Diffusion of High Molecular Weight Compounds through Sclera

Abstract

purpose. To determine the in vitro permeability of the sclera to high molecular weight compounds and the relationship between scleral permeability and molecular size.

methods. Fresh rabbit sclera was mounted in a two-chamber diffusion apparatus, and its permeability to sodium fluorescein, fluorescein isothiocyanate (FITC)–conjugated bovine serum albumin, FITC–IgG, and FITC dextrans ranging in molecular weight from 4 to 150 kDa was determined by fluorescence spectrophotometry. Electron microscopy was used to assess the impact of the experimental design on scleral ultrastructural integrity. The effect of the diffusion apparatus on scleral hydration was examined. Rabbit scleral permeability was compared with previously reported data for human and bovine sclera.

results. Scleral permeability decreased with increasing molecular weight and molecular radius, consistent with previous human and bovine data. Molecular radius was a better predictor of scleral permeability than molecular weight. The sclera was more permeable to globular proteins than to linear dextrans of similar molecular weight. The experimental apparatus did not alter scleral ultrastructure. Permeability of rabbit sclera was similar to human sclera but greater than bovine sclera.

conclusions. Large molecules, such as IgG, diffuse across sclera in a manner consistent with porous diffusion through a fiber matrix. Transscleral delivery of immunoglobulins and other large compounds to the choroid and retina may be feasible.

Avariety of chorioretinal disorders such as age-related macular degeneration, diabetic retinopathy, glaucoma, and retinitis pigmentosa may be amenable to specific anti-angiogenic or neuroprotective agents. ^{1 2 3 4 5 6} However, targeted delivery of drugs to the choroid and retina remains difficult and potentially dangerous. Systemic delivery of biologically active agents in doses capable of achieving therapeutic levels in the eye may cause unacceptable side effects. Repeated long-term intravitreous delivery, as would be required for these diseases, has been associated with significant complications such as cataract formation, endophthalmitis, retinal detachment, and vitreous hemorrhage. Furthermore, the internal limiting membrane of the retina impedes diffusion of linear molecules larger than 40 kDa and globular molecules larger than 70 kDa, ^{7 8 9 10} precluding intravitreous delivery for many of the antiangiogenic drugs under development. Topical delivery is not viable due to long diffusional path length, counter-directional intraocular convection, lacrimation, and corneal impermeability to large molecules. ¹¹  

It has been hypothesized that transscleral delivery may be a viable alternative. ^{12 13} The sclera has a large and accessible surface area, has a high degree of hydration rendering it conducive to hydrophilic molecules, is hypocellular with few protein-binding sites, has a paucity of proteolytic enzymes, and has a permeability that does not appreciably decline with age. ^{14 15 16 17} Previous in vitro experiments have demonstrated that the sclera is permeable to molecules as large as 70 kDa. ^{17 18} Knowledge of the diffusion properties of even larger molecules through sclera is desirable, because several candidate antiangiogenic drugs are 150 kDa antibodies. We also explored the relationship of scleral permeability to molecular weight and molecular (Stokes–Einstein) radius, because this information may aid in drug development. 

Methods

Isolation and Preparation of Fresh Rabbit Sclera

The experiments adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and guidelines developed by the Animal Care Committee of the Massachusetts Eye and Ear Infirmary. Dutch-belted rabbits (Pine Acres Rabbitry, Vermont, MA), each weighing 2 to 3 kg, were anesthetized with intramuscular 40 mg/kg ketamine (Abbott, N. Chicago, IL) and 10 mg/kg xylazine (Bayer, Shawnee Mission, KS). Scleral thickness was measured using a RK-5000 ultrasound pachymeter (KMI Surgical Products, West Chester, PA). The eyes were enucleated immediately before the rabbits were killed and immersed in Unisol (Alcon, Ft. Worth, TX) for 10 minutes or less. The adherent muscles were excised, and episcleral tissue was removed with a sterile gauze sponge. Areas free of nerve and vessel emissaries were used to obtain 7 × 12-mm slices of sclera under microscopic caliper guidance. Each piece of sclera was used on the day of isolation. 

In Vitro Diffusion Apparatus

A 5 × 10 window, 2 mm from the bottom, was created on one face of a spectrophotometry polystyrene cuvette (Sigma, St. Louis, MO) using a Bridgeport vertical milling machine (Bridgeport, CT), and a piece of sclera was blotted dry and placed over this window without stretching so as not to induce asymmetrical stresses. A small amount of cyanoacrylate tissue adhesive (Ellman International, Hewlett, NY) was applied to the entire boundary of the tissue rim to seal its cut surface to the cuvette and prevent leakage around the sclera, and a second identical cuvette was aligned with the first cuvette and glued in place over the tissue. After the glue polymerized, within 3 to 4 minutes, the cuvette facing the “orbital” surface of the sclera was filled with Unisol. The apparatus was discarded if leakage into the“ uveal” chamber was observed. Unisol was replaced with diffusion medium (see below), and the apparatus was incubated at 37°C in 5% CO₂ atmosphere for 1 hour to restore normal hydration and temperature. 

Diffusion Medium

Hanks’ balanced salt solution without phenol red, containing 1% glutamine–penicillin and streptomycin, tetracycline (48 μg/ml), and aprotinin (1.5 μg/ml; all from Sigma), was used as the diffusion medium. Tetracycline and aprotinin were excluded from the medium for certain experiments with fluorescein isothiocyanate (FITC)–bovine serum albumin (BSA) to evaluate the effect of these proteolysis inhibitors on scleral permeability. The pH of all solutions ranged from 7.41 to 7.45. 

Fluorescent Compounds

FITC–conjugated dextrans ranging in molecular weight from 4 to 150 kDa, FITC—BSA, FITC–rabbit IgG (all from Sigma), rhodamine–conjugated dextran of molecular weight 70 kDa (Molecular Probes, Eugene, OR), and sodium fluorescein (Akorn, Abita Springs, LA) were studied. At least 5 experiments were performed on each compound. To confirm that the parent compound was not cleaved from FITC, selected samples were subjected to protein precipitation with 20% trichloroacetic acid. ¹⁹ Samples were protected from light at all times before fluorescence measurements. 

Sample Collection

The medium in each “uveal” chamber was replaced by 4 ml of fresh medium at 37°C, whereas the “orbital” chamber was filled with an equal volume of diffusion medium containing 1 mg/ml of a fluorescent compound, freshly prepared and warmed to 37°C. Experiments were performed in a tissue culture incubator at 37°C in a 5% CO₂ atmosphere. Samples measuring 0.4 ml were removed from each chamber at 30-minute intervals for 4 hours and stored at −80°C. Solutions were stirred before each sample collection. 

Scleral Hydration

The water content of sclera was measured by comparing the wet weight of freshly obtained tissue to its dry weight, obtained by subjecting the tissue to drying at 100°C for 3 hours. The effect of the diffusion medium on scleral hydration ([wet weight − dry weight]/wet weight) was examined by comparing the water content of sclera exposed to the experimental apparatus for 4 hours to that of fresh sclera. 

Scleral Permeability Coefficient

Diffusion from the “orbital” chamber to the “uveal” chamber was characterized by means of a permeability coefficient (P _c), which is the ratio of steady state flux (the mass of solute crossing a planar unit surface normal to the direction of transport per unit time) to the concentration gradient. ²⁰ In these experiments, the concentration in the“ uveal” chamber, C _u, was a negligible fraction of the concentration in the “orbital” chamber, C _o, which did not change measurably over the course of the experiment. Within 30 minutes steady state diffusion was achieved; therefore, the permeability coefficient was calculated as follows:  

\[P_{\mathrm{c}}{=}\ \frac{(\overline{C_{\mathrm{u4}}}-\overline{C_{\mathrm{u0.5}}})V{\ast}}{AtC_{\mathrm{o}}}\]

where C _u0.5 and C _u4 are the concentrations in the“ uveal” chamber at 0.5 and 4 hours, respectively, estimated by linear regression on the concentration of the 8 collected samples, V ^* is the corrected chamber volume (4 ml divided by 3.6, to correct for the volume changes induced by sampling), A is the surface area of exposed sclera (0.84 cm²), and t is duration of steady state flux (3.5 hours). 

Analysis of Scleral Integrity

At the conclusion of selected experiments, the diffusion apparatus was thoroughly cleansed with Unisol, and the permeation of sodium fluorescein was observed and compared with diffusion kinetics of sodium fluorescein across fresh sclera to unmask possible damage to the sclera. The effect of cyanoacrylate tissue adhesive on scleral ultrastructure was examined by transmission electron microscopy. 

Fluorescence Measurements

Fluorescence was measured at room temperature (25°C) with a fluorescence spectrophotometer (model MPF-44A; Perkin–Elmer, Newton Center, MA) in a right-angle geometry. For FITC compounds, excitation and emission wavelengths were 492.5 and 520 nm, respectively. For rhodamine-conjugated dextran, excitation and emission wavelengths were 570 and 590 nm, respectively. Standard curves of fluorescence versus concentration were obtained by serial dilution of fluorescent compounds in diffusion medium. Concentrations in samples were determined by linear regression analysis within the linear portion of the standard curve. 

Statistics

An unpaired Student’s t-test was used to compare continuous variables. All probability values were two-tailed. An α level of 0.05 was used as the criterion to reject the null hypothesis of equality of means. 

Results

After the first 30 minutes of each experiment there was a constant flux of the fluorescent compound across the sclera. The permeability of sclera to the tracers studied is shown in Table 1 . Sodium fluorescein, the smallest compound, had the highest permeability coefficient (84.5 ± 16.1 × 10⁻⁶ cm/sec), whereas 150 kDa FITC–dextran, which had the largest molecular radius, had the lowest permeability coefficient (1.34 ± 0.88 ×10⁻⁶ cm/sec). The permeability coefficients of rhodamine-conjugated dextran 70,000 D and FITC-dextran 71,200 D were not significantly different (P = 0.88), strengthening the reliability of the paradigm. The sclera was more permeable to the two proteins tested (BSA and IgG) than to dextrans of comparable molecular weight. 

Scleral permeability declined exponentially with increasing molecular weight and molecular radius. Log-linear regression analysis demonstrated that molecular radius was a better predictor of permeability (r ² = 0.87, P = 0.001) than molecular weight (r ² = 0.31, P = 0.16; Fig. 1 ). 

Random samples containing FITC-BSA and FITC-IgG were subjected to protein precipitation with trichloroacetic acid after diffusion through sclera. The fluorescence of the resulting supernatants was not different from that of the diffusion medium, indicating that there was no significant dissociation of the FITC conjugate. 

The permeability coefficient of sodium fluorescein across fresh sclera (84.5 ± 16.1 × 10⁻⁶ cm/sec) was not significantly different from that across sclera previously used in a 4-hour in vitro diffusion apparatus (76.3 ± 24.1 × 10⁻⁶ cm/sec; P = 0.55). Transmission electron microscopy of sclera exposed to cyanoacrylate tissue adhesive demonstrated normal collagen fibrils in closely packed lamellae as well as normal banding patterns of collagen throughout the scleral stroma (Fig. 2) . There was no demonstrable difference between cyanacrylate-exposed and control sections, either in density of packing or in maximal width of individual collagen fibrils. There was no difference between tissue hydration of fresh sclera (69.5% ± 0.9%) and sclera exposed to diffusion medium for 4 hours (69.2% ± 0.4%; P = 0.66). Mean scleral thickness was 416 ± 21 μm. 

We did not find any significant difference in scleral permeability to FITC–BSA in media with (5.49 ± 2.12 × 10⁻⁶ cm/sec) and without (5.21 ± 1.85 × 10⁻⁶ cm/sec; P = 0.89) proteolysis inhibitors. 

Discussion

We measured the permeability of rabbit sclera to a series of fluorescein-labeled hydrophilic compounds with a wide range of molecular weights and radii. These data indicate that the sclera is quite permeable to high molecular weight compounds. In an ideal aqueous medium the Stokes–Einstein equation predicts that permeability declines as a linear function of molecular radius. However, in porous diffusion through a fiber matrix such as the sclera, permeability declines roughly exponentially with molecular radius, ^{15 21} as observed in these experiments. 

For all molecules studied, constant flux of compounds across the sclera occurred by 30 minutes, similar to observations in human sclera. ¹⁷ This is consistent with documented stability of FITC conjugation to the parent compound and the absence of aggregation. ²² In addition, protein precipitation revealed that the proteins studied, BSA and IgG, remained intact as they diffused across the sclera. The similarity in permeability coefficients of rhodamine–dextran 70,000 D and FITC–dextran 71,200 D reinforce the fidelity of the experimental design. 

Electron microscopy revealed no ultrastructural damage to the sclera caused by cyanoacrylate tissue adhesive (Fig. 2) . This was confirmed by the fact that permeability of fresh sclera to sodium fluorescein was not significantly different from that of used sclera. 

The use of proteolysis inhibitors (aprotinin and tetracycline) to limit tissue degradation and simulate in vivo sclera, which has a paucity of proteolytic enzymes in the absence of inflammation or injury, ¹⁴ did not alter scleral permeability. 

The reported permeabilities could be artificially low due to the resistance of unstressed layers along the scleral surface. However, the impact of possible boundary layers on high molecular weight tracers is not expected to be significant, especially when temperature fluctuations are minimal. ^{23 24}  

Our data are compared with the reported permeability of human and bovine sclera ^{17 18} (Fig. 3) . The permeability coefficients for rabbit and human sclera were converted to effective diffusivities (which are thickness invariant) by assuming thicknesses of 0.04 and 0.06 cm, respectively, and accounting for variations in scleral hydration using computer simulation of a mathematical model of transscleral diffusion. ¹⁵ Solute diffusivity in bovine sclera appears to be lower than in rabbit or human sclera, although the differences are less in magnitude than previously suggested, ¹⁷ because we have also attempted to account for differences in hydration. The residual differences may be due to differences in the scleral ultrastructure among species and differences in preparation and experimental techniques. 

In some tissues linear dextrans have greater diffusion rates than globular proteins of equivalent molecular weight. ²⁵ In contrast, our data, although limited to BSA and IgG, suggest that the sclera may be more permeable to proteins than dextrans. This disparity may stem from differential binding of proteins and dextrans to collagen fibers in the hypocellular sclera. ¹⁵ It is also interesting that proteins, despite having greater numbers of negative charges than dextrans, diffuse faster through sclera. This suggests that the molecular radius plays a greater role in determining scleral permeability than molecular weight or charge, similar to diffusion through extracellular tissue in brain. ²⁶ Scleral permeability to the 70-kDa dextran is not significantly greater than that to the 150-kDa dextran, suggesting that the hydrodynamic radii of these molecules within sclera is not identical with their molecular radii in aqueous solution. A similar situation exists in brain tissue, where values for diffusion of 40- and 70-kDa dextrans are not significantly different. ²⁷ These data are consistent with the existence of multiple diffusion passages in sclera with varying size limitations. The apparent plateauing above 70 kDa may have resulted from possible inhomogeneity in molecular weight of commercially available dextrans, but such dispersion would have similarly affected permeability even at lower molecular weight dextrans. In addition, the dispersion of the tracers used in this study is limited, with 90% of the dextrans being within 30% of the mean molecular weight. ²⁸  

Although the sclera appears to be permeable even to molecules as large as immunoglobulins, in vivo transscleral delivery faces other obstacles. Orbital clearance, intraocular pressure, uveoscleral outflow, choroidal blood flow, and the outer and inner blood–retinal barriers all may limit drug delivery to the posterior segment of the eye. However, our preliminary experiments have shown that transscleral delivery of protein-based reagents to the retina and choroid is possible. ²⁹ These data provide a rationale for further exploring the in vivo feasibility of transscleral delivery of high molecular weight compounds. 

 Supported by the Heed Ophthalmic Foundation (JA), the American Ophthalmological Society-Knapp Testimonial Fund (JA), a Foundation Fighting Blindness Career Development Award (JA), the Massachusetts Lions Eye Research Fund (APA), and the Roberta Siegel Research Fund (APA). E. S. Gragoudas is a Research to Prevent Blindness Scientific Investigator.

 Submitted for publication August 5, 1999; revised October 15 and December 7, 1999; accepted December 20, 1999.

 Commercial relationships policy: N.

 Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May, 1997.

 

Table 1.

 

The Permeability of Sclera to Tracers of Various Molecular Weights and Molecular Radii

Tracer	Molecular Weight (D)	Molecular Radius (nm)	Permeability Coefficient (×10−6 cm/sec, mean ± SD)
Sodium fluorescein	376	0.5	84.5 ± 16.1
FITC-D, 4 kD	4,400	1.3	25.2 ± 5.1
FITC-D, 20 kD	19,600	3.2	6.79 ± 4.18
FITC-D, 40 kD	38,900	4.5	2.79 ± 1.58
FITC–BSA	67,000	3.62	5.49 ± 2.12
Rhodamine D, 70 kD	70,000	6.4	1.35 ± 0.77
FITC-D, 70 kD	71,200	6.4	1.39 ± 0.88
FITC–IgG	150,000	5.23	4.61 ± 2.17
FITC-D, 150 kD	150,000	8.25	1.34 ± 0.88

 The molecular (Stokes–Einstein) radii were culled from the literature. ^{25 28 30 31} D, dextran; IgG, immunoglobulin.

Figure 1.

 

Scleral permeability versus molecular radius (A) and molecular weight (B). The least squares regression line is shown.

Figure 2.

 

(A) Transmission electron microscopy (TEM) of control sclera demonstrates normal intertwining collagen fibrils in closely packed lamellae. (B) TEM of sclera exposed to cyanoacrylate demonstrating collagen fibril packing similar to control sclera. (C) TEM of control sclera demonstrates normal collagen banding pattern. (D). TEM of sclera exposed to cyanoacrylate demonstrates collagen banding pattern and fibril diameters similar to those of control sclera. Scale bar, 5 μm for all.

Figure 3.

 

Scleral effective diffusivities (rabbit, human, ¹⁷ and bovine ¹⁸ ) versus molecular radius are depicted [rabbit, (♦); human, (▪); bovine, maximum value (•), minimum value (○)] for various FITC (F-) and rhodamine (R-) dextrans, bovine serum albumin (BSA), radioiodinated human serum albumin (RISA), hemoglobin (Hgb), and inulin. Effective diffusivities were calculated by multiplying permeability coefficients by tissue thickness (0.04 cm for rabbit and 0.06 cm for human). Because of differences in scleral hydration between studies, the data were also converted to yield the effective diffusivity using a mathematical model of transscleral diffusion. ¹⁵

The authors thank Tiansen Li at the Massachusetts Eye and Ear Infirmary for the use of a sonicator and Ambati M. Rao and Balamurali K. Ambati for critical review of this article.