In vitro modelling of extracellular matrix changes during skin aging: from static 2D to 3D dynamic microphysiological systems

Abstract

Skin aging results from a combination of intrinsic factors and exogenous stimuli, leading to changes in the structure and components of the extracellular matrix (including the skin basement membrane), which directly influence the aging process. In vitro models are powerful tools for exploring skin aging and overcoming inter-species differences and ethical issues associated with animal models, thus demonstrating powerful potential in skin aging research and anti-aging drug development. In this review, the advantages and disadvantages of in vitro models are discussed, including 2D monolayer models, 3D static reconstructed human skin models, 3D bioprinting models, organoid models, and Skin-on-Chip models for studying skin aging and anti-aging drug development. Finally, concepts and perspectives for the next-generation skin aging models are proposed. These models are expected to provide innovative tools for investigating the mechanisms of skin aging in depth, as well as skin aging repair and prevention.

Introduction

As the largest organ, skin is crucial for providing mechanical barrier and immune functions¹. However, aging process will impair these essential roles and trigger many health challenges. Skin aging can be classified into two categories: intrinsic aging that occurs over time, and photoaging caused by exposure to ultraviolet light². Structurally, it can be further divided into epidermal and dermal aging³. The changes in the core composition and mechanical properties of the basement membrane play a pivotal role in the skin aging process^4,5. It is of great significance to explore skin aging using skin models. Nevertheless, traditional animal models have exhibited clear genetic differences due to the interspecies differences between animals and humans. Furthermore, such experiments are costly, time-consuming, and lack specificity due to various confounding factors in the application of animal models. Moreover, ethical and moral considerations impose constraints on animal models. After the ban on cosmetic animal testing in Europe and the recommendations of the 3R principle^6,7, the development of alternative in vitro models that replicate the structural and functional characteristics of natural skin has become a top priority.

This review aimed to summarize the relationship between alterations in the extracellular matrix (ECM) during skin aging (especially the changes in core components and mechanical properties of the basal layer), and the overall skin aging process to identify the underlying causes of skin aging. Additionally, current status of in vitro alternative models was described to explore the molecular mechanisms of human skin aging, the methods for preventing and repairing skin aging, and future directions for the development of next-generation skin aging models (Fig. 1).

Fig. 1. Perspectives on the next generation of advanced in vitro skin aging models.

Overview of advanced skin microphysiological systems enabled by skin organoids, skin on chips, and 3D bioprinting. The scheme highlights key goals including recapitulating skin microanatomy, dynamically simulating aging processes, advanced aging detection and evaluation, and translation to scientific and industrial applications

Dynamic remodeling of ECM is a hallmark of skin aging

With aging, epidermal thickness decreases rete ridges flatten, and the skin’s moisturizing and barrier functions are compromised under the influence of gene-regulated cellular senescence⁸, extracellular matrix degradation due to oxidative stress⁹, and remodeling of the extracellular matrix by advanced glycation end products (AGEs)^10–12. Collagen fiber rupture and elastic fiber degradation in the dermis contribute to wrinkle formation and skin sagging¹³. Photoaging is a main contributor to facial skin aging¹⁴, and significant accumulation of thickened and curled pathological elastic fiber fragments lead to deeper wrinkles and reduced skin elasticity¹⁵. Furthermore,mechanical stresses, including tensile and compressive forces, have been shown to modulate aging trajectories, with sustained or excessive loading generally accelerating degeneration^16,17. In addition, persistent inflammation triggered by environmental pollutants, tobacco smoke, and cosmetic irritants promotes epidermal thinning, wrinkle formation, and hyperpigmentation—a constellation of features collectively referred to as inflammaging^18–21 (Fig. 2). Taken together, these structural, mechanical, and functional alterations reflect a continuous, stimulus-responsive remodeling of the ECM. This dynamic remodeling—rather than static abnormality—emerges as a central hallmark of skin aging, forming the conceptual basis for subsequent evaluation of how different in vitro skin aging models reproduce or partially mimic these processes.

Fig. 2. Skin-aging stimuli and manifestations, and the chronological evolution of in vitro skin-aging models.

Different in vitro skin models have been used to simulate the process of skin aging, investigate aging mechanisms, and facilitate anti-aging drug development, including skin explants (Yasuno et al., 1965⁷⁵; Boyer et al., 1992¹⁶⁵; Garre et al., 2018¹⁶⁶; Meunier et al., 2022¹⁶⁷), 2D skin models (Imokawa et al., 1998¹⁶⁸; Imokawa et al., 2009¹⁶⁹; Kovacs et al., 2010¹⁷⁰), 3D static reconstructed skin (Régnier et al., 1990¹⁷¹; Osborne et al., 1994⁸⁵; Liu et al., 2007⁸⁷; Diekmann et al., 2016⁶²), bioprinted skin constructs (Lee et al.¹⁷²; Min et al., 2018¹⁷³; Min et al., 2023⁶³; Sun et al., 2024⁶⁴), organoids (Rheinwald et al., 1975¹⁷⁴; Itoh et al., 2013¹⁷⁵; Lee et al., 2020⁶⁵; Kim et al., 2024⁶⁶), and skin-on-chips (Abaci et al., 2015¹⁷⁶; Mori et al., 2016¹¹¹; Lim et al., 2018¹⁶; Mori et al., 2018¹¹⁷; Li et al., 2023¹⁰⁹; Gu et al., 2024¹¹⁷)

Composition of the skin’s ECM

ECM of the epidermis

Epidermis is composed of a basal layer of epidermal stem cells that continuously differentiate into the stratum spinosum, stratum granulosum, stratum lucidum, and ultimately the stratum corneum. The epidermal extracellular matrix above the basal layer contains only trace amounts of polysaccharides and lipids. Basement membrane is a specialized extracellular matrix²², comprising core components such as COL17, COL4, COL7, and COL18²³, along with laminin, nidogen, and perlecan²⁴. COL4 and laminin 332 form a dense meshwork, with nidogen and perlecan acting as bridging proteins, thereby maintaining the structural integrity of the basement membrane. Anchoring complex is a highly organized assembly of α6β4 integrin, laminin isoform 332 (LM-332), COL7, and COL17, which connects basal keratinocytes upward, spans the basement membrane, and anchors the dermal papillary layer downward²⁵, thus forming a complex interactive interface between the dermis and epidermis²². Hemidesmosomes anchor keratinocytes to the basement membrane by linking cytoplasmic intermediate filaments to laminin in the basement membrane via integrin α6β4²⁶. COL7, a major component of the anchoring protofibrils²⁷, extends into the dermal collagen fiber network and binds to type I collagen and anchoring the basement membrane to the dermis²⁸. Anchoring complex confers the epidermis with the ability to resist mechanical stress and provides the basement membrane with robust adhesion and mechanical stability. This is essential for maintaining epidermal stem cells⁵.

ECM of the dermis

Dermis is derived from pluripotent mesenchymal stem cells (MSCs), which differentiate into fibroblast progenitors of the papillary and reticular layers, thereby subdividing the dermis into these two distinct layers²⁹. The ECM of the papillary layer is thin and loose and functions as a cushion²⁹, with a high density of fibroblasts surrounded by type I collagen, type III collagen, and elastin^30,31. Periosteal proteins are mainly expressed in the papillary layer, maintain the structural integrity of the papillary layer in combination with tenascin C and fibronectin, regulate collagen structure^32,33, and promote keratinocyte proliferation via the paracrine action of IL-6 from fibroblasts³⁴. Reticular layer constitutes the primary portion of the dermis. Despite the low density of fibroblasts²⁹, the thick and dense type I collagen bundles directly contribute to the tensile strength of skin³⁵. Reticular layer plays a key role in determining skin elasticity, with fibronectin 2 binding to protofibronectin 1 to regulate homeostasis within elastic fibers³⁶. Matricellular proteins, including elastin, fibronectin 4, and fibronectin 5, are primarily expressed in the reticular layer^37,38 and are essential for the formation of elastic fibers.

Changes in the ECM during skin aging

Changes in the ECM of the epidermis during aging

Changes in basement membrane stiffness and composition during aging impair the repair of epidermal damage¹⁴. AGEs increase basement membrane stiffness^39,40. Laminin-332 and COL17A1 are predominantly expressed in the uppermost protrusions of the rete ridges, where interfollicular epidermal stem cells (IFE-SCs) are located⁴¹. Increased ECM stiffness at the top of the rete ridge inhibits the expression of core components of the anchoring complex and the maturation of hemidesmosomes^5,42. The absence of anchoring structures leads to the flattening of the dermal–epidermal junction (DEJ) and diminishes the self-renewal capacity and stemness of epidermal stem cells. In contrast, the transcription factor KLF4 promotes keratinocyte differentiation to maintain epidermal barrier function, while the YAP1/TAZ-TEAD network maintains keratinocytes in an undifferentiated state^43,44. Rigid environment activates the expression of KLF4 and represses the transcription of the YAP1/TAZ-TEAD network⁵, inhibiting keratinocyte self-renewal and driving them into committed differentiation, and resulting in epidermal thinning. COL7 binding to elastic fibers is diminished, thereby leading to a weakened anchoring relationship between the epidermal basement membrane and the dermis^45,46, along with reduced elastic fiber tone and dermal elasticity. Additionally, perlecan recruits and maintains laminin 511/521, which is associated with keratinocyte regeneration^47,48. Perlecan expression is downregulated in aging keratinocytes⁴⁹, and further leads to reduced keratinocyte regeneration in the basal layer.

Changes in the dermal ECM during aging

Changes in the dermal ECM during aging include degradation of collagen fibers, decreased collagen production, and the development of an inflammatory microenvironment^13,50. Fibroblasts produce matrix metalloproteinases (MMPs) that degrade ECM proteins⁵¹. MMP-1 levels are significantly elevated in both intrinsically senescent and photoaged skin. The mechanical microenvironment of the dermal ECM profoundly affects skin aging⁵². Collagen degradation impairs the mechanical properties of the dermal ECM, leading to a senescence-associated secretory phenotype (SASP) in fibroblasts, and thereby reducing cell proliferation and activating MMPs^53,54. TGF-β/Smad pathway plays a critical role in ECM synthesis¹³, and TGF-β/Smad signaling is downregulated in the senescence-associated secretory phenotype^55,56. Mitochondrial dysfunction leads to the accumulation of reactive oxygen species (ROS), a common manifestation of both intrinsic aging and photoaging^57,58. ROS accumulation activates two signaling pathways, MAPK and NF-κB⁵⁹, which together inhibit collagen production by dermal fibroblasts and reduce collagen fiber density⁶⁰. Dermal ECM stiffness increases with age, while the YAP/TAZ pathways in fibroblasts are downregulated⁶¹, leading to the activation of the cGAS-STING pathway. This generates inflammation that damages the dermal ECM⁵². To provide an integrated overview, the key alterations in ECM components and mechanistic drivers during skin aging are summarized in Table 1.

Table 1.

Key ECM alterations in aging skin and mechanistic drivers

ECM components and compartments	Alterations with aging	Mechanisms and mediators	Reference
COL1 & COL 3 (dermis)	COL1 & COL3 ↓; neosynthesis ↓; fibril fragmentation ↑; network thickness ↓; misorientation	ROS → MAPK & NF-κB ↑; UV → AP-1 ↑ → MMP-1/-3/-9 ↑; fibroblast mechanical tension ↓ →TGF-β–Smad signaling ↓	54,59,60,144
COL4 & COL7 (BM)	COL4 & COL 7 ↓ (photoaging); anchoring fibrils ↓; rete ridges flatten	UV →MMP-2/-9/-13 → BM proteins ↓; AGE crosslinking ↑ → BM stiffness ↑; keratinocyte-driven BM assembly ↓	5,45,145,146
LM-332 (BM)	LM-332 function ↓; DEJ & BM continuity ↓	Plasmin, MMP-2/-14, etc ↑ → LM-332 processing ↓; α6β4-integrin ↓ → linear LM-332 deposition ↓	147,148
COL18 (BM)	COL18 assembly ↓; DEJ & BM continuity ↓	Cathepsin L & MMPs ↑ → endostatin ↑ & intact COL18 ↓ → BM deposition/assembly ↓	149,150
Perlecan (BM)	Perlecan HS chains ↓ (photoaging)	UV → heparanase ↑ → perlecan HS chain loss	49
FBN1 microfibrils & elastin fibers (dermis)	FBN1-microfibrils↓; elastin fibers thin/fragment ↑ (photoaging)	UV→ elastases/MMP-12, etc ↑ + oxidative/carbonyl damage + crosslinking changes → elastic-fibers ↓	151–153
Hyaluronic acid (epidermis & dermis)	HA total ↓ (esp. epidermis); low-molecular-weight-HA fraction ↑ (photoaging)	Dysregulation of HAS and HYAL; ROS-driven depolymerization ↑	154,155
Decorin (dermis)	GAG length ↓ (photoaging)	UV→Neutrophil-elastase ↑ → decorin cleavage ↑	156
Versican (dermis)	Intrinsic aging→ GAG length ↓; Photoaging→ versican ↑(in solar elastosis)	ADAMTS-mediated versican cleavage ↑; inflammatory and UV promote this remodeling	157,158
AGE crosslinking (dermis & BM)	AGEs ↑; irreversible crosslinks ↑; susceptibility to proteolysis ↓; insolubility and stiffness ↑	Maillard/glycoxidative pathways → AGE adducts (CML, CEL) & crosslinks (pentosidine, glucosepane) ↑	146,159

Note: Symbols: ↑/↓ denote increase/decrease; “→” indicates putative causal direction or functional consequence. Scope: Unless specified, “aging” covers intrinsic and/or photoaging; rows with site-specific differences explicitly note “intrinsic” or “photoaging”

BM basement membrane, DEJ dermal–epidermal junction, ECM extracellular matrix, ROS Reactive Oxygen Species, AP-1 Activator Protein-1, LM-332 laminin-332, AGE advanced glycation end products, CML carboxymethyllysine, CEL carboxyethyllysine, HAS/HYAL hyaluronan synthase/hyaluronidase, MMP matrix metalloproteinase, FBN1 fibrillin-1, GAG(s) glycosaminoglycan(s), ADAMTS A Disintegrin And Metalloproteinase with ThromboSpondin motifs

In summary, the mechanisms assembled in this section delineate what “faithful recapitulation” should look like in vitro: a credible skin-aging model should preserve a functional dermal–epidermal anchoring apparatus to transmit basement membrane mechanics to keratinocyte fate via the KLF4 and YAP/TAZ axes. Meanwhile, dermal compartment exhibit the canonical remodeling program in which MMP-1-driven collagen loss coexists with downregulated TGF-β/Smad signaling, ROS-activated MAPK/NF-κB, and age-associated stiffening accompanied by cGAS–STING-mediated inflammation. Contemporary platforms only partially reproduce this cascade—static reconstructed skin helps assess ECM mechanical properties and compositional changes as well as barrier function⁶² (Fig. 3d, e). Bioprinted tissues can reconstruct skin texture and surface microreliefs^63,64 (Fig. 4c, d), and skin organoids can realistically recapitulate microanatomy related to skin appendages^65,66 (Fig. 5). Besides, skin-on-a-chip can impose controlled mechanical loading and yield the expected molecular and structural outcomes^16,17 (Fig. 6c, d). Nevertheless, full recapitulation remains elusive across model classes; this shortfall reflects practical constraints: long-term, stable co-culture with resident immune cells is fragile; aged and site-matched primary cells are scarce. It remains challenging to reproduce basement-membrane mechanics together with DEJ microtopography and the correct protein composition at the relevant spatial scales. In addition, standardized and high-throughput readouts are still limited (despite recent progress toward harmonization). Guided by these benchmarks, the remainder of this review will systematically compare how different models recapitulate the mechanisms above and delineate current progress and remaining gaps on this basis.

Fig. 3. 3D static reconstructed human skin models.

a Schematic diagram of the construction of a full thickness human reconstructed skin model. b Static reconstructed full-thickness skin sections display a well-differentiated epidermis and a dermis containing fibroblasts⁸⁶. c The first static melanin skin model is presented, in which melanin is clearly visible⁸⁷. d The first static full-thickness skin aging model is presented: (i) Histological sections of young and aged skin equivalents; (ii–iii) Immunofluorescence staining for Filaggrin and Elastin in young and aged skin models, respectively⁶². e (i) Morphology of fibroblasts in young and senescent skin models; (ii) β-galactosidase activity; (iii) Proliferative capacity as determined by BrdU incorporation; (iv) Ki67 staining; (v) p53 activity; (vi) Reactive oxygen species (ROS) content, and (vii) Matrix metalloproteinase-1 (MMP-1) concentration and activity⁶²

Fig. 4. 3D bioprinted human skin models.

a 3D Printing of Skin Tissues Using Photopolymerized Bioinks Containing GelMA, SilMA, and Platelet-Platelet-Producing Growth Factor (PPRF)⁹⁴. b Schematic flowchart of a 3D bioprinted immunologically active skin model. The model includes layer-by-layer printing of multilayered skin tissue structures, air–liquid interface incubation (ALI), and inoculation of macrophages at the bottom of day 14 with exogenous stimuli applied for immune function studies⁹⁴. c (i) Acquisition of real skin surface microtexture structure using optical coherence tomography (OCT). (ii) Construction of an in vitro skin model with adjustable wrinkle depth in combination with collagen stamping technology⁶³. d Based on the Voronoi algorithm for modeling, the microrelief structure of the skin at different ages was simulated by printing with Digital Light Processing (DLP) technology⁶⁴

Fig. 5. Skin organoid models.

a Stepwise induction and structural characterization of human iPSC-derived skin organoids. (i) Schematic illustration of skin organoid induction from human pluripotent stem cells (hPSCs). By stage-specific modulation of TGFβ/BMP and FGF signaling pathways, hPSCs sequentially differentiate into surface ectoderm and cranial neural crest cells (CNCCs), ultimately forming skin organoids containing appendages such as hair follicles. (ii) Immunofluorescence image of a day-75 organoid showing stratified epidermal structure. The basal layer is marked by KRT5⁺/TFAP2A⁺, the intermediate layer by TFAP2A⁺, and the peripheral layer by KRT17⁺. (iii) Dark-field image of a skin organoid displaying pigmented hair shafts. (iv) Electron microscopy image showing mature epidermal stratification and melanocytes localized at the basal layer⁶⁵. b Structural, molecular, and cellular alterations in human iPSC-derived skin organoids (SkOs) following sUV exposure. (i) Schematic illustration of sUV-induced damage in skin organoids. (ii) Masson’s trichrome staining shows reduced collagen fiber density in the dermis after sUV exposure. (iii) Bright-field image reveals thinning of the hair shaft; asterisk indicates hair shaft, black arrows denote outer root sheath (ORS), and yellow arrows indicate dermal papilla (DP). (iv) Confocal imaging and quantitative analysis demonstrate decreased KRT5 expression and regionally increased Cleaved Caspase-3⁺ apoptotic cells in hair follicles. (v) Immunofluorescence and quantification show elevated IL-1β expression following sUV exposure. (vi) Immunofluorescence and quantification reveal upregulated expression of MMP-1⁶⁶

Fig. 6. Skin microphysiological systems.

a (i) Schematic of air–liquid interface-induced differentiation of human keratinocytes and melanocytes on a skin-on-a-chip platform, accompanied by histological images of the resulting skin model cultured on-chip¹⁰⁹. (ii) Photograph of the epidermis-on-a-chip system showcasing the integrated microfluidic system designed to regulate nutrient delivery and environmental conditions. b Schematic representation of a skin-on-chip model with integrated perfusable vascular channels, HE-stained sections of both perfused and non-perfused skin models, and immunofluorescence staining for the vascular endothelial cell marker CD31¹¹¹. c (i) Schematic representation of the flexible skin-on-a-chip architecture; (ii) Physical illustration of the compressive stress-driven system; (iii) Schematic diagram of cyclic compression induced by a mechanical stimulus-driven system; (iv) Normalized relative expression of the filaggrin, fibronectin, involucrin, keratin 10, β-galactosidase, and PSMD8 genes with respect to the duration of air exposure in both experimental and control groups; (v) Changes in epidermal thickness over the course of air exposure in both experimental and control groups¹⁷. d (i) Depictions of the width and depth of skin wrinkles, (ii) an image of a real wrinkle, (iii) A schematic illustration of wrinkle formation in a skin equivalent under uniaxial stretching, alongside a photograph of a skin-on-a-chip device designed to simulate wrinkle development¹⁶

The evolution of in vitro models promotes precise simulation of skin aging

Over recent decades, in vitro models of aging skin have progressed from two-dimensional monolayers to static three-dimensional constructs and to bioprinted tissues, organoids, and skin-on-a-chip platforms recently. The value of these systems lies in their capacity to elucidate the dynamic remodeling of the extracellular matrix (ECM) during aging, encompassing (i) temporal shifts in synthesis–degradation fluxes (e.g., deposition and fragmentation of collagen and elastic fibers, accumulation of advanced glycation end product-mediated cross-linking), (ii) mechano-biochemical coupling whereby matrix stiffness and loading rhythms reprogram cell states and thereby remodel the ECM, and (iii) region-specific changes driven by spatial gradients and microenvironmental heterogeneity (e.g., alterations in ECM composition and stiffness with consequent effects on barrier function). Collectively, these values underpin advanced models, thereby enabling immune cell incorporation and precise modulation of the skin aging microenvironment. In the following sections, we adopt the capacity to capture ECM dynamics as the organizing principle to synthesize model-construction strategies, delineate their domains of applicability, and provide targeted comparisons.

Strategies for constructing skin aging models

Skin aging development is complex, and it is extremely challenging to create in vitro skin aging models. For both 2D and 3D skin models, current methods for inducing aging in skin models can be categorized into the following five approaches:

1. Chemically induced cellular senescence: Cellular senescence is one of the nine hallmarks of organismal aging⁶⁷. However, the time required for fibroblasts to reach their normal division limit and mimic cellular senescence has been prolonged. A quicker alternative is to induce cellular senescence using mitomycin C, hydrogen peroxide, or adriamycin^62,68.

2. Extraction of primary skin cells from senescent donors: Studies have shown differences between primary cells extracted from senescent donors and senescent cells induced in vitro. Constructing models from primary skin cells of senescent donors better reflects the physiological environment of senescent skin⁶⁹. Additionally, in vivo dermal senescent fibroblasts account for only 20–60% of the fibroblasts under physiological conditions. Studies have shown that incorporating various proportions of senescent fibroblasts into the dermal layer of skin models can simulate different degrees of aging phenotypes⁶⁸.

3. Induction of AGEs: AGEs accumulate in the skin with age and accelerate organismal aging¹⁰. Hydrogels treated with sodium glyoxylate and sodium cyanide borohydride induce carboxymethyl lysine expression in the dermis, disrupting the ECM structure⁷⁰. Repeated glyoxal applications to skin models also increase CML levels, impairing barrier function, reducing ECM protein synthesis, and creating an inflammatory microenvironment⁷¹.

4. In vitro simulation of environmental exposure: Skin model is exposed to external stimuli to simulate the effects of environmental factors on the aging phenotype. For example, irradiation with UVA and γ-rays induces photoaging of the skin^72,73, and cigarette extracts have been used to simulate the aging effects of cigarette smoke on the skin⁷⁴.

5. Simulation of mechanical stress: Mechanical stress is a factor promoting skin aging. Simulations include intermittent vertical compressive stress applied to skin models to mimic circadian rhythms¹⁷ and uniaxial transverse stretching using magnets to replicate tensile stress¹⁶.

Skin explants and 2D skin senescence models

As shown in Fig. 2, in vitro skin models have evolved from simple to complex. Skin explants are one of the first skin models to be cultured in vitro⁷⁵. They contain all the cell types and extracellular matrix components of normal skin, making them simpler to culture than isolated cells or to mimic the skin ECM in vitro⁷⁶. However, obtaining large quantities of skin from donors is virtually impossible due to ethical constraints. Human skin samples vary in thickness, cell type, and cell number, affecting the reproducibility and consistency of studies⁷⁷. Skin explants lose tissue integrity after 14 days in culture, and the keratin-forming cells within the tissue gradually differentiate into keratin^78,79.

As an early static approach, 2D skin-aging models typically consist of monolayer cultures of fibroblasts, keratinocytes, or melanocytes and allow highly controlled assays of ECM component expression (e.g., collagen and elastin). Common readouts for aging or anti-aging efficacy include MMP activity, senescence-associated β-galactosidase, and expression of the aging-related gene p16^80–83. However, they cannot capture ECM spatial organization, dynamic cell–ECM interactions, or aging-associated shifts in ECM composition because these models do not recapitulate the stratified three-dimensional architecture or a native extracellular matrix^80,81. In addition, the lack of an intact barrier makes them poorly suited to evaluating permeability and transepithelial transport of anti-aging agents⁸².

3D static reconstructed human skin aging models

Reconstructed three-dimensional (3D) skin models, built on collagen or other natural/synthetic scaffolds, provide a tissue-like architecture that approximates the physical and biochemical context of the native extracellular matrix (ECM) enabling interrogation of ECM dynamic remodeling and matrix-dependent signaling during aging. In general, full-thickness equivalents comprise an epidermal compartment generated by air–liquid-interface differentiation of keratinocytes over a hydrogel/collagen scaffold⁸⁴ (Fig. 3a) and a dermal compartment formed by fibroblasts embedded within the matrix^85,86 (Fig. 3b); melanocytes can be incorporated to recapitulate pigmentation^87,88 (Fig. 3c).

With these features, 3D reconstructed skin models used for aging studies offer the following advantages: (1) recapitulate barrier function, enabling assessment of epidermal thickness (Fig. 3d-i), keratinocyte viability⁸⁹, and skin integrity/permeability; (2) replicate the basal layer⁹⁰, supporting evaluation of ECM compositional changes (Fig. 3d-ii, iii), and allow direct visualization of morphology, proliferative activity, and senescence-associated secretory phenotypes (Fig. 3e); and (3) are supported by established commercial platforms (Episkin, Sterlab, MatTek)⁹¹.

However, the model also has significant drawbacks: (1) long culture time (3–4 weeks) and short post-maturation viability (approximately 2 weeks)⁹², restricting long-term aging observations⁸⁴; (2) absence of perfusable vasculature or endothelial layers, limiting immune-cell transmigration studies; and (3) inability to replicate mechanical stress.

3D bioprinted skin aging models

3D bioprinting technology utilizes biomaterials, living cells, and growth factors to construct models that recapitulate organ phenotypes and functions⁹³. Compared with 2D/3D static models that lack the ordered architecture of native skin and are unable to replicate structural ECM changes such as wrinkle formation, 3D bioprinting is poised to advance the construction paradigm to a “programmable and controllable” stage. By spatially addressing cells and matrix in combination with multi-component bioinks, it enables fine regulation of the ECM microenvironment within an immune-involved framework and the introduction of parameterizable surface and mechanical features, providing a feasible engineering basis for integration with imaging or computational design methods to progressively achieve controllable reconstruction of aging-related phenotypes.

Demonstrations of this potential already exist in vascularization and immune competence. For instance, Bhar et al.⁹⁴ employed a layer-by-layer strategy with a photopolymerizable bioink containing platelet-derived growth factors to fabricate a pre-vascularized, immunocompetent model (Fig. 4a). After 14 days of air–liquid interface culture, human macrophages were introduced into the lower chamber of a Transwell and locally stimulated to elicit an immune response (Fig. 4b), thereby establishing an immune-involved platform for subsequent ECM microenvironment modulation.

Beyond vascularization and immune competence, 3D bioprinting can also reproduce, at the micrometer scale, surface microrelief and mechanical attributes associated with skin aging^63,64 (Fig. 4c, d). With advancing age, stiffening of the basement membrane and loss of anchoring-complex components flatten rete ridges and thin the epidermis; concomitant degradation of collagen and elastin compromises structural integrity, manifesting as wrinkles and laxity (Fig. 1). By extracting key parameters of surface microrelief—such as depth, pitch, and angle—via optical coherence tomography (OCT), digital models of aged skin can be generated⁶³ (Fig. 4c). Collagen stamping enables full-thickness models with tunable wrinkle width and depth in vitro, and CAD designs based on a Voronoi algorithm can map printing paths directly onto digital light processing (DLP) systems⁶⁴, thereby reconstructing age-specific microrelief meshes in hydrogels at high resolution (Fig. 4d).

Nevertheless, there are several limitations for 3D bioprinting: (1) expensive equipment, low economic efficiency, and high technical barriers⁹⁵; (2) poor printability and prolonged cross-linking of collagen-based materials⁹⁶, which complicate the fabrication of structures with specified shapes and compositions; (3) limited resolution for skin microstructures and potential loss of cell viability during printing⁹⁷; and (4) insufficient self-organization among multiple cell types in bioinks⁹⁸—for example, endothelial cells have difficulty undergoing directed self-assembly into ordered capillaries.

Skin organoid aging models

Skin organoid models represent a critical leap in skin aging simulation—from static reconstruction to dynamic self-assembly. Unlike manually assembled skin equivalents, organoids derived from pluripotent stem cells (ESCs, iPSCs) or adult stem cells (ASCs) undergo in vitro organogenesis within 3D matrices supplemented with defined growth factors⁹⁹, giving rise to structures that recapitulate the microanatomical organization and functions of native skin (https://www.fda.gov/science-research/advancing-alternative-methods-fda/about-alternative-methods) (Fig. 5a-i), including a stratified epidermis and dermis, as well as skin appendages (e.g., pigmented hair follicles and sebaceous glands)^65,99, (https://www.fda.gov/science-research/advancing-alternative-methods-fda/about-alternative-methods) (Fig. 5a-ii–iv). By virtue of intrinsic self-organization and coordinated multi-lineage morphogenesis, organoids can model, in vitro, the dynamic ECM–cell interactions characteristic of skin aging, providing a platform of heightened physiological relevance for dissecting aging mechanisms.

Organoids provide clear advantages for in vitro modeling of extrinsic drivers of skin aging and mechanistic interrogation of intrinsic processes. Using hair-bearing human organoids, Kim et al.⁶⁶ showed that sUV irradiation reproduced canonical photoaging hallmarks—collagen depletion, pro-inflammatory cytokine upregulation, and increased apoptosis (Fig. 5b)—and revealed hair shaft thinning and outer root sheath compromise (Fig. 5b-ii), a phenomenon not previously observed in vitro. Extending beyond photodamage, Miyake et al.¹⁶³ modeled IR-induced genotoxic stress in iPSC-derived keratinocyte organoids: 2 Gy γ-irradiation triggered nuclear 53BP1 foci in basal cells, indicating activation of the DNA damage response (DDR), while differentiated keratinocytes displayed markedly delayed repair kinetics, indicating reduced proliferative capacity. Together, these studies illustrate how organoids enable analysis of differentiation-linked DDR dynamics and aging-related genomic instability.

Conventional in vitro models induce cellular senescence via prolonged culture or stressors (e.g., mitomycin C, hydrogen peroxide), reproducing select phenotypes but not the in vivo-like physiological dynamics of tissue aging. By recapitulating organogenesis and lineage differentiation, organoids provide a more physiologically relevant setting—with better preservation of ECM–cell interactions—to study normal and pathological aging¹⁰⁰. When derived from iPSCs bearing senescence-associated variants, they can reconstruct aging trajectories at structural and molecular scales¹⁰¹; iPSC-based approaches further enable reprogramming of elderly donor cells to generate patient-specific models that preserve individual genetic and epigenetic signatures¹⁰².

Despite these advantages, several limitations hinder the current application of skin organoids in aging simulation. Under in vitro conditions, it remains uncertain whether organoids can faithfully recapitulate senescence-related features, largely due to the absence of aging niches that recapitulate the in vivo aging microenvironment^103–105. In addition, reprogramming somatic cells from aged donors to iPSCs is inefficient and may reset senescence-associated epigenetic marks^102,106, whereas direct isolation of adult stem cells (ASCs) from aged donors is technically difficult and low-yield. Finally, high inter-laboratory variability and the lack of standardized protocols present significant barriers to reproducibility and broader translational use⁹⁹.

Skin-on-a-chip aging models

Skin-on-a-chip is the most dynamic stage in the progression of in vitro skin aging models, integrating mechanical stimuli (stretching, compression, shear), microfluidic perfusion, and multicellular co-culture to recreate a time-varying, in vivo-like microenvironment. Risueño et al.¹⁰⁷ elaborated on the critical role of skin-on-a-chip in modeling dynamic skin changes, particularly in drug screening and physiological studies. This system enables programmable mechanochemical coupling under physiologic flow conditions and supports functional immune-cell integration, while precisely regulating the skin aging microenvironment. In this context, skin-on-a-chip—an organ-on-a-chip (OOC) format within the microphysiological systems family that recreates organ-level functions in engineered, perfusable microenvironments (https://www.fda.gov/science-research/advancing-alternative-methods-fda/about-alternative-methods)—uses microfluidic perfusion to continuously supply nutrients, remove metabolic waste, and apply controlled shear forces, thereby enhancing epidermal differentiation and stratification in vitro⁷⁷. Experimental evidence shows that Wang et al.¹⁰⁸ directly reconstructed full-thickness skin on a microfluidic system, measuring that the average stratum corneum thickness of the equivalent skin on the chip was 1.6 times that of the static skin model. Polarized columnar basal keratinocytes were observed in this microphysiological system, whereas only cuboidal basal keratinocytes were observed in static skin models. Additionally, the skin microphysiological system developed by Chen et al.¹⁰⁹ had an epidermal thickness exceeding 50 μm (Fig. 6a), closely resembling the epidermal thickness of human skin with an intact barrier function, while the epidermal thickness of a 3D static skin model from the same timeframe was only 28–43 μm¹¹⁰. These data confirm that microfluidic control not only operationalizes the dynamic microenvironment described at the outset but also yields a more defined epidermal structure than static models.

According to this dynamic framework, skin microphysiological systems operationalize this paradigm by establishing lumenized, perfusable vascular compartments^111–113 (Fig. 6b)—for example, Mori et al.¹¹¹ first created hollow dermal channels by removing nylon threads, then perfused an endothelial cell suspension to form endothelium-lined lumens, a configuration suitable for studying the vascular absorption of anti-aging drugs. These platforms enable immune–vascular integration^114–116; in a representative model, Ren et al.¹¹⁶ deposited an endothelial monolayer on a chip and used microfluidic chemokine gradients to mediate T-cell migration across the endothelium to defined skin layers during inflammation, providing a platform to examine immune system influences on aging. Crucially, skin-on-chips can apply defined tensile and compressive regimens—an advantage not achievable in static models. Across recent studies, mechanical cues shape aging phenotypes in a rhythm- and dose-dependent manner: prolonged circadian-like compression and rapid cyclic uniaxial stretch reproducibly yield pro-aging features (greater wrinkles/roughness, elevated ROS and β-gal, reduced TGF-β receptor signaling, and suppressed collagen synthesis), whereas low-magnitude periodic tensile loading produces mixed, context-dependent effects on barrier-associated readouts and matrix organization^{16,17,117,118}. Chronic or rapid cyclic stress tends to bias degeneration, while carefully titrated low-magnitude tensile regimens may help maintain barrier function under specific conditions. Taken together, these observations underscore that the trajectory of skin aging under mechanical loading is governed by key stimulus parameters—load waveform and frequency, strain amplitude, duty cycle, and exposure duration. To enhance reproducibility and foster broader adoption, standardization and throughput are advancing, Gu’s team (https://std.samr.gov.cn/gb/search/gbDetailed?id=25940C3CEF7E8A9AE06397BE0A0A525A) completed China’s first national standard in the field of organ-on-chips, the General Technical Requirements for Skin-on-Chips (GB/T44831-2024). This will aid in standardizing scientific research. High-throughput generation and analysis reduce randomness and variability, essential for biological experiments and clinical trials involving a wide range of parameters and processing steps¹¹⁹.

Despite these advantages, several limitations remain in the skin microphysiological system: Long-term co-culture of immune cells (e.g., mast cells and macrophages) in skin-on-a-chip models has yet to be achieved. Although Li et al.¹²⁰ incorporated fibroblasts and macrophages to study their interactions in skin trauma, these cells do not directly interact but instead communicate through secreted factors via microchannels with a 4 μm diameter, which does not fully replicate normal skin architecture where macrophages are typically found in the dermis. Additionally, skin appendages, including hair follicles, sweat glands, and sebaceous glands, have not yet to be incorporated into skin-on-chips¹²¹. However, integrating organoid culture iPSCs, and microfluidic control systems offers the potential to develop skin-organoid-on-chips that incorporate skin appendages. Moreover, the primary cell sources for skin-on-chips are currently limited to tissues such as boys’ foreskin and young female breast skin, with cells from other regions (e.g., face, neck) or from older populations not yet utilized, limiting the model’s ability to reflect the diversity of skin types across different populations. Finally, the complexity and cost of microfluidic devices and tissue culture systems remain a barrier¹²², although these costs are expected to decrease as the technology matures.

To provide a comprehensive overview, a comparative summary of current engineered 3D skin models used for aging research was compiled, including reconstructed skin, bioprinted models, organoids, and skin-on-a-chip systems. These models differ in their seed cells, scaffolding materials, structural complexity, culture methods, and aging induction strategies (Table 2).

Table 2.

Comparative table of engineered 3D skin models for aging research

Model type	Aging type	Seed cells	Bioscaffold	3D Structural features	Culture method	Aging induction method	Reference
3D Static Reconstructed Skin	Photoaging	NHKs, NHMs, NHDFs	Bovine type I collagen	Self-contracted dermis (HDF + bovine collagen) with NHK/NHM (10:1) epidermis	ALI, Static culture	UVA (2 J/cm²) + UVB (15 mJ/cm²), single exposure	Ko et al.73, 2024
	Photoaging	NHDFs, NHEKs	No exogenous scaffold	Three-layer self-assembled fibroblast sheets fused into dermis, overlaid with NHEKs	ALI, Static culture	3% cigarette smoke extract (7 days) + sequential UVA (5–20 kJ/m²)	Grenier et al.74, 2023
	Photoaging	NHDFs, NHEKs, THP-1 monocytes	Fibrin gel with type IV collagen coating (animal source unknown)	THP-1 cells embedded in fibrin-based dermis; overlaid with NHEKs	ALI, Static culture	UVA (365 nm, 80 J/cm²), single exposure	Phuphanitcharoenkun et al.160, 2024
	Photoaging	NHDFs, NHEKs	Bovine type I collagen	EpiDerm™ commercial full-thickness skin model; recapitulates collagen XVIII at DEJ	—	Repeated intermittent UV-B (280–315 nm, 0.56 mW/cm², 18 s) irradiation	Yonezawa et al.126, 2019
	Intrinsic aging	NHDFs, NHEKs	Type I & III collagen + chondroitin sulfate + chitosan (animal source unknown)	Dermis formed on collagen–chitosan scaffold; stratified NHEKs form epidermis	ALI, Static culture	Fibroblasts pretreated with MMC (200 nM, 48 h) prior to reconstruction	Diekmann et al.62, 2016
	Intrinsic aging	NHDFs, NHEKs	Bovine type I collagen	Dermis constructed with varying ratios of SIPS fibroblasts; overlaid with NHEKs	ALI, Static culture	Fibroblasts senesced via H₂O₂ (100 µM ×9) or Doxorubicin (200 nM ×2), then embedded	Weinmüllner et al.68, 2020
	Inflammaging	NHDFs, NHEKs	Bovine type I collagen	EpiKutis™ commercial model with stratified epidermis and functional dermis	—	0.1% SLS stimulation for 30 min to induce inflammation	Jiang et al.161, 2023
	Glycated ECM aging	HDFs, NHEKs	Bovine type I collagen (pre-glycated with specific AGEs)	Fibroblasts embedded in AGE-modified collagen; overlaid with NHEKs	ALI, Static culture	AGE pre-glycation of collagen: CML/CEL (100 mM, 4 h); MG-H1 (1 mM, 24 h); Pentosidine (ribose + N-acetyl-arginine, 14 days)	Pageon et al.10, 2015
	Glycated ECM aging	NHDFs, NHEKs	Bovine type I collagen	EpiDermFT model with dermal and stratified epidermal structure	ALI, Static culture	500 μM methylglyoxal (MGO) in medium for 7 days	Markiewicz et al.162, 2022
	Glycated ECM aging	NHDFs, NHEKs	Rat tail type I collagen (partially pre-glycated)	Fibroblasts embedded in glycated collagen; stratified keratinocytes on top	ALI, Static culture	0.25 mM sodium glyoxylate treatment for 24 h to pre-glycate collagen	Pennacchi et al.70, 2015
3D Bio-printed Skin	Wrinkle-mimicking aging model	NHDFs, NHEKs	Type I collagen + fibrinogen/thrombin composite (animal source unknown)	Dermal furrows reconstructed via collagen stamping to mimic aged skin morphology	ALI, Static culture	Physical stamping of dermal furrows to simulate skin sagging and wrinkling	Min et al.63, 2023
	Wrinkle-mimicking structural aging	None	Cell-free; no biomimetic scaffold	DLP-printed microreliefs with age-specific Voronoi geometry simulating aged skin surface	—	Simulated epidermal stiffening and compression over aged microrelief topography	Sun et al.64, 2024
Skin Organoid	Photoaging	hiPSCs (differentiated into keratinocytes and fibroblasts)	Type I collagen gel (animal source unknown)	Self-organized skin organoids with hair follicles, dermis, stratified epidermis	ALI, Static culture	sUV (UVA + UVB, total 50 kJ/m²), 3 exposures over 6 days	Kim et al.66, 2024
	Radiation-induced aging	hiPSCs (differentiated into keratinocytes), NHDFs	Type I collagen (animal source not stated)	Full-thickness skin organoid constructed with hiPSC-derived keratinocytes and HDFs showing stratified epidermis	ALI, Static culture	Single γ-ray exposure at 2 Gy	Miyake et al.163, 2019
Skin-on-a-chip	Mechanical stress-induced aging	HDFs, NHEKs	Type I collagen cross-linked with sulfo-SANPAH (animal source unknown)	Full-thickness model embedded in flexible PDMS chip with compression-driving system	ALI + gravity-driven microfluidic system	Circadian compression (12 h/day for 28 days)	Jeong et al.17, 2021
	Mechanical stress-induced aging	NHDFs, NHEKs	Porcine-derived type I collagen gel	Full-thickness model integrated into PDMS chip with electromagnet-driven wrinkle system	ALI + perfusion microfluidic system	Cyclic uniaxial stretching (10% strain, 0.01–0.05 Hz, 12 h/day, 7 days)	Lim et al.16, 2018
	Mechanical stress-induced aging	NHDFs, NHEKs	Rat tail type I collagen gel	Full-thickness skin model embedded in PDMS-based MSSC device with electromagnetic stretching system	ALI + gravity-driven microfluidics	Cyclic uniaxial stretching (5% strain, 0.01 Hz, 12 h/day, 28 days)	Kim et al.118, 2022
	Inflammaging	NHEKs, NHDFs	Rat tail type I collagen gel	Full-thickness skin equivalent integrated into pumpless PDMS chip with dual-channel perfusion and ALI	ALI + gravity-driven perfusion	IL-4/IL-13 (5–15 ng/mL), 3–14 days	Kim et al.164, 2022
	Radiation-induced aging	Human skin fibroblasts (HCA2), HUVECs	Type I collagen (porcine-derived)	Collagen-embedded fibroblasts with perfusable endothelial channel in PDMS chip	Perfusion-based flow culture	Fibroblasts pretreated with γ-irradiation (10 Gy) and BI2536 (100 nM) for 14 days	Pauty et al.72, 2021

ALI air–liquid interface, HDFs human dermal fibroblasts, NHEKs normal human epidermal keratinocytes, NHMs normal human melanocytes, hiPSCs human induced pluripotent stem cells, sUV solar ultraviolet (UVA + UVB)

Bioscaffold materials were identified from source publications. If the species origin of type I collagen (e.g., bovine, porcine, rat) was not specified, it is reported as “animal source unknown”.

Aging induction methods refer to either pre-treatment of seed cells prior to model construction, or direct stimulation of reconstructed models, as described in each original study.

Building on the benchmarks synthesized above, we contrast model types across structure-, mechanics-, and function-level fidelity and across practical axes—accessibility, throughput, and external validation (see Table 3 for a side-by-side summary). No single model fully recapitulates aging; model selection should be driven by the study question. In brief, reconstructed skin maximizes accessibility and throughput for routine functional readouts; 3D bioprinted tissues raise structural/mechanical fidelity at the cost of setup and printing complexity; organoids prioritize appendage biology while trading throughput for fidelity; and skin-on-a-chip provides the strongest control of dynamic regimens while remaining less accessible. These trade-offs, together with emerging in vivo/clinical correlations, frame study design.

Table 3.

Qualitative comparison of in vitro skin models: accessibility, fidelity, throughput, and validation

Model type	Structural fidelity	Mechanical fidelity	Functional readouts	Accessibility	Throughput	External validation
Reconstructed 3D skin	Moderate (epidermal layers)	Low– (static)	High (barrier, ECM composition)	High	High	Histology/barrier markers (strong)
3D bioprinted tissues	High (microrelief, tunable geometry)	Moderate–High (programmable context)	Moderate	Moderate	Moderate	Emerging, study-dependent
Skin organoids	High (appendages, microanatomy)	Low–Moderate	Moderate	Low	Low	Emerging (variable)
Skin-on-a-chip	Moderate–High (depends on build)	Highest for dynamics (defined loading/perfusion)	High (molecular/structural under load)	Low–Moderate	Moderate (advancing with standardization)	Growing, needs harmonized metrics

“Fidelity” refers to resemblance to in vivo aging within the evidence presented here; labels are qualitative (High/Moderate/Low). “External validation” denotes alignment with in vivo/clinical findings reported in cited studies

Evaluation of anti-aging substances using advanced in vitro models

The evaluation of anti-aging natural substances and cosmetic skincare products primarily relies on static reconstructed human skin models and skin microphysiological systems. Current evaluations of anti-aging effects include: (1) Overall evaluation: β-galactosidase gene expression and enzyme activity, expression of the P16 gene, and concentration of reactive oxygen species (ROS), among others¹²³; (2) Evaluation of epidermal aging: paraffin sections to determine epidermal thickness, and gene transcription analysis of filaggrin (FLG), involucrin (IVL), and keratin by qPCR¹⁷; ZO-1 protein expression in tight junctions (TJs) was detected by immunofluorescence staining, reflecting skin barrier function^109,124,125. (3) Evaluation markers of basement membrane aging: gene transcription analysis of laminin-332, COL IV, and COL XVIII by qPCR^74,126. (4) Evaluation indicators of dermal aging: skin thickness was determined by paraffin sectioning; gene transcription of COL I and COL III was assessed by qPCR. Semi-quantitative analysis of fibroblast number and collagen expression was performed by immunofluorescence staining for α-SMA, COL I, and COL III^73,74.

Microphysiological system is an engineered model that incorporates sensors and other advanced technologies, enabling the application of a wide range of evaluation tools for conducting tests. Potential future anti-aging detection methods applicable to skin-on-chips include: (1) Image and skin texture analysis: For instance, Min et al.⁶³ used swept-source optical coherence tomography (SS-OCT) imaging techniques to measure the width and cross-sectional depth of skin wrinkles (Fig. 4c). (2) Cell–ECM interaction force: Gu et al.¹²⁷ employed photonic crystal cell force microscopy (PCFM) to image and quantify vertical cell force; Pauty et al.⁷² used traction force microscopy to measure the traction stress exerted by fibroblasts on matrix gels. Cell force measurements offer an alternative to secretory characteristics for analyzing fibroblast senescence phenotypes. (3) ECM stiffness: Eve et al. used Atomic Force Microscopy (AFM) to measure the stiffness of the rete ridge⁵, Lei et al.¹²⁸ used PCFM to calculate the stiffness pattern of skin organoids and observed an increase in matrix stiffness at the DEJ of psoriatic lesions. These techniques can be further applied in skin microphysiological systems to monitor ECM stiffness in the basal lamina and dermis, reflecting the aging process. (4) Assessment of barrier function: Trans-epithelial electrical resistance (TEER) is a rapid, non-invasive test that assesses epithelial integrity^108,109,125. Furthermore, permeability tests conducted on skin-on-chips can assess the resistance of skin models to external stimuli^108,109,125.

Challenges and prospects: simulating skin aging based on dynamic extracellular matrix changes

Key challenges in skin-aging models cut across disciplines: on the engineering side, closed-loop stability and reproducibility of perfusion, mechanical loading, and environmental control remain limited; on the materials side, current ECM surrogates poorly capture age-dependent nonlinear viscoelastic stiffening and glycation^129,130; biologically, long-term steady-state co-culture of immune, vascular, and microbiome components is fragile; in tissue engineering, vascularization, scale, and innervation are constrained; and translationally, throughput, standardization, and regulatory readiness are uneven^131–133. Transferable solutions include sensorization and feedback control¹³¹, staged differentiation with standardized perfusion regimens^132,134,135, sacrificial templating to construct perfusable microvessels¹³⁶, coupled with unified quality-control (QC) panels¹³⁴, and AI-oriented data and metadata standards^132,133; systematic solutions on the materials side are still under active exploration.

Extracellular matrix (ECM) remodeling is both a key hallmark of aging and a shared effector pathway. Skin-on-a-chip technologies now provide an advanced platform for dynamically modeling this remodeling over the course of cutaneous aging. Building on these capabilities, future models should, under real-time monitoring, couple perfusion with programmable mechanical loading so that oxygen and nutrient delivery, metabolic clearance, and stress–strain dynamics are temporally coordinated^107,131, thereby sustaining ECM turnover and providing boundary conditions for immune-cell ingress and remodeling. Mechanical loads matched in intensity and frequency help delineate the boundary between repair-promoting stimulation and overload-induced degradation. Through closed-loop control, the spatiotemporal dynamics of ECM during aging can be recapitulated more faithfully¹⁰⁷, laying a stable foundation for subsequent multicellular microenvironmental coupling.

Aging cannot be simplified as a binary relationship between the ECM and cells, but rather reflects the multidimensional coupling between the ECM, cells, and their microenvironment. Under controlled mechanical conditions, three-dimensional bioprinting recapitulates progressive matrix stiffening and age-related remodeling of cutaneous microrelief. Introducing primary cells from elderly donors, or reprogramming somatic cells into iPSCs followed by directed differentiation into keratinocytes, fibroblasts, endothelial and immune cells, enables comparisons between young and aged microenvironments; sourcing cells from multiple donors increases individual-level physiological relevance. Integrating microfluidics to build perfusable endothelial layers and microvascular networks^111,115, together with resident immune cells and the skin commensal microbiota¹³⁷, allows mechanical cues and solute gradients to interact with immune–microbiota–vascular signaling, promoting stable evolution of the microenvironment within the model^77,136,138.

On this engineering and biological basis, the model should integrate biochemical (oxidative stress, accumulation of advanced glycation end products), mechanical (stretching, compression), and environmental (ultraviolet radiation, pollutants) stimuli along the temporal dimension to dynamically reproduce progressive ECM remodeling, particularly matrix stiffening and collagen degradation. To achieve quantification and external validation, high-resolution imaging should be paired with mechanical measurements to resolve skin-surface microrelief trajectories and cell–ECM interaction forces, with standardized QC readouts across barrier function, secretome, and mechanics^131,134. Although some studies report quantitative metrics, direct cross-study pooling is not methodologically rigorous given substantial differences across models in device physics, endpoint definitions, measurement units, and analysis pipelines. In particular, reports of matrix stiffness (kPa) and transepithelial electrical resistance (TEER) are inconsistent, and wrinkle metrics rely on study-specific thresholds and fields of view. Accordingly, this review emphasizes clear, standardized reporting of key parameters and readouts to support the reliability of future meta-analyses. In parallel, machine learning and image analysis should enable multimodal data fusion and predictive modeling, ultimately forming a closed loop from challenge identification to design, evaluation, and application to support anti-aging therapy development and personalized interventions.

As a concrete yet adaptable blueprint, we envisage constructing a perfusable skin-on-a-chip as follows: a bioprinted dermal scaffold is endowed with DEJ-like microtopography and embedded with sacrificially templated microchannels; after removal of the sacrificial material, fibroblasts are first seeded throughout the interstitial compartment to stromalize the matrix and deposit extracellular matrix (ECM). Endothelial cells are then perfused at low flow through the channel compartment to achieve endothelialization and establish a perfusable microvascular bed. Finally, epidermal organoid units are seeded at the air–liquid interface to generate a functionally stratified epidermis with regions enriched for appendage-associated markers and structures. Cell sources may be age-stratified donor primary cells, or induced pluripotent stem cell (iPSC)-derived keratinocytes, fibroblasts, endothelial cells, and immune cells produced via staged differentiation and banked under a unified quality-control (QC) panel (identity, senescence markers, barrier/ECM gene sets). Once perfusion is stabilized, resident immune cells and, where appropriate, commensal skin microbiota are introduced, and predefined loading regimens (e.g., low-frequency cyclic stretch or circadian-like compression) are applied in synchrony with flow. Inline sensors (e.g., transepithelial electrical resistance, TEER; oxygen) together with standardized readouts (barrier, secretome, mechanics, ECM composition) enable real-time monitoring; high-resolution imaging of surface microrelief and transcriptomic signatures allow external validation against young and aged human references. To place these measurements on shared benchmarks, harmonization can be framed along four axes—device/interface, loading regimens, core readouts, and validation/QA. GB/T 44831–2024 specifies device performance and biological acceptance windows (https://std.samr.gov.cn/gb/search/gbDetailed?id=25940C3CEF7E8A9AE06397BE0A0A525A)—e.g., optical transmittance and leaktightness/temperature tolerance on the device side, and histologic stratification, an MTT OD window (0.6–3.0; SD ≤ 18%), and barrier thresholds (TEER ≥ 1 kΩ·cm², SDS-IC₅₀ = 1.0–4.0 mg/mL, or ET₅₀ = 4–9 h) on the biology side—whereas detailed descriptors and acceptance criteria for mechanical/flow regimens remain to be standardized. Interoperability is addressed by ISO 22916 (microfluidic device dimensions, connections, and initial classification)¹³⁹; standardization priorities across devices, data, and validation are outlined in the CEN-CENELEC organ-on-chip roadmap¹³²; OECD Test Guidelines (TG 431/439/428/497)^{135,140–142} anchor validated skin endpoints for core readouts; and regulatory qualification is provided via the FDA ISTAND program for Drug Development Tools¹⁴³. This integrated design—organoid seeding on a perfusable, dynamically loaded chip, linked to a donor-to-chip pipeline—offers a tractable path for hypothesis-driven studies while retaining compatibility with diverse materials, microfluidic implementations, and immunobiological modules.

Conclusion

In vitro models are indispensable tools for elucidating the mechanisms of skin aging and developing targeted anti-aging interventions. The integration of microphysiological systems, organoid culture, and 3D bioprinting technologies into next-generation skin models offers a promising solution to address the limitations of current skin aging models. Moving forward, the development of advanced in vitro models that more faithfully replicate aging characteristics will be essential for reducing reliance on animal models, improving research efficiency, and minimizing costs, while ensuring compliance with ethical standards.

Author contributions

Y.Y.: Conceptualization, manuscript design, investigation, writing original draft and visualization. Z.Z.: Discussion, manuscript design and writing original draft. J.Z.: Discussion, manuscript design and writing-original draft. Y.G.: Discussion, manuscript design and writing original draft. X.L.: writing review and editing. K.Y.: writing review and editing. N.S.: Conceptualization, supervision, funding, writing, and manuscript review. Z.C.: Conceptualization, supervision, funding, writing, and manuscript review. Z.G.: Supervision, funding, writing, and manuscript review. N.Y.: Supervision, funding, writing, and manuscript review.

Competing interests

The authors declare no competing interests.