Brush-like branched biodegradable polyesters, part III

Journal of Controlled Release 73 (2001) 7–20 www.elsevier.com / locate / jconrel

Brush-like branched biodegradable polyesters, part III Protein release from microspheres of poly(vinyl alcohol)-graftpoly( D,L-lactic-co-glycolic acid) Karin Frauke Pistel, Armin Breitenbach, Regina Zange-Volland, Thomas Kissel* Department of Pharmaceutics and Biopharmacy, Philipps-University, Ketzerbach 63, D-35032 Marburg, Germany Received 21 June 2000; accepted 29 January 2001

Abstract Brush-like branched polyesters, obtained by grafting poly(lactic-co-glycolic acid), PLGA, onto water-soluble poly(vinyl alcohol) (PVAL) backbones, were investigated regarding their utility for the microencapsulation of proteins. Poly(vinyl alcohol)-graft-poly(lactic-co-glycolic acid), PVAL-g-PLGA, offers additional degrees of freedom to manipulate properties such as e.g. molecular weight, glass transition temperature and hydrophilicity. PLGA chain length was varied at a constant molecular weight (Mw ) of the PVAL backbone and secondly Mw of the PVAL backbone was varied keeping the PLGA chain lengths constant. The most striking feature of these polymers is their high Mw . Microencapsulation of hydrophilic macromolecules, such as bovine serum albumin, ovalbumin, cytochrome c and FITC-dextran using a w / o / w double emulsion technique was investigated. Surface morphology, particle size, encapsulation efficiencies and protein release profiles were characterized as well. Microencapsulation of model compounds was feasible at temperatures of 0–48C with yields typically in the range of 60–85% and encapsulation efficiencies of 70–90%. Both, encapsulation efficiency and initial protein release (drug burst) were strongly affected by the glass transition temperature, T g , of the polymer in contact with water, whereas the in vitro protein release profile depended on the PVAL-g-PLGA structure and composition. In contrast to PLGA, protein release patterns were mostly continuous with lower initial drug bursts. Shorter PLGA chains increased drug release in the erosion phase, whereas initial pore diffusion was affected by the Mw of PVAL backbone. Release profiles from 2 to 12 weeks could be attained by modification of composition and molecular weight of PVAL-g-PLGA and merit further investigations under in vivo conditions. The in vitro cytotoxicity of PVAL-g-PLGA is comparable to PLGA and therefore, this new class of biodegradable polyesters has considerable potential for parenteral drug delivery systems.  2001 Elsevier Science B.V. All rights reserved. Keywords: Branched biodegradable polyesters; Poly(vinyl alcohol)-graft-poly(lactic-co-glycolic acid); Microencapsulation; Protein release; Parenteral depot systems; In vitro cytotoxicity

1. Introduction

*Corresponding author. Tel.: 149-6421-282-5880; fax: 1496421-282-7016. E-mail address: [email protected] (T. Kissel).

Parenteral depot systems based on biodegradable polymers should provide continuous or ‘infusionlike’ drug release patterns in the range of weeks and months. Among biodegradable polymers, e.g. poly-

0168-3659 / 01 / $ – see front matter  2001 Elsevier Science B.V. All rights reserved. PII: S0168-3659( 01 )00231-0

8

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

(ortho esters) [1], poly(anhydrides) [2] or poly(esters) [3], the most frequently used carrier is poly(lacticco-glycolic acid) (PLGA) because of its well documented biocompatibility and safety [4]. Protein release from biodegradable depot systems is thought to occur by a combination of pore diffusion through water-filled channels and diffusional release from the device as a function of polymer degradation and erosion. The initial protein release from microspheres, MS, prepared from PLGA is characterized by a rapid burst release caused by drug located close to or on the surface of microspheres. After a lag phase following this pore diffusion phase, drug release recommences when mass loss of the devices becomes prominent (erosion-phase) [5]. Initial pore diffusion is mainly affected by parameters such as the protein / polymer ratio (loading), particle size and wettability of the carrier [6]. Lag and erosion phase are predominantly influenced by polymer degradation [7]. Factors, such as crystallinity, molecular weight, molecular weight distribution as well as co-monomer composition influence drug release in the erosion phase. Undesirable lag-phases, leading to discontinuous release of proteins from parenteral depot systems can be manipulated to a limited extent by formulation parameters [8]. To allow modification of the release pattern in a broader range, new biodegradable polymers with improved erosion properties are required. We have shown previously that ABA triblock copolymers consisting of PLGA A-blocks attached to central hydrophilic poly(ethylene oxide) (PEO) Bblocks provide a continuous release of proteins such as bovine serum albumin (BSA), ovalbumin, tetanus toxoid or erythropoietin over several weeks due to a more rapid water uptake and swelling of the polymer [9–11]. The incompatibility of several proteins with PEO, notably rh-erythropoietin, suggested further modifications of the hydrophilic component [11]. Grafting polyester chains onto hydrophilic backbones such as dextran or charge-containing dextransulfate [12,13] as well as poly(vinyl alcohol) (PVAL), PVAL-g-PLGA [14], yields water insoluble polyesters with branched structures. These polymers allow manipulation of their hydrophilicity and hence compatibility with proteins in a broad range. Moreover, polymer erosion can be controlled by the PLGA side-chain length. Formation of water soluble

PLGA degradation products can be attained after few cleavage steps shifting the degradation mechanism from bulk to surface erosion. This aspect is of particular importance for parenteral delivery systems containing acid-labile proteins. The mechanical and thermo-mechanical properties of PLGA are critical for preparation of parenteral delivery systems. Low Mw PLGA would be desirable from a protein release and device erosion point of view, but poor mechanical stability leads to manufacturing problems. Branched polyesters can be obtained with very high molecular weights, providing mechanical stability, and yet offer acceptable degradation rates compared to PLGA [13]. For the microencapsulation of peptides, frequently the w / o / w double emulsion has been used [15,16]. While process parameters are well characterized the influence of polymer properties on the w / o / w method is not completely clear. Therefore, another aim of this study was to evaluate a series of PVAL-gPLGA with different Mw and composition, for the preparation of protein-loaded microspheres to determine which polymer properties affect the microencapsulation and protein release from MS.

2. Materials and methods

2.1. Materials Bovine serum albumin (BSA, fraction V, 64,000 g / mol, IEP 4.7), cytochrome c (Cyt c, 12,500 g / mol, IEP 9.8), ovalbumin (OVA, 36,000 g / mol, IEP 4.6) and fluorescein isothiocyanate labeled dextran (FITC-dextran (in microsphere batches abbreviated as FD), 38,900 g / mol) were supplied by Sigma (Germany). The BCA-Assay was obtained from Pierce (USA). Poly(vinyl alcohol) (PVAL, 88% hydrolyzed, 130,000 g / mol, Mowiol  18-88) was a gift from Hoechst AG (Germany). All other materials of analytical grade were supplied by Merck (Germany). Poly(lactic-co-glycolic acid) (PLGA) (Resomer RG 503, LA:GA 50:50, 40,000 g / mol) was obtained from Boehringer Ingelheim (Germany). Poly(vinyl alcohol)-graft-poly(lactic-co-glycolic acid) (PVAL-gPLGA) was prepared by melt polymerization as described previously [14]. The PLGA co-monomer

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

9

Fig. 1. Schematic illustration of the structure variations of PVAL-g-PLGA and its influence on microsphere characteristics.

ratio of lactide to glycolide was 50:50. Table 1 summarizes the bulk properties of the polymers and Fig. 1 illustrates the different structural variations and its influence on microsphere characteristics.

2.2. Methods

17 h (Edwards Freeze dryer Modulyo, UK). Final products were stored at 148C in a desiccator. This standard protocol was varied concerning polymertype, encapsulated compounds (BSA, OVA, Cyt c and FITC-dextran) and drug loading. For each modification one microsphere batch was prepared.

2.2.1. Microsphere preparation Microspheres were prepared by a modified (w 1 / o / w 2 ) double emulsion technique [15]. The microencapsulation process was carried out at 48C. Briefly, 0.5 g of the respective polymer was dissolved in 2.5 ml dichloromethane (DCM). Into this organic phase (o), 250 ml of an aqueous drug solution (w 1 ) were emulsified using a high speed homogenizer (UltraTurrax TP18 / 10, IKA, Germany) operating at 20,000 rev. / min for 30 s to form the w 1 / o emulsion. This primary emulsion was injected into 200 ml of an aqueous phase containing poly(vinyl alcohol) (0.5% w / v) (external phase, w 2 ) and homogenized for 30 s (Ultra-Turrax T25, IKA, Germany) at 8000 rev. / min. The resulting w 1 / o / w 2 emulsion was stirred at 200 rev. / min for 2.5 h with a propeller stirrer to allow solvent evaporation and microsphere hardening. The microspheres were then isolated by centrifugation at 1000 rev. / min for 1.5 min, washed three times with distilled water and freeze-dried for

2.2.2. Protein content of microspheres Protein content was determined by a modified DMSO / NaOH / SDS method [17]. MS (10 mg) were completely dissolved (30 min) in 0.5 ml DMSO by occasional shaking. One ml of a NaOH / SDS-solution (0.25 N NaOH, 0.5% w / v SDS) was added and the mixture was agitated in a rotating bottle apparatus (Rotatherm, Liebisch, Germany) at 30 rev. / min and 378C for 6 h. The amount of encapsulated BSA, OVA and Cyt c was determined with the BCA-assay (Pierce Chemicals, USA) according to the producers instructions using microtiter plates (Microwell-plates, Nunc, Germany). Color development was measured at 570 nm with a Microplate Reader (Titertek Plus MS212, ICN, SLT Labinstruments Deutschland GmbH, Germany). All samples and standards were assayed in triplicate. Encapsulation efficiency, theoretical and actual drug loading are defined as follows (tot., total drug amount; enc., encapsulated drug amount):

10

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

drug(tot.) theoretical drug loading 5 ]]]]]] drug(tot.) 1 polymer drug(enc.) actual drug loading 5 ]]]]]] drug(tot.) 1 polymer actual drug loading encapsulation efficiency 5 ]]]]]]] theoretical drug loading 3 100

2.2.3. FITC-dextran content of microspheres Content of the FITC-dextran loaded microspheres was analyzed by an extraction method. Ten mg of the FITC-dextran loaded microspheres were dissolved in 0.5 ml of DCM, followed by addition of 4 ml of PBS buffer pH 7.4 and agitation in a rotating bottle apparatus for 15 h at 30 rev. / min and 378C (Rotatherm). After separating the two phases, the FITC-dextran concentration in the aqueous phase was determined fluorimetrically (excitation: 493 nm, emission: 515 nm, LS 50B Luminescence Spectrometer, Perkin Elmer, Germany) using an FITCdextran calibration curve. Each sample was measured in triplicate. 2.2.4. Particle size Particle size of microspheres was analyzed dispersing ca. 10 mg of the samples in an aqueous solution of Tween  20 (0.1% w / v). The measurements were carried out by laser light diffraction using a Malvern Mastersizer X (Malvern Instruments, UK). The 300-mm lens utilized covered a particle size range of 1.2–600 mm and the 100-mm lens covered a particle size range of 0.5–180 mm. The calculation of the particle sizes was carried out using the standard modus of the Malvern software according to the theory of Mie. The weighted average of the volume distribution D[4.3] was used to describe particle size. This parameter is defined as follows: D[4.3] 5 ond 4 / ond 3 (n is the number of particles in each area of particle sizes, d is the medium particle diameter in the area of particle sizes). Each sample was measured in triplicate. 2.2.5. Surface morphology Microspheres were dried in vacuo and subsequently sputter-coated with a gold layer at 25 mA in argon atmosphere at 0.3 hPa for 2 min (Edwards / Kniese

Sputter Coater S150, Edwards, Germany). The coating procedure was repeated three times. Morphology of the microspheres was analyzed by scanning electron microscopy (SEM) (Hitachi S510, Hitachi Denshi GmbH, Germany) in vacuo (0.001 mbar) and at a voltage of 25 kV.

2.2.6. In vitro release profiles of proteins and FITC-dextran from microspheres Forty mg of loaded microspheres were suspended in 4 ml of PBS-buffer pH 7.4 containing 0.05% NaN 3 and 0.01% Tween  20. The samples were agitated in a rotating bottle apparatus (Rotatherm) at 30 rev. / min and 378C. At defined time intervals the buffer was completely withdrawn after centrifugation and replaced by 4 ml of fresh buffer. The amount of protein released was determined with the BCAassay. The concentration of FITC-dextran in the supernatant was determined fluorimetrically (excitation: 493 nm, emission: 515 nm, LS 50B Luminescence Spectrometer, Perkin Elmer) using a calibration curve. Every batch was studied in triplicate. The release studies of the FITC-dextran MS were carried out over a period of up to 90 days whereas the release studies for protein-loaded MS were interrupted after 35 days. 2.2.7. DSC measurement of the polymers Polymer films cast from 5% (w / v) DCM or acetone solutions on Teflon  plates were allowed to dry for 72 h. Residual solvents were then removed in vacuum at room temperature, at 308C and finally 408C during 2 days until constant weights were obtained. Films were cut into 50310 mm 2 slabs with a thickness of 50 to 100 mm. Samples of known weight (ca. 50 mg, n52) were immersed in 5 ml of isotonic phosphate buffered saline solution (PBS, pH 7.4, 0.15 M) in sealed glass test tubes and stirred in a rotating thermostat (Rotatherm) at 30 rev. / min and 378C. At preset intervals samples were recovered. Differential scanning calorimetry (DSC) was conducted in nitrogen atmosphere using a DSC7 calorimeter (Perkin Elmer) in sealed aluminum pans, relative to indium and gallium standards. Thermograms covered a range of 220 to 2008C with heating and cooling rates of 108C / min. Glass transition temperatures (T g ) were determined from the second

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

run on film specimens after the water on the surface was removed by blotting.

2.2.8. Cell culture studies A mouse connective tissue fibroblast cell line, L929 (DSM, Germany) was grown in DMEM (Gibco, Germany) containing supplemental 10% fetal calf serum (Gibco) and 2 mmol glutamine in an incubator at 378C, 95% r.h. and 10% CO 2 . Polymer extracts were prepared according to the United States Pharmacopoeia using 100 mg powdered polymer per ml serum supplemented culture medium (DMEM) containing 2 mmol glutamine. The extracts were incubated at 378C (10% CO 2 ) for 24 h. If necessary the extracts were neutralized and sterilized by filtration. The L929 cells were seeded into 96-well microtiter plates at a density of 2000 cells / well. After 24 h the culture medium was replaced with different dilutions of the extracts. After an incubation period of 4 days the viability of the cells was evaluated by the MTT assay. Tin stabilized poly(vinyl chloride) (PVC) was used as positive control and poly(ethylene) powder as negative control (Riblene, Germany).

3. Results and discussion

3.1. PVAL-g-PLGA polymers The aim of this study was the evaluation of PVAL-

11

g-PLGA as biodegradable polymer for the preparation of protein loaded MS and to determine which polymer properties affect both the microencapsulation process itself and MS drug release. Polymer properties are summarized in Table 1. The comb architecture of the novel PVAL-g-PLGA offers additional degrees of freedom for manipulating polymer properties by the introduction of PVAL as hydrophilic polymer backbone. Factors contributing to MS protein release are, amongst others, balance of hydrophilic and hydrophobic components, thermomechanical properties, such as crystallinity and glass transition temperature (T g ), as well as degradation (loss of molecular weight) and erosion (mass loss) kinetics. Polymer glass transition temperature (T g ) has an important impact on the preparation of microspheres as well as on drug release [18]. As shown in Table 1 the difference of T g between the different PVAL-gPLGA was relatively small (658C) characterized by increasing T g with increasing PLGA chain length and decreasing PVAL Mw . Since the double emulsion method exposes polymers to water for several hours the T g of PVAL-g-PLGA films were analyzed after incubation in water as well. Polymer films were preferred to microspheres because of easier handling after incubation. In contrast to PLGA, the wet state of PVAL-g-PLGA led to a significant reduction in T g which occurred faster and to a higher extent for polymers with a higher PVAL / PLGA ratio. For example the T g of PVAL-g-PLGA-1 decreased from

Table 1 Physicochemical properties of the polyesters Polymer-batch a

Mw PVA (kg / mol)

PVA-OH:dimer (feed) (mol:mol)

End:chain (NMR) (%)b

Mn (kg / mol) (feed)c

Mn (kg / mol) (NMR)c

Tg (8C)

T g (wet)d (8C)

Intr. visc. (dl / g)

OH e (mol.%)

PVAL-g-PLGA-1 PVAL-g-PLGA-2 PVAL-g-PLGA-3 PVAL-g-PLGA-4 PVAL-g-PLGA-5 PVAL-g-PLGA-6 PVAL-g-PLGA-7 PLGA

15 15 15 15 6 20 20 –

2.6:100 13.0:100 26.0:100 78.5:100 23.6:100 21.9:100 2.2:100 0:100

1.60 5.99 9.60 19.68 8.20 10.07 1.0 –

1517 307.7 171.1 64.99 62.77 216.9 2011 –

1254 360.3 237.8 134.4 83.57 308.7 1135 –

39.7 37.0 34.0 34.2 37.0 35.0 41.1 45.0

29.8

0.60 0.31 0.17 0.11 0.50 n.d. n.d. n.d.

– – 53 46 68 75 – –

a

LA:GA in all batches |50:50. PLGA end to chain groups ratio determined from 1 H-NMR by signal intensity comparison. c Calculated assuming complete conversion of PVA hydroxyl groups. d Measured after a 24-h incubation in water. e Unesterified PVA-OH groups, calculated from inverse gated decoupling 13 C-NMR. b

14.6 ,0

12

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

39.78C to 29.88C, whereas T g of the polymer with the shortest PLGA chain lengths (PVAL-g-PLGA-4) already was below 08C. Due to the influence of the hydrophilic backbone and shorter PLGA chains, PVAL-g-PLGA have a higher affinity to polar solvents compared to PLGA resulting in a better wettability and a more rapid water uptake.

3.2. Microsphere preparation The preparation of MS at room temperature led mainly to aggregated MS (Fig. 2a) due to the plasticizing effect of the water. This effect resulted in insufficient mechanical stability of the PVAL-gPLGA during preparation of MS. Therefore, the temperature in all preparation steps was reduced to 148C and centrifugation duration and force had to be carried out in a gentle way as pointed out in

Section 2.2.1. These modifications stabilized the microsphere structure during manufacturing and separation resulting in a smooth, non-porous surface of spherical MS, similar to that observed with PLGA (Fig. 2b and c). Yields were in the range of 60–84%, which is acceptable for a laboratory scale encapsulation process (Tables 2 and 3). While the encapsulated protein and the theoretical drug loading did not affect the amount of microspheres obtained, the polymer type slightly influenced the yields. The lower yield of PVAL-g-PLGA MS could be due to the process conditions that could not totally prevent microsphere aggregation and to the slower rate of solidification or hardening. However, in comparison to PLGA, the preparation of MS from linear PLGA having a similar Mw as the PLGA side chains of the PVAL-gPLGA is impossible. Therefore the great advantage of PVAL-g-PLGA lies in the fact that through a

Fig. 2. Scanning electron micrographs of microspheres prepared from PVAL-g-PLGA (a) under standard conditions (room temperature), (b) at 148C, and (c) microspheres prepared from linear PLGA.

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

13

Table 2 Characteristics of FITC-dextran loaded microspheres prepared from PLGA and PVAL-g-PLGA under reduced temperature (148C)a Microsphere batch

Theoretical drug loading (%)

Yield (%)

Total drug loading (%)

Encapsulation efficiency (%)

D[4.3] (mm)

(I) Variation of polymer composition FD-1 1 FD-2 2 FD-3 3 FD-4 4 FD-5 5 FD-6 6 PLGA-1 PLGA

1.0 1.0 1.0 1.0 1.0 1.0 1.0

62 66 64 63 67 60 74

0.8260.04 0.9060.01 0.6960.01 0.9760.03 0.8860.04 0.3160.01 0.8560.04

8264 9061 6961 9763 8864 3161 8564

19.61 13.50 13.49 –b 11.40 13.00 30.12

(II) Variation of theoretical loading FD-3 3 FD-5 5 FD-7 3 FD-8 5 FD-9 3 FD-10 5 FD-11 3 FD-12 5

1.0 1.0 2.5 2.5 5.0 5.0 10.0 10.0

64 67 77 80 83 84 70 75

0.6960.01 0.8860.04 2.4960.16 2.4660.06 4.7860.17 4.9960.04 5.8360.06 7.3460.25

6961 8864 10066 9862 9663 10061 5861 7362

13.49 11.40 12.15 12.48 12.28 13.48 11.42 12.34

a b

PVAL-PLGA polymer

Variation of the microsphere batches is expressed with bold letters. Determination with laser light scattering impossible due to the formation of microsphere aggregates in the presence of water.

modification of the polymer structure by grafting PLGA on a PVAL backbone very short PLGA chains, becoming water-soluble after few cleavage steps, can be used for microencapsulation.

3.3. Particle size PLGA as well as PVAL-g-PLGA microspheres were prepared according to the same standard preparation protocol to exclude any influence of technological aspects. Tables 2 and 3 summarize the mean diameter of microspheres prepared from different polymers and with different encapsulated proteins as well as different protein loadings. The mean particle diameter (D[4.3]) of PLGA MS was found to be 30 mm, whereas PVAL-g-PLGA MS show much smaller mean diameters in the range of 11–19 mm. MS from polymer PVAL-g-PLGA-4 could not be measured by laser light scattering due to their aggregation tendency in contact with water. However, comparable particle sizes in the range of 15–20 mm were confirmed by SEM. The decrease of PVALg-PLGA microsphere size was caused by lower viscosity of the organic (o) phase which can directly

be attributed to the different polymer architecture. Although the Mw of PVAL-g-PLGA is much higher than that of linear PLGA, PVAL-g-PLGA exhibited lower solution viscosities (Table 1) as a consequence of the branched structure, which causes smaller hydrodynamic volumes in solution. Furthermore, an increase in Mw of linear polymers facilitates the formation of stable entanglements in solution, whereas only weak aggregates are likely for PVAL-gPLGA due to their structure. Thus, changes in the Mw of PVAL-g-PLGA did not lead to large differences in viscosity as seen for PLGA. The type of incorporated hydrophilic macromolecule or protein, had no effect on the average microsphere size.

3.4. Encapsulation efficiency The incorporation of proteins into microspheres depends on the partition coefficient of the protein between the dispersed phase and the continuous phase and the coefficient of protein diffusion through the dispersed phase [19]. In this study, the encapsulation efficiency of drug incorporated in the microspheres was strongly affected by the polymer com-

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

14

Table 3 Characteristics of protein loaded microspheres prepared from PLGA and PVAL-g-PLGA under reduced temperature (148C)a Microsphere batch

PVAL-PLGA polymer

Yield (%)

Total drug loading (%)

1 1 1 1 1 1 1

57 64 64 74 65 62 79

0.6660.05 0.7360.07 0.7260.05 0.8060.04 0.6960.08 0.3060.01 0.8660.05

6665 7367 7265 8064 6968 3061 8665

13.54 13.68 13.20 –b 12.72 13.64 32.35

(II) Variation of theoretical drug loading BSA-3 3 1 BSA-5 5 1 BSA-7 3 5 BSA-8 5 5 BSA-9 3 10 BSA-10 5 10

64 65 67 67 65 66

0.7260.05 0.6960.08 3.7160.35 4.4960.40 4.9260.58 6.4860.87

7265 6968 7467 9068 4966 6569

13.20 12.72 12.46 13.67 12.84 12.67

(III) Variation of protein molecular weight and isoelectric point Cyt c-1 3 1 Cyt c-2 5 1 OVA-1 3 1 OVA-2 5 1 BSA-3 3 1 BSA-5 5 1

71 69 65 66 64 65

1.0260.07 1.0460.02 0.6360.09 0.7560.10 0.7260.05 0.6960.08

10267 10462 6369 75610 7265 6968

13.68 12.99 13.97 11.26 13.20 12.72

(I) Variation of polymer composition BSA-1 1 BSA-2 2 BSA-3 3 BSA-4 4 BSA-5 5 BSA-6 6 PLGA-2 PLGA

a b

Theoretical drug loading (%)

Encapsulation efficiency (%)

D[4.3] (mm)

Variation of the microsphere batches is expressed with bold letters. Determination with laser light scattering impossible due to the formation of microsphere aggregates in the presence of water.

position (Tables 2 and 3). Surprisingly, we found that a decrease in PLGA chain length resulted in an increase of the incorporated drug from 82 to 97% for FITC-dextran and from 66 to 80% for BSA-loaded microspheres. As already mentioned, the T g of the PVAL-gPLGA decreased in the presence of water. Therefore, during microsphere formation the plasticized polymer matrix becomes sticky causing MS aggregation. Moreover, the increasing hydrophilicity of PVAL-gPLGA with decreasing PLGA chain lengths promotes hydration of the protein thus influencing the partition coefficient. The higher amount of encapsulated drug found for PVAL-g-PLGA with short PLGA chains can therefore be explained by a reduced extraction of protein into the continuous phase during the emulsification and hardening step. An increase in the PVAL Mw to 20,000 g / mol (PVAL-PLGA-6), however, yielded an encapsulation

efficiency of less than 32%, probably due to a higher hydrophilicity of the polymer, leading to a hydrogellike structure within the particles, facilitating protein release via diffusion already during MS solidification. The theoretical drug loading also had a substantial effect on the encapsulation efficiency. While the incorporation of up to 5% of protein yielded total drug loadings of more than 70%, less than 70% were encapsulated at a theoretical drug loading of 10%. This result could be the consequence of an inhomogeneous distribution of the protein within the polymeric matrix. Consequently, excess protein concentrates in the vicinity of the microsphere surface and easily diffuses into the external medium during solidification. In all polymers studied so far the physico-chemical properties of encapsulated substance, proteins or polysaccharides, did not influence the encapsulation

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

efficiency in a significant manner. This seems to indicate that the compatibility of hydrophilic drugs with PVAL-g-PLGA is improved compared to PLGA. Therefore, these new biodegradable polyesters could facilitate the development of parenteral depot systems for proteins. In particular, high encapsulation efficiencies are desirable for expensive drugs, such as peptides, proteins and oligonucleotides.

3.5. In vitro release properties of PVAL-g-PLGA microspheres containing proteins The influence of variations in the PLGA chain length on the cumulative protein release profile of BSA-loaded MS is shown in Fig. 3. The extent of the initial protein release, also designated as ‘drug burst’ from MS, was defined as drug release within the first 6 h. Immediately after incubation, water uptake into MS occurred, faster and to a higher extent for PVALg-PLGA with shorter PLGA chains [20], thus reducing the T g of the polymers (see Section 3.1). Consequently the MS became sticky and the pore structure of the MS surface was reduced. A porous structure of the surface, however, is necessary for the initial drug release phase as the burst release is normally caused by drug molecules located at or close to the surface of MS. In the absence of pores the very low initial burst of polymer PVAL-g-PLGA4 (1.7%) and the increasing burst of those MS prepared from polymers with longer PLGA chains (39.2% for PVAL-g-PLGA-2 and 24.7% for PVALg-PLGA-3) could be explained by this plasticizer effect. In contrast, the burst of linear PLGA microspheres increases with decreasing Mw and yielding a higher drug burst [19]. In the case of PLGA, this initial rapid release is generally followed by a lag-phase, characterized by a slow release rate of protein, and a phase with a rapid release rate resulting from polymer erosion [5]. The long hydrophobic PLGA chains are the dominating component in the polymer 1 leading to a PLGA-like release profile, avoiding burst-like profiles in the erosion phase. Drug release rates from the other PVAL-g-PLGA were significantly improved resulting in more continuous release profiles. Plotting the slope of the logarithmic rates of drug remaining in the microspheres (calculated without the initial burst) (Fig. 3b) against polymer composition demonstrated

15

that the release characteristics strongly depended on polymer PLGA chain length and hydrophilicity of the branched polyester, since a nearly linear relationship between these properties was observed (Fig. 3c). Therefore, the highest release rate was found for polymer 4, followed by the polymers with increasing PLGA chain length. The combination of low burst effects in combination with high release rates emphasizes the innovative aspects of PVAL-g-PLGA microsphere characteristics compared to linear PLGA for the encapsulation of therapeutic relevant proteins. At constant Mw of the PVAL backbone, shorter PLGA chain lengths resulted in a substantial increase of the release rates. By contrast, release profiles of MS prepared from PVAL-g-PLGA with increasing PVAL molecular weight showed a change from erosion controlled release profiles to a pore diffusion type release with increasing Mw of the PVAL backbone, documented by a substantial increase in the release rates (Fig. 4). In the case of polymer 6 nearly 80% of encapsulated protein was released within 14 days, whereas the same amount of drug was released after 55 days for PVAL-g-PLGA-3. The high Mw of PVAL-g-PLGA-6 even led to a profile being similar to that of star-branched block copolymers [11]. These release profiles can be explained by different hydrophilicities of these PVAL-g-PLGA. As demonstrated by inverse gated decoupling 13 C NMR ca. 50% of the hydroxyl groups of the 15,000 g / mol backbone were esterified, while in the case of the 20,000 g / mol PVAL nearly 75% free OH groups were found, allowing faster hydration and pore formation. The hydrated PVAL domains increased protein diffusion through these domains. Fig. 4 illustrates that similar release profiles can be achieved for BSA as well as for other hydrophilic macromolecules, such as FITC-dextran, indicating that interactions between the polymer matrix and encapsulated protein are not the predominating factor but the polymer matrix. Release rates of FITCdextran also demonstrated that release can be controlled over a wide range of release profiles lasting from 14 days up to 80 days by selecting appropriate polymer compositions of PVAL-g-PLGA. As protein release rates were similar to the release of FITCdextran all protein release studies were not carried out longer than 35 days. In addition to properties of the polymer matrix the

16

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

Fig. 3. In vitro release of BSA from microspheres prepared from PVAL-g-PLGA as a function of PLGA chain length. Theoretical drug loading 1%. (a) Cumulative release. (b) Semi-logarithmic plot of release data. (c) Dependency of initial burst and release rate on polymer composition.

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

17

Fig. 4. In vitro release profile of BSA and FITC-dextran from microspheres prepared from PVAL-g-PLGA of varying Mw of the core PVAL. Standard deviations were all below 2%. Theoretical drug loading 1%.

release depends on specific properties of the drug molecule such as diffusivity in the core material, solubility, the size and stability of the drug molecule, the distribution of drug in the microsphere matrix and specific interaction with the polymer [21]. For this reason, further release experiments were performed to probe differences in release characteristics caused by proteins of different Mw and IEP. As shown in Fig. 5, the release of OVA from MS ceased after 1 day presumably due to protein aggregation. Release profiles of BSA and Cyt c slightly differed in extent of the initial burst release, whereas the following release phase was very similar. Protein– polymer interactions seem to be negligible. To achieve comparable release rates for proteins with different molecular structure, Mw or IEP, polymers based on PVAL-g-PLGA offer a great flexibility for parenteral delivery systems. Depending on the theoretical drug loading, the initial drug release within 1 day or burst release could be varied from less than 16% to more than 50% (Fig. 6). The initial drug release of BSA from MS increased with increasing protein entrapment inside the particles. Moreover, PVAL-g-PLGA with higher Mw of the core PVAL showed a higher initial protein release in all cases of different drug loadings.

Only small differences were observed in the later phases, with a slightly higher release rate for MS with a low amount of encapsulated drug. The fraction of protein available for an immediate release from the microspheres depends on the compatibility of the protein with the polymeric matrix. High drug loadings, however, render a homogenous distribution within the microspheres more difficult thus leading to an increase in the fraction of protein located close to the surface. Furthermore, at the onset of polymer erosion a higher residual amount of protein is still available for release. These results demonstrate the flexibility of PVAL-g-PLGA because wide variations of the drug loadings are possible for controlled release devices. Specific aspects concerning protein stability within the MS and specific polymer–protein interactions, both probably leading to an incomplete release of protein are of interest in ongoing investigations.

3.6. PVAL-g-PLGA cytotoxicity in fibroblasts L929 To characterize the polymer cytotoxicity, extracts were prepared from powdered polymer samples, according to USP XXIII. Samples were incubated with medium at 378C for 24 h, whereby leachable

18

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

Fig. 5. In vitro release profile of BSA, Cyt c and OVA from microspheres prepared from PVAL-g-PLGA of different composition. Theoretical drug loading 1%.

toxic components, e.g. catalysts, residual monomers and degradation products, are extracted into the medium. The L929 fibroblasts were then incubated

for 4 days to ensure a contact of the extracts with the cells over several cell cycles in their exponential growth phase. The MTT assay was used to monitor

Fig. 6. In vitro release profile of BSA from PVAL-g-PLGA microspheres of varying theoretical drug loading (1%, 5% and 10%).

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

19

Fig. 7. Cytotoxicity of the carrier polymers by means of MTT assay.

cell viability. Extracts from poly(ethylene) (PE) and toxic poly(vinyl chloride) (PVC) were used as negative and positive control. The results obtained by the extraction method are shown in Fig. 7 and indicate that linear PLGA does not decrease the viability and proliferation of L929 fibroblasts confirming the well-documented biocompatibility of this polymer. The cytotoxicity of comb PVAL-g-PLGA was quite comparable. The highest concentration of 100 mg / ml led to a reduction of cell viability of ca. 20% similar to linear PLGA. Higher amounts of lactide in the side chains slightly increased cell toxicity, while shorter lactone chains decreased the toxic effects of these polymers (data not shown). Nevertheless, in all cases the extraction conditions recommended by the USP were found to be unsuitable to determine the IC 50 value (concentration that leads to a reduction of cell viability of 50%) since they were always higher than 100 mg / ml extraction medium. These results confirm the nontoxic character of the novel polyesters.

4. Conclusion From a technological point of view PVAL-g-

PLGA are suitable biodegradable polymers for the microencapsulation of hydrophilic macromolecules as well as for proteins using the w / o / w double emulsion method. Reduction of temperature to 148C is a prerequisite for the encapsulation process due to thermochemical properties of PVAL-g-PLGA in the presence of water decreasing the T g of the devices. Yields and encapsulation efficiencies were found to be in the same range as for PLGA microspheres. Apart from the release profiles, microsphere characteristics were strongly influenced by polymer properties such as the grafted structure, PLGA chain length and Mw of the PVAL backbone. The insertion of the hydrophilic backbone PVAL into hydrophobic PLGA chains significantly improved the release of drugs from microspheres towards a more continuous release profile with higher release rates. A correlation between PLGA chain length and release rate was observed. Drug– polymer interactions seem to be reduced, and the release characteristics were not significantly different when proteins of various structure, hydrophobicity and charge were microencapsulated. In summary, release data demonstrated the versatility of this new class of biodegradable polyesters, allowing modifications of the PLGA chain lengths on

20

K. Frauke Pistel et al. / Journal of Controlled Release 73 (2001) 7 – 20

the one hand and variations of the Mw of the hydrophilic backbone PVAL on the other hand. These polymer characteristics affect both the initial drug release and the following erosion phase, which can be utilized to design appropriate release properties for parenteral drug delivery systems by modification of the biodegradable polyester. The initial data on cytotoxicity seem to be promising for the design of biodegradable protein delivery systems.

[10]

[11]

[12]

Acknowledgements Financial support of the project Ki 592-I-I by Deutsche Forschungsgemeinschaft is gratefully acknowledged.

[13]

References

[14]

[1] J. Heller, Controlled release from poly(orthoesters), J. Control. Release (1985) 167–177. [2] K.W. Leong, J. Kost, E. Mathiowitz, R. Langer, Polyanhydrides for controlled release of bioactive agents, Biomaterials 7 (1986) 364–371. [3] D.H. Lewis, Controlled release of bioactive agents from lactide / glycolide polymers, in: M. Chasin, R. Langer (Eds.), Biodegradable Polymers as Drug Delivery Systems, M. Decker, New York, 1990, pp. 1–41. [4] F. von Recum, Handbook of Biomaterials Evaluation: Scientific, Technical and Clinical Testing of Implant Materials, Macmillan, New York, 1986. [5] D. Bodmer, T. Kissel, E. Traechslin, Factors influencing the release of peptides and proteins from biodegradable parenteral depot systems, J. Control. Release 21 (1992) 129–138. [6] R.A. Siegel, R. Langer, Mechanistic studies of macromolecular drug release from macroporous polymers. II: Model of slow kinetics of drug release, J. Control. Release 14 (1991) 153–167. [7] S.S. Shah, Y. Cha, C.G. Pitt, Poly(glycolic acid-co-DL-lactic acid): diffusion or degradation controlled drug delivery?, J. Control. Release 18 (1992) 261–270. [8] M. Morlock, H. Koll, G. Winter, T. Kissel, Microencapsulation of rh-erythropoietin using biodegradable poly( DL-lactideco-glycolide): protein stability and the effects of stabilizing excipients, Eur. J. Pharm. Biopharm. 43 (1997) 29–36. [9] Y.X. Li, T. Kissel, Synthesis and properties of biodegradable

[15]

[16] [17]

[18]

[19]

[20]

[21]

ABA triblock copolymers consisting of poly( L-lactic acid) or poly( L-lactic acid-co-glycolic acid) A blocks attached to central B blocks, J. Control. Release 27 (1993) 248–257. ¨ T. Kissel, Y.X. Li, C. Volland, S. Gorich, R. Koneberg, Parenteral protein delivery systems using biodegradable ABA-block copolymers, J. Control. Release 39 (1996) 315– 326. K.F. Pistel, B. Bittner, H. Koll, G. Winter, T. Kissel, Biodegradable recombinant human erythropoietin loaded microspheres prepared from linear and star-branched block copolymers: influence of encapsulation technique and polymer composition on particle characteristics, J. Control. Release 59 (1999) 309–325. Y. Li, J. Nothnagel, T. Kissel, Biodegradable brush-like polymers from poly( DL-lactide) or poly( D,L-lactide-co-glycolide) and charge-modified, hydrophilic dextrans as backbone — synthesis, characterization and in vitro degradation properties, Polymer 38 (1997) 6197–6206. Y. Li, C. Volland, T. Kissel, Biodegradable brush-like polymers from poly( DL-lactide) or poly( D,L-lactide-co-glycolide) and charge-modified, hydrophilic dextrans as backbone — in-vitro degradation and controlled release of hydrophilic macromolecules, Polymer 39 (1998) 3087–3097. A. Breitenbach, T. Kissel, Biodegradable comb polyesters. Part 1: synthesis, characterization and structural analysis of poly(lactide) and poly(lactide-co-glycolide) grafted onto water-soluble poly(vinyl alcohol) as backbone, Polymer 39 (1998) 3261–3271. Y. Ogawa, M. Yamamoto, H. Okada, T. Yashiki, T. Shimamoto, A new technique to efficiently entrap leuprolide acetate into microcapsules of polylactic acid or copoly(lactic / glycolic) acid, Chem. Pharm. Bull. 36 (1988) 1095– 1103. H. Okada, H. Toguchi, Biodegradable microspheres in drug delivery, Crit. Rev. Ther. Drug Carrier Syst. 12 (1995) 1–99. H. Sah, A new strategy to determine the actual protein content of poly(lactide-co-glycolide) microspheres, J. Pharm. Sci. 86 (1997) 1315–1318. R. Jalil, J.R. Nixon, Microencapsulation using poly( L-lactic acid) III: Effect of polymer molecular weight on the microcapsule properties, J. Microencaps. 7 (1990) 41–52. R.C. Mehta, B.C. Thanoo, P.P. DeLuca, Peptide containing microspheres from low molecular weight and hydrophilic poly( D,L-lactide-co-glycolide), J. Control. Release 41 (1996) 249–257. A. Breitenbach, K.F. Pistel, T. Kissel, Biodegradable comb polyesters. Part II: Erosion and release properties of poly(vinyl alcohol)-g-poly(lactic-co-glycolic acid), Polymer 41 (2000) 4781–4792. H. Okada, M. Yamamoto, T. Heya, Y. Inoue, S. Kamei, Y. Ogawa, H. Toguchi, Drug delivery using biodegradable microspheres, J. Control. Release 28 (1994) 121–129.