A convenient spectrophotometric test for screening skin-sensitizing chemicals using reactivity with glutathione in chemico

Abstract

To initiate skin sensitization, haptens react with endogenous proteins. During this process, skin sensitizers react with small endogenous molecules containing thiol or amino groups. In this study, a simple spectrophotometric method to identify skin sensitizers in chemico was developed using the reactivity of glutathione (GSH) with test chemicals in a 96-well plate. To quantitate the remaining GSH following the reaction with a skin sensitizer, 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB) was employed. The optimized experimental conditions included the pH- and time-dependent stability of GSH, stability of derivatized products of GSH with optimal concentration and incubation time of DTNB, incubation time of GSH with the test chemicals, and molar ratios of GSH to the test chemicals. With the optimized conditions with both acetonitrile and DMSO as vehicles and incubation of GSH with test chemicals in 1:10 and 1:15 ratios for 24 h at 4 °C, 23 skin sensitizers and 23 non-sensitizers, based on the local lymph node assay, were tested to determine the predictive capacity of individual conditions. The best result showed a predictive capacity of 95.2% sensitivity, 91.3% specificity, and 93.2% accuracy, with 93.2% consistency in three trials, when 5.8% depletion was used as a cut-off value in 1:10 of GSH:test chemicals in DMSO. It would be an economic and useful screening tool for determining the skin sensitization potential of small molecules, because the present method employs simple endogenous GSH as an electron donor for sensitizers with a spectrophotometric detection system in 96-well plates, and because the method requires neither experimental animals nor cell cultures.

Supplementary Information

The online version contains supplementary material available at 10.1007/s43188-023-00218-9.

Introduction

Due to the enforcement of European regulations prohibiting animal experiments on cosmetics and their ingredients, alternative test methods limiting the use of animals have been widely developed, particularly for skin sensitization tests. From the efforts, some verified methods, such as KeratinoSens™ assay, human cell line activation test (h-CLAT), local lymph node assay (LLNA), and direct peptide reactivity assay (DPRA), were adopted as OECD test guidelines [1, 2]. Among the tests adopted by the OECD, LLNA is the most predictable test method, and has been widely used as the gold standard for alternative tests of skin sensitization [3–9]. However, most currently available alternative test methods have several limitations. For example, LLNA, a method for measuring the proliferation of T-cells in draining lymph nodes after exposure to sensitizers, requires animal use [10–12]. In addition, some tests require mammalian tissue or cell cultures, as well as expensive and complex analyzers, such as HPLC and flow cytometry [2, 11]. Although the verified methods showed a fairly good predictive capacity, the use of cell cultures and high-cost sophisticated instruments limits their widespread use. To spread an alternative test method widely in practice, it is necessary to maintain the predictive power of the test while minimizing the experimental steps and performing it as simply as possible using convenient instruments.

It is generally accepted that skin sensitization is triggered by four key events in the adverse outcome pathways. Among these, the development of a test method that mimics the process of haptenization, a complex formation between a sensitizer and a protein, has attracted attention [11, 12]. For example, the direct peptide reactivity assay (DPRA) is one of the simplest methods currently available compared to other alternative tests for skin sensitization, requiring either mammalian cell cultures or animals [13]. In DPRA, the chemical reactivity of test chemicals with hepta-peptides containing either lysine or cysteine was evaluated [11, 12]. However, there are some disadvantages to using artificially synthesized high-cost hepta-peptides and the use of a sophisticated analytical instrument, HPLC [14–16]. In addition, this method is not applicable to high-throughput tests for many test chemicals. Therefore, in this study, we aimed to develop a simpler and more economical test method in a relatively high-throughput manner, while maintaining a high predictive capacity.

It is generally recognized that lysine and cysteine residues in skin proteins would covalently bind with skin sensitizers, and that the degree of chemical reactivity with either lysine- or cysteine-containing hepta-peptides might be the underlying basis of skin sensitization test using DPRA [15]. Likewise, it is also likely that, during haptenization, reactive skin sensitizers would interact well with numerous small molecular weight endogenous substances containing either amino or thiol groups, such as GSH, cysteamine, cysteine, and homocysteine. Although endogenous small molecules with nucleophilic functional groups do not play a role in skin sensitization, it has been well-demonstrated that these molecules react with electrophilic substances [17, 18], indicating the possibility of developing a simple and convenient test method for skin sensitization using the reactivity of these nucleophiles with test chemicals. An attempt to determine the reactivity of cysteamine and GSH with skin sensitizers was made using HPLC, which showed a high predictive capacity [1]. Recently, the reactivity of GSH with skin sensitizers was tested using monobromobimane, which can fluorometrically quantify the remaining GSH [19]. Based on these studies, much simpler and more cost-effective in chemico methods to identify skin sensitizers can be developed, if electron donors can successfully trap reactive skin sensitizers.

In the present study, we aimed to develop a more convenient test system employing spectrophotometry instead of either HPLC or fluorometry, and 96-well plates rather than a test tube for a high-throughput assay. A spectrophotometric test method complements a fluorometric test in the case of fluorescent sensitizers. In this assay, GSH was reacted with the test chemicals for 24 h to mimic adduct formation of haptens with endogenous proteins. Unreacted GSH was quantified spectrophotometrically by reaction with 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB), a well-known thiol-specific reagent [20]. DTNB is a dimeric form of 5-thio-2-nitrobenzoic acid (TNB) connected with a disulfide bond. When reacted with a thiol substance including GSH, TNB was released while forming a new disulfide bond, resulting in an absorption peak at 415 nm. After optimization of the essential experimental conditions, 46 test substances (23 sensitizers and 23 non-sensitizers) were tested to determine the predictive capacity of the developed test method.

Materials and methods

Materials

The test chemicals used are listed in Table 1, along with their suppliers. GSH (≥ 98% purity), cysteine (≥ 98.5%), cysteamine (≥ 98%), homocysteine (≥ 95%), and DTNB (98%) was obtained from Sigma-Aldrich (St. Louis, MO, USA). Acetonitrile and DMSO as vehicles of test chemicals were purchased from J. T. Baker (Philipsburg, NJ) and TCI (Tokyo, Japan), respectively. The followings for preparation of buffers were purchased from Duksan Pure Chemicals (Seoul, Korea): glacial acetic acid (99%), potassium phosphate monobasic (99%), and potassium phosphate dibasic (98%). Sodium acetate trihydrate (99%) was purchased from TCI (Tokyo, Japan). All chemicals were used as received.

Table 1.

General information on 46 test chemicals used in the study

No.	Test substances	CAS no.	Chemical supplier	Purity (%)	Physical state	EC3 (%)	LLNA category
1	2,4-Dinitrochlorobenzene	97-00-7	Sigma-Aldrich	97	Solid	0.08	Extreme
2	p-Benzoquinone	106-51-4	Sigma-Aldrich	> 98	Solid	0.01	Extreme
3	p-Hydroquinone	123-31-9	Sigma-Aldrich	99	Solid	0.11	Strong
4	2-Aminophenol	95-55-6	Sigma-Aldrich	99	Solid	0.40	Strong
5	Glutaraldehyde	111-30-8	Alfa Aesar	50	Liquid	0.20	Strong
6	Formaldehyde	50-00-0	Sigma-Aldrich	38	Liquid	0.40	Strong
7	p-Phenylenediamine	106-50-3	Sigma-Aldrich	> 99	Solid	0.06	Strong
8	Propyl gallate	121-79-9	TCI	> 98	Solid	0.32	Strong
9	Cinnamaldehyde	105-55-2	Sigma-Aldrich	95	Liquid	2.00	Moderate
10	Diethyl maleate	141-05-9	Sigma-Aldrich	97	Liquid	2.10	Moderate
11	Isoeugenol	97-54-1	TCI	> 97	Liquid	1.30	Moderate
12	Resorcinol	108-46-3	TCI	99	Solid	5.50	Moderate
13	2-Mercaptobenzothiazole	149-30-4	Sigma-Aldrich	97	Solid	1.70	Moderate
14	Benzylideneacetone	122-57-6	TCI	98	Solid	3.70	Moderate
15	1-Naphthol	90-15-3	TCI	> 99	Solid	1.30	Moderate
16	4-Amino-m-cresol	2835-99-6	TCI	> 98	Solid	2.15	Moderate
17	Methyl eugenol	93-15-2	TCI	> 98	Liquid	2.40	Moderate
18	Methyl methacrylate	80-62-6	Sigma-Aldrich	99	Liquid	90.00	Weak
19	Eugenol	97-53-0	TCI	> 99	Liquid	13.00	Weak
20	2,3-Butanedione	431-03-8	Sigma-Aldrich	97	Liquid	11.00	Weak
21	Ethyleneglycol dimethacrylate	97-90-5	Sigma-Aldrich	98	Liquid	35.00	Weak
22	Cinnamyl alcohol	104-54-1	TCI	97	Liquid	21.00	Weak
23	Phenyl benzoate	93-99-2	TCI	–	Solid	13.60	Weak
24	Methyl salicylate	119-36-8	TCI	99	Liquid	NC	Non-sensitizer
25	1-Butanol	71-36-3	DC chemical	99	Liquid	NC	Non-sensitizer
26	Salicylic acid	69-72-7	Sigma-Aldrich	≥ 99	Solid	NC	Non-sensitizer
27	Lactic acid	50-21-5	Sigma-Aldrich	≥ 85	Liquid	NC	Non-sensitizer
28	2-Propanol	67-63-0	Sigma-Aldrich	≥ 99.5	Liquid	NC	Non-sensitizer
29	4-Hydroxybenzoic acid	99--96-7	Sigma-Aldrich	> 99	Solid	NC	Non-sensitizer
30	Tween 80	9005-65-6	Sigma-Aldrich	100	Liquid	NC	Non-sensitizer
31	Sulfanilamide	63-74-1	TCI	99	Solid	NC	Non-sensitizer
32	Vanillin	121-33-5	Sigma-Aldrich	99	Solid	NC	Non-sensitizer
33	p-Aminobenzoic acid	150-13-0	TCI	> 99	Solid	NC	Non-sensitizer
34	Hexane	110-54-3	Sigma-Aldrich	95	Liquid	NC	Non-sensitizer
35	Chlorobenzene	108-90-7	Sigma-Aldrich	99.9	Liquid	NC	Non-sensitizer
36	6-Methylcoumarin	92-48-8	TCI	> 99	Solid	NC	Non-sensitizer
37	4-Methoxyacetophenone	100-06-1	Sigma-Aldrich	99	Solid	NC	Non-sensitizer
38	Diethyl phthalate	84-66-2	TCI	> 99	Liquid	NC	Non-sensitizer
39	n-Octanoic acid	124-07-2	TCI	> 98	Liquid	NC	Non-sensitizer
40	Nonanoic acid	112-05-0	TCI	> 98	Liquid	NC	Non-sensitizer
41	Benzaldehyde	100-52-7	Sigma-Aldrich	≥ 99	Liquid	NC	Non-sensitizer
42	Propylparaben	94-13-3	TCI	> 99	Solid	NC	Non-sensitizer
43	Methylparaben	99-76-3	TCI	> 99	Solid	NC	Non-sensitizer
44	1-Bromobutane	109-65-9	TCI	> 98	Liquid	NC	Non-sensitizer
45	2-Acetylcyclohexanone	874-23-7	TCI	> 97	Liquid	NC	Non-sensitizer
46	Propylene glycol	57-55-6	Sigma-Aldrich	≥ 99.5	Liquid	NC	Non-sensitizer

CAS No., Chemical Abstracts Service Number; NC, not calculated; –, purity not specified by suppliers; EC3, effective concentration of a test chemical required to induce a threefold increase in LLNA

Optimization of reaction conditions of GSH with DTNB

First, the reactivity of DTNB, a thiol-detecting agent, with four thiol candidates (cysteamine, homocysteine, cysteine, and GSH) was tested under three different pH conditions in 96-well transparent flat-bottom plates (Kennebunk, ME, USA). A thiol compound (100 μL) prepared in either 0.1 M sodium acetate buffer (pH 4.0), 0.1 M potassium phosphate buffer (pH 7.4), or 0.1 M ammonium acetate buffer (pH 10.0) was mixed with 50 μL of either acetonitrile or DMSO, vehicles for dissolving test chemicals. Thereafter, the samples were stored at 4 °C or room temperature for 0, 3, 6, 18, or 24 h. At a given time, 50 μL of DTNB in 0.1 M potassium phosphate buffer (pH 7.4) was mixed and reacted for 20 min, and the absorbance was measured at 415 nm. To minimize volume changes, the 96-well plate was tightly sealed with sealing tape and Parafilm. For the subsequent optimization of experimental conditions, 0.1 M ammonium acetate buffer (pH 10) was used for dissolving GSH and the incubation was done at 4 °C, because these conditions were proved to be the best among other conditions as described in the Results section.

For the optimization of vehicle concentration, 75 μL of either acetonitrile or DMSO at 0, 20, 40, and 100% (0, 10, 20, and 50% in final, respectively) was incubated with 75 μL of 600 μM GSH (300 μM in final) for 24 h. For the optimization of the concentration of GSH and DTNB, 120 μL GSH (0–0.5 mM, final) was mixed with 30 μL either acetonitrile or DMSO (20%, final), followed by an incubation for 24 h. The vehicles for the test chemicals were selected based on the test guidelines for DPRA [21]. To optimize the reaction time between GSH and DTNB, 120 μL of 375 μM GSH (300 μM, final), 30 μL of either acetonitrile or DMSO (20%, final), and 50 μL of 400 μM DTNB were mixed, and absorbance was measured at 415 nm after the given time.

Optimization for incubation of GSH with test chemicals

In order to optimize the molar ratio for the incubation of GSH with test chemicals, 375 μM GSH (120 μL) dissolved in 0.1 M ammonium acetate buffer (pH 10) was incubated with 45 mM (1:30), 22.5 mM (1:15) and 15 mM (1:10) solutions (30 μL) of 4 sensitizers and 2 non-sensitizers dissolved in either acetonitrile or DMSO in a 96-well plate and incubated for 24 h at 4 °C. During incubation, the 96-well plate was tightly sealed with sealing tape and Parafilm. Following the incubation, 50 μL of 400 μM DTNB dissolved in 0.1 M potassium phosphate buffer (pH 7.4) was added, and the absorbance was measured 20 min later at 415 nm. To optimize the reaction time of GSH with test chemicals, 120 μL of GSH at 375 μM was incubated with 30 μL of test chemicals prepared at 22.5 mM (1:15) in either acetonitrile or DMSO (20%, final) in 96-well plates. The mixture was incubated at 4 °C for 3, 6, or 24 h.

Main study with 46 test chemicals

The main experiments to determine the predictive capacity under the optimized test conditions were conducted using 46 test chemicals. First, each test chemical was dissolved in either acetonitrile or DMSO to be 15 mM, and 30 μL was dispensed into a 96-well plate. And then, 120 μL of either 375 μM or 250 μM GSH dissolved in 0.1 M ammonium acetate buffer (pH 10) were added into each well. The final GSH:test chemical ratios were 1:10 (300 μM:3 mM) and 1:15 (200 μM:3 mM). Following an incubation for 24 h at 4 °C, 50 μL of 400 μM DTNB was added and incubated at room temperature for 20 min. Absorbance was measured, using a Tecan Spark M10 plate reader. Appropriate controls were prepared with the same volume of vehicle, depending on the solvent used to dissolve the test chemicals. During incubation, the 96-well plate was tightly sealed with sealing tape and Parafilm to minimize the evaporation of volatile test chemicals and vehicles. No visible changes in volume were observed following incubation (data not shown).

Data analysis

The percent depletion of GSH by a test chemical was calculated from the absorbance observed by using the following formula: percent depletion of GSH = (X − Y)/X × 100, where ‘X’ was the mean absorbance of control group without test chemical but vehicle, and ‘Y’ was the absorbance of test group in the presence of test chemical. When the percentage depletion was lower than 0 or higher than 100, it was regarded as 0 or 100, respectively. All samples were prepared and analyzed in triplicate, and the mean percentage depletion of GSH ± S.D. was determined.

Modeling of cut-off criteria

Statistical modeling was used to predict skin sensitization to the 46 chemicals tested. For the analyses, the percent depletion of GSH by individual test chemicals dissolved in individual vehicles and incubated for 24 h was considered as four separate modeling units (i.e., GSH:test chemical dissolved in acetonitrile at 1:15 and 1:10, and GSH:test chemical dissolved in DMSO at 1:15 and 1:10). Each modeling group was subjected to receiver operating characteristics (ROC) analysis using the NCSS statistical software (NCSS 2020, LLC, version 20.0.1, Kaysville, UT, USA), and the appropriate cut-off values were generated from the individual modeling units. Sensitivity, specificity, and accuracy were calculated for individual conditions using appropriate cut-off values. In addition, the consistency of the percentage of test chemicals with three consistent judgements was calculated to estimate the within-laboratory reproducibility of the modeling units.

Results

Method optimization

To use thiol compounds as acceptors for skin sensitizers, a convenient method capable of quantitatively detecting thiol groups is required. Among many thiol-detecting agents, DTNB was used in this study. In a preliminary study, the reactivity of four endogenous thiol compounds, GSH, cysteamine, cysteine, and homocysteine, with DTNB was tested under three different pH conditions, and it was found that all thiol compounds dissolved at pH 10 reacted well with DTNB (data not shown). GSH reacted in a concentration-dependent manner with various concentrations of DTNB in the presence of acetonitrile and DMSO, which were the vehicles for dissolving the test chemicals (Supplementary Fig. S1A and S1B). When the absorbance of GSH, DTNB, and their reaction products was scanned in the range from 350 to 550 nm, the results showed that, in contrast to either GSH or DTNB alone (data not shown), the reaction product showed an absorption maximum at approximately 415 nm following the reaction of GSH with DTNB, indicating that GSH depletion by the test chemicals might be quantitatively determined by using DTNB reactivity (Supplementary Fig. S1C and S1D). Additionally, the product exhibited stable absorption for up to 1 h (Supplementary Fig. S1E and S1F).

To optimize the test conditions, thiol compounds including GSH at 400 μM were tested in buffers with pH 4.0, 7.4 and 10.0. The reactivity of candidate endogenous thiol compounds were incubated with either acetonitrile or DMSO for 24 h at either 4 °C or room temperature. All 4 thiol compounds revealed similar results that, no matter which vehicles were used, the reactivity of thiol compounds with DTNB was stable at the condition of pHs 4.0 and 10.0, and that the incubation at 4 °C would be better than that at room temperature (data not shown). The GSH results are shown in Supplementary Fig. S2. GSH at 400 μM prepared in 3 different pHs were incubated with 33% either acetonitrile or DMSO for up to 24 h at either 4 °C (Supplementary Fig. S2A and S2B) or room temperature (Supplementary Fig. S2C and S2D). And then, 500 μM of DTNB was added and incubated for 20 min at room temperature for the color development. The results suggested that, for the main study with test chemicals, GSH in 0.1 M ammonium acetate buffer, pH 10.0, would be optimal for the incubation with test chemicals up to 24 h at 4 °C, although the reactivity of GSH with DTNB were somewhat reduced when DMSO was used as a vehicle. Subsequently, the effects of vehicles on the reactivity of GSH with DTNB were tested, because the vehicle concentration required to dissolve the test chemicals is critical for the development of in vitro test methods. When 300 μM GSH (in final) dissolved in 0.1 M ammonium acetate buffer, pH 10, was incubated with various concentrations of vehicles at 4 °C for 24 h, followed by an incubation with 400 μM DTNB, both acetonitrile and DMSO did not affect the reactivity, except 50% concentration (Supplementary Fig. S3). Considering that the lower the vehicle concentration, the higher is the possible precipitation of test chemicals during incubation with GSH, we decided to use 20% of the vehicle for subsequent studies.

Subsequently, to determine the reaction conditions for GSH with the test chemicals, the GSH:test chemical ratio and incubation time were optimized (Fig. 1). Four well-known sensitizers and two non-sensitizers were tested to optimize the ratio of GSH to test chemicals. The ratios of GSH to the test chemicals were 1:10, 1:15, and 1:30 in acetonitrile (Fig. 1a) and DMSO (Fig. 1b). The results clearly showed that GSH depletion by the sensitizer increased as the reaction ratio increased. Based on these results, 1:30 was considered as the optimal ratio; however, DNCB showed a slight tendency to precipitate when reacted with GSH (data not shown). Therefore, 1:10 and 1:15 were used as the final reaction ratios in the main study. Subsequently, the time-dependent depletion of GSH was evaluated at 1:15 ratio (Fig. 1c, d). Among the time points tested, all sensitizers and non-sensitizers were well separated, based on the depletion of GSH after the 24-h incubation in both vehicles. Therefore, a 24-h incubation schedule was selected for this study.

Fig. 1.

Optimization of incubation concentration and time for GSH with 6 test chemicals. a, b Optimization of incubation concentration. GSH at 300 μM in 0.1 M ammonium acetate buffer at pH 10 was incubated with 1:10, 1:15 and 1:30 of test chemicals dissolved in either acetonitrile (a) or DMSO (b) for 24 h at 4 °C. The final concentration of each vehicle was 20%. c, d Optimization of incubation time. GSH at 300 μM in 0.1 M ammonium acetate buffer at pH 10 was incubated with 1:15 of test chemicals dissolved in either acetonitrile (c) or DMSO (d) (20%, in final) for 3, 6, and 24 h at 4 °C. And then, the mixture was incubated with 400 μM DTNB for 20 min at room temperature. Each bar represents the mean percent of GSH depletion + S.D. of triplicate determinations. The percent depletion < 0 was regarded as 0. DNCB, 2,4-dinitrochlorobenzene; p-PD, p-phenylenediamine; MEG, methyl eugenol; EG, eugenol; MS, methyl salicylate; IPA, 2-propanol

Main study with 46 test chemicals

Using the optimized four conditions, 46 chemicals (23 sensitizers and 23 non-sensitizers) were tested to evaluate their predictive capacity. The test chemicals dissolved in either acetonitrile or DMSO were incubated with GSH for 24 h at 4 °C in two different concentration ratios (i.e., 1:10 and 1:15). The results are presented in Supplementary Tables S1–S4. Except for glutaraldehyde and formaldehyde, most of extreme, strong and moderate sensitizers depleted GSH. Among the six strong sensitizers, p-hydroquinone and propyl gallate showed color interference, making classification difficult. Therefore, these two chemicals were excluded from the predictive capacity calculations. In the case of weak sensitizers dissolved in acetonitrile, three chemicals (2,3-butanedione, cinnamyl alcohol and phenyl benzoate) produced false negatives at high concentrations, whereas only one chemical (cinnamyl alcohol) showed false negatives at low concentrations. In contrast, in the case of DMSO, all the weak sensitizers were identified well under the conditions tested. Among the 23 non-sensitizers, Tween 80, which was classified as a non-sensitizer in LLNA, showed false positives under all four conditions. In addition, benzaldehyde, vanillin, 1-bromobutane and 2-acetylcyclohexanone showed false positives when specific cut-off values derived from receiver operating characteristic (ROC) analyses were applied.

Cut-off values and prediction modeling

Using the results from the tested chemicals, ROC analyses were performed to obtain the optimal cut-off values for the individual models (Fig. 2). All data obtained from the three runs of individual test chemicals were subjected to ROC analysis for each model. The sensitivity, specificity, and accuracy for the four test conditions were calculated from the individual models based on the obtained cut-off values, as presented in Table 2. Similarly, the consistency of the four individual test conditions was also obtained. All results are presented in Supplementary Tables S1–S4, based on the cut-off values obtained from the ROC analyses for the four individual conditions. The conditions under which the test chemicals were dissolved in DMSO at a 1:10 reaction ratio showed the best results, with 95.2% sensitivity, 91.3% specificity, 93.2% accuracy, and 93.2% consistency.

Fig. 2.

Receiver operating characteristic (ROC) curves of the prediction models with 4 different incubation conditions. 200 μM and 300 μM GSHs were incubated with 15-fold and 10-fold of test chemicals, respectively, dissolved in 20% either acetonitrile or DMSO for 24 h at 4 °C. And then, the reactant was incubated with 400 μM DTNB for 20 min at room temperature. a GSH at 200 μM incubated with 15-fold of test chemicals dissolved in acetonitrile. b GSH at 300 μM incubated with tenfold test chemicals dissolved in acetonitrile. c GSH at 200 μM incubated with 15-fold test chemicals dissolved in DMSO for. d GSH at 300 μM incubated with tenfold test chemicals dissolved in DMSO

Table 2.

Predictive capacity and consistency of individual prediction models for the depletion of GSH with 4 different incubation conditions

Solvent	Test chemicals	Cut-off (%)	Sensitivity (%)	Specificity (%)	Accuracy (%)	Consistency (%)
Acetonitrile	15-fold	≥ 8.5	76.2 (16/21)	87.0 (20/23)	81.8 (36/44)	90.9 (40/44)
	10-fold	≥ 6.0	90.5 (19/21)	91.3 (21/23)	90.9 (40/44)	72.7 (32/44)
DMSO	15-fold	≥ 8.7	90.5 (19/21)	82.6 (19/23)	86.4 (38/44)	90.9 (40/44)
	10-fold	≥ 5.8	95.2 (20/21)	91.3 (21/23)	93.2 (41/44)	93.2 (41/44)

The cut-off values were determined by analyzing corresponding receiver-operating characteristic (ROC) curves. To determine the cut-off value for these prediction models, total three results of each trial were subjected to ROC analysis. Optimal cut-off values were obtained by plotting true positive values (sensitivity) in y-axis against false positive values (1-specificity) in x-axis. The value that provided the highest Youden index and the point with the lowest distance to the upper-left corner of the ROC curve was chosen as the best cut-off value

Comparison with other alternative methods listed in OECD test guidelines

As shown in Supplementary Table S5 and Table 3, the predictive capacity of the current method was compared with some alternative test methods listed in the OECD test guidelines, based on the results from the 46 test chemicals tested in the present study. The individual results of the OECD-approved test methods were obtained from the literatures [22–24]. Although the simple reactivity of GSH with the test chemicals was adopted in the current test, requiring neither animal use nor cell culture, the predictive capacity of the current test was as high as that of other OECD-approved test methods.

Table 3.

Comparison of predictive capacity of current study with DPRA, KeratinoSens™ and h-CLAT

	Current study	DPRA	KeratinoSens™	h-CLAT
Sensitivity (%)	95.2 (20/21)	91.3 (21/23)	95.5 (21/22)	95.5 (21/22)
Specificity (%)	91.3 (21/23)	78.3 (18/23)	60.9 (14/23)	54.5 (12/22)
Accuracy (%)	93.2 (41/44)	84.8 (39/46)	77.8 (35/45)	75.0 (33/44)

Each value was derived from Supplementary Table S5

Discussion

Owing to the ban on animal testing for cosmetics and their ingredients and the implementation of European regulations on the registration, evaluation, licensing, and restriction of chemicals (REACH), alternative testing methods that limit animal use have been extensively developed. In particular, various in vitro and in chemico methods for skin sensitization have been developed using the three Rs concept, replacement, reduction, and refinement [1]. To induce chemically induced skin sensitization, the following four key steps of adverse outcomes should occur in sequence: covalent interaction of sensitizers with skin proteins, activation of keratinocytes to secrete inflammatory cytokines, activation and mobilization of dendritic cells to local lymph nodes, and activation of T cells to cause inflammation. Based on these key events, several alternative tests for skin sensitization have been developed and adopted as OECD test guidelines, such as DPRA for key event 1; KeratinoSens™ and LuSens™ for key event 2; and h-CLAT, U-SENS™, and IL-8 Luc assays for key event 3; Among these, DPRA is an alternative test method that mimics the hapten-protein complex formation of a chemical with proteins [14]. In addition, neither animals nor cell cultures are required to perform DPRA, making the test as convenient as possible.

The current study was designed to develop a more economical and simpler test method than conventional DPRA for distinguishing skin sensitizers by imitating the haptenization process, an early stage of skin sensitization, under conditions that require neither experimental animals nor animal cell cultures. Although not directly related to the process of skin sensitization, it is generally expected that reactive sensitizers form covalent complexes with GSH, cysteine, cysteamine, and homocysteine, which are thiol-containing compounds in cells. When this hypothesis is proven successful, it could be used as a screening test method for skin sensitizers to quickly determine the test chemical to proceed to the next stage of development for a large number of candidate substances. By measuring the reactivity of a chemical to a nucleophilic peptide in DPRA, it is possible to evaluate its potential for chemical-induced skin sensitization because a low molecular-weight sensitizer forms a complex with large molecules, including proteins [14].

GSH, an endogenous low-molecular weight substance, contains a thiol group that can easily form adducts with a variety of electrophiles, such as skin sensitizers [1]. The results of this study indicate that, to some extent, our assumption that GSH reacts with sensitizers is feasible. In this study, a convenient spectrophotometric method was developed to overcome the significant drawbacks of DPRA. First, spectrophotometry is considerably simpler than HPLC. Second, the present method has the advantage of enabling the experimenter to simultaneously measure multiple samples in the flat-bottom form of a 96-well transparent plate, with a maximum of 23 test chemicals when triplicate samples were tested with an appropriate control. When GSH reacted with DTNB, the adduct showed maximum absorption at 415 nm. More importantly, the adduct at 415 nm was not only distinguished from the absorbance of either the test chemical or GSH but was also quantitatively formed to determine the degree of depletion of GSH by the test chemicals (Supplementary Fig. S1). In this regard, spectrophotometric analysis of cysteine-containing hepta-peptides with DTNB, rather than HPLC, was also proven to be quantitative, as a modification of DPRA [25]. In addition, through the optimization process for various conditions, GSH was stable when incubated under refrigerated conditions (4 °C) in either acidic or basic buffer regardless of the vehicle used (Supplementary Figs. S2 and S3). In addition, in preliminary studies with several test chemicals, many compounds could not be correctly identified during incubation with GSH under acidic conditions (data not shown). Therefore, this study was conducted under basic conditions (pH 10.0). Although a GSH:test chemical ratio of 1:30 showed the best conditions for separating sensitizing chemicals from non-sensitizing chemicals (Fig. 1), 1:10 and 1:15 were selected for the main study, because of their tendency to be precipitated by certain test chemicals, such as DNCB (data not shown). Following optimization of several experimental conditions, 46 test substances, including 23 sensitizers and 23 non-sensitizers, were used to evaluate the predictive capacity of the current method. The reactivity of test chemical with GSH was measured after a 24-h incubation at 4 °C.

Initially, each reaction was performed in a glass test tube with a relatively large volume of reagent solution. Subsequently, to make the test system as convenient and simple as possible, and to reduce the required amount of test chemicals as much as possible, a 96-well plate was used. When using a 96-well plate, an error occurred in the absorbance measurement owing to the volume change by evaporation or volatilization of the solution after incubation for 24 h. To solve this problem, the plate was not only tightly sealed with a wide sealing tape, but also sealed again at the edge with Parafilm to prevent volume reduction. Thus, no significant volume change following a 24-h incubation was observed (data not shown).

To be adopted as a verified method in the OECD test guidelines, the predictive capacity, including sensitivity, specificity and accuracy, should be at least 80.0% [26]. As shown in Table 2, the current study demonstrated a high degree of predictive capacity for all four conditions tested. In particular, sensitivity (95.2%), specificity (91.3%), and accuracy (93.2%) were the highest, with 93.2% consistency when 5.8% depletion was applied as a cut-off value for sensitizers with DMSO, a vehicle, and a 1:10 ratio of GSH:test chemical. When compared with other three in vitro methods listed in the OECD test guidelines, the predictive capacity of present study was found to be comparable, as shown in Supplementary Table S5 and Table 3 [22–24]. In addition, the current method was judged to be simple and economical alternative test method, as neither sophisticated instruments such as synthetic peptides, HPLC, or flow cytometry were used, nor animals or cell cultures were required [27–31]. Based on the present results, a scheme of the current test method is shown in Fig. 3.

Fig. 3.

The final protocol for the method developed in this study

Meanwhile, some limitations of the current method have been observed. First, the current method cannot identify the grade of skin sensitizers as DPRA can. Second, test chemicals that would not be dissolved in either acetonitrile or DMSO, which would cause precipitation when a buffer is added, might not be applicable to this method. Third, test chemicals with self-absorption at 415 nm, a specific peak of the GSH-DTNB complex, might not be applicable to this method because of color interference. For example, some chemicals (p-benzoquinone, p-hydroquinone, 2-aminophenol, propyl gallate and 4-amino-m-cresol) exhibited various colors at a pH of 10. Among these chemicals, p-hydroquinone and propyl gallate were excluded from the calculation of predictive capacity in the present study, because they showed significant color interference at 415 nm. In this regard, our previously developed fluorometry method can be used to avoid color interference [19]. In addition, the maintenance of the test volume during the 24-h incubation is critical, because the reduction of the test volume would lead to errors in the quantitative measurement of GSH depletion. Therefore, the plate was sealed tightly with side-sealing tape during the incubation period. No significant changes in volume were observed for any of the tested chemicals (data not shown). To develop the current method as an alternative test for skin sensitization to unknown chemicals, the acceptance criteria would be helpful for evaluating the valid runs, which have not been performed in the study. In this context, based on the results of this study, DNCB and 2-propanol are recommended for evaluating valid runs as positive and negative control chemicals, respectively.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

TCJ contributed to the study conception and design. Material preparation, data collection and analysis were performed by DHC, GHK, RUN, and MRN. The first draft of the manuscript was written by DHC and all authors commented on the first version of the manuscript. All authors have read and approved the final manuscript, and TCJ finalized the manuscript.

Funding

This work was supported by a Grant from National Research Foundation, Korea (NRF-2022R1I1A3063957).

Data availability

The datasets generated and/or analyzed in the current study are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

The authors have no conflict of interest to disclose.