Simulation of drug distribution in the vitreous body after local drug application into intact vitreous body and in progress of posterior vitreous detachment.

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Simulation of Drug Distribution in the Vitreous Body After Local Drug Application into Intact Vitreous Body and in Progress of Posterior Vitreous Detachment CHRISTIAN LOCH,1 MALTE BOGDAHN,1 SANDRA STEIN,1 STEFAN NAGEL,1 RUDOLF GUTHOFF,2 WERNER WEITSCHIES,1 ANNE SEIDLITZ1 1 2

Institute of Pharmacy, Center of Drug Absorption and Transport, EMA University of Greifswald, Greifswald 17487, Germany Department of Ophthalmology, University of Rostock, Rostock 18057, Germany

Received 8 July 2013; revised 17 October 2013; accepted 18 November 2013 Published online 5 December 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23808 ABSTRACT: Intravitreal injections and drug-loaded implants are current approaches to treat diseases of the posterior eye. To investigate the release of active agents and their distribution in the vitreous body, a new test system was developed that enables a realistic simulation of eye motions. It is called the eye movement system (EyeMoS). In combination with a previously developed model containing a polyacrylamide gel as a substitute for the vitreous body, this new system enables the characterization of the influence of eye motions on drug distribution within the vitreous body. In the presented work, the distribution of fluorescence-tagged model drugs of different molecular weight within the simulated vitreous was examined under movement with the EyeMoS and without movement. By replacing a part of the gel in the simulated vitreous body with buffer, the influence of the progress of posterior vitreous detachment (PVD) on the distribution of these model substances was also studied. The results indicate that convective forces may be of predominate influence on initial drug distribution. The impact of these forces on drug transport increases with simulated progression of PVD. Using the EyeMoS, the investigation of release and C 2013 Wiley Periodicals, Inc. and the distribution from intravitreal drug delivery systems becomes feasible under biorelevant conditions. American Pharmacists Association J Pharm Sci 103:517–526, 2014 Keywords: vitreous body; intravitreal injection; posterior vitreous detachment; vitreous model; eye movement system; in vitro model; drug transport; hydrogels; distribution; diffusion

INTRODUCTION Many patients with eye diseases (glaucoma, macula edema, etc.) have to apply eye drops several times a day. The compliance of these patients is often poor because in most cases disease progression occurs painlessly and patients tend to forget the administration.1 An alternative to topically administered drugs are modern implants, loaded with active agents, which can be injected periocularly or intraocularly. Injections and implants can be placed, for example, subconjunctivally,2 in the sub-Tenon’s room3 or intrasclerally.4 The vitreous body is also a location for the administration that is attracting increasing attention.5 With Ozurdex (2010) and Iluvien (2013), intravitreal implants have been introduced into the market to treat macula edema.6–8 An important step in the development of intravitreal dosage forms is the in vitro characterization of drug release and distribution. To analyze drug release from intravitreal dosage forms in vitro, incubation methods can be used. Other approaches include the adaptation of the compendial flow-through apparatus as reported by Browne and Kieselmann.9 Such classical release setups in which the dosage form is immersed in a fully stirred compartment do not allow for the examination of evolving distributions. Knowledge regarding local distributions in the vitreous is however very important for these dosage forms R

R

Correspondence to: Anne Seidlitz (Telephone: +49-3834-864898; Fax: +493834-864886; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 103, 517–526 (2014) C 2013 Wiley Periodicals, Inc. and the American Pharmacists Association

as the site of action is located at the retina. The transport of substances in the gel-like vitreous body can occur through diffusion, sedimentation, or convection. Especially for substances with a low molecular weight diffusion has been discussed as the prevailing transport mechanism.10 For this reason, the examination of diffusion-driven distributions in a simulated vitreous compartment is desirable. For such studies, a suitable vitreous substitute is needed to mimic the gel-like vitreous body. The use of silicone-based oils or hydrogels for this purpose has been investigated.5,11–14 In situ cross-linking hydrogels, for example, consisting of polyacrylamide (PAA) or poly(ethylene) glycol, have been reported to be the most suitable to simulate the vitreous.12–14 The shape and volume of the human vitreous body was often simulated using a glass corpus.15 Furthermore, a potential impact of eye movement has been considered. Repetto et al.15 developed a model to simulate saccadic eye movements and investigate the effect on the distribution in glycerol as vitreous substitute. They showed that it is difficult to mimic the transport dynamics of the vitreous body because of the natural complexity. Nevertheless, they observed intensive effects on distribution processes in the vitreous cavity and shear stresses along the surrounding membranes caused by the simulated saccadic eye movement. It was the aim of this work to develop an in vitro model for the evaluation of drug release and distribution upon intravitreal administration combining some of the approaches listed above under further approximation of selected test conditions to the situation in vivo. Particularly, the simulation of different types of natural eye movements including pursuit movements and saccadic motions and the impact of

Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

517

518


posterior vitreous detachment (PVD) were to be evaluated. To the authors’ knowledge, these factors have not been included in in vitro studies before. PVD is a condition caused by the liquefaction of the vitreous beginning in childhood that progresses with age and may eventually cause the separation of the vitreous membrane from the retina. Because of this disease, the ratio of gel-like vitreous body and liquid vitreous portion changes intensively.16 The consequence should be an increase in the influence of convective transport in the vitreous cavity. It must be expected that the situation for drug release and distribution is also changing because of this phenomenon. This important aspect is often neglected and is most likely not accounted for in animal models either, as typically very young animals are used. The developed test system combines a previously reported gel-filled glass body called the vitreous model (VM)13 with a newly developed eye movement system (EyeMoS). With the combination of EyeMoS and VM, the shape, the volume, and four main movement types of the eye can be simulated. The concept allows for the investigation of the effects of different eye movement types on the distribution of an injected model substance in a modified PAA gel as vitreous substitute. Furthermore, the different compositions of a completely gellike and a partially liquefied vitreous body can be simulated by replacing certain fractions of the gel with ringer buffer. In the work presented here, this new setup is described in detail; the movement schemes achieved with the EyeMoS are characterized and the first experiments regarding the distribution of fluorescent model substances are described. In the therapy of ocular diseases, a wide range of drugs with various molecular masses are locally administered into the vitreous body. On the one hand, there are active agents such as dexamethasone (DX) with a comparably low molecular mass, and on the other hand, large molecules such as antibodies are available.17 Examples for large molecules are the IgG-antibody bevacizumab (150 kDa) as well as ranibizumab, a monoclonal antibody fragment with a molecular mass of 40 kDa, which are injected intravitreally to treat age-related macular degeneration.18 To account for this diversity, model substances, with different molecular weights (350–150,000 Da), antibodies and the therapeutically used drug substance DX were examined to obtain information on the influences of the molecular weight on distribution in the vitreous body.

MATERIALS AND METHODS Materials Dexamethasone, fluorescein sodium (FS), fluorescein isothiocyanate–dextran with molecular masses of 4 kDa (FITC 4), 40 kDa (FITC 40), and 150 kDa (FITC 150), and methanol were obtained from Sigma Aldrich Chemie GmbH (Steinheim, Germany). Rotiphorese , ammoniumperoxo disulfate (APS) and tetramethylethylenediamine (TEMED) were purchased from AppliChem GmbH (Darmstadt, Germany). The goat anti-human IgG (H+L)–FITC antibody (IgG antibody) was supplied by Dianova GmbH (Hamburg, Germany). All chemicals were of analytical grade. R

Methods

EyeMoS and VM The EyeMoS was created to mimic the natural eye movement. The system was designed to accommodate the previously introduced VM.13 The combination of these two systems allows the examination of drug distribution in the simulated vitreous body under simulation of eye movement. The VM consists of a spherical glass corpus with a vent (Fig. 1). The corpus can be disassembled along its equator to remove the content for analysis. The shape and volume (4 cm3 were adapted to the human vitreous body. For distribution experiments, the VM was mounted in the holder of the EyeMoS in a way that the longitudinal axis of the VM was oriented along the z-axis (Fig. 2b). To simulate drug distribution in the vitreous, the VM is filled with a modified PAA gel13 . Ten milliliter of the PAA gel consists of 9.23 mL standard ringer buffer, 0.67 mL of a 30% solution of acryl amide and bisacryl amide (Rotiphorese ), 0.1 mL TEMED, and 0.01 mL a 1% APS solution. These components are carefully mixed initiating the cross-linking of the acryl amide monomers, poured into the glass corpus of the VM and allowed to solidify for 15 min. To simulate natural movements of the eye, the VM can be embedded in a holder at the center of the EyeMoS (Fig. 2). At each motion shaft, servo motors (MKS DS 6125e and Savox SC 1257Tg) are assembled to move the holder periodically along the x- and y-axes (Fig. 1). The servo motors are processor controlled (Arduino Uno 65139). The open source software Arduino (www.arduino.cc) enables the definition of freely programmable R

Figure 1. The VM: (a) schematic view, (b) image with scheme of the injection position and separation into two parts (a + b) of the vitreous substitute for quantitative analysis. Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

DOI 10.1002/jps.23808


519

Figure 2. (a) The EyeMoS with (I) VM in the holder, (II) servo motor for the vertical movement, (III) servo motor for the horizontal movement, and (IV) control panel with display. (b) Movement axes of the EyeMoS.

movement schemes. In adaption to natural eye movements, slow to faster pursuit movements as well as saccadic and microsaccadic motions can be defined. The amplitudes and velocities of the various programmed movement schemes were measured using a potentiometer that was mounted centrally on top of the holder. The potentiometer was connected in a voltage divider circuit. A second processor was installed and connected with the potentiometer to record and digitize the position signals. The actual positions were depicted in real time with the software LogView (www.logview.info) and the amplitudes and velocities were calculated with the obtained time-position data.

Partition Coefficient of FS for the Vitreous Substitute To determine the partition coefficient (PC) of FS between the PAA gel13 used as vitreous substitute and Ringer buffer, 10.0 mL of the gel, loaded with FS (550 :g/L), were transferred into a beaker. After completed gelatinization, the gel was topped with 10.0 mL Ringer buffer and shaken on a plate shaker. Samples of 250 :L of the buffer phase were drawn periodically, replaced by the same volume of fresh Ringer buffer and quantified fluorimetrically as described below. When the steady state was achieved, the PC was calculated according to the following equation: PC =

V2 c0 − ct V1 ct

in which V1 and V2 denote the volumes of the model substanceloaded gel and the buffer phase, c0 the initial concentration in the gel, and ct the concentration of the model substance in the buffer phase at sampling times. The experiments were performed in triplicate.

Distribution Studies Using the VM and the EyeMoS The vitreous substitute (see above) was filled into the VM and allowed to solidify. Afterwards, solutions of the model drugs in ringer buffer were injected centrally into the VM using a 27 gauge needle. In Figure 1b, the site of injection is illustrated. For all experimental injections, the syringe was placed on a spacer, which was set on the top of the vent to ensure the reproducibility of needle penetration depth. Model drugs were 15 :L of a 1.25 mM solution of FS, FITC 40, or FITC 150. Additionally, 15 :L of a solution of the FITC-labeled IgG antiDOI 10.1002/jps.23808

body (1.3 mg/mL) was injected. For DX, an injection volume of 100 :L and a concentration of 30 mM were used. The higher injection volume was chosen for analytical reasons. After injection, the model was closed and mounted in the EyeMoS. A periodic movement (rotational oscillation along the x-axis with amplitude 17.5◦ , along the y-axis with amplitude 10◦ , maximum velocity for both rotations 60◦ /s, no phase shift, see Fig. 3a) was performed for 3 h simulating slow pursuit movements. Additional experiments were completed with FS and FITC 150 over a period of 24 h. All experiments were performed in triplicate. After the selected time span, the VM was removed from the EyeMoS. To gain insight into the initial distribution, experiments without incubation were also performed in which the gels were frozen immediately after injection. All gel samples were prepared for analysis as described below.

Simulation of PVD To simulate the progress of the PVD, 25 % (moderate PVD) and 50 % (advanced PVD) of the vitreous substitute were replaced by Ringer buffer solution. This was achieved by filling only 50% or 75%, respectively, of the total volume of the corpus with the gel-forming solution. After completed gelatinization, the complementary quantity of buffer was added. Initially, the buffer phase is located adjacent to the gel in the part A toward to the vent of the prepared VM (Fig. 2). The various filling levels are illustrated in Figure 4. The solutions of FS, FITC 40, and FITC 150 were always injected centrally into the VM. The experimental conditions remained the same as described above. In addition, experiments without movement during the intended incubation period were performed.

Quantitative Analysis The PAA gel samples as well as the samples containing gel and buffer, obtained in the simulation of the PVD, were prepared for quantification as previously reported.13 Briefly, after the termination of the experiment, the whole corpus of the VM was frozen. In the frozen state, the “globe” was divided along its equator in two parts (A and B). As per the definition, part A is the side with the vent from which the injection occurred (Fig. 1). The parts were separated, defrosted, and separately incubated in acetone for model substance extraction. After sufficient extraction time, the acetone phase was segregated from the gel phase. After evaporation of the acetone, the eluted model Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

520


Figure 3. Recorded position (blue) and speed (red) of the simulated slow pursuit (a) and microsaccadic eye movement (b).

drug was taken up with Ringer buffer. Afterwards, the content of fluorescence-tagged substances (FS, FITC 4–150, and IgG– FITC antibody) of two samples of 250 :L of the reconstituted extract were analyzed using a fluorescence reader (Varioskan Flash Multimode Reader; Thermo Scientific/Fisher,Waltham, Massachuetts, USA) at an excitation wavelength of 485 nm and an emission wavelength of 538 nm. The concentration range for calibration for the fluorescent model drugs amounted from 0.052 to 1.67 mg/mL. Using a linear calibration function, correlation coefficients ranging from 0.9958 to 0.9991 were obtained. For the limit of detection and quantification, values of 0.011 and 0.037 mg/L were determined. All experiments with fluorescent substances were performed under protection from light. Samples of DX were quantified by HPLC (Shimadzu CBM 20 A, degasser, binary pump, column thermostat, UV diode array detector) using an analytical column Phenomenex Luna 3u C18 (2) 100 A 150 × 4.6 mm2 (Aschaffenburg, Germany) at 30◦ C. The injection volume was set to 10 :L and the detection wavelength was set to 241 nm. Injection was performed two times per reconstituted extract. The retention time was 5.1 min. A mixture of 75% methanol and 25% water was used as mobile phase and pumped isocratically at a flow rate of 0.7 mL/min. The concentration range for calibration for DX amounted from 9.25 to 296 :g/mL with correlation coefficients ranging from 0.9955 to 0.9991 for a linear calibration function. For the limit of detection and quantification, values of 2.3 and 7.7 :g/mL were obtained. t-Tests were used for statistical comparisons where necessary, and the significance level was set to 0.05.

RESULTS Eye Movement System A range of various eye movements from slow to fast pursuit motions with angular velocities of 60◦ /s–124◦ /s along the x-axis and 60◦ /s along the y-axis, simulated saccadic movements (261◦ /s along the horizontal plain), and mimicked micro-saccadic movements along the x-axis with a velocity of 434◦ /s were simulated and characterized. The results of the performance evaluation of the EyeMoS using the potentiometer method are shown in Table 1. Figure 3 exemplarily illustrates graphs recorded for movement schemes whose amplitudes and velocities simulate pursuit eye movement (Fig. 3a) and microsaccadic movement (Fig. 3b) in the x-axis. Movement frequencies were not adapted to the situation in vivo.

R


PC of FS for the Vitreous Substitute After 24 h, a steady state was obtained in the determination of the PC. The PC for FS between the PAA gel and the Ringer buffer amounts to 1.5 ± 0.2. Comparison of the Distribution of Different Model Drugs For comparison, the distribution of FS, FITC 4, FITC 40, and FITC 150 as well as the model FITC-labeled antibody and the drug DX in the vitreous substitute after 3 h in the EyeMoS under simulated slow pursuit movements was studied. The results are depicted in Table 2. Significant differences in the distribution behavior were not observed after 3 h (p < 0.05). DOI 10.1002/jps.23808

521


Table 2. Distribution of Different (Model) Drugs in Part A of the Vitreous Substitute (Native State, 100% PAA Gel) After 3 and 24 h Simulated Slow Pursuit Eye Movement with EyeMoS Model Substance

FS FITC 150 kDa FITC 40 kDa FITC 4 kDa IgG antibody DX

Cumulative Amount in Part A After 3 h (%)

Cumulative Amount in Part A After 24 h (%)

96.2 ± 2.5 96.9 ± 1.3 98.5 ± 0.4 88.1 ± 4.7 86.5 ± 0.9 97.9 ± 1.3

75.4 ± 4.7 92.3 ± 3.8

FS, fluorescein sodium; DX, dexamethasone; isothiocyanate–dextran with respective molecular weight.

FITC,

fluorescein

representing the possible range of molecular masses of potential drug candidates.

Distribution in the Native State

Figure 4. Filling level of the VM to simulate different stadia of PVD [(a) native state without PVD, 0% buffer; (b) moderate PVD, 25% buffer (green); (c) advanced PVD, 50% buffer (green); buffer was stained with FS).

With amounts of 75.4 ± 4.7% for the FS and 92.3 ± 3.8% for the FITC 150 detected in the part A, different redistribution between both substances were obtained after the prolonged experimental time of 24 h (Table 2). FS, FITC 40, and FITC 150 were chosen as model substances for further experiments Table 1.

The results of the study of model substance distribution in the native state (100% PAA gel), representing a native vitreous body, are shown in Table 3. In addition, images of frozen and sectioned gel samples are depicted in Figure 5. As already mentioned above, no significant differences were observed between the tested substances despite their different molecular masses after 3 h (p < 0.05). Directly after central injection into the gel, model substance fractions of 42%–66% were detected in part A indicating that with slight deviations central injection was successfully performed. This corresponds to the observations shown in Figure 5I. Irrespective of whether the VM was moved with the EyeMoS or not, the main quantity of the model substances (96%–98%) was detected in part A after 3 h. This preferential distribution into part A may be explained with the observation that a certain fraction of the injected solutions flowed back along the puncture duct that runs through compartment A. This also becomes evident from the images of frozen and sectioned gel samples. The photograph of the frozen gel after 3 h without movement (Fig. 5II) shows a distinct coloration in the center of the gel and along a straight line across part A toward the vent. Near the vent, two more spots are detectable, possibly caused by injected substance that flowed back out of the puncture duct and spread at the interface between the glass body of the VM and the gel. In case of the moved sample (Fig. 5III), a similar phenomenon may have occurred; however, even though mainly located in part A, the model substance can be visualized in greater parts of the gel sample. This may possibly result from the movement in the EyeMoS. The observed distribution pattern is, however, more irregular.

Movement Schemes of the Distribution Experiments Realized with the EyeMoS

Scheme Slow pursuit movement Fast pursuit movement Saccadic movement Microsaccadic movement

Movement along the x-Axis Amplitude (◦ ) Velocity (◦ /s) 17 16 21 8

60 124 261 434

Movement along the y-Axis Amplitude (◦ ) Velocity (◦ /s) 10 10 10 0

60 60 60 0

Experiments were performed in triplicates for each mode. DOI 10.1002/jps.23808


522


Table 3. Influence of Simulated Slow Pursuit Eye Movement on the Distribution of FS, FITC 40, and FITC 150 in Part A of the Vitreous Substitute Under Consideration of Different Stadia of Posterior Vitreous, Using EyeMoS Stadia of PVD Native state (100% PAA gel)

Moderate PVD (75% PAA gel/25% buffer)

Advanced PVD (50% PAA gel/50% buffer)

Model Substance

t=0h

FS FITC 40 FITC 150 FS FITC 40 FITC 150 FS FITC 40 FITC 150

64.4 65.5 ± 0.9 42.2 72.2 ± 5.7 73.5 60.2 98.8 92.1 96.1

Three Hour Without Movement 99.4 99.4 98.8 82.6 98.1 99.0 58.5 74.2 65.6

± ± ± ± ± ± ± ± ±

0.3 0.4 0.4 10.8 1.3 0.2 3.4 13.9 8.6

Three Hour Movement 95.8 98.5 96.9 77.0 89.3 85.9 65.6 32.1 42.8

± ± ± ± ± ± ± ± ±

2.8 0.4 1.3 8.6 11.8 6.5 3.3 3.5 5.3

Data presented as mean values ± standard deviation, n = 3–5. FS, fluorescein sodium; FITC, fluorescein isothiocyanate–dextran, number indicating molecular weight in kDa.

Distribution Under Simulated PVD The results of the distribution studies under different PVD stadia simulated by replacing 25% (moderate PVD) and 50% (advanced PVD) of the PAA gel by Ringer buffer in the VM are shown in Table 3, respectively. Directly after injection, fractions of 60%–72% were detected in part A in the moderate PVD setup and 92%–99% when simulating advanced PVD. This increase in the amount initially detected in part A is most likely caused by the faster/easier back flow along the puncture duct or in case of the advanced PVD where injection theoretically occurs immediately at the border between gel and buffer, direct (back)flow into the buffer. Under simulation of the moderate PVD (25% of gel replaced by buffer), a location of the main quantity of the model substance in part A was detected similar to the model of the native state (Table 3, see above), even though the effect was less pronounced. After 3 h, 77%–99% of the model substances was detected in part A, irrespective of whether the VM was moved with the EyeMoS or not. When 50% of the gel was replaced by buffer (simulated advanced PVD), similar fractions of the model substances were found in part A and B under movement, with 32%–66% of the model substance located in part A. Without movement, a somewhat higher fraction (58%–74%) was detected in part A. These distributions are, however, far from homogenous distributions

in the respective parts. During the experiments, a rearrangement of the gel and buffer phase during the 3-h incubation in the EyeMoS was observed leading to a situation in which a spherical gel core is surrounded by the model substance containing buffer phase. A cross-section of a frozen “globe” (composed of 50% gel and 50% buffer) after 3 h of movement with the EyeMoS is depicted in Figure 6. As visible from the picture, the colored model substance is mainly located in the outer buffer phase oriented toward the glass walls of the VM. In the setup without movement, a rearrangement of the two phases was also observed. However, in this case, the gel was not located centrally in the body of the VM but slide along the glass walls of the VM reaching a position in which a smaller fraction of the gel was located in part A, whereas the main portion remained in part B in which it was completely located at the start of the experiment. These different rearrangements explain the observed model substance fractions in the respective parts of the vitreous substitute. Rearrangements in the location of the gel and buffer phase were not detected when simulating moderate PVD (25% buffer).

DISCUSSION Local drug delivery to the vitreous body is an important route for the administration of drug substances to certain parts of

Figure 5. Cross-section of the frozen vitreous substitute (native state, 100% PAA gel, FS): (I) shortly after injection, (II) after 3 h without movement, and (III) after 3 h with movement. Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

DOI 10.1002/jps.23808


Figure 6. Cros of the frozen vitreous substitute after 3 h movement under simulation of advanced PVD with 50% replacement of gel by buffer, injected model substance FS.

the eye, such as the retina. In vivo performance of locally administered injections is often complicated to analyze as the site of action is not available for concentration determination, and blood plasma levels, if measurable, are not representative for the local concentration. Local in vivo concentrations can, in most cases, only be determined in studies with animal models. For the eye, the rabbit model is often used.19 Such studies are time-consuming and expensive. Furthermore, differences regarding the anatomy and the transport properties have to be taken into account.20,21 Therefore, in early development phases of new dosage forms such as intravitreal implants, simple, fast, and cheap in vitro methods are desirable. Such test systems can be used to characterize the newly developed drug delivery systems and potentially enable the choice of an adequate prototype and animal model in preclinical studies. The EyeMoS and the VM were developed to examine drug distribution in the vitreous body in an in vitro model mimicking some of the hydromechanical properties of the situation in vivo. Transport of intravitreally injected substances in the vitreous body is essential to reach the site of action, which is, for most substances, located near the retina. As selective drug delivery to this site is not possible with the current dosage forms, homogenous distribution through the vitreous body is aspired. Transport in the vitreous may be affected through sedimentation, convection, or diffusion. All three types of movement can occur separately or can interfere with each other. Eye movement is a factor that can potentially influence drug transport, especially convection. In the past, different experimental and analytical in vitro models were developed to investigate the distribution of drugs and liquids in the vitreous body. Applying these methods, some aspects of the natural eye movement could already be examined.15,22 With the EyeMoS, it is now possible to perform various movement schemes with the aim to mimic the natural human eye movement. Angular velocities of 60◦ /s– 124◦ /s along the x-axis and 60◦ /s along the y-axis, slow and faster pursuit movements of the natural eye,23 respectively, can be achieved. These simulated movements are characterized by DOI 10.1002/jps.23808

523

amplitudes of about 17◦ along the x-axis and 10◦ on the y-axis. Also, saccades, which are denoted as movements with small amplitudes and high velocities,24 can be performed. To simulate the fastest natural eye movement, the microsaccadics, movements along the x-axis with short amplitude of 8◦ and a high angular velocity of 434◦ /s were tested. The obtained results confirm that the EyeMoS can simulate the natural eye movements ranging from pursuit to saccadic movements in their characteristic velocities and amplitudes. For the distribution studies presented here, we used a relatively slow periodic movement with a constant frequency. However, also more complex and nonperiodic movement schemes can be applied. The PAA gel chosen as the model vitreous substitute is easy and cheap to prepare and closely resembles the human vitreous body in some selected physical and chemical properties.13 Most of all, the shape of the gel body, the water content (99%), and the pH (7.4) of the PAA gel are equal to that of natural vitreous body. Furthermore, the parameters viscosity and density, which are of central importance for advective transport processes, are in good accordance with the properties of natural porcine vitreous body.13 The transport properties of the PAA gel and equilibrium distributions for the model substances were, however, not compared with those of natural vitreous humor. It might be most interesting to study the influence of different gel compositions regarding the transport and distribution in future experiments. Furthermore, convective transport that has been reported to occur in the vitreous in vivo because of the temperature and/or pressure gradients25–27 between the anterior and posterior part of the eye is not reproduced in the in vitro setup presented here but must be expected to significantly contribute to the distribution of large molecules in vivo. However, in spite of these limitations, the developed test system has some distinct advantages in early development phases. The drug release from prototypes of implants can be tested under more biorelevant conditions compared with standard test protocols, such as the immersion of the dosage form in a medium-filled stirred beaker. In addition to some biorelevant features, in vitro test systems possess a high degree of standardization and reproducibility. By the combination of the movement properties of the EyeMoS and the suitable vitreous substitute, the examination of the distribution of drugs in the vitreous body in a biorelevant in vitro setup becomes feasible. Furthermore, different situations that may arise during disease progression may be reproducibly simulated. In the presented experiments, the influence of movement on drug distribution in the vitreous body was investigated using the EyeMoS under simulation of the native state of the vitreous body and under simulated moderate and advanced PVD. Prior to conducting the experiments, the PC of FS, used as a model substance for small molecule drugs, between the PAA gel and the buffer was determined. A similar distribution of FS between the gel and the buffer was observed at steady state. Accordingly, substance-specific distribution phenomena, such as strong ionic interactions, which might hinder diffusion into the gel in the distribution experiments, can be excluded for FS and is also not to be expected for the FITCs. Furthermore, prior to the main experiments, the distribution of different model substances spanning a wide range of molecular masses and a drug candidate in the simulated vitreous body was compared under identical conditions (100% gel, 3 h of movement in the EyeMoS). As no significant differences were observed in this study, three model substances with different molecular masses Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

524


(FS 376 Da, FITC 40–dextran, and FITC 150–dextran) were chosen for further experiments. To get a first idea about the distribution process beyond the 3 h, the experimental time was prolonged to 24 h in an initial comparison between FS and FITC 150. After the extended time span, a different distribution could be obtained between the low-molecular FS and high-molecular FITC 150 in the parts A+B of the VM. As expected, the low molecular weight substance was detected in part B at a higher extent compared with the high molecular weight FITC 150 because of a higher diffusion velocity. As both compartments gel and buffer are highly hydrophilic (water content ≥ 99%), differences in the distribution behavior of the different substances because of their hydrophilicity are not to be expected. However, specific interactions cannot be ruled out for other nontested drugs. The results of the in vitro distribution studies presented here allow for some conclusions regarding drug injection into the vitreous body in vivo. Diffusion through the intact PAA gel occurred very slowly. In fact, the time span of the experiments of 3 h was too short to allow for the observation of remarkable redistributions because of diffusion. The results of the experiment over 24 h reinforce the thesis that diffusion in the gel needs more time than the 3 h. Concluding, the experimental time and timely resolution will be increased in future experiments. The site of injection and convective transport (even if only caused by retraction of the needle) seem to be the determining parameters/driving forces for the initial distributions in the simulated native state of the vitreous body. Distribution of injected substances along the puncture duct has also been observed in animals.26 Also, the needle size and the resulting diameter of the puncture duct may be relevant. These factors may also influence distribution in vivo. Outflow of fluid from a hole in the sclera produced by needle (30 gauge) penetration has also been reported.28 In a mathematical modeling setup, Friedrich et al.29 have shown the influence of the injection site on the obtained distributions. In their study, higher concentrations were achieved in the posterior regions of the vitreous after injections placed near the retina compared with injections administered centrally or in regions close to the lens. In this context, it must also be kept in mind that distribution toward the site of action is not only a matter of time, as administered drug will be cleared from the vitreous body over time in vivo. Two major pathways of clearance have been reported for the vitreous. Drug can either be cleared via the anterior route through the trabecular meshwork or be cleared posteriorly involving active transport through the retina.30 This “loss” of drug is not accounted for in the in vitro model used for the distribution studies presented here. If similar mechanisms of transport prevail in vivo including the back flow of injected substance into the puncture duct, drug transport via other routes than through the intact vitreous may be a major factor that must not be neglected. It is, however, unclear, whether such a route of transport will only clear the drug from the site or might even contribute to the drug transport to the site of action via other routes, as for example, passive or active elimination of the active agent over the blood vessels at the posterior interface between the vitreous body, the choroidea, and the retina following, the transport to the target organ via the blood are conceivable. Another purpose of the study presented here was to simulate PVD in an in vitro model by replacing defined fractions of the PAA gel with Ringer buffer. Age-related PVD is a disLoch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

ease resulting from progressive vitreous liquefaction in combination with progressive vitreoretinal adhesion.31 According to Johnson,31 liquid vitreous is present in the human eye from the age of 4 years on making up 20% of the vitreous by the late teenage years, presumably in the form of fluid-filled pockets. In the population older than 70 years, the vitreous body of most individuals is expected to be made up of at least 50% liquefied material. In patients younger than 60, the vitreous is expected to generally remain attached to the retina even in cases with extensive liquefaction. It is unclear whether such liquidfilled pockets might be connected and interchange fluid or even change their localization within the vitreous. It is also unclear whether the detachment once occurring will be limited to the posterior part of the eye. According to Wilson et al.,26 in case of a complete progressed PVD with approximately 50% portion of vitreous humor in vivo, the texture of the remaining vitreous body is feeble and the hold to the membrane of the outer orbital layers will be lost with time. This might even result in an arrangement similar to that observed in the VM under movement with the EyeMoS with a gelled core surrounded by a liquefied outer layer. Even though the exact portion of liquefied vitreous and the localization of this material in the vitreous body will inevitably be subject to great interindividual variation, the general presence of liquefied vitreous may have a major impact on drug distribution following administration into the vitreous. The results of the experiments presented here indicate that if drug reaches such a pocket, it will spread fast throughout the fluid because of the convective forces and may, from there on, slowly diffuse into the gelled portion or may also be cleared from the site. Besides the influence of the transport forces, specific affinities of drugs to portions of the gelled or liquefied vitreous, respectively, such as protein, may not be excluded and will influence the resulting distributions. This reinforces the need to further characterize and understand the in vitro and in vivo models used for the preclinical evaluations of dosage form performance, for example, regarding the state of the eyes used in animal studies that are typically conducted using young animals and how this relates to different characteristic patient groups. Hesse et al.32 and Tan et al.33 have shown that a PVD can be induced in an animal model. Tan et al.33 investigated the influence of the distribution of fluorescence-tagged substances in the vitreous body of rabbits. In the first hours of their experiments, they observed a slow transport of the injected model substances in the untreated vitreous body. Over time (5 h up to several days), the injected drug diffused in deeper regions of the vitreous body. This indicates that the transport in a gel-like vitreous body is slow and will require a longer period of time to reach equilibrium distribution. They also report a fast redistribution and elimination of intravitreally injected drugs with progressing PVD, presumably caused by an increase in convection. These results are, in general, in accordance with the results from the in vitro study presented here. However, exact translatability of the results obtained in the in vitro model to the different stadia of PVD in vivo cannot be expected. The EyeMoS in combination with the VM represents a novel approach to in vitro distribution testing of intravitreally administered dosage forms taking certain aspects of the physiological movement of the eye and the hydromechanical properties of the vitreous body that are responsible for the dominant transport forces into account. Nevertheless, the experimental setup as used here also poses some major drawbacks. First, it is unclear to what extent the arrangement of the liquefied and gelled DOI 10.1002/jps.23808


portion of the simulated vitreous corresponds to the situation in vivo in the simulated PVD stadia. Second, when assuming an in vivo positioning of the liquefied portion exclusively at the posterior part, the injection in the model occurs from the wrong side, meaning that the needle is driven through the liquefied portion in the in vitro setup. This may at least in the simulation of the moderate PVD (25% of gel replaced by buffer) lead to a different distribution behavior. In the case of the advanced PVD (50 % gel replaced by buffer) in which the injection occurs directly at the interface between gel and buffer, the side from which the injection occurs should not have a major influence as the injection is most likely subject to immediate convective transport. Third, in the currently performed experimental procedure, the timely and spatial resolution is not high enough. To adequately monitor the distribution behavior, more (including much longer) time points and a more sophisticated sectioning of the gel compartment is required. In future experiments, these shortcomings shall be addressed, where possible. New preparation techniques to more adequately section the gel compartment will be developed, allowing for a better analysis of the distribution than the division in two parts. Also, longer incubation periods as well as the analysis of gels at different time spans of incubation will be performed. A change of the setup to allow for injection via the simulated anterior side mimicking in vivo injection will be implemented while maintaining the already practiced injection using a space holder to allow for reproducible central injection. Hopefully, the refinement of imaging techniques such as magnetic resonance imaging and optical coherence tomography will allow for a better description of the state of the vitreous with respect to special patient groups that might than be translated to the in vitro model. Furthermore, in this first set of experiments, the possibilities provided by the EyeMoS were not fully exploited. The experiments were performed with fairly slow movements. The influence of the velocity and accelerations of the faster schemes (saccadics and microsaccadics) on the distribution remain to be examined. The influence of the frequency of saccadic eye movement on the dynamic flow in the vitreous cavity was reported to be large, while the amplitude should play a minor role.15,22 Through higher velocities and accelerations caused by the simulated saccadic and microsaccadic movements, major forces are expected to act on the model and its contents possibly resulting in differences in the distribution behavior when simulating PVD. Because of the less pronounced impact of convection within model filled with 100% PAA gel, it remains to be investigated whether the type of movement will influence distribution in these cases.

CONCLUSIONS With the EyeMoS and the VM, a new in vitro system was developed for the biorelevant investigation of the distribution of intravitreally injected solutions and drug released from implants. The VM contains a hydrogel that is intended to mimic selected important hydromechanical properties of the vitreous, which will likely impact drug transport. The EyeMoS enables defined movements schemes that are characteristic for certain types of eye movement in vivo. First results using this test system show that the substitution of parts of the gelled vitreous by buffer to simulate vitreous liquefication that also occurs in vivo greatly influences the observed distribution because of an DOI 10.1002/jps.23808

525

increase in convective transport. The experiments also show that much longer time spans of experiments are necessary to observe diffusive transport. Therefore, parameters associated with the injection procedure, such as the site of injection and the withdrawal of the needle may be of greatest importance for the initial distribution. The EyeMoS in combination with the VM are a novel tool for the reproducible analysis of such processes. Further development of the model, for example, regarding the location of the liquefied portion in the model of PVD, is desirable to obtain greater similarity with the situation in vivo and to be able to reliably predict the in vivo behavior of intravitreally applied medications.

ACKNOWLEDGMENTS This project was funded by the Federal Ministry of Education and Research (BMBF) within REMEDIS “H¨ohere Leben¨ durch neuartige Mikroimplantate” (FKZ: 03IS2081). squalitat The authors thank the staff members of the workshop of the Faculty of Mathematics and Natural Sciences for their excellent technical assistance in the realization of the EyeMoS and Thomas Brand for his support regarding HPLC analysis.

REFERENCES 1. Kholdebarin R, Campbell RJ, Jin Y-P, Buys YM. 2008. Multicenter study of compliance and drop administration in glaucoma. Can J Ophthalmol 43(4):454–461. 2. Hosoya K-i, Lee VHL, Kim K-J. 2005. Roles of the conjunctiva in ocular drug delivery: A review of conjunctival transport mechanisms and their regulation. Eur J Pharm Biopharm 60(2):227–240. 3. Silva AA, Wanczinski B, Oliveira AG, Cardillo JA, Paganelli F, Costa RA, Skaf M, Souza-Filho AA, Belfort R, Kuppermann BD. 2005. In vitro and in vivo release profile of a new ciprofloxacin hydrochloride biodegradable controlled release microspheres system optimized for sub-Tenon’s capsule delivery. Invest Ophth Vis Sci 46, E-Abstract 5395. 4. Thrimawithana TR, Young S, Bunt CR, Green C, Alany RG. 2011. Drug delivery to the posterior segment of the eye. Drug Discov Today 16(5–6):270–277. 5. Baino F. 2011. Towards an ideal biomaterial for vitreous replacement: Historical overview and future trends. Acta Biomaterialia 7(3):921–935. 6. Winterhalter S, Ruokonen P, van der Velden KH, Pleyer U, Joussen AM. 2011. Intravitreale implantate. Der Ophthalmologe 108(3):222– 229. 7. EMA. Accessed, at: www.ema.europa.eu/docs/de DE/document library/EPAR - Product Information/human/001140/WC500095499. pdf, last accessed 11/01/2013. 8. Alimera-Sciences. Accessed, at: http://investor.alimerasciences.com/ releasedetail.cfm?ReleaseID = 734042, last accessed 11/01/2013. 9. Browne DC, Kieselmann S. 2010. Low-level drug release-rate testing of ocular implants using USP apparatus 4 dissolution and HPLC end analysis. Dis Technol 17(1). S. 12–14. 10. Park J, Bungay PM, Lutz RJ, Augsburger JJ, Millard RW, Sinha Roy A, Banerjee RK. 2005. Evaluation of coupled convective–diffusive transport of drugs administered by intravitreal injection and controlled release implant. J Control Release 105(3):279–295. 11. Kleinberg TT, Tzekov RT, Stein L, Ravi N, Kaushal S. 2011. Vitreous substitutes: A comprehensive review. Surv Ophthalmol 56(4):300– 323. 12. Swindle KE, Hamilton PD, Ravi N. 2008. In situ formation of hydrogels as vitreous substitutes: Viscoelastic comparison to porcine vitreous. J Biomed Mater Res A 87A(3):656–665. Loch et al., JOURNAL OF PHARMACEUTICAL SCIENCES 103:517–526, 2014

526


13. Loch C, Nagel S, Guthoff R, Seidlitz A, Weitschies W. 2012. The vitreous model—A new in vitro test method simulating the vitreous body. Biomed Tech (Berl) 6(57) Suppl 1. 14. Tao Y, Tong X, Zhang Y, Lai J, Huang Y, Jiang Y-R, Guo B-H. 2013. Evaluation of an in situ chemically crosslinked hydrogel as a long-term vitreous substitute material. Acta Biomaterialia 9(2):5022–5030. 15. Repetto R, Stocchino A, Cafferata C. 2005. Experimental investigation of vitreous humour motion within a human eye model. Phys Med Biol 50(19):4729. 16. Johnson M, Tamm ER. 2010. Biomechanics of aqueous humor outflow resistance. In Encyclopedia of the eye; Darlene AD, Ed. Oxford, UK: Academic Press, pp 173–182. 17. Rodrigues EB, Farah ME, Maia M, Penha FM, Regatieri C, Melo GB, Pinheiro MM, Zanetti CR. 2009. Therapeutic monoclonal antibodies in ophthalmology. Prog Retin Eye Res 28(2):117–144. 18. Bandukwala T, Muni RH, Schwartz C, Eng KT, Kertes PJ. 2010. Effectiveness of intravitreal ranibizumab for the treatment of neovascular age-related macular degeneration in a Canadian retina practice: A retrospective review. Can J Ophthalmol 45(6):590–595. 19. Missel P, Horner M, Muralikrishnan R. 2010. Simulating dissolution of intravitreal triamcinolone acetonide suspensions in an anatomically accurate rabbit eye model. Pharm Res 27(8):1530–1546. 20. Durairaj C, Chastain JE, Kompella UB. 2012. Intraocular distribution of melanin in human, monkey, rabbit, minipig and dog eyes. Exp Eye Res 98(0):23–27. 21. Loch C, Zakelj S, Kristl A, Nagel S, Guthoff R, Weitschies W, Seidlitz A. 2012. Determination of permeability coefficients of ophthalmic drugs through different layers of porcine, rabbit and bovine eyes. Eur J Pharm Sci 47(1):131–138. 22. Stocchino A, Repetto R, Siggers JH. 2010. Mixing processes in the vitreous chamber induced by eye rotations. Phys Med Biol 55(2):453. 23. Rolfs M. 2009. Microsaccades: Small steps on a long way. Vision Res 49(20):2415–2441.


24. Martinez-Conde S, Macknik SL, Troncoso XG, Hubel DH. 2009. Microsaccades: A neurophysiological analysis. Trend Neurosci 32(9):463– 475. 25. Stay MS, Xu J, Randolph TW, Barocas VH. 2003. Computer simulation of convective and diffusive transport of controlled-release drugs in the vitreous humor. Pharm Res 20(1):96–102. 26. Wilson CG, Tan LE, Mains J. 2011. Principles of retinal drug delivery from within the vitreous drug product development for the back of the eye. In Kompella UB, Edelhauser HF, Eds. Springer, New York, USA, pp 125–158. 27. Xu J, Heys J, Barocas V, Randolph T. 2000. Permeability and diffusion in vitreous humor: Implications for drug delivery. Pharm Res 17(6):664–669. 28. Maurice DM. 1987. Flow of water between aqueous and vitreous compartments in the rabbit eye. Am J Physiol 252(1):F104– F108. 29. Friedrich S, Cheng Y-L, Saville B. 1997. Drug distribution in the vitreous humor of the human eye: The effects of intravitreal injection position and volume. Curr Eye Res 16(7):663–669. 30. Meredith TA. 2006. Intravitreal antimicrobials. In Intraocular drug delivery; Jaffe GJ, Ashton P, Pearson PA, Eds. New York: Taylor & Francis Group, S. 85–93. 31. Johnson MW. 2010. Posterior vitreous detachment: Evolution and complications of its early stages. Am J Ophthalmol 149(3):371– 382. 32. Hesse L, Nebeling B, Schroeder B, Heller G, Kroll P. 2000. Induction of posterior vitreous detachment in rabbits by intravitreal injection of tissue plasminogen activator following cryopexy. Exp Eye Res 70(1):31– 39. 33. Tan LE, Orilla W, Hughes PM, Tsai S, Burke JA, Wilson CG. 2011. Effects of vitreous liquefaction on the intravitreal distribution of sodium fluorescein, fluorescein dextran, and fluorescent microparticles. Invest Ophth Vis Sci 52(2):1111–1118.

DOI 10.1002/jps.23808

[Unexpected changes in the vitreous body after posterior detachment].

Retinal hemorrhages in posterior vitreous detachment.

Complications of acute posterior vitreous detachment.

Correlation between vitreous pigment granules and retinal breaks in eyes with acute posterior vitreous detachment.

Influence of posterior vitreous detachment and type of intraocular lens on lipid peroxidation in the human vitreous.

Posterior vitreous detachment and retinal detachment after implantation of the Visian phakic implantable collamer lens.

[Vitreous body and retinal vein occlusion].

A method for induction of posterior vitreous detachment during vitrectomy.

[Pre-basal vitreous body after subtotal vitrectomy of aphakic eyes].

Flow and diffusion in the vitreous body of the eye.

Posterior precortical vitreous pocket.

Rheology of the vitreous body. Part I: Viscoelasticity of human vitreous.

Interleukin-6 concentrations in the vitreous body of patients with retinal detachment.

Posterior vitreous detachment after cataract extraction in non-myopic eyes and the resulting retinal lesions.

VITREOMACULAR TRACTION AFTER DEXAMETHASONE INTRAVITREAL IMPLANT (OZURDEX) INJECTION: THE EFFECT OF ANOMALOUS POSTERIOR VITREOUS DETACHMENT.

Drug delivery implants in the treatment of vitreous inflammation.

Serous Effusion into the Vitreous.

Elevated vitreous body glial fibrillary acidic protein in retinal diseases.

Aspiration from the vitreous of a non-magnetic foreign body.

Relationship between variations in posterior vitreous detachment and visual prognosis in idiopathic epiretinal membranes.

The vitreous.

[Pars plana vitrectomy following accidental injection of depot steroids into the vitreous body].

[Electron microscopic investigations of vitreous collagen after treating the vitreous with liquefying substances (author's transl)].

Possible Relation between Lack of Posterior Vitreous Detachment and Severe Endogenous Endophthalmitis.