Recent advances in immunocompetent human skin-on-chip models: Construction strategies and biomedical applications

Abstract

Human skin, as the body's largest organ and primary barrier, possesses a sophisticated immune microenvironment crucial for its defensive functions. Recent advancements in organ-on-a-chip technology have facilitated the development of biomimetic skin-on-chip models, especially in the accurate recapitulation of complex immune responses for human skin diseases. This review provides a comprehensive overview of the construction strategies and biomedical applications of immunocompetent human skin-on-chip models. We systematically detail the essential components for building these systems, encompassing cell sources, biomaterial matrices, and microfluidic fabrications. A major focus is placed on the strategic integration of key immune components, specifically immune cells such as Langerhans cells, T cells, and macrophages, as well as critical immune factors. These immunocompetent skin models serve as powerful platforms for disease modeling of inflammatory conditions like psoriasis, atopic dermatitis, and allergic contact dermatitis. Furthermore, we highlight their applications in drug development, toxicity testing, and cosmetic safety assessments. Finally, the future perspective and challenges underscoring the role of immunocompetent skin-on-chip technology in advancing precision medicine and dermatological research are discussed.

Graphical abstract

There is a growing need for review articles that systematically summarize the construction strategies and biomedical applications of immunocompetent human skin-on-chip (SoC) models. This review outlines the components for building immunocompetent SoC, including cell sources, matrices, and microfluidic fabrications, with a major focus on the strategic integration of key immune components. The representative applications of immunocompetent skin models are also summarized. Finally, we discuss the future perspectives and challenges in the field.

Highlights

• •Presents immunocompetent 3D skin-on-chip models integrating immune cells and cytokines.
• •Enables dynamic simulation of skin-immune interactions and inflammatory responses.
• •Demonstrates applications in modeling psoriasis, atopic dermatitis, and contact dermatitis.
• •Serves as a platform for immune-related drug screening and cosmetic safety assessment.
• •Advances standardized in vitro testing and clinical translation of skin models.

1. Introduction

As the largest organ of the human body, the skin plays a critical role in defending against external stimuli and transmitting sensory information, serving as the primary barrier against environmental damage, pathogens, and chemical agents [1]. The skin consists of multiple layers, which are the epidermis, dermis, and hypodermis. Each characterized by distinct cell types and extracellular matrix (ECM) components that work together to sustain barrier function, immune surveillance, and tissue homeostasis. Conventional two-dimensional (2D) cell cultures and animal models exhibit significant limitations in recapitulating the structural complexity and physiological functions of human skin, thereby impeding progress in dermatological research [2]. Furthermore, increasing ethical and regulatory restrictions, such as the European Union's complete ban on animal testing for cosmetics in 2013 and its goal to achieve zero animal testing by 2026, have underscored the urgent need to develop alternative testing platforms that more accurately mimic human biology [3].

In recent years, advances in organ-on-a-chip technology have facilitated the development of biomimetic platforms that integrate microfluidics, tissue engineering, and immunology to emulate human organ functions at a microscale level [4,5]. Over the past decade, substantial progress has been made in the development of skin-on-chip (SoC) models, evolving from simple 2D cultures to more complex three-dimensional (3D) full-thickness structures [6,7]. Commercial skin equivalents, such as EpiSkin™, SkinEthic™, and EpiDerm™, which are based on co-cultures of epidermal and dermal cells under air-liquid interface conditions, have been widely adopted for standardized testing of phototoxicity, corrosion, and irritation [[8], [9], [10], [11]]. However, these models remain functionally rudimentary, such as the lack of vascular perfusion and integrated immune components, thereby failing to mimic the dynamic immune responses critical to human skin [12,13].

A critical transition from simulating skin structure to reconstructing its physiological and pathological functions depends on the effective integration of immune capabilities. In SoC, this integration is not an additive feature but a fundamental prerequisite for transitioning from morphological coculture to functional verification of immune responses, as well as for replacing animal models and achieving clinical translation [[14], [15], [16], [17]]. This integration is inherently challenging, as it requires recapitulating the dynamic multicellular interactions among structural cells and immune cells, including complex processes such as cell migration, antigen presentation, and cytokine cascades [18,19]. Moreover, the immune microenvironment demands precise regulation across both the temporal dimension and the spatial dimension [20,21]. Ultimately, the incorporation of immune cells must result in validated immune functions at the single or multi-organ level, generating models that are not only structurally coexistent but functionally active.

The skin immune system is remarkably complex, comprising both innate and adaptive components. Key cellular actors include Langerhans cells (LCs), T cells, neutrophils, and macrophages, which orchestrate both innate and adaptive immune responses. Dysregulation of this system is a central pathological mechanism in diseases such as psoriasis and atopic dermatitis (AD) [22,23]. Therefore, the effective integration of immune components into SoC models has emerged as a key strategy for enhancing their physiological relevance and translational value [24]. This includes incorporating key immune cells and cytokines such as IL-17, TNF-α, and IL-4/IL-13 into in vitro skin structures. Microfluidic technology is utilized to simulate dynamic substance transport, immune cell responses, and migration. These immunocompetent SoCs demonstrate significant advantages in inflammatory skin disease modeling, immune-related drug screening, and cosmetic sensitization assessment.

This review provides a comprehensive overview of the construction strategies and biomedical applications of immune-competent human SoC models (Fig. 1). We begin by detailing the essential components for building such systems, including cell sources, dermal matrices, and fabrication strategies. We then examine various model types, ranging from 2D and 3D full-thickness constructs to vascularized platforms. The integration of immune cells and their functional validation in disease modeling, drug development, and cosmetic safety assessment is also highlighted. Finally, future trends and challenges for practical translations in the field are outlined.

Fig. 1.

A multi-level framework for the design and applications of immunocompetent human skin models.

2. Construction of human skin-on-chip models

Human skin, a complex organ, is primarily composed of three layers: the epidermis, dermis, and hypodermis [25,26]. The epidermis, the outermost layer, consists mainly of keratinocytes, melanocytes, and LCs, which play key roles in barrier function, pigmentation, and immune surveillance, respectively [26]. The epidermis is anchored to the underlying dermis via the basement membrane zone, a specialized structure primarily composed of collagen and other extracellular matrix (ECM) components [27]. This membrane regulates the exchange of nutrients, signaling molecules, and metabolic waste between the epidermis and dermis. However, due to its structural complexity and challenges in replication, the basement membrane is often underrepresented or simplified in current in vitro skin models [28]. The dermis, a dense connective tissue layer beneath the epidermis, is structurally complex and primarily composed of Type I and Type III collagen fibers, along with elastic fibers (elastin), which provide strength and elasticity. The main cellular component of the dermis is fibroblasts, which are responsible for synthesizing and maintaining the extracellular matrix [29]. Within the deeper regions of the dermis, a network of blood vessels, macrophages, and adipocytes can be found, contributing to nutrient supply, immune response, and thermoregulation.

The development of in vitro skin organ models that accurately recapitulate the structure and function of native skin is of significant importance for applications in drug development, toxicity testing, disease modeling, and regenerative medicine. This section focuses on the methodologies and strategies used to construct such skin organ models (Fig. 2).

Fig. 2.

The general approach for the construction of human SoC models.

2.1. Components of skin-on-chip models

2.1.1. Cell resource

Immortalized cell lines, characterized by their robust viability and straightforward culture techniques, are commonly utilized in the development of in vitro skin models. Table 1 offers an overview of the diverse cell sources employed in constructing SoC. The HaCaT cell line, an immortalized human keratinocyte line with good differentiation capability, is widely used in constructing skin organ models [30]. Compared to primary human keratinocytes, HaCaT cells exhibit aneuploidy [31]. Although HaCaT-based epidermal models are capable of expressing terminal differentiation markers such as keratin 10 (KRT10) and involucrin, they exhibit an aberrant differentiation program and fail to develop a structurally and functionally competent stratum corneum [32]. In contrast, the N/TERT-1 and N/TERT-2G cell lines, which maintain a diploid state, are suitable for gene editing [33]. This functional disparity is further reflected in barrier integrity: HaCaT-derived skin equivalents typically exhibit transepithelial electrical resistance (TEER) values in the range of 500–1000 Ω cm². In contrast, models generated from N/TERT-1 cells or primary human keratinocytes (PHKs) consistently achieve higher TEER values exceeding 1000 Ω cm², often reaching 1–5 kΩ cm². These elevated values are indicative of a mature, functional epidermal barrier. [2,34,35]. With respect to differentiation kinetics, N/TERT-1 cells display accelerated maturation, with peak expression of terminal differentiation genes occurring by day 4 post-calcium induction, followed by a modest decline by day 7. While HaCaT cells offer practical advantages in long-term culture stability and experimental reproducibility-particularly in assays requiring prolonged stimulation or repeated sampling. Notably, under pro-inflammatory stimulation (e.g., TNF-α or IL-17A), PHKs secrete robust levels of cytokines and chemokines, with IL-8 production commonly exceeding 20,000 pg per 10⁶ cells. HaCaT cells, by comparison, produce markedly lower amounts (4000–6000 pg per 10⁶ cells). N/TERT-1 cells exhibit inflammatory responses that closely recapitulate, and in certain contexts surpass, the magnitude and sensitivity observed in primary keratinocytes [30,33]. Research has shown that N/TERT-2G cells are more sensitive to oxidative stress, while N/TERT-1 cells display higher immune activity upon pathogen-associated molecular patterns (PAMPs) stimulation. These differences highlight the necessity of functionally validating cell lines before using them in disease models or skin substitutes. Furthermore, they underscore the importance of selecting appropriate cell lines based on specific research goals, such as studying oxidative stress or immune responses. [36]. NCTC 2544 is a commercially available epithelial-like cell line derived from normal human skin. It is characterized by its robust expression of cytochrome P450-dependent enzymatic activities and is widely used in allergy research [37].

Table 1.

Cell sources for constructing SoC models.

Cell Type		Source and key characteristics	Ref
Keratinocytes	HaCaT	Immortalized human keratinocytes; easy to culture, but aneuploid.	[30,31], [[38], [39], [40]]
	N/TERT-1	Diploid, immortalized keratinocytes; high immune response to PAMPs.	[33,36]
	N/TERT-2G	Diploid, immortalized keratinocytes; highly sensitive to oxidative stress.	[33,36], [41]
	Primary Keratinocytes	Primary cells from skin/circumcision; superior differentiation and barrier formation.	[42], [43]
	NCTC 2544	Epithelial-like cell line; high cytochrome P450 activity, used in allergy/toxicity studies.	[37], [44]
Fibroblasts	HSF	Primary human dermal fibroblasts; produce collagen/elastin, support skin structure and repair.	[[45], [46], [47]]
	Primary Fibroblasts	Primary fibroblasts from human skin; physiologically responsive, support 3D tissue modeling.	[38,42], [41,43]

The human primary fibroblast cell line HSF is currently a commonly used cell line, widely applied in the construction of 3D skin models [45], and in research on skin aging [46], and wound healing [47]. HSF cells possess a strong capacity for synthesizing collagen and elastin, which not only aids in repairing damaged skin tissue and promoting wound healing but also helps maintain skin elasticity and firmness. Additionally, HSF cells can release bioactive substances such as cytokines and growth factors, which regulate immune responses and promote tissue repair, playing a significant role in the anti-inflammatory and anti-infective properties of skin tissue.

The main cell lines used to construct skin organ models with immunological functionality are the THP-1 and MUTZ-3 cell line [48]. The THP-1 cell line, derived from human monocytic leukemia, can be induced under appropriate conditions to differentiate into cells resembling epidermal LCs, which play a crucial role in immune surveillance within the epidermis. The MUTZ-3 cell line, also derived from human bone marrow, is capable of differentiating into various hematopoietic cell types, including dendritic cells (DCs) and LCs, under suitable culture conditions. By inducing these monocytic cell lines to differentiate into epidermal LC-like cells and integrating them into skin organ models, researchers can better simulate the human skin's immune response mechanisms. This is crucial for understanding skin barrier function, inflammatory responses, and for evaluating the safety and efficacy of drugs and cosmetics.

Primary keratinocytes and primary fibroblasts are typically derived from clinically accessible human tissues, such as skin specimens obtained during plastic surgery procedures or neonatal circumcision. Under ethical approval and informed consent, these tissue samples can be processed using enzymatic digestion (such as trypsin or collagenase) to isolate primary cells with high viability and proliferative capacity. Compared to immortalized cell lines, such as HaCaT cells, primary keratinocytes retain biological characteristics that more closely resemble their in vivo counterparts, exhibiting superior differentiation potential. They are capable of undergoing ordered terminal differentiation and stratification under air-liquid interface (ALI) culture conditions, forming a multilayered, physiologically relevant epidermis with distinct basal, spinous, granular, and cornified layers, which are key features for establishing a functional skin barrier. Furthermore, primary cells respond to growth factors, extracellular matrix cues, and microenvironmental changes in a more physiologically accurate manner. Consequently, they enable more realistic modeling of skin behaviors, such as wound healing, inflammatory responses, and drug permeability. As a result, skin models constructed with primary cells offer enhanced physiological relevance and predictive accuracy in applications including tissue engineering, disease modeling, and drug testing. Although primary cells are subject to limitations such as donor-to-donor variability, a finite number of passages in vitro, and more demanding culture requirements, their ability to generate high-fidelity, functionally mature skin constructs makes them the preferred cell source for advanced in vitro skin research [42].

2.1.2. Dermal matrix

Currently, the dermal matrix commonly used in the construction of skin organ models includes decellularized matrices, natural polymers such as alginate, collagen, hyaluronic acid, and elastin, as well as synthetic polymers. The relevant components of the dermal matrix are indicated in Fig. 3 and Table 2.

Fig. 3.

Natural polymers are commonly used to develop bioinks with excellent cell compatibility. In contrast, synthetic polymers, including flexible hydrogel precursors and rigid thermoplastic materials, can be tailored to either formulate printable bioinks or construct mechanically robust three-dimensional scaffolds, depending on the application requirements.

Table 2.

Core matrix materials for SoC platforms.

Material	Key functions and applications	Mechanical properties	Main limitations	Ref
Collagen	Supports epidermal stratification and barrier modeling; ideal for inflammation and wound healing studies	Soft, viscoelastic (0.1-20 kPa)	Weak mechanics; batch variability; contracts over time	[50,55]
Fibrin	Enables vascular network formation and immune cell trafficking under flow; responsive to shear stress	Tunable gel (0.1-100 kPa range)	Short-term stability; requires thrombin	[51]
GelMA	Photocrosslinkable bioink for 3D co-culture; supports spatially controlled tissue architecture	Stiffness tunable (1-70 kPa) via UV	Potential UV/photoinitiator toxicity	[38,56]
dECM	Provides tissue-specific biochemical cues; best for disease-mimetic models	Soft, heterogeneous; mimics native dermis	Poor printability alone; donor-dependent	[49]
Alginate	Structural support in composite bioinks; useful for modular assembly or controlled release	Elastic, shape-stable (1–hundreds of kPa)	Bioinert unless modified	[57]
PEG	Highly tunable synthetic hydrogel platform; easily functionalized with peptides (e.g., RGD) for cell adhesion	Stiffness widely adjustable (0.1-100 kPa range) via crosslinking	Lacks inherent bioactivity; non-degradable unless modified	[58]
PCL	Forms durable scaffolds or perfusable channel frames in multi-layer chips	Rigid, strong (MPa range); long-term stable	Hydrophobic; lacks bioactivity	[[59], [60], [61]]
PLA	Used for high-strength, biodegradable structural components; suitable for long-term chip housing or porous scaffolds	Rigid and brittle (hundreds of MPa to GPa); degrades over months	Acidic degradation byproducts may affect cells; poor cell adhesion	[62]

Decellularized extracellular matrix (dECM) retains various types of collagen, glycosaminoglycans, and growth factors after cellular components are removed. By effectively mimicking the biochemical composition and structural characteristics of the natural dermal matrix, it serves as an ideal biologically derived scaffold material and 3D printed bio-ink [49].

In the development of artificial skin substitutes, both natural and synthetic polymers are widely used to construct three-dimensional scaffolds. Commonly used natural polymers include alginate, collagen (such as rat tail collagen), chitosan, fibrin, hyaluronic acid, and elastin. Synthetic polymers include poly(ethylene glycol) (PEG), polycaprolactone (PCL), polyvinyl alcohol (PVA), polylactic acid (PLA), and photocrosslinkable gelatin methacryloyl (GelMA).

Collagen, the most abundant structural protein in the skin's extracellular matrix, provides excellent bioactivity and cell affinity. Combining collagen with other natural or synthetic materials can significantly improve the scaffold's structural stability and mechanical properties [50].

Fibrin gels are also an ideal material for constructing 3D scaffolds. They exhibit excellent biocompatibility, enable efficient diffusion of nutrients and oxygen, and support deep cell growth and sustained metabolic activity. Furthermore, fibrin gels can respond to mechanical stimuli such as fluid shear stress, promoting vascular maturation. Additionally, precoating glass surfaces with fibrinogen enhances the adhesion and long-term stability of fibrin gels in microfluidic devices. Thus, by integrating biological functionality with engineering adaptability, fibrin serves as a superior matrix for engineering functional vascular networks within skin-on-chip platforms [51].

Natural polymers offer good biocompatibility and structural similarity to native ECM, promoting cell adhesion, proliferation, and differentiation. However, they typically exhibit weak mechanical strength and less controllable degradation rates. In contrast, synthetic polymers generally have lower biocompatibility and may produce toxic degradation byproducts, but their mechanical properties are highly tunable and structurally designable, enabling precise 3D fabrication [50,52]. Natural and synthetic polymers offer distinct benefits and drawbacks. By combining them in appropriate ratios, we can balance these properties. The resulting mixture retains the biocompatibility of natural materials. It also gains the mechanical strength of synthetic polymers. This combination is highly effective for building complex 3D skin tissues. In recent years, numerous studies have explored various combinations of natural and synthetic polymers for applications in skin tissue engineering, aiming to achieve an optimal balance between bioactivity and structural performance [53,54].

2.1.3. Processing method of SoC models

The chip body serves as the physical platform that houses the skin organ model and enables dynamic culture conditions, such as fluid flow, mechanical stimulation, and real-time monitoring. Common fabrication techniques for the chip body include soft lithography, 3D printing, and laser cutting, each offering distinct advantages in resolution, scalability, and material compatibility (see Table 3).

Table 3.

Comparison of fabrication techniques for SoC platforms.

Technique	Suitability for biological applications	Resolution and precision	Material compatibility and scalability	Ref
Soft Lithography	High-resolution microchannels, live imaging. Limited by PDMS absorption in drug screening studies.	≤10 μm	Material: PDMS (Gas permeable, optical clarity, but absorbs hydrophobic drugs). Scalability: Low. Mold-dependent, labor-intensive, poor for mass production	[63,65]
3D Printing	Custom 3D architectures. Surface roughness & material toxicity can affect cell viability.	20-200 μm (<20 μm with high-res SLA/DL)	Material: Diverse (Resins, PCL, PLA). Potential cytotoxicity of uncured monomers. Scalability: High. Rapid prototyping, digital replication, no mold required.	[64,66]
Laser Cutting	Low-cost, fast prototyping of simple 2D chips. Cannot mimic intricate skin microvasculature.	>100 μm, limited by laser spot size; not for micro-scale features	Material: Thermoplastics (PMMA, PET). Opaque/translucent. Scalability: High. Very fast, low cost, simple operation.	[67]
Microcontact Printing (μCP)	Patterning ECM proteins or adhesion cues on 2D/2.5D surfaces to guide cell organization	1-10 μm	Material: Stamps (PDMS) for ECM protein transfer. Scalability: Moderate. Fast stamping, but limited to surface modification.	[68,69]
Hot Embossing	Mass production of thermoplastic chips (PMMA) with high fidelity. High upfront mold cost limits academic use.	≤10 μm, high fidelity replication from master mold	Material: Thermoplastics (PMMA). Requires high temp/pressure. Scalability: High. Excellent for mass production, but high upfront mold cost.	[70,71]

Soft lithography using polydimethylsiloxane (PDMS) is the most widely used method for fabricating microfluidic chip bodies in research laboratories (Fig. 4A). This mold-dependent technique involves replicating microscale features from a rigid master mold composed typically of silicon or SU-8 photoresist onto PDMS. The process includes casting liquid PDMS onto the mold, curing it thermally, demolding, punching inlet/outlet holes, and bonding the PDMS layer to a glass or another PDMS substrate to seal the microchannels. Due to its high resolution (down to micrometer scale), optical transparency, and gas permeability, PDMS soft lithography is ideal for prototyping skin-on-chip devices with complex microchannel networks [63]. However, its reliance on PDMS introduces a critical limitation: the absorption of hydrophobic small molecules. This characteristic can significantly alter the concentration of drugs or signaling molecules in the culture medium, potentially confounding results in pharmacological screening or toxicity studies. Furthermore, the master mold dependency makes it less suitable for mass production.

Fig. 4.

Processing methods of SoC models. (A) Fabrication of PDMS via soft lithography technology. (B) General fabrication process of SoC models via laser cutting technology. (C) Micro-contact printing method for fabricating SoC models [72]. Copyright 2018 American Chemical Society. (D) Using 3D printing technology: bioprinting of skin tissue models and 3D printing of the skin on chip main body using polymer plastics.

3D printing enables rapid, design-flexible fabrication of chip bodies without the need for a master mold. It operates on a material deposition principle, building the chip layer-by-layer using thermoplastic polymers (such as resins and PLA) or photocurable materials. This method is particularly useful for creating multi-layered or integrated fluidic systems with customized geometries. While resolution is generally lower than soft lithography, recent advances in high-resolution stereolithography (SLA) and digital light processing (DLP) have significantly improved the precision of 3D-printed microfluidic devices [64]. However, even with high-resolution SLA, surface roughness and the potential cytotoxicity of uncured resins or photoinitiators remain concerns for long-term cell culture.

Laser cutting is a cost-effective and operationally simple method for fabricating chip bodies, especially for larger-scale features (Fig. 4B). It employs a focused laser beam to remove material from polymer sheets (e.g., poly (methyl methacrylate) (PMMA), polyethylene terephthalate (PET) ), engraving microchannels and chambers through localized vaporization. The cut layers are then bonded to a transparent cover to form enclosed channels. Although laser cutting lacks the fine resolution required for sub-100 μm structures, it is well-suited for rapid prototyping and educational or low-cost applications. While laser cutting offers distinct advantages in terms of cost-effectiveness and operational simplicity, its resolution is fundamentally constrained by the laser spot size, typically limiting feature dimensions to greater than 100 μm. Consequently, this technique lacks the microscale fidelity required to recapitulate the intricate microvasculature of human skin. As a result, laser cutting is predominantly relegated to the fabrication of macro-scale perfusion systems or educational models, rather than serving as a viable strategy for constructing sophisticated, physiologically relevant skin-on-chip platforms.

Despite the enduring dominance of soft lithography in academic settings, alternative fabrication strategies are increasingly vital. Techniques such as microcontact printing, hot embossing, and 3D printing provide critical adaptability, enabling the selection of optimal manufacturing routes based on throughput demands, cost-efficiency, and specific architectural needs.

The biological component of a skin-on-chip device involves constructing a 3D skin model that mimics the epidermal and dermal layers, including cellular organization, ECM composition, and barrier function. Two primary strategies are employed: scaffold-based tissue assembly and bioprinting (Fig. 4D).

In traditional approaches, skin models are constructed by seeding human keratinocytes and fibroblasts onto porous scaffolds or hydrogel matrices, such as collagen, fibrin, or decellularized ECM. These cells are cultured under ALI conditions to induce epidermal stratification and cornification, resulting in a multilayered, barrier-functional epidermis. This method is widely used due to its reproducibility and compatibility with standard cell culture protocols.

Bioprinting represents a more advanced and precise method for skin model fabrication. It involves the spatially controlled deposition of bioinks, which are hydrogel-cell mixtures, using inkjet, extrusion, or laser-assisted printing technologies. Common bioinks include GelMA, extracellular matrix, and PEG-based materials, which provide structural support and biochemical cues for cell survival and differentiation [73]. Bioprinting allows for the layer-by-layer construction of stratified skin equivalents, including vascular networks or immune cell integration, enabling the creation of highly biomimetic and patient-specific skin models.

In summary, the integration of microfabrication techniques for the chip body and advanced tissue engineering methods for the skin model enables the development of sophisticated skin organ-on-a-chip systems. The synergy between these engineering and biological domains drives innovation in skin research, offering powerful platforms for drug screening, toxicity testing, disease modeling, and regenerative medicine.

2.2. 2D skin-on-chip models

2D cell cultures are the most common method for constructing skin models. In these models, the epidermis is typically represented by a monolayer of keratinocytes, enabling studies on skin permeability and compound screening. The high optical clarity and experimental simplicity of 2D cultures have greatly facilitated our understanding of cellular behavior and function. For example, Valentina et al. seeded NCTC 2544 cells into multi-well plates and co-cultured them with THP-1-derived DCs to investigate the interactions between keratinocytes and DCs under stimulation by different allergens. Their results showed that keratinocytes promote the full maturation of DCs in the presence of moderate and weak allergens, whereas strong allergens alone can induce complete DC maturation independently, without requiring keratinocyte involvement [44].

2D culture, although simple to operate and easy to observe, fails to accurately replicate the 3D physiological microenvironment that cells experience in vivo. This limitation significantly weakens the complex interactions between cells and between cells and the extracellular matrix, thereby affecting cell morphology, function, and signaling behavior. For instance, keratinocytes cultured in 2D often appear flattened rather than forming natural stratified structures, while fibroblasts' migration ability and matrix secretion functions are limited due to the lack of a 3D support structure.

2.3. 3D full-thickness skin-on-chip models

To overcome the limitations of 2D skin models, researchers have increasingly shifted their focus toward developing more physiologically relevant 3D full-thickness skin models. These models utilize biocompatible materials such as polymer scaffolds or hydrogels to mimic the mechanical support and physiological environment of the dermis. Keratinocytes are seeded on the surface of the dermal layer, self-organizing into a multilayered epidermal structure, thus faithfully reproducing the layered architecture of the epidermis and dermis. Researchers embed fibroblasts, adipocytes, melanocytes, among others, within the hydrogel network to simulate dynamic interactions between cells and the matrix in the dermis. Gijs et al. developed a more structurally complex 3D epidermal model based on a 2D system. They seeded a mixture of human primary fibroblasts and rat-tail collagen onto Transwell inserts and allowed it to stabilize. Subsequently, they seeded N/TERT-2G cells on top and induced their differentiation using an ALI culture method, resulting in a stratified, multi-layered epidermal structure. This 3D epidermal model not only exhibited good tissue integrity but also displayed a differentiation marker expression profile closer to that of native human skin, including key markers such as K10 and filaggrin. The researchers used this model to evaluate its immune responses to various skin-related bacteria and pathogens. The model was shown to sense microbial stimuli and release inflammatory cytokines and antimicrobial peptides, indicative of innate immune competence [41].

Zoio et al. built upon the foundation of multilayered epidermal reconstruction to advance model complexity. They developed a full-thickness skin model that incorporates not only keratinocytes and fibroblasts but also melanocytes, which are key contributors to skin physiology and pigmentation. They sequentially seeded primary fibroblasts, melanocytes, and keratinocytes onto porous scaffolds to achieve a layered architecture. This approach not only maintains cellular differentiation but also enhances extracellular matrix production—particularly collagen synthesis by fibroblasts, thereby better recapitulating the structural and functional properties of native skin [43]. Moreover, by tuning the porosity and mechanical properties of the hydrogel scaffold, the model can be adapted to mimic the altered microenvironments associated with pathological states such as scarring or tumor infiltration.

Due to its advantages such as high precision, precise control, and the ability to construct complex structures, 3D printing technology has been widely used in the construction of full-thickness skin models. In combination with microfluidic technology, researchers can engineer biomimetic vascular networks beneath the dermis to simulate nutrient delivery and waste removal, bridging the gap between in vitro models and in vivo physiology. Jin et al. proposed a method for constructing full-thickness skin models using 3D printing technology [38]. They used three different bioinks to create various skin structures. First, they printed vascular networks and scaffold structures using GelMA and human umbilical vein endothelial cells (HUVECs). Then, they constructed the dermal layer by mixing porcine acellular dermal matrix with human primary fibroblasts. Finally, they covered the dermal layer with GelMA hydrogel and seeded HaCaT cells on its surface. In vivo experiments demonstrated that the constructed full-thickness skin model effectively promoted angiogenesis. However, factors such as nozzle shape, nozzle diameter, and extrusion pressure during bioink extrusion can affect cell viability.

2.4. Skin-on-chip models under microfluidic control

Studies have shown that SoC models developed within microfluidic devices exhibit superior barrier functionality compared to traditional static two-dimensional culture systems, significantly enhancing the physiological relevance and overall quality of the skin model [2]. Microfluidic platforms provide continuous perfusion, ensuring a stable and sufficient supply of nutrients to the skin tissue while efficiently removing metabolic waste products. This dynamic environment helps maintain long-term cell viability and tissue homeostasis, preventing functional deterioration caused by nutrient depletion or the accumulation of toxic metabolites. Furthermore, the fluid flow promotes uniform distribution of oxygen and nutrients across the multilayered skin structure, supporting the normal proliferation and differentiation of keratinocytes, and thereby enhancing the integrity and functionality of the skin barrier. Rimal et al. developed a 3D vascularized dynamic skin model, in which dynamic flow culture was shown to restore tissue homeostasis by balancing the expression of matrix metalloproteinases (MMPs) and their inhibitors (TIMPs) and by modulating angiogenic activity (Fig. 5A). This led to improved skin barrier function. The researchers applied laser injury to both dynamically perfused and static skin models for comparative analysis. Their results demonstrated that under dynamic culture conditions, the wound site was rapidly infiltrated by keratinocytes, resulting in significantly accelerated wound healing compared to the static model [74,75].

Fig. 5.

Microfluidic control devices support the construction of SoC models and enable real-time monitoring. (A) Perfusion enhances the barrier function of the skin model in the organ-on-a-chip system [74]. Copyright 2021, Elsevier. (B) The LOC system enables stretch stimulation and real-time monitoring [76]. Copyright 2025, Elsevier.

Moreover, the shear stress generated by perfusion plays a significant role in the maturation of skin models. Human skin is continuously stretched and deformed by internal physiological processes, growth, and environmental influences. Microfluidic technology enables the application of mechanical stimulation to skin-on-a-chip models, along with real-time monitoring, allowing for more physiologically relevant in vitro studies [28,76].

Abdo et al. seeded human adult dermal fibroblasts into a chitosan-collagen composite hydrogel and cultured the construct in suspension between two Poly(octamethylene maleate (anhydride) citrate) (POMaC) wires on the Biowire II platform to engineer a dermal tissue (Fig. 5B). The team then induced fibrosis in the engineered tissue by adding TGF-β, mimicking scar formation during wound healing. By leveraging the mechanosensing capability of the POMaC wires, they continuously tracked the contractile forces generated during tissue contraction, enabling real-time, dynamic assessment of the skin construct's mechanical tensile properties. Using this platform, they further demonstrated the potential of Q-peptide to inhibit fibrotic responses, providing a valuable tool for screening anti-fibrotic therapeutics and studying skin repair mechanisms. However, the study was limited to a monoculture system using only dermal fibroblasts and did not incorporate keratinocytes, endothelial cells, or other relevant skin cell types. Consequently, a co-culture skin organ model with multilayered architecture and cellular crosstalk was not established, limiting its ability to fully recapitulate the physiological and pathological processes of full-thickness skin [76].

2.5. Vascularized skin-on-chip models

Vascularized skin organ models represent a significant advancement in skin tissue engineering and organ-on-a-chip technology, aiming to overcome the limitations of traditional avascular models by incorporating functional microvascular networks. The integration of blood vessels into skin constructs significantly enhances nutrient and oxygen delivery while facilitating waste removal. Furthermore, it enables physiologically relevant studies of immune cell recruitment, transendothelial migration, inflammatory gradient formation, and spatiotemporal drug distribution, processes that are critically dependent on a functional vascular compartment.

Vascularized SoC models are primarily constructed through three complementary strategies. The first involves pre-fabricated vascular channels created via soft lithography or sacrificial templates, which enable perfusable endothelial networks for studying immune cell transmigration and drug transport. The second strategy employs membrane- or compartment-based designs using porous membranes or side-by-side microfluidic chambers; these establish physiologically relevant tissue interfaces and controllable chemokine gradients for modeling transepithelial and transendothelial migration. Finally, self-assembled vasculature is generated by embedding endothelial cells, frequently co-cultured with pericytes, directly into dermal matrices to spontaneously form biomimetic, branched capillary-like networks, a process that is further advanced by 3D bioprinting. Pre-fabricated channels offer experimental control and seamless integration with microfluidic systems. However, they often lack the architectural complexity of native vasculature. In contrast, self-assembly better recapitulates the physiological structure, distribution, and functionality of skin microvessels [77]. The future of vascularized skin modeling lies in strategically combining these methods to achieve full-thickness, dynamically perfused, and immunologically competent skin equivalents with high physiological fidelity.

First, pre-fabricated vascular channels, created via soft lithography or sacrificial templates, enable the formation of perfusable endothelial networks that recapitulate key immune functions. Sun et al. (Fig. 6A) employed soft lithography and injection molding techniques to fabricate a grid of microchannels within the dermal layer. They then seeded primary human dermal microvascular endothelial cells (HDMECs) into these channels to form functional microvascular networks. By perfusing neutrophils through the vascular lumen, they demonstrated that upon HSV infection, neutrophils could transmigrate across the endothelial layer and infiltrate the dermal compartment. This behavior effectively mimics the in vivo inflammatory response, indicating that the engineered vessel wall functions as a dynamic signaling platform responsive to inflammatory cues [78]. Similarly, Mori et al. pre-embedded nylon sutures within the dermal matrix to create vascular channels. The two ends of each channel were connected to a perfusion system composed of a peristaltic pump and silicone tubing, allowing continuous flow through the channels (Fig. 6B) [79]. They cultured skin equivalents under perfusion conditions and found that the cell density in the perfused skin chips was significantly higher than in non-perfused controls. Furthermore, their study demonstrated the feasibility of using these engineered vascular channels as a model for studying vascular absorption and transport. Sacrificial printing and DLP are other methods for pre-embedding vascular structures [80,81]. Maggiotto et al. constructed a skin model with a biomimetic vascular network using 3D bioprinting technology and a GelMA/Pluronic system via sacrificial printing. Different concentrations of GelMA were used for cell loading, while Pluronic F127 was prepared as a sacrificial material to construct vascular channels. This model successfully solved the problem of nutrient supply in thick tissue engineering and has been proven to be an ideal in vitro platform for studying drug permeation and wound healing mechanisms [80]. These approaches highlight the potential of pre-engineered vascular networks in enhancing the physiological relevance and longevity of skin organ models.

Fig. 6.

Vascularized SoC models. (A) Microvascular channels were fabricated to construct a vascularized 3D skin organ model using soft lithography and injection molding [78]. Copyright 2022, Springer Nature. (B) Microvascular channels were fabricated with nylon fibers pre-embedded to form vascular channels [79]. Copyright 2016, Elsevier Ltd. (C) The vascularized skin organ model was constructed by seeding vascular endothelial cells onto a porous membrane [39]. Copyright 2020, Wiley. (D) The vascularized skin organ model was constructed by seeding vascular endothelial cells onto the surface of the gel [40]. Copyright 2012, The Royal Society of Chemistry. (E) Self-formed vascularized skin organ-on-a-chip model [82]. Copyright 2005, American Society of Transplantation & American Society of Transplant Surgeons.

Second, membrane- or compartment-based designs utilize porous membranes (such as PET) or side-by-side microfluidic chambers to establish spatially defined tissue–vascular interfaces. Guo et al. employed dual-layer nuclepore membranes to co-culture epidermal, dermal, and vascular layers, enabling studies of barrier integrity and immune cell trafficking across both endothelial and epithelial barriers (Fig. 6C) [39,83]. In a lateral compartmentalized hydrogel model, Ren et al. reconstructed a skin-vascular interface by seeding HaCaT and HUVEC cells on opposite sides of a collagen matrix, then quantified CCL20-driven transendothelial and transepithelial migration of T lymphocytes under controlled chemokine gradients. This represents a process impossible to model in non-perfused, non-vascularized systems that lack directional flow and endothelial signaling (Fig. 6D) [40].

Third, self-assembled vasculature leverages the intrinsic angiogenic potential of endothelial cells (often co-cultured with pericytes) embedded directly within dermal matrices to form biomimetic, branched capillary-like networks (Fig. 6E) [82]. Baltazar et al. incorporated endothelial colony-forming cells and pericytes into a 3D-bioprinted dermal bioink, yielding grafts that supported in vivo perfusion and enhanced epidermal maturation [84]. Jones et al. further demonstrated that such self-organized microvessels, when integrated into a tri-layered microfluidic chip, enable dynamic studies of vascular permeability, leukocyte adhesion, and inflammatory mediator diffusion under flow, which are functions that static, avascular skin equivalents cannot replicate [85]. Similarly, Kyunghee et al. utilized a PDMS-based platform to co-culture endothelial cells and fibroblasts. They applied gravity-driven low shear stress to facilitate the self-assembly of a vascularized skin equivalent. This model further demonstrated functional immune responses. These included immune cell recruitment and cytokine signaling [86].

Critically, these vascularized platforms reveal a fundamental functional gap. Non-vascularized models lack the endothelial interface, hemodynamic shear stress, and transmural transport. Consequently, they are unable to support realistic immune cell recruitment, gradient-driven migration, and pharmacokinetic profiling. In contrast, engineered vasculature actively participates in immune regulation by expressing adhesion molecules, responding to cytokines, and establishing spatiotemporal chemokine landscapes. Thus, the integration of functional microvessels transforms SoC systems from passive tissue mimics into immunologically competent, physiologically predictive platforms for modeling inflammation, infection, and therapeutic intervention.

The integration of functional microvasculature has significantly advanced the physiological relevance of SoC models by enabling perfusion, endothelial barrier formation, and basic immune cell trafficking. However, the faithful recapitulation of skin immunity requires far more than a mere vascular conduit. The skin harbors a sophisticated, multi-layered immune ecosystem. This system comprises tissue-resident cells (such as LCs, dermal DCs, resident memory T cells, and mast cells) as well as dynamically recruited circulating leukocytes. All of these cells engage in spatially and temporally coordinated crosstalk with keratinocytes, fibroblasts, and endothelial cells [7].

However, as highlighted in the previous section, current vascularized SoC platforms, despite their ability to mimic vascular architecture, still face substantial hurdles in achieving true immunological functionality. A major limitation lies in the difficulty of maintaining long-term immune cell viability and phenotypic stability within engineered microenvironments [87]. This challenge arises from several interconnected factors. First, there is a scarcity of primary human immune subsets, such as LCs (LCs) or tissue-resident T cells, coupled with significant donor-to-donor variability. Furthermore, immortalized cell lines or iPSC-derived alternatives often exhibit incomplete functional maturation. Second, immune cells, particularly short-lived subsets like neutrophils, rapidly lose their activation markers and effector functions under static or suboptimal flow conditions. Third, existing models lack precise spatial control over immune cell positioning. This control is critical for recapitulating compartment-specific roles, such as epidermal antigen surveillance versus perivascular immune memory. Fourth, it is technically difficult to sustain physiologically relevant dynamic cues over extended culture periods. These cues include the low-shear blood flow necessary for key processes like leukocyte rolling, adhesion, and transendothelial migration. Finally, many current models fail to faithfully respond to the complex, time-dependent inflammatory signals that characterize real-world skin pathologies.

Collectively, these barriers prevent current platforms from progressing beyond structurally vascularized constructs toward truly immune-competent skin models capable of emulating the dynamic multicellular crosstalk inherent to native skin immunity.

3. Integration of immune components into human SoC models

The limitations of structurally vascularized SoC models underscore a crucial paradigm: recapitulating skin immune function demands more than perfusable endothelial networks; it requires the incorporation of functional immune elements capable of sensing, responding to, and orchestrating inflammatory cascades. Early biomimetic skin platforms successfully constructed basic stratified structures composed of keratinocytes and fibroblasts, playing an important role in evaluating physical barrier function. However, the fundamental limitation of these static models lies in the complete absence of an immune microenvironment. Due to the lack of resident or circulating immune cells, early platforms could not recapitulate the complex physiological signaling cascades between immune cells and skin structural cells, preventing researchers from accurately simulating immune-mediated dermatoses or investigating immunotoxicological mechanisms [83,88].

To overcome these challenges, the field of tissue engineering has progressively shifted its research focus toward the development of immunocompetent SoC models. This chapter systematically reviews the engineering strategies to build functional immune-competent skin models, organized around the complete stepwise workflow for in vitro model construction framework that simultaneously adheres to the functional hierarchy of the native skin immune system while aligning with the practical implementation path of tissue engineering research. We open with core immune input strategies, the foundational step for model construction: we first detail cytokine-driven stimulation to mimic inflammatory milieus, and then describe the integration of immune cell subsets stratified by their native functional roles: antigen-presenting LCs, innate immune effector macrophages and neutrophils, and adaptive immune T cells. [78]. And finally, discuss the rational selection of cell sources to balance biological authenticity and experimental standardization. We then outline the engineering of immune-permissive microenvironments, which provides the structural and biochemical niche required for immune cell residence and function. Next, we review the biophysical regulation of immune dynamics to recapitulate the spatiotemporal behaviors of immune cells in vivo. We conclude with a standardized framework for multidimensional characterization and validation of immunocompetent skin models. Collectively, these progressive strategies transform SoC platforms from passive structural mimics into physiologically predictive systems for modeling inflammation, infection, and therapeutic intervention, bridging the gap between skin tissue engineering and human immunology [89,90]. The stepwise architectural evolution of these platforms and their emerging translational trends are shown in Fig. 7.

Fig. 7.

(A) Architectural evolution of immunocompetent SoC platforms, (B) and the emerging trends in multi-organ chip integration for high-throughput detection.

3.1. Immune input strategies: from immune factors to cells

Reconstructing the lesion inflammatory microenvironment in vitro is the fundamental prerequisite for modeling specific dermatoses. Traditional bioengineered skin models primarily focus on structural aspects, which are insufficient to capture the intricate network of interacting cytokines that characterize in vivo inflammatory states. Consequently, contemporary engineering efforts have prioritized the strategic introduction of immune components to bridge the gap between morphological mimicry and functional immune pathophysiology [89].

These immune input strategies span a broad continuum of biological complexity, ranging from the addition of isolated immune factors to the integration of complete cellular populations. At the foundational level, researchers utilize soluble mediators to chemically simulate the consequences of inflammation, an approach that successfully circumvents the complex coculture conditions required for incorporating viable immune cells. Conversely, at the more advanced end of the spectrum, the direct integration of physical immune cell components is imperative to reconstitute active immune defense mechanisms and dynamic inflammatory cascades. The selection between these methodologies requires balancing high throughput standardization against the need for profound biomimetic fidelity.

The detailed engineering strategies, model systems, and validation benchmarks of representative studies are systematically summarized in Table 4. Based on their spatial configurations and biophysical dynamics, these platforms can be broadly classified into two main paradigms: static 3D skin organ models and dynamic SoC systems, each presenting distinct advantages and intrinsic limitations.

Table 4.

Summary of engineering strategies and validation benchmarks for immunocompetent SoCs.

Immune input	Spatial engineering	Biophysical dynamics	Validation benchmarks	Ref
IL-17	FTS RHE transwell (ALI)	Psoriatic hyperplasia and altered differentiation	DEFB4A/SPRR2C↑	[91]
rhIL-17A	Scaffold-free transwell (ALI)	Hyperproliferation and abnormal differentiation	Chemokines↑, FLG↓ (restored by Anti-IL-17A)	[92]
IL-22	LSE transwell	Psoriatic differentiation and acanthosis	Epidermal thickness/S100A7/S100A8/S100A9↑	[93]
IL-36γ	Collagen gel transwell (ALI)	Hyperkeratosis and SC thickening	IL8↑, SC thickness↑	[94]
TSLP	Collagen gel transwell (ALI)	Barrier dysfunction and impaired differentiation	IL33↑, FLG/S100A7↓	[95]
IL-17	Human SE transwell	Tight junction impairment and aberrant filaggrin	ZO1/CLDN1/CLDN4↓, FLG monomer↑, SC thickness↑	[96]
TNF-α	3-layer vertical PDMS chip	Dynamic perfusion, fluid leakage and edema	IL1β/IL6/IL8↑, ZO-1 protein disrupted, permeability↑	[83]
IFN-γ	Transwell SE	Keratinocyte differentiation and proliferation	KRT16/IVL/FLG↑, Ki-67+ cell ratio↓	[128]
TNF-α	3D-printed SoC (ALI)	Inflammation and perfused drug response	IL6/IL8↑, LDH release↓	[123]
TNF-α, IL-17A, IL-22	RHE transwell (ALI)	Psoriasis-like inflammation	S100A7/DEFB4/PI3/IL6/IL8↑, KRT10/LOR↓	[89]
IL-1β, TNF-α	Primary keratinocytes transwell	Psoriatic tight junction alteration	OCLN/ZO1 broad upregulation, CLDN1/CLDN7↓, late TER↓	[97]
IL-4, IL-13	3D HSE transwell (ALI)	Spongiosis and keratinocyte apoptosis	CA2/NELL2/FAS↑, pSTAT6↑, apoptosis rate↑, spongiosis↑	[90]
IL-4, IL-13, IL-31, TNF-a	LEM transwell (ALI)	AD-like features, spongiosis and impaired barrier	TSLP↑, KRT10/FLG/LOR↓, FFAs/ceramides↓	[98]
IL-4, IL-13	Follicle 3D SE (ALI)	AD inflammation and impaired differentiation	CCL26/IL13RA2/CA2↑, spongiosis↑, KRT1/LOR↓	[99]
IL-4, IL-13	Pumpless chip 3D SE (ALI)	Gravity flow barrier dysfunction	Cytokines↑, CA2↑, spongiosis↑, FLG/LOR/IVL↓	[100]
IL-4, IL-13, IL-22, IL-23	3D skin biopsies (ALI)	Innate barrier alteration and spongiosis	TLR7/TLR9↑, intercellular distance↑, TLR2/TLR4/CLDN1↓	[118]
IL-1β, IL-6, TNF-α	Microfluidic chip (pneumatic)	Gradient-dependent microglial dynamics	Distance-latency correlation, peak Ca2+ intensity↓	[125]
TNF-α, IL-1α, IL-6, IL-22	De-epidermized dermis (ALI)	Psoriatic features and pro-inflammatory genes	SKALP/elafin/hBD-2/CK16/IL-8/TNF-α↑	[127]
IL-4, IL-13	Collagen gel transwell (ALI)	AD-like barrier disruption	MMP1↑, ZO-1/OCLN/claudins/FLG/IVL/LOR protein↓	[129]
Th1 cells (P), Th17 cells (P)	FTS transwell (ALI)	Psoriatic inflammation and disturbed differentiation	DEFB4/PI3/LCE3A/IL-6/IL-8↑, FLG↓	[103]
T cells (P)	FTS transwell (ALI)	Burn injury and adaptive immunity	DEFB4/PI3/LCE3A/IL-6/IL-8↑, FLG↓	[104]
Th1 cells (P), Th17 cells. (P)	Full-thickness HSC (ALI)	T cell infiltration and psoriatic hyperplasia	Migration rate↑, CD69/K16/hBD2/IL-17A/IFN-γ↑, PPP6C↓	[110]
Vγ9Vδ2-T cells (P)	Melanoma-in-skin (ALI)	T cell infiltration and melanoma trogocytosis	4-1BB/PD-1/PD-L1/MCSP uptake↑, tumor viability↓	[111]
T cells (P), Th1 cells (P)	Bioprinted perfusable SoC	T cell tethering/rolling under shear stress	Th1 cell attachment/retention↑, migration length↑	[81]
TH1-polarized CD4+ T cells (P) IL-2	Fibroblast matrix transwell	Psoriatic thickening and parakeratosis	Elafin↑, FLG↓, S. aureus growth↓, Ki-67+ T cells ratio↑	[55]
MUTZ-LCs (C)	FTS collagen gel (ALI)	Allergen-induced LC maturation and migration	Dermal LC/IL-1β/CCR7/CD83↑, Epidermal LC↓	[102]
LCs (S)	Pigmented skin (ALI)	LC differentiation and epidermal integration	CD1a/Lag/MHC-II↑, Birbeck granules (+)	[115]
MUTZ-LCs (C) or MoLCs (P)	Full-thickness RHS (ALI)	LC maturation/migration in sensitization	IL-6/IL-8/CD83/CD274/CXCR4/ATF3↑, cell mobility↑	[48]
MUTZ-LCs (C)	FTS transwell (ALI)	Allergen/irritant-induced LC plasticity	Allergen: CXCL12-migration↑; Irritant: CCL5-migration↑	[112]
CD14−/+ precursors (S) CD14+CD16− monocytes (P)	RHE transwell (ALI)	In vivo LC development	CD1a+/CD207+ LC count↑, TGFB1/CSF2/IL34↑	[116]
MUTZ-3-LCs (C)	Collagen matrix (ALI)	LCs integration during stratification	Suprabasal LCs, CD1a/CCR6↑, Birbeck granules (+)	[117]
MUTZ-LCs (C)	FTS (De-epidermized dermis)	Homeostatic trans-dermal LC immigration	Epidermal LC colonization (+), CCL5/CCL20↑	[124]
mDCs (P)	Microfibre scaffold (ALI)	Barrier formation and sensitizer-induced migration	CD86/HLA-DR/impedance↑, CK10/CK14 (+)	[130]
Neutrophils (P)	V Vascularized FTS chip (ALI)	Gravity perfusion, HSV infection	IL-6/IL-8↑, trans-endothelial migration rate↑	[78]
Neutrophils (P)	3-lane organoplate (3D ECM)	Bidirectional flow, fMLP gradient	ICAM1/IL-8/CXCL1/CCL2↑, migration rate↑	[107]
Neutrophils (P)	3D organotypic skin (ALI)	Psoriatic thickening and impaired differentiation	Epidermal thickness↑, IL-17C/CSF3/IL-8/CCL20↑	[108]
Neutrophils (P)	PHKs transwell	PMN-PHK crosstalk (S. aureus)	PMN lifespan↑, MPO/LCN2/IL-1β/IL-8/CXCL10↑	[119]
TNF-α–stimulated Neutrophils (P)	NHDF-Neutrophil transwell	Neutrophil-induced collagen damage	Collagen-Ⅲ degradation↑, MMP-9/elastase↑	[126]
HL-60 cells (C)	Perfusable 3D vascularized SoC	Inflammation, drug toxicity, transmigration	IL6↑, transmigration rate↑, CD31/KRT5/FLG↑, permeability↓	[39]
Neutrophils (P)	Collagen hydrogel 3D culture	Chemotactic gradients, NETs formation	Chemotaxis↑, ROS/NETs↑, cell viability↑, apoptosis rate↓	[120]
iPSC-DC (S)	iPSC 3D skin transwell (ALI)	Sensitization and DC migration/maturation	Migration rate↑/CD86/CD209/HLA-DR/IL-8/IL-1β↑	[121]
iDCs (C)	iDC-integrated FTS (ALI)	Dermal DC activation and sensitization	CD54/CD80↑, IL-8/IL-6↑, p38 MAPK activation↑	[131]
Macrophages (P)	Collagen 3D SE (ALI)	3D migration and cytokine stimulation	Macrophage viability↓, IL-6/IL-8/CCR7↓, CD206↑	[105]
Monocytes macrophages (P)	Self-assembled SE (ALI)	SSc fibrogenesis and tissue stiffening	CD163/CD204/α-SMA/collagen↑	[106]
THP-1 macrophages (C)	Fibrin FTS (ALI)	UVA DNA damage, oxidative stress	γ-H2AX/ROS/inflammation level↑	[113]
Macrophages (C)	Bioprinted fibrin tissue	Immune microenvironment and inflammation	Cell viability↑, cytokine secretion capacity↑	[114]
Macrophages (M1 and M2) (C)	3-channel wound-on-chip	Tri-culture, early wound inflammation	IL-1β/IL-6/IL-8↑, vascularization level↑	[88]
Macrophages (M1 and M2) (P)	Fibrin FTS (ALI)	ECM deposition (TGF-β1/LPS)	Collagen/α-SMA/fibronectin↑, IL-6/TNF-α↑	[122]
Macrophages (P)	3D Collagen-I co-culture	Early melanoma microenvironment, ECM remodeling	Cell motility↑, TAM phenotype (Arg-1/CD206)↑	[132]
LC, DDCs (C)	Collagen FTS (ALI)	Sensitization, DC migration	CD1a+ cell migration↑, CD83↑, IL-6/IL-8/IL-12B↑	[109]
Macrophages, LCs, Neutrophils (S)	hiPSC vascularized organoids	Vasculogenesis and immune migration	Perifollicular vascularization↑, CD68/CD207/SOX2↑	[18]

Note: Cell sources are annotated as follows: (C) Cell line; (P) Primary cells; (S) Stem cell-derived cells. Models are categorized by their spatial configuration and dynamic stimulation methods.

Traditional static 3D skin organ models, predominantly Transwell-based skin equivalents and collagen hydrogel systems, remain the gold standard for structural biomimicry. The primary advantage of these models lies in their robust capability to support ALI cultures, which is crucial for achieving high-fidelity epidermal stratification, terminal differentiation, and barrier formation. Furthermore, their standardized formats allow for high-throughput screening and excellent inter-laboratory reproducibility. However, their main drawback is the lack of a dynamic microenvironment. The inherent static nature of these models limits long-term nutrient exchange, drives the accumulation of metabolic waste, and cannot faithfully recapitulate physiological fluid mechanical forces, including shear stress, which are essential for systemic immune cell recruitment and continuous circulation.

Conversely, Soc platforms address these biophysical limitations by incorporating microfluidic technology. The prominent advantage of SoC systems is the introduction of dynamic perfusion and precise control over spatiotemporal biochemical gradients. This enables the authentic simulation of vascular fluid leakage, localized immune cell tethering, rolling, and active trans-endothelial migration under physiological shear stress events that static models cannot reproduce. Despite these capabilities, SoC systems face significant challenges. They are often hindered by complex and expensive fabrication processes, lower throughput, susceptibility to bubble formation or leakage, and difficulties in achieving the same macroscopic tissue thickness and robust ALI stratification seen in classic Transwell assays [91,92].

Within this methodological evolution, leveraging exogenous soluble factors emerged as the earliest and remains the most accessible approach. By introducing targeted biochemical signals into the culture systems based on the immunopathogenesis of specific dermatoses, investigators can precisely drive the pathological remodeling of keratinocytes or fibroblasts to reflect disease signatures. This foundational strategy provides a highly controlled environment to dissect specific signaling pathways, establishing the bedrock for understanding cytokine-driven stimulation techniques.

3.1.1. Cytokine-driven stimulation

Cytokine-driven strategies currently represent the most prevalent and standardized engineering approach for establishing immune-functionalized skin models. The core rationale of this strategy lies in environmental biomimicry; specifically, it involves introducing exogenous cytokine cocktails into culture systems based on the immunopathogenesis of specific dermatoses. This reconstructs the lesion inflammatory microenvironment in vitro, thereby driving the pathological remodeling of keratinocytes or fibroblasts. By circumventing the complex co-culture conditions required for incorporating viable immune cells, this approach enables the precise in vitro recapitulation of a spectrum of key pathological phenotypes, ranging from dysregulated gene expression to histomorphological alterations.

Early investigations predominantly employed single-factor stimulation strategies, aimed at precisely dissecting the independent contributions of specific immune pathways to cutaneous pathology. In the context of psoriasis modeling, research has centered on the Th17/Th1 axis. Chiricozzi et al. demonstrated that stimulation with IL-17 alone is sufficient to induce a characteristic psoriatic gene signature in RHE, including the upregulation of defensins and CXCL8 [91]. Building on these findings, Singh et al. established a chronic IL-17 model that successfully validated the therapeutic efficacy of Secukinumab [92]. Furthermore, Harvey et al. utilized mass spectrometry imaging to reveal the remodeling effects of IL-22 on lipid metabolism and protein distribution [93], while Ainscough et al. elucidated the critical role of IL-36γ within the inflammatory amplification loop (Fig. 8A) [94]. Collectively, these studies have precisely delineated the specific functions of individual pathogenic factors within the psoriatic molecular network. Regarding AD models, single-factor strategies have proven equally effective in uncovering the molecular mechanisms underlying barrier impairment. Dai et al. discovered that TSLP directly drives filaggrin downregulation by inducing nuclear IL-33 expression and STAT3 phosphorylation [95]. Complementing this, Yuki et al. highlighted the role of IL-17 in disrupting tight junctions in AD via the downregulation of Claudin-1 (Fig. 8B) [96]. Additionally, in the realm of microfluidic applications, Wufuer et al. successfully induced acute inflammation and increased vascular permeability using TNF-α alone, thereby validating the chip's capacity to simulate edematous phenotypes (Fig. 8C) [83].

Fig. 8.

Single-factor stimulation strategies in skin pathology models. (A) Psoriasis models. Distinct roles of Th17/Th1 axis cytokines: IL-17 upregulates defensins and CXCL8; IL-22 remodels lipid and protein profiles; IL-36γ drives inflammatory amplification loops. (B) AD models. Barrier dysfunction mechanisms: TSLP downregulates filaggrin through nuclear IL-33 expression and STAT3 phosphorylation; IL-17 disrupts tight junctions via Claudin-1 downregulation. (C) Microfluidic chips. TNF-α stimulation recapitulates acute vascular inflammation with endothelial cell contraction and elevated permeability. Created with BioRender.com.

However, the in vivo inflammatory microenvironment is not orchestrated by isolated signaling events but rather by an intricate network of interacting cytokines. While single-factor stimulation suffices to induce specific alterations in gene expression, it remains limited in its ability to recapitulate complex, tissue-level morphological changes. To overcome this limitation, combinatorial cytokine strategies have been developed to leverage synergistic effects among factors, thereby simulating a more authentic pathological microenvironment (Fig. 9A).

Fig. 9.

Combinatorial cytokine strategies for recapitulating complex inflammatory skin microenvironments in vitro. (A) Synergistic multi-factor cytokine networks drive inflammation, hyperproliferation, and barrier disruption, beyond the minimal tissue responses elicited by single cytokines. (B) Psoriasis models. TNF-α, IL-17A, and IL-22 cytomix cooperatively activate NF-κB and JAK/STAT pathways, inducing disease hallmarks of acanthosis, parakeratosis, and tight junction disruption. (C) AD models. Th2 cytokines IL-4 and IL-13 act synergistically to drive disease-characteristic spongiosis and intercellular edema. (D) IL-4, IL-13, TNF-α, and IL-31 driven inflammatory environment exacerbates barrier dysfunction via disruption of lipid metabolism, with altered lipid structure and reduced ceramide levels. Created with BioRender.com.

In the context of psoriasis modeling, Todorovic et al. established the canonical cytomix strategy, incorporating TNF-α, IL-17A, and IL-22. This synergistic multi-factor challenge not only activated the NF-κB and JAK/STAT pathways but also successfully induced epidermal acanthosis and parakeratosis, representing macroscopic phenotypes that remain elusive under single-factor stimulation (Fig. 9B) [89]. Kirschner et al. further demonstrated that this complex inflammatory milieu precipitates the aberrant distribution of tight junction proteins within the basal layer, thereby recapitulating the structural disruptions characteristic of early-stage disease [97].

The superiority of combinatorial strategies is even more pronounced in AD models. Kamsteeg et al. demonstrated that the combined stimulation of IL-4 and IL-13 is requisite for inducing typical spongiosis in vitro, specifically, the intercellular edema and widened intercellular spaces that constitute the histopathological hallmark of AD (Fig. 9C) [90]. Building on this foundation, Danso et al. incorporated TNF-α and IL-31, employing lipidomics to elucidate the profound perturbations in long-chain fatty acid and ceramide metabolism driven by the composite inflammatory environment (Fig. 9D) [98]. Furthermore, studies by Emmert et al. and Kim et al. have expanded the complexity of these models by introducing additional variables. They investigated the differential responses to multi-factor stimulation arising from distinct cell sources [99] and the impact of cytokine cocktails on barrier repair dynamics within microfluidic chips [100], respectively, thereby broadening the scope and depth of this strategic approach.

Despite the proven efficacy of cytokine-driven strategies in dissecting molecular pathways and enabling high-throughput drug screening, they fundamentally represent a static form of chemical biomimicry. Such unidirectional input systems are devoid of the quintessential biological attributes of the immune response, namely, the real-time sensing of tissue injury, physical recruitment, and transendothelial migration. While artificially formulated cytokine cocktails can mimic the consequences of inflammation, they fail to recapitulate the dynamic orchestration of its occurrence. Consequently, to reconstitute active immune defense mechanisms and inflammatory cascades on-chip, the incorporation of physical immune cell components is imperative. This necessity sets the stage for the immune cell integration strategies, which will be discussed in detail in the following section.

3.1.2. Immune cell integration

As the predominant antigen-presenting cells in the epidermis, LCs serve not only as sentinels of cutaneous immune surveillance but also as the pivotal interface between innate and adaptive immunity. However, the construction of skin models possessing this functionality has long been hindered by the scarcity of suitable seed cells, as primary LCs are notoriously difficult to expand in vitro and are prone to spontaneous maturation, resulting in model instability [101]. established the engineering bedrock for this field by demonstrating that the human acute myeloid leukemia cell line, MUTZ-3, when driven by a cocktail of GM-CSF, TGF-β, and TNF-α, differentiates into Langerhans-like cells that maintain a stable, immature phenotype. This provided a reliable cellular source for standardized immune modeling (Fig. 10A). Building on this foundation, Ouwehand et al. [102] advanced the focus from mere cell survival to dynamic functional verification. In full-thickness skin models, they confirmed that integrated MUTZ-LCs possess complete biological activity; upon exposure to sensitizers such as nickel sulfate, these cells rapidly upregulate co-stimulatory molecules and initiate a CCR7-dependent migration program, traversing the basement membrane to enter the dermis. This series of studies successfully reconstructed the entire physiological cascade from cell acquisition to functional homing, in vitro recapitulating the skin surveillance system from antigen capture to migratory activation.

Fig. 10.

Modular assembly and integration of next-generation immunocompetent skin models. (A) LCs engineering. MUTZ-3 derived LCs achieve stable epidermal colonization. (B) Sequential assembly of T cells. Pre-polarized Th1 and Th17 subsets integrate chronologically after resident macrophages to recapitulate physiological immune cascades. (C) Macrophage niche establishment. Fibroblast-derived matrices support tissue-resident macrophage homeostasis. (D) Dynamic neutrophil recruitment. IL-8 gradients drive neutrophil transendothelial extravasation and protease-cytokine axis activation in microfluidic platforms. (E) Holistic integration. Fusion of skin and vascular organoids generates vascularized full-thickness skin models incorporating four distinct immune cell types, marked with purple, blue, green, and red indicators. Created with BioRender.com.

If LCs function as resident sensing units, T cells represent the core effectors that execute immune responses and drive complex inflammatory pathologies. Unlike the resident nature of LCs, strategies for engineering T cell integration prioritize mimicking the dynamic process of infiltration from the circulation into tissue, as well as the functional polarization of specific subsets. The seminal work of Van den Bogaard et al. elucidated the critical importance of functional pre-conditioning; they demonstrated that the mere introduction of resting T cells fails to elicit effective pathological features [103]. Only by pre-differentiating naive T cells into Th1 or Th17 effector subsets in vitro could the model drive keratinocytes to manifest the histological alterations typical of psoriasis (Fig. 10B). To further reconstruct the intricate immune cascade in vivo, Mulder et al. proposed a physiologically sequenced assembly strategy. This involved prioritizing the in situ differentiation of monocytes within the dermal scaffold into resident macrophages to establish an innate defense line, followed by the introduction of T cells only after tissue maturation to mimic the intervention of adaptive immunity. This sequential construction logic faithfully restores the inflammatory evolution mechanism from innate immune residency to adaptive immune response, marking a shift towards sophisticated biomimetic capacity in SoC systems for dissecting multi-cellular synergy [104].

As dynamic orchestrators of the dermal microenvironment, macrophages complete the functional continuum of skin models by bridging immune defense with tissue repair and remodeling. However, the pronounced phenotypic plasticity of macrophages renders them highly susceptible to microenvironmental cues, posing severe challenges regarding culture system compatibility and niche reconstruction [105]. identified a significant metabolic incompatibility within co-culture systems: while traditional epidermal differentiation media maintain barrier function, they significantly suppress the viability and inflammatory responsiveness of primary macrophages. This finding underscores that constructing immune-competent skin models cannot rely solely on physical cellular coexistence but must involve the development of composite culture systems that simultaneously accommodate epithelial differentiation and immune cell metabolic homeostasis. Beyond solving basal survival issues, reproducing specific functions under pathological conditions represents a higher-order engineering objective [106] adopted an in situ self-assembly strategy to replace traditional mature cell implantation. By introducing patient-derived CD14⁺ monocytes into dermal scaffolds, they allowed the cells to differentiate in situ into tissue-resident macrophages within the matrix constructed by fibroblasts. This strategy, leveraging endogenous microenvironmental signals rather than artificial induction, successfully recapitulated the macrophage-driven skin fibrosis cascade seen in systemic sclerosis, confirming that engineering models can only precisely dissect macrophage-mediated tissue remodeling mechanisms when the correct tissue niche is established (Fig. 10C).

As the rapid response units of the innate immune system, the introduction of neutrophils aims to reproduce the acute eruptive phase of skin inflammation. However, their evanescent half-life and extreme sensitivity to the microenvironment dictate that they cannot be maintained as long-term resident components like LCs or macrophages. Consequently, the core engineering strategy has shifted from static physical mixing to dynamic introduction based on chemotactic signaling. Sun et al., using a HSV infection model, demonstrated that IL-8 released by pathological skin tissue is the critical signal driving neutrophil integration. In this model, cells were not embedded initially; instead, neutrophils were delivered to the tissue interface via a fluidic phase mimicking blood circulation. These cells initiated transendothelial migration and traversed the basement membrane into the epidermis only upon sensing the virus-induced chemokine gradient. This on-demand recruitment strategy not only circumvents lifespan limitations but also recapitulates, at the molecular level, the biological process of immune cell infiltration from peripheral blood into tissue during acute inflammation (Fig. 10D). Once neutrophils successfully enter the tissue, their migration within the matrix is strictly governed by the extracellular environment [78]. Riddle et al. revealed the decisive influence of matrix composition on neutrophil function, finding that composite matrices rich in basement membrane extracts support directional traversal and survival significantly better than collagen alone. This proves that introducing neutrophils requires not just the cells themselves, but also a matched biochemical milieu to sustain their motility and mechanics. More critically, neutrophils in these models act not merely as phagocytes but as molecular amplifiers of the inflammatory storm [107]. Henry et al. elucidated this core mechanism, confirming that infiltrating neutrophils, through the release of serine proteases like elastase and cathepsin G, specifically cleave and activate keratinocyte-derived IL-36 family cytokine precursors. This neutrophil-driven protease-cytokine axis drastically amplifies local inflammatory signals, signifying that modern skin models have achieved the capability to dissect the complex molecular events governing the acute inflammatory burst [108].

While the introduction of single immune cell types confers specific effector functions to the model, it fails to reproduce the complex immune communication network across different anatomical layers of the skin. To transcend the limitations of single-cell models, engineering strategies have shifted from single-point reinforcement to the spatiotemporal synergistic integration of multi-lineage immune cells. Hölken et al. proposed a dual-compartment surrogate strategy based on cell lines to address the scarcity and batch variability of primary cells. This study achieved precise localization of MUTZ-3-derived LCs in the epidermis and THP-1-derived dendritic cells in the dermis within a full-thickness skin model. This layered assembly not only physically replicated the skin's vertical immune axis but also functionally confirmed that signaling crosstalk between epidermal and dermal immune cells significantly lowers the detection threshold for sensitization, revealing that a complete longitudinal communication link is essential for capturing early, faint inflammatory signals [109]. Further breaking the traditional paradigm of sequential addition, Mostina et al. explored a self-assembly strategy based on developmental biology principles. Rather than seeding mature immune cells, they utilized the fusion culture of vascular organoids differentiated from induced induced pluripotent stem cells (iPSC) with skin organoids. In this pluripotent microenvironment, the vascular organoids provided not only a vascular network but also served as a hematopoietic niche, inducing the co-differentiation and in situ colonization of endogenous immune cells. This strategy successfully realized the spontaneous generation and spatial distribution of various tissue-resident immune cells, including macrophages and LCs, demonstrating that leveraging the developmental potential of stem cells for holistic tissue induction is a transformative trajectory for realizing next-generation immune skin models characterized by high complexity and microenvironmental self-consistency (Fig. 10E) [18].

3.1.3. Strategic selection of cell sources

In the pursuit of engineering high-fidelity Immunocompetent skin equivalents (ICSEs), the integration of primary immune cells is widely recognized as the paramount strategy for recapitulating the intricate physiological heterogeneity and pathological microenvironments characteristic of human tissue. In contrast to immortalized or genetically modified cell lines, primary cells isolated directly from peripheral blood or tissues, including CD4⁺ T cells, monocytes, and macrophages, preserve the intact genetic background, specific immunophenotypes, and native reactivity to environmental stimuli of the donor. This superior biological fidelity facilitates the development of patient-specific disease models and ensures the precise replication of clinical phenotypes [110,111]. Current research substantiates the indispensable role of primary cells in reconstituting complex immune-epithelial interactions. For instance, Shin et al. successfully recapitulated epidermal acanthosis and the IL-17A/IL-22-mediated inflammatory cascade within 3D scaffolds by incorporating polarized Th1/Th17 cells derived from patients with psoriasis, thereby establishing the pivotal role of primary T cells in driving psoriatic histopathology [110]. Furthermore, work by van den Bogaard et al. demonstrated that the dynamic bidirectional crosstalk between primary T cells and keratinocytes is sufficient to independently induce characteristic inflammatory phenotypes, highlighting a level of intercellular regulation that remains unattainable in monoculture systems [103]. Moreover, the capacity for in situ differentiation of primary cells within 3D microenvironments endows these models with profound biomimetic attributes. Huang et al. leveraged CD14⁺ monocytes from patients with systemic sclerosis (SSc) to model the differentiation of monocytes into tissue-resident macrophages, their recruitment, and the subsequent induction of dermal fibrosis, thereby elucidating the cellular mechanisms underlying increased matrix stiffness [106]. Similarly, Mulder et al. utilized a burn skin model to confirm that co-cultured primary monocytes and T cells spontaneously synergize in response to microenvironmental cues, accurately reflecting the cascade of innate and adaptive immune responses during tissue injury and repair [104]. Notably, while the utilization of primary cells inevitably introduces donor variation, this heterogeneity faithfully mirrors the diversity of clinical populations, thereby significantly enhancing the predictive value of these models in drug screening and translational research [48].

Given the inherent limitations associated with primary cells, including donor scarcity, restricted expansion capacity, and significant batch-to-batch variation, human immortalized cell lines have emerged as a pivotal alternative strategy for constructing reproducible, high-throughput immunocompetent skin models owing to their infinite proliferative potential, high genetic uniformity, and standardized culture advantages. In the context of epidermal immune reconstitution, the human acute myeloid leukemia cell line MUTZ-3 has established an irreplaceable status as an in vitro surrogate model for LCs. It remains the only known cell line capable of differentiating into functional LCs expressing characteristic phenotypes, such as CD1a, Langerin, and Birbeck granules, under the induction of cytokines including GM-CSF, TNF-α, and TGF-β1 [101,102]. Current research demonstrates that MUTZ-3-derived LCs can not only colonize 3D epidermal models at physiological densities but also undergo specific migration and activation upon exposure to sensitizers like nickel and dinitrochlorobenzene (DNCB). This capability allows for the effective discrimination between chemical sensitizers and irritants, providing an in vitro alternative platform with the potential to meet OECD guidelines for skin sensitization assessment [48,102,112]. In parallel, the monocytic cell line THP-1 is frequently utilized as a standardized precursor for dermal macrophages or DCs and is widely applied in the simulation of the dermal immune microenvironment. For instance, Phuphanitcharoenkun et al. successfully recapitulated the oxidative stress response and upregulation of inflammatory factors in macrophages within a UVA photoaging microenvironment using a THP-1-incorporated immunocompetent dermal model, revealing the molecular association between DNA double-strand breaks and the inflammatory cascade [113]. Recently, a breakthrough in full-thickness immune integration was achieved by Hölken et al., who successfully introduced both MUTZ-3-derived epidermal LCs and THP-1-derived dermal DCs into a single model. This dual-immune cell model not only reproduced the two-tiered immune defense system of the skin but also captured the synergistic activation characteristics of both cell types under sensitizer stimulation, significantly enhancing the analytical resolution for complex immune responses [109]. Furthermore, in wound healing and organ-on-a-chip research, the incorporation of cell lines such as KG-1 and THP-1 has greatly simplified the simulation of inflammatory microenvironments and drug screening workflows [88,114]. In summary, although cell line-based immune skin models may possess slightly lower biological fidelity compared to primary models, their superior standardization advantages render them ideal tools for large-scale toxicological screening and fundamental mechanistic research.

To surmount the inherent limitations associated with the restricted expansion of primary cells and the biological deviations of immortalized cell lines, the exploitation of the differentiation potential of stem cells for constructing immune-competent skin models represents a frontier in regenerative medicine. Hematopoietic stem and progenitor cells (HSPCs) served as the initial seed cells for reconstituting the epidermal immune microenvironment. The pioneering work of Régnier et al. demonstrated that by introducing cord blood-derived CD34⁺ hematopoietic progenitors into a co-culture system, these cells could be successfully induced to differentiate into melanocytes localized in the basal layer and LCs positioned suprabasally. This study provided the first confirmation that the microenvironment supplied by keratinocytes is sufficient to support lineage commitment and integration of immune cells, effectively resolving the challenge of expanding primary LCs in vitro [115]. Schuster et al. further advanced this understanding by validating that CD207 negative hematopoietic precursors, upon implantation into full-thickness skin models, spontaneously differentiated into phenotypically mature LCs without the requirement for exogenous cytokine stimulation. Relying solely on endogenous signals within the skin mimetic, this niche-driven differentiation mode precisely recapitulated the physiological process of LC colonization of the epidermis during human embryonic development [116]. In recent years, breakthroughs in iPSCs technology have opened new avenues for constructing higher-dimensional skin organoids. Mostin et al. utilized the co-culture of human iPSC-derived vascular organoids with skin organoids to successfully construct vascularized skin organoids (VSKOs) possessing a complete vascular network. This model not only achieved the self-assembly of endothelial cells but also introduced a multi-lineage immune cell population, including macrophages, neutrophils, and dendritic cells, via the vascular organoids. This stem cell-driven synergistic development strategy established a complex system comprising blood, immune, and skin multi-tissue interactions in vitro for the first time, providing an unprecedented biomimetic platform for the study of systemic immune responses [18].

3.2. Engineering immune-permissive microenvironments

The primary challenge in constructing immunocompetent skin models transcends the mere maintenance of cell viability or epidermal differentiation; rather, it centers on the engineering of high-fidelity recapitulations of the complex immune-permissive microenvironment found in human skin. Conventional static culture systems are frequently devoid of essential physical spatial cues and signaling gradients, resulting in immune cells that remain in a stagnant or stochastically distributed state, thereby failing to mimic authentic immune surveillance and response processes. To endow these models with dynamic immune functionality, contemporary engineering strategies are shifting from simple tissue stratification toward the precise modulation of the microenvironment, aiming to provide immune cells with recognizable homing signals, traversable physical barriers, and sustainable resident niches.

The engineering of such microenvironments predominantly centers on the optimization of spatial physical architecture and the establishment of biochemical gradients. On the one hand, vertical compartmentalization within multilayered architectures is required to simulate the physiological segregation and semi-permeability between the epidermis and dermis, thereby facilitating the directional migration and information transfer of immune cells across distinct tissue hierarchies. On the other hand, the utilization of advanced biofabrication technologies, such as 3D bioprinting and microfluidic chips, enables the construction of vascular network topologies or the site-specific implantation of inflammatory foci at defined coordinates with micrometer-scale precision. These strategies serve not merely to replicate the morphological appearance of skin but to reconstitute, in vitro, the physical boundary conditions and spatiotemporal navigation systems that govern immune cell behavior.

3.2.1. Vertical compartmentalization & intercellular crosstalk

In the engineering of immunocompetent full-thickness skin models, the rationale underlying vertical stratified architecture transcends the mere accumulation of cellular layers; rather, it aims to establish strict tissue compartmentalization essential for regulating intercellular communication and trafficking. The physical segregation between the epidermis and dermis not only recapitulates the physiological function of the basement membrane but also serves as an obligatory checkpoint for the trans-tissue migration of immune cells. Research indicates that stratified culture of keratinocytes and fibroblasts on porous membranes facilitates the formation of a distinct tissue interface; this boundary permits the bidirectional permeation of soluble factors, such as interleukins and chemokines, while effectively restricting the admixture of non-migratory structural cells [83]. Within this architectural configuration, Langerhans cell precursors, such as those derived from MUTZ-3 cells, have been demonstrated to integrate into the epidermal layer, exhibiting dendritic morphology and distribution densities consistent with in vivo conditions. This integration is primarily attributed to the maintenance of local concentrations of critical factors, including TGF-β1, secreted by keratinocytes [101,117].

The crux of interlayer crosstalk lies in the establishment of signaling circuits that traverse physical barriers. Upon specific stimulation of the epidermal layer with agents such as allergens or irritants, the vertical compartmentalization compels resident immune cells to traverse the basement membrane mimetic into the dermal layer, a process that highly recapitulates the antigen-presenting migration pathway of LCs in vivo (Fig. 11A) [102,112]. Experimental data reveal that the effective induction of a cascade release of inflammatory mediators, such as IL-6 and IL-8, in the dermal layer following epidermal application of chemical sensitizers occurs only in the presence of intact vertical stratification and fibroblast support. This evidences that vertical compartmentalization constitutes the structural basis for maintaining the spatial organization of immune responses [48]. Furthermore, the incorporation of CD4⁺ T cells within the dermal layer substantiates the critical nature of this interlayer dialogue; T cell-secreted IL-4 and IL-13 penetrate the matrix to modulate the expression of epidermal differentiation proteins, resulting in altered barrier function. This bottom-up regulation relies on strict control of vertical distance to establish effective concentration gradients [90,118].

Fig. 11.

Engineering strategies for constructing immune-permissive microenvironments. (A) Vertical compartmentalization and intercellular crosstalk. Left: steady-state skin model with resident epidermal LCs. Middle: chemical sensitizers induce downward migration of LCs. Right: dermal T cell-derived cytokines drive a trans-layer positive feedback loop to induce epidermal acanthosis and parakeratosis. (B) Precision patterning of inflammatory foci. Left: 3D-bioprinted bacterial pathogens in homogeneous dermal matrix. Middle: chemokine gradients generated from patterned inflammatory foci. Right: targeted immune infiltration toward inflammatory nodes. (C) Topological design of perfusable vascular networks. Left: biomimetic microfluidic vascular network surrounding hair follicles. Middle: ICAM-1 mediated rolling and adhesion of circulating leukocytes under dynamic perfusion. Right: membrane-free neutrophil extravasation via cytoskeletal reorganization. (D) Biomimetic integration of endogenous immune niches. Left: scaffold-free self-assembly of iPSC-derived structural and immune precursors. Middle: endogenous ECM remodeling forms a biochemical signaling repository. Right: mature integrated niche in vascularized skin organoids with resident macrophages and LCs. Created with BioRender.com.

Further investigations have leveraged this stratified architecture to dissect complex immune feedback mechanisms. In models simulating psoriasis or atopic dermatitis, vertical compartmentalization enables researchers to delineate the independent contributions of immune cells to distinct tissue layers. For instance, the introduction of activated Th1/Th17 cells or their secretory factors into the underlying compartment allows for the direct observation of pathological features such as acanthosis and parakeratosis in the epidermal layer; conversely, these pathological alterations induce the release of elevated levels of chemokines, such as CCL20, by epidermal cells, thereby establishing a trans-layer positive feedback inflammatory loop [103,110]. This design not only recapitulates the spatial pathology of the disease but also elucidates the decisive role of vertical structure in constraining and directing the ordered dialogue between immune and tissue cells, ensuring the vectorial nature of the immune response rather than chaotic diffusion.

3.2.2. Precision patterning of inflammatory foci

The application of 3D bioprinting technology marks a paradigm shift in the construction of immunocompetent skin models, transitioning from homogeneous cell seeding to the era of precise spatial patterning. The preeminent advantage of this technology lies not merely in the fabrication of macroscopic tissue geometry but in its capacity to deposit bioinks and cellular components with micrometer-scale resolution, thereby facilitating the engineering of heterogeneous inflammatory foci or infectious nidi within an otherwise homogeneous tissue matrix. By manipulating the extrusion trajectory and volume, researchers can precisely deposit hydrogel microspheres laden with high concentrations of bacteria, such as Staphylococcus aureus, or viruses, such as HSV, at specific coordinates within the dermal scaffold. These localized pathogen burdens function as targeted beacons for immune chemotaxis, directing the focal accumulation of effector cells, including neutrophils, from the surrounding medium (Fig. 11B) [78,119].

The patterning of inflammatory foci provides an unprecedented instrumental toolkit for interrogating the complex spatial dynamics of immune cells. For instance, wound models fabricated utilizing fibrinogen-laden bioinks represent more than mere physical tissue defects, which function as active emitters of biochemical signals. The deposition of fibrin recapitulates the provisional clot microenvironment characteristic of early-stage wounding, and this specific matrix composition engages in distinct interactions with subsequently introduced macrophages, thereby driving their polarization toward pro-inflammatory or epidermal-repair phenotypes [88,114]. Such precise spatial orchestration enables the simultaneous co-existence of healthy and compromised tissue domains within a single construct. This configuration facilitates the observation of edge effects and dynamic infiltration processes at the health-disease interface, a capability that remains unattainable via conventional bulk seeding methodologies.

Furthermore, the utility of precision patterning extends to the establishment of gradients for immune-inductive factors. By bioprinting carriers releasing cytokines, such as TNF-α, or discrete islands of stromal cells secreting specific mediators at defined loci, long-term stable chemotactic gradients can be established within the model interior. This heterogeneous signal distribution mimics the incipient phases of localized cutaneous inflammation, such as acne or folliculitis, compelling immune cells to modulate their migration velocity and directionality in response to signal intensity [107]. Investigations reveal that neutrophils exhibit heightened sensitivity to these spatially patterned inflammatory cues; they not only migrate toward regions of high concentration but also adapt their protease secretion strategies in accordance with the matrix stiffness, such as collagen concentration, of the lesion, thereby achieving precise clearance and remodeling of the pathological site [108].

3.2.3. Topological design of perfusable vascular networks

In immunocompetent skin models, the pivotal objective of microfluidic technology is the construction of perfusable vascular networks exhibiting defined topological architectures, which dictate the spatial distribution and transport efficiency of immune cells and soluble factors. Diverging from simplistic endothelial monolayers, topological design prioritizes the recapitulation of the geometric characteristics intrinsic to microvascular networks, including branching angles, luminal diameter variations, and three-dimensional network density. State-of-the-art models have demonstrated the capacity to generate perifollicular vascular plexuses surrounding hair follicle structures; this highly specialized vascular topology maximizes the interfacial surface area between immune cells and the surrounding tissue, thereby significantly enhancing the extravasation efficiency of circulating immune cells such as neutrophils [81].

The topological attributes of the vascular network directly govern the spatial coverage and temporal kinetics of immune surveillance. Microvascular networks characterized by patent luminal structures and continuous endothelial linings facilitate sustained perfusion driven by peristaltic pumps, effectively mimicking the physiological transport of immune factors via the blood circulation. Investigations reveal that within chip-based models featuring complex vascular branching, endothelial cells exhibit superior efficacy in upregulating luminal adhesion molecules, such as ICAM-1, upon inflammatory stimulation, creating specific exit sites that guide leukocyte transmigration [39]. This topology-driven functional vascularization ensures that circulating T cells or neutrophils undergo physiological rolling, adhesion, and transmigration along the vascular trajectory rather than random sedimentation, thereby precisely recapitulating the spatial sequence of the leukocyte recruitment cascade (Fig. 11C) [97,120].

Furthermore, the microstructural engineering of the vessel wall, specifically the establishment of membrane-free interfaces allowing direct contact between endothelium and matrix, is critical for mediating immune cell penetration behaviors. Self-assembled vascular networks devoid of artificial membrane barriers eliminate physical impediments, enabling immune cells to directly mechanosense extravascular matrix stiffness and detect chemotactic signals. In viral infection models, this membrane-free topological design permits neutrophils to rapidly reorganize their cytoskeletal architecture upon sensing extravascular viral cues and squeeze through endothelial junctions into the interstitial space, a process strictly dependent on the seamless topological integration of the vascular network with the surrounding matrix [39,78]. Consequently, the engineering of vascular networks centers on the construction, via topological design, of a biological thoroughfare that immune cells can recognize and utilize for efficient trafficking.

3.2.4. Biomimetic integration of endogenous immune niches

In contrast to the physical incorporation of exogenous cellular components, the biomimetic integration of endogenous immune niches constitutes the apex of strategy in engineering immunocompetent skin models. This approach prioritizes the application of developmental biology principles to orchestrate the autonomous self-assembly of cells into a native-like immune microenvironment. Leveraging iPSC technology, researchers have successfully achieved the simultaneous differentiation of keratinocytes, fibroblasts, and dendritic cells within a unified system. During co-culture, these isogenic cell populations autonomously construct highly biomimetic tissue architectures, driven by the secretion of their own extracellular matrix and signaling molecules [121]. This fully autologous strategy circumvents allogeneic rejection, thereby enabling immune cells to naturally colonize specific anatomical locations, such as the epidermal basal layer or the papillary dermis, effectively establishing stable resident niches comparable to those observed in vivo.

The synergy of biochemical signaling during the self-assembly process is pivotal for the establishment of functional immune niches. In the absence of forced intervention by artificial scaffolds, fibroblasts and macrophages orchestrate the construction of microenvironments possessing specific mechanical properties and biochemical modifications through the secretion of endogenous ECM components, including collagen and fibronectin (Fig. 11D) [106]. This endogenous matrix serves not merely as a physical scaffold for immune cells but also functions as a reservoir for growth factors and cytokines, thereby establishing a dynamic signaling repository. For instance, in models of systemic sclerosis, the self-assembled dermal matrix faithfully preserves pro-fibrotic signals originating from macrophages and reciprocally activates newly recruited monocytes. This bidirectional interactive niche is exclusively generated through the process of autonomous environmental remodeling by the cells [122].

Furthermore, the advent of vascularized organoids has further refined the integration of endogenous immune niches. Through the co-culture of iPSC-derived vascular organoids with skin organoids, it is possible to observe the natural engraftment of immune cells into the skin tissue concomitantly with the process of angiogenesis. This developmental synchrony ensures that the distribution of immune cells within the tissue is not artificially pre-determined but rather adheres to the natural laws governing angiogenesis and tissue development [18]. Models generated via this biomimetic integration strategy exhibit LCs with a more mature phenotype and enhanced antigen capture capacity, as their maturation occurs within a continuously developing, signaling-complete native-like environment, rather than being mechanically introduced into pre-formed tissues [115].

3.3. Biophysical regulation of immune dynamics

The construction of physiologically relevant immunocompetent skin models relies not merely on static cellular assembly but critically on the incorporation of biophysical signals capable of driving immune behaviors. In the in vivo milieu, immune cells are not situated in stagnant culture media but are perpetually exposed to dynamic variations in hemodynamic shear stress, chemical concentration gradients, and matrix mechanical properties. These physical signals constitute the instructional language of the immune system, dictating cellular activation thresholds, migration directionality, and functional phenotypes. Consequently, the ultimate objective of engineering microenvironments is to transcend from structural mimicry to functional emulation, achieved by the application of precisely controllable physical stimuli to recapitulate the spatiotemporal dynamic responses of immune cells within tissues.

This mechanism of biophysical regulation is predominantly manifested across the three dimensions of fluid dynamics, chemical navigation, and solid mechanics. Fluid shear stress serves as the primary intravascular signal responsible for priming endothelial adhesion functions and initiating the preliminary steps of immune surveillance. Subsequently, complex chemokine gradients established within the interstitial space provide precise spatial navigational maps that orchestrate the trans-tissue migration of immune cells within specific temporal windows. Ultimately, the mechanical interaction between cells and the extracellular matrix forms a closed-loop feedback system wherein immune cells alter environmental physical attributes via matrix remodeling, while reciprocal changes in environmental stiffness determine cellular fate. The synergistic action of these three elements endows in vitro models with physiological vitality, thereby manifesting authentic immune dynamics.

3.3.1. Hemodynamic forces and endothelial activation

In conventional static culture systems, endothelial cells typically remain quiescent and lack the capacity to capture circulating immune cells. The introduction of a biomimetic hemodynamic environment is therefore crucial for activating this functionality. Extensive research indicates that fluid shear stress within microfluidic chips serves not merely as a driving force for nutrient transport but functions fundamentally as a mechanical signal directly acting upon the apical surface of vascular endothelial cells. Stimulation via physiological levels of shear stress significantly upregulates the expression of key adhesion molecules, including ICAM-1, VCAM-1, and E-selectin, on the endothelial surface. This mechanically induced activation state remains profoundly challenging to replicate under static conditions [39,107]. Experimental data demonstrate that vascular networks within skin-on-a-chip models cultured under perfusion exhibit heightened sensitivity to inflammatory mediators such as TNF-α. Consequently, endothelial cells undergo a rapid phenotypic transition from an anticoagulant to a proinflammatory state, establishing the requisite physical conditions for leukocyte capture [81,123].

A central mechanism of hemodynamics is the mediation of the immune cell rolling and adhesion cascade. Investigations into neutrophil and vascular endothelial interactions reveal that neutrophils manifest classical rolling behavior exclusively when driven by specific flow velocities. This dynamic engagement facilitates comprehensive mechanical probing and subsequent binding between cellular surface receptors and corresponding endothelial ligands [107]. Conversely, immune cells introduced under static conditions typically undergo direct gravitational settling, which fundamentally fails to recapitulate the selective recruitment processes inherent to physiological states. Furthermore, fluid shear stress modulates endothelial cytoskeletal rearrangement and regulates vascular wall permeability, thereby exerting strict control over the rate of transendothelial migration of immune cells [39].

Advanced investigations utilizing vascularized models have further elucidated the critical role of the mechanical environment in host defense against pathogens. In microfluidic models simulating Staphylococcus aureus or HSV infections, the continuous flow provided by perfusion systems effectively mimics systemic circulation. Concurrently, these flow dynamics enhance the directional extravasation capacity of neutrophils by clearing localized metabolic waste and sustaining critical chemokine concentration gradients [78,119]. Quantitative data indicate that under flow conditions, neutrophils demonstrate a markedly higher efficiency in recognizing IL-8 signals originating from localized infection sites. Subsequently, they traverse the endothelial barrier into the interstitial space. This phenomenon substantiates that hemodynamic force acts as an indispensable physical switch for initiating acute inflammatory responses.

3.3.2. Spatiotemporal navigation via chemotactic gradients

The precise tissue localization of immune cells does not arise from stochastic diffusion; rather, it fundamentally depends on spatiotemporally regulated chemokine gradients for directed navigation. The prerequisite for reconstructing these gradients within in vitro models involves recapitulating the in vivo dynamic equilibrium between physiological sources and sinks, which establishes the directional vectors essential for guiding cellular locomotion. Regarding LCs, their epidermal retention and subsequent emigration are governed by rigorous gradient regulation. Investigations confirm that establishing concentration gradients of CCL5 and CCL20 within the dermal compartment can induce MUTZ-3-derived precursor cells to undergo directional transmigration across basement membrane models into the epidermis. This process effectively simulates the physiological seeding of LCs precursors [112,124]. Such migration transcends rudimentary chemoattraction, representing instead an acute cellular computation of the surrounding environmental gradient slope.

The spatiotemporal dynamics of these gradients dictate the highly compartmentalized and phasic characteristics of the immune response. During the initial stages of sensitization, epidermal keratinocytes stimulated by chemical sensitizers, such as DNCB, downregulate adhesion molecules, including E-cadherin, while simultaneously altering their chemokine secretion profiles. This shift generates a novel chemotactic gradient oriented toward the dermal lymphatic vessels. This gradient inversion compels mature LCs to upregulate CCR7 receptors and undergo emigration from the epidermis to the dermis along the resulting CCL19/CCL21 gradient, thereby completing the foundational step of antigen presentation [48,102]. This gradient-dependent bidirectional shuttling mechanism remains completely unobservable in static co-culture models. Consequently, this demonstrates that dynamic gradient fields serve as the central navigational controllers dictating the ultimate directional fate of immune cells.

Furthermore, concerning the recruitment of adaptive immune populations, meticulously designed gradients enable highly specific subpopulation sorting. Within psoriasis models, gradients established by specific chemokines, such as CXCL10 and CCL20, secreted by keratinocytes, have successfully induced the targeted transmural infiltration of Th1 and Th17 cells. Crucially, irrelevant T cell subpopulations fail to migrate under these identical conditions [110]. State-of-the-art microfluidic technologies now employ programmable pneumatic valves to construct dynamically fluctuating calcium ion or cytokine waves at the micron scale. These advanced platforms permit the real-time observation of microglia or macrophages as they continuously realign their cytoskeletal polarization in response to instantaneous gradient perturbations, thereby elucidating the fundamental biophysical mechanisms underlying immune navigation [125].

3.3.3. Mechano-immunological feedback loops

The interplay between immune cells and the microenvironment transcends unidirectional adaptation to constitute a bidirectional remodeling mechanoimmune feedback loop. Upon recruitment to tissues, immune cells secrete proteases and remodeling factors that alter the physical attributes of the extracellular matrix, including stiffness and porosity. Concurrently, these modified biomechanical properties retroactively regulate immune cell functionality to establish a closed loop. A quintessential example involves the secretion of elastase and proteinase 3 by neutrophils at sites of inflammation. Beyond degrading collagen to induce matrix softening and structural damage, these enzymes fundamentally execute specific cleavage and activation of interleukin 36 family cytokines. This process converts inert precursors into highly active inflammatory mediators, thereby amplifying inflammatory signaling and establishing a positive feedback circuit [108,126].

This biomechanical feedback is particularly pronounced in the pathogenesis of chronic inflammation and fibrotic disorders. Within full-thickness skin models of systemic sclerosis, the crosstalk between macrophages and fibroblasts elucidates a stiffness-driven phenotypic mechanism. Activated macrophages secrete profibrotic factors such as transforming growth factor beta to induce excessive collagen synthesis by fibroblasts, resulting in significant stiffening of the dermal matrix. This elevated matrix stiffness has been demonstrated to function as a physical ligand that signals through mechanotransduction pathways to stably lock macrophages in a profibrotic phenotype. This prevents their transition toward a reparative state and ultimately sustains the chronic progression of the pathology [106,122].

The establishment of this feedback circuit indicates that in vitro models possess the capacity for autonomous evolution. Investigations into ultraviolet A (UVA)-induced photoaging models reveal that macrophages do not merely sense the light stimulus directly but rather detect the matrix fragmentation and loss of physical integrity resulting from photodamage. The subsequent collapse of this biomechanical environment triggers intracellular oxidative stress and DNA damage responses within the macrophages. This in turn provokes the release of additional matrix metalloproteinases, which further exacerbates the deterioration of the microenvironment [88,113]. This vicious cycle of destruction, perception, and subsequent secondary destruction faithfully recapitulates the tissue remodeling processes observed under in vivo pathological conditions. Consequently, it highlights the central role of biophysical mechanics in either maintaining or disrupting immune homeostasis.

3.4. Multidimensional characterization and validation benchmarks

The primary objective of constructing immunocompetent skin models is to obtain an in vitro surrogate system capable of accurately predicting clinical responses. Therefore, establishing a standardized validation framework is of paramount importance. This framework must evaluate not only the physical barrier integrity of the model but also quantify its pathological responsiveness under specific immune stimulation. This section elucidates the characterization parameters and acceptance benchmarks required for a qualified immune skin model across three dimensions: physical function, immunological lineage, and cellular dynamics.

3.4.1. Barrier integrity and physical metrics

The physical barrier serves as the most fundamental quality control metric for skin models and the primary readout for assessing inflammation-induced barrier disruption. As the nondestructive gold standard for evaluating tight junction integrity, transepithelial or transendothelial electrical resistance directly reflects the ionic permeability and compactness of the tissue interface.

In the validation of traditional static models, a qualified epidermal layer typically needs to maintain a high baseline resistance value (>400 Ω cm²) to demonstrate the formation of tight junctions in the stratum corneum and stratum granulosum. However, in inflammatory models, the dynamic variation of this resistance holds greater diagnostic significance than its absolute value. Yuki et al. established this dynamic criterion in their study of atopic dermatitis models, confirming that the downregulation of Claudin 1 expression induced by Th2 type cytokines exhibits high temporal synchrony with a precipitous decline in electrical resistance values. Consequently, a significant and sustained decrease in resistance is universally recognized as the biophysical benchmark for determining whether a model successfully recapitulates a barrier compromised phenotype [96].

For vascularized models constructed via microfluidics, the validation criteria for the physical barrier extend from simple occlusion to mechanical responsiveness. The research by Kwak et al. established an acceptance benchmark for vascular barriers based on fluid dynamics. They indicated that statically cultured endothelial cells frequently exhibit a leaky phenotype, whereas a qualified chip model must demonstrate a significant escalation in electrical resistance under fluid shear stress stimulation. This mechanical environment-induced elevation in electrical impedance signifies that junctional proteins between endothelial cells have translocated from the cytoplasm to the cell membrane surface, thereby forming a functional vascular wall. If a vascularized model fails to exhibit this impedance strengthening effect under perfusion conditions, it indicates that the model has not yet met the physical requirements for immune cell trafficking [39].

Furthermore, for pathological models involving fluid dynamic abnormalities such as edema, the apparent permeability coefficient provides a more intuitive quantitative basis than electrical resistance. Wufuer et al. proposed a fluid extravasation validation method based on fluorescent tracers. Under acute inflammatory conditions induced by TNF-α, a qualified model should exhibit a substantial fold increase in the transmembrane diffusion rate of FITC dextran from the vascular channel to the tissue side. Dai et al. supplemented the physical validation criteria from the perspective of tissue structural integrity, establishing the continuous distribution of filaggrin in the upper epidermal layers as the fundamental morphological criterion for judging the soundness of the epidermal barrier. Any discontinuity or loss of expression is regarded as definitive evidence of successful pathological model construction [83,95].

3.4.2. Immunological profiling and secretome analysis

Validation at the biochemical level aims to confirm whether the model has generated the correct inflammatory signature, requiring the model to not only be morphologically similar but also align with clinical data in terms of molecular mechanisms. For cytokine-driven models, the core of validation lies in the consistency of the stimulus and response. Tjabringa et al. established a classical validation framework for psoriasis models, proposing that the model must not only express the IL-17 receptor but also secrete downstream effector molecules following stimulation. Studies indicate that a qualified psoriasis model should exhibit detectable high levels of CXCL8 and human beta defensin 2 in the supernatant, and this secretion should be blockable by specific antibody therapeutics. The ELISA readouts of such specific secreted proteins serve as the core biochemical basis for determining whether the model possesses a functional immune reflex arc and whether it is suitable for drug screening [127].

As detection technologies advance in dimensionality, validation criteria have expanded from the quantification of single proteins to profile matching at the omics level. Chiricozzi et al. elevated the validation standards to the transcriptomic fingerprint level, confirming that a qualified model should be able to recapitulate the characteristic gene expression profile found in the skin lesions of clinical psoriasis patients, including the synergistic upregulation of key genes such as S100A7 and DEFB4. In the metabolomics dimension, Harvey et al. introduced matrix-assisted laser desorption ionization mass spectrometry imaging as an advanced validation tool [91,93].

Distinct from traditional homogenization assays, this technology enables the in situ visualization of the spatial distribution of lipid metabolites. For instance, in IL-22-induced models, the depletion of ceramide subclasses in specific regions of the epidermal layer has been established as the validation endpoint for metabolic remodeling. Furthermore, Danso et al. confirmed that alterations in the composition of long chain free fatty acids constitute the specific metabolic fingerprint distinguishing atopic dermatitis models from healthy controls. The incorporation of these high-dimensional data ensures that the depth of model validation is no longer confined to phenotypic simulation [98].

3.4.3. Cellular dynamics and imaging endpoints

For advanced dynamic models integrating live immune cells, static metrics are no longer sufficient to encapsulate their functionality; dynamic validation parameters based on spatiotemporal behavior must be introduced. Within microfluidic chips, the recruitment efficiency of immune cells is the primary benchmark for evaluating the biomimetic degree of the model. Sun et al. proposed the rolling to adhesion ratio as a quantitative metric for neutrophil function in vascular chips [78].

In a qualified inflammatory model, circulating neutrophils should neither flow directly through nor nonspecifically lodge in the channel. Instead, they should exhibit a specific three-step cascade reaction on the endothelial surface encompassing capture, decelerated rolling, and firm adhesion. Only when this complete dynamic process is observed can the model be considered to have successfully recapitulated in vivo immune surveillance functions.

Once immune cells enter the tissue matrix, their migratory behavior parameters become crucial for validating the stability of the chemotactic environment. Riddle et al. established dual validation criteria of velocity and chemotactic index via single cell trajectory tracking technology. The research indicates that in models with a stably established gradient, the movement trajectories of neutrophils should exhibit high directionality with a chemotactic index approaching rather than random Brownian motion. If the immune cells within the model survive but merely exhibit place oscillation or disordered migration, it suggests a failure in microfluidic gradient construction or a mismatch in matrix pore size, rendering the model functionally unqualified [107].

For specific subpopulations such as LCs responsible for antigen presentation functions, the validation focus shifts to their cross-tissue homing capability. Ouwehand et al. established a functional validation protocol for LCs: following exposure to chemical sensitizers, cells in a qualified model should not only display morphological activation via dendritic extensions but also initiate a CCR7-dependent migration program. Monitoring whether they can successfully degrade the basement membrane and transit into the dermal layer via live cell imaging is the defining characteristic distinguishing static coculture from dynamic immune models [102].

In summary, the validation of modern immunocompetent skin models is no longer satisfied with endpoint dead cell staining but is evolving toward real-time, dynamic, and visualized modalities. Rhee et al. further incorporated pharmacokinetic parameters into the validation framework. By monitoring the drug clearance curve under perfusion conditions, they ensure the model possesses clinical predictive value not only immunologically but also pharmacologically. These multidimensional validation benchmarks collectively constitute a rigorous quality control matrix, ensuring the reliability and translatability of engineered model data [123].

4. Biomedical applications of immunocompetent human SoC models

The development of SoC technology has provided a robust platform for accurately modeling human skin physiology and pathological conditions. Compared to traditional 2D cultures that lack tissue polarity, multilayer architecture, and functional tissue organization, as well as animal models limited by species differences in immune cell composition and response pathways, SoC models offer a more physiologically relevant human system [133]. They enable the recapitulation of human-specific immune crosstalk, dynamic cell interactions, and barrier permeability assays for drug and allergen testing, closely mimicking key disease features such as immune cell infiltration in psoriasis, spongiosis and filaggrin downregulation in AD, and crosstalk among DCs, T cells, and keratinocytes in allergic contact dermatitis (ACD) [40,134,135]. However, current SoC platforms still face several limitations. These include the lack of skin appendages such as hair follicles and sweat glands, incomplete systemic or neuro-immune integration, and technical challenges in achieving reproducible construction of complex 3D immune microenvironments. Nevertheless, by integrating key immune components such as inflammatory factors, immune cells, vasculature, and bacteria, these immunocompetent SoC platforms can mimic the immune microenvironment in vivo.

These immunocompetent SoC platforms not only enhance the understanding of disease mechanisms and drug responses among researchers and clinicians but also support safety and efficacy assessment of cosmetics. Furthermore, it effectively narrows the gap between conventional in vitro models and in vivo responses, showing great potential to reduce reliance on animal testing and accelerate clinical translational research.

4.1. Disease modeling

In the field of pathological research, advanced skin organ models incorporating immune components provide a powerful in vitro research platform for investigating the mechanisms of various skin pathologies, owing to their ability to recapitulate the complex immune microenvironment. The integration of specific immune cell populations not only enables the recreation of hallmark histopathological features of diseases but also facilitates the analysis of dynamic intercellular signaling between immune cells and skin structural cells under pathological conditions. These offer a robust in vitro platform for gaining deeper insights into the pathophysiological mechanisms of diverse skin disorders. This section focuses on the applications of immune-integrated SoC models in studying immune-mediated skin diseases like ACD, psoriasis, and AD.

To establish a direct link between disease-specific immune pathways and SoC design strategies, it is essential to characterize the distinct immunological underpinnings of each condition. Psoriasis is driven by the TNF-α-IL-23-Th17 axis, with IL-17A and IL-22 mediating keratinocyte hyperproliferation and aberrant differentiation [136]. Consequently, psoriatic SoC models prioritize the incorporation of Th17 cells or IL-17A stimulation, along with microfluidic architectures that enable the study of T cell migration and keratinocyte immune crosstalk. In contrast, AD exhibits a biphasic immune response, necessitating SoC designs that can accommodate dynamic cytokine switching and incorporate Th2-associated effectors like IL-4, IL-13, and eosinophils, while also enabling the transition to Th1-dominated chronic inflammation [137,138]. ACD, as a classic T-cell-mediated delayed type hypersensitivity reaction, requires models that recapitulate both the sensitization and elicitation phases, focusing on LCs activation, hapten presentation, and subsequent T cell recruitment, with perfusable vascular channels to mimic circulating immune cell trafficking [139]. These design principles are reflected in the models summarized in Table 5.

Table 5.

Comparison of immunocompetent SoC models for inflammatory skin diseases.

Diseases	Key immune pathways	Exogenous immune components	Ref.
Psoriasis	TNF-α-IL-23-Th17 axis and Th17/Th22 polarization	IL-1α, TNF-α, IL-6, and IL-22	[127]
	CCR6/CCL20 axis, CXCR4/CXCL12 axis, and S1PR1/S1P axis	activated human peripheral blood T cells, human blood neutrophils, CXCL12, CCL20WT, S1P, CCL20LD, and fMLP	[40]
	Th1/Th17 pathways, CCR6/CCL20 axis, and CLA homing pathway	Th1 cells, Th17 cells, CD4+ T cells, and patient-derived CCR6+ CLA+ T cells	[110]
	Th17/IL-17 pathway, Th22/IL-22 pathway, and Th1/TNF-α pathway	IL-17A, IL-22, TNF-α, IL-1α, and IL-6	[140]
AD	Th2/IL-4/IL-13 pathway and IL-4Rα/JAK/STAT6 signaling axis	IL-4 and IL-13	[100]
	Th2/IL-4/IL-13/IL-31 pathway and TNF-α mediated inflammatory pathway	TNF-α, IL-4, IL-13, and IL-31	[98]
	IL-31/IL-31RA/OSMRβ signaling axis, eosinophil keratinocyte direct interaction pathway, and Th2 type inflammation associated chemokine network	eosinophil and IL-31	[141]
	Th2/IL-4/JAK-STAT signaling axis	IL-4	[142]
	microbial dysbiosis	Propionibacterium acnes ATCC 25746, Staphylococcus epidermidis NCTC 11047, Malassezia globosa CBS 7874, Staphylococcus aureus ATCC 29213, Corynebacterium jeikeium NCTC 11915, Escherichia coli ATCC 25922, and the native skin microbiome from human volunteers.	[143]
	microbial dysbiosis affecting TLR signaling pathway and antimicrobial peptide expression regulatory pathway	Staphylococcus epidermidis strain S9 and Staphylococcus aureus strain SH1000	[144]
ACD	LCs migration and activation	MUTZ-LCs and MoLCs	[48]
	antigen presentation and DCs maturation pathway, immune cell migration and homing pathway	MUTZ-3 cells	[87]

4.1.1. Psoriasis

Psoriasis is a chronic inflammatory skin disease triggered by the interplay of genetic predisposition and environmental factors, characterized by aberrant immune activation [145]. Clinically, it presents as well-demarcated erythematous plaques with silvery scales. Histopathological examination reveals hyperproliferation of keratinocytes, impaired epidermal differentiation, enhanced angiogenesis, and prominent inflammatory infiltration in lesional skin [146]. Immunologically, disrupted Th1/Th2 balance is observed in psoriatic lesions, with T-bet and GATA-3 mediated Th1 polarization leading to significantly elevated levels of IL-2, IFN-γ, and T-bet [147]. Notably, the TNF-α-IL-23-Th17 axis plays a pivotal role in the pathogenesis of T cells-mediated plaque psoriasis, accompanied by increased circulating Th17 and Th22 cell populations (Fig. 12A) [22,[148], [149], [150], [151], [152]].

Fig. 12.

Non-immunocompetent SoC models for investigating the pathogenesis of psoriasis. (A) Key cellular and signaling pathways in psoriasis immunopathogenesis [148]. Copyright 2021, MDPI. (B) H&E staining of psoriasis-like skin equivalents induced by Th1/Th17 cytokines at day 12. Scale bars: 50 μm [140]. Copyright 2024, Springer Nature. (C) Conceptual figure, (D) photographic image, and (E) side-view diagram of the dual-organ gut-skin chip, with skin module in air-exposure condition for studying immune crosstalk. Scale bar: 10 mm. Concentrations of hBD-2 in (F) 3D skin equivalents and (G) dual-organ gut-skin chip with different treatments [166]. Copyright 2022, John Wiley and Sons.

For mechanistic studies, various animal models have been established, including spontaneous mutants, genetically modified models, pro-inflammatory cytokine injection models, imiquimod (IMQ)-induced models, and humanized skin mouse models [[153], [154], [155], [156], [157], [158], [159]]. In vitro systems primarily utilize HaCaT cell lines, normal human epidermal keratinocytes (NHEKs), epidermal progenitors, HUVECs, T cells, and THP-1 monocytic cells [110,[160], [161], [162]]. Psoriasis skin equivalents can be generated either with lesional keratinocytes and fibroblasts isolated from patients or by supplementing healthy reconstructed skin with pro-inflammatory cytokines [24,156,163,164]. For instance, Tjabringa et al. constructed HSEs using mouse fibroblast cell line 3T3 and human primary keratinocytes, and induced psoriatic skin equivalents by stimulating with IL-1α, TNF-α, and IL-6 for 4 days [127]. This model effectively mimicked keratinocyte responses to inflammation and distinguished the effects of epidermal-targeting drugs such as all-trans-retinoic acid from immunomodulators such as cyclosporine A. However, given the human-specific nature of psoriasis, mouse models fail to fully recapitulate the human pathology, and the use of mouse fibroblasts further limits the clinical relevance of this model. To improve accuracy, Morgner et al. employed human primary dermal fibroblasts and keratinocytes with fibronectin as a dermal matrix, generating differentiated epidermis via air-liquid interface culture [165]. Subsequent stimulation with IL-17A, IL-6, IL-22, IL-1α, and TNF-α successfully established a psoriatic model. Compared to controls, the model exhibited classic psoriatic features, including downregulation of differentiation markers like CK10, FLG, and IVL and upregulation of proliferation-related proteins like CK16 and S100A7 accompanied by impaired epidermal barrier function, increased IL-6/IL-8 secretion, expansion of Ki67-positive basal cells, and parakeratosis. These results confirmed the model's reliability in simulating psoriasis-associated inflammation, hyperproliferation, and aberrant differentiation. In a separate study, Morgner et al. developed a skin model based on primary keratinocytes, which simulated the characteristics of psoriasis, including parakeratosis and abnormal differentiation, by subcutaneously adding a mixture of Th1/Th17 cytokines (Fig. 12B) [140]. Notably, this model utilized a fibroblast-derived matrix as the dermal layer, resembling human skin more closely and enhancing the accuracy of the simulation.

However, conventional single-tissue skin models are insufficient to investigate the systemic interactions between psoriasis and other organs. To address this, Lee et al. constructed a dual-organ gut-skin chip combining fibroblasts, HaCaT cells, and Caco-2 cells to study the expression of psoriasis-associated hBD-2 under gut inflammatory conditions (Fig. 12C–E) [166]. Fig. 12F and G showed that a pro-inflammatory cytokine cocktail (TNF-α, IL-6, IL-1α) or palmitic acid (PA) could induce hBD-2 secretion in the skin, while gut barrier dysfunction exacerbated skin inflammation by increasing the absorption of inflammatory molecules or fatty acids. This study provided the first in vitro evidence of the gut-skin axis, offering new insights into psoriasis pathogenesis.

Such models can mimic psoriatic inflammation, but they lack immune cells like DCs and T cells, limiting their ability to recapitulate immune-keratinocyte crosstalk [103]. To overcome this, Ren et al. designed a microfluidic chip to quantitatively analyze T cell migration in a simulated skin inflammatory microenvironment [40]. The chip featured a central collagen-embedded gel channel mimicking the dermis, flanked by chemical gradient channels, integrating HaCaT cells, HUVECs, and activated human peripheral blood T cells (ahPBTs). Upon stimulation with CXCL12, CCL20, S1P, or TNF-α, T cells exhibited directed migration, highlighting their critical role in psoriasis (Fig. 13A and B). As illustrated in Fig. 13C, Shin et al. developed a 3D psoriatic skin equivalent model incorporating patient-derived CCR6⁺CLA⁺ T cells, and demonstrated that the epidermis not only provides crucial chemotactic signals guiding the directional migration of T cells toward the epidermal layer but also significantly enhances their infiltration depth and activation status (Fig. 13D and E) [110]. This finding underscores the active role of the epidermal microenvironment in recruiting and activating immune cells during the pathogenesis of psoriasis. Beyond immune cell-mediated regulation of keratinocytes, the role of dermal fibroblasts in psoriasis progression has garnered significant attention. Jiang et al. established skin organoids incorporating epidermal cells, dermal cells, and DCs to investigate how DCs and dermal cells influence epidermal proliferation under psoriatic conditions [167]. It demonstrated that DC-secreted LGALS9 binds to CD44 on FBs, stimulating collagen deposition and ECM synthesis. This ECM remodeling increases tissue stiffness and subsequently enhances epidermal proliferation through the HMGB2-RRM2 axis. These findings not only elucidate the mechanistic role of fibroblasts in psoriasis pathogenesis but also suggest that disrupting the LGALS9-mediated DC-fibroblasts crosstalk may represent a novel therapeutic strategy. Compared to conventional models, immune cell-containing skin organoids better recapitulate intercellular interactions, demonstrating considerable potential for future psoriasis research.

Fig. 13.

Immunocompetent SoC models for investigating the pathogenesis of psoriasis. (A) CXCL12 induced T cells transmigration in the SoC model. (B) The displacement analysis of T cells in different experimental groups at 1 h. Scale bar: 100 μm [40]. Copyright 2021, The Royal Society of Chemistry. (C) Schematic diagram of preparing immunocompetent skin equivalents. (D) The spatial distribution of K14⁺ keratinocytes and CD3⁺ T cells in skin equivalents with an intact epidermis or following its removal. Nuclei are counterstained with DAPI. (E) The expression levels of the early activation marker CD69 and CD25 on CD3⁺ T cells isolated from skin equivalents with an epidermis or without an epidermis [110]. Copyright 2020, Springer Nature.

4.1.2. Atopic dermatitis

AD is a complex multifactorial disease involving diverse pathogenic mechanisms, including genetic predisposition, skin barrier dysfunction, microbial dysbiosis, food allergies, and sensitization to aeroallergens [168]. Clinically, affected areas present with eczematous lesions characterized by xerosis, erythema, and scaling, accompanied by intense pruritus leading to frequent scratching behavior [169,170]. Histopathological examination of AD induced by IL-4, IL-13, and IL-31 reveals marked epidermal hyperplasia with abnormal granular layer structure and reduced expression of FLG and claudins (Fig. 14A). The dermis is characterized by T cell infiltration, while impaired intercellular adhesion among keratinocytes results in widened intercellular spaces and intracellular edema. Immunologically, there is maturation of LCs and dermal DCs, along with significantly elevated TARC/CCL17 expression in both skin and peripheral blood. IL-4-mediated differentiation and expansion of Th0 to Th2 cells initiates a Th2-dominant immune response through cytokines including IL-4, IL-5, and IL-13. Increased circulating CD4⁺ T cells and CLA⁺ T cells further indicate Th cell-mediated immune dysregulation. Under the influence of IL-4 and other cytokines, plasma cells produce specific IgE that triggers mast cell degranulation and histamine release, exacerbating AD symptoms [171]. The acute phase is dominated by Th2 cell activation with impaired skin barrier function, while disease progression to the chronic phase shows gradual Th1 cell predominance, where cytokines like IFN-γ promote lichenification and skin thickening [24]. This dynamic from Th2 to Th1 immune shift has motivated researchers to develop multi-level model systems for investigation.

Fig. 14.

Non-immunocompetent SoC models for investigating the pathogenesis of AD. (A) Key cellular and signaling pathways in AD immunopathogenesis [189]. Copyright 2019, Springer Nature. (B) Schematic diagram of the process for preparing normal HSE and AD-HSE by adding IL-4/IL-13 [100]. Copyright 2022, MDPI. (C) Effect of transwell inserts on the induction of IL-6 and (D) CCL18 in co-culture of eosinophils and keratinocytes under IL-31 stimulation [141]. Copyright 2010, Oxford University Press.

In animal models, MC903 or DNCB induced murine models [169,172], NC/Nga inbred strains [173], and genetically engineered models with FLG, Tmem79/Matt, or Ctip2/Bcl11b knockout/mutations are widely utilized [174,175]. In SoC models, keratinocytes serve as primary targets, acting as key initiators of AD inflammatory cascades [176,177]. For constructing AD-like skin equivalents, researchers typically activate the STAT6/STAT3 signaling pathways by adding IL-4/IL-13 to suppress FLG expression in keratinocytes, reduce loricrin (LOR) production, and induce S100A7 overexpression, thereby mimicking pathological features such as epidermal hyperplasia [177,178]. For example, Kim et al. developed a skin equivalent using NHEKs and human fibroblasts with a collagen-fibronectin dermal matrix, successfully inducing an acute-phase AD model through Th2 cytokines (Fig. 14B) [100]. Danso et al. employed the Leiden Epidermal Model to recapitulate the AD inflammatory microenvironment using TNF-α, IL-4, IL-13, and IL-31 [98].

The development of in vitro models has provided crucial tools for elucidating the molecular mechanisms underlying AD pathogenesis. FLG, a crucial structural protein for epidermal terminal differentiation, promotes keratin filament aggregation and keratinocyte flattening, which has been conclusively linked to AD pathogenesis through its deficient expression [179]. To investigate the role of FLG in AD, researchers have employed various genetic manipulation approaches combined with SoC technology to construct sophisticated in vitro models. For example, FLG knockdown models have been established using small interfering RNA (siRNA) or short hairpin RNA (shRNA) techniques [[180], [181], [182], [183], [184], [185]]. However, these gene knockdown-based models still face limitations such as residual protein expression and potential off-target effects. To more accurately assess the consequences of complete FLG deficiency, Niehues et al. developed differentiated human epidermal equivalents (HEEs) by culturing FLG-null primary human keratinocytes on transwell inserts [186]. Notably, these HEEs failed to spontaneously exhibit immune abnormalities, consistent with observations in mouse models [187]. It should be emphasized that most existing models primarily focus on epidermal barrier dysfunction while inadequately recapitulating the synergistic destructive effects of immune cell infiltration.

To better mimic the pathological process of T cells infiltration in AD, Engelhart et al. constructed a skin equivalent incorporating activated CD45RO⁺ T cells [188]. This model utilized type I collagen and fibronectin as the dermal matrix, integrating HaCaT cells, human fibroblasts, and activated T cells to create a skin equivalent with activated T cells (SE-AT). Without requiring additional stimulation, this model successfully recapitulated key AD features, including epidermal spongiosis and barrier dysfunction, while promoting the secretion of IL-1α, IL-6, IL-8, and various chemokines, thereby providing an immunologically relevant tool for AD research. Beyond T-cell-mediated immune dysregulation, DCs as critical antigen-presenting cells have also been extensively studied in AD pathogenesis using SoC technology. Importantly, direct cell-cell contact between immune cells and skin cells plays a pivotal role in AD development. For example, IL-31-induced inflammatory responses depend on direct contact between eosinophils and keratinocytes. As shown in Fig. 14C and D, when this interaction was blocked using transwell inserts, the secretion of inflammatory cytokines IL-6 and chemokines CCL18 was significantly reduced [141]. These findings highlight the necessity of incorporating immune cell interactions in advanced AD models to better reflect disease pathophysiology.

In the investigation of AD pathological mechanisms, beyond inflammatory cells, abnormal vascular proliferation has emerged as another critical feature. Fig. 15A showed that Liu et al. employed 3D bioprinting technology to develop a vascularized skin model for AD research [142]. This model utilized fibrin-based hydrogel as the dermal matrix, incorporating NHEKs, NDFs, iPSC-derived ECs, and pericytes to recapitulate skin architecture. Results showed that IL-4 treatment led to a reduction in angiogenesis accompanied by increased release of IL-1α and TARC, while ECs secreted VEGF, ICAM-1, and VCAM-1 to promote immune cell migration, thereby exacerbating inflammatory responses. This study provided novel experimental evidence for understanding the synergistic interactions between vascular and immune systems during AD pathogenesis.

Fig. 15.

Immunocompetent SoC models for investigating the pathogenesis of AD. (A) Schematic of the 3D bioprinting process and structural characterization of the vascularized skin equivalent [142]. Copyright 2020, IOP Publishing. Contribution of National Institute of Health. (B) Downregulation of FLG and LOR expression in epidermal layer of skin organoids following S. aureus infection. Scale bar: 50 μm [190]. Copyright 2022, Elsevier.

The microbial community, as an integral component of the skin surface microenvironment, significantly influences AD progression. Previous studies have simulated pathogen over colonization and barrier defects in AD lesions by inoculating S. aureus onto skin organoid models [143,144]. These models have provided valuable insights into the complex interplay between microbial pathogens and skin barrier function, particularly in the context of AD. As shown in Fig. 15B, Jung et al. found a dose-dependent reduction in FLG and LOR expression after S. aureus infection leveraging the ALI skin organoid model, correlating with compromised barrier function in AD patients [190]. These findings underscore the utility in dissecting pathogen-barrier interactions and highlight the need to maintain a healthy skin microbial community to prevent AD pathogenesis. Jiao et al. established a co-culture system combining human basophils/eosinophils with dermal fibroblasts, subsequently infected with S. aureus. The findings showed that S. aureus-derived NOD2/TLR2 ligands compromised skin barrier function through innate immune activation, while stimulating fibroblasts to release chemokines including CXCL8, CCL2, CCL5, and IL-6, along with upregulated ICAM-1 expression, collectively exacerbating inflammatory infiltration [172]. These results suggest that microbial dysbiosis and skin barrier defects cooperatively contribute to AD development.

4.1.3. Allergic contact dermatitis

ACD is an inflammatory skin disorder mediated by T cells, belonging to type IV delayed hypersensitivity reactions [191]. The pathogenesis begins when the skin is exposed to low-molecular-weight chemicals, known as haptens [192]. These haptens can initiate immune responses through two pathways: binding to carrier proteins in the epidermis or directly activating DCs to form complete antigens. The resulting hapten-carrier protein complexes are taken up by DCs and presented to reactive T cells, with this sensitization process typically requiring 10 to 14 days [139,193]. Upon re-exposure to the same allergen, sensitized individuals experience migration of activated CD4⁺ and CD8⁺ T cells to the contact site, where they release pro-inflammatory cytokines such as IFN-γ, IL-2, and IL-17, ultimately leading to characteristic skin lesions.

ACD is characterized by delayed clinical manifestations, typically appearing 24 to 96 h after re-exposure to the allergen. Common sensitizing agents derive from diverse sources, including fragrances and preservatives in cosmetics, nickel, chromium, or cobalt in metal accessories, accelerators and carba mix in rubber products, as well as dyes in textiles [[194], [195], [196]]. The skin lesions usually begin with erythema and edema at the contact site, progressing to papules and vesicles accompanied by intense pruritus [197,198]. Histopathological examination reveals characteristic features such as eosinophilic spongiosis, acanthosis, and hyperkeratosis in the epidermis, along with dermal infiltration of lymphocytes, eosinophils, and multinucleate dendritic fibrohistiocytic cells [199]. The diagnosis of ACD relies on a comprehensive approach involving a detailed medical history, observation of typical clinical presentations, and patch testing, with the latter serving as the gold standard for identifying specific allergens [200,201].

To gain deeper insights into the pathogenesis of ACD and develop more effective diagnostic and therapeutic approaches, researchers commonly employ the murine local lymph node assay (LLNA) and guinea pig maximization test to evaluate the potential sensitization capacity of chemicals. While animal models offer complete skin appendages, spatial tissue organization, and physiological functions for disease simulation, they exhibit significant species differences in immune responses. An example is the murine Toll-like receptor 4 (TLR4). Its amino acid sequence naturally lacks the non-conserved histidines at positions 456 and 458. This structural difference prevents it from accurately replicating the specific human immune recognition and sensitization process to nickel ions [202]. Furthermore, with the promotion of the 3R principles (Reduction, Replacement, and Refinement) in animal experimentation, the European Union has completely banned animal testing for cosmetic ingredients since 2013. This trend was further reinforced by the REACH regulation in 2017, which explicitly stipulates that in vitro assays and computational methods should be prioritized for skin sensitization testing, with animal experiments permitted only when these alternative approaches fail to provide conclusive results regarding sensitization potential [3]. To better recapitulate human ACD pathogenesis, various in vitro models have been developed and utilized in ACD-related research, including THP-1 cells [44], LCs [102,112,117], human leukemic monocyte lymphoma cell line (U937) [203], human immature dendritic cells (iDCs) [204], mDCs [130], and MUTZ-3 cells [205,206]. However, a comparative evaluation of these models is essential for informed selection. Primary human LCs remain the gold standard for biological relevance for ACD, exhibiting extremely high sensitivity and specificity. Upon stimulation with sensitizers, they show significantly upregulated expression of CD86, PD-L1, and DCIR, and can predict the effects of strong, moderate, and weak sensitizers like DNCB, nickel sulfate, and cinnamaldehyde, demonstrating high clinical predictive value [207]. However, their use is limited by tissue availability, high variability, reproducibility issues, and challenges in large-scale culture and expansion. The human monocytic THP-1 cell line is often employed in standardized protocols such as the human Cell Line Activation Test (h-CLAT) due to its stable availability and low maintenance cost, yet it shows lower sensitivity and clinical relevance [208,209]. In contrast, the MUTZ-3 cells, which can be differentiated into a LC-like phenotype, provide an intermediate sensitivity between primary LCs and THP-1 [210]. It also maintains high specificity and improved clinical relevance, while mitigating the donor variability typical of primary cells. Therefore, selecting an appropriate model ultimately requires careful trade-offs between practical feasibility and biological fidelity.

Cottrez et al. employed Episkin™ to evaluate 20 sensitizers and 20 non-sensitizers, including 4-nitrobenzyl bromide, 1,4-phenylenediamine, 2,4,6-trinitrobenzenesulfonic acid, and formaldehyde [211]. By integrating data from the LLNA with qRT-PCR analysis, they examined expression changes in nearly 200 genes associated with skin sensitization following chemical exposure. The results demonstrated that sensitizers specifically upregulated ARE-dependent genes and a newly defined SENS-IS gene subset, and the combined analysis of these two gene groups enabled accurate classification of all tested sensitizers and non-sensitizers. Similarly, Saito et al. using the EpiDerm™ model identified significant upregulation of genes such as ATF3, DNAJB4, and GCLM after sensitizer treatment, suggesting their potential as biomarkers for skin sensitization [212]. Using epidermal equivalents VUMC-EE, epiCS, EpiDerm, and SkinEthic RHE as their experimental platform, Gibbs et al. advanced the understanding of skin sensitization mechanisms. Their study investigated protein-level functional changes by challenging these models with a range of chemical stimulants, including 17 known sensitizers like DNCB, oxazolone, and formaldehyde, alongside 10 non-sensitizers such as glycerol, phenol, and sodium lauryl sulfate [213]. By quantifying IL-18 secretion in culture supernatants combined with cytotoxicity assessment, the researchers developed an effective method for both sensitizer detection and potency ranking. However, these studies exclusively utilized epidermal models, lacking critical components including dermal fibroblasts, vascular endothelial cells, and immune cells. Consequently, such models cannot fully recapitulate the complex biological processes occurring in authentic ACD reactions in human skin.

To better emulate the immunocompetent microenvironment of human skin, more sophisticated models have subsequently been developed. As shown in Fig. 16A, existing studies have utilized HaCaT cells to simulate the epidermis and U937 cells to model immune cells, constructing a microfluidic SoC to evaluate responses to stimuli such as LPS, the sensitizer nickel sulfate, and physical stimuli like UV irradiation [203]. The chip integrated Ag/AgCl electrodes for real-time TEER monitoring and enabled multi-channel differential stimulation experiments, making it suitable for high-throughput screening. Bock et al. incorporated MUTZ-LCs and MoLCs into the epidermal layer of a reconstructed human skin (RHS) model, using DNCB and isoeugenol stimulation to mimic the immune response in ACD, with particular focus on LCs migration, cytokine release profiles, and gene expression alterations [48]. Fig. 16B showed that DNCB stimulation significantly enhanced the secretion of IL-6 and IL-8, while promoting the migration of LC from the epidermis to the dermis. Building upon the HUMIMIC Chip3plus platform (Fig. 16C), Koning et al. also incorporated MUTZ-LCs into a reconstructed human skin model. Following topical application of a NiSO₄ saturated patch to the organoid, they observed migration of LCs to the dermal compartment and an upregulation of mRNA markers associated with their activation and maturation. By recapitulating the activation of DC (AOP KE3) in skin sensitization, these studies offer a robust strategy for mechanistic investigation and safety assessment, serving as a strong alternative to animal experiments. Michielon et al. constructed a highly complex endothelialized immune SoC system (Fig. 16D) [87]. The model utilized a hybrid hydrogel of rat-tail type I collagen and fibrinogen as the dermal matrix, co-culturing keratinocytes, melanocytes, and dermal fibroblasts. Human dermal microvascular endothelial cells (HDMECs) were seeded on the underside of the transwell membrane to form a vascular barrier (Fig. 16E). In this study, the MUTZ-3 cells were perfused through the chip to serve as circulating immune cells, mimicking in vivo immune responses. NiSO₄ exposure confirmed its immunostimulatory effect within the chip, as evidenced by the upregulation of activation markers CD86 and CD83 in MUTZ-3 cells (Fig. 16F). Furthermore, the results indicated that the dynamic flow environment enhanced the activation response of immune cells to the sensitizer. This model, which is easily scalable and standardized, is suitable for high-throughput toxicological screening in the study of ACD pathological mechanisms, holding significant potential for replacing animal testing and advancing personalized medicine. These findings suggest that OoC technology with fully functional immune components may provide a more reliable experimental platform for investigating ACD mechanisms, ultimately facilitating the development of innovative diagnostic approaches and therapeutic strategies.

Fig. 16.

SoC models for investigating the pathogenesis of ACD. (A) Schematic of the microfluidic device and integrated TEER electrodes [203]. Copyright 2016, The Royal Society of Chemistry. (B) Application of DNCB for 24 h enhanced the migration of LCs-like cells from epidermis to dermis in reconstructed HSE [48]. Copyright 2017, Elsevier Ltd. (C) Microfluid chip for modeling systemic sensitizer exposure on the HUMIMIC Chip3plus platform [206]. Copyright 2022, Frontiers Media. (D) The complete micro-physiological system platform CubiX MVP2C for environmental control of the SoC model. (E) Schematic of the internal fluidic sealed structure in the immunologic function SoC model. (F) CD83 and CD86 expression in MUTZ-3 cells following exposure to NiSO₄ or H₂O under static and dynamic conditions [87]. Copyright 2024, Wiley.

4.2. Drug development

Drug development is a lengthy and complex process, with its greatest challenge being the limited predictive accuracy of preclinical models for drug efficacy and toxicity, resulting in extremely low regulatory approval rates. Data indicate that the clinical translation success rate for oncology drugs is only 5%, while the average success rate across all indications is merely 9.6% [214,215]. To improve predictive efficiency, preclinical modeling and simulation techniques are widely used to evaluate critical pharmacokinetic (PK) and pharmacodynamic (PD) parameters [216]. In recent years, OoC technology has emerged as a research focus due to its high controllability. Among these, SoC microfluidic platforms can accurately simulate the physiological environment of skin tissue to assess the permeability and mechanisms of drugs, chemicals, and cosmetics, significantly reducing reliance on animal testing [217].

For instance, Kim et al. constructed a pump-free microfluidic SoC model using human fibroblasts and keratinocytes with a gravity-driven system [75]. Experiments with curcumin long leaf extract stimulation demonstrated thickening of the stratum corneum and stratum spinosum, along with significant upregulation of barrier-related genes, including FLG, involucrin, laminin alpha-5, and keratin 10. These findings confirmed that the model not only promotes keratinocyte differentiation but also enhances the skin's physical and chemical barrier functions.

Recently, there have been advances in the construction of skin models based on 3D bioprinting technology. Kang et al. successfully fabricated a 3D skin structure containing HaCaT cells and NHDFs using GelMA/HAMA composite bioink through a layer-by-layer 3D printing deposition process [218]. By introducing TNF-α and hydrogen peroxide to simulate an inflammatory environment, the study effectively reproduced pathological processes including proinflammatory signaling pathway activation and oxidative damage, thereby evaluating the therapeutic effects of garlic extract N-benzyl-N-methyldodecan-1-amine.

4.3. Cosmetics testing

With the rapid development of the global cosmetics industry, a relatively comprehensive classification system for cosmetics testing has been established, primarily including safety testing, efficacy evaluation, and quality control [1,[219], [220], [221]]. Safety testing involves toxicological assessments such as skin irritation and sensitization; efficacy evaluation focuses on verifying functional claims like whitening and anti-aging, while quality control includes component analysis, microbial contamination testing, and stability tests.

The integration of SoC and microfluidic technology has enhanced the physiological relevance of models used for evaluating the toxicity and efficacy of cosmetics. Chen et al. constructed homogeneous dermal spheroids using a blend of matrigel and type I collagen as a dermal matrix substitute for ECM, combined with human neonatal dermal fibroblasts [222]. Vitamin C, a common cosmetic ingredient with antioxidant, anti-aging, and brightening properties, was continuously perfused for 48 h to assess its penetration behavior and remodeling effects on the dermal matrix. Beyond measuring protein expression, one study developed a gold nanodot-RGD patterned skin fibroblast chip to monitor redox currents in live cells, sensitively evaluating cytotoxicity and inflammatory responses induced by the preservative imidazolidinyl urea [223].

In efficacy evaluation, antioxidant and anti-aging ingredients can protect fibroblasts from oxidative stress and reduce collagen degradation. Kim et al. co-cultured rat tail type I collagen with human dermal fibroblasts and human epidermal keratinocytes to construct skin equivalents [224]. Using a pump-free SoC platform, they evaluated the effects of the anti-aging compound α-lipoic acid on epidermal structure formation and anti-aging efficacy. Results showed that α-lipoic acid treatment upregulated epidermal differentiation-related proteins such as FLG, involucrin, and keratin 10, enhanced type IV collagen expression at the dermal-epidermal junction, and improved matrix stability, skin barrier function, and structural integrity. Similar effects were observed when anti-aging agent coenzyme Q10 was applied to SoC [225]. Furthermore, Liu et al. developed an SoC model using HDFs and HaCaT cells, simulating inflammatory and UV-induced damage via lipopolysaccharide and UVB treatment to evaluate the anti-inflammatory effects of gentiopicroside [226]. These studies demonstrate that SoC enable efficient efficacy validation of cosmetic ingredients, with testing periods as short as one day, highlighting distinct advantages over animal testing.

Cosmetics can induce skin inflammation, and incorporating skin appendages such as vascular structures and immune-related cells into skin equivalents to construct immune-integrated and inflammation-mimicking SoC models can enhance physiological relevance. For instance, Wufuer et al. developed a triple-layered SoC comprising epidermis, dermis, and endothelial cells, which produced pro-inflammatory cytokines such as IL-6 and IL-8 under TNF-α stimulation. These cytokine levels were significantly reduced after dexamethasone (Dex) treatment, consistent with trends observed in human skin biopsies [83]. Fig. 17A and B showed a chip manufactured by Kwak et al., which integrated dermal fibroblasts, keratinocytes, and HUVECs to simulate immune responses of skin tissue exposed to sodium lauryl sulfate and ultraviolet radiation [39]. After UV irradiation, H&E staining (Fig. 17C) and fluorescent staining of HL-60 cells using CellTracker (Fig. 17D) showed that HL-60 cells migrated across the endothelial layer into the dermal region, adopting round or irregular morphologies, with a significant difference in the number of migrated cells as shown in Fig. 17E. Treatment with the anti-inflammatory drug Dex resulted in decreased levels of the inflammatory cytokine IL-6. This chip system was used to effectively evaluate inflammatory effects induced by chemical irritants and UV exposure, simulate leukocyte migration, and validate anti-inflammatory drug efficacy, providing an immune-functional platform for cosmetic safety testing.

Fig. 17.

SoC models for the application of cosmetics testing. (A) Schematic diagram and (B) picture of the microfluidic skin chip for studying the migration of HL-60 cells. Scale bar: 1 cm. (C) H&E staining images of the skin tissue in a chip treated (a) with UV and (b) without UV. Arrows indicate HL-60 cells. Scale bars: 100 μm. (D) Confocal microscope Z-section images of SoC treated (a) without UV and (b) with UV. Scale bars: 50 μm. (E) Comparison of number of migrated cells [39]. Copyright 2020, Wiley.

SoC technology has demonstrated distinct advantages in both safety assessment and efficacy verification. From an application perspective, it can be deeply integrated with automated downstream detection systems to enable parallel and rapid automated analysis of multiple indicators such as skin barrier function and cell viability, thereby improving detection efficiency and standardization. As illustrated in Fig. 18A, an SoC coupled with a cytokine detection system employs a disc-shaped cytokine detection chip with six independent channels that utilizes immunomagnetic bead-antibody-enzyme complexes to quantify inflammatory cytokines via fluorescence signals [226]. By integrating microfluidic technology with digital ELISA, detection can be completed within 90 s with a sensitivity of 0.41 pg mL⁻¹, achieving consistency comparable to conventional ELISA (Fig. 18B). Beyond cytokine detection, Zhang et al. developed an integrated epidermis-on-a-chip (iEOC) system using NHKs, which combines automated construction and in situ barrier function detection modules [227]. Through TEER attachment electrodes shown in Fig. 18C, the system enables real-time monitoring of transepithelial electrical resistance for evaluating the irritancy of cosmetic ingredients. The system was used to test the effects of ten substances, including isopropanol, methyl stearate, cyclamen aldehyde, heptyl aldehyde, and hexyl salicylate, on cell viability and TEER. Experimental results showed that the chip could accurately distinguish irritants like isopropanol and cyclamen aldehyde from non-irritants (Fig. 18D and E), with classification outcomes fully consistent with OECD Guideline 439. Such integrated and automated systems show high reliability for cosmetic safety assessment and are suitable for large-scale testing and irritancy evaluation of cosmetics and pharmaceuticals, highlighting considerable potential for industrial translation.

Fig. 18.

Detection systems integrated with skin models. (A) Schematic diagram of the SoC system integrated with a rapid multiplex inflammation detection system. (B) Performance of microfluidic-digital ELISA for IL-6 detection comparable to conventional ELISA [226]. Copyright 2024, Springer Nature. (C) Schematic diagram of the iEOC system for TEER detection. TEER values of the epidermis-on-chip models measured by iEOC system, before and after 42 h exposure to (D) isopropanol and (E) cyclamen aldehyde [227]. Copyright 2021, The Royal Society of Chemistry.

4.4. Multi-organ SoC models

Although microfluidic SoC models have advanced significantly, the skin does not function in isolation. For example, the simulation capability of single-organ chips remains limited due to the complex multi-tissue interactions involved in drug effects and hepatotoxicity or cardiotoxicity. Therefore, integrating skin chips with other tissue chips is essential to capture the complex interplay [[228], [229], [230]]. Kim et al. designed the binary tumor microenvironment chip utilizing HUVECs and breast cancer cell lines BT-474 and MCF-7 to simulate vascularized tumor spheroids [231]. This model revealed the anticancer potential of the natural product illudin S and its hepatotoxicity, providing critical data for dose optimization and the development of targeted formulations. Pires de Mello et al. developed a multi-organ platform integrating skin, heart, and liver tissues, which utilized Strat-M membranes to simulate skin barrier function for assessing both local and systemic toxicity of drugs, including diclofenac, ketoconazole, hydrocortisone, and acetaminophen [214]. Schimek et al. innovatively employed 96-well cell culture inserts (CCI) as substrates to construct full-thickness skin equivalents incorporating HDFs and NHKs [232]. This miniaturized design not only significantly reduced cell and reagent consumption but also supported high-throughput drug permeability testing and toxicity screening. More importantly, when combined with a dual-organ chip (2OC) dynamic microfluidic system, this platform can better mimic the physiological microenvironment of multiple organs in vivo, providing a novel testing solution for drug development that combines ethical compliance, high efficiency, and enhanced physiological relevance.

For cosmetic ingredients, which are often applied topically but may enter the circulation, multi-organ chips offer a powerful tool to assess systemic effects. As shown in Fig. 19A–C, skin-nerve and skin-liver models were established by co-culturing keratinocytes with neural stem cells and pluripotent stem cell-derived hepatocytes, respectively. These systems enabled integrated assessment of the keratolytic effect of retinoic acid, sensory irritation caused by exfoliating lactic acid, and anti-irritant strontium chloride, as well as hepatotoxicity of camphor after percutaneous absorption, achieving a three-level evaluation system encompassing skin, nerve, and liver [233]. To recapitulate the complete pathway from dermal absorption and hepatic metabolism to thyroid disruption, Tao et al. employed a HUMIMIC skin-liver-thyroid Chip3 model that integrated skin, liver, and thyroid tissues into a single microfluidic platform, simulating physiological blood and lymph flow [234]. The study focused on the inhibitory effects of the anti-aging and skin-whitening agents daidzein and genistein (Gen) on thyroxine (T4) and triiodothyronine. Fig. 19D demonstrated that pre-incubation with liver tissue significantly attenuated their hormone-inhibiting effects, highlighting the essential role of hepatic metabolism in reducing thyroid toxicity. By leveraging this multi-organ model incorporating exposure routes, skin and liver metabolism, and hormonal balance, a safe dose of 0.235 μg cm⁻² was established for Daidzein (Daid) in body lotion, underscoring the critical importance of accounting for inter-organ interactions in cosmetic safety assessment, which is a capability beyond single-organ models.

Fig. 19.

The construction of multi-organ SoC models. (A) Schematic of the hybrid SoC models. (B) Schematic of the skin-nerve and (C) skin-liver models [233]. Copyright 2022, The Royal Society of Chemistry. (D) Effects of Daid, Gen, and methimazole (MMI) on Thyroid Stimulating Hormone-stimulated T4 production in thyrocyte-derived follicles without and with 24 h pre-incubation with liver organoids [234]. Copyright 2023, Frontiers Media. (E) The schematic diagram of the gut microbe-skin chip [235]. Copyright 2025, The Royal Society of Chemistry.

Multi-organ chips are increasingly employed to model systemic inflammatory crosstalk between distant tissues. Ko et al. recently developed a gut microbe-skin chip that fluidically connects intestinal epithelial cells and human epidermal keratinocytes under gravity-driven flow, incorporating live gut microbes including Escherichia coli and Lactobacillus rhamnosus GG (LGG) together with a self-sustaining oxygen gradient to mimic the physiological gut environment (Fig. 19E) [235]. Although this model does not yet integrate immune cells, it successfully recapitulated the inflammatory gut-skin axis. Disruption of intestinal barrier integrity with DSS and LPS selectively reduced keratinocyte viability without affecting gut cells, while pretreatment with the probiotic LGG exerted protective effects by enhancing tight junction integrity and reducing epithelial permeability. Collectively, these multi-organ chips represent a crucial step toward recapitulating human physiology and improving the predictive power of in vitro models for drug and cosmetic testing.

4.5. Realistic potential and remaining limitations of immunocompetent SoC models

4.5.1. Clinical and regulatory applications

Immunocompetent SoC models have evolved from static structural mimics into dynamic, functional immunological assays, offering a highly biomimetic platform for disease modeling, drug development, and cosmetic testing. To accurately assess the translational readiness of immunocompetent SoC models, it is essential to distinguish between research-grade platforms and those approaching semi-standardized or pre-regulatory testing readiness. Research grade models are primarily designed to explore fundamental biological questions, including immune cells trafficking and disease mechanism elucidation, and often prioritize biological complexity over standardization. In contrast, platforms approaching pre-regulatory readiness are characterized by defined protocols, validated performance metrics, and compatibility with regulatory frameworks such as OECD guidelines.

In practical applications, SoC serves not only as a highly biomimetic platform for disease modeling but also as a tool for evaluating pharmacodynamics, toxicity, and pharmacokinetics during drug development. For instance, in the assessment of skin irritation indicators using SoC platforms, the experimental procedure generally involves three core steps: chip construction, cell culture, and test substance application. For immune-mediated skin diseases such as psoriasis and AD, models constructed based on patient-derived cells or cytokine stimulation can accurately replicate the unique T cell-keratinocyte interactions and drug responses observed in humans, with significantly higher predictive accuracy compared to rodent models. By applying the test substance to the chip surface to simulate exposure, the platform can strictly adhere to international standards such as OECD TG 439, integrating cell viability assays and TEER measurements to dynamically evaluate epidermal barrier function and tight junction integrity, thereby significantly improving the predictive accuracy of preclinical research [236,237]. For more complex skin sensitization assessments that incorporate antigen-presenting cells like DCs, after a defined period of stimulant exposure, SoC enables the parallel simulation of multiple KE like keratinocyte activation and DCs maturation within the skin sensitization AOP. By quantitatively detecting key biomarkers such as CD86, CD40, IL-8, and IL-18, a comprehensive judgment of the substance's sensitization risk can be made [131]. The technical guide indicates that the 2 out of 3 Defined Approach, advocated by OECD TG 497, combines in vitro assays such as the Direct Peptide Reactivity Assay, KeratinoSens™, and the h-CLAT to enable non-animal assessment [238]. This approach achieves a balanced accuracy of up to 97.4% in predicting human data, which is substantially higher than the 57.9% accuracy of the LLNA, demonstrating high reliability in simulating animal test outcomes and directly predicting human sensitization potential.

Their remarkable performance in practical applications has not only opened up new opportunities in fields such as disease modeling and drug development but also drawn the attention of regulatory authorities. As technology continues to advance, there has been a clear regulatory shift towards non-animal alternative methods. The Food and Drug Administration (FDA) Modernization Act 2.0 was signed into law in 2022, explicitly permitting the use of non-animal alternative models such as organoids and organs-on-chips for preclinical drug evaluation [239]. Concurrently, the regulatory landscape for cosmetics and chemicals has undergone significant reform. The Modernization of Cosmetics Regulation Act (MoCRA), enacted in 2022, granted the FDA expanded regulatory authority [240]. Although MoCRA does not explicitly prohibit animal testing, its provisions indicate that animal experiments are no longer a mandatory requirement. This policy direction, in alignment with animal testing bans enforced in the EU and other regions, has synergistically promoted the development and adoption of human-relevant in vitro models. Collectively, these regulatory and policy advancements provide formal international regulatory recognition for the industrial translation of SoC platforms.

Concurrent with these regulatory shifts, concrete pathways for industrial adoption and regulatory qualification have begun to take shape. The U.S. regulatory system has also undergone a historic transformation. In 2020, the Center for Drug Evaluation and Research (CDER) and the Center for Biologics Evaluation and Research (CBER) of the FDA officially launched the Innovative Science and Technology Approaches for New Drugs (ISTAND) pilot program, which has facilitated the application of cutting-edge methodologies, including OoC platforms, artificial intelligence-driven tools, and digital health technologies, in drug development [241,242]. The recent release of China's national standard GB/T 44831-2024 for general technical requirements of SoC technology marks a crucial step forward in standardization and quality control within this field [243]. Concurrently, the International Organization for Standardization (ISO) established the Technical Committee on Micro-Physiological Systems and Organ-on-a-Chip Technology (ISO/TC 276/SC 2) in 2024 [244]. The standard ISO/WD 25693 currently being developed by this committee remains in the drafting phase [245]. Technical validation against regulatory endpoints, formal policy endorsement, and the emergence of both qualification pathways and technical standards have collectively propelled the rapid translational advancement of immunocompetent SoC models, underscoring their growing potential to support both clinical efficacy testing and regulatory decision-making.

On this basis, only a small number of OoC platforms have been translationally applied to regulatory approval processes to date. For SoC models to obtain regulatory acceptance, the following categories of evidence are required. The first is biological relevance, meaning that the SoC can recapitulate key pathways of human skin immune responses. The second is reliability, which requires multi-batch and inter-laboratory validation to demonstrate consistent reproducibility across batches and laboratories. The third is predictivity, requiring that the SoC exhibits good sensitivity and accuracy when compared with human clinical data or traditional animal test results.

To effectively integrate SoC models into the drug development industry, the immunocompetent SoC can be used for early screening to rapidly exclude nonspecific immunostimulatory drugs and for drug optimization based on the relationship between drug concentration and immune cell activity and proliferation. When results from single cell models or other in vitro methods are inconsistent, the immunocompetent skin chip can serve as a high-weight complementary piece of evidence to provide additional judgment.

Regarding standardization, establishing quality control indicators related to immune cells is necessary. It includes baseline requirements for immune cells in a resting state, as well as expected outcome ranges for standard positive controls such as DNCB, nickel sulfate, and the cocktail of disease-modeling factors for psoriasis and AD. Expected ranges for standard negative controls such as culture medium and glycerol should also be defined.

Concerning alignment with existing regulatory testing frameworks such as those of the OECD, the immunocompetent SoC builds upon the EpiDerm™ model used in OECD TG 439 by incorporating immune cells, thereby enabling a more realistic assessment of immune cell activation events in diseases and allergic reactions. Specifically, after testing a substance for skin corrosion and irritation according to OECD TG 439, borderline samples can be further evaluated by DCs or T cells activation assays as complementary evidence. Moreover, compared with the OECD approved h-CLAT assay, the immunocompetent SoC offers multiple cell-cell interactions that more closely mimic the skin microenvironment, thus providing more physiologically relevant experimental results.

In summary, while immunocompetent SoC models have gained increasing regulatory attention and shown clear potential for integration into drug development and cosmetic testing pipelines. But further clinical validation and standardization are still required.

4.5.2. Remaining limitations

Despite significant advances, the maturity and translational potential of SoC platforms are collectively defined by three interlinked categories of limitations: biological, engineering, and translational.

4.5.2.1. Biological limitations

Biological limitations primarily stem from the inherent trade-off between physiological fidelity and experimental reproducibility. Several key biological mechanisms lead to the decline of function over time. The first is the shrinkage of the dermal matrix. Driven by the opposing forces between fibroblast traction and insufficient matrix mechanical strength, this process leads to macroscopic dimensional shrinkage as well as microscale alterations, including reduced porosity and aberrant ECM remodeling. These changes collectively result in an increase in Young's modulus, restricted nutrient diffusion, and a shortened tissue lifespan [[246], [247], [248]]. To address the shrinkage limitations inherent to pure collagen matrices, composite hydrogels incorporating fibrin, HAMA, and GelMA have been employed to enhance matrix stability via chemical crosslinking [2,249]. Additionally, structural reinforcement through the pre-insertion of inert porous polystyrene mesh scaffolds within microfluidic chambers, or by pre-compensating for contraction, are effective [35]. These strategies collectively enhance structural stability while preserving the biological advantages of collagen.

Second, primary human cells offer superior physiological relevance but are limited by high cost, restricted availability, substantial donor-to-donor variability, and poor long-term viability under microfluidic perfusion. This is especially true for primary DCs and T cells, which exhibit immune cell exhaustion and phenotypic drift more prominently [250]. To address these challenges, immune cells derived from induced pluripotent stem cells can be employed to circumvent the effects of donor heterogeneity, genetic background differences, and prior antigen exposure. The development of robust statistical normalization methods represents another viable strategy, with the potential to incorporate variability arising from individual differences into the acceptable range of results. Moreover, material optimization strategies such as tuning scaffold stiffness to mimic the mechanical environment of native skin can prevent macrophages and DCs from triggering polarization and spontaneous maturation programs induced by excessive mechanical stress [251]. Additionally, regulating flow rates to optimize fluid shear stress helps avoid pyroptosis via the Piezo1-NLRP3 pathway, thereby maintaining the structural and functional integrity of cell membranes and organelles, ultimately extending the functional lifespan of macrophages [252]. Cell lines offer advantages in terms of stability, scalability, and ease of use, but their phenotypic deviations from native human cells limit their physiological relevance. Engineered cells with defined genetic modifications that preserve core disease pathways have emerged as a promising alternative, improving both reproducibility and consistency, but their use remains confined to specialized laboratories and has not been broadly standardized [253].

In addition, it is commonly believed that increasing complexity through the incorporation of neurons [254], sweat glands [255], hair follicles [256,257], melanocytes [258,259], or even mechanical wrinkles [260,261] will yield more physiologically accurate models. However, each additional cell type introduces new variables and potential uncontrolled interactions, complicating data interpretation and experimental reproducibility. Skin immunity extends beyond localized cellular responses; it relies fundamentally on an integrated circulatory system. A prime example is the migration of LCs upon antigen capture. They traffic via lymphatic vessels to draining lymph nodes to prime naive T cells and initiate adaptive immunity. However, in current SoC platforms, LCs migration is typically driven either by chemokine-guided active migration or by passive fluid shear stress. The absence of a functional lymphatic network compromises the model's physiological fidelity, thereby limiting its predictive accuracy for drug efficacy and toxicity in complex immune response scenarios.

Currently, stable, perfusable, and branched vascular networks which can mimic human vasculature have been developed [[262], [263], [264]]. As the first line of human defense, the skin is continually subjected to mechanical stresses such as stretching, compression, and shear, which may influence immune cell trafficking and barrier function [265,266]. Reproducing these biomechanical cues requires miniaturized actuation systems compatible with centimeter or millimeter scale as well as high throughput SoC devices, which are still under development [267,268].

4.5.2.2. Engineering and manufacturing challenges

Conventional static culture often leads to cellular sedimentation. In contrast, dynamic environments enable immune cells to maintain prolonged viability and functionality, exhibiting activation patterns that more closely recapitulate physiological reality. Microfluidic actuation presents a persistent trilemma among precision, cost, and simplicity. For example, peristaltic pumps are affordable but introduce undesirable flow pulsations that disturb shear-sensitive cells. Pressure-driven pumps provide steady flow but are often cost prohibitive for most academic laboratories and industrial scale-out. Gravity-driven flow has emerged as a pump-free alternative that simplifies system operation and reduces costs, yet its flow control precision requires further optimization. To sustain long-term immune function, stable and precise fluid handling is essential for enhancing the stability and durability of chips under cell culture conditions [269].

Beyond chip design and fluid handling, there is a growing need for non-destructive, real-time monitoring systems. Current endpoint assays such as histology and qPCR preclude longitudinal tracking of individual samples and obscure the temporal dynamics of immune responses. Future efforts should focus on multi-parameter sensing technologies capable of continuously monitoring not only biochemical markers like cytokines and chemokines, but also physical and chemical parameters including TEER, extracellular matrix stiffness, and pH values, all while maintaining long-term biocompatibility and signal stability.

Material and manufacturing challenges directly hinder industrial translation. Conventional soft lithography using PDMS is costly and suffers from nonspecific absorption of hydrophobic molecules, thereby distorting the dose-response relationships. While 3D printing offers a more affordable alternative, extrusion-based methods face limitations in resolution due to nozzle swelling, and digital light processing printing imposes stringent rheological requirements. Thus, the development of novel materials that are biocompatible, mechanically robust, economically viable, and amenable to mass production constitutes a major engineering challenge. In this context, plastic processed by injection molding and hot embossing and paper-based substrates, which are already widely used in diagnostic strips, represent a promising direction for developing low-cost and disposable SoC platforms [[270], [271], [272]].

4.5.2.3. Translation and validation gaps

Looking beyond single organ models, multi-organ integrated systems face additional hurdles such as the development of universal culture media that support diverse tissue types and the balancing of flow rates and metabolic demands across different organ compartments [[273], [274], [275]]. Besides, automation and throughput remain below industrial requirements. Current manual protocols for cell seeding and medium exchange contribute to variability and low throughput. Although the integration of robotic liquid handling, high-content imaging, and artificial intelligence-driven phenotypic analysis holds promise for fully automated, high-throughput experimental workflows, such systems have not yet been validated for routine SoC operation. Even as technical progress accelerates, translational milestones must be met. Approximately two weeks differentiation timeline required for epidermal maturation currently limits throughput compared to conventional 2D cultures. In addition, reusability remains a major challenge. For SoC that have been used for disease modeling, they can be further applied to drug treatment evaluation and multi-round dynamic monitoring of therapeutic responses, thereby improving the utilization efficiency of organ-on-chip platforms. Finally, clinical validation remains essential to correlate SoC readouts with patient outcomes.

5. Conclusions and outlook

The SoC model, a microphysiological system constructed in vitro, accurately recapitulates the complex structure and immune microenvironment of human skin. It has been extensively applied in fields such as disease research, drug screening, and cosmetic testing, and holds promise for gradually replacing traditional 2D cell models and animal experiments. This review provides a systematic summary of the key components, model types, and representative applications of SoC with immune function. In terms of model construction, cell sources, dermal matrices, microfluidic systems, and manufacturing processes jointly determine the biological functions and stability of SoC. The introduction of immune-related components, including inflammatory factors, immune cells, and blood vessels, is a crucial strategy for enhancing the degree of simulation. Among them, inflammatory factors typically include IL-17, IL-4/IL-13, etc., which can establish the pathological states of immune-mediated skin diseases such as psoriasis and AD. The introduction of immune cells such as LCs, T cells, and neutrophils enables the reproduction of biological processes such as immune recognition, cell migration, and inflammatory response. Specifically, immunocompetent SoC models enable detailed investigation of modeling the dynamic processes of inflammation initiation, resolution, and tissue repair; investigating immune cell homing and tissue microenvironment interactions; and screening and evaluating immunomodulatory drug candidates. Moreover, the cosmetic testing system based on SoC has gradually achieved standardization and automation. Combined with downstream high-throughput detection technologies, it demonstrates enormous industrial potential in the safety and efficacy evaluation of cosmetics.

Based on current technological advancements and regulatory trends, establishing clear short-term goals and a development roadmap is crucial for the translation of SoC technology. In the near term, efforts should focus on the following objectives: developing standardized, low-cost, and disposable SoC platforms. Standardization efforts should not only cover chip design specifications, such as support membrane pore size, biocompatibility of chip materials, channel dimensions, and selection methods for extracellular matrix hydrogels, but also include functional performance metrics, such as epidermal differentiation markers and real-time TEER monitoring modules to support high-throughput preliminary safety screening of cosmetic ingredients [276,277]. Furthermore, for low-cost SoC platforms based on PMMA, PS, or paper substrates, injection molding and hot embossing have been demonstrated to be feasible for high-throughput manufacturing of organ chips. However, in terms of processing optimization methods, it is necessary to advance microfabrication techniques to investigate how to achieve uniform, defect-free bonding, avoid collapse and deformation of perfusion channels, and improve sealing integrity while ensuring optical transparency and channel fidelity. Such manufacturing reproducibility, underpinned by robust process optimization, is a prerequisite for meaningful standardization and cross-platform comparability. For functional validation, inter-laboratory validation studies using reference compounds such as sodium dodecyl sulfate and nickel sulfate are required to establish reproducibility benchmarks. Concurrently, patient-derived immunocompetent models should be optimized, with an emphasis on maintaining immune cell functionality and stability in co-culture systems. The goal is to achieve a preliminary assessment of the efficacy of individualized biologics within a two-week culture period.

Looking toward medium-term and long-term development, a multidisciplinary innovation roadmap is taking shape. First is the construction of intelligent materials and dynamic microenvironments. Intelligent materials refer to systems that respond to endogenous physiological or pathological signals as well as exogenous stimuli such as pH, mechanical force, and magnetic fields. “Smart” hydrogel materials, already widely used in piezoelectric, sensing, actuation, and controlled drug release, can be integrated into chips to dynamically simulate and regulate the skin microenvironment during repair, inflammation, or treatment, thereby providing a highly controllable tool for studying dynamic barrier repair mechanisms, analyzing spatiotemporal inflammatory responses, and developing targeted delivery strategies [278,279]. Second, in multi-organ integration and systems-biology analysis, coupling SoC with other organ chips, including lymphatic, neural, liver, and gut chips, enables the construction of interactive skin-immune-metabolism models. For instance, in atopic dermatitis, linking skin, neural, lymphatic, and microbial chips can systematically simulate the complex interplay among itching, immunity, and barrier function, facilitating an in-depth exploration of intercellular communication and key signaling pathways and bridging macroscopic phenotypes with microscopic mechanisms [280,281]. Third, further progress requires emphasis on the integration of automation and artificial intelligence. Incorporating robotic arms and automated protocols can achieve fully automated operation from chip fabrication, cell seeding, and culture to compound treatment, phenotype acquisition, and report generation. Specifically, in high-throughput skin-irritation screening, AI-based computer-vision algorithms can automatically quantify skin-barrier integrity from high-content real-time imaging, replacing manual scoring [[282], [283], [284]]. Moreover, leveraging databases and deep-learning algorithms to analyze features such as cell morphology can help predict results from traditional endpoint assays like cell-viability tests, shortening evaluation cycles. Furthermore, patient-specific skin-cell cultures represent a key technology for personalized medicine. For example, in patients with refractory atopic dermatitis, an autologous immune-cell-incorporated SoC can be established within one week post-biopsy to serve as an ex vivo avatar for personalized drug testing, enabling parallel evaluation of multiple treatment regimens such as JAK inhibitors and assisting in identifying the most effective therapeutic strategy before clinical intervention. Finally, to realize these prospects, data standardization and regulatory alignment are essential for successful translation. It is necessary to gradually establish public data platforms and reporting standards to promote systematic integration, sharing, and comparative analysis of data. On this basis, active collaboration with regulatory agencies should be pursued to jointly develop validation frameworks for SoC technology based on robust research outcomes and to explore its clinical pathway in personalized efficacy prediction, ultimately contributing to the update and adoption of relevant regulatory guidelines.

In summary, the future of SoC technology lies not only in more precise biomimicry of biological models, but also in constructing an integrated immune micro-system that incorporates biology, engineering, AI, and clinical medicine, thereby offering systematic solutions for human skin health and disease treatment.

CRediT authorship contribution statement

Yudi Pang: Writing – review & editing, Writing – original draft. Beiqin Liu: Writing – review & editing, Writing – original draft. Jiwei Jiang: Writing – review & editing, Writing – original draft. Chang-ou Wang: Writing – review & editing. Naiyuan Zhang: Supervision. Xuelian Guo: Writing – review & editing, Conceptualization. Yueping Zeng: Supervision. Hong Ma: Supervision. Zhimin Wang: Supervision.

Ethics approval and consent to participate

Not applicable. This research did not involve human participants, animals, or access to identifiable personal data.

Declaration of competing interest

The authors declare no conflicts of interest.