Efficient Intraocular Penetration of Topical Anti–TNF-α Single-Chain Antibody (ESBA105) to Anterior and Posterior Segment without Penetration Enhancer

Abstract

purpose. This study was designed to characterize ocular penetration pathways of ESBA105, a topically administered single-chain antibody (scFv) against tumor necrosis factor (TNF)-α, to the anterior and posterior segment of the eye.

methods. Fresh enucleated whole eyes and isolated corneas of rabbits mounted in perfusion chambers were used for ex vivo penetration studies. In vivo pharmacokinetics and ocular biodistribution of ESBA105 after intravitreal injection or topical administration as eye drops were investigated in rabbits.

results. After topical administration as eye drops, without a penetration enhancer, ESBA105 reached therapeutic levels in the anterior and posterior segment of the eye. ESBA105 migrated to aqueous humor via corneal penetration and vitreous and retina via intrascleral penetration pathways. In vivo, ESBA105 had a significantly prolonged elimination half-life in the vitreous of 25 hours compared with its serum half-life of 7 hours after i.v. administration. Therefore, based on frequency of topical dosing, a buildup of ESBA105 to distinct steady state levels in the vitreous could be achieved.

conclusions. Topically administered ESBA105 quickly reaches therapeutic levels in the anterior and posterior segment without any need for a penetration enhancer. Drug penetration and ocular biodistribution patterns of ESBA105 applied as eye drops appear highly attractive for clinical use to treat TNF-α dependant diseases of the eye.

Monoclonal antibodies have revolutionized modern medicine in many therapeutic areas. However, due to their nature as large macromolecules with molecular weights of 150 kDa, they have to be applied by injection. Significant systemic drug exposure is generally not justified in treatment of ocular diseases, although it is sometimes accepted due to lack of alternatives. ¹ Consequently, local antibody-based treatment of ocular diseases by means of intravitreal injection has been introduced to ophthalmology with the use of anti-VEGF agents for treatment of exudative age-related macular degeneration (AMD). ² Although this approach has offered considerable therapeutic options for AMD patients, intravitreal injection is a laborious and potentially risky procedure. ^{3 4} Thus, the challenge of convenient and safe ocular delivery of antibody therapeutics for posterior segment disease remains an important issue. 

Feasibility of topical administration of antibodies or antibody fragments to the eye would represent a major innovation in ophthalmology. Recent data have suggested that the single-chain antibody (scFv) format with a molecular weight of 26 kDa, consisting of the variable domains of conventional monoclonal antibodies interconnected by a peptide linker, may represent an optimal antibody format to reach this pharmacological goal. ^{5 6} However, the biophysical properties of common scFv antibodies, in particular limited protein solubility, required inclusion of penetration enhancers (sodium caprate) to the topical formulation to achieve measurable intraocular scFv levels. ^{5 6} Mechanistically, such penetration enhancers ease penetration of macromolecules through epithelial barriers, likely by negatively affecting integrity of intercellular tight junctions or possibly by modification of the epithelial plasma membrane. ⁷ Therefore, local toxicity of penetration enhancers may limit long-term clinical use in ophthalmology. ⁸ In particular, sodium caprate, previously shown to be the most effective penetration enhancer for topically applied scFvs, ⁵ is not an adequate penetration enhancing excipient for (long-term) clinical use in ophthalmology due to its toxicity on epithelial cells. Also, sodium caprate has so far not been approved by the FDA for topical use on the eye. 

To overcome the limitations of common scFvs, such as poor solubility and tendency to aggregate, we have generated ESBA105, a novel humanized scFv based on a fully human scFv framework. The framework used for ESBA105 was selected for optimal drug like properties from a human library, as described previously. ⁹ On this framework, complementarity determining regions of a murine monoclonal antibody directed against human tumor necrosis factor (TNF)-α were grafted, resulting in ESBA105. ⁹ ESBA105 has similar anti-TNF-α potency as the full IgG antibody infliximab (manuscript in preparation). 

TNF-α is a clinically validated therapeutic target for various ocular diseases. ¹ However, due to the requirement for systemic administration of TNF-α inhibitors, clinical use of drugs such as infliximab is limited to immediately sight-threatening diseases only. ¹  

The aims of this study were to investigate the effect of various penetration enhancers and increasing drug concentration on the ocular penetration of topically administered ESBA105; to characterize ESBA105’s different penetration pathways to and elimination kinetics from the anterior and posterior segment of the eye; and to evaluate the ocular bioavailability of topical ESBA105 in vivo in rabbits. 

Materials and Methods

Production and Formulation of ESBA105

ESBA105 consists of 246 amino acids and has a molecular weight of 26,255 kDa. ¹⁰ For the studies presented here, ESBA105 was expressed recombinantly in Escherichia coli BL21 (DE3), refolded from inclusion bodies, purified, and formulated at 10 mg/mL in 25 mM sodium phosphate (pH 6.5; PBS). For screening of penetration enhancers, 2 mg/mL ESBA105 were diluted 1:2 with PBS containing the following individual twofold concentrated excipients: benzalkonium chloride, sodium caprate, castor oil, DMSO, EDTA, saponin (Fluka, Buchs, Switzerland); Brij35, Brij78, taurodeoxycholate (Sigma-Aldrich, Buchs, Switzerland); Tween20, Tween80 (Applichem, Darmstadt, Germany); and chlorhexidine (Sigma-Aldrich). For animal experiments, buffer exchanges were performed by ultra- or diafiltration followed by sterile filtration. Monomer content of ESBA105 was determined by analytical size exclusion chromatography and was always above 95% (data not shown). ESBA105 is stable for at least one week at 37°C in serum, aqueous, and vitreous humor. ¹⁰ In all formulations used for in vivo experiments the lipopolysaccharide (LPS) content was below 0.1 EU. 

FITC Labeling of ESBA105

Fluorescein isothiocyanate (FITC; Sigma-Aldrich) was dissolved in anhydrous DMSO (Sigma-Aldrich) immediately before use and added to reach a ratio of 40 μg FITC per mg ESBA105. The reaction mixture was kept in the dark on a rotator wheel over night. Separation of unbound FITC from labeled ESBA105 was performed by dialysis using 3 mL dialysis-cassettes (Socochim, Lausanne, Switzerland) in 5 L PBS (pH 6.5) at 4°C. Dialysis with 4 changes of dialysis buffer was performed during 48 hours. 

After performance of the ex vivo penetration experiment (procedure described below) with ESBA105-FITC, the eyes were embedded in OCT medium (Sakura; Digitana, Horgen, Switzerland) and frozen. Eye sections of 16 μm thickness were cut with a cryostat microtome (HM560; Microme, Volketswil, Switzerland). Tissue sections were evaluated using a microscope (Leica DM RE; Leica Microsystems, Heerbrugg, Switzerland) and a digital camera (Leica DC500; Leica Microsystems). (The software used to acquire the pictures was Leica IM50.) 

Ex Vivo Penetration Experiments

Fresh enucleated whole eyes from New Zealand White rabbits were obtained from Metzgerei Schönbächler, Wädenswil, Switzerland, and were used within 20 hours. Enucleated eyes were transported in humid chambers at 4°C and subsequently stored in PBS (pH 7.4) at 4°C. Penetration experiments using intact whole enucleated eyes were carried out as described previously ⁵ with minor modifications. Briefly, eyes were incubated in spherical plastic wells of a diameter of 14 mm and a radius of curvature of 7.5 mm, thus limiting the exposed ocular surface to approximately 1.7 cm². To minimize adsorption of ESBA105, plastic wells were blocked with 500 μL/well of 5% low-fat milk diluted in PBS (pH 7.4) before incubation. Then, 125 μL of the respective formulation were added to wells and eyes were placed in the wells such that either cornea or sclera were in direct contact with the formulated drug. After drug exposure, eyes were washed 3× with PBS. Aqueous humor or vitreous fluid were collected with a 25-G or 22-G needle (Becton Dickinson, Allschwil, Switzerland), respectively. Samples were snap frozen in liquid nitrogen and stored at −80°C. 

Penetration experiments on perfused, isolated corneoscleral preparations were carried out as described previously by Thiel et al. 2002 ⁵ with minor modifications. Corneoscleral preparations were dissected from whole eyes and mounted in modified Ussing chambers consisting of a donor and receptor compartment (PermeGear Inc. Hellertown, PA). The penetration chamber consisted of a bottom and a lid block between which the cornea was mounted with the epithelium facing the donor compartment. The central cavity of the lid and the bottom block had a diameter of 10 mm limiting the area of tissue exposed to the formulation to 0.79 cm². Both lid and bottom block contained perfusion lines for in- and outflow connected to a peristaltic pump run at 0.5 mL/min for the receptor chamber. The total volume of the receptor compartment was 5 mL and contained physiologic saline (BSS-Plus; Alcon, Hünenberg, Switzerland). Approximately 1.0 mL of test formulation was added to the donor compartment before incubation of the entire installation at 37°C and 100% humidity. At hourly intervals, 0.1 mL of receptor chamber fluid was taken for analysis and replaced by physiologic saline. ESBA105 concentration in samples was assessed by quantitative ELISA (see below). Endothelium vitality was monitored by measuring cornea thickness with a hand-held pachymeter (SP-100; Medilas, Geroldswil, Switzerland) as described by Thiel et al. ⁵ Fluorescein staining of corneas was applied to assure integrity of the epithelial surface before mounting the corneas onto the chambers. Trypan blue staining of corneas was used to monitor integrity of corneas after experiments. Excised rabbit corneas were washed 3× in PBS and immersed for 1 minute in 0.2% Trypan blue solution (Sigma-Aldrich). Subsequently corneas were rinsed with PBS and washed for 4 minutes in 1% NaCl. Stained corneas were scanned and intensity of blue staining was quantified using analysis software (ImageQuant TL; Amersham Bioscience, Otelfingen, Switzerland). 

In Vivo Measurement of Penetration, Biodistribution, and Pharmacokinetics

In vivo studies in rabbits were conducted by RCC, Itingen, Switzerland, and were in line with Swiss regulatory and ethical guidelines for animal experimentation, as well as being conducted according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

To assess ocular and systemic pharmacokinetics, 0.05 mL of 10 mg/mL ESBA105 were administered as intravitreal injections into male New Zealand White rabbits (n = 10). Animals were killed 6, 24, 48, 72, and 96 hours after injection (n = 2 at each time point) and aqueous humor, vitreous humor and sera were collected. From two animals (96-hour time point), serum samples were collected at each of the preceding time points from ear veins. ESBA105 levels were measured by quantitative ELISA (see below). 

For topical administration studies, New Zealand White rabbits received one drop (0.03 mL) of 10 mg/mL ESBA105 to both eyes every 20 minutes for up to 10 hours. Drops were administered to the center of the corneal surface and eye lids were hold closed for a few seconds to allow for spill over of excessive fluid. At time points 3, 5, 8, 10, and 24 hours, two animals per group were killed. Aqueous and vitreous humor samples from both eyes as well as serum were collected for later analysis by ELISA. After cessation of the 10 hour treatment period two animals per group were kept alive for another 14 hours (wash-out period). From these two animals serum was collected at time points 0, 3, 5, 8, 10, 12, 16, and 24 hours. Original samples as well as aliquots and dilutions were stored at −80°C until analysis. For pharmacokinetics calculations, average values of all samples from a given time point were used. Calculations were done using analysis software (PK-solutions 2.0; Summit Research Services, Montrose, CO). 

Quantification of ESBA105 Concentrations in Liquids and Tissues

Ninety-six well plates (NUNC; Maxisorp) were coated overnight at 4°C with an affinity purified polyclonal rabbit anti-ESBA105 antibody (AKA3A), diluted 1:3000 in PBS. After immobilization plates were washed three times with 300 μL/well TBS-T (0.005% Tween20) using a micro plate washer (ASYS Atlantis, Salzburg, Austria). Unspecific binding sites were blocked by a two-hour incubation in blocking-buffer (PBS, 1% BSA, 0.2% Tween20). After blocking, plates were washed three times with TBS-T (0.005% Tween20). Predilutions of each sample in the respective matrix were generated to keep the matrix amount constant in each assay. For predilutions, ocular fluids from New Zealand White rabbit eyes obtained from Metzgerei Schönbächler were used. Samples were diluted 1:10 in blocking buffer and a volume of 50 μL was added to wells. Standard dilution series were generated in the respective matrix and added on each individual plate. After incubation for 90 minutes, plates were washed three times with 300 μL/well TBS-T (0.005% Tween20) and subsequently incubated for 90 minutes with 50 μL/well of a biotinylated affinity purified monoclonal mouse anti-ESBA105 antibody (Mono31, ESBATech, Schlieren, Switzerland) diluted to a final concentration of 100 ng/mL in blocking buffer. Mono31, in turn, was detected with poly-horseradish peroxidase streptavidin (Stereospecific Detection Technologies, Baesweiler, Germany) at a concentration of 0.2 ng/mL, diluted in blocking buffer. After incubation and three washing steps, 50 μL/well of peroxidase substrate (POD, Roche Diagnostics, Rotkreuz, Switzerland) was added. The color reaction was stopped after 2 to 20 minutes (depending on color intensity) by addition of 50 μL/well 1 M HCl. Absorbance was measured at 450 nm in a plate reader (GENios; Tecan, Maennedorf, Switzerland) and concentrations in samples were calculated by polynomial regression from a standard curve (GraphPad Prism 4.03 software; GraphPad Software, Inc., San Diego, CA). The lower limit of quantitation (LOQ) was defined for each individual plate as OD₄₅₀ of background ± 3 times its SD. 

Estimation of Apparent Penetration Coefficient

The apparent penetration coefficient P_app = (dQ/dt)*(1/A*C ₀), wherein dQ/dt is the time-dependant concentration change in the receptor chamber or the ocular compartment, A is the exposed area, and C ₀ the analyte concentration in the formulation, was calculated for the various experimental setups. 

Results

Cornea Penetration of ESBA105

Penetration of molecules through the cornea is mainly limited by the outermost epithelial cell layer containing tight junctions. ^{5 11} For dextrane derivatives, penetration has been shown to be molecular weight dependant. ¹² To confirm the ability of ESBA105 (a molecule of 26.3 kDa molecular weight) to penetrate across the cornea, freshly enucleated, intact rabbit eyes were incubated “cornea down” for 6 hours in spheric wells in PBS with 1.0 mg/mL of ESBA105, or a full-length IgG (infliximab), or a mixture of both, all supplemented with 0.5% sodium caprate as penetration enhancer. For these studies, infliximab was chosen as a full IgG control antibody because it is an established TNF-α inhibitor. ESBA105 efficiently migrated through all corneal tissue layers, and after 6 hours reached 30 μg/mL in the anterior chamber (Fig. 1) . In contrast, infliximab did not penetrate above LOQ. In presence of 0.5% sodium caprate, P_app for ESBA105 was 8.2 × 10 to 7 cms⁻¹, corresponding to a penetration rate of 4%. However, 0.5% sodium caprate exhibited strong toxic effects on corneal epithelium compared with PBS alone when evaluated by Trypan blue staining at 4°C or 37°C (data not shown). In addition, 0.5% sodium caprate affected endothelial cell viability resulting in increased corneal thickness (Fig. 2) . Benzalkonium chloride 0.05% enhanced ESBA105 penetration through the cornea in a dose-dependent manner, whereas EDTA 0.1% alone or in combination with 0.05% benzalkonium chloride had no effect on corneal penetration of ESBA105 (Fig. 3) . Use of 1% Brij78 or 1% Tween80 increased corneal penetration of ESBA105 approximately fourfold, whereas 10 mM taurodeoxycholate or 0.5% saponin were even more effective (10-fold increase). In addition, chlorhexidine at 0.2% enhanced penetration sevenfold (Fig. 4) . 

Efficient Cornea Penetration of ESBA105 in Absence of Absorption Enhancers

In the next set of “cornea down” experiments, we investigated the effect of increasing concentrations of ESBA105 on penetration without use of penetration enhancers. Penetration of ESBA105 into aqueous humor (Fig. 5A)and vitreous (Fig. 5B)increases linearly with increasing ESBA105 concentration. Although there was no striking effect of 0.02% chlorhexidine (the approved concentration of chlorhexidine as a preservative in lens care products), still, penetration of ESBA105 to aqueous and vitreous was driven by ESBA105 concentration. There was no tissue saturation effect on penetration up to the highest concentration of ESBA105 tested. Importantly, ESBA105 levels in the anterior chamber with topically administered ESBA105 at 10 mg/mL reached comparable levels as with 1 mg/mL ESBA105 in presence of sodium caprate (Fig. 5A) . In this 6-hour short-term experiment, ESBA105 concentrations measured in vitreous were lower than in the anterior chamber (Fig. 5B) . This finding is partially explainable by the higher volume (1.5 mL) of vitreous humor compared to anterior chamber (0.2 mL) of rabbit eyes. In addition, in this experiment, contact of limbus and sclera to solution was kept to a minimum. Most interestingly, in contrast to the anterior chamber, the effect of sodium caprate on vitreous levels was modest and already doubling the ESBA105 concentration more than compensated for sodium caprate in the formulation (Fig. 5B)which may be explained by penetration of ESBA105 at the limbus, which was in contact with the meniscus of the formulation in this spheric well setup (see Discussion). 

Cornea Penetration Kinetics of ESBA105 in Absence of Penetration Enhancers

To assess cornea penetration kinetics of ESBA105, isolated rabbit corneas with intact epithelium were mounted on modified Ussing chambers. Figure 6shows the time-dependant penetration of ESBA105 at 10 mg/mL in PBS without penetration-enhancing excipient through the rabbit cornea. After a lag phase of approximately 2 hours, ESBA105 levels rose linearly in the receptor chamber between 2 and 8 hours, when the experiment was terminated. In this experiment with the central cornea exposed, P_app was 9.2 × 10⁻⁸ cms⁻¹, with a penetration rate of 0.02% in 8 hours. 

The above results were confirmed in intact whole rabbit eyes by quantitative measurement of ESBA105 concentrations in the anterior chamber as well as monitoring transcorneal migration of FITC-labeled ESBA105 by fluorescence microscopy. FITC-ESBA105 migrated into corneal stroma and was evenly distributed after 4 hours (Fig. 7) . Interestingly, fluorescence intensity was higher in endothelium than in stroma, indicating partial enrichment potentially due to barrier function of this tissue layer that limits free diffusion. ESBA105 reached detectable levels in the anterior chamber after a lag time of 2 to 3 hours, which correlated with results from microscopy. P_app in the intact eye setup that exposed central and limbal cornea to ESBA105 was 10-fold higher as in the experiment with isolated corneas and exposure of the central cornea only (Fig. 6) . Again, these findings suggest efficient translimbal penetration of ESBA105 in whole eye experiments. 

Intraocular Migration Routes of ESBA105

Limbal drug penetration offers the possibility of direct drug delivery from the ocular surface to the posterior segment of the eye by intrascleral penetration. We measured efficiency of ESBA105 penetration to the posterior segment by comparing transcorneal and translimbal/intrascleral migration using the whole eye setup (Fig. 8) . While 0.5% sodium caprate significantly enhanced uptake to anterior chamber and also vitreous, there was no significant effect of sodium caprate on ESBA105 penetration to the retina. This indicates that delivery to the retina mainly occurs via penetration through the limbus, a penetration pathway that cannot be enhanced by penetration enhancers (see Discussion). For further confirmation the experiments were repeated, exposing an identical scleral area by placing the eye in the well with sclera instead of cornea in direct contact with ESBA105. There was a rapid penetration of ESBA105 to retina that again could not be enhanced by sodium caprate (Fig. 8B) . In fact, with sclera instead of cornea in direct contact with the formulation, ESBA105 was not detectable in aqueous, indicating different migration routes for delivery to anterior and posterior compartments. Finally, our results suggest that in an ex vivo setup where translimbal penetration is possible, ESBA105 levels in retina are even higher than in vitreous (Fig. 8) . Thus, ex vivo ESBA105 seems to migrate directly within the sclera rather than indirectly through the vitreous to the retina. Most importantly, our data suggest that for topical delivery of ESBA105 to posterior ocular compartments, absorption enhancers are not required. Ultimately, our data show that for such delivery, high solubility and monomeric state of a scFv antibody are prerequisites. 

Distribution of ESBA105 after Intravitreal Injection In Vivo in Rabbits

To assess local pharmacokinetics of ESBA105 single doses of 0.5 mg (0.05 mL of 10 mg/mL) of ESBA105 were intravitreally injected into rabbit eyes (Fig. 9) . After intravitreal injection, levels of ESBA105 in serum were found to be 10,000-fold lower than in vitreous during the entire observation period. The local half-life of ESBA105 in the vitreous was approximately 25 hours which is significantly higher than the systemic half-life of 7 hours after i.v. injection in rabbits (data not shown). After intravitreal injection ESBA105 concentrations in aqueous humor were 10- to 100-fold lower than in vitreous at every time point tested (Fig. 9) . 

Topical Application of ESBA105 to the Eyes of Rabbits

At 3 drops per hour topically to both eyes of rabbits, 10 mg/mL ESBA105 with or without 0.02% chlorhexidine resulted in detectable ESBA105 concentrations in aqueous as well as in vitreous humor after only 3 hours, the first time point tested (Fig. 10) . With this high frequency dosing schedule, concentrations increased first in the anterior chamber and slightly delayed in the vitreous. After cessation of dosing at 10 hours, ocular concentrations decreased with the expected half-life of approximately 25 hours, confirming results from direct intravitreal injections (Fig. 9) . Interestingly, anterior chamber concentrations were higher when 0.02% chlorhexidine was present in the formulation (Table 1) . In this in vivo setting, there was no positive effect of chlorhexidine on ESBA105 penetration to vitreous (Tables 1 and 2) . A likely explanation for this may be accelerated absorption of topical ESBA105 into the circulation by increased conjunctival absorption due to chlorhexidine, which would more rapidly diminish the amount of ESBA105 available on the surface of the eye for intraocular uptake. With chlorhexidine in the formulation, elimination from the circulation was considerably slower compared to the formulation lacking chlorhexidine, leading to a more than threefold higher exposure. Due to the application of ESBA105 to both eyes and the high frequency dosing performed in this study, serum concentrations were slightly higher than local concentrations during the 10 hour treatment phase. However, serum concentrations dropped rapidly after cessation of treatment when no chlorhexidine was present (Fig. 10) . In contrast, local ocular concentrations diminished with a significantly prolonged half-life of 29.3 hours. P_app in this in vivo experiment was 1 × 10⁻⁹ cms⁻¹, corresponding to a penetration rate of 0.00005%. This low P_app can be explained by the fast clearance from the ocular surface. However, high levels of ESBA105 (in the ng/mL range) are reached in all ocular compartments within a few hours in vivo, which needs to be set in relation to local TNF-α levels in the pg/mL range in uveitis patients (see Furrer et al. ¹⁰ ). 

Discussion

Here we describe ex vivo and in vivo data with ESBA105, a scFv antibody with excellent biophysical properties on topical application to the rabbit eye. Ex vivo, the use of Ussing chambers allowed assessing corneal penetration, whereas the use of whole enucleated eyes allowed assessing both, corneal and limbal/scleral penetration. In vivo studies allowed adding highly relevant elimination routes (loss from eye surface, systemic elimination, biodistribution/pharmacokinetics) to the experimental setup. 

The studies with penetration enhancers presented here put the usefulness of penetration enhancers for topical administration of suitable scFv antibodies to the eye in question. Firstly, effects of penetration enhancers on ESBA105 penetration were observed at low concentrations only, and secondly, the effects on penetration efficiency of ESBA105 could be achieved by simply rising the concentration of ESBA105 (Fig. 5) . In fact, penetration of ESBA105 could not be saturated at concentrations of ESBA105 as high as 10 mg/mL. Thus, rather passive diffusion than transporter-driven absorption accounts for corneal and translimbal penetration. This hypothesis was tested and supported by addition of brefeldin A or sodium azide, which block active transport across cellular membranes and which both were without effect on ESBA105 penetration to aqueous or vitreous (data not shown). Also, we observed toxic effects of penetration enhancers, in particular sodium caprate, at active concentrations on cornea (Fig. 2) . Similar toxic findings for a different set of penetration enhancers were reported by others. ⁸ In addition, systemic absorption of ESBA105 was enhanced when applied in combination with a penetration enhancer (Fig. 10 , Tables 1 2 ) and presence of a penetration enhancer had no impact on retinal levels reached (Fig. 8) . In summary, our studies strongly suggest that especially for long-term topical treatment with a scFv antibody, penetration enhancers are clearly not desirable. This is especially true for molecules such as ESBA105, which penetrate sufficiently without penetration enhancers and can be applied long-term without toxicity; in fact, ESBA105 (without penetration enhancer in the vehicle) has been administered topically at high concentration (10 mg/mL) to the eyes of rabbits at 16 drops daily for 7 days followed by 5 drops daily for 21 days without eye irritation or any sign of local or systemic toxicity (data not shown). Thus, instead of including a penetration enhancer in the vehicle, as illustrated in the series of studies presented here, the key to efficient penetration of a protein therapeutic after topical application to the eye are its biophysical properties such as molecular weight, solubility, monomeric state, and stability. 

We performed studies with and without sodium caprate as a penetration enhancer to evaluate penetration to anterior and posterior parts of the eye (Fig. 5) . In the spheric well setup used in these studies, the meniscus of the formulation can reach the limbus, although only the cornea is thought to be exposed to the solution. Alternatively, cornea or sclera can be exposed to the formulation (see Fig. 8 ). Since the average pore-size of the limbus epithelium is larger than that of the cornea epithelium, diffusion through this part of the exposed area is expected to be more efficient than penetration through the cornea. Thus, the effect of sodium caprate on translimbal migration could be expected to be of only minor importance for penetration of a molecule the size of a scFv. The substantial intravitreal ESBA105 levels described in Figure 8 , reached after only a few hours, are best explained by a translimbal/intrascleral migration pathway resulting in (non-exclusive) delivery to the vitreous. Migration of ESBA105 from the anterior chamber to the vitreous on the other hand appears rather unlikely to contribute significantly. Thus, topically administered ESBA105 apparently can enter the inner of the eye by at least two distinct pathways. First, transcorneal penetration leads to rapid increase of ESBA105 levels in the anterior chamber (Table 2) . Second, ESBA105 enters the eye at the limbus and migrates within the sclera to the back of the eye (Table 2) . Our data strongly suggest that ESBA105 shows efficient intrascleral transport to vitreous and retina (for additional data see Furrer et al. ¹⁰ ). ESBA105 levels build up in the vitreous to steady state and that ESBA105 is eliminated from the vitreous with a prolonged half-life of 25 hours. Our results suggest that intraocular ESBA105 steady state levels can be “dialed in” either by treatment schedule or by variation of the ESBA105 concentration in the eye drop (see also Furrer et al. ¹⁰ ). From this vitreal “reservoir” ESBA105 can then distribute to other ocular compartments including anterior chamber and maintain ESBA105 levels there. Most importantly, although some small molecules may reach the retina on topical application, they are rapidly eliminated from the vitreous due to their low molecular weight. Thus, our finding of high levels of topically applied ESBA105 measured in the vitreous on topical administration is entirely novel. The 25-hour half-life of ESBA105 (26.3 kDa) is in line with the described correlation of vitreal half-life with molecular weight for molecules delivered by intravitreal injection. ¹³ In other words, in contrast to molecules of low molecular weight, molecules of 25 kDa are still small enough to penetrate but sufficiently large to result in slower clearance from the vitreous. 

In these first in vivo studies of ESBA105 in rabbits (see also Ref. ¹⁰ ), intentionally high frequency topical administration regimens were chosen to maximize ESBA105 levels in the anterior chamber. Indeed, within a few hours ESBA105 levels rise to a several 100-fold excess over local TNF-α levels reported in patients suffering from acute anterior uveitis. ^{10 14} In the accompanying study, ¹⁰ excess of ESBA105 over TNF-α levels measured in ocular diseases are shown in detail. This confirms suitability of topical ESBA105 for clinical use in this indication. It is noteworthy, however, that the results presented here strongly point to the usefulness of scFv-based topical therapies for treatment of diseases affecting the back of the eye as well. 

Based on pharmacokinetic modeling, steady state levels in the vitreous should be reached after approximately 5 half-lives; that is, after 5 consecutive days of topical treatment. Due to the prolonged half-life in the vitreous, this compartment should act as a reservoir or “depot” that constantly releases ESBA105 into adjacent tissues during treatment free periods (e.g., overnight). This unique profile should allow for highly attractive dosing regimens and prolonged treatment intervals for maintenance therapy. In addition, due to the constant release of drug from the vitreal reservoir, stable levels of ESBA105 can be expected also in the anterior chamber. 

For clinical application, a variety of parameters could still be optimized to improve penetration efficiency and potentially to increase the ratio between local and systemic exposure. For example, here eye drops were administered to the center of the pupil and excess fluid was removed by holding the eye lids closed for a few seconds. In contrast, application of eye drops into the lower eye sac, as is standard in human therapy, would allow keeping most of the volume of the drop in the eye without risking loss of a vast amount due to spill out. Also, the cul-de-sac would present a reservoir to potentially increase residence time of the formulation on the ocular surface. Furthermore, administration in the lower eye sac would favor trans-limbal/intrascleral migration to posterior parts of the eye, which is as shown here more efficient than transcorneal penetration of ESBA105. High vitreal levels of ESBA105 would also counteract continuous loss of drug from the anterior chamber due to the high fluid turn-over rate of the aqueous. 

 Submitted for publication June 3, 2008; revised July 17, 2008; accepted December 22, 2008.

 Disclosure: M. Ottiger, ESBATech AG (E); M.A. Thiel, ESBATech AG (C); U. Feige, ESBATech AG (I, E, C); P. Lichtlen, ESBATech AG (E, P); D.M. Urech, ESBATech AG (E, P)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Penetration of ESBA105 and infliximab through corneas of intact rabbit eyes. Enucleated rabbit eyes (n = 4) were incubated for 6 hours “cornea down” in spheric wells, such that the cornea exclusively was exposed to ESBA105 (1 mg/mL), or infliximab (1 mg/mL), or both, all with 0.5% sodium caprate. After incubation, samples from aqueous humor were collected and assayed by quantitative ELISA. BLQ, below limit of quantitation.

Figure 2.

 

Pachymetry of intact rabbit corneas. To measure corneal thickness, freshly enucleated rabbit eyes (n = 4) were incubated “cornea down” with only the cornea exposed to the formulation in spheric wells for 6 hours. Formulations contained 1 mg/mL ESBA105 without or with 0.5% sodium caprate in PBS. Cornea thickness was assessed with a hand-held ultrasonic pachymeter.

Figure 3.

 

Enhancement of penetration of ESBA105 through corneas of intact rabbit eyes in presence of benzalkonium chloride (BzCl) and/or EDTA. Fresh enucleated rabbit eyes (n ∼ 6–9) were incubated for 4 hours in spheric wells “cornea down” such that the cornea exclusively was exposed to ESBA105 at 1 mg/mL in PBS without or with BzCl and/or EDTA at concentrations indicated. After incubation, samples of aqueous humor were collected and assayed for ESBA105 levels by quantitative ELISA.

Figure 4.

 

Enhancement of penetration of ESBA105 through corneas of intact rabbit eyes in presence of various penetration enhancers. Fresh enucleated rabbit eyes (n ∼ 7–13; except n = 3 for Tween80 and n = 4 for Tween20) were incubated “cornea down” with only cornea exposed to the formulation for 4 hours in spheric wells with 1 mg/mL ESBA105 in PBS and penetration enhancers as indicated. After incubation, samples of aqueous humor were collected and assayed for ESBA105 levels by quantitative ELISA. VC (vehicle control) was PBS without ESBA105.

Figure 5.

 

Penetration of ESBA105 with or without chlorhexidine (0.02%) through cornea to anterior chamber (A) and vitreous body (B) of intact rabbit eyes. Fresh enucleated rabbit eyes (n = 3) were incubated “cornea down” with only the cornea exposed in spheric wells for 6 hours. ESBA105 concentrations were determined in collected aqueous humor (A) and vitreous humor (B) by quantitative ELISA. ESBA105 at 1 mg/mL with 0.5% sodium caprate served as a positive control.

Figure 6.

 

Penetration of ESBA105 through isolated rabbit corneas. Corneas (n ≥ 5 per time point) were excised from intact rabbit eyes and mounted on penetration chambers. Ten milligrams per milliliter ESBA105 in PBS was added to the donor chamber. ESBA105 concentrations in receptor chambers were determined by quantitative ELISA.

Figure 7.

 

Penetration of ESBA105 through corneas of intact rabbit eyes. Fresh enucleated rabbit eyes (n = 5 per time point) were incubated “cornea down” in spheric wells with only the cornea exposed to 1 mg/mL ESBA105. (A) Monitoring of trans-corneal migration of FITC-labeled ESBA105 by UV-microscopy (×100). (B) ESBA105 concentrations in aqueous were determined by quantitative ELISA.

Figure 8.

 

Penetration of topical ESBA105 to aqueous, vitreous or retina. Fresh enucleated rabbit eyes (n = 4 per formulation) were placed for 4 hours in spheric wells such that either the cornea (A) or an equal area of the sclera (B) was exposed to 1 mg/mL ESBA105. ESBA105 concentrations in aqueous, vitreous, or retina homogenate were determined by quantitative ELISA.

Figure 9.

 

ESBA105 concentrations in different eye compartments and in serum after intravitreal injection of 500 μg ESBA105. Measured concentrations and exponentially fitted trend lines are shown. ESBA105 concentrations were determined by quantitative ELISA.

Figure 10.

 

Topical administration of ESBA105 to the eyes of rabbits. ESBA105 concentrations in ocular compartments and in serum after 10 hours of repeated administration (3 drops per hour to both eyes) in vivo are presented. ESBA105 concentrations were measured by quantitative ELISA. Concentrations of both eyes of an individual rabbit at each time point were averaged. Signals below LOQ are indicated as zero. Standard deviations of average concentrations are not shown for clarity of the figure but can be found in Table 2 . (A) Formulation containing 11.2 mg/mL ESBA105 and 0.02% chlorhexidine as penetration enhancer. (B) Formulation containing 10.4 mg/mL ESBA105 without penetration enhancing additive. Note that maximal concentrations in serum as well as in aqueous are about twice as high in the presence of chlorhexidine whereas the opposite was observed for vitreous.

Table 1.

 

Summary of Pharmacokinetic Calculations

Time	Formulation A (0.02% chlorhexidine)					Formulation B (PBS)
		Vitreous	Aqueous	Serum	Vitreous	Aqueous	Serum
Terminal half time (h)	4.8	36.3	4.8	29.3	18.6	nd
C max (pg/mL)	2,614	9,570	23,553	4,002	5,697	14,122
T max (h)	10	8	12	10	10	10
AUCobs (pg/mL · h)	25,095	109,922	256,611	61,254	94,609	95,489

 For experimental details see legend of Figure 9 . C _max, highest concentration reached in respective compartment; T _max, time point of highest concentration; AUC_obs, area under the concentration curve over the observed period.

Table 2.

 

Local and Systemic Concentrations after Repeated Administration of Eye Drops to Rabbits

Time (h)	ESBA105 (pg/mL)
		Aqueous	Vitreous	Serum
Formulation A (0.02% chlorhexidine)
3	383 ± 370	0 ± 0	3722 ± 5883
5	6675 ± 2746	0 ± 0	13481 ± 13796
8	9570 ± 13792	703 ± 818	20554 ± 29563
10	5118 ± 2572	2614 ± 2195	21799 ± 20790
12	nd	nd	23553 ± 33309
16	nd	nd	6610 ± 9348
24	3916 ± 3691	346 ± 692	2077 ± 2937
Formulation B (PBS)
3	2347 ± 1214	1051 ± 2101	2420.6 ± 3107
5	3335 ± 1128	985 ± 710	4752.5 ± 2027
8	4468 ± 1906	1610 ± 1881	13952 ± 76401
10	5697 ± 4406	4002 ± 2979	14122 ± 9440
12	nd	nd	4811 ± 3766
16	nd	nd	0
24	3380 ± 1580	2875 ± 2057	0

The authors thank Peter Steiner and Werner Atzenhofer for providing ESBA105 for these studies, and Daniela Binggeli, Simone Wuethrich, and Anja Marold for excellent technical assistance.