Functional Human Corneal Equivalents Constructed from Cell Lines

See allHide authors and affiliations

Science  10 Dec 1999:
Vol. 286, Issue 5447, pp. 2169-2172
DOI: 10.1126/science.286.5447.2169


Human corneal equivalents comprising the three main layers of the cornea (epithelium, stroma, and endothelium) were constructed. Each cellular layer was fabricated from immortalized human corneal cells that were screened for use on the basis of morphological, biochemical, and electrophysiological similarity to their natural counterparts. The resulting corneal equivalents mimicked human corneas in key physical and physiological functions, including morphology, biochemical marker expression, transparency, ion and fluid transport, and gene expression. Morphological and functional equivalents to human corneas that can be produced in vitro have immediate applications in toxicity and drug efficacy testing, and form the basis for future development of implantable tissues.

The cornea comprises three major cellular layers: an outermost stratified squamous epithelium, a stroma with keratocytes, and an innermost monolayer of specialized endothelial cells. The structure of the cornea allows it to serve as a barrier to the outside environment and as a major element in the optical pathway of the eye (1). The cornea is transparent, avascular, and immunologically privileged (2), making it an excellent candidate for tissue engineering for transplantation. Various researchers have attempted to fabricate artificial corneas or parts of corneas in vitro (3), but there have been no reports of successfully reconstructed human corneas that mimic the anatomy and physiology of the human cornea.

Our objective was to develop a morphological and functional equivalent of the human cornea. Human cell lines were developed from cells isolated from the individual cellular layers of the cornea. Most were immortalized (4) by infection with an amphotropic recombinant retrovirus containing HPV16 genes E6 and E7 (5, 6); others were immortalized by transfection (7) with mammalian expression vectors containing genes encoding SV40 large T antigen, pSV3neo (8), and adenovirus E1A 12S (9), separately or in combination. Immortalized cells had random chromosomal breaks, structural rearrangements, and, in several lines, aneuploidy (10); similar chromosomal anomalies associated with immortalization were reported in HPV E6/E7 immortalized vascular endothelial cells (6), although random structural rearrangements and aneuploidy are also present in normal human corneal cells (11). The immortalized cells also had significant telomerase activity (12) [associated with the immortalized phenotype (6)] compared to little or no activity in nonimmortalized cells.

Before use in corneal equivalents, cell lines were screened for morphological, biochemical, and electrophysiological similarities to freshly dissociated or low-passage corneal cells obtained from postmortem human corneas. Electrophysiological screening of epithelial cells by means of amphotericin-perforated patch clamping (13) showed that immortalized cells had whole-cell currents similar to those of cultured corneal epithelial cells (Fig. 1, A to F) (14). Cells with altered phenotypes and physiology (transformed cells), however, showed anomalous currents (Fig. 1, G to I). Patch clamping was also used in screening keratocyte and endothelial lines. Lines with whole-cell currents closest to those of corresponding normal cells were identified; phenotypes were confirmed by expression of appropriate biochemical markers (15).

Figure 1

Representative current tracings and current-voltage (I-V) records from cultured human corneal epithelial cells (A to C), an immortalized cell line (D to F), and a line found to be transformed rather than immortalized (G toI). A tail current protocol (tracings not shown) was used to determine reversal potentials. Tail currents were obtained by clamping cells at 0 mV, depolarizing to 100 mV, then hyperpolarizing to voltages shown on the x axis. Currents and I-Vrelations revealed similar nonselective cation and fenamate-activated K+ currents in both cultured and immortalized epithelial cell lines. Similar currents were observed in transformed cells, regardless of cell type. This current is likely a Cl or nonselective cation current, as inferred from reversal potentials of 0 mV in NaCl baths, and is identical in both NaCl (not shown) and KCl baths. Pipette solution: 145 mM potassium methanesulfonate, 2.5 mM NaCl, 2.5 mM CaCl2, 5 mM Hepes, and amphotericin B (240 mg/ml, Sigma). Bath solution: 149 mM NaCl, 5 mM KCl (145 mM KCl, 2.5 mM NaCl for KCl bath), 2.5 mM CaCl2, 5 mM glucose, and 5 mM Hepes. Insets (in A, D, and G) are voltage-clamp protocols.

After cell lines were screened and chosen, the corneal tissue layers were constructed. Other investigators have used natural and synthetic polymers to produce scaffolds for engineered tissues (3, 16). For a tissue matrix, we used a collagen–chondroitin sulfate substrate cross-linked with 0.02 to 0.04% glutaraldehyde, then treated with glycine to remove unbound glutaraldehyde. Stromal, epithelial, and endothelial layers were created by mixing cells into, layering cells below, and layering cells on top of this substrate, respectively (17) (Fig. 2A). Constructs were typically maintained in tissue culture medium 199 with 10% fetal bovine serum, 1% insulin-transferrin-selenium, and gentamicin. We also used a serum-free medium (18) for optimal growth under defined conditions. Both media contained protease inhibitor and ascorbic acid (19) to retard cellular degradation of the matrix and to stimulate collagen synthesis. Once the epithelium at the bottom of the insert was confluent, it was exposed to air to allow differentiation into multilayers. Corneal equivalents were also successfully constructed in the reverse order (with corneal endothelium at the bottom and the air-liquid interface on top) and from low-passage primary human corneal cells, allowing for future development of transplants. Assembled corneal equivalents were allowed to differentiate for 2 weeks before use.

Figure 2

(A) Postmortem human cornea with scleral rim (left) and human corneal equivalent with surrounding pseudo-sclera (right) after 2 weeks in culture. Both retained their transparency, as indicated by the clarity of the “E” beneath each cornea. Scale bar, 10 mm. (B and C) Cultured human eye bank cornea (B) and corneal equivalent (C). Both show well-defined epithelial (Ep), stromal (S), and endothelial layers (En), and acellular limiting Bowman's (Bm) and Descemet's (Dm) membranes. Interlamellar clefts in eye bank corneal stroma are storage and processing artifacts. Scale bar, 100 μm. (D) Corneal equivalent with surrounding pseudo-sclera containing microvessel-like structures (arrowheads), which bound acetylated low density lipoprotein labeled with the fluorescent probe DiI (red). The cornea (C), however, remained avascular. Scale bar, 500 μm. (E) Corneal equivalent visualized with epifluorescence by confocal microscopy. Epithelial (Ep) and endothelial (En) layers are distinct (red), and keratocytes (arrowheads) are present in the collagenous stromal matrix (S, green). Scale bar, 100 μm. (F) Cornea constructed with cell lines deemed transformed by electrophysiology. The three main layers are not readily distinguishable, and transformed epithelial cells (arrows) are seen invading the stroma. Scale bar, 100 μm. Insets (in E and F) are hematoxylin and eosin–stained sections of the same tissues.

The corneal equivalents resembled human corneas in gross morphology, transparency (Fig. 2A), and histology (Fig. 2, B and C). Transmission electron microscopy showed structures of healthy, actively metabolizing cells (electron-lucent nuclei, numerous mitochondria, extensive rough endoplasmic reticulum) (20). Modifications, such as incorporation of fibrin (21), produced matrix that supported angiogenesis so the avascular cornea could be surrounded with vascularized matrix to produce a single cornea-pseudosclera construct (Fig. 2D).

The most successful multilayered corneal equivalents (Fig. 2E) were constructed from immortalized cell lines with appropriate ion channel activities (Fig. 1). Cell lines with currents indicating abnormal physiological function (Fig. 1, G to I) resulted in unsuccessful constructs (Fig. 2F). Epithelial cells showing abnormal currents stained positively for epithelial-specific keratin 12, which demonstrates the need for functional testing of cell lines in addition to immunohistochemical screening. Electrophysiological screening was therefore effective for identification of cell lines with phenotypes similar to those of low-passage or freshly dissociated human corneal cells.

To test the physiological function of the corneal equivalents, we evaluated stromal swelling, gene expression, and tissue transparency. Optimal stromal hydration is maintained in the human cornea by the pumping action of the specialized endothelium (22). Exposure of human corneal endothelium to ouabain (100 μM) has been shown to disrupt pumping and increase stromal thickness [by 16% within 2 hours (23)]. Ouabain caused swelling of the corneal equivalents from 878 μm to 1077 μm, an increase of 22.7% (n = 5), as measured by optical coherence tomography (OCT) (24). Control corneal equivalents showed slight thinning by 44 μm or 5.4% (from 813 μm to 769 μm;n = 4), likely the result of an active pump. The corneal equivalents were thus similar to human corneas in both stromal swelling and a physiologically active endothelium.

We used the reverse transcription polymerase chain reaction (RT-PCR) to examine changes in gene expression in injured corneal equivalents and human eye bank corneas (25). Corneas and corneal equivalents were exposed to 5% SDS, a surfactant that causes mild injury to the cornea (26). SDS exposure resulted in increased mRNA for genes involved in corneal wound healing [c-fos; the genes encoding cytokines interleukin-1 (IL-1), IL-6, basic fibroblast growth factor (bFGF), and vascular endothelial growth factor (VEGF); and the gene encoding type I collagen (Coll I)] (27) in both fabricated and eye bank corneas (Fig. 3). Fabricated corneas showed greater sensitivity, possibly because they are healthier tissues than eye bank corneas available for research.

Figure 3

(A) Representative Southern blots of RT-PCR products from human corneal equivalents treated with 5% SDS or medium alone (Con). 18S internal standards are shown below each set. (B) Southern blots of RT-PCR products from three sets of human cornea donors (D1 to D3). One cornea in each pair was a control (a); the other was treated with 5% SDS (b). 18S internal standards are shown below each set. (C) Changes in mRNA expression in human corneas and corneal equivalents with SDS. The intensity of mRNA bands was normalized to corresponding 18S bands. Changes in gene expression are shown as relative increase over control. Corneal equivalents,n = 8 for control and n = 9 for treated groups; human corneas, n = 3 for control and treated groups. *P < 0.05 versus control (t test).

To further evaluate the corneal equivalents, we examined changes in light transmission after exposure to chemicals, comparing responses to those of human and rabbit corneas [the standard in ocular toxicology tests (26)]. We observed opacification (Fig. 4, A and B) with increased cell death (Fig. 4C) (28) within the treatment zone of all three types of corneas. Changes in the transparency of corneal equivalents in response to chemicals (29) were similar to those observed in human and rabbit corneas (Fig. 4, D to F). These results showed that the corneal equivalents had an active response to different grades of injury, an important functional characteristic of human corneas.

Figure 4

Human eye bank corneas (A) and corneal equivalents (B) treated on their epithelium for 5 min with 100 μl of DMEM (control; left cornea) or 70% dimethyl dialkyl ammonium chloride (right cornea), an ocular irritant. Treated corneas show opacification of treatment areas (arrowheads). Scale bar, 10 mm. (C) Treated corneal equivalent stained with live/dead stain. Red (arrowhead) indicates area of dead cells; green indicates live cells. Scale bar, 10 mm. (D to F) Changes in light transmission through (D) rabbit corneas, (E) corneal equivalents, and (F) human corneas treated for 5 min with 100 μl of DMEM (control); (a) artificial tears (0.3% hydroxymethylcellulose, 0.1% dextran, polyquad preservative); (b) anionic surfactant mixture (20% alkyl ethoxysulfate, 5% alkyl sulfate); (c) anionic surfactant mixture (30% alkyl benzene sulfonate, 15% alcohol ethoxylate); (d) cationic surfactant (50% trimethyl alkyl ammonium chloride); and (e) cationic surfactant (70% dimethyl dialkyl ammonium chloride). Historical rabbit low volume eye test (24) maximum average scores for these substances were 27.2 (b), 55.7 (c), 93.0 (d), and 109.3 (e), where a higher score indicates a greater degree of corneal opacification. Human corneas, n = 4 for each chemical; corneal equivalents, n = 5 for each chemical; rabbit corneas, control group, n = 12; all other groups, n = 17. *P < 0.05 versus control (ANOVA).

These engineered corneal equivalents have immediate applications as human ocular irritancy models for evaluating new chemicals and drugs, as alternatives to animals (30), and in drug efficacy testing such as for wound healing. The corneal equivalents can also be used in biomedical research, for example, to study wound healing and cell-matrix interactions. This technology provides a strong basis for the development of temporary or permanent cornea replacements with low rejection rates. Future research could lead to readily available, complex engineered tissues that reproduce their natural human counterparts and are suitable for implants, transplants, and biomedical research.

  • * To whom correspondence should be addressed. E-mail: mgriffith{at}


View Abstract

Navigate This Article