New Human Organotypic Corneal Tissue Model for Ophthalmic Drug Delivery Studies

Abstract

Purpose: The purpose of the current work was to develop a physiologically relevant, in vitro human three-dimensional (3D) corneal epithelial tissue model for use in ophthalmic drug development.

Methods: Normal human corneal epithelial cells were cultured at the air–liquid interface to produce the 3D corneal tissue model. Corneal barrier was determined by measuring transepithelial electrical resistance (TEER). Quantitative PCR arrays were utilized to investigate expression of 84 phase I/II metabolizing enzymes and 84 drug transporter genes. Permeability was evaluated using model compounds with a wide range of hydrophobicity, molecular weight, and excipients. Finally, different formulations of latanoprost and bimatoprost were administered and drug absorption and tissue viability and integrity were investigated.

Results: Histologic assessment and TEER of the corneal tissue model revealed tissue structure, thickness, and barrier formation (1000 ± 146 Ω·cm²) comparable to native human corneal epithelium. The 3D corneal tissue expressed tight junctions, mucins, and key corneal epithelial detoxification enzymes. Drug-metabolizing enzyme and transporter gene expression in 3D corneal tissue and excised human corneal epithelium were highly correlated (r² = 0.87). Coefficients of permeation for model drugs in the tissue model and excised rabbit corneas also showed a high correlation (r² = 0.94). As expected, latanoprost and bimatoprost free acids had much lower permeability (Papp = 1.2 × 10⁻⁶ and 1.9 × 10⁻⁶) than the corresponding prodrugs (Papp = 2.5 × 10⁻⁵ and 5.6 × 10⁻⁵), respectively. The presence of 0.02% benzalkonium chloride in ophthalmic formulations significantly affected tissue barrier and viability.

Conclusions: The newly developed 3D corneal tissue model appears to be very useful for evaluation of corneal drug permeability and safety during ophthalmic drug development.

The human cornea is a transparent, avascular mucosal tissue, which serves as an optical interface and the first line of defense against many types of injury, trauma, and infection. The human cornea is a complex barrier composed of five layers: epithelium, Bowman's layer, stroma, Descemet's membrane, and endothelium. The corneal epithelium is a stratified, nonkeratinized squamous epithelium containing five to seven cell layers. Although it accounts for only approximately 10% of the overall corneal thickness, the corneal epithelium plays a key role in conferring corneal barrier properties.¹ Permeation of topically applied eye formulations mainly occurs through the cornea, which serves as a gatekeeper to control and regulate inward and outward transport of fluids, solutes, and administered drugs.² 

According to a study by Prevent Blindness America and the National Eye Institute, the number of adults ages 40 and older with vision impairment and blindness has increased 23% since 2000 and is expected to double by the year 2050.^3,4 Vision disorders among Americans have an estimated annual economic burden of more than $35.4 billion.⁵ Due to its convenient method of administration, topical application is the desired route of administration for ophthalmic drugs to treat diseases of the anterior segment including glaucoma, inflammation, and infections. For many formulations, topical application is not only convenient and simple, but also the only means to achieve the required therapeutic drug concentrations in the eye due to the associated adverse effects.⁶ 

The ocular drug bioavailability of topically applied formulations is typically below 5%.^7,8 The low bioavailability is due to multiple factors, including fast clearing of the drug from the surface of the eye, drug solubility, and protective mechanisms and barriers of the cornea.^2,9 The apical layer of the corneal epithelium forms tight junctions and presents the main rate-limiting step to the majority of topically applied lipophilic drugs.^10,11 Since the stromal compartment, composed of hydrated collagen fibrils, mainly poses a diffusion barrier to lipophilic compounds, and a single layer of endothelium is not a significant barrier to drug permeation, three-dimensional (3D) corneal epithelial tissue models are considered to be very useful in studying drug permeation.^1,9,12,13 

Not only does the cornea serve as a diffusion barrier, it also expresses certain drug-metabolizing enzymes and transporters that are known to affect ocular drug availability and absorption into the eye.^14–16 For instance, robust esterase activity of the corneal epithelium allows the use of ester prodrugs for improved delivery.¹⁷ Also, efflux and uptake transporters can potentially hinder or facilitate ocular drug absorption.¹⁵ It has been suggested that multiple ophthalmic drugs (including antibiotics, antiviral, antifungal, and anti-inflammatory) can interact with efflux transporters that translocate substances across the cellular membranes and thereby restrict the intracellular accumulation of drugs.¹⁸ Therefore, expression of ABC (ATP[adenosine triphosphate]-binding cassette) superfamily of transporters, including P-glycoprotein (P-gp) and multidrug resistance proteins (MRP), as well as amino acid and organic anion transporters in the cornea, may regulate drug bioavailability in the eye.^16,18,19 

The assessment of corneal drug penetration is crucial for the development of effective ophthalmic medicines. Several approaches that are employed to improve ocular bioavailability (including pre-corneal drug retention, use of penetration enhancers, mucoadhesives, in situ forming gels, liposomes, micro- and nanospheres, and more) have been previously reviewed.^9,20 The limited availability of human corneas for research purposes makes them impractical for drug screening programs. Until recently, absorption studies were commonly performed in laboratory animals or in excised rabbit, bovine, or porcine corneas mounted in diffusion chambers, such as Franz diffusion cells or Ussing chambers.^9,21,22 Not only are these studies labor intensive, expensive, poorly standardized, and impractical for large-scale and rapid drug formulation screening; they also exhibit interspecies variation due to the differences in anatomy and morphology of the eyes.^9,23 Pig eyes are structurally the most similar to human eyes in terms of the size, corneal thickness, and the presence of Bowman's layer. Nevertheless, rabbit eyes, in spite of being smaller than human eyes, having a nictitating membrane, very low blinking frequency, and higher permeability, are considered to be the reference animal of choice mostly due to easy ex vivo–in vivo correlation, as reviewed by Agarwal and Rupenthal.⁹ Furthermore, human and animal corneas may significantly differ in metabolic enzymes and transporters present on their surface, affecting the bioavailability especially of actively transported molecules.^9,14,24 

Emerging technologies provide abundant opportunities to replace traditional animal-based experiments.^9,25 Reconstructed 3D tissue models derived from immortalized corneal cells are appealing due to the accessibility of the cell source.²⁶ Although several models proved to be helpful in permeability and drug transport studies, most of them were developed to serve as a tool for eye irritation testing.^27–30 Cell lines are known to be nonreproducible due to the phenotypic and genetic drifts as a result of continued passaging of the cell line. Therefore the advantages offered by the use of human cell lines or primary nonhuman cells are often overshadowed by incongruous gene expression making them nonreliable or irrelevant.^9,16,31–33 For instance, considerably higher expression of P450 and UGTIA1 enzymes and considerably lower expression of uptake transporters were documented for corneal tissues reconstructed from immortalized human corneal epithelial cells, thus limiting their utility for drug absorption studies.³⁴ 

Recent success in the isolation and expansion of primary human corneal epithelial cells has allowed for a reliable and reproducible source of primary human corneal epithelial cells for the development of novel organotypic corneal tissue models.^35,36 Primary cultured human corneal epithelial cells grown on permeable supports can address questions concerning drug transport mechanisms by avoiding species extrapolation and the influence of modified gene expression due to cell immortalization.^27,37 These model systems can be valuable for corneal penetration studies since they allow high sample throughput, access to both the apical and basolateral sides, and A to B and also B to A transport studies.^9,37 These models are desirable because they present an ethical alternative to in vivo experimentation, improve test accuracy and reproducibility, and allow testing with organotypic cultures derived from different donors, thus representing a large population, similar to that observed in clinical trials.^4,25 Thus, the development of in vitro reconstructed human corneas would serve an important, growing need. 3D corneal tissue models will play a key role in corneal absorption screening and will facilitate ophthalmic drug development, improve the accuracy of pharmacokinetic studies, and in vitro–in vivo correlations.^9,27,34,38 

The goal of the current study was to characterize a new in vitro organotypic human corneal epithelial tissue model with respect to tissue histology, barrier function, and permeation, and expression of cornea-specific genes, key drug-metabolizing enzymes, and transporters necessary for accurate prediction of drug transport and absorption in human cornea. The performance of the 3D reconstructed corneal tissues in drug transport and biocompatibility studies was demonstrated by several formulations of known antiglaucoma drugs, latanoprost and bimatoprost. 

Materials and Methods

In Vitro Reconstructed 3D Corneal Tissue Model

The 3D corneal tissues were cultured using primary normal human donor cells; cells from different donors were not pooled together. Human donor corneas were harvested within 24 hours postmortem and shipped to MatTek Corporation (Ashland, MA, USA) by the National Disease Research Interchange (NDRI, Philadelphia, PA, USA) in Optisol-GS medium (Bausch & Lomb, Rochester, NY, USA) at 4°C. Primary human corneal epithelial cells were isolated immediately upon arrival at MatTek Corporation (48 hours postmortem) using the tissue outgrowth technique.^39,40 Briefly, corneal epithelial cells were cultured for two to four passages on FNC-coated (AthenaES, Baltimore, MD, USA) cell culture flasks in supplemented NHCE-GM medium (MatTek Corporation). Corneal cells exhibited normal epithelial cobblestone morphology (Fig. 1A) and expressed tissue-specific cytokeratins CK3/12 and CK15 (Figs. 1B, 2C). 

Figure 1

 

Characterization of normal human corneal epithelial cells during expansion in a monolayer culture. (A) Phase contrast microscopy of corneal epithelial cells at confluence (passage 4, ×10 objective). (B, C) Immunohistochemical analysis of corneal epithelial cells at passage 3. (B) CK 3/12 (green) and Ki67 (white). (C) CK15 (red); nuclear staining, DAPI (blue), ×60 objective.

Figure 2

 

(A) Schematic of the 3D corneal tissue model grown in cell culture inserts at the air–liquid interface (ALI). (B) Handling of the 3D corneal tissue model grown in cell culture inserts at ALI. For drug permeability studies, culture medium is replaced by receptor medium (assay medium or Krebs-Ringer buffer).

To initiate culture of multilayered 3D tissue constructs, primary corneal epithelial cells were seeded in cell culture inserts on FNC-coated microporous membranes (surface area 0.6 cm², pore size 0.4 μm, Figs. 2A, 2B; part no. MILCEL-MTK, MatTek Corporation).^25,27,41,42 3D corneal constructs were grown using proprietary serum-free medium at standard culture conditions (SCC, 37 ± 1°C, 95 ± 3% relative humidity, and 5 ± 0.5% [vol/vol] CO₂) and were maintained under submerged conditions with medium contacting both the apical and basolateral sides of the tissue until the cells reached confluence. At confluence, the medium was removed from the apical side and the cells were exposed to an air–liquid interface (Fig. 2A) to stimulate tissue maturation and to form the organotypic 3D corneal tissue model. Under these culture conditions, the 3D corneal tissues mature to form an appropriate tissue barrier on day 10 (Figs. 3A, 4).⁴³ All cells used for tissue production were screened for infection with human immunodeficiency virus (HIV), hepatitis B and C, yeast, fungi, bacteria, and mycoplasma. 

Figure 3

 

Morphology of the 3D corneal epithelial tissue. (A, B) H&E-stained cross section of 3D corneal tissue model (A) and human corneal tissue (B). Morphologic structure of the in vitro tissues closely parallels that of native human corneal epithelium. (C, D) Immunohistochemical analysis of cross sections of 3D corneal tissues. Cytokeratins CK3/12 (red) and nuclear stain DAPI (blue) (×10 [C] and ×40 [D] objectives).

Figure 4

 

Transepithelial electrical resistance (TEER) of 3D corneal tissue model. Average TEER value of 18 different tissue lots. TEER assessment was performed at different days in culture (days 5–10) and after shipping and overnight equilibration (days 11–15). For each time point, TEER measurements were made from at least five different tissues.

Transepithelial Electrical Resistance Measurements

Transepithelial electrical resistance (TEER) was measured using an epithelial volt-ohm meter EVOM and the EndOhm-12 chamber (World Precision, Sarasota, FL, USA). TEER was used as an indicator of the tissue barrier integrity. Decreases in TEER are a sensitive measure of ocular barrier disruption.⁴⁴ 3D corneal tissues with TEER values > 600 Ω·cm² were used for drug transport studies and tissue characterization (Fig. 4).^42,44–46 

Light Microscopy and Hematoxylin and Eosin Staining

3D corneal tissues were harvested on day 10, rinsed in PBS, fixed in 10% neutral buffered formalin for 1 hour, dehydrated in a graded series of ethanol, and embedded in paraffin. Five-micrometer-thick hematoxylin and eosin (H&E)-stained cross sections were prepared using standard histologic techniques⁴⁷ and examined with a light microscope (Olympus BX41; Center Valley, PA, USA) or processed using an Olympus VS120 slide scanner at ×10 objective with a ×0.5 camera adjustment magnification. Human donor corneas were cut into wedges, fixed in 10% neutral buffered formalin, and processed in an identical manner. 

Immunofluorescent Staining

3D corneal tissue constructs were harvested on day 10, rinsed in PBS, and fixed in 10% neutral buffered formalin for 20 minutes at 4°C. Next, whole tissue constructs or deparaffinized tissue sections (see above) were permeabilized with 0.1% Triton X-100 for 30 minutes, blocked in 10% normal goat serum (Gibco, Gaithersburg, MD, USA) for 30 minutes, incubated with primary antibody for 60 minutes, and then with fluorochrome-conjugated secondary antibody for 60 minutes. All steps were performed at room temperature in a humidified chamber. Between each step, the samples were washed twice with PBS containing 0.1% Tween 20. The samples were analyzed by confocal microscopy using an Olympus FV1000 with FluoView imaging software. The following antibodies were used: tight junction sampler pack (ZO-1 [1:20], occludin [1:20], and claudin-1 [1:20]), MUC1 [1:50], and Texas Red (ThermoFisher Scientific, Waltham, MA, USA); cytokeratins CK3/12 [1:50], CK15 [1:50], Alexa Fluor 488 [1:400], and Alexa Fluor 555 [1:400] (Abcam, Cambridge, MA, USA); and DAPI [1.43 μM] (Sigma-Aldrich, Burlington, MA, USA). 

Transmission Electron Microscopy

3D corneal tissues were harvested on day 10, fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4), and postfixed in 2% OsO₄, per standard transmission electron microscopy (TEM) protocols.⁴⁸ Sections were examined and images digitally recorded with a JEOL 1200 EX electron microscope (Peabody, MA, USA) at the Harvard Medical School TEM facility (Cambridge, MA, USA). 

MTT Tissue Viability Assay

MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, Sigma) is a yellow tetrazole that is reduced to purple formazan only by the mitochondria of metabolically active, living cells, and it has been shown to be a robust marker of tissue viability.⁴⁹ Tissue viability was analyzed by colorimetric MTT assay after 3D corneal tissues were exposed to model compounds, ophthalmic drugs, or benzalkonium chloride (BAC)-containing solutions. Following exposure to model compounds or drug formulations, the culture inserts containing 3D corneal tissues were removed from the receiver solution, rinsed with PBS, and transferred into a 24-well plate containing 0.3 mL MTT solution (1 mg/mL; Sigma-Aldrich) in Dulbecco's modified Eagle's medium (DMEM, Gibco). 3D corneal tissues were exposed to MTT medium basolaterally for 3 hours at SCC that allowed the metabolically active cells to metabolize the MTT dye and to produce the purple formazan. The MTT assay was terminated by transferring the tissues into fresh 24-well plates containing 2 mL isopropanol. Purple formazan crystals were extracted at 4°C overnight, and analyzed spectrophotometrically at 575 nm on a plate reader (Molecular Devices, Palo Alto, CA, USA). Tissue viability was determined using the following equation:  

\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\unicode[Times]{x1D6C2}}\)\(\def\bupbeta{\unicode[Times]{x1D6C3}}\)\(\def\bupgamma{\unicode[Times]{x1D6C4}}\)\(\def\bupdelta{\unicode[Times]{x1D6C5}}\)\(\def\bupepsilon{\unicode[Times]{x1D6C6}}\)\(\def\bupvarepsilon{\unicode[Times]{x1D6DC}}\)\(\def\bupzeta{\unicode[Times]{x1D6C7}}\)\(\def\bupeta{\unicode[Times]{x1D6C8}}\)\(\def\buptheta{\unicode[Times]{x1D6C9}}\)\(\def\bupiota{\unicode[Times]{x1D6CA}}\)\(\def\bupkappa{\unicode[Times]{x1D6CB}}\)\(\def\buplambda{\unicode[Times]{x1D6CC}}\)\(\def\bupmu{\unicode[Times]{x1D6CD}}\)\(\def\bupnu{\unicode[Times]{x1D6CE}}\)\(\def\bupxi{\unicode[Times]{x1D6CF}}\)\(\def\bupomicron{\unicode[Times]{x1D6D0}}\)\(\def\buppi{\unicode[Times]{x1D6D1}}\)\(\def\buprho{\unicode[Times]{x1D6D2}}\)\(\def\bupsigma{\unicode[Times]{x1D6D4}}\)\(\def\buptau{\unicode[Times]{x1D6D5}}\)\(\def\bupupsilon{\unicode[Times]{x1D6D6}}\)\(\def\bupphi{\unicode[Times]{x1D6D7}}\)\(\def\bupchi{\unicode[Times]{x1D6D8}}\)\(\def\buppsy{\unicode[Times]{x1D6D9}}\)\(\def\bupomega{\unicode[Times]{x1D6DA}}\)\(\def\bupvartheta{\unicode[Times]{x1D6DD}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bUpsilon{\bf{\Upsilon}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\(\def\iGamma{\unicode[Times]{x1D6E4}}\)\(\def\iDelta{\unicode[Times]{x1D6E5}}\)\(\def\iTheta{\unicode[Times]{x1D6E9}}\)\(\def\iLambda{\unicode[Times]{x1D6EC}}\)\(\def\iXi{\unicode[Times]{x1D6EF}}\)\(\def\iPi{\unicode[Times]{x1D6F1}}\)\(\def\iSigma{\unicode[Times]{x1D6F4}}\)\(\def\iUpsilon{\unicode[Times]{x1D6F6}}\)\(\def\iPhi{\unicode[Times]{x1D6F7}}\)\(\def\iPsi{\unicode[Times]{x1D6F9}}\)\(\def\iOmega{\unicode[Times]{x1D6FA}}\)\(\def\biGamma{\unicode[Times]{x1D71E}}\)\(\def\biDelta{\unicode[Times]{x1D71F}}\)\(\def\biTheta{\unicode[Times]{x1D723}}\)\(\def\biLambda{\unicode[Times]{x1D726}}\)\(\def\biXi{\unicode[Times]{x1D729}}\)\(\def\biPi{\unicode[Times]{x1D72B}}\)\(\def\biSigma{\unicode[Times]{x1D72E}}\)\(\def\biUpsilon{\unicode[Times]{x1D730}}\)\(\def\biPhi{\unicode[Times]{x1D731}}\)\(\def\biPsi{\unicode[Times]{x1D733}}\)\(\def\biOmega{\unicode[Times]{x1D734}}\)\begin{equation}\tag{1}\rm \% Viability = \rm A_{sample} / \rm A_{NC} \times 100 , \end{equation}

where A_sample is the absorbance of the test solution of the treated tissues and A_NC is the absorbance of the solution of the tissues treated with placebo formulations (negative control). The MTT assay is routinely performed on various MatTek Corporation tissue models to determine tissue viability for quality control purposes and after experimental treatments.²⁹ 3D corneal tissues with viability above 1.0 optical density (OD) were used for tissue characterization and transport studies.  

Separation of Corneal Epithelium From Corneoscleral Button and RNA Extraction

Donor corneoscleral buttons were obtained from NDRI, rinsed in PBS, and excised using an 8-mm biopsy punch. Corneal buttons were cut into two sections and incubated with 2.5 U/mL Dispase II solution (StemCell Technologies, Cambridge, MA, USA) in DMEM (high glucose, Gibco) at 4°C overnight on a shaker. Next, corneal epithelium was carefully peeled with fine forceps, rinsed in three changes of PBS, and collected by centrifugation (1000g for 5 minutes). The pellet containing the corneal epithelial layer was then immediately homogenized in RNA lysis buffer, and total RNA was isolated using a RNAqueous kit according to the manufacturer's protocol (ThermoFisher Scientific). 

RNA Extraction From 3D Corneal Tissues and Real-Time Quantitative Polymerase Chain Reactions (qPCR)

3D corneal tissue constructs were harvested on day 11, and two 3D corneal tissues were homogenized together in RNA lysis buffer as above. Purity and concentration of the extracted total RNA were determined, based on the absorbance at 260 and 280 nm using the Experion system (BioRad, Hercules, CA, USA). Total RNA (1 μg) isolated from the 3D corneal tissues or from excised human corneal epithelium was used for synthesis of cDNA using the RT² First Strand Kit (Qiagen, Germantown, MD, USA). 

RT² Profiler Human Drug Metabolism (PAHS-002Z) and Human Drug Transporters (PAHS-070ZD) gene arrays were obtained from SABiosciences (Qiagen). Gene analysis was performed using the indicated RT² Profiler PCR arrays with a BioRad CFX96 PCR system according to the manufacturer's instructions. Eighty-four genes involved in drug metabolism and 84 genes involved in drug transport were analyzed. The expression was normalized to the value of GAPDH mRNA expression of the same sample. Average threshold cycle C_t, relative ΔC_t (ΔC_t = C_{t gene of interest} − C_{t GAPDH}), and ΔC_t expression (2^−ΔCt) were determined.^50,51 Based on the ΔC_t expression (2^−ΔCt) all genes were assigned into the following groups: (1) 1.0 ≤ 2^−ΔCt: very strong expression; (2) 0.1 ≤ 2^−ΔCt < 1.0: strong expression; (3) 0.01 ≤ 2^−ΔCt < 0.1: moderate expression; (4) 0.001 ≤ 2^−ΔCt < 0.01: low expression; and (5) 0.0001 ≤ 2^−ΔCt < 0.001: very low expression. If Ct > 35, gene expression was considered as not detected. 

The fold difference in expression of a target gene between the excised corneal epithelium and the 3D corneal tissues was determined using the following equation: Fold difference = 2^−Δ(ΔCt), where Δ(ΔCt) = ΔCt(3D corneal tissue) − ΔCt(corneal epithelium). 

Lucifer Yellow Assay

Tissue barrier integrity was analyzed using the Lucifer yellow (LY) leakage assay. Upon completion of the drug permeability experiments (see below), the inserts containing 3D corneal tissues were rinsed with Krebs-Ringer buffer (KRB, pH7.4; ZenBio, Research Triangle Park, NC, USA), blotted, and moved into a new 24-well plate prefilled with 300 μL KRB solution in each well. Then 300 μl 300 μM LY CH dilithium salt (Sigma) was applied topically to the tissues and tissues were incubated at SCC for 60 minutes. At the end of the incubation period, 100 μL of the receptor medium was transferred into a 96-well plate and the absorbance was analyzed on spectrophotometer (SpectraMax Plus, Molecular Devices) at 450/528-nm excitation/emission wavelengths. % Permeability was calculated using the following equation:  

\begin{equation}\tag{ 2}\rm \%\ Permeability = \rm (A_{test} - \rm A_{blank}) / \rm (A_{LY\ Don} - \rm A_{blank})\times 100 , \end{equation}

where: A_test is the absorbance of the test solution, A_blank is the absorbance of the blank solution, A_{LY Don} is the absorbance of the initial LY donor solution.  

Permeability Analysis of Reference Compounds

The following model compounds with different physicochemical properties were used in transport experiments: sodium fluorescein (Na-FL) and LY, hydrophilic markers with low permeability; rhodamine B (RhB), a lipophilic marker with high permeability; fluorescein isothiocyanate-dextran with molecular weight 4000 (FD-4), a high molecular weight compound with low permeability. In addition, we examined permeability of the common antibiotic ofloxacin (OFL, a synthetic antibiotic of the fluoroquinolone drug class) with relatively high corneal permeability.⁵² 

The permeability assays were performed in the cell culture inserts positioned inside wells of the 24-well plate (Fig. 2). Briefly, the inserts containing day 11 corneal tissues were transferred into 24-well plates containing 300 μL prewarmed KRB and equilibrated for 15 minutes. Next, 400 μL donor sample formulation was topically applied to the surface of three replicate tissues (Fig. 2). Tissues were incubated at SCC on a horizontal plate shaker and the receiver solution was collected at 30-minute intervals.²⁷ After 30 minutes and every 30 minutes thereafter, the tissues were moved into a new well containing 300 μL KRB and the plate was returned to the incubator. The tissues were moved after 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 hours of permeation time. All materials and ophthalmic solutions were prepared in KRB. The following parameters were used during tissue permeability analysis: permeation area: 0.6 cm²; donor volume: 400 μL; receptor volume: 300 μL; exposure time: up to 180 minutes. The concentration of the reference compound in the receiver solution was determined using a fluorescent plate reader (Synergy HT; BioTek, Winooski, VT, USA) or by UV absorbance. Each compound was tested in at least three different lots of 3D corneal tissues. At least three replicate tissues were used per compound. 

Plots of drug flux versus time were constructed, and the steady-state flux J [μg/(cm²·s)] was determined from the linear portion of the curve: J = dQ/dt × A. From the flux values, the apparent permeability coefficients (P_app) were calculated according to the equation P_app = (dQ/dt)/(C₀xA) where dQ/dt is steady-state flux (nmol/s), C₀ = initial donor solution concentration added topically, and A is the exposed surface area of the tissue (0.6 cm²). 

Permeation of Latanoprost and Bimatoprost Eye Drops

Three formulations of latanoprost-containing eye drops (50 μg/mL) were tested: (1) latanoprost eye drops (0.005%, Alcon, Fort Worth, TX, USA) in the commercial vehicle containing 0.02% BAC as a preservative; (2) latanoprost (0.005%, CAS# 130209-82-4, Tocris Bio-Techne, Minneapolis, MN, USA) in KRB (pH 7.4); and (3) latanoprost free acid (0.005%, CAS# 41639-83-2; Sigma-Aldrich Corp., St. Louis, MO, USA) in KRB. Three formulations of bimatoprost-containing eye drops (100 μg/mL) were also tested in parallel: (1) Lumigan eye drops (0.01%; Allergan, Irvine, CA, USA) in the commercial vehicle containing 0.02% BAC as a preservative; (2) bimatoprost (0.01%, CAS# 155206-00-1; Sigma-Aldrich Corp.) in KRB; and (3) bimatoprost free acid (0.01%, CAS# 38344-08-0; Cayman Chemical, Ann Arbor, MI, USA) in KRB. Primary stock solutions of latanoprost, latanoprost free acid, and bimatoprost were prepared in dimethyl sulfoxide (DMSO) (Sigma-Aldrich Corp.); the final concentration of DMSO was 0.5%. In addition, 0.5% DMSO was also added to KRB solution to serve as vehicle control (VC). 

For permeation experiments, the inserts with day 11 reconstructed corneal tissues were transferred into the wells of a 24-well plate containing 300 μL prewarmed KRB (receptor medium). Then 100 μL of the eye drop formulations was applied topically to the surface of three replicate tissues and incubated at SCC (37°C, 95% relative humidity, and 5% CO₂). Plates were kept on a horizontal plate shaker and the receiver solution was collected at 30-minute intervals. Each 30 minutes the tissues were moved into a new well containing 300 μL KRB, and the plate was returned to the incubator. Specifically, the tissues were moved after 0.5, 1.0, 1.5, and 2.0 hours of permeation time. To serve as VC, 100 μL vehicle solution (KRB containing 0.5% DMSO) was applied to the tissue surface of three replicate tissues. An additional set of three replicate tissues was left untreated and used as a negative control (NC) for tissue viability assays. NC was handled in parallel with tissues receiving treatment. Receiver solution was collected and analyzed using commercially available ELISA kits for latanoprost and 17-phenyl trinor prostaglandin F2α, according to the manufacturer's instructions (Cayman Chemical). Tissue barrier integrity was analyzed using TEER and the LY assay, and tissue viability was assessed using the MTT assay. 

The amount of latanoprost and bimatoprost free acid that permeated into the receptor fluid was calculated using the following equation: % Permeated = (A_test/C₀) × 100, where: A_test = amount of the free acid (pg/mL) in the receptor compartment, C₀ = initial donor solution concentration added topically. The steady-state period was determined from the linear portion (r² > 0.99) of the cumulative amount permeated versus time curve. The slope of the steady-state period represents the steady-state flux J [nmol/(cm²·h)]. The apparent permeability (Papp) was calculated by the following equation: Papp (cm/s) = J/(3600XC₀) where: J is the steady-state flux [nmol/(cm²·h)] and C₀ is the initial donor concentration (nmol/mL). 

Statistical Analysis

Tissue permeability was evaluated with regression data analysis and analysis of variance (ANOVA). Results for tissue permeability (Papp), barrier integrity (TEER and LY leakage), and viability (MTT assay) were expressed as the mean ± standard deviation. Statistical analysis was performed using a Student's paired t-test. Simple linear regression (r²) was used to determine the correlation coefficient for gene expression between the 3D corneal tissues versus excised human corneal epithelium and likewise for model drug permeation coefficients in 3D corneal tissues versus published permeability data obtained with excised animal corneas. 

Results

3D Corneal Tissue Morphology, Ultrastructure, and Protein Expression

The 3D corneal tissues exhibit in vivo–like morphology. Cross sections of the 3D corneal tissue model and a near limbal region of a normal human cornea are shown in Figures 3A and 3B. The thickness (50–70 μm) and tissue structure of the 3D corneal tissues closely parallel the morphologic features of human corneal epithelium: a single layer of columnar basal cells, two or three layers of polygonal wing cells, and two or three layers of squamous flattened cells with flattened nuclei. As expected, corneal epithelial cells in the 3D reconstructed tissue model expressed corneal epithelial cytokeratins CK3/12 (Figs. 3C, 3D), and formed tight intercellular junctions (Fig. 5) resulting in a tissue barrier similar to in vivo corneal tissues (Fig. 4).¹³ 

Figure 5

 

Transverse images of the 3D corneal tissue model—ultrastructure (transmission electron microscopy). (A, B, D) Apical cell layers of the 3D corneal epithelial tissue model. Open arrows point to microvilli on the surface of the tissue. Tight junctions (white arrows) and desmosomes (closed arrow) are observed in the apical layers. (C) Near the surface of the 3D corneal tissue; squamous cells. Glycogen granules (circle) and desmosomes (closed arrow) are observed. (E) Section going through almost full thickness of the 3D corneal tissue model; 𝔹, basal cells; 𝕎, wing cells; 𝕊, squamous cells. (A, B, D) 12,000× magnification, (C) ×2500 magnification, (E) ×30,000 magnification. n, nucleus.

Transmission electron microscopy confirmed an appropriate structure of corneal epithelial layer: The most apical cell layers were flat and contained microvilli, desmosomes, and tight junctions (Figs. 5A–C). The presence of tight junctions is imperative for tight barrier formation. The cytoplasm of apical cells contains flattened nuclei and many tonofilaments; glycogen granules were also observed (Figs. 5C, 5D). Wing cells have a polygonal shape and their cytoplasm contains rough endoplasmic reticulum cisternae, mitochondria, and Golgi's complexes. Large numbers of cytoskeletal tonofilaments and numerous interdigitations, as well as desmosomal junctions, were observed in wing cells (Figs. 5C, 5E). 

Immunohistochemical analysis confirmed expression of tight junction proteins ZO-1, occludin, and claudin-1 in the apical layers of the 3D corneal tissue construct (Figs. 6A–C).⁵³ ZO-1 and occludin staining was observed as a continuous pattern at the cell–cell borders, claudin-1 distribution appeared more localized and patchy. While the staining of claudin is more localized, it is likely more continuous than it appears since the 3D tissue constructs are not entirely flat when mounted as whole tissue on the microscope slide, thus preventing large areas from being in focus at the same time. Also, since the 3D tissue is a multilayer structure, tight junction proteins and cellular nuclei observed on the images are not in the same focal plane. The membrane-associated glycoprotein mucin 1 (MUC-1) was expressed on the surface of the 3D corneal tissues (Fig. 6D). 

Figure 6

 

Expression of tight junction and mucosal proteins in 3D corneal tissue model. Topical view of 3D corneal tissues, confocal microscopy. (A–C) Immunofluorescent staining of tight junction proteins. (A) ZO-1 (green) and nuclear stain DAPI (blue); (B) occludin (green); (C) claudin-1 (green) and DAPI (blue); (D) MUC-1 (red). (A, D) 200× magnification and (B, C) 100× magnification.

3D Corneal Tissue Transepithelial Electrical Resistance

When cultured at the air–liquid interface, the 3D corneal tissues developed an in vivo–like tissue barrier that was confirmed by measuring TEER (Fig. 4). Not fully mature organotypic 3D corneal tissues do not possess a mature barrier and display TEER ≤ 200 Ω·cm², but by day 10 of the culture period, TEER values reach 877 ± 200 Ω·cm². This coincides with flattening of the apical layers and formation of tight junctions (Figs. 3, 5, 6), and as a result, developing an in vivo–like diffusion barrier for drug absorption.¹³ When the 3D corneal tissues are cultured in COR-100-MM (MatTek Corporation) at SCC for up to 15 days, the barrier properties are maintained (1311 ± 215 Ω·cm², Fig. 4). Drug permeability experiments using the 3D corneal tissues were performed on day 11 when TEER measurements averaged 1000 ± 146 Ω·cm². Similar values of TEER have been reported in live rabbits and in human studies using intraocular electrodes.^43–45 

Expression of Drug-Metabolizing Enzymes

Since drug absorption can be significantly affected by drug-metabolizing enzymes, the expression of 84 phase I and phase II metabolizing enzymes, including cytochrome P450 (CYP450) superfamily, carboxylesterases, dehydrogenases, lipoxygenases, hydrolases, kinases, oxidoreductases, transferases, P-glycoproteins, and others in the 3D corneal tissues and human corneal epithelium was analyzed (Figs. 7, 8).³⁴ High correlation (r² = 0.87) between gene expression of the 3D corneal tissues and excised human corneal epithelium was observed (Fig. 7). 

Figure 7

 

Evaluation of phase I and phase II drug-metabolizing enzymes (A) and drug transporter (B) gene expression in the 3D corneal tissue model and human corneal epithelium. The scatter plots compare gene expression levels, log10 (2^−ΔCt), between 3D corneal and human corneal epithelium. High correlation of gene expression of drug-metabolizing enzymes (A) and drug transporters (B) was obtained for 3D corneal tissue model and the human corneal epithelium (R² = 0.87).

Figure 8

 

Gene expression of phase I and phase II drug-metabolizing enzymes in the 3D corneal tissue model and human corneal epithelium. ΔC_t (ΔC_t = C_{t gene of interest} − C_{t GAPDH}) and ΔC_t expression (2^−ΔCt) were determed. (A) Genes with very strong and strong expression (0.1 ≤ 2^−ΔCt). (B) Genes with moderate and low expression (0.001 ≤ 2^−ΔCt < 0.1); //, outside the range. See Table 1 for gene expression comments.

Out of 84 genes analyzed, 14 genes were below the limit of detection in the 3D corneal tissues and 16 genes were below the limit of detection in the excised human corneal epithelium (Table 1). Nine genes (10.7%) had ≥4-fold and eight genes (9.5%) had ≤4-fold lower expression in 3D corneal tissues when compared to excised human corneal epithelium. The 4-fold cutoff was chosen to obtain a meaningful data interpretation in the situation of small sample size (single donor for each data set).⁵⁴ 

Table 1

 

Gene Expression of Phase I and Phase II Drug-Metabolizing Enzymes in the 3D Corneal Tissue Model and Human Corneal Epithelium (Gene Array #PAHS-002Z)

 

In addition to aldehyde dehydrogenases, strong expression was detected for other phase II detoxifying enzymes involved in protection from oxidative stress: glutathione S-transferase, glutathione peroxidase, phospholipid hydroperoxidase, metallothionein 2A, pyruvate kinase, and catechol-O-methyltransferase (Table 1; Fig. 8A).⁵⁵ Out of all the genes analyzed, only three genes (3.6%) had more than 10-fold difference in expression compared to excised corneal epithelium. Glutathione S-transferases (GSTP1 and GSTM3), known to be involved in reactive oxygen species (ROS) detoxification, were significantly upregulated in 3D corneal tissues compared to excised corneal epithelium. In addition, CYP450 2J2 (CYP2J2), an oxygenase that catalyzes many reactions involved in the metabolism of drugs and other xenobiotics, was upregulated more than 12-fold in 3D corneal tissues (Table 1). 

Aryl hydrocarbon receptor (AHR), a ligand-activated transcription factor involved in the regulation of biological responses to aromatic hydrocarbons and regulation of xenobiotic-metabolizing enzymes such as CYP450, had strong expression (0.1 ≤ 2^−ΔCt < 1.0) in both excised corneal epithelium and the 3D corneal tissue model (Table 1; Fig. 8A). Moderate to strong expression (0.01 ≤ 2^−ΔCt < 1.0) was detected for genes responsible for lipid metabolism and genes involved in cellular defense against toxic, carcinogenic, and pharmacologically active electrophilic compounds: cytochrome b5 reductase 3 (CYB5R3), carboxylesterase 2 (CES2), microsomal glutathione S-transferase 2 (MGST2), biliverdin reductase B (BLVRB), and aminolevulinate dehydratase (ALAD). Cytochrome P450 2C19 (CYP2C19), an enzyme involved in the metabolism of xenobiotics and known to metabolize many drugs, had a moderate level of expression (0.01 ≤ 2^−ΔCt < 0.1) in both excised corneal epithelium and 3D corneal tissues (Table 1; Fig. 8). 

The expression of various additional CYP450 isoenzymes was also investigated. Our results indicated that CYP2D6 was moderately expressed in excised corneal epithelium and had low expression (0.001 ≤ 2^−ΔCt < 0.01) in 3D corneal tissues; CYP1A1, CYP2C8, CYP3A5, CYP2E1, and CYP2C9 had low expression (0.001 ≤ 2^−ΔCt < 0.01) in both 3D corneal tissues and excised corneal epithelium; CYP2J2 had very low expression (0.0001 ≤ 2^−ΔCt < 0.001) in corneal epithelium and moderate expression in 3D corneal tissues. Isoenzymes CYP11B2, CYP17A1, CYP19A1, CYP2B6, CYP2F1, and CYP3A4 were not detected in excised corneal epithelium and likewise were not detected or had very low expression in the 3D corneal tissues (Table 1). 

Fatty acid amide hydrolase (FAAH), an enzyme with both amidase and esterase activity that may be responsible for conversion of prostaglandin prodrugs in ocular tissue, had moderate and low level of gene expression in corneal epithelium and 3D tissues.⁵⁶ Also, UGT1A1, one of the esterases responsible for conversion of ester prodrug latanoprost,^56,57 was detected at very low/low levels in corneal epithelium and 3D corneal tissues (Table 1). 

Expression of Drug Transporters

Since drug absorption can be significantly affected by various drug transporter proteins, expression of 84 key drug transporters, including the main transporter classes present in corneal tissue, ABC, and solute carrier (SLC) transporters, as well as transporters with a lesser known role in the cornea, vacuolar H⁺-ATPases, copper pumps, aquaporins, voltage-dependent anion channels, and major vault protein, were evaluated in the 3D corneal tissue model in comparison to their expression in the excised human corneal epithelium (Figs. 7B, 9; Table 2).^14,34,58 A high correlation coefficient of r² = 0.87 was obtained between transporter gene expression of 3D corneal tissues and that of excised corneal epithelium (Fig. 7B). 

Figure 9

 

Gene expression of drug transporters in human corneal epithelium and 3D corneal tissue model. ΔC_t (ΔC_t = C_{t gene of interest} − C_{t GAPDH}) and ΔC_t expression (2^−ΔCt) were determed. (A) Genes with strong and moderate expression (0.01 ≤ 2^−ΔCt). (B) Genes with low expression (0.001 ≤ 2^−ΔCt < 0.01); see Table 2 for gene expression comments.

Table 2

 

Gene Expression of Drug Transporters in 3D Corneal Tissue Model and Human Corneal Epithelium (Gene Array #PAHS-070Z)

 

Of the 84 genes analyzed, 17 and 22 genes were below the limit of detection in the 3D corneal tissues and excised corneal epithelium, respectively (Table 2). Seven genes (8.3%) had ≥4-fold, and 2 genes (2.4%) had ≤4-fold expression in 3D corneal tissues when compared to excised corneal epithelium. Thus, expression in the 3D corneal tissue is remarkably similar to that of native corneal tissue. 

From the ABC transporter superfamily, the MRP subfamily is known to confer drug resistance when overexpressed in cancer tissues, and can have significant impact on drug transport in the cornea.^59,60 Excised corneal epithelium and 3D corneal tissues had moderate (0.01 ≤ 2^−ΔCt < 0.1) expression of ABCC1, ABCB6, and ABCC3 transporter genes, and low expression (0.001 ≤ 2^−ΔCt < 0.01) of ABCC3, ABCC5 transporter genes; the expression level of ABCB1 (P-gp) was very low (0.0001 ≤ 2^−ΔCt < 0.001), and ABCG2 (BCRP) was below detection level for both excised corneal epithelium and 3D corneal tissues (Table 2; Fig. 9). 

A number of SLC family members, including SLC38A2 and SLC3A2, and voltage-dependent anion channels, VDAC1 and VDAC2, had strong gene expression (0.1 ≤ 2^−ΔCt < 1.0) in both the 3D corneal tissue and the excised human corneal epithelium. Moderate gene expression (0.01 ≤ 2^−ΔCt < 0.1) was observed for other uptake transporters (SLCO3A1, SLC19A2, and SLC7A5) and multi-subunit vacuolar ATPase (ATP6V0C). Solute carrier family members, including SLC31A1, SLC7A6, SLC25A13, SLC22A1 (commonly known as OCT1), and SLCO2B1 and SLC15A1 (commonly known as PEPT1) had low/very low gene expression (0.0001 ≤ 2^−ΔCt < 0.01). Many SLC family members, including organic cation transporters SLC22A2 (OCT2), SLC22A6 (OAT1), SLC22A7 (OAT2), and SLC22A8 (OAT3), and anion transporters SLCO1A2 (OATP1A2), SLCO1B1 (OATP1B1), and SLCO1B3 (OATP1B3), known to be expressed in liver and other metabolically active tissues,¹⁴ were not detected in excised human corneal epithelium or 3D corneal tissues (Table 2). 

Permeation of Model Compounds

The lipophilic, high permeability marker RhB demonstrated high permeability, Papp = 3.5 ± 0.4 × 10⁻⁵ cm/s (Table 3). Similar values had been previously reported for other cell systems.^10,61 In contrast, the hydrophilic markers Na-Fl and LY displayed low permeability, Papp = 6.1 ± 1.1 × 10⁻⁷ and 6.8 ± 0.7 × 10⁻⁷, respectively, which was consistent with previous reports.^61,62 High molecular weight marker FD-4, as expected, demonstrated the lowest permeability, Papp = 5.7 ± 3.8 × 10⁻⁸, similar to in vivo.⁶¹ The permeability of the widely used topical antibiotic OFL (small, not charged molecule) was slightly lower than expected (Papp = 9.2 ± 2.1 × 10⁻⁶) from studies with excised rabbit corneas (Papp = 1.5 ± 0.1 × 10⁻⁵).^63,64 

Table 3

 

Permeation of Model Compounds Through 3D Corneal Tissue Model

 

Since in vivo and excised rabbit corneas are considered to be gold standards for ocular drug delivery analysis, correlation coefficients were determined for 3D corneal model versus published data (Table 3).^{10,52,61−66} The apparent permeability coefficients (P_app) obtained from the model drugs in the in vitro 3D corneal tissues were compared to P_app obtained in excised rabbit corneas. A high correlation coefficient of r² = 0.94 was achieved, confirming the physiological relevance of the data obtained from the in vitro 3D corneal model (Fig. 10).⁶⁷ 

Figure 10

 

Correlation of permeation coefficients (P_app) of 3D corneal tissue model and excised rabbit corneas. The 3D corneal tissue model displays a high correlation with excised rabbit corneas.

Permeability of Latanoprost and Bimatoprost Drug Families

All latanoprost and bimatoprost solutions reached a steady-state period in 0.5 hour and remained at steady state for the duration of the permeation reaction (2 hours) (Fig. 11). Steady-state flux reached by latanoprost in commercial formulation and latanoprost in KRB was 3.47 nmol/(cm²·h) and 2.04 nmol/(cm²·h), and Papp was 2.50 × 10⁻⁵ and 1.47 × 10⁻⁵, respectively. Steady-state flux for bimatoprost in commercial formulation (Lumigan) and bimatoprost in KRB was 18.33 nmol/(cm²·h) and 3.43 nmol/(cm²·h), and Papp was 5.58 × 10⁻⁵ and 1.04 × 10⁻⁵, respectively. Steady-state flux reached after application of free acid metabolites latanoprost free acid and bimatoprost free acid in KRB was 0.18 nmol/(cm²·h) and 0.53 nmol/(cm²·h), and Papp for both was 1.22 × 10⁻⁶ and 1.85 × 10⁻⁶ (Table 4). 

Figure 11

 

Latanoprost acid and bimatoprost acid permeation profiles in the 3D corneal tissue model. Permeation of latanoprost acid (A) and bimatoprost acid (B). Flux over time of latanoprost acid (C) and bimatoprost acid (D). Open circle and solid line, free acid in KRB; filled circle and solid line, commercial formulation; open circle and dashed line, formulation in KRB.

Table 4

 

Drug Permeability Following Application of Ophthalmic Formulations

 

At the end of the drug permeability experiments (2-hour incubation), changes in 3D corneal tissue barrier integrity and permeability were evaluated by measuring TEER and performing LY leakage assay. 3D corneal tissue viability was also evaluated using the MTT assay. Results were compared to the tissues treated with the VC (KRB) or untreated tissues (NC). Although the tissue barrier integrity was not significantly affected after 2-hour incubation with latanoprost and bimatoprost free acid formulations, incubation with the commercial formulations of latanoprost and Lumigan resulted in a statistically significant increase (2.9- and 1.8-fold) in tissue leakage when compared to the incubation with VC (Table 5). This also corresponded to a decline in TEER to 38.6% and to 55.0% in commercial latanoprost- and Lumigan-treated tissue, respectively (compared to NC). Based on MTT tissue viability assay, the 2-hour incubation with commercial formulations brought about a statistically significant reduction in 3D corneal tissue viability to 71.0% and 74.2% when compared to tissues dosed with VC. Two-hour incubation with free acid formulations caused a reduction in 3D corneal tissue viability to 95.1% (latanoprost free acid) and 70.8% (bimatoprost free acid) when compared to NC (Table 4). While bimatoprost in KRB did not have a significant effect on 3D corneal tissue barrier and viability, latanoprost in KRB significantly reduced TEER (down to 65.7%) and increased tissue leakage (2.6-fold). 

Table 5

 

Effects of the Latanoprost and Bimatoprost Ophthalmic Formulations on 3D Corneal Tissue Barrier Integrity (TEER), Permeability (LY leakage), and Viability (MTT)

 

To investigate how early the eye drop formulations can cause an effect on tissue barrier integrity, permeability, and viability, tests were also performed after 30-minute treatment. Thirty-minute treatment with commercial formulations caused similar results to those observed after 2-hour reaction, except that tissue viability after 30 minutes of exposure to Lumigan was not significantly affected. All other eye drop formulations did not have a significant effect after 30-minute incubation (Table 5). 

Since both commercial formulations used in the study contained 0.02% BAC as a preservative and since it was previously shown that BAC at 0.02% can have detrimental effects on corneal tissue in vivo,⁶⁸ we investigated the effect of different concentrations of BAC on the 3D corneal tissue. As seen in Table 6, BAC solution at concentrations 0.005% and 0.01% in KRB did not have a significant effect on tissue barrier integrity, permeability, or viability. However, BAC at 0.02% had a significant effect on TEER, and BAC at 0.04% had significant effects on barrier integrity, permeability, and viability (Table 6). 

Table 6

 

Effects of BAC Solutions on 3D Corneal Tissue Integrity (TEER), Permeability (LY Leakage), and Viability (MTT)

 

Discussion

The increased use of ophthalmic medications in recent years has led to an increased demand for absorption studies during the development and characterization of new ophthalmic substances and formulations.⁷ Analysis of topical ophthalmic drug absorption is one of the main steps in the identification of compounds with favorable pharmacokinetic properties since the physical and structural barriers of the cornea, which are vitally important for eye protection, present major challenges for delivery and bioavailability. Although in vivo animal studies can be used, anatomic and physiological differences make species extrapolation unreliable. In addition, the use of animals does not allow for high-throughput approaches, which are often keys to identifying new ophthalmics and to optimizing the final formulations. To this end, human cell-based tissue culture models can be employed to accelerate identification of new compounds by allowing multifactorial analysis. In addition to drug absorption, drug metabolism and biocompatibility with ocular tissues can be addressed on a large scale. Development of an in vitro organotypic normal human corneal tissue model that structurally and functionally resembles native corneal tissue is an important step in providing tools necessary for modern-day drug development.²⁸ 

The corneal epithelium, more precisely, the apical surface of the epithelium, has a major contribution to the overall barrier properties of the cornea and typically is the limiting factor relating to the permeability of ophthalmic formulations.^13,69 As reported herein, the barrier of the 3D corneal tissue model is well formed (Fig. 4) and closely parallels the structure of the human corneal epithelium (Fig. 3), as it contains tight junctions and desmosomes (Fig. 5), expresses tight junction proteins in the apical cell layers (Fig. 6), and develops TEER above 1000 Ω·cm².⁴³ Although there is not a lot of information regarding TEER in normal human corneal epithelium, recently published TEER levels of 750 ± 111 Ω·cm² and 690 ± 69 Ω·cm² for rabbits and humans, respectively, were reported when measured using a novel noninvasive technique.^45,46 Similarly, several earlier reports presented slightly higher TEER values (1122.3 ± 61.3 Ω·cm²) for excised rabbit cornea,²² which are approximated by the 3D tissue model. 

Proper expression of surface mucins by the corneal epithelium is important not only for ocular defense against invading microbes, but also for accurate adhesion of microparticles and nanomaterials used in the modern ophthalmic drugs.⁷ We have compared gene expression of surface mucins (MUC1, MUC4, and MUC16) in the 3D corneal tissue model and excised human corneal epithelium and demonstrated that all mucins were similarly expressed in both reconstructed 3D corneal tissues and excised human corneal epithelium, with MUC16 having the highest level of expression and MUC4 the lowest (Kaluzhny Y, unpublished data, 2018). 

The ultimate effect and clinical impact of drug transporters and drug-metabolizing enzymes on topically delivered ophthalmics are not fully understood. Many contradictory reports can be found in the literature describing gene expression and function of the enzymes and transporters affecting drug permeation through corneal tissue, and most often this is due to the diverse array of methods that have been used.^{2,14,34,58,70} To this end, it is important to utilize the most appropriate currently available prediction models to study eye-related bioavailability.⁴³ The 3D corneal reconstructed tissues should be able to correctly predict drug bioavailability in the target tissue since gene expression of key phase I and II drug-metabolizing enzymes (Figs. 7, 8; Table 1) and the most important efflux transporters (Figs. 7, 9; Table 2) had very high correlation to excised human corneal epithelium, r² = 0.87. As expected, excised corneal epithelium and 3D corneal tissues exhibit robust expression of detoxifying enzymes, enzymes involved in the protection from oxidative stress, and enzymes that are involved in lipid metabolism (Fig. 8; Table 1). Most CYP450 isoenzymes were expressed at low to very low levels and many were not detected at all, similar to findings in previously published studies.^14,34 Although the information about expression and function of these enzymes and transporters in ocular tissue is limited, based on our and several previously published studies,^14,34 it is evident that corneal epithelial cells have much lower expression of metabolizing enzymes when compared to principal sites of drug metabolism (e.g., liver, kidney, small intestine). 

Out of 84 genes tested in the drug-metabolizing enzyme gene array, only three enzymes had more than 10-fold difference in gene expression between 3D corneal tissues and excised corneal epithelium (Table 1). One gene that exhibited significantly higher expression in the 3D corneal tissue model was CYP450 2J2 (CYP2J2), an oxygenase involved in the metabolism of many drugs and other xenobiotics. The two other genes, GSTP1 and GSTM3, are glutathione transferases that play an important role in the detoxification of carcinogens, therapeutic drugs, environmental toxins, and products of oxidative stress.⁵⁵ It is not clear if this deviation is an accurate representation of modified gene expression since its expression in excised corneal epithelium was detected at very low/low levels that may skew results reported as fold difference. Additional studies are needed to determine if these discrepancies are due to the donor variability or a result of modified gene expression in in vitro culture. 

Out of 84 genes tested in the human drug transporter gene array, only 5 had a greater than 10-fold difference in gene expression between 3D corneal tissues and excised corneal epithelium (Table 2). Two genes, ATPase ATP7A and solute carrier SLCO2A1 were downregulated and solute carriers SLC16A1, SLC28A3, and SLCO4A1 were upregulated in 3D corneal tissues. As previously stated, these deviations may be a result of low levels of detection (0.0001 ≤ 2^−ΔCt < 0.001), donor variability, or the result of cell expansion in the in vitro culture. 

It has been reported that corneal epithelium expresses MRP5 (ABCC5) at mRNA and protein levels and that ABCC5 was strongly expressed in the basal layer of human corneal epithelium and limbus.¹⁶ We have observed expression of ABCC5 mRNA in both excised corneal epithelium and 3D corneal tissues, although at low level in both (0.001 ≤ 2^−ΔCt < 0.01, Table 2). The physiological function of ABCC5 is unclear, although it has been shown to transport important second messengers, cAMP and cGMP, and could confer resistance to antiviral and anticancer compounds.⁷¹ In addition to ABCC5, the corneal epithelium and 3D corneal tissues exhibited moderate expression of ABCC1 (0.01 ≤ 2^−ΔCt < 0.1) and low expression of ABCC3 (0.001 ≤ 2^−ΔCt < 0.01), while BCRP (ABCG2) was not detected in either (Table 2). It was shown that many ocular drugs that are known to be used as topical eye medications (including antibiotics OFL and erythromycin, antifungal clotrimazole, and immunomodulator CsA) can interact with P-gp (ABCC1/MRP1).^16,19,72 Published results on ABC transporter gene expression in human and animal corneas, as well as in immortalized human corneal epithelial cell–based tissue models, are often conflicting.^16,31,66,73 Only a limited selection of ABC transporters is known to be functionally expressed in human cornea and overall gene expression is typically low (when compared to the function of liver enzymes).^14,34 Previous publications clearly indicate a species-dependent expression of the efflux transporters and clear evidence of erroneous expression in immortalized cell–based tissue models.^16,34 In contrast, gene expression of key transporters in 3D corneal tissues (Figs. 7, 9; Table 2) had a high correlation (r² = 0.87) with expression in excised corneal epithelium, indicating that 3D corneal tissues should be able to more reliably predict drug bioavailability. 

This work is the most extensive study to date investigating gene expression of key drug-metabolizing enzymes and drug transporters in excised human corneal epithelium in comparison to in vitro reconstructed corneal tissues. Similar findings were demonstrated only for selected genes in human or animal corneas, or in corneal tissue models reconstructed from a cell line.^14,16,34,73 Furthermore, it is not always clear if corneal epithelial layer or whole corneas were used for RNA isolation in previously reported studies; therefore it is not easy to interpret and compare results from different studies since stromal keratocytes and endothelial cells may contribute differently to the gene expression profile. More donors will be investigated in future studies to determine the effect of donor variability on expression of genes that can influence corneal drug availability. 

Aqueous solubility and lipophilicity are two major factors that govern the rate of drug penetration through the cornea. The fluoroquinolones are one of the main classes of antimicrobial drugs that are widely used for topical application in the eye due to their broad spectrum and high therapeutic efficiency. They are often used to treat bacterial keratitis and conjunctivitis, but clinical failures that have been described can be due to reduced bioavailability. In this study, we have demonstrated that 3D corneal tissue model can be used to predict formulations with improved bioavailability. The fluoroquinolone antibiotic tested (OFL, Papp = 9.2 ± 2.1 × 10⁻⁶) in 3D corneal tissues had permeability similar to the highly permeable lipophilic transcellular marker substance RhB (Papp = 3.5 ± 0.4 × 10⁻⁵). Moderately permeable materials have Papp in the range of × 10⁻⁶ to 10⁻⁷ cm/s² (Na-FL and LY), and poor permeable compounds have Papp in the range of × 10⁻⁸ cm/s² (FD-4) (Table 3). Only limited data are available on in vitro drug permeability of the human cornea; therefore we compared 3D corneal tissue model permeability to literature data reported for excised rabbit corneas. The morphology of human and rabbit corneas is similar in that both have tight apical layers that play a major role in the barrier function, but they also have many distinctions, including tissue thickness and distinct gene expression of key drug-metabolizing enzymes and efflux transporters. There is a debate about what animal models best represent drug transport in the human eye.⁹ The most commonly used in ex vivo models include rabbit, porcine, bovine, and often goat and sheep corneas, all of which are readily accessible. Rabbit and porcine corneas can be useful for ex vivo/in vivo correlations; however, there is contradicting literature on the permeability and a good correlation to human studies is often not achieved.^9,74 Since at the present time, rabbit corneas are still considered to be a gold standard for drug availability studies, we compared the data obtained with the 3D corneal tissue model to the available data obtained with rabbits. A high correlation coefficient of r² = 0.94 was achieved that confirmed the physiological relevance of the data obtained from the in vitro 3D corneal model (Fig. 10).⁶⁷ 

Prostaglandins are potent hypertensive and anti-inflammatory medications that are topically applied to the cornea, but their ocular bioavailability can present an obstacle. Penetration through the cornea can be enhanced by formulating lipophilic prodrugs, a common strategy in ocular drug development, including the development of the latanoprost drug family. Studies performed in vitro and in vivo have shown that the ester hydrolysis of latanoprost is rapid and results in generation of the biologically active latanoprost acid.¹⁷ Another example is bimatoprost, which acquires its lipophilic properties through the ethyl amide group that can be hydrolyzed to its free acid (17-phenyl-PGF2a), a potent prostaglandin F receptor agonist, by ocular tissues.⁷⁵ The 3D corneal tissues rapidly converted the isopropyl ester and ethyl amide prodrugs into pharmacologically active metabolites as free acid forms that were readily detectable in the receptor samples after latanoprost- and bimatoprost-containing eye drops were applied (Fig. 11). Our results confirmed that 3D corneal tissues possess the functional esterase and amidase activity necessary for metabolic conversion of ester and amide prodrugs, and can be utilized for in vitro evaluation of transcorneal permeability of prodrug candidates. 

The 3D corneal in vitro tissue model proved to be functional and highly reproducible, and replicated the in vivo permeation profile of different topical ocular formulations. Latanoprost (Alcon) and Lumigan (Allergan) commercial formulations exhibited the fastest corneal permeation rate, and latanoprost and bimatoprost free acids dissolved in KRB exhibited the lowest corneal permeation rate (Fig. 11; Table 4). Both preservative-free prodrug formulations of latanoprost and bimatoprost (dissolved in KRB) exhibited corneal permeation rates much higher than the free acid metabolites, although significantly slower than the commercial formulations. 

Commercial formulations containing 0.02% BAC resulted in the lowest tissue viability at the end of the 2-hour incubation period (Table 5). BAC is known to increase tissue permeability and drug bioavailability by affecting tight junction integrity but it can affect tissue viability.^44,46 We confirmed that a significant part of the topical drug formulation's effect on tissue barrier integrity and viability was due to BAC (Table 6). However, since the drug formulation's effects were more pronounced than those observed after treatment with 0.02% BAC in KRB, other components of the commercial vehicle, as well as the drug itself, may contribute to or exacerbate the detrimental effect on tissue viability observed for BAC. Since the drug donor concentration in in vitro experiments is kept constant, the effect of formulation composition (e.g., lipophilicity, solubility, molecular size and shape, presence of permeation enhancers), the role of active and passive transports, the biocompatibility, and the mechanism of action can be studied without interference of variable physiological forces.⁹ 

In summary, we have demonstrated that the 3D in vitro reconstructed human corneal tissue model possesses similar tissue structure, barrier properties, and expression of cornea-specific markers to the in vivo human cornea. The expression pattern of drug-metabolizing enzyme and transporter genes confirmed that the model is equipped with enzymes and transporters necessary for proper tissue performance. Most importantly, permeability of model compounds and ophthalmic formulations with different properties and excipients was similar to that of the intact cornea. The use of the 3D corneal tissue model for ophthalmic drug optimization will be advantageous in that it will avoid species extrapolation and provide information related to drug permeability, toxicity, metabolism, and effects on barrier integrity within the same experiment. In addition, we anticipate that the model will reduce the number of animals needed for ocular drug permeation studies, be more cost-effective than current animal methods, and facilitate preclinical optimization of formulations. 

Acknowledgments

The authors thank the Harvard Medical School TEM facility and Susumu Ito for help and expertise with transmission electron microscopy. 

Disclosure: Y. Kaluzhny, None; M.W. Kinuthia, None; T. Truong, None; A.M. Lapointe, None; P. Hayden, None; M. Klausner, None