Fabrication of a Corneal Model Composed of Corneal Epithelial and Endothelial Cells via a Collagen Vitrigel Membrane Functioned as an Acellular Stroma and Its Application to the Corneal Permeability Test of Chemicals

A collagen vitrigel membrane (CVM) we developed can function as both a scaffold for cells and a pathway for chemicals. To extrapolate the corneal permeability of chemicals in vivo, we proposed six corneal models using the CVM. Thin and thick CVMs were used as models for Bowman’s membrane (BM) and an acellular stroma (AS), respectively. Models for a corneal epithelium (CEpi), a CEpi-AS, a CEpi-endothelium (Endo), and a CEpi-AS-Endo were fabricated by culturing corneal epithelial cells and/or corneal endothelial cells on the surface of CVMs. Subsequently, the permeability coefficient (Papp) value of each model was calculated using five chemicals with different molecular radii; cyanocobalamin and four fluorescein isothiocyanate-dextrans (FD) (FD-4, FD-10, FD-20, and FD-40). The slopes of Papp versus molecular radii of those chemicals in the both BM and AS models were almost similar to data using an excised rabbit corneal stroma. The ratios of Papp values in models for BM, CEpi, and CEpi-Endo against those in data using an excised rabbit cornea were calculated as 75.4-fold, 6.4-fold, and 4.5-fold for FD-4, and 38.7-fold, 10.0-fold, and 4.2-fold for FD-10, respectively. Similarly, those in models for AS, CEpi-AS, and CEpi-AS-Endo were calculated as 26.1-fold, 2.5-fold, and 0.6-fold for FD-4, and 26.1-fold, 1.5-fold, and 0.6-fold for FD-10, respectively. These results suggest that the CEpi-AS-Endo model with both the barrier function of corneal cell layers and the diffusion capacity of chemicals in thick CVM is most appropriate for extrapolating the corneal permeability of chemicals in vivo.


DMD#80820 Introduction
Corneal permeation studies required for ophthalmic drug development are generally performed by the test methods utilizing laboratory animals in vivo or ex vivo (Hahne et al., 2012;Dave et al., 2015).
However, animal experiments have disadvantages such as ethical issues on the sacrifice of life, high costs, poor reproducibility and the questionable extrapolation of animal results to humans (Acheampong et al., 2002;Reichl and Becker, 2008;Baranowski et al., 2014;Scott et al., 2010;Pescina et al., 2015). To overcome these issues on animal experiments, a novel corneal model in cell culture systems in vitro appropriate for extrapolating the corneal permeability of chemicals in vivo is required for promoting the efficient development of eye drops.
Human cornea is made of three different tissues. Corneal epithelium, stroma and endothelium in the order from the outside is formed by about 6 corneal epithelial cell layers with barrier function, keratocytes scattered in high density-collagen fibrillar layers in approximately 500 µm thick with diffusional inhibition and an endothelial cell monolayer with barrier function, respectively. The corneal permeability of chemicals is mainly dependent upon the barrier function of corneal epithelium and is collaterally regulated by the diffusion rate of corneal stroma and the barrier function of corneal endothelium (Reichl and Becker, 2008). Therefore, the single models of corneal epithelium (CEpi) and the combined models of corneal epithelium-stroma-endothelium in cell culture systems have been developed by utilizing various scaffolds.
CEpi models exhibited multi-layer architecture like CEpi in vivo and revealed the formation of epithelial barrier function with rising trans-epithelial electrical resistance (TEER-epi) values. However, the CEpi This article has not been copyedited and formatted. The final version may differ from this version. DMD Fast Forward. Published on May 29, 2018as DOI: 10.1124 at ASPET Journals on November 1, 2021 dmd.aspetjournals.org Downloaded from DMD#80820 6 models lack the effect of corneal stroma and endothelium. From this viewpoint, several corneal models composed of not only epithelium but also stroma and endothelium have been developed by utilizing three different types of corneal cells and unique scaffolds functioned as an artificial corneal stroma, e.g. crosslinked collagen-chondroitin sulfate (Griffith et al., 1999), an acellular corneal matrix derived from porcine (Xu et al., 2008) and a type I collagen gel (Reichl et al., 2004). Such corneal models in vitro well mimicked cornea in vivo in morphological and physiological features. Especially, the corneal model using the collagen gel showed the similar behavior with excised porcine cornea in the permeability of model drugs such as pilocarpine hydrochloride, befunolol hydrochloride, and hydrocortisone. However, none of the corneal models in vitro have been in widespread use as pharmaceutical routine works. One of the reasons is the limitation of producing same models due to disadvantages of previous scaffolds such as mechanical weakness, variation of animal-derived materials, complex preparation, etc. Therefore, the novel scaffold that allows the brief and reproducible fabrication of a corneal model is necessary for promoting the efficient development of eye drops.
A collagen vitrigel membrane (CVM) is a stable gel produced by rehydration after vitrification of a traditional collagen gel, and consequently it is formed of high density collagen fibrils comparable to connective tissues in vivo. The CVM is tough and transparent, and also bioactive chemicals with various molecular weight can penetrate it. Therefore, it is easily handled with forceps and functions as a scaffold to reconstruct useful tissues for the studies on regenerative medicine, drug developments and alternatives to experimental animals, etc. (Takezawa et al., 2004(Takezawa et al., , 2007a(Takezawa et al., , 2007b(Takezawa et al., , 2007c. Moreover, we developed a This article has not been copyedited and formatted. The final version may differ from this version. DMD Fast Forward. Published on May 29, 2018as DOI: 10.1124 at ASPET Journals on November 1, 2021 dmd.aspetjournals.org Downloaded from DMD#80820 7 mass fabrication technology of not only a thin CVM but also a thick one (Takezawa et al., 2012). We established a fabrication method of a CEpi model utilizing an air-liquid interface culture system. The culture system facilitates the induction of layering corneal epithelial cells cultured on the CVM scaffold that was prepared on a polyethylene terephthalate (PET) membrane of a Millicell chamber. (Takezawa et al., 2008(Takezawa et al., , 2011. However, this model is unsuitable for immunohistological analyses by freeze sectioning due to the existence of hard PET membrane. Therefore, we developed a new chamber merely accompanying a CVM without the PET membrane and established its mass production process (Takezawa et al., 2012).
From a viewpoint of developing an ideal scaffold functioned as an artificial corneal stroma, the CVM has several advantages such as mechanical toughness, reproducibility and brief preparation. Moreover, the CVM is composed of 10 to 25% collagen fibrils equivalent to corneal stroma (Takezawa et al., 2007c;Reichl et al., 2008). In this study, we aimed to define an optimum corneal model involving the CVM scaffold appropriate for extrapolating the corneal permeability of chemicals in vivo. The CVM scaffolds with two different thicknesses of about 20 μm as a replacement of Bowman's membrane and about 450 μm as an acellular corneal stroma were designed, and subsequently six corneal models composed of merely a thin CVM, CEpi on a thin CVM, corneal epithelium-endothelium (CEpi-Endo) via a thin CVM, merely a thick CVM, CEpi on a thick CVM, and CEpi-Endo via a thick CVM were fabricated (Fig. 1). Then, the permeability coefficient of each model was calculated by using five model chemicals with different molecular radii and was compared with that of excised rabbit cornea. This article has not been copyedited and formatted. The final version may differ from this version.  were purchased from Sigma-Aldrich.

Preparation of CVM chambers.
A chamber (ad-MED Vitrigel TM ) with a collagen xerogel membrane (CXM) containing 0.05 mg type-I collagen per 1.0 cm 2 was purchased from Kanto chemical Co., Inc.
(Tokyo, Japan) in which CXM chambers were manufactured in accordance with our previous report (Takezawa et al., 2012). The CXM chamber was set in the well of a 12-well plate. Then, the CXM was This article has not been copyedited and formatted. The final version may differ from this version. immersed in a culture medium by pouring 1.5 ml outside and 0.5 ml inside the chamber in the well for 10 min to convert the xerogel into a vitrigel immediately before use, resulting the preparation of a chamber with a thin CVM of about 20 μm-thickness. Here, the thin CVM of chamber can be used as a Bowman's membrane (BM) model because of the similarity of architecture and components. Regarding the chamber with a thick CVM of about 450 μm-thickness (i.e. an acellular-stroma (AS) model), we custom-ordered a chamber with a CXM containing 1.125 mg type-I collagen per 1.0 cm 2 to Kanto chemical Co., Inc. in the similar procedure of ad-MED Vitrigel TM .

Culture of HCE-T cells and BCD C/D-1b cells.
A SV40-immortalized human corneal epithelial cell strain (HCE-T cells; RCB #2280) was obtained from RIKEN BioResource Center (Tsukuba, Japan) (Araki- Sasaki et al., 1995). The cells were cultured and maintained in a 1:1 mixture medium of Dulbecco's Modified Eagle Medium and Nutrient Mixture F-12 supplemented with 5% heat-inactivated fetal bovine serum, 5 μg/ml recombinant human insulin, 10 ng/ml recombinant human epidermal growth factor, 0.5% dimethyl sulfoxide, 100 units/ml penicillin and 100 μg/ml streptomycin (DF-medium) at 37°C in a humidified atmosphere of 5% CO2 in air. Isolation and characterization of an endothelial cell-like clone from BCD C/D-1b cells. BCD C/D-1b cells suspended in the D-medium were seeded and cultured in a 96-well cell culture plate at a density of 0.5cells/well. Several endothelial cell-like clones were morphologically selected by phase-contrast microscopy. Each clone was expanded to about 1x10 7 cells and preserved in liquid nitrogen. One clone representing excellent proliferative performance was subjected to the following experiment. The clonal cells recovered from liquid nitrogen were defined as passage 1 and were sub-cultured until passage 28 to estimate their morphological stability.
To confirm the properties essential for the fabrication of a corneal endothelial layer, the endothelial cell-like clone was cultured in the chamber with a thin CVM and subjected to the time-dependent analyses for cell morphology, protein expression and barrier function. The clonal cells suspended in 0.5 ml of the D-medium at a density of 2.0×10 5 cells/ml were seeded onto the thin CVM of chamber pre-set in a well of a 12-well plate with 1.5 ml of the fresh medium in the well and cultured for 7 days. The media were changed at day 1, 4 and 7. The cell morphology and immuno-histology for ZO-1 and Na + -K + ATPase protein expressions were observed at day 1, 3 and 7 by a phase-contrast microscope (TE300; Nikon, Tokyo, Japan) and a laser scanning confocal microscope (FV1000; Olympus, Tokyo, Japan), respectively. Also, the barrier function was analyzed by measuring the electrical resistance values of each CVM chamber before (Rblank) and after (Rsample) culturing the cells at day 0 (2 hours), 1, 2, 3, 4 and 7 using an electrical This article has not been copyedited and formatted. The final version may differ from this version.  11 resistance meter (Kanto Chemical). Trans-endothelial electrical resistance (TEER-end) value was calculated by using the following formula: TEER-end = (Rsample -Rblank) × effective surface area Here, the effective surface area of CVM chamber was 1.0 cm 2 .
Fabrication of six corneal models. We prepared six corneal models as shown in Fig. 1. The preparation of BM model and AS model were described above. Corneal epithelium (CEpi) model, corneal epithelium-endothelium (CEpi-Endo) model, corneal epithelium-acellular stroma (CEpi-AS) model and corneal epithelium-acellular stroma-endothelium (CEpi-AS-Endo) model were fabricated as follows.
In the fabrication of CEpi model, HCE-T cells suspended in 0.5 ml of the DF-medium at a density of 1.2×10 5 cells/ml were seeded onto the thin CVM (i.e. a BM model) of chamber pre-set in a well of a 12-well plate with 1.5ml of the fresh medium in the well and the cells were cultured for 2 days.
Subsequently, the exterior medium of chamber was changed and the interior medium of it was removed to start the additional culture under the air-liquid interface for 4 days. The exterior medium of chamber was changed on the third day in the additional culture. Also, the fabrication of CEpi-AS model was achieved by using the thick CVM (i.e. an AS model) of chamber in the above cell seeding process.
In the fabrication of CEpi-Endo model, HCE-T cells were first cultured for 3 days in the same procedure for fabricating the CEpi model. Subsequently, the CVM chamber was transferred to a well of 6-well plate and flipped upside down in order to make a reverse-side compartment for cell culture by fixing a plastic This article has not been copyedited and formatted. The final version may differ from this version.

DMD#80820
12 cylinder with an appropriate size (e.g. inner-outer diameter: 11-16 mm; length: 8.5 mm) onto the bottom of CVM chamber. The endothelial cell-like clonal cells derived from BCD C/D-1b cells suspended in 0.5 ml of the D-medium at a density of 2.0×10 5 cells/ml were seeded into the reverse-side compartment. The clonal cells were cultured for 2 hours to induce the sufficient adhesion to the CVM, resulting the co-culture with HCE-T cells via the CVM. Subsequently, the endothelial cell medium was removed and the cylinder was detached from the CVM chamber. Then, the CVM chamber was set in a well of a 12-well plate with 1.5ml of the DF-medium and both cells were co-cultured for additional 3 days under air-liquid interface.
Also, the fabrication of CEpi-AS-Endo model was achieved by using the thick CVM (i.e. an AS model) of chamber in the above cell seeding process.
All models were subjected to histology and permeability examinations.
Histology and immunohistology. For clarifying morphological characteristics, corneal endothelial layers in a CVM chamber at day 1, 3 and 7 and three corneal models (CEpi-Endo model, CEpi-AS model and CEpi-AS-Endo model) except for the CEpi model as previously reported (Yamaguchi et al., 2013) were isolated from the plastic cylinder of the CVM chamber using an appropriate disposable biopsy punch and fixed for 5 min in methanol kept on ice immediately after sufficiently chilling it at -45 °C.
The fixed corneal endothelial layers were washed with phosphate buffered saline (PBS) three times.
Subsequently, they were incubated with PBS containing 1% normal goat serum for 30 min to block non-specific adsorption of antibodies. Then, the first antibodies against ZO-1 or Na + -K + ATPase prepared This article has not been copyedited and formatted. The final version may differ from this version. in PBS containing 1% normal goat serum at a concentration of 5 μg/ml were applied and incubated for 16 hours at 4°C, followed by washing them with PBS three times. The secondary antibodies against rabbit IgG or mouse IgG prepared in PBS containing 1% normal goat serum at a concentration of 4 μg/ml were applied and incubated for 3 hours at room temperature, followed by washing them with PBS three times.
Subsequently, cell nuclei were counterstained with Hoechst33342. Stained cells were observed by a laser scanning confocal microscope (FV1000; Olympus, Tokyo, Japan).
The fixed corneal models were embedded in an O.C.T. Compound after removing the excessive methanol around them with an absorbent paper towel, frozen in liquid nitrogen, and stored at -80°C. The samples were vertically cut into cross-sections with a thickness of 5 μm against the CVM using a cryostat (CM3050S; Leica Microsystems, Wetzlar, Germany). The frozen sections spread on a glass-slide were dried out for 60 min at room temperature. For histology, the sections were immersed in water for 5 min to remove the O.C.T. Compound, stained with hematoxylin and eosin, and observed by a light microscope (E600; Nikon, Tokyo, Japan). For immunohistology, the sections derived from CEpi-AS model were subjected to immunostaining using the first antibodies against ZO-1, occludin, connexin-43, cytokeratin 3, or MUC1 and counterstaining using Hoechst33342 in a similar procedure mentioned above.
Calculation of permeability coefficient. All 6 corneal models were subjected to the permeability test using the following 5 test chemicals; cyanocobalamin, FD-4, FD-10, FD-20 and FD-40 representing approximate 8.5, 14, 23, 33 and 45 Å in Stokes' radius, respectively. Also, a commercially available This article has not been copyedited and formatted. The final version may differ from this version.  The Papp values of FD-4 and FD-10 in excised male rabbit corneas are reported as 0.056 ×10 -6 and 0.031 ×10 -6 cm/s, respectively . The ratios of Papp values for FD-4 and FD-10 in all 6 corneal models against those in the report using excised rabbit corneas were calculated by dividing each Papp value of the former by that of the latter. Here, the ratio of 1.0 represents an ideal model equivalent for excised rabbit cornea. Next, the contribution of epithelium and endothelium in 6 corneal models were analyzed by the concept for reduction degree when the Papp values of CEpi (or CEpi-AS) models were compared with those of BM (or AS) models and when the Papp values of CEpi-Endo (or CEpi-AS-Endo) models were compared with those of CEpi (or CEpi-AS) models, respectively. Here, the reduction degree for epithelium was calculated using the following formula: 16 cellular shapes were not uniform polygon but different varieties such as small round, cobblestone, spindle and hypertrophy. Also, some interstices were observed especially around the fibroblast-like cells even in the confluent stage ( Fig. 2A). Several endothelial cell-like clones showing the morphology of cobblestone were successfully isolated from the BCD C/D-1b cells by the limiting dilution method. One clone representing excellent proliferative performance was subjected to the following experiment. The endothelial cell-like clonal cells at passage 1 formed a confluent monolayer and the individual cellular shapes represented cobblestone (Fig. 2B). Also, the clonal cells at passage 28 showed the same morphological properties as that at passage 1 (Fig. 2C), suggesting that the clonal cells could maintain the morphological stability for the following experimental period.
The clonal cells cultured in a CVM chamber for 1 day formed a confluent monolayer with loose cell-to-cell communications in which the individual cellular shapes represented non-uniform cobblestone (Fig. 3A). The cells expressed Na + -K + ATPase that is active transporter known as a marker of corneal endothelium (Fig. 3D), however they rarely did ZO-1 that is tight junction-associated protein (Fig. 3G).
Meanwhile, the cells on day 3 fabricated a confluent monolayer with tight cell-to-cell communications in which the individual cellular shapes represented hexagon and they well expressed not only Na + -K + ATPase but also ZO-1 (Fig. 3B, 3E and 3H). Subsequently, the cells on day 7 showed the more uniform thickness of a corneal endothelial layer with expressions of Na + -K + ATPase and ZO-1 (Fig. 3C, 3F and 3I).
Also, the TEER-end value of the clonal cells cultured in a CVM chamber increased for initial 3 days with significant differences, and subsequently it showed the nearly constant without significant differences This article has not been copyedited and formatted. The final version may differ from this version.  (Fig. 4).
These data suggest that a corneal endothelial layer with barrier function was fabricated by culturing the clonal cells in a CVM chamber for 3 days.  (Fig. 5A). The uppermost layer was covered with extremely flattened cells and the other layers were mostly composed of squamous cells (Fig. 5B). Regarding the five corneal epithelium-associated proteins, ZO-1 and occludin that are tight junction-associated proteins were abundantly expressed in the lateral and basal surfaces of cells in the superficial layer in comparison to the other layers (Fig. 5C, D). Connexin-43 consisting gap junctions and cytokeratin 3 that is a type II cytokeratin in corneal epithelium were expressed in the membrane and cytoplasm of cells in all layers, respectively (Fig. 5E, F). MUC1 that is a cell membrane-spanning mucin was merely expressed in the apical surface of cells in the superficial layer (Fig. 5G). These data demonstrated that the epithelial This article has not been copyedited and formatted. The final version may differ from this version.  Table 1. The Papp values of cyanocobalamin in CEpi-Endo, CEpi-AS and CEpi-AS-Endo models cannot be calculated because its concentration in each sample was lower than the detection limit.
In the comparison of acellular models with a common scaffold for fabricating corneal models, the slopes of permeability coefficients versus molecular radii of test chemicals in BM model, AS model and commercially available multi-porous PET membrane were -2.64, -2.04 and -0.84, respectively (Fig. 7).
These results reveal that the Papp value corresponding for each test chemical in the both models exponentially decreases as its molecular radius increases although that in the multi-porous PET membrane slightly does. In addition, the Papp values of cyanocobalamin, FD-4, FD-10 and FD-20 in the AS model were significantly low in comparison to that in the BM model, suggesting that the permeability coefficient of test chemicals was regulated by the thickness of CVM used for each model. FD-4 and 94.3% and 58.7% for FD-10, respectively. In particular, there was no significant difference between the Papp value of CEpi-AS-Endo model and that of excised rabbit cornea (Fig. 8). These results reveal that the architecture composed of not only both epithelium and endothelium but also acellular stroma is essential for fabricating a culture model equivalent to the excised rabbit cornea.

Discussion
The corneal permeability of drugs in vivo is mainly dependent upon the barrier function of epithelial cell layers on Bowman's membrane and is collaterally regulated by the diffusion capacity in stromal matrix and the barrier function of an endothelial layer on Descemet's membrane (Reichl et al., 2008). Here, the main component of Bowman's membrane, stromal matrix and Descemet's membrane is a network of high density collagen fibrils. In the current study, we focused on the network of high density collagen fibrils functioned as not only a scaffold for cells but also a pathway for drugs and newly prepared a thick CVM equivalent to acellular stroma in addition to a previous thin CVM equivalent to Bowman's membrane. This article has not been copyedited and formatted. The final version may differ from this version. Based on this concept, six corneal models possessing matrix-dependent diffusion capacity and/or cell layer(s)-dependent barrier function were fabricated by utilizing thin and thick CVMs, corneal epithelial cells and endothelial cells; BM model, CEpi model as previously described (Yamaguchi et al., 2013), CEpi-Endo model, AS model, CEpi-AS model and CEpi-AS-Endo model. Subsequently, an optimal corneal model in vitro useful for extrapolating the corneal permeability was determined.
The CVM can play a role of not only a bifacial scaffold in three-dimensional cell culture but also a pathway for proteins with a wide range of molecular weight (Takezawa et al., 2007a(Takezawa et al., , 2007b. From this viewpoint, we proposed six corneal models using chambers accompanying thin and thick CVMs (Fig.1).
The CVM potential as a scaffold for fabricating four culture models was examined as follows. First, we confirmed that the thin CVM provided an appropriate scaffold for corneal endothelial cell-like clonal cells and consequently that a corneal endothelium with barrier function and specific protein expressions was well reconstructed by culturing the clonal cells on it for at least 3 days (Fig. 2-4). Next, we fabricated 4 culture models as follows. A CEpi model with barrier function and specific protein expressions was prepared by culturing HCE-T cells on the thin CVM of chamber for 6 days (data not shown) as previously reported. 21 According to this fabrication procedure, a CEpi-AS model was prepared by culturing HCE-T cells on the thick CVM of chamber for 6 days. Subsequently, the model was confirmed to exhibit both epithelial barrier function (data not shown) and specific protein expressions (Fig. 5). Also, a CEpi-Endo model and a CEpi-AS-Endo model were prepared by culturing HCE-T cells in the thin and thick CVM chambers for 2 days and subsequently co-culturing corneal endothelial cell-like clonal cells on the opposite This article has not been copyedited and formatted. The final version may differ from this version. DMD Fast Forward. Published on May 29, 2018as DOI: 10.1124 at ASPET Journals on November 1, 2021 dmd.aspetjournals.org Downloaded from DMD#80820 21 surface of CVMs for 4 days, respectively (Fig. 6). These data suggest that the CVM chambers in comparison to the traditional scaffolds are appropriate for mass-production with high reproducibility and can provide an easy handling in cell culture manipulation and a short-term fabrication of not only tissue-type but also organ-type culture models.
The CVM potential as a pathway for penetrating five test chemicals was examined as follows. We calculated the Papp values in both acellular models and culture models from the viewpoint of the diffusion capacity in a network of high density collagen fibrils and the barrier function of cell layer(s), respectively (Table 1). First, in the acellular models, the slopes of permeability coefficients versus molecular radii of test chemicals in BM model, AS model and commercially available multi-porous PET membrane were investigated (Fig. 7). Meanwhile, the slope value in the excised rabbit corneal stroma was calculated as -2.52 from the previous report describing the Papp values of four test chemicals; phenylephrine, acebutolol, hemoglobin and albumin representing approximate 4.0, 5.0, 31 and 35 Å in Stokes' radius, respectively. (Prausnitz and Noonan., 1998). This work demonstrated that the Papp values decrease as the molecular radii increase in the excised rabbit corneal stroma. Also, our current data revealed that the Papp values decrease as the molecular radii increase in both a BM model and an AS model, and consequently their slope values were -2.64 and -2.04, respectively. Meanwhile, the slope of a multi-porous PET membrane was -0.84. This calculation demonstrated that the slope in the excised rabbit corneal stroma almost matched both a BM model and an AS model although it was far from that of a multi-porous PET membrane. These findings suggest that acellular models utilizing CVM chambers can be utilized as This article has not been copyedited and formatted. The final version may differ from this version. DMD Fast Forward. Published on May 29, 2018as DOI: 10.1124 at ASPET Journals on November 1, 2021 dmd.aspetjournals.org Downloaded from DMD#80820 22 alternatives to the excised rabbit corneal stroma due to their similarity in the diffusion rate of chemicals.
Next, in the culture models, we compared the Papp values of four culture models with that of an excised rabbit cornea to clarify the contribution of barrier function based on cell layers. Consequently, the ratios of Papp values of FD-4 in a BM model, a CEpi model and a CEpi-Endo model against an excised rabbit cornea were calculated as 75.4-fold, 6.4-fold, and 4.5-fold, respectively. Also, those of FD-10 were calculated as 38.7-fold, 10.0-fold, and 4.2-fold, respectively. These calculations mean that the effect of epithelium and endothelium on the reduction degree of Papp values is 91.5 % and 30.6% for FD-4 and 74.2% and 58.1% for FD-10, respectively. Similarly, those of FD-4 in an AS model, a CEpi-AS model and a CEpi-AS-Endo model against an excised rabbit cornea were calculated 26.1-fold, 2.5-fold and 0.6-fold, respectively. Also, those of FD-10 were calculated as 26.1-fold, 1.5-fold and 0.6-fold, respectively. Meanwhile, the effect of epithelium and endothelium on the reduction degree of Papp values is in the range of 74.2-94.3% and 30.6-76.4%, respectively. This suggests that both epithelial and endothelial cell layers contribute to the regulation of the permeability coefficient although the former effect is larger than the latter. Moreover, the Papp values of FD-4 and FD-10 in not a CEpi-Endo model but a CEpi-AS-Endo model were very close to those of an excised rabbit cornea (Fig. 8). These data suggest that not only the barrier function of both epithelial and endothelial cell layers but also the diffusion capacity in a network of high density collagen fibrils is indispensable to fabricate corneal models in vitro appropriate for extrapolating the corneal permeability of chemicals in vivo. The advantages of our CEpi-AS-Endo models utilizing CVM scaffold are that the fabrication period is short and the handling in This article has not been copyedited and formatted. The final version may differ from this version.

DMD#80820
23 permeability assay is easy in comparison to the traditional in vitro corneal models utilizing fragile scaffold (Reichl et al., 2004;Xu et al., 2008;Rönkkö et al., 2016). To investigate the feasibility of our CEpi-AS-Endo models in comparison to the traditional models in vitro, however, further studies using candidate chemicals for ophthalmic drugs are essential in not only corneal permeability test but also metabolic assay.
In this study, we fabricated a novel corneal model utilizing corneal epithelial cells, endothelial cells and a CVM functioned as both a bifacial scaffold and a pathway for chemicals, and subsequently the permeability of test chemicals with different molecular size were analyzed to evaluate the correlation with the data on an excised rabbit cornea. Consequently, we succeeded in developing a new technology for extrapolating the corneal permeability of test chemicals. Therefore, we hope that this study would provide a new concept for fabricating an artificial cornea as a research tool useful for the development of ophthalmic drugs.

Authorship Contributions
Participated in research design: Takezawa.   Table 1 and that in excised rabbit cornea. Each value This article has not been copyedited and formatted. The final version may differ from this version.