Validating Immune Responses in In Vitro Skin Sensitization Models: A Guide for Researchers and Developers

Abstract

This article provides a comprehensive overview of the strategies and challenges in validating immune responses for in vitro skin sensitization models. Aimed at researchers, scientists, and drug development professionals, it explores the foundational immunology based on the Adverse Outcome Pathway (AOP), details current methodologies from single-assay to complex 3D immunocompetent models, and addresses key troubleshooting aspects for complex mixtures and data variability. It further outlines the framework for model validation, including the use of Defined Approaches (DAs) and performance benchmarking against historical animal and human data, serving as a critical resource for advancing non-animal safety assessments in compliance with evolving global regulations.

The Immunological Basis of Skin Sensitization: From Adverse Outcome Pathway to In Vitro Modeling

The journey from covalent binding to T-cell proliferation represents a critical sequence of events in adaptive immunity, with profound implications for areas ranging from autoimmune disease to immunotherapy development. This process is central to skin sensitization, where low molecular weight chemicals act as haptens, initiating a cascade that culminates in antigen-specific T-cell responses. The Adverse Outcome Pathway (AOP) framework developed by the OECD formally delineates this process into discrete, measurable key events, providing researchers with a structured approach to evaluate immune responses using in vitro models [1] [2]. Understanding these mechanistic steps is essential for developing reliable non-animal methods for sensitization assessment and harnessing T-cell responses for therapeutic applications.

Molecular Initiating Event: Covalent Binding to Cellular Proteins

The sensitization process initiates when low molecular weight chemicals (haptens) penetrate the skin and form stable complexes with self-proteins. Most contact sensitizers contain electrophilic functional groups that form covalent bonds with nucleophilic residues on skin proteins, particularly cysteine and lysine side chains [2]. Some chemicals require activation (pre-haptens) or metabolic transformation (pro-haptens) to become immunologically reactive [2].

Experimental Assessment Methods:

• Direct Peptide Reactivity Assay (DPRA): This OECD-approved test method evaluates hapten reactivity by measuring depletion of synthetic peptides containing either cysteine or lysine.
• Amino acid Derivative Reactivity Assay (ADRA): An alternative approach that also measures chemical reactivity with nucleophilic amino acids.
• kinetics DPRA (kDPRA): A modified version that incorporates kinetic measurements for improved potency assessment.

Protocol Overview:

• Prepare solutions of test chemical in appropriate solvent
• Incubate with synthetic peptides (cysteine-containing and lysine-containing)
• Analyze peptide depletion via high-performance liquid chromatography (HPLC)
• Calculate percentage depletion for classification according to OECD guidelines

Key Event 1: Keratinocyte Activation and Inflammatory Response

Following covalent binding, the second key event involves keratinocyte activation and release of inflammatory mediators. This response is characterized by the activation of inflammasomes and subsequent release of interleukin-18 (IL-18) and interleukin-1α (IL-1α) [2]. These cytokines create a pro-inflammatory environment that facilitates dendritic cell maturation and migration.

Experimental Assessment Methods:

• IL-8 Luc assay: Measures IL-8 promoter activity in transfected keratinocyte cell lines
• U937 CD86 test: Monitors surface CD86 expression on human monocytic U937 cells
• IL-18 bioassay: Quantifies IL-18 release using primary human keratinocytes or reconstructed human epidermis models

Key Event 2: Dendritic Cell Activation and Maturation

Dendritic cells (DCs) play a pivotal role as sentinels of the immune system, capturing and processing hapten-protein complexes before migrating to draining lymph nodes. During this phase, DCs undergo maturation characterized by upregulation of surface markers including CD86, CD83, and CCR7 [1]. The expression of CCR7 enables DCs to follow CCL19 and CCL21 chemokine gradients to lymph nodes [1].

Experimental Models and Protocols: Researchers employ three primary DC models for in vitro assessment:

• CD34-DCs: Derived from CD34+ hematopoietic progenitor cells from cord blood
• Mo-DCs: Generated from peripheral blood monocytes
• BM-DCs: Derived from mouse bone marrow precursors

Detailed Mo-DC Generation Protocol:

• Isolate peripheral blood mononuclear cells (PBMCs) via Ficoll-Paque gradient centrifugation
• Enrich monocytes by plastic adherence or CD14+ magnetic selection
• Culture monocytes with GM-CSF (800-1000 IU/mL) and IL-4 (500-1000 IU/mL) for 5-7 days
• Confirm DC phenotype by flow cytometry (CD11c+, CD14-, HLA-DR+, CD86low)
• Expose to test substances for 24-48 hours
• Measure maturation markers (CD86, CD83, CCR7) via flow cytometry

Key Event 3: T-Cell Activation and Proliferation

The final key event involves antigen-specific T-cell activation and clonal expansion. In lymph nodes, mature DCs present processed hapten-peptide complexes via MHC molecules to naïve T-cells, leading to TCR engagement and activation. Recent research has revealed that covalent TCR-pMHC interactions can occur through disulfide bonds between cysteine residues in TCR CDR3 regions and peptide-MHC complexes, profoundly influencing T-cell activation thresholds and fate decisions [3] [4].

Experimental T-Cell Activation Methods:

• Magnetic bead-bound antibodies: Beads coated with anti-CD3 and anti-CD28 antibodies provide co-stimulation
• Plate-bound antibodies: Wells coated with CD3 and CD28 antibodies for controlled stimulation
• Soluble antibodies: Free antibodies in solution, yielding less differentiated cells
• Microbubble technology: Buoyant microbubbles coated with antibodies minimize overstimulation [5]

Quantitative Comparison of T-Cell Activation Methods:

Method	CD8+ Differentiation	CD4+ Expansion	Scalability	Risk of Exhaustion
Magnetic Beads	High terminal differentiation	Moderate	High	Moderate
Plate-Bound	Variable	Good	Low	Low
Soluble Antibodies	Low differentiation	Low	Moderate	Low
Microbubbles	Moderate	Good	High	Low

Protocol for Assessing T-Cell Proliferation:

• Isolate naïve T-cells from peripheral blood (negative selection)
• Label with cell proliferation dyes (e.g., CFSE)
• Activate using chosen method (beads, plates, etc.)
• Culture in IL-2 containing medium (50-100 IU/mL)
• Analyze proliferation by CFSE dilution via flow cytometry at 72-96 hours
• Measure cytokine production (IFN-γ, IL-2) by ELISA or intracellular staining

Covalent TCR-pMHC Interactions: A Specialized Activation Mechanism

Recent groundbreaking research has identified a unique T-cell activation mechanism involving disulfide bond formation between TCR and pMHC. Studies demonstrate that cysteine residues at the apex of TCR CDR3 regions can form covalent bonds with cysteine-containing peptide-MHC complexes, inducing strong Zap70-dependent signaling that redirects T-cell fate in the thymus [3] [4].

Experimental Evidence:

• Engineered 6218 TCR variants with cysteine at CDR3α or CDR3β apex (6218αC and 6218βC) demonstrated altered T-cell development, skewing fate away from conventional T-cells and toward intraepithelial lymphocytes or deletion [3]
• These covalent interactions facilitated T-cell activation even with low-affinity pMHC ligands [3]
• The effect was Zap70-dependent, as evidenced by aberrant development into conventional T-cells in Zap70mrd/mrt mice [3]

Signaling Pathways: From Covalent Binding to T-Cell Proliferation

Diagram Title: AOP for Skin Sensitization from Covalent Binding to T-cell Proliferation

The Scientist's Toolkit: Essential Research Reagents

Reagent/Cell System	Function/Application	Example Sources
Synthetic Peptides (Cys/Lys)	Assessing hapten reactivity (KE1)	Custom synthesis, commercial vendors
Reconstructed Human Epidermis (RHE)	Evaluating keratinocyte responses (KE2)	EpiDerm, EpiSkin, SkinEthic
CD34+ Hematopoietic Progenitors	Generating human dendritic cells	Cord blood, mobilized peripheral blood
GM-CSF & IL-4 Cytokines	In vitro DC differentiation from monocytes	Miltenyi Biotec, BD Biosciences
Anti-CD3/CD28 Antibodies	T-cell activation and expansion	Various commercial suppliers
CFSE Cell Proliferation Dye	Tracking T-cell division by flow cytometry	Thermo Fisher, BioLegend
MACS Cell Separation System	Immune cell isolation and enrichment	Miltenyi Biotec
IL-2, IL-7, IL-15 Cytokines	T-cell culture and maintenance	PeproTech, R&D Systems
Atrazine-d5	Atrazine-d5, CAS:163165-75-1, MF:C8H14ClN5, MW:220.71 g/mol	Chemical Reagent
Deisopropylatrazine	Deisopropylatrazine, CAS:1007-28-9, MF:C5H8ClN5, MW:173.60 g/mol	Chemical Reagent

The pathway from covalent binding to T-cell proliferation represents a sophisticated immune activation cascade that can be systematically evaluated using defined in vitro approaches. The AOP framework provides researchers with a validated structure for investigating these key events, while emerging technologies like covalent TCR-pMHC probes and improved DC maturation assays offer increasingly refined tools for mechanistic studies. Understanding these discrete biological events enables more predictive assessment of skin sensitization potential and supports the development of novel immunotherapies that harness or modulate T-cell responses. As research continues to elucidate the nuances of these processes, particularly the role of specialized covalent interactions, our ability to precisely control immune outcomes for therapeutic benefit will continue to advance.

The Role of the Adverse Outcome Pathway (AOP) in Guiding Test Development

The Adverse Outcome Pathway (AOP) framework is an analytical construct that describes a sequential chain of causally linked events at different levels of biological organization that lead to an adverse health or ecotoxicological effect [6]. An AOP is conceptually similar to a series of dominos, where a chemical exposure initiates a biological change within a cell, triggering a cascade of sequential key events along a toxicity pathway that can ultimately result in an adverse health outcome in a whole organism [7]. This framework serves as a critical knowledge assembly, interpretation, and communication tool designed to support the translation of pathway-specific mechanistic data into responses relevant to assessing and managing risks of chemicals to human health and the environment [8]. The AOP framework provides a structured approach for interpreting new data streams often not employed by traditional risk assessment, including information from in silico models, in vitro assays, and short-term in vivo tests with molecular endpoints [8].

Each AOP begins with a Molecular Initiating Event (MIE), which represents the direct interaction between a stressor (e.g., a chemical) and a molecular target within an organism, such as binding to a receptor, inhibition of an enzyme, or damage to DNA [7]. This MIE triggers a series of measurable Key Events (KEs) at the cellular, tissue, or organ level that eventually lead to an Adverse Outcome (AO) considered relevant for risk assessment or regulatory decision-making [7]. The relationships between these events are described through Key Event Relationships (KERs), which outline the likelihood and conditions under which a particular biological change will trigger the next key event in the sequence [7]. This structured approach allows scientists to predict biological outcomes by extrapolating from available data and provides a framework for developing and validating New Approach Methodologies (NAMs) that can reduce reliance on traditional animal testing [7].

The AOP for Skin Sensitization: A Case Study in Test Development

The development of an AOP for skin sensitization represents one of the most advanced and successfully implemented applications of the AOP framework in toxicology. Skin sensitization is a chemical-induced immune response that leads to allergic contact dermatitis, a health problem affecting an estimated 15-20% of the world's population [9]. The AOP for skin sensitization initiated by covalent binding to proteins has been formally described and reviewed by the Organisation for Economic Co-operation and Development (OECD) [10]. This AOP provides the mechanistic foundation for replacing traditional animal tests with a new generation of in vitro and in chemico testing strategies.

The skin sensitization AOP consists of a linear sequence of key events beginning with the molecular initiating event of covalent binding to skin proteins (haptenation), which is followed by keratinocyte inflammation response, then dendritic cell activation, and ultimately leading to T-cell proliferation and the adverse outcome of allergic contact dermatitis [9]. This well-defined pathway has enabled the development and validation of individual test methods that target specific key events within the pathway, creating opportunities for integrated testing strategies that can comprehensively address the entire AOP without using animal models [9] [8]. The regulatory adoption of these approaches has been particularly driven by legislation such as the EU Cosmetic Regulation (1223/2009), which implemented a total animal testing ban for cosmetics [9] [11].

Table 1: Key Events in the Skin Sensitization AOP and Correspondative Test Methods

AOP Key Event	Biological Process	OECD Validated Test Methods	Measurement Endpoints
Molecular Initiating Event (KE1)	Covalent binding to skin proteins	OECD TG 442C: Direct Peptide Reactivity Assay (DPRA)	Peptide depletion via HPLC [9]
Key Event 2 (KE2)	Keratinocyte inflammatory response	OECD TG 442D: ARE-Nrf2 Luciferase Test Method (KeratinoSens)	Luminescence measurement of gene activation [9]
Key Event 3 (KE3)	Dendritic cell activation	OECD TG 442E: human Cell Line Activation Test (h-CLAT)	Flow cytometry of CD86/CD54 surface markers [9]

Experimental Validation of AOP-Based Testing Strategies

Performance of Individual Test Methods

The scientific consensus indicates that no single non-animal test method can fully address the complexity of the skin sensitization AOP and replace animal tests [9]. Consequently, researchers have focused on validating testing strategies that combine multiple in chemico and in vitro methods. A 2020 study evaluating the performance of different test methods on "difficult to test" cosmetic ingredients with particular physicochemical properties revealed significant variations in predictive capacity [11]. The DPRA model demonstrated limited predictive capability for these challenging ingredients, resulting in many false negative responses compared to animal studies, or being unsuited to the mode of action of the selected ingredients [11]. In contrast, the SENS-IS assay, which assesses the first two AOP Key Events with consideration of dermal penetration, showed real capability to discriminate sensitizers from non-sensitizers [11]. The KeratinoSens model tended to overestimate the sensitization potential of tested ingredients, while the h-CLAT model tended to underestimate sensitizers [11].

These findings highlight the importance of understanding the applicability domains and limitations of individual test methods within an AOP framework. The performance variations become particularly evident when testing materials with complex properties, such as poorly water-soluble components, surfactants, or complex substances that may fall outside the optimal operating parameters of standardized tests [11]. This understanding has driven the development of more sophisticated testing strategies that leverage the complementary strengths of multiple assays to overcome the limitations of any single method.

Integrated Testing Strategies (ITS)

Research has demonstrated that integrated testing strategies combining multiple non-animal methods can achieve high predictive accuracy. A sequential testing strategy developed for "difficult to test" ingredients that combined the SENS-IS assay (assessing the first two AOP Key Events) with follow-up testing using h-CLAT (assessing Key Event 3) and potentially KeratinoSens (assessing Key Event 2) achieved an accuracy of 88% on challenging ingredients and minimized the risk of false negative conclusions [11]. This approach strategically covers the main key events of the skin sensitization AOP while addressing specific technical challenges posed by complex ingredients.

Other proposed strategies include the "2 out of 3" approach, which uses a combination of DPRA, KeratinoSens, and h-CLAT, where concordant results from any two tests determine the classification [9]. Additionally, Integrated Approaches to Testing and Assessment (IATA) have been developed using Bayesian networks and other computational approaches to weight and combine data from different test methods [9]. The OECD has developed guidance documents on defined approaches to testing and assessment, including a general document outlining principles for using these approaches within IATA and a second document focusing on specific case studies [9]. These frameworks provide structured, hypothesis-based approaches for integrating data from various sources to support regulatory decision-making.

Table 2: Comparison of AOP-Based Testing Strategies for Skin Sensitization

Testing Strategy	Components	Reported Accuracy	Advantages	Limitations
"2 out of 3" Approach	DPRA, KeratinoSens, h-CLAT	Varies by chemical space	Simple implementation; uses validated OECD methods	Limited performance with difficult-to-test substances [9] [11]
Bayesian Network	Multiple in chemico and in vitro inputs	High in published validations	Flexible weighting of tests; probabilistic output	Complex implementation; requires specialized expertise [9]
Sequential Strategy (2020)	SENS-IS followed by h-CLAT and/or KeratinoSens	88% (difficult substances)	Handles challenging ingredients; minimizes false negatives	SENS-IS not yet OECD-validated [11]

Advanced AOP Applications in Immune Response Research

Liver Organoid Models for Immune-Mediated Drug Reactions

The AOP framework is extending beyond skin sensitization to more complex immune-mediated reactions. Recent research has developed a human liver organoid microarray platform designed to predict which drugs might trigger harmful immune responses in susceptible patients [12]. This platform combines induced pluripotent stem cell (iPSC)-derived liver organoids with a patient's own immune cells (autologous CD8⁺ T cells) to create a human, immune-competent system that reproduces the genetic and immune variation found in patients [12]. This model successfully recreated liver injury caused by the antibiotic flucloxacillin, which affects only carriers of the HLA-B*57:01 risk gene, reproducing classic signs of immune-mediated liver toxicity including T cell activation, cytokine secretion, and hepatocyte damage [12].

This advancement addresses a critical gap in conventional toxicology testing, as standard laboratory tests and animal models cannot replicate complex, patient-specific immune mechanisms responsible for idiosyncratic drug-induced liver injury (iDILI) [12]. The platform demonstrates how AOP-informed models can incorporate human genetic variability and immune responses to better predict rare but serious adverse outcomes that may not be detected in conventional animal studies or simplified in vitro systems. This approach represents a significant step toward personalized toxicology and safety assessment.

Artificial Intelligence and AOP Development

The future of AOP development is being shaped by artificial intelligence (AI) and computational approaches. Recognizing that building AOPs remains a time-consuming, largely manual process, initiatives are now exploring how AI can accelerate AOP development and strengthen the bridge between mechanistic science and regulatory decision-making [13]. These efforts aim to leverage machine learning and natural language processing to rapidly synthesize toxicological literature and identify potential key event relationships, thereby speeding up the assembly and evaluation of AOPs.

The integration of AI into the AOP framework represents a paradigm shift in how toxicological knowledge can be organized and applied. By automating the extraction and synthesis of mechanistic information from the vast scientific literature, AI-powered approaches promise to dramatically expand the coverage and currency of the AOP knowledge base, making it an even more powerful resource for test development and chemical safety assessment [13]. This is particularly important as regulatory programs increasingly require consideration of the potential health effects of thousands of chemicals, a task that cannot be accomplished using traditional toxicological approaches alone [8].

Essential Research Reagent Solutions for AOP-Based Testing

The implementation of AOP-based testing strategies requires specialized research reagents and platforms. The following table details key materials essential for conducting research in this field.

Table 3: Essential Research Reagent Solutions for AOP-Based Testing

Reagent/Platform	Function	Application in AOP Testing
Synthetic Peptides (Cysteine/Lysine)	Measure covalent binding reactivity in DPRA	Assessing Molecular Initiating Event (KE1) in skin sensitization AOP [9]
ARE-Nrf2 Reporter Cell Lines	Detect antioxidant response element activation	Measuring keratinocyte response (KE2) in skin sensitization AOP [9]
Human Monocytic Leukemia Cell Line (THP-1)	Evaluate dendritic cell activation	Assessing CD86/CD54 expression changes (KE3) in h-CLAT [9]
iPSC-Derived Liver Organoids	Model human liver responses in a genetically defined system	Studying immune-mediated drug reactions and idiosyncratic toxicity [12]
Animal-Free Extracellular Matrices	Provide human-relevant scaffolding for 3D cell culture	Supporting organoid growth without animal-derived materials like Matrigel [13]
Microfluidic Organ-on-Chip Devices	Mimic tissue-level physiology and dynamic culture conditions	Advanced model systems for key event relationships in complex AOPs [13]

The Adverse Outcome Pathway framework has fundamentally transformed the approach to test development in toxicology, providing a structured, mechanistic foundation for creating and validating new assessment methodologies. The skin sensitization AOP case study demonstrates how a well-defined pathway can facilitate the replacement of animal tests with integrated testing strategies that combine multiple in chemico and in vitro methods targeting specific key events. The continued evolution of AOP-based approaches, including the incorporation of human organoid models, artificial intelligence, and patient-specific immune responses, promises to further enhance the human relevance and predictive capacity of safety assessment. As these approaches mature, they will enable more efficient, mechanistically informed chemical evaluation while reducing reliance on traditional animal testing methods.

AOP Framework for Skin Sensitization

Sequential Testing Strategy Workflow

In immunology research and drug development, the historical focus on isolated segments of the immune response has created a critical knowledge gap. The human immune system functions as an integrated network where innate and adaptive immunity continuously communicate to mount effective protection. This cross-talk begins when innate immune cells like dendritic cells (DCs) and macrophages recognize foreign antigens, process them, and present epitopes to T cells of the adaptive immune system, initiating a specific, targeted response [14]. In transplantation, for instance, evidence now indicates that not all rejection can be explained by traditional adaptive immune paradigms, with innate cell allorecognition playing a significant role in unexplained graft inflammation [15].

Capturing this integrated response is particularly crucial for validating in vitro skin sensitization models and therapeutic development. Traditional models that focus solely on either innate or adaptive components provide an incomplete picture, potentially missing key mechanisms of immunogenicity, adverse reactions, and treatment efficacy. This guide compares current methodological approaches for comprehensive immune assessment, providing experimental data and protocols to bridge this technological gap.

Comparative Analysis of Immune Assessment Platforms

The following table summarizes key in vitro platforms capable of capturing integrated immune responses, highlighting their applications and limitations in drug development and immunotoxicity testing.

Table 1: Comparison of Integrated Immune Response Assessment Platforms

Platform	Key Components	Applications	Advantages	Limitations
Whole Blood Assay (WBA) [14]	Minimally processed human blood retaining all immune cell types	Cost-effective initial vaccine/drug candidate screening; cytokine release profiling	Retains native immune cell populations and soluble factors; minimal processing artifact	Lower cell concentrations; limited granularity for rare cell populations; immediate processing required
Monocyte-Derived DC with T-cell Interface (MoDC-DTI) [14]	In vitro differentiated monocyte-derived DCs co-cultured with autologous T-cells	Gold standard for antigen-specific T-cell response evaluation; vaccine immunogenicity testing	Controlled antigen presentation environment; enables study of DC-T cell cross-talk	MoDCs may differ functionally from natural DCs; requires complex culture conditions
Human Tissue Construct (HTC) Assay [14]	Engineered tissue constructs mimicking physiological immune environments	Advanced translational studies; tissue-specific immune responses; spatial immunity analysis	Enhanced physiological relevance; captures spatial and temporal immune interactions	High technical complexity and cost; longer experimental timelines
Cytokine Secretion Assay (CSA) [16]	Bispecific antibodies capturing secreted cytokines on viable cell surfaces	Isolation of viable cytokine-secreting cells for functional characterization; low-frequency cell detection	Preserves cell viability and function; enables multiplexed cytokine detection	Signal reduction with multiplexing; requires optimization of capture reagents
Intracellular Cytokine Staining (ICS) [17] [18]	Cell permeabilization and staining of accumulated intracellular cytokines	Functional immunophenotyping; identification of cytokine-producing cell subsets	High multiparameter capability combined with cell surface markers	Requires cell fixation; eliminates possibility of subsequent functional assays

Key Methodologies for Comprehensive Immune Profiling

In Vitro Immunization (IVI) Assays

IVI assays represent a sophisticated approach to recapitulate the complete immune response cascade in a controlled laboratory setting. These systems typically follow three core steps: (1) immune cell isolation, (2) differentiation and antigen stimulation, and (3) immune response readout [14]. The most advanced IVI platforms include:

• Whole Blood Assay (WBA): This cost-effective approach utilizes diluted human blood containing all native immune cell populations, incubated with test antigens after collection in anticoagulant tubes. Following incubation, supernatant can be analyzed for secreted molecules, or cells can be processed for gene and protein expression analysis [14].

• MoDC-DTI System: This two-stage platform first generates monocyte-derived DCs, which are then pulsed with antigens and co-cultured with autologous T-cells. This setup directly models the critical innate-adaptive interface where DCs present antigen to T-cells, initiating adaptive immune activation [14].

• Human Tissue Construct (HTC) Assays: These engineered systems incorporate multiple cell types in three-dimensional architectures that better mimic in vivo tissue environments, providing enhanced physiological relevance for studying spatial aspects of immune responses [14].

Functional Cytokine Profiling Assays

Functional cytokine analysis provides crucial insights into immune cell activity, with distinct methodological advantages:

• Intracellular Cytokine Staining (ICS): This flow cytometry-based method involves cell stimulation with antigens or nonspecific activators like PMA/ionomycin in the presence of secretion inhibitors (Brefeldin A/monensin). Cells are then fixed, permeabilized, and stained with fluorescent antibodies against cytokines, allowing identification of cytokine-producing subsets when combined with cell surface markers [18]. Studies demonstrate ICS can detect drug-specific cytokine production in 75% of patients with drug hypersensitivity reactions [17].

• Cytokine Secretion Assay (CSA): This viable cell approach uses bispecific antibodies that bind to cell surface markers (e.g., CD45) on one end and capture specific cytokines on the other. During stimulation, secreted cytokines are bound by these capture reagents and detected with fluorochrome-conjugated anti-cytokine antibodies, enabling sorting of viable cytokine-secreting cells for downstream functional applications [16]. Research shows CSA performs equivalently or superiorly to ICS for detecting low-frequency cytokine-secreting cells like IL-10+ B cells (CSA: 2.22% ± 0.59% vs ICS: 0.81% ± 0.28%) [16].

Table 2: Performance Comparison of Cytokine Detection Methods

Parameter	Intracellular Cytokine Staining	Cytokine Secretion Assay
Cell Viability	Not preserved (fixed cells)	Preserved (viable cells)
Detection Sensitivity	75% for drug-specific responses [17]	Equivalent or superior to ICS for low-frequency cells [16]
Multiplexing Capability	High (with panel optimization)	Moderate (signal reduction with multiple cytokines)
Downstream Applications	Limited to molecular analysis	Functional assays, cell culture, adoptive transfer
Low-Frequency Cell Detection	Moderate	Excellent for frequencies as low as 2-5% [16]

Key Signaling Pathways in Integrated Immunity

Innate-Activated Adaptive Immunity Pathway

The critical signaling cascade that bridges innate and adaptive immunity begins with antigen uptake by innate immune cells and culminates in pathogen-specific adaptive responses. This pathway can be visualized as follows:

Diagram 1: Innate to Adaptive Immune Activation

This pathway illustrates how innate immune cells (macrophages, dendritic cells) initially respond to vaccine or pathogen antigens through phagocytosis [14]. These cells then digest and present antigen epitopes to T cells via major histocompatibility complex (MHC) proteins, providing the first of three necessary signals for T-cell activation: (1) T-cell receptor binding to MHC-peptide complex, (2) CD28 on T-cells binding with CD80/CD86 on antigen-presenting cells, and (3) cytokine production by activated innate cells [14]. Once activated, T cells stimulate B cells, leading to antibody production and generation of long-term immunological memory [14].

Missing Self Recognition Pathway

In transplantation immunology, the "missing self" pathway represents a crucial innate immune mechanism that can trigger tissue rejection independent of adaptive immune responses:

Diagram 2: Missing Self Recognition in Transplantation

This pathway explains how natural killer (NK) cells become activated when encountering donor cells lacking compatible human leukocyte antigen (HLA) class I molecules that normally engage inhibitory killer-cell immunoglobulin-like receptors (KIRs) on NK cells [15]. This "missing self" recognition results in lost inhibition and subsequent NK cell activation, causing endothelial damage and microvascular inflammation in transplanted tissues [15]. Notably, this innate mechanism can produce histologic patterns indistinguishable from antibody-mediated rejection but requires different therapeutic approaches, highlighting why comprehensive immune assessment is clinically essential [15].

Essential Research Reagent Solutions

The following research reagents and tools are fundamental for implementing comprehensive immune assessment protocols:

Table 3: Essential Research Reagents for Integrated Immune Monitoring

Reagent/Category	Key Examples	Research Applications
Cell Isolation Media	Peripheral Blood Mononuclear Cells (PBMCs), Density gradient centrifugation media	Isolation of primary immune cells from whole blood for in vitro assays [14] [16]
Cell Stimulation Reagents	PMA/Ionomycin, Antigenic peptides/proteins, LPS	Polyclonal and antigen-specific activation of immune cells for functional assays [17] [16]
Secretion Inhibitors	Brefeldin A, Monensin	Intracellular accumulation of cytokines for ICS detection by flow cytometry [18]
Cytokine Capture Reagents	Bispecific antibodies (anti-surface marker/anti-cytokine)	Viable cell cytokine secretion assay (CSA) for sorting functional subsets [16]
Detection Antibodies	Fluorochrome-conjugated anti-cytokine antibodies, Cell surface marker antibodies	Multiparameter flow cytometry analysis of immune cell phenotypes and functions [18] [16]
Cell Culture Media	Serum-free media, Cytokine supplements (GM-CSF, IL-4 for MoDC differentiation)	Maintenance and differentiation of primary immune cells for IVI assays [14]

The future of immunology research and drug development lies in embracing integrated assessment platforms that capture the dynamic interplay between innate and adaptive immunity. As evidenced by the methodologies compared in this guide, technological advances now enable researchers to move beyond siloed immune analysis toward systems that reflect biological reality. The clinical implications are significant – from explaining previously enigmatic transplant rejection episodes [15] to developing safer biologics with reduced immunogenicity risk [19]. As these integrated approaches become more accessible and standardized, they will accelerate the development of more effective immunotherapeutics, vaccines, and diagnostic tools that account for the full complexity of human immune responses.

The field of toxicology is undergoing a fundamental transformation, driven by a convergence of regulatory bans on animal testing and significant advancements in human-relevant biology and engineering. New Approach Methodologies (NAMs) represent a suite of innovative tools—including in vitro assays, computational models, and microphysiological systems—designed to deliver more human-predictive safety assessments while reducing reliance on traditional animal testing [20]. This shift is particularly evident in skin sensitization testing, where the complex immunobiology of allergic contact dermatitis has been deconstructed into measurable key events through the Adverse Outcome Pathway (AOP) framework, enabling the development of non-animal methods that can accurately assess this endpoint [2].

The regulatory impetus for this change is unmistakable. The EU Cosmetics Regulation (1223/2009) effectively banned animal testing for cosmetic ingredients, creating an urgent need for alternative approaches [2]. More recently, the U.S. Food and Drug Administration (FDA) announced a groundbreaking plan to phase out animal testing requirements for monoclonal antibodies and other drugs, encouraging the use of advanced computer simulations and human-based lab models instead [21]. This regulatory landscape has accelerated the development, validation, and implementation of NAMs, positioning them as the future cornerstone of chemical safety and drug development.

The Regulatory Landscape: From Animal Bans to NAMs Integration

Global Regulatory Drivers

The transition to NAMs is being shaped by evolving regulatory policies worldwide. These policies are increasingly mandating the reduction, refinement, and ultimate replacement (the 3Rs) of animal testing while creating pathways for the acceptance of human-relevant data.

Table 1: Key Global Regulatory Developments Driving NAMs Adoption

Region/Organization	Regulatory Action	Key Provisions & Impact	Timeline/Status
European Union	Cosmetics Regulation 1223/2009 [2]	Bans animal testing for cosmetic ingredients and finished products within the EU.	Fully in force
United States	FDA Modernization Act 3.0 & 2025 FDA Roadmap [21]	Phases out animal testing requirement for monoclonal antibodies; encourages NAMs data in regulatory submissions.	Implementation began 2025
International	OECD Test Guidelines [20]	Adopts Defined Approaches (DAs) combining NAMs for endpoints like skin sensitization (e.g., TG 497).	Ongoing; multiple guidelines adopted
International	REACH Regulation (EC) [22]	Prioritizes non-animal methods for sensitization potential; largest repository of toxicology data.	Updated 2016

The Scientific Framework: Adverse Outcome Pathway for Skin Sensitization

The adoption of NAMs has been facilitated by the development of the Adverse Outcome Pathway (AOP) framework, which breaks down complex toxicological responses into a sequence of measurable key events. For skin sensitization, the AOP outlines four critical key events that can be evaluated using specific NAMs [2].

Figure 1: The Adverse Outcome Pathway (AOP) for Skin Sensitization. This pathway delineates the sequence of biological events from the initial molecular interaction to the adverse health outcome, with each key event (KE) addressed by specific New Approach Methodologies (NAMs) [2].

Methodologies: Technical Approaches for Skin Sensitization Assessment

Defined Approaches and Integrated Testing Strategies

For regulatory decision-making, Defined Approaches (DAs) have been developed. These are fixed combinations of specific information sources (e.g., in chemico, in vitro) that are processed through a standardized data interpretation procedure to predict a hazard or potency [20]. The OECD Test Guideline 497 provides a framework for such defined approaches for skin sensitization, validating the use of integrated NAMs data without requiring animal testing [20] [2].

Experimental Protocols for Key Events in Skin Sensitization

Protocol for KE1: Direct Peptide Reactivity Assay (DPRA)

The DPRA is an in chemico method that evaluates a chemical's ability to covalently bind to proteins, the Molecular Initiating Event in the skin sensitization AOP [2].

• Principle: Test chemicals are incubated with synthetic peptides containing either cysteine or lysine. Sensitizers deplete peptide content by forming covalent adducts.
• Procedure:
  • Prepare peptide solutions: 0.667 mM cysteine peptide and 0.667 mM lysine peptide in phosphate buffer.
  • Incubate test chemical (liquid or solid) with each peptide solution for 24 hours at 25°C.
  • Analyze samples using high-performance liquid chromatography (HPLC) to quantify remaining peptide.
• Data Analysis: Calculate percent depletion for each peptide. Use a prediction model combining cysteine and lysine depletion to classify chemicals as sensitizers or non-sensitizers.

Protocol for KE2: Interleukin-18 Release Assay

The IL-18 release assay uses a reconstructed human epidermis (RHE) model to assess keratinocyte activation, the second key event [2].

• Principle: Sensitizers induce the release of the pro-inflammatory cytokine IL-18 from keratinocytes in a 3D skin model.
• Procedure:
  • Apply test substance topically to the RHE model (e.g., EpiDerm, EpiSkin) for 24 hours.
  • After a 42-hour post-incubation period, collect culture media.
  • Quantify IL-18 concentration in the media using a commercial enzyme-linked immunosorbent assay (ELISA) kit.
• Data Analysis: Compare IL-18 levels to established thresholds to classify the sensitization potential of the test substance.

Protocol for KE3: Human Cell Line Activation Test (h-CLAT)

The h-CLAT assesses dendritic cell activation by measuring changes in the expression of surface markers CD86 and CD54 [2].

• Principle: The test uses the human THP-1 monocytic leukemia cell line. Sensitizers induce increased expression of the surface antigens CD86 and CD54.
• Procedure:
  • Expose THP-1 cells to the test chemical at various concentrations for 24 hours.
  • Stain cells with fluorescently labeled antibodies against CD86 and CD54.
  • Analyze fluorescence intensity using flow cytometry.
• Data Analysis: Determine the relative fluorescence intensity (RFI) for each marker. A chemical is positive if it induces at least 150% RFI for CD86 and/or 200% RFI for CD54 at any tested concentration where cell viability is ≥50%.

Data Comparison: Performance of NAMs vs. Traditional Animal Models

The scientific validation of NAMs relies on rigorous assessment of their predictive capacity compared to traditional animal tests and, where available, human data.

Table 2: Predictive Performance of Selected NAMs for Skin Sensitization [20] [22] [2]

Method (OECD TG)	AOP Key Event	Endpoint Measured	Reported Accuracy	Human Relevance
DPRA (442C)	KE1	Peptide reactivity	~85% (vs. LLNA)	Direct measurement of chemical reaction
KeratinoSens (442D)	KE2	Nrf2-mediated gene activation	~83% (vs. LLNA)	Uses human keratinocyte cell line
h-CLAT (442E)	KE3	CD86/CD54 expression on dendritic cells	~87% (vs. LLNA)	Uses human monocytic cell line (THP-1)
EpiSensA (442D)	KE2	IL-18 release from RHE	~90% (vs. human data)	Uses reconstructed human epidermis
LLNA (Animal Test)	N/A	Lymph node proliferation in mice	40-65% (vs. human toxicity) [20]	Species differences limit translation

It is crucial to note that the benchmarking of NAMs against animal data presents a scientific challenge. While often treated as a "gold standard," the mouse Local Lymph Node Assay (LLNA) itself has a documented human toxicity predictivity rate of only 40-65% [20]. Therefore, superior performance of a NAM is not necessarily defined by its ability to replicate animal test results, but by its capacity to more accurately predict outcomes in humans.

The Scientist's Toolkit: Essential Reagents and Models

Successful implementation of NAMs requires specific biological reagents, cell models, and analytical tools.

Table 3: Key Research Reagent Solutions for In Vitro Skin Sensitization Research

Reagent / Material	Function & Application	Example Use Case
Reconstructed Human Epidermis (RHE)	3D model of human skin; assesses keratinocyte response (KE2) and chemical penetration.	EpiSensA test for IL-18 release [2].
THP-1 Cell Line	Human monocyte line; differentiates into dendritic-like cells to assess cell activation (KE3).	Human Cell Line Activation Test (h-CLAT) measuring CD86/CD54 [2].
Synthetic Peptides (Cys/Lys)	Mimic skin protein nucleophiles; measure direct peptide reactivity (KE1).	Direct Peptide Reactivity Assay (DPRA) [2].
Cytokine Detection Kits (e.g., IL-18 ELISA)	Quantify inflammatory mediators released by activated keratinocytes.	Quantification of KE2 response in RHE models [2].
Flow Cytometry Antibodies (CD86, CD54)	Detect cell surface markers indicative of dendritic cell activation.	Readout for h-CLAT and other dendritic cell activation assays [2].
1,3-Dimethylpyrazole	1,3-Dimethylpyrazole | High-Purity Reagent | RUO	High-purity 1,3-Dimethylpyrazole for research. A versatile heterocyclic building block for organic synthesis & medicinal chemistry. For Research Use Only.
d-AP5	5-Phosphono-D-norvaline | NMDA Receptor Antagonist	5-Phosphono-D-norvaline is a potent and selective NMDA receptor antagonist for neuroscience research. For Research Use Only. Not for human or veterinary use.

Challenges and Future Directions

Despite significant progress, barriers to the widespread adoption of NAMs persist. These include scientific and technical hurdles, such as accurately modeling complex endpoints like systemic toxicity [20], as well as regulatory and cultural obstacles related to familiarity with established animal methods and perceptions of regulatory acceptance [20] [23].

Future advancements will focus on increasing model complexity and integration. Technologies such as microphysiological systems (organs-on-chips) that incorporate immune components aim to better recapitulate the spatial and temporal interactions between skin and the immune system [2] [24]. Furthermore, the integration of omics technologies and AI-driven computational models promises to enhance the predictive power of NAMs, moving the field toward a more comprehensive and human-relevant framework for safety assessment known as Next Generation Risk Assessment (NGRA) [20]. This exposure-led, hypothesis-driven approach integrates data from various NAMs to deliver protective safety decisions without reliance on animal data [20].

A Toolkit for Researchers: From Single-Key Event Assays to Complex Immunocompetent Models

The assessment of skin sensitization potential, a critical endpoint in chemical safety, has undergone a paradigm shift with the adoption of the Adverse Outcome Pathway (AOP) framework. The AOP describes the sequence of biological events leading to allergic contact dermatitis (ACD), a T cell-mediated inflammatory skin condition affecting 15–20% of the general population [25] [2]. This mechanistic understanding has enabled the development and validation of New Approach Methodologies (NAMs) that target specific key events within the AOP, moving regulatory testing away from traditional animal methods like the murine Local Lymph Node Assay (LLNA) [25] [26].

The skin sensitization AOP, as formalized by the OECD, is built upon four essential Key Events (KE): KE1 is the Molecular Initiating Event, involving covalent binding of electrophilic chemicals to skin proteins [25] [27]. KE2 represents the inflammatory response of keratinocytes, KE3 is the activation of dendritic cells, and KE4 is the proliferation of antigen-specific T-cells [25] [28]. This guide provides a comparative analysis of four OECD-validated, single-Key Event assays—DPRA, KeratinoSens, h-CLAT, and ADRA—which address the first three key events of this AOP. These assays are cornerstone tools for researchers and regulators in the non-animal assessment of skin sensitization hazard [2] [26].

The following table provides a consolidated overview of the core characteristics of the four OECD-validated assays.

Table 1: Overview of OECD-Validated Single Key Event Assays for Skin Sensitization

Assay Name	Key Event Addressed	OECD Test Guideline	Principle & Measured Endpoint	Test System
DPRA (Direct Peptide Reactivity Assay)	KE1 (Molecular Initiating Event) [28]	TG 442C [28]	Measures peptide depletion via HPLC; assesses direct covalent binding to synthetic peptides containing cysteine or lysine [2]	In chemico (Synthetic peptides) [25]
ADRA (Amino Acid Derivative Reactivity Assay)	KE1 (Molecular Initiating Event) [28]	TG 442C [28]	Measures peptide depletion via spectrophotometry; assesses direct covalent binding to synthetic peptides [2]	In chemico (Synthetic peptides) [25]
KeratinoSens	KE2 (Keratinocyte Activation) [28]	TG 442D [28]	Measures luciferase gene activity under control of the Antioxidant Response Element (ARE); detects Nrf2 pathway activation [28]	In vitro (Genetically modified KeratinoSens cell line) [28]
h-CLAT (human Cell Line Activation Test)	KE3 (Dendritic Cell Activation) [28]	TG 442E [28]	Measures surface expression of CD54 and CD86 immunomarkers via flow cytometry; detects dendritic cell maturation [28]	In vitro (THP-1 human monocytic cell line) [28]
Daltroban	Daltroban | TP Receptor Antagonist | High Purity	Daltroban is a selective TP receptor antagonist for cardiovascular and inflammation research. For Research Use Only. Not for human or veterinary use.	Bench Chemicals
Fluoroglycofen	Fluoroglycofen | Herbicide | Research Compound	Fluoroglycofen is a PPO-inhibiting herbicide for plant biology research. For Research Use Only. Not for human or veterinary use.	Bench Chemicals

Detailed Assay Methodologies and Protocols

KE1 Assays: DPRA and ADRA

The Direct Peptide Reactivity Assay (DPRA) is an in chemico method that quantifies a chemical's direct reactivity, the Molecular Initiating Event in the AOP. The assay uses two synthetic heneicosapeptides containing either cysteine or lysine, mimicking nucleophilic centers in skin proteins [2]. The experimental workflow is as follows [2]:

• Incubation: The test chemical is incubated with each peptide separately in a phosphate buffer (pH 7.5 for cysteine peptide; pH 10.2 for lysine peptide) for 24 hours at 25°C.
• Analysis: The remaining peptide concentration is analyzed using High-Performance Liquid Chromatography (HPLC) with a UV detector (e.g., set at 220 nm).
• Calculation: The percentage of peptide depletion is calculated for each peptide. The final result is the mean peptide depletion, which is used for categorizing the sensitization potential.

The Amino Acid Derivative Reactivity Assay (ADRA) is a similar in chemico method also covered under OECD TG 442C. A key operational difference is that ADRA uses spectrophotometric analysis (e.g., using a microplate reader) instead of HPLC to measure the depletion of the cysteine-containing peptide, offering a potentially higher-throughput alternative [2] [28]. A modified version, the kinetics DPRA (kDPRA), introduces a time-course measurement to improve the accuracy of potency assessments [2].

KE2 Assay: KeratinoSens

The KeratinoSens assay addresses KE2 by measuring the activation of the Nrf2-mediated antioxidant pathway in a transfected keratinocyte cell line. This pathway is activated by electrophilic substances and oxidative stress [28]. The standard protocol is [28]:

• Cell Culture and Exposure: KeratinoSens cells are cultured and exposed to a range of concentrations of the test chemical for 48 hours. Cytotoxicity is assessed in parallel.
• Luciferase Measurement: After exposure, the cell lysates are prepared, and luciferase activity is measured using a luminometer.
• Data Interpretation: A chemical is considered positive if it induces a statistically significant increase in luciferase activity (≥1.5-fold over solvent control) at any concentration where cell viability is ≥70%.

KE3 Assay: h-CLAT

The human Cell Line Activation Test (h-CLAT) assesses KE3, the activation of dendritic cells, by measuring the upregulation of surface markers CD54 and CD86 on THP-1 cells (a human monocyte line that differentiates into dendritic-like cells) [28]. The detailed protocol involves:

• Cell Exposure and Viability: THP-1 cells are exposed to the test chemical for 24 hours. A preliminary range-finding test is conducted to determine the concentrations that cause 50-150% cell viability (CV75-CV150) using a viability marker like MTT.
• Immunostaining and Flow Cytometry: Cells from the main test are harvested, stained with fluorescently-labeled antibodies against CD54 and CD86, and analyzed using a flow cytometer.
• Classification: The Relative Fluorescence Intensity (RFI) is calculated. A chemical is classified as a sensitizer if, at any tested concentration where viability is ≥50%, it induces an RFI of CD54 ≥ 150% or an RFI of CD86 ≥ 200% in at least two of three independent runs [28].

Signaling Pathways and Experimental Workflows

The following diagrams illustrate the biological pathways and standardized experimental workflows for these assays.

Skin Sensitization AOP and Key Event Assays

Diagram 1: AOP and corresponding assays. This diagram maps the OECD-validated single-Key Event assays onto the specific key events of the skin sensitization Adverse Outcome Pathway (AOP) that they are designed to address [25] [28].

KeratinoSens KE2 Signaling Pathway

Diagram 2: KeratinoSens mechanism. The assay detects KE2 by measuring activation of the Nrf2-ARE pathway. Electrophilic sensitizers modify Keap1, leading to Nrf2 release, nuclear translocation, and ARE-driven luciferase expression [28].

h-CLAT Experimental Workflow

Diagram 3: h-CLAT workflow. The assay measures KE3 by quantifying surface marker expression on THP-1 cells. Cells are exposed to the chemical, stained for CD54/CD86, and analyzed by flow cytometry to determine sensitization potential [28].

The Scientist's Toolkit: Essential Research Reagents

Successful implementation of these assays requires specific, high-quality reagents and instruments. The following table details key materials and their functions.

Table 2: Essential Research Reagents and Tools for Single-Key Event Assays

Category	Specific Item	Function in the Assay
Cell Lines & Biochemicals	THP-1 human monocyte cell line	Differentiates into dendritic-like cells; used as the test system in the h-CLAT [28].
	KeratinoSens transfected keratinocyte line	Stably incorporates the ARE-luciferase reporter gene; test system for the KeratinoSens assay [28].
	Synthetic peptides (Cysteine & Lysine)	Mimic nucleophilic sites in skin proteins; react with test chemicals in DPRA/ADRA [2].
Antibodies & Detection	Fluorochrome-conjugated anti-human CD54 (ICAM-1)	Binds to and labels the CD54 surface protein for detection by flow cytometry in h-CLAT [28].
	Fluorochrome-conjugated anti-human CD86	Binds to and labels the CD86 surface protein for detection by flow cytometry in h-CLAT [28].
	Luciferase Assay Reagent / Substrate	Reacts with the luciferase enzyme to produce bioluminescent light, quantified in the KeratinoSens assay [28].
Key Instruments	Flow Cytometer	Essential instrument for quantifying fluorescence intensity of cell surface markers in the h-CLAT [28].
	Luminometer	Precisely measures the low-intensity light emitted by the luciferase reaction in the KeratinoSens assay [28].
	HPLC System with UV Detector	Separates and quantifies peptide concentrations in the DPRA [2].
	UV/Visible Spectrophotometer (Microplate Reader)	Measures peptide depletion spectrophotometrically in the ADRA [2].
2-methyl-5-HT	2-Methyl-5-hydroxytryptamine | High-Purity 5-HT Agonist	High-purity 2-Methyl-5-hydroxytryptamine, a selective 5-HT1 receptor agonist for neurological research. For Research Use Only. Not for human or veterinary use.
1,4-Cyclohexanedione	1,4-Cyclohexanedione | High-Purity Reagent | RUO	High-purity 1,4-Cyclohexanedione for research. A key building block in organic synthesis & materials science. For Research Use Only. Not for human or veterinary use.

The OECD-validated single-Key Event assays—DPRA, ADRA, KeratinoSens, and h-CLAT—represent a foundational toolkit for modern, human-relevant skin sensitization assessment. Each assay provides mechanistic insight into a specific key event of the well-defined AOP, from initial chemical reactivity to dendritic cell activation. While these methods are mature and regulatory-accepted, the field continues to advance with the development of integrated testing strategies that combine these assays, as well as more complex models like immunocompetent reconstructed skin tissues, to improve predictive accuracy and potency assessment without animal testing [25] [2] [27]. For researchers in drug and chemical development, mastering these protocols and understanding their place in the AOP is crucial for generating robust safety data that meets contemporary regulatory and ethical standards.

Integrated Testing Strategies (ITS) and Defined Approaches (DAs) for Potency Prediction

The evaluation of skin sensitization potency is a critical component of safety assessment for cosmetics, pharmaceuticals, and industrial chemicals. With the implementation of regulatory bans on animal testing for cosmetics in the European Union and other regions, the development and validation of non-animal methods has become imperative [2]. Skin sensitization is a complex process that can lead to allergic contact dermatitis (ACD), a T-cell-mediated inflammatory skin condition affecting approximately 20% of the European population [2]. The traditional animal-based methods, particularly the murine Local Lymph Node Assay (LLNA), have historically provided potency information through the EC3 value (the estimated concentration required to produce a stimulation index of 3) [29]. However, regulatory and ethical demands have accelerated the development of New Approach Methodologies (NAMs), including Integrated Testing Strategies (ITS) and Defined Approaches (DAs) that combine multiple non-animal data sources to predict sensitization hazard and potency [29] [30]. This review comprehensively compares the performance, experimental protocols, and applications of these evolving approaches within the framework of validating in vitro skin sensitization models for immune response research.

Foundational Concepts: From AOP to Regulatory Frameworks

The Adverse Outcome Pathway (AOP) for Skin Sensitization

The OECD's Adverse Outcome Pathway (AOP) for skin sensitization provides a conceptual framework that organizes existing knowledge about the linkage between a molecular initiating event and an adverse outcome at the organism level [31]. The AOP describes the sensitization process through four key events (KE):

• KE1 (Molecular Initiating Event): Covalent binding of haptens to skin proteins.
• KE2: Keratinocyte activation and inflammatory responses.
• KE3: Activation of dendritic cells in the skin.
• KE4: Proliferation of hapten-specific T-cells [27] [31].

This AOP framework enables the development of test methods that target specific biological events in the sensitization pathway, providing a mechanistic foundation for ITS and DAs [31].

Defining Integrated Testing Strategies and Defined Approaches

Integrated Testing Strategies (ITS) represent flexible, often tiered approaches that combine information from multiple sources (in chemico, in vitro, in silico) in a weight-of-evidence manner to address a specific regulatory need [31]. In contrast, Defined Approaches (DAs) are more formalized testing strategies that consist of "fixed data generation and interpretation procedures" [30]. According to OECD Guideline No. 497, a DA includes:

• Input data generated from specifically identified methods
• A fixed data interpretation procedure that translates data into a prediction [30]

This distinction is important for regulatory acceptance, as DAs provide standardized protocols that ensure consistency and reproducibility across different laboratories and contexts.

Major Defined Approaches and Integrated Testing Strategies

OECD Guideline 497 Defined Approaches

The OECD Guideline 497, first issued in June 2021 and updated in 2025, represents the first internationally harmonized guideline describing a non-animal approach that can replace animal tests for identifying skin sensitizers [30]. This guideline incorporates several DAs, including:

• ITSv1 DA: Integrates data from the Direct Peptide Reactivity Assay (DPRA), human Cell Line Activation Test (h-CLAT), and Derek Nexus (in silico tool) [29]
• ITSv2 DA: Utilizes DPRA, h-CLAT, and OECD QSAR Toolbox [29]
• 2 out of 3 Strategy: Considers consensus from any two of three tests (DPRA, KeratinoSens, h-CLAT) [32]
• KE3/1 Sequential Testing Strategy: Employs a tiered approach starting with h-CLAT (KE3) followed by DPRA (KE1) [32]

These DAs can predict skin sensitization hazard and potency subcategorization according to the United Nations Globally Harmonized System (GHS), classifying chemicals as Category 1A (strong), Category 1B (weak), or not classified [29].

Advanced ITS-Based Methods for Potency Prediction

ITSv1-Based Read-Across

A novel strategy incorporating ITSv1 DA into read-across (RAx) has been developed to refine potency prediction by estimating EC3 values with high confidence [29]. This ITSv1-based RAx approach follows a systematic workflow:

• Selection of suitable analogues using in silico tools (OECD QSAR Toolbox, ChemTunes.ToxGPS, AMBIT) based on structural similarity
• Comparison of ITSv1 DA scores, predictions, and LLNA data between target chemical and analogues
• Determination of predicted EC3 (pEC3) value for the target chemical based on analogue data [29]

In a case study on the fragrance material lilial, this approach determined a pEC3 value of 9.5%, which was close to the historical LLNA EC3 value of 8.6%, demonstrating its potential for reliable potency estimation [29].

Machine Learning Approaches

Machine learning models have been developed to predict skin sensitization potency using non-animal data. Strickland et al. implemented a two-tiered strategy using Support Vector Machine (SVM) that first classifies sensitizers from non-sensitizers, then further classifies sensitizers as strong or weak [33]. This approach demonstrated 88% accuracy for predicting LLNA outcomes and 81% accuracy for human outcomes, outperforming the LLNA's accuracy for predicting human potency categories (69%) [33].

Table 1: Performance Comparison of Potency Prediction Approaches

Approach	Basis	Accuracy (LLNA)	Accuracy (Human)	Advantages	Limitations
ITSv1-based RAx	Read-across with ITSv1 DA	Case study: pEC3 9.5% vs actual 8.6%	Not specified	Provides quantitative EC3 estimation	Limited validation on broad chemical domains
Machine Learning (2-tiered SVM)	DPRA, h-CLAT, KeratinoSens, physicochemical properties	88% (120 substances)	81% (87 substances)	High accuracy, automated classification	Requires extensive training data
LLNA	Animal test	Reference standard	69% (136 substances)	Historical benchmark	Ethical concerns, species differences
ITSv1 DA	DPRA, h-CLAT, Derek Nexus	Categorization only (no EC3)	Categorization only (no EC3)	OECD guideline, standardized	Cannot estimate exact EC3 values

Experimental Protocols and Methodologies

Key Test Methods in ITS and DAs

Direct Peptide Reactay Assay (DPRA)

The DPRA addresses KE1 of the AOP by measuring the covalent binding of chemicals to synthetic peptides containing either cysteine or lysine [30] [32]. The assay quantifies peptide depletion through high-performance liquid chromatography (HPLC) and classifies chemicals as having high, moderate, or low reactivity based on predetermined thresholds [33].

KeratinoSens Assay

This assay addresses KE2 by measuring the activation of the Nrf2 antioxidant response pathway in a transfected keratinocyte cell line [33]. The method detects luciferase activity as an indicator of pathway activation and determines a chemical's sensitization potential based on induction criteria and cytotoxicity measures [33].

Human Cell Line Activation Test (h-CLAT)

The h-CLAT addresses KE3 by quantifying the expression of CD54 and CD86 surface markers on the human monocytic THP-1 cell line after exposure to test chemicals [28] [33]. Flow cytometry is used to measure marker expression, with specific thresholds (CD54 ≥ 150% and CD86 ≥ 200% relative fluorescence intensity) indicating positive activation [28].

Workflow of ITSv1-Based Read-Across

The following diagram illustrates the experimental workflow for the ITSv1-based read-across approach:

Immunocompetent Skin Models

Advanced immunocompetent skin models represent a significant innovation in capturing the complex immune responses in skin sensitization. The ImmuSkin-MT model incorporates:

• Hair follicle-derived keratinocytes and fibroblasts forming a physiologically relevant epidermis and dermis
• Monocyte-derived Langerhans cells (MoLCs) positioned beneath the dermal layer
• CD4+ T-lymphocytes co-cultured in the lower chamber of a transwell system [27]

This model captures both KE3 (dendritic cell activation) and KE4 (T-cell proliferation) of the AOP, enabling differentiation between extreme, moderate, and weak sensitizers based on MoLC migration, CD86 expression, and T-cell proliferation [27].

Table 2: The Scientist's Toolkit: Essential Research Reagents and Methods

Reagent/Assay	Biological Target	Key Measurements	Application in ITS/DA
DPRA	Peptide reactivity (KE1)	Peptide depletion (%) via HPLC	Hazard identification, potency categorization
KeratinoSens	Nrf2 pathway (KE2)	Luciferase activity, cytotoxicity	KE2 activation assessment
h-CLAT	Dendritic cell activation (KE3)	CD54/CD86 expression via flow cytometry	KE3 activation assessment
THP-1 cells	Human monocytic cell line	Surface marker expression, cytokine secretion	h-CLAT, co-culture models
Reconstructed human epidermis	3D skin model	IL-18 secretion, tissue viability	Complex KE2 assessment
OECD QSAR Toolbox	In silico prediction	Structural alerts, read-across analogues	Analog identification, data gap filling

Performance Comparison and Validation

Accuracy of Defined Approaches

The performance of DAs has been extensively evaluated against both animal and human reference data. The 2 out of 3 DA demonstrated approximately 80-85% accuracy for hazard identification when compared to LLNA results [32]. For potency prediction, the ITSv1 DA correctly categorizes chemicals according to GHS classifications but cannot provide quantitative EC3 values without incorporation into read-across strategies [29].

Machine learning approaches that integrate data from DPRA, KeratinoSens, h-CLAT, and physicochemical properties have shown particularly strong performance. The two-tiered SVM model achieved not only high overall accuracy but also correctly classified a higher percentage of strong human sensitizers compared to the LLNA, which underclassified one-third of strong human sensitizers as weak [33].

Addressing Current Limitations

Despite these advances, significant challenges remain in potency prediction. A primary limitation is the insufficient dynamic range of many alternative test methods compared to the four orders of magnitude spanned by LLNA EC3 values [31]. Additionally, the expression of potency in weight-based units rather than molar units may compromise the robustness of predictions, particularly for quantitative structure-activity relationship (QSAR) models [31].

Advanced immunocompetent models show promise in addressing these limitations by providing a more physiologically relevant environment that captures cell-cell interactions critical for immune activation [27]. However, these complex models currently face challenges with variability and reproducibility, limiting their regulatory acceptance [27].

Integrated Testing Strategies and Defined Approaches represent a paradigm shift in skin sensitization potency assessment, moving away from animal testing toward mechanistic, human biology-based methods. The OECD Guideline 497 DAs provide standardized frameworks for hazard identification and potency categorization, while emerging approaches like ITSv1-based read-across and machine learning models offer promising pathways for quantitative potency prediction.

The ongoing development of increasingly complex immunocompetent skin models that incorporate multiple cell types (keratinocytes, dendritic cells, T-cells) will enhance our ability to capture the key immunological events in skin sensitization. However, for immediate regulatory applications, Defined Approaches that combine existing validated methods offer the most practical solution for skin sensitization potency assessment aligned with the 3Rs principles of replacement, reduction, and refinement of animal testing.

As the field evolves, future research should focus on expanding the chemical domain of applicability for these approaches, improving the prediction of potency for problematic chemical classes (e.g., pre- and pro-haptens), and enhancing the quantitative accuracy of EC3 value predictions to support robust risk assessment decisions.

The ban on animal testing for cosmetics in the European Union and similar regulatory shifts worldwide have catalyzed the development of advanced non-animal methods (NAMs) for skin sensitization assessment [2]. The complex immunobiology of allergic contact dermatitis (ACD)—a T cell-mediated hypersensitivity reaction—requires models that transcend traditional two-dimensional assays [2]. While the Adverse Outcome Pathway (AOP) for skin sensitization provides a structured framework for understanding key biological events, existing OECD-approved tests typically address only isolated key events [34]. This limitation has driven innovation toward three-dimensional immunocompetent skin models that incorporate key immune players—Langerhans cells (LCs) and T-lymphocytes—to better mimic the native human immune response in skin [34] [35]. These advanced constructs represent a paradigm shift, enabling simultaneous assessment of multiple AOP key events within a physiologically relevant architecture that includes a stratified epidermis and dermal compartment [34]. This guide compares the performance of these emerging complex models with established alternatives, providing researchers with experimental data and protocols to inform model selection for regulatory testing and mechanistic studies.

Model Comparison: Capabilities and Performance Metrics

Comparative Analysis of Skin Sensitization Testing Platforms

Table 1: Comparison of skin sensitization testing methods and their capabilities.

Model Type	Key Features	AOP Key Events Addressed	Sensitization Potency Discrimination	Throughput	Physiological Relevance
ImmuSkin-MT (Hair follicle-derived)	3D structure with MoLCs and CD4+ T-cells; transwell system	KE3 (DC activation) and KE4 (T-cell proliferation)	Differentiates extreme, moderate, and weak sensitizers [34]	Medium	High (incorporates multiple immune cell types and native tissue architecture) [34]
EpiSensA (OECD TG 442D)	Reconstructed human epidermis (RHE)	KE2 (Keratinocyte activation)	Limited	High	Medium (human tissue but no integrated immune components) [2]
Loose-fit Coculture-based Sensitization Assay (LCSA)	Co-culture of keratinocytes and PBMCs	KE2 or KE3 (depending on endpoints)	Moderate	Medium	Medium (cellular crosstalk but no 3D structure) [34]
Direct Peptide Reactivity Assay (DPRA) (OECD TG 442C)	In chemico peptide binding	KE1 (Molecular initiating event - covalent binding)	No	High	Low (non-biological system) [2] [34]
h-CLAT (OECD TG 442E)	Monocyte-derived dendritic cell line	KE3 (DC activation - CD86/CD54 expression)	Limited	High	Low (single cell type in 2D culture) [34]

Quantitative Performance Data of Advanced Models

Table 2: Experimental outcomes of the ImmuSkin-MT model when exposed to sensitizers of varying potency. [34]

Sensitizer Potency Category	CD86 Upregulation on MoLCs	CD4+ T-cell Proliferation	Cytokine Secretion Profile	Prediction Accuracy
Extreme	Strong increase (>2-fold)	Significant expansion	Pro-inflammatory cytokine surge	Correctly identified
Moderate	Moderate increase (1.5-2-fold)	Measurable expansion	Detectable inflammatory signals	Correctly identified
Weak	Mild but detectable increase	Low but significant proliferation	Baseline to mild elevation	Correctly identified
Non-sensitizer	No significant change	No expansion	No inflammatory profile	Correctly identified

Experimental Protocols for Advanced Immunocompetent Models

Protocol 1: Generation of the ImmuSkin-MT Model

The ImmuSkin-MT model represents a significant technical advancement by incorporating both MoLCs and T-cells within a hair follicle-derived skin equivalent [34].

Cell Sourcing and Isolation:

• Hair follicle-derived keratinocytes (HFDKs) and fibroblasts (HFDFs): Isolate from 30-35 plucked human hair follicles (donors aged 25-35) via enzymatic digestion (0.05% trypsin/0.02% EDTA) using a detach-time-selection strategy (5 minutes for HFDFs, 10 minutes for HFDKs) [34].
• Monocyte-derived Langerhans Cells (MoLCs): Isolate monocytes from anonymous donor buffy coats using adherence method. Differentiate in DC generation medium supplemented with 100 ng/mL GM-CSF, 20 ng/mL TGF-β, and 20 ng/mL IL-4 for 6 days, replacing half the medium every 3 days. On day 6, isolate CD1a+ cells using magnetic bead separation [34].
• CD4+ Naïve T-lymphocytes: Isolate from PBMCs using naive CD4+ T-Cell Isolation Kit II according to manufacturer specifications [34].

3D Model Assembly:

• Culture 10-12 hair follicles on poly-D-Lysine-coated transwell inserts containing postmitotic 3T3-J2 fibroblasts in outer root sheath medium until 80% confluence is reached.
• Seed MoLCs beneath the dermal layer of the established skin equivalent.
• Place CD4+ T-cells in the lower chamber of the transwell system to allow migration-mediated interaction [34].

Exposure and Analysis:

• Apply test chemicals for 24-48 hours.
• Assess MoLC migration and CD86 expression via flow cytometry.
• Measure T-cell proliferation using CFSE dilution or similar method.
• Quantify cytokine secretion (IL-1β, IL-18, etc.) via ELISA [34].

Protocol 2: Assessing Key Events 3 and 4 in a Single Model

This integrated protocol enables simultaneous evaluation of dendritic cell activation and T-cell proliferation, addressing a critical gap in existing test methods [34].

Dendritic Cell Activation (Key Event 3):

• Endpoint: CD86 surface marker expression on MoLCs.
• Method: Flow cytometric analysis after 24h chemical exposure.
• Controls: Use known sensitizers (e.g., DNCB) and non-sensitizers for benchmark responses.
• Threshold: ≥1.5-fold increase in CD86 MFI (mean fluorescence intensity) relative to vehicle control indicates positive activation [34].

T-cell Proliferation (Key Event 4):

• Endpoint: CD4+ T-cell proliferation in response to MoLC-presented antigen.
• Method: CFSE dilution assay or 3H-thymidine incorporation after 5-7 days of co-culture.
• Analysis: Percentage of proliferating T-cells in CD4+ population.
• Threshold: ≥2-fold increase in proliferation compared to vehicle control indicates sensitization potential [34].

Additional Endpoints:

• Measure secretion of IL-1β, IL-18, and other pro-inflammatory cytokines via multiplex ELISA.
• Assess morphology and viability through H&E staining and MTT assay.
• Evaluate barrier integrity through TEER measurement for penetration studies [2] [34].

Visualizing the Experimental Workflow and AOP Framework

Experimental Workflow for 3D Immunocompetent Skin Model Generation

AOP for Skin Sensitization and Model Coverage

The Scientist's Toolkit: Essential Research Reagents

Table 3: Key reagents and materials for constructing 3D immunocompetent skin models. [34]

Reagent/Material	Specification/Purpose	Function in Model Development
Hair Follicles	30-35 follicles from donors (25-35 years)	Primary cell source for keratinocytes and fibroblasts with improved differentiation capacity [34]
Transwell Plates	Corning Costar with permeable membranes	Physical support for 3D culture and compartmentalization of immune cells [34]
Poly-D-Lysine	Hydrobromide solution coating	Enhances cell attachment to membrane surfaces [34]
Cytokine Cocktail	GM-CSF (100 ng/mL), TGF-β (20 ng/mL), IL-4 (20 ng/mL)	Differentiation of monocytes into Langerhans-like cells (MoLCs) [34]
CD1a MicroBeads	Magnetic separation beads (Miltenyi Biotech)	Isolation of purified MoLC population after differentiation [34]
Naive CD4+ T-Cell Isolation Kit	Magnetic bead-based separation (Miltenyi Biotech)	Isolation of pure population of naive CD4+ T-lymphocytes [34]
Outer Root Sheath Medium	Specialized formulation with cholera toxin, EGF, insulin, adenine	Supports growth and differentiation of hair follicle-derived keratinocytes [34]
Flow Cytometry Antibodies	Anti-CD86, Anti-CD1a, Anti-CD4	Detection of cell surface markers for activation and proliferation assessment [34]
NAPQI	Acetimidoquinone | High Purity Research Chemical	Acetimidoquinone for organic synthesis & biochemical research. High-purity, For Research Use Only. Not for human or veterinary use.
Aklaviketone	Aklaviketone | High-Purity Research Compound	Aklaviketone: A key intermediate for anthracycline research. For Research Use Only. Not for human or veterinary diagnostic or therapeutic use.

Advanced 3D immunocompetent skin models incorporating Langerhans cells and T-lymphocytes represent a transformative approach in skin sensitization testing. The ImmuSkin-MT model and similar constructs demonstrate that simultaneous assessment of multiple AOP key events (particularly KE3 and KE4) is achievable, enabling more accurate hazard identification and potency discrimination compared to single-event assays [34]. While challenges remain in standardization and reducing donor variability, these models offer unprecedented physiological relevance for studying the complex cellular crosstalk in allergic contact dermatitis [2] [34]. The integration of additional advancements—such as organ-on-a-chip technologies, microbiome components, and real-time monitoring systems—will further enhance the predictive power of these platforms [2] [36]. For researchers and drug development professionals, these models provide not only a regulatory testing tool but also a powerful platform for mechanistic studies and the development of targeted therapeutics for inflammatory skin conditions [35].

The global ban on animal testing for cosmetics has catalyzed a paradigm shift in safety assessment, driving the development of advanced non-animal methods (NAMs) for evaluating skin sensitization potential [2] [37]. Within this evolving landscape, in silico tools, particularly machine learning (ML) and R-based models, have emerged as powerful computational approaches for hazard identification and potency assessment. These models leverage the Adverse Outcome Pathway (AOP) framework for skin sensitization, which describes a sequence of measurable key events from initial covalent binding to proteins (Molecular Initiating Event) through keratinocyte activation, dendritic cell activation, and ultimately T-cell proliferation [2] [37]. This guide provides an objective comparison of the performance and application of contemporary in silico models, offering researchers a critical evaluation of available tools for integrating computational toxicology into next-generation risk assessment (NGRA) paradigms.

Performance Comparison of In Silico Predictive Models

The following section provides a detailed, data-driven comparison of the performance, characteristics, and applications of leading in silico models for skin sensitization assessment.

Quantitative Performance Metrics

Table 1: Comparative Performance of Key In Silico Models for Skin Sensitization Prediction

Model Name	Model Type	Prediction Target	Key Input Data	Accuracy (r²)	Error (RMS)	Regulatory Status
R-based ANN Model [38]	Artificial Neural Network	LLNA EC3 Value (Potency)	DPRA, KeratinoSens, h-CLAT, Structural Alerts	0.889	0.434	Research Use
SARA-ICE [39]	Bayesian Statistical Model	Human ED01 (Potency)	Any combination of in vivo & in vitro data (DPRA, kDPRA, KeratinoSens, h-CLAT, U-SENS)	Not Specified	Not Specified	Incorporates OECD TG 497 DAs
Previous QwikNet ANN Model [38]	Artificial Neural Network	LLNA EC3 Value (Potency)	SH test, h-CLAT, ARE data	0.857	0.429	Research Use (Paid Software)
ITS-based Models [38]	Integrated Testing Strategy (e.g., Bayesian Network)	LLNA Potency Category	DPRA, KeratinoSens, TIMES-SS, QSAR Toolbox	Varies by approach	Varies by approach	OECD Guideline 497

Model Characteristics and Applications

Table 2: Characteristics and Practical Application of In Silico Models

Model Name	Key Advantages	Limitations	Ideal Use Case
R-based ANN Model [38]	High accuracy for potency prediction; uses free, open-source R software; handles complex non-linear relationships.	Requires specific in vitro input data; model training dataset of 134 compounds.	Quantitative risk assessment for cosmetic ingredients where LLNA EC3 values are needed.
SARA-ICE [39]	Predicts human-relevant point-of-departure (ED01); flexible input requirements; integrates with OECD TG 497.	Bayesian model may be less intuitive than other ML approaches.	Next-generation risk assessment (NGRA) for human safety evaluation without animal data.
Previous QwikNet Model [38]	Validated high performance; direct predictor of LLNA threshold.	Built on paid, proprietary software (QwikNet).	Historical comparison and validation of new, open-source models.
ANN with Structural Alerts [38]	Incorporates chemical structure-based alerts to improve prediction.	Complexity increases with additional input parameters.	Screening of new chemical entities with limited test data.

Experimental Protocols for Model Development and Validation

This section details the key methodological workflows for developing and validating the in silico models discussed, providing a roadmap for their implementation and critical evaluation.

Protocol for Developing an R-Based ANN Model

The development of the open-source R-based Artificial Neural Network (ANN) model follows a structured pipeline to ensure predictive robustness and regulatory relevance [38].

• Chemical Selection and Data Curation

  • Source a high-quality dataset of chemicals with reliable reference in vivo data, specifically LLNA EC3 values, from curated databases [38].
  • Ensure availability of corresponding in vitro and in chemico data for the same chemicals, specifically from DPRA (Key Event 1), KeratinoSens (Key Event 2), and h-CLAT (Key Event 3) assays [38].

• Data Preprocessing and Feature Selection

  • Normalize all input data (EC3 values and assay results) to a consistent scale to facilitate model training.
  • Optionally, incorporate in silico structural alert parameters as additional descriptors to capture intrinsic chemical reactivity [38].

• Model Architecture and Training

  • Implement a multi-layer perceptron (MLP) ANN architecture within the R statistical environment using packages such as nnet or neuralnet.
  • Partition the dataset into training and validation sets (e.g., 134 compounds for training, 28 for external testing) [38].
  • Train the model to establish complex, non-linear relationships between the input assay data and the continuous output variable (LLNA EC3 value).

• Model Validation and Performance Assessment

  • Validate the model's predictive accuracy on the held-out test set of chemicals.
  • Calculate performance metrics, including coefficient of determination (r²) and Root Mean Square (RMS) error, and compare them against existing models (e.g., QwikNet) and a defined accuracy threshold (e.g., ≥80%) [38].

Protocol for a Defined Approach (DA) According to OECD TG 497

The OECD Test Guideline 497 provides a framework for using Defined Approaches (DAs) for skin sensitization hazard classification, which often integrate in silico components [38] [39].

• Input Data Generation

  • Obtain data from at least two of the three AOP Key Events (KE1, KE2, KE3) from validated non-animal test methods.
  • Key Event 1 (Protein Binding): Perform the DPRA or its kinetic version, kDPRA [2] [38].
  • Key Event 2 (Keratinocyte Response): Conduct the KeratinoSens assay or the EpiSensA test using reconstructed human epidermis [2] [38].
  • Key Event 3 (Dendritic Cell Activation): Use the h-CLAT or U-SENS assay [38].

• Data Integration via a Prediction Model

  • Input the generated data into a DA prediction model, such as a Bayesian Network, an Integrated Testing Strategy (ITS), or the SARA-ICE model [38] [39].
  • These models use predefined rules or statistical algorithms to integrate the results from the individual Key Events.

• Hazard Classification and Potency Assessment

  • The DA model provides a binary prediction (sensitizer vs. non-sensitizer) corresponding to the UN GHS sub-categories.
  • For more advanced models like SARA-ICE, a point-of-departure (ED01) for human-relevant quantitative risk assessment can be derived [39].

The Adverse Outcome Pathway (AOP) and Signaling Pathways

The development and interpretation of in silico models are grounded in the AOP for skin sensitization. The following diagram illustrates the biological sequence of key events and the corresponding test methods that inform predictive models.

The Scientist's Toolkit: Essential Research Reagents and Models

This section catalogs key reagents, computational tools, and biological models essential for conducting research and testing in the field of in silico skin sensitization.

Table 3: Essential Research Reagents and Tools for In Silico Skin Sensitization Model Development

Category	Item / Solution	Critical Function & Application
In Chemico Assays	DPRA / kDPRA [38]	Measures hapten reactivity with synthetic peptides (Cysteine, Lysine); directly addresses AOP Key Event 1 (Molecular Initiating Event).
	Amino Acid Derivative Reactivity Assay (ADRA) [2] [38]	Alternative reactivity assay for KE1; adopted in OECD TG 442C.
In Vitro Assays (KE2)	KeratinoSens / LuSens [38]	Reporter gene assays measuring Nrf2-dependent gene activation in keratinocytes for KE2 assessment.
	Reconstructed Human Epidermis (RHE) Models [2]	3D tissue models (e.g., EpiSensA) used to measure IL-18 release; provide a more physiologically relevant platform for KE2 and beyond.
In Vitro Assays (KE3)	h-CLAT / U-SENS [38]	Measures surface marker expression (CD86, CD54) on dendritic-like cell lines to assess dendritic cell activation (KE3).
	GARDskin [38]	Genomic biomarker-based assay for KE3; adopted in OECD TG 442E.
Computational Tools & Data	R Statistical Software [38]	Open-source platform for developing and deploying custom predictive models (e.g., ANN).
	SARA-ICE Web Tool [39]	Publicly available Bayesian model for predicting human-relevant point-of-departure (ED01).
	OECD QSAR Toolbox [38]	Software for grouping chemicals and filling data gaps via read-across, used in Integrated Testing Strategies.
	TIMES-SS Platform [38]	In silico expert system that predicts sensitization potency by integrating metabolic activation and reactivity.
Reference Data	LLNA EC3 Value Database [38]	Curated dataset of historical murine Local Lymph Node Assay results; serves as a benchmark for training and validating new prediction models.
Coelenterazine hcp	Coelenterazine hcp, CAS:123437-32-1, MF:C25H25N3O2, MW:399.5 g/mol	Chemical Reagent
Solvent Yellow 16	Solvent Yellow 16 | High-Purity Research Dye	Solvent Yellow 16 is a lipophilic azo dye for industrial & materials science research. For Research Use Only. Not for human or veterinary use.

Navigating Challenges: Overcoming Hurdles in Model Performance and Applicability

The safety assessment of Botanical and Natural Substances (BNS) and Unknown or Variable Composition, Complex Reaction Products or Biological Materials (UVCBs) presents a significant challenge in modern toxicology. These substances, which can comprise over 20% of chemical registrations in Europe, defy conventional testing approaches designed for single chemical entities [40]. The inherent complexity of these materials—derived from variable plant compositions, manufacturing processes, or finished product formulations—places them outside the standard applicability domains of many validated testing methods [40] [41]. Within the cosmetic, personal care, and chemical industries, this creates a critical need for robust testing strategies that can accurately evaluate skin sensitization potential while aligning with the global regulatory trend toward animal-free safety assessment [2] [37].

The fundamental challenge lies in the chemical complexity of these substances. UVCBs may contain hundreds to millions of isomeric chemical constituents, while botanical extracts exhibit natural variation based on growth location, conditions, and harvest times [42] [41]. Furthermore, finished products represent complex mixtures of multiple ingredients, creating potential interactions that cannot be captured by testing individual components in isolation. This complexity necessitates innovative approaches that move beyond traditional single-chemical testing paradigms toward integrated testing strategies and Next Generation Risk Assessment (NGRA) frameworks [40] [41].

The Adverse Outcome Pathway Framework for Complex Mixtures

The Adverse Outcome Pathway (AOP) for skin sensitization provides a conceptual framework for organizing biological events leading to allergic contact dermatitis, comprising four key events (KEs): covalent binding to skin proteins (KE1), keratinocyte activation (KE2), dendritic cell activation (KE3), and T-cell proliferation (KE4) [2] [38]. While this framework was developed using single chemicals, it nonetheless provides a valuable structure for investigating complex mixtures by identifying which key events in the sensitization process are triggered by mixture components.

Adaptation of AOP for Complex Mixtures

For complex mixtures, the AOP framework must be applied with consideration of several unique factors:

• Component Interactions: Individual chemicals in a mixture may interact, potentially enhancing or inhibiting each other's effects on different key events
• Bioavailability: The vehicle or formulation matrix may affect the release and skin penetration of individual sensitizers
• Cumulative Effects: Multiple weak sensitizers acting on the same key event may produce a combined effect exceeding their individual contributions
• Novel Sensitizers: Previously uncharacterized components in UVCBs may trigger key events through novel mechanisms

The following diagram illustrates how complex mixtures interact with the established AOP for skin sensitization, highlighting points where mixture complexity introduces additional considerations:

Testing Strategies by Product Category

Botanicals and Natural Substances (BNS)

BNS present unique testing challenges due to their complex, variable composition and natural origins. A weight of evidence (WoE) approach that integrates multiple data sources has shown promise for these materials [41]. Case studies with 14 plant species demonstrated successful classification of sensitization potential by combining:

• History of human use and clinical patch test data
• Compositional analysis for known sensitizers
• New Approach Methodology (NAM) data from in chemico and in vitro tests
• Historical animal data where available

For BNS with sufficient data, a next generation risk assessment (NGRA) framework can be applied using a tiered approach that begins with exposure-based waiving. When exposure exceeds defined thresholds, a comprehensive WoE assessment is triggered [41].

UVCB Substances

Petroleum substances represent a well-studied category of UVCBs. Research on 141 petroleum substance extracts demonstrated that dose-response transcriptomic profiling in human induced pluripotent stem cell (iPSC)-derived hepatocytes, cardiomyocytes, neurons, and endothelial cells can successfully group these UVCBs by manufacturing class [42]. The transcriptional activity showed strong correlation with polycyclic aromatic compound (PAC) concentration, particularly in iPSC-derived hepatocytes, providing a mechanistic basis for biological responses [42].

For UVCBs, successful testing strategies often combine:

• Chemical characterization to identify dominant constituents
• Bioactivity profiling across multiple cell types
• Transcriptomic analysis to understand mechanistic pathways
• Dose-response relationships to establish potency rankings

Finished Products and Formulations

Finished products represent the most complex category due to the presence of multiple ingredients in a formulated matrix. The Skin Sensitization Prediction Model (SSPM) represents an innovative approach that leverages historical human repeat insult patch test (HRIPT) data from 1,274 unique product formulations containing 1,226 ingredients tested on 203,640 subjects [43]. This data-driven analytics approach predicts sensitization risk based on ingredient combinations and their historical performance.

For finished products, key considerations include:

• Ingredient interactions that may enhance or inhibit sensitization potential
• Vehicle effects on dermal penetration and bioavailability
• Cumulative exposure to multiple sensitizers in one product
• Matrix effects on assay performance in in vitro systems

Methodologies and Assay Performance

The following experimental protocols represent key methodologies adapted for testing complex mixtures:

Direct Peptide Reactivity Assay (DPRA) for Botanicals

Protocol Adaptation: The standard DPRA (OECD TG 442C) measures haptenation potential by quantifying depletion of synthetic peptides containing cysteine or lysine. For botanical extracts, modifications include:

• Sample Preparation: Prepare botanical extracts in DMSO or acetone at 100 mg/mL, followed by serial dilution in buffer
• Controls: Include method blanks with extraction solvent only to account for background reactivity
• Incubation: Mix 100 μL of peptide solution (0.667 mM) with 100 μL of test sample, incubate at 25°C for 24 hours
• Analysis: Use HPLC with UV detection to measure peptide depletion
• Interpretation: Cysteine peptide depletion >6.38% indicates sensitization potential, but positive results should be interpreted in context of other data [41]

Limitations: Botanical components may interfere with HPLC detection, and colored extracts can quench fluorescence in some assay variants. Negative results may be inconclusive due to potential assay interference [40].

Transcriptomic Profiling for UVCB Categorization

Protocol Adaptation: This approach uses gene expression changes to categorize UVCBs and understand mechanistic basis:

• Cell Models: Human iPSC-derived hepatocytes, cardiomyocytes, neurons, endothelial cells, plus MCF7 and A375 cell lines
• Dosing: Petroleum substance extracts prepared in DMSO using ASTM standard procedures to concentrate biologically active fractions
• Exposure: Dose-response treatment for 24 hours, with minimum of 3 concentrations and vehicle controls
• RNA Extraction: Isolate total RNA using column-based methods with DNase treatment
• Transcriptomic Analysis: Use targeted RNA sequencing or microarray platforms focused on stress response pathways
• Data Analysis: Apply pathway enrichment analysis and unsupervised clustering to identify patterns corresponding to manufacturing categories [42]

Applications: This approach successfully distinguished petroleum substances by manufacturing class and correlated transcriptional activity with PAC content [42].

Reconstructed Human Epidermis (RHE) Models

Protocol Adaptation: RHE models (EpiSensA, OECD TG 442D) provide a more physiologically relevant platform:

• Model Preparation: Use commercially available RHE models (EpiDerm, EpiSkin, SkinEthic)
• Dosing: Apply complex mixtures directly to the epidermal surface, using appropriate vehicles
• Endpoint Measurement: Assess cell viability (MTT assay) and specific biomarkers (IL-18, IL-1α)
• Biomarker Analysis: Collect culture media for cytokine measurement via ELISA
• Data Interpretation: Compare biomarker expression patterns to known sensitizers and irritants [2] [37]

Advantages: RHE models maintain barrier function and keratinocyte differentiation, providing a more realistic exposure scenario for topically applied mixtures.

Comparative Performance Data

Table 1: Performance of Testing Methods Across Complex Substance Categories

Method	Botanicals	UVCBs	Finished Products	Key Limitations
DPRA	Limited applicability; interference from colored compounds	Variable performance; depends on dominant constituents	Not recommended; matrix interference	High false negatives with complex mixtures [40]
RHE Models (EpiSensA)	Good for extracts; maintains barrier function	Shows promise; physiologically relevant environment	Suitable with formulation adjustments	Limited metabolic capacity; cost [2] [37]
Transcriptomics	Emerging application; identifies mechanistic patterns	Strong performance for categorization and potency ranking	Limited data; matrix effects may interfere	Complex data interpretation; standardization needed [42]
GARDskin	Limited published data	Limited published data	Limited published data	Requires specialized expertise [44]
Integrated Approaches (WoE)	Recommended strategy	Recommended strategy	Recommended strategy	Resource-intensive; subjective elements [41]

Table 2: Case Study Results for Botanical Substances Using Weight of Evidence Approach

Botanical Substance	Human Data	Animal Data	NAM Results	Compositional Analysis	Overall Classification
Poison Ivy	Strong clinical evidence	Positive in animal studies	Positive in multiple NAMs	Known sensitizers (urushiol)	Strong sensitizer [41]
Feverfew	Case reports	Limited data	Variable results	Sesquiterpene lactones	Weak-moderate sensitizer [41]
Green Tea Extract	Limited evidence	Negative animal data	Negative in NAMs	Catechins (not reactive)	Non-sensitizer [41]
Compositae Mix	Clinical evidence	Positive data	Positive in adapted NAMs	Sesquiterpene lactones	Strong sensitizer [41]

The Scientist's Toolkit: Essential Research Reagents

Table 3: Key Research Reagent Solutions for Complex Mixture Testing

Reagent/Model	Function	Application Notes
Reconstructed Human Epidermis	3D tissue model for topical application	Maintains barrier function; appropriate for extracts and finished products [2]
iPSC-Derived Cells	Physiologically relevant human cells	Hepatocytes show particular sensitivity for UVCB testing [42]
Synthetic Peptides	Measure haptenation potential (KE1)	Cysteine and lysine peptides for DPRA; may require adaptation for complex mixtures [38]
Cytokine ELISA Kits	Quantify inflammatory mediators	IL-18 and IL-1α for keratinocyte activation (KE2) [37]
Cell Line Activation Tests	Assess dendritic cell activation (KE3)	h-CLAT, U-SENS; may require extraction procedures for complex mixtures [40]
Gene Expression Panels	Pathway-focused transcriptomics	Targeted panels for stress response and immunomodulatory genes [42]
ZINC acetate	Zinc Acetate | High-Purity Reagent | RUO	High-purity Zinc Acetate for cell culture, biochemistry & catalysis research. For Research Use Only. Not for human or veterinary use.
Epiboxidine	Epiboxidine, CAS:188895-96-7, MF:C10H14N2O, MW:178.23 g/mol	Chemical Reagent

Integrated Testing Strategies and Decision Frameworks

For complex mixtures, no single method provides comprehensive assessment. Successful evaluation requires integrated testing strategies that combine multiple data streams. The following diagram illustrates a recommended workflow for testing complex mixtures:

Weight of Evidence Framework

A WoE framework for botanicals has been successfully demonstrated using 14 representative plant species [41]. This approach integrates:

• Human Data: Clinical patch test results, case reports, and human use history
• Animal Data: Historical LLNA, GPMT, or Buehler test results
• NAM Data: Results from adapted in chemico and in vitro methods
• Compositional Data: Presence of known sensitizers or reactive chemicals

Through expert judgment, these data streams are combined to reach conclusions regarding sensitization hazard and potency classification [41].

Defined Approaches for Regulatory Acceptance

For regulatory decision-making, defined approaches (DAs) that integrate multiple NAMs according to fixed rules provide transparency and reproducibility. The OECD Guideline 497 describes several DAs that combine:

• KE1 Assays: DPRA, kDPRA, or ADRA
• KE2 Assays: KeratinoSens, LuSens, or IL-8 Luc assay
• KE3 Assays: h-CLAT, U-SENS, or GARDskin
• In Silico Tools: DEREK, OECD QSAR Toolbox

These DAs have shown good accuracy for single chemicals, but require further validation for complex mixtures [38].

Testing complex mixtures for skin sensitization potential requires a paradigm shift from single-chemical approaches to integrated strategies that account for complexity, variability, and potential interactions. The methods and frameworks discussed here provide a foundation for assessing botanicals, UVCBs, and finished products without animal testing.

Key success factors include:

• Method Adaptation to address technical challenges like solubility and matrix effects
• Integrated Strategies that combine multiple data streams
• WoE Frameworks for transparent decision-making
• Mechanistic Understanding through transcriptomics and pathway analysis

As research continues, emerging technologies like organ-on-a-chip and microfluidic systems with integrated immune components promise to better recapitulate the complexity of skin sensitization, particularly for challenging substance categories [2]. Additionally, standardized testing frameworks specifically validated for complex mixtures will enhance regulatory acceptance and improve safety assessment for these materials.

In the evolving landscape of predictive toxicology, particularly for assessing immune-mediated responses like skin sensitization, the traditional binary classification of substances as simply "positive" or "negative" is increasingly recognized as insufficient. This recognition has led to the formalization of the Borderline Range (BR) concept—a defined zone around a test method's classification threshold where results are considered scientifically inconclusive due to inherent biological and technical variability [45]. The implementation of BRs represents a significant advancement in the interpretation of New Approach Methodologies (NAMs), providing a more nuanced and transparent framework for addressing data variability in regulatory decision-making.

The validation of in vitro skin sensitization models specifically benefits from this approach. Skin sensitization, a key endpoint in immune response research, follows a well-defined Adverse Outcome Pathway (AOP) involving molecular initiating events, keratinocyte responses, dendritic cell activation, and T-cell proliferation [46] [47]. By quantifying the uncertainty around classification thresholds, BRs enhance the reliability of integrated testing strategies that combine multiple NAMs, ultimately supporting more confident safety assessments for chemicals and drug candidates while reducing reliance on traditional animal testing [46].

Quantifying Borderline Ranges: Calculation Methods and Applications

Statistical Foundation of Borderline Ranges

The core principle behind establishing a Borderline Range is the statistical quantification of a test method's variability around its decision threshold. Rather than treating a single cutoff value as absolute, the BR defines an interval within which the distinction between positive and negative outcomes becomes uncertain. This approach acknowledges that all biological test systems exhibit inherent variability that can influence results near critical thresholds.

For skin sensitization methods, the log pooled median absolute deviation (MAD) method has been employed to calculate these ranges objectively [46]. This robust statistical approach characterizes the dispersion of test data around the median, making it particularly suitable for establishing BRs around a classification cutoff. The method involves analyzing historical validation data to determine the typical variability observed for each test method, then using this variability measure to set the upper and lower bounds of the borderline range.

Implemented Borderline Ranges in Key Assays

Table 1: Borderline Ranges in Prominent Skin Sensitization Assays

Test Method	Measured Parameter	Classification Cut-off	Borderline Range	Implications
U-SENS [46]	CD86 Stimulation Index (SI)	SI > 150%	128% ≤ SI ≤ 176%	Results in BR require confirmatory testing
h-CLAT [45]	Relative Fluorescence Intensity (RFI)	RFI ≥ 150%	135% ≤ RFI ≤ 165%	Inconclusive outcomes for values within BR
DPRA [45]	Peptide Depletion	6.38% (Cys) & 22.62% (Lys)	Defined variability bounds	Affects accuracy of potency subcategorization
LLNA [45]	EC3 Value	Variable based on chemical	Defined variability bounds	Impacts GHS potency subcategorization (1A vs. 1B)

The implementation of Borderline Ranges has demonstrated significant practical utility. In the U-SENS assay, which measures dendritic cell activation by assessing CD86 expression, applying the defined BR (128% ≤ SI ≤ 176%) changed the predictions for 35 of 191 chemicals in the OECD database, highlighting how substantial proportions of tested substances may fall into this uncertain zone [46]. Similarly, studies quantifying BRs in the DPRA, LuSens, and h-CLAT methods found that between 6% and 28% of tested substances were classified as borderline depending on the method [45].

Experimental Protocols for Borderline Range Determination

U-SENS Assay Methodology and BR Implementation

The U-SENS protocol represents a standardized approach for evaluating the activation of key event 3 in the skin sensitization AOP—dendritic cell activation. The experimental workflow follows a structured process:

Cell Culture and Preparation: The assay utilizes U937 human histiocytic lymphoma cells, which are maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 1% penicillin-streptomycin at 37°C in a 5% CO₂ atmosphere. Cells are passaged regularly to maintain logarithmic growth.

Chemical Treatment: Test chemicals are dissolved in appropriate solvents (typically DMSO or water) and serially diluted to achieve multiple concentrations. Cells are exposed to these concentrations for 48 hours, with viability assessed using the MTT assay to ensure testing occurs under non-cytotoxic conditions.

Flow Cytometric Analysis: Following exposure, cells are stained with fluorescently-labeled anti-CD86 antibodies and analyzed by flow cytometry. The CD86 expression is quantified as a Stimulation Index (SI) relative to solvent controls.

Borderline Range Application: The raw SI values are interpreted using the established classification scheme:

• Negative: SI < 128%
• Borderline: 128% ≤ SI ≤ 176%
• Positive: SI > 176% [46]

This tripartite classification system explicitly acknowledges the uncertainty in results falling near the historical 150% cutoff, providing more transparent interpretation of the assay data.

Integrated Testing Strategies Incorporating Borderline Ranges

The "2-out-of-3" (2o3) Defined Approach (DA) for skin sensitization assessment sequentially integrates results from three NAMs addressing different key events in the AOP [46]. When Borderline Ranges are implemented for the constituent tests, the interpretation strategy becomes more sophisticated:

Key Event 1 (Molecular Interaction): Typically assessed via DPRA or kinetic DPRA, measuring peptide reactivity that mimics hapten-protein binding.

Key Event 2 (Keratinocyte Response): Often evaluated using the KeratinoSens assay, assessing gene expression associated with antioxidant responses.

Key Event 3 (Dendritic Cell Activation): Measured by h-CLAT or U-SENS, quantifying cell surface marker expression (CD86, CD54) indicative of dendritic cell maturation.

When any individual test result falls within a predefined Borderline Range, the integrated assessment acknowledges this uncertainty. Research demonstrates that applying BR thresholds in the 2o3 DA improved balanced accuracy from 71% to 77% against LLNA data (n=142) and from 77% to 88% against human data (n=55) [46]. This enhancement in predictive performance underscores the value of transparently addressing methodological variability rather than ignoring it.

The Researcher's Toolkit: Essential Reagents and Materials

Table 2: Key Research Reagents for Skin Sensitization Assessment

Reagent/Material	Function in Assay	Specific Application Example
U937 Cell Line	Model dendritic cell system	U-SENS assay for KE3 assessment [46]
Anti-CD86 Antibodies	Detection of cell surface activation marker	Flow cytometric analysis in h-CLAT/U-SENS [46]
Cysteine/Lysine Peptides	Measurement of hapten-protein binding	DPRA for KE1 assessment [47]
HPLC System with UV Detection	Quantification of peptide depletion	Analytical component of DPRA [47]
Recombinant Cytokines	Positive control stimulation	Assay validation and quality control
MTT Reagent	Assessment of cell viability	Cytotoxicity determination in cell-based assays

The selection of appropriate reagents is critical for generating reliable data in skin sensitization assessment. The U937 cell line serves as a standardized model for dendritic cell behavior, providing a consistent biological system for evaluating potential sensitizers [46]. Detection reagents such as anti-CD86 antibodies enable quantification of cell surface markers indicative of dendritic cell activation, while synthetic peptides facilitate measurement of the initial molecular interaction between chemicals and skin proteins [47]. Analytical instrumentation including flow cytometers and HPLC systems provide the technical platform for objective endpoint measurement, and viability assessment reagents ensure tests are conducted under non-cytotoxic conditions to avoid artifactual results.

Implications for Drug Development and Regulatory Science

The formalization of Borderline Ranges represents a significant shift in how variability is addressed in toxicological assessment. For drug development professionals, this approach provides several important advantages:

Enhanced Decision Transparency: By explicitly identifying results falling within methodological uncertainty ranges, BRs support more informed risk assessments and resource allocation for follow-up testing [45].

Improved Predictive Accuracy: Incorporating BR thresholds in integrated testing strategies like the "2-out-of-3" DA has demonstrated measurable improvements in classification accuracy against both animal and human data [46].

Regulatory Adaptation: The inclusion of U-SENS and its borderline range thresholds in OECD Guideline 497 reflects growing regulatory acceptance of this nuanced approach to test interpretation [46].

Strategic Testing Optimization: When results fall within a borderline range, researchers can make conscious decisions about whether to conduct additional confirmatory testing, use alternative methods, or exercise scientific judgment in classification.

The growing adoption of BR concepts reflects a broader movement toward more sophisticated, probabilistic approaches in toxicological assessment that better represent the continuum of biological responses than traditional binary classification systems.

The global ban on animal testing for cosmetics, driven by the EU Cosmetics Regulation 1223/2009 and the principles of Replacement, Reduction, and Refinement (3Rs), has catalyzed the development of alternative non-animal methods (NAMs) for skin sensitization assessment [48] [2]. These methods are largely built upon the Adverse Outcome Pathway (AOP), which describes the mechanistic sequence of events from a chemical's initial contact with the skin to the clinical manifestation of allergic contact dermatitis (ACD) [48] [2]. However, significant limitations persist in accurately identifying and characterizing certain types of chemicals, namely prehaptens and prohaptens, within complex formulations. Prehaptens are chemicals that become sensitizing after activation by non-enzymatic processes (e.g., air oxidation), while prohaptens require enzymatic bioactivation to transform into a reactive state [49] [2].

This review objectively compares the performance of current testing strategies against these challenges. We synthesize experimental data to highlight where existing models succeed and where critical gaps in reliability remain, providing a crucial resource for researchers and safety assessors in pharmaceutical and cosmetic development.

The AOP Framework and Its Testing Strategies

The AOP for skin sensitization organizes the complex biological process into four Key Events (KEs), each associated with specific in chemico and in vitro testing methods [48] [2].

Table 1: The Adverse Outcome Pathway for Skin Sensitization and Associated Test Methods

Key Event (KE)	Biological Process	OECD Test Guideline (TG) & Method Names	Measured Endpoints
KE1: Molecular Initiating Event	Covalent binding of haptens to skin proteins.	TG 442C: DPRA, ADRA, kDPRA	Peptide depletion (%) [28] [2]
KE2: Keratinocyte Response	Keratinocyte activation & inflammatory response.	TG 442D: KeratinoSens, LuSens, IL-8 Luc, EpiSensA	Nrf2 pathway activation (e.g., Luciferase activity), IL-18 secretion [28] [2]
KE3: Dendritic Cell Activation	Dendritic cell maturation & migration.	TG 442E: h-CLAT, U-SENS, IL-8 Luc assay	Surface marker expression (CD54, CD86) [28] [50]
KE4: T-cell Proliferation	Proliferation of allergen-specific T-cells.	(Historically LLNA; in vivo)	T-cell clone expansion [48]

The following diagram illustrates this sequential pathway and the points where standard testing methods assess each key event.

The Core Challenge: Prehaptens and Prohaptens

The fundamental limitation of many in chemico and simple in vitro models is their inability to fully replicate the complex metabolic and oxidative processes that occur in human skin. This leads to a high risk of false negatives for prehaptens and prohaptens [49] [2].

• Prehaptens: These are chemicals that are not reactive in their original form but are transformed into potent sensitizers through non-enzymatic processes outside the body, most commonly air oxidation. For example, certain terpenes like limonene can form allergenic hydroperoxides upon exposure to air [51] [52].
• Prohaptens: These substances require enzymatic metabolism within the skin to convert into a reactive, sensitizing form (the hapten). This bioactivation typically involves enzymes like cytochromes P450, peroxidases, or other metabolic pathways [49] [2].

Table 2: Experimental Detection of Prohaptens and Prehaptens in Different Assays

Substance / Category	Example Compounds	LLNA (In Vivo) Result	Standard KE1 Assay (e.g., DPRA)	Modified Assay with Oxidation/Activation	Key Finding
Prohapten	2-methoxy-4-methylphenol (2M4MP)	Moderate sensitizer [49]	Likely false negative (non-reactive)	Positive in Peroxidase Peptide Reactivity Assay (PPRA) [49]	Activation by HRP/Hâ‚‚Oâ‚‚ generates reactive quinone methide.
Fragrance Prehaptens	Hydroperoxides of linalool, limonene	Sensitizers [51]	May fail if unoxidized	Positive after air oxidation; detected in clinical patch tests [51]	EU is mandating labeling of 56 additional such allergens [52].
General Prohaptens	Substances requiring CYP450 metabolism	Varies by substance	False negative (lacks bioactivation)	No standardized high-throughput in vitro method available	A major identified gap in current testing strategies [2].

Experimental Protocol: The Peroxidase Peptide Reactivity Assay (PPRA)

To address the prohapten challenge, researchers have developed modified versions of KE1 assays. The PPRA is a key experimental protocol designed to identify prohaptens that require peroxidase-mediated activation [49].

• Activation Step: The test chemical (e.g., 2M4MP) is incubated with Horseradish Peroxidase (HRP) and hydrogen peroxide (Hâ‚‚Oâ‚‚) in a suitable buffer (e.g., phosphate buffer, pH 7.0-7.4) for a defined period (e.g., 1-2 hours) at 37°C. This step mimics enzymatic oxidation in the skin.
• Reactivity Step: A synthetic peptide containing a nucleophilic amino acid (e.g., cysteine or lysine) is added to the reaction mixture. The reactive intermediate (e.g., a quinone methide) generated in step 1 covalently binds to the peptide.
• Detection and Quantification: Peptide depletion is measured using high-performance liquid chromatography (HPLC) with ultraviolet (UV) or mass spectrometry (MS) detection. The percentage of peptide depletion correlates with the sensitizing potency of the activated prohapten [49].
• Solubility Challenges: For hydrophobic chemicals, the assay can be adapted to run in a microemulsion system to maintain both enzyme activity and chemical solubility [49].

The Formulation Effect: The Complex Matrix Problem

Another critical limitation is the "formulation effect," where the sensitization potential of an ingredient can be altered—masked, enhanced, or mitigated—when it is part of a complex mixture like a final cosmetic or pharmaceutical product [28] [37]. Simple assays using single chemicals in buffer solutions may not predict this behavior.

Table 3: Limitations of Models in Assessing Formulation Effects

Testing Model	Typical Application	Limitations with Formulations	Supporting Experimental Evidence
In chemico assays (KE1)	Pure single chemicals in buffer [49]	Cannot account for ingredient interactions, partitioning, or bioavailability in a matrix.	Data generated on pure substances may not reflect reactivity in a complex cream or lotion.
Cell-based assays (KE2/KE3)	Single chemicals in culture medium [28] [50]	Surfactants, emulsifiers, or preservatives in a formulation can be cytotoxic at testing concentrations, interfering with readouts.	In a study on bacteriocins, cytotoxicity had to be ruled out before h-CLAT could be performed reliably [50].
Reconstructed Human Epidermis (RHE)	Skin irritation and corrosion [53]	Limited immunocompetence; lacks functional dendritic cells for a full KE3 response.	New co-culture models are being developed to address this [28].
Co-culture Models (RHE + THP-1)	More integrated assessment of KE2 & KE3 [28]	Shows promise but is not yet standardized or OECD-validated. Protocol transferability between labs needs verification.	A 2025 study used SkinEthic RHE co-cultured with THP-1 cells, measuring IL-18, CD54, and CD86 to screen 41 cosmetic formulations [28].

Experimental Protocol: In Vitro Co-culture Model for Formulation Testing

A promising advanced model involves co-culturing a 3D reconstructed human epidermis (RHE) with immune cells to better assess formulations [28].

• Model Preparation: A commercially available RHE model (e.g., SkinEthic, EpiDerm) is used. The model is equilibrated according to the manufacturer's instructions.
• Co-culture Setup: THP-1 cells (a human monocytic cell line acting as dendritic cell precursors) are introduced into the underlying culture compartment or directly co-cultured, allowing for cross-talk between skin and immune cells.
• Formulation Exposure: The final cosmetic formulation (e.g., a leave-on cream or rinse-off cleanser) is applied topically to the RHE surface at a defined dose (e.g., 10-16 mg/cm²) for a specified period (e.g., 24-48 hours).
• Endpoint Measurement: Multiple endpoints are analyzed to capture different KEs:
  • Viability (MTT assay): To ensure effects are not due to general cytotoxicity.
  • KE2 Marker (IL-18 secretion): Measured in the culture supernatant using ELISA.
  • KE3 Markers (CD54 and CD86 expression): THP-1 cells are analyzed by flow cytometry to determine the Relative Fluorescence Intensity (RFI). An RFI of ≥150% for CD54 or ≥200% for CD86 is considered a positive sensitization response [28] [50].
• Data Integration: Results are combined to predict the formulation's sensitization potential, offering a more holistic view than single-assay approaches.

The Scientist's Toolkit: Essential Research Reagents and Models

Table 4: Key Reagent Solutions for Investigating Sensitization Mechanisms

Research Reagent / Model	Function in Sensitization Research	Specific Application Example
Synthetic Peptides (Cys/Lys)	Nucleophilic targets for KE1 reactivity assessment [49].	Used in DPRA and PPRA to quantify haptenation potential of chemicals.
THP-1 Cell Line	Human monocyte line modeling dendritic cell activation (KE3) [28] [50].	Used in h-CLAT and co-culture models to measure CD54/CD86 upregulation.
Reconstructed Human Epidermis (RHE)	3D model of human epidermis for topical application [28] [53].	Used in EpiSensA (KE2), skin irritation tests, and advanced co-cultures with immune cells.
Horseradish Peroxidase (HRP)	Enzyme to metabolically activate prohaptens in modified assays [49].	Key component of the PPRA to study prohaptens like 2M4MP.
Interleukin 18 (IL-18)	Pro-inflammatory cytokine released by keratinocytes during KE2 [28] [2].	A key biomarker measured in RHE-based models and co-culture systems to assess keratinocyte response.
Antibodies (CD54, CD86)	Fluorescently-labeled antibodies for flow cytometry.	Essential for quantifying dendritic cell activation in h-CLAT and co-culture models [28] [50].

Current non-animal models for skin sensitization provide valuable mechanistic data within the AOP framework but possess critical, well-documented limitations. The inability of standard KE1 assays to reliably detect prehaptens and prohaptens without modification and the challenge of predicting effects within complex formulations represent the most significant gaps. These limitations necessitate the use of integrated testing strategies (IATA) that combine multiple methods rather than relying on a single stand-alone assay [48] [2].

The future of the field lies in developing and validating more complex, immunocompetent models. The integration of organ-on-a-chip technologies, microfluidics, and in silico models holds the promise of capturing the metabolic conversion of prohaptens and the complex ingredient interactions that define the formulation effect, ultimately leading to more predictive and human-relevant safety assessments [2] [37].

The ban on animal testing for cosmetics in the European Union and the global push for the Replacement, Reduction, and Refinement (3Rs) of animal experiments have propelled the development of non-animal methods (NAMs) for safety assessment, particularly in the field of skin sensitization [2]. Allergic Contact Dermatitis (ACD), the clinical manifestation of skin sensitization, affects approximately 20% of the population in European countries, a significant proportion of which is caused by ingredients in cosmetic products [2] [9]. Ensuring consumer safety while complying with these regulatory demands requires robust and physiologically relevant in vitro models that can accurately predict human responses.

The cornerstone of modern non-animal testing is the Adverse Outcome Pathway (AOP), which deconstructs the complex process of skin sensitization into a sequence of measurable Key Events (KEs), from the initial molecular interaction to the adverse organism-level response [2] [9]. While several alternative methods addressing individual KEs have been formally validated by the OECD, a scientific consensus holds that no single test can fully capture the intricate cellular crosstalk of the human immune response [2] [9]. This review will objectively compare the performance of advanced, physiologically relevant testing strategies, focusing on how the optimization of cell sources, co-cultures, and metabolic capacity enhances the accuracy of in vitro skin sensitization models.

The Adverse Outcome Pathway: A Framework for Validation

The skin sensitization AOP, as formalized by the OECD, provides a critical framework for validating the physiological relevance of any in vitro model [2] [54]. The pathway comprises four key events in the induction phase:

• KE1: Molecular Initiating Event – Covalent binding of electrophilic haptens to nucleophilic residues in skin proteins (e.g., cysteine or lysine) [2].
• KE2: Keratinocyte Response – Activation of keratinocytes, leading to the release of pro-inflammatory cytokines and the activation of specific gene pathways, such as the antioxidant/electrophile response element (ARE)-dependent pathway [2] [9].
• KE3: Dendritic Cell Activation – Maturation and activation of dendritic cells, characterized by upregulation of specific cell surface markers (e.g., CD54, CD86), enabling them to present the antigen [2] [9].
• KE4: T-cell Proliferation – In the lymph nodes, activated dendritic cells present the antigen to naïve T-cells, leading to their proliferation and differentiation into allergen-specific memory T cells [2] [55].

The final adverse outcome, ACD, manifests clinically upon re-exposure to the allergen [2]. This AOP underpins all subsequent discussions of model optimization and validation.

Performance Comparison of Testing Platforms

The following tables summarize the performance characteristics of various testing approaches, from single-key-event tests to complex, optimized models.

Table 1: Performance of Individual Non-Animal Tests for Specific Key Events

Test Method (OECD Guideline)	AOP Key Event Addressed	Measured Endpoint	Typical Application Context
Direct Peptide Reactivity Assay (DPRA), 442C [9] [56]	KE1: Molecular Initiating Event	Depletion of cysteine- and lysine-containing synthetic peptides	In chemico screening of haptenation potential
ARE-Nrf2 Luciferase Test (KeratinoSens), 442D [2] [9]	KE2: Keratinocyte Response	Activation of the ARE pathway, measured via luciferase gene reporter activity	In vitro assessment of keratinocyte activation
Human Cell Line Activation Test (h-CLAT), 442E [9] [56]	KE3: Dendritic Cell Activation	Upregulation of CD86 and CD54 surface markers on THP-1 cell line	In vitro assessment of dendritic cell activation
Reconstructed Human Epidermis (RHE) Models (EpiSensA), 442D [2]	KE2 & Tissue Context	Cytokine release and tissue viability in a 3D epidermal model	Physiologically complex platform for KE2

Table 2: Quantitative Performance Comparison of Defined and Integrated Testing Strategies

Testing Strategy	Components	Reported Accuracy vs. Human Data	Key Advantages	Key Limitations
"2-out-of-3" ITS [9]	DPRA, SENS-IS (or KeratinoSens), h-CLAT	High accuracy on 33-chemical set; resolved ~88% of chemicals with DPRA+SENS-IS first	Follows AOP logic; reduces number of tests needed with a strategic sequence	Performance can vary with the choice and sequence of assays
In Silico Consensus Model [55]	Rule-based (KE1), LLNA stats-based (KE4), GPMT stats-based (Adverse Outcome)	78% Balanced Accuracy (vs. human data)	Combines multiple KE/AO pathways; covers wide chemical space	Dependent on quality and size of underlying data sets
In Silico Tools (Individual) [54]	e.g., Toxtree, QSAR Toolbox, Derek Nexus	~70-80% Accuracy (vs. human data), comparable to LLNA	Very fast and low-cost; useful for high-throughput screening	Variable sensitivity/specificity; not all are reliable for standalone regulatory use

Optimizing Physiological Relevance in Model Systems

The limitations of single-cell-type assays have driven research into more complex systems that better mimic human skin physiology.

• 3D Reconstructed Human Epidermis (RHE) Models: These models represent a significant leap in physiological relevance. They provide a structured, multi-layered epidermis with a functional stratum corneum, more accurately modeling the barrier function and cellular microenvironment of human skin than 2D cultures [2]. The EpiSensA model, the first RHE-based test adopted in an OECD guideline (442D), leverages this complexity to assess the keratinocyte response (KE2) [2].

• Incorporating Immune Competence: A persistent gap in many standard models is the lack of active crosstalk between skin and immune cells. Next-generation approaches aim to incorporate dendritic cells (DCs) or their precursors into RHE models, or to create full skin equivalents that include a dermis with fibroblasts [2]. These co-culture systems are designed to directly model the critical interaction between antigen-presenting cells and keratinocytes (KE2-KE3), which is fundamental to the sensitization process [2].

• Organ-on-a-Chip and Microfluidics: These emerging technologies allow for dynamic perfusion of nutrients and test chemicals, as well as controlled introduction of immune cells. They can be used to create a more physiologically relevant microenvironment and to model the migration of dendritic cells from the skin to the lymph nodes, a crucial step in the AOP that is absent from static models [2].

Addressing Metabolic Capacity and Microenvironment

A cell's metabolic state is intrinsically linked to its function, and uncontrolled metabolic drift is a major source of experimental irreproducibility [57].

• The Problem of Metabolic Drift: Studies have shown that under standard, non-optimized culture conditions, cells can experience drastic nutrient depletion (e.g., glutamine, glucose) and accumulation of waste products (e.g., lactate) within hours [57]. This shifting microenvironment forces cells to rewire their metabolism, which can alter their response to toxicants and lead to highly variable and irreproducible data [57]. For instance, the effect of a glutaminase inhibitor was masked when cells depleted the media glutamine, a key substrate, too quickly [57].

• Strategies for Metabolic Optimization:

  • Rational Assay Development: Systematically varying parameters such as initial cell seeding density and the timing of inhibitor addition to ensure that nutrient levels remain in a stable, non-limiting range throughout the entire assay duration [57].
  • Dynamic Metabolic Modelling: Advanced computational approaches, like dynamic Flux Balance Analysis (dFBA), can be used to simulate the metabolism of individual cell types in a co-culture. These models integrate genome-scale metabolic networks to predict nutrient consumption, metabolite production, and microbial abundance, providing a rationale for optimizing culture media and conditions to maintain physiological metabolic states [58].
  • Monitoring Metabolites: Implementing routine monitoring of key extracellular metabolites (e.g., glucose, glutamine, lactate) can help identify metabolic shifts and validate that optimized conditions are effective [57].

The following diagram illustrates the logical workflow for developing a physiologically relevant model, integrating the optimization of both cellular components and the metabolic microenvironment.

Detailed Experimental Protocols

To ensure reproducibility and provide a clear basis for comparison, detailed methodologies for key experiments are provided below.

This protocol exemplifies a rigorous approach to modeling multi-species interactions, a methodology that can be adapted for co-cultures of human skin and immune cells.

• Objective: To investigate the metabolic behavior of a co-culture using the characterized individual organisms that constitute it, and to model the fermentation dynamics behind the microbial composition.
• Materials:
  • Strains: Lactococcus lactis subsp. cremoris MG1363, Lactococcus lactis subsp. lactis IL1403, Streptococcus thermophilus LMG 18311, Leuconostoc mesenteroides subsp. cremoris ATCC 19254.
  • Culture Medium: Chemically defined medium (CDM) as described by Otto et al. and modified by Poolman & Konings [58].
  • Equipment: 1-L stirred tank bioreactor (e.g., Biostat Q, B. Braun), anaerobic conditions, no pH control (initial pH 6.8).
• Methods:
  • Inoculum Preparation: Grow individual strains in CDM until late exponential phase.
  • Bioreactor Inoculation: Inoculate the bioreactor (0.6 L working volume) with 2% (vol/vol) inoculum. For co-cultures, use initial optical density (OD) ratios of 1:1 for two-species or 1:1:1 for three-species cultures.
  • Fermentation Conditions: Maintain anaerobic conditions by flushing with Nâ‚‚. Run mesophilic cultures (e.g., L. lactis and Leuconostoc) at 30°C and thermophilic cultures (e.g., L. lactis and S. thermophilus) at 33°C. Use slow mixing (50 rpm).
  • Sampling and Analysis:
    • Biomass: Determine total biomass concentration via OD600 and correlate with dry weight. For co-cultures, use qPCR with strain-specific primers to determine relative microbial abundance over time.
    • Metabolites: Analyze cell-free supernatant for glucose, organic acids, and amino acids using HPLC and DNS methods for reducing sugars.
  • Modeling: Use dynamic Flux Balance Analysis (dFBA) with strain-specific kinetic parameters estimated from pure culture data to simulate and predict co-culture metabolism and interactions.

This protocol describes a defined approach to hazard prediction that combines multiple AOP key events.

• Objective: To accurately predict skin sensitization hazard by integrating data from three non-animal methods following a strategic sequence.
• Test Methods:
  • DPRA (OECD 442C): Assesses covalent binding to peptides (KE1).
  • SENS-IS Assay: Assesses gene expression changes in a reconstructed epidermis model (KE2).
  • h-CLAT (OECD 442E): Assesses dendritic cell activation (KE3).
• Testing Strategy & Prediction Model:
  • First, perform the DPRA and the SENS-IS assay.
  • If both assays agree (both positive or both negative), a prediction can be made without the need for the third test.
  • If the results from the DPRA and SENS-IS are discordant (one positive, one negative), perform the h-CLAT as a tie-breaker.
  • Apply a "2-out-of-3" prediction model: A chemical is predicted as a skin sensitizer if at least two of the three tests yield a positive result.
• Validation: This strategy was validated on a set of 33 reference chemicals, showing high predictive accuracy when compared to in vivo (LLNA) and human data [9].

The Scientist's Toolkit: Key Research Reagent Solutions

The following table lists essential reagents and tools for developing and running optimized in vitro models for skin sensitization research.

Table 3: Essential Research Reagents and Tools for Advanced In Vitro Models

Reagent / Tool	Function / Application	Example in Context
Reconstructed Human Epidermis (RHE)	3D tissue model for KE2 assessment and barrier function studies	EpiSensA model for OECD TG 442D testing [2]
THP-1 Human Monocytic Cell Line	In vitro model for dendritic cell activation (KE3)	Used in the h-CLAT (OECD TG 442E) to measure CD86/CD54 upregulation [9]
Keratinocyte Reporter Cell Lines	In vitro model for keratinocyte activation (KE2)	KeratinoSens cell line with ARE-luciferase construct [9]
Chemically Defined Media (CDM)	Provides a consistent, serum-free environment for reproducible cell culture and metabolism studies	Used in co-culture metabolic modeling to precisely control nutrient composition [58]
Genome-Scale Metabolic Models (GEMs)	Computational frameworks to simulate and predict cellular metabolism	Used with dFBA to simulate co-culture behavior and optimize conditions [58]
STING and LTβR Agonists	Immune-activating agents to study and induce tertiary lymphoid structures	Used in mouse models to create "immune-hot" tumors for immunotherapy research [59]
In Silico Prediction Tools	Software for rapid, cost-effective screening of sensitization potential	Tools like Toxtree, Derek Nexus, and OECD QSAR Toolbox [54]

The transition to non-animal methods for skin sensitization testing is well underway, driven by a robust AOP framework. While single-key-event tests provide valuable data, the scientific and regulatory future lies in integrated, physiologically relevant models. As the comparative data and protocols in this guide demonstrate, optimizing cell sources through 3D models and co-cultures, coupled with a rigorous approach to controlling and modeling the metabolic microenvironment, is paramount to improving prediction accuracy. These advanced models, which more faithfully recapitulate the complex biology of human skin, are essential for ensuring consumer safety and driving innovation in the cosmetics and chemical industries in a post-animal-testing era.

From Bench to Regulation: Establishing Confidence through Validation and Benchmarking

In the field of toxicology and immunology, validating new methodologies requires rigorous benchmarking against established reference points. For skin sensitization research, the murine Local Lymph Node Assay (LLNA) and human repeated insult patch tests (HRIPTs) serve as these critical benchmarks, providing the foundation for assessing the predictive capacity of novel approaches [60] [61]. The LLNA functions as a pivotal animal test that quantitatively measures skin sensitizing potency through the EC3 value—the estimated concentration required to induce a three-fold increase in lymphocyte proliferation compared to vehicle controls [60] [62]. Meanwhile, human clinical data, particularly from HRIPTs, represent the ultimate ground truth for human responses, offering direct evidence of sensitization thresholds in humans [60]. This guide objectively compares the performance of established and emerging methods against these gold standards, providing researchers with a structured framework for evaluating non-animal testing strategies in the context of evolving regulatory landscapes and the global shift toward alternative methods [61] [2].

Gold Standards: Defining the Benchmarks

The LLNA EC3 Value: A Quantitative Animal Benchmark

The LLNA provides a quantitative assessment of skin sensitization potency by generating a dose-response curve, from which the EC3 value is interpolated [60] [62]. This value has demonstrated robust inter- and intra-laboratory reproducibility and has been extensively correlated with human sensitization experience [60]. The statistical properties of EC3 values are crucial for proper interpretation; recent analyses indicate that EC3 data follows a log-normal distribution rather than a normal distribution, which must be considered when comparing potency classifications and evaluating new testing approaches [62].

Human Data: The Clinical Ground Truth

Human repeated insult patch tests provide the most direct measurement of human skin sensitization thresholds [60]. In these controlled clinical studies, human volunteers are repeatedly exposed to sub-irritant concentrations of test materials to determine the threshold dose required to induce sensitization. Research has demonstrated a clear linear relationship between LLNA EC3 values and human sensitization thresholds when both are expressed as dose per unit area (µg/cm²), substantiating the utility of EC3 values for predicting relative human sensitizing potency [60].

Understanding "Gold Standard" vs. "Ground Truth"

In scientific validation, a critical distinction exists between these terms [63] [64]:

• Gold Standard: The best available method under reasonable conditions, which may have known limitations but serves as the benchmark for comparison. The LLNA occupies this position for skin sensitization potency assessment [64].
• Ground Truth: The absolute state of information; in skin sensitization, this represents the actual human response, which HRIPTs attempt to measure [63].

This distinction is particularly relevant given the recognition that even gold standard tests have uncertainties and limitations that must be accounted for in validation studies [62] [64].

Comparative Performance of Testing Methods

Established In Vitro and In Chemico Methods

Multiple non-animal methods have been developed and validated against the LLNA and human data, each addressing specific Key Events in the Adverse Outcome Pathway (AOP) for skin sensitization [61] [65]. The following table summarizes the fundamental characteristics of these methods and their relationship to the AOP framework:

Table 1: Skin Sensitization Testing Methods and Their Correlation to Gold Standards

Method	AOP Key Event	Measurement Endpoint	Correlation with LLNA/Human	Regulatory Status
Direct Peptide Reactivity Assay (DPRA)	KE1: Molecular Initiating Event	Peptide depletion via HPLC/SPE-MS/MS	Good correlation for reactivity; limited for potency	OECD TG 442C
KeratinoSens	KE2: Keratinocyte Response	Nrf2-mediated luciferase activation	Moderate correlation with LLNA; used in Defined Approaches	OECD TG 442D
h-CLAT	KE3: Dendritic Cell Activation	CD86/CD54 surface expression	Moderate correlation with LLNA; used in Defined Approaches	OECD TG 442E
SENS-IS assay	Multiple KEs	Genomic profiling in 3D epidermis	>93% reproducibility; correlates with human/LLNA potency	Validation for OECD TG 442D
Human Skin Explant Test	T-cell mediated response	T-cell proliferation & cytokine release	81% accuracy (13/16 mAbs) with clinical outcome	Research use

Quantitative Correlation with Gold Standards

The predictive performance of alternative methods is ultimately determined by their correlation with established benchmarks. Recent validation studies provide quantitative data on how well these methods perform against LLNA EC3 values and human data:

Table 2: Quantitative Performance Metrics of Alternative Methods Against Gold Standards

Method/Approach	Correlation with LLNA EC3	Correlation with Human Data	Reproducibility	Key Limitations
LLNA (Reference)	Reference	Linear relationship (dose/area)	Inter-lab variability characterized [62]	Animal use; statistical uncertainty near thresholds [62]
2-out-of-3 Defined Approach	75-85% accuracy [65]	Comparable to LLNA [61]	High (fixed rules)	Limited potency information; chemical applicability gaps
SENS-IS (Skin+ model)	Categorizes LLNA potency [66]	Predicts human potency categories [66]	>93% intra-/inter-batch [66]	Limited to specific 3D model types
qHTS Adaptation (KeratinoSens)	Screening capable [65]	Not fully established	Suitable for HTS	Part of battery, not stand-alone
Skin Explant Test	Not primary endpoint	81% accuracy with clinical outcomes [67]	Donor variability possible	Specialized for biologics; lower throughput

Experimental Protocols for Benchmarking Studies

Protocol for LLNA EC3 Determination

The standard protocol for establishing the LLNA EC3 value, which serves as the reference point for benchmarking alternative methods [62]:

• Animal Model: Female CBA/J mice (8-12 weeks old)
• Dosing Regimen: Apply 25 μL of test substance at three to four increasing concentrations to the dorsum of both ears daily for three consecutive days
• Control Groups: Concurrent vehicle controls included in each experiment
• Lymphocyte Proliferation Measurement: On day 6, inject mice intravenously with ³H-methyl-thymidine; five hours later, excise draining auricular lymph nodes
• Lymph Node Processing: Create single-cell suspension and measure ³H-thymidine incorporation using β-scintillation counting
• EC3 Calculation: Interpolate the concentration required to produce a three-fold increase in proliferation compared to vehicle controls using dose-response regression analysis
• Statistical Considerations: Account for log-normal distribution of EC3 values; calculate confidence intervals based on established uncertainty parameters [62]

Protocol for SENS-IS Assay Using 3D Epidermis Models

The recently validated protocol for the SENS-IS assay, which can be benchmarked against LLNA EC3 values [66]:

• Model Preparation: Use reconstructed human epidermis (RHE) models (Episkin or Skin+)
• Test Substance Application: Apply chemicals directly to the epidermal surface for 24 hours
• RNA Extraction and Quality Control: Isolve total RNA; ensure RNA Integrity Number (RIN) >7.0
• Gene Expression Analysis: Quantify expression of 17 biomarker genes using targeted RNA-Seq or qPCR
• Prediction Model Application: Apply proprietary algorithm to gene expression data
• Potency Categorization: Classify chemicals as non-sensitizer, weak, moderate, or strong sensitizer based on genomic signature
• Validation Against Benchmarks: Compare categorization with existing LLNA EC3 values and human potency data

Protocol for Defined Approaches (2-out-of-3)

Standardized protocol for the widely used Defined Approach that integrates multiple non-animal methods [65]:

• Test Battery Application:

  • Perform DPRA (OECD TG 442C): Measure peptide depletion for cysteine and lysine
  • Conduct KeratinoSens (OECD TG 442D): Assess Nrf2-mediated luciferase activation
  • Execute h-CLAT (OECD TG 442E): Quantify CD86 and CD54 expression in THP-1 cells

• Data Interpretation Procedure (DIP):

  • Score each assay result as positive (+) or negative (-) based on established thresholds
  • Apply the "2-out-of-3" rule: If at least two assays return positive results, classify as sensitizer
  • If only one or no assays positive, classify as non-sensitizer

• Performance Verification:

  • Compare classification results with LLNA and human data
  • Calculate accuracy, sensitivity, and specificity metrics

Signaling Pathways and Testing Strategies

Adverse Outcome Pathway for Skin Sensitization

The AOP provides the mechanistic framework for understanding skin sensitization and developing testing strategies [61] [2]. The following diagram illustrates the key events and their relationship to testing methods:

Integrated Testing Strategy Workflow

A defined approach workflow that integrates multiple non-animal methods for comprehensive assessment:

The Scientist's Toolkit: Essential Research Reagent Solutions

Table 3: Essential Research Reagents for Skin Sensitization Testing

Reagent/Model	Specific Examples	Research Application	Function in Assay
3D Reconstructed Epidermis	EpiSkin, Skin+, EpiDerm	SENS-IS assay, EpiSensA	Provides physiologically relevant barrier for chemical exposure
Peptide Reagents	Cysteine peptide (Ac-RFAACAA-COOH), Lysine peptide (Ac-RFAAKAA-COOH)	DPRA	Nucleophilic targets for haptenation measurement
Reporter Cell Lines	KeratinoSens (Nrf2-ARE Luciferase), IL-8 Luc THP-1	KE2 and KE3 assessment	Mechanism-specific activation readouts
Dendritic Cell Lines	THP-1, U937	h-CLAT, U-SENS	Measure cell surface marker changes (CD86, CD54)
Cytokine Detection	IL-8, IL-18, IL-1β HTRF kits	Cytokine profiling	Quantify inflammatory responses in various assays
qPCR/Omics Reagents	RNA isolation kits, targeted RNA-Seq panels	SENS-IS, genomic analysis	Gene expression profiling for sensitization signatures

The landscape of skin sensitization testing is undergoing a fundamental transformation, driven by the transition from animal models to mechanistically based non-animal methods [61] [2]. Successful benchmarking against gold standards requires acknowledging both the strengths and limitations of reference methods while recognizing that the scientific community is moving toward integrated approaches that may eventually surpass the predictive capacity of any single test [61]. The correlation between LLNA EC3 values and human data provides a crucial bridge for translating between animal and human responses, while emerging technologies like 3D epidermis models and high-throughput screening platforms offer promising avenues for more human-relevant, ethical, and efficient safety assessment [66] [65]. As research advances, the continued refinement of these methods and their correlation with human responses will further solidify their role in next-generation risk assessment paradigms.

The ban on animal testing for cosmetics in the European Union and similar regulatory shifts globally have accelerated the development and validation of non-animal methods (NAMs) for skin sensitization assessment [2]. For these methods to gain regulatory acceptance and be deployed reliably in safety decisions, a rigorous evaluation of their performance metrics—primarily predictive accuracy, sensitivity, and specificity—is essential. These metrics provide researchers and regulators with a standardized framework to quantify the reliability and applicability of novel testing strategies. This guide objectively compares the performance of key defined approaches and standalone assays, providing supporting experimental data to inform their use in immune response research.

Performance Comparison of Key Assays and Defined Approaches

The table below summarizes the performance metrics of several prominent non-animal testing strategies, as reported in formal validation studies and scientific literature.

Table 1: Performance Metrics of Skin Sensitization Testing Strategies

Test Method / Defined Approach (DA)	Sensitivity (%)	Specificity (%)	Accuracy (%)	Balanced Accuracy (%)	Reference Data
EpiSensA (RhE-based assay)	92.6	63.0	82.7	77.8	LLNA [68]
GARDskin Dose-Response (Potency prediction)	NESIL prediction error: 2.75-3.22 fold change				LLNA EC3 & Human NOEL [69]
2 out of 3 (2o3) DA (DPRA, KeratinoSens, h-CLAT)	81.0	75.0	79.0	78.0	LLNA [70]
Integrated Testing Strategy (ITSv2)	86.0	50.0	76.0	68.0	LLNA [70]
Machine Learning Model (Two-tiered, SVM)	-	-	88.0 (LLNA), 81.0 (Human)	-	LLNA & Human [33]
KE 3/1 Sequential Testing Strategy	67.0	44.0	59.0	56.0	LLNA [70]

Performance varies significantly across methods. The EpiSensA assay demonstrates high sensitivity, correctly identifying most true sensitizers, though its specificity is more moderate [68]. Defined Approaches that integrate multiple data sources, such as the "2 out of 3" DA, show a more balanced profile of high sensitivity and specificity, leading to strong overall balanced accuracy [70]. Advanced computational models, like the support vector machine (SVM) model, can achieve high accuracy in categorizing substances, in some cases outperforming animal tests like the LLNA in predicting human outcomes [33].

Experimental Protocols for Key Methods

EpiSensA (Reconstructed Human Epidermis-Based Assay)

The EpiSensA method addresses Key Event 2 (keratinocyte activation) in the Adverse Outcome Pathway (AOP) for skin sensitization [2] [68].

• Test System: Uses a Reconstructed Human Epidermis (RhE) model, which features a stratified stratum corneum, allowing for topical application of chemicals akin to in vivo exposure [68].
• Chemical Application: Test chemicals, including lipophilic substances and pre-/pro-haptens, are applied topically to the RhE surface. Vehicles like acetone-olive oil or 50% ethanol are used to ensure proper delivery [68].
• Endpoint Measurement: The core endpoint is the quantification of changes in the expression levels of a defined set of gene biomarkers following chemical exposure. This is typically measured using reverse transcription quantitative polymerase chain reaction (RT-qPCR) [68].
• Data Interpretation: The fold-change in gene expression is calculated relative to vehicle-treated controls. A chemical is classified as a sensitizer if the expression of the biomarker genes meets or exceeds a pre-defined threshold [68].

GARDskin Dose-Response Assay

The GARDskin assay addresses Key Event 3 (dendritic cell activation) and is approved in OECD TG 442E [69] [71].

• Test System: Utilizes the Senzacell cell line, a human dendritic cell line engineered to monitor immune responses [69].
• Chemical Exposure & Viability Assessment: The assay begins with a cytotoxicity test to determine a concentration range that induces low-to-non-toxic conditions. The highest concentration tested is the "GARD input concentration" [69].
• Dose-Response Design: The test chemical is evaluated over a series of concentrations, typically using a dilution series. This generates a dose-response curve [69].
• Genomic Endpoint: At each concentration, the expression of a Genomic Prediction Signature (GPS) of nearly 200 genes is measured. A support vector machine (SVM) algorithm analyzes this expression data to output a Decision Value (DV) for each concentration [69] [71].
• Potency Estimation (cDV0): The cDV0 value is the main readout for potency. It is the estimated lowest concentration predicted to induce a positive classification (DV ≥ 0), determined by linear interpolation between concentrations that bracket the classification threshold [69].

Defined Approaches (DAs)

DAs integrate results from multiple, mechanistically complementary tests.

• 2 out of 3 (2o3) DA: This is a simple hazard identification strategy. It requires data from three OECD-approved test methods: the DPRA (KE1), KeratinoSens (KE2), and h-CLAT (KE3). The final classification is based on the majority vote, or "2 out of 3" concordant results [30] [70].
• Integrated Testing Strategy (ITS): More complex DAs, like ITSv2, integrate data from the h-CLAT and DPRA with an in silico hazard prediction from the OECD QSAR Toolbox. These inputs are processed through a pre-defined data interpretation procedure (DIP) or scoring system to generate a hazard classification and/or a potency categorization according to GHS [30] [70].

Visualizing the Testing Workflow and AOP Framework

The following diagram illustrates the general experimental workflow for a genomic-based in vitro test method, such as GARDskin, and its alignment with the AOP.

Diagram 1: Genomic assay workflow in AOP context.

The Scientist's Toolkit: Essential Research Reagents and Materials

The table below details key reagents and materials essential for conducting the in vitro tests discussed in this guide.

Table 2: Key Research Reagents and Materials for In Vitro Skin Sensitization Testing

Item Name	Function / Application in Assay
Reconstructed Human Epidermis (RhE)	3D tissue model mimicking human skin structure; used for topical application in assays like EpiSensA [2] [68].
Senzacell Cell Line	Human dendritic cell line used in the GARDskin assay to model the immune response to sensitizers [69].
THP-1 or U937 Cell Lines	Human monocytic cell lines differentiated into dendritic-like states; used in the h-CLAT and U-SENS assays to measure cell surface markers [33] [70].
Synthetic Peptides (containing Lysine & Cysteine)	Nucleophilic targets for test chemicals in in chemico assays like the DPRA and ADRA to assess protein binding (KE1) [2] [30].
ARE-Luciferase Reporter Cell Lines (e.g., KeratinoSens)	Keratinocyte-based cell lines used to measure activation of the Nrf2-ARE pathway, a key event in keratinocyte response (KE2) [33] [70].
Gene Expression Panels (e.g., for GARDskin GPS or EpiSensA biomarkers)	Pre-defined sets of genes whose expression is quantified via RT-qPCR or RNA-Seq to serve as a biomarker signature for classification [69] [71] [68].
Flow Cytometry Antibodies (e.g., anti-CD86, anti-CD54)	Used in h-CLAT to detect and quantify the upregulation of cell surface activation markers on dendritic cells following chemical exposure [70].

The landscape of skin sensitization safety assessment is firmly anchored on a foundation of robust NAMs. The performance data presented in this guide demonstrates that modern Defined Approaches and advanced standalone assays can achieve a level of predictive accuracy that is fit for regulatory purpose, with some methods even rivaling or surpassing the performance of historical animal tests. For researchers, the choice of method depends on the specific need—whether for simple hazard identification, potency categorization, or quantitative risk assessment—and should be guided by the respective performance strengths and limitations of each approach. The continued integration of high-content data, such as genomics, with sophisticated machine learning models promises to further refine these metrics, enhancing the biological relevance and predictive power of next-generation risk assessments.

The assessment of skin sensitization potential is a critical component of the safety evaluation for chemicals, cosmetics, and pharmaceuticals. The global regulatory landscape for this endpoint has undergone a profound transformation, shifting from traditional animal-based tests like the murine Local Lymph Node Assay (LLNA) toward New Approach Methodologies (NAMs) that are more human-relevant and ethically aligned with the principles of Replacement, Reduction, and Refinement (3Rs) of animal testing [72] [2]. This evolution is framed by the Adverse Outcome Pathway (AOP), a conceptual model that deconstructs the complex biological process of skin sensitization into a sequence of measurable key events (KEs) [2] [40]. For researchers and drug development professionals, navigating this new paradigm requires a clear understanding of two interconnected pillars: the detailed OECD Test Guidelines (TGs), which provide standardized, internationally accepted test methods, and the specific acceptance criteria of major regulatory bodies like the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA) [72] [73] [74]. This guide provides a comparative analysis of these frameworks, offering a detailed overview of validated protocols and their application in regulatory decision-making for immune response research.

The Adverse Outcome Pathway (AOP) for Skin Sensitization

The AOP for skin sensitization provides a mechanistic framework that links a molecular initiating event to an adverse outcome at the organism level—allergic contact dermatitis (ACD) [2]. This conceptual model is foundational to the development and validation of all modern NAMs.

• Molecular Initiating Event (KE1): A low-molecular-weight chemical (a hapten) penetrates the skin's stratum corneum and forms a stable covalent bond with nucleophilic residues on skin proteins (e.g., on cysteine or lysine), a process known as haptenation [2] [65].
• Cellular Response in Keratinocytes (KE2): The hapten-protein complex induces an inflammatory response in keratinocytes, including the activation of the Keap1-Nrf2 pathway, leading to the release of pro-inflammatory cytokines such as interleukin-18 (IL-18) and IL-1α [2] [65].
• Dendritic Cell Activation (KE3): Epidermal dendritic cells (e.g., Langerhans cells) recognize, internalize, and process the hapten-protein complexes. They then mature and migrate to the draining lymph nodes, a process characterized by the upregulation of cell surface markers like CD54 and CD86 [2] [65].
• T-Cell Proliferation (KE4): In the lymph nodes, dendritic cells present the antigen to naïve T-cells, triggering their proliferation and differentiation into allergen-specific memory T-cells [65].
• Adverse Outcome (Elicitation): Upon subsequent exposure to the same allergen, these memory T-cells are recruited to the skin, initiating an inflammatory cascade that results in the clinical symptoms of ACD (erythema, eczema) [2].

The following diagram illustrates the key events and relationships within the skin sensitization AOP.

OECD Test Guidelines: Standardized Methods for Each Key Event

The Organisation for Economic Co-operation and Development (OECD) Test Guidelines are internationally recognized as the standard methods for chemical safety testing. The following table summarizes the key OECD TGs for skin sensitization, each targeting a specific KE in the AOP [2] [73] [75].

Table 1: OECD Test Guidelines for Skin Sensitization Assessment

OECD TG	Test Method Name	Target AOP Key Event	Measured Endpoint	Brief Principle
442C	Direct Peptide Reactivity Assay (DPRA)	KE1 (Molecular Initiating Event)	Peptide depletion	Measures the covalent binding of a chemical to synthetic peptides containing cysteine or lysine in a test tube [76] [65].
442D	ARE-Nrf2 Luciferase Test Methods (e.g., KeratinoSens)	KE2 (Keratinocyte Response)	Luciferase induction	Uses a reporter gene cell line to measure the activation of the antioxidant response element (ARE) pathway, indicative of electrophilic stress [76] [65].
442E	In Vitro Skin Sensitisation Assays Addressing KE3 (e.g., h-CLAT, U-SENS)	KE3 (Dendritic Cell Activation)	CD86/CD54 expression	Measures the upregulation of cell surface markers associated with dendritic cell activation in a human cell line (e.g., THP-1) [76] [65].
497	Defined Approaches for Skin Sensitisation	Integrated AOP	Hazard & Potency	Provides fixed data interpretation procedures (DIPs) for combining results from multiple TGs (e.g., 442C, D, E) to predict hazard and potency without animal data [76] [75].

Recent Updates in OECD Test Guidelines

The OECD TGs are continuously updated. In June 2025, several key guidelines were revised [73] [75]:

• TG 497 (Defined Approaches): Updated to allow the use of in vitro and in chemico methods in TGs 442C, 442D, and 442E as alternate information sources and to include a new Defined Approach for determining the point of departure for skin sensitization potential [75].
• TG 442C (DPRA): Updated to address borderline ranges for the assay [75].
• TG 442D: The first reconstructed human epidermis (RHE)-based test method, EpiSensA, was incorporated into this guideline [2].

Regulatory Acceptance: FDA and EMA Frameworks

FDA Acceptance of NAMs

The FDA has actively promoted the integration of NAMs into regulatory decision-making. The agency's New Alternative Methods (NAM) Program is intended to spur the adoption of methods that can replace, reduce, and refine animal testing [77].

• Policy and Submission Acceptance: The FDA has announced that NAM data is now welcome within Investigational New Drug (IND) applications [72]. In a significant move, the FDA has also stated it will be phasing out animal test requirements for monoclonal antibodies and other drugs, encouraging the use of NAMs such as in silico models, AI, organoids, and organ-on-chip technologies [72].
• Qualification Programs: The FDA employs a formal qualification process for alternative methods, which evaluates a method for a specific Context of Use (COU). This provides developers with confidence that the method is acceptable for the qualified purpose. Relevant programs include the Drug Development Tool (DDT) and Medical Device Development Tool (MDDT) qualification programs [77].
• Specific Guideline Recognition: The FDA accepts alternative methods from OECD guidelines for some product types. Examples include OECD TG 439 (3D reconstructed human epidermis model for skin irritation) and OECD TG 437 (reconstructed human cornea-like epithelium for eye irritation) [77].

EMA Acceptance of NAMs

The European Medicines Agency (EMA) has implemented several mechanisms to support the incorporation of NAMs into regulatory submissions [72].

• Support Pathways: The EMA provides multiple avenues for developers:
  • Innovation Task Force (ITF): Offers informal, early-stage discussions on innovative methods.
  • Scientific Advice: The Scientific Advice Working Party provides formal advice on the use of NAMs in clinical trial or marketing authorization applications.
  • CHMP Qualification: A procedure for developers with robust data to seek a positive qualification opinion on a novel method [72].
• Guidance and Frameworks: EMA's scientific guidelines, while not always NAM-specific, provide the framework for evidence required for marketing authorization. Applicants can deviate from guidelines with full justification, often after seeking scientific advice [74]. The EMA also encourages voluntary data submission to build confidence in novel approaches [72].

Table 2: Comparison of FDA and EMA Acceptance for Skin Sensitization NAMs

Aspect	U.S. FDA	European EMA
Overall Stance	Active encouragement; NAM data "welcome" in INDs; phasing out animal mandates for some products [72].	Supportive; established procedures for qualification and scientific advice [72].
Key Initiative	New Alternative Methods Program; Drug Development Tool (DDT) qualification [77].	Innovation Task Force (ITF); CHMP Qualification Procedure [72].
Basis for Acceptance	Qualification for a specific Context of Use (COU); OECD Test Guidelines [77].	Scientific advice and qualification; adherence to OECD Test Guidelines and defined approaches [72] [74].
Data Integration	Welcomes data from in silico, MPS, and OMICS in submissions, especially when backed by real-world evidence [72].	Promotes defined approaches (e.g., under OECD TG 497) that integrate results from multiple TGs [72] [76].

Experimental Protocols for Key Assays

For researchers designing studies, understanding the core methodology of key assays is essential. Below are detailed protocols for two fundamental OECD TG methods.

Direct Peptide Reactivity Assay (DPRA) - OECD TG 442C

The DPRA is an in chemico method that addresses the Molecular Initiating Event (KE1) by quantifying a chemical's reactivity with model peptides [76] [65].

• Principle: The test quantifies the depletion of two synthetic peptides—one containing cysteine and the other lysine—after incubation with the test chemical. Peptide depletion is a proxy for the chemical's ability to form covalent bonds with skin proteins [65].
• Detailed Procedure:
  • Preparation: Separate solutions of the cysteine-containing peptide and the lysine-containing peptide are prepared in phosphate buffer.
  • Reaction: The test chemical is incubated with each peptide solution in the dark at 25°C for 24 hours.
  • Analysis: The remaining peptide is quantified using high-performance liquid chromatography (HPLC) with ultraviolet (UV) detection. The percentage depletion for each peptide is calculated.
  • Prediction Model: The average of the cysteine and lysine peptide depletion values is used to classify the chemical as a sensitizer or non-sensitizer [65].

Human Cell Line Activation Test (h-CLAT) - OECD TG 442E

The h-CLAT is an in vitro assay that addresses Dendritic Cell Activation (KE3) by measuring changes in surface marker expression [65].

• Principle: The test assesses the upregulation of the surface markers CD86 and CD54 on the human monocytic leukemia cell line THP-1 following exposure to a test chemical. This mimics the activation of dendritic cells [65].
• Detailed Procedure:
  • Cell Culture: THP-1 cells are maintained in standard culture medium (e.g., RPMI 1640 with 10% FBS).
  • Chemical Exposure: Cells are exposed to a range of sub-cytotoxic concentrations of the test chemical for 24 hours.
  • Staining and Analysis: Cells are stained with fluorescently labeled antibodies against CD86 and CD54. A flow cytometer is used to measure the fluorescence intensity, which corresponds to the expression level of each marker.
  • Prediction Model: A chemical is classified as a sensitizer if it induces at least a 150% increase in CD86 expression or a 200% increase in CD54 expression relative to the solvent control at any tested concentration [65].

The Scientist's Toolkit: Essential Research Reagents and Materials

Successful implementation of these test guidelines requires a specific set of reagents and materials. The following table details key solutions for setting up and conducting these assays.

Table 3: Essential Research Reagent Solutions for In Vitro Skin Sensitization Testing

Reagent / Material	Function / Application	Example Assays
Synthetic Peptides	Model nucleophilic targets (cysteine, lysine) for measuring haptenation in in chemico assays.	DPRA (TG 442C) [65]
ARE Reporter Cell Lines	Genetically engineered keratinocyte lines (e.g., KeratinoSens) for detecting Nrf2 pathway activation.	ARE-Nrf2 Luciferase Test (TG 442D) [65]
THP-1 Cell Line	A human monocyte cell line used as a surrogate for dendritic cells to measure activation markers.	h-CLAT (TG 442E) [65]
Fluorescent Antibodies	Antibodies against CD54 and CD86, conjugated to fluorophores, for flow cytometric analysis.	h-CLAT (TG 442E) [65]
Reconstructed Human Epidermis (RHE) Models	3D human skin equivalents that provide a more physiologically relevant platform for testing.	EpiSensA (TG 442D) [2]

The regulatory landscape for skin sensitization assessment is firmly anchored in the Adverse Outcome Pathway framework and the OECD Test Guidelines that operationalize it. Both the FDA and EMA are actively fostering an environment where New Approach Methodologies are not only accepted but increasingly preferred. For researchers, the path forward involves leveraging defined approaches under OECD TG 497, which integrate data from multiple, mechanistically informative tests to provide a comprehensive and human-relevant assessment of skin sensitization potential. Mastery of the standardized protocols and a deep understanding of regulatory expectations are now indispensable for successfully navigating drug and chemical development in this new era.

The global ban on animal testing for cosmetics, coupled with the ethical imperative to reduce animal use in all toxicological research, has catalyzed a paradigm shift in safety assessment. For complex endpoints like skin sensitization—an immunological process that can lead to Allergic Contact Dermatitis—the scientific and regulatory communities have turned to New Approach Methodologies (NAMs). These methods, which include in chemico, in vitro, and in silico approaches, are anchored in the Adverse Outcome Pathway (AOP) framework, which deconstructs the sensitization process into a sequence of measurable key events [2].

This review presents a critical analysis of case studies demonstrating the successful validation and application of Defined Approaches (DAs) and Next-Generation Risk Assessment (NGRA) for skin sensitization. NGRA is characterized as a human-relevant, exposure-led, and hypothesis-driven approach [78]. Framed within a broader thesis on validating in vitro models, this article examines how these innovative methodologies are being integrated into robust testing strategies to meet regulatory requirements and advance immune response research.

The Mechanistic Foundation: The Adverse Outcome Pathway for Skin Sensitization

The AOP for skin sensitization provides the essential mechanistic foundation for developing NAMs. It outlines a sequence of biological events from the initial chemical exposure to the adverse outcome, Allergic Contact Dermatitis [2]. The following diagram illustrates this causal chain and the key events (KEs) targeted by NAMs.

• Molecular Initiating Event (KE1): The process begins when a low-molecular-weight chemical (hapten) penetrates the skin and forms a stable covalent bond with self-proteins, a process called haptenation [2]. Some chemicals, known as pre- and pro-haptens, require activation (e.g., via air oxidation or host metabolism) to become immunogenic.
• Keratinocyte Response (KE2): The hapten-protein complex triggers the activation of keratinocytes, leading to the release of pro-inflammatory cytokines such as interleukin-18 (IL-18) and interleukin-1α (IL-1α). This response may involve the activation of inflammasomes [2].
• Dendritic Cell Activation (KE3): Skin-resident antigen-presenting cells (e.g., Langerhans cells) capture and process the hapten-protein complexes. They then mature and migrate to the draining lymph nodes [2].
• T-Cell Activation (KE4): In the lymph nodes, dendritic cells present the antigen via the Major Histocompatibility Complex (MHC) to naïve T-cells, leading to the proliferation and differentiation of hapten-specific T-cells [2]. This event generates immunological memory.
• Adverse Outcome: Upon subsequent exposure to the same allergen, these memory T-cells are recruited to the skin, causing the clinical symptoms of ACD, including inflammation, erythema, and eczema [2].

Case Studies in Defined Approaches and NGRA

The following case studies showcase the practical application of NAMs, moving from individual tests to integrated strategies for hazard identification and risk assessment.

Case Study 1: Bayesian Network for Potency Assessment

• Objective: To develop a Defined Approach (DA) that classifies skin sensitization potency without animal data by integrating results from multiple NAMs [79].
• Experimental Protocol: This DA employs a Bayesian network, a computational model that calculates the probability of an outcome based on input data. The model integrates in chemico and in vitro data streams targeting the first three Key Events of the AOP [79]:
  • KE1 Data: Input from the Direct Peptide Reactivity Assay (DPRA) or its kinetic version (kDPRA).
  • KE2 Data: Input from the KeratinoSens assay or similar.
  • KE3 Data: Input from the human Cell Line Activation Test (h-CLAT) or U-SENS.
• Results and Validation: The Bayesian network processed these inputs to predict sensitization potency in four categories, aligning with classifications derived from the murine Local Lymph Node Assay (LLNA). The model also provided a Point of Departure (POD) for risk assessment and indicated a confidence level for each prediction, demonstrating that risk assessment can be performed effectively without new animal data [79].

Table 1: Summary of Bayesian Network DA Performance

Model Feature	Description	Regulatory Advantage
Data Integration	Combines 3+ NAMs targeting KE1, KE2, KE3 [79]	Moves beyond stand-alone methods; provides a more comprehensive assessment.
Potency Classification	Predicts LLNA-equivalent potency categories (e.g., weak, strong) [79]	Provides crucial information for ingredient classification and labeling.
Point of Departure (POD)	Generates a toxicity value for risk assessment [79]	Enables safety evaluation and margin of safety calculations.
Confidence Indication	Provides a measure of certainty for each prediction [79]	Supports transparent and weighted decision-making.

Case Study 2: Validation of a 3D Epidermis Model (SENS-IS Assay)

• Objective: To validate the performance of a new 3D reconstructed human epidermis (RHE) model, Skin+, against the established EpiSkin model within the SENS-IS genomic assay [66].
• Experimental Protocol:
  • Model Culture: The Skin+ and EpiSkin RHE models were maintained according to manufacturers' protocols.
  • Chemical Exposure: Both models were topically exposed to 19 proficiency chemicals with well-characterized sensitization potency in humans and the LLNA.
  • Endpoint Analysis: The key endpoint was the measurement of changes in the expression of a panel of genomic biomarkers associated with the skin sensitization pathway. Barrier function integrity was also tested using sodium lauryl sulfate (SLS) [66].
• Results and Validation: The Skin+ model demonstrated over 93% intra- and inter-batch reproducibility, matching the performance of the EpiSkin model. Gene expression profiles and barrier function responses to benchmark chemicals were highly consistent between the two models. This study successfully broadened the options for laboratories using 3D models for sensitive and mechanistic-based sensitization testing [66].

Case Study 3: Tiered NGRA for Cumulative Pyrethroid Exposure

• Objective: To implement a tiered NGRA framework for the cumulative risk assessment of a mixture of pyrethroid insecticides, moving beyond traditional, substance-by-substance approaches [80].
• Experimental Protocol: A five-tier framework was employed, integrating bioactivity and toxicokinetic (TK) data:
  • Tier 1: Bioactivity Indicators: AC50 values for six pyrethroids were gathered from the ToxCast database, categorized by gene and tissue pathways [80].
  • Tier 2: Combined Risk Assessment: The hypothesis of a common mode of action was tested by calculating relative potencies and comparing them to Acceptable Daily Intakes (ADIs) and No-Observed-Adverse-Effect Levels (NOAELs) [80].
  • Tier 3: Margin of Exposure (MoE) Analysis: TK modeling was used to estimate internal doses at realistic exposure levels, which were compared to in vitro bioactivity thresholds to calculate MoEs [80].
  • Tiers 4 & 5: Refinement and Confirmation: Bioactivity indicators were refined using TK models to compare in vitro and in vivo results, confirming that dietary exposures in adults were close to, but below, levels of concern [80].
• Results and Validation: The tiered approach rejected the simple hypothesis of a common mode of action, revealing a more complex interaction. It successfully identified tissue-specific pathways as critical risk drivers and demonstrated that NGRA could provide a nuanced, regulatory-relevant framework for assessing combined chemical exposures [80].

The workflow of this tiered, hypothesis-driven approach is visualized below.

Case Study 4: The Skin Sensitization Prediction Model (SSPM) for Product Development

• Objective: To develop a data-driven in silico model that predicts the risk of skin sensitization in humans for finished product formulations, minimizing the need for new Human Repeat Insult Patch Tests (HRIPT) [43].
• Experimental Protocol: The SSPM is an algorithm based on a historical database of 1,274 unique product formulations and 1226 ingredients, representing results from over 203,640 human subjects [43].
  • Data Input: The formulation's ingredients and their concentrations are entered into the model.
  • Algorithmic Analysis: The model calculates a risk score based on the historical HRIPT outcome data associated with each ingredient and their combinations.
  • Output: The model provides an estimate of the formulation's sensitization potential [43].
• Results and Validation: The SSPM was validated through internal case studies. For instance, it correctly identified a face moisturizer with an anti-aging ingredient as being within the acceptable risk threshold, while a face mask with an anti-acne agent was correctly flagged as exceeding the threshold. This model allows for pre-market screening and decision-making, significantly reducing the reliance on new human testing [43].

Table 2: Comparison of Featured Case Studies

Case Study	Methodology Type	Key AOP Events Covered	Primary Output	Regulatory Readiness
Bayesian Network [79]	Defined Approach (Computational)	KE1, KE2, KE3	Potency category & POD	High (Animal-free)
SENS-IS with Skin+ [66]	In Vitro (3D RHE)	KE2 (Genomic biomarkers)	Hazard & Potency	Validated alternative
Tiered Pyrethroid NGRA [80]	NGRA (Integrated NAMs)	Multiple (Bioactivity)	Cumulative Risk Assessment	Framework for regulatory use
SSPM [43]	In Silico (Data Analytics)	Empirical human data	Formulation risk score	Internal use for product development

The Scientist's Toolkit: Essential Research Reagents and Platforms

Successfully implementing these advanced approaches requires a specific set of tools and reagents. The following table details key solutions for researchers in this field.

Table 3: Essential Research Reagent Solutions for Skin Sensitization NAMs

Tool Category	Example	Function in Research
3D Reconstructed Human Epidermis (RHE)	EpiSkin, Skin+ [66]	Provides a physiologically relevant platform for topical application and assessment of keratinocyte response (KE2); used in assays like SENS-IS and EpiSensA.
In Chemico Assay Kits	DPRA, kDPRA [2]	Quantifies a chemical's reactivity with synthetic peptides containing cysteine or lysine, directly measuring the Molecular Initiating Event (KE1).
Cell-Based Assay Kits	KeratinoSens, h-CLAT [79]	KeratinoSens measures Nrf2-dependent gene expression in keratinocytes (KE2). h-CLAT measures surface marker expression (CD86, CD54) in dendritic cells (KE3).
Genomic Profiling Tools	SENS-IS Assay Panel [66]	Identifies and quantifies changes in a panel of genomic biomarkers associated with the skin sensitization pathway, providing a mechanistic potency estimate.
Computational Platforms	OECD QSAR Toolbox, Bayesian Network Models [79] [81]	Integrates data from multiple sources (e.g., chemical structure, in vitro results) to predict hazard and potency using (Q)SAR and statistical models.
Toxicokinetic Modeling Tools	High-Throughput TK (httk) Models [80]	Predicts in vivo internal exposure concentrations based on in vitro bioactivity data, bridging the gap between in vitro assays and human risk.

The case studies presented herein provide compelling evidence that Defined Approaches and Next-Generation Risk Assessment are no longer theoretical concepts but are practical, validated tools for skin sensitization assessment. The success of these integrated strategies hinges on their foundation in the AOP framework, which allows for the systematic generation and interpretation of mechanistic data.

The validation of new 3D models like Skin+ expands laboratory options [66], while Bayesian networks and tiered NGRA frameworks demonstrate how combining in chemico, in vitro, and in silico data can reliably predict potency and assess risk for single substances and complex mixtures without animal data [79] [80]. Furthermore, data-driven approaches like the SSPM showcase the potential to leverage existing human data to minimize future testing [43].

For researchers and drug development professionals, the path forward involves the continued refinement of these methodologies, increased complexity of models (e.g., incorporating immune-competent organ-on-a-chip systems), and broader regulatory adoption. The collective progress in this field marks a significant advancement in validating in vitro models for immune responses, ensuring both human-relevant safety assessments and ethical scientific practice.

Conclusion

The validation of immune responses in in vitro skin sensitization models has matured significantly, moving from single-key event assays to sophisticated, integrated Defined Approaches and complex immunocompetent models that more accurately reflect human biology. The successful adoption of these New Approach Methodologies hinges on a robust validation framework that benchmarks performance against high-quality human and historical animal data. Future directions will focus on enhancing the physiological relevance of models to better capture immunoregulatory mechanisms, expanding their applicability to the most challenging substances like complex mixtures, and fully integrating these tools into next-generation, animal-free risk assessment paradigms. This evolution promises not only to ensure consumer safety but also to accelerate the development of safer products across the cosmetic, chemical, and pharmaceutical industries.

References

• 1. Models of Dendritic Cells to Assess Skin Sensitization - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC8971372/]
• 2. In Vitro Skin Models for Skin Sensitisation: Challenges and ... [https://www.mdpi.com/2079-9284/12/4/173]
• 3. Covalent TCR-peptide-MHC interactions induce T cell ... [https://www.nature.com/articles/s41467-022-32692-4]
• 4. Covalent TCR-peptide-MHC interactions induce T cell ... [https://pubmed.ncbi.nlm.nih.gov/35999236/]
• 5. Common T Cell Activation Methods [https://www.akadeum.com/t-cell/activation/methods/?srsltid=AfmBOoq7aSLKd25AUL9uk2ZtzRWZO6IDRPMhBDAgpj35KHMY7Gmw36ZG]
• 6. Adverse Outcome Pathways [https://www.oecd.org/en/topics/sub-issues/testing-of-chemicals/adverse-outcome-pathways.html]
• 7. Adverse Outcome Pathways | US EPA [https://www.epa.gov/chemical-research/adverse-outcome-pathways]
• 8. The Adverse Outcome Pathway: A Multifaceted Framework ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5906804/]
• 9. Comparison and validation of an in vitro skin sensitization ... [https://www.sciencedirect.com/science/article/abs/pii/S0887233317301303]
• 10. The Adverse Outcome Pathway for Skin Sensitisation ... [https://www.oecd.org/en/publications/the-adverse-outcome-pathway-for-skin-sensitisation-initiated-by-covalent-binding-to-proteins_9789264221444-en.html]
• 11. In vitro testing strategy for assessing the skin sensitizing ... [https://pubmed.ncbi.nlm.nih.gov/32001296/]
• 12. Stem cell liver model reproduces rare immune drug reactions [https://www.drugtargetreview.com/news/189683/stem-cell-liver-model-reproduces-rare-immune-drug-reactions/]
• 13. Boosting the AOP Framework with AI, First-ever NAM to be ... [https://www.proanima.fr/en/blog/boosting-the-aop-framework-with-ai/]
• 14. Methods integrating innate and adaptive immune ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1584852/full]
• 15. Sensitization in Transplantation Assessment of Risk 2025 ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12331380/]
• 16. Multiplexed detection and isolation of viable low-frequency ... [https://www.nature.com/articles/s41598-020-71750-z]
• 17. In vitro detection and characterization of drug ... [https://pubmed.ncbi.nlm.nih.gov/19796221/]
• 18. Intracellular Cytokine Staining Assays - Functional ... [https://kcasbio.com/blogs/intracellular-cytokine-staining-assays-functional-assessment-using-flow-cytometry/]
• 19. Immunogenicity in Clinical Practice and Drug Development [https://pmc.ncbi.nlm.nih.gov/articles/PMC7070797/]
• 20. New approach methodologies (NAMs): identifying and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10964841/]
• 21. FDA Announces Plan to Phase Out Animal Testing ... [https://www.fda.gov/news-events/press-announcements/fda-announces-plan-phase-out-animal-testing-requirement-monoclonal-antibodies-and-other-drugs]
• 22. Analysis of Publically Available Skin Sensitization Data from ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5546098/]
• 23. New Approach Methodologies – Where are we, what ... [https://www.understandinganimalresearch.org.uk/news/new-approach-methodologies-where-are-we-what-opportunities-are-we-currently-pursuing-and-what-risks-must-we-avoid]
• 24. New Approach Methodologies (NAMs): Advances in ... [https://www.axionbiosystems.com/new-approach-methodologies-nams-advances-preclinical-safety-studies]
• 25. Alternative Methods for Skin-Sensitization Assessment - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9783525/]
• 26. New Horizons in Skin Sensitization Assessment of ... [https://www.mdpi.com/2305-6304/13/8/693]
• 27. A hair-follicle reconstructed in vitro immunocompetent skin ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12454623/]
• 28. Application of an in vitro reconstructed human skin ... [https://www.sciencedirect.com/science/article/abs/pii/S0378427425026840]
• 29. Incorporating integrated testing strategy (ITSv1) defined ... [https://www.sciencedirect.com/science/article/abs/pii/S0273230023000260]
• 30. Defined Approaches to Identify Potential Skin Sensitizers [https://ntp.niehs.nih.gov/whatwestudy/niceatm/test-method-evaluations/skin-sens/da]
• 31. Integrated Testing Strategies for Skin Sensitization Hazard ... [https://www.mdpi.com/2079-9284/3/2/16]
• 32. Integrated Testing Strategies for Skin Sensitization [https://iivs.org/assays/sensitization/integrated-testing-strategies/]
• 33. Prediction of skin sensitization potency using machine ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5435511/]
• 34. A hair-follicle reconstructed in vitro immunocompetent for... skin model [https://link.springer.com/article/10.1007/s00204-025-04130-z]
• 35. Generating Immunocompetent 3-Dimensional Full- ... [https://www.sciencedirect.com/science/article/pii/S0022202X25035146]
• 36. Frontiers | From 1D microbiological assays to 3 ... D advanced skin [https://www.frontiersin.org/journals/cellular-and-infection-microbiology/articles/10.3389/fcimb.2025.1676663/full]
• 37. Requirements for Alternative In Vitro and In Silico Skin Models ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12318915/]
• 38. An R-based predictive model for skin-sensitizing potential ... [https://www.sciencedirect.com/science/article/abs/pii/S2468111325000106]
• 39. SARA-ICE: Skin Sensitization Risk Assessment [https://ntp.niehs.nih.gov/whatwestudy/niceatm/test-method-evaluations/skin-sens/da/SARA-ICE]
• 40. New Horizons in Skin Sensitization Assessment of Complex ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12390330/]
• 41. A botanical reference set illustrating a weight of evidence ... [https://www.sciencedirect.com/science/article/abs/pii/S0278691523008165]
• 42. Grouping of UVCB Substances with Dose-Response ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9344966/]
• 43. The skin sensitization prediction model: an algorithm for ... [https://www.sciencedirect.com/science/article/abs/pii/S0273230025001734]
• 44. GARDâ„¢skin Models for Identifying Sensitizers [https://ntp.niehs.nih.gov/whatwestudy/niceatm/test-method-evaluations/skin-sens/gard]
• 45. The borderline range of toxicological methods [https://pubmed.ncbi.nlm.nih.gov/28230889/]
• 46. Determination of U-SENSâ„¢ borderline range thresholds to ... [https://pubmed.ncbi.nlm.nih.gov/40413735/]
• 47. Prediction of skin sensitization potency sub-categories ... [https://www.sciencedirect.com/science/article/pii/S088723331730245X]
• 48. The Skin Sensitisation of Cosmetic Ingredients [https://pmc.ncbi.nlm.nih.gov/articles/PMC10146005/]
• 49. In chemico evaluation of prohapten skin sensitizers ... [https://www.sciencedirect.com/science/article/abs/pii/S0378427413000775]
• 50. In vitro assessment of skin sensitization, irritability and ... [https://www.nature.com/articles/s41598-022-08441-4]
• 51. What is New in Contact Allergy To Cosmetics ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12552389/]
• 52. Labeling of 56 additional allergens in cosmetic products [https://www.bisafe.si/en/the-eu-requires-the-labeling-of-56-additional-allergens-in-cosmetic-products/]
• 53. Application of an in vitro reconstructed human skin on ... [https://pubmed.ncbi.nlm.nih.gov/33053260/]
• 54. Evaluation of the Global Performance of Eight In Silico Skin ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11316520/]
• 55. Consensus model for skin sensitization assessment using ... [https://www.sciencedirect.com/science/article/pii/S2468111325000088]
• 56. In chemico, in vitro and in vivo comparison of the skin ... [https://pubmed.ncbi.nlm.nih.gov/28958912/]
• 57. Rational cell culture optimization enhances experimental ... [https://www.nature.com/articles/s41598-018-21050-4]
• 58. Dynamic co‐culture metabolic models reveal the fermentation ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7971941/]
• 59. Johns Hopkins Researchers Show Novel Immune System ... [https://www.hopkinsmedicine.org/news/newsroom/news-releases/2025/09/johns-hopkins-researchers-show-novel-immune-system-boost-helps-fight-cancer-cells]
• 60. Predictive identification of human skin sensitization ... [https://pubmed.ncbi.nlm.nih.gov/16283904/]
• 61. Skin sensitization: Uncertainties, challenges, and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6587935/]
• 62. Estimating uncertainty in LLNA EC3 data and its impact on ... [https://www.sciencedirect.com/science/article/abs/pii/S0273230023000259]
• 63. What is gold standard and what is ground truth? - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC4296658/]
• 64. Gold standard (test) [https://en.wikipedia.org/wiki/Gold_standard_(test)]
• 65. Use of in vitro methods combined with in silico analysis to ... [https://www.frontiersin.org/journals/toxicology/articles/10.3389/ftox.2024.1321857/full]
• 66. Validation of a new 3D epidermis model for the SENS-IS ... [https://pubmed.ncbi.nlm.nih.gov/40054703/]
• 67. An In Vitro Human Skin Test for Predicting ... [https://www.mdpi.com/2305-6304/12/6/401]
• 68. The inter-laboratory validation study of EpiSensA for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11503481/]
• 69. In Vitro Prediction of Skin-Sensitizing Potency Using the ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11435491/]
• 70. Application of Defined Approaches for Skin Sensitization to ... [https://www.frontiersin.org/journals/toxicology/articles/10.3389/ftox.2022.852856/full]
• 71. Skin Sensitization Testing—What's Next? - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6387141/]
• 72. NAMs: Scientific Foundations and Evolving Regulatory ... [https://www.dlrcgroup.com/nams-scientific-foundations-and-evolving-regulatory-excellence/]
• 73. Guidelines for the Testing of Chemicals [https://www.oecd.org/en/topics/sub-issues/testing-of-chemicals/test-guidelines.html]
• 74. Scientific guidelines | European Medicines Agency (EMA) [https://www.ema.europa.eu/en/human-regulatory-overview/research-development/scientific-guidelines]
• 75. OECD Releases 2025 Test Guideline Programme Updates [https://www.icapo.org/news/oecd-releases-2025-test-guideline-programme-updates]
• 76. Case studies on skin sensitization risk assessment [https://www.sciencedirect.com/science/article/abs/pii/S0273230025002107]
• 77. Implementing Alternative Methods [https://www.fda.gov/science-research/advancing-alternative-methods-fda/implementing-alternative-methods]
• 78. Next generation risk assessment and new approach ... [https://pubmed.ncbi.nlm.nih.gov/40490209/]
• 79. Skin Sensitization Testing Using New Approach ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12115617/]
• 80. A tiered next-generation risk assessment framework ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12198285/]
• 81. Advancing human health risk assessment: the role of new ... [https://www.frontiersin.org/journals/toxicology/articles/10.3389/ftox.2025.1632941/full]