sedimentation field-flow fractionation and granulometric analysis of plga microspheres

10
Original Paper Faisant, Battu, Senftleber, Benoit, Cardot 1407 J. Sep. Sci. 2003, 26, 1407 – 1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Nathalie Faisant 1 Serge Battu 2 Fred Senftleber 3 Jean-Pierre Benoit 1 Philippe J. P. Cardot 2 1 ERIT M0104, IBT, 10 rue AndrØ Boquel, 49100 Angers, France 2 Laboratoire de Chimie Analytique et Bromatologie, FacultØ de Pharmacie, UniversitØ de Limoges, 2 Rue du Dr. Marcland, 87025 Limoges Cedex, France 3 Department of Chemistry, Jacksonville University, 2800 University Blvd. N., Jacksonville, Florida, USA Sedimentation field-flow fractionation and granulometric analysis of PLGA microspheres Sedimentation field flow fractionation operated in the steric hyperlayer mode was used to obtain fractions of defined characteristics from crude samples of poly(D,L-lac- tic-co-glycolic acid) microspheres which were polydisperse in size. In less than ten minutes, Sedimentation Field Flow Fractionation (SdFFF) separation yielded three analytical fractions of very different size and particle size distribution (PSD) character- istics, as determined by granulometric analyses (Coulter Counterm and image analy- sis of SEM). A crude sample (average size = 45 lm, 105% size polydispersity index) was separated into fractions of 73 lm, 56 lm, 8 lm average diameters which showed a PSD of 39%, 33%, 30%, respectively. Our results demonstrated that SdFFF used in conjunction with particle size analysis offers a new approach to laboratory scale production of drug vectors of a specified average size and reduced size dispersity. In the future, this could be used to select the most convenient particles for drug loading and release. Key Words: Sedimentation field flow fractionation; PLGA microspheres; Granulometry; PSD; Received: September 21, 2002; revised: April 1, 2003; accepted: April 1, 2003 DOI 10.1002/jssc.200301516 1 Introduction New drug carriers have been developed that utilize micron-sized particles to transport and release a drug directly to its target [1]. The field of microencapsulation includes several technologies and several synthetic or natural polymer support materials that all lead to different types of microparticles (microcapsules, microspheres) [2, 3]. Among polymeric materials, poly(lactide-co-glycolide) (PLGA) is fully accepted for the fabrication of microspheri- cal implants in humans [4 – 6]. PLGA microspheres offer various advantages such as the possibility of controlling the resulting drug release over prolonged periods of time, easy administration, good biocompatibility, and complete biodegradation [7]. Such microspheres are usually pre- pared by extraction or emulsion evaporation, whereby the internal phase is removed by stirring [1, 3]. The micro- sphere populations are characterized by their mean aver- age size, size distribution, or higher statistical moments [8] such as skew or excess. They are also characterized according to their drug loading and drug release pro- files [9]. Unfortunately, because of the emulsification step, current manufacturing techniques lead to polydispersity in size and drug loading, affecting mass and therefore drug release profiles [10, 11]. Such dispersities in size, density, and mass complicate pre-industrial development, as well as regulatory compliance processes. There is, therefore, a need to clearly determine at least the polydispersity in size of microspheres within each batch in order to better control their characteristics according to given specifica- tions. A better description of such drug carriers must be assessed according to carrier size characteristics as well as to drug content per particle. 1.1 Field flow fractionation of drug vectorization objects Sedimentation field flow fractionation (SdFFF) [12] using an analytical system operating in the “Steric/Hyperlayer mode” [13] offers a rapid (minute scale) and efficient method for producing fractions of controlled size and reduced size dispersity. This approach has been applied to submicron devices such as liposomes [14 – 16], solid lipid nanoparticles [17], and nanospheres (300 nm) [14] for determination of the average size. In the case of micron sized particles, other FFF techniques have been used including dielectrophoretic/gravitational FFF to separate polystyrene microbeads [18], and simple Gravi- tational FFF to obtain narrow size distributions of polyhy- droxybutyrate microspheres [19]. However, no literature is, to date, available concerning micron sized PLGA parti- cle separation using either Flow FFF, hydrodynamic chro- matography, or SdFFF. Studies with loaded microparti- cles were not performed, probably due to the long analysis time necessary that can result in release of the drug dur- ing the separation. Correspondence: Philippe J.P. Cardot, Analytical Chemistry Dept., FacultØ de Pharmacie, UniversitØ de Limoges, 2 Rue du Dr. Marcland, 87025 Limoges Cedex, France. Phone: +33 5 55 43 58 57. Fax: +33 5 55 43 58 59. E-mail: [email protected].

Upload: independent

Post on 16-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Ori

gin

alP

aper

Faisant, Battu, Senftleber, Benoit, Cardot 1407

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Nathalie Faisant1

Serge Battu2

Fred Senftleber3

Jean-Pierre Benoit1

Philippe J.P. Cardot2

1ERIT M0104, IBT, 10 rue Andr�Boquel, 49100 Angers, France2Laboratoire de Chimie Analytiqueet Bromatologie, Facult� dePharmacie, Universit� deLimoges, 2 Rue du Dr. Marcland,87025 Limoges Cedex, France3Department of Chemistry,Jacksonville University, 2800University Blvd. N., Jacksonville,Florida, USA

Sedimentation field-flow fractionation andgranulometric analysis of PLGA microspheres

Sedimentation field flow fractionation operated in the steric hyperlayer mode wasused to obtain fractions of defined characteristics from crude samples of poly(D,L-lac-tic-co-glycolic acid) microspheres which were polydisperse in size. In less than tenminutes, Sedimentation Field Flow Fractionation (SdFFF) separation yielded threeanalytical fractions of very different size and particle size distribution (PSD) character-istics, as determined by granulometric analyses (Coulter Counterm and image analy-sis of SEM). A crude sample (average size = 45 lm, 105% size polydispersity index)was separated into fractions of 73 lm, 56 lm, 8 lm average diameters which showeda PSD of 39%, 33%, 30%, respectively. Our results demonstrated that SdFFF usedin conjunction with particle size analysis offers a new approach to laboratory scaleproduction of drug vectors of a specified average size and reduced size dispersity. Inthe future, this could be used to select the most convenient particles for drug loadingand release.

Key Words: Sedimentation field flow fractionation; PLGA microspheres; Granulometry; PSD;

Received: September 21, 2002; revised: April 1, 2003; accepted: April 1, 2003

DOI 10.1002/jssc.200301516

1 Introduction

New drug carriers have been developed that utilizemicron-sized particles to transport and release a drugdirectly to its target [1]. The field of microencapsulationincludes several technologies and several synthetic ornatural polymer support materials that all lead to differenttypes of microparticles (microcapsules, microspheres) [2,3]. Among polymeric materials, poly(lactide-co-glycolide)(PLGA) is fully accepted for the fabrication of microspheri-cal implants in humans [4–6]. PLGA microspheres offervarious advantages such as the possibility of controllingthe resulting drug release over prolonged periods of time,easy administration, good biocompatibility, and completebiodegradation [7]. Such microspheres are usually pre-pared by extraction or emulsion evaporation, whereby theinternal phase is removed by stirring [1, 3]. The micro-sphere populations are characterized by their mean aver-age size, size distribution, or higher statistical moments[8] such as skew or excess. They are also characterizedaccording to their drug loading and drug release pro-files [9]. Unfortunately, because of the emulsification step,current manufacturing techniques lead to polydispersity insize and drug loading, affecting mass and therefore drugrelease profiles [10, 11]. Such dispersities in size, density,

and mass complicate pre-industrial development, as wellas regulatory compliance processes. There is, therefore,a need to clearly determine at least the polydispersity insize of microspheres within each batch in order to bettercontrol their characteristics according to given specifica-tions. A better description of such drug carriers must beassessed according to carrier size characteristics as wellas to drug content per particle.

1.1 Field flow fractionation of drug vectorizationobjects

Sedimentation field flow fractionation (SdFFF) [12] usingan analytical system operating in the “Steric/Hyperlayermode” [13] offers a rapid (minute scale) and efficientmethod for producing fractions of controlled size andreduced size dispersity. This approach has been appliedto submicron devices such as liposomes [14–16], solidlipid nanoparticles [17], and nanospheres (300 nm) [14]for determination of the average size. In the case ofmicron sized particles, other FFF techniques have beenused including dielectrophoretic/gravitational FFF toseparate polystyrene microbeads [18], and simple Gravi-tational FFF to obtain narrow size distributions of polyhy-droxybutyrate microspheres [19]. However, no literatureis, to date, available concerning micron sized PLGA parti-cle separation using either Flow FFF, hydrodynamic chro-matography, or SdFFF. Studies with loaded microparti-cles were not performed, probably due to the long analysistime necessary that can result in release of the drug dur-ing the separation.

Correspondence: Philippe J.P. Cardot, Analytical ChemistryDept., Facult� de Pharmacie, Universit� de Limoges, 2 Rue duDr. Marcland, 87025 Limoges Cedex, France.Phone: +33 5 55 43 58 57. Fax: +33 5 55 43 58 59.E-mail: [email protected].

1408 Faisant, Battu, Senftleber, Benoit, Cardot

In this work, as a preliminary step, we sorted blank micro-spheres into subpopulations by SdFFF and analyzedthem by granulometry.

In order to develop a fast, analytical batch purification pro-tocol for PLGA microspheres, the particles must be elutedin SdFFF at moderate field intensity because of their highdensity (d 1.28). The collected fractions must then be sub-mitted to an analysis of their precise average sizes andsize distribution patterns [20], which are of critical impor-tance because density distributions, as well as size versusdensity functions, are unknown.

Thus, the main objective of this work was to apply theunique particle size separation effectiveness of FFF toproduce fractions with uniform characteristics in terms ofaverage size and polydispersity, with the future goal ofstudying the correlation between size and either drugloading or drug release on active microspheres.

1.2 Principle of sedimentation field flowfractionation

The general concept of field flow fractionation wasinvented by Giddings in the late sixties of the past century[13]. Sedimentation FFF techniques, which employ multi-gravitational external fields, are remarkably well suited forthe separation of colloids and micron-sized particles,including species of biological origin such as cells [21].The specific elution model for micron-sized species (diam-eter >1 lm) is described as the “Steric-Hyperlayer”model [22, 23].

According to this elution model, under a constant externalfield, species of identical densities are eluted according totheir size, the bigger ones being eluted first. In the case ofspecies of identical size but different densities, the heavierones are eluted last. There is therefore a size and densitybalance driving the separation process. As a conse-quence, fractions obtained with low size dispersity may beassociated with a relatively low density distribution. Apractical example using standard latex particles of con-trolled size is shown in Figure 1 in a channel of reducedthickness (80 lm). The gentle non-destructive propertiesof FFF enables fractions to be collected for subsequentgranulometric analyses.

Polydispersity is defined as follows

P% ¼�

r

m

�6100 ð1Þ

where r is the standard deviation of the concerned param-eter, and m its mean value. Because of the large polydis-persity in size of the microspheres analyzed in this paper(average size = 45 lm, 105% size polydispersity), allseparations were performed in 250 lm channels.

2 Experimental

2.1 Preparation of microspheres

Resomerm 506; PLGA (Poly(D,L lactic-co-glycolic acid);containing 25% D-lactic units, 25% L-lactic units, and 50%glycolic units; with a molecular weight of 104000) wasobtained from Boehringer Ingelheim (Paris, France).Dichloromethane (Rectapur) and polyvinyl alcohol 4/125(Rhodoviol) were obtained from Merck Eurolab (Gra-dignan, France). Microparticles were prepared on a 5 gscale using an O/W solvent extraction technique accord-ing to the method previously described [24]. Basically, asolution of PLGA in dichloromethane was emulsified intoan aqueous phase containing polyvinyl alcohol (10%). Byadding a large volume of water, dichloromethane wasextracted so that the droplets solidified into microspheres.The final suspension was then filtered under nitrogen andthe microparticles were freeze dried. To avoid any un-emulsified polymer fragments, the final product wassieved on 125 lm mesh.

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Figure 1. Sedimentation field flow fractionation of micronsized standard latex beads in channel of reduced thickness.Carrier phase: bidistilled water/Tween 20 0.1% (w/w); chan-nel dimensions: 78061560.08 mm; rotation axis to channelradius: 14.5 cm. Injection volume: 20 lL of 0.1% wt/wt sus-pension. Photometric detection at 254 nm. A.U. = Absor-bance Unit.

Microsphere size dispersity analysis 1409

2.2 Microsphere particle size distributiondeterminations

2.2.1 Scanning electron microscopy

The external morphology of the microparticles was char-acterized by Scanning Electron Microscopy (SEM) JEOL6301F Field Emission Microscope (JEOL-France, Paris).The microparticles were transferred onto adhesive paperand covered by a fine layer of carbon (10 nm) vapordeposited by a MED 020 (Baltec, Balzers, Lichtenstein)coating system. Examinations were done at 5 kV.

2.2.2 Coulter counter

Coulter counter measurements were performed on theoriginal microspheres sample: Particle size distributions(PSD) and mean diameters were determined using a Mul-tisizer (Coultronics, Margency, France). Microparticles(10 mg) were suspended by sonication for 10 min in 2 mLof balanced electrolyte solution (Isotonm II, Coultronics,Margency, France) containing 1% (w/v) Tween 80 andassayed after dilution in 200 mL Isotonm II. Size determi-nation accuracy was controlled by means of Duke (DukeScientific Corp, Palo Alto, CA, USA) calibrated particles of50.28 lm, 2% CV (w/w).

2.2.3 Image analysis algorithm

Size distributions of crude sample and of the 3 fractionsobtained from FFF elution were determined on opticalimages obtained from a Nikon inverted microscope modelTMS F (Nikon Corp, Tokyo, Japan, 10 and 206magnifica-tion) using a computerized image analyzer (Granix 2.1,Microvision Instruments, Evry, France). Size calibrationwas obtained by analysing sets of slides prepared frommonodisperse spherical latex beads (Duke) with averagecalibrated diameters: 4.99, 7.00, 10.24, and 50.28 lmrespectively. Imageanalysiswasperformedonthreeseriesof slides. Each series contained approximately 1000 to1500 particles. The very spherical particle shape observedfor the microspheres allowed simplified particle averageFeret diameters to be determined for every particle. Dia-meters, which were expressed in arbitrary dimensions,werethenusedtocalculate individualparticlevolumes [25].Cumulative arbitrary volumes, expressed in percentage,versus arbitrary size curves were then constructed. Oncethe 10%, 25%, 50%, 75%, 90% quartiles and minimum andmaximumsizeshadbeendetermined,acalibrationfunctionwas established using the quartiles of calibrated particles.The calibration functions obtained from Coulter and imageanalysis wereaveraged todetermine thesizedistributionofmicrospheresineverysampleorfractionassayed.

2.3 Field flow fractionation

The sedimentation FFF system used in this work consistsof the FFF separator, described as the “FFF channel” con-

nected to a UV detector and to a classical HPLC pump viaa chromatographic-like sample injector. The channel wasmanually cut from a Mylar band and inserted between twopolystyrene plates, one described as the depletion walland the other called the accumulation wall. The thicknessof the Mylar band defines the channel’s height. Thedimensions of the ribbon-like channel were 780 mm long,16 mm wide, and 0.25 mm thick, with two V-shaped ends(50 mm), giving a channel volume of 3100 lL. Thisassembly was sealed into a centrifuge basket, with spe-cial care being taken to avoid deforming the channel’s par-allelepiped shape during sealing. The channel diametermeasured after sealing in the centrifuge basket was27.8 cm (Mylar to Mylar). This channel was connected tothe other devices by means of laboratory designed rotat-ing seals and connecting tubes. The almost zero voidvolume rotating seals were made of two symmetrical pla-nar disks of different composition (metal to polymer)drilled to connect the tubes in their center. Sealing wasdone under mechanical pressure (10 bars). The sampleinlet tube (ID = 0.508 mm, 2.03 lL/cm) was connected tothe channel via the accumulation wall [21, 26]. Thevolumes of inlet tube and rotating seal were experimen-tally measured and found to be 150 lL. Similarly the voidvolume of the channel outlet tube (ID = 0.508 mm,2.03 lL/cm) and rotating seal was found to be 102 lL.Total void volume (channel volume + connection tubes +((injection and detection device volumes)/2) was3360 l 4 lL (n = 15) measured using acetone 1% (v/v). Itmust be pointed out that the volume at the inlet tube wasless than the 10% of the system void volume. Sedimenta-tion fields were expressed in units of gravity (1 g =980 cm/s2) and were calculated using both the measuredrotational speed (rpm: rotations per minute) and the chan-nel radius, r (cm), according to the classical equationdescribed elsewhere [21].

A Waters 590 (Waters France, St Quentin en Yvelines,France) programmable HPLC pump was used to producethe carrier liquid flow. In each experiment, the flow rateswere systematically measured at least every ten minutesand the experimental values used for data interpretation.

The mobile phase was phosphate buffered saline (PBS)of physiological ionic strength supplemented with 0.1%(w/w) Tween 20 (Atlas Chemie, Essen, Germany). Thesample suspension (30 mg of microspheres per mL of car-rier phase) was introduced via a 7525 Rheodyne (Cotati,CA, USA) valve with a loop of 50 lL. The detector was aUV-Vis Spectroflow 783 (ABI-Kratos, Ramsey, NJ, USA)operating at 254 nm, with a 0.2 second rise time constant.Signals were systematically recorded by means of a 14Bytes Analog to Digital converter already described [27].The signal acquisition rate was set at 2 Hz. Channeldecontamination was performed by injecting solutions of amixture of ethanol and water when the field was stopped.

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1410 Faisant, Battu, Senftleber, Benoit, Cardot

3 Results and discussionThe success of any separation strategy for assay or evenqualitative purposes depends on the choice of appropriatedetection or characterization methodology. The elution ofparticles by field flow fractionation uses classical liquidchromatography pumps and spectrophotometric detec-tors. Recent developments, however, have led to the con-cept of on-line hyphenation with particle characterizationtechniques like Multi Angle Light Scattering systems(M.A.L.S.S.) for sub-micron colloids [28] or Coulter count-ing and flow cytometry for micron-sized species [26]. Off-line detection is also possible by means of these tech-niques or by using microscopic observation and subse-quent image analysis. In the present work, two off-linetechniques based on very different principle were used tocharacterize PLGA particles in order to assess the accu-racy of size measurements: Image Analysis and CoulterCounting.

3.1 PLGA granulometric analysis techniques andprocedures

Coulter counting techniques can handle size diversity butrequire large sample populations to yield accurate andreproducible results. Thus, if small amounts of particlesare available for any given sample, the Coulter countingprinciple is limited and can lead to discrepancies. Thisproblem can be solved using appropriate image analysismethodologies.

Particle size analysis is a complex procedure whose gen-eral characteristics have been described in detail in a spe-cial publication issued by the National Institute of Stan-dards and Technology (NIST) [29]. For highly polydis-perse particle populations the NIST requirements foraccurate PSD assignments require large numbers of indi-vidual particles [30]. For instance, it can be deduced fromNIST recommendations that at least 10,000 particlesmust be counted using microscopic techniques, while20,000 are required for automatic Coulter counting.

The PLGA microspheres of interest in the present studywere characterized by scanning electron microscopy(SEM) as shown in Figure 2. The particles appearedremarkably spherical but exhibited a large size dispersity.Low magnification procedures (Figure 2.A) show largenumbers of particles with sizes ranging from 3 to 90 lm.At higher magnification (Figure 2.B) the shape stillappears spherical with surface characteristics (smooth-ness) that seem to be homogeneous regardless of particlesize. For accurate size determination, complementaryand independent techniques should be used.

Coulter counter techniques offer a suitable alternative toSEM in such situations. Thus quantitative particle charac-teristics were also determined using this method. Obtain-ing accurate particle size distributions expressed in terms

of particle volume or particle number requires proper cali-bration. Since the PLGA microspheres were shown to beremarkably spherical through scanning electron micro-scopy, Coulter calibration was performed using a set ofspherical calibrated latex particles.

Reproducible sampling during Coulter counting requiresagitation of the particle suspension, a process whoseeffectiveness is intimately linked to the size/density of theparticles. In order to obtain an accurate sampling of themicrospheres, a homogeneous and stable suspensionmust be provided. This is made possible by reducing parti-cle-medium density differences as well as increasingmedium viscosity. Although polystyrene latex particlescan be suspended homogeneously in Coulter Isotonmmedium, this is not the case for PLGA microspheres.PLGA and latex samples were diluted in Tween 80 solu-tions (1% w/v in Isotonm) to avoid measurement bias.

3.2 Crude particle size distribution

Particle size analysis data can be presented in differentways. Cumulative curves can be produced as shown inFigure 3. Here the abscissa represents the size while they axis represents the cumulative volume fraction. It is pos-sible to correlate data obtained by Coulter counting withthe size-based PSD established by image analysis mea-surements as shown in Figure 3. The image analysis algo-

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Figure 2. Scanning electron microscopy of PLGA microparti-cles. Sample treatment for image capture: Carbon vapor.Magnifications given in Figure. Voltage: 5 kV.

Microsphere size dispersity analysis 1411

rithm minimizes small particle volumes compared to Coul-ter counting while the opposite is observed for large parti-cle sizes. Differences can arise from many origins by com-parison between Coulter sizing and image analysis algo-rithms: for example, a sample in suspension for Coulteranalysis and sedimented on a surface for image analysis.Sample volume also can be involved. In order to obtain arelatively accurate size estimation, information providedby both techniques will be taken account. Therefore thedata series obtained by these two independent techni-ques was used to obtain an average particle size distribu-tion curve as shown in Figure 3.

In assessing the particle size distributions (PSD) fromboth data sets, it must be kept in mind that the data werenot obtained under the same conditions as are given inTable 1. These differences also could explain the sizingdifferences.

By using both granulometric techniques, an accuratedetermination of the microspheres PSD was undertaken.However, nothing is known concerning particle density ormass distribution, although the latter can be evaluated bymeans of velocimetric methods [20]. Attempts to studythese parameters were unsuccessful because of both thevery large average size and the high density of the sam-ple. So far, only the average density determination can beperformed using density gradient centrifugation tech-niques. If there is a density distribution among the sampleparticles, it is first necessary to fractionate the sampleaccording to its buoyant mass, then to characterize thefraction PSD, and finally to determine the average densityof this fraction. Field flow fractionation is a separationtechnique particularly well suited to isolation of particlesas a function of their size and density, especially if operat-ing in the hyperlayer mode. If fractions can be isolatedthey can be submitted to PSD determination, and correla-tion with retention time may indicate density differences. Ifthe fractions are very homogeneous in size, the “sterichyperlayer” elution model predicts that their density will

also be homogeneous. However, for fractions having ahigh retention ratio, large size polydispersity will not sig-nify high density distribution.

3.3 Optimization of SdFFF separation

The immediate consequence of the steric hyperlayer [13,22, 23] elution mode, in a channel of given thickness, isthat retention and separation depend on two main param-eters: the external field intensity and the average flow ratein the separator. Figure 4 shows the SdFFF separationsetup for microsphere separation. Initial conditions aredetermined using two parameters, the first being density.Heavy micron sized particles do not require high externalfield intensities, particularly if they are of large averagesize. The second parameter is linked to the generation oflift forces, which are determined by the linear velocity ofthe carrier phase flowing through the channel. The thickerthe channel, the greater the flow rate required to establisha particular linear velocity. The carrier phase is chosen tolimit particle/wall interactions. Preliminary experimentsare described in the sequence of fractograms shown inFigure 4.A. The first run was performed with channel rota-tion as low as 200 RPM and 5 mL/min flow rate. Themicrospheres sample is completely eluted in 12 minutes.In a second run, the external field was increased in turnincreasing microsphere retention as well as band spread-ing. However, it was necessary to check for the presenceof trapped particles. This is why, in the third sequence, thefield was stopped, leading to a signal appearing at 35 min-utes. Such a signal denoted a non-negligible release ofreversibly trapped particles. Systematic recovery studies(n = 5) showed that 78 l 6% of the injected sample waseluted in the fractogram while 15l7% was released as thebasket rotation was stopped. By inference between 8 and12% of the sample was irreversibly trapped or lost in theFFF system.

With the objective to limit the trapping of sample materialin the SdFFF system, high flow rates (9 mL/min) and

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Table 1. Granulometric analysis of microspheres. Coultermeasurement characteristics given in text. Image analysisprocedure: Feret diameter determination, individual particlevolume calculation assuming spherical shape, size/versusvolume histogram rebuilt at table given resolution. Resolutionis given after calibration with Spline interpolation processusing crude sample data convolution algorithm.

Volume% method Granulometrictechnique

Resolution Count

Crude sample Coulter 0.523 lm 79856Crude Sample Granix 1.72 lm 5010Fraction 1 Granix 4.67 lm 3702Fraction 2 Granix 2.65 lm 4641Fraction 3 Granix 0.25 lm 3072

Figure 3. Cumulative particle size distributions. Volume%cumulative curves, comparison of coulter data versus imageanalysis. Coulter (1), Image analysis (f), Average (*).

1412 Faisant, Battu, Senftleber, Benoit, Cardot

reduced external field (186 RPM–50 lL uninterruptedinjection) were used. These conditions reduced irreversi-ble trapping and enhanced recovery; 82 l 15% (n = 5)recovery was obtained. Reproducibility tests showed ori-ginal characteristics: only the measured eluted samplearea was disperse as shown in Figure 4.B. Elutions wereassociated with very constant peak profile parameters(retention and band spreading). Such an area dispersitycan be explained by differences in PLGA microspherequantities (number or mass) injected. To overcome this

limited quantitative reproducibility, sample were injectedaccording to the following procedure which required someskill. The sample was sonicated for 1 minute, thenvibrated for 30 seconds and injected, with the syringe in ahorizontal position, in the following 15 seconds. Optimisedconditions were found by slightly increasing the externalfield to enhance retention as shown in Figure 4.C (270RPM–9 mL/min). Elution recovery (n = 5) was measuredby means of fraction collection of the eluted peak and ofthe “released one” which showed that 85 l 10% of the

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Figure 4. SdFFF separation of microspheres. A) Initial conditions. Fractogram elution parameters described in Figure. “Release”corresponds to the signal obtained when external field is stopped. B) High density sample elution reproducibility. C) External fieldoptimized fractogram.

Microsphere size dispersity analysis 1413

sample (in particle number) was present in the elutedpeak, 10 l 7% released and 5 l 3% lost (irreversiblytrapped). To optimise the retention vs. recovery balance,the external field was slightly reduced and led to optimisedparticle recovery. FFF conditions associated to the fracto-gram shown in Figure 5.A. Three one-minute scaled frac-tions corresponding to 9 mL each were taken and labeledas fractions 1, 2, and 3. A fourth fraction was collected dur-ing the release procedure, that is at stopped rotation. It isobvious from the separation point of view that sharperfractions could be collected leading to enhanced numbersof average size particles versus retention time data. Thiscould be assayed if only average size were needed asalready published by Moon et al. [31, 32] for other sampletypes. However, to be able to properly describe not onlythe fraction average size but its associated PSD, numer-ous particles per fraction must be counted in accordancewith NIST requirements [29, 30]. With such a goal onlythree fractions were possible. Collected 9 mL fractionswere centrifuged, supernatant removed and fraction batchdiluted in 900 lL PBS prior to inverse phase microscopicobservation at 106magnification. Slides of the crude

sample and the fractions are shown in Figure 5.B forcomparison.

Fractions contained very different sized microspheres.Fraction 1 showed a relatively homogeneous populationof large particles. Fraction 2 was composed of smallerparticles while Fraction 3 had the smallest. The amount ofcollected fractions does not allow Coulter counter analy-sis; this is why image analysis was performed on the threefractions. Particles collected in the “release” fraction wereso few that even image analysis algorithm application wasnot possible. Quantitatively, the particle size distributionof the released particles appeared similar to that of thecrude sample.

3.4 Granulometric analysis of collected fractions

Crude samples, as well as fraction slides, were submittedto the same image analysis algorithm and volume or num-ber dependent particle size distribution was calculated asshown in Table 2. Fraction 1 was enriched in large sizedparticles while Fraction 3 contained the smallest sizedparticles. The pattern of Fraction 2 showed particles ofintermediate size. Not only are the 50% volume average

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Figure 5. Microsphere separation by SdFFF. A) Microsphere fractogram. Elution conditions described in Figure. One minutefraction collection, 4 fractions collected: 1,2,3, Release. B) Inverse phase microscopic pattern of crude sample and 4 FFF elutedfractions. Magnification610 after 10/1 concentration.

1414 Faisant, Battu, Senftleber, Benoit, Cardot

sizes different but the quartile values show that size dis-persities differ. The cumulative quartile data of Fraction 3indicated small size dispersity while those of Fractions 1and 2 were associated with much larger size dispersity.Compared to the crude sample, Fraction 1 encompassedthe larger particles.

Here, all results derive from image analyses. Such resultsare in complete qualitative accordance with the hypoth-esis of the size dependent elution order of the “steric-hyperlayer” elution mode if it is assumed that particle aver-age densities of the fractions are comparable or sizedependent. It appears essential to describe PSD usingquartile parameters for further comparison with drug-loaded microspheres.

The PSD differential curves obtained for the three frac-tions and the original sample can be drawn as shown inFigure 6. The y-axis was normalized to the mode (modevolume% = 100).

Figure 6 patterns compare particle size not only in termsof average and standard deviation dimensions but alsoPSD. Profile shape allowing accurate determination offraction and crude sample characteristics is complemen-tary to the data shown in Table 2. Fraction 3 appears to bethe sharpest; polydispersity calculations using volumepercentage dimensions indicated a polydispersity index of30%. Fractions 2 and 1 containing sizes were also asso-ciated with increased polydispersities of 33 and 39%. Thenumber dependent and volume dependent sizing datashown in Table 2 are very different. This is obvious, differ-ent dimensions led to different results, indicating the needto clearly define the sizing procedure used.

In the particle size purification performed by SdFFF, itappears that the fraction of smallest average size is alsothe least disperse. With only three collected fractions oflarge volume and wide polydispersities, it is not possible tobuild a rigorous elution time (or fraction) dependent size

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Table 2. FFF fractions and granulometric analysis. Granulometric analysis of microsphere crude sample and SdFFF collectedfractions. %Vol express size versus particle volume histogram; %NB express size versus particle number histogram. The aver-age is calculated by means of particle volume or number histogram versus size first moment, standard deviation is the squareroot of the centered second moment. Histogram resolution in size as given in Table 1

ElutionTime(min)

AverageDiameter

(lm)

Mode(lm)

StandardDeviation

(lm)

Mini-mum

Diameter(lm)

Maxi-mum

Diameter(lm)

Quartile10% (a)

Quartile25% (a)

Quartile50%(a)

Quartile75%(a)

Quartile90%(a)

Crude (Granix)%VOL 47.15 58.49 18.01 2.58 83.43 23.3 32.73 47.05 61.15 71.96Crude (Granix)%NB 12.76 1.72 13.45 2.58 83.43 1.56 2.62 7.39 18.81 31.3Crude (Coulter)%VOL 50.2 32.68 30.05 – – 18.08 26.81 42.91 67.52 96.48Crude (Coulter)%NB 13.15 4.968 10.19 – – 4.975 6.085 9.955 16.66 25.14Fraction 1%VOL 1.5 72.97 98.05 28.56 2.58 130 38.87 50.95 68.85 96.67 119.26Fraction 1%NB 16.51 1.72 21.18 2.58 130 1.41 2.24 5.14 27.79 47.65Fraction 2%VOL 2.5 35.56 30.96 11.75 2.58 71.39 22.17 27.50 34.00 42.28 51.77Fraction 2%NB 15.43 1.72 12.75 2.58 71.39 1.63 3.07 13.37 25.18 32.78Fraction 3%VOL 3.5 8.25 10.32 2.52 2.58 14.62 4.78 6.32 8.32 10.24 11.6Fraction 3%NB 5.01 3.44 2.47 2.58 14.62 1.99 3.15 4.63 6.49 8.66

Figure 6. Particle size distribution of crudesample and FFF eluted fractions. Y axis 100%scale centered on PSD mode (determined withthe volume% dimension). Note that number%one will lead to a different PSD pattern.

Microsphere size dispersity analysis 1415

selectivity curve [33]. Moreover, complete density distri-bution determination of every fraction appears, so far,relatively complex to perform, especially in the light of thehypothesis that density may be covariant with size [34]. Ifcomplete NIST-compliant granulometric techniques canbe associated with FFF elution, opportunities are open todevelop and validate large panels of physical characteris-tics of crude and purified samples including surface prop-erties, shape, rigidity, and density.

4 Concluding remarksTaken together, our results demonstrated that SdFFF is asuitable technique for preparing, in an analytical range, orin a post production stage, microsphere fractions of char-acterized size and PSD. SdFFF can overcome or consid-erably reduce the size dispersity as well as the polydisper-sity barriers described in the introduction. SdFFF used onan analytical level is a technical and methodological basisfor laboratory-scale preparative purposes or a basis fordeveloping SPLITT preparative separations. The couplingof retention characteristics with one or several NIST-com-pliant PSD determinations offers the possibility of highlyaccurate size characterization. This is of major impor-tance if microspheres are to be used for drug vectoriza-tion.

It must be noted here that microspheres loaded with anti-cancer drugs are to be injected directly into a patient’sbrain. From a practical point of view, a low number ofmicrospheres are effectively inserted, lying in the range ofquantities obtained on this FFF analytical scale. If pre-liminary pharmacological experiments are performed onanimal models using anticancer drugs under development(mice), even lower quantities are required. If pilot-scaleproduction of calibrated microspheres is needed, SPLITTtechnology [35] could be employed. Some analyticalSdFFF to SPLITT technology upgrades have alreadybeen reported [35, 36].

Acknowledgements

Contracts 200REC41 (Aide R�gionale � La Recherche enLimousin), and ACI 2001-TS9 From (Fond National de laTechnologie) funded this work. Jeanne Cook-Moreau (La-boratoire de Biochimie M�dicale, Facult� de M�decine,Universit� de Limoges, Limoges, France), Maria TeresaGalceran and Rams�s Sanz (Departament de QuimicaAnalitica, Universitat de Barcelona, Barcelona, Spain),Meyon Hee Moon (Department of Chemistry and Chemis-try Institute for Functional Materials, Pusan National Uni-versity, Pusan, S. Korea,) and Karin Caldwell (BiomedicalCenter, Upsala, Sweden) are gratefully thanked for helpfuldiscussions in preparing the original manuscript and forEnglish corrections.

References[1] J.P. Benoit, H. Marchais, H. Rolland, V. Vande Velde, in

Microencapsulation, S. Benita (Ed.). Marcel Dekker, Inc.,New York, USA, 1996, pp. 35–72.

[2] C. Thies, in Microencapsulation, S. Benita (Ed.). MarcelDekker, Inc., New York, USA, 1996, pp. 1–19.

[3] J. Richard, J.P. Benoit, in Trait� de G�nie des proc�d�s,Techniques de l’Ing�nieur, 2000 p. Doc J 2 210 211–220.

[4] L. Brannon-Peppas, Int. J. Pharm. 1995, 116, 1–9.

[5] J.M. Anderson, M.S. Shive, Advanced Drug DeliveryReviews 1995, 28, 5–24.

[6] R.A. Jain, Biomaterials 2000, 21, 2475–2490.

[7] P. Menei, V. Daniel, C. Montero-Menei, M. Brouillard, A.Pouplard-Barthelaix, J.P. Benoit, Biomaterials 1993, 14,470–478.

[8] N. Dyson, Chromatographic integration methods. RCSChromatography Monograph Series. RCS, Cambridge, UK,1990.

[9] C. Washington, in Microencapsulation, S. Benita (Ed.). Mar-cel Dekker, Inc., New York, USA, 1996, pp. 156–181.

[10] M. Dunne, O.I. Corrigan, Z. Ramtoola, Biomaterials 2000,21, 1659–1668.

[11] J. Akiki, N. Faisant, J. Siepmann, P. Oury, R. Troude, E.Bruna, J. Haffner, J.P. Benoit, 16th Annual Meeting of theGTRV, Lyon, France, 2001.

[12] E. Assidjo, T. Chianea, M.-F. Dreyfuss, P.J.P. Cardot, J.Chromatogr. B 1998, 709, 197–207.

[13] J.C. Giddings, Science 1993, 260, 1456–1465.

[14] K.D. Caldwell, G. Karaiskakis, M.N. Myers, J.C. Giddings, J.Pharm. Sci. 1981, 70, 1350–1352.

[15] J.J. Kirkland, W.W. Yau, F.C. Szoka, Science 1982, 215,296–298.

[16] R. Dreyer, E. Hawrot, A.C. Sartorelli, P.P. Constantinides,Anal. Biochem. 1988, 175, 433–441.

[17] W. Mehnert, K. Mader, Advanced Drug Delivery Reviews2001, 47, 165–196.

[18] X.-B. Wang, J. Vykoukal, F.F. Becker, P.R.C. Gascoyne,Biophys. J. 1998, 74, 2689–2701.

[19] A.C. Kassab, K. Xu, E.B. Denkbas, Y. Dou, S. Zhao, E.Piskin, J. Biomater. Sci., Polym. Ed. 1997, 8, 947–961.

[20] J.S. Reed, Introduction to the principle of ceramic proces-sing, Wiley-Interscience, NewYork 1988.

[21] S. Battu, W. Elyaman, J. Hugon, P.J.P. Cardot, Biochim.Biophys. Acta-Gen. Subj. 2001, 1528, 89–96.

[22] J. Chmelik, J. Chromatogr. A 1999, 845, 285–291.

[23] M. Martin, P.S. Williams, NATO ASI Ser., Ser. C 1992, 383,513–580.

[24] N. Faisant, J.P. Benoit, P. Menei, Wo 00/69413: FR9906207, 1999.

[25] P. Robert, J. Melcion, F. Le Deschault De Monredon, Pow-der Technol. 1997, 90, 141.

[26] P. Cardot, S. Battu, A. Simon, C. Delage, J. Chromatogr. B2002, 768, 285–295.

[27] P.J.P. Cardot, P. Trolliard, E. Guernet-Nivaud, Chromato-graphia 1992, 33, 361–368.

[28] B.A. Korgel, J.H. van Zanten, H.G. Monbouquette, Biophys.J. 1998, 74, 3264–3272.

[29] A. Jillavenkatesa, S.J. Dapkunas, L.-S.H. Lum, Particle sizecharacterization. National Institute of Standards and Tech-nology, Washington, DC 2001.

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1416 Faisant, Battu, Senftleber, Benoit, Cardot

[30] Certificate of Analysis, Standard Reference Materialm1982,Zirconia Thermal Spray Powder – Particle Size Distribution,Standard Reference Materials Program, National Institute ofStandards and Technology, Gaithersburg, MD, 1996.

[31] M.H. Moon, J.C. Giddings, Anal. Chem. 1992, 64, 3029–3037.

[32] M.H. Moon, J.C. Giddings, Ind. Eng. Chem. Res. 1996, 35,1072–1077.

[33] M.R. Schure, J. Chromatogr. A 1999, 831, 89–104.

[34] K.D. Caldwell, in Field-flow fractionation handbook, M.E.Schimpf, K.D. Caldwell (Eds.). John Wiley & Sons, Inc., NewYork 2000, pp. 295–312.

[35] J.C. Giddings, Sep. Sci. Technol. 1985, 20, 749–768.

[36] C. Contado, P. Reschiglian, S. Faccini, A. Zattoni, F. Dondi,J. Chromatogr. A 2000, 871, 449–460.

J. Sep. Sci. 2003, 26, 1407–1416 www.jss-journal.de i 2003 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim