IntroductionAs the primary interface with the external environment, the skin maintains homeostasis through coordinated barrier, immune and sensory functions1,2. Structurally, it comprises a stratified epidermis, fibroblast-rich dermis and a dynamic vascular network that regulates nutrient transport, thermoregulation and immune-cell trafficking3,4. Owing to this structural and functional complexity, human skin has become a central focus in biomedical, cosmetic and pharmaceutical research, serving as both a biological interface and a translational testbed for therapeutic discovery5,6.Over the past decades, multiple in vitro strategies have been developed to emulate human skin architecture and function. Two-dimensional keratinocyte monolayers and reconstructed three-dimensional cultures have been widely used for studies of epidermal differentiation and barrier physiology, whereas engineered human skin equivalents have enabled pharmacological testing and toxicological screening7,8,9. However, these systems lack key physiological features, including vasculature and immune-associated components, which are important for modelling inflammatory and vascular-associated skin diseases. Although animal models provide systemic complexity, interspecies differences, ethical considerations and limited scalability constrain their translational relevance10. Recent advances in tissue engineering, such as microfluidic skin-on-chip and bioprinting platforms, have partially addressed these challenges, yet still fall short in fully capturing the multicellular organization, dynamic signalling and tissue–immune interactions of native human skin11,12. In particular, the absence of vascular-associated components can restrict nutrient exchange, immune-cell trafficking and inflammatory crosstalk, thereby limiting long-term tissue maintenance and physiological modelling.Human induced pluripotent stem cell (hiPSC) technology has opened up new opportunities for modelling organogenesis through the self-organization of multiple cell types. Skin organoids (SkOs) derived from hiPSCs recapitulate epidermal and dermal differentiation and generate appendages such as hair follicles, melanocytes and neurons13,14. Despite these advances, SkOs exhibit notable limitations compared with native human skin. They often display inverted polarity, in which the epidermis is positioned internally and the dermis externally, and frequently undergo ectopic cartilage differentiation14,15. These abnormalities compromise physiological fidelity and standardization, hindering their scalability for high-throughput applications such as disease modelling or drug screening. Moreover, the lack of vascular and immune-associated components remains a major constraint, as endothelial and immune networks are important for establishing microenvironmental gradients, mediating inflammatory responses and evaluating pharmacological responses.Several recent approaches have attempted to overcome these limitations. Transient activation of WNT signalling during organoid development suppresses ectopic cartilage formation and promotes the generation of large, correctly oriented skin organoids suitable for modelling solar UV-induced damage and repair16. However, these models remain avascular and lack immune integration, which restricts their use for studying complex inflammation. In parallel, advances in the generation of hiPSC-derived blood vessel organoids (BVOs) have yielded self-assembled endothelial networks with lumina and mural support cells capable of forming perfusable vascular structures. Although these systems represent major progress towards physiologically relevant vascular complexity15,17, a modular and experimentally accessible skin model that enables controlled integration of epithelial, vascular and immune components, epidermal-side allergen stimulation and downstream pharmacological testing remains limited. The concept of assembloids, which modularly integrate lineage-specific organoids to reconstruct tissue–tissue interfaces, offers a promising strategy for overcoming these barriers18. Assembloids have recently been applied to model complex inter-organ interactions, including vascularized tumour systems and brain–immune interfaces, thereby enabling functional readouts such as immune infiltration and therapeutic response19,20. Inspired by these developments, we sought to create a human skin assembloid system that incorporates vascular and immune components while preserving experimental control over component assembly, tissue accessibility and inflammatory perturbation. In this context, the modular assembloid strategy is not simply a method for combining tissue components, but provides a controllable platform in which separately generated SkOs and BVOs can be assembled, matured under air–liquid interface conditions, exposed to epidermal allergens and subsequently analysed for coordinated epithelial, vascular and immune-associated responses.In this study, we established vascularized skin assembloids (VSAs) by modularly integrating hiPSC-derived SkOs and BVOs under air–liquid interface culture. VSAs displayed compartmental organization comprising a stratified epidermis, a fibroblast-rich dermal compartment, hair follicle- and neural-associated elements, and interconnected endothelial networks with luminal structures. Unlike conventional organoid models, VSAs supported immune co-culture and provided an experimentally accessible architecture for epidermal-side allergen exposure, enabling analysis of vascular-associated inflammatory responses and pharmacological modulation. Functionally, VSAs modelled selected features of house dust mite (HDM)-induced inflammatory remodelling relevant to atopic dermatitis and responded to pharmacological modulation of JAK–STAT signalling, supporting their use as a structurally complex, vascularized and immune-integrated human skin assembloid model. Collectively, these findings position VSAs as an advanced platform for mechanistic dissection, disease modelling and preclinical pharmacological evaluation in human inflammatory skin biology.Materials and methodsHuman induced pluripotent stem cell cultureThis study was carried out in accordance with the approved guidelines of the Seoul National University Institutional Review Board (IRB-No.2504/003-001). The hiPSC line CMC003 was obtained from the National Stem Cell Bank of Korea. HiPSCs were maintained in StemFlex (Gibco, USA) on vitronectin (Gibco)-coated dishes, and passaged with ReLeSR reagent (STEMCELL Technologies, Canada).Generation of induced pluripotent stem cell-derived human skin organoidsBased on previous reports13,16, iPSCs were differentiated into SkOs. Briefly, on day 0, a total of 1200 dissociated iPSCs/well were seeded in StemFlex with 20 μM ROCK inhibitor Y-27632 (Tocris, UK) in an ultra-low attachment 96-well plate (Corning, USA) to generate uniform embryoid bodies (EBs). When the size of the EBs became 250–300 μm, the EBs were transferred into individual new ultra-low attachment 96-well plates in Essential 6-based differentiation medium containing 2% Matrigel (Corning), 10 μM SB431542 (Tocris), 4 ng ml−1 bFGF (R&D Systems, USA), and 2.5 ng ml−1 BMP4 (Peprotech, USA) to initiate non-neural ectoderm induction. After 3 days of culture in non-neural ectoderm induction medium, 200 ng ml−1 LDN (Tocris), 50 ng ml−1 bFGF, and 15µM CHIR99021 (Tocris) were added at 25 μl per well to induce cranial neural crest cell formation. On day 6 of differentiation, 75 μl of fresh E6 medium was added. Half of the medium was changed on days 8 and 10. On day 12, to induce self-assembly of the epidermis, all organoids were transferred into ultra-low-attachment six-well plates and placed on an orbital shaker (SPL Life Science, Republic of Korea) in 3 ml of skin maturation medium containing a 1:1 mixture of advanced DMEM-F12 and Neurobasal medium (Gibco) supplemented with 1X N-2 supplement (Gibco), 0.5X B-27 without vitamin A supplement (Gibco), 2-mercaptoethanol (Sigma-Aldrich), GlutaMAX (Gibco) and Primocin (InvivoGen, USA). All cells were incubated in a humidified CO2 incubator at 37 °C.Generation of induced pluripotent stem cell-derived human blood vessel organoidsBVOs were generated and cultured following the protocol as previously described21. Briefly, hiPSC aggregates were generated by seeding in a low-attachment plate in StemFlex with 50 μM Y-27632. The next day, hiPSC aggregates were collected and the medium replaced with N2B27 medium supplemented with 12 μM CHIR99021 and 30 ng ml−1 BMP4. On day 3, the medium was replaced with N2B27 supplemented with 100 ng ml−1 VEGF-A (Peprotech) and 2 μM forskolin (STEMCELL Technologies). On day 5, aggregates with a diameter of 250–300 μm were selected and embedded in Collagen–Matrigel mixture (3:1 ratio) in 12-well Transwell inserts, with the aggregates placed in the centre of the extracellular matrix (ECM) solution. The cultures were maintained in StemPro-34 SFM medium (Gibco) supplemented with 15% fetal bovine serum (Gibco), 100 ng ml−1 VEGF-A, and 100 ng ml−1 FGF2.Fused skin-vascular assembloid culturesAt approximately 85 days, each cyst-like SkO was cut into eight evenly sized portions and placed onto the ECM, on which BVOs (day 12 of BVO generation) had been formed. The VSAs were maintained for 3–4 weeks in an incubator. The skin maturation medium was changed every 3 days. To further induce epidermal maturation, the VSAs were transferred to a dry-conditioned incubator at 37 °C for an additional 6 days, and the medium was replaced daily.Immunofluorescence assayThe cryo-sectioned organoids were fixed with 4% paraformaldehyde. After washing with PBS, the samples were blocked with 5% normal goat serum and incubated with the primary antibodies overnight at 4 °C (Supplementary Table 1). After PBS washing, fluorescence-conjugated secondary antibodies (Invitrogen, USA) were applied for 1 h at room temperature. For nuclei staining, 4,6-diamidino-2-phenylindole dihydrochloride (DAPI, Invitrogen) was applied for 10 min at room temperature. The stained samples were mounted with fluorescent mounting medium (DAKO, Denmark) and visualized using a confocal microscope (Eclipse TE2000, Nikon, Japan) with EZ-C1 3.8 software (Nikon) or a THUNDER Imaging System (Leica Microsystems; NFEC-2025-09-308232). The mean fluorescence intensity or the number of cells expressing the protein of interest was quantified by ImageJ software.Single-cell RNA sequencing and analysisSingle-cell transcriptome profiling of SkOs and VSAs generated from hiPSCs was performed. Organoids and assembloids were dissociated using the Whole Skin dissociation Kit, human (Miltenyi Biotec) for 4 h at 37 °C in a 1.5-ml microcentrifuge tube. Next, the organoids and assembloids were carefully washed with PBS, and residual tissue fragments were gently triturated to obtain single-cell suspensions. Single-cell libraries were constructed using Chromium Next Gem Single Cell 3′ Reagent Kits v4 (10x Genomics) and sequenced on an Illumina NovaSeq X Plus platform (paired-end 150 bp). FASTQ files were processed with Cell Ranger (v9.0, 10x Genomics) to generate gene-expression matrices. Downstream analyses were performed using the Seurat package (v5.0.0, R), Loupe Cell Browser (10x Genomics), and Winseurat (ebiogen, Republic of Korea). Low-quality cells were filtered using standard thresholds, and datasets from SkO and VSA samples were normalized, integrated and clustered. Dimensionality reduction was conducted with principal component analysis and uniform manifold approximation and projection. Cell-type annotation was guided by canonical lineage markers, and visualization was performed using Seurat plotting functions, ggplot2 and Loupe.Lucifer Yellow permeability assayTo assess epidermal barrier-associated permeability, Lucifer Yellow CH dipotassium salt (Invitrogen) was applied to the epidermal surface of SkOs and VSAs. After incubation for 4 h at 37 °C, samples were washed with PBS, fixed with 4% paraformaldehyde, embedded, cryosectioned and analysed by fluorescence microscopy. Lucifer Yellow penetration into the epidermal and dermal regions was quantified using ImageJ software.Histological analysisFor haematoxylin and eosin (H&E) staining, sections were stained with haematoxylin solution for 5 min, differentiated in 0.5% HCl in ethanol, and counterstained with eosin for 1 min. For Masson’s trichrome staining, sections were re-fixed in Bouin’s solution overnight at room temperature. The sections were then stained with Weigert’s haematoxylin for 10 min, followed by Biebrich scarlet acid fuchsin for 5 min. Differentiation was performed using a mixture of phosphotungstic acid, phosphomolybdic acid and distilled water (1:1:2) for 10 min. Subsequently, the slides were directly transferred to 2% aniline blue for 5 min and washed with 1% acetic acid. Following dehydration through a graded ethanol series, the samples were mounted with Canada Balsam. Stained sections were visualized with an Olympus microscope and ProgRes CapturePro software (Olympus, Japan).In vivo transplantation and vascular accessibility analysisAll animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of Seoul National University (IACUC No. SNU-250310-3-1). 6-week-old female BALB/c-Foxn1nu/Arc mice were purchased from JA BIO Co., Ltd. and maintained under pathogen-free conditions at the Animal Center for Pharmaceutical Research of Seoul National University. To evaluate the in vivo persistence and vascular accessibility of VSA-derived vascular structures, VSAs were transplanted into mice. Briefly, mice were anaesthetized, and a full-thickness dorsal skin wound was generated using a 5-mm biopsy punch. VSAs were placed onto the wound bed and secured by suturing the surrounding host skin. Grafts were allowed to engraft for 4 weeks before analysis. To assess vascular accessibility, rhodamine-conjugated Ulex europaeus agglutinin I (UEA-1) was intravenously injected before tissue harvest. Graft-containing skin tissues were collected, fixed with 4% paraformaldehyde, embedded, sectioned and analysed by histological and immunofluorescence staining. Human nuclear antigen, human CD31, mouse CD31 and UEA-1 staining were used to assess graft-derived human tissue, host–graft vascular association and vascular accessibility.Umbilical cord blood-mononuclear cell isolationUmbilical cord blood (UCB) was obtained from the Seoul metropolitan Government with the approval of the Seoul National University Institutional Review Board (IRB-No.E2405/002-021). Human mononuclear cells (MNCs) were isolated as described previously22. In brief, UCB samples from independent healthy donors were diluted in PBS at a 1:1 ratio, and human MNCs were isolated from the supernatants by Ficoll gradient centrifugation (Ficoll-Paque PLUS, GE Healthcare). MNCs from each donor were independently incorporated into VSAs to assess the reproducibility of immune responses.Co-culture of vascularized skin assembloids and mononuclear cellsFor the establishment of co-culture systems, 12-well Transwell inserts (Corning) were pre-coated with 150 μl of collagen solution and allowed to gel at 37 °C. MNCs (3×105 cells) were suspended in 500 μl of MNC medium and seeded onto the collagen gel within the inserts. After 24 h of incubation, the culture medium was removed, and the VSAs were subsequently placed onto the inserts. Co-cultures were maintained in a mixed medium composed of skin maturation medium and MNC medium at a 1:1 ratio. The MNC medium consisted of RPMI1640 (Gibco) supplemented with 10% FBS and Primocin.Immune stimulation with lipopolysaccharidesTo evaluate the immune responsiveness of immune-supplemented VSAs, cultures were stimulated with lipopolysaccharides (LPS) (100 ng ml−1; Sigma-Aldrich). LPS stimulation was performed twice at 2-day intervals, and assembloids were harvested 2 days after the final treatment for downstream analyses and histological assessment.Allergen-induced inflammatory modellingTo establish an HDM-induced inflammatory model, VSAs were stimulated with HDM extract (Dermatophagoides pteronyssinus; Citeq, Netherlands) at a dose of 5 µg per sample (in 5 µl). HDM stimulation was performed three times at 2-day intervals, and samples were harvested 3 days after the final treatment for cytokine analysis, immunostaining and transcriptomic profiling.Flow cytometric analysis of immune cells in vascularized skin assembloid–mononuclear cell co-culturesTo characterize immune-cell populations incorporated into VSAs, VSA–MNC co-cultures were dissociated into single-cell suspensions at the indicated time points using the Whole Skin Dissociation Kit, human (Miltenyi Biotec). The cell suspensions were filtered, washed with FACS buffer and stained with the StainExpress Immune Cell Composition Cocktail, human (Miltenyi Biotec), or with fluorophore-conjugated antibodies against CD45, CD3, CD11c, HLA-DR and CD1c for myeloid/APC-like cell analysis. Dead cells were excluded using a viability dye. Samples were analysed using a MACSQuant Analyzer 10 flow cytometer (Miltenyi Biotec). Compensation was performed using MACS Comp Beads and the MultiColor Compensation function. Flow cytometry data were analysed using MACSQuantify software.Quantitative real-time PCROne sample of SkOs from each group was washed with PBS and lysed in TRIzol (Macherey-Nagel, Germany) for total RNA extraction. Complementary DNA synthesis was performed using a Superscript III First-Strand synthesis system kit (Invitrogen) according to the manufacturer’s instructions. Quantitative real-time PCR analysis was performed with SYBR Green PCR Mix (Applied Biosystems, USA) on a 7500 Real-Time PCR system (Applied Biosystems). The relative mRNA expression of each target gene was quantified by the 2-ΔΔCt method and normalized to the housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Supplementary Table 2).RNA sequencing and bioinformatics analysisTotal RNA was extracted from organoids using TRIzol reagent and libraries were prepared with a commercial RNA-sequencing (RNA-seq) library preparation kit. Sequencing was performed on the Illumina NextSeq 2000 platform with paired-end reads. Raw data were aligned to the human reference genome (GRCh38), and differentially expressed genes were identified based on fold-change and statistical thresholds. Heatmaps and volcano plots were generated to visualize gene-expression patterns. Gene ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses were performed, and gene set enrichment analysis was conducted to evaluate functional pathways. To benchmark the HDM-induced VSA transcriptomic response against human atopic dermatitis (AD), a publicly available bulk RNA-seq dataset of skin biopsies from healthy controls and patients with AD was used (GSE277961).Enzyme-linked immunosorbent assayTo analyse the secretion of VSAs under both baseline and HDM-exposed conditions, the supernatant of conditioned media was collected and subjected to enzyme-linked immunosorbent assay (ELISA). To assess endothelial-specific functions, the concentrations of human VEGF and nitric oxide were quantified using the Human VEGF Quantikine ELISA Kit (DVE00, R&D Systems) and the Nitric Oxide Detection Kit (iNtRON Biotechnology, Republic of Korea), respectively. To evaluate immune-related responses in the HDM-induced inflammatory model, the levels of thymic stromal lymphopoietin (TSLP; DY1398, R&D Systems), TARC/CCL17 (DY364, R&D Systems), IL-4 (DY204, R&D Systems) and IL-13 (DY213, R&D Systems) were measured using the corresponding human ELISA kits according to the manufacturer’s instructions.Western blotSample lysates were extracted by homogenization with PRO-PREP (iNtRON Biotechnology), followed by sonication. The 20 μg of proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. The membranes were blocked with 3% bovine serum albumin at room temperature for 1 h and then incubated on an agitator with the primary antibodies at 4 °C overnight (Supplementary Table 3). Membranes were washed and incubated with secondary antibodies (Invitrogen) at room temperature for 1 h. Each band was detected with an enhanced chemiluminescence detection kit (Cytiva, USA).Proteome profiler arraysConditioned media collected from VSAs were subjected to cytokine profiling using the Proteome Profiler Human XL Cytokine Array Kit (R&D Systems) in accordance with the manufacturer’s protocol. Membranes were incubated with the samples, and bound cytokines were detected by chemiluminescence. Signal intensities were quantified using ImageJ software, and relative levels were compared across experimental groups.Statistical analysisSample size (n) represents the number of biologically independent replicates, with at least three independent experiments performed for each condition. All values are presented as mean ± standard deviation (SD), and all graphs were generated using GraphPad Prism version 10.0. Statistical analyses were performed using unpaired two-tailed t tests, one-way analysis of variance with Tukey’s post-hoc test, or two-way analysis of variance with Bonferroni’s post-hoc test. Statistical significance was denoted as *P