An in vitro skin sensitization assay termed EpiSensA for broad sets of chemicals including lipophilic chemicals and pre/pro-haptens

Kazutoshi Saito ⁎, Osamu Takenouchi, Yuko Nukada, Masaaki Miyazawa, Hitoshi Sakaguchi
Kao Corporation, R&D, Safety Science Research, 2606 Akabane Ichikai-Machi Haga-Gun, Tochigi 321-3497, Japan

Article history: Received 25 July 2016
Received in revised form 7 December 2016 Accepted 8 December 2016
Available online 10 December 2016
To evaluate chemicals (e.g. lipophilic chemicals, pre/pro-haptens) that are difficult to correctly evaluate using in vitro skin sensitization tests (e.g. DPRA, KeratinoSens or h-CLAT), we developed a novel in vitro test termed “Epi- dermal Sensitization Assay: EpiSensA” that uses reconstructed human epidermis. This assay is based on the in- duction of multiple marker genes (ATF3, IL-8, DNAJB4 and GCLM) related to two keratinocyte responses (inflammatory or cytoprotective) in the induction of skin sensitization. Here, we first confirmed the mechanistic

Skin sensitization Alternative method
Reconstructed human epidermis (RhE) model Gene expression
Lipophilic chemicals Pre/pro-haptens
relevance of these marker genes by focusing on key molecules that regulate keratinocyte responses in vivo (P2X7 for infl ammatory and Nrf2 for cytoprotective responses). The up-regulation of ATF3 and IL-8, or DNAJB4 and GCLM induced by the representative sensitizer 2,4-dinitrochlorobenzene in human keratinocytes was signifi- cantly suppressed by a P2X7 specific antagonist KN-62, or by Nrf2 siRNA, respectively, which supported mecha- nistic relevance of marker genes. Moreover, the EpiSensA had sensitivity, specificity and accuracy of 93%, 100% and 93% for 29 lipophilic chemicals (logKow ≥ 3.5), and of 96%, 75% and 88% for 43 hydrophilic chemicals includ- ing 11 pre/pro-haptens, compared with the LLNA. These results suggested that the EpiSensA could be a mecha- nism-based test applicable to broad sets of chemicals including lipophilic chemicals and pre/pro-haptens.


Allergic contact dermatitis (ACD) resulting from skin sensitization is a common occupational and environmental health problem (Peiser et al., 2010). To date, the evaluation of skin sensitizers has been mainly de- pendent on animal tests such as the local lymph node assay (LLNA) (Kimber et al., 2002). However, the ethical issues involved, and the European Union ban on animal testing for cosmetic ingredients (Cos- metics Regulation EC 1223/2009) have accelerated the development of non-animal tests for evaluating the skin sensitization potential of chemicals.
ACD is an adaptive immune response caused by skin sensitizers (haptens). Skin sensitizers that penetrate through the skin form hapten–protein complexes (haptenated proteins) and are captured by skin-resident dendritic cells (DCs) (Sasaki and Aiba, 2007). These DCs mature and migrate to skin draining lymph nodes, where the DCs present the peptides from haptenated proteins to naive T cells. Finally, antigen-specific T cells proliferate and disseminate into the peripheral circulation (Kimber et al., 2011). Repeated exposure to the same skin sensitizers after induction of sensitization can lead to the elicitation and symptoms of ACD (Toebak et al., 2009).

Abbreviations: ACD, allergic contact dermatitis; AOO, acetone:olive oil = 4:1; AOP, Adverse Outcome Pathway; ARE, antioxidant response element; ATF3, activating transcription factor 3; BKC, benzalkonium chloride; DCs, dendritic cells; DMSO, dimethyl sulfoxide; DNAJB4, dnaJ (Hsp40) homolog, subfamily B, member 4; DNCB, 2,4- dinitrochlorobenzene; DPRA, Direct Peptide Reactivity assay; DW, distilled water; ECVAM, European Centre for the Validation of Alternative Methods; EHA, 2-ethylhexyl acrylate; EpiSensA, Epidermal Sensitization Assay; GCLM, glutamate-cysteine ligase, modifier subunit; h-CLAT, human-Cell Line Activation Test; IL, interleukin; IM, isopropyl myristate; KCs, keratinocytes; LA, lactic acid; LLNA, local lymph node assay; MBT, 2- mercptbenzothiazole; MS, methyl salicylate; MTT, methylthiazolydiphenyl-tetrazolium; 4-NBB, 4-nitrobenzyl bromide; NHEK, normal human epidermal keratinocyte; Nrf2, nuclear factor E2-related factor 2; OA, octanoic acid; OXA, oxazolone; PPD, p- phenylenediamine; RhE, reconstructed human epidermis; SA, squalic acid; SLS, sodium lauryl sulfate.
⁎ Corresponding author.
E-mail address: [email protected] (K. Saito).
0887-2333/© 2016 Elsevier Ltd. All rights reserved.
To date, a number of in vitro tests that focus on key events in the skin sensitization Adverse Outcome Pathway (AOP) (OECD, 2012a) have been developed. Among these tests, the Direct Peptide Reactiv- ity assay (DPRA), which focuses on the first key event (molecular inter- action with skin proteins), and the KeratinoSens™ assay, which is based on the second key event (keratinocyte response), were adopted as OECD Test Guidelines in February 2015 (OECD, 2015a, 2015b). Addi- tionally, the human-Cell Line Activation Test (h-CLAT), which focuses on the third key event (dendritic cell activation), was approved by the OECD Working Group of the National Coordinators of the Test Guide- lines Programme in April 2016 (ECVAM, 2016). It has been reported that these three tests show approximately 80% accuracy compared to the LLNA for over 140 chemicals (Takenouchi et al., 2015; OECD,2015a, 2015b). However, since all of these tests are performed in an aqueous system (buffer or culture medium), it is often difficult to cor- rectly evaluate substances with high lipophilicity (i.e. with logKow N 3) (OECD, 2012b) due to their precipitation or phase separation (Takenouchi et al., 2013; Natsch et al., 2013). It is also diffi cult to evaluate pre/pro-haptens, which form protein reactive haptens through either air-oxidation or metabolic conversion (Aptula et al., 2007), due to the limited metabolic systems in the assays (OECD, 2015a, 2015b and OECD, 2015C). Thus, to fully replace in vivo skin sensitization tests, a novel test that can evaluate lipophilic chemicals and pre/pro-haptens is necessary.
To overcome the above assay limitations, a three dimensional recon- structed human epidermis model (RhE model), which consists of nor- mal human-derived epidermal keratinocytes (NHEKs), could be a useful tool. Since RhE models are cultured at an air–liquid interface, test chemicals are directly applicable to RhE models in a manner similar to real skin, suggesting that lipophilic chemicals could be applied and tested correctly. Moreover, it has been reported that RhE models exhibit metabolic capability similar to that of human skin (Luu-The et al., 2009; Götz et al., 2012a, 2012b; Eilstein et al., 2014), suggesting that RhE models would successfully metabolize pre/pro-haptens for correct eval- uation. On the other hand, mechanistically relevant biomarkers that re- flect keratinocyte responses in the early phase of skin sensitization are necessary in RhE-based assays. The two main important keratinocyte responses that occur in the skin sensitization AOP are; 1) Inflammatory response (e.g. activation of inflammatory cytokines) and 2) Induction of cytoprotective gene pathways (e.g. antioxidant/electrophile response element (ARE/EpRE)-dependent pathways) (OECD, 2012a). Key mole- cules that regulate the two pathways, P2X7, a purinergic receptor in- volved in the production of pro-inflammatory cytokines such as IL-1β and IL-18 (Kaplan et al., 2012), and nuclear factor E2-related factor 2 (Nrf2), a transcription factor that is involved in the activation of antiox- idant and cytoprotective genes by binding to ARE (Uruno and Motohashi, 2011), have been reported to modulate skin sensitization in vivo either positively (P2X7) or negatively (Nrf2) (Weber et al., 2010; et al. et al., 2013).
Based on these data, some RhE-based skin sensitization assays that are focused on the skin sensitization AOP have been developed and reported by several research groups: The Epidermal equivalent (EE) assay is based on the production of infl ammatory cytokines such as IL-18 in multiple RhE models such as epiCS® (Gibbs et al., 2013); the SenCeeTox® assay is mainly based on eight Nrf2-dependent genes in the EpiDerm™ and SkinEthicRHE assays (McKim et al., 2012); and the SENS-IS assay focuses on 24 Nrf2-dependent genes and 41 ad- ditional genes (SENS-IS genes) in the EpiSkin assay (Cottrez et al., 2016). The biomarkers of these assays such as IL-18 and Nrf2 dependent genes are closely related to keratinocyte responses in the skin sensitiza- tion AOP (OECD, 2012a), which strongly supported the idea that these assays could be mechanistically relevant for evaluation of skin sensitiza- tion. However, the EE assay uses only one biomarker, IL-18, which is re- lated to inflammatory responses, and the SenCeeTox® assay relies on Nrf2-dependent genes that are related to cytoprotective responses, im- plying that these assays have lower sensitivity for sensitizers that can- not induce IL-18 production or activate the Nrf2 pathway. Also, the capability of these two assays to predict the skin sensitization potential of lipophilic chemicals as well as of pre/pro-haptens has yet to be fully assessed. On the other hand, the SENS-IS assay can detect both inflam- matory and cytoprotective responses by using 65 marker genes and showed 100% accuracy for 150 chemicals when compared to the LLNA (Cottrez et al., 2016). However, the use of large sets of genes would complicate an assay system (e.g. procedure and prediction) and lacks transparency in the rationale used for interpretation for a prediction.
We have developed a novel skin sensitization assay, termed “EpiSensA”, that is based on the expression of marker genes related to cellular stress responses in an RhE model; EpiDerm™ (Saito et al., 2013a). Our previous study (Saito et al., 2013a) demonstrated that three marker genes, ATF3, DNAJB4 and GCLM, in the EpiSensA could correctly predict the skin sensitization potential of 16 reference chemicals recommended by the European Centre for the Validation of Alternative Methods (ECVAM) (Casati et al., 2009). The 16 chemicals in- cluded four pre/pro-haptens (e.g. eugenol), suggesting that these three marker genes would be useful for evaluation of the skin sensitization potential of chemicals including pre/pro-haptens. Of the three marker genes in the EpiSensA, GCLM has been reported to be regulated through the Nrf2/ARE pathway in the HaCaT human keratinocyte cell line (MacLeod et al., 2009), suggesting that it is a mechanistically relevant marker for the prediction of skin sensitization. Also, it has been sug- gested that DNAJB4 might be related to a “cytoprotective response” based on its original function (i.e. a molecular chaperone) (Qiu et al., 2006), whereas up-regulation of DNAJB4 by sensitizers has been report- ed to be independent of Nrf2 (Emter et al., 2013). On the other hand, ATF3 might be involved in inflammatory pathways based on the previ- ous finding that the expression of ATF3 in NHEKs was up-regulated by ATP (Ohara et al., 2010), which is sensed through P2X7 and which mod- ulates the “inflammatory pathway” in the early phase of skin sensitiza- tion (Kaplan et al., 2012). Based on the combined data, it is expected that the EpiSensA would have higher sensitivity in predicting skin sen- sitization due to the fact that it has gene markers that reflect two differ- ent keratinocyte responses (infl ammatory and cytoprotective). However, there has been no direct evidence that ATF3 and DNAJB4 are involved in keratinocyte responses in the skin sensitization AOP.
The activation of infl ammatory cytokines has also been shown to play an important part in the inflammatory response of keratinocytes (OECD, 2012a). An earlier study that used several sensitizers and irri- tants suggested that IL-8, which is a chemokine that promotes the re- cruitment of polymorphonuclear leukocytes in the initiation of cutaneous inflammation (Barker et al., 1991), is an indicator of skin sen- sitization in a RhE model (Coquette et al., 2003). Recently, several inves- tigators demonstrated that IL-8 could be a good skin sensitization marker in human keratinocytes (McKim et al., 2010; Bae et al., 2015). Although all of the above studies used dozens of test chemicals with rel- atively high water solubility, it is tempting to hypothesize that IL-8 would also be a good biomarker of skin sensitization for lipophilic chemicals in an RhE model.
In this study, in order to further support the mechanistic relevance of the EpiSensA marker genes to predict skin sensitization, we first deter- mined whether the up-regulation of four marker genes (ATF3, IL-8, DNAJB4, GCLM) in NHEKs is regulated by two important molecules, P2X7 and Nrf2, that are involved in key keratinocyte responses of in- fl ammation and cytoprotection, respectively. We then examined the utility of the four marker genes for predicting the skin sensitization po- tential and potency of 29 chemicals with relatively high lipophilicity (logKow ≥ 3.5) as well as of 43 chemicals with relatively low lipophilic- ity (logKow b 3.5) including 11 pre/pro-haptens in a RhE model to as- sess whether the EpiSensA can overcome the limitations of existing in vitro tests.

2.Materials and methods
2.1. Chemicals and sample preparation
For Assessment of the regulation mechanism of marker genes in NHEKs (Section 2.2), a representative sensitizer 2,4-dinitrochlorobenzene (DNCB) purchased from Sigma-Aldrich (St. Louis, MO, USA) was evalu- ated. DNCB was dissolved in dimethyl sulfoxide (DMSO; Sigma- Aldrich). For Assessment of the predictive performance of marker genes (Section 2.3), a total of 72 test chemicals that are shown in Table 1 were selected based on their skin sensitizing properties reported in the literature (Gerberick et al., 1992; Gerberick et al., 2005; Kimber et al., 1998 and Kern et al., 2010), their diverse chemical structures, and their availability from commercial sources. This collection includes 29 chemicals with logKow ≥ 3.5 (“lipophilic chemicals” in this study),which have been reported to show a relatively high false negative ratio in the h-CLAT (Takenouchi et al., 2013), as well as 43 chemicals with logKow b 3.5 (“hydrophilic chemicals” in this study). In addition, 11 putative pre/pro-haptens were also included based on the previ- ous literature referred to in Urbisch et al., 2015. Most of the tested chemicals were purchased from Sigma–Aldrich, except for tetrachlorosalicylanilide (TCSA; Kanto Chemical Co. Inc., Tokyo, Japan), methyldibromoglutaronitrile (Alfa Aesar Ward Hill, MA, USA), lillial and undecylenic acid (Wako Pure Chemicals, Osaka, Japan), and 2-ethylhexyl acrylate (EHA; Nacalai Tesque, Kyoto, Japan). Tested chemicals were dissolved in AOO (acetone (Sigma- Aldrich):olive oil (Kanto Chemical) = 4:1), distilled water (DW; Otsuka Pharmaceutical Factory, Inc., Tokyo, Japan) or 50% ethanol in DW (50% EtOH; ethanol was purchased from Kanto Chemical) when applied to the RhE model, since all of these vehicles have often been adopted in dermal application in animal testings and failed to affect cytotoxicity under the testing condition used in this study. The CAS No., LLNA Estimated Concentration of a test substance needed to produce a stimulation index of three (EC3) values (%) (OECD, 2010), logKow calculated by KOWWIN ver.1.68 in EPI suite™ (Environmental Protection Agency, Washington, DC, USA), and vehicle for each chemical are listed in Table 1.
2.2.Assessment of the regulation mechanism of marker genes in NHEKs

2.2.1.Cell culture
NHEKs (KURABO, Osaka, Japan) were cultured in HuMedia–KG2 (KURABO) supplemented with insulin, bovine pituitary extract, epider- mal growth factor, hydrocortisone, kanamycin, and amphotericin B. When used for experiments, cells were seeded in 12-well plates at a density of 1.0 × 105 cells per well in 1 mL of culture medium.

2.2.2.siRNA experiment
The transfection mix was prepared by mixing 940 μL of culture me- dium, 0.2 μL of 10 μM control siRNA (sc-37007, Santa Cruz Biotech, Santa Cruz, CA, US) or Nrf2-siRNA (sc-37030, Santa Cruz Biotech) (final con- centration, 5 nM) and 60 μL of Hiperfect transfection reagent (Qiagen, Valencia, CA, USA). This transfection mix was incubated at room tem- perature for 10 min. Subsequently, the cell culture medium of NHEKs that were pre-seeded in 12-well plates overnight was exchanged for 1 mL of the transfection mix. Twenty four hours later, the cells were washed with 1 mL of HEPES buffer (Kurabo) and exposed to vehicle (DMSO; final concentration, 0.2%) or DNCB (2.2 μg/mL) in 1 mL of cul- ture medium for 6 h in triplicate.

2.2.3.P2X7 receptor blocking experiment
For blocking the P2X7 receptor, KN-62 (Santa Cruz Biotech), a human selective P2X7 antagonist (Anderson and Nedergaard, 2006) was used. KN-62 dissolved in DMSO was diluted in culture medium to a concentration of 30 μM and was then applied to NHEK cells in 12- well plates. Fifteen minutes later, the cells were washed with 1 mL of HEPES buffer and exposed to the vehicle (DMSO; final concentration, 0.2%) or DNCB (2.2 μg/mL) in 1 mL of culture medium for 6 h in triplicate.

2.2.4.RNA isolation
Following 6 h of treatment with chemicals, the cells were lysed with RLT buffer (Qiagen) containing 2.5% DL-Dithiothreitol (Sigma-Aldrich). The RNA was purified using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions and was then quantified using an ND- 1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and stored at – 80 °C until use.

2.2.5.cDNA synthesis
The Superscript™ III First-Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad, CA, USA) was used to prepare cDNA. The mixture consisted of 1 μL of a 10 mM dNTP mix, 1 μL of Oligo (dT) (0.5 μg/μL), RNase free water (the above 3 reagents were supplied by the manufacturer) and 0.5 μg of total RNA (variable volume) was added to achieve a total volume of 10 μL. The mixture was incu- bated at 65 °C for 5 min and then on ice for 1 min. A mixture of 10× RT Buffer (2 μL), 25 mM MgCl2 (4 μL), 0.1 M DTT (2 μL), and RNase Out (1 μL) was added to the reaction and incubated at 42 °C for 2 min. Subsequently, 1 μL of Super Script III was added and incubated at 50 °C for 50 min. The reaction was terminated by incubation at 85 °C for 5 min. The cDNA was treated with 1 μL RNase H for 20 min at 37 °C and was then stored at – 20 °C.

2.2.6.Real-time PCR
The primers and probes for fi ve genes: activating transcription factor 3 (ATF3); DnaJ (Hsp40) homolog, subfamily B, member 4 (DNAJB4); glutamate-cysteine ligase, modifi er subunit (GCLM); in- terleukin-8 (IL-8); and Nrf2, and for one endogenous control gene (glyceraldehyde 3-phosphate dehydrogenase (GAPDH)) were de- signed by Assays-by-Design Service from Applied Biosystems and the sequence information remains confi dential. The primers and probes were delivered as a 20× Taqman Gene Expression Assay mix (Applied Biosystems). A total volume of 20 μL sample, which consisted of 1 μL TaqMan® Gene Expression Assay 20×, 10 μL TaqMan Universal PCR Master Mix (Applied Biosystems), 1 μL cDNA template and 8 μL dH2O was prepared and applied to an opti- cal reaction plate (96-well plate; Applied Biosystems). Real-time PCR reactions were performed using the Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems). Relative gene expression levels versus control (fold change) were calculated using the 2- ΔΔCt method (comparative CT method; Livak and Schmittgen, 2001).

Student’s t-test was used to evaluate statistical signifi cance. p- Values b 0.01 were considered to be statistically signifi cant.

2.3.Assessment of the predictive performance of marker genes for 72 chemicals
2.3.1.Tissue culture
The RhE model “LabCyte EPI-MODEL (24 well format)” (Japan tissue Engineering Co. Ltd., Aichi, Japan) was used in this study. This model consists of NHEKs whose biological origins are neonatal foreskins, and which construct a multilayer structure consisting of a fully differentiat- ed epithelium with features of a normal human epidermis, including a stratum corneum (Katoh et al., 2009). This RhE model was adopted in the OECD test guideline 439; In vitro Skin Irritation: Reconstructed Human Epidermis Test Method (OECD, 2013). The tissues were pre- cultured overnight at 37 °C (5% CO2) in 0.5 mL/tissue of culture media provided by the manufacturer.

2.3.2.Chemical exposure
For dose fi nding, solid test chemicals were dissolved in an ap- propriate vehicle (AOO, DW or 50% EtOH) at the maximum soluble concentration (from 50%) and 4-fold serial dilutions were then made to prepare test the chemical solutions at doses ranging from the maximum soluble concentration to a minimum concentration of 0.02%. Liquid test chemicals were serially diluted 4-fold with an appropriate vehicle to prepare test chemical solutions at doses ranging from 100% (neat chemical) to a minimum dose of 0.02%. A 5 μL aliquot of each test chemical solution was applied to the sur- face of the LabCyte EPI-MODEL (1 well per group) and was incubat- ed for 6 h at 37 °C (5% CO2). An untreated and a killed control tissue exposed to 10 μL of 10% triton (1 well per group) were prepared as control tissues for cell viability measurement. Cell viability was assessed using the lactate dehydrogenase assay (LDH assay) (see Section 2.3.4).For the main study, each test chemical was dissolved at the minimum concentration that showed b 80% cell viability in the dose fi nding study. The test chemical solutions were prepared by making 2-fold serial dilutions of the maximum concentration that showed N 90% cell viability in the dose fi nding study. Generally, four to fi ve different concentrations were prepared for each test chemical. If no cytotoxicity was observed in the dose fi nding study, at least three test chemical solutions of different concentra- tions (e.g. 1×, 1/2× and 1/22 × maximum soluble concentration) were prepared. A 5 μL aliquot of each test chemical solution was ap- plied to the surface of the LabCyte EPI-MODEL (3 wells per group) and was incubated for 6 h at 37 °C (5% CO2). An untreated tissue, a killed control tissue exposed to 10 μL of 10% triton, and vehicle- treated tissues (3 wells per group) were prepared as control tissues.

2.3.3.RNA isolation, cDNA synthesis and Real-time PCR
Following 6 h treatment with chemicals, the tissue surfaces were rinsed three times in 500 μL pre-warmed D-PBS(-) (Life Technolo- gies, Carlsbad, CA, USA). The tissues were then gently removed, placed into a 1.5 mL microtube containing 0.5 mL TRIzol® (Invitrogen) and homogenized by vortex mixing. Chloroform (100 μL; Tokyo Chemical Industry, Tokyo, Japan) was added to each microtube containing a homogenized sample. The samples were centrifuged at 12,000 × g for 15 min at 4 °C. The aqueous phase containing RNA was transferred to a 1.5 mL microtube and RNA was purifi ed using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. The RNA was quantifi ed using a ND-1000 spectrophotometer and was stored at – 80 °C until use. cDNA synthesis and real-time PCR were performed in accordance with the protocol described in Sections 2.2.5 and 2.2.6.

2.3.4.LDH assay
To determine cell viability, LDH activity in the culture media was measured using an LDH cytotoxicity detection kit (Takara Bio, Inc., Tokyo, Japan) in accordance with the manufacturer’s instructions. Upon LDH release, this kit creates a red formazan dye that absorbs light at 490 nm. The absorbance of each well was measured at 490 nm and 620 nm (the reference wavelength) using a plate reader (BMG LABTECH GmbH, Offenburg, Germany). ΔAbsorbance was calculated by subtracting the absorbance at 620 nm from the absorbance at 490 nm. Media from an untreated tissue or a tissue treated with 10% Tri- ton X-100 was used for calculation of minimal LDH release (negative control) or maximal LDH release (killed control), respectively. Cell via- bility was calculated using the following formula:exposed to culture media was set to represent 100% cell viability and the results are expressed as percentage of an untreated control.

2.3.6. Data analysis and prediction
The mean value (3 wells per group) of maximal fold induction (Imax) of each gene was determined using data obtained from chemical con- centrations that displayed over 80% cell viability. When the Imax of at least one out of the four marker genes exceeded the cut-off value of that gene (15-fold for ATF3, 2-fold for DNAJB4 and GCLM, and 4-fold for IL-8) at over 80% cell viability, the chemical was judged as positive (The EpiSensA prediction). Also, the estimated concentration (EC) that showed a fold induction of each cut-off value (ATF3 EC15, DNAJB4 EC2, GCLM EC2 and IL-8 EC4) at over 80% cell viability was calculated using linear interpolation from the values above and below the induc- tion thresholds. If fold inductions at all tested concentrations exceeded the cut-off values, the EC values were calculated by linear extrapolation from the values at the minimum two tested concentrations.

3.1.Assessment of the regulation mechanism of marker genes in NHEKs We fi rst checked the involvement of Nrf2 in the up-regulation of
DNAJB4 and GCLM by a representative sensitizer, DNCB, in NHEKs. To this end, Nrf2 expression was inhibited by knock-down of Nrf2 with siRNA. The up-regulation of DNAJB4 and GCLM by DNCB was signifi – cantly suppressed (67% inhibition of DNAJB4 and 64% inhibition of GCLM) when Nrf2 expression was strongly down-regulated (b 10% compared with the normal condition) (Fig. 1A). On the other hand, the up-regulation of ATF3 and IL-8 DNCB was not significantly inhibited by Nrf2 siRNA treatment (Fig. 1B). The induction level of ATF3 by DNCB was 1.9-fold higher under the Nrf2 knock-down condition, but there’s no signifi cant difference (p = 0.02). These results suggested that the up-regulation of DNAJB4 and GCLM by sensitizers could be partly regu- lated by Nrf2. We next checked the involvement of P2X7 in the up-reg- ulation of ATF3 and IL-8 by DNCB, by blocking P2X7 with the P2X7 specific antagonist KN-62. The up-regulation of both ATF3 and IL-8 by DNCB was significantly inhibited by KN-62 (64% and 77% inhibition, re- spectively) (Fig. 2A), whereas the up-regulation of DNAJB4 and GCLM was not (Fig. 2B). These results indicated that P2X7 could be partly in- volved in the up-regulation of ATF3 and IL-8 by sensitizers in NHEKs.
For chemicals that are known to affect the LDH assay (e.g. lactic acid, salicylic acid), the MTT assay was alternatively performed (see Section 2.3.5).

2.3.5. MTT assay
A methylthiazolydiphenyl-tetrazolium bromide (MTT, Sigma- Aldrich Co.) solution was prepared at a concentration of 0.5 mg/mL in culture medium (provided by Japan tissue Engineering Co. Ltd). The tissue surfaces were washed three times with D-PBS (-) and 300 μL of MTT solution was placed on top and incubated for 3 h at 37 °C (5% CO2). After incubation, the MTT solution was discarded and the tissues were gently removed and placed into 200 μL isopropanol (Sigma-Aldrich). The extraction process was performed for 2 h at room temperature. The absorbance of the extract was mea- sured at 570 nm using a plate reader (BMG LABTECH GmbH, Offenburg, Germany). The absorbance of the untreated control tissue

3.2.Predictive performance of marker genes to 72 chemicals
Next, we evaluated the utility of the four marker genes, ATF3, DNAJB4, GCLM and IL-8, for prediction of the skin sensitization po- tential of chemicals. A total of 72 chemicals (43 hydrophilic chemicals; logKow b 3.5, and 29 lipophilic chemicals; logKow ≥ 3.5) were evaluated using the EpiSensA protocol described in Materials &
Methods (Section 2.3). Fig. 3 shows the fold induction of the four marker genes (vs. vehicle control) and the cell viabilities calculated for four of the lipophilic sensitizers: bisphenol A diglycidyl ether; BADGE; undec-10-enal, EHA; and abietic acid. Each sensitizer up- regulated the expression of marker genes at non- to slightly toxic concentrations (cell viability was N 80%). The induction levels of four genes were dose-dependently increased by each of the four li- pophilic sensitizers. The maximum induction levels of ATF3 (ranging from 40.4-fold for BADGE to 174-fold for EHA) and IL-8 (ranging from 25.3-fold for EHA to 115.7-fold for undec-10-enal) at over 80% cell viability tended to be higher than the maximum induction levels of DNAJB4 (ranging from 2.8-fold for Abietic acid to 83.9-fold for EHA) and GCLM (ranging from 3.0-fold for undec-10-enal to 28.9- fold for EHA).
eugenol. As observed with the lipophilic sensitizers, each pre/pro-hap- ten up-regulated the expression of marker genes at non- to slightly toxic concentrations. Dose response patterns of ATF3 and IL-8 were sim- ilar for the four pre/pro-haptens, whereas the induction levels of ATF3 were higher than those of IL-8. On the other hand, the dose response patterns of DNAJB4 and GCLM were similar for three of the pre/pro-hap- tens (PPD, metol and eugenol), but were different for ethylene diamine (Fig. 4). Also, PPD, metol and eugenol induced higher up-regulation of the expression of DNAJB4 and GCLM compared to the up-regulation of ATF3 and IL-8 at over 90% cell viability. These results suggested that the four marker genes could have the potential to detect lipophilic skin sensitizers as well as pre/pro-haptens. Additionally, up-regulation of infl ammatory genes (ATF3 and IL-8), or cytoprotective genes (DNAJB4 and GCLM) would contribute more strongly to the detection of lipophilic sensitizers or pre/pro-haptens, respectively.
Table 1 shows the IC20 (inhibitory concentration showing 20% decrease of cell viability), applied dose ranges in main study, Imax and EC values of the four marker genes, and the EpiSensA prediction (positive, P, or negative, N, based on the prediction model described in Section 2.3.5) for the 72 test chemicals. The Imax values of the four marker genes, which depends on the test chemicals, varied from 0.30(sulfanilamide) to 704.8 (pentachlorophenol) for ATF3, from 0.8 (glycerol) to 202.7 (damascone) for DNAJB4, from 0.7 (glycerol) to 49.5 (damascone) for GCLM and from 0.8 (LA and sulfanilamide) to 147.1 (farnesol) for IL-8, regardless of their lipophilicity. Forty eight out of the 54 tested sensitizers showed an Imax of over 4-fold for ATF3, which was a cut-off value set in our initial study using EpiDerm™ (Saito et al., 2013a). However, 9 out of the 18 tested non-sensitizers were also observed to have an Imax of over 4-fold for ATF3; these non-sensitizers included non-sensitizing, non-irri- tating chemicals such as glycerol (Imax of 12.3-fold) and propylene glycol (Imax of 13.7-fold). On the other hand, 38 out of the 54 tested sensitizers showed an Imax of over 2-fold for DNAJB4 whereas 17 of the tested non-sensitizers, except for octanoic acid (OA), did not.
Also, 26 of the tested sensitizers, but none of the tested non-sensi- tizers, showed an Imax of over 2-fold for GCLM. Regarding IL-8, which was a newly added marker gene, 41 of the tested sensitizers showed an Imax of over 4-fold, whereas 14 out of the 18 tested non- sensitizers did not. Three irritants; benzalkonium chloride (BKC); methyl salicylate (MS) and sodium dodecyl sulfate (SLS); as well as octanoic acid (OA), also showed an Imax of N 4-fold, although the values were relatively low (b 10-fold).
We then determined the cut-off value of each marker gene based on the following criteria; 1) best predictive performance (sensitivity, specifi city and accuracy) when a test chemical was judged as positive for at least one out of two genes related to infl ammatory re- sponses (ATF3 and IL-8), or cytoprotective responses (DNAJB4 and GCLM), and 2) over 2-fold induction (to ensure the reliability of gene expression data by Real-time PCR). Regarding the infl ammato- ry response genes, a 15-fold increase in ATF3 and a 4-fold increase in IL-8 provided the best predictive performance of sensitivity (82%), specifi city (78%) and accuracy (81%). On the other hand, regarding cytoprotective genes, a 2-fold increase in both DNAJB4 and GCLM, which was the same value as that in our initial study (Saito et al., 2013a), showed the best predictive performance (sensitivity (72%), specifi city (94%) and accuracy (78%)) (Table 2). In both cases, the combination of two marker genes (ATF3 and IL-8, or DNAJB4 and GCLM) showed better sensitivity and accuracy than a single marker gene in each keratinocyte response (Table 2). Also, the predictive performance of infl ammatory genes (ATF3 and IL-8) to lipophilic chemicals was higher (86% accuracy) than that of cytoprotective genes (DNAJB4 and GCLM), whereas the predictivity for pre/prohaptens was higher (91% accuracy) for cytoprotective genes than for infl ammatory genes.
Finally, when each test chemical was judged as positive if at least 1 marker gene related to inflammatory or cytoprotective responses pro- vided positive results (EpiSensA prediction model), excellent predictive values of sensitivity (over 90%) and accuracy (approximately 90%) were obtained for the 72 test chemicals including for the lipophilic chemicals and the pre/pro-haptens (Table 2). These data suggested that the com- bination of four genes related to two keratinocyte responses could achieve higher sensitivity and accuracy in predicting the skin sensitiza- tion potential of broad sets of chemicals.
We also compared the predictive performance of the EpiSensA with those of three existing in vitro tests (the DPRA, KeratinoSens™ and h- CLAT). Table 1 shows that 11 tested lipophilic sensitizers (logKow ≥ 3.5) including dibutyl aniline, which were judged as false negative in at least two existing in vitro tests, provided positive results in the EpiSensA (Table 1). In addition, three hydrophilic sensitizers (benzoyl peroxide,of b 0.06, whereas 18 out of the 20 moderate or weak sensitizers showed EC2 values of N 0.06 (Table 1 and Fig. 5). On the other hand, IL-8 EC4 and ATF3 EC15 values showed a moderate correlation with LLNA EC3 values (R = 0.59 for IL-8 (N = 41); R = 0.50 for ATF3 (N = 35)). Extreme sensitizers provided lower IL-8 EC4 values (0.006 for OXA, 0.035 for TCSA, 0.034 for 4-nitrobenzylbromide (NBB), and 0.002 for DNCB) and/or ATF3 EC15 values (0.027 for TCSA, 0.05 for NBB and 0.026 for DNCB) than sensitizers with other potency categories (Table 1 and Fig. 5). DNAJB4 EC2 values showed a weak correlation with the LLNA EC3 values (R = 0.37, N = 38), suggesting this value might not contribute to the potency estimation. Based on these results, the follow- ing criteria were set to predict skin sensitization potency with the EpiSensA. If the EC value(s) of test chemicals met at least one of follow- ing three cases; A) GCLM EC2 ≤ 0.06, B) IL-8 EC4 ≤ 0.01 or C) ATF3 EC15 ≤ 0.04, the test chemicals were judged as extreme or strong sensi- tizers. If not, the test chemicals judged as positive in the EpiSensA were accuracy of the four in vitro tests are shown in Table 3. Regarding lipo- philic chemicals, the EpiSensA showed much better sensitivity and ac- curacy (93%) than the other three in vitro tests (ranging from 44% to 67%). Furthermore, the sensitivity of the EpiSensA to hydrophilic chemicals (96%) or pre/pro-haptens (100%) was greater compared to the other in vitro tests (ranging from 70 to 81%, or 55 to 80%, respective- ly) (Table 3).
We then assessed the potency predictive capacity of the EpiSensA based on the EC values of the four marker genes shown in Table 1. Fig. 5 shows scatter plots of log (LLNA EC3) versus log (ATF3 EC15), Log (DNAJB4 EC2), log (GCLM EC2) or log (IL-8 EC4) for sensitizers judged as positive for each marker gene.Table 4 shows the number of test chemicals that were categorized in each potency category of ex- treme or strong (E or S), moderate or weak (M or W), or non-sensitizer (NS). Eight out of the 12 extreme or strong sensitizers in the LLNA were correctly predicted as extreme or strong in the EpiSensA (Table 4). Also, 37 out of the 42 moderate or weak sensitizers in the LLNA were correct- ly categorized in the EpiSensA. Consequently, the over prediction rate, under prediction rate and accuracy of the EpiSensA were 8.3%, 9.7% and 82%, respectively (Table 4). These results suggested that EpiSensA would be used to predict potency classification based on the LLNA EC3.


In this study, we found that the up-regulation of DNAJB4 and GCLM, or ATF3 and IL-8 in NHEKs by a skin sensitizer was partly under the con- trol of Nrf2 or P2X7, which are key molecules for skin sensitization. This finding suggested that the four marker genes of the EpiSensA could be mechanistically relevant for skin sensitization. We also demonstrated
that combination of these four marker genes in the RhE model provided excellent predictive performance compared to the LLNA (sensitivity, specifi city and accuracy were 94%, 79% and 90%, respectively) for 72 chemicals including 29 lipophilic chemicals and 11-pre/pro-haptens. Finally, we showed that potency classification based on the LLNA EC3 (extreme/strong or moderate/weak) would be predicted based on the threshold concentrations of three marker genes (ATF3, GCLM and IL- 8). These data suggested that the EpiSensA, which is based on the ex- pression levels of four marker genes including the new marker IL-8, is a mechanism-based test with good predictive performance of skin sen- sitization to broad sets of raw materials.

4.1.Assessment of the regulation mechanism of EpiSensA marker genes We showed that Nrf2 was involved in the up-regulation of GCLM
and DNAJB4 by sensitizers in NHEKs (Fig. 1). Nrf2 is a key transcription factor that has been reported to negatively regulate the induction of skin sensitization (et al. et al., 2013; van der Veen et al., 2013). Also, multiple Nrf2-dependent genes have been reported to be key markers in dis-

On the other hand, the up-regulation of ATF3, as well as of IL-8, was suppressed by KN-62, a specifi c antagonist of the ATP receptor P2X7 (Fig. 2). Our data showing that the up-regulation of IL-8 was mediated through the P2X7 receptor is consistent with previous reports (Montreekachon et al., 2011; Qiu et al., 2014). On the other hand, our study is the first report to show that the up-regulation of ATF3 by hap- ten in NHEK cells was mediated through P2X7. Since P2X7 has been re- ported to be an important positive regulator for the induction of skin sensitization (Weber et al., 2010), our fi nding suggests that ATF3 might be positively involved in the induction of skin sensitization. This notion is supported by a previous finding by Nguyen et al. (2014) who showed that the production of pro-infl ammatory cytokines such as TNF-α, which are involved in the induction of skin sensitization (Kaplan et al., 2012), was impaired in bacteria-infected macrophages of ATF3 KO mice. Our combined data suggested that expression of the four marker genes of ATF3, DNAJB4, GCLM and IL-8 was regulated, at least in part, through the key molecules, Nrf2 and P2X7, of skin sensiti- zation in NHEKs and that these genes could therefore be mechanistically relevant gene markers.

4.2.Comparison between the LabCyte EPI-MODEL and EpiDerm™
In this study, we assessed the predictive performance of four marker genes using the LabCyte EPI-MODEL, which is one of the RhE models adopted in OECD TG439 (OECD, 2013). Our previous study using EpiDerm™ demonstrated that three marker genes (ATF3, DNAJB4 and GCLM) showed good predictive performance for 16 ECVAM reference chemicals (12 sensitizers and 4 non-sensitizers) (Saito et al., 2013a). The expression levels of GCLM genes induced by 16 ECVAM reference chemicals (1.1- to 15.2-fold increase; shown in Table 1) were compara- ble to those in our previous study (1.0- to 13.8-fold increase) (Saito et al., 2013a). On the other hand, the induction levels of ATF3 and DNAJB4 in the LabCyte EPI-MODEL (1.0- to 77-fold increase in ATF3, 1.0- to 14.7-fold increase in DNAJB4) were slightly lower compared to those in the EpiDerm™ assay (1.5- to 165-fold increase in ATF3; 1.2- to 51-fold increase in DNAJB4). Also, relatively higher induction levels of ATF3 were observed for four reference non-sensitizers (glycerol, lactic acid, salicylic acid and sodium lauryl sulfate), which provided Imax values from 1.0- to 12.3-fold in this study (Table 1), compared to the 1.5- to 4.0-fold induction levels observed in our previous study (Saito et al., 2013a). It is unclear why such differences in the expression of some marker genes were observed between the two different RhE models. However, it has been reported that the proper- ties of the stratum corneum barrier (e.g. lamella layer, metabolic en- zyme activity) are different between the EpiDerm™ and the Labcyte

EPI-MODEL (Kano et al., 2011), which would affect the responsive- ness of keratinocytes to topically applied chemicals. Another investi- gation reported similar fi ndings in which they showed a difference in the production of the infl ammatory gene IL-18 between different RhE models (i.e. the EpiDerm™ and the “epiCS® Epidermis Equiva- lent model”, a commercially available RhE model) (Gibbs et al., 2013). Regarding IL-8 induction, McKim et al. (2012) reported that the maximum induction levels of IL-8 induced by the four ECVAM reference skin sensitizers DNCB, 2-mercptbenzothiazole (MBT), isoeugenol and eugenol in the EpiDerm™ were about 10-fold, over 20-fold, 3.5-fold or 2.5-fold, respectively. These values are similar to the Imax values of IL-8 obtained in the present study (7.0-fold induction by DNCB, 16-fold by MBT, 2.3-fold by isoeugenol and 2.1- fold by eugenol) (Table 1), suggesting that the induction pattern of IL-8 by sensitizers might be similar in the two RhE models. The com- bined data suggest that it is possible that the marker genes used in the EpiSensA might be compatible between RhE models, although optimization of the prediction model for each RhE model would be needed to correctly predict skin sensitization.


In this study, we newly added IL-8 as a marker gene for the EpiSensA and showed that most of the tested sensitizers including li- pophilic chemicals (i.e. logKow ≥ 3.5) up-regulated the expression of IL-8 (Fig. 3 and Table 1). Although previous fi ndings showed that IL-8 could be a good predictive marker for skin sensitization in monolayer keratinocytes (McKim et al., 2010; Bae et al., 2015), in this paper, we are the first to demonstrate that IL-8 could be a good indicator of skin sensitization for larger sets of chemicals, including lipophilic chemicals, in an RhE model. Weber et al. (2015) recently reported that neutrophils are important immune cells for both the induction and the elicitation of skin sensitization. This finding implies that IL-8, which is a chemokine that mainly promotes the recruitment of neutrophils, could be an im- portant mediator of skin sensitization. On the other hand, IL-8 produc- tion can also be observed in other skin infl ammations such as in irritant contact dermatitis (Lee et al., 2013), suggesting that IL-8 might be less specific to the induction of skin sensitization. Nevertheless, it is also a fact that in our study five sensitizers (OXA, benzoyl peroxide, ben- zyl cinnamate, lillial, tridecane) were positive only for induction of IL-8, and three out of these five chemicals were lipophilic chemicals (Table 1), which affected the relatively high sensitivity of IL-8 (Table 2). Thus, our data suggested that IL-8 could be a good marker gene of skin sensitization, especially for decreasing false negatives.

4.4.The advantage of the EpiSensA for the prediction of skin sensitization potential

We showed that the combination of two genes related to inflamma- tory (ATF3 and IL-8) or cytoprotective (DNAJB4 and GCLM) responses provided better sensitivity and accuracy than a single marker (Table 2). These results suggested that the use of multiple marker genes related to one of two key responses in keratinocytes (infl ammatory or cytoprotective) provided better predictivity. Other studies have similar- ly shown that a combination of multiple genes related to the same sig- naling pathway (i.e. the Nrf2 pathway) could improve the predictive performance in RhE-based assays (McKim et al., 2012; Cottrez et al., 2016). Finally, we demonstrated that sensitivity and accuracy for 72 chemicals compared to the LLNA were 94% and 90%, respectively by the combination of four marker genes (Table 2), and that these values were relatively higher than those of existing in vitro tests (DPRA, KeratinoSens™ and h-CLAT) (Table 3). This higher predictivity was mainly attributed to a higher capability of detecting lipophilic chemicals (Table 3), which strongly supported our initial idea that the EpiSensA could overcome one of the major limitations of existing in vitro tests, i.e. the inability to assay lipophilic chemicals. Furthermore, the EpiSensA provided better sensitivity (96%) to hydrophilic chemicals compared to the existing three tests (DPRA, KeratinoSens™ and h-CLAT) (ranging from 70 to 81%, Table 3). The first reason for this higher sensitivity is the improved detection of pre/pro-haptens, whose detection is also a limitation of existing in vitro tests (OECD, 2015a, 2015b and OECD, 2015C). In this study, we demonstrated that EpiSensA could detect all of the 11 tested pre/pro-haptens, which included 6 hydrophilic pre/pro-haptens (i.e. isoeugenol, ethylene diamine, diethylene triamine, resorcinol, cinnamic alcohol and eugenol) that showed false negative results in at least one out of three other tests (Table 1). Although we have not shown that the metabolites of these pre/pro-haptens were produced in our test system, these data suggested that the RhE-based assay EpiSensA, has the capability of predicting pre/pro-haptens. This idea was supported by previous studies, which showed that other RhE based assays could also correctly identify several hydrophilic pre/pro- haptens (McKim et al., 2012; Gibbs et al., 2013; Cottrez et al., 2016). The second reason for the higher sensitivity of the EpiSensA is its ability to detect chemicals whose evaluation is technically difficult using cell- based systems. For instance, benzoyl peroxide, which is a sensitizer that showed a false negative in the cell-based KeratinoSens™ and h- CLAT assays, was correctly predicted in the EpiSensA (dissolved in a li- pophilic vehicle, AOO) (Table 1). It has been reported that this chemical would be immediately hydrolyzed to a non-sensitizing chemical, benzoic acid, under aqueous conditions (Aptula et al., 2005). These data indicated that EpiSensA is applicable to chemicals that are unstable in cell-based systems with aqueous media. The third reason for the higher sensitivity of the EpiSensA is that the EpiSensA used four marker genes that reflect two different keratinocyte responses (inflammatory and cytoprotective). EpiSensA is an RhE-based assay, the main constitu- ent of which is keratinocytes. KeratinoSens™ is also a keratinocyte- based assay that is based on Nrf2/ARE activation (cytoprotective response) in the keratinocyte cell-line HaCaT (OECD, 2015b). Some sen- sitizers such as 3-propylidenephthalide and penicillin G, both of which are acyl transfer agents (Roberts et al., 2007), were positive in the EpiSensA, but were negative in the KeratinoSens™ assay (Table 1). It has been reported that KeratinoSens™ has relatively lower sensitivity to acyl transfer agents because this category of sensitizers cannot acti- vate the Nrf2/ARE pathway (Urbisch et al., 2015). On the other hand, EpiSensA could detect penicillin G by its induction of ATF3 and IL-8 (Table 1), which could be involved in inflammatory pathways (Fig. 2). Also, 3-propylidenephthalide was detected by induction of ATF3, IL-8 and DNAJB4, although the induction level of DNAJB4 was relatively weak (2.3-fold increase). These data indicated that the EpiSensA based on two different keratinocyte responses could provide higher sensitivity than the other keratinocyte-based assays, which focus on only one keratinocyte response. Combining all of these data, it is strongly sug- gested that the EpiSensA can overcome the limitations of existing in vitro tests in terms of the testing system (RhE-based) and endpoints (four marker genes that relate to two different keratinocyte responses).
Moreover, although some RhE-based skin sensitization assays such as the EE assay and the SENS-IS assay showed N 95% accuracy for dozens of chemicals including several pre/pro-haptens, the predictivity of these assays for lipophilic chemicals has not been clearly depicted (Gibbs et al., 2013; Cottrez et al., 2016). Thus, this study is the first report to clear- ly demonstrate that lipophilic chemicals can be correctly evaluated by a RhE-based skin sensitization assay with high sensitivity (93%). On the other hand, it is also the fact that only two lipophilic non-sensitizers without cytotoxicity were tested in this study. To further assess the pre- dictive performance of the EpiSensA for lipophilic chemicals, we need to test more lipophilic non-sensitizers in the future.

4.5.False negatives and positives

There were three false negatives (tocopherol, squalic acid (SA) and isopropyl myristate (IM)) and four false positives (BKC, MS, OA and SLS) in the EpiSensA, which need to be discussed for better definition of the applicability domain of this assay.
Regarding the three false negatives, it has been reported that to- copherol and IM are essentially free from skin sensitizing activity in humans (Basketter et al., 2014), suggesting the EpiSensA could correctly predict skin sensitization potential in humans and that the LLNA results were probably false positives as suggested by other investigators (Natsch et al., 2013; Urbisch et al., 2015). On the other hand, it has been reported that SA cannot permeate the cell membrane in aqueous vehicles because, at physiological pH it exists as its dianion in aqueous conditions (Aptula et al., 2005). SA was not soluble in AOO (a lipophilic vehicle), but was soluble in 50% EtOH (an aqueous vehicle) at a maxi- mum of 5%, and showed no cytotoxicity in the EpiSensA. These data sug- gested that there was insuffi cient exposure of the cells to SA. Similar false outcomes have been observed in other in vitro skin sensitization assays such as the KeratinoSens™ and h-CLAT assays that use an aque- ous culture medium (Urbisch et al., 2015). Thus, due care should be taken with interpreting negative outcomes in the EpiSensA when easily ionized chemicals are tested using aqueous vehicles (DW or 50% EtOH in DW).
Regarding false positives, three out of the four false positives (BKC, SLS and MS) induced only IL-8 expression at levels slightly exceeding the cut-off value of 4-fold induction (BKC, 6.2-fold; SLS, 8.6-fold; and MS, 6.0-fold induction). These three chemicals are irritants and the fi rst two are surfactants, suggesting that some irritating chemicals such as surfactants might cause a false positive result in the EpiSensA, by induction of only IL-8 expression. Similar observations have been re- ported by other investigators using an RhE model. Coquette et al. (2003) reported that SLS, BKC and benzoic acid (an irritant) significantly up- regulated IL-8 mRNA expression in an RhE model, although the induc- tion levels were relatively weak compared to the levels induced by skin sensitizers (e.g. DNCB). Also, it has been reported that SLS weakly induced IL-8 mRNA in the EpiDerm™ assay (McKim et al., 2012). Con- sidering the fact that IL-8 production was observed in irritant contact dermatitis (Lee et al., 2013), relatively weak up-regulation of IL-8 by ir- ritants (surfactants) would be inevitable using RhE models. Thus, when irritating test chemicals such as surfactants up-regulate only IL-8 ex- pression to a level that slightly exceeds the cut-off value (e.g. b 10-fold induction), care would need to be taken to judge whether the test chemicals were true positives or not. On the other hand, OA up-regulat- ed the expression of ATF3, DNAJB4 and IL-8 to a level that exceeded the cut-off values (38.1-fold induction of ATF3, 3.5-fold of DNAJB4 and 5.9- fold of IL-8). It remains unclear why this chemical up-regulated three marker genes; however, previous reports showed that OA as well as other fatty acids have the potential to induce ATF3 and DNAJB4 in rat cardiomyocytes (Lockridge et al., 2008), or IL-8 in chicken macrophages (Sunkara et al., 2012), suggesting the fatty acids might have the poten- tial to up-regulate these three marker genes. To further define the appli- cability domain of the EpiSensA, other middle chain and long chain fatty acids need to be evaluated.

4.6.Potency assessment

In addition to assessment of skin sensitization potential, potency prediction is important for the GHS classification, which goes beyond the hazard identification by defining two classes of sensitizers, weak and strong. For potency prediction, LLNA EC3 values could be useful be- cause of the close correlation with the relative ability of contact aller- gens to cause skin sensitization in Human (Basketter et al., 2000; Griem et al., 2003). Although human sensitization potency of some fra- grance materials (e.g. hexyl salicylate) or nickel might be over- or under-estimated in the LLNA (Urbisch et al., 2015), it has been proposed that LLNA EC3 could be used to estimate human acceptable exposure levels of sensitizers as part of quantitative risk assessment (Loveless et al., 2010). Also, it is important to distinguish between the two potency categories of extreme/strong and moderate/weak, since chemicals with relatively weak (moderate/weak) skin sensitization potency might be used in a product at very limited concentrations with special care (e.g. based on the sensitization threshold (Keller et al., 2009; Safford et al., 2011)). In this study, we showed that the GCLM EC2 value corre- lated well with the LLNA EC3 value (Fig. 5). We also demonstrated that the rank categorization of LLNA EC3 values (extreme/strong or moder- ate/weak) of 72 tested chemicals including 29 lipophilic chemicals and 11 pre/pro-haptens was predicted with good accuracy (82%), based on ATF3 EC15, GCLM EC2 and IL-8 EC4 values (Table 4). Among 12 extreme or strong sensitizers tested in the EpiSensA, hexyl salicylate has been reported to be a very weak human sensitizer with no induction in human tests (Human Repeated Insult Patch Test and Human Maximi- zation Test) at over 20,000 μg/cm2 (Urbisch et al., 2015). Excluding hexyl salicylate, the predictivity of the EpiSensA for 11 extreme and strong chemicals were 73% (8/11), which was comparable to that of the h-CLAT (74%) for extreme and strong chemicals (Nukada et al., 2012). Thus, these data suggested that the threshold concentrations of EpiSensA marker genes (ATF3, GCLM, IL-8) might be useful to estimate the potency classification based on the LLNA EC3. It has been reported that other RhE-based assays are applicable to the estimation of skin sen- sitization potency. For instance, Gibbs et al. (2013) showed that the out- come of the EE assay for 11 hydrophilic sensitizers including 5 pre/pro- haptens showed a good correlation with skin sensitization potency in humans and with the LLNA data. Also, Cottrez et al. (2016) reported that the SENS-IS assay provided N 90% accuracy for the categorization of 150 tested chemicals into the 5 LLNA potency categories of extreme, strong, moderate, weak and non-sensitizer. In addition to these RhE- based assays, our data showed that the EpiSensA could also provide in- formation regarding the potency classification. On the other hand, it is also a fact that a 1:1 potency assessment (i.e. estimation of the LLNA EC3 values) cannot be made based solely on the result of the EpiSensA. To solve this problem, incorporation of the EpiSensA data into a global ITS system such as the Bayesian network Integrated Testing Strategy (Jaworska et al., 2015), which integrates all of the information (e.g. in vitro data, and physic-chemical properties) for all chemical classes into one model to predict skin sensitization potency might be useful, since the training set of the global ITS system contained limited data of lipophilic chemicals (Natsch et al., 2013). Further study is necessary to precisely predict the skin sensitization potency of broad sets of chemicals.

5.Conclusion and future plans

In summary, the EpiSensA, which is based on the expression levels of four marker genes that are related to keratinocyte responses in the early phase of skin sensitization, is a mechanism-based test with good predic- tive performance of skin sensitization to broad sets of raw materials. To further examine the utility of EpiSensA as part of a test battery for non- animal skin sensitization assessment, we will expand the dataset to fur- ther clarify the applicability domain of the EpiSensA. Also, intra/inter- laboratory reproducibility will be assessed towards establishment of the EpiSensA.

Transparency document

The Transparency document associated with this article can be found, in the online version.

Anderson, C.M., Nedergaard, M., 2006. Emerging challenges of assigning P2X7 receptor function and immunoreactivity in neurons. Trends Neurosci. 29, 257–262.
Aptula, A.O., Patlewicz, G., Roberts, D.W., 2005. Skin sensitization: reaction mechanistic applicability domains for structure-activity relationships. Chem. Res. Toxicol. 18, 1420–1426.
Aptula, A.O., Roberts, D.W., Pease, C.K., 2007. Haptens, prohaptens and prehaptens, or electrophiles and proelectrophiles. Contact Dermatitis 56, 54–56.

Bae, O.N., Ahn, S., Jin, S.H., Hong, S.H., Lee, J., Kim, E.S., Jeong, T.C., Chun, Y.J., Lee, A.Y., Noh, M., 2015. Chemical allergens stimulate human epidermal keratinocytes to produce lymphangiogenic vascular endothelial growth factor. Toxicol. Appl. Pharmacol. 283, 147–155.
Barker, J.N., Mitra, R.S., Griffiths, C.E., Dixit, V.M., Nickoloff, B.J., 1991. Keratinocytes as ini- tiators of inflammation. Lancet 337, 211–214.
Basketter, D.A., Balikie, L., Dearman, R.J., Kimber, I., Ryan, C.A., Gerberick, G.F., Harvey, P., Evans, P., White, I.R., Rycroft, R.J., 2000. Use of the local lymph node assay for the es- timation of relative contact allergenic potency. Contact Dermatitis 42, 344–348.
Basketter, D.A., Alépée, N., Ashikaga, T., Barroso, J., Gilmour, N., Goebel, C., Hibatallah, J., Hoffmann, S., Kern, P., Martinozzi-Teissier, S., Maxwell, G., Reisinger, K., Sakaguchi, H., Schepky, A., Tailhardat, M., Templier, M., 2014. Categorization of chemicals accord- ing to their relative human skin sensitizing potency. Dermatitis 25, 11–21.
Casati, S., Aeby, P., Kimber, I., Maxwell, G., Ovigne, J.M., Roggen, E., Rovida, C., Tosti, L., Basketter, D., 2009. Selection of chemicals for the development and evaluation of in vitro methods for skin sensitization testing. Altern. Lab. Anim 37, 305–312.
Coquette, A., Berna, N., Vandenbosch, A., Rosdy, M., De Wever, B., Poumay, Y., 2003. Anal- ysis of interleukin-1alpha (IL-1alpha) and interleukin-8 (IL-8) expression and release in in vitro reconstructed human epidermis for the prediction of in vivo skin irritation and/or sensitization. Toxicol. in Vitro 17, 311–321.
Cottrez, F., Boitel, E., Ourlin, J.C., Peiffer, J.L., Fabre, I., Henaoui, I.S., Mari, B., Vallauri, A., Paquet, A., Barbry, P., Auriault, C., Aeby, P., Groux, H., 2016. SENS-IS, a 3D reconstituted epidermis based model for quantifying chemical sensitization potency: reproducibility and predictivity results from an inter-laboratory study. Toxicol. in Vitro 32, 248–260.
ECVAM, 2016. EURL ECVAM validated methods adopted by the OECD paved the way for the revision of regulatory requirements for skin sensitisation under REACH [WWW document]. URL mechanistically-based-non-animal-test-methods-to-assess-the-potential-of- chemicals-to-cause-skin-allergy
Eilstein, J., Léreaux, G., Budimir, N., Hussler, G., Wilkinson, S., Duché, D., 2014. Comparison of xenobiotic metabolizing enzyme activities in ex vivo human skin and reconstruct- ed human skin models from SkinEthic. Arch. Toxicol. 88, 1681–1694.
El Ali, Z., Gerbeix, C., Hemon, P., Esser, P.R., Martin, S.F., Pallardy, M., Kerdine-Römer, S., 2013. Allergic skin inflammation induced by chemical sensitizers is controlled by the transcription factor Nrf2. Toxicol. Sci. 134, 39–48.
Emter, R., van der Veen, J.W., Adamson, G., Ezendam, J., van Loveren, H., Natsch, A., 2013. Gene expression changes induced by skin sensitizers in the KeratinoSens™ cell line: discriminating Nrf2-dependent and Nrf2-independent events. Toxicol. in Vitro 27, 2225–2232.
Erickson, A.M., Nevarea, Z., Gipp, J.J., Mulcahy, R.T., 2002. Identifi cation of a variant antioxidant response element in the promoter of the human glutamate-cysteine ligase modifi er subunit gene. Revision of the ARE consensus sequence. J Biol. Chem. 277, 30.
Gerberick, G.F., House, R.V., Fletcher, E.R., Ryan, C.A., 1992. Examination of the local lymph node assay for use in contact sensitization risk assessment. Fundamen. Appl. Toxicol. 19, 438–445.
Gerberick, G.F., Ryan, C.A., Kern, P.S., Schlatter, H., Dearman, R.J., Kinber, I., Patlewicz, G.Y., Basketter, D.A., 2005. Compilation of historical local lymph node data for evaluation of skin sensitization alternative methods. Dermatitis 16, 157–202.
Gibbs, S., Corsini, E., Spiekstra, S.W., Galbiati, V., Fuchs, H.W., Degeorge, G., Troese, M., Hayden, P., Deng, W., Roggen, E., 2013. An epidermal equivalent assay for identifica- tion and ranking potency of contact sensitizers. Toxicol Appl Pharmacol. 272, 529–541.
Götz, C., Pfeiffer, R., Tigges, J., Blatz, V., Jäckh, C., Freytag, E.M., Fabian, E., Landsiedel, R., Merk, H.F., Krutmann, J., Edwards, R.J., Pease, C., Goebel, C., Hewitt, N., Fritsche, E., 2012a. Xenobiotic metabolism capacities of human skin in comparison with a 3D epi- dermis model and keratinocyte-based cell culture as in vitro alternatives for chemical testing: activating enzymes (Phase I). Exp Dermatol. 21, 358–363.
Götz C., Pfeiffer R., Tigges J., Ruwiedel K., Hübenthal U., Merk H.F., Krutmann J., Edwards R.J., Abel J., Pease C., Goebel C., Hewitt N., Fritsche E., 2012b. Xenobiotic metabolism capacities of human skin in comparison with a 3D-epidermis model and keratinocyte-based cell culture as in vitro alternatives for chemical testing: phase II enzymes. Exp Dermatol. 21, 364–369.
Griem, P., Goebel, C., Scheffler, H., 2003. Proposal for a risk assessment methodology for skin sensitization based on sensitization potency data. Regulatory Toxicol. Pharm. 38, 269–290.
Inbaraj, J.J., Chignell, C.F., 2004. Cytotoxic action of juglone and plumbagin: a mechanistic study using HaCaT keratinocytes. Chem Res Toxicol. 17, 55–62.
Jaworska, J.S., Natsch, A., Ryan, C., Strickland, J., Ashikaga, T., Miyazawa, M., 2015. Bayesian integrated testing strategy to assess skin sensitization potency: from theory to prac- tice. Arch Toxicol. 89, 2355–2383.
Jeong, J.B., Choi, J., Baek, S.J., Lee, S.H., 2013. Reactive oxygen species mediate tolfenamic acid-induced apoptosis in human colorectal cancer cells. Arch Biochem Biophys. 537, 168–175.
Kano, S., Todo, H., Furui, K., Sugie, K., Tokudome, Y., Hashimoto, F., Kojima, H., Sugibayashi, K., 2011. Comparison of several reconstructed cultured human skin models by microscopic observation: their usefulness as an alternative mem- brane for skin in drug permeation experiments. Alternatives to Animal Testing and Experimentation 16, 51–58.
Kaplan, D.H., Igyártó, B.Z., Gaspari, A.A., 2012. Early immune events in the induction of al- lergic contact dermatitis. Nat. Rev. Immunol. 12, 114–124.
Katoh, M., Hamajima, F., Ogasawara, T., Hata, K., 2009. Assessment of human epidermal model LabCyte EPI-MODEL for in vitro skin irritation testing according to European Centre for the Validation of Alternative Methods (ECVAM)-validated protocol. J Toxicol Sci. 34, 327–334.

Keller, D., Krauledat, M., Scheel, J., 2009. Feasibility study to support a threshold of sensi- tization concern concept in risk assessment based on human data. Arch Toxicol. 83, 1049–1060.
Kern, P.S., Gerberick, G.F., Ryan, C.A., Kimber, I., Aptula, A., Basketter, D.A., 2010. Local lymph node data for the evaluation of skin sensitization alternatives: a second com- pilation. Dermatitis 21, 8–32.
Kimber, I., Hilton, J., Dearman, R.J., Gerberick, G.F., Ryan, C.A., Basketter, D.A., Lea, L., House R.V, Ladics, G.S., Loveless, S.E., Hastings, K.L., 1998. Assessment of the skin sensitiza- tion potential of topical medicaments using the local lymph node assay: an interlaboratory evaluation. J. Toxicol. Environ. Health A. 53, 567–579.
Kimber, I., Dearman, R.J., Basketter, D.A., Ryan, C.A., Gerberick, G.F., 2002. The local lymph node assay: past, present and future. Contact Dermatitis 47, 315–328.
Kimber, I., Basketter, D.A., Gerberick, G.F., Ryan, C.A., Dearman, R.J., 2011. Chemical aller- gy: translating biology into hazard characterization. Toxicol. Sci. 120 (Suppl. 1), S238–S268.
Lee, H.Y., Stieger, M., Yawalkar, N., Kakeda, M., 2013. Cytokines and chemokines in irritant contact dermatitis. Mediators Inflamm. 2013, 916497.
Livak K.J., Schmittgen T.D., 2001. Analysis of relative gene expression data using real-time quantitative PCR and the 2(- Delta Delta C(T)) Method. Methods 25, 402–408.
Lockridge, J.B., Sailors, M.L., Durgan, D.J., Egbejimi, O., Jeong, W.J., Bray, M.S., Stanley, W.C., Young, M.E., 2008. Bioinformatic profiling of the transcriptional response of adult rat cardiomyocytes to distinct fatty acids. J Lipid Res. 49, 1395–1408.
Loveless, S.E., Api, A.M., Crevel, R.W., Debruyne, E., Gamer, A., Jowsey, I.R., Kern, P., Kimber, I., Lea, L., Lloyd, P., Mehmood, Z., Steiling, W., Veenstra, G., Woolhiser, M., Hennes, C., 2010. Potency values from the local lymph node assay: application to classification, labelling and risk assessment. Regul Toxicol Pharmacol. 56, 54–66.
Luu-The, V., Duche, D., Ferraris, C., Meunier, J.R., Leclaire, J., Labrie, F., 2009. Expression profiles of phases 1 and 2 metabolizing enzymes in human skin and the reconstruct- ed skin models Episkin and full thickness model from Episkin. J. Steroid Biochem. Mol. Biol. 116, 178–186.
MacLeod, A.K., McMahon, M., Plummer, S.M., Higgins, L.G., Penning, T.M., Igarashi, K., Hayes, J.D., 2009. Characterization of the cancer chemopreventive NRF2-dependent gene battery in human keratinocytes: demonstration that the KEAP1-NRF2 pathway, and not the BACH1-NRF2 pathway, controls cytoprotection against electrophiles as well as redox-cycling compounds. Carcinogenesis 30, 1571–1580.
McKim Jr., J.M., Keller 3rd, D.J., Gorski, J.R., 2010. A new in vitro method for identifying chemical sensitizers combining peptide binding with ARE/EpRE-mediated gene ex- pression in human skin cells. Cutan Ocul Toxicol. 29, 171–192.
McKim Jr., J.M., Keller 3rd, D.J., Gorski, J.R., 2012. An in vitro method for detecting chem- ical sensitization using human reconstructed skin models and its applicability to cos- metic, pharmaceutical, and medical device safety testing. Cutan. Ocul. Toxicol. 31, 292–305.
Montreekachon, P., Chotjumlong, P., Bolscher, J.G., Nazmi, K., Reutrakul, V., Krisanaprakornkit, S., 2011. Involvement of P2X(7) purinergic receptor and MEK1/2 in interleukin-8 up-regulation by LL-37 in human gingival fibroblasts. J Periodontal Res. 46, 327–337.
Natsch, A., Ryan, C.A., Foertsch, L., Emter, R., Jaworska, J., Gerberick, F., Kern, P., 2013. A dataset on 145 chemicals tested in alternative assays for skin sensitization undergo- ing prevalidation. J Appl Toxicol. 33, 1337–1352.
Nguyen, C.T., Kim, E.H., Luong, T.T., Pyo, S., Rhee, D.K., 2014. ATF3 confers resistance to pneumococcal infection through positive regulation of cytokine production. J Infect Dis. 210, 1745–1754.
Nukada, Y., Ashikaga, T., Miyazawa, M., Hirota, M., Sakaguchi, H., Sasa, H., Nishiyama, N., 2012. Prediction of skin sensitization potency of chemicals by human Cell Line Acti- vation Test (h-CLAT) and an attempt at classifying skin sensitization potency. Toxicol In Vitro. 26, 1150–1160.
OECD, 2010. Guideline for Testing of Chemicals No. 429: Skin Sensitization: Local Lymph Node Assay, Adopted 22nd July.
OECD, 2012a. The Adverse Outcome Pathway for Skin Sensitisation Initiated by Covalent Binding to Proteins. Part 2: Use of the AOP to Develop Chemical Categories and Inte- grated Assessment and Testing Approaches. Series on Testing and Assessment No. 168. ENV/JM/MONO(2012)/PART2.
OECD, 2012b. Guideline for Testing of Chemicals No. 305: Bioaccumulation in Fish: Aque- ous and Dietary Exposure, Adopted 2nd October.
OECD, 2013. Guideline for Testing of Chemicals No. 439: In Vitro Skin Irritation – Reconstructed Human Epidermis Test Method, Adopted 26th July.
OECD, 2015a. Guideline for Testing of Chemicals No. 442C: In Chemico Skin Sensitisation: Direct Peptide Reactivity Assay (DPRA), Adopted 5th Feb.

OECD, 2015b. Guideline for Testing of Chemicals No. 442D: In Vitro Skin Sensitisation: ARE-Nrf2 Luciferase Test Method, Adopted 5th Feb.
OECD, 2015c. Guideline for Testing of Chemicals. Draft Proposal for a New Test Guideline: In Vitro Skin Sensitisation: human Cell Line Activation Test (h-CLAT).
Ohara, H., Saito, R., Hirakawa, S., Shimada, M., Mano, N., Okuyama, R., Aiba, S., 2010. Gene expression profiling defines the role of ATP-exposed keratinocytes in skin inflamma- tion. J Dermatol Sci. 58, 143–151.
Peiser, M., Tralau, T., Heidler, J., Api, A.M., Arts, J.H., Basketter, D.A., English, J., Diepgen, T.L., Fuhlbrigge, R.C., Gaspari, A.A., Johansen, J.D., Karlberg, A.T., Kimber, I., Lepoittevin, J.P., Liebsch, M., Maibach, H.I., Martin, S.F., Merk, H.F., Platzek, T., Rustemeyer, T., Schnuch, A., Vandebriel, R.J., White, I.R., Luch, A., 2010. Allergic contact dermatitis: epidemiol- ogy, molecular mechanisms, in vitro methods and regulatory aspects. Current knowl- edge assembled at an international workshop at BfR, Germany. Cell Mol. Life Sci. 69, 763–781.
Qiu, X.B., Shao, Y.M., Miao, S., Wang, L., 2006. The diversity of the DnaJ/Hsp40 family the crucial partners for Hsp70 chaperones. Cell Mol. Life Sci. 63, 2560–2570.
Qiu, Y., Li, W.H., Zhang, H.Q., Liu, Y., Tian, X.X., Fang, W.G., 2014. P2X7 mediates ATP-driv- en invasiveness in prostate cancer cells. PLoS One 9, e114371.
Roberts, D.W., Patlewicz, G., Kern, P.S., Gerberick, F., Kimber, I., Dearman, R.J., Ryan, C.A., Basketter, D.A., Aptula, A.O., 2007. Mechanistic applicability domain classification of a local lymph node assay dataset for skin sensitization. Chem Res Toxicol. 20, 1019–1030.
Safford, R.J., Aptula, A.O., Gilmour, N., 2011. Refi nement of the Dermal Sensitisation Threshold (DST) approach using a larger dataset and incorporating mechanistic chemistry domains. Regul Toxicol Pharmacol. 60, 218–224.
Saito, K., Nukada, Y., Takenouchi, O., Miyazawa, M., Sakaguchi, H., Nishiyama, N., 2013a. Development of a new in vitro skin sensitization assay (Epidermal Sensitization Assay; EpiSensA) using reconstructed human epidermis. Toxicol In Vitro 27, 2213–2224.
Saito, K., Miyazawa, M., Nukada, Y., Sakaguchi, H., Nishiyama, N., 2013b. Development of an in vitro skin sensitization test based on ROS production in THP-1 cells. Toxicol In Vitro 27, 857–863.
Sasaki, Y., Aiba, S., 2007. Dendritic cells and contact dermatitis. Clin. Rev. Allergy Immunol. 33, 27–34.
Soeur, J., Eilstein, J., Léreaux, G., Jones, C., Marrot, L., 2015. Skin resistance to oxidative stress induced by resveratrol: from Nrf2 activation to GSH biosynthesis. Free Radic Biol Med. 78, 213–223.
Sunkara, L.T., Jiang, W., Zhang, G., 2012. Modulation of antimicrobial host defense peptide gene expression by free fatty acids. PLoS One 7, e49558.
Takenouchi, O., Miyazawa, M., Saito, K., Ashikaga, T., Sakaguchi, H., 2013. Predictive per- formance of the human Cell Line Activation Test (h-CLAT) for lipophilic chemicals with high octanol-water partition coefficients. J Toxicol Sci. 38, 599–609.
Takenouchi, O., Fukui, S., Okamoto, K., Kurotani, S., Imai, N., Fujishiro, M., Kyotani, D., Kato, Y., Kasahara, T., Fujita, M., Toyoda, A., Sekiya, D., Watanabe, S., Seto, H., Hirota, M., Ashikaga, T., Miyazawa, M., 2015. Test battery with the human cell line activation test, direct peptide reactivity assay and DEREK based on a 139 chemical data set for predicting skin sensitizing potential and potency of chemicals. J Appl Toxicol 35, 1318–1332.
Thimmulappa, R.K., Mai, K.H., Srisuma, S., Kensler, T.W., Yamamoto, M., Biswal, S., 2002. Identification of Nrf2-regulated genes induced by the chemopreventive agent sulfo- raphane by oligonucleotide microarray. Cancer Res. 62, 5196–5203.
Toebak, M.J., Gibbs, S., Bruynzeel, D.P., Scheper, R.J., Rustemeyer, T., 2009. Dendritic cells: biology of the skin. Contact Dermatitis 60, 2–20.
Urbisch, D., Mehling, A., Guth, K., Ramirez, T., Honarvar, N., Kolle, S., Landsiedel, R., Jaworska, J., Kern, P.S., Gerberick, F., Natsch, A., Emter, R., Ashikaga, T., Miyazawa, M., Sakaguchi, H., 2015. Assessing skin sensitization hazard in mice and men using non-animal test methods. Regul Toxicol Pharmacol. 71, 337.
Uruno, A., Motohashi, H., 2011. The Keap1-Nrf2 system as an in vivo sensor for electro- philes. Nitric Oxide 25, 153–160.
van der Veen, J.W., Gremmer, E.R., Vermeulen, J.P., van Loveren, H., Ezendam, J., 2013. In- duction of skin sensitization is augmented in Nrf2-deficient mice. Arch Toxicol. 87, 763–766.
Weber, F.C., Esser, P.R., Müller, T., Ganesan, J., Pellegatti, P., Simon, M.M., Zeiser, R., Idzko, M., Jakob, T., Martin, S.F., 2010. Lack of the purinergic receptor P2X7 results in resis- tance to contact hypersensitivity. J Exp Med. 207, 2609–2619.KN-62
Weber, F.C., Németh, T., Csepregi, J.Z., Dudeck, A., Roers, A., Ozsvári, B., Oswald, E., Puskás, L.G., Jakob, T., Mócsai, A., Martin, S.F., 2015. Neutrophils are required for both the sen- sitization and elicitation phase of contact hypersensitivity. J Exp Med. 212, 15–22.