Journal of Controlled Release 172 (2013) 1002–1010

Contents lists available at ScienceDirect

Journal of Controlled Release journal homepage: www.elsevier.com/locate/jconrel

A multifunctional bilayered microstent as glaucoma drainage device Christian Wischke a,b, Axel T. Neffe a, Bui Duc Hanh a, Christine F. Kreiner c, Katrin Sternberg d, Oliver Stachs e, Rudolf F. Guthoff e, Andreas Lendlein a,b,⁎ a

Institute of Biomaterial Science, Helmholtz-Zentrum Geesthacht, Kantstrasse 55, 14513 Teltow, Germany Berlin-Brandenburg Center for Regenerative Therapies, Teltow, Germany S&V Technologies AG, Neuendorfstr. 20a, 16761 Hennigsdorf, Germany d Institute for Biomedical Engineering, University of Rostock, F.-Barnewitz-Str. 4, 18119 Rostock, Germany e Department of Ophthalmology, University of Rostock, Doberaner Str. 140, 18057 Rostock, Germany b c

a r t i c l e

i n f o

Article history: Received 28 June 2013 Accepted 17 October 2013 Available online 24 October 2013 Keywords: Multifunctional implant Glaucoma drainage microstent Spatially directed drug release Degradable polymer Suprachoroidal implant

a b s t r a c t Commercial non-degradable glaucoma implants are often associated with undesired hypotony, fibrosis, long term failure, and damage of adjacent tissues, which may be overcome by a multifunctional polymeric microstent for suprachoroidal drainage. This study reports the design and fabrication of such devices with tailorable internal diameters (50–300 μm) by solvent-free, continuous hot melt extrusion from blends of poly[(ε-caprolactone)-coglycolide] and poly(ε-caprolactone) [PCL]. A spatially directed release was supported by bilayered microstents with an internal drug-free PCL layer, and a quantitative description of release kinetics with diclofenac sodium as model drug was provided. Furthermore, the slow degradation pattern (N 1 year) was analyzed and potential effects of 1–5 wt.% drug loading on material properties were excluded. Translational aspects including sterilization by γ-irradiation on dry ice, in vitro biocompatibility, and in vivo implantation were addressed. The promising results support further functional analysis of long-term in vivo performance and suppression of disadvantageous capsule formation. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Effective and safe drug delivery remains a major challenge in the treatment of eye diseases [1,2]. For many types of pathologies, surgical intervention is required, which inherently creates a need for a local drug therapy to modulate cellular response on surgical traumata [3]. Accordingly, regenerative medicine requires multifunctional implants, which can be translated into a therapy and address simultaneously the diverse clinical needs, which is not given for existing systems. An example of eye diseases with high incidences throughout the world and a strong impact in health economics are glaucomas. Glaucomas are typically associated with increased intraocular pressure (IOP) and a disturbed outflow of steadily produced aqueous liquid due to morphological changes (open-angle glaucoma) or obstruction (closed-angle glaucoma) of tissues. The resulting pressure-induced structural and functional defects of the retinal ganglion cells and the optical nerve head can lead to blindness [4]. Surgical intervention may be required, e.g. if topical medication fails [5] and laser trabeculoblasty [6] will not provide a suitable IOP reduction in open-angle glaucoma. In such cases, penetrating glaucoma surgery is recommended to create ⁎ Corresponding author at: Institute of Biomaterial Science, Helmholtz-Zentrum Geesthacht, Kantstrasse 55, 14513 Teltow, Germany. Tel.: +49 3328 352 450; fax: +49 3328 352 452. E-mail address: [email protected] (A. Lendlein). 0168-3659/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jconrel.2013.10.021

new functional outflow pathways, aiming to establish a status where patients do not need additional medication [7]. Besides trabeculectomy with decreasing case numbers and clinical success [8,9], glaucoma drainage devices (aqueous shunts) as a biomaterial based approach can be used [10,11]. Such devices enable percolation of aqueous liquid through a tube to a filtering plate in the subconjunctival space (e.g., Molteno®; Baerveldt®), to the Schlemm's canal (e.g., Eyepass™), or to the suprachoroidal space (e.g., Solx®). However, while the indications for aqueous shunts have broadened, there are a number of unsolved clinical drawbacks of existing systems [12]. To better fulfill distinct requirements as described below, a number of scientific, conceptual, technological, and surgical challenges need to be addressed by an aqueous shunt device that exhibits multifunctionality. A first requirement is a suitable regulation of aqueous outflow that does not cause hypotony after implantation as is the case for the available devices, mostly due to their large internal tube lumina (e.g., 300 μm) [13]. The occlusion of the device tube with a degradable suture as presently performed intraoperatively by ocular surgeons [11,14] is a difficult to standardize process. It also bears the risk of increasing IOP if the suture is not removed in time. Depending on the IOP of the patient and the aimed site of drainage (subconjunctival, suprachoroidal), inner diameters of 40–70 μm are required [13], which is a challenge for polymer processing. Second, fibrous capsule formation in response to surgical trauma needs to be avoided. Otherwise the outflow may be ruled by the fibrous

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

capsules with interindividual, time-, and product-specific differences in thickness rather than by the device [15,16]. Capsule formation can presently not be prevented and is addressed by additional medication, needling of the capsule, capsule excision [17], or placement of an additional implant. All these actions contradict the desired regenerative approach. Therefore, a modulation of cellular responses needs to be achieved by bioactive molecules, which should i) be delivered locally rather than orally [18], ii) be available directly at the implant surface rather than in a separate ocular depot [19], and iii) be released in a sustained rather than rapid fashion [20,21]. Additionally, we propose that the drug release should be spatially directed to the site of cellular responses, i.e., to the outside of the tube rather than to the lumen. Drug embedment, however, may adversely affect the properties of the implant's matrix materials, which is a scientific challenge to be addressed by proper selection of the matrix polymers. Third, the long-term presence of drainage implants with mechanical properties that cannot match all surrounding tissues, particularly when having to fulfill a supportive role, is known to result in tissue damage. Therefore, the device should be degradable rather than biostable, ideally leaving back a functional flow pathway as aimed for in a regenerative approach. Importantly, hydrolytic matrix degradation needs to be well controlled and slow. Considering possible interference of different levels of embedded drugs with diffusion processes of water and polymer degradation products, the selection of a suitable polymer again is an essential prerequisite to establish multifunctionality. Fourth, a device designed for intrascleral implantation with a drainage of aqueous to a suprachoroidal rather than a subconjunctival location should allow the minimization of device-induced damage to the conjunctiva, particularly their erosion and eventual penetration as in the case of tubes of existing commercial subconjunctival drainage devices [12]. Finally, as a fifth point, the implants should be biocompatible, suitable for handling during implantation, sterilizable, and compatible with industrial production processes (e.g., hot-melt extrusion rather than dip coating), which are all essential preconditions for a translation in regenerative therapies. Based on these scientific requirements and compared to previous approaches [22,23], multifunctionality of a polymeric suprachoroidal glaucoma implant might be achieved by establishing i) a tubular structure, herein called microstent, with small orifice sizes as required for an optimal drainage functionality, ii) an integrated drug release functionality without impeding material properties desired for structural function, iii) a spatially oriented direction of drug release to the implant surface, which should be realized by a microstructuring of the microstent with a challenging bilayered design, iv) hydrolytic degradation functionality for slow implant removal to enable tissue regeneration, and v) biocompatibility and feasibility for implantation as sterilized devices (Fig. 1).

Fig. 1. Scheme of multifunctional microstent as glaucoma drainage device.

1003

This concept was evaluated using poly(ε-caprolactone) [PCL] and its blend with poly[(ε-caprolactone)-co-glycolide][PCG]. This selection was based on the recently observed interesting mechanical properties [24], the improved PCG degradability due to glycolide diads acting as weak links [25,26], and its proposed semi-crystallinity that should allow reducing possible interference of drug payload and material properties and degradation. This study covers the microstent design and fabrication with tailorable orifices, the analysis of spatially directed release and quantitative description of release kinetics with diclofenac sodium as model drug, the analysis of degradation pattern, potential drug loading effects on material properties, and translational aspects including sterilization, in vitro biocompatibility, and in vivo implantation. 2. Material and methods 2.1. Materials Diglycolide (≥ 99%), ε-caprolactone (N99%), (Bu)2SnO (98%, all Aldrich), 1,8-octanediol (N99%, Fluka), and ethyl acetate (N99%, Acros) were used in polymer synthesis. The employed poly(ε-caprolactone) was Capa® brand material from Solvay (now Perstorp, Warrington, UK). Diclofenac sodium (99.3%) was purchased from Fagron (Barsbuettel, Germany). For high performance liquid chromatography (HPLC), solvents of isocratic grade were used (Merck). All other chemicals were of analytical grade. 2.2. Polymer synthesis and coextrusion for blending and drug incorporation Copolymerization of purified diglycolide (recrystallization from ethyl acetate) was conducted in the melt with ε-caprolactone as comonomer, 1,8-octanediol as starter molecule, and (Bu)2SnO as catalyst to yield PCG with a weight average molecular weight Mw of ~23 kDa, a polydispersity of 1.4, and a glycolide content of 8 wt.%. PCG was blended with commercial PCL powder with a Mw of ~80 kDa (Capa® 6808, Perstorp, Cheshire, UK) at a 50:50 weight ratio and subjected to coextrusion at 100 °C with a co-rotating twin screw extruder (Prism Eurolab 16, L/D 25, Thermo Electron, Karlsruhe). A rod die was used with a die temperature of ~55 °C and an extruder speed of 50 rpm. The product was cooled in a water bath at ambient conditions. For drug loading, the required quantity of diclofenac sodium (wt.%) was premixed with the polymer powder and homogenized in the extruder. 2.3. Processing by hot melt extrusion and melt compression Single layer stents were obtained by extrusion of the blend through a vertical catheter die with a width of 1.8 mm and a 1.0 mm core at a die temperature of ~60 °C using a single screw extruder (Gimac TR 14, L/D 24, Castronno, Italy). The extruded stents were stabilized by filling their core with water at room temperature and by immersion in a water bath for cooling. For double layer stents, two single screw extruders were connected to a custom made double layer catheter die (width 1.4 mm, core 0.7 mm, Erocarb, Giez, Switzerland). PCL (Capa® 6800 granules) was fed to the inner stent layer by a Polylab system at 130 °C (Haake Rheocord with Rheomex 19, L/D 25, Thermo Electron, Karlsruhe), while the Gimac TR 14 extruder (80 °C) was used for forming the outer layer consisting of the blend material. In order to stabilize the extruded double layer stents, support air with an excess pressure of 30 mbar was applied to the stent core. In both cases, single and double layer stents, the diameter of stents was controlled by adjusting the extrusion and the haul-off speeds. Polymer films were prepared between hot metal plates in a Collin P200E platen press (Collin, Ebersberg, Germany) at 100 °C. Samples were pressed to thin films at 50 bar with ~100 μm spacers (3 min) and subsequently cooled down to 20 °C with persisting exposure to a pressure of 50 bar.

1004

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

2.4. Characterization of polymer properties

t lag ¼

Thermogravimetric analyses (TGA) were conducted with a Netzsch TG 209 instrument (Netzsch, Selb, Germany) between 20 and 600 °C with a heating rate of 10 K · min− 1. Differential scanning calorimetry (DSC) was performed in aluminum pans with pierced lids between − 20 and + 80 °C with heating and cooling rates of 4 K · min− 1 on a Netzsch DSC 204. Tensile tests of samples were conducted with polymer films cut into dumbbell-shaped test specimens at ambient conditions using a Zwick Z2.5 tensile tester (Zwick GmbH & Co. KG, Ulm, Germany) with a 50 N load cell at a pre-load of 0.05 N. The maximum stress σmax as well as the elastic modulus E from the initial slope of the stress–strain curves were obtained. Polymer film samples or pure drug powder (in a polycarbonate foil) were analyzed using a D8 Discovery (Bruker, Karlsruhe, Germany) at 40 kV and 40 mA with a two-dimensional detector (Hi-Star), a wavelength λ = 0.154 nm, a sample-to-detector distance of 150 mm, and an exposure time of 300 s. From isotropic two-dimensional raw data, one-dimensional curves were obtained by integration over an angle χ = 120° and the 2θ range of 7.7 to 42.3° after background subtraction. Curve fitting was performed with the Bruker TOPAS® software using a parameter-free curve for the amorphous phase and the Pearson VII model for the crystalline phase [27]. A crystallinity index Xc was determined from the area of crystalline peaks compared to the sum of peak areas of crystalline and amorphous phases. Scanning electron microscopy (SEM) analysis was conducted with a Gemini Supra™ 40 VP SEM (Carl Zeiss NTS GmbH, Oberkochen, Germany). If the task was to study cross-sections, cryofracturing was performed prior to sputtering with Pt/Pd. Secondary electron detectors were used as standard detectors. By employing a backscattered electron (BSE) detector, drug could be visualized in the polymer by contrast differences in the image. Also, an energy dispersive X-ray (EDX) detector was employed to map the distribution of elements in the samples. 2.5. Drug release, diffusion and degradation studies Drug release studies from 1.5 cm pieces of stents (sample weights: 16–44 mg depending on stent diameter) were conducted at 37 °C in a horizontal shaker (60 rpm, Certomat® IS, Sartorius BBI Systems GmbH, Melsungen, Germany). In 2 ml test tubes, samples were incubated with 1.5 ml of phosphate buffered saline pH 7.4 (PBS; 5.8 mM NaH2PO4, 5.8 mM Na2HPO4, 150 mM NaCl + 0.02 wt.% NaN3), from which 1 ml was withdrawn for HPLC analysis and replaced with fresh medium. Additionally, drug diffusion through initially drug-free, ~100 μm thick polymer films were conducted at 37 °C in side-by-side glass cells (Type 5G-00-00-11.28-05, PermeGear, Hellertown, PA, USA) with a 10 mg · ml−1 drug solution in PBS in the donor chamber and PBS in the receptor chamber. Isocratic HPLC with a 125-4 RP-18 column (LiChroCART® 125-4, LiChrospher® 100, 5 μm; Merck, Darmstadt, Germany) was performed on an Agilent 1200 series HPLC (Agilent Technologies Deutschland GmbH, Böblingen, Germany) at 1.2 ml · min−1 (10:35:55 acetonitrile/methanol/20 mM sodium acetate buffer pH 5.6) with diclofenac detection at 275 nm. The drug permeability coefficient P was obtained by linear regression of curves of the permeation flux dQ/dt under steady state conditions [28] according to Eq. (1), with c0 being the concentration of saturated drug solution in the donor compartment (10 mg · ml−1) and A being the permeated surface area of the films (1 cm2). Additionally, the apparent diffusion coefficient D was estimated following Eq. (2). The extrapolation of the regression curve to its intercept with the abscissa provided the lag time tlag, with h being the thickness of the individual film [28]. dQ ¼ P
c0 A
dt

ð1Þ

h2 6
D

ð2Þ

Degradation studies were conducted with ~ 6 · 20 · 0.12 mm3 film samples with a typical weight of 15–20 mg, which were placed in 15 ml screw cap tubes and incubated with 12 ml PBS in a shaking water bath (GFL 1083, GFL, Burgwedel, Germany) at ~30 rpm and 37 °C. The mass loss during degradation was determined by weighing samples after drying. 2.6. Sterilization The final sterilization protocol involved gamma (γ) sterilization of samples on dry ice with a 28 kGy dose. Additionally, γ-sterilization with 18kGy and 28kGy doses at ambient conditions, steam sterilization (121 °C, 2 bar, 20 min), and ethylene oxide (EO) sterilization (6 vol.% EO in CO2 at 42 °C/180 min or 37 °C/240 min with vacuum degassing) were evaluated. 2.7. Cell culture studies with human fibroblasts Primary human fibroblasts (40 000 cells; 93% viability; Provitro GmbH, Berlin, Germany) were cultured on film samples (13 mm discs) of drug loaded or drug free blend in 1 ml Dulbecco's modified eagle medium supplemented with 10 vol.% fetal calf serum (Biochrom, Berlin, Germany) in 24 well microtiter plates. Living cells were stained with an adapted fluorescein diacetate method (25 μg FDA/ml cell culture medium [29]) before analysis by confocal laser scanning microscopy (LSM 510, Axiovert, Zeiss, Jena, Germany). 2.8. Ocular implantation studies Animal experiments were approved by the Ethics Committee of the University of Rostock. All experiments were performed in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. To evaluate the feasibility of surgical techniques and microstent placement in the eye, a prototype implantation study was performed by an approved animal protocol (LALLF M-V/TSD/7221.32.1-024/06). Samples were implanted in the left eye of three New Zealand White Rabbits (Charles River GmbH, Sulzfeld, Germany), which were anesthetized with a mixture of 30 mg/kg ketamine hydrochloride (Bela-pharm GmbH & Co. KG, Vechta, Germany) and 5 mg/kg xylatine hydrochloride (Rompun; Bayer Health Care, Leverkusen, Germany). After preparing a conjunctival flap at the upper circumference to isolate and cut the superior straight (rectus superior) muscle, a scleral lamellar flap of 3 · 3 mm2 with the basis towards the limbus was created using a diamond knife. By a limbus-parallel incision at the distal end of the scleral flap through the whole sclera, the suprachoroidal space was accessed to insert the microstent. Then, through a small incision under the proximal end of the scleral flap at the limbus, the PCL microstent was introduced into the anterior chamber. Subsequently, the flap was closed with sutures, the muscle was refixed at the insertion, and the conjunctival flap was closed with sutures. The animals were observed for 100days to monitor healing, tissue response, and other potential long-term effects. 3. Results 3.1. Polymer selection and bilayered microstent design with tailored orifices The envisioned glaucoma implant should combine several functions, i.e., a control of drainage rates by a microstent design with independently adjustable inner and outer diameters, a local drug release that is sustained and spatially directed, a controlled degradation, and biocompatibility of sterilized devices. The selection of a suitable polymeric matrix is crucial for all these features. This matrix should be semi-

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

crystalline to exclude plasticization and major alteration of mechanical properties by drug molecules. The polymer should be suitable for melt-based processing to facilitate solvent-free, continuous preparation as desired in medical device industrial fabrication. Based on these requirements, a recently reported blend of poly(ε-caprolactone) [PCL] with a Mw of 23 kDa and poly[(ε-caprolactone)-co-glycolide) [PCG] with a Mw of 80kDa was selected as a semi-crystalline matrix, which advantageously combines the good mechanical properties of PCL with the tailored degradation pattern of PCG [24]. Furthermore, this blend exhibited a suitable melting and solidification pattern for processing by hot melt extrusion. TGA analyses supported the stability of all involved compounds (PCl, polymer blend, model drug diclofenac sodium) at 250 °C with a yield N99 wt.%, as relevant for the involved extrusion or compression techniques. First, microstents consisting of a single layer polymer wall were obtained from hot melt extrusion, where an in-line cooling of the microstent lumen with water allowed the prevention of microstent collapse and the facilitation of rapid solidification. Importantly, the internal and external diameters could be controlled independently by carefully adjusting the instrument specific process parameters like the extrusion speed, haul-off speed, and water filling of the stent lumen during extrusion (Fig. 2A–C). The minimum internal diameters were in the range of 50 μm, while the external diameters should at least be 300 μm to ensure easy handling and mechanical stability of the microstents. In the second step, a custom made double layer catheter die was employed and drugfree PCL was used as inner layer. This resulted in a microstructure with two well-connected polymer layers (Fig. 2D) and no need for water cooling of the stent lumen due to its faster solidification. 3.2. Drug embedment and spatially directed drug release functionality Based on their higher solubility and thus a larger concentration gradient towards the release medium, hydrophilic rather than hydrophobic

1005

model drugs typically result in faster release. Therefore a hydrophilic drug should be used in this study in order to face the more challenging condition. The incorporation of diclofenac sodium as model compound (solubility in PBS ~12 mg · ml−1 at 37 °C) into the matrix material was conducted by coextrusion of drug powder with the PCG and PCL material during polymer blending. SEM analysis was applied to evaluate a homogenous drug distribution as required for reproducibility of the payload of each individual microstent and exclusion of uncontrolled release from large, poorly embedded drug aggregates. Cryofractured films of the drug loaded blend illustrated a homogeneous drug distribution at 1 wt.% payload with only few visible drug particles, which were smaller than 1μm in all cases. When increasing the drug payload to 5wt.%, some islets of drug aggregates (up to 15 μm) were observed (Fig. 3A–C). Similar results were found for microstents, where bright spots or regions of drug aggregates were visualized by material contrast imaging using the backscattered electron detector (see arrows in Fig. 3D). By mapping the distribution of elements in the samples using an EDX detector (Fig. 3C), these contrast differences in SEM images could be correlated with high concentrations of elements derived from diclofenac sodium such as chlorine and sodium. The low number of drug aggregates compared to the drug loading of 5 wt.% suggests that the majority of drug was homogeneously dispersed in the polymer matrix. Quantitative SEM–EDX analysis of areas without visible drug aggregates supports this conclusion by roughly estimating N80% of the 5wt.% drug payload to be homogeneously dispersed in films and microstents. This determination was based on the quantity of drug-specific chlorine atoms related to the total number of atoms in these areas. Higher temperatures for blending and drug loading of the polymers, i.e., 120°C instead of 100°C, improved disaggregation of the drug. However, such treatment was associated with yellow coloration of samples and therefore had to be avoided. In a systematic series of release experiments, first the impact of drug diffusion length in a homogeneous matrix as directly related to the wall thickness should be studied with single layer stents with 5 wt.% drug

Fig. 2. Exemplary SEM images of single layer (A−C) or double-layer (D) microstents loaded with 5 wt.% of diclofenac sodium (in the polymer blend layer) as prepared by hot melt extrusion with a variable adaptation of internal (ID) and external diameter (ED) to clinical requirements [(A) ID: 52 μm, ED: 312 μm; (B) ID: 54 μm, ED: 426 μm; (C) ID: 278 μm, ED: 523 μm; (D) ID: 236 μm, diameter of internal PCL layer: 346 μm, ED: 507 μm].

1006

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

Fig. 3. SEM analysis of drug loading homogeneity in films and stents for 5 wt.% diclofenac sodium payload in PCG/PCL blend polymer. Arrows exemplarily indicate the presence of aggregated drug particles. (A) Cryofractured film samples. (B) Aggregated drug particles. (C) EDX-mapping of chlorine distribution in panel B. (D) Cryofractured stent; drug particles visualized by material contrast (BSE detector).

loading. The stents were of similar internal diameter (~50 μm) and, as expected, exhibited in a more sustained in vitro release pattern over up to 24 h with increasing external diameters (Fig. 4A). When considering the surface of the microstent edges as negligible in a simplified approach, two polymer–water interfaces are available for drug diffusion out of the matrix polymer: i) the inner surface facing the lumen of the stent and ii) the outer surface (see inset of Fig. 4B). In the in vitro release studies, the outer surface should contribute strongest

to the overall diffusion controlled drug release due to its larger size and easier accessibility. However, as illustrated in Fig. 3D, surface-proximal drug particles are also accessible from the lumen of the microstents. Moreover, drug release to the lumen of these stents should become more pronounced in vivo upon permanent rising with the drained aqueous liquid (produced at about 2 μl · min−1 [13]), which removes the boundary layers and increases the drug diffusion rate to the lumen. Introducing an internal drug-free PCL layer (double layer stents) clearly

Fig. 4. Drug release from (A) single layer microstents of ~50 μm internal diameter and different external diameters prepared from the PCG/PCL blend with 5 wt.% drug loading [n = 6, mean, S.D.], (B) 500 μm single vs. double layer microstents (with additional internal PCL layer; see inset) [n = 6, mean, S.D.], (C) drug permeation in side-by-side diffusion cells (see inset, scheme provided by PermeGear Inc., USA) through initially drug-free films of either PCL or blend [normalized plot for comparison of triplicates], and (D) representative fitting of drug permeation curves in the steady-state by linear regression to obtain the permeability coefficient (Eq. (1)) and the apparent drug diffusion coefficient (Eq. (2)).

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

retarded the in vitro drug release, e.g., ~64 wt.% after 12 h compared to ~90 wt.% for single layer stents (Fig. 4B). The complex microstent geometry with curved internal and external surfaces of different sizes as well as contributions of the stents' cutting edges to the release pattern limited a quantitative description of the effect of the PCL layer on the spatial direction of drug diffusion rates. Therefore, initially drug free PCL or blend films were analyzed for drug diffusion kinetics in side-by-side glass cells with planar, well defined sample geometries (Fig. 4C). Since small variations in film thickness result in altered diffusion length, the experimental drug permeation data were normalized to the film thickness for comparability of replicates [30]. The curves showed an exponential increase in drug permeation until a steady state of permeation was achieved. By linear regression of the steady state phase of each individual curve from plots of drug permeation over the time (Fig. 4D), the slope of the curves as well as the lag time tlag could be determined. Based on Eq. (2), the apparent drug diffusion coefficients D were found to be 5.2 ± 0.93 · 10−9 cm2 · s−1 (blend) and 2.7 ± 0.5 · 10−9 cm2 · s−1 (PCL). By Eq. (1), the permeability coefficients P were determined to be 14.9 ± 1.8 · 10−7 cm · s−1 (blend) and 8.8 ± 1.2 · 10−7 cm · s−1 (PCL).

3.3. Drug effects on material properties and hydrolytic degradation Generally, the polymer morphology is strongly affecting the properties and functions of polymeric materials, e.g. mechanical properties and degradation behavior. WAXS analysis illustrated, that the scattering profile of diclofenac sodium exhibited some peaks that overlay with the characteristic scattering of semi-crystalline PCL (Fig. 5A) as well as its blend with PCG. Thus both, drug and polymer contributed to the calculated crystallinity index (Fig. 5B). The drug-free polymer blend showed a higher crystallinity than pure PCL. Loading the blend with either 1 wt.% or 5 wt.% of diclofenac sodium clearly reduced the crystallinity to values similar to that of PCL with and without drug loading. In addition to mass loss and changes of thermal properties studied during hydrolytic degradation, the effects of drug loading on the mechanical properties were explored with standardized film samples suitable for tensile testing. As shown in Fig. 6A, the mass loss was generally slow. Samples loaded with 5 wt.% diclofenac sodium initially showed a faster mass loss, which can be assigned to the drug release. The subsequent curves run parallel to those of drug-free samples. DSC analysis indicated for the second DSC heating run, i.e., after controlled polymer crystallization, an initial increase in melting temperature Tm during the first week of degradation. This similarly occurred for drug-loaded and drug-free materials (Fig. 6B). Subsequently, a slight trend towards a Tm reduction was observed. In contrast, the first DSC run, which corresponds to the thermal properties of native test

1007

specimens as subjected to thermomechanical analysis, revealed a trend towards higher Tm and a larger statistical spread of data points compared to the second DSC heating run (Fig. 6B). The mechanical properties of the films at ambient conditions were determined after up to 12 weeks of degradation, until experimental difficulties occasionally occurred due to the brittleness of samples. When considering the common variation and precision of tensile tests, no clear changes in the E modulus could be concluded (Fig. 6C). For σmax there was a trend to lower values for the unloaded blend, while no clear trend could be observed in the case of increasing drug loading (Fig. 6D).

3.4. Translational aspects: sterilization, biocompatibility, in vivo pilot study Sterility is a key requirement for further exploration of microstents in vitro and in vivo. Considering the thermal properties of the material, steam or heat sterilization was unsuitable. For the explored ethylene oxide sterilization, the temperature had to be reduced from commonly used 42 °C to 37 °C to exclude partial melting of PCG. Since ethylene oxide sterilization at 37 °C was associated with long process times for sterilization and subsequent degassing, efforts were focused on γsterilization. γ-Sterilization at 28 kGy or 18kGy doses resulted in yellow coloration of the polymer samples indicating chemical alterations, which could be avoided by freezing samples in dry ice prior to their irradiation. No Tm alteration could be detected by DSC for drug-loaded and drug-free samples after irradiation in the frozen state, but also not for discolored samples from room temperature irradiation. This indicates that the visually observed degradation products may be present at a rather low extent, which may be the reason why they could not be identified by 1H NMR (data not shown). However, irradiation on dry ice avoided such degradation products, did not change the mechanical properties, and, as confirmed by GPC, did not alter the molecular weight distribution of the material. Primary human fibroblasts were used to evaluate the biocompatibility of the drug-loaded and drug free polymer blend by direct culture on film samples for several days. The material could be categorized as noncytotoxic since the cells showed normal proliferation and the characteristic growth pattern of fibroblasts (Fig. 7A). When exploring the effect of polymer drug loading, an only slightly reduced cell confluence was observed for 1 wt.% drug loading. A drug payload of 5 wt.% clearly effected fibroblast proliferation (Fig. 7B). Microscopic images at higher resolution showed blebs of the cell membrane, which are characteristic for apoptotic cells (Fig. 7C). In a pilot study aiming to prove the feasibility of the microstent implantation in vivo, PCL prototypes were successfully implanted in rabbit eyes and provided the desired connection of the anterior chamber and

Fig. 5. Scattering profile (A) and crystallinity index (B) of diclofenac sodium loaded and drug-free films of PCL and PCG/PCL polymer blend as observed by WAXS analysis (n = 9−12, mean, S.D.).

1008

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

Fig. 6. Properties of films from PCG/PCL polymer blend with and without diclofenac sodium payload during hydrolytic degradation. (A) Mass loss (n = 2−6, mean, S.D.). (B) Melting temperature Tm determined by DSC during the first (native sample) and second (with controlled thermal history) heating run. (C) E modulus of native samples (n = 4−6, mean, S.D.). (D) σmax of native samples (n = 4−6, mean, S.D.).

the suprachoroidal space. One day postoperatively, all eyes showed signs of mild inflammation with conjunctival hyperemia without any fibrin in the anterior chamber. With treatment of topical corticosteroid and antibiotic eye drops, the inflammation was well controlled and disappeared within a few days. The wound healing was normal in the further postoperative course. A slight scleral vascularization was observed (Fig. 8). Due to the dislocation of the implant in one case, it should be noted that a fixation device of the tubes is stringently recommended to avoid dislocation into the anterior chamber. In addition to generally proving the suitability of the surgical routine, the in vivo study confirmed that microstents could be intraoperatively cut to the required size so that providing one-size tubes of 20mm length would be suitable.

4. Discussion 4.1. Multifunctionality of glaucoma microstents Glaucoma drainage surgery may be improved by polymer-based implants that i) have small internal diameters suitable to prevent massive outflow of aqueous humor, ii) enable drug release that is controlled, local, and spatially directed to the surrounding tissue to avoid extensive formation of fibrous capsules, iii) degrade slowly, leaving back a new flow pathway for aqueous humor, iv) facilitate suprachoroidal rather than subconjunctival drainage in order to avoid erosion of the conjunctiva, and v) match the various requirements that critically effect their

Fig. 7. Cell studies with primary human fibroblasts cultured on (A) drug-free blend films and (B−C) films loaded with 5 wt.% diclofenac sodium. Cells were analyzed by confocal laser scanning microscopy after selectively staining vital cells with fluorescein diacetate after 3 days of culture.

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

Fig. 8. Prototype ocular implantation of a diclofenac sodium loaded PCL microstent (highlighted by arrows) for drainage to the suprachoroidal space in rabbit eyes. Images were taken 100 days postoperatively.

translation such as industrial processing, sterilization, biocompatibility, and implantability. Combining these demands into a polymer microstent leads to a multifunctional device. The different properties and functions should be preserved when adding new functionalities such as drug release or subjecting the material to required postpreparation processes such as sterilization. For suprachoroidal drainage, a suitable design would be thin hollow polymer tubes with internal diameters as small as 40–70μm [13], which is a technological challenge. By employing hot melt extrusion as an industrially relevant, efficient, continuous, and solvent-free technique, the goal of small diameters could be achieved (Fig. 2). The processing requirements for stabilization of the microstent lumen were a consequence of the crystallization kinetics of the matrix polymers. While pure PCG could not be extruded to stable tubular structures due to its low Tm of 34 °C, its blend with PCL with a higher Tm facilitated the stabilization of the tube lumen by water. Importantly, for bilayer stents with an inner, faster crystallizing pure PCL layer, the lower heat transmission provided by air compared to water was already suitable for device cooling. In this way, the use of water possibly remaining at traces in the microstents and initiating degradation during storage could be avoided. Generally, fibroblast proliferation around glaucoma implants is a natural response included in the wound healing process after surgical interventions, which results in the formation of scar tissue rather than tissue regeneration. The biofunctionalization of microstents with drugs may enable the modulation of post-surgical cell responses, if their release is local and sustained as explained above. An important criterion for reproducible release pattern is a homogeneous drug distribution rather than large drug aggregates located close to the device surface, as studied in here for diclofenac sodium as a model compound that could be well detected in SEM–EDX analysis (Fig. 3). The physicochemical properties of the drug are largely affecting the release kinetics, e.g., hydrophobic drugs with a low solubility in aqueous media are often released much slower than hydrophilic molecules. This is, e.g., due to potential drug–polymer hydrophobic interactions, low diffusion coefficients, and/or slow dissolution also under sink conditions due to drug saturation in the hydrated polymer matrix [31]. Therefore, a hydrophilic model drug was selected in here as the most challenging condition. It was indicated that even considering the low diffusion length in the thin microstent walls, a tailored release is possible but appears limited in time (Fig. 4A). It should be noted that the volume of release medium for sink conditions was large compared to the aqueous humor and the available interstitial fluid. Therefore, a slower release is expected in the tissue due to drug saturation in boundary layers. When changing

1009

to other drugs such as glucocorticoids, their lower aqueous solubility will support an extension of the release period. It can be expected that the continuous flushing of the single-layer microstent lumen with aqueous humor in vivo will be the major control of drug concentration gradients, resulting in a shortened release period and diminished drug availability at the outer stent surface. Since obstructions of the posterior lumen by fibroblast reaction [22] were an occasionally observed failure of previous non-degradable suprachoroidal drainage devices, it should be noted that drug release to microstent lumen is not generally undesired, but should be controlled as achieved by the double layered design. The additional, initially drug-free PCL layer showed to be an effective tool to adjust release rates (Fig. 4B) already in vitro, where the majority of release occurs to the large outer stent surface that is covered with continuously agitated release medium. The bilayered design can be expected to have even larger effects in vivo. The decreased release may be due to a lower diffusivity of drug molecules in pure semi-crystalline PCL compared to the more hydrophilic glycolide containing PCG/PCL blend and/or an effect of diffusion length, since only the outer blend layer was loaded with drug. Mechanistic insights were provided by diffusion studies in side-byside cells (Fig. 4C–D), which proved a reduced apparent drug diffusion coefficient D and permeability coefficient P for PCL compared to the blend material. This finding suggests that the applied concept of internal layers can limit drug diffusivity and may allow for spatially directed control of the release functionality even for microsized devices. Further studies with other types of drugs including hydrophobic molecules should be performed to confirm the capability for spatially directed release. Polymer morphology, defining the materials' mechanical and thermal properties and drug diffusibility, is of fundamental importance for establishing many device functions and should therefore be preserved. WAXS analysis indicated improved crystallization of the blend compared to pure PCL (Mw = 80 kDa) due to the presence of shorter (23 kDa) and more flexible PCG chains (Fig. 5). The reduced crystallinity index of the PCG/PCL blend after drug loading to values comparable PCL suggested that the drug cannot be well excluded from nascent PCL crystallites in the presence of glycolide-containing, shorter PCG chains. However, these drug effects on the crystallinity index and semicrystalline morphology were marginal, as supported by an absence of changes in the polymers melting enthalpy determined by DSC. Importantly, drug loading had no adverse effect on sample degradation as can be seen from well matching slopes of mass loss curves after initial drug release, congruent data of Tm in DSC studies, and in most cases very similar thermomechanical properties (Fig. 6). Despite the initially observed altered σmax, drug loading was not relevantly affecting this property during degradation. Similarly, no relevant alterations of material properties were observed after γ-sterilization when conducted on dry ice. As it is true for all new drug–polymer combinations, this would need to be verified when alternative drug substances are to be used. Degradability, as a device function, based on the hydrolytic scission of ester bonds, is expected to occur at higher rates for PCG due to glycolide diads acting as a weak link in the polyester backbone [25]. Therefore, a stepwise degradation from outside to inside may be proposed for the double layered microstent, which might support the regenerative approach. While the Tm of samples with a controlled thermal history (second DSC heating run) remained constant, the increasing Tm of the first DSC heating run (also reflecting the state of the samples subjected to mechanical testing) suggested an increasing size and perfection of PCL crystallites (Fig. 6). This supports the expected preferential hydrolysis of glycolide diads leading to better crystallizable PCL-rich polymer segments. 4.2. Demands for clinical translation in regenerative medicine The clinical applicability of novel biomaterial devices holds complex requirements, which are technologically as well as scientifically challenging. It was mentioned before that the suitability for high throughput

1010

C. Wischke et al. / Journal of Controlled Release 172 (2013) 1002–1010

industrial production, preferentially without the use of toxic solvents, is one relevant aspect addressed in this study by melt processing. It should be noted that any increase in the complexity of the aqueous shunt such as a valve technology, at the same time also bears new technological challenges and increases the risk for mechanical failure of the device [32]. Sterility is another very important requirement that demands technologies, which neither affects the device shape and morphology e.g. by heat treatment nor alters the polymer structure such as by chain rupture or crosslinking [33]. γ-Sterilization on dry ice as suggested herein allowing preserved material properties could be further explored compared to ethylene oxide sterilization. Biocompatibility is a fundamental and very complex requirement that certainly cannot be reflected by an in vitro test with one cell type. Still, the absence of any material cytotoxicity on human primary cells (Fig. 7A) represents an important result for continuing the efforts towards a clinical translation. The impairment of fibroblast proliferation at a high dose release of diclofenac sodium (Fig. 7B–C), despite probably being unspecific, indicated in principle that a drug loading could translate to a biofunctionality of the reported devices. Further studies should include the analysis of immunocompatibility and the cell modulation by a microstent payload with antiproliferative drugs. Finally, the successful proof-of-concept implantation of a device prototype in vivo (Fig. 8) on the one hand confirmed the suitability of the surgical technique for these filigree devices and provided first insights into the requirements of device length and its tailoring to the specific needs of the treated eye. On the other hand, the study illustrated that the devices were generally well tolerated in vivo. The observed vascularization, which was outside the optical path and therefore not critical, might be reduced by a further tailoring of the devices' mechanical properties or a pre-bending of the implants to better reflect the curvature of the eye. A functional analysis of intraocular pressure reduction remains an important goal for future comprehensive in vivo studies. 5. Conclusion Single and bilayer microstents have been prepared in a continuous extrusion process with and without drug loading to provide thin diameter, sterilizable, slowly degrading, biocompatible devices for glaucoma filtration surgery. The introduced drug release functionality could support long-term performance in suprachoroidal drainage by impeding disadvantageous capsule formation. Based on their multifunctionality that matches so far unmet key challenges of glaucoma implants and their first successful in vivo study, these microstents are promising candidates for further functional analysis of long-term performance. Acknowledgment The authors are grateful for the technical support by Mrs. Pfeiffer, Dr. Nöchel, Dr. Wagermeier, Mr. Schossig, Dr. Kosmella, Mrs. Schwanz, Dr. Klein, Mr. Jeziorski, and Dr. Boese as well as for the funding by the Bundesministerium für Bildung und Forschung, Germany, grant No. 0312123. References [1] C.C. Peng, M.T. Burke, B.E. Carbia, C. Plummer, A. Chauhan, Extended drug delivery by contact lenses for glaucoma therapy, J. Control. Release 162 (2012) 152–158. [2] S.R. Chennamaneni, C. Mamalis, B. Archer, Z. Oakey, B.K. Ambati, Development of a novel bioerodible dexamethasone implant for uveitis and postoperative cataract inflammation, J. Control. Release 167 (2013) 53–59. [3] V.R. Kearns, R.L. Williams, Drug delivery systems for the eye, Expert Rev. Med. Devices 6 (2009) 277–290. [4] B.T. Gabelt, P.L. Kaufman, Changes in aqueous humor dynamics with age and glaucoma, Prog. Retin. Eye Res. 24 (2005) 612–637.

[5] R.J. Noecker, The management of glaucoma and intraocular hypertension: current approaches and recent advances, Ther. Clin. Risk Manag. 2 (2006) 193–206. [6] T. Realini, Selective laser trabeculoplasty: a review, J. Glaucoma 17 (2008) 497–502. [7] European Glaucoma Society, Terminology and Guidelines for Glaucoma, third ed. Editrice Dogma, Savona, 2008. [8] P.Y. Ramulu, K.J. Corcoran, S.L. Corcoran, A.L. Robin, Utilization of various glaucoma surgeries and procedures in Medicare beneficiaries from 1995 to 2004, Ophthalmology 114 (2007) 2265–2270. [9] S.J. Gedde, D.K. Heuer, R.K. Parrish II, Review of results from the tube versus trabeculectomy study, Curr. Opin. Ophthalmol. 21 (2010) 123–128. [10] R.S. Ayyala, J.L. Duarte, N. Sahiner, Glaucoma drainage devices: state of the art, Expert Rev. Med. Devices 3 (2006) 509–521. [11] C.H. Hong, A. Arosemena, D. Zurakowski, R.S. Ayyala, Glaucoma drainage devices: a systematic literature review and current controversies, Surv. Ophthalmol. 50 (2005) 48–60. [12] D.S. Minckler, B.A. Francis, E.A. Hodapp, H.D. Jampel, S.C. Lin, J.R. Samples, S.D. Smith, K. Singh, Aqueous shunts in glaucoma: a report by the American Academy of Ophthalmology, Ophthalmology 115 (2008) 1089–1098. [13] R.F. Guthoff, W. Schmidt, D. Buss, C. Schultze, U. Ruppin, O. Stachs, K. Sternberg, D. Klee, B. Chichkov, K.P. Schmitz, Development of a glaucoma microstent with drainage into the suprachoroidal space: fluid mechanical model approach, Ophthalmologe 106 (2009) 805–812. [14] S.R. Sarkisian Jr., Tube shunt complications and their prevention, Curr. Opin. Ophthalmol. 20 (2009) 126–130. [15] M. Wilcox, O.A. Kadri, Force and geometry determine structure and function of glaucoma filtration capsules, Ophthalmologica 221 (2007) 238–243. [16] D.S. Minckler, A. Shammas, M. Wilcox, T.E. Ogden, Experimental studies of aqueous filtration using the Molteno implant, Trans. Am. Ophthalmol. Soc. 85 (1987) 368–392. [17] A. Rosentreter, A.C. Mellein, W.W. Konen, T.S. Dietlein, Capsule excision and Ologen implantation for revision after glaucoma drainage device surgery, Graefes Arch. Clin. Exp. Ophthalmol. 248 (2010) 1319–1324. [18] J. Valimaki, P.J. Airaksinen, A. Tuulonen, J. Risteli, Postoperative systemic corticosteroid treatment and Molteno implant surgery: a randomized clinical trial, Acta Ophthalmol. Scand. 77 (1999) 50–56. [19] Y.E. Choonara, V. Pillay, M.P. Danckwerts, T.R. Carmichael, L.C.Du. Toit, A review of implantable intravitreal drug delivery technologies for the treatment of posterior segment eye diseases, J. Pharm. Sci. 99 (2010) 2219–2239. [20] V.P. Costa, A. Azuara-Blanco, P.A. Netland, M.R. Lesk, E.S. Arcieri, Efficacy and safety of adjunctive mitomycin C during Ahmed glaucoma valve implantation: a prospective randomized clinical trial, Ophthalmology 111 (2004) 1071–1076. [21] J.T. Jacob, O.J. Lacour, C.F. Burgoyne, Slow release of the antimetabolite 5-fluorouracil (5-FU) from modified Baerveldt glaucoma drains to prolong drain function, Biomaterials 22 (2001) 3329–3335. [22] J.F. Jordan, B.F. Engels, S. Dinslage, T.S. Dietlein, H.D. Ayertey, S. Roters, P. Esser, W. Konen, G.K. Krieglstein, A novel approach to suprachoroidal drainage for the surgical treatment of intractable glaucoma, J. Glaucoma 15 (2006) 200–205. [23] M. Löbler, K. Sternberg, O. Stachs, R. Allemann, N. Grabow, N. Roock, C.F. Kreiner, D. Streufert, A.T. Neffe, B.D. Hanh, A. Lendlein, K.P. Schmitz, R. Guthoff, Polymers and drugs suitable for the development of a drug delivery drainage system in glaucoma surgery, J. Biomed. Mater. Res. B Appl. Biomater. 97 (2011) 388–395. [24] C. Wischke, M. Löbler, A.T. Neffe, D.B. Hanh, M. Zierke, K. Sternberg, K.P. Schmitz, R. Guthoff, A. Lendlein, A blend of poly(ε-caprolactone) and poly [(ε-caprolactone)-co-glycolide] with remarkable mechanical features and wide applicability as biomaterial, Macromol. Symp. 309 (2011) 59–67. [25] A. Lendlein, P. Neuenschwander, U.W. Suter, Tissue-compatible multiblock copolymers for medical applications, controllable in degradation rate and mechanical properties, Macromol. Chem. Phys. 199 (1998) 2785–2796. [26] S.M. Li, P. Dobrzynski, J. Kasperczyk, M. Bero, C. Braud, M. Vert, Structure–property relationships of copolymers obtained by ring-opening polymerization of glycolide and epsilon–caprolactone. Part 2. Influence of composition and chain microstructure on the hydrolytic degradation, Biomacromolecules 6 (2005) 489–497. [27] K. Pearson, Mathematical contributions to the theory of evolution. XIX. Second supplement to a memoir on skew variation, Philos. Trans. R. Soc. Lond. Ser. A 216 (1916) 429–457. [28] W. Khunawattanakul, S. Puttipipatkhachorn, T. Rades, T. Pongjanyakul, Chitosan– magnesium aluminum silicate nanocomposite films: physicochemical characterization and drug permeability, Int. J. Pharm. 393 (2010) 219–229. [29] L. Netuschil, Vitalfluoreszenzfärbung mit FDA/EB in der Zellkulturtechnik, Z. Zahnärztl. Implantol. 4 (1988) 246–250. [30] F. Lecomte, J. Siepmann, M. Walther, R.J. MacRae, R. Bodmeier, Blends of enteric and GIT-insoluble polymers used for film coating: physicochemical characterization and drug release patterns, J. Control. Release 89 (2003) 457–471. [31] C. Wischke, A.T. Neffe, S. Steuer, A. Lendlein, Evaluation of a degradable shape-memory polymer network as matrix for controlled drug release, J. Control. Release 138 (2009) 243–250. [32] S. Siewert, C. Schultze, W. Schmidt, U. Hinze, B. Chichkov, A. Wree, K. Sternberg, R. Allemann, R. Guthoff, K.P. Schmitz, Development of a micro-mechanical valve in a novel glaucoma implant, Biomed. Microdevices 14 (2012) 907–920. [33] M.B. Sintzel, A. Merkli, C. Tabatabay, R. Gurny, Influence of irradiation sterilization on polymers used as drug carriers - a review, Drug Dev. Ind. Pharm. 23 (1997) 857–878.