Evaluation of Skin Sensitization Induced by Four Ionic Liquids

Abstract

Ionic liquids (ILs) are synthetic solvents used as replacements for volatile organic solvents. Human exposure occurs through dermal or oral routes. In rodents several ILs were reported to induce dermal toxicity, irritation, and sensitization. Due to the potential for occupational exposure, and industrial use as non-volatile solvents, 1-ethyl-3-methylimidazolium chloride (EMIM, 6.25–50% v/v), 1-butyl-3-methylimidazolium chloride (BMIM, 3.12–12.5% v/v), 1-butyl-1-methylpyrrolidinium chloride (BMPY, 0.825–6.25% v/v), and N-butylpyridinium chloride (NBuPY, 0.825–12.5% v/v) were nominated to the National Toxicology Program and evaluated for skin sensitization. The test compound was applied to the ears of female BALB/c mice daily for three days in a primary irritancy (IRR)/local lymph node assay (LLNA). Sensitization was assessed in vitro in the Direct Peptide Reactivity Assay (DPRA), KeratinoSens^™ assay, and Human Cell Line Activation Test (h-CLAT). In the LLNA, the butylated ILs, BMIM and BMPY were more potent than NBuPY (butylated), or EMIM (ethylated), which was neither an irritant nor a sensitizer. NBuPY induced skin irritation in vivo at ≥3.12% (p≤0.01), and sensitization in vitro in the KeratinoSens^™ assay and h-CLAT, but was negative for sensitization in vivo and in the DPRA. Although SI3 was not achieved, dermal treatment with 12.5% BMIM or 6.25% BMPY increased (p≤0.01) lymph node cell proliferation in the LLNA. In vitro, BMIM was positive for sensitization in the h-CLAT, and BMPY was positive in the h-CLAT and KeratinoSens^™ assay; both were negative in the DPRA. Integrated data analyses, weighted toward in vivo data, suggested that BMIM and BMPY may induce weak to mild sensitization.

Short Abstract

Due to human dermal occupational exposure, and rodent dermal toxicity, 1-ethyl-3-methylimidazolium chloride, 1-butyl-3-methylimidazolium chloride (BMIM), 1-butyl-1-methylpyrrolidinium chloride (BMPY), and N-butylpyridinium chloride (NBuPY) were evaluated for skin sensitization. In vitro, BMIM was positive in the Human Cell Line Activation Test (h-CLAT); BMPY and NBuPY were positive in the h-CLAT and KeratinoSens^™ assay. NBuPY induced skin irritation; BMIM and BMPY increased lymph node cell proliferation in the local lymph node assay, collectively indicating that BMIM and BMPY may induce weak to mild sensitization.

Introduction

Ionic liquids (ILs) are synthetic solvents that are used in chemical production processes, with applications in pharmaceutical, metallurgical, chemical, and energy industries. ILs typically consist of 3 major components: a cation, an anion, and a side chain, different combinations of which could generate approximately 10¹⁸ different ILs (Cvjetko Bubalo M et al., 2014). ILs are distinguished by low-to-negligible vapor pressure, volatility, and flammability; melting point below 100°C; high thermal stability and ionic conductivity; and powerful solubility properties. The negligible vapor pressure allows separation of the solubilized product and the IL for recycling and reuse. These characteristics have led to the investigation of ILs as “green chemistry” replacements for volatile organic solvents (VOCs) in catalysis, synthesis, separation, and polymerization processes (Landry T et al., 2005; Sipes I et al., 2007; Cvjetko Bubalo M et al., 2014). Recent investigations have shown that ILs may be effective as solvents for gene transfection (Chen H et al., 2015); electrolytes in sodium ion batteries (Hagiwara R et al., 2018), lubricants, and corrosion inhibitors (Cvjetko Bubalo M et al., 2014); delivery systems for drugs (Goindi S et al., 2014; Wang C et al., 2018); and antimicrobial agents (Mester P et al., 2015; Dobler D et al., 2016; Aditya A et al., 2018; Martini Garcia I et al., 2019; Shevachman M et al., 2020; Young GR et al., 2020), depending on the chemical structure. Reports of dermal toxicity in humans are very limited, however, dermal contact is a potential route of human exposure to ILs, especially in an occupational setting where these compounds are used in chemical engineering, plasticization, chemical separation, and manufacturing (Xie F et al., 2015; Wang C et al., 2019; Chen K et al., 2020). Consumers can be exposed through the use of IL-containing personal care products applied to the skin, such as hand sanitizer and orthodontic adhesives (Martini Garcia I et al., 2019; Shevachman M et al., 2020). Due to the low vapor pressure, IL-based products, specifically hand sanitizer, do not evaporate as quickly as conventional products, increasing the length of dermal contact (Shevachman M et al., 2020). Based on the potential for human exposure, use as replacements for VOCs, and lack of toxicity data, four ILs, 1-ethyl-3-methylimidazolium chloride (EMIM), 1-butyl-3-methylimidazolium chloride (BMIM), 1-butyl-1-methylpyrrolidinium chloride (BMPY), and N-butylpyridinium chloride (NBuPY) (Figure 1), were nominated to the National Toxicology Program (NTP) by the University of Alabama Center for Green Manufacturing for toxicological characterization, including subchronic toxicity, genetic toxicity, immunotoxicity, and absorption, distribution, metabolism, and elimination (ADME) studies (Sipes I et al., 2007; CEBS, Genetic Toxicity Data).

Figure 1.

Chemical structure of ionic liquids

The structure of an IL determines the stability, toxicity, bioavailability, and possible contamination of the natural environment by the compound. Published literature indicates that the most significant factor in the toxicity of ILs is the alkyl chain length, in particular, the interaction of the alkyl chain with the cation (Cvjetko Bubalo M et al., 2014). Short (2–3) and moderate (4–6) carbon chain length ILs exhibit more limited toxicity (Mester P et al., 2015; Mendonça CMN et al., 2018; Kusumahastuti DKA et al., 2019). Longer carbon chains, though more susceptible to degradation, enhance lipophilicity of the cation to more easily interact with both the phospholipid and protein parts of the cell membrane leading to cell membrane disruption, and are more cytotoxic (Ranke J et al., 2007; Cvjetko Bubalo M et al., 2014; Tsarpali V et al., 2015). At chain lengths >10 carbons IL toxicity plateaus (Kusumahastuti DKA et al., 2019).

Approximately 40% of occupational diseases are skin-related; allergic and irritant dermatitis account for about 90% of those diseases (Martin S et al., 2018). Allergic dermatitis is a systemic, T-cell mediated immune response, while irritant dermatitis involves temporary, localized damage to epithelial cell populations (Martin S et al., 2018). Imidazolium ILs, including BMIM, are reported to disrupt the negatively charged cell membrane structure and lipid bilayer, leading to membrane permeation and potential leakage of cytokines and other molecules (Gal N et al., 2012; Galluzzi M et al., 2018) which may contribute to the activation of the innate immune response and initiation of skin sensitization (OECD 168, 2012; Martin S et al., 2018). Although evidence is limited, rodent studies indicate that NBuPY and BMIM can induce irritation, and that BMIM may be a contact sensitizer (Sipes I et al., 2007; Cheng Y et al., 2009). The objective of this study was to evaluate the potential of BMIM, EMIM, BMPY, and NBuPY to induce allergic dermatitis (hypersensitivity) and/or dermal irritation via the traditional local lymph node, irritancy, and mouse ear swelling test assays. BMIM, EMIM, BMPY, and NBuPY were chosen to represent the most common cation classes (imidazolium, pyridinium, and pyrrolidinium) of ILs, are the starting materials for other ILs, and are produced commercially. These representative ILs were also evaluated in 3 OECD (Organization for Economic Co-operation and Development)-validated in chemico/in vitro assays: Direct Peptide Reactivity Assay (DPRA) (OECD 442C, 2015), KeratinoSens^™ assay (OECD 442D, 2018), and human Cell Line Activation Test (h-CLAT) (OECD 442E, 2018). Collectively, these assays provide information on the four key events in the OECD Adverse Outcome Pathway for Skin Sensitization: covalent interaction with skin proteins, activation of keratinocytes, activation of dendritic cells, and T-cell proliferation and activation (OECD 168, 2012).

Materials and Methods

Test Substances

BMPY (C9H20NCl, 99–100%, CAS # 479500–35-1, in vivo Lot #99/831, Solvent Innovation, GmbH, Germany, in vitro Lot #20100610, Promy Chemical, LLC, El Sobrante, CA); NBuPY (C9H14ClN, 100%, CAS #1124–64-7, in vivo Lot #99/830, Solvent Innovation, GmbH, in vitro Lot #20100610, Promy Chemical, LLC,); BMIM (C8H15N2Cl, 98%, CAS #79917–90-1, in vivo Lot #99/787, in vitro Lot #99/787, Solvent Innovation, GmbH); and EMIM (C₆H₁₁ClN₂, 99–100%, CAS #65039–09-0, in vivo Lot #S37784, in vitro Lot #STBB3624, Sigma-Aldrich, St Louis, MO) (Figure 1) were obtained through a DNTP analytical chemistry contract at Midwest Research Institute (Kansas City, MO). Chemical identification and purity analyses are summarized in the supplemental materials (CEBS, In Vitro Supplemental Data).

Direct Peptide Reactivity Assay (DPRA)

The in chemico assay DPRA was conducted in accordance with the OECD guideline (OECD 442C, 2015). The DPRA as an individual test is reported to have sensitivity, specificity, and accuracy of 75–86% relative to human data and the LLNA (Roberts DW & Patlewicz G, 2018; OECD 442C, 2019). All in vitro studies were conducted at Burleson Research Technologies (BRT, Morrisville, NC). Reactive chemicals that have the potential to cause skin sensitization bind to peptides resulting in their depletion. Test chemicals (250 μl of 100 mM in acetonitrile) were mixed with either a cysteine-containing peptide (Ac-RFAACAA-COOH, 750 μl, 0.667 mM stock in phosphate buffer, RS Synthesis, Louisville, KY) or a lysine-containing peptide (Ac-RFAAKAA-COOH, 750 μl, 0.667 mM stock in ammonium acetate buffer, RS Synthesis), and incubated at 22.5–30°C for 24 hr in the dark. The interaction of the test compound with the peptide was measured by high performance liquid chromatography (HPLC) with UV detection at 220 nm, following the flow conditions in OECD guideline 442C (2015). The peptide concentration in solution determined the percent peptide depletion and the degree of peptide reactivity. All samples and controls were analyzed in triplicate (N=3 vials/sample). The positive control for the assay was cinnamic aldehyde (10 mM in acetonitrile, Sigma-Aldrich, CAS #104–55-2). The HPLC instrument used in these studies was a Shimadzu LC-2010A_HT (Columbia, MD), with a Zorbax SB-C18 2.1 mm × 100 mm × 3.5 μm HPLC column (Agilent, Santa Clara, CA), and a Security Guard C18 4 mm × 2 mm guard column, (Phenomenex, Torrence, CA). Data analysis was performed with LabSolutions Software version 5.97 (Shimadzu). The percent depletion of each of the peptide peaks and the mean percent peptide depletion of each peptide were interpreted for each test chemical and the positive control using the prediction models outlined in the OECD testing guideline. The following OECD acceptance criteria were applied to verify the validity of the assay: a.) positive control % cysteine depletion 60.8–100% (±14.9% standard deviation), b.) positive control % lysine depletion 40.2–69.0% (±11.6% standard deviation), c.) reference control A mean peptide concentration 0.5±0.05 mM, d.) reference controls B and C coefficient of variance <15%, and e.) test chemical maximum standard deviation ≤14.9% for cysteine and ≤11.6% for lysine. A single run of both cysteine and lysine peptides was used to classify each test chemical as a non-sensitizer (mean % cysteine+lysine depletion ≤6.38%), or as a low (6.38–22.62%), moderate (22.62–42.47%), or high (42.47–100%) reactivity sensitizer.

KeratinoSens^™ Assay

The in vitro KeratinoSens^™ assay was conducted in accordance with the OECD guideline 442D (2018). The KeratinoSens^™ assay as an individual test is reported to have sensitivity, specificity, and accuracy of 76–78% relative to human data and the LLNA (OECD 442D, 2018). KeratinoSens^™ cells (Givaudan, SA, Switzerland), immortalized human adult low calcium temperature (HaCaT) keratinocyte cells (Boukamp P et al., 1988; Grafe F et al., 2004) transfected with a plasmid containing the luciferase gene whose expression is under control of an antioxidant/electrophile response element (ARE), were exposed to each test compound at 0.98–2000 μM (final) in dimethyl sulfoxide (DMSO, Sigma-Aldrich). The positive control was cinnamic aldehyde (4–64 μM (final) in DMSO). The cells (10,000 cells/well, in Dulbecco’s Modified Eagle’s medium (Gibco, Gaithersburg, MD) were cultured in 96-well plates for 48 hr at 37ºC/5% CO₂ following OECD guideline 442D (2018). After incubation, the cells were washed with phosphate buffered saline (PBS), lysed with Passive Lysis Buffer (50 μl, Promega, Madison, WI), and Steady-Glo^® luciferase substrate was added (50 μl, Promega). Activation of the Keap1-Nrf2-ARE-dependent pathway was assessed by measuring the amount of luminescence in each well using a luminometer, Molecular Devices SpectraMax^® i3 and i3x with a LUMI cartridge (Molecular Devices, San Jose, CA), and data analysis was performed with SoftMax^® Pro GxP (version 6.5.1 and 7.0.3, Molecular Devices), following the OECD guideline 442D (2018); luciferase expression represents a measure of keratinocyte activation. Cell viability was evaluated concurrently using the methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay (Sigma-Aldrich). Briefly, MTT (200 μl of a 0.6 mg/mL solution) was added to treated cells and incubated for 4 hr at 37°C. MTT solution was removed at the end of the incubation period and 50 μl of isopropanol was used to solubilize formazan while shaking on an orbital plate shaker for 30 minutes. Absorbance at 600 nm was measured with a spectrophotometer (SpectraMax^® i3 or i3x) and data analysis performed with SoftMax^® Pro GxP (version 6.5.1 and 7.0.3). Luciferase activity and cell viability for each test chemical and the positive control were analyzed using the prediction model outlined in the OECD testing guideline (OECD 442D, 2018). The OECD acceptance criteria were applied to verify the validity of the assay, and to classify each test chemical as a non-sensitizer or sensitizer (OECD 442D, 2018). A valid assay was one in which the positive control (cinnamic aldehyde) induced ≥1.5 fold (relative to the vehicle) increase in luciferase activity, with an average induction at 64 μM between 2 and 8, EC_1.5 value (concentration yielding a 1.5 fold induction of luciferase) between 7 and 30 μM, and a clear dose response. A chemical was classified as positive for keratinocyte activation if the maximal fold induction (I_max) was ≥1.5 fold and statistically significant relative to the vehicle control, with cell viability >70% and a clear dose response. Statistical significance of the fold induction of luciferase activity for each concentration and repetition of test chemical versus vehicle control was determined by performing a two-tailed Student’s T-test assuming homoscedastic variance (two-sample equal variance). Induction was considered significant at p<0.05. According to the OECD guideline, the EC_1.5 value (concentration for which luciferase induction is 1.5 fold) must be <1000 μM. If the EC_1.5 is >1000 μM the assay is considered negative and repeated. The overall prediction for the chemical is derived from the concurrence of two independent assays (OECD 442D, 2018).

Human Cell Line Activation Test (h-CLAT)

The in vitro h-CLAT assay was conducted in accordance with the OECD guideline (OECD 442E, 2018). The h-CLAT as an individual assay is reported to have sensitivity and accuracy of 76–93%, and specificity of 66–68% relative to human data and the LLNA (OECD 442E, 2018; Roberts DW & Patlewicz G, 2018). Human acute monocytic leukemia (THP-1) cells (American Type Culture Collection, ATCC, Manassas, VA), a surrogate for dendritic cells, were treated with each test chemical at a range of concentrations. Dendritic cell (DC) activation was determined by measuring expression of CD86 and CD54 on the cell surface by flow cytometry. Briefly, a dose finding assay was conducted in which THP-1 cells that had passed the specified reactivity check were exposed to the test article at a range of doses to measure cytotoxicity and determine the concentration of test article resulting in 75% cell viability (CV75), as described in the OECD guideline (OECD 442E, 2018). In the DC activation assay, 1×10⁶ THP-1 cells were cultured in RPMI-1640 medium with GlutaMax (Gibco, Gaithersburg, MD) and with 10% Fetal Bovine Serum at 37ºC/5% CO_2. The cells were exposed to the test article at concentrations ranging from (0.335 × CV75) to (1.2 × CV75) in PBS for 24 hr. The positive control was 4.0 μg/ml 2,4-dinitrochlorobenzene (DNCB) in DMSO (Sigma-Aldrich). At 24 hr, the cells were washed in PBS with 0.1% bovine serum albumin (BSA), blocked with PBS with 0.1% BSA and 0.01% (w/v) globulin (Cohn fraction II, Human (MP Biomedicals, Irvine, CA)), and split into three aliquots. Each aliquot was stained with 3 μg/ml of either FITC-labelled anti-CD86 (BD-Biosciences, San Jose, CA), or anti-CD54 (Agilent DAKO, Santa Clara, CA), or anti-mouse IgG1 (isotype control, Agilent DAKO) antibodies, as described in the OECD guideline (OECD 442E, 2018). The cells were subsequently washed, resuspended in PBS with 0.1% BSA, and stained with 0.0625 μg propidium iodide (BD Biosciences) to measure cell viability. CD86 and CD54 expression, and cell viability were analyzed using flow cytometry (BD Accuri^™ C6, BD Biosciences) and data analysis performed with CFlow Plus (version 1.0.264.21, BD Biosciences); the data were analyzed using the calculations, the acceptance criteria, and the prediction model outlined in the OECD testing guideline to classify each test chemical as a sensitizer or non-sensitizer (OECD 442E, 2018). A minimum of two independent runs with a minimum cell viability of 50% was required for analysis of each cell surface marker in the prediction model. Each run was evaluated independently, and a single prediction was derived based on concurrent results in two runs for at least one marker. A valid assay was one in which the positive control (DNCB) increased the relative fluorescence intensity (RFI) ≥150% for CD86 and ≥200% for CD54 with viability >50%. A chemical was considered positive if the RFI value was ≥150% in at least one concentration for CD86 or ≥200% in one concentration for CD54. The RFI value represents the percentage increase in cell population mean fluorescence intensity (MFI) in the test chemical treated cells relative to the MFI of the vehicle control treated cells after the MFI for the isotype control labeled cells is subtracted from each.

Animals

BALB/c specific pathogen-free mice (female, 6–8 weeks of age, 17–20 grams) were obtained from Charles River Laboratories (Raleigh, NC). The animals were maintained on a 12-h light/dark cycle at 18–26°C in an AAALAC-accredited facility and received NTP-2000 diet (Ziegler Brothers, Inc, Gardner, PA) and tap water ad libitum. At 7–9 weeks of age, the mice were randomized to treatment groups by an electronic program, identified by tattoo, and placed on the assigned treatment (N=8/group for the in vivo assays). All in vivo studies were conducted at Virginia Commonwealth University under an approved institutional Animal Care and Use Committee protocol.

In vivo Dose Selection

Each test article was evaluated to determine solubility in organic solvents recommended by the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM) and the OECD guideline (OECD 429, 2010; ICCVAM, 2011), and the most effective solvent was selected for each compound. BMIM and EMIM were prepared in acetone:olive oil (AOO, 4:1). BMPY and NBuPY solutions were prepared daily in dimethylformamide (DMF), as they were not fully soluble in AOO (Table 1). The exposure levels for each chemical were selected based on published reports of acute toxicity and irritancy following dermal application of BMIM in BALB/c mice (Landry T et al., 2005), toxicokinetic evaluation of BMIM in F-344 rats and B6C3F1 mice (Sipes I et al., 2007), and similarity in dermal absorption among the four test articles (NTP pilot study, data not published). Due to the 2-carbon alkyl chain, EMIM was not expected to induce significant adverse effects, therefore, the highest dose was based on solubility. Landry et al. demonstrated that acute toxicity was induced following dermal exposure to 50% BMIM; in the current study the highest dose was reduced to 25% BMIM. In the absence of published literature evaluating dermal toxicity, the doses selected for BMPY and NBuPY were based on similarity to BMIM, but were reduced relative to BMIM to limit potential toxicity. The specific doses are shown in Table 1. The doses selected for the mouse ear swelling test (MEST) were based on the results of the IRR and LLNA assays. In preliminary studies, treatment on the shaved back resulted in more acute toxicity than exposure via ear painting. Therefore, the highest dose used in the MEST was reduced by 50% relative to the LLNA for BMIM, BMPY, and NBuPY. The positive controls for this study were 0.15% 2,4-dinitroflourobenzene (DNFB, Sigma-Aldrich, and Wako Chemical USA, Inc., Richmond, VA) and 25% hexyl cinnamic aldehyde (HCA, Sigma-Aldrich) which were evaluated in both AOO and DMF to provide appropriate comparisons for the test articles.

Table 1.

Dosing Strategy

Chemical	Solvent	IRR/LLNA	MEST
			Sensitization	Challenge
BMIM	AOO	0, 3.12, 6.25, 12.5, 25%	3.12, 6.25, 12.5%	6.25%
EMIM	AOO	0, 6.25, 12.5, 25, 50%	25, 50%	50%
BMPY	DMF	0, 0.825, 1.65, 3.12, 6.25, 12.5%	1.56, 3.12, 6.25%	3.12, 6.25%
NBuPY	DMF	0, 1.65, 3.12, 6.25, 12.5%	3.12, 6.25, 12.5%	12.5%
DNFB	AOO/DMF*	0.15%	0.15%	0.15%
HCA	AOO/DMF*	25%	N/A	N/A

BMIM = 1-Butyl-3-methylimidazolium chloride, EMIM = 1-Ethyl-3-methylimidazolium chloride, BMPY = 1-Butyl-1-methylpyrrolidinium chloride, NBuPY = N-Butylpyridinium chloride, DNFB = 2,4-Dinitroflourobenzene, HCA = Hexyl cinnamic aldehyde LLNA = Local Lymph Node Assay, IRR = Primary Irritancy Assay, MEST = Mouse Ear Swelling Test AOO = Acetone:Olive Oil 4:1, DMF = dimethylformamide

^*The vehicle for DNFB and HCA was AOO for the EMIM and BMIM studies, and DMF for the BMPY and NBuPY studies.

Primary Irritancy and Local Lymph Node Assays

The sensitization phase of contact hypersensitivity and the potential for ionic liquids to induce non-specific irritation were evaluated using a modification of the combined primary irritancy (IRR) and local lymph node (LLNA) assay to reduce the number of animals required (Woolhiser M et al., 1998; Anderson P et al., 2003; Auttachoat W et al., 2011). Ear thickness was measured prior to dosing (pre-measurement) and reported as mm X 10⁻². Twenty-five μl of the appropriate dilution of the test article (Table 1), vehicle (DMF or AOO), or the positive controls DNFB (0.15%) or HCA (25%) was applied to the dorsum of each ear for three days (days 1–3). On day 4, 24 hr after the last treatment, measurements of ear thickness were made (post-measurement) to assess treatment-induced irritancy (Figure 2). The percent ear swelling was calculated as follows: [(post-treatment measure / pre-treatment measure)*100]-100. For the LLNA, the same mice were injected intravenously with 20 μCi ³H-thymidine (Perkin Elmer, Inc., Waltham, MA) in 250 μl phosphate buffered saline (PBS, pH 7.4) on day 6. Five hr later the mice were euthanized via carbon dioxide asphyxiation and the cervical (auricular) lymph nodes were excised, single cell suspensions were prepared in PBS, and the cells were incubated overnight in 5% trichloroacetic acid (w/v in distilled H₂O). DNA synthesis was assessed via incorporation of [³H]-thymidine into proliferating cells. Samples were counted using the PerkinElmer 1450 Counter (PerkinElmer Instruments, Shelton, CT).

Figure 2.

Study designs for the combined Local Lymph Node Assay and Irritancy Assay, and Mouse Ear Swelling Test

Mouse Ear Swelling Test (MEST)

The MEST, which incorporates both the sensitization and challenge aspects of the response to contact allergens, was used to further identify and characterize whether the four ILs evaluated here were sensitizers at non-irritating doses. The dorsal lumbar surface of each animal was shaved, and 50 μl of test article (Table 1), 0.15% DNFB, or vehicle (DMF or AOO) was applied to the site daily for three days (days 1–3, sensitization phase), as described by Gad et al. (Gad SC et al., 1986) (Figure 2). The mice were rested on days 4–7. On day 8, ear thickness was measured, and mice were challenged on the right ear with 25 μl of test article (Table 1), 0.15% DNFB or vehicle (DMF or AOO) (challenge phase). Ear thickness was measured 24- and 48- hr post-challenge, and the percent swelling was calculated as described above. Four controls were included to ensure the validity of the assay: vehicle irritancy control (VHIC, no sensitization, challenge only with test article), vehicle only control (VH), DNFB positive control, and DNFB challenge only control (PCCO). The VHIC was analyzed relative to the VH control to test for irritancy of the test article. Ionic liquid treated groups were analyzed relative to the VHIC to eliminate irritancy as a confounding factor. The DNFB positive control was analyzed relative to the PCCO.

Individual Assay Analysis and Defined Approaches for Determination of Skin Sensitization

All calculations and graphing for the individual in vitro/in chemico assays were performed in Microsoft Excel 2016 (Redmond, WA). All analyses were conducted according to the OECD models outlined in the respective guideline documents. The accuracy, sensitivity, specificity, and positive and negative inapplicability domains for the individual in vitro assays are not sufficiently robust for use as a single predictor of sensitization therefore published defined approaches (Bauch C et al., 2012; Nukada Y et al., 2013; Takenouchi O et al., 2015; Urbisch D et al., 2016) were applied to integrate in vitro and in chemico data to determine sensitization potential. The “2 out of 3” model (Bauch C et al., 2012; OECD 256, 2016; Urbisch D et al., 2016) and the “KE 3/1 STS” model (Nukada Y et al., 2013; Takenouchi O et al., 2015) have both been reviewed by the OECD (OECD 256, 2016; Kleinstreuer NC et al., 2018), and have been accepted for regulatory use by the EPA (EPA, 2018). In the “2 out of 3” model, non-animal methods, representing key events 1–3 in the skin sensitization AOP, are evaluated in an undefined order. In the current study, the DPRA, KeratinoSens^™ assay, and h-CLAT assay were conducted on all samples. The final prediction is based on concordant results from any two methods (Bauch C et al., 2012; OECD 256, 2016; Urbisch D et al., 2016) (Bauch C et al., 2012; OECD 256 A, 2016; Urbisch D et al., 2016). The “KE 3/1 STS” DA is a decision tree that evaluates the h-CLAT and DPRA assays in a strict order (Nukada Y et al., 2013; Takenouchi O et al., 2015; EPA, 2018; Roberts DW & Patlewicz G, 2018). If one or both assays identify the test article as a sensitizer, the final prediction is sensitizer. A final prediction of non-sensitizer is assigned only if both assays identify the test article as a non-sensitizer.

All in vivo results are presented as mean ± SE for 7–8 animals per group. All statistical analyses were conducted using SAS software (SAS Institute, Cary, NC). Jonckheere’s Test (Jonckheere A, 1954) was used to test for dose-related trends. Body weights (BW) and BW gains, which typically exhibit a normal distribution, were analyzed using a parametric multiple comparison procedure. If a significant trend was observed in the Jonckheere’s test, Williams’ test (Williams D, 1971; Williams D, 1972) was used to determine significant difference from control; if the trend was not significant Dunnett’s test (Dunnett W, 1955) was used. Positive control and vehicle control data were compared by Student’s T-test. IRR, LLNA and MEST data were assumed to be non-normally distributed and were analyzed using a non-parametric multiple comparison procedure. If a significant trend was observed Shirley’s test (Shirley E, 1977) was used to determine significant difference from control; if the trend was not significant Dunn’s test (Dunn O, 1964) was used. Positive control and vehicle control data were compared by a Wilcoxon Rank Sum Test (Wilcoxon F, 1945). Data that were different from control at p≤0.05 were considered statistically significant. For the LLNA assay, the Stimulation Index (SI) was calculated as the ratio of the mean proliferation in each treated group to that in the concurrent vehicle control group. The SI3 threshold for sensitization is reached when the SI value is ≥3.

Results

Supplemental materials that support the findings of this study are available in Chemical Effects of Biological Systems (CEBS) at https://doi.org/10.22427/NTP-DATA-002-03253-0003-0000-8.

IN VITRO/IN CHEMICO STUDY

DPRA

Protein reactivity of BMIM, EMIM, BMPY, and NBuPY was evaluated by the chemical capability to bind to and deplete cysteine- or lysine-containing peptides. BMIM resulted in 6.9% depletion of cysteine (Table 2, (CEBS, In Vitro Supplemental Data)), relative to the control, but EMIM, BMPY, and NBuPY each resulted in 0% depletion of cysteine. In addition, EMIM did not deplete the lysine peak (0.6% depletion). However, BMIM, BMPY, and NBuPY caused a shift in retention time of a portion of the lysine peptide and resulted in HPLC peak splitting that increased with incubation time (triplicate samples were analyzed approximately 10 hours apart and the last replicate had the largest proportion of lysine peptide with shifted retention time). Separate integration of the peptide peaks was not possible. The primary prediction model (cysteine 1:10/lysine 1:50) requires both cysteine and lysine data to determine sensitization potential. The total lysine peptide peak area was integrated to determine lysine percent depletion. BMIM, BMPY, and NBuPY resulted in 0.8%, 0.2%, and 0% depletion of lysine, respectively (Table 2). Based on the primary prediction model, the mean cysteine+lysine % depletion values were EMIM 0.3%, BMIM 3.9%, BMPY 0.1%, and NBuPY 0% (Table 2). Per the OECD guideline, if this factor is <6.4%, the prediction is negative; between 6.4% and 100%, the prediction is positive. Therefore, all four chemicals were classified as non-sensitizers in the DPRA (Table 2). The positive control, cinnamic aldehyde, reacted appropriately with both peptides, resulting in 52.7–54.0% depletion of lysine and 72.3–72.8% depletion of cysteine, and indicating that the assay performed as expected (Table 2).

Table 2.

Direct Peptide Reactivity Assay Summary

Chemical	Mean Cysteine % Depletion	Mean Lysine % Depletion	Mean Cys+Lys % Depletion	Reactivity Class	Prediction
BMIM*	6.9%	0.8%	3.9%	Minimal Reactivity	Non-sensitizer
EMIM	0.0%	0.6%	0.3%	Minimal Reactivity	Non-sensitizer
BMPY*	0.0%	0.2%	0.1%	Minimal Reactivity	Non-sensitizer
NBuPY*	0.0%	0.0%	0.0%	Minimal Reactivity	Non-sensitizer
Cinnamic Aldehyde	72.3–72.8%	52.7–54%			Sensitizer

^*Lysine peptide peak splitting was observed, and peaks could not be integrated separately.

Total peptide peak area was integrated to determine lysine percent depletion.

BMIM = 1-Butyl-3-methylimidazolium chloride, EMIM = 1-Ethyl-3-methylimidazolium chloride, BMPY = 1-Butyl-1-methylpyrrolidinium chloride, NBuPY = N-Butylpyridinium chloride, Cys = cysteine, Lys = Lysine

KeratinoSens^™ Assay

KeratinoSens^™ cells were cultured with the test chemicals to evaluate the potential for the chemical to induce activation of the Keap1-Nrf2-ARE-dependent pathway. The ILs tested in this study did not induce cytotoxicity, and testing was possible at all concentrations recommended in the guideline (CEBS, In Vitro Supplemental Data). At concentrations that resulted in 70% or higher cell viability, exposure to BMPY and NBuPY induced a 1.6–2.1 fold increase in expression of the luciferase reporter gene at 500 and 1000 μM in two assays, indicating induction of the pathway and keratinocyte activation (Table 3, (CEBS, In Vitro Supplemental Data)). The prediction model outlined by the OECD (OECD 442D, 2018) classified BMPY and NBuPY as sensitizers (Table 3). BMIM induced luciferase expression at 2000 μM; however, the prediction model requires induction at ≤1000 μM for classification as positive for sensitization potential; whereas EMIM failed to induce the luciferase reporter gene and activate keratinocytes at any concentration. Therefore, both BMIM and EMIM were classified as non-sensitizers. The positive control, cinnamic aldehyde, induced a 3.3–10.6 fold increase in expression of the luciferase reporter gene at 64 μM and activated keratinocytes, indicating the assay performed as expected.

Table 3.

KeratinoSens^™ Assay Summary

	Run 1					Run 2
Chemical	Viability ≥70%	Prediction	EC1.5 (μM)	Imax	IC50 (μM)	Viability ≥70%	Prediction	EC1.5 (μM)	Imax	IC50 (μM)
BMIM	Yes	Non-sensitizer	1629.04	1.64	No IC50	Yes	Non-sensitizer	1012.23	3.93	No IC50
EMIM	Yes	Non-sensitizer	1542.47	1.82	No IC50	Yes	Non-sensitizer	NI	1.30	No IC50
BMPY	Yes	Sensitizer	412.55	2.31	No IC50	Yes	Sensitizer	342.00	4.73	No IC50
NBuPY	Yes	Sensitizer	181.61	4.00	No IC50	Yes	Sensitizer	451.83	2.34	No IC50
Cinnamic Aldehyde	Yes	Sensitizer	8.18	10.57		Yes	Sensitizer	10.52	5.18

EC_1.5 = Concentration yielding a 1.5 fold induction of luciferase, IC₅₀ = Concentration yielding a 50% reduction in cell viability.

I_max = Maximum induction factor of luciferase activity

EC_1.5 >1000 μM is classified negative per European Union Reference Laboratory European Centre for the Validation of Alternative Methods Database on Alternative Methods to Animal Experimentation Protocol No. 155.

NI = No Induction of luciferase expression above the 1.5 threshold.

BMIM = 1-Butyl-3-methylimidazolium chloride, EMIM = 1-Ethyl-3-methylimidazolium chloride, BMPY = 1-Butyl-1-methylpyrrolidinium chloride, NBuPY = N-Butylpyridinium chloride

^*Cinnamic Aldehyde was evaluated in three assays, with 2 runs each time. A representative evaluation is shown. The EC_1.5 values ranged from 11.28 to 17.26, and the I_max values ranged from 3.25 to 6.45 at 64 μM, in the other studies.

h-CLAT

Human THP-1 cells, a surrogate for DC, were exposed to the test chemicals; labelled with FITC-anti-CD86, FITC-anti-CD54, or FITC-anti-mouse IgG1; and evaluated via flow cytometry for expression of cell surface markers of DC activation. Treatment of THP-1 cells with 4167 and 5000 μg/ml BMIM and 5000 μg/ml NBuPY resulted in viability below 50%; those samples were excluded from analysis in the prediction model. At concentrations that resulted in 50% or higher cell viability (≤ 5000 μg/ml for BMPY, ≤ 4167 μg/ml for BMIM and NBuPY), BMIM, BMPY, and NBuPY upregulated CD54 in THP-1 cells, but failed to modulate CD86 (Table 4, (CEBS, In Vitro Supplemental Data)). The CD54 RFI values (defined in methods) for BMIM, BMPY, and NBuPY exceeded 200 in two runs each, leading to EC200 values (the concentration at which the RFI reached 200) of 962 μg/mL for BMIM, 4110–4375 μg/ml for BMPY, and 1455–2194 μg/ml for NBuPY (Table 4). As such, the OECD prediction model (OECD 442E, 2018) classified BMIM, BMPY and NBuPY as sensitizers, based on concordant results in 2 runs (Table 4). The RFI for CD86 did not reach 150 in any of the ILs. EMIM did not induce either CD54 or CD86 in the assay, leading to classification as a non-sensitizer. The positive control, 4.0 μg/ml DNCB, induced both CD54 (RFI 444–577) and CD86 (RFI 532–694) indicating that the assay performed as expected (CEBS, In Vitro Supplemental Data).

Table 4.

Human Cell Line Activation Test Summary

	Run 1				Run 2
Chemical	Viability ≥50%	Prediction	EC150 (μg/ml)	EC200 (μg/ml)	Viability ≥50%	Prediction	EC150 (μg/ml)	EC200 (μg/ml)
BMIM	Yes	Sensitizer	(>5000)	962	Yes	Sensitizer	(>5000)	*
EMIM	Yes	Sensitizer	(>5000)	4781	Yes	Non-sensitizer	(>5000)	(>5000)
Run 3	Yes	Non-sensitizer	(>5000)	(>5000)
BMPY	Yes	Sensitizer	(>5000)	4375	Yes	Sensitizer	(>5000)	4110
NBuPY	Yes	Sensitizer	(>5000)	2194	Yes	Sensitizer	(>5000)	1455
DNCB**	Yes	Sensitizer	4.0	4.0	Yes	Sensitizer	4.0	4.0

NI = No induction. EC150 = Effective Concentration, the concentration yielding a relative fluorescent intensity (RFI) of 150 for CD86, EC200 = The concentration yielding a relative fluorescent intensity (RFI) of 200 for CD54,

^*Did not meet guideline criteria to calculate the EC200. The RFI value at the lowest dose was above the positive criteria and no higher doses (up to the fourth lowest dose) resulted in an RFI value ≥10% of the RFI value at the lowest dose.

EMIM = 1-Ethyl-3-methylimidazolium chloride, Due to inconsistent results in the first two runs of the assay, a third run was conducted.

BMIM = 1-Butyl-3-methylimidazolium chloride, BMPY = 1-Butyl-1-methylpyrrolidinium chloride, NBuPY = N-Butylpyridinium chloride

^**The positive control, DNCB, was run at one concentration only, 4.0 (μg/ml). This concentration is higher than the EC150 and EC200 for DNCB, however, the actual EC150 and EC200 cannot be calculated from a single data point.

Defined Approach Predictions

The “2 out of 3” DA model classified BMPY and NBuPY as sensitizers based on a positive outcome in the KeratinoSens^™ and h-CLAT assays; BMIM and EMIM were classified as non-sensitizers. The “KE 3/1 STS” DA model classified BMIM, BMPY and NBuPY as sensitizers based on the outcome of the DPRA and h-CLAT assays; EMIM was classified as a non-sensitizer.

IN VIVO STUDY

BMIM

No biologically relevant differences were observed in BW or BW gain at 3.12–12.5% BMIM (CEBS, In Vivo Supplemental Data). There was one occurrence of increased BW at 3.12% BMIM and one occurrence of increased BW gain at 12.5% BMIM in a single assay, but these increases were not observed at higher doses and were not repeatable. All mice treated with 25% BMIM on the ear (IRR/LLNA assay) exhibited ruffled fur at the site of treatment; 4 out of 16 total mice (two studies with N=8 in each study) became moribund, or died between 24 and 72 hr post-treatment, and were removed from that study. In addition, all animals treated with 12.5% BMIM on the shaved back (MEST) in a preliminary study became moribund and were removed from that study. All moribund animals were humanely euthanized. The larger surface area and better absorption on the back may account for the acute response in the MEST at 12.5% BMIM, compared to the IRR/LLNA in which no adverse effects were observed at that dose. Due to the adverse clinical observations, and in accordance with NTP Laboratory Animal Management guidelines (NTP, 2011), all mice that received 25% BMIM on the ear or 12.5% BMIM on the back were excluded from statistical evaluation. All data below are based on doses that did not lead to acute toxicity.

BMIM induced statistically significant pairwise increases in lymph node cell proliferation at 12.5% BMIM, when compared to vehicle control (Figure 3, (CEBS, In Vivo Supplemental Data)) in both of the LLNA assays. The assay was run twice to confirm the observed increase in lymph node cell proliferation. There was a small, statistically significant increase observed at 3.12% BMIM in one assay only; there was no significant effect on lymph node cell proliferation at 6.25% BMIM. The SI3 threshold (SI≥3), classically associated with sensitizers, was not achieved at any exposure level (3.12–12.5%). BMIM did not induce ear swelling in the IRR/LLNA assay at any treatment level (Figure 4, (CEBS, In Vivo Supplemental Data)). The 0.15% DNFB control was positive for both sensitization (SI>3) and irritancy in both IRR/LLNA assays, when compared to vehicle control; the 25% HCA control achieved the SI3 level for sensitization and induced ear swelling in one assay. Based on the IRR/LLNA data, the MEST assay was conducted at doses that did not induce acute toxicity. Significant ear swelling was observed in mice that were sensitized with 3.12% BMIM and challenged with 6.25% BMIM, relative to the VHIC group, but no effect was observed in mice sensitized and challenged with 6.25% BMIM (Figure 5, (CEBS, In Vivo Supplemental Data)).

Figure 3. Contact hypersensitivity in BALB/c mice following dermal exposure to ionic liquids – Local Lymph Node Assay.

(A) 1-Butyl-3-methylimidazolium chloride (BMIM), (B) 1-Ethyl-3-methylimidazolium chloride (EMIM), (C) 1-Butyl-1-methylpyrrolindium chloride (BMPY), (D) N–Butylpyridinium chloride (NBuPY). Mice were exposed to ionic liquids via skin painting on the ear days 1–3; incorporation of [³H]-thymidine into the draining lymph nodes was quantitated as a measure of DNA synthesis, as described in Methods. The solid line indicates the SI3 (stimulation index=3 threshold) value of run 1; the dashed line indicates the SI3 value of run 2.

Statistically significant trend values: BMIM ‡ run 1 p=0.006; BMPY ‡ run 1 p=0.003, ‡ run 2 p=0.002; NBuPY + run 1 p=0.025.

Values = mean±SEM, N=7–8 mice/group, * Value significantly different from vehicle control *p≤0.05, **p≤0.01

VH=vehicle, BMIM and EMIM vehicle = Acetone/Olive Oil (AOO), BMPY and NBuPY vehicle = dimethylformamide (DMF)

The positive controls for this study were 0.15% 2,4-dinitroflourobenzene (DNFB) and 25% hexyl cinnamic aldehyde (HCA).

Figure 4. Contact irritancy in BALB/c mice following dermal exposure to ionic liquids.

(A) 1-Butyl-3-methylimidazolium chloride (BMIM), (B) 1-Ethyl-3-methylimidazolium chloride (EMIM), (EMIM), (C) 1-Butyl-1-methylpyrrolindium chloride (BMPY), (D) N–Butylpyridinium chloride (NBuPY). Mice were exposed to ionic liquids via skin painting on the ear days 1–3. Ear swelling was measured on day 4 to assess treatment-induced irritancy.

Statistically significant trend values BMIM + run 1 p=0.045; NBuPY ‡ p=0.001

Values = mean±SEM, N=7–8 mice/group, * Value significantly different from vehicle control *p≤0.05, **p≤0.01

VH=vehicle, BMIM and EMIM vehicle = Acetone/Olive Oil (AOO), BMPY and NBuPY vehicle = dimethylformamide (DMF),

The positive controls for this study were 0.15% 2,4-dinitroflourobenzene (DNFB) and 25% hexyl cinnamic aldehyde (HCA).

Figure 5. Contact hypersensitivity in BALB/c mice following dermal exposure to ionic liquids – Mouse Ear Swelling Test (MEST).

(A) 1-Butyl-3-methylimidazolium chloride (BMIM), (B) 1-Ethyl-3-methylimidazolium chloride (EMIM), (C) 1-Butyl-1-methylpyrrolindium chloride (BMPY), (D) N–Butylpyridinium chloride (NBuPY). In the MEST, mice were sensitized with vehicle, 1–50% of the respective ionic liquid, or 0.15% DNFB for 3 days, and challenged with 6.25% BMIM, 50% EMIM, 3.15% BMPY, 12.5% NBuPY, vehicle, or 0.15% DNFB, as described in Methods.

Statistically significant trend values: NBuPY ‡ 24 hr p=0.006

Values = mean±SEM, N=7–8 mice/group, * Value significantly different from vehicle control *p≤0.05, **p≤0.01

VH=vehicle, BMIM and EMIM vehicle=Acetone/Olive Oil (AOO), BMPY and NBuPY vehicle=dimethylformamide (DMF)

The positive control for this study was 0.15% 2,4-dinitroflourobenzene (DNFB).

The vehicle irritancy controls (VHICs) were sensitized with VH and challenged with 6.25% BMIM, 50% EMIM, 3.12% BMPY, or 12.5% NBuPY.

EMIM

No significant differences were observed, subsequent to treatment, in BW or BW gain in mice treated with 6.25–50% EMIM (CEBS, In Vivo Supplemental Data). Topical application of 6.25–50% EMIM did not affect lymph node cell proliferation, or induce ear swelling in the IRR/LLNA assay (Figures 3–4, (CEBS, In Vivo Supplemental Data)). Similarly, in the MEST, there were no significant changes in ear swelling in mice that were sensitized with 25% and 50% EMIM and challenged with 50% EMIM, relative to the VHIC group, at either of the time points examined (Figure 5, (CEBS, In Vivo Supplemental Data)). The positive controls, 0.15% DNFB and 25% HCA, produced the expected sensitization and irritancy responses in all assays.

BMPY

All mice that were treated topically on the ears with 12.5% BMPY (LLNA), and approximately 50% of the mice treated with 6.25% BMPY on the shaved back (MEST), became moribund within 72 hr of exposure. Due to the adverse clinical observations, and in accordance with NTP Laboratory Animal Management guidelines (NTP, 2011), these mice were humanely euthanized and all mice treated with 12.5% BMPY in the LLNA or with 6.25% BMPY in the MEST were excluded from the study. No significant differences were observed in BW or BW gain in mice treated with BMPY at doses that did not induce acute toxicity (CEBS, In Vivo Supplemental Data); all data below are based on doses that did not lead to acute toxicity.

BMPY induced a statistically significant pairwise increase in lymph node cell proliferation at 6.25% BMPY, when compared to vehicle control. The assay was run twice to confirm the observed increase in lymph node cell proliferation. However, the SI3 threshold was not achieved (Figure 3, (CEBS, In Vivo Supplemental Data)). BMPY was not an irritant at doses ranging from 0.825–6.25% when compared to the vehicle control (Figure 4, (CEBS, In Vivo Supplemental Data)). In the MEST, there were no significant changes in ear swelling in mice that were sensitized with 1.65–3.12% BMPY and challenged with 3.12% BMPY, relative to the VHIC group at either of the time points examined (Figure 5, (CEBS, In Vivo Supplemental Data)). The positive controls, 0.15% DNFB and 25% HCA, produced the expected sensitization and irritancy responses in all assays.

NBuPY

There were no differences in BW or BW gain between animals treated with NBuPY or the vehicle control (CEBS, In Vivo Supplemental Data). Topical application of 1.65–12.5% NBuPY had no effect on lymph node cell proliferation when compared to the vehicle control in either assay (Figure 3, (CEBS, In Vivo Supplemental Data)). Although exposure to NBuPY did not induce a sensitization response, an increasing trend (p<0.01) in percent ear swelling (irritancy) was observed 24 hr post-exposure compared to the vehicle control, with statistically significant increases at ≥3.12% NBuPY, indicating that NBuPY was an irritant (Figure 4, (CEBS, In Vivo Supplemental Data)). The DNFB positive control produced the expected sensitization and irritation responses. The HCA positive control induced statistically significant increases in lymph node cell proliferation in both assays, but the response reached the SI3 level only in the second assay; HCA was an irritant. The second IRR assay was disqualified because the vehicle response was outside of the historical control range. Interestingly there was a statistically significant decreasing trend observed in percent ear swelling in the MEST at 24 hr post-challenge in mice sensitized with 6.25–12.5% NBuPY and challenged with 12.5% NuBPY (Figure 5, (CEBS, In Vivo Supplemental Data)). There were no significant differences observed at 48 hr post-challenge. The pre-treatment measurement for ear thickness was significantly higher in NuBPY treated mice, compared to the VHIC group which may have contributed to the discrepancy in ear swelling post-treatment. In addition, all doses used for sensitization and challenge in the MEST were positive in the irritancy assay. The VHIC control was increased compared to the VH control, reaching statistical significance at 24 hr, further indicating irritancy by NuBPY. The irritant effects confounded the interpretation of the sensitization data for NuBPY.

Discussion

Allergic and irritant dermatitis account for approximately 36% of occupational diseases (Martin S et al., 2018). Human exposure to ILs occurs primarily through skin contact, and research suggests that some ionic liquids have the potential to cause skin irritation and dermal toxicity, particularly in keratinocytes (Landry T et al., 2005; Cheng Y et al., 2009; Hwang J-H et al., 2018). A recent clinical study demonstrated that an antimicrobial, IL-based hand sanitizer effective against Staphylococcus aureus and human coronavirus remained in contact with the skin considerably longer than more conventional products, increasing protection, but also increasing dermal contact. Although the specific ILs used in the hand sanitizer are proprietary, eight of 52 people in the study developed skin irritation (Shevachman M et al., 2020), indicating that ILs as a class may induce dermal toxicity in sensitive individuals. Dermal exposure to NBuPY induced a statistically significant, dose-dependent irritancy response in BALB/c mice in the current study, suggesting that NBuPY might induce irritant dermatitis in humans. Repeated exposure to irritant chemicals, e.g. in an occupational setting by workers without proper protective equipment, can lead to skin inflammation and epidermal damage. Exposure to irritant compounds in combination with skin allergens can enhance percutaneous absorption, and amplify DC activation and the sensitization response (Sipes I et al., 2007; Martin S et al., 2018).

Although dermal irritation and sensitization share some common symptoms, the physiologies are different, and irritation is not required for sensitization to occur. In the Skin Sensitization Adverse Outcome Pathway (AOP) (OECD 168, 2012; Strickland J et al., 2016; Kleinstreuer NC et al., 2018), the molecular initiating event is the covalent binding of an electrophilic molecule to nucleophilic amino acids in skin proteins, measured by the depletion of lysine and cysteine peptides (Lalko JF et al., 2012; OECD 168, 2012; OECD 442C, 2015). In the current study, EMIM had no effect on either cysteine or lysine peptides and was classified as a non-sensitizer in the DPRA prediction model. BMIM, BMPY, and NBuPY, which were also classified by the prediction model as non-sensitizers, did not deplete the cysteine peptide, but did alter the retention time of a portion of the lysine peptide resulting in a split peak. Complications, such as non-covalent interactions, weak binding, or similarity in column retention times between the test article and the peptide in the HPLC/UV-based DPRA can lead to interference with the peptide peak, and false prediction/classification of test chemicals (Zhang F et al., 2018). The split peak observed with all of the butylated ILs in this study suggests a weak interaction with the lysine peptide, as the proportion of peptide impacted increased with incubation time.

Induction of the cytoprotective Keap1-Nrf2-ARE-dependent pathway by activated keratinocytes upregulates genes involved in skin sensitization and serves as the second key event in the skin sensitization AOP (Natsch A, 2010; OECD 168, 2012; OECD 442D, 2018). In the current study, BMPY and NBuPY upregulated the Keap1-Nrf2-ARE pathway, and were classified as sensitizers, while EMIM and BMIM were classified as non-sensitizers. The Keap1-Nrf2-ARE pathway contains reactive cysteine residues. Compounds that do not react with cysteine may lead to false negative classification in the KeratinoSens^™ assay (Natsch A, 2010; OECD 442D, 2018). Natsch (2010) reviewed the mechanism of the Keap1-Nrf2-ARE pathway extensively, focusing on the role of cysteine and cytokine induction in activation of the pathway. A small number of known sensitizers were identified that preferentially activate lysine, rather than cysteine, and appear not to activate the Keap1-Nrf2-ARE pathway. The author hypothesized that these sensitizers may act through a novel toxicity pathway to induce sensitization (Natsch A, 2010). Under certain conditions, for example partial cytotoxicity or dermal irritant properties, a compound may induce false positive results in the KeratinoSens^™ assay, possibly due to the generation of endogenous Keap-1-ligands (Emter R et al., 2010). The positive result observed for NBuPY may have been confounded due to the irritancy of the compound or activation of a novel toxicity pathway (Emter R et al., 2010; Natsch A, 2010).

Activation of monocytes and DC, evaluated through expression of CD54 and CD86 cell surface markers in the h-CLAT assay, is the third key event in the skin sensitization AOP (OECD 442E, 2018; Roberts DW & Patlewicz G, 2018). In the current study, BMIM, BMPY, and NBuPY upregulated CD54 and were classified as sensitizers. Published reports have indicated that some dermal irritants may upregulate CD54 and CD86 through an alternate pathway, non-specific activation, danger signals, or activation of interleukin-1β and tumor necrosis factor-α, leading to a false positive result (Aiba S et al., 1997; Emter R et al., 2010; Mitachi T et al., 2018; OECD 442E, 2018). It is possible that NBuPY, an irritant, may have induced CD54 expression through one of these alternate mechanisms. Consistent with the other in vitro/in chemico assays, EMIM failed to modulate CD54 or CD86, and was classified as a non-sensitizer.

In the current study, the in vitro/in chemico data were evaluated using two defined approach (DA) models. The “2 out of 3” model (Bauch C et al., 2012; Urbisch D et al., 2016; OECD, 2020), which is reported to be 70–77% accurate compared to human data and the LLNA, identified BMPY and NBuPY as sensitizers; EMIM and BMIM were identified as non-sensitizers (Table 5). The “KE 3/1 STS” (Key Events 3 and 1 sequential testing strategy) DA model (Nukada Y et al., 2013; Takenouchi O et al., 2015; OECD, 2018), reported to be 78–80% accurate compared to human data and the LLNA, classified BMIM, BMPY, and NBuPY as sensitizers; EMIM was identified as a non-sensitizer (Table 5). The sole discrepancy between the “2 out of 3” DA and the “KE 3/1 STS” DA in the current study was the prediction for BMIM. The KeratinoSens^™ assay categorized BMIM as a non-sensitizer, resulting in two assays indicating that BMIM is a non-sensitizer; the KeratinoSens^™ assay is not a component of the “KE 3/1 STS” DA. The limitations of the individual assay prediction models, the applicability domains, and the confounding factors described above, carry over to the DA analysis and can lead to false positive/false negative predictions (Roberts DW & Patlewicz G, 2018; Kolle SN et al., 2019). It is possible that by not capturing the interaction of test chemicals with the lysine peak in the DPRA, noted above, the predictivity of the DA may have been altered in this study.

Table 5a.

Analysis of In vitro and In chemico Assays

Evaluation	EMIM	BMIM	BMPY	NBuPY
Individual Assay Predictions
DPRA	Non-sensitizer	Non-sensitizer	Non-sensitizer	Non-sensitizer
KS	Non-sensitizer	Non-sensitizer	Sensitizer	Sensitizer
h-CLAT	Non-sensitizer	Sensitizer	Sensitizer	Sensitizer
“2 out of 3” DA (Bauch C et al., 2012)	Non-sensitizer	Non-sensitizer	Sensitizer	Sensitizer
KE 3/1 DA (Takenouchi et al. 2015)	Non-sensitizer	Sensitizer	Sensitizer	Sensitizer
Binary Combination (Roberts DW & Patlewicz G, 2018)
DPRA (1) + h-CLAT (2)	Non-sensitizer	Sensitizer	Sensitizer	Sensitizer
DPRA (1) + KS (2)	Non-sensitizer	Non-sensitizer	Sensitizer	Sensitizer
KS (1) + h-CLAT (2)	Non-sensitizer	Sensitizer	Sensitizer	Sensitizer

The activation of T-lymphocytes is the fourth key event in the skin sensitization AOP; as a whole animal model, the LLNA includes the cellular and molecular events of the entire AOP. In the current study, EMIM and NBuPY were not sensitizers. However, exposure to BMIM and BMPY induced a statistically significant increase in lymph node cell proliferation at 12.5% and 6.25%, respectively, in two assays, but the SI3 level classically associated with sensitizers (OECD 429, 2010) was not achieved. Exposure to BMIM and BMPY was restricted due to limitations on tolerated dose. While it may have been possible to achieve an SI3 level using higher doses, the relevance of the information would be confounded by the overt toxicity. Under the OECD LLNA guideline, these ILs would not be sensitizers. With the development of modifications to the LLNA and in vitro analysis methods, a more integrated weight of evidence approach has been recommended, with human studies having the most weight, followed by animal studies, and with all non-animal studies weighted lowest (CPSC, 2014). Further, issues regarding potential false negative interpretation based on SI3 values alone have been identified. For example, the shape of the dose response curve, the solvent and chemical solubility, irritancy properties, structural alerts, and metals have all been shown to affect the response and interpretation (Basketter DA et al., 2007; Basketter DA et al., 2009; OECD 429, 2010; Roberts DW & Patlewicz G, 2018). Given the statistical significance in multiple assays, the dose response of the LLNA, and the supporting evidence of the KeratinoSens^™ and h-CLAT assays, BMIM and BMPY could plausibly be considered weak sensitizers (Swaen G & van Amelsvoort L, 2009; OECD 429, 2010; Fedak KM et al., 2015). When evaluated independent of the alkyl chain, aromatic cations (e.g. imidazolium and pyridinium) are reported to be more toxic than non-aromatic cations (e.g. pyrrolidinium), or linear cations (e.g. ammonium), particularly in aquatic toxicity models (Larson JH et al., 2008; Hernandez-Fernandez FJ et al., 2015; Wang C et al., 2015; Pham T et al., 2016; Gomez-Herrero E et al., 2020). However, in the current study, BMIM (aromatic) and BMPY (non-aromatic) demonstrated similar degrees of sensitization in the IRR/LLNA and acute toxicity in the MEST assays.

There was a discrepancy in the determination of sensitization for NBuPY between the in vivo (non-sensitizer) and in vitro (sensitizer) analyses, and for BMIM between the LLNA (sensitizer) and “2 out of 3” DA model (non-sensitizer). The DA predictions must be interpreted with caution due to the irritancy of NBuPY and the potential interference in the cell-based KeratinoSens^™ and h-CLAT assays, and to the chemical interference in the DPRA, described above. Kolle, et al. (2019) and Leontaridou, et al. (2019) have proposed that the strict, rule-based evaluation of the “2 out of 3” DA model can lead to false negative predictions, due to the uncertainty of the borderline range, the range above and below the classification threshold in which there is nearly equal likelihood of a positive or negative result. Weak sensitizers are particularly vulnerable to false negative classification (Roberts DW & Patlewicz G, 2018; Kolle SN et al., 2019; Leontaridou M et al., 2019). Murine LLNA and human sensitization databases serve as the reference points against which all non-animal models are evaluated. Therefore, in the overall assessment of sensitization potential, greater weight was afforded the LLNA data, and a final identification of non-sensitizer was ascribed to NBuPY, and of sensitizer to BMIM.

Ionic liquids are differentially toxic, depending on the nature of the dermal exposure (Landry T et al., 2005; Sipes I et al., 2007; Cheng Y et al., 2009). In previously conducted studies (Sipes I et al., 2007; Cheng Y et al., 2009), BMIM and NBuPY were not well absorbed by male F-344 rats following dermal exposure at a dose of 5 mg/kg (~1 mg for a 200 g rat, single exposure). Total absorption for BMIM and NBuPY in dimethylformamide (DMF)/water (55:45, v/v) was 12.8% and 35%, respectively, and was even lower in ethanol/water (3.2% and 16%, respectively). In addition, only the parent compound was detected in the urine, demonstrating that BMIM and NBuPY were not metabolized in vivo (Sipes I et al., 2007; Cheng Y et al., 2009). Landry, et al. (2005) reported that dermal application of 2% BMIM (~1 mg/day, 3 day exposure) to the ears of BALB/c mice resulted in lymph node cell proliferation (SI=2.1), although not considered significant (SI <3). At moderate concentrations, application of 10% BMIM in DMF (~5 mg/day, 3 day exposure) to BALB/c mice resulted in mortality in 1 mouse, moderate to severe erythema of the ear, and increased lymph node cell proliferation (SI=4.3), indicating dermal irritation and sensitization. In female F344 rats 200–800 mg of BMIM in DMF (~40–160 mg for a 200 g rat, single exposure) induced dermal irritation (Landry T et al., 2005). Higher doses, 2000 mg BMIM (~200 mg for a 200 g rat, single exposure), induced dermal irritation in male and female F344 rats in water, but resulted in mortality in male (40%) and female (100%) rats when dissolved in DMF. At 50% and 75%, BMIM in DMF (~25 and 37.5 mg/day, 2 day exposure) resulted in mortality in BALB/c mice (Landry T et al., 2005). Although different solvents were not evaluated in the current research, it is possible that high dose exposure in a more permissible solvent, or chronic dermal contact with BMIM and BMPY at moderate doses, such as might result from occupational exposure or use of IL-containing personal care products, could potentially result in sensitization in humans.

In summary, consistent with published in vivo and in vitro work indicating that short chain ILs demonstrate very limited dermal penetration and cell damage (Mester P et al., 2015; Losada-Pérez P et al., 2016; Mendonça CMN et al., 2018; Kusumahastuti DKA et al., 2019), EMIM (an ethylated IL) (6.25–50%) was negative for sensitization and irritation in the current study. BMIM is reported to induce limited morphological changes and membrane damage in cultured cells at low concentrations, but, at high concentrations can induce membrane disruption, cytotoxicity, apoptosis, and necrosis in vitro (Galluzzi M et al., 2018), and dermal irritation and sensitization, and mortality (Landry T et al., 2005) in vivo. Similarly, under the conditions of this study, the butylated ILs BMIM and BMPY were weak skin sensitizers at low to moderate concentrations (12.5% and 6.25% respectively), but were acutely toxic at higher concentrations (25% and 12.5%, respectively). Although NBuPY (a butylated IL) was not a sensitizer in vivo, it did induce contact skin irritation (3.12–12.5%) in a dose dependent manner. As skin irritants have been shown to induce false positive reactions in the KeratinoSens^™ and h-CLAT assays (Emter R et al., 2010; Mitachi T et al., 2018), the in vitro/in chemico evaluation of the sensitization potential of NBuPY may have been confounded by its irritancy as measured in the in vivo irritancy assay. While ILs are purported to be safer than volatile organic solvents, these data demonstrate that ILs can induce adverse health effects, specifically irritant and allergic dermatitis. ILs may pose a health hazard in the occupational setting, and health factors and worker protection (minimally avoiding dermal contact) need to be considered in the design and selection of ILs for industrial use.

Table 5b.

Analysis of Local Lymph Node Assay

Evaluation	EMIM	BMIM	BMPY	NBuPY
Significance (doses) +	Non-sensitizer	Sensitizer 12.5% *	Sensitizer 6.25% **	Non-sensitizer
Trend p-value	0.748	0.006, 0.075	0.003, 0.002	0.093, 0.025
Stimulation Index ++		1.9, 1.5	2.0, 1.9

Table 5c.

Analysis of Irritancy Assay

Evaluation	EMIM	BMIM	BMPY	NBuPY
Significance (doses) +	Non-irritant	Non-irritant	Non-irritant	Irritant 3.12–12.5% **
% Ear Swelling				4.8–10.3%
Trend p-value ++	0.519	0.045, 0.389	0.081, 0.416	0.001

BMIM = 1-Butyl-3-methylimidazolium chloride, EMIM = 1-Ethyl-3-methylimidazolium chloride, BMPY = 1-Butyl-1-methylpyrrolidinium chloride, NBuPY = N-Butylpyridinium chloride

DPRA = Direct Peptide Reactivity Assay, KS = KeratinoSens^™, h-CLAT = Human Cell Line Activation Test, KE = Key event, DA = Defined Approach,

^*p≤0.05,

^**p≤0.01

⁺Sensitizer/Non-sensitizer based on weight of evidence. Doses based on statistical significance.

⁺⁺Stimulation indices and % ear swelling were reported only for those groups that were statistically significant.

Abbreviations

AOO

Acetone and olive oil

AOP

Adverse outcome pathway

ARE

Antioxidant/electrophile response element

BMIM

1-Butyl-3-methylimidazolium chloride

BMPY

1-Butyl-1-methylpyrrolidinium chloride

BW

Body weight

CV

Cell viability

DA

Defined approach

DC

Dendritic cell

DMF

Dimethyl formamide

DNFB

2,4,Dintrofluorobenzene

DPRA

Direct peptide reactivity assay

EC

Effective concentration

EMIM

1-Ethyl-3-methylimidazolium chloride

HaCaT

Human adult low calcium temperature (HaCaT) keratinocyte cells

HCA

Hexylcinnamic aldehyde

h-CLAT

Human cell line activation test

HPLC

High pressure liquid chromatography

IC

Inhibitory concentration

ICCVAM

Interagency Coordinating Committee on the Validation of Alternative Methods

IL

Ionic liquids

Imax

Maximum induction

IRR

Irritancy

kg

Kilogram

KS

KeratinoSens assay

LLNA

Local lymph node assay

MEST

Mouse ear swelling test

mg

Milligram

mM

Millimolar

MTT

Methylthiazolyldiphenyl-tetrazolium bromide assay

NBuPY

N-Butylpyridinium chloride

NTP

National Toxicology Program

OECD

Organization for Economic Co-operation and Development

PBS

Phosphate buffered saline

PCCO

Positive control challenge only

SI

Stimulation Index

SI3

Stimulation Index ≥3 threshold

μCi

Microcurie

μl

Microliter

μM

Micromolar

v/v

Volume to volume ratio

VH

Vehicle

VHIC

Vehicle irritancy control

w/v

Weight to volume ratio