlation between time of retention, tr, and the diffusion coefficient is given by eq 1,

tr )

t0Vcw2 6DV0

(1)

where tr is the retention time, t0 is the retention time of an unretained solute, Vc is the cross-flow rate, V0 is the volume of the separation channel, w is the channel thickness, and D is the diffusion coefficient of the separated sample species. The plotting of intensity of the detector signal versus elution time or elution volume is referred to as a fractogram. Multiangle Light Scattering. Particles and macromolecules in a liquid scatter light from an incident beam in all directions. Thereby, the scattering intensity at various angles is used to determine the average particle size. In most cases, the RayleighGans-Debye (RGD) theory is used, which can be expressed as

K*c 1 ) R(θ) MwP(θ)

(2)

where R(θ) is the excess Rayleigh ratio (the light scattered by a solution at an angle θ in excess of that light scattered by the pure solvent, divided by the incident light intensity) at a scattering angle θ, K* represents an optical constant, c is the sample concentration, (16) Wyatt, P. J. J. Colloid Interface Sci. 1998, 197, 9-20. (17) Coelfen, H.; Antonietti, M. Adv. Polym. Sci. 2000, 150, 67-187. (18) Field-Flow Fractionation Handbook; Schimpf. M., Caldwell, K., Giddings, J. C., Eds.: Wiley-Interscience: New York, 2000. (19) Janca, J. J. Chromatogr. Libr. 1992, 51, 449-479. (20) Wahlund, K. G.; Giddings, J. C. Anal. Chem. 1987, 59, 1332-1339. (21) Wahlund, K. G.; Litze´n, A. J. Chromatogr. 1989, 461, 73-87. (22) Litze´n, A.; Wahlund, K. G. Anal. Chem. 1991, 63, 1001-1007.

Mw is the weight-averaged molecular weight of the eluted sample species, and P(θ) is the scattering form factor which describes the angular dependence of the intensity of scattered light. Equation 2 shows that the higher the sample concentration and the higher the molecular weight of the sample specimen, the larger the amount of scattered light. The hydrodynamic diameter can be calculated independently of the molecular weight. Equation 323 illustrates the calculation of the hydrodynamic diameter of a sphere from the radius of gyration (the latter being accessible by using multiangle light scattering)

d)

x203〈R



2 G

(3)

with 〈RG2〉 as the mean square radius of gyration. Theory and underlying principles of light scattering are extensively discussed elsewhere.24 The coupling of fractionating techniques, such as SEHPLC or field-flow fractionation, with MALS detection are also described in the literature.25-28 EXPERIMENTAL SECTION Gelatin Nanoparticles. Gelatin type A from porcine skin (175 Bloom), glutaraldehyde (25%), EDC (N-[3-dimethylaminopropyl]N’-ethylcarbodiimide hydrochloride) and cholaminchloride hydrochloride ([2-aminoethyl]-trimethylammonium chloride hydrochloride) were purchased from Sigma-Aldrich (Taufkirchen, Germany); acetone was purchased from VWR (Ismaning, Germany). Nanoparticle Preparation. Gelatin nanoparticles were prepared as described previously.29 A 1.25-g portion of gelatin was dissolved in water under gentle heating (5% w/w). The first desolvation step was initiated by adding 25 mL of acetone. After sedimentation of precipitated gelatin fractions over 60 s, the supernatant was discarded, and the sediment was redissolved in water under constant heating and magnetic stirring at 500 rpm. Subsequently, the pH was adjusted to 3.0 using 1 M HCl. By dropwise addition of acetone, the gelatin was desolvated again and stirred for 10 min. Finally, 200 µL of glutaraldehyde (25%) was added in order to stabilize the gelatin nanoparticles by intraparticular cross-linking of gelatin amino groups. Surface Modification. The surface modification was carried out similarly as previously described.30 A 50-mg portion of EDC and 50 mg of cholaminchloride hydrochloride was added, and the batch was incubated for 14 h. The resulting nanoparticles were purified by triple centrifugation (16000g for 20 min) and redispersed in Milli-Q water. Plasmid Loading. A 20-µg portion of DNA plasmid, containing the photinus luciferase gene (under control of the cytomegalovirus enhancer/promoter) was incubated with surface-modified gelatin (23) Roessner, D. Ph.D. Dissertation, University of Hamburg, Hamburg, Germany, 1994. (24) Wyatt, P. J. Anal. Chim. Acta 1993, 272, 1-40. (25) Roessner, D.; Kulicke, W. M. J. Chromatogr., A 1994, 687, 249-258. (26) Wittgren, B.; Wahlund, K. G. J. Chromatogr., A 1997, 760, 205-218. (27) Striegel, A. Anal. Chem. 2002, 74, 3013-3018. (28) Schure, M. R.; Palkar, S. A. Anal. Chem. 2002, 74, 684-695. (29) Coester, C.; von Briesen, H.; Langer, K.; Kreuter, J. J. Microencapsulation 2000, 17, 187-194. (30) Coester, C. New Drugs 2003, 1, 14-17.

nanoparticles in Milli-Q water (1.0 mg/mL). After incubation for 2 h at 37 °C by using a Thermomixer comfort (Eppendorff, Hamburg, Germany), the samples were centrifugated, and the loading efficiency was quantified spectrophotometrically (260 nm) in the supernatant. Oligonucleotide Loading. A 50-µg portion of oligonucleotide was incubated for 2 h at 37 °C with surface modified gelatin nanoparticles in Milli-Q water (1.0 mg/mL). The oligonucleotide (18-mer, 5451 Da) was purchased from MWG Biotech GmbH (Ebersberg, Germany). Nanoparticle Sizing. Loaded and unloaded particles were characterized using a Zetamaster (Malvern Corp., U.K.). Each assigned size and the corresponding polydispersity index values were based on 10 individual measurements. ζ-Potential Measurements. ζ-Potential values were determined using the Zetamaster. Milli-Q water (conductivity < 0.04 mS/cm) was used as analysis diluent. Pharmaceutical Proteins. A 100-kDa immonuglobulin fragment (F(ab′)2 fragment of isotype G3K immunoglobulin, used as marker protein for SE-HPLC) was donated by Abbott GmbH & Co. KG, Ludwigshafen, Germany. Granulocyte colony stimulating factor (G-CSF, 20 kDa) was donated by Roche Diagnostics GmbH, Penzberg, Germany. Instruments. The AF4 experiments were carried out with equipment purchased from Postnova Analytics (Landsberg/Lech, Germany), consisting of an HRFFF-10.000 AF4 system, autosampling system (PN 5200 sample injector) and in-line degaser (PN 7505 vacuum degaser). An in-line solvent filter (0.1 µm, poly(tetrafluoroethylene), Postnova Analytics, Gemany) was placed between the laminar flow pump and the channel to reduce detector background signals. The channel thickness was 350 µm, the applied ultrafiltration membranes were polyethersulfone membranes with 1 kDa cutoff and regenerated cellulose membranes with 5 kDa cutoff (both Nadir Filtration GmbH, Wiesbaden, Germany) and regenerated cellulose membranes with 1, 5, and 10 kDa cutoff (Postnova Analytics, Germany). Axial channel flow rates were 1 mL/min for all AF4 experiments unless stated otherwise, and cross-flow rates were controlled via NovaFFF 3.00 software (Postnova Analytics, Germany). For example, stated conditions of a 10% cross-flow rate and a 1 mL/min channel outlet flow volume imply that elution medium entered the channel with a 1.11 mL/min flow volume. While passing the channel, 10% of the elution medium was pumped off through the ultrafiltration membrane to generate a cross-flow volume of ∼0.11 mL/min. The resulting axial flow at the channel outlet results in a 1.0 mL/min volume. For each AF4 experiment, 100 µL samples were injected into the channel. AF4 was coupled with MALS (DAWN EOS, Wyatt Technology Corp., Santa Barbara, CA), refractive index detection (∆n - 1000, λ ) 620 nm, differential refractometer, WGE Dr. Bures, Dallgow, Germany; dn/dc measurements performed at ambient temperature) and UV spectrophotometry (λ ) 220, 260, and 280 nm; Spectra System UV1000, Thermo Separation Products, Germany). With the DAWN EOS, scattered light was detected by an array of 15 photodiodes arranged at various angles relative to the incoming laser beam. According to light scattering theory, nanoparticles scatter light in an anisotropic way (i.e. in a more forward direction). Consequently, the photodiodes placed at lower angles will detect more scattered light than photodiodes Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

1911

placed at higher angles. To enable accurate size determination of the nanoparticles, the sensitivity of eight photodiodes was dimmed by a factor of 100 (photodiode angles: 14, 26, 35, 43, 90, 121, 142, and 163°). Thus, overexcitation of the photodiodes by huge intensities of light scattered by nanoparticles was avoided. Since small sample species (e.g., 100 kDa were 38% in bulk, 29% in supernatant, 58% in sediment. (b) molecular weight distribution (b) of gelatin in bulk solution, calculated via MALS, spanning a total range from 10 to 1000 kDa; note that a MW of ∼100 kDa clearly marks off two gelatin fractions with MW greater (hmw) and lower (lmw) than 100 kDa, respectively. (c) Flow 0.3 mL/min; curves represent UV signals; ratios of hmw gelatin specimen >100 kDa were 15% in bulk, 13% in supernatant, 24% in sediment.

decreasing the elution flow volume from 0.7 mL/min to 0.3 mL/ min and changing, thereby invariably, other separation parameters, such as sample species residence time and system pressure, results in noticeable data change. It is generally accepted that especially for hmw components, the applied SEC conditions (i.e.,

column packing, flow profiles) have to be optimized and calibrated to avoid a systematic underestimation of sample Mw and D values.35 When analyzing well-characterized globular proteins, such as human serum albumin or immunoglobulins, under varying SEC conditions (i.e., different elution flow volumes) the determined monomer, dimer, and aggregate ratios are not fundamentally affected. On the contrary, the molecular weight profile of gelatin is influenced by parameters such as time-dependent temperature effects, spontaneous reformation of lmw gelatin components to hmw species, impetus of R chain association, and other variables outlined above. It is important to note that SEC conditions influence those parameters during gelatin separation. Furthermore, as a consequence of the industrial production processes, a molecular weight heterogeneity is innate to gelatin. Therefore, fractionation conditions have to be carefully selected and should be retained unchanged in order to enable proper comparison of data from experiments performed at different times. When characterizing the gelatin fractions with AF4, the gelatin molecular weight heterogeneity was confirmed. In contrast to SEC, the overall molecular weight range of gelatin specimen in bulk solution analyzed by AF4/MALS ranged from ∼20 kDa to over 10 000 kDa, outranging maximal molecular weight values yielded with SEC/MALS by more than 1 order of magnitude (Figure 2a). Due to the low field strengths exerted (5% cross flow), the elution profile of the gelatin species corresponds to a characteristic molecular weight order without separating individual fractions. We hypothesized that the moderate conditions during the AF4 experimentssmainly due to package material free separation within the AF4 channelsresults in nondestructive fractionation. In contrast, hmw polymers in SEC experiments are prone to degradation due to abrasive shear forces exerted by column package material, and this may consequently lead to data misinterpretation.36 Our results are in good agreement with former studies of molar mass distributions of natural rubbers in microgels, wherein the largest molecules detected via field-flow fractionation were found to be more than 3 orders of magnitude greater than indicated by SEC.37 More generally, the suitability of flow-FFF for the characterization of soluble polyelectrolyte-gelatin complexes of high molecular weight and approaching the three-digit nanometer range has been demonstrated.38 Thereby, the hydrodynamic radii derived by the retention times of the complexes revealed a reasonable correlation with those determined by dynamic light scattering, indicating that no analytes’s structural artifacts were induced in the course of the AF4 fractionation. Comparable to SEC data, the ratio of (ultra)-hmw components fractionated via AF4 increases in the following order: gelatin in supernatant < gelatin in bulk solution < precipitated gelatin. In this concern, it is to be outlined that the haphazard comparison of data derived from SEC and AF4 is not admissible. Generally, there is the risk that ultra-hmw components would be eluted close to the exclusion limit of the SEC columns and be difficult to differentiate from various noise signals. Furthermore, such components may be trapped in the column filter frits and not be (35) Mendichi, R.; Schieroni, A. G. Polymer 2002, 43, 6115-6121. (36) Kulicke, W. M.; Boese, N. Colloid Polym. Sci. 1984, 262, 197-207. (37) White, R. J. Polym. Int. 1997, 43, 373-379. (38) Tan, J. S.; Harrison, C. A.; Li, J. T.; Caldwell, K. D. J. Polym. Sci., Part B: Poly. Phys. 1998, 36, 537-542.

Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

1913

Table 1. Nanoparticle Characterization via PCS, SEM, and MALS nanoparticles

sizea,b

(nm, via PCS) PIc,d ζ potential (mV) sizea (nm, via SEM) Rfrms (nm, via MALS) d (nm, calcd from Rrms)

unloaded

loaded

256 ( 1.80 0.028 ( 0.012 +42 150-300 60-98 155-253

269 ( 7.37 0.062 ( 0.038 +9 e 68-105 175-271

a Referring to nanoparticle size distribution, note that the data obtained by PCS are noticeably inconsistent with data delivered by SEM and MALS. b PCS experiments were performed with n ) 10; all other experiments were performed at least in triplicate. c Determined via method of cumulants (see ref 42). d PI ) polydispersity index. e Not determined. f Rrms ) z-average root-mean-square radius.

Figure 2. AF4 analysis of various gelatin fractions, detection via UV 280 nm and MALS. Curves represent UV symbols: gelatin in bulk solution (grey line), gelatin in supernatant fraction after addition of acetone/centrifugation (pale line), gelatin in sediment fraction (black line); hmw ) gelatin species eluting >30 min. (a) A cross flow of 5% does not sufficiently separate hmw species. (b) Cross flow 10%; ratios of hmw gelatin: 7% in bulk, 3% in supernatant, 20% in sediment. (c) cross flow 55%; ratios of hmw gelatin: 17% in bulk, 10% in supernatant, 27% in sediment.

eluted from the column, consequently avoiding detection. Due to the lack of such drawbacks, AF4/MALS proved suitable for the detection of ultra-hmw components, as shown for ethylhydroxyethyl cellulose.39 Moreover, other shear-sensitive samples, such (39) Andersson, M.; Wittgren, B.; Wahlund, K.-G. Anal. Chem. 2001, 73, 48524861.

1914 Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

as certain proteins, can be fractionated without running the risk of structural alterations.40 As outlined, the amount of higher-order species which can be retained within the AF4 system and which can be fractionated within an individual peak is dependent on the extent of the cross flow exerted. For example, 10% cross-flow conditions are not sufficient to separate higher-order components (Figure 2b), in contrast to the situation when 55% cross-flow rates are applied (Figure 2c). In that case, an assessment of the different gelatin batches in terms of hmw species content is possible. Characterization of Gelatin Nanoparticles. The results of nanoparticle size characterization by PCS and ζ-potential measurements are summarized in Table 1. The ζ-potential values of both loaded and unloaded particles indicate the presence of interparticular repulsive forces, by this opposing agglomeration tendencies. Emanating from dimensions of ∼256 nm, the size increase due to plasmid loading is almost negligible. Thereby, the plasmid strands enjoy a great steric freedom, propelling a tight packing of the twisted strands on the surface of the spherical particles. Furthermore, the noticeable positive ζ potential of unloaded nanoparticles is attracting a diffuse layer of surrounding medium, thus increasing the hydrodynamic layer encasing the particles. Conversely, if the positive surface charge is countervailed by negatively charged plasmid strands or oligonucleotides, both ζ potential and thickness of the surrounding diffuse medium layer are decreased, resulting in reduced hydrodynamic diameters determined by PCS.41 SEM images of nanoparticles confirmed the spherical dimensions of the nanoparticles and unveiled their surface to be smooth and unruffled (Figure 3). In contrast to the unimodal size distributions suggested by PCS, SEM analysis, demonstrating smaller and larger unloaded nanoparticles to be present at random, provides reason to critically reconsider PCS data: whereas SEM reveals unloaded particle dimensions to encompass 150-300 nm, PCS assigns sizes of unloaded and loaded nanoparticles as ∼256 and 269 nm, respectively. Of course, one explanation for the differences can be ascribed to the underlying principles of both methods,43 exempli(40) Li, P.; Hansen, M.; Giddings, J. C. J. Liq. Relat. Technol. 1997, 20, 27772802. (41) Mueller, R. H., Zeta potential and particle load in laboratory practice; Wissenschaftliche Verlagsgesellschaft: Stuttgart, 1996. (42) Koppel, D. E. J. Chem. Phys. 1972, 57, 4814-4820.

Figure 3. SEM picture of gelatin nanoparticles, revealing a broad nanoparticle size distribution, because smaller (∼150 nm) and larger (∼300 nm) nanoparticles are detectable in random quantities.

fied earlier with synthetic polymer-based nanoparticles:44 the contrast of SEM pictures allows only the visualization of the nanoparticle core; conversely, PCS assesses the hydrodynamic radius of analytes. Due to the data inconsistency of SEM with PCS, a demand for accessorial sizing techniques appears on the analytical agenda. To minimize adhesion/adsorption phenomena of nanoparticles on the AF4 ultrafiltration membrane, various buffer media were investigated in pilot AF4 experiments. Using a buffer medium at pH 7.4 resulted in maximum sample recovery, whereas buffer media at pH 4.8 and 6.0 led to a sample recovery decrease of 3.8 and 3.1%, respectively, when compared to pH 7.4 conditions. This feature, when using more acidic buffers, is probably due to interactions of nanoparticles with equipment surfaces (e.g., ultrafiltration membrane, channel and tubing material); however, we cannot explain this phenomenon satisfactorally. When using acidic buffer media of pH 4.8, even 20 min after the nanoparticles should have left the channel completely (indexed by UV and MALS signals), the light scattering detector revealed a noisy basis signal, indicating that nanoparticles still were eluting. Generally, all membranes in use revealed good applicability, except for regenerated cellulose membranes with a 1-kDa cutoff, which showed low reproducibility of AUCUV values. As outlined in Figure 4a, the UV signal intensity corresponds linearly with nanoparticle concentration of the sample batch. If AF4 experiments aim at providing great resolution in size separation, to apply high cross-flow rates is considered to be stateof-the-art. Because high cross-flow rates may cause the sample concentration at the ultrafiltration membrane to increase, the risk of inducing sample-sample and sample-membrane interactions gains a higher probability. To provide optimal resolution in order to measure veritable size distributions, it was decided to apply low cross-flow levels below 20% and to maintain them constant for the time periods necessary (e.g., 20 min). By this, size distributions of unloaded nanoparticles were determined to span a 155-253-nm diameter range. Increasing the injected sample species mass by a factor of 1.8 definitely did not affect size distribution results (Figure 4a), thus showing the AF4 operating parameters to be appropriate. The dependency of nanoparticle elution profiles on even slight modifications of cross-flow rates is illustrated in Figure 4b. (43) Finsy, R.; de Jaeger, N.; Snyers, R.; Gelade, E. Syst. Charact. 1992, 9, 125137. (44) Hoffmann, F.; Cinatl, J.; Kabickova, H.; Kreuter, J.; Stieneker, F. Int. J. Pharm. 1997, 157, 189-198.

Figure 4. AF4/MALS analysis of gelatin nanoparticles, detection via UV 280 nm and MALS. Curves represent UV signals. (a) Nanoparticle concentration 0.5 mg/mL (pale line), nanoparticle concentration 0.9 mg/mL (black line); at both concentrations, rootmean-square radii (rms r) span 60-98 nm. (b) and (c) Characterization of plasmid-loaded nanoparticles; symbols assigning UV signal and rms radii, respectively, pertain to 5% cross-flow rate; rms radii 80-107 nm (pale lines); 10% cross-flow rate; rms radii 68-105 nm (grey lines); 15% cross-flow rate; rms radii 68-105 nm (black lines).

Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

1915

Exerting a 5% cross-flow level leads to a ∼6 min delay in the elution of the first nanoparticle fraction. Enhancing the cross flow to 10 and 15%, respectively, further increased this delay, as expected (Figure 4b,c). Objectionable is the considerable deceleration of sample elution, reflected by a suboptimal peak symmetry. If the cross-flow strength would be carried to the extremes, a hmw sample specimen may agglomerate and become manifest at the ultrafiltration membrane surface at the worst. To minimize this phenomenon, the addition of surfactants, for example 0.002-0.02% Tween or SDS, to the elution medium is well-established, especially when nanocolloids, such as liposomes, microgels or nanoparticles, are to be separated.45,46 One drawback of this procedure may be the undesired decrease of signal-to-noise ratios. When adding surfactants to buffer media in concentrations exceeding the critical micelle concentration, nanomicelles arise as sort of “stealth analytes”. The micelles, commonly of 2-10 nm in dimension, accumulate near the ultrafiltration membrane when exposed to high cross-flow rates. In case of fast changes of the resolution power within short time lags, that is, operating with >30% cross-flow intensity decays within a few seconds, the micelles are flushed out of the channel and flood the detector cells, thus inducing artifact signals and facilitating data misinterpretation (i.e., the overassessment of nanoparticle dimensions due to false attribution of scattered light intensities). Typically, refractive index detectors and light scattering detectors are by far more sensitive to such stealth analytes than UV detectors when coupled with AF4.47 Figure 4b reveals that exerting cross-flow rates of 5% is insufficient for the accurate measurement of nanoparticle size distributions: the nanoparticle fractions are not eluted in total accordance with their size, but nanoparticles with the smallest dimensions are eluted contemporarily with larger nanoparticles right at the beginning of the elution process. Consequently, the MALS signal of the smallest nanoparticles is superimposed by the signal of larger nanoparticles, and thus, MALS averages the dimensions of both eluted fractions to a mean radius of 80 nm. Due to the insufficient resolution of 5% cross-flow conditions, nanoparticle aggregation phenomena cannot be circumvented. If 10% cross-flow rates are applied, the resolution is sufficient to achieve an elution order in strict accordance with increasing particle sizes. At those conditions, the MALS signal of the smallest nanoparticles is not superimposed, and the radii of the nanoparticles, which are eluted first, are calculated to be 68 nm. To control, if 10% cross-flow rates really provide sufficient resolution, the resolution simply can be increased by raising the cross flow to a 15% level (Figure 4c). This results in further delay of the nanoparticle elution profile when compared to 10% crossflow situations. Yet, it is to be outlined that as far as the calculated nanoparticle size distributions are concerned, there is no data difference, no matter whether 10 or 15% cross-flow rates are applied. In both cases, the distribution of the nanoparticle radii is calculated to span 68-105 nm. (45) Jungmann, N.; Schmidt, M.; Maskos, M.; Weis, J.; Ebenhoch, J. Macromolecules 2002, 35, 6851-6857. (46) Thang, N. M.; Knopp, R.; Geckeis, H.; Kim, J. I.; Beck, H. P. Anal. Chem. 2000, 72, 1-5. (47) Fraunhofer, W. 4th World meeting on pharmaceutics, biopharmaceutics and pharmaceutical technology, Florence, Italy, 2002.

1916

Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

Considering the particle sizes, MALS clearly substantiates SEM data: from both analytical techniques, that is, SEM and MALS, a broad particle size distribution within a diameter range of 150300 nm is to be concluded. Thus, monodisperse gelatin nanoparticle size distributions suggested by PCS measurements will have to be reconsidered in future experiments. Determination of Nanoparticle Loading Efficiency. Hitherto, the assessment of drug-loading efficiency of colloidal drug carriers, such as gelatin nanoparticles, is a time-consuming and somewhat tedious procedure. After incubation of drug carrier and designated payloads, the sample species are subjected to several washing and centrifugation steps. Subsequently, the amount of unbound payload, for example, plasmid or a designed oligonucleotide, are determined by UV spectrophotometry in the supernatants.48 However, this modus operandi is delicate. If the centrifugal force exerted upon the sample is too intense, the nanoparticles may be destroyed or aggregated, thus opposing proper redispersing. On the other hand, a too gentle centrifugation would inevitably give rise to the possibility of not quantitatively removing nanoparticles from the supernatant. Generally, the risk of DNA degradation is to be avoided, as fragmentation of DNA is deemed to affect the transfectivity of loaded nanoparticles.49 In this concern, the wide applicability of AF4 provides the prospect of an alternative way to assess loading efficiency. Being principally capable of separating entities such as plasmids, DNA, or dissolved drugs from undissolved drug carriers, AF4 may allow almost real-time analysis of sample concentrations in the incubation vials. Ideally, after incubation, the AUCUV decrease of the DNA fraction designated for loading should comply with the AUCUV increase of the (then loaded) nanoparticles. Consequently, the difference in signal intensities of unloaded and loaded nanoparticles is due to drug payload. Whereas AF4 was shown in this study to be able to characterize nanoparticles qualitatively, that is, to assess size and size distributions, the quantification of nanoparticle concentration in a reproducible way is a requirement for the monitoring of the nanoparticle loading step. As outlined in Figure 5, the correlation of nanoparticle concentration with assessed AUCUV values can be performed with AF4 in a reproducible and accurate way, that is, the quantification of nanocolloidal drug carriers is manageable. This is not to be taken for granted, considering a variety of pitfalls described in the literature associated with the AF4 analysis of sample dimensions of ∼100 nm. For example, the characterization of polyorganosiloxane nanoparticles in aqueous dispersion by AF4 was shown to involve void peaks accounting for >50% of total UV detector signals, owing to certain circumstances.50 Similarly, the AF4 analysis of Janus micelles, which exhibited dimensions of 40-250 nm, was coming along with comparable problems.51 Due to a contingent sample loss, obligatory with both the challenging application and the arranged separation parameters, the sample recovery was assumed to be 100% in order to render a weight(48) Prabha, S.; Zhou, W.-Z.; Panyam, J.; Labhasetwar, V. S. Int. J. Pharm. 2002, 244, 105-115. (49) Walter, E.; Moelling, K.; Pavlovic, J.; Merkle, H. P. J. Controlled Release 1999, 61, 361-374. (50) Jungmann, N.; Schmidt, M.; Maskos, M. Macromolecules 2001, 34, 83478353. (51) Erhardt, R.; Zhang, M.; Boeker, A.; Zettl, H.; Abetz, C.; Frederik, P.; Krausch, G.; Abetz, V.; Mueller, A. H. J. Am. Chem. Soc. 2003, 125, 3260-3267.

Figure 5. Quantification of gelatin nanoparticle concentration via AF4/UV280 (experiments performed in triplicate). The assessed AUCUV values were in sound agreement with the individual sample concentrations: 476 ng/mL, AUC 1.966; 238 ng/mL, AUC 0.901; 119 ng/mL, AUC 0.459.

Figure 6. Separation of G-CSF protein (∼20 kDa) from nanoparticles applying a decay of initial constant 20% cross flow to 1% after 13 (pale line), 16.3 (grey line) and 19.7 min (black line), respectively. The reproducibility of the separations are reflected by the overlay of detector signals of void peak and protein.

averaged molecular weight determination possible, although the signals of detectors aiming at concentration assessment, such as RI detectors, may be prone to noisiness and unsettled baselines when analyzing sample species with three-digit nanometer dimensions. This may be due to the high sensitivity of large sample species on even minor cross-flow conditions, forcing the samples in the close vicinity to the ultrafiltration membrane and facilitating adhesion phenomena on the membrane. The task of separating nanoparticles from unbound DNA drug load was approached by surrogate experiments, for example, the fractionation of nanoparticle-protein mixtures. Therefore, G-CSF as a model compound (∼20 kDa) was added to nanoparticle batches, and the influence of cross-flow variations on the nanoparticle elution pattern was investigated (Figure 6). During the first stages of the experiments, the cross flow was held constant at a 20% level and finally reduced to a 1% degree within 60 s. This parameter proved to be sufficient for a baseline separation of void fraction and protein fraction. According to the underlying principles of AF4, G-CSF is eluted prior to the nanoparticle fraction, being separated from the drug carrier in baseline quality. Variations in the duration of the applied field strength render a modified

elution profile of protein possible, giving analytical tolerance if a change in sample elution pattern were desired. This flexibility of AF4 separation parameters is an essential prerequisite for both accurate sample quantification and size determination of the individual sample components, because superimposing of UV or MALS detector signals would inevitably lead to the assessment of incorrect concentrations or size distributions. In the case of the fractionation of an oligonucleotide (5.4 kDa) from gelatin nanoparticles, the attention is directed toward a sufficient separation of void peak from oligonucleotide (Figure 7). This was realized by starting the AF4 run and applying a 65% cross-flow level for ∼3 min before decreasing the field strength in a linear way until a 1% level was reached after ∼14 min. Being exposed to marginal cross-flow, the nanoparticles are eluted readily out of the channel and can be quantified. In this concern, it is important to note that by subjecting the nanoparticles to a constant cross flow of high intensity (e.g., >30%) throughout the whole experiment, the nanoparticles are immobilized upon the membrane, because no UV280 and no light-scattering signal can be monitored for over 60 min elution time. Analytical Chemistry, Vol. 76, No. 7, April 1, 2004

1917

Figure 7. Fractionation of a sample containing oligonucleotide (ODN) and nanoparticles. The overlay of the depicted three individual UV signals indicates a reproducible manner of quantification (standard deviation of AUCUV280,