The Journal of Toxicological Sciences
Online ISSN : 1880-3989
Print ISSN : 0388-1350
ISSN-L : 0388-1350
Original Article
Optimisation of Photo-KeratinoSens™ for in vitro photoallergenicity assessment
Tomomi AtobeSuttinont ChawapunYuri HatakeyamaShiho OedaToshiyuki OhtakeTakao AshikagaHirokazu KouzukiMorihiko HirotaAkiko Tamura
Author information
JOURNAL FREE ACCESS FULL-TEXT HTML

2025 Volume 50 Issue 8 Pages 399-412

Details
Abstract

The aim of this work was to develop and validate a cell-based in vitro assay for predicting photoallergenicity by expanding the scope of our previously reported in vitro method, photo-KeratinoSens™, which is a luciferase-based assay dependent on activation of the Keap1-Nrf2-ARE pathway. First, we increased the maximum starting test concentration from a fixed 2000 µM in the original KeratinoSens™ to either 5000 μg/mL or 4 times the concentration providing a cell viability of 75% (under UV irradiation), depending on the cytotoxicity. Then, we established that 0.5% ethanol, 0.5% acetone and 0.1% tetrahydrofuran are available as solvents in addition to 1% DMSO, which is used in the standard KeratinoSens™ method (OECD TG442D). We confirmed that a representative photoallergen, 6-methylcoumarin, gave reproducible results. To validate the developed assay, we used it to evaluate a library of 90 chemicals consisting of 60 known photoallergens and 30 non-photoallergens. The accuracy, sensitivity, specificity and balanced accuracy of photo-KeratinoSens™ were 76.7% (69/90), 66.7% (40/60), 96.7% (29/30) and 81.7%, respectively. When we excluded chemicals with no UVA absorption, the accuracy, sensitivity, specificity and balanced accuracy were improved to 81.8% (45/55), 80.0% (36/45), 90.0% (9/10) and 85%, respectively. Our results suggest that photo-KeratinoSens™, which combines the OECD TG442D in vitro skin sensitization test detecting key event 2 in the adverse outcome pathway (AOP) of skin sensitization with exposure to UV iradiation, may be useful as a contributory input in a weight-of-evidence approach for evaluating photoallergenicity potential without animal testing.

INTRODUCTION

Photoallergy is provoked as a consequence of an immunotoxicological response to photoallergens under light irradiation, and is an important safety concern in the development of cosmetics and pharmaceuticals for application to the skin. The photosensitizing potential of chemicals has been evaluated by using animal models and the adjuvant and strip method (Sato et al., 1980) and the Herber method (Herber, 1969), but recent changes in regulatory requirements, driven by ethical considerations, mean that the safety assessment of cosmetic ingredients now is desired to be conducted without animal tests.

Phototoxicity and photoallergy can be triggered by exposure to ultraviolet (UV)/visible (VIS) light (290-700 nm), especially UVA, after oral or dermal application of chemicals and is related to the production of reactive oxygen species derived from the chemicals. In the case of UV/VIS spectral analysis (OECD, 1981), the OECD test guidelines No. 432 and 498 suggest that no further photosafety testing is needed if the molar extinction coefficient (MEC) of a chemical is less than 1,000 L·mol-1·cm-1 (OECD, 2019a, 2019b). In the case of in chemico ROS assay which is adopted as a photoreactivity test method (OECD, 2019b) is negative, we can conclude that there is no concern about photosafety, including photosensitization (OECD, 2019b; Guidance, 2022). On the other hand, if the ROS assay result is positive, follow-up testing is necessary though ROS assay has high sensitivity but low specificity. However, there is still no effective test method to assess photoallergic potential, because the mechanism of photoallergenicity is very complex, as in the case of skin sensitization (Onoue et al., 2017).

A photoallergenic compound is thought to trigger an immune response similar to that involved in sensitization via the formation of photoreaction products such as protein adducts (Tokura, 2018). In skin sensitization, the adverse outcome pathway (AOP) and three key events (KEs) (KE1: binding of haptens to intrinsic skin proteins, KE2: activation of keratinocytes, KE3: activation of dendritic cells) have been proposed (OECD, 2012), and an integrated approach to testing and assessment (IATA) combining several test methods (in silico, in chemico and in vitro) based on the AOP KEs is considered desirable (OECD GD256). Therefore, it is also important to combine tests detecting the three KEs for skin sensitization under conditions of UV irradiation. Photo-direct peptide reactivity assay (DPRA) (Nishida et al., 2021) and photo-amino acid derivative reactivity assay (ADRA) (Yamamoto et al., 2020) have been reported as in chemico methods to detect KE1 under UV irradiation, and DPRA and ADRA were adopted in the OECD Test Guidelines TG 442C (OECD, 2021) for skin sensitization evaluation. A photo-SH/NH2 test which detects changes of cell-surface thiols and amines due to hapten-protein binding under UV irradiation was also reported (Oeda et al., 2016).

For KE2, KeratinoSens™ assay was adopted in the OECD Test Guideline TG 442D (OECD, 2018a). KeratinoSens™ is a reporter transgenic cell developed to evaluate the Nrf2-Keap1-ARE pathway. This pathway is known to be activated by many skin sensitizers and serves as a cellular stress response mechanism. We previously reported the photo-KeratinoSens™ assay, which incorporates a UV irradiation process into KeratinoSens™ assay. The assay evaluates photoallergenicity in terms of the relative augmentation of luciferase activity in test-chemical-treated cells under conditions of UV irradiation versus non-irradiation. This assay was suggested to be effective to predict photoallergenicity, showing accuracy of 56%, sensitivity of 50% and specificity of 67%, but the number of test chemicals was only 9 (Tsujita-Inoue et al., 2016). Another test, PhotoSENSIL-18 (Nguyen et al., 2023), detects IL-18 secretion from a chemical-treated reconstructed human epidermis (RHE) model under UV irradiation. To address KE3, photo-h-CLAT, which incorporates the UV irradiation process into h-CLAT (OECD TG442E), was developed (Hino et al., 2008; Hoya et al., 2009).

In this study, we focused on photo-KeratinoSens™, which we previously reported, and reviewed the protocol in order to extend its scope. We then validated the ability of the updated assay to predict photoallergenicity using a previously reported extended dataset consisting of 60 known photoallergens and 30 non-photoallergens (Onoue et al., 2017).

MATERIALS AND METHODS

Chemicals

The 60 known photoallergens and 30 non-photoallergens used in this study are listed in Table 1 (Onoue et al., 2017). Dimethyl sulfoxide (DMSO) (≥99.7%, Sigma-Aldrich Corp.), distilled water (DW) (Otsuka Pharmaceutical Co., Ltd., Tokyo, Japan), ethanol (≥99.5%, Wako Pure Chemical Industries, Ltd.), acetone (≥99.0%, Wako Pure Chemical Industries, Ltd.) or tetrahydrofuran (THF) (≥99.5%, Wako Pure Chemical Industries, Ltd.) was employed as a solvent. The cLogP value of each chemical was calculated using KOWWIN V.1.68 in the EPI suite™ (Environmental Protection Agency, DC, USA).

Table 1. Chemicals used in this study.


Cell culture

KeratinoSens™ is a stable cell line derived from HaCaT human keratinocytes (OECD, 2018a). KeratinoSens™ cells were obtained from acCELLerate GmbH, Hamburg, Germany and maintained in Dulbecco’s modified Eagle’s medium containing Glutamax supplemented with 9% (v/v) fetal bovine serum (FBS) and 500 μg/mL G418 (Calbiochem, CA, USA) at 37°C under an atmosphere of 5% CO2 in air. Cells were used for the assay until passage 25.

UV light source

A SOL500 Sun Simulator (Dr Hönle AG, Grägelfing, Germany) equipped with a metal halide lamp (500 W) and an H1 filter to attenuate UVB was used in the same as used in the previous report. The filter was included because UVB is cytotoxic. The irradiation was carried out with an irradiance of ca. 2.0 mW/cm2 as determined with a calibrated UVA detector UVR-300 (Topcon, Tokyo, Japan). SXL-3000V2 (Seric, Saitama, Japan), which blocks UVB and is listed in the phototoxicity test guideline (OECD, 2019a) was used to confirm that there was no effect between devices.

Pre-study cytotoxicity assay

Cells were seeded in 96-well plates at a density of 30,000 cells/well in 125 μL of growth medium without G418 and pre-cultured for 24 hr at 37°C under 5% CO2. The test chemicals were dissolved in DMSO, DW, culture medium, ethanol, acetone or THF. The solvent that best dissolved the test chemical or formed a stable dispersion of the chemical at the highest concentration of 500 mg/mL was selected as the appropriate vehicle. For each test chemical, 12 two-fold dilutions were made, and these were further diluted 25-fold for DMSO, 50-fold for EtOH or acetone or 125-fold for THF with phosphate-buffered saline without Ca2+ and Mg2+ (PBS).

After pre-culture, 50 µL of diluted test chemical was added to cell culture containing 150 µL of PBS in place of the medium, so that the final concentration of DMSO was 1%. Cells were exposed to various dilutions of the test chemicals for 1 hr at 37°C under 5% CO2 and then irradiated with UV (5 J/cm2). Control samples were kept in the dark under the same conditions (non-UV). After UV irradiation, test chemical dilutions were replaced with fresh culture medium containing 1% FBS, and cell culture was continued for 48 hr at 37°C under 5% CO2.

Cells were washed twice in PBS, and the cell viability was determined using CellTiter-Glo Luminescent Cell Viability Assay Reagent with a GloMax Luminometer (Promega KK, Tokyo, Japan). The concentration providing a cell viability of 75% (CV75) was calculated for each chemical by interpolation on the log-linear dose-response curve.

Main study luciferase assay

Photo-KeratinoSens™ was performed as described by Tsujita-Inoue et al. (2016) with some modifications. Briefly, cells were seeded and treated with chemicals in the same way as described above. The test doses were set from 1/512 × CV75 to 4 × CV75 (final concentrations). If the CV75 could not be determined, the test doses were set at 12 two-fold dilutions from the maximum concentration of 5000 μg/mL (for DMSO-soluble chemicals), 2500 μg/mL (for ethanol-soluble or acetone-soluble chemicals) or 1000 μg/mL (for THF-soluble chemicals) or from the concentration corresponding to the maximum solubility in each solvent.

UV irradiation and subsequent medium exchange were carried out as in the pre-study cytotoxicity assay. At 48 hr after UV irradiation, cells were washed twice in PBS, and luciferase induction was measured using a Steady-Glo Luciferase Assay System (Promega KK, Tokyo, Japan) with a GloMax Luminometer. The cell viability was tested in parallel assays run under the same conditions and with the same test concentrations as described above.

Data analysis

The assay was repeated three times for each chemical. Each experiment was considered an individual replicate. The Imax ratio (UVA+/UVA-) was determined as the maximal fold induction of luciferase activity under UV irradiation versus non-irradiation measured at any concentration for each run. The EC1.5 value was determined as the threshold concentration when 1.5-fold luciferase induction was set as the criterion for positive or negative judgment or the maximal test concentration if the luciferase induction did not exceed 1.5-fold at any concentration for each run. The final EC1.5 value, highlighted in bold in Table 2, is the median value of 3 repetitions. A chemical was assessed as positive if the Imax ratio (UVA+/UVA-) was higher than 1.5 in at least two out of three repetitions. When cell viability was less than 70% the relative luciferase activity was not used in the determination, as described in the original KeratinoSens™ protocol.

Table 2. Summary of results of UV/VIS spectral analysis and photo-KeratinoSens™.


RESULTS

Distribution of chemicals in the database

The chemicals shown in Table 1 were selected from the previous papers about photosafety. The chemicals in the database were separated into photoallergens and non-photoallergens, grouped according to molecular weight (MW) and cLogP (Fig. 1). The majority of test chemicals, both photoallergens and non-photoallergens, fall in the range between MW 100 and MW 400 (Fig. 1(A)), and the database contains few compounds with MW below 100 or above 500. The cLogP for the majority of the tested compounds lies between -1 and 3, though the range extends from -2 up to >7 (Fig. 1(B)).

Fig. 1

Distribution of (A) molecular weight and (B) cLogP of the total of 90 chemicals.

Confirmation of applicable solvents and concentrations

We investigated whether ethanol, acetone and THF could be used as applicable solvents for photo-KeratinoSens™ because some photoallergens were not soluble in DMSO which was usable in KeratinoSens™ (OECD TG442D). In terms of relative luciferase induction and cytotoxicity, 0.5% ethanol, 0.5% acetone, 0.1% THF were confirmed to be equivalent to 1% DMSO (Fig. 2). In this study, ethanol was used for 9 chemicals.

Fig. 2

Effect of UV irradiation in the presence of four different test chemical solvents on luciferase activity and cell viability. Cells were exposed to each solvent and irradiated with 5 J/cm2 UV. The concentrations assayed were: 1% DMSO, 0.5% EtOH, 0.5% acetone, 0.1% THF. The UV symbols "-" and "+" indicate cells irradiated or not irradiated with UV, respectively. Each bar represents the mean luciferase unit ± S.D. of three independent experiments.

Collection of historical data for the positive control

6-Methylcoumarin (6-MC) is known to induce photosensitivity in both humans and animals, and is an easy-to-handle fragrance in terms of availability, solubility and stability. Furthermore, it showed stable relative luciferase induction with its low cytotoxicity. Thus, it was selected as a positive control. Fig. 3 summarizes the results obtained for serially diluted 6-MC in the concentration range from 20 to 320 μg/mL. The fold induction of luciferase activity under UV irradiation or non-irradiation were similar, and the ratio of fold induction under UV irradiation versus non-irradiation were significantly higher than the threshold of 1.5 in any dose regardless of which solvent was used (Fig. 3(A)). The ratio of fold induction of luciferase activity under UV irradiation versus non-irradiation at each concentration using DMSO solvent showed concentration dependence in the range of 20 to 320 μg/mL (Fig. 3(B)). The average threshold concentration for the 1.5-fold relative luciferase induction (EC1.5 value) of 79 test data using DMSO solvent was 23.8 μg/mL.

Fig. 3

(A) Effect of solvents on luciferase activity and viability of cells treated with 6-MC under UV irradiation. Each bar represents the mean relative fold induction ± S.D. from three independent experiments. DMSO without UV (open bar), DMSO with UV (filled bar), EtOH without UV (open hatched line), EtOH with UV (filled hatched line), acetone without UV (open crosshatched line), acetone with UV (filled crosshatched line), THF without UV (open vertical line), THF with UV (filled vertical line). Each line represents the mean cell viability of three independent experiments. DMSO without UV (open circle), DMSO with UV (filled circle), EtOH without UV (open triangle), EtOH with UV (filled triangle), acetone without UV (open square), acetone with UV (filled square), THF without UV (open diamond), THF with UV (filled diamond). The dotted line shows the 1.5-fold criterion for positive result. Statistical analysis was performed with t-test. *P < 0.05, **P < 0.01. (B) Distribution of fold induction ratios in irradiated and non-irradiated cells treated with 5 concentrations of 6-MC (79 tests). Boxplots show the first quartile, median, and third quartile, with markers at the minimum and maximum values.

Predictive performance of photo-KeratinoSens™, and comparison with photo-ARE assay

The results of photo-KeratinoSens™ are summarized in Table 2. Among the 60 photoallergens, 40 were detected as positive, while all but one of the 30 non-photoallergens were detected as negative. The overall accuracy of the assay was therefore 76.7% (69/90), the sensitivity was 66.7% (40/60), the specificity was 96.7% (29/30) and the balanced accuracy was 81.7% (Table 3(A)). UVB irradiation was blocked in this assay, so when the 35 chemicals (15 photoallergens and 20 non-photoallergens) whose maximal MEC value (320-700 nm) was under 1000 were excluded from the calculation of predictive performance, the calculated accuracy, sensitivity, specificity and balanced accuracy were 81.8% (45/55), 80.0% (36/45), 90.0% (9/10) and 85% (Table 3(B)). The improvement in sensitivity is particularly large.

Table 3. Predictive performance of photo-KeratinoSens™ with (A) a dataset consisting of 90 chemicals and (B) the same dataset excluding 35 chemicals that absorb only UVB.


We previously reported a photo-ARE assay based on activation of the Keap1-Nrf2-ARE pathway as photo-KeratinoSens™ (Tsujita-Inoue et al., 2016). Comparison of the results of the two test methods showed similar predictive performance for 21 chemicals which both data were available (Table 4).

Table 4. Comparison of results in photo-KeratinoSens™ and photo-ARE assay.


Possibility of using SXL-3000V2

In this study, SOL500 was used for UV irradiation. For SXL-3000V2 (Seric, Saitama, Japan), which is listed in the phototoxicity test guideline (OECD, 2019a), the predictive performance was confirmed using it in place of SOL500. The results of relative luciferase induction and cytotoxicity showed no significant difference from SOL500 for the serially diluted 6-MC with DMSO in the concentration range from 20 to 320 μg/mL (Fig. 4 compared to Fig. 3(A)).

Fig. 4

Effect of using SXL-3000V2 on cellular responses. Cells were irradiated with 5 J/cm2 UV using an SXL-3000V2 apparatus. Each value of luciferase unit is the mean ± S.D. of five independent experiments. The dotted line shows the 1.5-fold criterion for positive result. Statistical analysis was performed with t-test. *P < 0.05, **P < 0.01.

DISCUSSION

The ARE-Nrf2 luciferase test method using the human keratinocyte cell line KeratinoSens™ has been developed and validated as an in vitro sensitization test (OECD, 2018a). Here, in order to confirm that Keratino Sens™ cells can be used to predict photoallergenicity, as well as allergenicity, we measured relative luciferase induction under UV irradiation versus non-irradiation in KeratinoSens™ cells treated with photo-allergens. We previously reported that 4 of 7 photoallergenic/phototoxic chemicals were judged as positive in this assay (Tsujita-Inoue et al., 2016), but the number of chemicals used in that evaluation was very small. In this study, therefore, we evaluated a library of 90 chemicals, including the photoallergens used in our previous paper (Onoue et al., 2017).

In terms of the distribution of MW and cLogP, the physico-chemical properties of photoallergenic/phototoxic chemicals are different from those of skin sensitizers. The distributions of MW and cLogP in the photoallergenic/phototoxic chemicals were in higher ranges than those of skin sensitizers reported previously (Natsch et al., 2013). The reason why the MW is skewed towards the high MW side is that our test chemicals contain specific functional groups and structures that absorb light and convert energy. Furthermore, the reason for the bias towards high cLogP is that the UV absorption of chemicals is affected by the number of conjugated double bonds, so that compounds with high UV absorbance might tend to have high cLogP. The newly optimized test method is able to evaluate chemicals with a wide range of physical properties.

The starting dose was set to 5000 µg/mL or 4 times CV75 (under UV irradiation), taking into account the cytotoxicity, instead of the fixed maximum starting concentration of 2000 µM in the original KeratinoSens™ for assessing skin sensitization. For many chemicals, the EC 1.5 concentrations showed some cytotoxicity, and concentrations that were not cytotoxic or were highly cytotoxic showed lower relative luciferase induction. Many photosensitive/phototoxic chemicals have relatively high MW, so from the perspective of safety assessment, where it is desirable to reduce the number of false negatives, a concentration of 2000 µM may not be the most effective. In fact, our assay with 5000 µg/mL decreased the number of false negatives by one (5-aminosalicylic acid (mesalazine)) and increased the number of false positives by one (benzocaine) compared to the original 2000 µM condition (data not shown).

As for the available solvents in this assay, we found that the addition of ethanol enabled the evaluation of 9 chemicals in this study. We also confirmed that acetone and THF can be used. We selected 6-MC as a positive control for this assay and confirmed that it dose-dependently induced ARE-dependent luciferase activity under UV irradiation with high reproducibility. Using 6-MC, we also confirmed that the SXL-3000V2 instrument, which is recommended for UV irradiation in ROS assay (OECD TG495), can be used for UV irradiation in the present test, in place of the SOL500. Therefore, it can be said that this is a very versatile measurement method which does not have any dependence on measurement equipment.

Regarding predictive performance, the accuracy of this assay was 76.7% for our extended dataset of 90 chemicals, whereas it was 56% for the previous limited preliminary dataset of 9 chemicals. The photo-ARE assay is based on activation of the Keap1-Nrf2-ARE pathway, like photo-KeratinoSens™ (Natsch and Emter, 2008). Comparison of the results of the two test methods showed similar predictive performance for 23 chemicals (Table 4).

According to OECD TG442D, the accuracy/sensitivity/specificity of KeratinoSens™ was 77% (155/201)/78% (71/91)/76% (84/110). Compared to KeratinoSens™ for assessing skin sensitization, the sensitivity of photo-KeratinoSens™ for assessing photo-allergenicity was low. This can be attributed to (i) the applicability domain and (ii) assessment with a single test method, as described below.

(i) Applicability domain

Some photoallergens which predominantly absorb in the UVB region but have no absorption in the UVA region are unlikely to be detected by this method, since we use an H1 filter that cuts off the UVB region in order to avoid the potent cytotoxicity of UVB. The 20 false negatives include 11 photoallergens that have no absorption in the UVA region.

Photoallergens with low water solubility are also unlikely to be detected since all of the tests are performed in solution and do not mimic topical application. In general, test molecules with cLogP < 5 can be easily tested, while extremely hydrophobic molecules with cLogP > 7 are considered inapplicable for testing. However, the false negatives in this study did not have particularly high values of cLogP.

When the 35 chemicals (15 photoallergens and 20 non-photoallergens) with no absorption in the UVA region were excluded, the calculated sensitivity and the overall accuracy increased to about 80%, and the predictive performance was similar to that of KeratinoSens™ for detecting skin sensitization potential. Thus, chemicals which have no absorption in the UVA region lie outside the applicability domain.

(ii) Assessment with a single test method

Among the 9 false negatives in the photo-KeratinoSens™ for which photo-DPRA data were available, 7 were positive in photo-DPRA (Nishida et al., 2021) (Table 5). Of the only two photoallergens not detected by photo-DPRA, sulfasalazine was positive in photo-KeratinoSens™, while isoniazid was not. Chemicals that could not be assessed in photo-DPRA because of co-elution could be evaluated in photo-Keratinosens™; for example, lomefloxacin HCl was judged to be positive in photo-Keratinosens™. Though additional testing (e.g., for KE3) would be needed, our results indicate that photo-KeratinoSens™ can be used for the assessment of photo-allergenicity for a wide range of chemicals as part of an integrated testing strategy.

Table 5. Comparison of results in photo-KeratinoSens™ and photo-DPRA.


In this study, we optimized photo-KeratinoSens™ as a method for assessing the photoallergenicity potential of UV absorbing chemicals whose MEC is more than 1,000 L·mol-1·cm-1. Validation of photo-KeratinoSens™ with a library of 90 test chemicals afforded accuracy, sensitivity, specificity and balanced accuracy of 76.7%, 66.7%, 96.7% and 81.7%, respectively.

We confirmed that predictivity improved to more than 80% when the chemicals with a maximum MEC value (320-700 nm) less than 1000 L·mol-1·cm-1 were excluded, and found the appropriate range of application. For the chemicals that absorb UVB, other tests such as DPRA may be needed.We believe that it is important to develop a mechanism-based IATA that uses not only photo-Keratinosens™ for KE2 but also other in vitro tests for KE1 and 3, as well as UV/VIS spectral analysis and ROS assays for predicting the photoallergenicity of a broad range of chemicals.

ACKNOWLEDGEMENT

The authors are grateful to W.R.S. Steele for his critical reading of the manuscript.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES
 
© 2025 The Japanese Society of Toxicology
feedback
Top