Predictive performance of the Vitrigel-eye irritancy test method using 118 chemicals.

We recently developed a novel Vitrigel-eye irritancy test (EIT) method. The Vitrigel-EIT method is composed of two parts, i.e., the construction of a human corneal epithelium (HCE) model in a collagen vitrigel membrane chamber and the prediction of eye irritancy by analyzing the time-dependent profile of transepithelial electrical resistance values for 3 min after exposing a chemical to the HCE model. In this study, we estimated the predictive performance of Vitrigel-EIT method by testing a total of 118 chemicals. The category determined by the Vitrigel-EIT method in comparison to the globally harmonized system classification revealed that the sensitivity, specificity and accuracy were 90.1%, 65.9% and 80.5%, respectively. Here, five of seven false-negative chemicals were acidic chemicals inducing the irregular rising of transepithelial electrical resistance values. In case of eliminating the test chemical solutions showing pH 5 or lower, the sensitivity, specificity and accuracy were improved to 96.8%, 67.4% and 84.4%, respectively. Meanwhile, nine of 16 false-positive chemicals were classified irritant by the US Environmental Protection Agency. In addition, the disappearance of ZO-1, a tight junction-associated protein and MUC1, a cell membrane-spanning mucin was immunohistologically confirmed in the HCE models after exposing not only eye irritant chemicals but also false-positive chemicals, suggesting that such false-positive chemicals have an eye irritant potential. These data demonstrated that the Vitrigel-EIT method could provide excellent predictive performance to judge the widespread eye irritancy, including very mild irritant chemicals. We hope that the Vitrigel-EIT method contributes to the development of safe commodity chemicals.

Introduction

The prediction of eye irritation following chemical exposure is required for the development of not only cosmetics and consumer products but also drugs. Test methods in vivo, ex vivo and in vitro have been developed to predict eye irritation (Yamaguchi et al., 2013). Concerning eye irritation tests (EITs) in vivo, the Draize rabbit EIT has been mainly utilized for evaluating cosmetic ingredients (Draize et al., 1944; OECD, 2012a) and pharmaceutical agents (Uematsu et al., 2007). Regarding those ex vivo, excised bovine corneas and chicken eyes successfully contributed to the establishment of Organisation for Economic Co‐operation and Development (OECD) Test Guidelines as the bovine corneal opacity permeability test (OECD, 2013) and isolated chicken eye test (OECD, 2009), respectively. In addition, concerning those in vitro, various kinds of two‐ and three‐dimensional cell culture systems have been proposed. The fluorescein leakage test method was adopted as an OECD test guideline (OECD, 2012b). The draft test guideline of short time exposure test using a monolayer culture system of Statens Seruminstitut rabbit cornea cells was published by OECD (ICCVAM, 2013; OECD, 2014; Sakaguchi et al., 2011; Takahashi et al., 2011). In vitro test methods using three‐dimensional culture models have an advantage that these models can directly expose water‐insoluble chemicals (Cotovio et al., 2010; Jung et al., 2011; Katoh et al., 2013). The EpiOcular‐EIT using a tissue culture model reconstructed by culturing normal human epidermal keratinocytes is currently under peer review process aiming for an OECD test guideline (Kaluzhny et al., 2011; Pfannenbecker et al., 2013). However, no ex vivo and in vitro test methods have the performance to replace fully the Draize‐EIT that has been adopted for regulatory purposes. Therefore, a tiered approach of combining several test methods that used different complementary indicators was proposed (Scott et al., 2010). In ex vivo test methods such as the bovine corneal opacity permeability and isolated chicken eye tests, the degree of tissue damage based on change in corneal opacity, corneal permeability of fluorescein and thickness of cornea was utilized as an indicator. In in vitro test methods using three‐dimensional culture models fabricated in various culture inserts, cellular viability measured by the MTT assay has been used as a major indicator. However, the MTT assay has some limitations. The MTT solution could penetrate only two or three layers from the basal side because this solution was placed under the basal side of the models. This procedure may underestimate the mild irritancy that has toxic effects only in the superficial layer of the corneal epithelium (Pauly et al., 2009). To overcome this issue, different indicators such as occludin, interleukin‐8 and MUC1 have been proposed (Meloni et al., 2010; Song and Joo, 2004). The most apical part of the lateral membrane in the superficial epithelial cells contains the junctional complex, including tight junctions, which thus directly contribute to the first line of defense in the cornea. Therefore, the structural and functional change of tight junction‐associated protein such as occludin and ZO‐1 was reported as an early marker of eye irritancy (Meloni et al., 2010).

A collagen vitrigel membrane (CVM) we previously developed is composed of high‐density collagen fibrils equivalent to connective tissues in vivo and is easily handled with tweezers. In addition, it possesses excellent transparency and permeability of protein with high molecular weight and consequently the various studies utilizing it as a cell culture scaffold advances so well (2004, 2007a, 2007b, 2007c). We established a preparation method of a corneal epithelium model utilizing an air–liquid interface culture system that facilitates induction of layering rabbit corneal epithelial cells cultured on the CVM scaffold (Takezawa et al., 2008). To overcome species differences between human and rabbit in sensitivity to exogenous chemicals, we developed a human corneal epithelium (HCE) model by three‐dimensionally culturing HCE‐T cells on the CVM scaffold (Takezawa et al., 2011a). Here, the scaffold was fabricated on a polyethylene terephthalate (PET) membrane of a Millicell chamber appropriate for the transepithelial electrical resistance (TEER) assay of epithelial cells. TEER is known as a suitable method for evaluating the integrity of the tight junction of corneal epithelium in vivo (Uematsu et al., 2007). We examined four chemicals using the HCE model, and consequently demonstrated that the time‐dependent relative changes of TEER are useful indicators to assess ocular irritancy effects of chemicals even in the middle category (Takezawa et al., 2011a). However, this model is inappropriate for immunohistological analyses due to difficulty in preparing its frozen sections, including the PET membrane. To overcome this inconvenience we recently developed a novel chamber merely accompanying a CVM without the PET membrane and established its mass production process (Takezawa et al., 2011b, 2012). Recently, we established a new test method to extrapolate the widespread eye irritancy by briefly analyzing the time‐dependent profile of TEER after exposing chemicals to a HCE model reconstructed in a CVM chamber. Here, we named the new test method as a “Vitrigel‐EIT method.” Thirty chemicals were successfully classified into the irritant or non‐irritant category without false negatives by the Vitrigel‐EIT method (Yamaguchi et al., 2013).

In this study, we aimed to estimate the predictive performance of the Vitrigel‐EIT method by testing a total of 118 chemicals, including the previous 30 ones. In addition, we intended to clarify the mechanism‐based reason for raising false‐negative and false‐positive reactions by measuring the pH level of the test chemical solution and observing the immunohistology of HCE models after exposing test chemicals, respectively. The immunohistological observation was performed for ZO‐1, a tight junction‐associated protein and MUC1, a cell membrane‐spanning mucin.

Materials and methods

Antibodies and reagents

The rabbit polyclonal antibody for ZO‐1 and mouse monoclonal antibody for MUC1 were purchased from Life Technologies Corp. (Grand Island, NY, USA) and Sanbio BV (Uden, the Netherlands), respectively. A goat Alexa Fluor 555‐conjugated secondary antibody for rabbit IgG and a goat Alexa Fluor 488‐conjugated secondary antibody for mouse IgG were purchased from Life Technologies Corp. Hoechst33342 was purchased from Dojindo Laboratories (Kumamoto, Japan). Normal goat serum was purchased from Sigma‐Aldrich (St. Louis, MO, USA). Tissue‐Tek optimal cutting temperature (OCT) compound was purchased from Sakura Finetek Japan (Tokyo, Japan). All other reagents not specified above were of the highest grade.

Human corneal epithelium T‐cell culture

A SV40‐immortalized HCE cell strain (HCE‐T cells, RCB no. 2280) was obtained from the RIKEN BioResource Center (Tsukuba, Japan). The cells were maintained in the following culture medium: 1 : 1 mixture of Dulbecco's modified eagle medium and nutrient mixture F‐12 supplemented with 5% heat‐inactivated fetal bovine serum, 5 µg ml–1 recombinant human insulin, 10 ng ml–1 recombinant human epidermal growth factor, 0.5% dimethyl sulfoxide 100 units ml–1 penicillin and 100 µg ml–1 streptomycin (Araki‐Sasaki et al., 1995; Yamasaki et al., 2009). Cells were grown at 37 °C in a humidified atmosphere of 5% CO2 in air.

Preparation of collagen vitrigel membrane chambers

A collagen xerogel membrane chamber (ad‐MED VitrigelTM) was purchased from Kanto Chemical Co., Inc. (Tokyo, Japan). The collagen xerogel membrane chamber was set in the well of a 12‐well plate. Then, the collagen xerogel membrane was immersed in the above 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.

Reconstruction of a human corneal epithelium model

A culture medium outside the chamber in the well of a 12‐well plate was changed to 1.5 ml of the fresh medium. The medium inside the chamber was removed and 0.5 ml of a cell suspension in a culture medium at a density of 1.2 × 105 cells ml–1 was poured on to the CVM of the chamber and cultured for 2 days at 37 °C. Subsequently, the medium inside the chamber was removed and the cells were cultured for 4 days under the air–liquid interface to fabricate a HCE model. The medium outside the chamber was changed every day or the third day during the culture period on the air–liquid interface to simplify the previous culture procedure (Yamaguchi et al., 2013).

Immunohistology of human corneal epithelium models after exposing test chemicals

The HCE models after exposing test chemicals were isolated from the plastic cylinder of the chamber and fixed for 5 min in methanol kept on ice immediately after sufficiently chilling it at –45 °C. Then, they were embedded in an OCT 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. The frozen sections were immersed in phosphate‐buffered saline (PBS) for 5 min to remove the OCT compound, and 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 MUC1 in PBS containing 1% normal goat serum were applied and incubated for 16 h, followed by washing them with PBS three times. Alexa fluor 555‐ or Alexa fluor 488‐conjugated secondary antibodies were applied and incubated for 3 h, followed by washing them with PBS three times. Subsequently, cell nuclei were counterstained with Hoechst33342. Sections were observed by a laser scanning confocal microscope (FV1000; Olympus, Tokyo, Japan).

Calculation of transepithelial electrical resistance values of a human corneal epithelium model

Each HCE model in a CVM chamber as a sample, a CVM chamber as a blank were subjected to the measurement of electrical resistance (R sample and R blank, respectively) by the method as previously described (Yamaguchi et al., 2013). The TEER value was calculated using the following formula:

Exposure experiment of test chemicals in the Vitrigel‐eye irritancy test method

Eighty‐eight test chemicals were selected according to the globally harmonized system of classification and labeling (GHS) classification for eye irritation (United Nations, 2013). The information on the total 118 test chemicals, including the previously tested 30 chemicals is shown in Table 1. Every test chemical solution was prepared in a culture medium at a concentration of 2.5 (weight/volume) % appropriate for measuring TEER values without being influenced by the test chemical‐dependent electrical resistance. Here, the chemicals were dissolved in the medium by using an appropriate technique(s) as follows: vortex mixing within 1 min, sonication within 20 min and/or heating in a water bath <70 °C. In case test chemicals are insoluble or immiscible by the above technique(s), the test chemical solution was prepared as a homogeneous suspension that the chemical was mixed well in the medium by vortex within 1 min immediately before use. The pH level of each 2.5 w/v % test chemical solution was measured using Universal pH test paper from ADVANTEC (Tokyo, Japan).

Table 1

List of the 118 test chemicals

Chemical	Class	CAS no.	Supplier	GHS classa	EPA classb	Draize score
Methoxyethyl acrylate	Acrylates	3121‐61‐7	Sigma‐Aldrich	1	> III	45c
Cyclohexanol	Alcohols	108‐93‐0	Sigma‐Aldrich	1	I	79.8c
2,5‐dimethyl‐2,5‐hexanediol	Alcohols	110‐03‐2	Sigma	1	I	28.3c
Diethylethanolamine	Amines	100‐37‐8	Sigma	1	I	94.7e
m‐phenylenediamine	Amines	108‐45‐2	Wako Pure	1	I	–
Acetic acid	Carboxylic acids	64‐19‐7	Wako Pure	1	I	68d, f
2‐methylbutanoic acid	Carboxylic acids	116‐53‐0	Sigma	1	I	–
Imidazole	Heterocyclics	288‐32‐4	Sigma	1	I	59.3c
Promethazine hydrochloride	Miscellaneous	58‐33‐3	Sigma	1	I	71.3c
Sodium salicylate	Organic salts	54‐21‐7	Wako Pure	1	I	83.7d
Lactic acid	Carboxylic acids	50‐21‐5	Alfa Aesar	1	I	102.7e
Pyridine	Heterocyclics	110‐86‐1	Sigma‐Aldrich	1	I	48c
Sodium hydroxide	Inorganic chemicals	1310‐73‐2	Wako Pure	1	I	108c, f
Potassium laurate	Surfactants (anionic)	10124‐65‐9	Wako Pure	1	I	33.7e, f
di(2‐ethylhexyl) sodium sulfosuccinate	Surfactants (anionic)	577‐11‐7	Sigma‐Aldrich	1	I	57d, f
Cetyltrimethylammonium bromide	Surfactants (cationic)	57‐09‐0	Sigma	1	I	96e, f
Stearyltrimethylammonium chloride	Surfactants (cationic)	112‐03‐8	Wako Pure	1	I	91.3d, f
Benzalkonium chloride	Surfactants (cationic)	8001‐54‐5	Sigma‐Aldrich	1	I	108c, f
Distearyldimethylammonium chloride	Surfactants (cationic)	107‐64‐2	Wako Pure	1	I	96.3d
Cetylpyridinium bromide	Surfactants (cationic)	140‐72‐7	TCI	1	I	89.7c, f
Domiphen bromide	Surfactants (cationic)	538‐71‐6	Sigma‐Aldrich	1	I	96.3e, f
Cetylpyridinium chloride	Surfactants (cationic)	6004‐24‐6	Sigma‐Aldrich	1	I	94.7d, f
Triton X‐100	Surfactants (non‐ionic)	9002‐93‐1	Sigma‐Aldrich	1	I	68.7c, f
Butyl cellosolve	Alcohols	111‐76‐2	Sigma	1	II	68.7c
Monoethanolamine	Alkanolamines	141‐43‐5	Sigma‐Aldrich	1	III	23.3d, f
Sodium lauryl sulfate	Surfactants (anionic)	151‐21‐3	Wako Pure	1	III	59.2c, g
3‐methyl‐pentynol	Alcohols	77‐75‐8	Sigma	1	–	–
Nonylphenyl‐polyethylene glycol	Polyols	9016‐45‐9	Wako Pure	1	–	–
Tetrahydrofuran	Furans	109‐99‐9	Sigma‐Aldrich	1	–	–
Benzethonium chloride	Surfactants (cationic)	121‐54‐0	Sigma‐Aldrich	1	–	67e
Benzyl alcohol	Alcohols	100‐51‐6	Sigma	2	I	31d
Acid red 92	Color additives	18472‐87‐2	Wako Pure	2	I	71d
Sucrose fatty acid ester	Polyols, esters	‐	TCI	2	II	28.3d
2‐ethoxyethyl acetate	Esters (acetate)	111‐15‐9	Sigma	2	III	15c
Glycolic acid	Carboxylic acids	79‐14‐1	Wako Pure	2	III	17.3e, f
Sodium 2‐naphthalenesulfonate	Organic salts	532‐02‐5	Sigma‐Aldrich	2	III	–
Diisopropanolamine	Alcohols	110‐97‐4	Sigma‐Aldrich	2	–	9.7e, f
Butanol	Alcohols	71‐36‐3	Wako Pure	2A	I	60.8c
Ethanol	Alcohols	64‐17‐5	Wako Pure	2A	I	24c
Isobutyl alcohol	Alcohols	78‐83‐1	Wako Pure	2A	I	60.3e
n‐hexanol	Alcohols	111‐27‐3	Aldrich	2A	II	64.8c
2‐ethyl‐1‐hexanol	Alcohols	104‐76‐7	Wako Pure	2A	II	51.3c
1‐octanol	Alcohols	111‐87‐5	Wako Pure	2A	II	41c
Cyclopentanol	Alcohols	96‐41‐3	Aldrich	2A	II	21.7d
2‐benzyloxyethanol	Alcohols, ethers	622‐08‐2	Wako Pure	2A	II	–
Methyl acetate	Esters	79‐20‐9	Sigma‐Aldrich	2A	II	39.5c
Methyl cyanoacetate	Esters, nitrile compounds	105‐34‐0	Sigma‐Aldrich	2A	II	27.7c
Butyrolactone	Lactone	96‐48‐0	Sigma‐Aldrich	2A	II	43c
Acetone	Ketones	67‐64‐1	Wako Pure	2A	II	65.8c
Isopropyl alcohol	Alcohols	67‐63‐0	Wako Pure	2A	III	30.5c
Myristyl alcohol	Fatty alcohols	112‐72‐1	Sigma‐Aldrich	2A	III	4c
Methyl ethyl ketone (2‐butanone)	Ketones	78‐93‐3	TCI	2A	III	50c
Hexyl cinnamic aldehyde	Aldehyde	101‐86‐0	Wako Pure	2A	–	–
Citric acid	Carboxylic acids	77‐92‐9	Sigma‐Aldrich	2A	–	–
Potassium sorbate	Organic salts	24634‐61‐5	Sigma‐Aldrich	2A	–	–
Calcium thioglycolate	Organic salts	814‐71‐1	Wako Pure	2A	–	52.3e
Propasol solvent P	Alcohols	1569‐01‐3	Sigma	2B	II	–
3,3'‐dithiodipropionic acid	Acids	1119‐62‐6	Wako Pure	2B	II	31.7c
2‐methyl‐1‐pentanol	Alcohols	105‐30‐6	TCI	2B	III	13c
n‐butanal	Aldehydes	123‐72‐8	Sigma	2B	III	–
Ethyl acetate	Esters (acetate)	141‐78‐6	Sigma	2B	III	18c
Camphene	Hydrocarbons	79‐92‐5	Sigma	2B	III	–
Isobutanal	Aldehyde	78‐84‐2	Wako Pure	2B	III	–
Di(propylene glycol) propyl ether	Alkoxylated alcohols	29911‐27‐1	Aldrich	2B	III	–
Ethyl‐2‐methylacetoacetate	Esters	609‐14‐3	Sigma‐Aldrich	2B	III	18c
Ethyl 2,6‐dichloro‐5‐fluoro‐beta‐oxo‐3‐pyridinepropanoate	Esters	96568‐04‐6	TCI	2B	III	–
3‐chloropropionitrile	Nitriles	542‐76‐7	Wako Pure	2B	III	13.7c
Ammonium nitrate	Organic salts	6484‐52‐2	Sigma	2B	III	18.3c
Sodium monochloroacetate	Organic salts, halogen compounds	3926‐62‐3	Aldrich	2B	III	–
n‐lauroylsarcosine sodium salt	Sarcosine derivatives	137‐16‐6	Sigma‐Aldrich	2B	III	–
6‐methylpurine	Bases	2004‐03‐7	Sigma‐Aldrich	2B	–	–
Xylene	Aromatics	1330‐20‐7	Wako Pure	NC	II	9c
Toluene	Hydrocarbons	108‐88‐3	Wako Pure	NC	III	9c
1,5‐hexadiene	Alkanes	592‐42‐7	Sigma	NC	III	4.7c
Triethanolamine	Alkanolamines	102‐71‐6	Wako Pure	NC	III	8d
N,N‐Dimethylguanidine sulfate	Organic salts	598‐65‐2	TCI	NC	III	6.7c
Styrene	Aromatics	100‐42‐5	Sigma‐Aldrich	NC	III	6.8c
Methyl cyclopentane	Cycloalkanes	96‐37‐7	TCI	NC	III	3.7c
Butyl acetate	Esters	123‐86‐4	Sigma‐Aldrich	NC	III	7.5c
Ethyl trimethyl acetate	Esters	3938‐95‐2	Sigma‐Aldrich	NC	III	3.8c
2,2‐dimethyl‐3‐pentanol	Fatty alcohols	3970‐62‐5	Sigma‐Aldrich	NC	III	8.3c
1,2,3‐trichloropropane	Hydrocarbons	96‐18‐4	Aldrich	NC	III	8.7c
Dodecane	Hydrocarbons	112‐40‐3	Sigma‐Aldrich	NC	III	2c
Methyl isobutyl ketone	Ketones	108‐10‐1	TCI	NC	III	4.8c
Methyl pentyl ketone	Ketones	110‐43‐0	Wako Pure	NC	III	–
Cyclohexanone	Ketones, hydrocarbons (cyclic)	108‐94‐1	Sigma‐Aldrich	NC	III	–
Tween20	Surfactants (non‐ionic)	9005‐64‐5	Sigma‐Aldrich	NC	III	4c
Polyoxyethylene 23 lauryl ether	Surfactants (non‐ionic)	9002‐92‐0	Sigma‐Aldrich	NC	III	0e
Dimethyl sulfoxide	Thioethers	67‐68‐5	Sigma‐Aldrich	NC	III	7.3e
2,4‐pentandiol	Alcohols	625‐69‐4	Sigma	NC	IV	1.3c
3‐methoxy‐1.2‐propanediol	Alcohols	623‐39‐2	TCI	NC	IV	0c
Isopropyl bromide	Hydrocarbons	75‐26‐3	Sigma	NC	IV	2.7c
n‐octyl bromide	Hydrocarbons	111‐83‐1	Sigma	NC	IV	0c
Gluconolactone	Lactone	90‐80‐2	TCI	NC	IV	2e
Glycerol	Polyols	56‐81‐5	Wako Pure	NC	IV	1.7c
Propylene glycol	Polyols	57‐55‐6	Wako Pure	NC	IV	1.3c
Polyethylene glycol 400	Polyols	25322‐68‐3	TCI	NC	IV	0c
Iso‐octyl acrylate	Acrylates	29590‐42‐9	Sigma‐Aldrich	NC	IV	0.7c
3,3‐Dimethylpentane	Alkanes	562‐49‐2	Aldrich	NC	IV	0c
1,9‐decadiene	Alkenes	1647‐16‐1	Sigma‐Aldrich	NC	IV	2c
Polyoxyethylene hydrogenated castoroil (60E.O.)	Alkoxylated alcohols, polymeric ethers	61788‐85‐0	Wako Pure	NC	IV	0d, f
1,3‐di‐isopropylbenzene	Aromatics	99‐62‐7	Sigma‐Aldrich	NC	IV	2c
Isopropyl myristate	Esters	110‐27‐0	Sigma‐Aldrich	NC	IV	0d
Ethylhexyl salicylate	Esters, ultraviolet absorbing agents	118‐60‐5	Sigma‐Aldrich	NC	IV	–
2‐methylpentane	Hydrocarbons	107‐83‐5	Sigma‐Aldrich	NC	IV	2c
Diisobutyl ketone	Ketones	108‐83‐8	Sigma‐Aldrich	NC	IV	0.7c
Tween80	Surfactants (non‐ionic)	9005‐65‐6	Sigma‐Aldrich	NC	IV	0e, f
2‐ethylhexyl p‐dimethyl‐amino benzoate	Ultraviolet absorbing agents	21245‐02‐3	Aldrich	NC	IV	0c
Cyclopentasiloxane	Silicon compounds	541‐02‐6	Sigma	NC	–	–
EDTA, di‐potassium	Amines	25102‐12‐9	Sigma‐Aldrich	NC	–	10.3e
Betaine monohydrate	Amino acids	590‐47‐6	Sigma‐Aldrich	NC	–	5.3e
1,2,4‐trimethylbenzene	Hydrocarbons	95‐63‐6	Sigma‐Aldrich	NC	–	4.7e
Petroleum ether	Hydrocarbons	8032‐32‐4	Sigma‐Aldrich	NC	–	2e
Hexane	Hydrocarbons	110‐54‐3	Sigma‐Aldrich	NC	–	0e
Silic anhydride	Inorganic chemicals	7631‐86‐9	Wako Pure	NC	–	2.7d
2,4‐pentanedione	Ketones	123‐54‐6	Sigma‐Aldrich	NC	–	14e
3‐glycidoxypropyltrimethoxysilane	Organosilicon compounds	2530‐83‐8	Sigma‐Aldrich	NC	–	2e
Polyethylene glycol monostearate (10E.O.)	Surfactants (non‐ionic)	9004‐99‐3	Wako Pure	NC	> III	0d

–, unknown; 1, category 1 (irreversible effects on the eye); 2, category 2 (irritating to eyes); 2A, category 2A (irritating to eyes); 2B, category 2B (mildly irritating to eyes); NC, not classified; I, category I (corrosive (irreversible destruction of cornea tissue) or corneal involvement or irritation persisting for more than 21 days); II, category II (corneal involvement or other eye irritation clearing in 8–21 days); III, category III (corneal involvement or other eye irritation clearing in 7 days or less); IV, category IV (minimal effects clearing in less than 24 h).

GHS category (United Nations, 2013).

EPA category (EPA, 1998).

ECETOC (1998).

Ohno et al. (1999).

Takahashi et al. (2011).

Data from 10% exposure condition.

Data from 15% exposure condition.

The HCE models on day 6 were subjected to the exposure experiment of test chemicals. At first, 500 µl of culture medium was poured in the chamber and the value of the R sample, before chemical exposures, was measured to obtain the initial TEER value of each model. Next, the medium inside the chamber was changed to 500 µl of test chemical solution and the periodical values of R sample were measured by the TEER recorder at intervals of 10 s for 3 min after exposure of each test solution. Three independent models were subjected to the exposure experiment for each test solution to plot the average time‐dependent profile of TEER values on a chart. The chemical exposure experiment was conducted in the ambient temperature of 28 ± 2 °C.

Eye irritant potential of test chemicals using the Vitrigel‐eye irritancy test method

The average time‐dependent profile of TEER values after exposing each test chemical solution in three‐independent experiments was analyzed by three indexes for time lag, intensity and plateau level. The score of each index was calculated by the formula as previously described (Yamaguchi et al., 2013). The eye irritant potential of test chemicals was classified into two categories, irritant and non‐irritant, according to the criteria for the scores of three indexes shown in Table 2.

Table 2

Criteria for the judgment using the Vitrigel‐eye irritancy test method

Judgment	Criteria
Irritant	Score of time lag ≤ 180 or score of intensity ≥ 0.05 or score of plateau level > 5
Non‐irritant	Score of time lag > 180 and score of intensity < 0.05 and score of plateau level ≤ 5

Subsequently, the correlation with the GHS classification of 118 test chemicals was estimated by calculating sensitivity, specificity and accuracy in accordance with the following formula. (The correlation with the US Environmental Protection Agency [EPA] classification of 98 test chemicals was also estimated in a similar manner except for 20 chemicals unknown as EPA classification.)

Here, A, B, C and D represent the number of chemicals categorized as irritant by both the traditional GHS or EPA classification and judgment of the Vitrigel‐EIT method, irritant by the traditional GHS or EPA classification and non‐irritant by judgment of the Vitrigel‐EIT method, non‐irritant by the traditional GHS or EPA classification and irritant by judgment of the Vitrigel‐EIT method, and non‐irritant by both the traditional GHS or EPA classification and judgment of the Vitrigel‐EIT method, respectively.

Results

Predictive performance of the Vitrigel‐eye irritancy test method

The final judgment of the 118 test chemicals using the Vitrigel‐EIT method and the pH level of each test chemical solution are shown in Table 3. Here, 80 test chemicals were classified as irritant and the other 38 chemicals as non‐irritant.

Table 3

Summary data of the results by Vitrigel‐eye irritancy test method for 118 test chemicals

Chemical	pH a	Scoreb
		Lag time	Intensity	Plateau level	Final judgment
Methoxyethyl acrylate	7	0	0.24	42.7	I
Cyclohexanol	7	0	0.31	56.0	I
2,5‐dimethyl‐2,5‐hexanediol	7	80	0.15	18.0	I
Diethylethanolamine	10	0	0.66	66.0	I
m‐phenylenediamine	8	10	0.35	62.0	I
Acetic acid	4	>180	0.00	–51.0	NI
2‐methylbutanoic acid	4	>180	0.02	–0.7	NI
Imidazole	9	100	0.26	22.0	I
Promethazine hydrochloride	6	0	0.69	69.0	I
Sodium salicylate	7	0	1.01	60.0	I
Lactic acid	3	>180	–0.30	0.0	NI
Pyridine	7	10	0.18	32.7	I
Sodium hydroxide	≥11	0	13.27	133.0	I
Potassium laurate	7	0	0.40	71.0	I
Di(2‐ethylhexyl) sodium sulfosuccinate	7	0	0.24	35.0	I
Cetyltrimethylammonium bromide	7	0	0.35	63.0	I
Stearyltrimethylammonium chloride	7	0	0.32	57.0	I
Benzalkonium chloride	7	0	1.00	90.0	I
Distearyldimethylammonium chloride	7	90	0.03	1.0	I
Cetylpyridinium bromide	7	0	1.16	81.0	I
Domiphen bromide	7	0	0.32	57.1	I
Cetylpyridinium chloride	7	0	0.33	59.0	I
Triton X‐100	7	0	0.92	83.0	I
Butyl cellosolve	8	0	0.48	58.0	I
Monoethanolamine	≥11	0	0.36	65.4	I
Sodium lauryl sulfate	7	0	0.70	84.0	I
3‐methyl‐pentynol	8	0	0.60	36.0	I
Nonylphenyl‐polyethylene glycol	7	40	0.20	30.1	I
Tetrahydrofuran	7	0	0.22	40.0	I
Benzethonium chloride	7	0	0.38	68.3	I
Benzyl alcohol	7	0	0.27	49.0	I
Acid red 92	‐	0	0.74	82.0	I
Sucrose fatty acid ester	7	0	0.24	42.0	I
2‐ethoxyethyl acetate	7	0	0.20	37.0	I
Glycolic acid	4	0	0.37	3.7	I
Sodium 2‐naphthalenesulfonate	7	0	0.54	59.0	I
Diisopropanolamine	9	0	0.24	43.1	I
Butanol	8	0	0.89	53.0	I
Ethanol	7	10	0.14	26.0	I
Isobutyl alcohol	7	10	0.24	44.1	I
n‐hexanol	7	0	0.33	59.0	I
2‐ethyl‐1‐hexanol	7	0	0.23	41.0	I
1‐octanol	7	10	0.16	29.0	I
Cyclopentanol	7	10	0.20	35.7	I
2‐benzyloxyethanol	7	10	0.32	57.7	I
Methyl acetate	7	10	0.16	29.2	I
Methyl cyanoacetate	7	20	0.07	14.0	I
Butyrolactone	7	60	0.11	16.1	I
Acetone	7	0	0.21	10.0	I
Isopropyl alcohol	7	0	0.30	27.0	I
Myristyl alcohol	7	>180	–0.03	0.0	NI
Methyl ethyl ketone (2‐butanone)	7	0	0.21	37.0	I
Hexyl cinnamic aldehyde	7	0	0.57	6.0	I
Citric acid	3	30	0.20	15.7	I
Potassium sorbate	7	0	0.72	21.7	I
Calcium thioglycolate	10	0	0.53	53.4	I
Propasol solvent P	8	0	0.38	57.0	I
3,3'‐dithiodipropionic acid	4	>180	–0.02	0.0	NI
2‐methyl‐1‐pentanol	7	0	0.77	46.0	I
n‐butanal	7	80	0.15	18.0	I
Ethyl acetate	8	0	0.29	52.0	I
Camphene	7	100	0.04	4.0	I
Isobutanal	6	30	0.17	27.5	I
Di(propylene glycol) propyl ether	7	0	0.22	40.4	I
Ethyl‐2‐methylacetoacetate	7	10	0.16	29.7	I
Ethyl 2,6‐dichloro‐5‐fluoro‐beta‐oxo‐3‐pyridinepropanoate	5	>180	0.00	0.0	NI
3‐chloropropionitrile	5	10	0.31	56.0	I
Ammonium nitrate	8	0	2.07	62.0	I
Sodium monochloroacetate	7	0	0.78	31.4	I
n‐lauroylsarcosine sodium salt	6	0	0.35	63.6	I
6‐methylpurine	7	>180	–0.05	0.0	NI
Xylene	7	>180	–0.01	0.0	NI
Toluene	7	140	0.02	0.0	I
1,5‐hexadiene	7	>180	–0.01	0.0	NI
Triethanolamine	9	0	0.17	31.0	I
N,N‐Dimethylguanidine sulfate	7	0	1.35	41.0	I
Styrene	7	>180	–0.01	0.0	NI
Methyl cyclopentane	7	>180	–0.02	0.0	NI
Butyl acetate	7	10	0.15	26.3	I
Ethyl trimethyl acetate	7	110	0.07	7.2	I
2,2‐dimethyl‐3‐pentanol	7	10	0.20	35.7	I
1,2,3‐trichloropropane	7	80	0.11	14.2	I
Dodecane	7	>180	–0.02	0.0	NI
Methyl isobutyl ketone	7	0	0.25	32.0	I
Methyl pentyl ketone	7	50	0.20	9.0	I
Cyclohexanone	7	10	0.26	48.0	I
Tween 20	7	>180	–0.03	0.0	NI
Polyoxyethylene 23 lauryl ether	7	>180	–0.01	0.0	NI
Dimethyl sulfoxide	7	>180	–0.11	0.0	NI
2,4‐pentandiol	8	70	0.07	8.0	I
3‐methoxy‐1,2‐propanediol	7	>180	–0.10	0.0	NI
Isopropyl bromide	8	>180	0.00	0.0	NI
n‐octyl bromide	8	>180	–0.05	0.0	NI
Gluconolactone	6	>180	0.00	0.0	NI
Glycerol	7	0	0.34	20.4	I
Propylene glycol	7	>180	0.00	0.0	NI
Polyethylene glycol 400	7	>180	–0.01	2.0	NI
Iso‐octyl acrylate	7	>180	–0.02	0.0	NI
3,3‐dimethylpentane	7	>180	–0.02	2.0	NI
1,9‐decadiene	7	>180	–0.01	0.0	NI
Polyoxyethylene hydrogenated castoroil (60E.O.)	7	>180	–0.01	0.0	NI
1,3‐di‐isopropylbenzene	7	>180	0.00	0.0	NI
Isopropyl myristate	7	>180	0.00	0.0	NI
Ethylhexyl salicylate	7	>180	–0.02	0.0	NI
2‐methylpentane	7	>180	–0.01	0.0	NI
Diisobutyl ketone	7	>180	0.00	0.0	NI
Tween 80	7	>180	–0.02	0.0	NI
2‐ethylhexyl p‐dimethyl‐amino benzoate	7	>180	–0.04	0.0	NI
Cyclopentasiloxane	8	>180	–0.02	0.0	NI
EDTA, di‐potassium	5	0	0.38	37.8	I
Betaine monohydrate	7	0	0.26	21.1	I
1,2,4‐trimethylbenzene	7	>180	–0.01	0.0	NI
Petroleum ether	7	>180	–0.03	0.0	NI
Hexane	7	>180	–0.01	0.0	NI
Silic anhydride	7	>180	–0.05	0.0	NI
2,4‐pentanedione	6	10	0.15	28.2	I
3‐glycidoxypropyltrimethoxysilane	7	0	0.11	20.2	I
Polyethylene glycol monostearate (10E.O.)	7	>180	–0.02	0.0	NI

–, not tested; I, irritant; NI, non‐irritant.

pH 2.5 (w/v)% test chemical solution.

These scores were calculated from the average time‐dependent profile of transepithelial electrical resistance values in three independent experiments.

The classification by this test method was in accordance with the GHS categories on 95 test chemicals in a total of 118 chemicals. Meanwhile, seven chemicals were predicted as non‐irritant among the 71 irritant chemicals classified in categories 1, 2, 2A and 2B by the GHS, indicating a 9.9% false‐negative rate. In addition, 16 chemicals were predicted as irritant among the 47 non‐irritant chemicals using the GHS classification, indicating a 34.1% false‐positive rate. Therefore, the sensitivity, specificity and accuracy were 90.1%, 65.9% and 80.5%, respectively.

In addition, the classification using this test method was in accordance with the EPA categories on 82 test chemicals in a total of 98 chemicals. Meanwhile, 14 chemicals were predicted as non‐irritant among the 79 irritant chemicals classified in to categories I, II and III by the EPA, indicating a 17.7% false‐negative rate. In addition, two chemicals were predicted as irritant among the 19 non‐irritant chemicals using the EPA classification, indicating a 10.5% false‐positive rate. Therefore, the sensitivity, specificity and accuracy were 82.3%, 89.5% and 83.7%, respectively.

Distribution of pH levels in test chemical solutions

The pH levels of 118 test chemical solutions could be measured except for one solution that was inappropriate for the pH test paper and found distributed in the wide range from 3 to more than 11 as shown in Table 3. Here, in the nine chemical solutions falling under the pH ranges of less than 5, four chemicals were irritant whereas five chemicals were non‐irritant and false‐negative by the Vitrigel‐EIT. The other two false‐negative chemicals judged as non‐irritant by the Vitrigel‐EIT were pH 7 in their solution.

Immunohistological characteristics of human corneal epithelium models after exposing test chemicals

In the HCE model after exposing polyoxyethylene 23 lauryl ether, Tween 80 or polyoxyethylene hydrogenated castor oil (60E.O.), the time‐dependent relative changes of TEER values were almost nothing (Fig. 1A–C). Therefore, these chemicals were all classified as non‐irritants using the Vitrigel‐EIT method. Here, ZO‐1 was abundantly expressed in the lateral and basal surfaces of cells in the superficial layer in comparison to the other layers (Fig. 1D–F). MUC1 was merely expressed in the apical surface of cells in the superficial layer (Fig. 1G–I). These immunohistological observations demonstrated that the HCE models maintained healthy morphology even after exposing the test chemicals classified in to category NC by the GHS and classified as non‐irritant by the Vitrigel‐EIT method.

Figure 1

The average time‐dependent profile of TEER values in three independent experiments and immunohistological characteristics of human corneal epithelium models using test chemicals that were classified in category NC by the globally harmonized system of classification and labeling and classified as a non‐irritant using the Vitrigel‐eye irritancy test method. The average time‐dependent profile of TEER values after exposing polyoxyethylene 23 lauryl ether (A), Tween 80 (B) and polyoxyethylene hydrogenated castor oil (60E.O.) (C). Cross‐sections of the human corneal epithelium models after exposing polyoxyethylene 23 lauryl ether (D,G), Tween 80 (E,H) and polyoxyethylene hydrogenated castor oil (60E.O.) (F,I) were stained with antibodies for ZO‐1 (D–F) and MUC1 (G–I). Nuclei of cells were stained with Hoechst 33342. Scale bars represent 50 µm. TEER, transepithelial electrical resistance.

In the HCE model after exposing potassium laurate, butanol or propanol solvent P, the time‐dependent relative changes of TEER values were rapidly decreased (Fig. 2A–C). Therefore, these chemicals were all classified into irritant by the Vitrigel‐EIT method. Here, ZO‐1 and MUC1 expressions were remarkably disappeared in the HCE models (Fig. 2D–I). These immunohistological characteristics demonstrated that the HCE models lost their barrier function after exposing the test chemicals that were classified in to category 1, 2A and 2B by the GHS and classified as irritant by the Vitrigel‐EIT method.

Figure 2

The average time‐dependent profile of TEER values in three independent experiments and immunohistological characteristics of human corneal epithelium models using test chemicals that were classified as category 1, 2A and 2B using the globally harmonized system of classification and labeling and classified as an irritant by the Vitrigel‐eye irritancy test method. The average time‐dependent profile of TEER values after exposing potassium laurate (A), butanol (B) and propasol solvent P (C). Cross‐sections of the human corneal epithelium models after exposing potassium laurate (D,G), butanol (E,H) and propanol solvent P (F,I) were stained with antibodies for ZO‐1 (D–F) and MUC1 (G–I). Nuclei of cells were stained with Hoechst 33342. Scale bars represent 50 µm. TEER, transepithelial electrical resistance.

Meanwhile, in the HCE model after exposing triethanolamine, methyl isobutyl ketone or glycerol, the time‐dependent relative changes of the TEER values were slowly decreased (Fig. 3A–C). Therefore, these chemicals were all classified into irritant by the Vitrigel‐EIT method. Here, the ZO‐1 and MUC1 expressions were partially disappeared in the HCE models (Fig. 3D–I). These immunohistological characteristics demonstrated that the HCE models lost their barrier function after exposing the false‐positive test chemicals classified in to category NC by GHS whereas they were classified as irritant by the Vitrigel‐EIT method.

Figure 3

The average time‐dependent profile of TEER values in three independent experiments and immunohistological characteristics of the human corneal epithelium models using test chemicals that were classified as category NC using the globally harmonized system of classification and labeling, whereas they were classified as an irritant using the Vitrigel‐eye irritancy test method. The average time‐dependent profile of TEER values after exposing triethanolamine (A), methyl isobutyl ketone (B) and glycerol (C). Cross‐sections of the human corneal epithelium models after exposing triethanolamine (D,G), methyl isobutyl ketone (E,H) and glycerol (F,I) were stained with antibodies for ZO‐1 (D–F) and MUC1 (G–I). Nuclei of cells were stained with Hoechst 33342. Scale bars represent 50 µm. TEER, transepithelial electrical resistance.

Discussion

We developed the Vitrigel‐EIT method by measuring the time‐dependent profiles of the TEER values for 3 min after exposing 30 test chemicals as previously reported (Yamaguchi et al., 2013). However, generally more than 100 chemicals should be tested for estimating the predictive performance of an EIT method. In this study, we tested a total of 118 test chemicals with a variety of physical and chemical properties. Consequently, this test method indicated the good predictive performance comparable to other test methods currently in development aiming for the OECD test guidelines. For example, the sensitivity, specificity and accuracy of the EpiOcular‐EIT were 98.1%, 72.9% and 84.8%, respectively (Kaluzhny et al., 2011). In addition, the sensitivity, specificity and accuracy of the short time exposure test in a bottom–up approach were 88%, 80% and 85%, respectively (ICCVAM, 2013). However, some chemicals were classified as false‐negatives or false‐positives. It is important to clarify the mechanism‐based reason for raising false‐negative and false‐positive reactions, particularly for establishing an in vitro test method that can truly extrapolate an in vivo reaction after exposing a chemical. To overcome this issue, we tried to clarify the mechanism for raising false‐negative and false‐positive reactions.

Regarding the false‐negative reactions, five of the seven false‐negative chemicals were acidic and their 2.5 (w/v)% solutions for exposure experiments indicated the pH level lower than 5. In addition, the TEER values of the HCE models after exposing the five acidic false‐negative chemical solutions were increased from the initial TEER values. Interestingly, it was reported that isolated rabbit esophageal mucosal epithelium and normal human bronchial epithelial cell layers in culture increased their TEER values when they were exposed to weak acidic solutions (Farré et al., 2008; Oshima et al., 2012). In the case of the nine test chemicals showing pH 5 or lower in their exposure solutions that were excluded from total 118 chemicals, the sensitivity, specificity and accuracy on the 109 chemicals were improved from 90.1%, 65.9% and 80.5% to 96.8%, 67.4% and 84.4%, respectively. One of two non‐acidic false‐negative chemicals was myristyl alcohol. Myristyl alcohol is a rugged waxy solid and water‐insoluble at room temperature. In the Draize‐EIT, solid chemicals have the potential to injure eyes due to their mechanical stress inducing scratching on living tissues (Kaluzhny et al., 2011; Takahashi et al., 2011; York and Steiling, 1998). In the Vitrigel‐EIT method, solid chemicals were applied to the HCE models as its homogeneous suspension and kept stationary for 3 min without stirring or shaking. Therefore, one speculation for the false‐negative result on myristyl alcohol is the deference of test conditions between the Draize‐EIT and the Vitrigel‐EIT method. Another non‐acidic false‐negative chemical was 6‐methylpurine. It is a powdery solid and water‐soluble. We are currently investigating the reason for the false‐negative judgment with 6‐methylpurine; however, it is still unknown at present.

Regarding the false‐positive reactions, Draize scores provided an interesting viewpoint. The average of Draize scores for 14 false‐positive chemicals and that for 27 non‐irritant chemicals revealed 6.5 and 2.2 except for six chemicals unknown as Draize scores, respectively. These values suggest that the eye irritant potential of false‐positive chemicals is stronger than that of non‐irritant chemicals. Moreover, one of the 16 false‐positive chemicals and eight of those were classified into categories II and III by the EPA, respectively. Interestingly, it was reported that about 30% chemicals classified into category III by the EPA were labeled as not‐classified by GHS (ICCVAM, 2010). The high rate of reducing eye irritant chemicals by using the criteria of GHS compared to that of EPA is attributable to the difference in the classification system between GHS and EPA as briefly mentioned below. In the classification system, each chemical was tested using at least three animals in both the GHS and EPA. The eye irritancy of test chemicals in animals was estimated using the criteria for four indexes, i.e., corneal opacity, conjunctival redness, swelling and iritis. Here, each index has three endpoints for the time, i.e., 24, 48 and 72 h after the test chemical administration. In the case of the GHS classification, test chemicals were evaluated in two steps, i.e., each test chemical was first categorized according to the value of the three endpoints for the four indexes in each animal, and next the category involving the over half result in all tested animals was finally judged as the classification of the test chemical (United Nations, 2013). On the other hand, in the case of the EPA classification, test chemicals were evaluated in one step, i.e., the classification of each test chemical was determined based on the maximum score for each endpoint in any animal (EPA, 1998).

The HCE model appropriate for the Vitrigel‐EIT method possessed about six cell layers, expressed the HCE‐related proteins and developed the barrier function, suggesting the model well reflects HCE in vivo (Yamaguchi et al., 2013). In particular, ZO‐1 is one of the tight junction‐related proteins and associated with the principal barrier that separates the eye from the outside environment (Yi et al., 2000). In addition, MUC1 is one of the cell membrane spanning mucin families expressed in the superficial layer of corneal epithelium and plays a protective role against the adherence of pathogens (Song and Joo, 2004). Those reports suggest that ZO‐1 and MUC1 expressions are essential for maintaining healthy corneal epithelium.

In the current study on immunohistological analyses, ZO‐1 and MUC1 expressions in HCE models were maintained after exposing chemicals judged as non‐irritant (Fig. 1D–I) and disappeared after exposing chemicals judged as irritants (Fig. 2D–I) by the GHS classification and the Vitrigel‐EIT method. By contrast, those expressions disappeared in the model after exposing chemicals judged as non‐irritant by the GHS classification and as irritant by the Vitrigel‐EIT method, i.e., false‐positive chemicals (Fig. 3D–I). These data demonstrated that such false‐positive chemicals induced the unhealthy conditions for the HCE models, suggesting that the chemicals have an eye irritant potential.

In addition, gluconolactone is classified as a non‐irritant chemical using the GHS. However, gluconolactone is hydrolyzed by water to form gluconic acid, which is a severe eye irritant chemical. In case the test chemical solution of gluconolactone is left for more than 6 min before exposing it to HCE models, it was judged as irritant by the Vitrigel‐EIT method due to the coexistence of gluconic acid (data not shown). This suggests that such a test chemical solution revealing hydrolyzability should be prepared immediately before the chemical exposure experiment.

In this study, we demonstrated that the Vitrigel‐EIT method could estimate the widespread eye irritancy of various chemicals with very few false‐negatives. The validation study of the Vitrigel‐EIT method has been conducted by the international validation management team organized in association with the International Collaboration on Alternative Test Methods. Hereafter, we intend to decide the applicability domain of the Vitrigel‐EIT method and to describe it in the final protocol for the validation report. We hope that such an effort for registering the Vitrigel‐EIT method as a new test guideline for the OECD contributes to the development of safe commodity chemicals.

Conflict of interest

Mr. Yamaguchi and Dr. Takezawa have a patent Cell culture chamber, method for producing same, tissue model using cell culture chamber, and method for producing same issued.